Happiness, Freedom and Control

Abstract

How do people value freedom of choice? Drawing on economics and psychology the paper provides an hypothesis and empirical evidence on how individuals may value freedom of choice and derive utility from it. It is argued that the degree of perceived control that individuals have over choice – a construct known as the locus of control in psychology – regulates how we value freedom of choice. People who believe that the outcome of their actions depends on internal factors such as effort and skills (the ‘internals’) have a greater appreciation of freedom of choice than people who believe that the outcome of their actions depends on external factors such as fate or destiny (the ‘externals’). We find some evidence in support of this hypothesis using a combination of all rounds of the World and European Values Surveys. A variable that measures freedom of choice and the locus of control is found to predict life satisfaction better than any other known factor such as health, employment, income, marriage or religion, across countries and within countries. We show that this variable is not a proxy of happiness and measures well both freedom of choice and the locus of control. ‘Internals’ are found to appreciate freedom of choice more than ‘externals’ and to be happier. These findings have important implications for individual utility, social welfare and public policies.

Full text

This article appeared in a journal published by Elsevier. The attached
copy is furnished to the author for internal non-commercial research
and education use, including for instruction at the authors institution
and sharing with colleagues.
Other uses, including reproduction and distribution, or selling or
licensing copies, or posting to personal, institutional or third party
websites are prohibited.
In most cases authors are permitted to post their version of the
article (e.g. in Word or Tex form) to their personal website or
institutional repository. Authors requiring further information
regarding Elsevier’s archiving and manuscript policies are
encouraged to visit:
http://www.elsevier.com/copyright

Author's personal copy
Journal of Economic Behavior & Organization 71 (2009) 146–161
Contents lists available at ScienceDirect
Journal of Economic Behavior & Organization
journal homepage: www.elsevier.com/locate/jebo
Happiness, freedom and control
Paolo Vermea,b,∗
aDepartment of Economics “S. Cognetti de Martiis”, University of Torino, Via Po 54, 10124 Torino, Italy
bSchool of Management SDA-Bocconi, Via Bocconi 8, 20136 Milano, Italy
a r t i c l e i n f o
Article history:
Received 13 January 2008
Received in revised form 17 April 2009
Accepted 17 April 2009
Available online 3 May 2009
JEL classification:
D1
D6
D7
D9
H4
O1
Keywords:
Happiness
Utility
Freedom of choice
Locus of control
abstract
How do people value freedom of choice? Drawing on economics and psychology the paper
provides an hypothesis and empirical evidence on how individuals may value freedom
of choice and derive utility from it. It is argued that the degree of perceived control that
individuals have over choice – a construct known as the locus of control in psychology –
regulates how we value freedom of choice. People who believe that the outcome of their
actions depends on internal factors such as effort and skills (the ‘internals’) have a greater
appreciation of freedom of choice than people who believe that the outcome of their actions
depends on external factors such as fate or destiny (the ‘externals’). We find some evidence
in support of this hypothesis using a combination of all rounds of the World and European
Values Surveys. A variable that measures freedom of choice and the locus of control is found
to predict life satisfaction better than any other known factor such as health, employment,
income, marriage or religion, across countries and within countries. We show that this
variable is not a proxy of happiness and measures well both freedom of choice and the
locus of control. ‘Internals’ are found to appreciate freedom of choice more than ‘externals’
and to be happier. These findings have important implications for individual utility, social
welfare and public policies.
© 2008 Elsevier B.V. All rights reserved.
1. Introduction
That people are in constant quest of happiness is not a novelty of our times. As noted repeatedly by happiness researchers,
Greek and Roman philosophers since Aristotle have been concerned about the causes of happiness although progress in this
field has been hard to come. Seneca in his opening statement of the De Vita Beata writes to his brother: “Brother Gallio, all
want to be happy, but when it comes to see clearly what makes life happy they are shadowed by obscurity”.1
What distinguishes modern from ancient times in this respect is that we have begun to have some empirical evidence
about what may determine happiness. The last four decades have provided a stream of contributions to happiness research in
several disciplines such as psychology, sociology and economics that significantly changed the way we understand happiness.
We are starting to lift the “shadow of obscurity” by finding elements that seem to explain well fluctuations in self-perceived
happiness.
Drawing on economics and psychology, the paper follows this recent tradition by focusing on one possible predictor
of happiness: Freedom of choice. It is generally accepted that freedom of choice increases happiness but it is unclear how
more freedom of choice turns into more happiness. In this paper, we hypothesize that the appreciation of freedom of choice
∗Correspondence address: Via Po 54, 10124 Torino, Italy.
E-mail address: paolo.verme@unito.it.
1“Vivere, Gallio frater, omnes beate volunt, sed ad pervidendum quid sit quod beatam vitam efficiat caligant.” Seneca (1996, p. 32).
0167-2681/$ – see front matter © 2008 Elsevier B.V. All rights reserved.
doi:10.1016/j.jebo.2009.04.008

Author's personal copy
P. Verme / Journal of Economic Behavior & Organization 71 (2009) 146–161 147
depends on one aspect of personality known as the locus of control. We argue that people who believe that the outcome of
their actions depends on internal factors such as effort and skills (the internals) have a greater appreciation of freedom of
choice than people who believe that the outcome of their actions depends on external factors such as fate or destiny (the
externals). If this is the case, we should find that a measure that combines freedom of choice with the locus of control predicts
happiness better than measures of freedom alone.
An empirical investigation that covers over 260,000 individuals from 84 countries during a period of 25 years finds
evidence in support of this hypothesis. A very strong association between life satisfaction and a variable that measures both
freedom of choice and the locus of control is found controlling for country and individual characteristics, personal values and
social attitudes. This association is stronger and more consistent than the association between life satisfaction and any of the
other known predictors of life satisfaction in a cross-country and within country context. Two tests show that the variable
freedom and control is not a proxy of life satisfaction and that both concepts of freedom of choice and locus of control are
captured by the variable. A third test confirms that the ‘internals’ have a greater preference for freedom of choice than the
‘externals’. These very preliminary findings open an interesting agenda for future research on freedom and happiness and
have important implications for public policies.
We start in Section 2by outlining the main hypothesis of the paper building on theory and empirics drawn from economics
and psychology. Section 3reviews some of the main contributions to happiness research and suggests how this paper
can contribute to such literature. Section 4presents data, model and variables used and Section 5discusses the results.
Section 6provides various tests to check on the robustness of our hypothesis. Section 7concludes by discussing the possible
implications of the findings for public policies.2
2. Freedom of choice, the locus of control and happiness
We can simply define freedom of choice as the size of an opportunity set with mutually exclusive alternatives. The larger
is the set of alternatives (choices) the more is freedom of choice. A restaurant menu listing ten alternatives provides more
freedom of choice than a restaurant menu listing five alternatives.
The appreciation of freedom of choice and the utility derived from freedom of choice may depend on individual pref-
erences. Some people may appreciate freedom of choice more than others. Mary may be happier with ten choices on a
restaurant menu while John may be happier with five choices. We can list at least four possible views on how people may
appreciate freedom of choice:
(1) One view is that the size of the choice set does not matter. What really matters is that the choice set contains the utility
maximizing solution. If the same utility maximizing solution is found in two or more choice sets of different sizes, these
choice sets are equivalent in terms of utility. Neoclassical utility theories, for example, focus on utility maximization and
do not attribute to freedom of choice an intrinsic value. They also tend to ignore individual heterogeneity and assume
that all individuals are equal. In such a framework, increasing the size of the choice set matters only if the probability of
capturing a utility maximizing solution increases with size. For example, with more competitors in the market we should
expect the likelihood that prices will decrease to be higher. However, it is also possible that increasing the choice set leads
to a decreased probability of finding an optimal solution. The voting paradox is one example. We could call this view the
heterotonic/homogeneous view where heterotonic refers to possible outcomes in terms of utility and homogeneous refers
to the characteristics of the agents. According to this view, increasing the choice set (freedom of choice) may lead to more
or less utility (heterotonic outcomes) but the impact will be the same for all agents (homogeneous individuals).
(2) A second view is that freedom of choice is always good for individuals, the larger the choice set the better for individuals,
and this is the same for all individuals. We can call this view the monotonic/homogeneous view. Increasing the choice set
leads invariably to more utility and this applies equally to all individuals.
(3) A third view is what we could call the monotonic/heterogeneous view. In this case, individuals are different in preferences
and an increase in choice has a different impact on individuals but this impact is always positive. Happiness is non-
decreasing in choice. One example would be Sen’s capability theory where freedom of choice contributes to define utility
in a world of heterogeneous individuals. Sen (1987) and others have argued that the size of the choice set or the degree
of freedom of choice has an intrinsic value for individuals.3Expanding the range of possible freedoms such as political
and economic freedoms should be valuable to individuals even if people do not vote or do not profit from the economic
possibilities offered.
(4) A fourth view is that preferences for freedom of choice change across individuals so that increasing the choice set may
have positive or negative consequences on utility. We can call this the heterotonic/heterogeneous view. If some people
have a taste for ease of choice rather than for freedom of choice, an increase in the set of options may lead to reduced
utility. Various explanations have been offered for such kind of attitude. One is that enlarging the choice set leads to
an increased computational cost for individuals so that – at some point – individuals self-restrict the choice problem
2Note that this paper will use the concepts of utility and happiness as one concept and measure it with life satisfaction as in Easterlin (2001) or Alesina
et al. (2004).
3See Gravel (1994) and Bavetta (2004) for critical reviews of this literature.

