Citizens,

Families

& Reform
Citizens,

Families

& Reform
Stein Ringen
With a new introduction by the author
Originally published in 1997 by Oxford University Press

Published 2006 by Transaction Publishers

Published 2017 by Routledge

2 Park Square, Milton Park, Abingdon, Oxon OX14 4RN
711 Third Avenue, New York, NY 10017, USA

Routledge is an imprint of the Taylor & Francis Group, an informa business

New material this edition Copyright © 2006 by Taylor & Francis.

All rights reserved. No part of this book may be reprinted or reproduced or

utilised in any form or by any
electronic, mechanical, or other means, now
known or hereafter invented, including photocopying and recording, or
in
any information storage or retrieval system, without permission in writing
from the publishers.

Notice:
Product or corporate names may be trademarks or registered trademarks,
and are used only for identification and
explanation without intent to
infringe.

Library of Congress Catalog Number: 2005043675

         Library of Congress Cataloging-in-Publication Data

Ringen, Stein.
 Citizens, families, and reform / Stein Ringen ; with a new introduction by
 the author.
   p. cm.
 Originally published: Oxford : Clarendon Press : New York : Oxford
 University Press.
 Includes bibliographical references and index.
 ISBN 1-4128-0498-1
  1. Social policy. 2. Family policy. 3. Child welfare. 4. Public welfare.
 I. Title.
HN18.R47 2005
361.6'1—dc22

      2005043675

ISBN 13: 978-1-4128-0498-1 (pbk)

In the sentiment running through this text which links goodness and
restraint, I owe much to my parents. Analyses
which started from a
methodological interest in measurement have lifted family to the forefront
of attention, more
strongly than I had anticipated. I dedicate this book to my
father and to the memory of my mother.
 CONTENTS

Introduction to the Transaction Edition

Acknowledgements
Preface: Reason and Restraint

PART I WELL-BEING

Introduction: Hard Facts about Soft Values

1. Persons
2. Families
3. Children

PART II REFORM

Introduction: Soft Means for Hard Goals

4. Class
5. Income
6. The Civic Spirit
Postscript: A Statement on Political
Philosophy
References
Index
 INTRODUCTION TO THE

TRANSACTION EDITION

SCIENTIFIC DELIBERATION

The twentieth century made itself the century of social science. In the
process it also produced what was,
until close to its end, the authoritative
answer to the question of what the social sciences are for. That
answer was
contained in the idea of scientific management: once we understand human
behavior and
organization well enough we can manage human affairs with
scientific rationality.
For all the hopes that were invested in that idea, it finally proved itself a
monstrosity. It paid too little
attention to ordinary people. They are, it
turned out, not simple cogs in the machinery of society who will be
happy if
only managed well. Far from it; they are wilful individuals, often with
priorities different from
those of their fellows, and usually capable of
rationality on their own. Scientific management paid too much
attention to
extraordinary people, the custodians of scientific knowledge. They, the
experts – social engineers
in the work place, psychologists in the family,
economists in the economy, and the gentlemen of the central
committee in
governance – were the ones who were going to do the managing.
The twentieth century was also the century of democracy. But while
democracy prevailed, scientific management
collapsed. Indeed, scientific
management collapsed because democracy prevailed. The idea of
democracy
is that people should manage themselves because they know
best how they want their affairs to be arranged. The
idea of scientific
management is that they should be managed by others who know more than
they do. While still
believed in, that idea contributed to many ills and failed
experiments, such as taylorism, collective child
rearing, euthanasia, central
planning, autocracy, and dictatorship. As a result, scientific management
has
fallen into deserved disrepute.
 That is all for the good, but it has left those of us who work in the social
sciences with a bit of a hangover.
If the idea that was, it must be confessed,
our idea was disgraced and if the social sciences were
capable of being
dangerous, what then?
In 1990, I was invited to take up the newly created chair in sociology and
social policy at Oxford University,
its first chair in sociology. I used my
inaugural lecture to make my contribution to the healing of our
collective
hangover. I am in no doubt that we are justified in defending our business as
a scientific
one.1 Nor about the importance of
this business to modern life.
But how should we make ourselves useful given that we want modern life
to be
lived democratically?
For my own answer to that question I took inspiration from an idea of
restraint. “Neither in politics
nor in science – no more than in love, family
or friendship – can we succeed for ourselves or be of use for
others if we
always insist on holding our ground.” Scientific management was social
science unrestrained, the
social scientist as the arbiter of truth and wisdom.
We social scientists have, I think, good reasons to be
assertive about the
contribution we can make towards order in society. My reason for saying
that is not miles
away from scientific management: there are some things
we should know better than others. But there is also
much we cannot know
better than others and therefore we should offer our services in a spirit of
restrained
assertiveness.2
My argument was essentially a democratic one. Since then, we have seen
some important advances in democratic
theory under the label of
“deliberative democracy.”3 That enables me now to give the program I was
suggesting then a proper name. It is
diametrically opposed to scientific
management, yet not miles apart. I call it scientific
deliberation.
The theory of deliberative democracy says that rationality in collective
decision-making and governance depends
on more than the aggregation of
preferences in elections and similar arrangements, it rests as well on the
honing of those preferences through continuous processes of deliberation.
What we want to do in those processes
is to cut through the fog, identify the
matters that really deserve disagreement and then try to work our way
through to some kind of shared platform or reasonable compromise where
we recognize the inevitability that
there is no single compromise that is the
best one for all concerned. Deliberative democracy is an
eminently
idealistic model of citizens of good will working earnestly with each other
to make themselves
enlightened and work their way through their inevitable
differences in a civilized manner. It is the kind of
idealism that appeals to
someone like myself who believes that reason comes from restraint.
Productive deliberation depends crucially on information. Obviously, it
depends on much more—effort
and honest good will—but information is
essential. Although no sharp dividing line can be drawn between factual
and moral questions, we are in a much better position to concentrate debate
on the predominantly moral ones,
which are the questions that deserve
debate, if we know as much as possible about the predominantly factual
ones. Information is obviously no less necessary in actual decision-making
than in deliberations before and
after decision-making, but the theory of
deliberative democracy helps us to understand better the significance
of
informed democracy.
Informed democracy, in turn, depends on conditions of its own. A first
one is that citizens have access to the
best quality of relevant information.
For that they should turn to the scientists; science is not
omni-
knowledgeable nor infallible but the scientific method is the best we have.
The scientists, therefore,
have the capacity to be of service by contributing
to clearing away the fog of ignorance. A second condition is
that
information be free and not under the control of parties or interests who
might use it manipulatively. The
scientists, therefore, have a duty, in this
regard, to inform democracy objectively.
Information, I said in that lecture, is “the common ground on which
science and politics can meet without
scientists playing politicians or
politicians playing scientists; information is what politics needs and
science
can provide.” That led me to a three-item program for the contribution of
social research to scientific
deliberation: First, as social scientists we should
accept responsibility for informing the democratic
process. Second, we
should not only accept that responsibility but also claim this as our
authority.
Third, although we should claim the domain of information as our
own, we should not, as scientists,
transgress into the domain of decisions.

TRUTHS AND MYTHS

The lecture that contains this program is reproduced as the last chapter in
this book. The rest of the book is
dedicated to doing what I said there
should be done. I take on two issues on which I think we social scientists
have not done our part of the job towards scientific deliberation. I could
have called the book “The
Truth About Families and the Truth About
Inequality.”
When I say “we social scientists,” I’m afraid I mean “we sociologists.”
Although the family is the sociological
institution par excellence, family
sociology has been wholly unsuccessful in ridding political
deliberation of
one of its great myths. That is the belief that modern families – or the
family as is
often the terminology – are typically small nuclear families,
that the nuclear family has arisen from the
demise of the large extended
family, and that while extended families were multi-functional, nuclear
families
are “reduced” to specializing in catering to the emotional needs of
family members. In my native language,
Norwegian, the concept is
funksjoristϕmming, the family bereft of functions.4
I am sure it would be unfair to accuse family sociology of having created
this myth, but it is regrettable that
it has not been able to kill it. Still today,
it is impossible to listen to any debate about family policy
without the
mythology of “the nuclear family” being offered up as if it were obvious
truth.5
The matter is important. If the nuclear family story is believed, the
recommendation in social policy would be
to build up alternative
institutions that could take on the functions families can no longer perform.
If that
were to happen, social policies might contribute to the family demise
that is thought to have already happened
and we would be in a spiral of self-
fulfilling prophesies.
Since I had become distrustful of family sociology – there is altogether
too much “discourse” and too little
hard-nosed empiricism – I turned
instead to the tools of economic analysis. That allowed me to ask and
answer
the question of what families do. And what I found was that modern
families in modern societies in
fact do astonishingly much, that much of
what they do is elementary economic production, and that therefore
they are
not sites of “only” social and emotional life.
While family is a sociological favorite, inequality is another one. One
myth about inequality, very widely
believed in European sociology today, is
that changes in social structure which we often associate with
progress in
our part of the world, such as economic growth or the expansion
of access to higher education, contrary to expectations tend not to filter
down to less inequality
between persons or groups in well-being or
opportunity.
If that story were true, its influence in public policy could be drastic and
unfortunate. We would, for
example, have to take it that a policy of opening
higher education up to more young people would probably make
little or no
contribution towards a more open society. Exactly that scenario is presently
unfolding in Britain.
The government has announced that it wants to extend
university-level education enough to admit 50 per cent of
graduates from
secondary school (from short of 40 per cent today). That policy is meeting
massive resistance (a
resistance which is really only understandable in light
of a deep skepticism to education in British culture).
The resistance comes,
predictably, from taxpayers and some of those who see themselves as their
representatives, but also, less predictably, from a mixed lobby of
commentators who sometimes argue that the
new graduates would not be
able to benefit from university training and sometimes that the
consequences would
be a levelling down of standards.
The myth about inequality that I have referred to plays a minor part in
this drama. If it cannot be expected
that more places in university education
will contribute to less inequality, then from that point of view there
is no
reason to go down that road.
I have for years been concerned with issues of public policy, inequality,
and welfare. In that career, during
the last twenty years or so, I’ve done
battle with a parade of propositions that by and large see inequalities
to be
so entrenched in economic and social organization that they are beyond the
reach of reformist public
policy. As a result of this feisty warfare, I believe
that I have been able to pick off those propositions one
by one.
First proposition: The distribution of income in industrial societies tends
towards stability and is not
sensitive to (weak) policy interventions, at least
in the long run. My reply (in The Possibility of
Politics): Advances in
comparative research on income since about 1970 have shown that income
distributions fluctuate strongly within nations and differ strongly between
nations, and that such fluctuations
and differences are associated with
changes and differences in government policies.6
Second proposition: Although the distribution of income is not stable, the
incidence of poverty is, the
implication being that reformist policies are not
sufficiently powerful to rectify injustices at the bottom of
the distribution.
My reply (in “Direct and Indirect Measures of Poverty”): This proposition
has been shown to
rest on weak definitions of poverty that make empirical
results non-interpretable, not sufficiently robust for
alternative methods of
measurement, and unable to stand up to stricter and more appropriate
criteria of
definition and measurement.
Third proposition: Although the distribution of income and the incidence
of poverty are not insensitive to
reformist policies, class inequalities of
opportunity are, the implication being that deep, hard-core
inequalities
persist unabated. My reply (in chapter 4 below): Support for this last stand
of the
anti-reformists has been sought in elaborately estimated statistics that
propose to show long-term stabilities
in class inequality. These statistics
have, however, proved no less fragile than those on stable income
distributions and poverty rates. Stability results rest on excessively
relativized methods of measurement that
are biased in favor of the preferred
hypothesis, the results are not robust enough for tests by alternative
methods, and are disconfirmed by methods of unbiased measurement.
The first two propositions have been successfully despatched and those
matters can be considered settled. The
third one, however, is still believed
to be true and therefore deserves another round of attention.

THE TRUTH ABOUT FAMILIES

It turns out that economic analysis is ideally suited to uncover what family
sociology has found it so
difficult to describe: what families do.7 Of course
that analysis can only tell us what goes on economically within
families, but
that is just what we need to know for our purpose. The theory of
funksjonstϕmming will
have it that less is done in families precisely
because they no longer perform economic functions.
What made it possible to analyze within-family economies with a detail
and authority that was innovative at the
time, and perhaps still is, was the
availability of two British data sets that Stephen Jenkins, Jay Gershuny,
Brendan Halpin, and I were able to combine through a process of synthetic
merging, one on family income and one
on family time-use. I was able to
use those merged data to measure the volume of within-family production
and
to add the value of that production to the value of goods families could
buy in the market for their cash
income. I was also able to introduce
assumptions about economic co-operation in families and to impute values
to the gains that are made through economies of scale when total goods
available to families are metamorphosed
into consumption value for family
members.
All of this is, of course, quite technical, as can be seen in chapters 1 to 3,
but the long and the short of it
was a conclusion, based on cautious and
conservative assumptions and understating rather than overstating,
that, in
these modern families in this modern society, within-family economies of
production and co-operation
add as much value to final consumption as the
value of goods and services consumers can buy in the market for
the
income available to them. If the national economy is what produces a
population’s capacity for consumption,
half of the nation’s economy is in
the nation’s families.
That conclusion, which I think is absolutely robust, was and is, in my
opinion an astonishing one. If we think
modern families are bereft of
functions, we are simply wrong. If we think modern families are no longer
economic institutions, we are simply wrong. More generally, if we think
that our well-being, even defined in
narrow economic terms, is generated in
formal settings “out there” in society, such as in markets, and then
taken
home to the informal setting of the family for enjoyment, we are also
wrong. What we take home from
markets are raw materials that are
processed further in the family before we sit down and consume the
combined
product of market and family economies. That processing in the
family is no-nonsense economic enterprise: it is
adding value to raw
materials by applying labor force and capital to them and it is sharing and
distributing
the produce.
According to my program of scientific deliberation, my job is to establish
facts such as these but not to go on
and draw conclusions for what should
be done in social policy. That’s for deliberation. What the facts say is
that
modern families are not bereft of functions and that they are what families
have always been: vibrant and
essential institutions in our quest for well-
being. If, however, I were to disregard my own rules of scientific
deliberation and transgress into the domain of decisions, which I of course
do, I would say for social policy:
invest in families! The truth about
families is that they are the most productive institutions known to
mankind.

THE TRUTH ABOUT INEQUALITY

When I joined Oxford University in 1990, I encountered a small, tightly
knit community of sociologist who were
dedicated to the analysis of class
inequality. (An exception was Brian Wilson, the eminent sociologist of
religion.) The question that preoccupied my colleagues was whether
industrial modernization follows through to
societies of more open
opportunity. That is the assumption of liberal class theory, but the matter is
not
obvious and therefore needs to be tested empirically. The Oxford
sociologists, mainly in Nuffield College, had
been exploring this problem
for years with much determination and quite exceptional skills in what must
count
among the most impressive projects ever undertaken in empirical
sociology.
It was also a project that by now was producing a firm answer to its main
question. That answer was to reject
the liberal assumption of the withering
away of the rigidity of class. This conclusion was based on the
observation
in Britain that during the half-century leading up to 1980, in spite of a
strong upward thrust of
social change that brought increasing numbers from
all backgrounds of social class into higher education and
middle-class jobs,
“no significant reduction in class inequalities was in fact achieved.”8 This
became “the stability thesis of
class inequality.” Soon enough, that thesis
was found to apply not only over time in Britain, but also (almost
without
exception) from country to country in Europe, and this included
comparisons between the advanced
capitalist countries and the more
backward ones whose post-war experience had been of communist
regimes.9 Years of empirical
research had established a high “degree of
temporal constancy and cross-national communality,” something that
“might be regarded as the main achievement to date of class analysis,” so
much so that “the focus of
theoretical effort in the field should now be to
explain” that constancy and communality.10
This, for me, was sociology worth taking seriously: persistent, excitingly
original, patient, earnest and hard
working, deeply scientific in its
orientation, committed to the testing of hypotheses with the careful use of
the best available data and advanced methods of measurement. All this
effort had paid off. To observe class
inequality is to look down into the
foundations of the social architecture. What the Nuffield sociologists saw
there, in the careful way they looked at it, gave rise to an outlook on the
very nature of present-day
social organization. Beneath all that has
obviously been changing in the process of industrial modernization,
there
is, deep down, a persistent stability in social conditions, at least as observed
through class
inequality. In history the question is always, continuity or
change? This research strongly tilted the
interpretation of modern social
history towards a message of continuity. Everything seems to be
changing,
but really it’s all the same. There are big questions asked and answered
here.
In the community I came into, the stability thesis was gospel. It was
backed up by years of outstanding
research on social mobility, educational
attainment, class politics, and more. It was celebrated year after
year in a
string of brilliant seminars, often by invited speakers who added solidity to
the thesis from their
own research. Followers elsewhere took it up and
proved and re-proved its validity.11 The only problem for me was that I did
not believe a word of it.
There were also other disbelievers. One was Professor Ray Pahl, another
outstanding British sociologist who
wrote a sharp and perceptive criticism
in a paper he titled “Is the Emperor Naked?” He was invited to give a
seminar on his criticism, which he did, possibly to his regret. The research
in question is nothing if not
highly sophisticated in statistical methodology.
Professor Pahl, although in full control of the substance of
the matter, is not
a quantitative methodologist and was not in a position to debate the
methodological basis of
the research he was assessing. His criticism could
therefore easily be dismissed as based on misunderstandings,
which indeed
it was on that memorable occasion.12
By happy coincidence, my own teacher in quantitative methodology from
the University of Oslo, who by now had
become a colleague, Professor
Ottar Hellevik, was at about this time in Oxford for a period of research. We
found that we had a shared interest. He, a first-rate methodologist and the
author of several textbooks on
statistical methods for the social sciences,
was skeptical of some uses of log linear regression analysis in
sociology,
which he saw as a complicated and impenetrable way of doing what could
usually be done by simpler
and more transparent techniques. Log linear
analysis is the methodology on which everything stands or falls in
this
branch of class analysis. I, from my own work on income inequality, was
skeptical of the stability
conclusion, which was simply totally at odds with
what I had been finding about trends in inequality.
Ottar Hellevik and I resolved to sort this out. We took the case of class
background and educational
attainment, one of the associations that had
been found to have remained roughly stable over decades in
Britain. And
we decided to do it by reanalyzing the British mobility data in which our
colleagues had read
stability of class inequality, obviously using a different
methodology than theirs. That took a bit of doing.
It is in fact, even with the
assistance of the best available methodological competence and persistent
work,
very difficult, to put it mildly, to grasp just what log liner regression
analysis does in the kind of analysis
in question. It also took time to devise
and test out the alternative techniques. Result: about three years of
work
producing the chapter on class in this book and a separate paper authored by
Ottar Hellevik.13
Empirical research on class inequality and mobility consists in the
reading of mobility tables. If we are
interested in trends in inequality, which
is the issue here, we need to observe consecutive mobility tables to
see if
the most recent ones display less or more inequality. The research that has
produced the stability
thesis relies for readings of inequality on the use of
log linear regression analysis to estimate what is known
as “odds ratios.”14
What we did in our reanalysis was to keep everything the same except
the statistical technique. We read the
same mobility tables with the same
data for the same country for the same period. But instead of odds ratios we
adapted the standard technique in the study of inequality of gini-indices
estimated from Lorenz curves to the
analysis of mobility table data.
Two ways of reading the same facts, then. With diametrically opposite
results, as it turned out. Where our
colleagues had seen stability, we saw a
straight, sharp, and remarkable reduction in class inequality of
educational
attainment. (See table 4.2 in chapter 4.)
This is the place to try to explain the guts of this mystery in as
nontechnical a way as possible (and then
those who want the technical
explanation can go to chapter 4 and to other works referred to).
A mobility table contains two sets of distributional data. The first set
consists of two marginal
distributions, say the proportion of the population
with a working-class background, a middle-class background
and so on,
and the proportions that finish education at the basic level, that go on to
secondary education and
so on. The second set consists of the conditional
distributions in the space defined by the two marginals, say
those
originating in the working class and finishing education at the basic level,
those originating in the
middle class and going on to university education
and so on.
 Marginal and conditional distributions are related. If something changes
in the marginal distributions,
something has to change also in the
conditional ones. If for example, as in the British case, the number of
places
in university education is extended and those places are taken up, the new
students have to come from
somewhere so that a higher proportion of
people from at least one social class, possibly several or all, will
end up
having a university education.
Everyone agrees that mobility tables contain data on inequality, but the
story of inequality cannot be read
directly out of the table, the data need to
be interpreted. For example, if the British government extends the
number
of places in university education and new students are recruited from all
classes by background, is the
result more or less inequality? It is exactly for
that kind of change we have two diametrically opposite
readings.
The interpretation of mobility tables that is behind the stability thesis is
based on the following logic. The
marginal distributions are seen to display
social structure, the class structure and the structure of education
in our
case. Inequality, or at least some kind of inequality, rests not on the social
structure but how people
are distributed within that structure. Therefore,
only the conditional distributions contain data on the
relevant kind of
inequality. Furthermore, some of the information contained in the
conditional distributions is
really structural data in disguise, namely the
effect in the conditional distributions that follows directly
from changes in
the marginal distributions. Therefore, the correct reading on trends in
inequality is to
identify changes in the conditional distributions after
subtracting those changes that follow directly from
changes in the
marginals. For example, if the British government extended university
places there would, by
this logic, be no effect on the relevant kind of
inequality unless one could identify changes in the
conditional distributions
beyond those that come directly as a result of that extension. The odds ratio
technique is said to be “margin-insensitive.”15
The great methodological achievement of this class research is to have
solved the mathematical riddle of how to
practically read mobility tables in
the way the relevant theory of inequality asks that they be read. The
methodology for doing that corresponds to what became the outlook on
social organization that was extracted
from the research: it is a way of
observing social conditions beneath all that is apparent for all to see on
the
surface so as to tease out what is, in the first instance, invisible in the
foundations. Consistency of
theory and methodology does not come better
than this.
Even so, I found that way of reading mobility tables mysterious and
bizarre. If we are interested in changes in
inequality, why remove from our
vision certain changes in inequality because they are changes that have a
certain cause, in this case changes in inequality which follow directly from
changes in the social structure?
Why not do the obvious thing and just read
the whole picture? That is what we proceeded to do. We interpreted
the
mobility tables by taking note of all the distributional information in them,
both in marginal and
conditional distributions in other words, and that’s
what made the difference.
Now then, if my reading, as I think, tells the true story of class inequality,
what is the story that is told
by the alternative reading? The answer to that
question has been a bit unclear and shifting. It used to be that
that particular
reading on class inequality was seen to be sufficient for general
pronouncements on
social inequality. That’s why this research was seen to
be powerful enough to allow generalizations
from immediate research
results to a general outlook on social organization and history. This was still
the
case in the first version of Social Mobility and Class Structure in
Modern Britain (1980), the book
that established the stability thesis. This
position was then, however, abandoned, including in the second
edition of
that book (1987). Class inequality now became, more modestly, a matter of
precisely class
inequality. In the next stage, following the results of Ottar
Hellevik’s and my work, and in the exchange over
those results with
Gordon Marshall and Adam Swift, that position also was abandoned in
favor of a still more
modest view according to which there are several kinds
of class inequalities and that both ways of reading
mobility tables are
correct, albeit in respect to different kinds of inequality.16 That escape I
rejected as a form of
relativism which suggests there be “one truth for me
depending on my interests and another truth for you
depending on your
interests and that nothing more can be said about inequality as such.”17 The
reason I rejected that invitation
to consensus is that I thought my colleagues
needed to make up their minds. Either they are studying inequality
in a big
way so as to be able to support a general outlook based on their research. Or
they are studying a
limited aspect of inequality, in which case they should
limit their pronouncements on stability and change to
just that thing on
which they have a carefully cultivated expertise. They can’t have it both
ways.
There appears to be something that emerges from the reading of mobility
table data through the prism
of odds ratios that seems to have something to
do with class inequality and that seems to be pretty stable in
time and space.
But what is it? This is where it gets really tricky to grasp just what evidence
the odds ratio
analysis yields.18 I suggested
that it could have something to
do with how people of different social backgrounds might be “treated”
along the
way from class origins to educational destinations. If so, what was
being observed was a contributing
cause of inequality, but not inequality as
such.19
No sooner had the dust settled after the Hellevik-Ringen-Marshall-Swift
skirmishes – and to us it seemed that
the opposing camp would just push on
with business as usual as if no one had questioned anything – that all
this
was revisited all over again in a summary article by Adam Swift in which
he wanted to set the record
straight on just what “odds ratios does and does
not tell us.” In this article he underlines the “limitations”
in this kind of
class analysis and how it is a “narrower, or more carefully specified,
research programme than
some had realised.”20 In the end,
the odds ratio
reading does not appear to tell us much at all of any normative significance
about inequality:
“everybody can be getting better of, there can be more
chances of upward mobility, the gaps between the
positions that members
of the…groups tend to end up in can be getting smaller, the distribution of
opportunities to achieve absolute levels of goods can be getting more equal.
More equal also can be both the
distribution of opportunities to achieve [a
specified standard of living] and the distribution of opportunities
to buy
goods. All this can happen without any increase in social fluidity between
class positions.’21 That, as I read it, is exactly what
Ottar Hellevik and I had
been arguing all along. It now turned out, according to Swift, that the
program of the
Nuffield sociologists was never about inequality in a
normative meaning at all, or at least not primarily so,
it was always “an
explanatory project.” That, again, was exactly our argument. We did not
claim that there was
noting in the alternative reading – I called it
“treatment” – only that it was difficult to understand just
what it was and
that in any event it was not something that could tell us very much about
inequality as such or
in a normative meaning. With Swift’s reassessment,
therefore, it would seem that agreement was established.
 From this, Adam Swift seeks again to explain what that elusive
something is that emerges from the odds
ratio readings: “I have attempted to
justify the normative significance of the findings of mobility
analyses…
involving odds ratios. But I’ve not sought to conceal the complexities
involved in such a
justification.…To those who care about equality, but who
also care about the distribution of things like
opportunities to buy goods,
social fluidity is an important part of the story. But it is only a part. A
complete analysis of the distribution of those opportunities would involve
reference to the size of
the gaps between class positions as well as the
extent of movement between them.”22 In other words, the “complete
analysis”
would be the kind of analysis Ottar Hellevik and I had conducted
which consists of reading all the
distributional data in the mobility table.
For the rest, the “important part of the story” is that odds ratio
readings tell
us something about some differences in some kinds of opportunity. Not
about social inequality as such, not even about class inequality as such, not
even about class inequality of
opportunity as such, but about one “carefully
specified” form of relative class inequality of opportunity. If
cutting that
narrow slice from the whole pie of inequality is “the main achievement to
date of class analysis,”
so be it.
My concern in this debate was never really about class analysis, but
always about the stability thesis. That is
not a thesis about “relative class
inequality of opportunity” at all, but about recent social history, public
policy, and the welfare state. It represents an outlook. That thesis was used,
when it was established, as
evidence of the futility of “seeking to attack
social inequalities via legislative and administrative measures
of a
piecemeal kind,” of the welfare state in other words. That quote is from the
first edition of Social
Mobility and Class Structure in Modern Britain.23
And it is used today as evidence against the liberal view that modernization
comes
with progress and increasing openness in society. If we go to another
summary of the research program, that of
John Goldthorpe in On Sociology,
what is “the main achievement of class analysis to date” is not the
ability to
measure a carefully specified kind of class inequality but the boast of
having successfully
dismissed liberal theory.24
What we now know following the Hellevik-Ringen-Swift agreement is
that if anyone asks us about trends in social
inequality or in class inequality,
our answer cannot be “temporal constancy” or for that matter “cross-
national
communality.” There may well be a tiny subplot of constancy and
communality in the bigger story but if so it is
of very little significance for
real-world matters. The truth about class inequality – like the truth about
income inequality and the truth about poverty – is that changes in social
structure which we often associate
with progress in our part of the world
such as economic growth or the expansion of access to higher education,
do
in fact tend to filter down to less inequality in the distribution of well-being
and opportunity. The
liberal theory stands.
Again, getting these facts right should by my programme of scientific
deliberation be the end of my job. But
Britain is suffering political
stalemate in its predicament over higher education. Were I again to
disregard my
own rules and presume to tell the British what they should do,
I would say: open up higher education to more
young people! There are
many reasons for doing that. One good reason is that the best available
sociological
evidence suggests that the extension of university level
education to more young people is likely to follow
through to less
inequality in educational opportunity and attainment.

SCIENTIFIC DELIBERATION AGAIN

We social scientists are justified in defending our business as a scientific

one. It is of course not easy to
live up to the ideal of working scientifically,
but as scientists we must start from a commitment to a
scientific orientation.
What that means is that we work from an assumption that there is an
identifiable reality in society that is
there to be uncovered and that we aspire
as best as is possible to describe and measure and explain that
reality. It is
not enough to think about it or to hypothesize about it or to pronounce on it
with the help of
even the most elaborate theory. We need to go out there and
investigate. A commitment to a scientific
orientation is to be determined
that facts trump discourse, always, and that discourse does not trump facts,
ever.
In the social sciences, we must face the difficulty that our research is
intimately woven into the societies we
live in. That makes it extra difficult
to honor the ideal of working scientifically; it presents us with a huge
temptation to let discourse trump facts. This is a dilemma not least in
democratic societies where we want
irrelevant authority – say the authority
of money or of fame or of title or of science – to have no sway in
politics.
The answer to that dilemma, I have suggested, is not for the social scientist
to withdraw from
society but to engage with it by balancing assertiveness
and restraint in the service of scientific
deliberation.
I have told some stories about variable success in social science
enterprises, stories about myths and truths.
There are lessons here about
what creates myths and what it takes to expose them. One elementary and
infallible
way of creating myths is to allow discourse to trump facts. That’s
what has kept alive the myth of the nuclear
family. Against such myths the
obvious medicine is to do what should have been done in the first place, to
go
out there and investigate.
But myths can arise not only because discourse is allowed to trump facts
but also in spite of a firm commitment
that facts shall trump discourse. That
happens obviously when we get the facts wrong or they get muddled or
distorted. That’s what I think happened in the class analysis that persuaded
itself of the truth of the
stability thesis.
It’s odd how that could have happened. I think it is right to describe this
as a case of a grand research
effort of supreme quality and a firm dedication
to a scientific orientation that somehow lost its way. Having
sat in on all
those Nuffield seminars, in which “limitations” were never the order of the
day, I don’t for a
moment believe my colleagues would be content to see
their program defended as a “narrow” one. After much
reflection, I think
what went wrong was on the side of methodology. Not necessarily in the
choice of
methodology – although possibly that too; just what odds ratios
does and does not tell us is still not clear –
but above all in the conviction
that there is one and only one methodology that can tell the story truthfully
and that the story that is revealed by that methodology is the true one. It is
usually unwise to put all one’s
eggs in one basket.
I do not think a scientific orientation means that research is scientific in
proportion to the apparent
sophistication of its methodology. I’m all for the
use of the most advanced research techniques, including
statistical
techniques (as I hope is in evidence in this book). But science is not
methodology; it is about
answering questions. Methods are mere tools. With
tools it must be that the right ones are the right ones for
the job at hand.
This leads me to two recommendations on methodology. First, don’t put
all your eggs in one basket. Don’t trust
any single methodology. if you think
you have proven the truthfulness of something by testing it in one way,
you
should try to test your test by doing it in another way. A truth that
evaporates when looked at from a
different angle is probably a myth.
Second, don’t choose your methodology by how fancy it is. Choose it
according to the job you want it to do. The most scientific way of working
is probably to always use the
simplest method that is capable of solving the
problem in front of you.
Oxford, January 2005

NOTES
1 But in my own university that was only formally accepted when, in a
reorganization in 2000, what had been the faculty of social studies became
the division of social
sciences.
2 This idea of restraint I had applied previously in an analysis of good

government (in Ringen 1987) and applied again later in an analysis of

freedom (in Ringen 2005).
3 Gutman and Thompson 2004
4 For an overview of the state of knowledge and ignorance about

families, see,
e.g., Mason et al. 2003.
5 This is the true story: “The delusion about marriageable age is,
perhaps, the
most conspicuous of all the errors we seem to want to make
about everyday life in the old world, but is not the
only one. More far
reaching in its consequences for the view we take of ourselves, and the
plans we make for our
society and its welfare, is the supposition that the
family group in the pre-industrial world was large. It is
erroneously
believed that it was large because it contained whole groups of kinsfolk
living together… In any
case, and this seems to be the most deep-seated and
important generalisation of all, in the familial, patriarchal
world the family
was seen as the source of welfare. Sickness, unemployment, bereavement
were all the
responsibility of the family, and so to a large extent was
education. To fulfil all these functions the family in
the old world would
have to be large. Something like this seems to be the general impression of
family life among
our ancestors. But their families were not large, at least
the average family was relatively small. In
fact the evidence we now have
suggests that household size was remarkably constant in England at all
times from
the late sixteenth until the early twentieth century.” (Laslett
1971, p. 93).
 6 The Possibility of Politics was first published in 1987 and is to be
republished with a new introduction in 2006 by Transaction Publishers.
7 Another case, regrettably, of economics, with its commitment to

working
scientifically, crowding out sociology, with its sometimes
orientation to “discourse,” even on sociology’s home
turf.
8 Goldthorpe et al. 1987, p. 328.
9 Erikson and Goldthorpe 1992, Marshall et al. 1997.
10 Goldthorpe 2000. p. 163, 257.
11See, e.g., Shavit and Blossfeld 1993.
12 That seminar was in Nuffield College in 1992. I chaired it. Professor
Pahl, a
distinguished colleague, was treated in a shoddy manner. I have
blamed myself for allowing a seminar I chaired to
disintegrate the way it
did. The experience helped me to decide that I would try to get to the
bottom of what my
intuition told me had to be an unresolved confusion in
this branch of class analysis and that the only way to do
that would be to
penetrate the complexity of its methodological basis.
13 Hellevik 1997, later followed up in Hellevik 2002 and in further as

yet not
published work.
14 On the methodology, see Erikson and Goldthorpe, 1992, ch. 2.
15 The concept of ‘margin-insensitivity’ turns out to be rather mysterious

both
in origin and meaning, cf. Hellevik 2002.
16 See Marshall and Swift 1999, replies by Hellevik 2000 and Ringen

2000, and
counter-reply by Marshall and Swift 2000.
17 Ringen 2000, p. 84.
18 To see just how murky and complicated it is, consult Hellevik 2002.
19 In the literature, this “something” is called by a range of more or less

mysterious names, some of which are listed in chapter 4.

20 Swift 2000, p. 664.
21 Swift 2000, p. 675.
22 Swift 2000, p. 675.
23 1980, p. 252.
24 Goldthorpe 2000, in particular pp. 161ff. There is a sub-story here that

merits being told. The research program to which Goldthorpe attributes “the
main achievement to date of class
analysis” has in fact come under strong
and sustained criticism. Adam Swift in his summary article very properly
takes on that criticism and ends up salvaging the research program by
underlining its limitations and narrowness.
In On Sociology, Goldthorpe
deals with that same criticism by simply not mentioning it. Swift refers and
responds to criticism by Pahl, Sorensen, Holton, Crompton, Saunders,
Hellevik, and Ringen. None of the relevant
works are mentioned by
Goldthorpe. The introduction (to what is a collection of essays) does not
take the
opportunity to respond to this criticism and continues to generalize
from findings that according to Swift are
“narrower or more carefully
specified than some had realised” to conclusions about class inequality in
all
generality: “Despite the expansion of educational provision and
increases in general levels of attainment that
are apparent in all modern
societies, class differentials have in most cases remained rather little altered
over
many decades” (p. 23). Pahl is accorded the same respect as at his
seminar. He gets one mention, not his critical
article but a short “reply to
Goldthorpe and Marshall.”

REFERENCES

Erikson, R. and J. Goldthorpe (1992): The Constant Flux: A Study of Class

Mobility in Industrial
Societies. Oxford: Oxford University Press.
Goldthorpe, J. (2000): On Sociology. Oxford: Oxford University Press.
Goldthorpe, J. et al. (1987): Social Mobility and Class Structure in Modern
Britain (2nd edn.).
Oxford. Oxford University Press.
Gutmann, A. and D. Thompson (2004): Why Deliberative Democracy?
Princeton, N.J.: Princeton
University Press.
Hellevik, O. (2002): ‘Inequality versus association in educational
attainment research: Comment on Kivinen, Ahola
and Hedman.’ Acta
Sociologica 45 (2): 151–158.
Hellevik, O. (2000): ‘A less biased allocation mechanism.’ Acta
Sociologica 43 (1): 80–83.
Hellevik, O. (1997): ‘Class Inequality and Egalitarian Reform.’ Acta
Sociologica 40 (4):
377–397
Laslett, P. (1971): The World We Have Lost. London: Methuen.
Marshall, G. and A. Swift (2000): ‘Reply to Ringen and Hellevik.’ Acta
Sociologica 43 (1): 85.
Marshall, G. and A. Swift (1999): ‘On the Meaning and Measurement of
Inequality.’ Acta
Sociologica 43 (1): 84.
Marshall, G., A. Swift and S. Roberts 1997: Against the Odds: Social Class
and Social Justice in
Industrial Societies. Oxford: Oxford University
Press.
Mason, M., A. Skolnick and S. D. Sugarman, eds. (2003): All Our Families.
Oxford: Oxford
University Press.
Pahl, R. (1989): ‘Is the Emperor Naked? Some Questions on the Adequacy
of Sociological Theory in Urban and
Regional Research.’ International
Journal of Urban and Regional Research 13 (4): 709–720.
Ringen, S. (2005): ‘Liberty, Freedom and Real Freedom.’ Society 42,
March/April (Part I),
May/June (Part II).
Ringen, S. (2000): ‘Inequality and Its Measurement.’ Acta Sociologica 43
(1): 84.
Ringen, S. (1988): ‘Direct and Indirect Measures of Poverty.’ Journal of
Social Policy 17 (3)
351–365
Ringen, S. (1987): The Possibility of Politics. Oxford. Oxford University
Press.
Shavit, Y. and H.-P. Blossfeld, eds. (1993): Persistent Inequality: Changing
Educational Attainment in
Thirteen Countries. Boulder, Col.: Westview
Press.
Swift, A. 2000: ‘Class analysis from a normative perspective.’ British
Journal of Sociology 51
(4): 663–680.
 ACKNOWLEDGEMENTS

I started the studies that have trickled into this book when I joined the
University of Oxford in 1990. I have
been able to discuss the work in its
various stages with my students, and their sharp and critical, and often
unexpected, responses and suggestions have been helpful more than they
may know.
I wish to express my gratitude to the many colleagues who have offered
advice, and sometimes material, often
taking time to read and comment on
drafts, including Tony Atkinson, Raymond Boudon, Jerry Cohen, Roger
Crisp,
Partha Dasgupta, Maurizio Ferrera, Diego Gambetta, Anne Gauthier,
Jay Gershuny, Anne-Marie Guillemard, Erik
Jϕrgen Hansen, Anthony
Heath, Michael Hechter, Duncan Ironmonger, Stephen Jenkins, Nanak
Kakwani, Jon
Eivnd Kolberg, Robert Lane, Michael Lockwood, Claus Offe,
Nigel O’Leary, Ivo Možný, Martin Potůček, Natalie
Rogoff, Tor Rϕdseth,
Peter Saunders, Hannu Uusitalo, Mariken Vaa, Jiří Večerník, Claire
Wallace, Kari
Wærness, and my brothers Anders and Knut. The work on
class inequality behind Chapter 4 was undertaken jointly with Ottar
Hellevik, of the
University of Oslo. He found the ingeniously simple
devices for getting to the core of the terribly difficult
matter of analysing
inequality from mobility tables. The chapter here is one of several products
of that
cooperation. We owe special thanks to Anthony Heath, who, with
exceptional generosity, helped us with access to
the reanalysis of his own
data on class and education. Brendan Halpin was my research assistant
from 1993 to 1995
and undertook the empirical work for Chapters 2 and
3,
including solving many of the methodological problems
in the Appendices
to Chapter 3. Many more than I can
mention have contributed ideas and
suggestions, not least in a range of seminars I have had opportunity to give
far and wide during the process. Susan Dyson, Yvonne Painting, and Susan
Walters have been my outstanding
assistants in all matters of University
work and made it possible for me to do some research and writing and
helped me in countless ways in those efforts. Early ideas for the works
contained in Part I were presented in several seminars at the Social Policy
Research Centre of the University of New South Wales in Sydney while I
was a Visiting Fellow there in 1991. The
final editing of the manuscript was
undertaken in the beginning of 1996 while I enjoyed some peaceful weeks
as
professeur invité at the University of Paris, Panthéon-Sorbonne. This is
an opinionated book—it even
offers the opinion that social scientists should
not be opinionated—and it is therefore in place to state the
obvious that
these friends and colleagues carry no responsibility for opinions, nor for
such defects as there may
be in analyses.
Chapter 1 is a somewhat revised version of a paper
published previously
in Acta Sociologica (1995, no. 1). Chapter 5 contains material also used in a
paper in the Czech Sociological Review (1996, no.
1). Chapter 6 is, with a
few cosmetic changes, my
inaugural lecture at the University of Oxford
(delivered 27 October 1992, published separately by OUP in 1993). I
am
grateful to editors and publishers for permission to reprint and use
previously published material.
The work reported in Part I was supported by a grant from the
Economic
and Social Research Council (grant R000234427).
S.R.

