HOME PRODUCTION—A FORGOTTEN INDUSTRY
Reuben Gronau*
I. Introduction
R ECENT years have witnessed an awakened
interest in the economic activity taking
place outside the market, and in particular the
activity taking place at home. This interest
spurred by the new consumption theory of Beck-
er and Lancaster and by the estimates of the
Measure of Economic Welfare of Nordhaus and
Tobin (1973) has taken two distinct forms: an
increased number of studies on the economics of
household behavior and a renewed effort to place
a money value on the household home activity.
However, while the major thrust of the first type
of studies is in the field of microeconomics, the
estimates of home production refer, in general, to
the economy as a whole. These estimates, crude
as they are, indicate that home production is far
from being a negligible part of the economic ac-
tivity. Even in an advanced economy such as the
United States the value added generated by the
home sector seems to account for over one third
of the output produced at the market (Haw- young children, the loss of home output when the
rylyshyn, 1976). In less advanced economies this wife joins the labor force equals almost her in-
fraction is presumably even higher. It seems, creased money earnings. Finally, home produc-
therefore, of interest to repeat the question in a
microeconomic context and examine the role of
home production at the household level, rather
than in the aggregate.
In contrast to past studies which have focused
on the labor inputs going into home production
(Sirageldin, 1969; Walker and Gauger, 1973), the sources is discussed in section III. The paper
emphasis in this paper is on the measurement of closes with some concluding remarks.
productivity and total home output. The ques-
tions I try to answer are: What are the factors
Received for publication August 24, 1978. Revision ac-
cepted for publication July 27, 1979.
* Hebrew
University.
This paper was written while on sabbatical at the National
Bureau of Economic Research. It has not undergone the full
critical review accorded the NBER studies.
I benefited from comments by Y. Ben-Porath, V. Fuchs,
G. Hanoch, J. Heckman, R. Michael, S. Rosen, T. P. Schultz,
and F. Welch, and two anonymous referees. I am thankful to
Kris Chinn, Kyle Johnson, and Henry Moore for their compu-
tational assistance. The research reported herein was sup-
ported by grants to NBER from the Rockefeller Foundation
and the Alfred P. Sloan Foundation.
The paper was first presented at the ADC's Household
Studies Workshop in Singapore in August 1976.
[ 408
determining the wife's productivity at home?
What is the value of home production and how
does it compare with the family's money in-
come? How does the value of home production
differ among families with different socioeco-
nomic backgrounds? How is it affected by the
wife's labor force participation and by the exis-
tence of young children? How does it change
over the family's life cycle?
It is found that the value of home production
associated with the work at home of U.S. wives
in 1973 exceeded 60% of the family's money in-
come before taxes, and 70% of the family's
money income after taxes. It was lower for fami-
lies with no preschool children and almost equal
to the family's money income after taxes when
the family had young children. Home productiv-
ity increases with education but at a lower rate
than market productivity. Home production is.
only slightly affected by the wife's employment
in the market when the family does not have
young children. However, when the family has
tion tends to peak at a younger age (35—39) than
money income and drops significantly thereafter.
The paper opens with a discussion of the esti-
mation of household productivity—the model,
the data and the estimates. The role home output
plays in comparison with other material re-
II. The Estimation of Home Productivity
Following conventional theory, it is assumed
that welfare (U) is a function of consumption (X)
and leisure (L),
U = U(X,L). (1)
Goods can be either purchased in the market
(XM) or produced at home (Z). Measuring the
home-produced goods and services in terms of
the price charged for them in the market,
X=XM+Z. (2)
 HOME PRODUCTION—A FORGOTTEN INDUSTRY 409
Home goods are produced using market inputs productivity of time in home production, and
(XH) and time inputs (i.e., work at home H), generate the value of home production by inte-
Z = f(XH,H) grating the marginal productivity function.
To estimate the marginal productivity function
subject to decreasing marginal productivity f > fH one can proceed in one of two ways: assume a
