Subjective well-being, social trust and spatial dependence
Author: Lucas E. van den Akker
1Supervisor: Dr. Viola Angelini
University of Groningen
15-07-2012
Abstract
This paper investigates cross-country differences in the effect of social trust on life satisfaction. A two-step procedure is used. The first step estimates a life satisfaction function for each country present in a pooled cross-sectional sample from the World Values Survey for the years 1990-2007. The second step creates a new database with the coefficients estimated for the effect of social trust on life satisfaction and applies a spatial model (SARAR) together with a number of socio-economic variables to explain the variance. The results imply that the effect of social trust on life satisfaction is negatively affected by economic freedom and income inequality. Public spending on education is positively related. Spatial dependence is found in the error term.
JEL: D12, I31, J10
Keywords: Life satisfaction, social trust, spatial analysis
1. Introduction
Research in the field of well-being has provided us with many insights on its determinants. For example, we now know that marriage (Frey and Stutzer, 2000), income (Blanchflower and Oswald, 2004) and good health (Smith, 2003) have a positive influence on the happiness someone
experiences in his life. However, preferences on the determinants vary between countries and so do their influences on well-being. Income, for instance, has a larger effect on well-being in countries with a lower GDP per capita (Stanca, 2010). Many studies pool data from various countries or only look at a number of countries to draw an analysis. By doing so, they fail to take the cross-country differences in the determinants of life satisfaction into account.
This paper will analyze the cross-country differences for one determinant of life satisfaction in particular: social trust. Social trust is closely linked to life satisfaction. Those who feel that in general they can trust others have much more satisfaction in life (Helliwell and Wang, 2011). Virtually every social setting requires a certain degree of trust, whether it involves doing a transaction, going bungee jumping or having an intimate relationship. Trust is thus a very important factor in people’s lives. Yet levels of trust vary greatly between countries, due to different preferences and socio-economic conditions. To illustrate, Bjørnskov (2005) shows that more income inequality substantially lowers the level of trust in a nation. The causes of the varying trust levels may also be able to explain the cross-country differences in the effect of social trust on life satisfaction.
To investigate this issue, I employ a two-step procedure that has also been used by Stanca (2010). In the first step, a life satisfaction function is estimated per country on a pooled cross-sectional sample from the World Values Survey for the years 1990-2007. In the second step, the coefficients estimated for the effect of social trust on life satisfaction are used to create a new database with 73 countries, representing 80% of the world population. I try to explain the variance in the coefficients on social trust with variables that are known to affect levels of social trust, such as economic freedom, income inequality, public spending on education and religion. Furthermore, I introduce spatial interactions in the dependent variable and in the error term to account for norms and traditions that are hard to observe but do extend across borders. The paper thus makes two contributions to the literature. Firstly, it looks at cross-country differences in a determinant of well-being. Secondly, it models spatial dependences between countries. Both approaches have rarely been used and can provide new insights into the determinants of the effect of social trust on life satisfaction. Policymakers can benefit from these insights to achieve a higher well-being.
The paper proceeds as follows. Section 2 summarizes the literature in the three fields that are important to this study: subjective well-being, social trust and spatial economics. It also discusses the assumptions made on the effects of the explanatory variables. Section 3 presents the methodology behind the two-step procedure. Section 4 describes the datasets and their sources. In section 5, the results of the first and second step are discussed. The section also presents a number of robustness checks. Section 6 provides concluding remarks. In the rest of the paper, the terms happiness, life satisfaction and subjective well-being are used interchangeably, as is common in the literature. The same goes for the expressions social trust and interpersonal trust.
2. Literature review
2.1 Subjective well-being
In microeconomics, utility is an important concept for measuring the well-being people obtain from consuming goods. Increasing one’s utility is the driving force to explaining individual choices in many models. However, it is not directly clear how we can measure utility itself. Early thinkers on the topic, such as Jeremy Bentham (1789), were convinced that utility can be measured as a number. This is what we now call cardinal utility. The difficulty is that a person’s utility is affected not only by his choices between various goods, but also by culture, religion and personality among others, which makes it inherently hard to measure utility.
Economists therefore opted for the use of ordinal utility, in which revealed preferences, the choices actually made, are used to infer utility. To analyze revealed preferences, Von Neumann and
Morgenstern (1944) developed four axioms of rational choice. These axioms are completeness, transitivity, continuity and independence. Any agent who satisfies the axioms has a utility function. Also, by using the axioms it was demonstrated that demand theory can be based on observed choices only, as long as people behave rationally (Becker, 1974).
Recent research has shown that people are in fact not rational. Tversky and Kahneman (1981) show that preferences depend on framing; the way in which decision problems are formulated. The majority of people is risk averse in choices regarding gains, and risk seeking in choices regarding losses. For example, when a decision problem is framed as a choice between a certain gain of €100 and a 0.11 chance on €1000, people prefer the certain gain, even though a 0.11 chance on €1000 has a higher expected value. But, when the problem is to choose between a certain loss of €100 and a 0.1 chance to loose €1000, the large majority will prefer the second option, even though the expected value of both is equal. This certainty effect leads to preferences that are inconsistent with the axioms of rational choice in which choices are consistent.
frequency or probability of an event. This results in overestimation of low-probability events such as nuclear disasters, and underestimation of high-probability events such as dying of lung cancer if you are a regular smoker. Anchoring means that people start thinking from an initial position and then adjust to obtain a final decision. The problem here is that different initial positions yield different final decisions.
The heuristics may also cause markets to work inefficiently. Shiller et al. (1984) presents evidence on the influence of social dynamics on stock prices. Investor sentiments lead to stocks that are
fashionable to buy, even though the underlying fundamentals are unfavorable. This effect creates bubbles in the stock market.
Assuming rationality works well in many economic models, but may also fail when decisions are framed or when heuristics are involved. It is therefore necessary to look for alternatives when assessing utility. One such alternative can be found in the extensive literature on well-being in the field of psychology.
Easterlin (1974) introduced the idea of using subjective happiness as a measure for utility. He investigated whether economic growth improved happiness by linking psychological to economic data. The result was that happiness does not increase when countries experience economic growth. A finding that stands in stark contrast to what one might expect from the microeconomic literature, which associates higher income with a higher level of happiness at the individual level. Later studies on this so-called ‘Easterlin paradox’ have found that this is due to the presence of relative income in the utility function. (Pollak, 1976) People compare their income to a reference group which may consist of family or friends, for example, and derive utility from their relative position in the
reference group. Another explanation states that people adapt their aspirations to a higher income (Easterlin, 2001; Brickman et al., 1978). They experience a temporary rise in happiness following an increase in income, but once they get used to this rise they return to their natural level of happiness. From Easterlin’s 1974 paper onwards, the literature on so-called subjective well-being has expanded greatly and now covers several concepts. According to Diener (1984), these concepts of subjective being can be grouped into three categories, each of which is used in the literature. Firstly, well-being is not seen as a subjective state, but rather as possessing certain qualities that are desirable. This is a normative definition of well-being, in which one should strive for virtues such as success in life. The second category focuses on life satisfaction and lets people themselves decide what determines a good life. A person chooses his own criteria and these may be normative or positive. I will use this definition in the rest of the paper. A third category stresses emotions as the root of well-being, most notably the predominance of positive emotions over negative emotions. This definition comes closest as to how the term happiness is used in everyday life.
require more time to complete. Which scale to use depends on the purpose of the study (Frey and Stutzer, 2002). If the purpose is to do a study on psychological issues, a more detailed approach is necessary. However, economists are more interested in the overall picture for which a
representative and sufficiently large sample is important. This sample can more easily be obtained with a brief single-item question.