Oxford

29 February 1996
PREFACE:
REASON AND RESTRAINT

Suddenly I had the absurd feeling that at last I understood Lenka Silver’s
enigmatic essence: in spite of her
daring bikini, in spite of her painted
fingernails, she was actually living at the turn of the century, at a
time
before all values had become relative.
Josef Škvorecký, Miss Silver’s Past

Liberty is at the core of dignity and quality in the lives of citizens;

democracy safeguards liberty. Democracy
also protects citizens against
destitution, such as avoidable famine.1 Democracies do not wage war
against each other. Democracies do not leave the
envi- ronment
unprotected.2 The
experience of democracy is that it delivers on
fundamental values.
There are good reasons for confidence in democracy. Its record of
performance where it has been tried shows it
to be among the supreme
achievements of human ingenuity. Where it has not been tried, the yearning
of peoples
is for its arrival. When given the possibility, populations turn
away from despotic regimes and choose
democracy. Towards the end of the
twentieth century, we experience a world-wide democratic revolution.
Yet confidence is not widespread, and there is much despond- ency. In
the established democracies, many are
complacent or indifferent, prone not
to participate in voting, eager to see the shortcomings of their polities,
and
reluctant to appreciate their strengths. In observing contemporary
democracy at work, one might suspect
there to be a conspiracy of bad
leadership and bad taste in politics and press to discredit the idea of
democracy. In the new democracies, expectations about what democracy
can do, and how
fast, are vastly unrealistic—understandably so, no doubt,
but still unrealistic, and it is already fashionable
for learned commenta- tors
in seductively subtle ways to pinprick what they now have in the place of
what they
wanted to get away from.
 Real democracies are necessarily imperfect. This is an exceed- ingly
elementary observation, but also one of
great importance for the
understanding of democracy, and that appreciation again is of great
importance for
democracy itself. If we want democracy to be without flaws,
we shall be disappointed. If we want democracy to
produce prosperity and
order over night, we shall be disillusioned. If disappointment and disillusion
are
prevailing sentiments among citizens, the vitality of democracy will be
undermined. If, instead, we remember
that democracy is predominantly
about liberty, we can appreciate its greatness in spite of imperfection.
Democracy delivers liberty, and it delivers more, but it does not do
everything and it does not do everything
at once.
Perfection is a dangerous idea: it is the idea that there is an end towards
which history is, or should be,
moving; it is an au- thoritarian idea, an idea
alien to the spirit of democracy. Demo- cracy comes in many
shapes and
forms. It is always in the making, and will never be made. Democracies
today are different from
democracies in the last century—think only of
voting rights— and democracies in the next century will be
different again.
Democracy is not about agreement, but about living with, indeed
encouraging, disagreement,
which is to say liberty. Real life democracy is
and is supposed to be dynamic, experimental, and changing, and
is always
quirky; democratic institutions shudder under the stress of shifting powers,
participants in the
struggle with power react in strange ways. If we are
quick to cultivate distaste for the squabbles of
democracy, the next station
may be to see virtue in dictatorial order. Democracies are not bigger than
people—they are not supposed to be—and as we live our lives by trial and
error, so democracy is an eternal
process of trial and error. If we see the fact
of error as an essential flaw, the next station may be to look
for alternatives
to democracy rather than for improve- ments in democracy. To live with
democracy is to live
with imper- fection, but living with imperfection is
difficult. In particular for the intellectual mind, there
is always the terrible
temptation towards perfection.
Imperfection means responsibility. Since democracy cannot be ‘a perfect
system’, it
cannot work by itself on its own devices. It works the way its
citizens make it work. Good institutions are
important, but not enough;
democracy also depends on democratically minded persons who behave
democratically.
It is a dilemma for each of us that the things we do
individually, on their own matter next to nothing, but
that we are
nevertheless members of the citizenry that matters everything. Unless
citizens are democrats, the
citizenry will not be democratic. It is the
responsibility of the citizen to under- stand democracy and hence
to support
it with critical confidence, and it is the responsibility of the citizen not to
misunderstand
democracy and to undermine it with uncritical despondency.
There is no more one can do than to support
democracy as best one can, but
doing that is not trivial. He who insists that he will only partici- pate if his
participation is worth his while in power is not demo- cratically minded and
does not behave democratically. It
is altogether fitting that each has
responsibility, but none power. The understanding of the fact of
imperfection in democracy, and the acceptance of the responsibility arising
from imperfection, I call ‘the
civic spirit’.
One of the imperfections in real democracies is that not all citizens have
the same say in public affairs, in
spite of the one-person, one- vote ideal.
Some do not have the vote and some do not have the power to make
their
voices heard. The basic form of democratic participation is to vote. The
vote gives citizens (although of
course not individually) much more power
than is often acknowledged. At elections it gives them the power to
change
their politicians. Be- tween elections it gives them the power to remind
politicians that they can be
rejected. By exercising these powers, citizens
create a mechanism for the selection of issues for political
attention and of
solutions to issues under consideration.
Children are excluded from the democratic process, for both reasons
mentioned above. Most democratic
countries have over the last decades
extended voting rights by lowering the voting age, now usually to 18. This
has enfranchised large numbers of citizens who were previously considered
too immature to be entrusted with the
responsibility of voting, but children
remain excluded.
Political parties and representatives will perforce keep the inter- ests of
voters in mind since they know they
will have to present themselves to the
voters at the next election. The interests of
children will not in the same
way force themselves to the attention of parties and candidates for political
office since children will not be voting until they are no longer children.
Obviously, there are other sources
of power than the vote, such as
organization, but children are also weak in organization and other power
resources; and in any event it is the vote that represents the ultimate
sanction in democratic politics.
Parents, of course, think about their children
when they vote, and politicians know that they do, but parents
are a
minority in the electorate and have many other interests than those of
children to think about; they
think about themselves and, for example, about
themselves as old persons in the future and about their own
elderly parents
here and now. Although children are increasingly becoming recognized as
citizens with ordinary
citizenship rights—for example in the UN
Convention on the Rights of the Child—the interests of children are
terribly
poorly represented in the electorate. If democracy is a system of equal
rights for citizens, which it
is, there is a remaining democratic deficit
caused by the exclusion of children and the non-representation of
their
interests.
In a presidential address to the Population Association of America in
1984, Samuel Preston suggested that the
power of the elderly and children’s
lack of power might cause public policies to be distorted to the
disadvantage of children.3 There is
empirical evidence in support of the
proposition that children are relatively disadvantaged in modern
industrial
societies, including in public policies. Professor Preston laid out compelling
data to that effect
in his address, including comparisons showing a growing
disparity in well-being between children and the
elderly. International
studies of income inequality and social spending show that the democratic
welfare state
is universally a good bargain for the elderly, whereas the
treatment of children is more varied.4 In the USA, children have a much
higher
probability than the population at large of being in the lowest income
strata, for example approximately a 50
per cent higher probability than the
population average of living below the official poverty line.5 Also in
Britain, according to official
statistics, children have a higher than average
probability of belonging to the lowest
income groups.6 In France, recent
analyses of social trends show a tendency after 1970, and increasingly so
after 1980, for younger generations
to be excluded from the benefits of
economic growth.7
A comparative standard of living for children on the lower side in their
society could, perhaps, be seen, at
least in an affluent society, as a reflection
of ‘natural’ family life-cycle fluc- tuations in income and
prosperity.
Children, after all, are pre- dominantly in young families, and we probably
want life-course
distributions that enable us to experience improvements as
we go through life. (Even so, it is hardly ‘natural’
that one in four children
should live below an officially specified minimum.) The study of the
standard of
living of children in this book covers a period in Britain when
the trend in the overall distribution of income
had shifted from a long
period of moving towards equality—from about 1950 to the mid-1970s—to
moving rather
dramatically towards inequality. This was also a period of
increas- ing prosperity; hence, there was no
absolute shortage but ‘only’ a
redistribution of standards of living. Increasing inequality means a rougher
struggle over resources. There is much to be learned by observing who the
winners are and who are the losers
when the social conflict intensifies.
This study shows that children, as a group, have a lower standard of
living than both adults and parents, and a
higher probability of living in
relative deprivation. The study also shows that, during the period of
increasing inequality, the standard of living of children increased less than
that of adults/parents, and that
children’s prob- ability of being in relative
deprivation increased more. Children fell behind, simply. It was
not that
children fell dramatically into desti- tution; the drama lies in what did not
happen: in a
period of increas- ing prosperity society did not see to it that
children were the first to benefit, or even
that children maintained their
share. The low standard of living of children results from families with
children living on a relatively low standard; over the period the relative
deprivation of families with
children increased. Even if it were ‘natural’ for
children and for families with children to be on the lower
side of the
distribution, there is nothing ‘natural’ about children and families raising
children being
disproportionately among the losers in the redistribution
which occurs when inequality
increases.
Are children losing out because they do not have political power? The
British experience suggests that they
are, in at least two ways. First, the
finding that children were losers when inequal- ity increased is clearly
a
power problem of some kind, although the mechanisms are no doubt
complex. When the competition intensi- fied,
children were not able to—
did not—maintain their position. Secondly, if disproportionate relative
deprivation
among children, and/or increasing relative deprivation among
children, were recog- nized and accepted to be a
problem, it would be an
eminently simple problem to solve. Government could eliminate the
problem
overnight by increasing transfers to (needy) families with children,
for which purpose all industrial
democracies have the machinery in place.
Generally, affluent democracies have instituted social secur- ity
systems
which give the elderly a standard of living roughly on a par with that of the
population
average.8 The same societies have,
however, not instituted
‘social security for children’ which would allow families to give their
children, the
second large group of dependents in society besides the
elderly, an even level standard of living, or, as seen
in the British case,
which would have protected the standard of living of children when the
conflict over
resources intensified. It could easily have been done, but it has
not been done.
Should the political power of children be strengthened by ex- tending the
vote to children? If the
question appears odd, it might be useful to
remember that the extension of the vote to new groups has been the
major
force in the development, the improvement, of democracy during the last
century or two, and that all
extensions came to be seen as obvious after the
fact although they were all before the fact regarded as
irresponsible, by
those who already had the vote. If children are disproportionately deprived,
this is evi-
dence that they are not able to hold their ground in the rough
game of democracy and that they are not
adequately protected by those in
power; there are material consequences of the democratic defi- cit. The
combination in a social group of deprivation in power and deprivation in
standard of living is a problem in
democracy, a situation that is as a matter
of principle not acceptable. An old-age bias in the electorate gives
too much
power to those whose interests are short term and in what they can get out
of
society here and now. If the old-age bias were removed and the electorate
made more representative, the
political power of those citizens who have
the greatest stake in the future of society would be much increased.
The
democratic ideal is clear: one person, one vote. If any group of citizens is to
be excluded from this
ideal, it is the case for exclusion that needs to be
made. There is hardly a case to be made for the exclusion
of children as a
matter of principle, and there are good instrumental arguments in favour of
their inclusion.
Could the vote be extended to children? Again, it might be useful to
remember that it has always been
argued against the extension of the vote
that those to whom it might be extended did not have the competence to
vote responsibly. Are children com- petent to vote? Democracies have
answered in the affirmative by lowering
the voting age to include age-
groups that were previously regarded precisely as children in respect to the
responsibility of voting. However, democracies also maintain that not all
children are competent to vote. How
far down can the age of voting be
taken? If 18, why not 17? If 17, why not 15, why not 13, why not 12? Why
not
11? No, 11 is too low; perhaps also 12 and 13? Somewhere there is a
limit, and although children mature
differently it would most likely be
determined by age. I do not know where the line should be drawn, but no
doubt it is well below 18.
Perhaps the question is not all that important. It is fully possible to
enfranchise children who are too
immature to vote for them- selves. These
children could be given the vote, but so that their vote was
administered on
their behalf by parents. This should then probably be the mother. Mothers
are, often, more
altruistic than fathers towards children; with mothers acting
as the custodians of the votes of children, the
chances would be that the
votes would in fact be used on behalf of children rather than being
expropriated by
parents for their own use. This same logic is the reason
child allowance is usually paid to the mother.
The solution is not ideal since there can be no presumption that parents
and children always have common
interests and since also mothers will be
inclined to protect their own interests, and have every right to do
so. But
real democracies are not perfect and do not require ideal solutions. If the
best is not possible, the
second best may do and be better than an original
situation that is already identified as unsatisfactory. It is
not ideal to go
beyond the principle of one person, one vote by allowing some persons
to
control several votes, but this could still be an acceptable way of extending
the vote to the remaining 25
per cent of the citizenry which would
otherwise continue to be excluded from basic democratic power and
representation.9
Reform is the political project of democracy. In democracy, it is not
acceptable to ignore avoidable problems,
but nor is it acceptable to use any
means to solve problems—ambitions in goals but hands tied in the use of
means.
Some eminent studies, in particular of class and poverty in Brit- ain, have
found social inequality and
deprivation to have remained stable during the
decades of reform after 1945 in spite of continu- ous economic
growth,
transformation of the occupational division of labour, educational
expansion, and consolidation of the
welfare state. These findings have been
taken in support of the proposition that social inequality has proved
immune
to reform.
 Other studies, which have looked to the distribution of income, have
come to totally different conclusions,
finding both differences between
nations and changes over time within nations—including Britain in the
same
period when class and poverty studies have found stability—and that
these variations reflect, at least to some
degree, reformist policy
interventions. This has been taken as evi- dence of effectiveness in reform.
One group
of studies has estab- lished a link between reform and inequality;
other studies have found independence.
Both stories could be true, but they are not. It turns out that the difference
in interpretation
results not from different realities be- ing displayed but
from different methodologies being applied. Once
methodological
difficulties are sorted out, it is the story as told in the income studies that
prevails.
There are philosophical issues underlying this methodological dispute;
there is the question of change and
stability. History can always be
described as a combination of these two elements, but some analysts are
more inclined to see change and some more inclined to see stability. I
belong
to those who see change. I believe that the affluent democracies
have undergone far-reaching and profound
change in all aspects of their
social structure in this cen- tury, that the process of change is continuing
and
accelerating towards the end of the century, and that this is a correct
analysis. I interpret the work of my
class-analysis colleagues to be of a
differ- ent disposition. They, too, see a combination of change and
stabil-
ity, but in their basic analysis they see stability as the essential element.
This interpretation is
influenced by a philosophy which tends to see change,
or signs of change, as superficial. No one, of course,
questions that the
nature of class has been changing, but it is the proposition of this literature
—as I read
it—that at its elemen- tary core Britain remains a class society
as it has been, and that in its essentials,
including inequality, class has been
stable, unchang- ing, and resistant to policy intervention.
The choice of methodology in the class studies is consistent with this
philosophical position. Class inequality
is defined and measured, at its
elementary level, in what is described as relative terms, and change
in class
inequality is, hence, measured ‘net’ of changes in class structure, which is
to say that not
everything that looks like change is change. There is a
parallel in the measure- ment of
poverty. Some analysts have taken the
relativization of the concept of poverty so far that it becomes
independent
of the level of prosperity and hence a straight function of its distribu- tion.
With this device,
it can be ‘proved’ that the extent of poverty has been
unchanging in periods of increasing prosperity and that
class inequality has
remained unchanged in spite of social mobility. Class inequality and
poverty? Unaffected
by prosperity, mobility, and social reform? If so, what
is left in the political project of democracy?
These analyses I believe to be flawed. Class inequality is not, even in its
most elementary core, a residual
independent of struc- ture, any more than
poverty can be independent of affluence. (There is, obviously, no
mystery in
finding stability when one has first controlled away what changes.) Class
inequality is a result of
how individuals position themselves in a structure
of class and opportunity, and change in class inequality is
a function of the
changing ways individuals move over the life course from origins to
destinations and
changing structures of class and opportunity. To ‘control
away’ these changing
structures is to tell only half the story. Britain is a
class society, but so is the rest of Europe and so is
North America. What is
peculiar about Britain is not the fact of class, but the preoccupation with
class (and
the symbols of class, such as manners and language).
In a powerful polemic, Raymond Boudon has turned against the
influence in contemporary intellectual life of
le relativism, which he
describes as le nihilisme moderne—the idea that there is no objec-
tivity,
that it is for each to find his own truth, that values are no more than
products of ‘culture’, that
standards of judgement are but illusions, that it’s
all relative.10 When I observe the low esteem in which democracy stands
today—which I believe
to be utterly unjustified—I cannot but share this
sentiment. The superiority of democracy over authoritarianism
is not a
matter of opinion, it is true superiority. It is true superiority because
democracy is based on
liberty and also performs with efficiency on other
fundamental values. Why then, in democratic Europe, has
there been so
much tolerance of and even fascination with authoritarianism, and why was
it so difficult to
reject authoritarian real existierender Sozialismus until after
the regimes had fallen? This, I am
sure, remains a mega-question for
European intelligentsia which we have yet to come to terms with. And we
are
not finished with the misery; to this day the flirtation with perfection
continues, as was, for example, so
shamefully displayed at the state visit of
Fidel Castro, the President of Cuba, to France in the spring of 1995
at the
end of the Mitterrand presidency. Why in the established democra- cies is it
so easy to get away with
uncritical despondency about democracy and why
is critical confidence in such short supply? Why is there not
more
confidence, more self-confidence?
The social sciences are often criticized for having little of wisdom to say
about contemporary society, at
least wisdom others are interested in
hearing. The problem of too much relativism is found not only in general
intellectual life, but also specifically in much of social research: reluctance
to offer conclusions and stand
by them, failure to distinguish between
science and opinion, and, when going to society, the inclination to
offer
opinion rather than knowledge. The propositions that poverty is unaffected
by increasing prosper- ity and
that class inequality is unaffected by upward
leaps in social mobility are wildly
outlandish but have nevertheless been
widely accepted. These propositions come from eminent research: careful
theoretical consideration, elaborate methodology, excellent data. Yet, the
best of sociological analysis got
the interpretation of con- temporary society
wrong because it let go of all firm footing and allowed itself to
slip into
total relativity. Of course, there is no final truth about well-being,
inequality, and poverty. These
are relative phenomena, and their
understanding must be relative to time, place, and circumstance. But
they
are not totally relative, and although every truth is subject to
reinterpretation, some stories are more
truthful than others.
The observation that class inequality (or poverty) is resistant to reform
has not been taken to imply that
egalitarian change is impossible, but only
that powerful policies are needed. Since inequalities in the
distribution of
goods reflect ‘the distribution of social power and advantage [they will]
change only as the
distri- bution itself changes, which can hardly be
expected to occur in a conflict-free way’.11 Here, again, the terrible
temptation towards
perfection, the temptation to forget that although state
power is necessary, there is also always in state
power a danger of abso-
lutism, the flirtation with policies which are not constrained by the
imperatives of
democracy: if there is inequality and something should be
done about it, it is no good fiddling around on the
surface with piecemeal
redistributions of the symptoms of deeper forces, one needs to shake up
those very
forces! Fortunately, that devastating interpretation rests on faulty
analysis. The conclusion to be drawn from
the evidence on class inequality
is not that reform is impossible because effective strategies are
unacceptable, but that reform is possible because unacceptable strategies
are not necessary.
It is a matter of some importance that the social sciences are able to
develop reliable methods of measurement
in order to monitor trends in
well-being and performance in government. It is the job of applied social
research to offer the citizenry truthful information about these matters so
that citizens can evaluate the
quality of their society, consider the need for
policies, and pass judgement on their representatives.
In the same way that there is a democratic method of aggregating
preferences in
society—one person, one vote—there is a demo- cratic
method for the measurement of well-being. Only if measure-
ment is
organized with the individual as the unit of analysis does each person have
the same weight in
aggregate measurement out- comes. If we compare
groups, for example city folk and country folk, this method
sees to it that
each person, irrespective of which group she belongs to, matters equally,
and when we look to
distri- butions that each person counts the same. Were
we to use, for example, the family as the unit, persons
belonging to large
families would count less than those belonging to small families, as if
voting were by one
family, one vote.
In the chapters on measurement, I try to cultivate this methodo- logical
individualism by taking
children as the individuals I want to observe and
treating children as persons in their own right—as
citizens—rather than as
attributes of their families, and I then drive the argument home by insisting
on
measuring the well-being of children by the income method, the method
of individualism par excellence.
This experiment in measurement has some
remarkable consequences. It emerges that the well-being of children,
who
have little or no income of their own, can indeed be measured through the
income method. This I was greatly
encouraged to find in em- pirical results
since I argue theoretically that the income method has more merit
than it is
often granted.
But, more importantly, by pulling the measurement in the direc- tion of
individualism, the
concept of well-being is pushed in a ‘social’ direction.
The main finding, I believe, in this study of
chil- dren, is of the power and
importance of the family as an institution in the production of well-being. It
is sometimes believed that the family has become a marginal institution in
advanced economies or has been
reduced to a setting in which people seek
the satisfaction of mainly emotional needs. I have looked to what the
family
does in respect to material well-being only. This is not the most impor- tant
aspect of family life, but
even in this limited perspective, the family turns
out to be of paramount importance. The family, in a modern
society like
Britain, produces much more than emotional well-being. Each family is a
potent little factory for
the production of material well-being. I try to
quantify the value added in the family. That is all but an
objective exercise,
but this is a reasonable approximation on reasonable assumptions: if we do
the thought
experiment that the family vanished and we all lived on our
own individually, with
the same volume of goods that are presently
available in the market, we would sacrifice about a half of the
consumption
well-being we now enjoy. This ‘social’ result is the outcome of strict
individualism in the method
of measurement. I take this as a strong
argument against the criticism of methodologi- cal individualism that
it is
necessarily a reflection of some form of anomic philosophical
individualism.
There is here, obviously, a substantive result and not only a
methodological one. The family matters, it
matters a lot, it pro- duces
copious well-being. The economic importance of the family is increasingly
recognized in the study of development.12 The family also remains an
economic force in developed economies—half of the national
economy is
inside the nation’s families. With weaker families, individuals suffer and the
economy suffers. Any
rational society would go to lengths to protect,
strengthen, and support families.
Well-being is subjective; ultimately only you know what is good for you,
and I for me, and it is not
for me to say what is good for you. Does it follow
that the measurement of well-being should be about
happiness? I am
concerned with information for the political process, with signposts for
political judgement
and political action; for that purpose I do not
recommend ‘subjectivism’ in the measure- ment of well-being,
but argue
‘objectivism’. It counts that I am inclined to see happiness as elusive and, at
best, difficult to
meas- ure, but it is more important that I do not think
politics is or should be about happiness. Happiness is
personal, it comes
from inside and from close relations with other people; it is independent of
material
conditions and of what politics can influence. People are not so
simple-minded as to perpetually run around
after ‘utility’. Happiness is not
a right. The authors of the American Declaration of Independence (i.e.
Thomas
Jefferson, then at the age of 33!) understood this in their oh, so
impressive wisdom. The ‘inalienable rights’
they listed were life, liberty,
and the pursuit of happiness— life? yes; liberty? yes; happiness? no!—the
pursuit of happiness. Politicians, however democratically accountable,
should not even start to think
of themselves as responsible for the happiness
of citizens, and citizens should not contemplate surrendering
the
responsibility for their own happiness. Yet again, the terrible tempta- tion
towards perfection, the temptation to see on the horizon per- fect politics in
perfect societies where perfect
people live perfect lives.13 Democracy
is
about ordinary people living ordinary lives, and ordinary people living
ordinary lives are sometimes happy
and sometimes miserable, and it is for
each to sum up these experiences in the quiet contemplation of old age.
The social sciences are notoriously lacking in firm conclusions. Absolute
conclusions are no more possible in
the social sciences than in other
sciences, if anything less since experimentation is more difficult. But how
relative? Many social scientists—and this is perhaps the case in particular
among sociologists—make a virtue of
qualifying to destruction any
conclusion and seek refuge in the pretence that the object of the whole
business
is to raise questions more than to offer answers. In this book, I
carefully—and on occa- sion not so
carefully—pull in the other direction. I
suggest that it is possible to measure well-being ‘objectively’
without doing
violence to ‘subjective’ preferences. I suggest that equivalent income,
hitherto used exclusively
as an instrument of comparison, can on reasonable
assumptions be given an absolute interpretation. I sug- gest
that the value of
production and co-operation in the family can be estimated in monetary
terms. I suggest that
comparisons of standards of living which ignore family
economies are wrong. I suggest that there is a correct
method for the
measurement of class inequality, and that once we get the measurement
right, we can correctly
evaluate the relationship between policy reform and
that last bastion of social inequality, that between the
classes. I suggest that
we in the social sciences are not doomed to sink ever deeper into the
oblivion of
relativism.
1 Sen 1984.
2 On democracy, war, and environment, see Gleditsch 1995; Gleditsch

and Sverdrup
1995.
3 Preston 1984.
4 Palmer et al. 1988.
 5 House of Representatives 1993; in 1991, almost one in four American
chil-
dren—22 per cent—lived in poverty thus defined.
6 DSS 1994.
7 INSEE 1996.
8 Smeeding et al. 1990, although Britain is an exception to this rule.
9 I see no grave problems with extending the vote to children, except for

this:
in some societies, children make up much more than about a quarter of
the population. Even I would be reluctant
to accept a below-18 majority in
the electorate. This, of course, is not an argument against resolving the
democratic deficit arising from the exclusion of children, but it might be an
argument for further compromise in
how to arrange the political
representation of children.
10 Boudon 1995: 46–7.
11 Goldthorpe et al 1987: 329.
12 Cf e.g. Dasgupta 1993.
13 While waiting for proofs. I am reading Isaiah Berlin’s The Crooked

Timber
of Humanity and there find this (p. 18): ‘… the search for perfection
does seem to me a recipe for bloodshed
...’
PART I

Well-Being
 INTRODUCTION: HARD FACTS
ABOUT SOFT VALUES

[U]ne sociologie de l’action non confondue avec la tradition

‘militariste’.
Raymond Boudon, Dictionnaire critique de la sociologie

Welfarist approaches to well-being are attractive because they respect

preferences; they appear genuinely
liberal. However, we cannot redistribute
happiness (unless we are willing to accept brain-washing), we can only
redistribute goods. If we are egali- tarians and welfarists, we must seek to
equalize happiness. Since we can
only redistribute goods, we must then
give to Peter the goods he requires in order to be as happy as is Paul. If
Peter says he needs a Bentley to be happy, while Paul is content to walk, we
must give Peter the Bentley.
Welfarism gives those who are demanding the
power to deprive those who are modest, and is hence libertarian
rather than
liberal; the principle of due concern for others is thrown overboard.
The chapters in this part are an essay in the analysis of well-being
without the concept of utility. This works
well. I find that the problem of
preferences can be handled through preference neu- trality and without
resort to
utility. ‘Objectivism’ in the measure- ment of well-being is
sometimes said to be dictatorial because it ignores
difference in preferences,
but this rests on too broad a concept of preferences. Utility is a function not
of
goods and pref- erences, but of goods and tastes, and tastes depend not
only on preferences but also on
expectations. You may have a taste for the
high life, but you cannot claim the right to any lifestyle by
insisting that it is
your misfortune that you have expensive tastes. The rest of us are not that
potty.
I wish to recapture the excellent liberal notions of equality of opportunity
and freedom of
choice from the libertarian rascals. I see opportunity, or for
that matter
freedom, as a function of re- sources with the individual and
options in the ‘arenas’ in which individuals are
active. This brings me to a
broad concept of well-being in which social options are brought directly
into the
definition. It is not only that well-being is produced in arenas, my
well-being is lodged directly,
although not exclusively, in the options I am
offered in my environment. Well-being is social.
This understanding of well-being is applied to the analysis of children. I
open up what has been a black box in
the empirical study of well-being: the
family, and look into the meaning of within- family processes for family
members.
This study was inspired by the question of within-family in- equality.
Papers on income distribution have for
some time been making the point
that we do not know the ‘real’ distribution be- cause we assume there to be
equity within families, which everyone knows there is not, at least not
always. The study simulates the effects
of hypothetical within-family
inequities between children and parents, but these simulations reveal that it
does
not matter much for the measurement of inequality in society whether
we assume this or that distribution within
families. (It matters only for the
composition of the population below poverty lines, but then the head-count
method is a crude measure of the extent of poverty. It does not matter much
in poverty-gap measures.)
Within-family economies consist not only of distribution, but also of
production and co-operation, and these
processes turn out to be of a very
different methodological importance. Through pro- duction and co-
operation,
very considerable value is added to the consumption that reaches
family members, compared to the vol- ume of
goods the family can
purchase in the market; it roughly doubles. The twist that makes it possible
to estimate
this value added in income terms is the absolute interpretation of
the equival- ent income parameter. While
within-family distributions do not
affect measurement results much, production and co-operation change the
results dramatically. All results—levels, distributions, trends—are totally
different after within-family
production and co- operation is considered,
compared to the results when these econo- mies are ignored. These
differences are so large and so systematic that there is no other conclusion
than to see it as established that
income method comparisons of well-being
which ignore the family, for example per capita comparisons, simply get
it
wrong. This would be the case for comparisons between groups, for
example
families with children and families without children, for compari-
sons over time, even as relatively a short
period as ten years in an industrial
society like Britain, and for comparisons between socie- ties, for example
developed and underdeveloped economies. From standard international
statistics, in which the family is usually
con- sidered to be ‘outside’ of the
economy, we simply have no idea of relative standards of living in poor and
rich countries. Incorporat- ing within-family economies is notoriously
difficult and results are strongly
assumption-dependent. In comparisons
which do consider the family, the final truth may therefore remain elusive,
but for the measurement of consumption well-being, such comparisons are
more truthful than those in which the
family is simply disregarded.
 1

PERSONS

People have different preferences for arranging their lives. Prefer- ences
should be respected by researchers who
want to compare how well people
live and by politicians who want to plan for a good society. The researcher
is in
no position to decide for people how they ought to live their lives and
should not offer criteria for government
policy with a built-in dictatorial
logic.
Approaches to well-being can be welfarist or non-welfarist. In welfarist
approaches, the individual’s experience
of how well he or she lives is taken
as the criterion of well-being. This is the approach of mainstream
theoretical welfare economics and of psychological quality-of-life research.
In non-welfarist approaches,
well-being is established from the ‘objective’
circumstances within which people live rather than from their
‘subjective’
utility, satisfaction, or happiness. Most empirical research on inequality in
economics and
sociology is in this sense non-welfarist.
It is often implied that welfarist approaches respect preferences while
non-welfarist approaches assume
preference agreement. The cost of
respecting preferences is taken to be the impossibility of valid interpersonal
welfare comparison, while the benefit of ignoring preferences is the
possibility of getting on with empirical
measurement.
This gives rise to two tensions. On the theoretical level, a tension is seen
between welfarist approaches, which
are taken to have the virtue of being
non-dictatorial, and non-welfarist approaches, which are considered to lack
this virtue. The second tension is between theory and measurement; while
the non-dictatorial virtue can be
preserved in theory, it is sacrificed in
measurement, either because experienced well-being (utility) is assumed
not to be directly observable or because data are available on people’s cir-
cumstances only.
A possible solution is to stick to welfarist approachs in theory, but give
up the ambition of developing
empirical measures which are at least close
to capturing underlying well-being. In this
case, a study of income
inequality, for example, would be simply a study of income inequality and
not of the
distribution of well-being. This solution is, however, not only
theoretically shallow but has also been overtaken
by events in empirical
research. For example, when household income is re-estimated to
equivalent income, which
is now customary, we are no longer studying
physical income but well-being as measured by a highly theoretical,
adjusted income measure (see Chapter 2). A second
solution is to develop
methods for measuring experienced well-being directly. This is possible, as
demonstrated
in quality-of-life research,1 but I have
reservations about such
measures, for reasons which will be explained below. I propose a third
solution in which
the welfarist/non-welfarist tensions are resolved without
resort to measures of experienced well-being.
I argue along four lines. I start from a notion of well-being as freedom,
which I hold to be both
ethically sound and methodologi- cally simple in
the sense that only modest assumptions are needed about what
people seek
in order to live well. I introduce the con- cept of arenas between resources
and outcomes,
and offer a speci- fication of the circumstances of choice as a
function of personal resources and arena options.
I draw a distinction
between prefer- ences and expectations and with the help of that
distinction
show that welfarist approaches do not have a special quality above non-
welfarist approaches of
respecting preferences. I offer a typology of
approaches to the measurement of well-being and with the
help of this
typology show that non-welfarist measurement can be or- ganized so as to
be neutral with regard to
preferences.
I propose to demonstrate that non-welfarist/non-dictatorial approaches to
(the measurement of) well-being are
possible. The most powerful non-
welfarist approaches in the literature are in terms of goods (Rawls and
others),
resources (Dworkin and others), and more recently capabilities
(Sen). All have deficiencies. Goods approaches pay
too little attention to
preferences. Re- sources approaches pay too little attention to structural
constraints on
the individual. The capabilities approach has the potential of
overcoming these problems, but is shot through
with ambiguities which
need to be sorted out.

CHOICE AND WELL-BEING

Welfarist approaches to well-being have their ethical basis in the logic of

utility: what determines the
goodness or badness of a person’s existence is
how good or bad he or she experiences it to be. This is clearly
a powerful
position. In non-welfarist research, which tends to be mainly empirical and
less philosophical in
orien- tation, ethical assumptions or implications are
not often considered explicitly. There is, however, an
alternative ethical
principle which is at least as powerful as that of utility and which can
underpin
non-welfarist approaches, namely the logic of freedom: the good-
ness or badness of a person’s existence
depends on how free he or she is.
In Sport on 4 (BBC Radio 4), on Saturday 6 February 1993, there was an
interview with Miss F. C.
Christie who had been a fan of rugby football
since 1914 when her father had taken her to see her first game.
Since then,
she had watched a game almost every Satur- day. She explained that she
had lived alone since she
was 28 and that her interest in rugby had saved
her from turning ‘into a cab- bage’. And then she said, ‘If you
have an
interest, however silly it may appear to other people, provided it doesn’t
upset them or impinge on
their things, why not?’ In On Liberty (1859), J. S.
Mill wrote: ‘If a person possess any tolerable
amount of common sense and
experience, his own mode of laying out his existence is the best, not
because it is
best in itself, but because it is his own mode.’ These two
elegant statements contain the perspective on
freedom which I adopt here.
There is an assumption of common sense and experi- ence. There is the
conclusion
that there is no need, under normal conditions, for others to be
concerned with how individuals lay out their
existence. Indeed, since there
is no common way of living which is good or bad for all, others should
not
be concerned with how individuals lay out their existence. Applied to
measurement, we need not, indeed
should not, measure the good life, since
there is no universal good life aspiration common to all; we should
meas-
ure the circumstances within which people can lay out their own existence.
Finally, there is a
constraint, that doing your thing does not impinge on
others doing their things.
 I start from a loose assumption of rational behaviour. People are thought
of as actors who seek well-being
through the choices they make in life and
are assumed, generally, to know what is good for themselves and to try to
make the most out of their circumstances in relation to whatever it
is they
want to achieve. The central element is choice: choice in itself, not what
things one chooses
but the ability to choose. I say simply that people live
well or badly to the degree they themselves can choose
how to live, or,
conversly, to the degree to which their choices in this respect are limited or
constrained.
Attention is focused not on how people live, but on how they
can live; not on what they choose, but on
what they can choose; not on
what they do, but on what they have opportunity to do.
This perspective is related to the ‘capabilities approach’ as devel- oped
by Sen in its emphasis on
freedom.2 It is similar to that on equity
suggested
by Le Grand in which equity is defined in terms of ‘choice sets’ rather than
choice outcomes.3 It derives from a more general social theory in
which
individual action and choice are the central organizing principles, what
Boudon has called la
sociologie de l’action.4 It is a sound
perspective
politically in that it invites political concern for individual well-being but at
the same time
preserves a domain of personal authority, akin to the ‘pursuit
of happiness’ principle in the American
Declaration of Independ- ence. It is
a liberal perspective in which the whole question of how to live is left
to
people themselves, and the concern of others—be it the researcher who
wants to measure well-being or the
politician who wants to advance it—is
limited to the domain of opportunities. I here follow the warning by
Dahrendorf that as we move beyond a concept based on opportunity, we
move into a conceptualization in which the
scope of freedom is increasingly
threatened.5
The perspective is liberal, not libertarian. Choice is specified as exercised
by persons in social settings.
Freedom is a function of personal resources
and structural options. It is defined positively in terms of real
opportunities
—the range of choice available and accessible—and not negatively as only
the absence of certain
speci- fied restrictions on the individual. This
perspective is neutral with regard to government, e.g. market
interventions
and social policy, and leaves the question of whether and when government
action is good or bad
for freedom to be explored empirically.
Obviously, there are assumptions. A first one is that people want choice. I
take it to be given that
we do. There would not be much left of liberal
theory if we gave up the idea that
people want to make their own choices.
Nevertheless, it is not obvious that any extent of choice is a good thing. It
is elementary that some choices
are more important than others. Most
people would probably hold the freedom to speak out against coercion to be
more important than the freedom to choose between fifty brands of
washing-powder (which is the favourite
example in these discussions).
From this, however, it does not follow that the choice between fifty brands
of
washing-powder is unimportant or certainly not that it is negative. Trivial
choices are only trivial, they are
not oppressive.
In a utility perspective, it is arguable that there might be too much choice
if the burden and cost of
searching outweighs the benefit of having much
(or more) to search from. My perspective, however, is not
utility but
freedom, and the question here is hence whether there may be a problem of
too much choice in
respect to freedom. Sen does not dismiss this as a
possibility, but is not clear on the matter, in particular
because he does not
distinguish the freedom logic from the utility logic.6
My answer is as follows. First, in terms of freedom I do not see a
problem of too much choice. For example,
hardly anything could be more
confusing and difficult to manage than the mass of books published these
days.
This may impose transaction costs to be taken into the utility account,
but could freedom be served by limiting
the number of books published?
Clearly not. If some range of choice were to cause paralysis or extreme
waste of
time, there might be a problem, but this is not the way we organize
choice. As Simon has shown, people
‘satisfice’, i.e. they find ways of
avoiding paralysis and of saving time.7 We do not start afresh each day by
going through the whole range of available
choice; probably 99 per cent of
the choices we make each day are taken care of with little or no thought at
all
by the habits we have established through pre- vious experience.
(Generally, the element of experience is much
underestimated in rational
choice models.)
The value of choice in terms of freedom is not to go through the whole
range of choice again and again, but the
knowledge that you can go through
it if you want to. Freedom lies not so much in the act of choosing as in the
availablity of choice. Even if we in daily affairs should do all we can to
avoid
having to make choices all the time, it does not follow that we do not
want to be able to choose. Access to
health care is not a good thing because
anyone wants to consume health care, but because we want to know it is
there if it should be needed (and then never having to use it). I would not
want to live by Stalin’s dictate
even if Stalin were to dictate me to live in
the way I would myself have chosen to live. Why? Because I would
not
have had the privilege of choosing and because I could not change my
mind; I would not be free.
Second, even in the unlikely event that too much choice could in some
way be seen to limit freedom, this would
still not be a difficulty with
respect to the measurement of well-being, as will be developed below. It
would
be treated as another limitation in real opportunities to be
incorporated into the measurement of well- being
along with other
constraints and limitations.
In practice choice cannot be unlimited. There will be budget constraints.
Choice may have to be limited today
in the interest of preserving choice in
the long run, or it may have to be limited out of other concerns than
well-
being, such as environmental con- siderations. Above all, as Miss Christie
explains, each person’s freedom
of choice may have to be limited out of
concern for the security and freedom of others.
Here, I think it is important to preserve a conceptualization in which such
limitations are recorded as costs
rather than to broaden the concept of well-
being in order to perform the magic of making limitations in choice
add to
well-being. If we choose to limit our consumption today in order to
consume more tomorrow, there is a
cost today, although one we may be
perfectly happy to accept. If government imposes restrictions on the use of
motor cars for envi- ronmental reasons, there is a cost in the form of
restricted choice, although again one we
may gladly live with. If I limit my
range of choice out of concern for others, there is a sacrifice, although
possibly a reasonable one. (Restrictions on choice in one area, for example
the use of motor cars, may add
choice in another area, for example access to
undisturbed nature, in which case there may be no overall cost
when it all
adds up.) In this concept of well-being, we preserve the conceptual
sharpness which is necessary
for the analysis of dilemmas and trade-offs,
for example between well- being and environmental awareness. If we
broaden the concept and turn everything into well-being, including costs
and sacrifice, we muddle concepts which should be kept apart and blunt the
concept of well-being as an analytic
tool.
A second assumption is that people can choose, i.e. the ration- ality
assumption. This carries the
usual problems that people may not fully
understand their choice sets and often display irrational behaviour.
Differences in skills and competence do not represent a difficulty, since
these can be seen as a difference in
resources and incorporated into the
measurement; opportunities we are not aware of or which we cannot
manage
are not real accessible opportunities. Nor does the fact of irrational
behaviour represent a problem of
principle. However much choice is
stressed as basic to a liberal society, some people will be recognized as
needing more or less help with arranging their lives and some choices will
be widely held to be unacceptable,
for example violence or the misuse of
drugs or alcohol (although the dividing line between acceptable and
irrational preferences will necessarily be difficult to draw).