O,fH> < 0, andf11 < 0. The maximization certain form of the home production function f
of welfare is subject to two constraints: (a) the (e.g., Cobb-Douglas or constant elasticity of
budget constraint specifying that in this one-per- substitution (CES)), derivefH and estimate it, or
son, one-period model, market consumption alternatively make an explicit assumption about
cannot exceed money income. the functional form offH and derivef. I adopted
XM + XH = WN + V the second of these methods.
It is assumed that the marginal productivity of
where W denotes the wage rate, N denotes work the home production function is of the semi-log
in the market, and V other sources of income; variety2
and (b) the time constraint specifying that time lflfH = — cx1H + a2Y
(T) is a scarce resource to be allocated among its (8)
three uses—work at home, work in the market where Y denotes a vector of variables affecting
and leisure: the value of marginal productivity at home.
L + H + N = T. Given this specific function and the equilibrium
(5)
condition (6) one can derive the work at home
The maximization of welfare subject to these function for labor force participants:
constraints yields two kinds of equilibria: (a) the
case where the person works in the market H = (a0 — In W + a0Y)/a1. (9)
(N> 0) in which case Estimating this function,
fH = S W (6) Ha0—a1lnW+a2Y, (10)
where fli is the marginal productivity of work one derives the estimates of the parameters a,.:
at home and s denotes the marginal rate
of substitution between leisure and goods (1/a1) = est(a1),
[s = (aU/aL)/(aU/ax)], and (b) the case where (ao/a1) = est(a0),
the person does not participate in the labor force (11)
(N = 0) where only the first of these equalities is (a2/a1) = est(a0).
satisfied:
Using these parameters one can estimate the
fH = (7) value of home production
The estimation of the home goods production z = Jf(t)dt
function (equation (3)) is abound with difficul-
ties. Not only is the outputZ unknown, but there (H
= exp(a0 — a11 + a0Y)dt
is no direct way of separating XM from XH and Jo
measuring the market inputs (XH) that are used in = {exp(a0 + a0Y)[l — exp(—a1H)]}/a1,(12)
the process. An indirect approach is, therefore,
clearly needed. The point of departure of this assurtling there is no home production when
indirect approach is equation (6) rather than equa- there is no work at home.
tion (3). Its purpose is to estimate the marginal Data on time use are hard to come by, and
sources where such data are accompanied by
I The qualifications and ramifications of a similar model are adequate information on the household's socio-
discussed at length in Gronau (1977), and so I shall not economic characteristics are even more scarce.
discuss them here. It is sufficient to say that the optimum One of the few sources that satisfies our require-
conditions (6) and (7) could have been generated also by
other structures and, hence, my method of estimation is not
crucially dependent on the validity of this model. Two crucial 2 The assumption on the semi-log form follows Heckman
assumptions are that there is no joint use of time for work at (1974). Tests indicate that this functional form is superior to
home and leisure activities, and that work at home and work the linear and double log form. A more elaborate test of
in the market generate the same direct utility. consistency is described in the appendix.
410 THE REVIEW OF ECONOMICS AND STATISTICS
ments is the Michigan Study of Income Dynam- The wage is imputed as a function of the wife's
ics. In this study the head of the household was education (EDUCW), her labor force experience
asked about the number of weeks the head and since the age of 18 (EXPRNW), the experience
his wife worked in the market in the previous variable squared and the husband's education
year and the number of hours worked per week. (EDUCH). The estimated equation is of the semi
People were also asked how much time they and log form:
their wives spent on housework in an average ln W = — .4237 + .0827 EDUCW
week. Housework was not very strictly defined
and the examples mentioned in the questionnaire (3.83) (8.34)
such as cooking, cleaning and other work around
+ .0186 EDUCH + .0331 EXPRNW
(2.25) (5.48)
the house allow for a lot of ambiguity. Thus, it is
.0006(EXPRNW)2 -
not clear whether the respondents included in
their answers time spent on activities such as (3.39)
R2 = .16 (13)
shopping (which does not take place within the
home) or childcare (which may not be regarded where the values in parentheses are the corre-
as housework). The reported annual hours spent sponding t-values.4