The validity of measures of subjective well-being can be evaluated by analyzing its correlations with other factors that are associated with a higher level of life satisfaction or happiness. Kahneman and Krueger (2006) have collected results from several studies and found that high life satisfaction is correlated with a faster recovery after injury or exposure to a virus, a higher smiling frequency, better sleep quality and having an outgoing personality among others. Low life satisfaction and happiness are correlated with chronic pain, unemployment and recent negative changes of circumstances. Moreover, Oswald and Wu (2010) compare an objective quality-of-life ranking, including sunshine hours among others, with a life satisfaction regression with controls and find a correlation of 0.598. These findings suggest that data on subjective well-being contains valid information on the quality of people’s lives.
Many studies have been performed on the determinants of subjective well-being. It goes beyond the scope of this paper to discuss each determinant in detail, but table 1 provides a summary of the relevent ones. Dolan et al. (2008), Clark et al. (2008) and MacKerron (2011) provide detailed reviews of the literature.
[Table 1 about here]
An effort to formalize measures of subjective well-being has been made by Blanchflower and Oswald (2004). They propose the following well-being function:
R = H[U(z, t)] + e
Where R is a self-reported number or level on a well-being scale (for example, ‘not very happy’). U is the true level of well-being. Socio-demographic factors, such as income, employment and marital status are captured by z, t is the time period, and e is the error term. True well-being U and reported well-being R are linked by the function H, which is a continuous, non-differentiable function.
(2011) investigate these different types of scoring by using anchoring vignettes. They observe that divergence in life satisfaction ratings between countries can partly be explained by different interpretations of the response scale. Oishi (2010) identifies more biases. He discerns number-use, item functioning, self-representation, memory, positivity and reference group effect biases, but concludes that these seem to play a relatively small role in explaining national differences in mean levels of life satisfaction.
Studies that compare well-being across nations have found many differences. Following Maslow (1943), the higher a nation is on his pyramid of needs, the higher the average well-being of that nation. Veenhoven (1991) tested whether people who live in wealthier nations have higher life satisfaction and found a positive correlation. Income is linked to basic needs (e.g. food), but probably also to higher needs such as safety. As for higher needs, Inglehart et al. (2008) show that the extent to which a society allows free choice has a major impact on differences in happiness. Futhermore, self-esteem is more strongly associated with life satisfaction in more individualistic nations, such as the United States, than in more collectivist nations, such as India (Oishi et al., 1999). Some studies compare cross-country differences in the determinants of well-being. Helliwell et al. (2009) estimates life satisfaction functions for 105 countries and finds for example that marriage is valued less in terms of life satisfaction when an inadequate amount of food is available. Apparently the gains from marriage come from being able to provide for basic needs better. In addition, Stanca (2010) shows that being unemployed is worse for life satisfaction in countries with a higher GDP per capita or a higher unemployment rate. Socio-economic differences thus seem to explain cross-country variance in the determinants of well-being.
With the growing attention for the field of subjective well-being, a general sense has developed that our current measures of human welfare are inadequate. The most widely used measure of human welfare is an economic one: GDP. GDP mainly measures market production by adding up the monetary value of goods and services, making it easy to compare quantities of a very different nature. However, GDP ignores the value for human welfare of for instance leisure and longer life spans. Some scholars such as Diener (2000) and Stiglitz et al. (2009) have argued that governments should instead use indicators of subjective well-being to assess the effects of public policy. They recommend using standards such as the Index of Economic Well-Being (Osberg and Sharpe, 2002) and the Genuine Progress Index (Cobb et al., 1995). These focus more on the quantitative
measurement of well-being and sustainability.
2.2 Social Trust
An important determinant of well-being discussed in the previous section is social trust. Social trust can be defined as the trust that one has in other people. It is usually measured in surveys by asking the question: “In general, do you think that most people can be trusted?”. Many studies have found that the results of this measure correlate highly with other ways of asking people about the
Trust is essential in the cooperation between people and some scholars even argue that a society cannot function without a certain level of trust (Putnam, 1993; Fukuyama, 1995). Few market transactions would be made if there did not exist some trust between the involved parties (Arrow, 1972). For example, think of the simple act of ordering a meal in a restaurant. As a customer, you trust the chef not to have tempered with the meal such that it is unsafe to eat. As a waiter, you trust the customer to pay the bill when the time comes. This sense of trust is not explicitly articulated and no contracts are signed before entering the restaurant; it is simply an expectation of what will happen based on common social and ethical norms and traditions. Of course, legal institutions are in place to implement sanctions when one of the parties does not adhere to this implicit agreement. But these institutions would not function properly without the same social and ethical norms and traditions that foster trust (Fukuyama, 1995). Similarly, contracts let people do transactions without a basis of trust, yet these can never be complete and are entered into much faster when trust exists. Therefore, trust is essential to transactions.
Research on social trust is part of the larger literature on social capital. Coleman (1990) defines social capital as “the ability of people to work together for common purposes in groups and organizations”. It may also refer to trust, norms, and networks that can improve the efficiency of society by
facilitating cooperation. Like conventional capital, it is productive, can be used as a kind of collateral and those who have it tend to accumulate more of it (Putnam, 1993).
For many years, economists have relied on the concepts of physical and human capital to explain economic phenomena, but recently models have been extended with social capital. Following Adam Smith (1776), people’s economic decision-making is highly complex and grounded in broader social habits and mores. He already recognized that we cannot fully describe human behavior without taking social capital into account. Efforts to do this have been made by Glaeser et al. (2002), who incorporate the decision to accumulate social capital into an optimal investment model. They find that differences between social capital and other forms of capital stem from the interpersonal externalities it generates, through networks and status among others. Social capital seems to primarily be a by-product of other social activities.
The following paragraphs describe factors that can explain the level of social trust in a nation. These factors will also be used to explain cross-country differences in the effect of social trust on life satisfaction.
Income inequality
Education
Education may influence trust through a number of channels (Knack and Keefer, 1997). Firstly, education makes people less ignorant. To the extent that ignorance brings forth distrust, this will be reduced. Secondly, students can be taught to behave more cooperatively. The more people
cooperate, the more likely they are to trust each other. Thirdly, educated people are better able to perceive risks and potentials, thereby reducing uncertainty in dealing with others. Studies analyzing these effects have generally found a positive correlation (Knack and Keefer, 1997; Zak and Knack, 2001), although Bjørnskov (2005) presents an outcome in which trust and education are unrelated. Furthermore, theoretically education is positively related to well-being, but here results are mixed as well. Again, the effect of education on well-being works through different channels. For example, those with higher levels of education also need higher levels of income to attain a certain level of well-being (Hagenaars, 1986), which may be hard to realize. Therefore, education might also decrease well-being in certain cases. I analyze the effect of education on well-being through social trust and expect this effect to be positive.
