Journal of Comparative Economics xxx (xxxx) xxx–xxx

Contents lists available at ScienceDirect

Journal of Comparative Economics

journal homepage: www.elsevier.com/locate/jce

Empirical linkages between good governance and national well-

being☆
⁎
John F. Helliwella, Haifang Huangb, Shawn Groverc, Shun Wangd,
a
Vancouver School of Economics, University of British Columbia, Canada
b
Department of Economics, University of Alberta, Canada
c
Infrastructure Canada, Canada
d
KDI School of Public Policy and Management, 263 Nansejongro, Sejong, Korea 30149

AR TI CLE I NF O AB S T R A CT

JEL classifications: This paper brings together the largest available sets of national-level data, covering 157 countries
H11 over the years 2005–2012, to assess the extent to which governance quality contributes to life
I31 evaluations. Our most significant new finding is that changes in governance quality within a
P50 policy-relevant time horizon can lead to significant changes in the quality of life. For example,
Keywords: the ten most-improved countries, in terms of changes in government service delivery quality
Subjective well-being between 2005 and 2012, when compared to the ten most-worsened countries, are estimated to
Happiness have average life evaluations higher by 0.4 points on a 0 to 10 scale. The results also confirm
Worldwide Governance Indicators
earlier findings that service delivery quality generally dominates democratic quality in sup-
Democracy
porting better lives until delivery quality has reached sufficient levels. The situation changes as
Corruption
Government effectiveness development proceeds, with democratic quality showing a positive influence among countries
Gallup World Poll that have already achieved higher quality of service delivery.

1. Introduction

This paper argues that people's own evaluations of the quality of their lives provide reliable and inexpensive new ways to show
how much the quality of government matters. These well-being measures complement and encompass more established indicators of
economic and social progress. Their encompassing capacity is based on their focus on life as a whole, thereby permitting economics,
health, trust, freedom and social relations to be consistently taken into account, using survey-based life evaluations as the research
base to establish what matters most.
The paper then brings together the largest available sets of national-level measures of the quality of governance, including
indicators of government institutions’ effectiveness and those of democratic rights, and assesses the extent to which they contribute to
explaining the levels and changes in life evaluations in 157 countries over the years 2005–2012, using data from the Gallup World
Poll.
Previous evidence, reported in Bjørnskov et al. (2010), Helliwell (2003), Helliwell and Huang (2008), Ott (2010a,b, 2011), shows
that people are more satisfied with their lives in countries having better governance quality. But all these studies estimate the effect of
governance quality, in full or in part, by comparing international differences in levels of happiness and governance. This is mainly a
consequence of data limitations. Until the recent arrival of Gallup World Poll (starting from 2005), the integrated World Value

☆
The authors are especially grateful for the helpful advice of the referees and editor, and to the Canadian Institute of Advanced Research for its Social Interactions,
Identity and Well-Being Program, within which this research was conceived and carried out.
⁎
Corresponding author.
E-mail address: swang@kdis.ac.kr (S. Wang).

https://doi.org/10.1016/j.jce.2018.01.004
Received 31 August 2017; Received in revised form 12 January 2018; Accepted 17 January 2018
0147-5967/ © 2018 Association for Comparative Economic Studies. Published by Elsevier Inc. All rights reserved.

Please cite this article as: Helliwell, J.F., Journal of Comparative Economics (2018), https://doi.org/10.1016/j.jce.2018.01.004
J.F. Helliwell et al. Journal of Comparative Economics xxx (xxxx) xxx–xxx

Surveys and European Surveys (WVS/EVS) had the biggest international coverage of happiness measures. They had fewer than 50
countries in their first two waves (1981–1993), and were expanded to about 55 countries in wave 3 (1994–1998), about 75 in wave 4
(1999–2004), and about 100 in wave 5 (2005–09). Helliwell and Huang (2008) used 160 observations from 75 countries in waves
1–4; it is fair to say that most of the variations were cross-sectional differences in levels. Bjørnskov et al. (2010) used waves 1–5; but
they too estimated the effects of institutions using pooled-OLS estimations without country fixed effects, which means that cross-
sectional differences in levels played a role, likely a non-trivial one, in their identification. Ott (2010a,b) and Ott (2011) used cross
sections in 2006 covering about 130 countries, drawing data from the World Database of Happiness (Veenhoven, 2008), which in turn
combines data from multiple sources, likely including the WVS/EVS and the early waves (2005/2006) of the Gallup World Poll.
The existing evidence, though informative, is still limited. One drawback is the primary reliance on international comparisons in
levels. A range of country-specific historical or cultural factors may have led some countries to have low quality governance and high
levels of well-being. For example, Latin American countries tend to report levels of subjective well-being that are higher than average
(for example, Bjørnskov et al., 2010) or higher than model predictions (for example, Helliwell et al., 2015, p.29). But Latin American
countries have low levels of governance; in the Gallup World Poll, the group of Latin America and Caribbean countries on average
have 80% of their respondents perceiving corruption in business and governments, vs the average of 76% in about 150 countries; only
42% express confidence in their national governments, vs the world average of 49%. It follows that the inclusion of Latin American
countries weakens the relation between institutional quality and happiness in cross-sectional studies, unless a regional dummy in
included to force the regressions not to make level comparisons between Latin American countries and other countries. The use of
regional dummies assumes however that the within-region variations are “good” variations that can help pin down the governance-
happiness relationship, while the between-region variations are the “bad” ones that should be avoided. Such an assumption, although
reasonably justifiable on historical, linguistic and cultural grounds, invites further analysis. Stronger evidence can be obtained by
looking at changes over time within the same countries, so that we are looking at changes in governance quality for the same national
governments, instead of differences in governance quality between different countries in the same geographic or cultural group.
Analysis of changes thus permits us to eliminate the effects of long-standing country-specific factors that might otherwise confound
the governance-happiness relationship.
Analysis of changes has at least one more benefit. A positive relationship in cross-sectional analysis does not tell us whether
institutional quality can be changed in a relatively short time, in years instead of scores of years, and whether the change can have an
impact on subjective well-being. Analysis using changes thus provides a greater degree of policy relevance, as it can potentially reveal
where improvements of governance have been made, and perhaps to establish the extent to which they provide useful lessons for
others.
Change analysis still has limitations. We cannot preclude the possibility that some other underlying forces are driving both
changes in the perception of life quality and changes in the indicators of governance quality, some of which can be influenced by
perception as well, such as the perceived levels of corruption. We try to alleviate these concerns by drawing data from different
sources, with happiness measures coming from the Gallup World Poll and indicators of the quality of government from elsewhere
(mostly the World Bank's WGI data, but we will present findings from alternative sources as well). We will also experiment with
regressions that have other subjective variables from the Gallup surveys to absorb some of the perception changes. But these lim-
itations are equally or more present in the cross-sectional evidence. The change analysis thus provides a net increase in the strength of
the evidence linking governance quality to subjective well-being.
We also recognize that improvements in governance can contribute to national well-being in multiple ways. Control of corruption,
for example, potentially promotes economic growth and helps maintain social trust at the same time, both beneficial to national well-
being. Part of our research objective is thus to estimate the economic and non-economic channels linking changes in governance and
changes in happiness, and to evaluate the relative contributions.
Existing results also show that the delivery or technical quality of government services dominates democratic quality in sup-
porting better lives (Helliwell and Huang, 2008; Ott, 2010a,b, 2011). There is, however, some difference in relative importance as
development proceeds, with democratic quality having a positive influence among countries that have already achieved reasonably
high per capita incomes. This paper revisits the evidence using an expanded dataset, and explores whether similar hierarchical
relationships are found when countries are grouped by the level of service-delivery quality instead of by GDP per capita.
The rest of the paper is organized as follows: Section 2 reviews the theoretical links between good governance and national well-
being and reviews the literature. Section 3 describes the data and our empirical models. Section 4 presents our empirical findings.
Section 5 concludes.

2. Theory and literature review

2.1. What connects good government to well-being?

There are several types of theory that have been used to underlie linkages between good governance and well-being. Good
governance may improve life evaluations either directly, because people are happier living in a context of good government (Frey and
Stutzer, 2005; Ott, 2010b), or indirectly, because good governance permits people to achieve higher levels of something else that is
directly important to their well-being. Mixed cases are likely, as the implicit definition of good government is likely to include the
capacity to provide instrumental support should the need arise. Hence an important question that we ask in our later empirical
exercise is whether good government improves well-being directly, or does so mainly or entirely indirectly. Examples of both pos-
sibilities may be found. Consider the control of corruption, which has been shown to affect well-being both directly and indirectly.