in housework may, therefore, underestimate the The estimated work at home function of em-
extent of work at home. ployed wives is
At this stage of the analysis I focused on white H = 852.51 + 7.526 AGEW
married couples, and since husbands reported
relatively few hours spent in work at home (less (4.63) (3.35)
+ 46. 168 EDUCW + 25.813 EDUCH
than 200 hours in the case of employed males and
about 350 hours in the case of the not employed), (2.25) (2.58)
- 4.614 WAGEH - 1.879 OTINCM
I narrowed the analysis even further to married
women. The Michigan Study of Income Dynam- (0.56) (1.22)
+ 190.080 CLD — 17.494 AGEYC
ics is a panel study encompassing seven years,
1968—1974. In this paper I shall report the (9.58) (3.56)
+ 30.617 ROOMS
findings relating to l973. (1.80)
The set of explanatory variables Y includes the
wife's age and education, the husband's educa-
- 1009.743 EXPWAGE
(5.18)
tion and wage rate, the family's non-earned in- R2 = .158
come, the number of children, the age of the (14)
youngest child, and the number of rooms in the where AGEW denotes wife's age, WAGEH—
house. Education and on-the-job training (i.e., husband's wage, OTINCM—non-earned in-
age) are expected to increase the wife's marginal come, CLD—number of children, AGE YC—age
productivity at home the same way they affect of youngest child, ROOMS—number of rooms,
productivity in the market. Children increase the and EXP WAGE—the imputed wage (in ln).
value assigned to the wife's services and, hence, The estimate confirms our expectations. The
should increase her marginal productivity at wife's age and education have a positive effect on
home, though this increase may taper off as the the wife's work at home. The wife's work at
child grows older. Similarly, the value of margi- home tends to increase with her husband's edu-
nal productivity may increase with the size of the cation but seems to be insensitive to changes in
house as measured by the number of rooms. The his wage rate and non-earned income.5 Children
husband's wage and education and the family's
non-earned income are proxies for the other in- The imputation should reduce the measurement error in
puts in the home production process. wage. However, I ignore throughout this paper the effect of
selectivity biases in the wage and labor supply functions
(Gronau. 1974; Heckman, 1974).
The data are derived from the 1974 panel. Estimates The husband's education may stand for his long-term
based on the 1972 panel yield results that are very similar to earnings and market inputs (and in particular consumer dura-
the ones reported here. An attempt to apply my method to the bles), but it may also serve as a proxy for the wife's ability
1969 panel was less successful. The 1974 sample contained (allowing for selective mating). A third explanation ties the
1,990 observations, out of which 1.022 wives reported that sign of the husband's education with substitution between the
they were employed in 1973. husband's and the wife's time in the production of home
 HOME PRODUCTION_A FORGOTTEN INDUSTRY 411
are associated with an increase in work at home, $10,000 when it had. On the average, it equaled
but this effect diminishes with the age of the two-thirds of total family money income and
youngest child. Similarly, the number of rooms reached 86% of family money income for families
in the house exerts a positive effect on work at with preschool children.5 This value by far ex-
home. ceeded the wife's money earnings, and at least in
the case of families headed by a person with
III. The Value of Home Production elementary education, it is almost as large as the
To estimate the structural parameters off1, I earnings of the husband. Home production is, of
used the estimates of the restricted simultaneous course, tax exempt, and the importance of home
equations method described in the appendix'
production, therefore, is even greater when we
compare it with the family's money income after
lnf11 = 0.599 + .005467 AGEW taxes (the two are almost of equal importance
+ .05829 EDLJCW + .02358 EDLJCH when the family has preschool children).
- .004990 WAGEH Given the positive correlation between the
— .004250 OTINCM + .1934 CLD husband's and the wife's education, it is ob-
— .01875 AGEYC + .05546 ROOMS served that the value of home production in-
— .000926 H. (15) creases with the husbands education (the in-
crease in productivity due to the increase in the
My method does not allow the estimation of wif&s education more than offsets the decline in
the standard errors, but it seems that our expec- the number of hours she works at home). It in-