A TYPOLOGY OF APPROACHES

Measurement approaches can be classified by the type and amount of

information used.
Starting from the concept of choice, well-being can be measured either
with information which describes the
outcome of the choices people have
made, or information which describes the choices they can
make. I call this,
in the first case, direct measurement, and in the second case, indirect
measurement. This terminology follows the distinction by Sen between the
‘direct method’ and the ‘income
method’, and corresponds to his concepts
of capabilities and functionings and to the distinction by Dasgupta
between
‘constitu- ents’ and ‘determinants’.8
It differs from the common input-
output distinction, as we shall see, in that information on what things inputs
are put into is incorporated.
In both direct and indirect measurement, we may find that a limited
amount of information is sufficient or that
a broader range of information is
required. Measures of well-being can, hence, be classified as more or less
narrow or broad.

Direct and indirect

Choice is always exercised within limitations and constraints. These are of

two main kinds: in the resources
persons command and in the possibilities
they are offered in their socio-economic environ- ment for making
productive use of resources.
 Resources are things which can be used to produce or acquire things of
value. In a market economy,
money is the basic resource. Other resources
are skills, competence, physical strength, endur- ance, charm,
‘connections’,
and so on (and negative resources such as handicaps). The greater a
person’s resources, the
greater her range of choice. Of two persons, the one
with more money can make more choices. Of two persons
with the same
amount of money, the one who is more knowledgeable about the market can
get more out of his
means.
The socio-economic environment can be described as a set of arenas in
which people work, live, buy,
or are active in other ways, such as the
family, the consumer market, the labour market, the neighbourhood,
or the
political arena. The more arenas are open to you, the more options these
arenas offer, and the less
risk they impose (including, if relevant, the risk of
too much choice), the greater your range of choice.
Having money will do
no good if there are no commodities to be bought or if stores are closed to
you because
of racial prejudice. Even the most skilled worker can only get a
good job if there are openings in the labour
market for his particular skills.
Family formation becomes difficult in communi- ties in which HIV is
endemic.9
People make choices by using resources in arenas, for example money in
the consumer market, skills in the
labour market, shrewd- ness in the
political arena, and so on. This produces outcomes of various kinds, for
example consumption, a job, a political position, or other things people may
seek. What outcomes a person
achieves depends on her resources, the
arenas which are open to her and the options these offer, and the
productivity of her particular resources in these particular arenas.
If we use information on outcomes, we measure well-being directly. If
we use information on resources and/or
arenas, we measure well-being
indirectly.

Narrow and broad

The issue of broadness is simply a question of how much informa- tion we

require before we are willing to
trust or accept what we are offered as a
valid measure of well-being. Measures of well-being are at their
narrowest
when they are based on a single indicator. While the direct/indirect
distinction is a dicotomy,
narrowness/ broadness is a continuum.
Indirect approaches

By far the most used measure is (disposable) income, which is a single-

indicator indirect measure. This I
call the income approach.
In order to take income as a valid measure of well-being, i.e. to accept
that two persons with the same
income both have the same range of choice,
we need to accept two assumptions, the first one being that income
is the
only resource which matters for what choices people can make, and the
second one that all persons
operate in identical arenas (or markets). If these
assumptions are considered too strong, we can broaden the
measure by
including more information.
The first possibility is to include information on other resources than
income. This would eliminate the
problems caused by the first of the above
assumptions. This is the solution suggested by Titmuss in
Income
Distribution and Social Change (1962). He demonstrated a number of
biases in the income
approach and suggested instead an approach based on
‘command over resources-in-time’. This I call the
resource approach.
The second possibility is to include information on arenas in addition to
resources. This would eliminate, in
addition, the prob- lems caused by the
second assumption. This solution has been suggested, as I understand
it, by
Sen in the concept of ‘entitle- ments’ in Poverty and Famines (1981) and in
the related
concept of ‘capabilities’ in later works.10
These concepts can be
interpreted as expressing a need to take into consideration not only
resources but also
what use can be made of resources. Money is of no use if
there is nothing to buy with it; nor is the
availability of things to buy useful
if I am denied the right to buy them. Only if there are things to buy
and I
have the resources which are required to buy these things, do I have an
entitlement. Capabilities, in Sen’s terminology, express a person’s freedom
to achieve valuable functionings.
In order to function, we must have both
resources to make use of the opportunities our environment offers and
opportunities in our en- vironment for making productive use of the
resources we have. This I call the
capabilities approach.

Direct approaches

The simplest direct measure of well-being is consumption expend- iture,

which is a single-indicator direct
measure. This I call the expenditure
approach.
As with the income approach, the expenditure approach rests on two
strong assumptions, namely that what we
buy is a valid expression of what
we consume and that what we consume is a valid expression of well-being.
Again, these assumptions may easily be considered too strong, and again
the problems caused by imposing such
assumptions can be eliminated by
introducing more information.
First, we can include information on non-market consumption, such as
consumption produced within the
household,11 the con- sumption of ‘free’
services transferred from outside of the house- hold, notably from
government,12 and even the consumption of leisure.13 This would eliminate
the problems caused by the
first assumption above. This I call the
consumption approach.
Secondly, we can broaden the measure beyond consumption as we
usually understand that term by including more
information on how people
actually live, e.g. on health, leisure, social relations, and the like, thus
eliminating, in addition, the problems caused by the second assumption.
This I call the way of life
approach.

Typology

We now have a typology of approaches to the measurement of well-being

with three types of direct and three
types of indirect approaches, as
summarized in Table
1.1.

Table 1.1 A typology of approaches to the measurement of

well-being

Indirect
Narrow (1) The income approach
(2) The resource approach
Broad (3) The capabilities
approach
Direct
(4) The expenditure approach
(5) The consumption approach
(6) The way-of-life approach

PREFERENCE NEUTRALITY
The argument for welfarism is that people should have the au- tonomy to
decide what is good for themselves, and
that personal satisfaction is
therefore the best criterion of well-being. The criti- cism of non-welfarism
is
that this autonomy is violated and that a dictatorial element is introduced
by implicitly imposing a common
utility function on all persons. Neither
point is well taken.
Every person has a ‘filter’ in his or her mind through which the
‘objective’ circumstances he or she lives in
are translated into ‘sub- jective’
utility, satisfaction, or happiness. There are two elements in this filter,
on
the one hand expectations with regard to the level of consumption, and on
the other hand
preferences with regard to the composition of consumption
(be it consumption in the narrow
‘economic’ understanding or taken
broadly as way of life). If two persons have different expectations with
regard to the level of consumption, they will need different levels of income
to achieve the same level of
satisfaction. People may, however, want to live
very different lives if they have different preferences for the
com- position
of consumption without, therefore, requiring different levels of income.
Utility, then, (or
satisfaction or happiness) is a product of circumstances,
expectations, and preferences; not of circumstances
and preferences only.
The two components I here call preferences and expectations are often
treated as a single category, then called
tastes. If the issue of autonomy is
seen in terms of tastes, one encounters the problem of expensive tastes,
i.e.
that some people insist they can only be happy if they drink champagne and
drive Rolls-Royces (and, for
that matter, a problem of inexpensive tastes,
i.e. that some people will be happy even if they live in
squalor). True
welfarists will have to give in to expensive tastes. Dworkin has
argued that
this makes nonsense of welfarist approaches to well-being and that tastes
should be disregarded in
favour of a pure resource approach.14 Cohen
has
argued that expensive tastes which people have through no choice of their
own should be respected and
therefore that some semi-welfarist approach is
needed.15
I suggest that this broad category of tastes is not very meaningful. It is a
totally different matter to want a
higher level of consumption compared to
wanting a different composition of consumption. Nor do I think the
notion
of expensive tastes as chosen or not chosen has much to it (and what about
inexpensive tastes that are
not chosen?). Where tastes come from, we do not
know, but tastes are hardly things we simply choose.
 The broad category of tastes needs to be broken down into preferences
and expectations. While preferences
should indeed be treated as tastes,
expectations should be treated as demands or interests.
Within a liberal agenda, each person should have the right to choose a
composition of consumption (or a way of
life) according to his or her own
preferences. This must, however, be done within some budget limitation (or
more generally, with due concern for the rights of others to choose their
way of life). While preferences
relate to the composition of consumption
within a budget, expecta- tions relate to the size of the budget
itself. The
reason the choice of composition should be free is that my choices on how I
spend a given budget do
not (in principle) affect your budget. The reason
the choice of budget cannot be free is that the more I take
the less you can
have. The distinction between preferences and expec- tations thereby allows
for a liberal
theory in which preference autonomy is respected within the
limitation of concern for the preferences of
others, and reveals true
welfarism to be not liberal but libertarian, implying the acceptance of free
choice
without concern for the consequences for others.
Since personal satisfaction is, in part, a product of expectations, we
cannot accept satisfaction as the
criterion of well-being without also
accepting the individual’s right to determine what level of consumption
(or
budget) he should have. If we do not accept the latter, as I think we should
not, neither should we accept
the utility approach to well-being. Unless you
are willing to be blackmailed by the
rich kid who demands himself to be
rich because he says he cannot be happy without the style of life he has
been accus- tomed to, or to ignore that the inability of the deprived to
articulate their deprivation is often
part of their deprivation, you abandon
utility as the test of well-being. The rejection of utility does not
imply a
rejection of the individual’s autonomy to follow his or her own preferences.
Utility is rejected
because of the problems caused by expectations, and not
because of any difficulty with preferences.
Having now dismissed welfarism as the approach associated with
autonomy of preferences and having demonstrated
that this does not imply
the introduction of a dictatorial element, my next question is whether non-
welfarist
measurement is possible without violating preference autonomy.
The alternative to measur- ing well-being as the
product of circumstances
and preferences (which, anyway, cannot be done since utility is
‘contaminated’ by
expectations) while still respecting personal preferences,
is to use information on circumstances alone but to
organize this informa-
tion so that no priority is given to certain preferences above others. The
ability of a
measure not to violate personal autonomy with regard to the
composition of consumption, or more generally, with
regard to how people
choose to arrange their lives, I call preference neutrality.
With indirect measurement, well-being is measured by the choices people
can make. In this, no
assumption is imposed about what choices people
should make or about some choices being superior to others. In
principle,
therefore, these approaches are preference-neutral.
With direct measurement, well-being is measured by the out- comes of
the choices people have made.
This requires stronger assumptions about
preferences. In order to evaluate how well people live on the basis of
the
choices they have made, we need assumptions about how they want to live,
or should want to live, against
which how they in fact live can be compared.
For example, if we measure well-being by consumption, we give
priority to
con- sumption preferences over the preferences of misers for saving and
givers for charity. Direct
approaches, therefore, are in principle not
preference-neutral.
Although the distinction between direct and indirect measure- ment is
logical when starting from the concept of
choice, only indirect measurement
approaches are ultimately compatible with a liberal
choice perspective on
well-being.
Some qualifications are needed. In the income approach, it is implicitly
assumed that people agree about the
importance of in- come and the relative
importance of goods which can be bought relative to goods which are not
bought. There are, therefore, limitations in the preference neutrality of this
approach.
On the side of indirect measurement, preference neutrality is improved
when the approach is broadened. In the
capabilities ap- proach, there is in
principle perfect preference neutrality since the full range of choice is
described in terms of both personal resources and structural options, i.e.
actual, accessible opportunities.
In the expenditure and consumption approaches, consumption
preferences are given priority over non-consumption
preferences, but these
approaches are still neutral with regard to different pref- erences within the
range of
preferences considered. If, for example, total consumption is the
same, people who spend their money on the
town are considered no better
or worse off than people who spend their money on home and family.
On the side of direct measurement, preference neutrality is re- duced
when the approach is broadened. The
way-of-life approach can be
interpreted as a specification of the arguments in the welfare function which
is
assumed shared by all persons. However, even this is not to assume full
agreement on preferences. Agreement is
assumed on the arguments in the
welfare function but not necessar- ily on any specific combination of
arguments. If well-being is meas- ured by considering, consumption,
wealth, and social integration, there may
be a range of different but
equitable combinations. It is perhaps not a very radical assumption that
there is
in any culture a fairly broad agreement on what basic factors are
important for well-being, but there will not
be total agreement, and in multi-
cultural societies there may be much disagreement. The prefer- ences of
those
who are not part of mainstream agreement will be disregarded, and
this is perhaps what respect for personal
preferences is all about.

DISCUSSION

Utility is not a particularily good criterion of well-being since it is a

function not only of circumstances
and preferences, but also of
expectations. In the measurement of well-being, respect
for per- sonal
preferences is best sought in non-welfarist approaches which have the
quality of preference
neutrality. Although non-welfarist approaches differ
considerably on preference neutrality, the meas- urement
of well-being with
the use of information on circumstances alone, arranged so as to be neutral
with regard to
preferences, is possible. Both tensions—the theoretical one
between welfarist and non-welfarist approaches and
the one between theory
and measurement—are resolved.

The concept of well-being

The meaning of well-being belongs to the eternal questions in moral

philosophy.16 For the purpose of measurement, a simple
perspective is
sufficient based on the ethics of freedom and with choice seen as exercised
within limitations
and constraints. In the sociological literature, there has
been much controversy over the concept of
well-being, notably on
‘objective’ vs. ‘subjec- tive’ and ‘narrow’ vs. ‘broad’ concepts.17 As soon
as preferences are brought in, the
concept of well-being cannot but be
subjective. The objective/subjective distinction is neither necessary nor
useful for penetrating the meaning of well-being, and the notion of
objective well-being as a concept does
not make much sense. The issue of
objectivity relates to data. It is suggested in this chapter that
well-being can
be measured with ‘objective’ data without violating the ‘subjective’ quality
of the concept.
As to narrowness vs. broadness, this is not an issue in
definition at all but only a technical question with
regard to the information
requirement.

Interpersonal comparison

In welfare economics, interpersonal comparison of well-being is usually

considered impossible because
preferences, and hence utility, cannot be
observed directly. In non-welfarist approaches, interpersonal
comparison is
technically no problem, but such com- parisons may be rejected as
dictatorial because of
assumptions of shared preferences.
This objection against non-welfarist approaches is not generally valid. It
applies
to direct measurement approaches, in particular the broader ones,
which perform badly on preference
neutrality, but not to indirect
approaches, which do well in this respect. Two persons with the same
income
(or the same capabilities) can be said to have the same well-being
without (almost) any assumption being made
about preference agreement
because uses of income (or the use of resources in arenas) are not
considered. We
ask what choices people can make, not what choices they
have made. Under these terms,
therefore, non-dictatorial interpersonal
comparison of well- being is possible.

Redistribution

In The Economics of Welfare (1920), Pigou proposed a proof that

redistribution from the rich to the
poor was a social good on the argument
that it would increase aggregate well-being: it follows from the law
of
falling marginal utility that a certain income increment would give more
additional utility to the poor
than its utility value to the rich; hence,
redistribution would increase total utility in society. In An
Essay on the
Nature and Significance of Economic Science (1932), Robbins proved
Pigou’s proof wrong, on
the argument that the law of falling marginal utility
applies to individuals but not to com- parisons between
individuals:
although a certain income incre- ment is worth less and less utility to any
person as his income
increases, it does not follow that the same income
increment is worth less utility to a rich man than to a
poor man. We cannot
know.
Robbins was right that Pigou was wrong. Since then, welfare economics
has had nothing to say about
inequality. However, wel- fare economics went
wrong in the conclusion that was drawn from Robbins’s proof.
Instead of
turning itself into the science of welfare that is indifferent to inequality (and
also to liberty,
indeed to any issue of intrinsic value in processes, which are
always mere means towards the single goal of
utility), it should have
thrown out welfarism. Pigou was right in substance, he was only wrong on
the
theoretical concept, and Robbins was only right within the terms of the
wrong concept. It was not a knock-out
and the match should not have been
stopped.
If we substitute choice for utlity, Pigou’s proof holds. A given income
increment gives a poor man more additional choice than its choice value to
a rich man. The poor man has
virtually no freedom of choice since all his
income is absorbed in necessities. Even a very small amout of
additional
income would matter greatly; it could bring him out of the dictatorship of
necessity and into the
realm of some freedom of choice. The rich man is
leagues beyond necessity in his consumption. If he were to
give up the
small amount of income that would matter greatly to the poor man, his
scope for choice would be
hardly affected. He would have a bit less to accu-
mulate (and hence have a utilitarian reason to resist
redistribu- tion), but he
would have no less (to speak of) freedom of choice. Once we shed
welfarism,
redistribution is a social good. A society without poverty is a
better society not only because it
is charitable and does good to those who
might be poor—that too, but not only—it is a better society also
because
there is more goodness all together.

Resources and arenas

The conceptual pair of resources and arenas has very considerable power. It
offers an
operational conceptualization of the circum- stances of choice in
terms of real, available, and accessible
oppor- tunities, and thereby rescues
the ‘freedom of choice’ slogan from the narrow understanding it has
taken
on in much of the literature as little more than the absence of political
interventions in markets.
This is a conceptualization with broad usefulness,
equally applic- able to poor and rich societies and to the
situation of poor
and rich people, and for capturing the life chances of women and men, the
young and the
old, and the handicapped and the strong. It shows up the
limitation in pure resourse approaches—that
structural con- straints are
ignored—and also how that limitation can be over- come. It gives a
framework for
monitoring real opportunities without, as in goods
approaches, having to rely on information on outcomes
after people have
made their choices on how to organize their lives.

The income approach

The classical measure of well-being is income, and to a large extent the

debate over measurement approaches
is a dialogue between the income
approach and alternative approaches.
The income approach comes out fairly well in this review. This is not
because it is an exceptionally precise
measure of well-being, but because
the alternatives are not much better. First of all, it performs well in
terms of
respecting preferences. Some alterna- tives—those of direct measurement—
are possible only at the
price of sacrificing preference neutrality. The
income approach also has a number of practical advantages to
recommend
it. Income data tend to be easily available, measurement results are
unambiguous, and
interpretation is easy. Since income information makes
sense intuitively to people, the power of persuasion
is considerable. The
measure is convenient as a dependent variable in causal analysis. Contrary
to broader
approaches which combine monetized and non-monetized
information, well-being is captured in a single in- dex,
and it can always be
established for any member of a popula- tion whether his or her measured
well-being is
equal to or different from that of any other member of the
population. Income can be measured in both micro
and macro and there is a
clear logic be- tween the two.
Its convenience is an important argument in its favour and prob- ably
goes a long way towards explaining the
endurance of income as a measure
of well-being. However, because this convenience rests on strong
assumptions, the income approach is also open to criticism for a lack of
trustworthiness. The response to
this criticism is to broaden the measure by
including more information. Broader approaches are in theory more
trustworthy because assumptions are relaxed and more information used,
but there are problems in handling the
broader range of information which
give rise to new uncertainties or make the alternative approaches
difficult to
manage.

Capabilities

The concept of capabilities, as developed by Sen, has proved outstandingly

fruitful in broadening the
perspective of welfare economics and bringing
economic and sociological theories of well- being closer
together. The
capabilities approach is the most prom- ising of the alternatives to the
income approach. It
does not suffer from the limitations of resources and
goods approaches discussed above. It overcomes the
weakness of the
income approach—narrowness—without throwing out its strength—
preference neutrality. Only the problem of information overload remains.
Yet, as has been pointed out by, for example Williams and Cohen, the
approach is unclear and
ambiguous.18 This remains true even after
Sen’s
later contributions, notably his comprehensive re- examination of
inequality.19 The specification I have offered, based on the
distinction
between resources and arenas, may, I suggest, be taken as an operational
definition of the
concept of capabilities which could contribute to making
the approach clearer and more applicable for
empirical purposes.
The great attraction of the capabilities approach is that informa- tion on
persons and structures is
combined. All alternative ap- proaches in my
typology—more generally, all resources and goods approaches—use
only
information on persons. The limitation in these approaches of having to
assume that everyone lives and
acts within identical environments is
overcome in the capabilities ap- proach. This makes it possible to
consider
not only market imper- fections but also fundamental values such as
political rights and civil
liberties. These are obviously of the greatest
importance for the quality of the lives we lead, but have
proved difficult to
bring into theories of well-being, possibly because they are secondary in the
utility
perspective and elusive in resources and goods perspec- tives. In the
schema used here, such liberties—or
their absence— will be legal or
structural properties of the political arena.
 My specification, however, is apparently not one that Sen would have
chosen. Mine is a ‘range of choice’ and
‘equality of opportunity’
perspective, both of which Sen rejects in Inequality Reexamined. ‘Range of
choice’ is rejected on the argument that it is necessary to take outcomes into
consideration since these
throw light on what freedoms are in fact enjoyed.
This seems to me to be an argument about information: we
may not be able
to describe actual freedoms fully and may therefore need to use information
on achievments as
a proxy. That may be a correct observation for practical
research but is not much of a theoretical argument.
If we were able to
describe freedom fully, we would not need additional information on
achievements to throw
light on what had already been established
satisfactorily. ‘Equality of opportunity’ is rejected because that concept has
taken on a specific and restricted meaning in the literature. This,
again, may
be true as an observation, but is only a matter of terminological
convenience. If equality of
oppor- tunity is a good concept, which it is, why
not take it over and give it a proper definition now that
we have the
conceptual apparatus to do so?
To me, the capabilities approach looks very much like a range of choice
and equality of opportunity
approach.20 With the help of the conceptual
pair
of resources and arenas, I believe it can be specified properly as such in
both a practical and
theoretically sound way. The ambiguities in Sen’s own
exegesis comes from a failure to make some necessary
conceptual
distinctions, in particular be- tween personal and structural elements
(resources and options) and
also between freedom and utility, and from a
resulting overconceptualization. The clearing away of
ambiguities requires
conceptual simplification. The introduction of the arena concept, and the
distinction
between arenas and resources, does the job and makes it possible
to specify a clear-cut approach to the
measure- ment of well-being which is
non-welfarist, transcends resources and goods, and is independent of
outcomes, achievments, and functionings.

Opportunities and results

There is some dispute in the literature on well-being and inequality between

opportunity theories and result
theories. Is it enough to ascertain what
opportunities people have, or do we need to know what outcomes they
achieve after having made use of their opportunities?
 The usual objection to opportunity theories is that they disregard
differences in the structural
circumstances within which people act, such as
market imperfections. This is a valid criticism if the concept
of opportunity
is defined in terms of personal resources only. That objection can, however,
be met without
resorting to result theories and their strong assumptions
about preferences. Opportunity can be defined as a
combination of personal
resources and arena options. This definition meets the standard objection
against
opportunity theories without violating the principle of preference
neutrality.
The standard objection against opportunity theories often leads to a
recommendation of direct measurement,
and this has hence come to be
regarded as the ‘liberal’ approach. If, however, the conceptual objection is
met by improving the definition of opportu- nity rather than switching to a
result theory, the liberal
recommen- dation with regard to measurement
would be to stay with indirect approaches. To the broader concept
of
opportunity, there is a corresponding measurement solution—the
capabilities approach. This solution should
be preferable to direct
approaches, in which the standard objection against opportunity theories is
met at
the cost of imposing a dictatorial element.
1 Cf. e.g. Strack et al. 1991.
2 Sen 1992.
3 Le Grand 1991.
4 Boudon 1984.
5 Dahrendorf 1979.
6 Sen 1992: 62–4.
7 Simon 1957.
8 Sen 1981, 1992; Dasgupta 1990.
9 The resources/arenas terminology follows Coleman (1971) and

Levekårs-
undersøkelsen (1976).
10 e.g. Sen 1985, 1992.
11 Cf. e.g. Goldschmidt-Clermont 1982, 1993.
12 Cf. e.g. Gillespie 1965.
13 Cf. Becker 1965.
14 Dworkin 1981a, b.
15 Cohen 1989.
16 Cf. e.g. Griffin 1986.
17 Cf. e.g. Erikson and Uusitalo 1987.
18 Williams 1987; Cohen 1989.
19 Sen 1992.
20 As also Arneson (1989) has argued.
 2

FAMILIES

I take well-being to be defined in terms of consumption, i.e. in principle as

objectively observable, and a
person’s well-being to be equal to the value of
the consumption to which he or she has access. Consumption is
generated, in
part, in families, and the well-being of persons is hence determined, in part,
by economies within
the family.1
The problem: how to measure consumption well-being with the use of
income information, given the fact of family
economies. I am not at all
concerned with income or the distribution of income as such, only with
income
information (income data) organized (some would say
‘manipulated’) so as to reflect con- sumption well-being.

FAMILIES AND PERSONS

Family economies consist of production, co-operation, and distribution.

Families command goods; persons consume services from goods. In some
cases the good is the
service, e.g. a chocolate bar. In other cases there is a
distinction, e.g. we do not consume the house we live
in but the housing it
offers. Some goods are private and some more or less public within the
family. Goods that
are capable of being public, only become public when
several persons co-operate in their use, as they do when
they form families.
A house is a public good only if several persons live in it.
By family production I understand activities in the family through which
goods are created. The family adds goods of its own making to the goods it
can acquire in the market (or
receive through transfers). Examples are
maintenance, care, cleaning, cooking, and the like. In these
activities, goods
from the market or from transfers may enter as production inputs. Families
may be differently
inclined towards family production and differently
effective in such activities.
By family co-operation I understand the joint use by family mem- bers of
goods that are public within
the family. Through co- operation, consumption
efficiencies result. Public goods can de- liver services again
and again to
several person. In co-operating, the family exploits this possibility and can
generate additional
services without acquiring additional goods. These
efficiencies are determined by family structure (size and
composition) and
by behaviour. Other things being equal, larger families are more
consumption-efficient than
smaller families because there is more potential
for co-operation (although there may be diminishing marginal
returns
because of congestion). The more the members of a family choose to co-
operate, or the better they are
able to do so, the more consumption-efficient
they will be. If co-operation breaks down or turns into conflict,
there may be
no or negative consump- tion efficiencies.
By distribution I understand the allocation of services to family members.
Individuals may have
different needs, e.g. children com- pared to adults.
This means, for example, that two persons may
each have the same
consumption well-being from different joint volumes of available services.
Thus, the value
to a person of the services to which he or she has access is
relative to his/her needs. Differences in needs
are not specific within
families, but the struc- ture of needs within the family influences the well-
being that
can be derived from given services. Hence, in addition to
efficiency effects between goods and services, there
are needs effects
between ser- vices and well-being.
If what matters ‘at the end of the day’ is well-being to persons, small-
family populations are more ‘expensive’
than large-family populations, as
are populations with relatively more adults com- pared to populations with
relatively more children (assuming chil- dren have lower needs).

The unit problem

There are two units of analysis: the family and the individual. Analyses of
goods and well-being move back
and forth between these two types of units.
Goods are the raw materials of well-being. Goods are not con- sumed
directly by persons, but are used by and
distributed in families. Hence, in the
analysis of goods, the family is the first-order unit of analysis. (We
may in
the next round turn to the individual as the unit, for example if we want to
know the volume of goods
per person.)
Well-being is observed in the status of family members, after family
economies. In the analysis of
well-being, the individual is the first-order unit
of analysis. There are three reasons. First, empiric-
ally, the end outcome of
family economies are not known until it is known how these outcomes are
absorbed by
family members. Sec- ond, philosophically, the phenomenon of
well-being resides with individuals and with
individuals only. Groups, from
families and upwards to populations, do not have well-being in any other
mean-
ing than as some aggregate of the well-beings of the members of the
group. Third, normatively, only with the
individual as the unit of analysis (or
by weighting each family by the number of members, which is the same
thing) can each person in the population be given the same weight (count
equally) in determining the social
condi- tion. This is the democratic
approach to the measurement of well- being, akin to the one-person,
one-
vote principle in politics.
The processes of family economies between market goods and well-being
are illustrated in Figure 2.1. In single-person house- holds, production and
needs may intervene between market goods and well-being. In multi-person
households, also co-operation and
distribution apply. Generally, family
economies add value, al- though theoretically value might be lost if
families
waste rather than produce goods or fight rather than co-operate. (Various
possible feed-back processes are not considered, for example from co-
operation or distribution to
production.)

Fig. 2.1 Household economies from market goods to well-being

Implications

Some consequences for measurement now follow. In all measure- ment of

well-being, the correct procedure is to
first establish indi- vidual well-beings
and then aggregate. This is the case whether we want to know the
level of
well-being in a family, group, or popula- tion, or the degree of inequality of
well-being. In no
measure of this kind is there any issue of weighting: each
person counts equally.
Since the individual condition is established after family econo- mies, and
since we need to know individual
conditions before we can know the social
condition, we can say nothing about the social condition before we
have
carried the analysis through family economies.
In multi-person households, as long as efficiencies are positive, the value
of aggregate well-being is higher
than the market value of available goods.
The difference between the value of goods and aggregate family
well-being
is the value added through the transfor- mation of goods into services and
services into
well-being. This carries through to populations: aggregate well-
being in a popula- tion is higher than the
sum value of goods available to
families.
In comparative analyses, comparisons in terms of goods and in terms of
well-being may, and usually will, give
different results. Between societies,
for example, if family structures are different, there might be less
difference
in well-being than in goods. Over time in a given society, if there are
changes in family
structure, trends in well-being would be different from
trends in goods (eco- nomic growth).
In the measurement of poverty, there are several relevant meas- ures
which are ambiguously related depending
on family econo- mies. If persons
in poverty are defined as persons with a well-being below a specified
minimum, families in poverty would be defined as families which do not
command sufficient goods to protect
all their members against poverty in
well-being. Since poverty among per- sons can result from distributions
within families, there could be poverty among persons without poverty
among families (but not the other way).
Since efficiency and needs effects
intervene be- tween goods and well-being, the relationship between the
incidence of poverty among persons (well-being) and among families
(goods) would
depend on family structures and behaviours. A final measure
is the poverty gap, which would be defined as the
addi- tional goods needed
by families in poverty to pull poor persons out of poverty in well-being. The
poverty gap would then not be a function of the incidence of poverty among
persons and the degree of poverty
among the poor, but of these factors and
family struc- ture and behaviour.

Equality and equity

A society that values equality would seek equality of well-being between
persons. Because of different
consumption efficiencies in families (beyond
differences in size), equality thus defined would result from
some unequal
distribution of goods between families. A distribution of goods between
families which could
generate equality of well-being between persons might
be said to represent equity between families. Similarly
within families. If
family mem- bers have different needs, equality of well-being within the
family would
result from some unequal distribution of services between
persons. A distribution of services between persons
which would generate
equality of well-being could be said to represent equity between persons.
Hence,
equality of well-being in society would be achieved with equity in
the distribution of
goods between families and equity in the distribution of
services within families.

EQUIVALENT INCOME

The accepted specification of income in the analysis of well-being is

equivalent income (family income divided
by the family’s equival- ence
factor as determined by the equivalence scale applied).
Unadjusted income is a measure of goods, ‘cash income’ of the value of
goods that can be bought in the market,
and ‘full income’ of the value, in
addition, of goods produced in the family and other non-market goods.
Equivalent income is usually interpreted as a relative measure, and is then
a purely theoretical
estimate which enables compari- sons between families
of different composition or between indi- viduals
belonging to families of
different composition, in which the difference in family
composition is
allowed for. In the relative interpretation, equivalent income measures do not
necessarily ex-
press any real value; equivalization is a matter of
normalizing across families.
Equivalent income can, however, also be interpreted as an abso- lute
measure, namely of consumption
well-being, meaning the real value to the
individual of the consumption services to which he or she has access
after
family economies, given the family’s goods and relative to the market value
of those goods.
Equivalization is now a matter of estimating real values for
individuals (and subsequently families), and not
only of normalizing across
families.
 The absolute interpretation requires a method of equivalizing which
mirrors the processes in the family of
transforming goods into well-being.
Depending on composition, families require dif- ferent volumes of
additional
goods for each additional member in order to generate the same well-being
to all members. This is
the efficiency effect. With increasing family size, this
effect is usually positive in that less
additional goods are needed for each
additional member. The needs of the additional person may be higher or
lower than the needs of original family members, and this will obviously
influence the volume of additional
goods required. This is the needs effect.
Per capita adjustments of income are not valid in the measure- ment of
well-being because they recognize
neither efficiency nor needs effects in
family economies. Therefore, some scaled adjust- ment is needed. A wide
range of equivalence scales are in use in current empirical research and the
choice of equivalence scale
matters strongly for measurement results.2
Under the absolute in- terpretation of equivalent income, the range of valid
equivalence scales is narrower
than under the relative interpretation. To
enable absolute interpretation, the scale must, first, be normalized
around a
‘reference household’ in which there are no efficiency effects between goods
and services and in which
needs are ‘standard’. This might be the single
adult household. Second, in order to read the equivalent income
estimate as
expressing a real value, the scale must assign the weight of 1.0 for the
single/first adult and
weights for subsequent family members expressed as
fractions thereof. Hence, in single-adult households
equivalent income
would be equal to income (well-being equal to goods). In other
families,
equivalent income would be different from income, and this differ- ence
would reflect the value added
between goods and well-being through family
economies. Finally, to reflect both efficiency effects and needs
effects, the
scale must differentiate between both the number of family members and the
type of family member
by needs. Given the absolute interpretation of
equivalent income, there is, in other words, a ‘correct’
structure for the
equivalence scale. Scales which are not normalized around the weight of 1.0
for the
single-adult household are excluded (e.g. scales which assign the
weight of 1.0 to a couple), as are scales
based only on the number of family
members (which could recognize efficiency effects but not needs effects).
What remains would be to determine the precise weights in the scale to
reflect efficiency and needs effects.
This is unresolved in the literature
because not enough is known about real family econo- mies.3 However, with
the absolute interpretation, the
relevant al- ternatives are limited to scales
which assume different strengths in the efficiency and needs
effects, i.e. to
scales which are more or less ‘flat’ or ‘steep’ in the profile of weights. The
consequence is
that with the application of any equivalence scale with a
‘correct’ struc- ture, it would usually be
straightforward in which direction
relevant alternative specifications would pull measurement results.
A scale which satisfies the structural criteria for absolute inter- pretation is
the original ‘OECD scale’,
which uses these weights: single person/first
adult = 1.0; subsequent adults = 0.7; children = 0.5.4 The assumptions are:
no value added (and no waste)
in single- person households, differences in
needs only between adults and children, and that each additional
person
represents the same re- quirement for additional goods irrespective of the
number of family members and
the order in which additional members are
introduced. With this equivalence scale, a family of, for example,
two adults
and two children would have an equivalence factor of 2.7 and the equivalent
income would be family
income divided by 2.7. That would, then, be a
measure of the well-being of each person in the family, given
family income
and assuming equity in the distribution of services, i.e. to each in proportion
to needs (in this
case: all adults the same, all children the same but less than
the adults). A
‘flatter’ equivalence scale would give a higher equivalence
factor (e.g. 3.0 if the scale were 1.0–0.8–0.6) and
a ‘steeper’ scale a lower
factor (e.g. 2.4 if the scale were 1.0–0.6–0.4). The more consumption-
efficient the
family, the lower the equival- ence factor and the higher the
equivalent income relative to family income.
Some extensions are possible towards more sophisticated differ- entiations
in the equivalence scale within the
recommended struc- tural logic.
Differences in needs might be specified more carefully, e.g. between
categories
of adults and categories of children. In this case, one would have
to define more carefully the reference
house- hold, which might be, for
example, the single gainfully employed male aged 30–50. Negative
economies of
scale are possible, e.g. if the second person in a family is a
handicapped person. A more careful distinction
could be introduced between
population catego- ries on the assumption that efficiency and/or needs
effects are
socially differentiated, in which case scales with different degrees
of steepness or flatness would apply to
different categories of families, e.g.
by age, income, level of education, rural/urban resi- dence, and
obviously
between countries, and to family and non- family households. In
comparisons between countries (or
regions) price structures could be
relevant.5
If we apply the same equival- ence scale to families with the same
structure, we also assume that they have
identical markets. If markets are
different, e.g. in the relative prices of goods that are public within the
family,
there might be different scale economies. For example, if housing is cheaper,
efficiency effects would
be weaker. If in addition to family structure we
consider family behaviour, there might again be dif- ferent
efficiency and
needs effects in families with the same struc- ture, which might for instance
be the case
between one-earner and two-earner families. If behaviour
changes over time, efficiency and needs effects may
change independently
of family structure. For example, if the power of children to demand quality
goods in-
creases (new labels rather than old hand-downs), needs effects
might weaken. Eventually, things get pretty
complicated and in the end we
might need changing tailor-made equivalence scales for each and every
household.
TABLE 2.1 Goods, well-being, and
inequality
 DEFINITIONS

1Sum of household income (disposable), divided by the number of households.

2Sum of household income (disposable), divided by the number of persons.
3Equivalent income allocated to each household member, summed up across persons and divided by
the
number of persons.
4Equivalent income multiplied by the number of household members, summed up across households,
divided by the number of households.
5The difference between equivalent income and income (per person or per household), relative to
income.
6Equivalent income allocated to persons, distributed across persons.
7Equivalent income allocated to households, distributed across households.
8Income per person, distributed across persons.
9Income per household, distributed across households.
10Equivalent income multiplied by the number of household members, allocated to households,
distributed across households.

EMPIRICAL ILLUSTRATIONS

A range of estimates for Britain in 1976 and 1986 are given in Table 2.1,
using Family Expenditure Survey data, adjusted for over-time
comparability.6 Equivalization is by the OECD scale. All estimates
are based
on disposable income (1986 £s). Household production is disregarded.
Within-household distributions are assumed equit- able. Precise definitions
are given in the table.
The intention is to illustrate (1) value added, (2) the difference between
economic growth and trends in
well-being, and (3) the difference between
valid and invalid measures of inequality of well-being.

Value added

Household income gives the value of goods available to households (free

goods, taxes, savings, and
periodicity disregarded) and equival- ent income
the value of well-being to persons. In 1986, mean house-
hold income was
£9,708 (and income per person £3,799). This translates into a mean value of
well-being for
persons of £4,870 and mean household well-being of
£12,443. The difference between income and well-being is a
measure of
value added in the processes between goods and well-being. In 1986, the
value added was 28 per
cent of income. In 1976, the value added was higher,
31 per cent.
These are the results with the OECD equivalence scale. A ‘flatter’
equivalence scale, assuming less
consumption efficiency, would give less
value added; and a ‘steeper’ scale more. In 1986, the value added
according
to the two closest alternative scales (1.0–0.8–0.6 and 1.0–0.6–0.4) would be
19 and 38 per cent
respectively. Hence, although the range of valid
equivalence scales is narrowed down by the absolute
interpretation, the
sensitivity of results to the choice of equivalence scale remains strong.
A value added in the household of roughly 30 per cent is a remarkably
high figure, considering that this
reflects only the trans- formation of given
goods into well-being. In addition, there is the production of
goods in the
household (which is considered in detail in the next chapter). Estimates,
based on conservative
assumptions, suggest that household production may
typically add 50 to 60 per cent to the value of market
goods and that the
value added includ- ing production may typically be in the order of 100 per
cent of market
goods.
The reduction in value added from 1976 to 1986 reflects a move towards
less consumption efficiency in
households as a result of reduced average
household size. Change over time in the value added can be
interpreted as a
measure of the ‘cost’ to society of moving towards smaller
households (or
the ‘gain’ if household size increased). If the movement towards smaller
households were a
result of voluntary choice (young people establishing own
house- holds earlier, old persons managing longer
on their own), this might
be considered just another way of taking out increasing prosperity, in which
case
there would be a cost in consumption well-being but possibly no
welfare loss. In effect, the change may be a
combination of choice and non-
choice (ageing, divorce, lower fertility), in which case there might be a
welfare loss but possibly not identical to the cost.

Economic growth

Economic growth is usually measured by per capita income, i.e. before

household economies. The rate of growth
thus measured is, in the British
data, 27 per cent over the ten years. Our preferred measure of well-being is
by per capita equivalent income, i.e. after household economies. This
measure shows a lower rate of change,
approximately 10 per cent lower, the
same, obviously, as the reduc- tion in value added.
We could say that the value of economic growth (as convention- ally
measured) to persons has been undermined
to the effect of 10 per cent by a
weakening of aggregate efficiency effects in house- holds resulting from
the
move to smaller households. With this magnitude of effect in a society like
the British one—a typical
industrial market economy in which the economic
role of the family is often regarded as having become
marginal—and over no
more than ten years, one must assume there are all the greater effects in the
comparison
of well-being between societies with rad- ically different
household structures and/or market-household
divisions of labour, e.g.
industrial and developing countries. Since household structure is usually not
stable in time and space, meas- ures derived before household economies
would be seriously sus- pect if
intended to reflect well-being, be it in
comparisons between societies or over time within societies. (The
estimates
incorporating household production show, for the same period, a reduction in
the value of household
production of 10 to 12 per cent, and that all together
about a third of economic growth is ‘lost’ to persons
as a result of less
production and less consumption efficiency within households.)