Religion
Inglehart (1999) states that the level of trust within a nation is determined by historical economic and religious factors. He shows that historically Protestant nations exhibit far more interpersonal trust than historically Catholic or Islamic nations. An explanation is that Protestant churches are usually small-scale organizations that operate relatively decentralized, whereas Catholic churches are part of a bishopric that is controlled by the Vatican. Since horizontal, locally-controlled organizations stimulate interpersonal trust and distant hierarchical organizations tend to be detrimental to it (La Porta et al., 1997), differences in interpersonal trust can be explained by religion. Horizontal networks are also referred to as the driving force for developing trust by Putnam (1993). However, Knack and Keefer (1997) demonstrate that horizontal networks - measured as membership in groups - do not have an effect on trust. Perhaps the effect of religion does not work through the network channel, but is simply a proxy for some other historical factor such as social mobility. The evidence on the influence of religion on well-being is contradictory. For instance, Smith (2003) finds positive effects for the USA and Germany, while in Belarus and Russia religion negatively affects well-being. Studies on religion and well-being have not looked at separate types of religion, but simply at religion as a pooled variable. However, the type of religion does seem to have an impact on the level of interpersonal trust. Since interpersonal trust has an effect on well-being, it may be that Protestants and Catholics compensate for each other in the effect of religion on well-being, but that in fact Protestants are much happier than Catholics. I will look into this issue through the social trust channel and expect that Protestantism increases the effect of social trust on life satisfaction, while Catholicism and Islam decrease it.
Economic freedom
freedom is also positively related to life satisfaction. Thus, I expect a positive effect of economic freedom on well-being through social trust.
2.3 Spatial dependence
As Inglehart (1999) noted, interpersonal trust reflects a people’s history. A history that is connected by the interaction people have in economic, cultural, political, religious and other settings. The idea that economic agents interact has led to the development of many models that are able to explain social norms, peer effects, spatial spillovers and network effects among others (Anselin, 2005). Since interpersonal trust is to a great extent about cooperation, models that contain interaction effects are also useful in explaining trust.
One way of modeling interaction effects is found in the spatial economics literature, in which spatial dependences arise between cross-sectional units, for example between households or states.
Concerning trust, this means that people who live in different geographical regions interact with each other to form similar social norms and traditions. These interactions may have happened in the past or in the present, and may have occurred by trading, by sharing a language or religion or by other factors. A number of these interactions is easy to observe and measure, while others are not. To avoid an omitted variables bias in these cases, the spatial dependence effect can be used as a general proxy for interaction between different peoples.
Studies in the field of well-being that take the spatial dimension into account usually do so by generating location dummy variables (see, for example Di Tella et al., 2003 and Clark et al., 2005). Others, like Aslam and Corrado (2007) take a multilevel approach to the issue. While Brereton et al. (2008) link location data of the respondent to climate, environment and urban variables from the same area. All these studies are useful for analyzing the direct influences that a location has on a person’s well-being, but they fail to allow for the spatial externalities that people exert on others. Stanca (2010) does examine these externalities. He looks at the effects of income and unemployment on well-being across countries and finds a positive spatial dependence. Since spatial dependence plays a role in both trust and well-being, I expect that there exists spatial dependence in the relationship between social trust and subjective well-being.
3. Methodology
To assess the effects noted above, I use a two-step procedure as described in Stanca (2010). Firstly, I estimate a subjective well-being function for each country in which life satisfaction is regressed on several variables that are known to affect well-being at the individual level. Secondly, I analyze the relationship between economic, social and cultural factors and the effect of social trust on well-being at the country level. By doing so, I explain cross-country differences in the amount of life satisfaction people attach to trusting others.
well-being, most studies assume that it runs from trust to well-being, although they recognize that the effect is probably simultaneous (Helliwell, 2003; Hudson, 2006). The exact direction of the effect is not the focus of this study. The results presented here are merely an indication of the correlations between economic, social and cultural factors on the one hand, and the effect of trust on subjective well-being on the other.
Let us denote with the well-being of individual i in country j (j = 1, ..., 73) and assume that it depends in a linear fashion on the main explanatory variable social trust ( ) and a vector of other personal characteristics of the respondent ( ).
(1) Where is the error term, which is normally distributed and containing unobserved influences on well-being. The personal characteristics include economic, demographic and social conditions, such as income, gender and relational status, which will be described in the next section. When several survey waves have been carried out in a country, a year dummy is included to account for year-specific circumstances.
To estimate this type of equation, psychologists usually assume cardinality and perform an OLS regression, whereas economists assume ordinality and use ordered latent response models, such as ordered probit. Ferrer-i-Carbonell and Frijters (2004) compare these two assumptions and find that it makes hardly any difference for the results. I use OLS to estimate equation 1 at the country-level, since OLS models are more straightforward to interpret than ordered probit models.
In the second step, I assume that the country-level effect of social trust on well-being found in equation 1 can be explained by differences in macroeconomic and social conditions and by geographic location. I use the so-called spatial-autoregressive model with spatial-autoregressive
disturbances or SARAR model, described by Drukker et al., (2011a) to analyze this topic. Their
procedure estimates the parameters of the following cross-sectional model:
(2)
(3)
Where is a vector of observations on the dependent variable social trust. and are spatially-weighted contiguity matrices that are assumed to be equal. Contiguous units are known as neighbors and are assigned a weight of 1. is a parameter that measures the spatial lag, is a parameter that measures the spatial error. is a matrix of exogenous variables that includes variables on economic freedom and education levels among others, is the corresponding parameter vector. is a spatially autocorrelated error term, is a normally distributed error term, containing all other unobserved influences.
generalized spatial two-stage least squares estimator is used. This type of estimator can produce results for models that contain both a spatial lag as well as a spatial error, and is computationally simple (Kelejian and Prucha, 1998).
To assess the spatial dependencies, I estimate four alternative specifications of the SARAR model. The first model is a benchmark model in which and . This reduces the SARAR model to an OLS model:
(4)
The second specification is a spatial lag (SAR) model, under the assumption that . This means that the dependent variable is affected by the dependent variables of countries that lie in the proximity.
(5)
The third specification is a spatial error model, under the assumption that . Hereby I test whether there are any spatial disturbances in the error term.
(6)
(7)
The fourth specification is a pure SARAR model as described in equation 2 and 3, including both the spatial lag and spatial error term.
4. Data description
For the analyses in the first step, I use data from the World Values Survey (WVS). The WVS has conducted five waves of surveys from 1981 to 2007, of which four waves contain variables that can be used in this study. Only a few countries appear in all four waves, making the data unsuitable for a panel study. I thus created a repeated cross-sectional sample that covers 73 countries and represents 80% of the world population from 1990 to 2007. On average, a national sample contains about 2000 respondents who are randomly selected from the population. Generally, this was done by using the following procedure: interviewers listed all urban and rural areas within each region and then randomly selected suburbs to visit. In each suburb, 6 to 8 interviews were done using a random walk procedure to select the starting point. Every fourth household was then visited. Within a household, all males or all females older than 16 years were listed on a grid. Together, the questionnaire number and grid decided which household member was to be interviewed. Male and female interviews in the selected suburb were alternated to obtain an equal share of interviews for each gender.
analysis and therefore it was not used in this study. Table 2 reports descriptive statistics for all variables employed in the first step. Since most variables are binary, I calculate country means for each of them and present the minimum and maximum values of the country means here.
Life satisfaction is taken as the main measure for subjective well-being. Following standard practice, the WVS measures life satisfaction in all five waves of the survey by asking:
All things considered, how satisfied are you with your life as a whole these days? Please use this card to help with your answer [1 dissatisfied (…) 10 satisfied]
This single-item question has been shown to be a valid measure of life satisfaction in a broad number of domains (Andrews and Withey, 1976; Kahneman and Krueger, 2006; Oswald and Wu, 2010). Also, people are assumed to be the best judge of their own well-being (Frey and Stutzer, 2002). The other measure of subjective well-being is happiness, which is based on a 1-4 scale [1 not at all happy (…) 4 very happy]. To ease the interpretation of the results section, life satisfaction and happiness are multiplied by 10.
The focus of this study lies on the explanatory variable social trust. The WVS measures this concept with the following question:
Generally speaking, would you say that most people can be trusted or that you can’t be too careful in dealing with people?