tations are confirmed—education has a positive
creases. however, at a much slower rate than the
effect on marginal productivity at home (its ef-
family's money income. The value of home pro-
fect is about 70% of that it has on productivity in
the market, as reflected by the wage function duction
families
equals almost 80% of money income in
headed by males with less than eight
(l3)). Children and the number of rooms in the years of schooling hut is less than 60 of the
house have a positive effect on the value of mar-
money income of families whose head had 13
ginal productivity, but the effect of children dis-years of schooling or more.
sipates with their age. It has been observed (Gronau, 1977) that em-
Given these parameters of the household pro-
duction function a, I estimated for each woman ployed married women work less at home than
in the sample her value of home production (us- the not employed. It is worth examining, there-
ing equation (12)). The findings are summarized
f'ore, how much market employment affects
in table 1, where households are classified by the home production. As tables 2 and 2A indicate.
husband's education and the existence of pre- the value of home production in families where
school children. The first four rows of this table the wife is not employed exceeds that where the
follow conventional accounting methods, de- wife is employed by over 20%. Given the differ-
scribing the husband's and wife's earnings and ence in income, home production equals about
the familys money income before and after 80% of money income when the wife is not em-
taxes. The fifth line contains my estimates of the ployed but only 50% when she is employed. The
value of home production. As these figures indi- increase in money earnings due to the wife's
cate, this value is far from negligible. According work in the market exceeds by far the loss of
to these estimates, the average value of home home production when there are no young chil-
production of a U.S. household in 1973 was over dren at home, but this difference almost disap-
$7,500. It was close to $6,500 when the family did pears when the family has preschool children
not have preschool children, and reached almost (the not employed spending much more time in
work at home and the employed working rela-
services. It is surprising, however, in this case that the hus-
tively few hours in the market). It is worth not-
band's wage has no significant effect on the wifes work at ing, however, that even in the case of employed
home.
These estimates should be more efficient than those de- Mv measure ignores the value of home production due to
rived from the unrestricted OLS estimates of equation (14). the work at home of other members of the household )spe-
This estimate is Consistent with Michaels estimates of crfical)y. the husband). However, given the small number of
the effect of education on the efficiency of consumption hours husbands report they work at home, this bias should he
(Michael, 1972. pp. 324—325). relatively small.
412 THE REVIEW OF ECONOMICS AND STATISTICS
women with young children whose husbands
+ —, N N
'C
-
— N r.1 Os r- . ,. 0. have a college education, the value of home pro-
duction exceeds the wife's earnings.
Given the important effect of young children
— e'l
00 —00—'C r—
'C
o _.
+ 00
os
05 00 r-l and the wife's employment status on the value of
——— —
home production, and the changes in these two
factors over the family's life cycle, I compared in
'C 'C r'
' C" '0—fl
— 00 — 'C
CO N 00 05 Os C—
4 'C 00 figure 1 the changes taking place over time in
home production with the age-income profiles
(the age is that of the husband). The age-home
05 00
00 05 r— r-1
production profiles resemble in many respects
+ 00
'C
Cf 'C r- • 'C
those of age-income. They have the same in-
z
verted U shape, they are flatter the lower the
CO 'C C" 00
Cu 'C
00
r-
05 r husband's education, and they tend to funnel out
0.
Os 05 — with age. Still, there are some significant differ-
C ences between the two profiles. The value of
N_ N"05 'Cr 05
z — N 'C 'C ')
C
- 0005
CO
e 'C '0
C' 05 e-i 'C N home production tends to increase with age at a
U
E2 faster rate than does money income during the
C early phases of life (at least when the husband
N 'C C5 05 rfl u- s—
>. + N r- c-i.i Cs —
'C
c-i
N N C' has nine years of schooling or more), it peaks
CO
earlier (during the ages 35—39) and, usually,
'0
MS
N
z
05
—C
—I
ZL)
00
I '-505
N c-i 'C 'C 05 N
00
drops more sharply.9
FIGURE 1.—FAMILY MONEY INCOME AND THE VALUE OF
HOME PRODUCTION OVER THE LIFE CYCLE
—e5"105-C
0N c-i
0
c-i m
_ 0 000 N c--
CO Os c-i
00 —
2 'd— osu-c