Inequality

A valid measure of inequality in well-being is between persons, after

household economies. This is given in
specification (6) in Table 2.1.
Four additional specifications are estimated, which are all in use in the
literature, but which are not valid
as measures of inequality in well-being
according to the logic argued presently, either be- cause of
misspecification
of the unit (household rather than indi- vidual), misspecification of well-
being (unadjusted
rather than adjusted income), or both. Specification (7) is
invalid because it is based on the household as
the unit, specification (8)
because it uses unadjusted income, specification (9) for both reasons, and
specifi- cation (10) again because the household is the unit. Specification
(10) is valid as a measure of
aggregate well-being, but not for the
measurement of inequality of well-being since it distributes across
households and not persons.
The different specifications give noticeably different results. All show
inequality to have increased (which
is consistent with known trends), but
they differ on the extent of inequality and the rate of change. The
valid
specification gives a lower level of inequality but a higher rate of change.
The difference is least between the preferred specification (6) and
alternative (7) which measures the same
range of variation but over a smaller
number of units. In specification (7), the household is the unit and
each
household is weighted by one. If, instead, households were weighted by the
number of persons, we would
be back to the preferred specification. Hence,
in this case, the weight- ing or not of households may not
matter much for
measurement results.

Alternative efficiency assumptions

Since the valid alternative equivalence scales are limited to those with the
same structure but either
flatter or steeper profiles, the consequences of
applying alternative relevant equivalence scales are clear
enough. The use of
a flatter scale would in all cases reduce the difference between the results of
preferred
and alternative measures—less value added, higher rates of change
in average well-being, higher levels of
inequity, and lower rates of change in
inequity. The use of a steeper scale would have the opposite results.

CONCLUSIONS

Income is a measure of goods, not of well-being. Given the fact of within-

family economies, the use of income
data in the measure- ment of well-
being requires radical manipulations of the data from the form in which they
are available. Once the data are in the appropriate form, one is no longer
observing income or the distri-
bution of income, but only using manipulated
income information to reflect underlying well-being. There may be
advantages in avoid- ing the income terminology in the reference to well-
being and limiting its use to the
reference to goods.

Absolute interpretation

An absolute interpretation is argued of the equivalent income measure as the

value to each family member of
his or her equitable share of consumption
services, given the goods available to the family. Guidance is
suggested on
the ‘correct’ structure of equival- ence scales, when applied to the
measurement of well-being.
This narrows the range of valid alternative
scales, compared to the valid range under the more conventional
relative
interpretation of the equivalent income measure. However, the specific
weights in equivalence scales
remain unresolved and the choice of weights
continues to matter strongly for measurement results.
In empirical studies it is now common to test the sensitivity of results by
applying alternative equivalence
scales to the population under observation.
However, one and the same equivalence scale may not be ‘correct’
for all
categories of households. A different strategy, which has so far not been
much explored, could be to
test for different equivalence scales within the
population, assuming, for example, different scale
economies in elderly
families and in young families with children.

Family economies

Once the ‘black box’ of the family is opened, the economic life within
families turns out to be remarkably
complex. There is pro- duction and
distribution, and between goods and well-being value is added through the
joint use of goods, as mediated through needs. In economic parlance, the
term ‘production’ is often used for
the entire process of translating goods
into well-being. In families, however,
there are at least three distinct
processes which need to be identified separately: a process of creating
goods
(which I call production), and a process making goods public (which I call
co- operation), and a
process of distributing services from goods.
In terms of consumption well-being to persons, as much value may be
generated in the average family as the
value that flows to the family through
market activities.

Inequality in Britain

It is argued that inequality of well-being should be measured be- tween

persons, after family economies. This
specification is applied to Britain for
the period 1976 to 1986, a period of rising income inequality.
Results are
compared to the results of other specifica- tions which are considered less
valid. The preferred
specification shows the level of inequality to be lower
but the rate of rising inequality to be faster. Since
the alternative
specifications are com- mon in the literature, it is possible that the body of
literature on
income inequality in Britain may exaggerate the extent of in-
equality and underestimate the increase in
inequality if what one has in mind
is the distribution of well-being.

Economic growth

Economic growth is usually measured by per capita income. It is argued that

this measure is seriously suspect
if intended to reflect trends in well-being,
because family economies are ignored. In the case of Britain, it
is
demonstrated that the per capita measure exaggerates the rate of growth in
well-being. It is suggested
that the per capita measure would also tend to
exaggerate differences in well-being between societies, in
particular between
developed and developing economies.

Children

If policy goals are in terms of well-being but means limited to influencing

goods (for example if politicians
are reluctant to inter- vene in internal
family affairs), the analysis of policy effectiveness would have to
take in
family economies. For example, in order to determine what additional goods
(economic growth) would be needed to increase the level of well-being in
the population by a certain
fraction, it would be necessary to take the
analysis ‘back- wards’ from individual well-being to family
goods. The
additional goods necessary would then depend on the structure and
behaviour of families and the
distribution of goods between families.
If society seeks to maximize aggregate well-being, additional goods
should be allocated to the most
consumption efficient fami- lies. If society
seeks to equalize well-being, additional goods should be
allocated to poor
families. If society wants both, it could face a dilemma of contradictory
goals if poor
families were relatively consumption-inefficient.
Families with children are ‘market-inefficient’ because parents (mothers)
must devote time to children and
are less free to pursue work careers. There
is, hence, an opportunity cost of children in the form of income
foregone.
Families with children are, therefore, on average relatively poor in terms of
market goods.
Families with children are, however, comparatively
consumption-efficient be- cause they are larger and have
more members
with lower needs than non-child families. From a situation in which families
with children are
located towards the lower end of the income distribu- tion,
which is generally the situation in contemporary
industrial economies,
redistribution of income to families with children would be a recommended
policy
whether society seeks to maximize or to equalize well-being.
1 I use the term ‘family’ rather than ‘household’ for consistency with the

family
theme of the book, except in the empirical illustrations in this chapter,
in which I use household data. Much of
what is said here, would apply
equally to families and households, although there might be some
differences in
the patterns of economic co-operation between families and
non-family households.
2 Buhmann et al. 1988; Coulter et al. 1992.
3 Cf. e.g. Deaton and Muellbauer 1980; Nelson 1992.
4 OECD 1982.
5 Atkinson and Michlewright 1992.
6 Coulter et al 1994.
 3

CHILDREN

Children are vulnerable citizens; they are dependent on the good- will of
others and have little power of their own. Children are also important
citizens; they represent the future of their society. How a society treats its
children has much to say about its charitable quality and its rational
organization.
This is a study in the well-being of British children. It covers the period
from 1976 to 1986, which were years of rapidly increasing income
inequality. According to the most authoritative study of changes in the
distribution of income in Britain, the year 1977 was a turning-point
whereafter inequality increased steadily at least up to the beginning of the
1990s.1 The substantive questions in the study are about how children and
families with children fared during a period of overall increasing inequality.
It is also a study in measurement. The income approach to the
measurement of well-being is an important and much used approach which
has a number of strengths, in spite of well-known weaknesses. One
limitation in standard income analysis is that the family is treated as a black
box and that economic processes within the family are disregarded or treated
superficially. The study deliberately stays within the theoretical conventions
of standard income analysis, but moves beyond standard conventions in the
method of measurement by opening up the family black box. The
methodological questions are about variations in measurement within an
established theoretical framework.
The concept of well-being is derived from the consumption of goods.
Goods are bought in the market or produced in the family (or come through
transfers). In the translation of the goods avail- able to the family into
consumption available to family members, there is potentially both co-
operation and conflict. Through cooperation, value may be added; through
conflict, inequities may arise. The measurement of well-being should reflect
the result for individuals ‘after’ within-family production, co-operation, and
distribution.
The well-being of children depends on two processes, first how the
families in which children live fare in society, and then how children
specifically fare in their families. The study is organized accordingly; it
starts with the situation of families with children and then moves inside
families to the situation of children. Of course, there are children who do not
live in families and some of these may be among the most vulnerable of all
children, but the present study is limited to children who live in ‘normal’
family circumstances.

WELL-BEING AND MEASUREMENT

Well-being is estimated from the sum of income from sources outside of the
family (cash income) and the value added through within-family economies.

Assumptions

The consumption measured is that available to the family and to family

members. This is not necessarily the same as actual consumption. Actual
consumption may in a period be higher than income if the family is digging
into savings or drawing on loans. It may also be lower, if the family is saving
or transferring income to others. The use of potential rather than actual
consumption is appropriate on the argument that standard of living is
determined by available resources whereas specific uses of available
resources are (within certain limits) a matter of preferences which should not
influence the comparison of standard of living. As long as we are dealing
with the family as a unit, the issue of positive or negative savings is only a
matter of periodicity, which we can assume to balance out in groups of
families. Transfers to others do not represent a problem as long as they are
voluntary (unless we attach lower welfare values to giving than to
consuming, which we should not do). Involuntary transfers, however, could
cause confusion; what they add to the well-being of the receiver is lost to the
well- being of the sender but only the former is recorded in standard income
data. There is nothing to do about this except to hope that the problem is
minor and more or less absorbed within groups of families. Cash transfers
from others are included. Non-cash income from outside of the family, such
as ‘free’ services from government, is not considered. Incorporating non-
cash income represents another expansion of standard income measurement
on which a good deal of work has been done elsewhere.2
Goods available from the market are measured in cash income, and cash
income is recorded by established survey procedures.
Goods produced in the family are measured from the volume of household
work, as recorded in time-use data. In turning information on time in
household work into consumption value, we have to decide which household
activities to regard as production and what value to put on the time used in
these activities.
On activities only those are included that generate consumption which is
subject to being shared within the family, which is the within-family
equivalent of from-outside-of-family income. This covers housework proper,
but excludes activities the benefits of which are allocated directly to certain
family members. The latter is the case for child care, which is consumed
entirely by children, leisure, which is consumed by adults/parents
(disregarding child leisure, which may or may not be a meaningful concept),
and personal care, which is ‘consumed’ by the ‘producer’.
Valuation is by the housekeeper wage, on the assumption that what
families ‘earn’ by producing goods of their own is what they would have had
to pay for someone else to produce the same goods for them.3 This is
compatible with the consumption approach; it is the value of available
consumption that is measured. Paying for housework is in most cases a real
and practical option, subject to affordability, and a much used option by
families with children. The main alternative method is to use opportunity
cost. This, how- ever, is less compatible with the consumption approach
since this method measures some welfare value to the individual of his or
her housework rather than the ‘objective’ value of the consumption which is
produced. It also assumes a freedom of choice in the amount of paid work
which is hardly realistic, at least not for parents, who have to adjust their
activities to the needs of their children. The opportunity-cost method would
give higher estimated values for within-family production; the housekeeper-
wage method is, hence, the cautious approach. (A third possibility would be
to use the market value of outputs rather than the work input, e.g. the price
of meals, but this would require very different data.)
 As is common with the use of income as a measure of well-being, the
effort involved in generating income and consumption is disre- garded and
the analysis is hence not one of final welfare.
Economies of scale in consumption are assumed, arising from co-
operation so that given goods go further in consumption terms the more
persons share in their use.
What the actual economies of scale in families are, is subject to much
discussion (cf. Chapter 2). No doubt, there are different economies of scale
in different families, for example by location, structure, position in the life
cycle, level of income, and way of life. These complex relationships are
poorly known, and students of income inequality have been confined to
choosing equivalence scales to some degree by intuition.
One way out is to choose scales that have some political or administrative
authority, for example an equivalence scale which is implicit in the structure
of social security payments, but this does not really make the scale more
‘scientific’. In Chapter 2, I argue, from an absolute interpretation of
equivalized income, that there is a ‘correct’ structure for the equivalence
scale. This goes some way towards reducing the degree of arbitrariness in
the choice among theoretically possible scales. The original ‘OECD scale’
has this structure, and I therefore apply this scale in the main body of
estimates. Also, this is a scale which has been much used in the literature,
which I have used in previous work, and with which I am comfortable. It is
in the middle of the range among equivalence scales in use in the literature
in terms of the assumed strength in economies of scale and hence neither
exaggerates nor underestimates the power of such economies.4 The best
interpretation of the resulting equivalent income measure is that it gives the
value of the consumption available to each family member, given the goods
available to the family, assuming economies of scale are captured
realistically and the distribution between family members is equitable.
By measuring household production through time-use, the same
productivity is implicity assumed in household work in all families. By
using a single equivalence scale across families, the same quality of co-
operation is assumed in all families. Economies of scale are assumed to be
the same in respect to goods purchased in the market and goods produced in
household work. In these assumptions, a wide range of variation in actual
family arrangements is no doubt being overlooked.
 The remaining problem, that of within-family distribution, is not resolved
by observation since neither income nor time-use data give actual
information on such distributions. The issue of within- family distribution
turns out to be extremely complicated theoretically, as will be explained
below. Suffice it to say for now that within-family distributions are dealt
with without actual observations of such distributions.

Data

The study draws on two types of data. First, the 1976 and 1986 Family
Expenditure Surveys, using data sets which have been harmonized for
comparability.5 Similarly harmonized later FES data were not available.
These data give socio-economic information for relatively large samples of
families, including cash income. The information on within-family
economies is based on time-use data from two surveys, the 1987 Social
Change and Economic Life Survey of six British towns (and their ‘travel-to-
work’ areas), and the BBC 1974–5 time-use survey.6 These surveys give
information on the use of time in paid work and household activities.
The income definition used is disposable family income (earnings from
work and assets, plus transfers from the state, less taxes and National
Insurance contributions). All monetary quantities are expressed in December
1986 £s.
Housework time is defined as time spent by adults on core housework
(e.g. cooking, cleaning), odd jobs (e.g. gardening, maintenance, DIY), and
shopping, but excludes child care.
The income surveys and time-use surveys are not undertaken with
overlapping samples. It has therefore not been possible to match data from
these two sources directly. Instead, the most likely pattern of time-use for
each family in the FES samples has been inputed on the basis of known
patterns of time-use for similar families in the time-use surveys. In this we
are obviously limited to working with those variables which happen to be
present and identically defined in both relevant data sets. The procedure has
been to estimate a predictive model of time-use based on the time-budget
data sets, separately for the 1970s’ and the 1980s’ observations, and to use
the resulting parameter estimates to time-use for the FES families. Appendix
B explains the imputation procedure in more detail.

Samples
Since our concern is with children and family economies, and in order to
keep the study within manageable proportions, the samples for analysis have
been limited to the ‘normal’ child/family range. Both young and old families
have been excluded, namely those in which the head of household is either
19 years or younger or 50 years or older. This greatly simplifies the analysis,
in particular the over-time comparisons in which the restriction to the
narrower age-band effectively removes the problem of differences in age
structure at the two points of observation. What is lost is information about
‘non-normal’ child families, but that is beside the main concern in this
particular study. Also excluded are households with more than two adults,
households headed by married couples in which one of the spouses is absent,
and in 1986 one household which is recorded with an exceptionally high
income. The latter exclusions are made on the basis of previous experience
with these income data sets which suggests that these smaller categories in
the sample may be exceptional or non-representative so that their inclusion
might distort the analyses.
This leaves families in the age range 20 to 49 years (age of head of
household), headed by a married couple or single person, with or without
children. ‘Marriage’ includes cohabitation, ‘children’ are under 16, or
dependent and under 19, the ‘head of household’ is the male parent/partner if
present, else the female.7
A remaining ‘exceptional’ category is that of single-father families. The
number of single fathers is in most cases too small for multivariate analysis
and some results for single-father families as a group are surprising in ways
that suggest that this may not be a very meaningful or homogeneous
category sociologically. Some results for single-father families are given in
the tables, but not much will be said about these results in the analyses.

Terminology

I use the term family rather than household since non-family households are
not included in the analysis (but will use household production and head of
household). By income is understood income from sources outside of the
family (cash income). Income enters the measure of well-being as
disposable income. The sum of disposable income and goods from
household production is referred to as consumption resources. The term
standard is used for equivalized values, so that income standard is
equivalized disposable income and consumption standard is the equivalized
value of consumption resources.

Robustness of results

A study of this kind requires a range of choices to be made on

methodological procedures where there are alternative solutions which may
be valid from a theoretical point of view, for example on the concept of
household work and its valuation, on equivalence scales, and on measures of
inequality. In matters of this kind, solutions are chosen which are well
known and established in the literature and/or otherwise cautious and
conservative and which help to keep the analysis down to earth and within
common-sense boundaries.
Ideally, estimates ought to be made with alternative specifications which
are theoretically equitable, but it would not be practically possible to do this
throughout. In the underlying analyses, however, alternative solutions have
in many cases been tried out, including alternative regression models,
alternative methods for valuating household work, and alternative
equivalence scales. Some sensitivity tests are referred to briefly in footnotes
and the technical appendices. Generally, these tests have not revealed
specific problems with the results or the interpretations they offer.

FAMILIES

This section analyses the relative economic situation of families with

children, bringing the analysis forward to a comparison of family types by
consumption standard. The next section takes the analysis forward to the
economic situation of persons.

The sources of income

In 1986, mean gross income for all British households was £11,918 and
mean disposable income £9,708. Compared to 1976, gross income was up by
14.6 per cent and disposable income by 19.1 per cent (in real terms).
Households with children had a mean gross income in 1986 of £14,088 and a
mean disposable income of £11,545, up by 13.3 and 16.8 per cent
respectively.
 Tables 3.1 and 3.2 show the distribution of gross income by source and
the level of disposable income relative to gross income for the sample of
families to which the present study is limited. (All tables are in Appendix
A.) While earnings are and remain by far the most important source of
family income, there are considerable differences in the composition of
income between family types and considerable changes over time.
Families with children, compared to families without children, generally
have a larger share of their income from state benefits and a correspondingly
lower share from earnings. The share of state benefits in total income
increases when the number of children is above two. Single-female-headed
families with children are strongly dependent on state benefits and other
transfers.
The main changes to observe from 1976 to 1986 are in the relative share
of earnings by heads of households (down) and in the relative share of state
benefits (up). The relative importance of state benefits specifically for
families with children increased quite dramatically, to about double. The
main non-change to observe is in the share of family income from wife’s
earnings in couple families, which is up only marginally.
In the trade-off with the state, as far as cash transactions are concerned,
families did very well during this period, receiving more in benefits and
paying less in taxes and contributions. Also, tax levels are down, bringing
disposable income up to a higher level relative to gross income.

Paid work and household activity

Tables 3.3 and 3.4 show the distribution of time-use on average per day for
adults by the type of family and the number of children, in 1987 and 1974–5.
Couples for which time-use diaries have failed to be recorded for both have
been excluded. The observation covers 21 hours of the day, the period from
2 to 5 a.m. having been excluded for reasons of over-time comparability
since this period was left out in the 1974–5 survey. This adds up to 1,260
minutes (with a small margin of rounding-off error).
The pattern of time-use is roughly as follows, reading first the 1987
observations:
1. On average for all persons, irrespective of family type and children
(except for couples without children), men use more than twice as much
time as women in paid work, and women about twice as much as men in
housework and three times as much in child care.
 2. Married women and men (all families considered) have about the same
patterns of time-use as all women and men.
3. Single men have a pattern of time-use which differs a good deal from
that of all men: less time in paid work and child care, and more leisure.
4. Single women have a pattern of time-use which is only moderately
different from that of all women: somewhat less in paid work and
housework, the same in child care, and some more leisure. Single women
with children use less time than other women in paid work. Single women
without children have a ‘male-type’ pattern of time-use.
5. In families with children, compared to those without children, women
spend less time in paid work and more time in housework and child care;
men spend about the same or marginally less time in paid work, significantly
more time in child care, and some more in housework.
6. For women, the pattern of time-use depends strongly on whether or not
they have children: with children significantly less time is in paid work and
more in housework and notably child care. For women who have children,
the number of children matters less for the pattern of time-use.
7. For men, the pattern of time-use depends less on children, except that
men with children spend more time on child care. Time in paid work
decreases marginally and in housework increases marginally with children
and the number of children.
All this is straightforward and intuitive, except possibly for the
observation that men’s time in paid work decreases with an increasing
number of children.
In the 1974–5 observation, the pattern of time-use has many similarities to
that summarized above but there are also some changes over the period.
First, the amount of time devoted to child care has increased. This is the case
for both women and men, and irrespective of family type and the number of
children. Women continue to put in much more time in child care than do
men, but the gap has narrowed, quite significantly so for married couples
(down from the range of 4 or 5 to 1, towards 3 to 1). Second, the distribution
of time in housework has changed. Women have moved to spending
marginally less time in housework and men considerably more. The addition
for men is greater than the reduction for women, so that in the aggregate
there is an increase in the time invested in housework. Women continue to
do more housework than men, but, again, the gap has narrowed. Third,
women, including married women specifically, have moved to more time in
paid work, although not much more, and probably not at all so for the
minority of women with three or more children. Men in all categories spend
less time in paid work. The increase for women is slightly less than the
decrease for men so that in the aggregate time in paid work has decreased
marginally. The observations on time in personal care and leisure may be
uncertain (because of the missing hours) but suggest for both women and
men less time in personal care, for both women and men without children
more time in leisure, for (married) women with children the same or
marginally less time in leisure, and for men with children the same or
marginally more time in leisure.
Overall, the two main features are a shift in time-use from paid work to
housework (in spite of the shift in the opposite direction for women) and a
considerable increase in the amount of time spent in child care, including per
child. The pattern of time-use has changed a great deal for both men and
women, and hardly less for men than for women.
In recent years, there have been very considerable changes in labour force
participation, and this is often considered to be among the most significant
recent social trends. While more women, notably married women, have
entered the labour force, the rates of economic activity of men have
decreased. The participation rates of married women have increased more
than the rates for men have decreased. For example, from 1973 to 1986,
participation rates increased for married women from 55 to 66 per cent and
declined for men from 94 to 89 per cent.8 The decline in male labour force
participation is reflected in the observations here in a decline in the
contributions of male earnings to family income and a decline in men’s time
in paid work. The stronger increase in the labour force participation of
married women is, however, not reflected in a similarly strong increase in
time in paid work or in the relative contribution of wives’ earnings to family
income. While aggregate participation rates have increased, aggregate time
in paid work has decreased. In married-couple families, the relative
contribution of male earnings to gross income declined from 76 to 68 per
cent while the contribution of female earnings increased only from 17 to 20
per cent. These seemingly contradictory results may be explained by
differences in wages for women and men, the (increasing) frequency of part-
time work for women, and other factors.

Consumption resources
Tables 3.5 and 3.6 show the sources of consumption resources by disposable
income and the value of women’s and men’s housework. (Remember that
housework does not include child care.)
1. On average in all families, two-thirds of consumption resources, as that
term is defined and measured here, is from income and one-third from
household production. In other words, adding household production, even
defined as conservatively as is done in the present exercise, increases the
value of available resources by upwards of 50 per cent.
2. Housework contributes more to total consumption resources in families
with children than in families without children. The added importance of
housework in families with children is the result almost totally of additional
housework by mothers.
3. In families with children, there is virtually no difference by the number
of children in the relative contributions of income and household production.
The presence of children matters strongly but the number of children hardly
at all.
4. In single-mother families, more than 40 per cent of total consumption is
from housework, which is a significantly higher proportion than the sum of
housework by mothers and fathers in married-couple families with children.
(But remember that single-mother families have the lowest absolute level of
income and consumption.)
In the period from 1976 to 1986, there was not all that much change in
these relationships. The relative contribution from cash income increased
slightly, as did the relative contribution from men’s housework, whereas the
relative contribution from women’s housework was slightly down. The
increasing importance of cash income might be unexpected since aggregate
time in paid work went slightly down, but is probably explained by higher
levels of transfers and lower taxes.

Effects of children

The presence of children in a family influences its economic situation

through the costs of rearing children.9 This is taken into consideration in the
analysis of standards in the next section. There may, however, be an earlier
influence already in the generation of income and consumption, which is
analysed here, i.e. the opportunity cost of children.
In these processes, the presence of children might produce a range of
effects which could pull and push in different directions. While the needs of
children represent a pressure on parents to increase earnings, parents must
also devote time and attention to children and may hence have less time for
paid work and less ability to take advantage of market opportunities. These
influences may affect mothers and fathers differently and may work in
different ways during different periods of the family life-cycle. On parents’
household production, the effect of children could hardly be other than to
increase this, but as we have seen above, time devoted to housework does
not increase all that much with children and may not be sensitive at all to the
number of children. Again, the effects may be different for mothers and
fathers and in different periods of the life cycle.
Tables 3.7 and 3.8 give estimates for the aggregate outcomes of these
various influences. The tables read as follows. The figures for the non-child
categories (top left-hand cells) show the expected income and consumption
resources of ‘typical’ married couples in the relevant age-band, with no
children. The rest of the cells give the differences between these estimates
and expected disposable income and consumption resources of families with
children, other things being equal. The differences are in absolute figures
(1986 £s) and in per cent of the value for the relevant typical non-child
couple. The tabular presentation is made for reasons of interpretability, in
particular because of complex interactions between number of children, age
of youngest child, and lifecycle period, which cause the relationships that
appear in the tables to be obscured in a general regression model. The
technicalities of the regression are explained in more detail in Appendix C.
Main results, starting with the 1986 disposable income estimates (Table
3.7):
1. The effect of children on family disposable income is generally
negative, and in some categories strongly negative (henceforth referred to as
the ‘income gap’).
2. The strongest effect is found in the one-child-under-5 categories,
suggesting the greater significance in this respect of any children vs. no
children than of subsequent/older children after the first child.
3. The income forgone because of children is higher (both in absolute and
relative terms) the younger the parents.
4. The effect of additional children, after the first child, is modest, in
particular while the youngest child is still under 5, except in families with
four or more children in which there is a considerable additional income gap.
 5. In the youngest family category, the income gap is higher when the
youngest child is over 5, no doubt reflecting an effect of early family start.
6. In the two older family categories, the negative effect on family income
weakens with the age of children.
7. In the oldest household category (and just barely so in the middle
category), the effect eventually turns positive with increasing age and
number of children.
8. In no category with a child younger than 5, and in no category with a
single child, is the effect found to be positive. However, with increasing age
of head of household, increasing age of youngest child, and increasing
number of children, the effect turns from negative to positive and eventually
turns strongly positive.
The second part of Table 3.7 reproduces the same type of esti- mates for
1976. There are both similarities and differences in the results for these two
years of observation. Findings 1, 2, and 3 above also apply to the 1976
results. Finding 4 applies as to the modest effect of additional children after
the first child, but not as to the additional income gap in families with three,
four, or more children. Here, the effect of additional children is generally to
reduce the income gap, although moderately so, except in the oldest family
category. Families with four or more children do not stand out as a special
case. Finding 5 does not hold in the 1976 data: there is no trace of a
‘penalty’ for early start in the youngest family category—rather the opposite
—whereas in 1986 there is a considerable penalty. Findings 6 and 7 hold
also in 1976, but it appears that the weakening of the negative effect and the
movement towards a positive effect occurs earlier in the family life-cycle in
the 1976 than in the 1986 data. This is also reflected in a difference in
respect to finding 8 in that the effect turns positive also in the ultimate one-
child category, i.e. with head of household over 40 and the single child over
10.
The relative figures in the table give the magnitude of the effects in
relation to the expected income in the comparison categories. Comparing
1986 to 1976, it appears generally, but not without exception, that the effect
of children has moved so as to become more strongly negative in the
categories in which it is negative and less strongly positive in the categories
in which it is positive (in addition to turning positive ‘later’ in the family
life-cycle, as observed above). In the youngest category of families, the
relative income gap increased from the order of 20 to 30 per cent in 1976 to
30 to 40 per cent ten years later. In the oldest category, the negative effects
are generally higher and the positive effects generally lower in 1986 than in
1976. The picture is mixed in the middle age-group. In all categories with a
youngest child under 5, the relative income gap is either equal to or greater
in 1986 than in 1976, whereas in the categories with older children the
recorded effects move equally in both directions.10
Table 3.8 gives the results for consumption resources. The patterns in this
table are almost exactly the same as those we have observed above as to the
direction of effects, both within the agebands and with respect to life-cycle
effects between age-bands. In no single cell does the direction of the effect
change from Table 3.7 to Table 3.8 and there is no difference between the
two tables in the direction of change in the effects when moving between
age-bands. However, there is considerable difference in the magnitude of
effects. In absolute terms, the value of consumption resources is obviously
higher than disposable income. In spite of this, all negative effects are lower
in absolute terms in consumption resources than in disposable income, and
much lower in relative terms. The positive effects are (with only one single
exception) higher in absolute terms and in relative terms in some cases
higher and in some lower.
This analysis of the ‘effects’ of children suggests that something happened
to the family over the relatively short period that is covered in these
observations. In the 1976 observation, there is virtually no difference
between the gap in cash income and the gap in consumption resources. In the
over-time comparison, the gap in cash income has increased while the gap in
consumption resources has remained stable. Families have, in other words,
become more sensitive to children in their income behaviour but not in their
total consumption behaviour. While families in the most recent observation
‘lost’ a great deal of income by having children and recuper- ated some of
that loss through additional household production, families ten years earlier
lost less income but also found that loss to carry directly through to a similar
loss in consumption resources; there was no compensation in the form of
additional household production. A further change appears to be that while
the 1986 observation shows an economic ‘penalty’ for early start with
children as well as for many children (three or more), there is no trace of
such penalty effects in the 1976 observation.
These findings reflect considerable changes in the economics of child-
rearing. Families have, in a sense, become more economically ‘vulnerable’
to children. In the end, the economic sacrifice involved in having children is
about the same at the recent observation as ten years earlier, but this implies
more drastic adjustments in the behaviour of parents. In terms of the gap in
cash income, childrearing has become more expensive. Parents have,
however, met this additional expense themselves by shifting back into
household production.
The transition that is observed here, over no more than ten years, is really
quite astonishing. The present analysis does not allow a full explanation. A
range of factors may be relevant—changes in family size, the timing of
children, more single-parent families, unemployment—but it seems likely
that part of the explanation must lie in changes in the role of women, in
particular in the increasing labour-force participation of married women.
This has made families more dependent on female earnings for their total
income (although, as observed above, the contribution of wives’ earnings to
family income has increased less than has female employment). Since
women adjust more than men to children in their labour-market behaviour,
this has made family income more sensitive to children. Without children,
women have come to contribute more income. With children, there is more
income in the form of female earnings to lose. Women, moving out of the
labour market and into the family as a result of children, cause increasing
household production which compensates for decreasing income. With
families being more dependent on female earnings, an early start with
children, and a higher number of children, will be more disruptive for family
income.

Consumption standard

So far, the analysis has been in terms of income and consumption per family.
We now turn to translating these values into estimates of standard after
adjusting for family size and composition. This is done by re-estimating
absolute values per family to equivalent values per family, thus producing
measures of standard which are comparable across families of different size
and composition.
Tables 3.9 and 3.10 show these estimated standards, relative to the mean
standard for all families. The main results are, reading first the 1986
observations:
1. Families with children have a lower than average standard, families
without children a higher than average standard, and these differences are
generally quite large.
2. Single-female-headed families have a lower standard than both
married-couple families and single-male-headed families. This is the case
among both families with and without children.
3. Among families with children, relative standards are in all cases lower
the higher the number of children.
4. When measured with the consumption standard, as compared to the
income standard, the differences between family types are in the same
direction, but of much less magnitude. Families with children are still below
average, but less so; families without children are still above average, but
less so; and single-femaleheaded families are still at the bottom of the
distribution, but at a higher relative position.
In 1976, these relationships were much the same: families with children
lower than families without children, single-female- headed families lowest,
and less difference in consumption standard than in income standard.
However, the observations also show some considerable changes over the
period:
1. The relative standard of families with children is lower in 1986 than it
was ten years earlier.
2. Among married couple families, the relative standard is about the same
in 1986 as it was in 1976 for those who are without children and with one or
two children, but lower for those with three or more children.
3. In single-female-headed families, the relative standard is much lower in
1986 than it was in 1976, both for those with and without children. These
changes over time are about the same in income standard and in
consumption standard.

How much is the family worth?

Goods purchased in the market are, in a sense, the raw materials for the
family from which it generates consumption for family members. To these
raw materials, the family itself makes two additions. It produces additional
goods of its own, and family members co-operate in translating the goods
available to the family into consumption for family members.
The value of goods from outside of the family is set equal to family
disposable income. The value of household production is estimated from the
time invested in such production. The effects of co-operation is captured in
the equivalence scale, and with the absolute interpretation of equivalent
income, it is possible to estimate the value added through co-operation. Such
estimates are given in Table 3.11. The results are, reading first the 1986
estimates:
1. Disposable income per person is £3,515. Through the process of co-
operation, this is translated into an income standard per person of £4,744,
which represents a value added of 35 per cent. In other words, even if the
family produced no goods of its own, it would still, on average, add about a
third to the value of the consumption available to its members, relative to the
market value of available goods. This factor would obviously be higher in
larger families and lower in smaller families, and could be higher in
harmonious families and lower in families in conflict.11
2. Through household production, the family adds, as we have already
seen above, 56 per cent to the goods available to it from the market.
3. Taking into consideration the effects of both production and co-
operation the aggregate value added in the family is estimated to 113 per
cent on top of disposable income.12
These estimates of value added appear remarkably high. The family itself
produces more value than the value of the goods its income can purchase in
the market. Even so, the estimates are conservative. The equivalence scale is
modest and does not impute exceptional economies of scale compared to
what is conventional in the literature. Household production is valuated
conservatively by the housekeeper-wage method, and only housework
proper is included.13
The 1976 estimates are slightly different, giving higher estimates of value
added. This is a result of two factors, the increasing relative importance of
cash income (as we have seen above, in spite of reduced time in paid work)
and the transition towards smaller families. The latter has the effect of
weakening the economies of scale since available goods are shared between
fewer family members.
The change from 1976 to 1986 in income per person is akin to much-used
measures of economic growth, such as GDP per capita. This shows a rate of
‘growth’ of 31 per cent. That, however, is not a good measure of the trend in
consumption well-being since it does not take into account within-family
economies and changes in these. The results here suggest that per capita
income estimates of the standard of living strongly exaggerate the pace of
growth (in Britain during this period). The standard of living has increased,
but less so when measured after within-family economies than if measured
by per capita income.
1. Consumption resources have increased considerably less than has
income. This is a result of the declining relative importance of household
production whereby families add less goods through production of their own
relative to what they can buy in the market.
2. The increase in equivalent values is less than the increase in per capita
values. This is a result of the move towards smaller families.
The introduction of household production reduces the growth rate by 5
percentage points, and equivalizing reduces it by another 3 percentage
points. All in all, this brings the estimated increase in standard of living
down from 31 per cent when measured in per capita disposable income to 23
per cent when measured in the value of consumption available to persons
after within-family economies. This difference could be taken as a measure
of the ‘cost’ to society of the changes that have taken place over this period
in family arrangements and the economies of the family. For example, the
‘cost’ of the reduction in average family size alone is in the magnitude of 10
per cent of economic growth as conventionally measured.
There is no mystery or statistical magic in any of these results. The costs
in question are very real costs. If relatively less work is done in the family,
relatively more goods need to be purchased in the market, for example more
processed or ready-made food or more paid help. If average family size goes
down, the population will need more houses, more cars, more washing-
machines, stoves, TV sets, and so on, and will hence have less consumption
per person from given aggregate resources.
These changes in family arrangements may to a greater or lesser degree be
the result of voluntary choices, in which case the estimated ‘costs’ could be
regarded more or less as another way of taking out increasing prosperity. We
may be buying more goods in the market, and doing less housework,
because we can afford to. Families may be becoming smaller because young
persons can afford to establish families of their own earlier and elderly
persons are able to manage longer on their own. No doubt, other factors are
also in play, some of which are clearly involuntary (ageing of the
population) and some of which may be more or less involuntary (divorce,
single motherhood, the family effects of increasing female labour-force
participation).
 PERSONS

Equivalized measures of income or consumption standard can be interpreted

as the value of the consumption available to each member of a family, given
the family’s total resources, on the assumption of equitable distribution in
the family. The equity assumption has long been recognized as unrealistic.14
What are the consequences for the measurement of standard of living of
relaxing this assumption?

Within-family inequity

The family is in one respect an ordinary social system in which collective

solutions are worked out through conflict, disagreement, and negotiation,
and in another respect a special social system: being small, co-operation is
less difficult than in larger systems: being intimate, there is more than usual
sharing. Families, obviously, differ a great deal in these respects.
Within families, there may be various kinds of differences in the economic
situation of family members. Such differences are in recent literature
described mainly in terms of gender relations. There are differences in who
does what, for example in who brings in what income and who does what
housework; there are differences in who decides what, for example in the
management of money;15 and there are differences in who gets what, for
example in the distribution of food;16 other consumption;17 or income.18
While the existence of such differences inside families is now well
documented, it is all but straightforward to work out how such differences
translate into, first, differences in well-being, and, second, into differences in
well-being that are also inequities.
Information on who does what or on who decides what is not in itself of
much use for resolving questions of who gets what. There is no reason, for
example, why there could not be equity in consumption in families with only
one breadwinner or with a single person controlling the purse. These
distributions may be related—if we find consumption to be inequitable, it
would be near at hand to try to explain that inequity as a result of earlier
distributions in earning or decision-making power—but the inequity to be
explained would have to be established independently in consumption.
There is some discussion in the literature on the issue of income ‘pooling’.
One extreme assumption is no pooling, i.e. that the economic situation of
each person is determined by his or her own earnings. This assumption does
not have much power in the analysis of well-being since by necessity all
members of a family have some consumption even if they have no earnings.
The opposite assumption is that all income is pooled and used jointly. This is
the (implicit) assumption in standard analysis of equivalent income, the
assumption which is usually resorted to out of necessity or convenience
although recognized as unrealistic. There are intermediate solutions in which
some income is assumed pooled and the rest is assumed distributed.19
In the discussion of pooling, it seems useful to draw a distinction between
who decides what and who gets what. One question is how the various bits
of income that flow to the family are managed (or for that matter how the
consumption produced in the family is managed). This is of great interest in
the study of family arrangements and power relations. In the study of
consumption well-being, however, money management is only an
independent variable; it is not a factor in the function in which consumption
well-being is defined. The question in respect to housing is not whose
income pays for it or who manages the bills, but who consumes what
‘services’ from the house and its equipment.
If our concern is with the distribution of consumption, we should observe
consumption directly. For that purpose it is not necessary to go via any
assumption about pooling in the management of income or goods. Indeed,
for the analysis of well-being, the very notion of ‘the distribution of income
within the family’ is not very meaningful. Even if we know that mother
manages 80 per cent of family income as the household budget, while father
keeps 10 per cent and children 10 per cent as pocket money, we know
nothing about the distribution of consumption.
In direct observation of consumption within the family, there are a range
of measurement problems. If we use income data for the valuation of
consumption, there is the question of how to treat income that is not put into
consumption. In the introduction, this problem was dismissed in respect to
gifts on the argument that giving is just another way of spending. Once we
move inside the family, this remains true only if all family members approve
equally of such generosity. The treatment of saving may depend on what we
assume its purpose to be. If it is for security, it might reasonably be
considered to be for the benefit of all current family members, in which case
we would know at least between whom the benefit is to be distributed,
although not necessarily how. If it is for some specific future purpose, the
circle of those benefiting might be smaller, for example if parents save for
their own old age or to support their children’s future education. Under this
assumption there would also be uncertainty about how things might actually
work out in the future.
Consumption goods have very different properties in terms of what
consumption they generate. Some goods are entirely ‘private’. A £’s worth
of a chocolate bar generates a £’s worth of consumption to whomever eats
the chocolate. Other goods are ‘public’ in the sense that one person’s
consumption in it does not exclude other persons from also consuming in it.
Housing is the case in point (along with tools and equipment). If a single
person lives in a house, he or she consumes the services of that house. If a
second person moves in, he or she also consumes services from the same
house (with the only difference that each can have only slightly less than the
single person had alone because the house has now become more crowded).
The ‘multiplier effect’ we can take care of with an equivalence scale, but
that does not tell us who in fact consumes what.
If we were able to overcome these (and other) problems of data and
measurement, we might see a range of differences in actual consumption,
e.g. between children and adults and possibly between women and men or
younger and older adults. Even this, however, would not answer the question
of the distribution of wellbeing.There would then be the problem of deciding
whether these differences are also inequities. In doing that, there are some
relevant criteria in respect to the family as an ordinary social system and
some in relation to it as an extraordinary system.
The first standard criterion for translating inequality into inequity is to
control for differences in needs. This is what is attempted through
equivalizing in the comparison between families. However, there are
obviously further differences in needs within families, clearly between
children and adults, probably between adults of different age, and possibly
also by gender. Other standard criteria of equity are differences in desert
(e.g. those bringing in more income might be considered to deserve more
goods) and in rights (e.g. those carrying more responsibility might be
considered to have a right to more benefits). An equitable distribution of
consumption within the family would usually not be an equal distribution of
consumption.
What makes the family an extraordinary social system is its intimacy. This
brings in two additional criteria of equity which weigh more heavily in the
analysis of the family than in the analysis of more ordinary social systems:
those of sharing and of agreement. The meaning of sharing is that family
members benefit not only from their own consumption but also from each
other’s consumption. This is clearly the case within a utility framework, as
for example when the husband takes pride in the elegance of the wife’s dress
or the wife in the husband’s display of wealth in expensive cuff-links or tie-
pins. The same applies within a consumption framework. A mother who is
breast-feeding consumes not only for herself but also for her child. A meal
can be a shared experience in which what is consumed is not only food but
the ‘services’ of this experience. In a poor family with one breadwinner, it
may be in the interest of all family members to allocate disproportionate
food consumption to the breadwinner to strengthen his or her earning
capacity.
Related is the matter of agreement. If a family agrees on some unequal
distribution of consumption, this distribution should not be regarded as
inequitable. This criterion might apply to any social system, but certainly
applies to the family since there the problem of establishing agreement is
manageable, at least in terms of the number of persons who need to agree.
The concept of agreement is, however, difficult. For example, women might
be allocated lessconsumption than men and accept that as just because that is
the way it has always been. Is this agreement or institutionalized
discrimination? Parents may be tight with pocket money for their children,
and buy them less things than they could afford, because they believe it is
best for children not to grow up spoiled. Children might resent this, but we
ought not classify the results of responsible child-rearing, if that it be, as
discrimination and inequity. It would probably be generally accepted that
children and parents do not have the same say in establishing agreement in
the family (i.e. that the family cannot but be an unconditional democracy).
It has been argued that within-family distributions are a private matter
which there is no need for outsiders—be they researchers or politicians—to
be concerned with.20 I would not agree with that; it must be equally
important to protect children against injustice inside the family as outside of
the family. I would, however, accept that the family is not just any social
system, but a system with special qualities of intimacy, and that there is for
that reason broad scope for different treatments of family members,
including in consumption, which arise through some form of agreement and
which are therefore entirely a matter for the family and should not be taken
as evidence of inequity.
 The distribution of well-being in the family should ideally be measured
directly in the distribution of consumption. That is, however, not easy in
practice, because of problems of measurement and because there are
exceedingly intricate forms of sharing in the intimacy of the family. Indeed,
for the time being at least, I suspect that when all is said and done the
distribution of consumption wellbeing within the family may not be
observable at all without radically simplifying assumptions which violate the
complexities of family arrangements.
Since we are at present not in a position to establish inequities empirically,
we shall in the following start by first assuming equity (all family members
benefit equally from the resources available to the family) and then simulate
the effects on measurement results of theoretical inequities.