[0 can’t be too careful, 1 most people can be trusted]
To test the validity of this question, Knack (2001) performs a field experiment in which the return frequency of deliberately dropped wallets is measured in several cities in 14 countries in Western Europe and in the United States. He finds that at the national level, the return frequency has a correlation of 0.65 with the responses to the social trust question in the WVS. In a laboratory setting, Naef and Schupp (2009) show that the outcome of a trust game is significantly correlated with social trust, but not with trust in institutions and trust in known others. The trust question used in this study can therefore be seen as a valid measure of social trust.
The other explanatory variables are also taken from the WVS. Income is measured by a 10-step scale per country, so that it is expressed in relative terms and comparable across nations. A respondent’s employment status is described by a dummy variable from one of the following categories: full-time, part-time, self-employed, retired, housewife, students, unemployed and other employment.
being divorced or having separated. Therefore, I excluded these categories from the regression for these countries.
A number of beliefs are taken into account as well. Firstly, a person’s view on income inequality is transformed into a dummy variable. On a 1-10 scale where 1 means that people think ‘incomes should be more equal’ and 10 means that ‘we need larger income differences’, I create a dummy variable for more equality for values 1-4, a dummy variable for a neutral position for values 5-7 and a dummy for larger inequality for values 8-10. A similar approach is taken for how much freedom of choice and control an individual experiences in his life. A ‘low freedom’ dummy is created for values 1-4, a ‘neutral freedom’ dummy for values 5-6, a ‘more freedom’ dummy for values 7-8 and those who experience a great deal of freedom are represented by the ‘most freedom’ dummy for values 9-10. A respondent’s subjective state of health is included in a 1-4 scale [1 very good (…) 4 poor], where ‘very poor’ has been added to ‘poor’ because of its low number of observations. People’s proudness of their nationality is identified by the variable ‘chauvinism’ [1 not very proud (…) 3 very proud]. The category ‘not at all proud’ has been merged with the category ‘not very proud’, because of an insufficient number of observations. A respondent’s self-positioning on a 1-10 political scale is used to create two dummy variables. People who score themselves between 1 and 4 are considered to be left-wing, while those who score themselves between 7 and 10 are on the right of the political spectrum. For China, Iraq, Malaysia and Saudi Arabia, this variable is not reported and therefore excluded from the regression. The importance of God in one’s life is measured by the variable ‘god’ [1 not at all important (…) 10 very important]. In Pakistan, all respondents report God to be very important in their lives. Other dummy variables that are included are sex, age, age squared and year of birth (deducted by 1900). Since the various waves took several years to complete, I use year dummies to identify year-specific circumstances during the interview.
[Table 2 about here]
The aim of this study is to investigate the determinants of the relation between social trust and life satisfaction. Figure 1 shows a scatterplot for the average values of life satisfaction and the average values of social trust per country. A clear positive trend can be distinguished, a higher average value of social trust is associated with a higher average value of life satisfaction.
[Figure 1 about here]
another denomination. The spatial matrix was created using a shapefile from
http://aprsworld.net/gisdata/world/ that includes an outline of the world. Drukker et al. (2011b) describe the spmat-command in STATA that was subsequently applied to the shapefile.
All variables are expressed as long-run averages over the sample period 1990-2008, except for the homicide rate (2003-2008) and the data on religion (1999). The long-run averages are assumed to be representative of the fundamental socio-economic conditions of the nations under analysis. Table 3 provides descriptive statistics for all variables used in the second step, as well as for some robustness check variables that will be explained in the results section.
[Table 3 about here]
5. Results
5.1 Estimation results
In this section I present the results from the estimations. Firstly, I analyze the results of step 1 of the procedure and check whether these results are robust to different specifications and estimation techniques. Secondly, I describe the coefficients of the effect of social trust on life satisfaction and identify spatial clusters. I then proceed by explaining cross-country differences in the coefficients. Lastly, I perform a number of robustness checks on the results of the second step.
In step 1 the model is estimated per country, making it cumbersome to exhibit all the outcomes individually. Instead, table 4 presents the results as if step 1 was performed on the entire sample. This is done to get a feeling for the impact of the various variables on subjective well-being. To check the robustness of the model used in step 1, I perform the same estimation on the dependent variable ‘Happiness’. Life satisfaction and happiness are considered similar concepts in the field of subjective well-being. In this sample, they have a correlation of 0.46. Furthermore, I look at the effects of excluding political orientation from the regression, since many observations do not contain this variable.
The results presented in table 4 largely confirm the results generally found in the literature. All four specifications show qualitatively similar effects, making them robust to the use of different measures of subjective well-being and to the exclusion of the variables on political orientation. Here, I will discuss the results of column 2 only, as this is the estimation from which the social trust coefficients for the second step will be taken. Again, the table only provides an indication of the correlations between the dependent and explanatory variables, causality is not implied.
[Table 4 about here]
Income is positively associated with both measures of subjective well-being. Moving up one step in the income distribution increases life satisfaction by 1.31 points on a 10-100 scale, which is a
to be happier than those who want more equality. A finding that is confirmed by the ideologically related variable on political orientation. Respondents who report to be on the right of the political spectrum are more satisfied with their lives than respondents with a left-wing orientation. Excluding political orientation from the regression in columns 1 and 3 only slightly changes the coefficients of other variables and is thus of no major concern for this study. Concerning marital status, living as if you are married is best for your well-being, while being a widow is worst. Also, males are slightly less happy than females. People who find God important in their lives are more satisfied than those who do not. Social trust is positively related to well-being. Thinking you are in good health greatly
increases your well-being. Another important factor for well-being is a respondent’s feeling of how much freedom of choice and control he experiences in his life. Experiencing more freedom improves well-being by many points in all specifications. Being prouder of one’s country increases life
satisfaction by 2.99 per step on a 1-3 scale of chauvinism. Age is negatively associated with well-being, although the relation is U-shaped, with an implied low-point at age 70.
Table 5 demonstrates the results of four ordered probit regressions performed on the same variables as in table 4, which was estimated using OLS. The coefficients of both table 5 and table 6 are
qualitatively similar and thus robust to the estimation technique, as predicted by Ferrer-i-Carbonell and Frijters (2004).
[Table 5 about here]
Table 6 presents the effects of social trust on life satisfaction per country. The coefficients vary from a minimum of -6.48 for Trinidad and Tobago to a maximum of 10.37 for Ghana. The aim of this study is to explain these large differences, although no logical explanation can be given for the alternating signs of the coefficients. All five countries with the highest positive effect of social trust on life satisfaction lie in Africa, indicating that African culture probably attaches much satisfaction to being able to trust others. Coefficients that take a value between -1 and 1 are not statistically different from 0. The relatively low t-values are caused by including many explanatory variables in the first step. An experiment in which many of these explanatory variables were left out caused the t-values of social trust to rise significantly without changing the qualitative effects. Coefficients obtained in the experiment and those used here have a correlation of 0.78.
[Table 6 about here]
The coefficients listed in table 6 are plotted on a map of the world in figure 2. As could already be seen in the ranking in table 6, a spatial cluster of countries in which the effect of social trust on life satisfaction is large can be distinguished in North-West Africa. Similarly, a low effect cluster is identified in Middle America. Central Europe forms another cluster of which the effects are in the medium range of the ranking. These spatial patterns are exploited in step 2 of this study.