2
J.F. Helliwell et al. Journal of Comparative Economics xxx (xxxx) xxx–xxx

Absence of corruption has often been shown to increase the efficiency of public and private enterprise and thus create favorable
conditions for economic growth, but there is also lots of evidence that the higher levels of general and specific trust make people
happier above and beyond higher incomes.
To help to distinguish direct from indirect linkages between good governance and well-being at the national level, we shall
consider two main types of correlational evidence, first the simple relations and then models that control for intermediating factors to
sort out the likely channels of influence. As an empirical matter, we cannot hope to control for all intermediating factors linking
governance quality to happiness. The main factor we control for is economic outcomes, which allow us to compare the economic vs
noneconomic channels whereby governance affects the quality of life.
Before broader measures of well-being were sufficiently widely measured and understood, the effects of good governance were
usually assessed by searching for linkages between governance and economic outcomes, and treating these economic outcomes as
proxies for well-being more generally. The sufficiency of economic growth as an indicator of well-being has been questioned in two
quite different ways, one emphasizing the uncertainty of the links between income and well-being, and the second arguing that other
factors than income cannot be ignored, since together they might be more important than income. Especially relevant to our study of
governance and well-being, governance may affect income and these other factors in quite different ways.
First, there is the proposition most famously presented by Richard Easterlin and known as part of the Easterlin Paradox
(Easterlin, 1974), that as countries become richer, they might not become happier, because increased incomes cause associated
increased material norms. Other scholars have found little strong evidence of a direct connection between growth and SWB in middle-
income countries and theorized that this could be because increased income also is associated with “negative indicators of life quality
such as increased pollution, the social costs of economic transformation, the importance of positional goods and the dominance of
country- or locally-based relative income concerns” (Kenny, 2005). While the Easterlin Paradox has been qualified by a variety of
studies showing that when income differences are represented in proportionate terms they are frequently found to have quite
comparable effects in and among countries at all stages of development,1 whether national-level average incomes and happiness
move in the same or different directions depends a great deal on the countries and time periods chosen for comparison (see for
example, the contrasting results of Easterlin and Sawangfa, 2010 and Sacks et al., 2012).
Second, the evidence and arguments supporting the use of subjective well-being do not rest on the size or constancy of the link
between income and happiness, even if much literature has been devoted to that issue. The primary reason for looking beyond
economic outcomes is not that the linkage between incomes and happiness is insecure and variable, but that it excludes too many
other things that are fundamentally important for better lives.
Another way of putting this second point is that the most fundamental explanation for the Easterlin paradox - if this paradox is
represented by cases where income has risen a great deal yet average happiness has declined or been stagnant - is not that the effects
of income on happiness are absent, or even differ among countries, but because other things critical to happiness have changed in the
opposite direction. Thus it has been argued that Easterlin's widely quoted example showing much post-1970 economic growth in the
United States without any increases in average happiness can be well explained by a model in which income has significant positive
impacts on happiness that have been offset by declines in other key supports for well-being, including especially in social trust and
the quality of social connections.2 In the same vein, when average changes in life evaluations from 2005–07 to 2010–12 are examined
for 130 countries, three-quarters of the explained changes in average life evaluations were due to factors other than income, with one-
quarter due to income.3 And this period spans the largest (and unevenly distributed) economic recession in the past seventy-five
years. Income changes were indeed a significant part of the story, as would be expected in such circumstances, but nonetheless other
factors together played a much larger part.
Additionally, better governance improves well-being beyond any impacts it has on income or transaction costs. One potential
channel is through “procedure utility” developed in Frey and Stutzer (2005), the notion that people care about more than just
outcomes but also the decision making process leading to those outcomes. Frey and Stutzer (2005), by comparing the life satisfaction
of citizens and foreigners living in different cantons (states) in Switzerland, report evidence of people gain utility from having the
rights to participate in the political decision making process. Ott (2010b), focusing on the technical quality of governments (effec-
tiveness, regulatory quality, rule of law and control of corruption), suggests that such quality can contribute directly to well-being, on
top of indirect "output utility", because it affects the quality of contacts between government and citizens.
Good governance not only affects contacts between citizens and governments, it also affects the quality of contacts among
citizens. One such channel is by enhancing social trust in general. Evidence in the literature has shown that lives are happier in
communities where people feel that they can trust others, including police, neighbors, work colleagues and strangers (Helliwell and
Putnam, 2004; Helliwell and Wang, 2011; Helliwell et al., 2014b). Those perceptions of trustworthiness, in turn, can be influenced by
the quality of governmental institutions (see experimental evidence reported in Rothstein and Daniel, 2009). There is further

1
The Gallup World Poll data for more than 150 countries have been found to show income effects that are quite similar for countries at different levels of
development (Deaton, 2010; Sacks et al., 2012; Helliwell et al., 2010). This stands in contrast to earlier findings based on the World Values Survey that suggested
income effects to be smaller in OECD than in non-OECD countries (Helliwell, 2003).
2
See Bartolini et al. (2013), Lane (2001), and Layard (2005). Their analysis considers the negative influences from declines in trust and social connections, and
relative income effects of the sort emphasized by Easterlin.
3
The changes from 2005-07 to 2010-12 in national average life evaluations are shown in Fig. 2.5 and 2.6 of Helliwell and Wang (2013). When these changes are
explained by changes in the six factors used in Table 2.1 of the same chapter, 25% of the variance is predicted by the model, with one-quarter of that coming from
changes in income and the other three-quarters coming from the combined effects of changes in the other five factors. A similar model is used later in this paper to
provide an updated assessment of the linkages between governance and well-being.

3
J.F. Helliwell et al. Journal of Comparative Economics xxx (xxxx) xxx–xxx

evidence, based on natural experiments of migration from countries with lower-quality institutions to those with higher quality ones,
that improvements in institutional quality enhance social trust (Dinesen, 2013), and that institutional differences dominate cultural
differences in determining the levels of social trust (Nannestad et al., 2014). There is also substantial evidence that international
differences in the quality of governance play directly into these well-being consequences of social trust, as suggested by the fact that
when people are asked about the chances of a lost wallet being returned intact if found by a police officer, there is a much bigger
difference in answers across nations for this variable than there is about the likelihood of a lost wallet being returned if found by a
stranger or a neighbor.4 Furthermore, differences of social trust have been found to also play a significant role in explaining other
outcomes that affect well-being through a variety of non-income channels. For example, differences across countries in social trust are
significantly correlated with differences in both traffic fatalities and suicide rates.5

2.2. Existing evidence on the link between governance and subjective well-being

At the aggregate level, several studies have compared the well-being links between two major sets of government characteristics
and average life evaluations. The first set of characteristics relates to the reliability and responsiveness of governments in their design
and delivery of services, referred to here as the quality of delivery. The second set of characteristics relate to the presence and
pervasiveness of key features of democratic electoral elections and representation. This we refer to as democratic quality.
The first multi-country evidence was from three waves of the World Values Survey, based on 46 countries with a heavy re-
presentation from OECD industrial countries. The key results, using a model that included both individual-level and national-level
variables, were that cross-country differences in life satisfaction were very strongly linked to international differences in an average
of the six World Bank governance quality measures: government effectiveness, regulatory quality, rule of law, and the control of
corruption, voice and accountability, and political stability and absence of violence. In these fully specified models, this significant
effect of government quality suggested an importance beyond that channeled through other outcomes. By contrast, per capita GDP
and education levels, although significantly correlated to life satisfaction in simple terms, dropped out of the more complete models,
suggesting that their impact was mostly instrumental. These initial results were, however, based on relatively few countries, and the
model also included significant regional effects that were possibly also indirectly picking up some of the effects of differences in
income and other factors (see Table 6 in Helliwell, 2003).
Subsequent work extended the number of countries and years being considered, and split the six World Bank governance mea-
sures into two separate quality measures. The quality of delivery was defined as the average of four measures: government effec-
tiveness, regulatory quality, rule of law, and the control of corruption. The quality of a country's democratic processes was based on
the average of the remaining two World Bank measures: voice and accountability, and political stability and absence of violence. The
results showed that for all countries taken together the quality of delivery mattered more for well-being than did the presence or
absence of democracy. The quality of delivery was strongly important for all groups of countries, while the democracy variable had a
zero effect for all countries as a group, with a positive effect among richer countries offset by a negative effect among the poorer
countries (Helliwell and Huang, 2008).
Subsequent studies using ever-larger country samples, and a variety of survey sources and life evaluations, have generally sup-
ported this ranking of the relative effects of the delivery and democratic aspects of government quality (see, for example, the
comparison between democratic and technical quality in Ott, 2010a,b, 2011, and the findings in Bjørnskov et al., 2010). Ott (2010a,
b) use the term “technical quality” to emphasize the non-controversial aspect of government effectiveness, regulatory quality, rule of
law, and the control of corruption, relative to the democratic quality or the size of governments.
One thing that has changed over the past decade, as country samples have embraced more countries and more years, is that GDP
per capita has become a stronger element of the full model than it was in the results based on the first three rounds of the World
Values Survey. Since the more recent work has been based on the Gallup World Poll, which uses the Cantril ladder form of life
evaluation, it was first thought that this difference was because the ladder form of life evaluation was responsible for the higher
weight found for per capita incomes, when compared to the life satisfaction responses in the World Values Survey (Diener et al.,
2010). However, inclusion of both forms of evaluation in the same Gallup surveys has revealed that both attach the same weight to
income.6 Thus the larger role given to income differences in more recent studies must instead be attributed to the greater number of
countries involved, which has correspondingly broadened the range of national income differences under review.
One complication needs to be addressed at the outset. While much of the literature argues or assumes that causal forces run from
the quality of governance to levels or changes in subjective well-being, there is the possibility of causal arrows running also in the
other direction. For example, there is a large literature showing that those who are happier tend thereafter to have longer, healthier,
better-connected and generally more successful lives (see Guven, 2011 and the recent survey by De Neve et al., 2013). If these
feedbacks are always positive, then the total effects of a policy improvement may exceed the initial direct effects, although there is a

4
Table 1 of Helliwell and Wang (2011) shows the international share of the variance to be about twice as high for answers about the police as it is for the parallel
questions relating to neighbours and strangers.
5
See Helliwell and Wang (2011, p. 50-51). Traffic fatalities and suicide are roughly tied as the tenth leading cause of deaths around the world, and effects of
international trust differences are substantial. For both suicide and traffic fatalities, a change in social trust of 0.1 on a scale running from 0 to 1.0 is associated with a
10% change in both death rates.
6
See Helliwell et al. (2010), Tables 10.1 and 10.3 for the matched analysis. Indeed, even though the means and distribution shapes of the answers to the two forms
of question (the life ladder and life satisfaction) differ significantly (Helliwell et al., 2010, Figs. 10.1 and 10.2), the determinants are so similar that stronger models are
obtained by averaging the answers to the two questions (as shown in Tables 10.1 and 10.3).