Author's personal copy
148 P. Verme / Journal of Economic Behavior & Organization 71 (2009) 146–161
Fig. 1. Happiness, freedom of choice and the locus of control.
to be able to take a decision (Simon, 1955). Others have argued that increasing the choice set increases the likelihood
of disappointment for choosing a wrong alternative (Bell, 1985) or the regret for foregone options (Bell, 1982). Indeed,
various experiments have shown that consumers may be adverse to excessive choice. For example, Iyengar and Lepper
(2000) and Sethi-Iyengar et al. (2004) have shown that some consumers prefer not to make a choice if the choice set is
too large.
These last explanations of why some individuals may not favour an increase in the choice set have to do with the degree
of control that individuals think they exercise on the outcome of choices. If John believes that he cannot cope with more
than five choices in a restaurant menu and Mary believes that she can cope with up to ten choices, six choices will result in
less happiness for John and more happiness for Mary. It is therefore important to understand how expectations in relation
to control over choice are formed. Why does John believe that he cannot cope with six choices while Mary thinks she can?
Social and personality psychology offer one interesting concept that could help to explain how people shape expectations
about the outcome of their own choices. This concept is known as the locus of control and was initially proposed by Rotter
(1954).
Rotter (1966, 1990) has distinguished between people who attribute the outcomes of their actions to internal factors
such as their own efforts and skills (the ‘internals’) and people who tend to attribute the outcome of their own actions to
external factors such as fate or destiny (the ‘externals’). Rotter remarked that individuals can be ranked according to the
locus of control and devised a scale (known as the Rotter scale) to measure how close are personalities to an external type
as opposed to an internal type. The locus of control has become a very popular concept since its introduction and is now
accepted in psychology as one of the useful constructs that help to describe personalities.4
How does the locus of control relate to freedom of choice? Our hypothesis is that the locus of control acts as a regulator of
the intrinsic value that people attribute to freedom of choice. The ‘internals’ should attribute more importance to freedom of
choice than the ‘externals’. If I believe that fate alone is managing my life I will not consider having an opportunity to choose
among alternatives as an asset that could improve my life. Vice-versa, if I feel in control of my life and trust that my own
choices will have an impact on my future life I will give a greater value to freedom of choice. Thus, more freedom of choice
should deliver more happiness to internals than to externals.
We can model this hypothesis with a simple graph (Fig. 1). Suppose that we have two agents, John and Mary. John is an
‘external’ who scores low on the Rotter’s scale of control and Mary is an ‘internal’ who scores high on the Rotter’s scale of
control. According to our hypothesis, internals have a greater appreciation of freedom implying that Mary will derive greater
happiness than John at all levels of freedom. Both John and Mary will reach a point where more freedom will turn into
disutility rather than utility but this point will be higher for Mary as compared to John.5
4Note that the locus of control is a very different construct from self-control. Self-control is the ability to control the manifestation of emotions. It is
generally regarded as a facet of conscientiousness, one of the Big Five constructs popular in personality psychology (Goldberg, 1981). The locus of control
is a much earlier construct, has never really found a proper location in the Big Five and is clearly different from self-control in that is related to the inner
and self-evaluation of individuals, not the external manifestation of emotions. There may be a relation between self-control and the locus of control but
the two constructs are clearly different. Skinner (1996) provides a comprehensive review and taxonomy of the various constructs related to control.
5An analogy may illustrate further this point. We could think of agents as sailing boats, freedom as the wind in the ocean, control as the size and strength
of the sails and happiness as the speed of the sailing boat. The stronger is the wind the faster the boat can go. A boat (agent) with larger and stronger

Author's personal copy
P. Verme / Journal of Economic Behavior & Organization 71 (2009) 146–161 149
The question of personality is generally ignored in the economic discourse. In the neoclassical decision utility framework,
all individuals are considered equal in terms of personal characteristics. In Sen’s capabilities theory individual characteristics
are seen as measurable characteristics such as age and education but not as personality. Even in Kahneman’s moment based
framework (Kahneman et al., 1997; Kahneman, 2000), personality is not explicitly treated.
By contrast, it is well known in psychology that personality is associated with happiness. For example, it has been shown
that pleasant and unpleasant affects have a strong genetic basis (Lykken and Tellegen, 1996) and that optimism, self-esteem,
extraversion and neuroticism are all aspects of personality correlated with happiness (Diener et al., 1997; Myers and Diener,
1995).
It is also known that the locus of control is related to happiness. Lower order constructs of personality that include the
locus of control have been found to be closely associated with both job satisfaction and life satisfaction (Judge et al., 1997,
1998)6and internals are consistently found to be happier than externals (Langer, 1983; Strickland, 1989). Research on the
locus of control has evidenced how an internal locus of control is associated with a variety of positive outcomes in adults
and children (Lefcourt, 1982).
Moreover, according to Haworth et al. (1997), there is an established relation between freedom of choice and leisure:
“Freedom of choice in the activity being undertaken has been regarded as a critical regulator of what becomes leisure in people’s
minds. (...) Obtaining intrinsic rewards from engaging in freely chosen activities has been almost unquestionably accepted by
researchers (...).” (pp. 347–348).
In substance, happiness is strongly rooted in personality and the locus of control as well as freedom of choice seem to play
a relevant role in explaining happiness. Both freedom of choice and the locus of control have a direct impact on happiness
and the locus of control may regulate the impact of freedom of choice on happiness. It seems natural therefore to argue
that the combination of the notions of freedom of choice and the locus of control can deliver a very powerful predictor of
happiness.
The policy implications of our hypothesis are multiple. If the locus of control plays a pivotal role in the determination of
happiness and the appreciation of freedom of choice, then we should be concerned about the evolution of the locus of control
from childhood to adulthood and about the intergenerational transmission of the locus of control. Can parents, teachers and
governments contribute to improve the likelihood of happiness in future adults by encouraging the development of internal
personalities as opposed to external personalities?
Existing research across the social sciences would suggest that this is the case. We know from studies conducted by Heck-
man and various co-authors how relevant are early childhood interventions in the cognitive and non-cognitive spheres for
the development of successful adults. We also know how primary and secondary education build on early childhood inter-
ventions to improve individual abilities and capabilities. “The best evidence suggests that learning begets learning. (...) Learning
is a dynamic process and is most effective when it begins at a young age and continues through to adulthood”(Heckman, 200 0). We
have evidence that success in the labor market is partly determined by behavioral traits and that these traits are genetically
and socially transmitted (Bowles et al., 2001a, b). A few studies have also found a positive association between parents and
children locus of control suggesting that the locus of control can be transmitted across generations (see Morton, 1997 for the
results of an experiment and a literature review). As discussed more in detail in Section 5of this paper, research is attributing
increasingly more importance to the role of personality in explaining life outcomes and there is increasingly more evidence
that personality, beliefs and values can be shaped through social policies and are relevant for individuals and nations alike.
3. Predicting happiness
Research on happiness over the past four decades has made tremendous progress in identifying predictors of happiness.
The World Database of Happiness,7which makes an effort to catalogue empirical findings, lists hundreds of variables that
have been found to be correlated with various measures of happiness. For some of these variables, there is quasi unanimous
recognition of their importance. For example, there is a wealth of evidence and little disagreement about the fact that
unemployment and poor health tend to reduce happiness while marriage and religion increase it(Wilson, 1967; Veenhoven,
1996; Diener et al., 1997; Clark and Oswald, 1994; Blanchflower and Oswald, 1997; Winkelmann and Winkelmann, 1998). It
is also generally accepted that individuals or countries with a higher income tend to be happier on average (Blanchflower
and Oswald, 2000; DiTella et al., 2001; Inglehart, 1990; Diener et al., 1995).
More controversial is the relation between happiness and income in longitudinal and life-cycle studies. Easterlin (1974)
was one of the first to note that the increase in GDP per capita in the United States since the 1950s had not been accompanied
by an increase in self-perceived happiness. This finding was confirmed by later studies on the part of the same author (1995,
sails (control) will be able to go faster and further (happiness) but eventually any sail will reach its breaking point, beyond which speed (happiness) will
inevitably decrease. Note that this same idea could be applied to societies rather than individuals. We may think of societies with an internal as opposed
to an external locus of control and we may think of a freedom level where more freedom turns into anarchy and delivers social disutility rather than social
utility.
6Judge et al. (1997) proposed the concept of core evaluations, a core set of conclusions that individuals reach about themselves. These include the locus
of control as well as self-esteem, self-efficacy and neuroticism. Judge et al. (1997) found these four core evaluations to be closely related to job satisfaction
and Judge et al. (1998) found them to be related to both job satisfaction and life satisfaction.
7R. Veenhoven, World Database of Happiness, Correlational Findings: http://worlddatabaseofhappiness.eur.nl (2007).