[Figure 2 about here]
[Table 7 about here]
Economic freedom has a negative coefficient and is strongly significant. People in countries with a larger economic freedom attach less life satisfaction to being able to trust others. At first sight, this result seems counter-intuitive since economic freedom enhances both social trust (Berggren and Jordahl, 2006) and life satisfaction (Lelkes, 2006). However, having more economic freedom means that institutions function better, so that transactions require less social trust. This decreased
necessity also leads to a lower valuation of social trust in terms of life satisfaction. A line of reasoning that is supported by Knack and Keefer (1997), who have found that social trust seems to be more important in facilitating economic activity when good institutions are unavailable.
The proxy used for social distance, the Gini index, is negatively associated with the effect of social trust on life satisfaction and statistically significant at the 5% level. More income inequality results in lower level of social trust (Bjørnskov, 2005). Not being able to trust others generally lowers well-being and thus income inequality can partly explain cross-country differences in life satisfaction through the social trust channel.
A higher public spending on education as a percentage of government expenditure is positively related to the effect of social trust on life satisfaction. By spending a higher amount of their budget on education, countries can increase social trust (Knack and Keefer, 1997). This on its turn raises well-being, as could also be seen in table 4.
None of the variables on religion are significant. This probably has to do with the way the religious variables are measured. For example, La Porta et al. (1999) report that in the Netherlands, 42.4% of the population is Protestant, 42.6% Roman Catholic, 1% Muslim and 14% consider themselves to have another denomination. Meanwhile Statistics Netherlands (CBS) reports that in 1999 only 21% of the population considers themselves Protestant, 31% Roman Catholic, 8% of other denominations and 41% has no denomination. This divergence between sources can easily cause results to be insignificant. A test was performed in which countries are assigned a dummy variable for the proportionally largest religion in their country. These dummies are not significant either and do not alter the predictions for the other variables by much. Another reason for the non-significance might be that religions other than Protestantism, Roman Catholicism or Islam are pooled under the heading ‘other denomination’. Asian countries included in the sample have large shares of the population who adhere to Buddhism or Hinduism and these people may have very different norms and
traditions than for example atheists who are put in the same category. Furthermore, previous studies on the effects of religion on well-being have found mixed results. This study tried to resolve the issue by distinguishing between different types of religions and by analyzing their effects through the social trust channel. Unfortunately, this approach also fails to decide on the matter.
religion are excluded. The spatial matrix may thus capture deep historical factors. These may be the result of people moving between countries and influencing the culture of the original population. The spatial lag term is not significant in any specification, meaning that the effect of social trust on life satisfaction in one country does not depend on the effects in neighboring countries. Using the spatial matrix demonstrates that spatial dependencies exist across countries, although these do not
qualitatively alter the coefficients of the other explanatory variables.
5.2 Robustness checks
Uncertainty of coefficients
To check whether the outcomes of the second step are not the result of a series of coincidences, I carry out robustness checks. A number of coefficients of the effect of social trust on life satisfaction, estimated in the first step and reported in table 6, are not statistically different from 0. Therefore, I perform estimations using the upper and lower bound of the 95% confidence interval of the effect of social trust on life satisfaction as dependent variables. By doing so, I take into account the
uncertainty that exists around certain coefficients.
Table 8 presents the results of the robustness check using the coefficients of the lower bound of the 95% confidence interval. Economic freedom has a negative effect on the social trust sensitivities in all specifications, but is only statistically significant in the spatial lag model (column 2) and the SARAR model (column 4) which includes both a spatial lag and spatial error. The negative and significant impact of the Gini index is a robust result in all estimations. Public spending on education has the right, positive sign, but fails to become significant. Just as in the original estimations in table 7, the proportional religious variables remain non-significant. Interestingly, spatial dependence can now also be found in the dependent variables of neighboring countries, as well as in the spatial error term.
[Table 8 about here]
Table 9 analyzes the upper bound of the 95% confidence interval of the effect of social trust on life satisfaction. All relevant variables show a qualitatively similar result in comparison with table 7 and are statistically significant. Overall, one can say that the results presented in table 7 are robust to the non-significance of some coefficients of the effect of social trust on life satisfaction.
[Table 9 about here]
Other explanatory variables
Next to the variables used in the previous section, other variables may be relevant to explain the cross-country differences in the effect of social trust on life satisfaction. Here, I will test whether including three other variables affects the results presented in table 7.
of the state by elites and private interest’. The higher the value of the indicator, the less corruption a nation experiences. Secondly, as Putnam (1993) pointed out, regions with higher levels of trust also have more economic development. Economic development may thus be able to help explain cross-country differences in the effect of social trust on life satisfaction. GDP per capita is taken from the World Development Indicators database of the World Bank (2012) and measured in 1000’s of constant 2005 US dollars. Lastly, Lederman et al. (2002) demonstrate a negative link between violent crime and the level of trust in a community. High levels of crime discourage people to go to meeting places such as town squares; it prevents especially women from going to work and expanding their social network and it increases the number of students that drop out of (night) school. All these factors negatively affect social trust and may also be able to explain differences in the effects of social trust on life satisfaction. As in Lederman et al. (2002), I use the homicide rate as a measure of violent crime. Internationally comparable numbers on the homicide rate are collected by the United Nations Office on Drugs and Crime (2012) and are measured as the number of intentional homicides per 100,000 inhabitants. Descriptive statistics for these variables are shown in table 3.
A potential problem with the type of variables I add here lies in multicollinearity. When variables are highly correlated, individual predictors in a model may not give valid results. Multicollinearity does not reduce the predictive power or reliability of the model as a whole. Table 10 presents a
correlation matrix. The table shows that economic freedom is correlated with control of corruption and GDP per capita and that these last two variables are also strongly interconnected. Furthermore, a high homicide rate is correlated with more income inequality. Various specifications will therefore have to be used to test whether the variables are able to explain differences in the effect of social trust on life satisfaction.
[Table 10 about here]
Focusing on the homicide rate, one might expect that in countries with a high crime rate, being able to trust others is more important and thus associated with a higher satisfaction with life. The opposite is true. An explanation can be that the result we see here is an indirect effect of crime on life satisfaction through the social trust channel. We know that fear of crime reduces life satisfaction (Adams and Serpe, 2002), but this effect probably works through several channels, the social trust channel being one of them. As for the spatial interactions, spatial lags are not significant in any specification, while a positive spatial error effect can be found in columns 2 and 4, both of which do not include the homicide rate and control of corruption. Perhaps these two variables are spatially correlated and were therefore partly captured by the spatial error term. More research will have to be done on this issue to draw conclusions.
This paper makes no claim that all relevant variables explaining cross-country differences in the effect of social trust on life satisfaction were included in the analyses. Nevertheless, the results found in table 7 are robust to the inclusion of a number of other variables.
[Table 11 about here]
6. Concluding remarks
In this paper, I have investigated cross-country differences in the effect of social trust on life satisfaction. A number of studies have shown that social trust has a positive impact on life
satisfaction (Bjørnskov, 2003; Helliwell and Putnam, 2004), but none have analyzed the determinants of this impact.