4
J.F. Helliwell et al. Journal of Comparative Economics xxx (xxxx) xxx–xxx

related risk of over-estimating these direct effects while underestimating the combined direct and indirect effects.
Another theoretical complication is posed by the variety of individual tastes about what constitutes good government. Citizens
and voters may differ a lot amongst themselves about the purposes of government, and about how well any given government is
doing its job. It has been shown that people in OECD countries are happier when and if the government in power is one that matches
their own political preferences (Di Tella and MacCullough, 2005), presumably because the views of those voters and their govern-
ments are more likely to be closely aligned. Theory would suggest that average happiness is likely to be higher where different groups
of citizens have fairly consistent views about what sort of government they want. There will likely also be differences across nations
and cultures in what constitutes good government, and hence some ambiguity in measures of government quality that are inter-
nationally comparable, and that have similar effects on well-being in all countries. This increases the value of finding links between
changes in governance and changes in well-being, but it will remain true that we are measuring the average size of the effect. It will
be therefore useful to see how these linkages differ among groups of countries at different stages of development, and among different
income groupings within nations. Some assurance of cross-cultural comparability of life evaluations and their determinants has been
provided by individual-level estimation of well-being relationships in each of more than 150 countries showing a high degree of
consistency on what determines subjective well-being across the world.7

3. Data and methodology

3.1. Data and variables

To confirm and extend earlier results reported in the literature, we make use of a larger set of years and countries than has
previously been available. This involves data from 157 countries, with each country represented by as many observations as there
were Gallup World Poll surveys conducted in that country between 2005 and 2012. By combining time series and cross-country
evidence for a large number of countries and a reasonable number of years, we have a real chance of identifying linkages between
changes in the quality of governance and changes in subjective well-being.

3.1.1. Subjective well-being

Our primary interest is in measures of subjective well-being, and especially in how people value their own lives. Within the general term
‘subjective well-being’ there are three main categories: life evaluations, measures of positive affect (or emotion) and measures of negative
affect (or emotion). The most widely available comparison of these three types of subjective well-being is provided by the Gallup World Poll,
and summarized in the World Happiness Report and the World Happiness Report 2013. The main data utilized in this paper are Gallup World
Poll for the survey years from 2005 to 2012. The Gallup World Poll life evaluation makes use of the Cantril ladder scale, asking respondents to
think of their lives as a ladder, with the worst possible life for them as 0, and the best possible life as 10. The English wording of the question is
“Please imagine a ladder, with steps numbered from 0 at the bottom to 10 at the top. The top of the ladder represents the best possible life for
you and the bottom of the ladder represents the worst possible life for you. On which step of the ladder would you say you personally feel you
stand at this time?” They are then asked to use this scale to evaluate their current lives. Measures of positive and negative affect, on the other
hand, usually come from survey questions about current or recent emotions.
The OECD has recently issued guidelines for the collection of subjective well-being data by national statistical offices. The
guidelines advise collecting all three types of subjective well-being data (see OECD, 2013 and Durand and Smith, 2013). If there can
only be a single measure, preference is given for a life evaluation, in the form of a life satisfaction question on a 0 to 10 scale. The
reasons for concentrating on a life evaluation are the same as those presented in the first World Happiness Report. The first reason is
that while emotions vary a lot among individuals, variations are relatively smaller at the national level.8 Second, and even more
importantly, a much higher proportion of the international variation in life evaluations, compared to either positive or negative
affect, is explained by differences in established measures of the quality of life.9
Thus for the purposes of assessing the links between good governance and well-being, we rely, as has most earlier research, on life
evaluations as the preferred measure of well-being. We shall not attempt to survey the vast literature linking various types of
institutions, including especially government ones, to either the levels or rates of change of GDP per capita (for a wide-ranging review
of that literature, see Acemoglu and Robinson, 2012). We shall, however, when explaining the correlations between governance and
subjective well-being, try to assess the extent to which economic growth is a mediating factor.
Subjective well-being, and especially the notion of happiness as a sufficient measure of well-being, has been criticized as being too
subject to adaptation, peer effects, survey context,10 and to lack of due attention to the basics of a good life. These are all important

7
Helliwell et al. (2010) Fig. 10.3 shows the cross-country distributions of the estimates of the life-evaluation effects of key variables.
8
Thus of all the global variation among individuals in their responses to the Cantril ladder life evaluation, 22% was among countries (and hence 78% among
individuals in the same country), compared to 7% among countries for positive affect and 4% for negative affect. See Table 2.1 in the first World Happiness Report.
9
For a sample of 732 national observations, comprising several annual average observations from each of 149 countries, three-quarters of the pooled variance of the
Cantril ladder is explained by six key variables. This is to be compared with less than one-half for positive affect and less than one-quarter for negative affect. See
Table 2.1 of World Happiness Report 2013. Table 3.1 of the first World Happiness Report shows a similar result for cross-sections of national averages.
10
Deaton (2012) emphasizes the potentially swamping effect of these effects by reference to whether life evaluations in the Gallup Daily Poll were asked after a set
of political questions that sharply reduced life evaluations. Similarly, Bonikowska et al. (2013) show a variety of contextual effects on answers to the life satisfaction
question in several large Canadian surveys. However, that paper also shows how it is relatively easy to measure and adjust for these contextual effects, and thus to
effectively combine data from different surveys and survey contexts.

5
J.F. Helliwell et al. Journal of Comparative Economics xxx (xxxx) xxx–xxx

points, but recent research tends to suggest that none of these problems are fundamental.11 Hall (2013) has argued that the human
development approach and the HDI are complementary with subjective life evaluations, since the latter are influenced by the key
underlying supports for the human development approach, while also providing an umbrella measure that avoids the need for experts
to choose weights on the components of the HDI or any alternative well-being index.

3.1.2. Quality of government

For measures of government quality, we use mainly data from the World Bank's Worldwide Governance Indicators (WGI)
Project.12 Following Bjørnskov et al. (2010), Helliwell and Huang (2008), and Ott (2010a,b, 2011), we define the quality of delivery
as the average of four WGI indicators: government effectiveness, regulatory quality, rule of law, and the control of corruption. The
quality of democracy as the average of two WGI indicators: voice and accountability, and political stability and absence of violence.
We also supplement the findings with governance indicators from other sources, such as Political Rights and Civil Liberties Indices
from the Freedom House, the Corruption Perception Index from Transparency International, the World Economic Freedom Index
from Wall Street Journal and Heritage Foundation, and Gallup World Poll's own measure of confidence in national governments. The
confidence measure is the percentage of respondents who answered “yes” to the question “Do you have confidence in each of the
following, or not? How about the national government?”

3.1.3. Covariates
In our regression analysis, we will first regress our measure of happiness on quality measures of governments alone to estimate the
total effects of government. We then regress happiness on an increasingly larger set of explanatory variables. The purposes are to
tease out the indirect effects as well as to address a potential concern that some unobserved perception changes may affect both the
measure of happiness and the measure of government quality.
The set of control variables, similar to those used in the World Happiness Report 2013, include log GDP per capita, healthy life
expectancy, social support, freedom to make life choices, and generosity.13 GDP per capita in purchasing power parity (PPP) adjusted
in constant 2005 dollars from the September 2013 release of the World Development Indicators (WDI). The GDP figures for Taiwan
and Zimbabwe are from the Penn World Table 7.1. The statistics of healthy life expectancy at birth are from the World Health
Organization (WHO), and are available for most countries for 2007. But the data are not available for other years, and are missing for
some countries. To impute the missing values, we obtain data for non-health adjusted life expectancy at birth from the WDI, which
provides good availability in terms of time and country coverage. We compute the ratio of healthy life expectancy to life expectancy
in 2007 for countries with both variables, and assign countries with missing data the world average of the ratio. We then use the ratio,
together with the non-health adjusted life expectancy, to generate the healthy life expectancy data. Social support (or having
someone to count on in times of trouble) from the Gallup World Poll, is the national average of the binary responses (either 0 or 1) to
the question “If you were in trouble, do you have relatives or friends you can count on to help you whenever you need them, or not?”
Freedom to make life choices, from the Gallup World Poll, is the national average of responses to the question “Are you satisfied or
dissatisfied with your freedom to choose what you do with your life?” Generosity, from the Gallup World Poll, is the residual of
regressing national average of the binary response to the question “Have you donated money to a charity in the past month?” on GDP
per capita.

3.1.4. Summary statistics

Tables 1 and 2 report the summary statistics of the variables that will be used in our analysis. Table 1 shows the average levels of
the variables over the pooled 2005–2012 sample. Table 2 shows the changes in the levels from the 2005–07 sub-period to the
2010–2012 sub-period. These data on changes will be used for our first difference analysis (see the details in the following section).
As shown in Tables 1 and 2, there are observations with missing values in some of the covariates. In our analysis, we follow the
principle that we never give up an observation unless forced to do so. If a country, in a particular year, does not have the observation
of the sense of freedom in the Gallup World Poll, we have no choice but to exclude it in regressions that use freedom as a covariate,
but this country-year would still be present in other regressions that do not use the information. As a result there are variations in
sample size and coverage from one regression to another. Our biggest samples include 157 countries in the period from 2005 to 2012.
Not all countries appear in all years; there are only 836 country-year observations. This biggest sample appears in regressions that
regress subjective well-being on WGI measures of government quality. As we include more variables to the right-hand side of the
regressions, the sample size becomes smaller. In Tables 3 and 5, which report our results based on the WGI governance indicator, our
smallest sample covers 154 countries, or 3 less than the maximum. The sample gets smaller in regressions that use alternative
governance measures (in Tables 4 and 6). The biggest drop occurs when Gallup World Poll's measure of confidence in national
government is used. There are only 147 countries with usable information on this measure. Yet a bigger drop happens in Table 7
when we regress changes in the subjective well-being from 2005–07 to 2010–12 to changes in governance quality over the same
period. The reason is that the Gallup World Poll in the earlier years covered fewer countries than it does in recent years; thus we do

11
For a summary of the evidence on each of these issues, see Helliwell and Wang (2012). See also Diener et al. (2009), especially Chapter 5.
12
The Worldwide Governance Indicators (WGI) are a research dataset summarizing views on the quality of governance provided by a large number of enterprise,
citizen and expert survey respondents in industrial and developing countries. These data are gathered from a number of survey institutes, think tanks, non-govern-
mental organizations, international organizations, and private sector firms. See Kraay et al. (1999) for a discussion of the related econometric issues, and
Thomas (2010) for a more general discussion of the WGI methodology. For the latest data, see http://info.worldbank.org/governance/wgi/index.aspx#home
13
We exclude the perceived corruption variable because control of corruption is already one of the key components of the delivery quality variable.