Author's personal copy
150 P. Verme / Journal of Economic Behavior & Organization 71 (2009) 146–161
2001) and by other authors for the USA (Diener et al., 1999) and for other countries as diverse as Japan (Veenhoven, 1993), the
Philippines (Mangahas, 1995), Russia (Ravallion and Lokshin, 2000) and the UK (Clark and Oswald, 1994).Easterlin (2001)
also noticed that income and happiness do not move together over the life-cycle. People tend to recall that they were worse
off in the past and generally forecast that they will be better off in the future while in fact they report the same level of
happiness at different times during their life-time.
The inconsistent relation between happiness and income in longitudinal studies is generally explained with theories of rel-
ative deprivation or rising expectations. Similar theories have been elaborated by psychologists, sociologists and economists
alike and seem to explain well why happiness does not increase consistently with income over time. In substance, peo-
ple make judgments on the relative position they occupy within a reference group (Runciman, 1966) or adjust quickly to
changed circumstances (Diener et al., 1997; Brickman et al., 1978; Brickman and Campbell, 1971).Easterlin (2001) explains
the finding that happiness does not seem to vary much over the life-cycle arguing that aspirations move upwards together
with income during the life-cycle. This finding not only reinforces what the literature on longitudinal studies finds about
income and happiness but is also consistent with the finding in psychological research that people are not generally good
in either remembering or forecasting feelings and that they tend to undervalue the past and overvalue the future (Gilbert et
al., 1998; Loewenstein and Schkade, 1999).
A model of utility which includes the locus of control could also contribute to explain the lack of covariance between
income and happiness over time or over the life-cycle for individuals. Income expands freedom of choice by definition and we
suggested that the appreciation of freedom of choice (the intrinsic value of freedom of choice) is partly affected by the locus of
control. At very low levels of income, more income turns into more freedom and more happiness for internals and externals
alike. But above a certain level of income and freedom, more income and more freedom can turn into more happiness only
if individuals become more ‘internals’. If the locus of control in adults who live in wealthy countries changes very little over
the working life, more income and more freedom have little effect on happiness.
For countries and in the long run this may be very different. When countries move from autocracies to democracies,
improve their educational system and try to empower their people they are in fact fostering the development of internals
over externals. Vice-versa, if countries are authoritarian and encourage obedience rather than critical attitudes they tend to
reward and prefer externals over internals. Countries try intentionally to shape the personality of their citizens via public
policies such as educational policies. Economic development often (but not always) coincides with the transition from
autocracy to democracy and the development of internal over external personalities allowing more and more people to
enjoy the benefit of more freedom. However, this effect can only be observed if a transition from autocracy to democracy is
occurring, over the very long period when generational changes occur and for countries rather for individuals. Individuals
may have rather stable personalities over the life-cycle but the average personality in a given country may change significantly
across generations. We will discuss this point in greater detail in Section 5at the end of the paper.
In our knowledge, research on happiness, freedom of choice and the locus of control has only one precedent, a study
by Veenhoven (2000) that focused on the relation between freedom and happiness. The author devised two measures of
freedom, one based on the opportunity to choose and the second based on the capability to choose. In particular, capability
to choose is measured with two variables, one capturing individualistic work values and the other measuring what the
author defines as ‘perceived fate-control’, which is what social psychologists define as the locus of control. The author finds a
positive and significant correlation between happiness and each of the components of freedom described including perceived
fate-control. The relation seems to be linear and richer nations are shown to be happier and freer as compared to poorer
nations.
The ‘perceived fate-control’ variable used by Veenhoven has been taken from a question present in the World Values
Survey.8The question asked was: “Please use this scale where ‘1’ means “none at all” and ‘10’ means “a great deal” to indicate
how much freedom of choice and control you feel you have over the way your life turns out.” More recently, Inglehart et al. (2008)
have used this same variable as a measure of freedom of choice. In our view, the question combines information on freedom
of choice with information on the locus of control. It is therefore the ideal instrument to start our investigation on the relation
between happiness and freedom.
4. Data, model and variables
We use a large data set compiled from the European and the World values surveys.9These surveys have been carried
out since the early 1980s and question individuals worldwide on happiness, personal values, social attitudes and individual
8Available at: http://www.worldvaluessurvey.org/. Veenhoven refers to the World Values Survey 2, item 95 but this same question has been asked in all
rounds of the World and European Values Surveys.
9Values surveys 1981–2004, integrated questionnaire version 20060423. Data can be freely downloaded from: http://www.jdsurvey.net. We are grate-
ful to the Values Study Group and World Values Survey Association for creating and making accessible the EUROPEAN AND WORLD VALUES SURVEYS
FOUR-WAVE INTEGRATED DATA FILE, 1981–2004, (v.20060423, 2006). Aggregate File Producers: Análisis Sociológicos Económicos y Políticos (ASEP) and
JD Systems (JDS), Madrid, Spain/Tilburg University, Tilburg, The Netherlands. Data Files Suppliers: Analisis Sociologicos Economicos y Politicos (ASEP)
and JD Systems (JDS), Madrid, Spain/Tillburg University, Tillburg, The Netherlands/Zentralarchiv fur Empirische Sozialforschung (ZA), Cologne, Germany.
Aggregate File Distributors: Análisis Sociológicos Económicos y Políticos (ASEP) and JD Systems (JDS), Madrid, Spain/Tillburg University, Tilburg, The
Netherlands/Zentralarchiv fur Empirische Sozialforschung (ZA) Cologne, Germany.

Author's personal copy
P. Verme / Journal of Economic Behavior & Organization 71 (2009) 146–161 151
attributes. The version of the data set we use contains 267,870 observations on individuals from 84 countries surveyed
between 1981 and 2004 where each country has been surveyed from a minimum of one to a maximum of four times.
The full data set contains 913 variables most of which could be used as predictors of life satisfaction. In an effort to learn
from the data as much as possible and avoid missing on important variables we first run OLS bivariate regressions between
life satisfaction and all the possible regressors of life satisfaction present in the database. We then ranked variables on the
basis of the R squared values. As expected, the variables with the highest R squared were proxies of life satisfaction such
as happiness or satisfaction with income, family or job. This type of variables occupied the top ten positions in terms of R
squared in a list of over 800 variables. With one exception. This was the freedom and control variable used by Veenhoven
(2000) and describe d in the previous section, which ranked in 7th place. Subjective health and income were also very relevant
in this classification with the first variable in 15th place and the second variable in 25th place but the real surprise of this
exercise was the variable freedom and control which had an R squared of 0.165 as compared to an R-squared for subjective
health of 0.0873 and of 0.0412 for income rank. Freedom and control seemed to explain life satisfaction twice as well as
subjective health and four times as well as income rank.
Many variables in the database were only present for some years or for some country and the number of observations
available varied significantly across variables. This made the R squared comparison across variables difficult as we compared
different sets of observations. Restricting the possible predictors of life satisfaction to only those variables with at least
100,000 observations reduced the database to about a fourth of the original number of variables. Among these variables,
freedom and control ranked 3rd in terms of R squared after two proxies of life satisfaction (happiness and satisfaction with
the financial situation of the household). Subjective health followed in 4th place and relative income in 8th place. Restricting
further the database to variables with at least 200,000 observations reduced the data set to a further half of the variables
leaving approximately 100 variables. If we exclude the proxies of life satisfaction which occupied the top two positions, the
top three variables in order of importance were freedom and control, subjective health and income rank in this order. In a
bivariate context, the variable freedom and control emerged as the best explanatory factor of life satisfaction.
On the basis of the happiness literature discussed in the previous section and on the basis of the bivariate exercise
described above we defined the multivariate equation as follows:
Hi=˛+!Fi+"Cc+ˇEi+#Pi+ıVi+$Si+%i
where His subjective happiness; Fis the variable that measures freedom of choice and control over one own life (freedom and
control for short); Cis a vector of macroeconomic country variables; Eis a vector of individual entitlements such as income
and work; Pis a vector of personal and family characteristics, Vis a vector of variables standing for individual values; Sis a
vector of variables standing for individual social attitudes;˛, !,",ˇ,#, ı and $are the parameters to be estimated and εis the
error term. The subscript istands for individuals and the subscript cstands for countries. The regression is estimated first for
the pooled world sample and in a second stage for all countries available omitting the macroeconomic country variables. For
all estimations, we use an ordered logit model, the robust Huber-White sandwich estimator and regional cluster estimates.10
Subjective happiness (H) is measured with a question on life satisfaction. The question asked is: “All things considered, how
satisfied are you with your life as a whole these days?” Answers include a ten steps ladder where ‘1’ is equal to “Dissatisfied”
and ‘10’ is equal to “Satisfied”. This question is a rather standard question used in happiness research and validation studies
in various disciplines have shown that answers to this question are reliable (Lepper, 1998; Sandvik et al., 1993; Fordyce,
1988; Inglehart, 1990; Saris et al., 1996).
The variable freedom (F) is the variable already described where the question asked is: “Please use this scale where ‘1’
means “none at all” and ‘10’ means “a great deal” to indicate how much freedom of choice and control you feel you have over the
way your life turns out.” From the formulation of the question we derive that this variable captures two aspects which we said
are closely related: Freedom of choice and the locus of control. Personality being equal, two persons who enjoy a different
degree of individual freedom should provide different scores to this question. Vice-versa, freedom of choice being equal, two
persons with a different locus of control should provide different answers. Further in the paper we will test whether this
variable captures effectively both aspects of freedom and control.
We use two macroeconomic variables (C) to account for country economic heterogeneity. The first variable is GDP per capita
estimated at Purchasing Power Parity (2000 prices). This variable is extracted from the World Bank Indicators database11 and
is the only variable which is exogenous to the database used. The second variable is the country employment rate calculated
as the number of employed people divided by the working age population. This was preferred to the unemployment rate
because unemployment is used already as an individual variable and because the ILO unemployment definition is not suited
for informal and developing economies which are largely represented in our database.12
10 Regional cluster estimates are indicated in our case for at least tworeasons. One is that regressing summary country measures such as GDP on individual
measures such as life satisfaction may provide bias estimates(Moulton, 1986). And the second is that within regions the number of observations is generally
small and interviews may have been concentrated in restricted spatial areas failing to capture the full within region variance.
11 Available at: http://www.worldbank.org.
12 Where unemployment benefits are non-existent and in rural areas the real poor cannot afford to seek employment and engage themselves in survival
activities. In such situations the ILO unemployment rate is a very poor indicator of labor market status. On the other hand, the employment rate is affected
by variations in the working age population and provides no information about the quality of employment. Both GDP and the employment rate are also
introduced in the equation in squared form.