I use a two-step procedure to analyze these determinants, as described by Stanca (2010). Firstly, I estimate a life satisfaction function for each country present in a pooled cross-sectional sample from the World Values Survey for the years 1990-2007. Secondly, I use the coefficients estimated for the effect of social trust on life satisfaction to create a new database with 73 countries, representing 80% of the world population. These coefficients are used as dependent variables in a
spatial-autoregressive model with spatial-spatial-autoregressive disturbances (SARAR), so that spatial interactions in the dependent variable and the error term can be taken into account. The spatial interactions are the result of people moving and traveling across borders and influencing norms and traditions in
neighboring countries. Furthermore, variables on economic freedom, income inequality, public spending on education and religion are included as explanatory variables. To my knowledge, this is the first study that examines the cross-country differences in the amount of satisfaction people attach to trusting others. Another contribution is the inclusion of spatial dependencies, which are rarely used in this type of research.
trust more. Contrary to the finding of Inglehart (1999) no significant effect can be found for the impact of any type of religious variable on social trust. The discrepancy with respect to my results may be explained by the use of different sources for religious data. Moreover, the literature on religion generally presents mixed results, making this study no exception. As expected, spatial dependencies between countries exist and can be found in the error term, indicating that
unobserved influences are spatially correlated and that cross-border cultural influences are present. This logical result means that the spatial dimension should be taken into consideration in well-being research. Overall, these results are robust to the use of different specifications, explanatory variables and measures of well-being, as well as to the uncertainty that exists around certain coefficients of the effect of social trust on life satisfaction.
Next to the scientific significance of this study, it also holds some economic significance for
policymakers. An increase in economic freedom by itself leads to an increase in life satisfaction, but this effect is moderated by its negative impact on the effect of social trust on life satisfaction. Policymakers should be aware that an increase in economic freedom does not automatically lead to an increase in well-being. Furthermore, increased public spending on education can raise life satisfaction through the social trust channel, although life satisfaction may be decreased through other channels such as the higher income people with higher education desire and which may be hard to attain (Hagenaars, 1986). Income inequality reduces well-being and this effect partly runs through social trust. Government policy aimed at reducing income inequality through redistributive policies can kill two birds with one stone. Not only does it decrease social distance, it also increases social trust and life satisfaction.
Future research should exploit the growing number of longitudinal surveys in the field of well-being. Panel studies can filter out individual-specific, time-invariant effects in the first step of the analysis and therefore capture the complexity of human behavior better than a pooled cross-sectional study such as this one. Also, the coefficients obtained in the first step are available for more years and thus allow for a time-series approach in the second step of the analysis. The second step could also be expanded by including more determinants of social trust. Furthermore, researchers could use the two-step spatial procedure employed here to analyze cross-country differences in other
References
Adams, Richard E. and Richard T. Serpe. (2002) “Social Integration, Fear of Crime, and Life Satisfaction.” Sociological Perspectives, 43(4), pp. 605-629.
Alesina, Alberto, Rafael di Tella and Robert MacCulloch. (2004) “Inequality and Happiness: Are Europeans and Americans Different?” Journal of Public Economics, 88, pp. 2009-2042.
Andrews, Frank M. and Stephen B. Withey. (1976) Social Indicators of Well-Being: American
Perceptions of Life Quality. New York: Plenum.
Angelini, Viola, Danilo Cavapozzi, Luca Corazzini and Omar Paccagnella. (2011) Do Danes and Italians rate life satisfaction in the same way? Using vignettes to correct for individual-specific scale biases. HEDG Working Paper 11/20, University of York.
Anselin, Luc. (1988) Spatial Econometrics: Methods and Models. Boston: Kluwer Academic Publishers. Anselin, Luc. (2005) “Spatial Econometrics.” Journal of Geographical Systems, 4(4), pp. 405-421. Arrow, Kenneth. (1972) “Gifts and Exchanges,” Philosophy and Public Affairs, 1(4), pp. 343–362. Aslam, Aqib and Luisa Corrado. (2007) “No Man is an Island: the Inter-Personal Determinants of Regional Well-Being in Europe.” Cambridge Working Papers in Economics no. 0717.
Becker, Gary S. (1974) “A Theory of Social Interactions.” Journal of Political Economics, 82(6), pp. 1063-93.
Bentham, Jeremy. (1789/1907) An introduction to the Principles of Morals and Legislation. Library of Economics and Liberty. Retrieved May 9, 2012 from:
http://www.econlib.org/library/Bentham/bnthPML18.html
Berggren, Niclas and Henrik Jordahl. (2006) “Free to Trust: Economic Freedom and Social Capital.”
Kyklos, 59(2), pp. 141-169.
Bjørnskov, C. (2003) “The Happy Few: Cross-Country Evidence on Social Capital and Life Satisfaction.”
Kyklos, 56(1), pp. 3-16.
Bjørnskov, C. (2005) “The determinants of trust.” Ratio Institute Working Paper no. 86.
Bjørnskov, Christian, Axel Dreher and Justina A.V. Fischer. (2008) “Cross-Country Determinants of Life Satisfaction: Exploring Different Determinants Across Groups in Society.” Social Choice and Welfare, 30, pp. 119-173.
Blanchflower, David G. and Andrew J. Oswald. (2004) “Well-being Over Time In Britain and the USA.”
Blanchflower, David G. and Andrew J. Oswald. (2008) “Is Well-being U-shaped Over the Life Cycle?”
Social Science and Medicine, 66, pp. 1733-1749.
Brereton, Finbarr, J. Peter Clinch and Susana Ferreira. (2008) “Happiness, Geography and the Environment.” Ecological Economics, 65, pp. 386-396.
Brickman, Philip, Dan Coates and Ronnie Janoff-Bulman. (1978) “Lottery Winners and Accident Victims: Is Happiness Relative?” Journal of Personality and Social Psychology, 36(8), pp. 917-927. Clark, Andrew E. (2003) “Unemployment as a Social Norm: Psychological Evidence from Panel Data.”
Journal of Labor Economics, 21(2), pp. 323-351.
Clark, Andrew E. Fabrice Etilé, Fabien Postel-Vinay, Claudia Senik and Karine Van der Straeten. (2005) “Heterogeneity in Reported Well-Being: Evidence from Twelve European Countries.” The Economic
Journal, 115, pp. 118-132.
Clark, Andrew E., Paul Frijters and Michael A. Shields. (2008) “Relative Income, Happiness, and Utility: An Explanation for the Easterlin Paradox and Other Puzzles.” Journal of Economic Literature, 46(1), pp. 95-144.
Cobb,Clifford, Ted Halstead and Jonathan Rowe. (1995) “The Genuine Progress Indicator: Summary of Data and Methodology.” San Francisco, CA: Redefining Progress.
Coleman, James S. (1990) Foundations of Social Theory. Cambridge: Harvard University Press. Diener, Ed. (1984) “Subjective Well-Being.” Psychological Bulletin, 95, pp. 542-575.
Diener, Ed, Robert A. Emmons, Randy J. Larsen and Sharon Griffin. (1985) “The Satisfaction With Life Scale.” Journal of Personality Assessment, 49(1), pp. 71-75.
Diener, Ed, Suh M. Eunkook, Heidi Smith and Liang Shao. (1995) “National Differences in Reported Subjective Well-Being: Why Do They Occur?” Social Indicators Research, 34(1), pp. 7-32.
Diener, Ed, Suh M. Eunkook, Heidi Smith and Liang Shao. (1999) “Subjective Well-Being: Three Decades of Progress.” Psychological Bulletin, 125(2), pp. 276-302.
Diener, Ed. (2000) “Subjective Well-Being: The Science of Happiness and the Proposal for a National Index.” American Psychologist, 55(1), pp. 34-43.
Di Tella, Rafael, Robert J. MacCulloch and Andrew J. Oswald. (2003) “The Macroeconomics of Happiness.” The Review of Economics and Statistics, 85(4), pp. 809-827.
Di Tella, Rafael and Robert J. MacCulloch. (2008) “Gross National Happiness as an Answer to the Easterlin Paradox.” Journal of Development Economics, 86, pp. 22-42.