6
J.F. Helliwell et al. Journal of Comparative Economics xxx (xxxx) xxx–xxx

Table 1
Summary statistics: averages over the pooled 2005–2012 sample.

Variable Mean Std. Dev. Min. Max. N

Life Ladder 5.46 1.11 2.81 8.02 836

GDP per capita 12,978 13,526.97 275.84 72,650.87 836
Democratic qualityj −0.14 0.86 −2.24 1.53 836
Delivery quality 0.01 0.96 −1.69 2.20 836
Confidence in the national government 0.48 0.19 0.07 0.98 734
Healthy life expectancy 58.95 10.94 28.05 75.39 834
Freedom to make life choices 0.71 0.15 0.26 0.97 821
Generosity - adjusted for GDP 0 0.16 −0.33 0.54 792
Social support 0.82 0.12 0.29 0.98 829

Table 2
Summary statistics: changes from 2005–07 to 2010–2012.

Variable Mean Std. Dev. Min. Max. N

Δ Life Ladder 0.07 0.43 −1.07 1.27 130

Δ Logged GDP per capita 0.09 0.12 −0.53 0.41 126
Δ Democratic quality −0.01 0.21 −0.85 0.70 129
Δ Delivery quality 0.02 0.14 −0.43 0.66 130
Δ Confidence in the national government −0.01 0.13 −0.39 0.25 117
Δ Healthy life expectancy 1.13 0.67 −2.86 4.04 130
Δ Freedom to make life choices 0 0.08 −0.20 0.24 126
Δ Generosity - adjusted for GDP −0.02 0.09 −0.30 0.20 124
Δ Social support −0.01 0.06 −0.27 0.23 129

not have a base to calculate the change. As a result, the analysis covers only 122 countries.
Finally we note that, for the two measures of governance quality constructed by adding up the WGI indicators of governance, the
means are both close to zero and the standard deviations close to 1. The original WGI indicators are normalized to have mean 0 and 1
in the full WGI sample. Our sample is necessarily smaller than WGI's because of missing values. But our sample means and standard
deviations are close to the original sample, indicating that our sample is quite representative of the original one.

3.2. Methodology

We estimate two models to establish the key relationships between changes in governance and changes in well-being at national
level. The first model is to allow for two-way fixed effects using our full panel of countries and years
Yit = α + QoGit′ β + Xit′ Ψ + τi + θt + uit , (1)

where i indexes countries, t years. The variable Yit is the average happiness (Cantril ladder) of country i in year t. QoGit is a set of
measures of the quality of government. Xit is a set of covariates. τi denotes country fixed effects, which are included to capture
country-specific unobserved factors. Allowing for country fixed effects means that our estimates are based on the within-country
correlations between changes in governance and changes in life evaluations. θt denotes time fixed effects, and ɛit is the error term. Our
coefficient of interest is β, the estimated relationship between the quality of government and average happiness.
As discussed in the introduction, the evidence reported in the existing literature is entirely or in part based on international level
comparisons Bjørnskov et al., 2010; Helliwell, 2003; Helliwell and Huang, 2008, and Ott, 2010a,b, 2011). Model ((1), with the
presence of the country effects on the right-hand side, does not make such comparisons when estimating β, the relationship between
the quality of government and happiness. Instead, each country has its own average levels, and those level differences are not
utilized. The model instead relies on changes in governance quality and changes in national average happiness over time within the
same countries to estimate β. The existing evidence in the literature, on the other hand, comes from cross-sectional regressions or
pooled regressions, fully or partly relying on international level differences to estimate β.
To better connect our findings to those in the literature, we will also report estimates from alternative models that make use of
international level comparisons. Doing so serves two purposes. First, it tells us whether we can replicate the positive β reported in the
literature, using our new data but the same model. Second, it tells us how removing international level differences influences the
estimated effects of governance on happiness.
The first alternative model removes country fixed effects (τi) from the right-hand side. Such a model estimates the relationships
between government quality and happiness (β) by looking at international levels differences as well as changes within same coun-
tries. The second alternative replaces τi with regional fixed effects. Each of the regions, namely Western Europe, Central and Eastern
Europe, Commonwealth of Independent States, Southeast Asia, South Asia, East Asia, Latin America and Caribbean, North America
and ANZ, Middle East and North Africa, and Sub-Saharan Africa, is allowed to have its own average levels of happiness and gov-
ernance quality. Those inter-regional level differences are thus excluded from the estimation of β. This approach is an alternative to

7
J.F. Helliwell et al. Journal of Comparative Economics xxx (xxxx) xxx–xxx

Table 3
Subjective well-being and quality of government measured by WGI Indicators of Governance, sample period 2005–2012.

(1) (2) (3) (4) (5) (6) (7) (8) (9)

Pooled regressions without regional or Pooled regressions with regional or country Country fixed effects regressions
country fixed effects fixed effects

Democratic quality 0.007 0.01 -0.02 -0.02 -0.04 -0.09 0.13 0.10 0.06
(0.14) (0.12) (0.11) (0.13) (0.11) (0.10) (0.18) (0.18) (0.15)
Delivery quality 0.81*** 0.25* 0.06 0.62*** 0.31** 0.24** 0.84*** 0.74*** 0.57**
(0.13) (0.14) (0.12) (0.14) (0.14) (0.11) (0.23) (0.23) (0.23)
Logged GDP per capita 0.52*** 0.33*** 0.41*** 0.32*** 0.65** 0.84***
(0.06) (0.07) (0.08) (0.07) (0.27) (0.26)
Healthy life expectancy 0.02** 0.005 -0.06
(0.007) (0.01) (0.05)
Freedom to make life 1.16*** 0.66** 0.95***
choices
(0.34) (0.29) (0.23)
Generosity 0.96*** 0.55* 0.25
(0.26) (0.29) (0.19)
Social Support 2.12*** 1.97*** 1.46***
(0.44) (0.39) (0.33)
Regional fixed effects, with Western Europe as the omitted group
Central and Eastern -0.95*** -0.96*** -0.67***
Europe
(0.19) (0.18) (0.18)
Commonwealth of -0.50 -0.48* -0.33
Independent States
(0.33) (0.29) (0.24)
Southeast Asia -0.55** -0.35 -0.43***
(0.22) (0.22) (0.17)
South Asia -0.91*** -0.49 -0.24
(0.29) (0.31) (0.45)
East Asia -0.91*** -0.92*** -0.79***
(0.20) (0.19) (0.23)
Latin America and 0.23 0.25 0.24
Caribbean
(0.23) (0.22) (0.19)
North America and ANZ 0.29*** 0.34*** 0.20*
(0.10) (0.12) (0.11)
Middle East and North -0.41 -0.49** -0.30
Africa
(0.25) (0.23) (0.22)
Sub-Saharan Africa -1.27*** -0.70*** -0.48
(0.23) (0.25) (0.30)
Year fixed effects Yes Yes Yes Yes Yes Yes Yes Yes Yes
Country fixed effects Yes Yes Yes
Number of observations 836 836 776 836 836 776 836 836 776
Number of countries 157 157 154 157 157 154 157 157 154
R2 for pooled regressions 0.51 0.63 0.74 0.72 0.76 0.80
Within R2 for panel 0.10 0.11 0.20
regressions

Notes: 1) Standard errors in parentheses. *, **, and *** indicate statistical significance at 10%, 5% and 1% levels. All standard errors are cluster-adjusted at the country
level. 2) The r-squared from pooled regressions are not comparable to those from country fixed effects regressions. The calculation of the latter does not take into
account the variations across countries, and focuses solely on variations within a country over time. 3) All regressions include year fixed effects. In general, removing
the year fixed effects increases the estimated effect of delivery quality slightly.

deal with the possible objection that there are linguistic and other cultural differences at the continental level that might affect either
the reported happiness or the institutions in ways that are biasing the coefficients. If we do not have within-country and over-time
variations for a large enough number of countries, this approach would be our preferred way to deal with the concern about
linguistic/cultural influences. Here we report the estimates from the model as a robustness check as well as a stepping stone towards
our preferred model that uses within-country variations over time. These two sets of estimates are presented in Table 3, the table of
regression estimates, under the block headings “Pooled regressions without country or regional fixed effects”, and “pooled regressions
with regional fixed effects”, respectively. The estimates from model (1) are presented in the block “Country fixed effects regressions”.
There are pros and cons for each of the models. Pooled regressions without country effects use a fuller set of information. But part
of the international correlation between the levels of governance quality and happiness may reflect long-standing country-specific
factors that do not reflect the true effect of institutional quality on well-being. Adding country fixed effects reduces the risk, but at a
cost of giving up potentially valuable information. As we have argued in the introduction, we believe that using changes within same
countries eliminates potential contaminations from those country-specific factors. At the minimum, model (1) can test whether the
robustness of the existing evidence to the removal of international level variations in the empirical analysis. Furthermore, estimates

8
J.F. Helliwell et al. Journal of Comparative Economics xxx (xxxx) xxx–xxx

Table 4
Subjective well-being and quality of government measured by indicators of governance from alternative sources, country fixed effects regressions, sample period
2005–2012.