Author's personal copy
152 P. Verme / Journal of Economic Behavior & Organization 71 (2009) 146–161
Two variables were selected to capture individual economic status (E). These are income and unemployment. Income is
measured as self-positioning in a ten-steps income scale where the income brackets have been measured in local currency in
each country. This is not self-perceived income but the positioning of individuals into income brackets. In some sense, this is
a more accurate indicator than self-reported income which is known to be underreported in household surveys worldwide.
That is because people are not asked to tell how much they earn but simply to say to which income brackets they belong
to. A categorical variable constrains the variance of the income variable as compared to a continuous variable but this is
not a great shortcoming considering that the dependent variable is categorical (also based on a ten-steps ladder) and that
coefficients are estimated with an ordered logit model. We call this variable ‘income rank’ because it measures the income
rank of individuals rather than the value of income. The unemployment status is the self-reported unemployment status
measured with a binary variable.
A set of variables measures individual attributes (P). These are sex (1= female and 0= male), age (continuous with the
addition of age squared), a dummy for tertiary education and marriage status (dummy where ‘1’ includes: “married” and
“living together as married”).
Personal values (V) are taken into account with four variables. These include the importance attributed by individuals
to family, work, religion and politics. All these variables were originally measured on a scale from one to four where one
was “Very important” and four was “Not important at all”. We created dummies for each variable with one equal to “Very
important” or “Rather important” and zero equal to “Not very important” and “Not at all important”. Values matter for at least
two reasons. One is that they contribute to define individual personalities as they are partly an expression of personality.
And the second is that they contribute to determine how much importance people give to the different attributes they have.
For example, being married or being unemployed have an impact on life-satisfaction but we should expect these variables
to have a different impact depending on the importance that people give to family or to work.
Another set of variables captures what we call social attitudes (S). One variable measures on a scale from one to ten how
people think is justifiable to cheat on taxes where one corresponds to “Never” and ten to “Always”. This seemed an important
control for social cooperation and also an aspect which may contribute to define personality. Another variable measures the
political orientation of people on a scale from one to ten where one corresponds to “Left” and ten corresponds to “Right”. This
variable has been used in the past and found to be an important predictor of happiness (Alesina et al., 2004). A third variable
measures the degree of trust in institutions that people have. The surveys asked respondents to rank from one to four the
degree of trust in various types of national institutions where one was equal to “A great deal” and four to “None at all”. We
calculated the individual average trust for the institutions of the army, police, justice, parliament, civil service, press, private
companies and trade unions and we then reversed the score to make trust increasing in happiness. Thus, this variable ranges
from one to four but entered the equation as a continuous rather than a categorical variable. A last variable is trust in people
measured with a dummy variable where one is “Most people can be trusted” and zero is “Can’t be too careful”. The trust vari-
ables account for the mutual trust present in society and can be considered as a measure of social capital as in Helliwell (2003).
5. Results
In Table 1 we report the multivariate results for the life satisfaction equation estimated on the world pooled sample with
an ordered logit model, robust standard errors, regional clusters and year fixed effects. The world sample for which the
specified equation could be estimated includes 75 countries, 1,119 regions and 160,405 observations. The sample is reduced
vis-à-vis the original sample given that not all variables have a full set of observations. Selection bias can be checked in
Table A1 which provides descriptive statistics for each variable and for the full and reduced sample used in Table 1. As it can
be seen from the table, means and standard deviations are very close between the reduced sample and the full sample and
we should exclude that our reduced sample is significantly biased vis-à-vis the full sample of 84 countries.
The variable freedom and control is by far the most significant predictor of life satisfaction. It shows the highest coefficient,
the highest odds ratio, the highest z-score and one of the lowest standard errors. For a one step increase in the one to ten
freedom and control scale, happiness is expected to change by about 36 percent of a step on the one to ten happiness scale
(considering the ordered log-odds scale with the other variables held constant).
Individual economic status. Income rank has also a positive effect but with decreasing marginal effects as rank increases.
This conforms to previous results on various income variables. Income is a powerful predictor of life satisfaction at low levels
of income but its predicting capacity decreases as income increases. Also, as shown by previous studies, unemployment is a
strong predictor of unhappiness.
Individual characteristics. Across the world sample, females seem to be happier than males on average while increasing
age decreases happiness up until a certain age when the trend reverses. Tertiary education marginally increases happiness
and being married is a very strong predictor of happiness as it is well known in the literature.
Individual social attitudes. In societies where people trust other people and the national institutions people are happier
while individuals who have a lax attitude towards tax cheating seem to be more unhappy. This conforms to and reinforces
what we know about social capital and its role for happiness.
Individual values. Including a high importance attributed to family, work and religion are all good predictors of happiness
with a positive sign. Religion in particular seems to be the strongest predictor of happiness among the ‘values’ variables.
Instead, individuals who attribute a great importance to politics seem to be less happy on average, although the effect is
rather weak. These are all results consistent with previous literature.

Author's personal copy
P. Verme / Journal of Economic Behavior & Organization 71 (2009) 146–161 153
Table 1
Life satisfaction equation—pooled world sample. a
Coeff. Std. Err. Odds ratio Std. Err. z
Freedom and Control
Freedom and control 0.362 0.010 1.436 0.014 37.6
Individual Economic Status
Income rank 0.164 0.021 1.178 0.025 7.8
Income rank squared −0.008 0.002 0.992 0.002 −4.8
Unemployed −0.431 0.026 0.650 0.017 −16.9
Individual Characteristics
Female 0.052 0.011 1.053 0.011 4.9
Age −0.054 0.003 0.947 0.003 −20.3
Age squared 0.0 01 0.000 1.0 01 0.000 17.4
Education-tertiary 0.105 0.022 1.110 0.024 4.8
Married 0.292 0.018 1.339 0.024 16.4
Individual Social Attitudes
Tax cheat −0.033 0.004 0.967 0.003 −9.4
Trust in people 0.127 0.021 1.135 0.023 6.2
Trust in institutions 0.212 0.021 1.236 0.026 10.2
Individual Values
Family importance 0.351 0.040 1.421 0.057 8.8
Work importance 0.142 0.022 1.153 0.026 6.4
Religion importance 0.302 0.023 1.353 0.031 13.3
Politics importance −0.047 0.016 0.954 0.015 −3.0
Country Economic Status
GDP (000) 0.064 0.006 1.067 0.007 10.2
GDP squared (000) −0.001 0.000 0.999 0.000 −3.7
Employment rate 0.016 0.006 1.016 0.006 2.9
Employment rate squared 0.000 0.00 0 1.000 0.0 00 −3.7
aOrdered logit estimations with robust standard errors, regional cluster and year fixed effects. 75 countries, 1119 regions, 160,405 observations.The odds
ratio is computed as ‘e’ to the power of the logistic coefficient.
Country economic status. Both GDP per capita in purchasing power parity and the employment rate have a positive effect
on life satisfaction and both with decreasing returns. At low levels of GDP, a rise in output generates a significant rise in life
satisfaction. This effect disappears as GDP per capita reaches high values. The effect for the employment rate is also positive
at low levels of employment and diminishes for high levels. Thus, both GDP and the employment rate can help to improve
happiness in poor countries but improving happiness simply with increases in these two measures becomes a very hard task
for rich nations. Again, these results are largely consistent with the existing literature.
The pooled sample we used in Table 1 took into account some aspects of the economic country situation captured by the
country variables described but could not take into account the full country heterogeneity. Deriving lessons for individual
countries from a pooled world sample is also difficult as economic policies are still largely made within countries. With very
large samples is also easier to detect covariances among variables but these covariances are not necessarily valid for each
country.
We decided therefore to run the same equation we used for the world pooled sample (excluding the country economic
variables) for all 75 countries considered in Table 1. Full results cannot be shown for all countries. In Table 2A we report, as
an illustration, full results for ten representative countries.13 In Table 2B, we report only the number of times each predictor
is significant across the 75 countries and whether significant predictors take a negative or a positive sign. As before, we
use ordered logit estimations, robust standard errors, regional clusters and year fixed effects to make results as robust as
possible.
Freedom and control is the only variable that is consistently significant with a positive sign across all ten countries in
Table 2A. The coefficient and the z-score is always very high ranging from 0.548 in Canada to 0.242 in Nigeria. We can also
observe that the coefficient of the freedom and control variable tends to decrease as we move from developed to developing
countries which is in line with the hypothesis that the locus of control is likely to become more important as countries
develop and improve freedom. Across the full sample of 75 countries (Table 2B) the freedom and control variable is always
significant with one exception (Turkey) and varies in size between 0.08 (Egypt) and 0.712 (New Zealand), always with a
positive sign.
13 The selection of the ten countries was made on the basis of cultural diversity, population size and geographical location. In terms of number of
observations, the ten countries selected represent over a quarter of the sample used in Table 1 and in terms of population they represent over half of the
world population.