Standard of living under family equity

Table 3.12 gives estimates of relative income and consumption standards for
adults and children, under the assumption of equity within families. This
works so that in any given family, adults and children are recorded as having
the same standard; each person is allocated the equivalent income and
equivalent value of consumption resources, as estimated from family
income/consumption resources.
In spite of the assumed equity within families, it turns out that the
standard of living is different for children and adults as groups. These
differences do not result from any discrimination against children in their
families, but from the position of families with children in the overall
distribution of income. Results:
1. Children have a lower standard of living than do adults. In 1986, the
average income standard of children was 75 per cent of the average for
adults.
2. Children also have a lower standard of living than do parents, but this
difference is much less than between children and all adults. For example,
children in married-couple families have an income standard which is 94 per
cent of that of married parents. The difference is, however, persistent. The
average for all children is lower than the average for all parents, that of
children in married-couple families lower than that of married parents, and
that of children in single-mother families lower than that of single mothers.
3. The difference between children and adults is less in consumption
standard than in income standard. The difference between children and
parents is about the same in consumption standard as in income standard.
4. From 1976 to 1986, the standard of living (income standard as well as
consumption standard) improved less for children than for adults and slightly
less for children than for parents. Both children and adults had a slightly
better development in income standard than in consumption standard.
5. Among children, the relative standard is slightly higher than average for
children in married-couple families and considerably lower for children in
single-mother families. Children in marriedcouple families had a slightly
better development from 1976 to 1986 than the average for all children;
children in single-mother families lagged further behind compared to other
children.
6. Along the family life-cycle, the relative standard of children improves
with the age of their parents.
7. Among adults, the standard is much lower for those who have children
in the family than for those who are without children (but less so in
consumption standard than income standard), and those with children had a
slightly slower increase in standard of living than those without children.

Standard of living under family inequity

Assumptions about inequity within the family are now introduced.

First assumption: that children are discriminated against compared to
adults/parents. This is justified on the argument that the family is a normal
social system where collective solutions are the result of distributional
conflict and outcomes more or less reflect differences in power. Children are
taken to be the weaker party. Patterns of family discrimination against
children are observed in shortage economies.21 The effects of three degrees
of inequity are estimated, resulting from discriminations against children to
the effects of 5, 10, and 20 per cent (all children in all families). The method
by which discrimination is implemented in the estimates is explained in
Appendix D.
As a sub-plot, we will consider the possibility of persistent discrimination
against children but to a lesser degree in the most recent observation than ten
years earlier. This is justified on the grounds that there may have been
changes in this period in power relations within the family which are to the
benefit of children. The position of women may have become stronger,
either as a result of increasing female labour-force participation and earnings
(combined with reductions in the same or men) or as a result of general
changes in attitudes in society to male and female authority. There is
evidence in the literature that women (mothers) may have stronger
preferences than men (fathers) for spending to the benefit of children and
that this carries through to indicators of child welfare.22 We might therefore
expect some improvements for children if the position of mothers in the
family is strengthened.
Second assumption: generally, children are treated equitably within
families, except that parents in poor families protect their children against
poverty by sacrificing their own consumption. This is justified on the
argument that families are extraordinary social systems where normal
conflicts are not in play and that parents are altruistic when their children are
in need. There is evidence of such preferences and behaviour among poor
parents from survey research.23 Children whose standard is below 50 per
cent of the median for all persons are assumed protected to the effect of 10
per cent.
The first stage in the analysis was to re-estimate tables of the 3.12 kind
under the assumed discriminations against children.24 This had the obvious
effect of reducing the measured relative standard of living of children. For
example, the income standard of all children relative to all adults in 1986
was reduced from 75 per cent under equity to 70, 64, and 55 per cent under
the three assumed degrees of discrimination. Their relative consumption
standard was reduced from 81 per cent to 75,70, and 59 per cent
respectively. Since the magnitude of discrimination is assumed to be the
same in both years, there is no change in the trends over time compared to
those in Table 3.12.
If a reduction in the magnitude of discrimination from 1976 to 1986 is
introduced, this will obviously affect the trend figures. For parents and
children in married-couple families it works out like this: under the equity
assumption (Table 3.12), income standard increased by 25 per cent and
consumption standard by 22 per cent for children, and by 28 and 23 per cent
respectively for parents. If we assume discrimination against children to the
magnitude of 10 per cent in 1976 and 5 per cent in 1986, the trend figures
become, for children, 36 per cent in income standard and 31 per cent in
consumption standard, and for parents 25 and 20 per cent respectively.
Hence, as would be expected, the movement from a higher to a lower
magnitude of discrimination against children causes the measured standard
of living of children to increase considerably more, and that of parents
slightly less compared to the results under both the equity and the stable
discrimination assumptions.
We now move from comparisons of relative standards to measures of
overall distribution and poverty. Table 3.13 gives two meas- ures of
inequality, the Gini-coefficient and half the coefficient of variation squared
(HCVS). These both record the shape of the entire distribution so that the
higher the value the more inequality, but with the difference that the HCVS
may be more sensitive to partial redistributions within the overall
distribution, for example in the bottom part of the distribution from the very
poor to the less poor. (As it happens, in this case both measures give the
same picture.) Distributions are estimated in equivalent values between
persons. These are the main results:
1. There is much less measured inequality in consumption standard than in
income standard.25
2. There is considerably more inequality in 1986 than in 1976, as
measured in both standards.
3. The introduction of within-family inequities has little or no effect on
the measures of overall inequality. With discrimination, the degree of
measured inequality is increased, but only with a discrimination to the
magnitude of 20 per cent is there a real change in the measures of inequality.
The 20 per cent assumption must be said to be extreme, both in its
magnitude and in applying to all children. The effect of moving from the
equity assumption to the protection of children assumption is negligible in
the measures of overall inequality. The reason these effects are modest is
obviously that all redistribution as a result of assumed discrimination or
protection occurs within the family, which for individuals means marginal
shifts up and down around the level of income or consumption given by the
economic situation of the family.
The measurement of poverty is by the head count method. Those who are
below half the median, specified in alternative ways, are counted as poor.26
Table 3.14 gives the results under the equity assumption. The overall
incidence of poverty may here be relatively low compared to similar results
in some other studies since the sample is restricted so that both the young
and the old are not included.
The first set of poverty estimates are in terms of income standard and
measure the proportion of persons whose income standard is less than half
the median income standard of all persons. This shows, in 1986, an
incidence of 11 per cent among children, 9 per cent among parents, and 7 per
cent among all adults. The incidence of child poverty is the same among all
children in married-couple families, about 11 per cent, but strongly
differentiated by family life-cycle, being as much as 21 per cent in the
youngest family category. Poverty is on a slightly higher level among
children in single-mother families, about 13 per cent, and here on about the
same level in all three age/life-cycle groups.
The 1976 figures are in all respects different. Overall, relative poverty was
on a much lower level, 5 per cent among children and 3 per cent among
adults. It was marginally lower among children in married-couple families,
about 4 per cent, but here with virtually no difference by age/life cycle.
(Further evidence, hence, that something happened to the family over this
reltively short period.) Among children in single-mother families, poverty
was on a clearly higher level than among all children, about 9 per cent, and
much higher in the youngest category. The increasing rates of poverty as
measured here result directly from the increasing overall inequality.
The second set of estimates is by consumption standard. Poverty by this
measure is universally much lower than by the income standard measure,
reflecting the lower level of inequality in consumption standard than in
income standard; about 1 per cent among children (4 per cent among
children in single-mother families), and less than 1 per cent among adults (2
per cent among single mothers). In 1976, poverty thus measured is about 0.5
per cent among children and 0.3 per cent among adults. (Poverty rates by
consumption standard, but keeping the income standard poverty line, were
also estimated, and this entirely wiped out measured poverty among children
and reduced the incidence to a fraction of 1 per cent among adults.)
Table 3.15 summarizes the poverty results, including after the imputation
of assumed discriminations against children and the assumed protection of
children in poor families. Again, we see higher incidences in 1986 than in
1976 and higher incidences in income Standard than in consumption
standard. These results are, as would be expected, sensitive to assumptions
on within-family distributions, but not in an entirely straightforward manner.
First, the total poverty rates are only moderately sensitive to within-family
distributions. This is for the same reason that the overall inequality measures
are only moderately sensitive; assumed within-family redistributions move
persons marginally up and marginally down along the overall distribution.
There is a further consequence for the measurement of poverty. Had we used
the poverty-gap method, rather than the head-count measure, we would
again have found the total measure (the aggregate poverty gap) to have been
relatively unaffected by alternative assumptions of within-family
distributions since the additional poverty for some persons (e.g. children)
would more or less be balanced out by less poverty for other persons (e.g.
parents).
Second, the results for child and adult poverty are strongly sensitive to the
assumptions. This shows that while the introduction of within-family
inequities does not matter much for measures of the magnitude of poverty in
the population, it matters strongly for results in terms of who the poor are
(when poverty is measured by the head-count method).

The standard of living of children

The standard of living of children depends on the position of families with

children in society and the position of children in their families. If we on the
latter assume that children are treated equitably with parents, we get the
results in Tables 3.12 and 3.14. These tables show that children as a group
have a standard of living which is lower than that of both adults and parents
specifically, and that children have seen less increase in their standard of
living than have adults/parents. This does not mean, obviously, that all
children have a lower than average standard of living—there is, for example,
a clear life-cycle pattern to the effect that the relative standard of living of
children improves noticeably with the age of the family—but on average
children are worse off than adults/ parents.
These are the results we get when children are treated with scrupulous
fairness in their families in terms of access to consumption, i.e. when in each
family children and parents are assigned the same standard of living. The
interpretation of these results must hence rest on how families with children
are treated in society.
Children are costly for families. This causes families with children to be
located towards the lower end of the overall distribution of well-being in the
population; this is why children have a lower standard than adults.
Additional children represent an additional opportunity cost (at least while
families and children are relatively young). This causes families with more
children to be located towards the lower end of the distribution among all
families with children; this is why children have a lower standard of living
than parents. Since families with children have these locations in the overall
pattern of distribution, they will suffer disproportionately in a period of
increasing inequality. This shows up in our results in the lower rates of
increase in the standard of living of children than of adults/parents.
The relative position of families with children can be assumed to result
from a combination of voluntary choices and involuntary adjustments on the
part of parents. From the point of view of children this might work out as
follows:
1. The lower rates of increase in the standard of living of children than of
adults/parents is a result of increasing overall inequality. In this, there is no
element of voluntary choice on the part of parents. It is therefore established,
I suggest, that in terms of consumption well-being children suffered more
than adults/parents from the increasing degree of inequality over the period;
they found themselves on the losing end of the redistribution.
2. As for the relatively low level of resources available to families with
children at any point in time, it would seem difficult to interpret this entirely
as the result of voluntary adjustments on the part of parents. There is a very
considerable opportunity cost of children on family income. This appears to
result mainly from the withdrawal of women/mothers from gainful
employment. This effect is stronger in Britain than in some other European
countries, partly due to the relative lack of child-care facilities and other
forms of family support.27 It is therefore established, I suggest, that to some
degree the relatively low standard of living of children is a result of society
not offering families with children the access to resources which they would
need to obtain an equitable standard of living.
3. To some degree, however, the lower level of material standard in
families with children may nevertheless be a result of voluntary adjustments
by parents, for example in order to be able to spend more time with and for
children. In the period under consideration, time used in child care has
increased quite substantially. This is true for fathers as well as mothers. If we
take it that parents are doing what they in the circumstances see as best for
their children, there would be a residual in the difference in standard of
living between children and adults/parents which we should not count as
relative deprivation of children. What the magnitude of this residual may be
we cannot know, but in the final interpretation the differences expressed in
our statistics may need in some measure to be tempered as resulting from
voluntary choice.
 There are very considerable consequences in the measurement of standard
of living of incorporating within-family production. This applies to
measured levels, as well as to distributions and trends. The consequences of
incorporating within-family distributions, however, are much more modest.
First, we have found that children are a relatively deprived group, both
compared to adults/parents at any point in time, and with respect to trends in
the standard of living over the period observed. This conclusion does not rest
on any specific assumption about within-family inequity and is compatible
with an assumption of equitable treatment of children within their families.
Second, we have found that commonly used measures of inequality in
society are remarkably insensitive to different assumptions about within-
family distributions and that measured inequality in society is about the
same irrespective of whether within-family distributions are assumed to be
equitable or inequitable, unless within-family inequality is assumed to be
extreme. Only when it comes to the ranking of persons (or groups) in the
structure of inequality, including the composition of the below-poverty-line
population, do different assumptions about within-family distributions
appear to be of real significance for measurement results.

Conclusions

Three stories emerge from this study. There is a story about the family, a
story about children, and a story about measurement.

The family

Production within the family adds very considerable resources to the goods
families are able to buy in the market with available cash income. Using
conservative definitions and methods of valuation, household production is
found to add as much as just over a half to disposable income for all families
and about two-thirds for families with children.
Production of goods is only the first job the family does in respect to the
consumption available to family members. It also translates its aggregate
bundle of goods (be they bought in the market or produced in the family)
into consumption for individual family members. In this process the family
does a second job through co-operation. This (normally) enables family
members to each have a higher value of consumption than they could each
have had on their own from the same aggregate bundle of goods. The degree
of co-operation is without doubt very different from family to family, but a
family would not be a family if there was no co- operation (although in
extreme cases of destructive family conflict there may be ‘negative co-
operation’). With reasonable assumptions it can be suggested that the
average sum value of consump- tion available to family members ‘after’ co-
operation is between 30 and 40 per cent higher than the per capita market
value of the consumption resources available to the family. The value added
through co-operation depends strongly on the number of family members
and will be higher in families with children than in families without children.
These estimates should not be taken as precise statistics, but rather as
indications of the order of magnitude. It is clear enough, however, that in
terms of consumption well-being, cash income and the goods cash income
can buy in the market are only the raw materials that go into final
consumption. The average family adds as much or more value on its own as
it can bring in from the outside. This is not to say, obviously, that the family
could make do without the goods and services it buys—it would have
nothing to process further through its own activity—but it firmly establishes
the family as a substantial institution of production.
Families are sensitive in their economic arrangements to the presence of
children. Mothers are much less likely than other (married) women to have
paid work, or, if they have paid work, are likely to work less. Both mothers
and fathers devote more time than other adults to housework and (obviously)
child care, in particular mothers. Because of these influences, families with
children have less cash income than do families without children, other
things being equal. The ‘opportunity cost’ of children depends very much on
the age and number of children, and on the ‘age’ of the family in the family
life-cycle. As a rough approximation for young families with young
children, the opportunity cost may be in the range of 30 per cent of expected
income, expected meaning other things being equal but no children. (This is
‘opportunity cost’ only and does not include outlays to cover the cost of
child-rearing.) Hence, although consumption value is added through family
economies, families with children on average start from a lower income
base.
Families do not take the loss of income sitting down. They adjust (i.e. in
the main mothers adjust) by increasing the production of goods within the
family. The loss of income is to some degree compensated by additional
household production. This reduces the opportunity cost but far from
eliminates it. For young families with young children the value of
consumption resources (from the mar- ket and household production) is in
the order of 20 per cent less than the expected value without children.
These effects of children on the resources available to their families carry
through to differences in the standard of living of families with and without
children. The standard of living in families with children, whether measured
by income or total consumption resources, is considerably lower than the
average standard of all families, and is lower the higher the number of
children.
In addition to the level of consumption resources, families with children
differ from other families in the composition of resources by source. Not
only are families with children relatively more dependent on household
production, they are also in terms of cash income relatively more dependent
on state benefits. This is as one would expect since some state benefits are
aimed directly to compensate for the cost of child-rearing. However, state
benefits are nowhere near enabling families with children to maintain a
standard which is on a par with the average standard in the population.
Over the period covered in this study—1976 to 1986—there were very
considerable changes in these aspects of family economies. In time-use,
there was a movement ‘into’ the family, with more time invested in
housework and child care and less in paid work. In terms of cash income, in
families with children, the relative contribution of earnings went down while
the relative contribution of state benefits increased quite strongly. The value
of household work relative to cash income decreased marginally and the
value added through co-operation was reduced substantially. Although
economies of scale in the family are assumed to have remained unchanged,
their impact was reduced because average family size was reduced. This
movement towards smaller families brought the value added down by 2 to 3
percentage points. In order of magnitude, this represents about 10 per cent of
economic growth as measured by the conventional per capita income
method.
All this adds up to a decline in the relative standard of living of families
with children. In 1986, these families were further behind the average
standard of living in the population than had been the case ten years earlier.
The brunt of this relative decline was carried by single-mother families,
while married-couple families roughly maintained their relative position.
 Do these analyses suggest that the family has become more important or
less important as a basis for economic well-being? Only in the sense that
families have become smaller is there a sign of ‘decline’ in the importance
of the family; in all other respects the family seems to have become more
rather than less important. The direction of shift in economic activity is from
market work to household work. The direction of shift in income
dependency is away from market earnings. The direction of shift in the
opportunity cost of children is towards increasing opportunity costs in terms
of cash income, while the opportunity cost in terms of consumption
resources has remained stable. This is interpreted as families having become
increasingly ‘vulnerable’ to children in terms of income but that this
increasing vulnerability is neutralized by compensating adjustments in
within-family activity.

Children

Families with children have a lower standard of living than do families

without children. This difference may to some degree be explained by age
(families with children are on average younger), but is to a significant degree
also the result of income forgone and additional needs. Families with
children lag less behind other families in consumption standard than in
income standard, but are by both standards clearly behind.
This difference between families translates into a difference in standard of
living between children and adults. The standard of living of children is
lower than than of adults, and also of parents specifically. This is a result of
the distribution of families with children in the overall structure of inequality
(families with children are towards the lower end of the distribution).
Children do relatively better the further advanced their families are in the life
cycle.
Over the period covered by this study, the relative situation of families
with children (compared to all families or families without children)
deteriorated, as did the relative standard of living of children. The standard
of living of children increased less than did the standard of living of
adults/parents. The incidence of poverty among children increased, and
increased more than among adults. This was a period of increasing overall
inequality; the cost of increasing inequality in terms of relative standards of
living was suffered disproportionately by families with children and hence
by children.
 The point of departure of this study was the question of how children have
fared in Britain over a period of increasing inequality. The answer to this
question is that although the absolute standard of living of families with
children and of children increased, in relative terms children fared badly.
Because of the position of families with children in the structure of
inequality, children suffered disproportionately from the increasing degree of
inequality.

Measurement

The income approach represents a powerful approach to the empirical study

of well-being and inequality. The present study strengthens that view; it
shows that the income approach can be applied to the measurement of the
consumption well-being of children, although children have little or no
income of their own, and that it can give insight into family life and the
significance of family life for family members.
There are well-known shortcomings in the income approach, but the
approach can be improved so as to overcome, to some degree, these
shortcomings. The present study strengthens also this view; it shows that it is
possible to take into account within-family production and distribution while
otherwise staying within the conventions of standard income analysis.
The key to improving the income approach is to think of income as
information, as an indicator, which needs to be rearranged and adjusted so as
to reflect as closely as possible the underlying con- cept of well-being. Raw
income data as they are available to us need to be adjusted quite radically.
Such adjustments necessarily rest on a range of assumptions, both theoretical
and technical. The more sophisticated the treatment of income data becomes,
the more results become assumption-dependent (although to some degree
they only appear to be more assumption-dependent because one is forced to
make assumptions more explicit).
Taking into account within-family production radically affects almost all
measures. It changes the estimated values of consumption well-being
(strongly upwards), it changes all comparisons of well-being (generally by
reducing differentials, in many cases strongly so), it changes the estimated
opportunity costs of children (reducing these), it changes the estimated value
added to consumption in the family (radically increasing the magnitude), it
changes the estimated differences in well-being between children and adults
(reducing these), and it changes measures of inequality and poverty
(reducing both). Finally, it changes all trend estimates, generally reducing
the rates of growth and upward movement in consumption well-being over
the period observed.
In contrast, imputing various redistributions within families between
children and parents in most cases has relatively modest effects on
measurement results. It does change both measures of total inequality and
poverty, but not much. Only in the composition of the poverty population is
there a real effect. Hence, if our question is, ‘how much inequality is there in
society?’ or ‘how much poverty?’, it does not matter much if we neglect
possible within-family inequities. Only if our question is, ‘who are the
poor?’, is this factor of real significance for the measurement result.

APPENDIX A: TABLES
TABLE 3.1 Income by source, 1986
TABLE 3.2 Income by source, 1976

See Table 3.1.

TABLE 3.3 Time-use by activity, 1987
TABLE 3.4 Time-use by activity, 1974/5

TABLE 3.5 Consumption resources by source, 1986

TABLE 3.6 Consumption resources by source, 1976
TABLE 3.7 Expected family income by children and cohort
TABLE 3.8 Expected family consumption resources by children and cohort
TABLE 3.9 Family income standard and consumption standard, 1986
TABLE 3.10 Family income standard and consumption standard, 1976
TABLE 3.11 Value added through within-family economies

TABLE 3.12 Income standard and consumption standard, children and adults
TABLE 3.13 Inequality under alternative assumptions about within-family
distributions

TABLE 3.14 Incidence of poverty, children and adults (%)

TABLE 3.15 Incidence of poverty under alternative assumptions about within-
family distributions (%)

1 Goodman and Webb 1994.

2 e.g. Ruggles and O’Higgins 1981.
3 The notional housekeeper wage is £2.80 per hour. This approximates
the Dec. 1986 rate as intermediate between the values for Apr. 1986 and
1987 (respectively £2.75 and £2.93) for the category chefs/cooks for adult
females given in the New Earnings Survey 1986 and 1987.
4 Coulter et al 1992.
5 Coulter et al. 1994.
6 Gershuny, forthcoming.
7 Sample inclusions and exclusions:
 8 Women aged 16–59, men 16–64, employed and unemployed; General
Household Survey 1986.
9 Cf. e.g. Cigno 1991, ch. 6.
10 While the absolute income differences in the table are independent of
the choice of reference category, the relative differences are not. Estimates
with alternative reference categories show the relative differences to be
sensitive to this choice, notably to geographical location within or outside
the south-east. The refer- ence categories in Tables 3.7 and 3.8 are located in
the south-east. Substituting reference categories outside the south-east for in
the south-east does not change the pattern of difference in the relative
estimates between the two years of observation, but reduces the relative
difference in many cases.
11 The estimated value added depends on the equivalence scale. The 1.0–

0.7–0.5 scale gives the 35% estimate. A flatter scale would result in a lower
estimate of value added and a steeper scale in a higher estimate. For
example, a scale of 1.0–0.8–0.6 would give the value added at 24%, and a
scale of 1.0–0.6–0.4 at 48%. This is a bit different from the similiar results
reported in Ch. 2 because the analysis here is on a more restricted sample.
12 With the alternative scales given in note 11, this estimate would shift to
95% and 135%, respectively.
13 However, the value added is estimated relative to disposable income,
which excludes non-cash resources from outside the family, e.g. ‘free’
services from government or ‘fringe benefits’ from employers. This may
have the effect of exaggerating the relative importance of within-family
economies.
14 Cf. e.g. Young 1952.
 15 Brannen and Wilson 1987; McRae 1987; Pahl 1989; Vogler 1989.
16 Sen 1984; Harriss 1990.
17 Glendinning and Millar 1987.
18 Piachaud 1982; Dwyer and Bruce 1988; Lazear and Michael 1988;
Haddad and Kanbur 1989; Jenkins 1994.
19 For example, Davies and Joshi 1994.
20 e.g. Dilnot et al 1984.
21 Harriss 1990.
22 Brannen and Wilson 1987; Thomas 1991; Hoddinot and Haddad 1995.
23 Easton 1986.
24 These tables are not reproduced.
25 This result is consistent throughout the study: there is less difference

and inequality in consumption standard than in income standard. The result

is influenced by the method of valuating household work, which is by the
housekeeper- wage method. This gives a relatively flat distribution of the
monetized value of household work since this value is determined by time-
use only. With the alternative method, using the opportunity cost, the value
would be determined by both timeuse and wage level, and this would give a
less flat distribution of the monetized value of household work. Gini-
coefficients for the distribution of consumption resources were estimated
using both shadow wage and housekeeper wage. This does not change the
main conclusion: irrespective of the method of valuation, there is much less
inequality in the distribution of consumption resources than of disposable
income. However, with the shadow-wage method the Ginis are reduced less
than with the housekeeper-wage method, as a general indication by about
half as much. (Shadow-wage estimates are based on actual wages for all
persons in work, and using either mean wages or ‘Heckman’ estimates of the
wage where the person’s wage-rate is not known due to non-employment or
self-employment. These two methods give roughly the same results.)
26 The term ‘poverty’ is used only as a terminological convenience. It is
defined in purely relative terms, and is hence derived directly from the
distribution.
27 Joshi and Davies 1993.
 APPENDIX B: THE IMPUTATION OF
TIME-USE IN THE INCOME DATA

The imputation of time-use in the income data is based on regression models

estimated on data from the Social Change and Economic Life survey (1987)
and the BBC 1974/5 time-use survey. Using a set of variables common to
these surveys and the harmonized Family Expenditure Surveys, a predictive
regression model of time-use was developed. The parameter estimates from
this model were then used in the FES surveys to generate predicted time-use
values for FES households. The procedure is reported in greater detail in
Gershuny and Halpin (1995).
The models estimated on the 1974/5 data are used to predict time-use in
the 1976 FES, and the models estimated on the 1987 SCEL data are used in
the 1986 FES. Separate models were estimated for men and women.
The set of relevant variables available in all four data sets is small. They
consist of:

1. Sex
2. Age group
3. Age of youngest child
4. Number of children
5. Employment status (i.e. full-time, part-time, not in the labour market)
6. Paid work time

Definitions must match for all the variables listed above. The first four are
straightforward. Employment status reflects full-time/part-time conditions of
employment, and is not directly calculated from actual or usual hours
(though it will correspond very closely with the latter). This can be matched
in the FES. Paid work time is more problematic. The diary work time is
accurate work time for the period of observation and includes travel-to-work
time. The best we can get in the FES is reported usual hours. The
relationship between these is not extremely close, as hours reported (rather
than recorded) may be notional or contractual hours (see Robinson and
Gershuny 1994). However, the mean levels for FES reported hours and time-
budget diary hours are very close, once respondents with zero scores are
removed.
A number of other available variables have been tested and excluded as
they were not significant. Conversely, some significant variables have been
excluded because they are not available over all files—for instance, income
source (but, importantly, not income level, which has very little effect on
time-use).
The simple model performs relatively well, compared with fuller models
which were estimated on the 1987 SCEL data, and which included detailed
information about the effects of income amount and type, and wage level, on
time-use.

APPENDIX C: THE ‘EFFECTS OF

CHILDREN’ REGRESSION

The goal of this analysis is to examine the effects of children on the

disposable income and consumption resources of the family. This is done by
estimating a regression model, with family income or consumption resources
as the dependent variable, a number of control variables, and variables
describing the number of children and the age of the youngest child. The
intention is only to flush out the influence of the single factor we are
concerned with, the presence of children, and not to explain family
income/resources more comprehensively.
We use a straightforward least-squares regression model which is
described below. Before arriving at the final model, a range of variables and
model forms was examined. In the end, a relatively simple form (an
untransformed dependent variable and ordinary least-squares) was preferred,
simply because more complex models did not perform better in terms of fit
or interpretability.
The final model contains relatively few variables. Had our purpose been
to model the determinants of family income/consumption, we would have
added more variables and obtained a better-fitting model. However, since the
intention is only to estimate the effects of children, we have excluded
intermediate variables through which children affect income/resources (e.g.
the employment status of parents), and limited the model to child variables
and genuine control variables. We wanted to include the education level of
the head of household (and did so for some analyses of the 1986 data, where
it contributed significantly), but could not do this in the comparative analysis
since this information was not collected in 1976.
In fitting the model on the whole data set it became apparent that the
effect of age, even controlling for other factors, was non-linear, and the
effect of other variables changed with age. To take account of this in a single
model would be rather unwieldy and the results would be difficult to
interpret. Instead, we split the sample into three age-bands, 20–9, 30–9, and
40–9 by head of household, and fitted a simpler model equation to each
separately (age is entered only in linear form, and the only interaction fitted
is that between age of youngest child and number of children). This strategy
removes the non4inearity of age (within each band the effect is more or less
linear, and the overall non-linearity is reflected by differing age parameter
estimates in each ba2nd) and also allows the parameter estimates for the
other variables to differ with age.
The dependent variables are family disposable income and family
consumption resources (untransformed and unequivalized). The presence of
children is captured by two independent variables, number of children
(categorical: none, 1, 2, 3, 4 or more), and age of youngest child
(categorical: none present or under 5, 5–9, 10 and over). Control variables
are number of adults (continuous), age of head of household (continuous),
residence (binary: London and the south-east versus elsewhere), family type
(categorical: couple, single male, single female), and a dummy variable
identifying single-female parents. The interactions between number of
children and age of youngest child are included. The raw regression results
are in Table 3.16.
This model fits well within the age-bands. R2 is not very high but most
parameter estimates are highly significant; overall fit is constrained by our
wish to isolate the relationship between children and the dependent
variables, which requires us to exclude other powerful explanatory variables
such as employment status because the effect of children will operate in
large part through such variables.
 The interaction between number of children and age of youngest is
significant only in some of the models but is included in all panels for
consistency. Where it is insignificant its inclusion has very little effect on the
predicted values.
Tables 3.7 and 3.8 are generated from the raw regression results by
summing the parameters for age of youngest and number of children
relevant to each cell. For instance, in Table 3.8 (1986, the 20–9 age-band)
the value of –3,898 for families with a single child aged 5–9 is the sum of –
3,064 (the raw effect for number of children = 1), –707.70 (the raw effect of
age of youngest child being 5–9), and –126.8 (the interaction term).
Other formulations of the regression model have been examined, notably
the transformation of the dependent variable, and the square and square-root
transformations. None offers particular improvement in fit over
untransformed income.
Years of education of head of household is available only for 1986 and is
therefore not included in the presented models. Its effect is generally to
reduce the range of difference between categories. In particular it reduces the
penalty associated with higher numbers of children. However, it does not
materially alter the pattern of the relationships. Its effect differs across the
three age-bands, suggesting that at least part of its effect is through the
timing of family formation.
TABLE 3.16 Estimates of the ‘effects’ of children regression on family income
and consumption resources
 APPENDIX D: PARAMETERIZING
UNEQUAL SHARES

Making a rule to define unequal sharing within families is problematic. How

do we characterize something that is dependent on the structure of the family
(i.e. number of adults and children) in terms of a single number? What could
it mean to say that children are discriminated against to the extent of, say, 10
per cent?
In, for example, the case of a one-child, two-adult family, if the child
benefits from only 90 per cent of a ‘fair share’, then each adult gets 105 per
cent. If there are five children, each adult gets 125 per cent, at least if we use
a rigid rule that each child loses by 10 per cent and the adults absorb the
surplus.
One alternative is to give each person a score, indicating his relative
ability to acquire a share of his family’s income, and to distribute the income
according to the ratio between the individual’s score and the sum of the
scores within the family. Children can be assigned a different (lower) score.
Let π1 be the portion of the family income consumed by individual i:
τ
πi = Y ρi = Y
∑ τj
j

1 − 10% if i is a child
τ = {
1 + 10% if i is an adult.

Σi πiis equal to Y, the family income, since Σiρi, equals one. If we express
the actual share achieved as a proportion of a perfectly equal distribution, we
find this varies with the family composition. The actual share can be defined
as:
βi = ρi × N ,

where N is the number of persons in the family. For a ‘bias’ of 10 per cent,
in a one-child/one-adult family, this will be:
1+10% 1.1
β1 = 2 × = 2 × = 1.1
(1+10%)+(1−10%) 2

1−10% 0.9
β2 = 2 × = 2 × = 0.9
(1+10%)+(1−10%) 2

for the adult and the child respectively. For other family compositions the
figures move away from 10 per cent but not extremely so. Table 3.17 gives
the proportional shares in families with one or two adults and one to five
children. Children’s shares vary from a low of 87 per cent (single child with
two adults) to a high of 96 per cent (five children with a single adult). It may
seem counterintuitive that a group of many children will do better than a
single child, but bear in mind that (i) in a large family each child will be
competing against other children as well as against adults, and (ii) large
families will tend to be worse off at the level of equivalent income.
This strategy improves over the technique of simply reducing children’s
income by a fixed proportion in that it takes account of family composition,
and in having an interpretation in terms of relative ability to demand shares
of consumption.