(1) (2) (3) (4)

Corruption Perception Index 0.14** 0.10* 0.10* 0.04

(0.06) (0.06) (0.06) (0.06)
Political Rights 0.03 0.02 0.02 0.05
(0.05) (0.05) (0.05) (0.04)
Civil Rights 0.06 0.05 0.05 0.09
(0.07) (0.06) (0.06) (0.06)
Economic Freedom Index -0.01 -0.01 -0.01 -0.02**
(0.01) (0.01) (0.01) (0.01)
Logged GDP per capita 0.80*** 0.79*** 1.00***
(0.30) (0.29) (0.28)
Healthy life expectancy -0.04 -0.10**
(0.04) (0.04)
Freedom to make life choices 1.07***
(0.25)
Generosity 0.32
(0.22)
Social support 1.55***
(0.35)
Year fixed effects Yes Yes Yes Yes
Country fixed effects Yes Yes Yes Yes
Number of observations 816 816 814 756
Number of countries 153 153 153 150
Within R2 0.07 0.09 0.09 0.20

Notes: Standard errors in parentheses. *, **, and *** indicate statistical significance at 10%, 5% and 1% levels. All standard errors are cluster-adjusted at the country
level.

based on changes over time can better inform us whether institutional quality can be changed in a relatively short time (less than 10
years in our data), and whether the change can have an impact on subjective well-being.
Throughout the three alternative models, we include year fixed effects θt in the regressions. Having year fixed effects alleviates
the concerns that prominent contemporaneous time trends, such as the recent worldwide financial crisis, or systemic changes in
survey designs and measurement of governance, if there are any, create spurious correlations in changes over time. But it turns out
that the inclusion or exclusion of year fixed effects has little impact on our estimates. When it does cause changes in the estimated β,
the changes are always small, and the inclusion of year effects lowers the estimates. Having the year effects thus helps our analysis
stay on the conservative side and not over-state the effects of governance on happiness.
As stated earlier, our analysis will proceed in a gradual manner with an increasingly bigger set of explanatory variables. There are
three sub-specifications within each of the three models. Each serves a purpose and we do not regard the last specification as our
preferred specification. The first specification regresses happiness on measures of governance quality alone to estimate the total effect
of well-being. The second adds GDP per person as a control variable to estimate government's well-being effects on top of those that
flow through the income channel. The third specification adds other explanatory variables used in the series of World Happiness
Reports, namely healthy life expectancy, perceptions of social support and freedom to make life choice, and a measure of generosity in
making donations. If it is true that good institutions can foster social trust and other well-being enhancing characteristics of the
society (see our earlier theoretical discussion), the inclusion of these new control variables will further lower the estimated β. We note
a caveat that some of the variables (sense of freedom, perceived social support) come from the same source as the happiness measure
(Gallup World Poll), and may contain common perception components as the dependent variables. This raises concerns of econo-
metric interpretation, which the World Happiness Report 2017 addresses using random selections of subsamples to make sure that the
subjective variables on the right-hand side are drawn from a different sample than the subjective happiness. The result in its statistical
appendix shows that the bias thereby eliminated is of the expected direction but small enough not to have material effect on the
results. But we caution nevertheless that in the last specification, the measures of government quality are likely placed in a dis-
advantaged position that raises the bar of finding statistical correlations between them and the measure of happiness.
The second model, similar to that adopted in World Happiness Report 2013, is to measure the difference in average Cantril ladder
in each country from before the global financial crisis (based on averages in the years 2005–07) until the most recent period (based on
averages from the years 2010–12), and then do a cross-country analysis to see to what extent improvements and losses in the quality
of governance were matched by changes in average life evaluations. In other words, it is a first difference model, wherein 2005–07
and 2010–12 are treated as the first and second period respectively. The model is given by

ΔYi = γ + ΔQoGi′ δ + ΔXi′ Φ + ɛi, (2)

where i indexes countries. The variable ΔYi is the change in average Cantril ladder from the period 2005–07 to 2010–12. ΔQoGit and
ΔXi denotes the corresponding changes in measures of the quality of government and the set of covariates respectively. ɛi is the error
term. Our coefficient of interest in this model is δ, the estimated relationship between the quality of government and the national-

9
J.F. Helliwell et al. Journal of Comparative Economics xxx (xxxx) xxx–xxx

level Cantril ladder. These estimates are reported in column (3) of Table 7. Before that column, however, we will also report cross-
sectional regressions of levels in columns (1) and (2) of the table, in order to see if we can replicate the positive relationship between
governance and happiness reported in the literature, and to see how moving to change-based regressions affect the estimates.

4. Empirical findings

Table 3 reports the estimates from model (1) and its alternatives. Columns 1–3 are from models without country or regional fixed
effects. Columns 4–6 are from those with regional fixed effects, while columns 7–9 are from country fixed-effect regressions. The first
two blocks (columns 1–6) in Table 3 are from pooled estimations that rely in part on international comparisons of level differences to
estimate the governance effects. All the six columns show that the coefficients of delivery quality are positive and statistically
significant (except in column 3), while the coefficients of democratic quality are small and not significant. The results are consistent
with earlier findings, e.g. in Helliwell and Huang (2008). These pooled OLS results are included here to set the stage for the third and
final block with two-way fixed effects: country fixed effects and year fixed effects. The country fixed effects are the essential part; year
fixed effects are secondary and can be removed with little consequence (to do so slightly strengthens the estimated effect of delivery
quality).
Comparing columns 1–6 to columns 7–9, it is clear that the estimated well-being effects of delivery quality are either similar or
higher in the last block than in earlier ones. For example, when measures of governance quality are the only control variables (namely
columns 1, 4 and 7). The estimated effect is 0.84 in column 7 (with country fixed effects), 0.81 in column 1 (without country or
regional effects) and 0.62 (with regional fixed effects). After controlling for income per person, the estimated effect of delivery quality
is 0.74 in column 8, 0.31 in column 5, and 0.25 in column 2. A similar pattern is observed when comparing columns 9, 6, and 3. This
suggests that our analysis, using within-country changes to estimate the well-being effect of governance, strengthens existing findings
that rely in part on international level comparisons. In our later discussion, we will concentrate on columns 7–9, those that estimate
the effects of governance using changes over time in governance and happiness within same countries.
The first equation in the third block (column 7 of Table 3) shows that changes in delivery quality, but not democratic quality, are
significantly correlated with changes in well-being for the Gallup World Poll sample of 157 countries over the 2005 to 2012 period.
This confirms, in a larger and more current sample, and with full allowance for country fixed effects, the same pattern of results found
using smaller samples of earlier data from the World Values Survey. The estimated effects of delivery quality are very substantial. The
governance variables are in standardized form, with a mean of zero and a standard deviation of 1.0. So the coefficient on delivery
quality of 0.84 in column 7 of Table 3 suggests that an improvement in delivery quality equal to one standard deviation would lead to
an increase in average life evaluations of 0.84, which corresponds to a change of about three-quarters of a standard deviation of the
distribution of national averages of life evaluations (Table 1 shows the summary statistics including the standard deviations). This
estimate is from a regression that includes country-fixed effects, so the sources of variation behind the estimate are within-country
changes in governance and life evaluations relative to changes in other countries. As shown in Table 14 of Helliwell et al. (2014a), the
ten countries with the largest increases in delivery quality are Peru, Brazil, Macedonia, Taiwan, Paraguay, Poland, Myanmar,
Georgia, Rwanda and Palestinian Territories; the average increase is 0.3. Life evaluations rose in eight of the ten countries (except
Myanmar and Rwanda), the average increase for the entire group of ten is 0.15. On the opposite end, the countries with the largest
declines in delivery quality are Madagascar, Greece, Venezuela, Yemen, Austria, Kuwait, Tanzania, Nepal, South Africa and Hungary.
All except for Venezuela, Austria and Kuwait saw declines in life evaluations. The average changes are −0.26 for delivery and −0.18
for life evaluation. When we compare the ten-most-improved countries with the ten most worsened, in terms of delivery quality, the
associated well-being difference is 0.4 points on the 0 to 10 scale – about one tenth of the 4-point gap separating the happiest and
unhappiest 10 countries among more than 150 countries surveyed by the Gallup World Poll.
Column 8 of Table 3 adds GDP per capita to the equation, permitting us to estimate the extent to which the quality of government
is influencing happiness by improving the economic productivity of the economy. The coefficient on delivery quality drops from 0.84
to 0.74, suggesting that some, but only a fraction, of the life-improving effects of better governance are flowing through its effects on
GDP per capita. Of course, better governance will affect the structure as well as the size of the economy, and thus enable increases in
the amount of well-being provided by a given level of GDP. For example, if levels of corruption are lower, streets and communities
safer, and trust levels higher, people are more likely to get the goods and services they want, and the required levels of defensive
expenditures will be lower.14 The equation shows changes in GDP per capita and governance to have contributed significantly to
changes in national-average life evaluations over the 2005–2012 period, with delivery quality being slightly the more important
contributor, since changes in delivery quality have slightly greater variance than changes in GDP (0.14 to 0.12 in terms of standard
deviations of changes from 2005–07 to 2010–12, as shown in Table 2) and the former have a larger estimated coefficient (0.74 to
0.65). Furthermore, the delivery effect is, in principle, net of the effects flowing to well-being indirectly through changes in GDP per
capita.
The last column of Table 3 extends the list of explanatory variables, and therefore adds to the number of channels through which
good governance might be influencing life evaluations. The variables added are those that were used to explain a similar set of Gallup

14
The utility-wasting effects of defensive expenditures have been central to many influential complaints about using GDP to represent levels of welfare. See
especially Stiglitz et al. (2009). If the quality of government has a large and systematic effect on the amount of life satisfaction provided by any given level of GDP per
capita, then we might expect to find an interaction term between GDP per capita and delivery quality. To test for this, we added a second GDP variable for observations
where delivery quality was above average. The estimated coefficient was small and negative, suggesting no significant interaction effect between the quality of
delivery and the life satisfaction benefits of measured GDP.