Author's personal copy
154 P.Verme / Journal of Economic Behavior & Organization 71 (2009) 146–161
Table 2A
Life satisfaction equations—selected countries. a
USA Canada Germany Spain South Africa Mexico Russia China India Nigeria Signif. (#) Change sign if
signif.?
Freedom and control 0.496 0.548 0.50 6 0.517 0.385 0.405 0.260 0.341 0.267 0.242 10 No
(26.76)** (27.96)** (21.11)** (29.99)** (9.26)** (5.92)** (19.14)** (10.34)** (6.57)** (13.00)**
Income rank −0.039 0.017 0.055 0.186 0.44 0.112 0.062 0.169 −0.029 −0.053 2 No
−0.81 −0.31 −1.15 (3.64)** (10.59)** −1.23 −0.84 −1.71 −0.21 −0.57
Income rank squared 0.01 0.0 01 −0.0 02 −0.01 −0.026 −0.006 0.003 −0.001 0.016 0.02 4 Yes
(2.76)** −0.16 −0.44 (2.13)*(6.85)** −0.86 −0.55 −0.09 −1.14 (2.38)*
Unemployed −0.268 −0.559 −1.432 −0.586 −0.539 0.09 −0.37 0.286 −0.165 0.019 6 No
(2.08)*(3.63)** (12.69)** (7.27)** (6.72)** −0.38 (3.51)** −1.33 −1.42 −0.2
Female 0.002 0.045 0.129 −0.036 0.064 0.167 −0.019 0.188 0.055 0.168 4 No
−0.03 −1.84 (2.41)*−0.94 −1.14 (2.12)*−0.21 (2.24)*−1.31 (2.72)**
Age −0.026 −0.042 −0.049 −0.06 −0.062 −0.062 −0.064 −0.06 −0.0 05 −0.065 9 No
(2.46)*(4.28)** (6.15)** (5.55)** (6.86)** (2.82)** (3.91)** (3.82)** −0.38 (6.43)**
Age squared 0.00 0 0.001 0.000 0.0 01 0.001 0.001 0.001 0.001 0.0 0 0.001 9 No
(3.07)** (5.58)** (5.92)** (5.05)** (6.30)** (2.82)** (3.41)** (4.05)** −0.25 (6.37)**
Education – tertiary 0.002 −0.121 0.252 0.253 −0.14 −0.068 0.335 −0.001 0.127 0.253 6 Yes
−0.02 (2.29)*(2.59)** (2.71)** (2.81)** −1.23 (4.48)** −0.02 −1.88 (5.04)**
Married 0.589 0.691 0.506 0.554 0.302 0.35 0.29 0.509 0.118 0.217 9 No
(5.15)** (11.61)** (9.61)** (7.54)** (5.43)** (6.18)** (4.02)** (4.76)** −1.76 (3.19)**
Tax cheat −0.058 −0.013 −0.02 −0.029 −0.022 −0.049 −0.001 −0.094 −0.029 0.014 5 No
(5.16)** −0.88 (2.11)*(2.83)** −1.42 (4.04)** −0.06 (3.63)** −1.15 −0.58
Trust in people 0.156 0.054 0.427 0.082 0.193 −0.091 0.22 0.23 0.079 −0.041 4 No
−1.78 −0.56 (10.30)** −1.02 (3.00)** −1.13 (3.43)** (2.99)** −0.65 −0.91
Trust in institutions 0.182 0.31 0.416 0.078 0.096 −0.018 0.418 0.337 0.127 0.241 7 No
(2.20)*(2.47)*(4.82)** (1.98)*−1.16 −0.43 (6.21)** (4.02)** −1.17 (6.04)**
Family importance 0.371 0.58 0.195 0.32 0.315 0.367 0.178 0.274 −0.18 0.215 3 No
−1.42 (2.97)** −1.18 −0.96 −1.25 (3.38)** (2.18)*−1.24 −0.6 −0.43
Work importance −0.222 −0.074 0.268 0.251 −0.017 0.3 0.069 0.265 −0.076 0.08 4 Yes
(2.57)*−0.58 (3.22)** (2.77)** −0.17 (2.6 6)** −1.21 −1.76 −0.51 −0.58
Religion importance 0.267 0.165 0.165 0.151 0.288 0.064 0.135 −0.119 0.343 0.668 8 No
(3.76)** (3.39)** (2.89)** (3.05)** (5.96)** −0.8 (2.29)*−1.36 (3.06)** (4.08)**
Politics importance 0.091 −0.059 0.037 −0.087 −0.1 −0.11 −0.038 0.173 0.03 0.044 2 Yes
−1.66 −1.15 −0.74 −1.74 (2.20)*−1.42 −0.63 (2.47)*−0.33 −0.95
Observations 4071 3104 6016 5521 6 848 4344 4980 2755 5053 4321
Pseudo R-squared 0.08 0.09 0.1 0.08 0.09 0.07 0.05 0.07 0.06 0.04
GDP/capita PPP (000, aver.) 27.532 23.777 22.611 16.984 9.719 8.086 7.506 2.825 2.045 0.868
aOrdered logit estimations with regional clusters and year fixed effects. zstatisticsin parentheses.
*Significant at 5%.
** Significant at 1%.

Author's personal copy
P. Verme / Journal of Economic Behavior & Organization 71 (2009) 146–161 155
Table 2B
Life satisfaction country regressions—number and sign of significant predictors.
Variable Sign Total Percentage
(+) (−)
Freedom and control 74 0 74 98.7
Age 1 57 58 77.3
Married 54 0 54 72.0
Trust institutions 48 1 49 65.3
Unemployed 3 43 46 61.3
Age squared 43 0 43 57.3
Religion important in life 36 3 39 52.0
Income rank 29 2 31 41.3
Trust people 30 0 30 40.0
Tertiary education 18 11 29 38.7
Justifiable: cheating on taxes 1 27 28 37.3
Family important in life 23 2 25 33.3
Female 20 2 22 29.3
Income rank squared 7 10 17 22.7
Work important in life 11 5 16 21.3
Politics important in life 4 6 10 13.3
Total countries 75 100.0
Ordered logit estimations with regional clusters and year fixed effects. Variables significant at 1% or 5%.
In terms of cross-country consistency, age follows freedom and control with nine countries where this variable is sig-
nificant in Table 2A and 58 countries in Table 2B. With one exception, age takes always a negative sign. Age squared is also
significant in 43 countries and always with a negative sign indicating that this variable is concave. Happiness tends to increase
with age but only up to a point when it starts to decline.
Marriage comes in third place in terms of importance with nine countries where this variable is significant in Table 2A
and 54 countries in Table 2B, always with a positive sign. The only country in Table 2A where marriage is not significant
is India. This country is the only of the ten countries considered in Table 2A that uses the practice of arranged marriage
extensively. Despite evidence that arranged marriages can work, it could be that – on average – non-arranged marriages are
more successfull. However, India is not the only country in Table 2B where marriage is non-significant and other factors such
as the role of women in society may well be at work.
Trust in institutions is significant and with a positive sign in seven of the ten countries in Table 2A and in 49 of the
75 countries in Table 2B. With one exception, the sign of this variable is always positive. Trust in people is also signif-
icant and with a positive sign in 30 countries. In Table 2A, the countries where trust in institutions is not significant
are South Africa, Mexico and India, three very large, culturally diverse, democratic and developing nations. On the con-
trary, trust in institutions is positive and significant in Russia and China, two countries also very large and caught in
a process of development but more autocratic and less culturally diverse. Social capital is very relevant overall but not
everywhere and it is unclear from our data what are the factors that make social capital a good predictor of life satisfac-
tion.
The status of unemployed is found to be a significant predictor of happiness in six countries in Table 2A and in 46
countries in Table 2B. Together with religion, this is the only other variable which is significant in over 50% of the countries
considered.
The importance of religion is significant in eight countries in Table 2A and in 39 countries in Table 2B. It is worth noting
in Table 2A that the importance of religion is not significant in China as we may expect and in Mexico which is instead a
deeply devoted Catholic country. In Table 2B we also show that in three countries the sign of this variable is significant and
negative which is contrary to what is expected. Therefore, despite the quasi universal consensus on the part of researchers
in accepting the relevance of religion for life satisfaction, we find this variable significant in only half of the countries and
not always with the expected sign.
Income rank is significant in only two countries in Table 2A and in 31 countries (41.3% of the sample of countries)
in Table 2B. The effect on happiness is positive with the exception of two countries where the effect is negative. This is
surprising particularly in the light of the fact that both income and relative income have been found in the past to be relevant
in most countries studied and especially in poor countries. The variable we used is neither income nor relative income but
we should expect income rank to show a consistent positive sign. Instead, in Table 2A, income rank is significant for only
Spain and South Africa and among the 31 countries where this variable is significant we find both rich and poor nations and
also two negative signs.
Our results indicate that countries heterogeneity is remarkable and reading in world data or in single countries data
universal findings can be very misleading. However, this statement does not apply to the variable freedom and control
which is a remarkable stable predictor of life satisfaction in all countries. If we had to bet on what variables best predict life
satisfaction anywhere in the world our money would certainly go on freedom and control.