Table 3.17 Shares of the cake where children suffer a 10% ‘bias.’
PART II

   Reform
 INTRODUCTION: SOFT MEANS FOR
HARD GOALS

When the necessaries of life are once provided, everyone should seek to
increase the beauty of things in his
possession rather than their number
or their magnificence.
Alfred Marshall, Principles of Economics

Liberty is the central value in democracy. Therefore, democratic

government is bound and constrained: it cannot
take action which
trespasses on liberty.
Liberty is, however, not the only value in democracy; democracy is also
about equality. There may be
differences about equality of what—rights,
opportunity, resources, property, living conditions— but equality is
an
intrinsic value in the very idea of democracy. Therefore, democratic
government is condemned to action: it
can- not choose to do nothing about
persistent injustice.
The libertarian answer to the double imperative is that there is very little
government can do without
overstepping on liberty. The revolutionary
answer is that the equality imperative is overriding and that
liberty must be
sacrificed. In the classical theory the expectation was that liberty would be
sacrificed in the
short term in order to create ‘true democracy’ in the longer
term, but we now know that the revolutionary
answer is undemocratic not
only in the short term, but totally.
Between the libertarian and the revolutionary answers, there are two
reformist positions, a weak and a strong
version. The strong reformist
position shares the vision of the revolutionaries of a ‘true democracy’ in the
future to which society should be moved, but also accepts the liberty
imperative and therefore holds that
society can only move gradually
towards that goal. This is the original social democratic theory. The weak
reformist position is that de- mocracy does not arise as some ultimate result,
but sits in the process, and that reform is not a project moving towards a
specific goal, but a way of living
permanently with the double imperative.
This is the Popperian theory of piecemeal trial and error, or,
for that matter,
the Gandhian axiom: there is no road to peace, the road is peace.
In this menu of theories, there are two deep dividing lines. One is
between the libertarian position and the
rest. Libertarians believe
(essentially) that there is no dilemma, that liberty and equal- ity are one, and
that liberty is advanced when government retracts as much as possible. The
other dividing line is between the
weak reformist position on the one hand,
and the strong reformist and revolutionary positions on the other
hand.
These two latter positions have in common the idea that there is an end goal
to which society should be
taken—the terrible temptation towards
perfection. That agreement is more essential than is the disagreement on
means. The weak reformist position is philosophically different, it has no
vision of an ultimate goal, it is
open, it is the theory of the open society. In
practical affairs, the strong reformist theory will encourage
strong policy
interventions because it has a strong goal to justify strong means. The weak
reformist theory
will encourage action, but within restraint because the
imperative of liberty will weigh more heavily.
The Scandinavian countries have often been seen as the standard bearers
of social democracy, but that is hardly
correct if social democracy is
associated with the strong reformist theory. Only in Sweden has this theory
been
a direct influence. That is explained by the strength of the social
democratic movement in Sweden and the
strength of ideology in Swedish
social democracy. There has, in my view, been no other experiment than the
Swedish one in strong reform. That experiment has now collapsed.
Libertarian theory is today a powerful influence in the democratic world.
This, I believe, is unfortunate and
harmful. It allows inequalities to thrive
which have nothing to do with liberty, it en- courages indifference
to
deprivation and unfairness, it infuses all aspects of social life with ‘the
enterprising spirit’ and other
forms of egoism, it ridicules the idea of public
service, and it makes good currency out of ‘efficiency’ void
of values.
With the demise of strong reform, never mind revolution, the alternative
to libertarianism is weak reform. Much
therefore hinges on the credibility
of this theory, the very survival of reform
democracy. The libertarian
counterfactual is repugnant. The need for reform is as strong as ever. The
modern
period of reform in the industrial democracies, roughly the three
decades after 1945, did succeed in making
capitalism palatable, but did not
elevate society to a higher plateau of spontaneous harmony where further
reform would be unnecessary. When reform lapsed, in the two subsequent
decades, society fell back towards
repugnance, with conspicuous
consumption and the re-emergence of old inequities—the income gap, mass
unemployment—and with new aberrations emerging, in particular the risk
of social exclusion among the young as a
result of combined family failure
and job failure.1 Democracy itself is suffer- ing because many see it as
unable to reconcile the
complexities of modern society, and because the
have-nots see nothing in it for them and turn away, as they are
already
doing in large numbers by not voting. Is, then, as seemingly a pitiful theory
as that of weak reform a
credible alternative to the theoretical splendour of
libertarianism, which holds up to us the promise of good
society arising
from individual egoism?
In Passions and Constraint, Stephen Holmes argues that limited
government can be powerful government
because restraint can be enabling.
He is surely right, and also right to take this as an argument in favour of
liberal democracy. Liberal democracy is government with hands tied, but
because citizens then know that
government is bound by law and not
omnipotent, government may gain legitimacy and through legitimacy
effectiveness. Critics of weak reform will argue, with tantalizingly simple
logic, that government should use
the most powerful possible means
because more powerful means must produce more effective policies. They
overlook, however, the danger that if government does not observe restraint,
it loses legitimacy and becomes
totally impotent, at least eventually (or else
authoritarian, which also does not last).
In The Civic Spirit, I look down from government behaviour to
individual behaviour. Again, the theme
is restraint: self-interest tempered.
Utilitarian theory of rationality predicts that citizens will not vote
because
there is no power in the individual’s single ballot and voting is therefore not
worth anyone’s time.
However, citizens do vote. Some political scientists
are agitated by this ‘paradox of voting’—but there is no paradox, only a
clash between theory and observation, a
straightforward case, in other
words, of a theory disconfirmed. The theorists are right that democracy
would
not work if citizens were in it only for themselves and that the
utilitar- ian theory does not explain
democratic behaviour. But rational
people are not simple-minded about goals and values; they give careful
consideration to questions such as, ‘what is good?’ and ‘what is right?’ and
are serious about duty and
responsibility. Rational citizens vote, I believe,
and take pleasure in so doing, not for reward but because
they understand
democracy. Because they understand democracy, they choose not to give in
to the instinct that
participation is a waste of effort.
The concept of restraint is central to liberal political theory. Liberalism is
distinguished from
libertarianism by ‘due concern for others’. Respect for
others is, obviously, not irrational, but utilitarian
theory, the theory that
claims to be at the core of rational choice, does not predict behaviour based
on
respect; it predicts that the individual will seek a free ride and avoid
giving away anything unless he sees
more benefit than cost to himself. This,
however, is not the way people decide what to do and what not to
do—not
in family, not in friendship, not in co-operation. Of course, people look after
themselves and do so
vigorously, but not single-mindedly, not without
compromise. In the same way that people vote because they
understand
democracy, they co-operate because they understand that without co-
operation there is no community.
Years ago, when I first started to think about inequality and society, I
wanted to believe that it was really
in the interest of the rich to help the
poor because they, the rich, would then be able to enjoy living in a
decent
society. Subsequent experience led me, to my regret, to give up that
beautiful theory. If you are rich,
it is to your benefit that others are poor
because your richness then goes all the further, certainly in
material terms
and perhaps also psychologically. Not all rich people celebrate inequality,
and many do help the
poor, but that is not necessarily because they have
calculated that inequality is to their own disadvantage; it
may just be that
they understand or believe that (some) inequalities are indecent. It is
unfortunately not to
be expected that everyone would see decency in society
—be it clean streets or universal access to health
care—to be in their own
interest if they only thought things through properly. Community and co-
operation require at least some degree of restraint in self-interest.
Where does restraint come from? It can be imposed through constitution,
laws, rules, and conventions, but such
institutions are not enough. Good
laws are necessary but not sufficient for good government, and certainly not
sufficient to encourage active, balanced behaviour by citizens. We can
impose it on ourselves through decisions
today which make it impossible
for us to give in to temptation tomorrow. This we do, but truly binding
decisions are difficult to come by, and we cannot predict all temptations. It
can be imposed through social
pressure, but determined free-riders do not
take good advice. In addition to imposed constraint, there must be,
in the
individual, understanding, flexibility, and compromise: a spirit of voluntary
restraint. There must be
acceptance of responsibility. Ultimately, restraint
cannot be only what is imposed or encouraged from outside,
it must come
from the sensitivity of each individual. Rationality is not fully explained by
external influences
on the individual. If we want to conduct ourselves with
reason, we must decide to apply reason.
A peculiar form of utilitarian indulgence is the temptation, and
convenience, for social scientists to see
themselves as observers only. This
is both too modest and too escapist. If political philosophy is any good,
the
philosopher is a participant, and towards this we should strive. But here,
too, I apply the themes of
responsibility and restraint, and I do this in a
double message to the philosophers to speak up and to shut up;
the former
when we can speak on the basis of legitimate authority as scientists, the
latter when we have
nothing but opinions to offer.
All reformists share the belief that society unchecked—unregu- lated
markets, only voluntary social insurance
and private provisions in education
and health care—would generate more than acceptable inequality. The
question is then what kind of interven- tion is needed, and in particular
whether a strategy which is com-
patible with the weak theory of reform is
sufficient. This strategy has the double attraction of promising to
modify
inequality—not to create equality; that is for the strong theory—and to do
this without infringing on
liberty.
Weak reform strategies seek to increase opportunity for choice without
regulating
the processes of choice. It is far from obvious that such strategies
would result in a more equal distribution
of well-being since new
opportunities could be ‘confiscated’ by those already in advantage. The
effectiveness
of reform strategies is therefore a matter of experience. For
research, the question is whether links can be
established between the
broadening of opportunity and the equalization of well-being.
Income redistribution from the rich to the poor is a typical weak reform
strategy. If income is successfully
thus shifted, more additional opportunity
for choice is given to the poor than the loss of opportunity the rich
have to
accept. The instruments of redistribution are transfers and taxes:
government redistributes income
‘after the fact’ without regulating the
deeper economic processes which generate inequality in the first place.
Of
course, in practice the line between redistribution and regulation is not clear
and democratic governments
certainly regulate the economy, although
usually for other than distributional reasons. Incomes policies are
not
typical and after-the-fact redistribution remains the basic strategy for
modifying inequalities of income.
Does this work, or do the rich manœuvre
around attempts to get at their privilege?
Another example of a weak reform strategy is to expand higher education
while maintaining free choice and
access in the hope that less scarcity of
opportunity will equalize attainment. Does this work, or does the
middle
class run away with it all?
The link or not between reform and post-reform distributions has been
sought, in particular, in studies of
income, poverty, and class. An observer
arriving from Mars might well conclude, after reading the biggest and
most
impressive books, that the income studies have established a link, but that
the poverty and class studies
have found independence—a two-to-one
majority, in other words, that weak reform is not enough. Taken together,
and on face value, the results of these various approaches might seem to
suggest that weak reform may shift
some income around on the surface, but
that there are ‘real’ deprivations and inequalities underneath which
persist.
However, the Martian observer might find, to his surprise, that his
balanced summary of findings from more big
books than he had thought
would be needed to clarify the matter, was rejected by all concerned. Critics
of the
work on income redistribution would call it ‘simplistic’ (and might
under their breath whisper of ‘naïvety’)
and reject the established link as
spurious. Critics of the work on poverty and
class would complain of
‘excessive relativism’ (and might under their breath whisper of ‘black
magic’) and be
equally dismissive. The score would seem to be up in the
air.
It is, indeed, difficult to establish conclusively the power of transfer and
tax policies for the
redistribution of income, and there are shortcomings of
methodology in studies that are taken as evidence of
the link. However, it
turns out to be relatively easy to sort out those shortcomings and to see
what the
resulting consequences are. That exercise is interesting mainly for
being uninteresting: new and better
research has not led to new and
different conclusions. The link on income inequality remains established.
 In Britain, statistical research on poverty has a distinguished tradition
originating in the work of Seebohm
Rowntree. He published his first major
study of poverty in 1901, and his third and final one (with Lavers)
fifty
years later. In these studies, poverty was shown to have been almost
extinguished during the first half
of the century. Britain seemed to be on the
way to becoming a society in which poverty was a problem of the
past.
The core of Rowntree’s methodology was a strict definition of the
concept of poverty; poverty meant ‘living in
obvious want and squalor’ and
those so living were counted as poor. This strictness in the concept
Rowntree
applied deliberately in order to protect the integrity of his work.
He wanted to be certain that he could not
be criticized for having
exaggerated the extent of the problem with which he was concerned by
applying a
frivolous definition.
Rowntree’s studies, in the next round, came under criticism for being too
strict in the definition and for
failing to capture the true extent of, and real
trends in poverty. It was argued that he did not sufficiently
consider the
changing nature of poverty in a changing society. A relative deprivation
theory of poverty was
proposed according to which poverty is not simply a
matter of want and squalor but of living below the standard
which is
customary and considered to represent a decent minimum in the
circumstances of the society in
question.2 This standard, obviously, is
more
sensitive to social change than that of Rowntree, for example to increasing
levels of overall prosperity;
if society becomes more prosperous the
standard of poverty will have to be adjusted upwards. From this
theory, new
‘relative’ measures of poverty were devised, and these measures gave
very
different results from those of Rowntree’s studies, results showing not only
that the extent of poverty in
Britain was high, but also that it was stable or
rising. Hence the conclusion of policy failure.
However, subsequent work resulted in a re-revision of the revision. First,
the revisionist theory was seen as
unfair in its criticism of Rowntree. In an
exceptionally fine paper, John Veit-Wilson demonstrated, after
having gone
back and carefully reread Rowntree’s work, how badly this pioneer had
been treated by those
following in his footsteps.3 The label of
‘absolute’
poverty was attached to the Rowntree type of approach, thereby suggesting
an unsophisticated theory.
However, ‘absolute’ was not a term Rowntree
used; since Adam Smith it had been fully understood that poverty is
relative
to socio-economic circumstances, no one has ever suggested otherwise, and
Rowntree’s approach was
clearly within a relative frame- work, both
theoretically and in its empirical application.
Second, the relative deprivation theory, as it came to be applied in British
research, was itself shown to be
flawed in the direction of being too
relative. Although poverty is indisputably relative, it is not totally
relative.
The concept of poverty will have no real meaning unless an absolute core is
preserved in it of real
hardship in the way people live.4
Finally, rigorous measurement under the relative deprivation theory
proved to be extremely difficult. The
extent of poverty continued to be
measured mainly by the income-poverty-line method, but the theory
contains no
element to determine where to draw the poverty line. The
combination of too much relativism in the theory and
arbitrariness in the
method of measurement gave little credence to measurement results. Also,
the
income-poverty-line method is not really the appropriate approach
under the rela- tive deprivation theory.
Rather, this theory calls for some
form of direct measurement of deprivation in way of life. Attempts show
direct measurement to give different results from the conventional
approach, including declining trends in
deprivation under stable trends in
relative income poverty.5
These difficulties in theory and application have caused measurement to
err in the direction of exaggeration.
While Rowntree was cautious about
integrity, excessively so perhaps, the revisionist
theory went too far in the
other direction, exposing itself to just criticism for—out of political
expediency?—having exaggerated the extent of poverty. The proposition
that the problem of poverty could remain
unaffected through rising
prosperity is so bold, so improbable, however much one accepts the
relativity of
poverty, that it would need very firm empirical support to be
taken as confirmed. It does not have firm
support, in fact, since income-
poverty-line results must be considered theoretically arbitrary, it has little
support at all, firm or not. In the absence of necessary support for the
improbable hypothesis, the probable
hypothesis stands. Neither from theory
nor from empirical results can it credibly be argued that poverty has
remained unaffected by economic growth and welfare state reform, and the
conclusion of independence cannot be
sustained.6
Thereby, the majority position is shifted two-to-one in favour of reform.
Remains only the ultimate question of
class inequality.
 The argument is that inequality can be modified without the processes of
choice being (fundamentally)
regulated. No doubt, further equality could be
possible if one were willing to regulate choice, such as in
access to
education or jobs. In communist Czechoslovakia, for example, impressive
equality was achieved with
the use of such means,7 but such draconian
regulations, the weak reform theory will have it, are not necessary in order
to modify inequality.
This theory has been called, correctly, ‘the liberal
theory’ in respect to class inequality.8 The liberal theory pro- poses that
class
inequality should be expected to be reduced in the process of
economic-political modernization. Economic
modernization creates new
opportunities, for example more jobs in higher- status occupations and
increased
access to higher education. Political modernization encourages
democracy, which, in return, does not remain
indifferent to inequality. The
effective use of human resources becomes more valued, and this results in a
shift towards meritocracy in the recruitment to education and occupation.
These forces combine to push back the
power of class and hence to modify
class inequalities. This is done mainly through
expanded opportunities and
by eliminating old biases in the recruitment to attractive positions in
society,
rather than by introduc- ing new ‘egalitarian’ regulations of choice.
Do modern societies work according to this optimistic theory? No, is the
answer in the best of British class
analysis. Right through all embracing
economic-political modernizations in the second half of the twentieth
century, class inequality is said to have remained constant. I disagree! This
class analysis has defined class
inequality ‘net’ of the structure of
opportunity, and has measured change in class inequality by what remains
after changes in structure are controlled for, ‘once all effects of expansion
per se are properly
allowed for’.9 The liberal theory, however,
proposes that
it is through changes in the structure of opportunity that class inequality is
modified.
In order to test that theory, we need analytic tools which keep
structure on board. One cannot test a theory
about relationships between
structure and choice with statistical measures which are estimated ‘net’ of
structure. That is exactly the mistake of the extreme relativists in poverty
research, who have proposed to
test poverty by measures which are
estimated ‘net’ of the level of prosperity.
The liberal theory on class inequality is found in its purest form in
education: expansion of opportunity
without regulating choice (with the
exception of some tentative positive discriminations on gender and
ethnicity). When class inequality of educational attain- ment is tested with
measures which are not specified
independently of opportunity, we do, in
fact, find vast changes in the degree of class inequality, as predicted
in the
liberal theory. Again, the independence hypothesis falls.
Income inequality, poverty, class inequality—the evidence is the same in
all three areas. The final score is
three to nil in favour of reform.
1 Ringen 1996.
2 Townsend 1979.
3 Veit-Wilson 1986.
4 Sen 1983.
5 Ringen 1988; Jencks and Torrey 1988.
6 Income poverty measures of various descriptions may be useful for a

range of
purposes—and are used elsewhere in this book, although mainly
for convenience as a way of summarizing
distributional information—but
are, paradoxically, not very useful as a basis for comparative conclusions on
‘true’ extents or trends in poverty.
7 Atkinson and Michlewright 1992.
8 Erikson and Goldthorpe 1992.
9 Goldthorpe 1995: 10.
 4

CLASS

‘[T]he post-war project of creating in Britain a more open society, through

economic expansion, educational
reform, and egalitarian social policies, has
signally failed to secure its objective.’1 Why? Because, in spite of a strong
upwards thrust of
social change which brought increasing numbers from all
backgrounds of social class into higher education and
middle-class jobs, a
pattern of difference persisted in favour of those from higher-class
backgrounds. This
difference remained stable amidst the expansion of higher
positions in education and occupation. Hence, ‘no
significant reduction in
class inequalities was in fact achieved;’2 ‘the terms of the competition
between the classes has shown little sign of change…
Neither the
meritocratic reforms of the 1944 Act nor comprehensive reorganization can,
in this respect at
least, be said to have succeeded.’3 ‘[The
consequence] of
post-war economic growth was not that it facilitated egalitarian reform but
rather that it
obscured its failure. [The reformers underestimated] the
resistance that the class structure can offer to
attempts to change it; or, to
speak less figuratively, the flexibility and effectiveness with which the more
powerful and advantaged groupings in society can use the resources at their
disposal to preserve their
privileged positions.’4 ‘[T]he advantaged
social
classes have been able to outmanoeuvre the social reformers… Legislation
may change the rules of the
game, but the players who are most motivated to
succeed may be able to adapt their strategies so that they
succeed in the new
game just as they did in the old.’5
This is the dismal story from an impressive body of research on class
inequality (which I for convenience will
refer to as the HGM- studies).
Implications, indeed (as G observes), which must count ‘as rather grave
ones’
for social reform.
 The HGM-story is the truth—it is a story of stable differences in treatment
between
the classes—but it is not the whole truth. While certain differences
of treatment have remained stable, much
else has changed, in particular the
structure of class and opportunity within which differential treatment has
been levied.6
Only one part of the story of class inequality is considered in the
conclusions with the grave
implications—only the difference in treatment
itself, but not how this difference interacts with the structure
of class and
opportunity to generate a distributional out- come. I will consider the entire
story and
demonstrate how this yields different conclusions and, thus,
different implications, which, as it happens, must
count ‘as rather
encouraging ones’ for social reform. This I will do without challenging the
analyses of the
HGM-studies, nor their basic assumptions, nor their results,
nor their data; in fact, I shall use the very same
data. I will challenge, not the
specific analytic results, but the broader interpretations.

THE SOCIAL COMPETITION

The HGM-studies take their conclusions from the analysis of social

mobility: the movement of persons through
the life course from class of
origin to position of destination. Social mobility is seen as a competition
which
participants enter with certain background characteristics, such as
class, and exit in more or less attractive
positions. Inequality is observed at
the outcome of the social com- petition when one can record who has
attained the goods the competition has been about, and who not. There are
two basic assumptions (which
henceforth I refer to as the B-assumptions).
First, everyone participates in the competition; everyone enters
or at least all
are (initially) equally motivated to compete. Second, there is agreement
about which positions
are the attractive ones; everyone competes (at least
initially) for the same goods. Our discussion concentrates
on education; it is
assumed that everyone has the same desire for higher education (or, if in the
course of
events that desire becomes socially differentiated, that this reflects
previous deprivations, inequalities, or
discriminations).
The social competition is analysed in mobility tables which show the
constellation
of participants by class of origin and position of destination. In
order to produce the mobility table, the
destination of the participants must
be known. Hence, the mobility table shows the constellation after
the
participants have gone through the com- petition and attained or not attained
attractive positions, as the
fact may be—after it is all over, so to speak.
(Simple mobility tables are displayed in Table 4.1.)
The social competition is played out within a framework and through
processes of allocation.
The framework consists of a structure of origins and
a structure of destinations. The structure of origins is
given in the
constellation of participants by the class in which they originated (as
measured, for example, by
father’s occupation). This shows where the
participants came from, their social position on entering
competition. The
structure of destinations is given in the constellation of participants by
position of
attainment, for example higher or lower education. This shows
where the participants ended up, their social
position after having done
competition.
The processes of allocation are the ones through which participants move
from their social background and into
the positions which are on offer. These
processes are located between origins and destinations. Participants
leave
their origins, pass through the processes of allocation, and enter their
destinations. In these
processes, people may be treated equally or differently;
the processes of allocation may, for example, be said
to be biased by class if
participants of certain class origins are systematically favoured. Class bias in
allocation has been likened to a ‘loaded dice’ of social opportunity.7
The analysis of the framework of the competition is straightfor- ward. It is
just a question of recording the
class background and the destination of each
participant and of counting it up. The analysis of the processes
of allocation,
however, is extremely complicated. In respect to education, this is a question
of accounting for
a variety of decisions and constraints in processes that
stretch over 20 or 30 years, or more: the attitudes,
motivations, and abilities
of young persons who enter the education system and who subsequently
choose to leave
it or to continue to higher levels, the attitudes and
encouragements of parents, the
resources of parents, the cost of education
and the availability of financial support, neighbourhood and
friendship
effects, rules of recruitment, grading and examinations, the influence of
teachers, school culture,
anticipations of future work and income, and a
range of other factors, and of how all these factors change,
interact, and
interact in changing ways over the years. The term ‘treatment’ is an
abstraction which is used as
a catch-all for the combined influence of all
relevant factors, known and unknown. What the exact processes of
allocation are, we cannot say, except that they are manifold and complex.
Since it is assumed, however, that
everyone enters the competition, we must
assume that something happens during the competition that allocates
some
of the participants to the attractive positions and sorts other participants out.
This ‘something’ I call
‘treatment’.
In mobility tables, the framework of the social competition is given
directly in the marginals, the structure
of origins in one marginal and the
structure of destinations in the second marginal. These data can be
interpreted without ambiguity (except for usual questions of data reliability).
In principle, the mobility table also recommends itself to the analysis of
the distributional outcome, at least
in the sense that this outcome and the
data in the table are contemporaneous. However, the distributional
outcome
cannot be read directly from the raw mobility table; the data in the table need
to be analysed further.
In this, obviously, there may be problems of
interpretation related to how this further analysis is taken
forward.
The processes of allocation, however, are not described in the mobility
table. We have data on the outcome of
the competition, but no actual
information about the process of competition itself; we can only infer post
hoc from what we know about the outcome back to what may have
happened in the process. Given the B-
assumptions, it is assumed that such
inference may be made from the information inside the mobility table.
Between the two marginal constellations, the mobility table gives the
conditional con- stellation of the
participants within the prevailing structures
of class and opportunity. This conditional constellation is taken
to reflect
previous treatments in the processes of allocation during the competition.
The process would be
assumed to have been without class bias if each class
of origin had attained a share of the good which is
identical to its size in the
structure of origins. If this is not the case, it is
assumed that there must have
been some bias. That bias can then be measured by how far the actual
constellation
deviates from the non-biased constellation.
The HGM-studies have measured differential treatment by class and from
that drawn conclusions about class
inequality (and from that again gone on
to the interpretation of failure in reform). There are two possible
objections
to this analysis. First, one could object to the post hoc interpretation from
mobility
table data back to the processes of allocation. Second, since the data
in the mobility table are subject to
further analysis before one can draw
conclusions about inequality, one could object to the specific further
analysis. Both objections are represented in the British literature.
The B-assumptions are in effect very strong. In several papers, Murphy,
for example, describes them as ‘a most
respectable prejudice’. Class
differences in educational attainment, he suggests, re- flect a lack of
aspiration among youth from the working class, and this attitude stems not
from earlier inequality or
deprivation, but from culture.8
If the B-assumptions are given up, nothing can really be said about the
processes of allocation from mobility
table information. One would instead
need a different research strategy and try to observe the processes
directly,
for example by focusing on decision points during the process rather than on
distributional
outcomes.9
The second objection is that the mobility studies may be technically
correct, but that what they measure is not
inequality. The result might be
both different analytical conclusions and different policy interpretations.
Hence, the data on class and education have shown ‘little change in the
relative share of places taken by
different social groups. Indeed, the recent
figures for the 1980s show more entrants from the higher social
classes. The
pessimistic con- clusion is drawn that there has been no change or a
worsening in the relative
chances of children from manual social
backgrounds getting into university. However, there is a major flaw in
that
argument. The social composition of the population has been changing.
There are
more families who fall into the non-manual groups and fewer in
the manual. Stability in the shares going to
university implies a relative
growth in the chances of manual-class children going to university… What
has
happened is that the number of entrants to higher education from the
non-manual groups has risen, but not so
fast as their representation in the
population the reforms of the 1960s are having their effect in the 1970s
and
1980s.… A considerable degree of equity in educa- tion was achieved.’10
The criticism of mobility research for starting from unrealistic- ally strong
assumptions is persuasive. In the
end, therefore, mobil- ity tables probably
do not have all that much at all of substance to tell us about how
people
actually move or are treated on their way through life towards their social
destinations. However, this
is a meta-question for mobility research in
general which I gladly leave to others to sort out.11 In the present discussion
we stay with the
B-assumptions and within the conventional framework of
mo- bility analysis. The question then comes down to a
problem of
measurement, to the further analysis of the data in the mobility table.

INEQUALITY

I understand the social competition for education, as analysed through

mobility tables, as follows. Persons
compete for goods which are in short
supply, in this case places in higher
education. Success is determined by the
attainment of the good. The phenom- enon of inequality resides in the
distribution of attainment. The framework of the competition and the
processes of allocation together make up
the causal ‘system’ which
generates the dis- tributional outcome. Differential treatment in the processes
of
allocation is a part of this system, which, however, since it is cause and
not effect and since on its own it
does not determine the distributional
outcome, is not to be confused with the phenomenon of inequality itself.
Information about differential treatment, for example stability or change,
tells us something about the causal
system, but not everything about that
system (it is necessary to also take in the framework) and nothing about
the
distributional outcome; it tells us something about how inequality is
generated, but not how much
inequality is generated. In order to draw
conclu- sions about inequality, we need information on the
distributional
outcome.
The concept of inequality is used in different ways in the litera- ture,
including to describe differential
treatment. This causes con- fusion, and that
confusion is substantive rather than merely terminological. There
are three
objections against identifying the concept of inequality with differential
treatment. First, if the
label of inequality is attached to differential treatment,
we would per- force be using the same concept to
describe both the
distributional outcome and one element in the causal system behind that out-
come, both cause
and effect. This is a recipe for analytic mess; concepts are
tools, a main function of which is to separate the
elements of analysis.
Second, in a competition for a good, the concept of inequality cannot
meaningfully be
divorced from the distribution of that good; it must be the
distributional outcome that has the priority
‘right’ to the concept of
inequality. Third, if the concept of inequality were tied to differential
treatment,
one would be at risk of drawing conclusions about inequality
which were at odds with the evidence on the
distributional outcome. Hence,
there seems to be overwhelming arguments for reserving the concept of
inequality
for the description of the distributional outcome. My primary
position would be to look to the
distributional outcome for the description of
inequality and to the causal system (framework and processes) for
the
explanation, and to carefully separate the two. This position ought to be
entirely obvious.
However, since there are different usages, it might still be diffi- cult to
claim monopoly for that specific definition. If so, I would nevertheless hold
it to be an irrefutable
minimal condition of valid- ity in any conclusion about
inequality (irrespective of definition) that
the conclusion is not contrary to
the evidence on distributional outcomes. If a certain definition of
inequality
were to result in conclusions about trends in class inequality of education
which were not supported
by available and solid evidence on the class
distribution of educational attainment (the distributional out-
come), it would
be the definition and not the evidence that would have to yield.

ANALYSIS

Table 4.1 gives a series of simple mobility tables for

an imaginary
population (or sample) of 1,000 persons. Consider the first panel. The
middle class (by origin) is
a bit larger than the higher and lower classes. In
the entire population, 40 per cent achieve the higher level
of education. Of
those originating in the lower class, only 17 per cent attain the higher level
of education,
as compared to 67 per cent in the higher class. The other panels
represent ‘later’ constel- lations in the same
population, to which we return
below.
It is clear to the naked eye that we here see class inequality. The
attainment of higher education is strongly
differentiated by class of origin. It
is, however, not clear to the naked eye what the degree of inequality is.
If we
wanted to compare this class-education constel- lation to some other
constellation, for example one of
the other panels in Table 4.1, and establish
in which
there is more inequality, we would need some way of pulling the
information together into a standardized
summary measure.
One possibility is to use a measure of association between class of origin
and position of
destination. We enter the social competi- tion with the
‘baggage’ of our class background. If the dice are
loaded against people of
our class, our chances of attaining the higher level of education will be
below the
average chances in the population, and much below the chances of
those in the classes towards which the dice
are favourably loaded. Measures
in this category give the strength of association between class of origin and
position of destination within the prevailing structure. This can be
interpreted as
a measure of the degree of class differentiation of treatment in
the processes of allocation.
A second possibility is to use a measure of distribution. We enter the
social competition with the
intention of attaining higher educa- tion; higher
education is the good we are competing for. If we want to
describe the
distribution of this good, we must record the result attained after the
competition is resolved
and then analyse the distribution of that result.
A range of statistical measures of association are available. In recent
mobility research, odd ratios are much
used. I follow suit.12
For the distribution as well, there is a wide choice of statistical measures,
although there is not much
experience in the literature in applying these
measures to mobility-table information. I use the
Gini-coefficient which is
the most frequently used measure in, for example, studies of income
distribution. The
Gini measures the distance between the actual distribution
of a good and a theoretical distribution of perfect
equality. Since the
application of the Gini to mobility-table information is unconventional, an
illustration of
the procedure is given in Figure 4.1.13
In the first panel in Table 4.1, the association
between class and education
turns out to be measured to a value of λ = .58, and the distribution of higher
education to a value of the Gini = .26. So far, these coefficients on their own
tell us next to nothing, except
that there is a positive association and that the
distribution is different from perfect
equality. These measures, however,
allow us to make comparisons between alternative social situations,
for
example over time, which is the issue presently under consideration.

TABLE 4.1 Hypothetical constellations by class and

education
Fig. 4.1 Lorenz Curve Distribution of Education by Class

ANALYTIC RESULTS

The first panel in Table 4.1,1 call the original

situation. This shows a great
deal of class inequality in education. Let us now say that government finds
this to be too much inequality and decides to remedy that problem by
expanding the number of places in higher
education—a typical case of
educational reform, in other words. The three subsequent panels show
possible
alternative situations after the reform.
For all three alternatives, the same increase is assumed in the provision of
higher education (from 40 to 60
per cent), and it is further assumed that there
is (yet) no change in class composition. What sets the
alternatives apart, are
different new conditional constellations of education by class.
 In alternative A, the strength of association between class and education is
substantially weakened. This is
the kind of result which the HGM-studies
would require in order to conclude that inequality had diminished.
That
conclusion is supported also by the distribution measure; the Gini is reduced
to less than a third from
the original situation.
Alternative B gives a different constellation in which the associa- tion
between class of origin and
educational attainment remains the same as in
the original situation. This corresponds to the observations
from which the
HGM-studies have reached their conclu- sions of stable inequality.
In this
case, however, the distribution measure gives a very different result; it is
substantially reduced
(although not as much as in alternative A).

Table 4.2 Class and education in Britain

Source: Heath and Clifford 1990, Tab. 7

Class: Higher = class 1+2; Middle = class 3 + 4 + 5; Lower = class 6 + 7
Higher education = O level equivalent and higher
Association (λ) is estimated between the two extreme classes, higher and lower
Alternative C describes a third possible outcome after the same increase in
educational provision, namely one
in which the distribution measure is held
constant. Again, the two measures give different results; while the
distribution is stable, the strength of association has increased a great deal.
In these simulations, only the provision of higher education is
manipulated. In actual fact, of course, the
framework of the social
competition has changed in the class structure as well. In other simulations
(which
are not reported), also the class structure was manipulated. This,
again, gives different results in the two
meas- ures. For example, if the
proportion of the population belonging to the privileged class is increased
while both the association between class and education and the provision of
higher education is kept
constant, the degree of inequality in the distribution
measure is reduced.
We now turn from simulations to real data, namely reanalysed HGM-data
on trends in
educational attainment by class in Britain during the postwar
period. Table 4.2 shows the structure and out- come of the competition over
higher education in
four British cohorts, going back to those born in the
period 1930–9. We see, first, that the class structure
has changed a great
deal; the propor- tion of those of higher-class origin is up and those of lower
class
down. The provision of higher education (O level or equivalent and
higher) has changed quite remarkably, up
from 28 per cent in the older
cohort to 68 per cent in the younger cohort.14
The further analyses of these data give, as in the simulations, very
different results for association and
distribution. The strength of association
between class and education is relatively stable, at least there is
no linear
downward trend. The distribution, however, has changed dramatically,
displaying a strong and
uniform move- ment towards less inequality. This is
the same kind of result as in alternative B in the
simulations.

Conclusion

There are two results, one of stability and one of change. Both the analysis
of actual data and the
simulations show measures of asso- ciation and
measures of distribution to behave differently to the same
changes in the
social competition.
Both analytic results are correct; there is at one and the same time stable
association and changing
distribution, stable difference of treatment, and
changing inequality of distributional outcome. The question
is not which
finding is the correct one, but what this combination of stability and change
means for
the question of class inequality.
 In the face of this evidence, the proposition of stable class in- equality
falls. There is no ambiguity about
this conclusion accord- ing to the primary
position on the concept of inequality. If there is
ambiguity in the concept of
inequality, the proposition nevertheless falls
according to the minimal
condition.

Interpretation

Associations and distributions are related. Differential treatments in the

process result in inequalities in
the distributional outcome. There could not
be class equality of outcome while there was still class bias in
the process.
However, associations and distributions are not mechanically linked.
Stability of association
does not neces- sarily translate into stability of
distribution, and change of distribu- tion does not
necessarily require change
in association. How a stable association between class and education carries
through to the distribution of education by class, depends on what happens
to the structure of class and
education. If there is stable association and also
no change in the structural constellations, there would be
no change in the
distribution. But if there is stable association within changing structures, the
distribution cannot remain stable.15 If the
structure of class and education
‘improves’ while the association between class and education within that
structure re- mains stable, the degree of inequality in the distributional
outcome will be reduced. This is
precisely what we have seen in the British
case.
If the dice of educational opportunity are loaded against children from a
working-class background, these
children face an uphill struggle in the
competition for education, compared to children from more advantaged
backgrounds. That would be the same whether there were many or few
children from the disadvantaged class.
However, the social effect of that
differential treatment, how it aggregates, would depend on the relative
size
of the disadvan- taged class (as it would on the supply of the good for which
there is competition).
Theoretically, if there was only a single person left in
the disadvantaged class, and the dice were the same,
that person could still
face the same uphill struggle, but the social problem would long since have
been overcome.
Difference of treatment, as that is conceptualized in this discussion and
summarized in the association measure, does not occur on its own in a
vacuum: it occurs within a structure.
Treatment and structure interact to
produce some distributional outcome. The treatment is not a thing in
itself, it
is only one half of the ‘system’ which is made up of—indeed defined by—
treatment and
structure. We cannot conceive of a structure of class and
education as existing independently of processes of
allocation; there would
be no chan- nels through which individuals could move from origins and
into the
available positions. Equally, we cannot conceive of processes of
allocation independently of a structure;
there could not be pro- cesses of
allocation if there were no positions to which individuals could be
allocated.
Young people entering education do not meet a process of allocation, nor
loaded dice, nor terms of
competition, nor treatments; they meet these things
and a structure of educa- tional positions; the
one cannot be divorced from
the other. There might be few or many attractive positions, the difference of
treat- ment may be the same in both cases (as measured by the strength of
association), but the
meaning of that differential treatment would not be the
same. The statistical measure could be
stable, but the sociological thing
would be different. The job a given differen- tial treatment does in
society—
which is not the same as what it does to individuals—depends indivisibly on
the structure within
which it operates.
To illustrate, we return to the simulations. Table
4.3 gives a different
display of the changes involved in moving from the ori- ginal situation to
each of
the three after-reform alternatives. The additional places in higher
education are assumed to be made
available in one go and filled with no
delay (and, again, that there is yet no change in the class
structure). All three
alternatives show the same constellation of pre-reform higher education, but
differ-
ent post-reform constellations, corresponding to the different val- ues
of association and distribution as
given in Table 4.1.
Alternative A displays the change which is compatible with weaker
association and (much) less inequality.
This example illus- trates the kind of
change that would be needed in the distributional outcome (when the
volume
of the good is expanded) in order for the strength of association to be
reduced. The distribution of
new places is the mirror image of the original
distribution, and the post- reform distribution is radically
equalized
compared to the pre- reform distribution. One reason why the association
between class and education has remained fairly stable in the British case
could be that when the
provision of higher education increases as strongly as
it has, very radical changes are in fact needed
elsewhere in the system for the
strength of association to be relaxed. One might think that expanding higher
education would make it easy to reduce the strength of association between
class and education, but in fact
(as association is measured in mobility
tables) this is not at all the case.
Table 4.3 Hypothetical pre- and post-reform distributions of higher
education by class.
Proportion with higher education
 Alternative C displays the change which is compatible with stable
distribution and (much) stronger
association. This example illustrates the
degree of change that is needed in the strength of association
(when the
volume of the good is expanded) in order for the distributional outcome to
remain stable. This is
achieved by holding constant the relative difference in
the proportions with higher education
(between the lower and the higher
classes), which is to say that the absolute difference has increased.
Again,
one can see how stability in one element requires very considerable change
elsewhere in the system.
Both these examples are extreme. Alternative B displays a more modest
and realistic kind of change. The
strength of association is stable but,
because the new educational places are distributed dif- ferently from
the
original ones, there is less inequality in the out- come post-reform than pre-
reform. In this case, the
absolute difference (between the lower and the
higher classes) in the pro- portion with higher education is
the same, but on a
higher level, hence less relative difference. If we take the stable absolute
differ- ence
to reflect stable difference of treatment, we would also have to
say that the power of this differential
treatment as a social force had
diminished. In the same way that an age difference between a 60-year-old
and
a 58-year-old is not the same thing as the differ- ence between a 6-year-
old and a 4-year-old—although it is
the same absolute difference, it matters
less—a 10 percentage point difference in the proportions
with higher
education in two classes is not the same thing if it is between 50 and 60 per
cent as if it is
between 10 and 20 per cent. The thing has changed, it matters
less; it does less of a job in terms
of generating inequality.
What the combination of stability and change then means, is that there has
been stability in one half of the
‘system’ which generates the distributional
outcome (i.e. in treatment), but that the effect of this
stability on the
distributional outcome has been weakened since there has not been
stability
in the other half of the ‘system’ (i.e. in structure).
In order to use mobility table information as a basis for conclu- sions on
inequality, it is necessary to
take the analysis all the way through to the
distributional outcome. This outcome results from the
interaction between
stable or changing structural constellations and stable or changing
differential
treatments. Differential treat- ment is part of the process of social
mobility but is on its own not an
indicator of the trend in the distributional
outcome. To observe the distributional outcome, we must observe
the
situation after the combined effects of treatment and structure.
What we see in the British data is not stable inequality as a result of stable
difference of treatment, but
massively reduced inequality in spite of stable
difference of treatment. The story that is told in Table 4.2 (and in alternative
B in the simulations) is a
story of a society which has moved from a high
level of class inequality to a lower level of class inequality
in educational
attainment. There is no ambiguity in this story about class inequality of
education, un- less
the concept of inequality were to be defined
independently of the distributional outcome, i.e. unless the
minimal
condition is violated. Is it also a story of a more open society? That is
perhaps more a
question of definition, but it would appear to be difficult not
to say yes: there is more higher
education available, all classes have
benefited more or less equally in absolute terms, the classes which
were
disadvantaged at the outset have benefited more in relative terms, there has
been no deterioration of
differential treatment. This is a more equal society,
it must also be a more open society
by any sensible understanding of that
term. It is not a classless society but it is a less class-ridden
society.