C'm050500
0 N c-i 00 0 'CC r5c-iMS
— '1' 1'0 'C
c-i

H +
'C
1'

CC
— MS0\—OMS N 00
r-ic-l0000 0500'C
CO i 00 N —'0 '0 '0
— .-'—c-10s 0 0
N MS
0 =

Li.
C
z —
COc-I 05
Os 'CC'
c-i QN —00 c-i
C '0 — 'C '5 's 'C'C
N— 'f C
0 MS
F-
'C I0
-J
'C
>

CO
-J
CO
-C
1-
CC
Co
C)
CC
'C 'C
E
00
0 C)00
C •u
i.... C
U —
0.0.
C C I- C)
CO .C
.0 U
us
00
> 'C
0 C
0 Husbands Age
C) I-

'
C— C C 0. E
C
C) C) . . CC
00 - C E E
C
Cu EC)
Computing the ratio (value of household production)/
C 0 us i_ CC 0.0.0 0.0. 'C- CON (family money income) for each household and averaging
C u -—CC over all the households in the age-education group, it is ob-
4- > COO)
served that this ratio increases up to ages 35—39 and declines
C
0
C)
U
C C) CC 10 0
E .0.0 E-'E
00. thereafter. On the other hand, had I computed this ratio using
'C 00 C'OCOCO OCO Cd CO CO CO
U, the means in each group, one would not have observed an
 HOME PRODUCTION—A FORGOTTEN INDUSTRY 413

'C
1 C-I 0' fl 00
C- 00 . C-
+ C-flf-l0C,
— I-I—
t 1 r'
C-I 00
—
'C
C-I +—
,
Ir
000'C
Cl C-I
f-I '
CI 'C

— C-I-
't 0
'C C- ' f-I 00
'C
rC- —

a' a' 'C C-

00
'C +
00
a' 00
0Io_o.
C-I ' - 00
00
C-
0 ' 'C CI0
00
— — 0 0 'C 0 —

H
—
C-fr— a' '- a'r
C- 'C 'C —
CO
'C a'—00
00 C- r—
jC-C- 'C. ' fl
C
00
00 00
C-I

-oooo 'r.
L
C—)

+ 0rC-C
'o
00 - V
00
'1 Cl
t ',
C-I '— >- * .—
00
a'
I—°--
C10
Cl 'C 'C
C—
r-
—
rI- 00
00
C-I
'C a' r— o
C—

z C
C Cl C- ' r— C-I 0 C-I
II.
Cl
a' — 0sr-i--—r-- 0
a' 'C 'C C- 00
'C
a' ('I
:; r— C-I
-i —
a'
CIa'0C
a'
--
C-I a' —
'0— 00

I
—
—c
C C-00000 C-I
2'C 0 r- 'f
f-I r- f-I — -' C-Clr
I'°v'
a'
0'
CI — 'C 't a'
0
0 ",? 'C C- 'C