Author's personal copy
156 P. Verme / Journal of Economic Behavior & Organization 71 (2009) 146–161
6. Tests
We have established with a certain degree of confidence that freedom and control is the best predictor of life satisfaction
worldwide among the variables we dispose of. In this section we want to address three questions which may challenge the
validity of our hypothesis.
(1) The first question is whether freedom and control and life satisfaction are, in fact, proxies (we call this the ‘proxies’
hypothesis). The questions asked are different and relate to different objects but it could be that people perceive the
two questions as the same. We have already shown that psychologists found freedom of choice and leisure to be closely
related in people’s mind (Haworth et al., 1997) and it could be that people consciously or unconsciously reply to the two
questions as if they were answering the same question. We had this question posed on several occasions when this paper
was presented in seminars and conferences.
(2) The second question is whether freedom and control is a variable that relates only to freedom of choice or only to the
locus of control or to both (we call this the ‘double role’ hypothesis). The formulation of the question would suggest
that people considers both components when answering the question but we did not provide evidence of that. Also, as
already discussed, research in psychology has shown how close freedom of choice and the locus of control are (Langer,
1983; Strickland, 1989). It is difficult to separate freedom of choice from the locus of control but we can check if the
variable freedom and control is correlated to both sentiments or to just one of the two.
We address these two questions with the estimations proposed in Table 3. This time we regressed the same set of
variables on life satisfaction and on freedom and control separately. We tested the ‘proxies’ and ‘double role’ hypotheses
as follows. First, among the variables already used, we picked two variables that we expected to have a positive impact on
life satisfaction but a negative effect on freedom. These are ‘being married’ and the ‘importance of religion’. We know the
institutions of marriage and religion to enhance happiness but we also expected these two institutions to limit freedom
of choice.
Second, we picked two other variables which could be considered as correlates of the locus of control but with opposite
signs. These are the importance attributed to child obedience and the importance attributed to child independence. We
expected child obedience to be a feature that would be most appreciated by the externals, those who think that what
happens to them depends on factors outside their control, and child independence to be a feature most appreciated
by the internals, those who think that they can determine their own future. We expected these two variables to have
opposite signs within each equation and between the two equations. If these expectations are met, then life satisfaction
and freedom cannot be considered as proxies and the freedom variable would show to have elements of both freedom
of choice and the locus of control.
In the two equations we also added a number of controls including income rank, income rank squared, unemployed,
female, age, age squared and tertiary education. We also included freedom as regressor in the life satisfaction equation
and life satisfaction as regressor in the freedom equation so as to remove all noise due to other factors unrelated to
our four variables of interest. Estimates were conducted on the world pooled sample using ordered logit estimates with
robust standard errors, country and year fixed effects and regional clusters.
Table 3
Life satisfaction vs. freedom and control.
Lifesat Freedom
Constraints to freedom
Married 0.338 −0.137
(20.33)** (9.84)**
Religion importance 0.201 −0.04
(15.26)** (2.10)*
Important child qualities
Obedience 0.071 −0.042
(5.73)** (2.86)**
Independence −0.05 0.09
(4.32)** (7.79)**
Freedom and control 0.319
(15.12)**
Life satisfaction 0.344
(18.62)**
Observations 187,198 187,198
Ordered logit estimates with robust standard errors, country and year fixed effects, regional cluster
and a set of controls. Controls are income rank, income rank squared, unemployed, female, age, age
squared and tertiary education. Robust zstatistics in parentheses.
*Significant at 5%.
** Significant at 1%.

Author's personal copy
P. Verme / Journal of Economic Behavior & Organization 71 (2009) 146–161 157
As shown in Table 3, signs are all as expected. Being married and the importance of religion have both a positive
and significant coefficient in the life satisfaction equation while they have a negative and significant coefficient in the
freedom equation. Child obedience has the expected negative and significant sign in the freedom equation while child
independence has a positive and significant sign. All coefficients in the table are significant and they all show opposite
signs for the life satisfaction and freedom equations. We conclude that life satisfaction and freedom and control cannot
be interpreted as proxies and that the variable freedom and control is related to both aspects of freedom of choice and
the locus of control.
(3) Next we want to check whether the hypothesis that internals have a greater appreciation of freedom than externals is
actually true (we call this the ‘freedom lovers’ hypothesis). The locus of control is generally measured with questionnaires
aiming to capture personality traits typical of internals and externals. For example, two popular questionnaires are
Rotter’s and Duttweiler’s questionnaires which are extensively used in psychology (see Fischer and Corcoran 2007 for
examples of these questionnaires). Questions identical to those used in the named questionnaires are not available
in the database we use and we cannot construct a precise locus of control scale. However, some of the questions we
have measure personality traits very similar to those generally attributed to internals such as self-confidence, positive
attitudes towards responsibilities and a taste for hard work. Using these questions, we could construct two variables able
to capture internal personalities: an ordinal scale ranging from external to internal personalities and a dummy variable
for internals.14
We also disposed of questions on the appreciation of various forms of freedom. For example, we had questions asking to
respondents whether they preferred freedom over equality or freedom over order or the importance attributed to individ-
ual economic freedom and to freedom of speech.15 We could therefore check whether internals have effectively a greater
appreciation of freedom than externals by regressing the dummies constructed to capture the appreciation of freedom on
the two constructed measures of internal personality.16
Results of these estimations for the pooled world sample are presented in Table 4. As expected, internals show a signifi-
cantly greater appreciation of freedom as compared to externals. Individuals who have a greater appreciation for economic
individual freedom and who have a preference for freedom over equality tend to score high on the internal scale that we
constructed (Columns 1 and 2). And individuals who have a greater appreciation of freedom over order and a greater appre-
ciation of freedom of speech tend to be internals rather than externals (columns 3 and 4). All the appreciation of freedom
variables constructed show a positive and significant sign at the one percent level.
7. Some implications for public policies
We have established that freedom of choice combined with the locus of control is a very powerful predic-
tor of life satisfaction. But does this matter for public policies and why? We think it matters in many different
respects.
Personality, or at least one of the aspects of personality – the locus of control – seems to contribute to shape the preference
attributed by individuals to freedom of choice and this, in turn, has an impact on utility. Utility theory and modern critiques
of utility theories have largely ignored the question of personality whereas we know from psychology and confirmed by this
study that personality has a great role in explaining choice and utility. It is not sufficient to have more choice, we need to
feel in control of these choices to be happier.
Moreover, personality seems to matter not only for individuals but also for nations as if countries had personalities.
Transitional economies provide a concrete example of this phenomenon. The European and the World Values Surveys show
that transitional economies were almost invariably at the bottom of the happiness league at the end of the 1990s and beyond.
These economies went through a deep recession during the 1990s and this may explain the low scores on happiness. However,
freedom of choice has increased in many respects and transitionaleconomies continue d toscore very low on happiness during
the more recent growth phase. Happiness levels are much lower than in other countries with a similar level of income per
capita. Opinion polls across transitional economies also indicate that the majority of citizens still expresses a preference for
the old socialist times. However, this is true for the old generation but not for the new generation and this is precisely what
our model predicts. The old generation, trained to delegate responsibilities for family and work to the state, has experienced
the transition to more freedom as a negative rather than a positive shock whereas the new generation may be better equipped
to make use of more freedom.
Several recent studies seem to come to the same conclusion. Inglehart et al. (2008) found that countries where liberties
have increased have also been countries where the perception of freedom of choice has increased, and this has been an
important factor in explaining increased happiness. In the words of these authors: “Happiness reflects not only people’s
objective experiences, but also how they evaluate these experiences. (...) In recent years, economic growth, democratization, and
14 See Table 4 for a description of these variables.
15 See Table 4 for a description of these variables.
16 Note that some of the questions selected were available for only two of the surveys included into the database and that the sample used is small.
Estimations cannot be considered as representative of the full sample of 84 countries.