REFORM

The alternative transitions from pre-reform to post-reform constel- lations

displayed in Table 4.3 represent very different policies. Alternative A is
too
good to be true. The new places of higher education are distributed in a
totally different way from the old
ones. This kind of transition might be
possible gradually over a long period of time, but would otherwise not
be
reform, but revolution. To require this kind of policy shift before one is
willing to accept
that reform works is to specify the concept of reform in
such a way that it could
hardly be achieved in the real world, at least not
democratically.
Alternative C is equally unrealistic. New educational places are made
available, and those previously in
advantage take all they can physically
swallow, leaving only what they have no use for to the others. Here,
indeed,
the reformers are outmanoeuvred and those in advantage continue to succeed
in the new game as they did
in the old, but that has required a coup d’état for
the distribution of the new educational places.
 Alternative B is the picture of reform. There are benefits to all, but
relatively more to those previously in
disadvantage. Who outmanœuvres
whom? The more powerful and advantaged group- ings in society would
have to do
much better than this to preserve their privileged positions. This
is not a display of economic growth
obscuring the failure of reform, but one
of growth (in this case in education) as an effective instrument of
reform.
Differential treatment in the process produces inequality in the outcome.
If we are concerned about such
inequality, there are two ways of dealing
with the process. We can regulate the mechanisms of allocation, or we
can
make unfair mechanisms ineffective by changing the framework of the
competition. What has been dem-
onstrated in the HGM-studies is that class
bias in the process of social mobility has been remarkably
entrenched. Class
inequality, however, has not been equally entrenched. It may be difficult
through cautious
reform to kill the beast of class bias, but it is entirely
possible through cautious reform to pull its teeth.
It might perhaps be suggested that the reformist intention has been not
only to modify class inequality, but to
do so through the specific strategy of
changing the mechanisms of allocation. If so, one could from the results
of
the HGM-studies argue that that specific strategy had failed. However, this
line of argument does not have
much power in respect to the problem of
effectiveness in the politics of reform (and not only because
politicians
hardly think in terms of association vs. distribution or ORs vs. Ginis). First,
why would reformist
politicians prefer a costly conflict strategy when a less
costly and more consensual strategy is available and
effective? Indeed, it is
in the very meaning of reform to prefer the less costly strategy; reformist
politicians should avoid, and clearly not prefer, conflict strategies. Second, it
seems to be an
empirical fact that the main strategy of reform, certainly in
education, has been to
work through the framework of the competition rather
than through the processes of allocation. The reformers
may have hoped for
a softening of the association between class and education, but what they
have done
is mainly to expand the provision of higher educa- tion, and to
good effect.
In education, the abstraction I have called ‘treatment’ is in real- ity for the
most part a matter of choice:
the choices parents make and the choices
children and young people make; choice within constraints, obviously,
but
nevertheless choice. It is not impossible to regulate the processes of
allocation, and in some measure this
is being done—quotas, affirmative
action, positive discrimination— but such regulation may easily come into
conflict with the principle of free choice in education. There are, therefore,
in a liberal soci- ety,
considerable limitations on this strategy. That, however,
does not define the limits of egalitarian reform
since it is possible to modify
distributional outcomes without (strong) regulations in the processes of
choice.
1 Marshall et al 1988: 138.
2 Goldthorpe et al 1987: 328.
3 Heath and Clifford 1990: 15.
4 Goldthorpe et al 1987: 328.
5 Heath et al 1992:220, 241.
6 Not all differences have remained stable, for example that between
women and
men, measured in the same way as between the classes, has been
reduced, cf. Halsey 1992.
7 Halsey 1977.
8 Murphy 1981, 1990, 1992; also Saunders (1990) has claimed that the
mobility
studies rest on unrealistic assumptions about equal participation in
the social competition.
9 For an analysis of this kind in respect to educational careers, see
Gambetta
1987.
10 Glennerster and Low 1990: 71–2, 76.
11 It is, in fact, extremely difficult to get a firm grip on just what is

contained in the conditional constellation in the mobility table, indeed what

this thing I call ‘treat- ment’
really is. Technically, it is simple enough: it is
the statistical association between position of origin and
position of
destination within the prevailing structure as laid out in mobility table data.
The substance is,
however, everything but simple. Since the two elements of
the structure given in the mobility table are in
reality decades apart, the
association in question is one that stretches over years and years, possibly
entire
life courses. That kind of association must obviously be the result of a
range of complex influences. It may be
of interest to observe whether these
influences sum up to stability or change, but knowing that does not tell us
much about actual influences. For example, a situation of fully open access
to education but class-differentiated
motivations could give the same result
as one of equal motivation but discriminatory rules of recruitment. It is,
therefore, perhaps not unexpected that the literature is not always crystal
clear in its exposition of the
meaning of the observed associations, mixing a
variety of concepts and metaphors, such as ‘loaded dice’, ‘terms
of
competition’, ‘opportunity’, ‘advantage’, ‘relative chances’, ‘relative
success’, and also ‘relative
inequality’.
12 But prefer to express the strength of association by the A parameter,

which is
a log transformation of the OR. Its numerical value is simply more
attractive than the unwieldy raw ORs. In all
references to association, we
compare the two extreme classes, higher and lower. ? is one-fourth of the
natural logarithm of the OR, when four frequencies are used in the
calculation.
13 The analysis of inequality in Table 4.1 takes the class of the individual

into account. Usually, the Gini is calculated from the

distribution of a good
among members of a population without considering other variables. This
‘univariate’ Gini
is a measure of simple dispersion, while our ‘bivariate’
Gini reflects the dispersion of education by class. It
has been suggested (in
comments to a previous version of this chapter) that this measure is a
‘hybrid’ which
‘confounds’ changes in marginal and conditional
constellations. Our argument, however, is that class inequality
(of edu-
cation) is a function of conditional and marginal constellations and that this
is precisely why
‘marginal insensitive’ measures are not the appropriate ones
for the purpose.
14 Following Heath and Clifford 1990, we use here a simple unadjusted
definition
of higher education as O level/equivalent and higher. It can
reasonably be argued that when the provision of this
level of education
increases the way it has over the period of observation, its ‘social value’
diminishes and
that we are, consequently, over the period not measuring the
same thing. This, however, applies equally to the
analysis of association and
the analysis of distribution. Since we consistently apply both approaches to
the same
data, such problems are of no consequence for the present
methodological discussion, although they may be
significant in other types
of analyses.
15 Theoretically, there could be different structural changes pulling in
different directions and cancelling each other out. In fact, though, as in the
British case, changes in both
class and educational structure would usually
pull in the same direction.
 5

INCOME

Surprisingly, the existence of a link between tax/transfer policy and the

distribution of income has been cast
in doubt. Although a large body of
research has concluded with great consistency that the distribution of
income in advanced economies is indeed linked to the level and structure of
transfers and taxes, this
conclusion has not become generally accepted as an
established finding.
Comparative studies of income distribution have produced two findings
which can now be characterized as firmly
based. The first is that there are
considerable differences in the distribution of income between nations. This
has been demonstrated above all in the Luxembourg Income Study.1 Before
the co-ordinated LIS estimates, there was uncertainty about this and the
tendency was to emphasize similarity more than difference. The second
finding is that there have been
considerable and systematic shifts within
countries in income inequality over relatively short periods of
time.2
Previously, the tendency was to regard
income inequality as relatively
stable and such changes as were observed as mainly fluctuations around a
stable
underlying trend. Through these findings, income inequality has
become established as a sensitive indicator of
social life.
However, the third conclusion, that transfer and tax policies are a factor
in accounting for variations and
changes in income inequality remains in
question.
The independence hypothesis is not implausible: the distribution of
income reflects
social inequalities which are determined by ‘deeper’
pressures, transfers and taxes are but a way of absorbing
and
accommodating these pressures, and the resulting distribution is not much
different from the counterfactual
under other conditions. But then the
alternative hypothesis is equally plausible: transfers and taxes
indisputably
do shift income between households, changes in transfers and taxes could
redirect income, for
example from the rich to the poor, and this could have
lasting effects. This competition of hypotheses matters.
If there is no link,
the implications would be, on the one hand, that transfers and taxes could
be seen mainly
as fiscal policies which are more or less neutral in terms of
distribution, and, on the other hand, that an
intention to modify inequality
would need stronger policy interventions which were capable of regulating
the
deeper pressures. If there is a link, the implications would be, on the one
hand, that we would have in
transfers and taxes an effective instrument of
modifying inequality, if we so wished, and, on the other hand,
that we could
not expect to be able to change such policies without social consequences,
if we should so wish
for other than distributional reasons.
We might well want to believe the independence hypothesis. There is a
strong temptation (among politicians) to
believe that transfer and tax
policies do not matter much for the distribution of income since this
liberates
fiscal policies from distributional constraints. There is a strong
temptation (among academics) to believe that
transfers and taxes are too
simple a mechanism to have much power in respect to fundamental social
structures,
such as inequality, that society will ‘strike back’ against attempts
to impose an arbitrary order on it,
certainly against attempts with the use of
such apparently superficial means which do not penetrate economic
processes and power relations, and that those in privilege will confound
such attempts by escaping the intended
consequences, for example with the
help of tax evasion or by shifting from cash income to fringe benefits. The
research on income redistribution has for the most part had to rely on
methodological approaches which go
relatively far in simplifying the
processes which generate the distribution of income, and has therefore,
when
challenged, not had sufficient power of persuasion to cut through the
temptations to dismiss transfers and
taxes as factors of much importance for
income inequality. When other
considerations—lower economic growth,
public budget deficits, unemployment—conspired towards a shift in
economic policy which would be facilitated by the belief that inequality is
not, or at least not strongly,
linked to transfers and taxes, that belief
prevailed.
In The Possibility of Politics, I asked whether the modification of income
inequality was doable with
the help of redistributive policies which are
acceptable under the liberty imperative. From as full a review as
possible of
‘standard method’ research, I found that it was, and I do not think that
conclusion is in question
as far as the relevant evidence goes. However,
although the research in support of that conclusion is
voluminous and solid,
there are shortcomings in the standard method which do leave some valid
questions about
the firmness of conclusions.
The international literature on income distribution has recently been
carefully reviewed in an OECD
study.3 That review confirms the
descriptive
findings on differences and non-stability, but also shows that a
review of actual experiences can hardly lead
to any single specific
conclusion on tax/transfer redistribution; there is now just too much
difference between
countries in policies and distributions. It does not really
make much sense to ask whether transfers and taxes
equalize income; it
depends on the transfers and taxes. But it does make sense to ask if transfers
and taxes
can equalize income. The question of what can be done will not
be answered from a review of
what is done; although what is done can be
done, what can be done is not
necessarily what is done.
There is no reason to believe that actual transfer and tax policies are
necessarily progressive in
distributional effects, nor that more transfers or
higher taxes are necessarily better in distributional terms,
nor that there are
no cases of regressive systems, nor that transfers and taxes could not
intentionally be used
for regressive redistribution. Rather than reviewing
again all matters of actual experience, a better strategy
for answering the
question of potentiality—could transfers and taxes be a credible method if
there
were a wish to modify income inequality?—would seem to be to go
back to the standard method, identify its
shortcomings, and try to sort out
what significance those shortcomings actually have for conclusions.
Standard
method research has established a link between tax/transfer policy
and the
distribution of income. Is that a conclusion which has emerged
because of shortcomings in the methodology, or
is it a robust finding which
stands in spite of shortcomings?

THE STANDARD METHOD

In the ‘standard method’, pre- and post-tax/transfer income distributions at

one point in time are compared and
the difference is taken as a measure of
the redistributive effect of transfers and taxes. This method has many
merits. It reflects real flows of market income, and of transfers and taxes,
and establishes correctly the
distribution of post tax/transfer income. It
shows the distributive profile of the tax/transfer regime and how
these
policies affect the resulting income distribution. Standard method results
should, therefore, not be
dismissed lightly, and there is nothing frivolous in
standard method conclusions which link distributions and
policies.
The weak points in the methodology are mainly two. One is the
narrowness of the factors and effects which are
considered. Direct effects of
transfers and taxes on disposable income are estimated (i.e. on income
‘after’
transfers and taxes), but all effects on factor income are ignored (i.e.
on income ‘before’ transfers and
taxes). This includes changes in economic
and social structures in society, policies which work directly on
factor
income (e.g. incomes policies), and indirect second-order tax/transfer
effects. Implicitly, the standard
method assumes that the factor income
distribution is the distribution ‘before policies’. This is, of course, a
considerable simplification. The factor income distribution itself is not
stable and is subject to policy
pressures. Taxation may have incentive or
disincentive effects which could influence the distribution of factor
income;
the same might be the case for transfer benefits, for example if they
encourage some people to work
less than they would have done if they had
not had access to ‘free’ income.
The second weakness is that the standard method is static; it gives cross-
section evidence on redistribution at
one point in time. In a sense, this is not
very meaningful since redistribution must necessarily occur over
time. A
certain redistributional result at one point in time may be ‘corrected’ by an
opposite result later,
for example as a result of second-order effects, so that
a trend analysis would show
stability or only arbitrary fluctuations around a
stable long-term distribution. Cross-section results may be
spurious in that
what appears to be redistribution between persons here and now, is in effect
the result of
redistributions over the life course for the same persons, in
which case there may still be no redistribution
of lifetime income.4

MARGINAL REDISTRIBUTION

One of the difficulties which arises from the static quality of the standard
method is that redistribution is
estimated from the profile of the tax/transfer
system as a whole. Of course, we do not keep constantly starting
afresh
with a virgin distribution of income upon which a system of transfers and
taxes is imposed. At any point
in time, we start from prevailing
distributions and policies and then introduce, if we so choose, some policy
change. In a dynamic perspective, redistribution is captured more
realistically in marginal effects on
inequality of marginal policy changes
than by estimating, more theoretically, global effects of entire policy
systems.
In Britain, the distribution of income moved in the direction of equality
from around 1950 to the mid-1970s;
that trend was then broken and the
distribution started to move back towards inequality. By the mid-1980s, the
1950 distribution had been restored and towards 1990 it had become yet
more unequal. This shift in the
distribution of income offers a sharp test on
the politics of inequality: if there
is a link between policies and distribution,
we should expect any shift in the distribution, including towards
inequality,
to be associated with changes in policies.
Atkinson, in addition to documenting the trends, has analysed the
movement towards inequality.5 He has pointed to three factors. First, a shift
from
work in the sense that fewer households are headed by persons in
work (down from about 70 per cent in 1970 to
less than 60 per cent in
1985) and a smaller proportion of household income is from work (down
from more than
80 per cent in 1975 to just over 70 per cent in the mid-
1980s). This factor, he finds, accounts for about half
of the increase in
inequality from 1975 to 1985. In addition, the income differential between
the in-work and
the not-in-work groups increased, notably after 1985.
Second, a rise in earnings inequality among those in
work. Around 1975,
among full-time workers, the earnings of those in the top decile were about
three times the
earnings of those in the bottom decile, whereafter the ratio
increased steadily to about 3.4:1 in 1990. Third,
the distribution of income
among those not in work. During the 1980s, while the value of non-state
pensions
continued to follow earnings trends, the relative value of state
pensions and benefits declined when the basis
for upgrading changed from
earnings to prices, the result being a relative deterioration in the income
position
of those depending most heavily on state pensions and benefits.
The story is complicated, with a range of economic, demographic, and
political factors contributing to the end
result. Policies clearly play a role.
Among the relevant policy factors is that of changing rules for the
upgrading of social security benefits. This accounts for a good deal of the
increasing inequality among those
not it work, and for some of the
increasing difference between those in work and those not in work. How
much
exactly this factor contributes to the total explanation is not clear, but
for the present purpose it is enough
to say that it is significant. The new
British trend towards inequality cannot be put down to ‘automatic’
forces of
long-term stability in the distribution of income; it is to some considerable
degree induced by
changes in policies. The change from earnings to prices
as the basis for upgrading state benefits was seen,
when it was made, as a
marginal adjustment. Over time, the consequences have
accumulated to a
very significant effect on inequality in Britain. This was possibly not
anticipated, and this
experience would then seem to demonstrate in a rather
illuminating way that the distribution of income is
indeed sensitive to even
marginal adjustments in policy.
In a comparison of income inequality trends in five countries during the
1980s, Fritzell found stability in
Germany and Canada and a trend towards
increasing inequality in Sweden, Britain, and the USA.6 Redistributive
policies are found to contribute to
explaining both stability and change. In
all five countries, an increase in factor income inequality was
observed; in
Germany and Canada this was counterbalanced by changes in policies
which pulled in the opposite
direction, hence final stability; in Sweden,
Britain, and the USA, changes in the distribution of factor income
carried
through to the distribution of disposable income, and redistributive policies
were reshaped so as to
add to this effect, rather than, as in the two other
countries, to neutralize it.
In an earlier study, I compared income redistribution in Norway and
Sweden, using carefully co-ordinated survey
data for 1982.7 That
comparison turned out to
have a quasi-experimental flavour: while factor
income was identically distributed in both countries,
disposable income was
more egalitarian in Sweden than in Norway. This suggests that the
explanation of the
difference in the end distribution lies ‘between’ factor
and disposable income, i.e. in political forces which
redistribute factor
income ‘afterwards’, rather than in economic forces which determine the
original
distribution of factor income (although some uncertainty about
second-order effects on factor income remains).
There also turned out to be a simple difference in the structure of
policies. In Sweden, transfers as well as
taxes were on a higher relative
level; in Norway, both had a more progressive distribution across income
groups. Through cross-simulation, the Swedish levels of transfers and taxes
were imputed into the Norwegian
system while the Norwegian
distributional profiles were maintained, and the Norwegian profiles were
imputed
into the Swedish system while the Swedish levels were maintained.
Both simulations, i.e. including the one in
which only the levels of transfers
and taxes were manipulated, had the effect of changing the distribution of
disposable income, compared to the pre-simulation distributions. As in the
actual experience in Britain, this theoretical simulation again shows that the
distribution is indeed sensitive
to marginal changes in policies, in this case
also to marginal changes in the levels of transfers and taxes
within constant
profiles.

STRUCTURAL FACTORS

From the mid-1960s to the mid-1980s, Finland experienced what is perhaps

the most dramatic case of reduced
income inequality in any industrial
nation and was transformed in this respect into the most egalitarian of the
Nordic countries. During the period, the Gini-index for the distribution of
equivalent disposable income fell
from about 0.32 to about 0.20. Most of
this change occurred during no more than ten years, from 1966 to 1976,
whereafter the distribution has remained stable. This, then, is a remarkable
case of a very considerable shift
in the distribution of income, occurring
over a short period of time, and with the new distribution becoming
settled
and stable at its new level, at least for a considerable time.8
Having established this remarkable development, Uusitalo has sought to
explain the change in inequality in
Finland through an analysis which
considers three sets of explanatory factors: transfer and tax policies,
incomes policies, and structural changes in society. This study covers a long
period, observes trends in both
factor income and disposable income, takes
in the effects of structural change in society, and considers policy
effects on
the factor income distribution (incomes policies) along with the effects of
transfers and taxes.
Second-order effects ‘back’ to factor income would be
captured in this analysis, although not identified
specifically. Through its
richness, this is an exceptionally informative study in respect to both the
major
problems in the standard method, its narrowness and the static
limitation.
 There were important changes in socio-economic structure in Finland
during the period, such as in the
occupational composition of the work-
force, the regional pattern of residence, and
household size and
composition. These structural changes turn out to be of relatively little
importance in
explaining the change in income inequality. They changed
the relative income position of some socioeconomic
groups, but these
changes tended to pull in different directions and to balance each other out.
Finland did not
become more egalitarian in income terms as a result of
Finnish society itself becoming more egalitarian in some
structural sense.
Rather, income equality was generated without structural pressures pushing
systematically in
that direction, something which is an important finding in
its own right.
The change in final inequality was a result of changes both in the
distribution of factor income and of
redistributions of factor income into
disposable income, both elements pulling in the direction of equality.
The
change in the distribution of factor income is attributed in large measure to
incomes policies. In a period
of ‘solidaristic wage policy’, factor income
inequality was reduced; when this policy was relaxed, factor
income
inequality stabilized. Here, there are three additional important findings.
First, when incomes policies
were relaxed, factor income did not revert
back to its previous distribution but stayed at its new
distribution. Second,
at no time was a trend of increasing inequality of factor income observed.
In a period of
strong redistributive pressures, there were in other words not
countervailing forces of sufficient strength to
undermine redistribution in
disposable income by pulling factor income in the direction of more
inequality.
Third, in the period of reduced inequality of factor income, the
explanation is not underlying socio-economic
change, but instead policy
interventions in market processes, in this case in the form of incomes policy
agreements between government and labour-market partners. Not only
changes in final inequality, therefore, but
also in factor income inequality
turn out to be politically induced rather than structurally determined.
Although changes in the distribution of factor income did push final
income in the direction of equality, the
most important of the explanatory
factors considered is transfer and tax policy. Changes in these policies had
a
strong equalizing impact. This effect was at work while factor income
inequality was reduced and continued to
work after factor income
inequality had stabilized. Redistribution occurred through taxes as well as
transfers,
between the working and non-working populations, and within
the working population.
As with the story of inequality in Britain, the story of equality in Finland
is a complex one with a range of
forces interacting. Had we studied this
story within a standard method framework, we would have found the
change in inequality to have been linked to policy. Having studied the story
within a broader framework that
overcomes the limitations of the standard
method, Uusitalo still finds the same link between policies and
distributions. The result of the more sophisticated methodology is to
strengthen, and not to weaken, standard
method conclusions.

LIFETIME REDISTRIBUTION

Equalization of income implies a transfer in favour of those in need. Being

in need, however, is not
necessarily a permanent situation for persons but
may be limited to certain periods of the life course, for
example during
unemployment or in old age. To the degree that such periods were
predictable, the relevant needs
could be accommodated through insurance-
type arrangements to which people in some periods make contributions
and
from which they in other periods draw benefits. Modern welfare-state social
security systems are in
considerable measure insurance-type arrangements,
although not necessarily explicitly so.9 If social security systems were
organized as perfect
systems of insurance, it would all be a matter of
intraperson redistribution over the life course with no
interperson
redistribution except what might result from accidental misfortune. If all
redistribution were
intraperson, there would be no redistribution between
persons in lifetime income. Standard method results add
up the cross-
section manifestations of interperson and intraperson over-time
redistributions.
Obviously, it is not possible to observe future effects of current policies
directly. What has proved possible,
however, is to simulate life-course
effects implicit in present policies. Harding has developed a
‘dynamic
micro-simulation model’ for this purpose applied to the case of Australia.10
The model simulates the life of a cohort ‘born’ in
1986 through a life span
of 95 years with regard to events such as birth, marriage,
divorce, death,
children leaving home, labour force entry and exit, and so on. The
simulations are made year by
year over the life span of the cohort, on the
basis of known probabilities from existing statistics and today’s
rules in
transfers and taxes. A simulation of this kind does not predict what will
happen in the future—that
obviously depends on future decisions and
events—but shows what would happen under a steady-state assumption.
The question the study seeks to answer is the following: if the demographic,
labour force, income, and other
characteristics of the population, and all
government policies existing in 1986, remained unchanged for 95
years,
what would the distribution of income be like at different points in time and
what income redistribution
would be achieved by government programmes
over the life course of the cohort?
The characteristics and limitations of the study are essentially those of
standard method analysis, except for
the introduction of the time dimension:
the analysis is limited to cash income and to first-order
redistribution,
income is recorded as equivalent income for persons, only cash transfers are
considered. On the
tax side, the simulations incorporate the level of income
taxation which is needed to finance actual cash
transfers. The simulation
does not break new ground on the concept of redistribution, but it
overcomes the
cross-section limitation of standard method research and
thereby shows the degree to which cross-section
results can be generalized
to a lifetime perspective.
Main findings: the tax-transfer system in Australia has a profound effect
on lifetime income. This works in two
ways. Income is redistributed both
over the life course of persons (from the years of work to the years of
retirement) and between persons during their life courses. On average for
males, about 45 per cent of the
income tax paid was returned to the same
persons at another point in the life cycle. The remaining 55 per cent
was
absorbed by interperson redistribution. The winners are, first of all, women
compared to men. While men on
average get less back in transfers than they
pay in taxes, women on average break even up to about the age of
60 and
then become net winners.
When groups defined in terms of lifetime standard of living are compared
and transfers and taxes are
accumulated over the life course, those with a
lower standard of living are the winners and those with a higher
standard of
living the losers.
In a simulation for Britain, Falkingham and others separate interperson
from
intraperson lifetime redistribution of cash transfers and taxes.11 Their
conclusions are similar to those of Harding. The British system is found to
redistribute income both in the interperson and intraperson sense. Of all
direct transfers received over the
life course, 62 per cent are found to have
been paid for by the same persons at various times during the life
course,
while 38 per cent represent transfers beyond what the recipients have
contributed. The profile of the
net benefit across lifetime income groups
(and of gross benefits and gross taxes for that matter) is
progressive: the
lower the lifetime income, the relatively higher the lifetime transfers and
the relatively
lower the lifetime taxes.

CONCLUSION

My question in this chapter is elementary: is there a link between

tax/transfer policy and the distribution of
income, or is there independence?
We do not start from scratch in suggesting that there is a link. This
interpretation is supported by a mass of research. The trouble, however,
with much of this research is that it
has been confined to a methodology in
which the actual distribution of income, which is recorded correctly
‘after’
transfers and taxes, is compared to a hypothetical (and, admittedly,
unrealistic) distribution ‘before’
transfers and taxes, rather than to the true
counter-factual distribution under an alternative policy regime,
which,
obviously, cannot be observed empirically. There are good reasons to
believe that this methodology might
tend to exaggerate the magnitude of
effects, but there are not good reasons to believe that the
standard method
should consistently or systematically be wrong as to the existence of effects.
It can be argued that the standard method is unrealistic since it considers
redistribution as a matter of
global rather than marginal effects. However,
there is ample evidence of effects of marginal changes in
policies. It can be
argued that the standard method is unrealistic since it ignores the process of
income
formation ‘before’ transfers and taxes. However, the dramatic case
of Finland shows that lasting changes in the
distribution can be made
mainly through policies and without underlying structural
trends towards
equality. It can be argued that the standard method is unrealistic since it
confuses
cross-section and life-course effects and fails to distinguish
between interperson and intraperson
redistributions. However, dynamic
simulations show that although effects recorded in cross-section analyses do
contain both types of redistribution, and hence may exaggerate true
redistribution, the recorded effects are
not only intraperson but also contain
an element of ‘real’ interperson redistribution of lifetime income.
It used to be a broadly held view among economists that the distribution
of income in society tends towards
stability. The whole issue of income
distribution was then not considered to be of much interest at all, not
for
research and not for policy. That theory has now been refuted as to its
descriptive implications. However,
although income inequality has become
established as a sensitive indicator, it may not have become regarded as
established that it is sensitive to policy intentions. There is remaining
uncertainty about
how sensitive income inequality might be to reform
strategies, but there should be no remaining good
reasons to doubt the
existence of a link. It is not true that the distribution of income is merely a
surface manifestation of ‘deeper’ structural forces and that this distribution
could only change as a result of
changes in ‘deeper’ structures. Modern
economies are complex, but things are not that complicated.
Should income inequality be modified? Perhaps, perhaps not—depends
on the inequality and on other
circumstances. That, however, is not my
concern here.12 The question for the moment is not should, but could? To
that question, the answer is
that if income should be equalized, it could be
equalized.
1 Smeeding et al. 1990. It has also been shown in more limited

comparisons, for example of the Nordic countries (Ringen 1986, 1991;

Gustafsson and Uusitalo 1990a,
1990b; Ringen and Uusitalo 1992) and of
Australia and New Zealand (Saunders et al 1991).
2 I refer to studies by Jenkins (1991), Atkinson (1993), Coulter et al
(1994), and Goodman and Webb (1994) for Britain, to the works cited
above for Finland, Sweden, Australia, and New
Zealand, to Ringen (1993)
on the composition of household income in Norway, to Fritzell (1993) for a
five-country
comparison, and to ‘official’ poverty estimates in Britain and
the USA.
3 Atkinson et al 1995.
4 There are also questions of social efficiency: even if transfers and taxes

were
successful in redistributing income, there could be costs in the form of
inefficiencies which could make the
policies prohibitive. Such costs could
arise through behavioural responses (e.g. work inefficiency) or through
attitudinal responses (e.g. cultures of dependency). I have considered such
‘moral hazards’ elsewhere (Ringen
1987, 1995). Although the matter is
complex, the following seems to be a good rule of thumb. By ‘cost’, I
understand ‘welfare loss’. (A budget expenditure is not in itself a cost to
society.) There are costs; cost-free
redistribution would be too good to be
true. Therefore, the cost side would have to be considered in the
formulation of policies. However, aggregate social costs are often low, for
example because income and
substitution effects balance each other out. (I
would not accept a definition of cost which includes only
substitution
effects.) Likewise, while some policies may encourage dependency, others
may stimulate
self-reliance. If costs are unavoidable, policies can be
designed so as to reduce potential costs. Nevertheless,
there may be trade-
offs between redistribution and social efficiency. Although cost-free
redistribution is
unlikely, effective redistributive reform (which is about
modifying inequality and not about creating equality)
is possible without
prohibitive costs.
5 Atkinson 1993.
6 Fritzell 1993.
7 Ringen 1986.
8 Uusitalo 1989, and personal communication on trends from 1985 to

1993. In the
most recent observation, from 1992 to 1993, there has been a
modest movement in the direction of inequality,
which may or may not be
the beginning of a new trend.
9 Hills et al. 1993.
10 Harding 1992.
11 Falkingham et al 1993, using a model population rather than a cohort.
12 There are four situations in which policy intervention for the

equalization of
income would unquestionably be correct: (i) the existence of
poverty which could be reduced or eliminated through
redistribution; (ii)
the existence of indefensible pockets of relative deprivation which could be
reduced or
eliminated through redistribution—relative deprivation among
children would be a case in point; (iii) the
existence of degrees of inequality
which are objectionable to community or threatening to cohesion or
efficiency;
and (iv) strong and rapid shifts towards inequality, or threats
thereof.
 6

THE CIVIC SPIRIT

At the end of 1989 and the beginning of 1990 I found myself in Chile,
immediately after the successful
congressional elections and during the
days when the Chileans were finally able to feel confident that
democracy
would be restored. We who have grown up in Western Europe after the
Second World War have never
experienced the loss of freedom. That is our
good fortune, of course, but we should also be aware that we do
not know
the physical reality of oppression. This was brought home to me during a
long evening in the
company of a dignified Chilean, an elderly lady whose
husband had disappeared during the dictatorship, whose
son had been killed
by having his throat cut in broad daylight in central Santiago, and whose
daughter had been
in exile until the year before, and who was now herself
elected Diputado for the Partido por la
Democracia, one of the parties in
the Concertacion which beat the Pinochet crowd in the
elections. As she
told her story, in a soft-spoken way which was unreal to me, this visiting
North European
learned something about democracy, and in particular the
danger of its absence.
In The Possibility of Politics, I tried to analyse the welfare state in such a
way as to comment on
the feasibility of piecemeal, step-by-step, reformist
policies. In intellectual circles, reform is suspect as
seen from both left and
right. I was therefore delighted, being a middle-of-the-road man myself, to
be able to
pass a favourable verdict on the power of reform.
I have always associated political reform with political democracy. I
believe that democracy generates reform
because it will not be accepted to
let problems lie where they fall. I also believe that democracy limits
politics
to reform since majorities will accept revolutions to democracy but not
revolutionary shifts within
democracy. In a sense, therefore, the analysis I
was conducting was about reason in the muddled arrangements of
the
mixed-economy democracy. I did not feel, however, that I could extend the
analysis at that time to the issue of democracy. ‘If it were not’, I wrote in an
introductory section,
‘that I already have more than a reasonable share of
broad questions on my plate I might have been tempted to
add, Is
democracy possible?’
I resisted the temptation then, wisely I’m sure. Today, however, I intend
to give in and to take up the
question I then felt I had to set aside. After all,
this is an occasion on which it is accepted—perhaps even
encouraged—to
address issues which would in other circumstances be frivolous. So, then,
does democracy work?
I have a professional interest in the application of social science to issues
of concern in society. Applied
research is intended to contribute to
something practical. What social research contributes to is perhaps not
always clear. I want to argue that we have a role to play in the workings of
democracy and that a programme for
applied social science can be
formulated in terms of fulfilling this role.
The last few years have seen revolutions of democracy throughout the
world. I have in mind, obviously, Eastern
Europe, but also the Phillipines,
Argentina, Zambia, South Africa (hopefully), Nicaragua, and certainly
Chile—once the democratic pride of South America, then the world’s most
despised dictatorship, and now again a
working democracy. I think of the
fall from grace of the outlandish idea in Africa that one-party rule is
necessary because of African tradition. I have in mind the eradication
(almost) of military rule in Latin
America and the many attempts, although
some unsuccessful so far, in the direction of democracy in, for
example,
China, Nigeria, Burma, Bangladesh, Kenya, Côte d’Ivoire, Algeria, and not
least the tragedy of Haiti,
where the first free elections after almost 200
years of independence were soon enough set aside. Having
visited that
country several times and been captured by the cultural richness of a proud
people suffering no
end of political corruption and material deprivation—
this is a land where the very physical soil has been
plundered—I feel that I
have experienced the degradation and destructiveness of bad government in
its extreme
form. Those who need to see for themselves the price of non-
democracy in the elementary life chances of
ordinary people, let them go to
Haiti.
 The logic of applied research is that the researcher puts to use his or her
tool-box for resolving some problem
which arises from the conduct of
practical affairs. There are odd problems which serious scholars would not
be bothered with and there are wicked problems which scholars of integrity
would not touch, but generally, in applied research, we take on problems
such as they are presented to us.
In the social sciences, problems are often presented to us by ‘decision-
makers’. That is as it should be, but
as a programme it is too narrow. The
act of decision-making, certainly in a democracy, is only the final step
in a
long process which leads up to the moment of truth when it is decided to do
A and not to do B. A proper
ambition for the social sciences is to respond to
the whole range of problems which arise in this entire
process and not only
to those of the final stage. The clients of applied social research are not only
decision-makers, but rather, and primarily, the public—the women and men
who participate in politics as
citizens. We should take up not only those
problems which are articulated by decision-makers but any problems
which
present themselves from the process itself, however they may or may not be
articulated. We should work
for the political process rather than for the
politicians.
The democratic revolution is enormously positive, but also chilling. The
most troubled peoples of the world are
turning to democracy for solutions
to their problems. I am quite convinced that they are right in so doing, but
the burden the democratic system of government is now asked to carry is
staggering. More is expected than can
be delivered. The coming of
democracy to the former Soviet Union is not exactly celebrated as a
triumph. In
Czechoslovakia, the Velvet Revolution tore the Federation
apart. In the first free elections in Poland in 1991,
and likewise in Zambia,
only about 40 per cent of the electorate chose to use the right to vote which
they had
just won.
In the established democracies we may feel safe, but I suggest we should
not be complacent. It has happened
before that democracy is lost. We
cannot take it for granted that our democracies are strong enough to stand if
the democratic experiments around us fail. On the European scene, some
politicians seem to believe in
supernational power without democratic
institutions. ‘[D]emocracies have always been, and still are, failure
prone.
They were short-lived, and by all accounts ill suited for survival.’ Thus we
are reminded by Giovanni
Sartori of the historical experience. Horrors we
thought were confined to the scrap-heap of history in our part
of the world,
have proved themselves to be alive and well. We need at this time to
be
deadly serious about the importance and difficulty of democracy and to give
careful thought to its meaning
and practical mechanics.

DEMOCRACY

As soon as people live in society, there will be a need for collective

decisions on things which have to be
done communally and on norms for
co-operation. This is true of small societies, such as families, and it is
true
of large societies, such as those we call nations—and for that matter of the
global society.
In large societies we use the term politics to denote the system for
making and implementing
collective decisions. Democratic systems have
one thing in common with all other political systems and are in
one respect
different from all alternatives. What all political systems have in common is
that they are
mechanisms for producing decisions. What distinguishes
democracies from other political systems is the way in
which decisions are
produced.
Both aspects are important. We often, and rightly, judge democracies by
their specific quality, how decisions
are made. But we also need to consider
them in terms of what they have in common with other systems, their
ability to produce decisions. Democracy is not just a way of sharing
responsibility, it is also an arrangement
for getting on with the job.
Let me try to explain further through some simple and rather formalistic
observations. I split the larger
process in two, first the process in which
collective decisions are made, and then the process whereby they are
implemented. Both these processes can be understood as movements, so to
speak, back and forth between the
individual and collective levels in
society. In the first instance, individual preferences aggregate into a
collective will, as when citizens cast votes in a referendum. In dictatorships,
this is simple because the
preferences of a single person, or a small clique,
are imposed as the collective will. In a democracy, it is
complicated because
the collective will has to be forged from an enormous range of individual
preferences, all
of which ideally count equally, and many of which will be
in discord.
 In the second round, the collective will disaggregates into individual
actions to carry it out. The simple
example is taxation. We decide
collectively on a certain level and distribution of
taxes and this decision is
carried out when each of us pays the tax which is expected. The collective
will is
usually followed by various forms of encouragement and control to
comply, but in the final analysis it comes
down to the individual.
Parliaments and governments can only make two kinds of decisions, about
rules of
conduct and about the spending of money. Such decisions will
come to naught unless citizens adapt their
behaviour to the collective
expectation. No amount of government spending on health services could
have any
effect whatsoever if innumerable doctors, nurses, technicians, and
administrators were unwilling to carry to
the patients the services which the
spending is intended to buy.
In the process of making collective decisions, democracies give each
citizen the same say. This is the
democratic ideal, pure and simple. It is a
simple ideal to express but strong assumptions are required in order
to see
how it could be reality. The most important of these assumptions are, first,
that all citizens are well
and equally informed, secondly, that they have the
same opportunity and ability to make their voices heard, and
thirdly, that
the decision-making process is without distortions which amplify some
voices and mute others.
As strong as these assumptions are, they are not Utopian. They are the
rules we apply to voting: one person one
vote, free access to the ballot, and
meticulous care in the counting of votes so that all votes are counted and
each only once. Through political campaigns, we go to great lengths to
inform voters about their options, and
we accept that everyone is
sufficiently informed to make a responsible choice (except that we persist in
denying children the right to vote). Very considerable effort, organization,
and expense is invested to protect
the integrity of the vote. However,
collective decision-making is not limited to the few occasions on which we
vote, but occurs every day, in economic as well as political life. In all other
situations than voting, we are
not remotely near giving everyone the
possibility, never mind the same possibility, of making their voices
heard.
There are differences in access to education, eloquence, connections,
money, and—the top price—the
30-seconds slot in the evening news on
television. The decisionmaking process is contaminated with no end of
tricks to make some voices sound larger than life, in the form of organized
interests, political action
committees, and bishops pushing particular points
of view from behind purple
dog-collared shirts. Beyond the formal elegance
of the vote, the democratic record of even the best of
democracies is dismal.
In the process of implementation we should expect decisions
democratically made to be accepted and complied
with by all. Since
everyone has had the same say, each decision reflects the best possible
compromise. Our
disagreements we bring to bear in the process of reaching
collective decisions and we can then fight as lions.
Once the battle is over
and a decision has been reached, we agree to accept the verdict because the
fight has
been fair, and we move our disagreements on to new issues which
are not yet resolved. ‘In a democracy,’
explained Max Weber in an
exchange with General Ludendorf, ‘people choose a leader in whom they
trust. Then the
chosen leader says, “Now shut up and obey me.” Later the
people can sit in judgement. If the leader has made
mistakes—to the
gallows with him!’
To a remarkable degree we do accept what is expected of us. We drive on
the same side of the road, we go to
work, we do not beat our children. But
non-compliance is still a problem, so much so that vast, unpleasant, and
wasteful bureaucracies are needed for the purpose of control. It may not
take elaborate controls to keep
drivers on the same side of the road, but it
does to impose speed limits, not to mention bringing in taxes. The
reason is
obvious. Collective decisions are compromises between conflicting
interests. It will always be
advantageous for some to avoid the
consequences. To ask for compliance, is to ask some people to act contrary
to their own benefit.
The two processes of democracy are closely linked. If we want
compliance, we must make sure that decisions are
made fairly. How can we
ask people to comply if what they are asked to comply with reflects special
interests
rather than the general will? If we want fairness in decision-
making, we must assure compliance. How can we ask
people to participate
in decision-making if what is decided is treated with indifference?
However, although
there is much interdependence in this system, it does not
propel itself to perfection. Compliance and fairness
are each other’s
necessary but not sufficient conditions. For democracy to work, citizens
must want to make it
work, and they must want that rather desperately.
Some general points follow from this discussion. First, the importance of
the vote. This institution is
important because when we use the vote, we do
in fact all have the same say, and it
is important because it symbolizes the
reality on earth, if I may so say, of the heavenly ideal. The low voting
participation in many elections is a sorry sign for democracy. Social
scientists sometimes do not have much
patience with formal procedures, but
instead want to understand the realities behind the formalities, so much
so
that formalities may be ignored. Democracy is a complicated system, but its
core is the elementary
simplicity of the universal vote.
Second, the importance of equality. Democracy is based on equality, be it
in the distribution of resources
which influence decisions or the distribution
of resulting benefits and burdens. ‘The assumption’, writes
Robert Dahl,
‘that a rough equality in the distribution of resources facilitates a rough
equality in the
distribution of power or control over government is common
in political theory since the classical Greek era.’
Tocqueville, in De la
démocratie en Amerique, found a high degree of social and economic
equality to
be necessary for democracy. Powerful egalitarian arguments
have been advanced in the philosophical literature,
for example by Tawney,
based on the moral principle of the equal worth of all persons, or by Rawls
in the
logical experiment of the ‘veil of ignorance’ which suggests that
opposition to equality is necessarily based
on vested interests. Still, the old
observation that democracy does not work unless everyone is included is
perhaps the strongest of all arguments for the principle of equality.
Third, the importance of the individual citizen. Democracy is a big
system, but it is not a mass phenomenon in
which individuals disappear.
Citizens depend on democracy for their freedom and democracy depends on
the
participation of citizens. There will be no democracy unless citizens
vote and otherwise take part in the
struggle over decisions. Nor will there
be democracy unless citizens accept and carry out the decisions which
are
made. Democracy is not possible unless citizens want it, and it is not
possible unless citizens behave
democratically.