10
J.F. Helliwell et al. Journal of Comparative Economics xxx (xxxx) xxx–xxx

World Poll life evaluations in World Happiness Report 2013. The most significant changes appear to have been in perceived freedom to
make life choices, and in the proportion of respondents who report having someone they can count on in times of trouble. When these
variables are added, there is a decrease in the direct governance effect, and an increase in that for GDP, suggesting that changes in
perceived freedom and social support were more likely to be positively correlated with changes in delivery quality than with changes
in GDP per capita.15 We note that some of the newly added variables, such as the sense of freedom and social support and the self-
reported generosity come from the same Gallup surveys as do subjective well-being. Their inclusion thus places the external gov-
ernance indicators at a possible empirical disadvantage. Yet delivery quality still largely retains its predictive power.
We next consider alternative indicators of governance quality. There are intrinsic complexities in the measurement of the quality
of governance. The WGI indicators are summaries of views on the quality of governance provided by enterprise, citizen and expert
survey respondents. Even though the subjective well-being and the WGI indicators are from different sources, one cannot preclude the
possibility that there may be some common driving forces behind the life evaluations and governance perceptions. In addition, the
WGI indicators are constructed to have a constant world average at zero in each year; they thus measure relative standing instead of
absolute levels. This does not make any difference to our key results, because our preferred specification, columns 7–9 of Table 3, is a
two-way fixed effects model that has both country and year fixed effects. Year fixed effects are used to prevent potential bias from
contemporaneous time trends impacting a large number of countries, such as the worldwide financial crisis in 2008 and 2009, or
changes in survey questionnaire framing. In our model the year fixed effects help us to err on the conservative side, because removing
the year fixed effects slightly strengthens the estimated effect of delivery. They also remove sample means, for each year, from the
dependent and independent variables; in other words, the use of yearly fixed effects would have removed the yearly mean of WGI
indicators even if they were not set at zero already. Two-way fixed effects estimate the governance effect in a procedure akin to
making difference-in-differences comparisons, asking whether a country that experiences a greater-than-average increase in gov-
ernance quality also tends to experience a greater-than-average increase in SWB over the same time. Finally, one cannot ignore the
considerable margin of errors in the WGI indicators. All these concerns, together with the well-established availability of alternative
governance indicators, suggest the need for robustness tests.
Table 4 uses alternative indicators of governance quality in the two-way fixed effects model as shown in model (1). These include
the Political Rights and Civil Liberties Indices from the Freedom House, the corruption perception index from Transparency Inter-
national, and the World Economic Freedom Index from Wall Street Journal and Heritage Foundation. There are four key findings.
First the Indices of Political Rights and Civil Liberties have little effect on happiness with or without the inclusion of GDP and other
controls. Second, the World Economic Freedom Index has a negative coefficient across all specifications.16 Third, the Corruption
Perception Index, highly correlated with our measure of delivery quality, has a positive effect on well-being with borderline sig-
nificance even after GDP per capita is controlled for. The estimate becomes smaller and insignificant only after the sense of freedom,
social support and generosity from the Gallup World Poll are included on the right-hand side. Overall, the results from these al-
ternative measures are consistent with our finding that delivery quality is the best measure of good governance in the worldwide
sample.
In Table 5, we divide the sample into two groups of countries divided by their average quality of service delivery. Within each
subsample, we repeat Table 3’s panel regressions with country and year fixed effects. The first two columns show the results based
only on the two quality of government measures, and show that the effects of delivery quality are highly significant, and to an equal
extent, for both groups of countries, while democratic quality has a positive impact only for those countries that have achieved
sufficiently high delivery quality. These results remain when GDP per capita is added, in columns 3 and 4. Table 5’s last two columns
report the split-sample estimates for the most fully specified model. A stark contrast emerges. For countries that have already
achieved better-than-average delivery quality, it is democratic quality that has a positive and significant effect on life evaluations; the
effect of delivery quality is essentially zero and statistically insignificant. The exact opposite is true for the countries with low delivery
quality: the effect of delivery is positive and significant, while the effect of democratic quality is zero. We make similar observations
when splitting the sample by GDP per capita instead of by government delivery quality (see Table 11 in Helliwell et al., 2014a). These
findings are not entirely surprising. They are consistent with earlier findings reported in Helliwell and Huang (2008), but we now
have many more countries, use a different survey source, test alternative ways of splitting the sample, and cover a more recent time
period.17 This difference in source, scope and years covered makes the new results strongly confirming of the pattern that had been
appearing in earlier studies based only on the World Values Survey.
In Table 6 we extend the analysis by adding a measure of confidence in government, derived from survey responses in the Gallup
World Poll, to see to what extent it supplements or substitutes for the independent estimates of the actual quality of delivery. Once

15
It is also worth noting that the variables measuring freedom and having someone to count on are drawn from the same survey respondents as the life evaluations,
while GDP per capita and delivery quality are drawn from different sources. Thus any changes from year to year in the representativeness of the population samples
may lead to an upward bias in the size of the effects estimated for freedom and social support, although not for GDP per capita and delivery quality.
16
Ott (2016) shows that the internal and predictive validity of the measurement of economic freedom by the Heritage Foundation and the Fraser Institute is limited,
because general levels of government activities, as expressed in taxation, transfers and subsidies, and government expenditures and consumption, are used as negative
indicators for economic freedom. The correlation between economic freedom and happiness in countries is at least neutral or positive if such levels are left out as
negative indicators, or if they are used as positive indicators. He also shows that the relation between such levels and happiness depends heavily on the quality of
governments.
17
In Helliwell and Huang (2008), we split the sample by GDP per capita, with the split between richer and poorer countries was set at a value of real GDP per capita
equal to half of that in the United States. In the parallel work reported in Helliwell et al. (2014a, Table 11), our usual split is at a GDP per capita equal to one-quarter of
that in the United States. In both cases the splits were done so as to give reasonably equal sample sizes to the two groups of countries. In each case the results are close
to what we report here using delivery quality as our preferred splitting variable.

11
J.F. Helliwell et al. Journal of Comparative Economics xxx (xxxx) xxx–xxx

Table 5
Subgroup analysis comparing countries with high delivery quality and countries with low delivery quality, country fixed effects regressions, sample period 2005–2012.

(1) (2) (3) (4) (5) (6)

High delivery Low delivery High delivery Low delivery High delivery Low delivery

Democratic quality 0.52* 0.05 0.52* 0.02 0.68** -0.06

(0.28) (0.20) (0.27) (0.22) (0.28) (0.17)
Delivery quality 0.84** 0.81** 0.60* 0.79** 0.33 0.61**
(0.34) (0.32) (0.34) (0.32) (0.37) (0.30)
Logged GDP per capita 0.94* 0.31 0.89* 0.88**
(0.53) (0.39) (0.48) (0.36)
Healthy life expectancy -0.09* -0.03
(0.05) (0.09)
Freedom to make life choices 0.79** 0.96***
(0.40) (0.30)
Generosity 0.74*** 0.04
(0.27) (0.25)
Social Support 1.52*** 1.45***
(0.53) (0.38)
Year fixed effects Yes Yes Yes Yes Yes Yes
Country fixed effects Yes Yes Yes Yes Yes Yes
Number of observations 336 500 336 500 303 473
Number of countries 62 95 62 95 61 93
Within R2 0.15 0.10 0.17 0.11 0.23 0.21

Notes: 1) The sample is split by whether the measure of delivery quality for a country, averaged across all years in the full sample, is greater or lower than zero. 2)
Standard errors in parentheses. *, **, and *** indicate statistical significance at 10%, 5% and 1% levels. All standard errors are cluster-adjusted at the country level.

Table 6
Comparing Gallup World Poll's measure of confidence in national government to quality measure of government based on WGI Indicators of Governance, country fixed
effects regressions, sample period 2005–2012.

(1) (2) (3) (4) (5) (6)

Confidence in the national government 0.74*** 0.68*** 0.58*** 0.45*** 0.44** 0.51***
(0.16) (0.15) (0.15) (0.17) (0.18) (0.17)
Logged GDP per capita 1.46*** 1.36*** 1.25*** 1.17*** 1.31***
(0.29) (0.27) (0.24) (0.25) (0.27)
Democratic quality -0.32*** -0.22**
(0.10) (0.10)
Delivery quality 0.78*** 0.59** 0.50**
(0.22) (0.25) (0.23)
Healthy life expectancy -0.07 -0.07 -0.08
(0.05) (0.05) (0.05)
Freedom to make life choices 0.52** 0.56** 0.54**
(0.26) (0.25) (0.26)
Generosity 0.22 0.23 0.26
(0.20) (0.21) (0.22)
Social Support 1.67*** 1.69*** 1.73***
(0.35) (0.36) (0.37)
Year fixed effects Yes Yes Yes Yes Yes Yes
Country fixed effects Yes Yes Yes Yes Yes Yes
Number of observations 734 734 734 696 696 696
Number of countries 147 147 147 146 146 146
Within R2 0.11 0.16 0.19 0.24 0.24 0.23

Notes: Standard errors in parentheses. *, **, and *** indicate statistical significance at 10%, 5% and 1% levels. All standard errors are cluster-adjusted at the country
level.

again, our analysis uses the specification in model (1), including both country and year fixed effects, so that we are estimating the
extent to which the variables are influencing the within-country changes. Although the samples are slightly different, the two effects
contribute independently to life evaluations.18 For both groups, confidence in government has a positive effect on life evaluations and
does so with comparable effect sizes.19 Presumably the implicit coverage of the confidence measure embraces some elements of both

18
In principle we would expect that the two measures would be correlated, and hence that the coefficient on the delivery variable would fall when the confidence
variable was added. However, over the 886 cases where both variables are measured, the correlation is essentially zero (-0.02).
19
For example, in Eq. (3), the most fully specified model in Table 4, the coefficients are 0.45 on confidence in government and 0.59 on delivery quality. As shown in
Table 2, the standard deviations of the changes in the two variables are also very similar, 0.13 for confidence in government and 0.14 for delivery quality.