1'
—

z
CC 'C
C- ''t I'Cr00 fl.,-
oo,tc— 'C
00
., 0CC-I
f-I
C-
'C C- 0'
o--
Cl
a'
a'
0a'C-fl
Cl 'C
0 C-
o-i- '
00
. 00 0 00 C-00
, rf-0' — r—

C
00 CI
00C
— 'Cra' 0 a'—
eC rfl "l' — (-I LCJ.. —
00
I
r— 'C C-I 0'
0
I 00 — C- rf- 0' -
0 C-i ? rC a'
r,
C-
0
'0

'
— 00 0' a' — '
—
aor-
'C'f-I—' 0
0
00
r a'
—
C-I 7-"-
C
— a't a' 00 0
a'r
0 0r
0 a'
'C
f-IC-C-I
0'

fl — C—

a' f-l
0 00 0

+
'C
C-''O
— 00
'Cf'l'C '
Cl
' Cl
C- C- rfl
C-I
00
—
C- + 00
'C a'
—.. —
oo- C-
C- f-I

—
'C"C-I— o — 'C — — C-0'
',
CC '0
0 00
— Cl
CO
.,- a' C-
'C- C-I a'
— 0Cl00 C-0 00 CO
0'
C0- F— —
I
-

CI-
J-
C 0C
— 'C 00 f-I
0Ca''C0C
OC
Cl C-
o
'C C- •1f-I
C-j 0
0
— 0 C-0C-
CO C- rC- 00 'C 0
00
'C
00
C-
'C 00
'C
0'
—
C- 0
C
C
z-
z
C
'C
-C
'C
> LL