SCIENCE

In 1936, Ragnar Frisch, the pioneer in the theory of planning, who was,
with Jan Tinbergen, to be the first
Nobel Prize winner in economics, wrote a
small memorandum which he entitled Den prinsipielle sondring mellom sak
og vurdering. There is a richness of nuance in this
Norwegian title which it
is difficult to capture in English, so when I offer The theoretical distinction
between fact and value as an English translation I am not doing full justice
to the original.
Professor Frisch was concerned with the relationship between science
and politics and with what social science
could do for politics. He
proceeded to define this relationship around the distinction between fact
and value.
Values enter science, he found, through the assumptions upon
which analysis rests. Value problems can and
should be resolved in
assumptions. The remaining problems are then factual, and scientific
analysis proper can
start. Once assumptions are given, it is possible,
although difficult, to achieve objectivity of analysis in
the sense that results
properly derived must be accepted by all who are competent to understand
the research.
The politician also, still following Frisch, confronts both types of
problems. The choice of goals is a value
problem, as is the decision about
which means are legitimate. Once this is established, however, the choice of
means among those which are allowed is a problem of optimality which can
be resolved through analysis.
From this Frisch went on to suggest what amounted to a division of
labour between politics and science.
Politicians should take care of the
value element, they should set goals and establish the boundaries for what
can and what cannot be done. The experts should then step in, take the
value premisses handed to them by the
politicians as their assumptions, and
find the most effective means.
Frisch stuck to this model throughout his life and never lost his
confidence in the ability of science to find
the right means once goals were
given. But he became increasingly impatient with the politicians who never
seemed able to do their part of the job. In his despair he constructed an
elaborate questionnaire which he
intended to administer to the members of
Stortinget, the Norwegian parliament, and through which he
would force
them to choose a consistent set of goals so that he and his colleagues could
get to work on their
computers and find the best means. This project never
came to fruition, and no doubt that was just as well.
Any social science student of my generation who suffered years of agony
at university under the wretched
‘positivism debate’ will shy away from
Frisch’s distinction between fact and value, but that is a trivial
matter which
there is no need to brood over here. Frisch wrote his memorandum when
he
was asked by a group of leading men in the business community to direct
an ambitious study of ‘economic
structures’. There had been an element of
expert rule thinking in the deliberations in this group and perhaps a
suggestion that the economic conditions of the nation might have been
better had expertise been trusted more
and politics less. Such ideas were
hardly surprising given the impotence of political authorities at the time
to
face up to the overwhelming problems of stagnation and unemployment,
but any hint of superpolitical
authority was, of course, an extremely
sensitive matter in Europe in the mid-1930s. Frisch was a radical in
politics
and he felt a need to formulate a position in which he disassociated himself
from any suggestion of
expert rule. With our eyes it might seem that he
handed over half of the political process to the experts, but
in the context of
the time, what he did was to defend the primacy of politics over expertise.
As all first-year students know, the strict distinction between fact and
value is not philosophically tenable,
and the idea that the value problem in
the social sciences can be cleared away by being confined to assumptions
had been rejected at the time Frisch wrote his memorandum, notably by his
colleague in Sweden, Gunnar Myrdal,
in a dissertation published in 1929
(and republished in revised form in English in 1953 under the title The
Political Element in the Development of Economic Theory). Those who
want to criticize my hero on this
point will have a field-day, but they will, I
think, miss the essential. By suggesting a division of labour
between
politics and science, Frisch claimed not only that there are certain problems
which only science can
solve and for which scientists should take
responsibility, but also that there are problems which science
cannot solve
and which scientists should therefore not take on. Exactly where to draw the
line of demarcation
may be in dispute, but that there is a line was right then
and remains right today.
All forms of political decisions are political, be it the choice of goals or
the use of means. No line of
demarcation there. There is, nevertheless, a
factual element in politics which is, in principle, non-political
and upon
which rational political action depends. This element is information. For
citizens to
participate in politics, they must be informed. Before political
decisions can be made, there must be
information on which to act. Having
abandoned the line of demarcation between fact
and value, I suggest an
alternative line drawn between information and decision; between, on the
one hand, the
questions of how things are and what can be done, and on the
other hand, the question of what
should be done. Let me, further, push to
the background the politician and pull to the foreground the
political
process. Let me suggest information as the common ground on which
science and politics can meet
without scientists playing politicians or
politicians playing scientists; information is what politics needs
and science
can provide. And let me, on this basis, and otherwise in Frisch’s spirit of a
division of labour,
try to redefine the job society should expect the social
sciences to do.
First, as social scientists we should accept responsibility for informing
the democratic process. We should
accept this responsibility (although
shared with, for example, journalists and educators) because we are the
custodians of the methodology which is best suited for the job of
establishing how things are and what can be
done. As has been pointed out
by Sten Johansson, the political methods of discourse, struggle, conflict,
and
compromise are useless for resolving questions such as: Is public health
improving or deteriorating? Is crime,
violence, and child-abuse on the
increase? What are the effects of changes in social security legislation on
poverty? How widespread is the problem of stress in the workplace? Do tax
cuts stimulate effort and initiative?
No debate, however sophisticated, can
produce reliable answers to questions such as these. These are empirical
problems and empirical problems should be resolved by analysis. To great
benefit for enlightened politics, we
do have generally accepted ways of
answering some persistent questions of this kind, for example about
unemployment and inflation, but we need authoritative information on a
much wider range of social and economic
trends.
I want it to be added that there should be no delusion about the limits of
analysis for producing irrefutable
answers to empirical questions. The
answers we produce will always be marred by uncertainty and there will
always be more questions asked than answers given. What I believe we can
and should say, however, is that there
is no alternative method to that of
careful analysis which can do the job better. Had we wanted, in 1957, when
Harold Macmillan told the British people, ‘You’ve never had it so good’, to
ascertain whether he was right, we
should not have organized a round table,
we should have consulted the statistics or
started a research project.
Second, we should not only accept responsibility for informing the
political process, we should claim this as
our authority. The political
struggle does not stop at the domain of information. The concern of the
politician
is, rightly, to be elected or re-elected, to win power or to stay in
power. Politicians who can doctor the data
to their advantage will do so, and
it is in the nature of competitive politics to spread disinformation. Since
we
cannot outlaw such behaviour without outlawing democracy—indeed, since
strong democracy depends on
ruthlessly aggressive politics—there is a need
to assure that the political process is independently informed.
Decisions
taken from distortions will be the wrong decisions, but much more
importantly, citizens’ control over
politics is not possible unless citizens are
informed. Hence, I do not agree with the great Henrik Ibsen who,
when
challenged, in a poem in 1875, wrote, ‘I prefer to ask, it is not my calling to
answer.’ Our calling is,
precisely, to answer.
In Britain today, government agencies retain ultimate control over the
publication of research results from
projects they have contracted. In the
social sciences we are dependent on government contracts and we therefore
do not have freedom to publish our results. This is a miserable state of
affairs which should be brought to an
end. It should be for the scientific
community to decide, through its system of peer review, what and how to
publish, and the power of the purse should make no difference to this.
Contract research for government
agencies is on behalf of the public and
the public should know what the research finds. It does not matter that
difficulties over publication usually do not arise, because it matters very
much in the few cases when
difficulties do arise—and there is always the
possibility of self-censorship.
Third, although we should claim the domain of information as ours, we
should, as social scientists, not
transgress into the domain of decisions. As
citizens we have the same rights as other citizens, but as
scientists we have
no special rights. We have some expertise with regard to the questions of
how things are and
what can be done, and therefore responsibility to serve,
but we have no competence on questions of what should
be done, and
therefore no special right to influence. If we use our expertise to inform the
political process,
we strengthen democracy on one of its central
assumptions, that of an informed
citizenry. If we claim authority which we
do not have in order to gain influence, we weaken democracy on one of
its
other assumptions, that of a decision-making process without distortions. It
is easy to be critical of
others who use unfair means to win influence, for
example money. We need to be equally critical of our own
ways. The
academic community is not a nobility with special privileges. We should be
activists, certainly, and
when necessary prepared to confront politicians in
the most forceful manner, but activists on the basis of
legitimate authority.
As social scientists we have no other expertise than information to offer; we
cannot do
surgery, we cannot build bridges, we cannot explode bombs.
Personally, I feel that the job of information is
important and sufficient, but
even if I did not, I should be made to understand that I have no business
trying
to convince the world that I am an expert in the rights and wrongs of
political decisions. Hence, I do not
agree with the great Karl Marx who, in
his Theses on Feuerbach in 1845 said, The philosophers have
only
interpreted the world; the task is now to change it.’ Our job is, precisely, to
interpret the world, and, precisely not, to change it.

THE CIVIC SPIRIT

Above I have sketched the outlines of perfect democracy: informed citizens

participating on equal terms and
accepting the outcome of fair compromise.
Could we make democracy work perfectly? The answer is that we could
not, and, further, that we should not so much as think the thought.
Perfect democracy is an illusion, as is the idea of perfection generally.
There will never be equality in the
distribution of means of influence. Even
if equality were our only concern, it would be beyond our power to
create.
And there are other concerns which may stand in the way of the egalitarian
ideal, such as the danger to
liberty in the attempt to achieve it, as was
Tocqueville’s warning, the economic cost which might result, as
was Arthur
Okun’s warning, or the avoidance of the kind of dreariness in society which
is a remembrance of what
used to be East Berlin.
Not only is the idea of perfection illusory, it is also wrong and terribly,
terribly dangerous. Imperfection is
not a temporary inconvenience on the
road to the promised land; it is a normal and permanent feature of all things
people do. Visions of perfection make today insignificant
because what we
do now matters only to the degree that it advances the ultimate goal. Under
such theories,
democracy is unnecessary and science has no chance. From
the experience of the collapse of communism it must be
possible to say with
authority, and from the experience of the rise of religious fundamentalism it
must be
right to say with emphasis, that the quest for perfection in human
affairs is destructive of civilization and
of dignity. ‘My dogma’, writes
Anthony Storr, ‘is that all dogma is suspect.’
 Rather than trying to turn democracy into dogma, we should seek
practical, reasonable, working democracy. I
have suggested that success in
this depends much on citizens. We must want democracy and we must
behave
democratically. We need democratic laws and democratic
institutions but these are only a framework. It is then
for citizens to take this
framework and breathe life into it, or destroy it, by the way they use it.
There are theories, according to which the pursuit of self-interest on the
part of individuals adds up to good
society. I do not know if there are
practical circumstances in which this works—I suspect not, it is too good
to
be true. Democracy is certainly not this kind of system. If citizens were to
approach democracy in the
what’s-in-it-for-me spirit, it would collapse
outright. There would be no reason to vote since each vote
matters next to
nothing; there would be no reason for those in advantage not to manipulate
the policy process;
there would be no reason for compliance. There would
not even be a semblance of collective will, since, as
Kenneth Arrow has
shown, individual preferences do not add up to a single collective choice.
No institutions
can make voting a good investment of time, and no
institutions can eliminate the element of sacrifice in
compliance.
Well-organized society depends on rules and the policing of rules, but the
role of rules is ambiguous. In a
dictatorship, rules are a part of the
repressive apparatus. In an open society, rules should reflect more than
impose agreement and have as their main function to prevent disruptive
free-riding problems on the fringes. In
democracy, where the whole idea is
that citizens regulate themselves, the role of rules must be limited. Some
countries have made voting compulsory but this will do no good if people
do not want to vote. Rules about compliance are necessary but there would
not be much left in the democratic
ideal if there were no more compliance
than was enforced.
In a nutshell the problem is this: democracy assumes equality. Equality
we shall never have. Reasonable,
working democracy therefore depends on
the full force of inequality not being brought to bear on the democratic
process. This cannot be done with rules alone; it would not work and would
anyway be self-defeating. It must in
great measure be done through restraint
on self-interest. We must have integrity in political leadership; we
must
expect those in wealth to refrain from using the power of their wealth as a
political tool; we must agree
not to organize for political blackmail; we
must request editors to allow free representation of opinion; we
must ask
pop stars and professors for modesty in being opinionated.
Am I then saying that democratic man cannot be rational man? I am not.
Rational man wants liberty. Democracy
protects liberty. Nothing could be
more rational than to act in support of democracy. What I am rather saying
is that rational man needs not be egoistic man. Thoughtful action on the part
of free persons cannot mean only
the effective use of means for short-term
satisfaction, it must also include the choice of goals. It is not
beyond the wit
of those who are concerned with the consequences of what they do to grasp
that the effectiveness
of their own actions depends on how they interact
with the actions of others, and the co-ordination of action
is not the mystery
it is sometimes made out to be once we allow for experience.
‘In every scientific venture’, wrote Joseph Schumpeter in History of
Economic Analysis, ‘the thing
that comes first is Vision.’ Is there a vision
for the social sciences? There are no doubt many. Read welfare
economics,
and you might think efficiency to be the vision. But efficiency cannot be
vision, it can only be in
relation to vision; welfare economics has gone
utilitarian and hence given up on the vision thing. In the broad
church in the
social sciences in which I feel at home, the vision is justice.
In what sense is justice a vision? It is first of all a normative vision,
something we believe in as a good
thing. On this occasion, however, I want
to argue that justice is also an instrumental vision. If we take
democracy as
the value we are ultimately concerned with, the call for justice—or
equality
—is no longer only a statement of opinion, it is also an analytic result,
reasonable equality being
among the preconditions of practical democracy.
I would not want to say that democracy is dependent on the social
sciences; of course not. However, I do not
think it is immodest to suggest
that high-quality social research inspired by the vision of justice can give
support to democracy by contributing to satisfying, in Claus Moser’s words,
our need for an informed society.
On the side of strategy, I have suggested that the social scientist who
wants to render service to democracy,
should, as should the citizen
participating in democracy, seek a balance between assertiveness and
restraint.
In this day and time of the me-generation, and of ideology
crowding out pragmatism in government, and bearing
in mind the over-
opinionated nature of much that is said and written in my own field of
research—that of social
policy—I want to underline restraint. Neither in
politics, nor in science—no more than in love, family, or
friendship—can
we succeed for ourselves or be of use for others if we always insist on
holding our ground.
Here my story ends. Does democracy work? My remarks have been
eminently old-fashioned ones. I have spoken of
persons rather than of
structures. I have suggested that we as scientists have responsibility in our
society. I
have argued that democracy depends on democratically minded
citizens and that we should apply restraint if we
want to do good.
 POSTSCRIPT: A STATEMENT ON
POLITICAL PHILOSOPHY

A good society is a society in which citizens can live good lives. The good
life is subjective to the
individual, but rests on objective conditions for its
pursuit. These conditions are the domain of politics. The
subjectivity of the
good life is the responsibility of free citizens and is outside the business of
the state.
The conditions which make the good life possible are liberty, law, and
community. Liberty is the availability
of free choice, including reasonable
resources and options for choice in lifestyle. Richness is trivial.
Community
is social guidance for wisdom in choice.
Good society depends crucially (but not exclusively) on good
government. Good government is necessarily
democratic. Democracy is
superior to autocracy in respect both to justice and to efficiency.
In democracy, the state has duties towards citizens. These include the
safeguarding of liberty (from overall
security to minority rights); the ability
to enact and execute adequate law; and the distribution of wealth so
as to
make choice a universal reality.
Equally, citizens have duties to each other. These include to make use of
liberty, and when participating to do
so with integrity; to co-operate in
community; and to comply with decisions of the state properly made.
Contradictions between liberty and duty, liberty and law, or liberty and
community are possible, but not
necessary and intolerable as long as
democracy is sufficiently vibrant to temper absolutism in the
interpretation
and practice of duty, law, and redistribution.
 REFERENCES

Andvig, J. C. (1986): Ragnar Frisch and the Great Depression. University

of Oslo: Department of
Economics.
Arneson, R. J. (1989): ‘Equality and Equal Opportunity for Welfare’,
Philosophical Studies 56: 77–93.
Arrow, K. (1963): Social Choice and Individual Values. New York: Wiley.
Atkinson, A. B. (1993): ‘What is Happening to the Distribution of Income
in the UK?’ Proceedings of the
British Academy 82: 317–351.
Atkinson, A. B. and J. Michlewright (1992): Economic Transformation in
Eastern Europe and the Distribution of
Income. Cambridge: Cambridge
University Press.
Atkinson, A. B. L. Rainwater, and T. M. Smeeding (1995): Income
Distribution in OECD Countries. Paris:
Oecd.
Becker, G. S. (1965): ‘A Theory of the Allocation of Time’, Economic
Journal 1965: 493–517.
Berlin, I. (1991): The Crooked Timber of Humanity. London: Fontana Press.
Boudon, R. (1974): Education, Opportunity, and Social Inequality. London:
Wiley.
Boudon, R. (1984): La place du désordre. Paris: Presses Universitaires de
France.
Boudon, R. (1995): Le juste et le vrai. Paris: Fayard.
Brannen, J., and G. Wilson (1987): Give and Take in Families: Studies in
Resource Distribution. London:
Allen & Unwin.
Buhmann, B., L. Rainwater, G. Schmaus, and T. M. Smeeding (1988):
‘Equivalence Scales, Weil-Being, Inequality,
and Poverty: Sensitivity
Estimates across Ten Countries Using the Luxembourg Income Study
(LIS) Database’,
Review of Income and Wealth 34: 115–142.
Cigno, A. (1991): The Economics of the Family. Oxford: Clarendon Press.
Cohen, G. A. (1989): ‘On the Currency of Egalitarian Justice’, Ethics 99:
906–944.
Coleman, J. (1971): Resources for Social Change. New York: Wiley.
Coulter, F. A., F. A. Cowell, and S. P. Jenkins (1992): ‘Differences in Needs
and Assessment of Income
Distributions’, Bulletin of Economic Research
44: 77–124.
Coulter, F. A. (1994): ‘Family Fortunes in the 1970s and 1980s’, in R.
Blundel, I. Preston, and I. Walker:
The Measurement of Household
Welfare. Cambridge: Cambridge University Press.
Dahl, R. A. (1985): A Preface to Economic Democracy. Cambridge:
Cambridge University Press.
Dahl, R. A. (1989): Democracy and Its Critics. New Haven: Yale
University Press.
Dahrendorf, R. (1979): Life Chances. Chicago: Chicago University Press.
Dahrendorf, R. (1988): The Modern Social Conflict. London: Weidenfeld &
Nicolson.
Dahrendorf, R. (1990): Reflections on the Revolution in Europe. London:
Chatto & Windus.
Dasgupta, P. (1990): Well-Being in Poor Countries. Cormandel Lecture 16.
Dasgupta, P. (1993): An Inquiry into Weil-Being and Destitution. Oxford:
Clarendon Press.
Davies, H., and H. Joshi (1994): ‘sex, Sharing and the Distribution of
Income’, Journal of Social Policy
23(3): 301–340.
Deaton, A., and J. Muellbauer (1980): Economics and Consumer
Behaviour. Cambridge: Cambridge University
Press.
Dilnot, A. W., J. A. Kay, and C. N. Morris (1984): The Reform of Social
Security. Oxford: Clarendon
Press.
DSS (1994): Households Below Average Income. London: Hmso.
Dworkin, R. (1981a): ‘What is Equality? Part 1: Equality of Welfare’,
Philosophy and Public
Affairs 10: 185–247.
Dworkin, R. (19816): ‘What is Equality? Part 2: Equality of Resources’,
Philosophy and Public Affairs
10: 283–345.
Dwyer, D., and J. Bruce (eds.) (1988): A Home Divided: Women and
Income in the Third World. Stanford
Calif.: Stanford University Press.
Easton, B. (1986): Wages and the Poor. Sydney: Allen & Unwin.
Erikson, R., and J. H. Goldthorpe (1992): The Constant Flux. Oxford:
Clarendon Press.
Erikson, R. and H. Uusitalo (1987): ‘The Scandinavian Approach to
Welfare Research’, in R. Erikson, E. J. Hansen,
S. Ringen, and H.
Uusitalo: The Scandinavian Model. New York: M. E. Sharpe.
Falkingham, J., J. Hills, and C. Lessof (1993): William Beveridge versus
Robin Hood: Social Security and
Redistribution over the Lifecycle.
London: LSE Welfare State Programme.
Frisch, R. (1962): Innledning til produksjonsteorien. Oslo: Norwegian
University Press.
Fritzell, J. (1993): ‘Income Inequality Trends in the 1980s: A Five-Country
Comparison’, Acta
Sociologica 36: 47–62.
Gambetta, D. (1987): Were They Pushed or Did They Jump? Cambridge:
Cambridge University Press.
Gershuny, J. I. (forthcoming): Changing Times. Oxford: Oxford University
Press.
Gershuny, J. I. and B. Halpin (1995): ‘Imputations of Household Income
from Unpaid Work Time’. ESRC Research
Centre on Micro-Social
Change (mimeo).
Gerth, H. H., and C. Wright Mills (1946): From Max Weber: Essays in
Sociology. New York: Oxford
University Press.
Gillespie, W. I. (1965): ‘The Effects of Public Expenditures on the
Distribution of Income’, in R. A. Musgrave:
Essays in Fiscal Federalism.
Washington, DC: Brookings.
Gleditsch, N. P. (1995): ‘Geography, Democracy, and Peace’, International
Interactions 20: 297–323.
Gleditsch, N. P. and B. O. Sverdrup (1995): ‘Democracy and the
Environment’. Oslo: International Peace
Research Institute (mimeo).
Glendinning, C., and J. Millar (1987): Women and Poverty in Britain.
Brighton: Wheatsheaf.
Glennerster, H., and W. Low (1990): ‘Education and the Welfare State:
Does it Add Up?’, in J. Hills: The
State of Welfare. Oxford: Clarendon
Press.
Goldschmidt-Clermont, L. (1982): Unpaid Work in the Household. Geneva:
ILO.
Goldschmidt-Clermont, L. (1993): ‘Monetary Valuation of Non-Market
Production Time’, Review of Income and
Wealth 39: 419–433.
Goldthorpe, J. H. (1995): ‘Class Analysis and the Reorientation of Class
Theory: The Case of Persisting
Differentials in Educational Attainment’.
Oxford: Nuffield College (mimeo).
Goldthorpe, J. H., et al. (1980): Social Mobility and Class Structure in
Modern Britain. Oxford:
Clarendon Press.
Goldthorpe, J. H. (1987): Social Mobility and Class Structure in Modern
Britain (2nd edn.). Oxford:
Clarendon Press.
Goodman, A., and S. Webb (1994): For Richer, for Poorer: The Changing
Distribution of Income in the United
Kingdom, 1961–91. London:
Institute for Fiscal Studies.
Griffin, J. (1986): Weil-Being. Oxford: Clarendon Press.
Gustafsson, B., and H. Uusitalo (1990a): ‘Income Distribution and
Redistribution during Two Decades:
Experiences from Finland and
Sweden’, in J. Persson: Generating Equality in the Welfare State. Oslo:
Norwegian University Press.
Gustafsson, B. (1990b): ‘The Welfare State and Poverty in Finland and
Sweden from the Mid-1960s to the
Mid-1980s’, Review of Income and
Wealth 36: 249–266.
Haddad, L., and R. Kanbur (1989): ‘How Serious is the Neglect of Intra-
Household Inequality?’, Economic
Journal 100: 868–881.
Halsey, A. H. (1977): ‘Towards Meritocracy? The Case of Britain’, in J.
Karabel and A. H. Halsey: Power and
Ideology in Education. Oxford:
Oxford University Press.
Halsey, A. H. (1992): Decline of Donnish Dominion. Oxford: Clarendon
Press.
Halsey, A. H. A. F. Heath, and J. M. Ridge (1980): Origins and
Destination: Family, Class and Education in
Modern Britain. Oxford:
Clarendon Press.
Harding, A. (1992): Income Distribution and Redistribution Across the
Lifecycle: Evidence from
Australia. London: LSE Welfare State
Programme.
Harriss, B. (1990): ‘The Intrafamily Distribution of Hunger in South Asia’,
in The Political Economy of
Hunger, 1: Entitlement and Well Being.
Oxford: Oxford University Press.
Heath, A. F., and P. Clifford (1990): ‘Class Inequalities in Education in the
Twentieth Century’, Journal of
the Royal Statistical Society, series A 153:
1–16.
Heath, A. F. C. Mills, and J. Roberts (1992): ‘Towards Meritocracy? Recent
Evidence on an Old Problem’, in C.
Crouch and A. Heath: Social
Research and Social Reform. Oxford: Clarendon Press.
Hills, J., et al (1993): Investigating Welfare: Final Report of the ESRC
Welfare State Programme.
London: LSE Welfare State Programme.
Hoddinot, J., and L. Haddad (1995): ‘Does Female Income Share Influence
Household Expenditures? Evidence from
Côte d’Ivoire’, Oxford Bulletin
of Economics and Statistics 57: 77–96.
Holmes, S. (1995): Passions and Constraint: On the Theory of Liberal
Democracy. Chicago: Chicago
University Press.
House of Representatives (1993): Overview of Entitlement Programmes
(Greenbook 1992).
Washington, DC: Government Printing Office.
INSEE (1996): Le niveau de vie par générations. Paris: Insee.
Jencks, C., and B. B. Torrey (1988): ‘Beyond Income and Poverty: Trends
in Social Welfare among Children and the
Elderly Since 1960’, in J. L.
Palmer et al. : The Vulnerable. Washington, DC: Urban Institute Press.
Jenkins, S. P. (1991): ‘Income Inequality and Living Standards: Changes in
the 1970s and 1980s’, Fiscal
Studies 12: 1–28.
Jenkins, S. P. (1994): ‘The Within-Household Distribution and Why It
Matters: An Economist’s
Perspective’, Discussion Paper Series 94–105,
University College of Swansea: Dept. of Economics.
Johansson, S. (1979): Mot en teori för social rapportering. Stockholm:
Swedish Institute for Social
Research.
Joshi, H., and H. Davies (1993): ‘Mothers’ Human Capital and Childcare in
Britain’, National Institute
Economic Review 1993: 50–63.
Lancaseter, K. (1966): ‘A New Approach to Consumer Theory’, Journal of
Political Economy 74: 132–157.
Lazear, E. P., and R. T. Michael (1988): Allocation of Income within the
Household. Chicago: University
of Chicago Press.
Le Grand, J. (1991): Equity and Choice. London: HarperCollins Academic.
Levekårsundersøkelsen (1976): Sluttrapport. Oslo: Universitetsforlaget
(NOU).
McRae, S. (1987): ‘The Allocation of Money in Cross-Class Families’,
Sociological Review 35(1).
Marshall, A. (1949): Principles of Economics (8th edn.). London:
Macmillan.
Marshall, G., H. Newby, D. Rose, and C. Vogler (1988): Social Class in
Modern Britain. London:
Hutchinson.
Mill, J. S. (1859): On Liberty. (In On Liberty and Other Writings.
Cambridge University Press
1989.)
Miller, D. (1976): Social Justice. Oxford: Oxford University Press.
Moser, C. (1990): ‘Our Need for an Informed Society’, Presidential Address
to the British Association
for the Advancement of Science 1990.
Murphy, J. (1981): ‘Class Inequality in Education: Two Justifications, One
Evaluation but no Hard Evidence’,
British Journal of Sociology 32: 82–
101.
Murphy, J. (1990): ‘A Most Respectable Prejudice: Inequality in
Educational Research and Policy’, British
Journal of Sociology 41: 29–
54.
Murphy, J. (1992): ‘A Most Respectable Prejudice—Boudon’, British
Journal of Sociology 43: 485–488.
Myrdal, G. (1953): The Political Element in the Development of Economic
Theory. London: Routledge &
Kegan Paul.
Nelson, J. A. (1992): ‘Methods of Estimating Equivalence Scales: An
Empirical Investigation’, Review of
Income and Wealth 38: 295–310.
Nussbaum, M. C., and A. Sen (eds.) (1992): The Quality of Life. Oxford:
Clarendon Press.
OECD (1982): The OECD List of Social Indicators. Paris: Oecd.
Okun, A. M. (1975): Equality and Efficiency. The Big Trade-Off.
Washington, DC: Brookings.
Pahl, J. (1989): Money and Marriage. London: Macmillan.
Palmer, J. L., T. Smeeding, and B. Boyle Torrey (eds.) (1988): The
Vulnerable. Washington, DC: Urban
Institute Press.
Piachaud, D. (1982): ‘Patterns of Income and Expenditure within Families’,
Journal of Social Policy 11:
467–482.
Pigou, A. C. (1929): The Economics of Welfare (3rd edn.). London:
Macmillan.
Preston, S. H. (1984): ‘Children and the Elderly: Divergent Paths for
America’s Dependents’, Demography
21: 435–457.
Przeworski, A. (1991): Democracy and the Market. Cambridge: Cambrige
University Press.
Rawls, J. (1971): A Theory of Justice. Cambridge, Mass.: Harvard
University Press.
Ringen, S. (1986): Difference and Similarity: Two Studies in Comparative
Income Distribution. Stockholm:
Swedish Institute for Social Research.
Ringen, S. (1987): The Possibility of Politics. Oxford: Clarendon Press.
Ringen, S. (1988): ‘Direct and Indirect Measures of Poverty’, Journal of
Social Policy 17: 351–365.
Ringen, S. (1991): ‘Households, Standard of Living, and Inequality’,
Review of Income and Wealth 37:
1–13.
Ringen, S. (1993): ‘Income Dependency in Norway’, in E. J. Hansen, S.
Ringen, H. Uusitalo, and R. Erikson:
Welfare Trends in the Scandinavian
Countries. New York: M. E. Sharp.
Ringen, S. (1995): ‘Trygdevekst, omfordeling og moralsk risiko’, in A.
Kjønstad, A. Hatland, and B. Halvorsen
(eds.): Det norske
trygdesystemet. Oslo: AdNotam Gyldendal.
Ringen, S. (1996): ‘Det sosiale spørsmål i dag: Fra utstøting til
eksklusjon’. Oslo: Norsk Gerontologisk
Institutt.
Ringen, S. and H. Uusitalo (1992): ‘Income Distribution and Redistribution
in the Nordic Welfare States’, in J.
E. Kolberg: The Study of Welfare State
Regimes. New York: M. E. Sharp.
Robbins, L. (1932): An Essay on the Nature and Significance of Economic
Science. London: Macmillan.
Robinson, J. P., and J. Gershuny (1994): ‘Measuring Hours of Paid Work:
Time-Diary vs. Estimate Questions’,
Bulletin of Labour Statistics
1994(1).
Rowntree, S. (1901): Poverty: A Study of Town Life. London: Macmillan.
Rowntree, S. (1941): Poverty and Progress. London: Longmans.
Rowntree, S. and G. R. La vers (1951): Poverty and the Welfare State.
London: Longmans.
Ruggles, P., and M. O’Higgins (1981): ‘The Distribution of Public
Expenditure among Households in the United
States’, Review of Income
and Wealth 27(2).
Sartori, G. (1987): The Theory of Democracy Revisited, I–II. Chatham:
Chatham House.
Saunders, P. (1990): Social Class and Stratification. London: Routledge.
Saunders, P., H. Stott, and G. Hobbs (1991): ‘Income Inequality in Australia
and New Zealand: International
Comparisons and Recent Trends’,
Review of Income and Wealth 37.
Schumpeter, J. (1954): A History of Economic Analysis. Oxford and New
York: Oxford University Press.
Sen, A. (1981): Poverty and Famines. Oxford: Oxford University Press.
Sen, A. (1983): ‘Poor, Relatively Speaking’, Oxford Economic Papers 35:
153–169.
Sen, A. (1984): Resources, Values and Development. Oxford: Basil
Blackwell.
Sen, A. (1985): Commodities and Capabilities. Amsterdam: North-Holland.
Sen, A. (1992): Inequality Reexamined. Oxford: Clarendon Press.
Simon, H. (1957): Models of Man. New York: Wiley.
Smeeding, T. M., M. O’Higgins, and L. Rainwater (eds.) (1990): Poverty,
Inequality and Income Distribution in
Comparative Perspective. London:
Harvester Wheatsheaf.
Storr, A. (1991): Human Destructiveness. New York: Grove Weidenfeld.
Strack, F., M. Argyle, and N. Schwarz (eds.) (1991): Subjective Well-Being.
Oxford: Pergamon Press.
Tawney, R. H. (1931): Equality. London: Allen & Unwin.
Thomas, D. (1991): ‘Gender Differences in Household Resource
Allocations’, LSMS Working Paper 79,
Washington, DC.: World Bank.
Titmuss, R. M. (1962): Income Distribution and Social Change. London:
Allen & Unwin.
Townsend, P. (1979): Poverty in the United Kingdom. London: Penguin.
Uusitalo, H. (1989): Income Distribution in Finland. Helsinki: Central
Statistical Office.
Veit-Wilson, J. H. (1986): ‘Paradigms of Poverty: A Rehabilitation of B. S.
Rowntree’, Journal of Social
Policy 15: 69–99.
Vogler, C. (1989): ‘Labour Market Change and Patterns of Financial
Allocation with Households’, Social Change
and Economic Life
Initiative Working Paper 12, University of Oxford: Nuffield College.
Williams, B. (1987): ‘The Standard of Living: Interests and Capabilities’, in
G. Hawthorn: The Standard of
Living. Cambridge: Cambridge University
Press.
Young, M. (1952): ‘The Distribution of Income within the Family’, British
Journal of Sociology 3:
305–321.
 INDEX

applied research 163, 164

arenas 18, 21, 27, 36, 38–9
Arneson, R. J. 39 n.
Arrow, K. 174
Atkinson, A. B. 48 n., 127 n., 149
n., 151 n., 154
authoritarian 10
Becker, G. S. 29 n.
Berlin, I. 14 n.
Boudon, R. 10, 23
Brannen, J. 76 n., 82 n.
Bruce, J. 76 n.
Buhmann, B. 46 n.
Castro, F. 10
child care 58
child rearing 80
children 3, 56
discrimination against 82–3
and family income 67–72
opportunity cost of 90
‘penalty’ for early start 69, 71
political representation of 3–8
and power 6
protection of against poverty 82
rights of 4
social security for 6
standard of living of 4–6, 54–5, 81–2, 86–9, 92
choice 23–6, 31, 33, 36, 38, 177
Christie, F. C. 22, 25
Cigno, A. 67 n.
class 130
and action vs. structure 9–10
and British society 9, 10, 129, 141–2
and education 133–4
and social inequality 8, 9, 124–8, 143–6
class bias 131, 132–3, 142, 147
Clifford, P. 129 n., 140, 141
n.
Cohen, G. A. 31, 38
Coleman, J. 27 n.
community 177
consumption resources 62, 67
effects of children on 70–2
consumption standard 62, 72–3,
76, 80–2
and poverty 85
Coulter, F. A. 46 n., 49 n., 59
n., 60 n., 149 n.
coup d’état 147
Dahl, R. 168
Dahrendorf, R. 23
Dasgupta, P. 13 n., 26
Davies, H. 77 n.
Deaton, A. 47 n.
Declaration of Independence, American 13, 23
democracy 1, 119, 162–3, 165–8
and agreement 2
and compliance 167, 177
confidence in 1, 10
deficit of 4, 6
and equality 119, 168, 175
and happiness 13–14
imperatives of 11
and imperfection 2
and liberty 1–2, 119
revolution of 1, 163, 164 see also reform
dictatorial logic 20, 21, 32
Dilnot, A. W. 80 n.
double imperative 119, 120
DSS 5 n.
Dworkin, R. 21, 31
Dwyer, D. 76 n.
Easton, B. 83 n.
economic growth 51, 54
economies of scale 48, 59
efficiency 175
equality 45, 119, 168
equality of opportunity 17, 38–9
equity 45, 79
assumed in families 76
and needs 79
equivalence scales 46–8, 52,
53, 59
equivalent income 18, 21, 45, 59–60,
76
absolute interpretation of 46–7, 53
Erikson, R. 34 n., 127 n.
expectations 21, 30, 31
Falkingham, J. 160
family 12, 76
agreement within 79–80
co-operation within 18, 76, 89
distribution of income in 77
and household 41 n., 62
importance of 12–13, 91
income pooling in 77
inequality within 18, 80
inequity within 60, 76–80
movement ‘into’ 91
production within 18, 89
sharing within 76, 79
‘vulnerable’ to children 71, 91
family economies 14, 18, 43, 53–4,
89–91
and co-operation 42, 73
and distribution 42, 60
efficiency effects in 46, 48
and needs 42
needs effects in 46, 48
and production 42, 58–60, 73–4
and savings 57, 78
family failure 121
final welfare 59
fiscal policies 150
freedom 18, 21, 22, 34, 39, 162
and choice 23, 24–5
freedom of choice 17, 36
Frisch, R. 168, 169, 170
Fritzell, J. 149 n., 155
full income 45
Gambetta, D. 133 n.
Gandhi 120
Gershuny, J. I. 60 n., 109
Gillespie, W. I. 29 n.
Gleditsch, N. P. 1 n.
Glendinning, C. 76 n.
Glennerster, H. 134 n.
Goldschmidt-Clermont, L. 29 n.
Goldthorpe, J. H. 11 n., 127 n., 128
n., 129 n.
Goodman, A. 56 n., 149 n.
goods 17, 41, 43, 53, 56, 58, 73
public within family 41–2, 78
services from 41, 54, 77
Griffin, J. 34 n.
Gustafsson, B. 149 n.
Haddad, L. 76 n., 82 n.
Halpin, B. 109
Halsey, A. H. 130 n., 131 n.
happiness 13, 17, 23
Harding, A. 158
Harriss, B. 76 n., 82 n.
head of household 62
Heath, A. F. 129 n., 140, 141
n.
Hills, J. 158 n.
Hoddinot, J. 82 n.
Holmes, S. 121
House of Representatives (USA) 4 n.
housework 58, 60, 67, 76
Ibsen, H. 172
imperfection 173–4
and responsibility 3
income 62
effects of children on 67–70
and social inequality 8
sources of 63–4, 90–1
stability of distribution 161
income gap 68–9, 70
income method 12, 33, 36–7, 53,
56, 92–3
income standard 62, 72–3, 80–2
and poverty 85
incomes policy 124, 157
independence hypothesis 150
inequality 52, 122, 135–6
and association 136, 142
in Britain 5, 8, 54, 56, 141, 146, 154–5
and children 4–5
and class 8, 11, 136
comparative 151, 155–6
and distribution 137, 142
and family inequity 83–4
in Finland 156–8
and income 8
measurement of 83
and poverty 8, 11, 122
and reform 8–9, 11
relative definition of 9, 11
information 11, 13, 41, 93, 170–1, 173
INSEE 5 n.
Jefferson, T. 13
Jencks, C. 126 n.
Jenkins, S. P. 76 n., 149 n.
job failure 121
Johansson, S. 171
Joshi, H. 77 n.
justice 175–6
Kanbur, R. 76 n.
labour force participation 66, 71
Lazear, E. P. 76 n.
legitimacy 121
Le Grand, J. 23
Levekårsundersøkelsen 27 n.
liberal democracy 121
liberal theory, the 127, 128
libertarian 17, 119, 120
and liberalism 124
liberty 1, 2, 119, 177
loaded dice 131, 142
Low, W. 134 n.
Macmillan, H. 171
McRae, S. 76 n.
Marshall, G. 129 n.
Marx, K. 173
measurement
democratic method of 12, 43
direct vs. indirect 26–7
and family economies 18–19, 88, 93
and family inequity 82–6, 88,
93
and information 28–9, 93
and methodological individualism 12, 43
narrow vs. broad 28
meritocracy 127
Michael, R. T. 76 n.
Michlewright, J. 48 n., 127 n.
Mill, J. S. 22
Millar, J. 76 n.
Mitterrand presidency 10
mobility tables 131–2, 134,
136
and Gini-coefficient 137
and inequality 146
Moser, C. 176
Muellbauer, J. 47 n.
Murphy, J. 133 n.
Myrdal, G. 170
Nelson, J. A. 47 n.
odds ratios 137, 147
OECD 47, 50, 59, 151
O’Higgins, M. 58 n.
Okun, A. 173
opportunity 18, 23, 130
and results 39–40
Pahl, J. 76 n.
Palmer, J. L. 4 n.
paradox of voting 121–2
perfection 2, 14 n.
and democracy 2, 3, 173–4
flirtation with 10
terrible temptation towards 2, 11, 14,
120
Piachaud, D. 76 n.
Pigou, A. C. 35
Popper, K. 120
Population Association of America 4
positivism debate 169
poverty 125
and family inequity 84–6
head count method 84
measurement of 44–5, 126
poverty-gap method 45, 86
and relative deprivation 125–6
and social inequality 8, 9, 124–8
preference neutrality 32–3, 34,
35, 37, 39
preferences 17, 20, 30, 31
and expectations 21, 32
Preston, S. 4
quality of life 20, 21
rationality 22–3, 26, 121–2,
175
Rawls, J. 21, 168
real existierender Sozialismus 10
redistribution 124–5, 152–3, 161
n.
and children 55
 lifetime 158–60
and tax/transfers 149–50, 151, 152, 160
total vs. marginal 153–4
and utility vs. choice 35–6
reform 119–20, 147, 162
and democracy 8, 9, 11
and efficiency 153 n.
possibility of 11, 124, 128, 129, 147–8, 151
and social inequality 8, 14, 123
strong position 119–20
weak position 119–20, 123–4
reform democracy 121
regulation 124, 127
relativism 10
and democracy 10
and social sciences 10–11, 14
resources 18, 27, 36, 38–9
responsibility 3, 123, 171–3
restraint 121, 122–3, 176
revolution 119, 120, 146
Ringen, S. 121 n., 126 n., 149 n., 153 n., 155
n.
Robbins, L. 35
Robinson, J. R. 109
Rowntree, S. 125, 126
Ruggles, P. 58 n.
rules 174–5
Sartori, G. 164
Saunders, P. 133 n.
Saunders, P. 149 n.
Schumpeter, J. 175
Sen, A. 1 n., 21, 23 n., 24 n., 26, 28, 38, 76 n., 126 n.
Simon, H. 24 n.
Smeeding, T. M. 6 n., 149 n.
Smith, A. 126
social change 9
social competition 130–2, 134–5
and treatment in 130, 132, 137, 142–3, 145,
147, 148
social conflict 5
social democracy 119, 120
social exclusion 121
social mobility 10–11, 130
social security 158
Stalin 25
standard method, the 151, 152–3,
159, 160–1
Storr, A. 174
Strack, F. 21 n.
Sverdrup, O. 1 n.
tastes 30–1
Tawney, R. H. 168
Thomas, D. 82 n.
time-use 60, 64–6
Tinbergen, J. 168
Titmuss, R. M. 28
Tocqueville, A. de 168, 173
Torry, B. B. 126 n.
Townsend, P. 125 n.
true democracy 119
UN Convention on the Rights of the Child 4
Utilitarianism 121–2
utility 17, 20, 22, 30, 31–2, 33,
39
and too much choice 24
Utopian 166
Uusitalo, H. 34 n., 149 n., 156
value added 12, 44, 50–1, 73–6, 89
Veit-Wilson, J. H. 126
Vogler, C. 76 n.
voting 3, 121, 166, 167–8
and children 6–8
and mothers 7
Webb, S. 56 n., 149 n.
Weber, M. 169
welfare economics 35–6, 37,
175
and interpersonal comparisons 34
welfare state 162
welfarism 17, 20–1, 30, 32, 35
well-being
aggregation of 44
and capabilities 21, 29, 37–9
and consumption 29, 41, 56, 80
and costs 25–6
dictatorial 17, 21, 30, 34
and economic growth 51
and expenditures 29
and family 12, 18, 91
and freedom 18, 21, 34
and goods 21, 31, 54
and income 28, 33, 36–7
measurement of 12, 26–30
as potential consumption 57
and preferences 20, 34
and resources 21, 28
social concept of 12, 18
subjective vs. objective 13, 17, 34
and way of life 29
Williams, B. 38
Wilson, G. 76 n., 82 n.
Young, M. 76 n.