12
J.F. Helliwell et al. Journal of Comparative Economics xxx (xxxx) xxx–xxx

Table 7
Alternative specifications: cross-sectional regressions of levels in period 2005–2012 (Columns 1 and 2) and cross-sectional regression of changes from 2005–07
to 2010–12 (Column 3).

(1) (2) (3)

Democratic quality -0.02 -0.11 -0.03

(0.10) (0.11) (0.18)
Delivery quality 0.06 0.26** 0.79**
(0.12) (0.12) (0.32)
Logged GDP per capita 0.29*** 0.29*** 0.25
(0.08) (0.07) (0.30)
Healthy life expectancy 0.02*** 0.003 -0.04
(0.007) (0.009) (0.05)
Freedom to make life choices 1.15*** 0.60 1.46***
(0.40) (0.39) (0.46)
Generosity 1.13*** 0.67** 0.51
(0.33) (0.33) (0.38)
Social Support 2.49*** 2.25*** 2.07***
(0.51) (0.46) (0.65)
Regional fixed effects, with Western Europe as the omitted group
Central and Eastern Europe -0.69***
(0.19)
Commonwealth of Independent States -0.36
(0.23)
Southeast Asia -0.55**
(0.23)
South Asia -0.31
(0.28)
East Asia -0.77***
(0.21)
Latin America and Caribbean 0.20
(0.18)
North America and ANZ 0.18
(0.24)
Middle East and North Africa -0.36*
(0.20)
Sub-Saharan Africa -0.59**
(0.26)
Number of countries 154 154 122
R2 0.80 0.86 0.27

Notes: 1) Columns (1) and (2) show estimates from cross-sectional regressions that use observations that are the 2005–2012 averages at the country level.
Column (3), on the other hand, is cross-sectional regression of changes, specifically changes from the 2005–07 period to the 2010–12 period. 2) Standard
errors in parentheses. *, **, and *** indicate statistical significance at 10%, 5% and 1% levels.

delivery and democratic quality, perhaps with weights that shift as one goal is achieved and attentions turns to another. The con-
fidence in government measure must cover more than what is in the delivery and democracy variables, as it has an importance above
and beyond their effects.20
Table 7 presents findings from two additional statistical specifications, and demonstrates that the positive influence of good
governance on life evaluations exists both across countries and over time, thus greatly increasing the likelihood that changes in
governance quality are leading to changes in well-being. The table's first two columns are from “pure” cross-sectional regressions, in
which each country appears once; the observations are average levels by country covering all surveys during the 2005–2012 period.
From this it is seen that delivery quality has a significant effect even after allowing for regional fixed effects. Column 3 is from a
regression based on changes from 2005–07 to 2010–12 as specified in model (2). Again a country appears only once; but the
observations are changes instead of levels. So the regression utilizes an entirely different source of variation compared to column 2.
But again, delivery quality is found to have a positive and significant effect on life evaluations, with an estimated magnitude that is
greater than the one found in the level regression. It thus appears that, while both levels and changes contribute to the estimated
effect, the contribution from longitudinal variation is greater.

5. Conclusions

This paper first makes an evidence-based case that people's own evaluations of the quality of their lives provide reliable and
inexpensive new ways to show how, and how much, good governance matters. It then brings together a large set of national-level
data to assess the extent to which governance quality contributes to explaining the levels and changes in life evaluations in 157
countries over the years 2005–2012. The results confirm earlier findings that the delivery quality of government services dominates

20
See columns 3 and 4 of Table 5, showing the impact of confidence in government to be strong even when the other two variables are included.

13
J.F. Helliwell et al. Journal of Comparative Economics xxx (xxxx) xxx–xxx

democratic quality in supporting better lives. The situation changes as development proceeds, with democratic quality having a
positive influence among countries that have already achieved a reasonable level of delivery quality.
The new results reported in this paper show not just that people are more satisfied with their lives in countries with better
governance quality, but also that actual changes in governance quality since 2005 have led to large changes in the quality of life. For
example, the ten most-improved countries, in terms of delivery quality changes between 2005 and 2012, when compared to the ten
countries with most worsened delivery quality, are estimated to have higher average life evaluations by one-tenth of the overall life
evaluation gap between the world's most and least happy ten countries.
This provides much stronger evidence that governance quality has changed, and that these changes have much larger effects than
those flowing simply through a more productive economy. When we explain changes in average life evaluations over the 2005 to
2012 period, just as much was explained by changes in governance quality as by changes in GDP, even though some of the well-being
benefits of better governance are delivered through increases in economic efficiency and hence GDP per capita. Our new results thus
confirm that quality of governance affects lives via many channels beyond those captured by GDP per capita, and also that important
improvements can be achieved within policy-relevant time horizons.
Our new results, based primarily on the analysis of changes in governance quality and making use of large global samples, strengthen the
welfare case for improving the quality of governance. But even these results are limited by their inability to be sure about causal direction,
and offer no specific advice about which aspects of governance are most important for well-being, and how they might be changed. There are
no quick fixes. Take corruption control as an example, Mungui-Pippidi (2016) finds that despite the ubiquity of anticorruption agencies
around the world, there is no statistical evidence linking their presence to improvements in the control of corruption (page 106). Fukuyama
(2014) describes the challenges as the problem of “getting to Denmark”, something that the international community lacks a good under-
standing of. Perhaps it is best to hope that our results are strong enough to motivate further experimentation and applied research designed to
deal with both of these limitations. Both natural and deliberate experiments can help to identify the causal directions more clearly, and
typically involve changes to specific elements of governance. Beyond this natural concern with what governments do, a much longer earlier
version of this paper (Helliwell et al., 2014a) focused on the implications that well-being research has for the ‘how’ as well as the ‘what’ of
government service delivery. Once the social context of the design and delivery of government services is taken more seriously, then it
becomes clear that giving citizens more opportunities to contribute to the design and delivery services offers substantial improvements in the
quality of governance and levels of national well-being. That research lies beyond the empirical analysis in this paper; our hope is that the
empirical evidence is strong enough to encourage those essential next steps to be taken.

Appendix - Data sources

Subjective well-being (SWB): A country-year panel of average survey measures of SWB derived from the October-2013 release
of the Gallup World Poll (GWP) for the survey years from 2005 to 2012. The SWB measure, unless stated otherwise, is the national
average response to the question of life evaluations. The English wording of the question is “Please imagine a ladder, with steps
numbered from 0 at the bottom to 10 at the top. The top of the ladder represents the best possible life for you and the bottom of the
ladder represents the worst possible life for you. On which step of the ladder would you say you personally feel you stand at this
time?” This measure is referred to as Cantril life ladder, or just life ladder in our analysis. The October-2013 GWP release also has
observations from the on-going 2013 survey. But those observations from the incomplete surveys are not used in the analysis.
Quality of governance - Source 1: A country-year panel of governance indicators from the Worldwide Governance Indicators
(WGI) project (Kaufmann et al., 2010). According to the source, the WGI “are a research dataset summarizing the views on the quality
of governance provided by a large number of enterprise, citizen and expert survey respondents in industrial and developing countries.
These data are gathered from a number of survey institutes, think tanks, non-governmental organizations, international organiza-
tions, and private sector firms.” The WGI project provides data for 215 economies over the period 1996 - 2012. For our analysis we
use observations that overlap with our panel of subjective well-being. There are six dimensions of governance in the WGI: Voice and
Accountability, Political Stability and Absence of Violence, Government Effectiveness, Regulatory Quality, Rule of Law, Control of
Corruption. The indicators are on a scale roughly with mean zero and a standard deviation of 1. In places where we need to further
reduce the number of dimensions, we use the simple average of the first two measures as an indicator of democratic quality, and the
simple average of the other four measures as an indicator of delivery quality, following Helliwell and Huang (2008).
Quality of governance - Source 2: Gallup's National institutions Index and its components from the 2005–2012 Gallup World
Poll (October 2013 release). We use primarily WP139, confidence in the national government. According to Gallup's Worldwide
Research Methodology and Codebook (June 2013), “[t]he national institutions index reflects citizens’ confidence in key institutions
prominent in a country's leadership: the military, the judicial system, the national government, and the honesty of elections.” The
index questions are “Do you have confidence in each of the following, or not? How about the military? (WP137)”, “Do you have
confidence in each of the following, or not? How about the judicial system and courts? (WP138)”, “Do you have confidence in each of
the following, or not? How about the national government? (WP139)” and “Do you have confidence in each of the following, or not?
How about honesty of elections? (WP144)”. Our analysis uses primarily WP139, confidence in the national government.
GDP per capita in purchasing power parity (PPP) and constant 2005 dollars from the September 2013 release of the World
Development Indicators (WDI). The GDP figures for Taiwan and Zimbabwe are from the Penn World Table 7.1. A small number of
observations are missing in the September-2013 WDI release but were present in the April-2013 release. In such cases, we use the
numbers from the earlier release.
Healthy Life Expectancy: We took the data used in the World Happiness Report (WHR) 2013. The statistics of healthy life
expectancy at birth are from the World Health Organization (WHO), and are available for most countries in the year of 2007. But the

14
J.F. Helliwell et al. Journal of Comparative Economics xxx (xxxx) xxx–xxx

data are not available for other years, and are missing for some countries. To impute the missing values, we obtain the data of non-
health adjusted life expectancy at birth from the WDI that has very good availability in terms of time and country coverage. We
compute the ratio of healthy life expectancy to life expectancy in 2007 for countries with both data, and assign countries with missing
data the world average of the ratio. We then use the ratio, together with the non-health adjusted life expectancy, to generate the
healthy life expectancy data.
Social support (or having someone to count on in times of trouble), from the Gallup World Poll, is the national average of the
binary responses (either 0 or 1) to the question “If you were in trouble, do you have relatives or friends you can count on to help you
whenever you need them, or not?”
Freedom to make life choices, from the Gallup World Poll, is the national average of responses to the question “Are you satisfied
or dissatisfied with your freedom to choose what you do with your life?”
Generosity, from the Gallup World Poll, is the residual of regressing national average of response to the question “Have you
donated money to a charity in the past month?” on GDP per capita.