0?
Cl
CO

I
'C
V I
'C

--
C
F—

CO .C

C0 I),
00
. 0-'0
0
C
.0 U
CO

.0
CU

C
o CU 'C
(I

l 0C1_ V
C
9 > C C 0
o > CO
- C— C
CU
V
-c
•00?
0?
CU 0
414 THE REVIEW OF ECONOMICS AND STATISTICS
IV. Some Concluding Comments puts in home production are assigned the prices
the household would have to pay had it pur-
This paper is a first expedition into an un- chased the services in the market. According to
explored land. Thus, it naturally has its the second approach (Sirageldin, 1969; Gronau,
shortcomings. Several reservations may have 1973), the criterion used is the costs the house-
been too easily brushed aside; others may have hold assigns to the marginal unit of its members'
been completely overlooked. Selectivity biases time. Both approaches seem equally unsatisfac-
may play a much more prominent role than ad- tory.
mitted. Entry costs (e.g., search and transporta- The market prices approach is deficient in the
tion costs) have been completely ignored. So has case of corner solutions—the case when the
an issue that has plagued followers of the new household does not consume the market ser-
household economics for a long time, namely, vices. In this case, the market prices approach
joint production (in our case the simultaneous assigns to home production prices which have
usage of time for work at home and leisure ac- been explicitly rejected by the household as a
tivities). Our data are far from ideal, and we tend true measure of its productivity. The family
to underplay the importance of husbands and could have bought the home services in the mar-
other family members in home production. ket but preferred not to do so, either because it
Still I feel that the method described in this found their prices too high, or because it found
paper provides a powerful tool to identify and their quality wanting. The "opportunity costs"
estimate the value of household production. My approach, on the other hand, ignores the fact that
first attempts in this area are highly encouraging. a person working at home is, in essence, self
At this stage I have focused on married white employed, contributing to the production pro-
families. A lot still has to be done. One should be cess both his labor services and his entrepreneu-
able to extend this work to compare the home rial capacity. The "producer surplus" generated
production of married and single persons, whites by homemakers is ignored in the conventional
and nonwhites. More important, the household "opportunity costs" calculation.
sector (or the nonmarket sector) plays a much To what extent can our estimates improve cur-
more important role in the less developed coun- rent estimates of the value added? The contribu-
tries than in the developed countries. The appli- tion of this paper to the improvement of the
cation of the method to less developed countries aggregate estimates can be only modest. To ob-
should yield better measures and understanding tain estimates of the value added generated by
of the process of growth and the gap in material the home production process, one has to subtract
resources between the developed and the less from the estimates of total output the value of
developed states.'° market inputs. These inputs are, however, un-
Finally, the focus of this paper was on produc-known. Still, for comparative reasons I listed in
tivity and the value of home production. A the tables my estimates of the value of labor
natural question that comes to mind is, What is inputs in home production according to the "op-
the value added in this production process and portunity costs" approach (the wage assigned to
what can he learned from our estimates on the the woman's time was generated by equation
estimates of the aggregate output of the home (13)).'' These estimates are only about one-half
sector? Existing estimates of the contribution of as large as our estimates of total output. They
the home sector to total output are based on two seem to underestimate the effect of young chil-
different approaches. According to the first ap- dren on home production (in absolute as well as
proach (Walker and Gauger, 1973), the time in- in relative terms). They understate the loss of
home output when the woman works in the mar-
ket (in particular when she has young children),
increase in this ratio over the first phases of the life cycle
(hence. it is difficult to detect this tendency in figure I).
0 In some of the less developed countries the wage sector
plays only a minor role. In this case, one should replace the A recent survey of estimates of aggregate home produc-
wage variable by a variable representing the marginal produc- tion (Hawrylyshyn. 1976) has not found significant differ-
tivity in marketable farm products (or in cottage industries). ences between the "market prices" and the "opportunity
and the wage function by an agriculture production function. costs'' estimates.
 HOME PRODUCTION—A FORGOTTEN INDUSTRY 415
and the changes in home output over the life Walker, Kathryn E., and W. H. Gauger, "The Dollar Value
cycle. of Household Work," New York State College of
The difference between our estimates of the Human Ecology, Cornell University, 1973.
value of home production and the labor inputs
can be attributed to market inputs, the capital
services of consumer durables and to the wife's APPENDIX