Dolan, Paul, Tessa Peasgood and Mathew White. (2008) “Do We Really Know What Makes Us Happy? A Review of the Economic Literature on the Factors Associated With Subjective Well-being.” Journal
of Economic Psychology, 29, pp. 94-122.
Dorn, David, Justina A.V. Fischer, Gebhard Kirchgässner and Alfonso Sousa-Poza. (2007) “Is It Culture or Democracy? The Impact of Culture and Democracy on Happiness. Social Indicators Research, 82(3), pp. 505-526.
Drukker, David M., Ingmar R. Prucha and Rafal Raciborski. (2011a) “Maximum-Likelihood and Generalized Spatial Two-Stage Least-Squares Estimators for a Spatial-Autoregressive Model with Spatial Autoregressive Disturbances.” Technical report. URL: http://econweb.umd.edu/~prucha/ Papers/WP_spreg_2011.pdf.
Drukker, David M., Ingmar R. Prucha and Rafal Raciborski. (2011b) “Creating and Managing Spatial-Weighting Matrices Using the spmat Command.“ Technical report. URL:http://econweb.umd.edu/ ~prucha/Papers/WP_spreg_2011.pdf.
Dynan, Karen E. and Enrichetta Ravina. (2007) “Increasing Income Inequality, External Habits and Self-Reported Happiness.” The American Economic Review, 97(2), pp. 226-231.
Easterlin, Richard A. (1974) "Does Economic Growth Improve the Human Lot? Some Empirical Evidence," in Nations and households in economic growth: Essays in honour of Moses Abramovitz. Edited by Paul A. David and Mel W. Reder. New York: Academic Press, pp. 89-125.
Easterlin, Richard A. (2001) “Income and Happiness: Towards a Unified Theory.” The Economic Journal, 111, pp. 465-484.
Easterlin, Richard A. (2006) “Life Cycle Happiness and Its Sources: Intersections of Psychology, Economics and Demography.” Journal of Economic Psychology, 27, pp. 463-482.
Ferrer-i-Carbonell, Ada. (2005) “Income and Well-being: an Empirical Analysis of the Comparison Income Effect.” Journal of Public Economics, 89, pp. 997-1019.
Ferrer-i-Carbonell, Ada and Paul Frijters. (2004) “How Important Is Methodology for the Estimates of the Determinants of Happiness?” The Economic Journal, 114(497), pp. 641-659.
Frey, Bruno S. and Alois Stutzer. (2000) “Happiness, Economics and Institutions.” The Economic
Frey, Bruno S. and Alois Stutzer. (2002) Happiness and Economics. Princeton: Princeton University Press.
Fukuyama, Francis. (1995) Trust: The Social Virtues and the Creation of Prosperity. New York: Free Press Paperbacks.
Glaeser, Edward L., David Laibson and Bruce Sacerdote. (2002) “An Economic Approach to Social Capital.” The Economic Journal, 112, pp. 437-458.
Gwartney, James, Joshua Hall and Robert Lawson. (2011) “2011 Economic Freedom Dataset” published in Economic Freedom of the World: 2010 Annual Report. Calgary: Fraser Institute. Hagenaars, Aldi J.M. (1986) The Perception of Poverty. Amsterdam: North Holland Publishing Co. Haller, Max and Markus Hadler. (2006) “How Social Relations and Structures Can Produce Happiness and Unhappiness: An International Comparative Analysis.” Social Indicators Research, 75(2), pp. 169-216.
Helliwell, John F. (2003) “How’s Life? Combining Individual and National Variables to Explain Subjective Well-being.” Economic Modelling, 20, pp. 331-360.
Helliwell, John F. and Robert D. Putnam. (2004) “The Social Context of Well-being.” Philosophical
Transactions of the Royal Society B: Biological Sciences, 359, pp. 1435-1446.
Helliwell, John F., Haifang Huang and Anthony Harris (2009) “International Differences in the Determinants of Life Satisfaction.” In New and Enduring Themes in Development Economics. Edited by Tridip Ray, E. Somanathan and Bhaskar Dutta. Singapore: World Scientific.
Helliwell, John F. and Shun Wang. (2011) “Trust and Wellbeing.” International Journal of Wellbeing, 1(1), pp. 42-78.
Hudson, John. (2006) “Institutional Trust and Subjective Well-Being across the EU.” Kyklos, 59(1), pp. 43-62.
Inglehart, Ronald. (1999) “Trust, Well-Being and Democracy” in Democracy and Trust. Edited by Mark E. Warren. Cambridge: Cambridge University Press, pp. 88-120.
Inglehart, Ronald, Roberto Fao, Christopher Peterson and Christian Welzel. (2008) “Development, Freedom and Rising Happiness.” Perspectives on Psychological Science, 3(4), pp. 264-285.
Kelejian, Harry H. and Ingmar R. Prucha. (1998) “A Generalized Spatial Two Stage Least Squares Procedure for Estimating a Spatial Autoregressive Model with Autoregressive Disturbances.” The
Journal of Real Estate Finance and Economics, 17(1), pp. 99-121.
Knack, Stephen. “Trust, Associational Life and Economic Performance. (2001)” MPRA paper no. 27247. Paper presented at the HRDC-OECD International Symposium on The Contribution of Investment in Human and Social Capital to Sustained Economic Growth and Well-Being, Canada. Knack, Stephen and Philip Keefer. (1997) “Does Social Capital Have an Economic Payoff? A Cross-Country Investigation.” The Quarterly Journal of Economics, 112(4), pp. 1251-1288.
La Porta, Rafael, Florencio Lopez-De-Silanes, Andrei Shleifer and Robert W. Vishny. (1997) “Trust in Large Organizations.” AEA Papers and Proceedings, 87(2), pp. 333-338.
La Porta, Rafael, Florencio Lopez-De-Silanes, Andrei Shleifer and Robert W. Vishny. (1999) “The Quality of Government.” Journal of Law, Economics and Organization, 15, pp. 222-279.
Lederman, Daniel, Norman Loayza and Ana-Maria Menendez. (2002). “Violent Crime: Does Social Capital Matter?” Economic Development and Cultural Change, 50, pp. 509-539.
Lelkes, Orsolya. (2006) “Tasting Freedom: Happiness, Religion and Economic Transition.” Journal of
Economic Behavior and Organization, 59, pp. 173-194.
MacKerron, George. (2011) “Happiness Economics from 35000 Feet.” Journal of Economic Surveys, article in press, doi: 10.1111/j.1467-6419.2010.00672.x.
Maslow, Abraham H. (1943) “A Theory of Human Motivation.” Psychological Review, 50(4), pp. 370-396.
Naef, Michael and Jürgen Schupp. (2009) “Measuring Trust: Experiments and Surveys in Contrast and Combination.” IZA Discussion Paper No. 4087.
Neumann, John von and Oskar Morgenstern. (1944) Theory of Games and Economic Behavior. Princeton: Princeton University Press.
Oishi, Shigehiro, Ed Diener, Richard E. Lucas and Suh M. Eunkook. (1999) “Cross-Cultural Variations in Predictors of Life Satisfaction: Perspectives from Needs and Values.” Personality and Social
Psychology Bulletin, 25(8), pp. 980-990.
Oishi, Shigehiro. (2010) “Culture and Well-Being: Conceptual and Methodological Issues” in
International Differences in Well-Being. Edited by Ed Diener, Daniel Kahneman and John F. Helliwell.
Oxford: Oxford University Press, pp. 1-39.
Osberg, Lars and Andrew Sharpe. (2002) “An Index of Economic Well-Being for Selected OECD Countries” Review of Income and Health, 48(3), pp. 291-316.