References

Acemoglu, D., Robinson, J., 2012. Why Nations Fail: The Origins of Power, Prosperity, and Poverty. Crown Publishers, New York.
Bartolini, S., Bilancini, E., Pugno, M., 2013. Did the decline in social connections depress Americans’ happiness? Soc. Indic. Res. 110 (3), 1033–1059.
Bjørnskov, C., Dreher, A., Fischer, J., 2010. Formal institutions and subjective well-being: revisiting the cross-country evidence. Eur. J. Polit. Econ. 26 (4), 418–430.
Bonikowska, A., Helliwell, J.F., Hou, F., Schellenberg, G., 2013. An assessment of life satisfaction responses on recent Statistics Canada surveys. Statistics Canada,
Analytical Studies Branch Research Paper Series. http://www.statcan.gc.ca/pub/11f0019m/11f0019m2013351-eng.pdf.
Deaton, A., 2010. Income, aging, health and well-being around the world: evidence from the Gallup World Poll. In: Wise, D.A. (Ed.), Research Findings in the
Economics of Aging. The University of Chicago Press, Chicago, pp. 235–263.
Deaton, A., 2012. The financial crisis and the well-being of Americans. Oxf. Econ. Pap. 64 (1), 1–26.
De Neve, J., Diener, E., Tay, L., Xuereb, C., 2013. The Objective Benefits of Subjective Well-Being. In: Helliwell, J., Layard, R., Sachs, J. (Eds.), UN Sustainable
Development Solution Network, New York, pp. 54–79 World Happiness Report 2013.
Diener, E., Lucas, R., Schimmack, U., Helliwell, J.F., 2009. Well-Being for Public Policy. Oxford University Press, New York.
Diener, E., Kahneman, D., Tov, W., Arora, R., 2010. Income's association with judgments of life versus feelings. In: Diener, E., Helliwell, J.F., Kahneman, D. (Eds.),
International Differences in Well-Being. Oxford University Press, Oxford, pp. 3–15.
Di Tella, R., MacCullough, R., 2005. Partisan social happiness. Rev. Econ. Stud. 72 (2), 367–393.
Dinesen, P.T., 2013. Where you come from or where you live? Examining the cultural and institutional explanation of generalized trust using migration as a natural
experiment. Eur. Sociol. Rev. 29 (1), 114–128.
Durand, M., Smith, C., 2013. The OECD Approach to Measuring Subjective Well-Being. In: Helliwell, J.F., Layard, R., Sachs, J. (Eds.), UN Sustainable Development
Solution Network, New York, pp. 112–137 World Happiness Report 2013.
Easterlin, R., 1974. Does economic growth improve the human lot? Some empirical evidence. In: David, P.A., Reder, M.W. (Eds.), Nations and Households in Economic
Growth: Essays in Honor of Moses Abramovitz. Academic Press, New York, pp. 89–125.
Easterlin, R., Sawangfa, O., 2010. Happiness and economic growth: does the cross-section predict time trends? Evidence from developing countries. In: Diener, E.,
Helliwell, J.F., Kahneman, D. (Eds.), International Differences in Well-Being. Oxford University Press, Oxford, pp. 166–216.
Frey, B.S., Stutzer, A., 2005. Beyond outcomes: measuring procedural utility. Oxf. Econ. Pap. 57 (1), 90–111.
Guven, C., 2011. Are happier citizens better citizens? Kyklos 64 (2), 178–192.
Hall, J., 2013. From Capabilities to Contentment: Testing the Links between Human Development and Life Satisfaction. In: Helliwell, J.F., Layard, R., Sachs, J. (Eds.),
UN Sustainable Development Solution Network, New York, pp. 138–153 World Happiness Report 2013.
Helliwell, J.F., 2003. How's life? Combining individual and national variables to explain subjective well-being. Econ. Model. 20 (2), 331–360.
Helliwell, J.F., Barrington-Leigh, C.P., Harris, A., Huang, H., 2010. International evidence on the social context of well-being. In: Diener, E., Helliwell, J.F., Kahneman,
D. (Eds.), International Differences in Well-being. Oxford University Press, New York, pp. 291–350.
Helliwell, J.F., Huang, H., 2008. How's your government? International evidence linking good government and well-being. Brit. J. Polit. Sci. 38 (4), 595–619.
Helliwell, J.F., Wang, S., 2012. In: Helliwell, J.F., Layard, R., Sachs, J. (Eds.), The State of World Happiness. The Earth Institute, Columbia University, New York, pp.
10–57 World Happiness Report.
Helliwell, J.F., Huang, H., Grover, S., Wang, S., 2014a. Good governance and national well-being: what are the linkages? In: OECD Working Papers on Public
Governance. OECD Publishing No 25.
Helliwell, J.F., Huang, H., Wang, S., 2014b. Social capital and well-being in times of crisis. J. Happiness Stud. 15 (1), 145–162.
Helliwell, J.F., Huang, H., Wang, S., 2015. The Geography of World Happiness. In: Helliwell, J.F., Layard, R., Sachs, J. (Eds.), UN Sustainable Development Solution
Network, New York, pp. 8–37 World Happiness Report 2013.
Helliwell, J.F., Putnam, R.D., 2004. The Social context of well-being. Philos. Trans. R. Soc. B 359 (1449), 1435–1446 Reprinted in: F. A., Huppert, B., Keverne, & N.,
Baylis (Eds.), 2005. The Science of Well-Being., London: Oxford University Press, pp. 435–459 (2004).
Helliwell, J.F., Wang, S., 2011. Trust and wellbeing. Int. J. Wellbeing 1 (1), 42–78.
Helliwell, J.F., Wang, S., 2013. World Happiness: Trends, Explanations and Distribution. In: Helliwell, J.F., Layard, R., Sachs, J. (Eds.), UN Sustainable Development
Solution Network, New York, pp. 8–37 World Happiness Report 2013.
Kaufmann, D., Kraay, A., Mastruzzi, M., 2010. The Worldwide Governance Indicators: Methodology and Analytical Issues. World Bank, Washington, D.C Policy
Research Working Paper No. 5430.
Kenny, C., 2005. Does development make you happy? Subjective wellbeing and economic growth in developing countries. Soc. Indic. Res. 73 (2), 199–219.
Kraay, A., Zoido-Lobaton, P., Kaufmann, D., 1999. Governance Matters. World Bank, Washington, D.C Policy Research Working Paper No. 2196.
Layard, R., 2005. Happiness: Lessons from a New Science. Penguin Press.
Lane, R.E., 2001. The Loss of Happiness in Market Democracies. Yale University Press, New Haven.
Nannestad, P., Svendsen, G.T., Dinesen, P.T., Sonderskov, K.M., 2014. Do institutions or culture determine the level of social trust? The natural experiment of
migration from non-western to western Countries. J. Ethn. Migr. Stud. 40 (4), 544–565.
OECD, 2013. OECD Guidelines on Measuring Subjective Well-being. OECD Publishing.
Ott, J.C., 2010a. Good governance and happiness in nations: technical quality precedes democracy and qualitybeats size. J. Happiness Stud. 11 (3), 353–368.
Ott, J.C., 2010b. Greater happiness for a greater number: Some non-controversial options for governments. J. Happiness Stud. 11 (5), 631–647.
Ott, J.C., 2011. Government and happiness in 130 nations: good governance fosters higher level and more equality of happiness. Soc. Indic. Res. 102 (1), 3–22.
Ott, J.C., 2016. Measuring economic freedom: better without size of government. Soc. Indic. Res. 1–20.
Rothstein, B., Daniel, E., 2009. Political corruption and social trust - an experimental approach. Ration. Soc. 21 (1), 81–112.
Sacks, D.W., Stevenson, B., Wolfers, J., 2012. In: The New Stylized Facts about Income and Subjective Well- Being. Forschungsinstitut zur Zukunft der Arbeit
Discussion Paper Series No. 710. http://www.econstor.eu/bitstream/10419/69372/1/733999808.pdf.
Stiglitz, J., Sen, A., Fitoussi, J.P., 2009. Report of the Commission on the Measurement of Economic Performance and Social Progress. Commission on the Measurement
of Economic Performance and Social Progress. http://www.stiglitz-sen-fitoussi.fr/en/index.htm.
Thomas, M.A., 2010. What do the worldwide governance indicators measure? Eur. J. Dev. Res. 22, 31–54.
Veenhoven, R., 2008. World Database of Happiness. Erasmus University Rotterdam, The Netherlands. http://worlddatabaseofhappiness.eur.nl.

15