Author's personal copy
158 P. Verme / Journal of Economic Behavior & Organization 71 (2009) 146–161
Table 4
The appreciation of freedom and the locus of control.
Dep. Var.: Internal scale (A) Internal scale (A) Internal dummy (B) Internal dummy (B)
Estimation: Ordered logit Ordered logit Probit Probit
Freevseq (1) 0.0661 ***
(0.0249)
Freepeople (2) 0.118 ***
(0.0246)
Freespeech (3) 0.359 ***
(0.0514)
Freevsorder (4) 0.113 ***
(0.0317)
Income rank 0.0255 0.0188 −0.0467 −0.0438
(0.0216) (0.0216) (0.0327) (0.0281)
Income rank squared 0.00450 ** 0.00525 *** 0.00592 ** 0.00687 ***
(0.00179) (0.00179) (0.00281) (0.00247)
Unemployed −0.0730 −0.0854 *−0.0541 −0.0393
(0.0493) (0.0483) (0.0626) (0.0494)
Female −0.233 *** −0.227 *** −0.235 *** −0.233 ***
(0.0244) (0.0237) (0.0342) (0.0313)
Age 0.00140 −0.000345 −0.0153 ** −0.0160 ***
(0.00516) (0.00548) (0.00651) (0.00568)
Age squared −2.70e-05 −2.64e-06 0.000177 ** 0.000196 ***
(5.22e-05) (5.49e-05) (6.91e-05) (6.08e-05)
Education-tertiary 0.214 ** 0.172 ** 0.381 *** 0.391 ***
(0.0838) (0.0721) (0.0450) (0.0380)
Married −0.0163 −0.0264 0.0102 0.0107
(0.0244) (0.0256) (0.0349) (0.0304)
Tax cheat 0.0146 *** 0.0129 ** −0.0536 *** −0.0521 ***
(0.00542) (0.00527) (0.00704) (0.00651)
Trust in people 0.0262 0.0227 0.0275 0.0321
(0.0197) (0.0206) (0.0432) (0.0436)
Trust in institutions 0.142 *** 0.145 *** 0.220 *** 0.245 ***
(0.0275) (0.0266) (0.0392) (0.0348)
Family importance −0.0703 −0.0157 0.146 0.215 **
(0.0751) (0.0803) (0.116) (0.105)
Work importance 0.279 *** 0.256 *** 0.268 *** 0.286 ***
(0.0461) (0.0497) (0.0701) (0.0554)
Religion importance 0.0457 ** 0.0435 ** 0.0775 *0.0923 ***
(0.0219) (0.0213) (0.0400) (0.0300)
Politics importance 0.222 *** 0.216 *** 0.216 *** 0.222 ***
(0.0249) (0.0298) (0.0417) (0.0359)
Constant 0.337 0.0850
(0.245) (0.206)
Observations 37,625 37,000 10,547 14,194
(A) 0–8 Scale. One point is given for each of the following statements: (1) I usually count on being successful in everything I do; (2) I enjoy convincing
others of my opinions; (3) I serve as a model for others; (4) I am good at getting what I want; (5) I own many things others envy me for; (6) I like to assume
responsibility; (7) I am rarely unsure about how I should behave; (8) I often give others advice. Zero is given if respondents did not subscribe to any of the
eight statements above. (B) Dummy variable= 1 if respondents mentioned that important in a job is responsibility and the opportunity to use initiative and
if they considered hard work to bring success (score 1–5 on a 1–10 scale where 1 = Hard work brings success and 10 =Hard work does not bring success).
Dummy variable =0 if the two above mentioned job qualities were not mentioned and if respondents did not believe that hard work brings success (6–10 on
the hard work scale above). (1) Dummyvariable. 1 = Agree completely or agree somewhat and 0= Neither agree nor disagree, disagree somewhat or disagree
completely with the following statement: “We are more likely to have a healthy economy if the government allows more freedom for individuals to do as
they wish”. (2) Dummy variable. 1= I find that both freedom and equality are important. But if I were to choose one or the other, I would consider personal
freedom more important, that is, everyone can live in freedom and develop without hinderance; 0 = Certainly both freedom and equality are important.
But if I were to choose one or the other, I would consider equality more important, that is, that nobody is underprivileged and that social class differences
are not so strong. (3) Dummy variable. 1= Very important; 0= Not very important or not at all important. Answers to the question of whether protecting
freedom of speech is a national goal. (4) Dummy variable. 1= To respect freedom for the individual; 0 =To maintain order in society. Answers to the question
of what is the most important responsibility for the government.
*Significance level: p<0.1.
** Significance level: p<0.05.
*** Significance level: p<0.01.
these changing cultural strategies actually seem to have raised happiness levels in much of the world. The evidence indicates that
these factors were conducive tohappiness mainly through their common tendency to increase human freedom” (p. 279). Transitions
in preferences also require a generational shift. Alesina and Fuchs-Schündeln (2007) found East-Germans to have different
preferences in relation to the redistribution of income as compared to West-Germans (controlling for socioeconomic factors).
However, they note that this effect is the most evident among the older generations and is expected to disappear within one
or two generations determining a convergence in preferences between East and West Germans. A recent study on Central

Author's personal copy
P. Verme / Journal of Economic Behavior & Organization 71 (2009) 146–161 159
Europe (Varnum, 2008) remarks that: “Since the collapse of communism, Central Europeans have a more internal sense of control
and make more dispositional attributions for others’ behavior” (p. 1). A report by the South African Center for International and
Comparative Politics (CICP, 2007) noted that: “South Africans feel that they have much more control over their lives than they
did in 1990; especially the black population whose lives were controlled by repressive laws under Apartheid. This may account for
the previously noted rise in happiness” (p. 8).
The institutional setting of a nation has an important role in shaping preferences. A country that forms its educational
system around values such as obedience evidently produces pupils who are different from those of a country that encourages
independence and creativity. The Japanese government is struggling to reform its educational system in the direction of
encouraging more creativity and independence of thought as opposed to obedience. On the contrary, Italy now thinks that
undiscipline and the lax educational policies that are the heritage of the 1960s have gone too far and is now trying to
reverse the trend. Governments try actively to ‘form’ citizens with public policies and by doing so they shape personalities.
This paper provided some additional evidence that these policies may well have an impact on the future well-being of
individuals.
Funding
The research is self-funded. An extension of the research has been commissioned and financed by Oxford Analytica. I am
also grateful to the Legatum Institute for an invitation to a symposium in London that allowed me to share and discuss some
of the findings of this work.
Acknowledgments
I am very grateful for detailed comments provided by Andrew Clark. John Helliwell, Avner Offer, Robert Putnam, Jacques
Silber, Guido Tabellini, Ruut Veenhoven and two anonymous referees provided very useful suggestions for improving the
paper. I have also benefitted by a number of comments from participants to seminars and conferences where the paper was
presented including the Department of Economics, University of Turin; IGIER, Bocconi University; the conference on “Policies
for Happiness” held in Siena 14–17 June 2007, the conference on “Happiness in Global Perspective” held in Bangkok 18–19
July 2007 and the Legatum Symposium held in London 20–22 June 2008. All remaining errors are mine.
Appendix A.
See Table A1.
Table A1
Descriptive statistics—full and reduced samples.
Variable Full sample Reduced sample (Table 1) Reduced sample (Table 2B)
Obs Mean Std. Dev. Obs Mean Std. Dev. Obs Mean Std. Dev.
Life satisfaction 224857 6.56 2.50 160405 6.58 2.49 187198 6.51 2.51
freedom and control 212083 6.61 2.45 160405 6.71 2.41 187198 6.62 2.47
Income rank 228825 4.68 2.47 160405 4.67 2.46 187198 4.63 2.45
Unemployed 228825 0.07 0.26 160405 0.08 0.27 187198 0.08 0.27
Female 228825 0.51 0.50 160405 0.51 0.50 187198 0.51 0.50
Age 227545 41.34 16.15 160405 41.28 15.92 187198 41.27 15.98
Edutert 228825 0.15 0.36 160405 0.18 0.38 187198 0.17 0.38
Married 228825 0.64 0.48 160405 0.65 0.48 187198 0.65 0.48
Tax cheat 211751 2.39 2.32 160405 2.40 2.32
Trust in people 228825 0.28 0.45 160405 0.28 0.45
Trust in institutions 222826 2.45 0.59 160405 2.44 0.58
Family importance 206642 0.98 0.13 160405 0.98 0.13
Work importance 205139 0.92 0.27 160405 0.93 0.26
Religion importance 203450 0.65 0.48 160405 0.63 0.48 187198 0.64 0.48
Politics importance 203136 0.39 0.49 160405 0.40 0.49
GDP (000) 214439 12.27 8.96 160405 12.11 9.32
Employment rate 227883 58.95 13.03 160405 59.32 12.46
Important child qualities: obedience 226699 0.36 0.48 187198 0.37 0.48
Important child qualities: independence 226706 0.44 0.50 187198 0.46 0.50