Oswald, Andrew J. and Nattavudh Powdthavee. (2008) “Does Happiness Adapt? A Longitudinal Study of Disability with Implications for Economists and Judges.” Journal of Public Economics, 92, pp. 1061-1077.
Oswald, Andrew J. and Stephen Wu. (2010) “Objective Confirmation of Subjective Measures of Human Well-Being: Evidence from the U.S.A.” Science, 327, pp. 576-579.
Pollak, Robert S. (1976) “Interdependent Preferences.” The American Economic Review, 1976, 66(3), pp. 309-320.
Putnam, R. (1993) Making Democracy Work: Civic Traditions in Modern Italy. Princeton: Princeton University Press.
Radcliff, Benjamin. (2001) “Politics, Markets and Life Satisfaction: The Political Economy of Human Happiness.” The American Political Science Review, 95(4), pp. 939-952.
Shiller, Robert J., Stanley Fischer and Benjamin M. Friedman. (1984) “Stock Prices and Social Dynamics.” Brookings Papers on Economic Activity, 2, pp. 457-510.
Smith, Adam. (1776/1976) The Wealth of Nations. Oxford: The Claredon Press.
Smith, Kenneth. (2003) “Individual Welfare in the Soviet Union.” Social Indicators Research, 64(1), pp. 75-105.
Stanca, Luca. (2010) “The Geography of Economics and Happiness: Spatial Patterns in the Effects of Economic Conditions on Well-Being.” Social Indicators Research, 99(1), pp. 115-133.
Stevenson, Betsey and Justin Wolfers. (2008) “Economic Growth and Subjective Well-being: Reassessing the Easterlin Paradox.” Brookings Papers on Economic Activity, Spring, pp. 1-87 Stiglitz, Joseph E., Amartya Sen and Jean-Paul Fitoussi (2009) Report by the Commission on the
Measurement of Economic Performance and Social Progress. Retrieved June 25, 2012 from
www.stiglitz-senfitoussi.fr.
Tversky, Amos and Daniel Kahneman. (1974) “Judgment under Uncertainty: Heuristics and Biases.” Science, 185(4157), pp. 1124-1131.
Tversky, Amos and Daniel Kahneman. (1981) “The Framing of Decisions and the Psychology of Choice.” Science, 211(4481), pp. 453-458.
Wolfers, Justin. (2003) “Is Business Cycle Volatility Costly? Evidence from Surveys of Subjective Well-Being.” International Finance, 6(1), pp. 1-26.
Appendix
Table 1: Influences of determinants of subjective well-being
Factor Influence References
Personal Situation
Income level Positive, but diminishing Blanchflower and Oswald (2004); Diener et al. (1999) Relative income Positive Ferrer-i-Carbonell (2005)
Age Negative Blanchflower and Oswald (2008); Easterlin (2006)
Age squared Positive Blanchflower and Oswald (2008); Easterlin (2006)
Gender Mixed Alesina et al. (2004); Oswald and Powdthavee (2008)
Education Mixed Blanchflower and Oswald (2004); Clark (2003)
Subjective Health Positive Okun and George (1984); Smith (2003) Objective Health Positive, but weak Okun and George (1984)
Unemployment Negative Di Tella et al. (2001); Clark (2003) Community Involvement Positive Helliwell and Putnam (2004) Religious Activities Positive Helliwell (2003)
Social Trust Positive Bjørnskov (2003); Helliwell and Putnam (2004) Political View Positive for socialists Radcliff (2001)
Religion Mixed Smith (2003); Helliwell (2003)
Marriage Positive Blanchflower and Oswald (2004); Frey and Stutzer (2000)
Separated Negative Helliwell (2003)
Children Mixed Frey and Stutzer (2000); Haller and Hadler (2006)
Social Contact Positive Lelkes (2006)
Economic, Social and Political situation
Economic Growth Positive Di Tella et al. (2003); Stevenson and Wolfers (2008) Unemployment Rate Negative Di Tella et al. (2003); Alesina et al. (2004)
Inflation Negative Di Tella et al. (2003); Wolfers (2003)
Income Inequality Mixed Haller and Hadler (2006); Alesina et al. (2004) Degree of Democracy Positive Frey and Stutzer (2000); Dorn et al. (2007)
Pollution Negative Di Tella and MacCulloch (2008)
Table 2: Descriptive statistics of variables used in step 1
Variable Mean Mean min Mean max Obs
Life satisfaction 64.689 39.147 84.157 148909 Happiness 30.570 24.787 35.068 148561 Income 4.600 2.249 6.563 148909 Full-time employment 0.354 0.121 0.615 148909 Part-time employment 0.073 0.014 0.208 148909 Self-employed 0.114 0.010 0.616 148909 Unemployed 0.098 0.011 0.355 148909 Retired 0.118 0.004 0.343 148909 Housewife 0.149 0.007 0.413 148909 Students 0.071 0.012 0.194 148909 Other employment 0.022 0 0.363 148909
Inadequately completed elementary education 0.129 0 0.626 148909 Completed elementary education 0.141 0.005 0.439 148909 Incomplete technical/vocational school 0.073 0 0.324 148909 Complete technical/vocational school 0.181 0 0.481 148909 Incomplete university-preparatory school 0.087 0 0.529 148909 Complete university-preparatory school 0.161 0 0.490 148909 Some university without degree 0.070 0 0.311 148909 University with degree 0.157 0 0.421 148909
Married 0.593 0.169 0.864 148909 As married 0.064 0 0.501 148909 Divorced 0.033 0 0.122 148909 Separated 0.018 0 0.063 148909 Widowed 0.057 0.006 0.129 148909 Single 0.236 0.068 0.505 148909 No children 0.273 0.108 0.587 148909 1 child or 2 children 0.412 0.142 0.750 148909 3 or more children 0.315 0.070 0.523 148909 More income equality desired 0.316 0.066 0.667 148909 Neutral position towards income inequality 0.315 0.131 0.594 148909 Larger income differences desired 0.369 0.012 0.710 148909 Importance of God 7.854 3.525 10 148909
Male 0.491 0.374 0.592 148909
Bad subjective state of health 2.168 1.722 2.812 148909
Table 2: Descriptive statistics of variables used in step 1 (continued)
Variable Mean Mean min Mean max Obs
Age 40.468 28.685 49.502 148909 Age squared 1881.869 901.549 2734.088 148909 Year 1990 0.022 0 0.285 148909 Year 1991 0.007 0 0.298 148909 Year 1995 0.064 0 0.578 148909 Year 1996 0.135 0 1 148909 Year 1997 0.044 0 1 148909 Year 1998 0.057 0 1 148909 Year 1999 0.021 0 1 148909 Year 2000 0.073 0 0.524 148909 Year 2001 0.140 0 1 148909 Year 2002 0.023 0 1 148909 Year 2003 0.015 0 1 148909 Year 2005 0.083 0 1 148909 Year 2006 0.142 0 1 148909 Year 2007 0.122 0 1 148909 Year 2008 0.034 0 1 148909
Table 3: Descriptive statistics of variables used in step 2
Variable Mean Std. Dev. Min Max Obs
Social trust 1.625 2.897 -6.480 10.367 73 Economic freedom, 1:10 6.610 0.812 4.753 8.844 73 Gini index, 0:100 38.463 8.676 22.650 60.273 73 Public spending on education, % of gov. expenditure 15.514 4.521 8.435 27.460 73 Protestant, % of population 11.805 21.245 0 97.8 73 Roman Catholic, % of population 30.868 36.262 0 96.9 73 Muslim, % of population 19.198 32.675 0 99.4 73
Robustness check variables