entrepreneurial capacity. Casual observation in- To test the assumption that the marginal productivity-at-
dicates that the variability in market inputs with home function (equation (9)) is semi-log, let it be assumed
that the marginal rate of substitution of goods for leisure .5 is
age of children and education is relatively small, also a semi-log function (of leisure and other variables)1
the differences in home production originating
presumably in the other two factors. Conven- Ifls=$0—/31L+/32y. (Al)
tional measures tend, therefore, to understate Given the equilibrium condition (6)
considerably the value added, not giving (A2)
sufficient credit to the entrepreneurial capacity which can be estimated in the sample of employed women by
required at home.'2 the semi-log regression
L=h0—b,lnW+h2y (A3)
REFERENCES where
Gronau, Reuben, 'The Measurement of Output of the Non- (1,/b1> = est(/31).
market Sector: The Evaluatipn of Housewives'
Time," in Milton Moss (ed). The Measzire,nc'nt / (h0/h) = est(f30), (A4)
Economic and Social Performance (New York: Na-
tional Bureau of Economic Research, 1973), 163—189. (h2/b ) = est(/32 ).
Wage Comparisons—a Selectivity Bias," Journal Furthermore, when a person does not work in the market
of Political Economy 82 (Nov/Dec. 1974), 1119—1144. N= H+ L=
Who is the Family's Main Breadwinner'—The 0, T, and by equation (7).
Wife's Contribution to Full Income," NBER Working H = [($T + ao —
/3)
Paper No, 148, Sept. 1976.
Leisure, Home Production and Work—The Theory
+ (a2 — )Y}/(a - (AS)
of the Allocation of Ti me Revisited," Journal of Pu/it- This equation can be estimated in the sample of the not
ic-al Economy 85 (Dec. 1977). employed
Hawrylyshyn, Oh, "The Value of Household Services: A H=c+
Survey of Empirical Estimates," The Revieic of In- (2Y (A6)
caine and Wealth, series 22. no' 2 (June 1976), 101- where
131,
Heckman. James J.. "Shadow Prices, Market Prices and = est[(/31T + ao
—
(3)//a +
Labor Supply." Econometrica 42 (July 1974), 679— and
694.
Michael, Robert T., "Education in Nonmarket Production," = est[(a2 —
f32)/(a1 + (3)].
Journal of Political Eeono,ns- 83 (Mar/Apr. 1973), (A7)
306—327. There is no way of isolating the parameters a and 13, from the
Nordhaus, William D., and James Tobin, "Is Growth Oho- estimates of c0 and c. These estimates provide us. however,
lete?'' in Milton Moss (ed. ), 1/ic Mea,uure,nent of wi a test of consistency. Deriving the estimates of a, and (3,
Economic and Social Performance (New York: Na- using equations (10) and (A3), then by equations (II) and
tional Bureau of Economic Research, 1973). 509—531. (A4),
Sirageldin, I. A., Nonmarket Co,nponeot oJ'5'./olio,ia/ inco,ne + — ab0
(Ann Arbor: University of Michigan Press, 1969). -—
— a1T b1a0
=est ( (31T + -— A \
a1+b, cs+/3
(A8)
2
ba2—a,h2 =est'°°'(32
To illustrate, assume that market inputs into home pro- 13
duction are represented by the family's consumption at food,
Under this assumption, when food and labor inputs are sub- aod one can compare the coefficients of equation (A6) (the
tracted from home production the residual would, on the work at home equation of the nonparticipants) with their
average, come to about one fifth of our estimate of home predicted values (equation (A8)) to generate a powerful test
production: for a mother of preschool children the residual is for the validity of the estimates and our assumption concern-
around 30% (irrespective of education and employment ing the functional form.
status) and it is about 10% for women whose youngest child is
aged 6 or more (the percentage increasing slightly with wife's
education). Ignoring the services of consumer durables, this
would imply that entrepreneurship at home is more important The vector I in (Al) and (8) need not necessarily have
when there are young children, and the more schooling the the same components (i.e., some of the elements °a2 and 132
wife has. may be assumed to be zero).
416 THE REVIEW OF ECONOMICS AND STATISTICS
Alternatively, one can pool the sample of participants and Hence,
nonparticipants and estimate the following set of simulta-
neous equations
A0 + B0 = T, A2 + B, = , A'0/A', = B'0/B'1,
= B'2/B1. (All).
H=A0+A2Y+ Ix (A'0—A'1lnW+ A'2Y) (A9) To derive the estimates of a and one should estimate the
L = B0+ B2Y+ Ix (B'0— B'1lnW+ B'2Y) two equations of (A9) simultaneously subject to the non-
linear restrictions imposed by (All).
Comparing the explanatory power of the restricted simul-
taneous system with the unrestricted ordinary least squares
where I is a dummy variable denoting that the person partici- (OLS) equations should yield an indication of how binding
pates in the labor force. By equations (9), (A2), and (AS), the theoretical restrictions are.
A0 = est[(f31T0 + a0 — f1)/(a1 + (3)], Applying the two consistency tests to our data it is found
A2 = est[(a. — /32)/(aI + f3)I, that they withstand these tests surprisingly well.2 The esti-
A0 + A'0 = est(tz/a1),
mates of equation (15) are derived from (A 10):
A'1 = est(l/a1), 1/A'1 = est(a1),
A2 + A'2 = est(a2/a1), (AlO)
B0 = est[(a1T0 + 0 —
B2 = est[( — 2)/(1
)/(1 + )J,
+
(A0 + A'0)/A'1 = est(a0), (A12)
p1)1, (A2 + A'2)/A'1 = est(a2).
B0 + B'0 = est(f30/),
B'1 = est(l//31), 2 For details see an earlier version of this paper (Gronau,
B2 + B'2 = est(fl2/1). 1976).