Author's personal copy
160 P. Verme / Journal of Economic Behavior & Organization 71 (2009) 146–161
References
Alesina, A., Di Tella, R., McCulloch, R., 2004. Inequality and happiness: Are Europeans and Americans different? Journal of Public Economics 88, 2009–2042.
Alesina, A., Fuchs-Schündeln, N., 2007. Goodbye Lenin (or not?): the effect of communism on people. American Economic Review 97, 1507–1528.
Bavetta, S., 2004. Measuring freedom of choice: an alternative view of a recent literature. Social Choice and Welfare 22, 29–48.
Bell, D.E., 1982. Regret in decision making under uncertainty. Operations Research 30, 961–981.
Bell, D.E., 1985. Disappointment in decision making under uncertainty. Operations Research 33, 1–27.
Blanchflower, D.G., Oswald, A.J., 1997. A study of labour market and youth unemployment in Eastern Europe. In: Warwick Economic Research Papers 499,
University of Warwick, United Kingdom.
Blanchflower, D.G., Oswald A.J., 2000. Well-being over time in Britain and the USA. Mimeo. Available at: http://www.dartmouth.edu/blnchr/
papers/Wellbeingnew.pdf.
Bowles, S., Gintis, H., Osborne, M., 2001a. Incentive-enhancing preferences:personality, behavior, and earnings. The American Economic Review91, 155–158.
Bowles, S., Gintis, H., Osborne, M., 2001b. The determinants of earnings: a behavioral approach. Journal of Economic Literature 39, 1137–1176.
Brickman, P., Campbell, D.T., 1971. Hedonic relativism and planning the good society. In: Apley, M.H. (Ed.), Adaptation-level Theory: A Symposium. New
York Academic Press, New York, pp. 287–301.
Brickman, P., Coates, D., Janoff-Bulman, R., 1978. Lottery winners and accident victims: Is happiness relative? Journal of Personality and Social Psychology
37, 917–927.
CICP 2007. World values Surveys 1981–2006. Mimeo. Available at: http://www.kas.de/wf/doc/kas12577–544-2-30.pdf.
Clark, A.E., Oswald, A.J., 1994. Unhappiness and unemployment. Economic Journal 104, 648–659.
Diener, E., Diener, M., Diener, C., 1995. Factors predicting the subjective well-being of nations. Journal of Personality and Social Psychology 69, 851–864.
Diener, E., Suh, E.M., Oishi, S., 1997. Recent findings on subjective well-being. Indian Journal of Clinical Psychology 24, 25–41.
Diener, E., Suh, E.M., Lucas, R.E., Smith, H.L., 1999. Subjective well-being: three decades of progress. Psychological Bulletin 125, 276–303.
DiTella, R., MacCulloch, R.J., Oswald, A.J., 2001. Preferences over inflation and unemployment: evidence from surveys on happiness. American Economic
Review 91, 335–341.
Easterlin, R.A., 1974. Does economic growth improve the human lot? In: David, P.A., Reder, M.W. (Eds.), Nations and Households in Economic Growth: Essays
in Honour of Moses Abramovitz. New York Academic Press, New York, pp. 89–125.
Easterlin, R.A., 2001. Income and happiness: towards a unified theory. The Economic Journal 111, 465–484.
Fischer, J., Corcoran, K., 2007. Measures for Clinical Practice and Research. A Sourcebook. Volume 2, Adults. Oxford University Press, New York.
Fordyce, M.A., 1988. A review of research on happiness measures: a sixty second index of happiness and mental health. Social Indicators Research 20,
355–381.
Gilbert, D.T., Pinel, E.C., Wilson, T.D., Blumberg, S.J., Wheatley, T., 1998. Immune neglect: a source of durability bias in affective forecasting. Journal of
Personality and Social Psychology 75, 617–638.
Goldberg, L.R., 1981. Language and individual differences: the search for universals in personality lexicons. In: Wheeler, L. (Ed.), Review of Personality and
Social Psychology 2. Sage, Beverly Hills, CA.
Gravel, N., 1994. Can a ranking of opportunity sets attach an intrinsic importance to freedom of choice? American Economic Review 84, 454–458.
Haworth, J.T., Jarman, M., Lee, S., 1997. Positive psychological states in the daily life of working women. Journal of Applied Social Psychology 27, 345–370.
Heckman, J.J., 2000. Policies to foster human capital. Research in Economics 54 (3), 56.
Helliwell, J.F., 2003. How is life? Combining individual and national variables to explain subjective well-being. Economic Modelling 20, 331–360.
Inglehart, R.F., 1990. Culture Shift in Advanced Industrial Society. Princeton University Press, Princeton, NJ.
Inglehart, R., Foa, R., Peterson, C., Welzel, C., 1981–2007. Development, freedom and rising happiness: a global perspective. Perspectives on Psychological
Science 3, 264–285.
Iyengar, S., Lepper, M., 2000. When choice is demotivating: Can one desire too much of a good thing? Journal of Personality and Social Psychology 79,
995–1006.
Judge, T., Locke, E.A., Durham, C.C., 1997. The dispositional causes of job satisfaction: a core evaluations approach. Research in Organizational Behavior 19,
151–188.
Judge, T., Locke, E.A., Durham, C.C., 1998. Dispositional effects on job and life satisfaction: the role of core evaluations. Journal of Applied Psychology 83,
17–34.
Kahneman, D., 2000. Experienced utility and objective happiness: a moment-based approach. In: Kahneman, D., Tversky, A. (Eds.), Choices, Values and
Frames. Cambridge University Press and The Russell Sage Foundation, Cambridge, pp. 673–692.
Kahneman, D., Wakker, P.P., Sarin, R., 1997. Back to Bentham? An exploration of experienced utility. Quarterly Journal of Economics 112, 375–405.
Langer, E.J., 1983. The Psychology of Control. Sage, Beverley Hills, CA.
Lefcourt, H.M., 1982. Locus of Control: Current Trends in Theory and Research. Wiley, New York.
Lepper, H.S., 1998. Use of other-reports to validate subjective well-being measures. Social Indicators Research 44, 367–379.
Loewenstein, G., Schkade, D., 1999. Wouldn’t it be nice: Predicting future feelings. In: Kahneman, D., Diener,E., Schwarz, D. (Eds.), Well-being: The Foundation
of Hedonic Psychology. Russell Sage, New York, pp. 85–108.
Lykken, D., Tellegen, A., 1996. Happiness is a stochastic phenomenon. Psychological Science 7, 186–189.
Mangahas, M., 1981. Self-rated poverty in the Philippines 1992. International Journal of Public Opinion Research 7, 40–55.
Morton, T.L., 1997. The relationship between parental locus of control and children’s perceptions of control. Journal of Genetic Psychology 158, 216–225.
Moulton, B.R., 1986. Random group effects and the precision of regression estimates. Journal of Econometrics 32, 385–397.
Myers, D.G., Diener, E., 1995. Who is happy? Psychological Science 6, 10–19.
Ravallion, M., Lokshin, M., 2000. Identifying welfare effects from subjective questions. Economica 68, 335–357.
Rotter, J.B., 1954. Social Learning and Clinical Psychology. Prentice-Hall, New York.
Rotter, J.B., 1966. Generalised expectancies of internal versus external control of reinforcements. Psychological Monographs 80, 1–28.
Rotter, J.B., 1990. Internal versus external locus of control of reinforcement: a case history of a variable. American Psychologist 45, 489–493.
Runciman, W. G., 1966. Relative deprivation and social justice. Reports of the Institute of Community Studies, Routledge and Kegan Paul, London, Boston
and Henley.
Sandvik, E., Diener, E., Seidlitz, L., 1993. Subjective well-being: the convergence and stability of self-report and non-self-report measures. Journal of
Personality 61, 317–342.
Saris, W.E., Scherpenzeel, A.C., Veenhoven, R., Bunting, B., 1996. A Comparative Study of Satisfaction with Life in Europe. Eötvös University Press, Budapest.
Sen, A., 1987. Commodities and Capabilities. Oxford University Press, Oxford Indian Paperbacks, Cambridge and New Delhi.
Seneca, L.A., 1996. Sulla Felicita’ (De Vita Beata). Edizioni BUR, Milano.
Sethi-Iyengar, S.S., Jiang, W., Huberman, G., 2004. How much choice is too much?: Contributions to 401 (k) retirement plans. In: Mitchell, O.S., Utkus, S.P.
(Eds.), Pension Design and Structure: New Lessons from Behavioural Finance. Oxford University Press, New York, pp. 83–97.
Simon, H.A., 1955. A behavioral model of rational choice. Quarterly Journal of Economics 69, 99–118.
Skinner, E.A., 1996. A guide to constructs of control. Journal of Personality and Social Psychology 71, 549–570.
Strickland, B.R., 1989. Internal–external control expectancies: from contingencies to creativity. American Psychologists 44, 1–12.
Varnum, M.E.W., 2008. Rapid adaptation to social change in Central Europe: changes in locus of control, attribution, subjective well-being, self-direction,
and trust. Slovak Sociological Review 3, 215–235.
Veenhoven, R., 1993. Happiness in Nations: Subjective Appreciation of Life in 56 Nations, 1946–1992. Erasmus University Press, Rotterdam.

Author's personal copy
P. Verme / Journal of Economic Behavior & Organization 71 (2009) 146–161 161
Veenhoven, R., 1996. Developments in satisfaction research. Social Indicators Research 37, 1–46.
Veenhoven, R., 2000. Freedom and happiness: a comparative study in 46 nations in the early 1990s. In: Diener, E., Suh, E. (Eds.), Culture and Subjective
Well-being. MIT press, Cambridge, pp. 257–288.
Veenhoven, R., 2007. World Database of Happiness. Available at: http://worlddatabaseofhappiness.eur.nl.
Wilson, W., 1967. Correlates of avowed happiness. Psychological Bulletin 67, 294–306.
Winkelmann, L., Winkelmann, R., 1998. Why are the unemployed so unhappy? Evidence from panel data. Economica 65, 1–15.