Keywords

1 Introduction

Individual well-being is a central concept of Economics, and there are several approaches for its measurement and inclusion in studies investigating its determinants and consequences on behaviour. In this chapter we will focus on the analysis of the contributions in the so-called income distribution literature and economics of happiness. The two are related since they examine what are there known as “objective” and “subjective” aspects of well-being, that are respectively what the analyst thinks individual well-being is, based on command over economic resources, and what the individual herself says it is, when asked to report on it. The distinction between objective and subjective well-being is by no means a distinction between reality and imagination, as well-being is always an experience made by the individual. Economists use these two terms to simply distinguish between measures of well-being computed by the analyst from the command the individual has over economic resources, and measures of well-being that are stated by the individual directly when asked to report on it. In this chapter devoted to Economics we will analyse both in details and refer to it as custom in the literature.

Traditionally, economists have dealt with the question of well-being through the lens of the concept of utility. Assuming that individuals are rational and fully informed, and seek to maximise utility, economists inferred the well-being of individuals from the decisions that they make (the revealed preferences) and from their behaviour. However, a burgeoning strand of the literature in Economics utilizes self-reported well-being as an indicator of economic and social progress. In this chapter, we provide a detailed review of this literature. We first outline the different measures of subjective well-being used by the literature and we assess the nature and cross-cultural validity of these measures. Second, a theoretical framework that serves as a basis for empirical analysis of subjective well‐being is set. We then review the most influential articles exploring the determinants of subjective well-being. The last part of our chapter presents a summary of the evidence regarding the behaviour of happy people. We conclude with measures of objective well-being from the income distribution literature, such as unidimensional and multidimensional poverty.

2 “Subjective” Well-being in Economics: Measurement, Objectives and Limits

2.1 Measurement of Subjective Well-being in Economics

There are two main types of measures of subjective well-being in Economics. The first type of subjective well-being measures is cognitive (or evaluative).Footnote 1 Individuals are directly asked to make statements about how well their life is going. The objective of the cognitive measures of well-being is to capture the reflective process of an individual judging herself, her satisfaction with respect to her life, or a particular aspect of it. The most popular cognitive measures of well-being in Economics are the Cantril Ladder and the life-satisfaction question. The first one is usually stated as follows: “Please imagine a ladder with steps numbered from 0 at the bottom to 10 at the top. The top of the ladder represents the best possible life for you and the bottom of the ladder represents the worst possible life for you. On which step of the ladder would you say you personally feel you stand at this time?” The life-satisfaction question is generally stated as follows: “How satisfied are you with your life, all things considered?” The scale the respondents use to answer this question varies across surveys. For instance, in the British Household Panel Survey, the life-satisfaction question’s scale goes from one to seven while it goes from zero to ten in the German Socio Economic Panel. Similar questions are also used to measure the satisfaction of individuals regarding different domains of life such as work, finance or family life.

The second type of self-reported well-being measures is referred to as affective (or hedonic) well-being and refers to a specific point in time. There are two types of affective measures of well-being: positive affect and negative affect. The former includes elements such as the frequency of experiencing positive feelings or smiling; the latter encompasses negative emotions (stress, anxiety etc.). Contrary to evaluative measures, affective well-being are not supposed to capture a reflective process but a personal emotional state at a precise moment. Questions measuring affect usually are of the following kind: “From 0–6, where a 0 means you were not stressed at all and a 6 means you were very stressed, how stressed did you feel during this time?” The U-index is one of the most-used measures of affect in the literature. This index, going from 0 to 1, represents the proportion of time spent in activities where negative feelings were more intense than positive feelings during a given day. As they usually concern a particular moment or activity, affective well-being is more sensitive to variations in the short-run than evaluative measures.

2.2 Measurement Issues and Reliability

The conditions of the collection of subjective well-being data has been shown to be important. In Conti and Pudney (2011), the level of satisfaction reported by survey respondents is on average higher during oral interviews as compared to computer-assisted self-interviews and when children are present during the interview. It confirms that self-reported scores of well-being are likely to be inflated because of a social desirability bias. In the same study, Conti and Pudney (2011) demonstrate that the presence of the interviewee’s partner during the interview lowers the level of reported satisfaction. This is seen by the authors as a way for the respondent to maintain a strong bargaining power in the household. The date of the interview as well as the question ordering (Deaton 2012) are also crucial and potential sources of biases. One may also raise conceptual questions such as the issues of cardinality or inter-personal comparisons.

Economists, however, have extensively discussed these various issues and the validity of the well-being measures.

There are two important elements to note regarding the reliability of measures of subjective well-being. First, both cognitive and hedonic measures of subjective well-being are highly correlated (Clark 2016). Albeit they are measured using different questions, different scales and they refer to different time horizons, the relatively high levels of correlation between these two families of measures mean that they are arguably capturing the same concept, i.e. well-being. Note also that the different measures of subjective well-being also share a large set of determinants. Nevertheless, cognitive and hedonic measures of well-being are never perfectly correlated; this is expected as they are meant to measure aspects of well-being that are arguably different.

Second, experimental studies confirm that measures like life satisfaction capture the concept of well-being. Using a series of hypothetical pairwise-choice scenarios, Benjamin et al. (2012) show that options that are perceived by individuals as the better choice in terms of life satisfaction score are also those they would choose in real life in roughly 80% of the cases.

Convergent validity tests have been performed too. One of the most famous examples is the cross-rating exercise: in Pavot and Diener (1993), the level of correlation between the interviewer ratings and self-ratings goes from 0.43 to 0.66. In the same study, one can find that family members and friends are also able to predict accurately the happiness of the main respondent.

The question of the interpersonal comparability of subjective well-being has been addressed in various ways. First, most of the researchers in Economics make use of longitudinal dataset to run within-regressions. This estimation method focuses on individual changes (within-variation) and neutralises differences between individuals (between-variance). It is also possible to neutralise a variety of reporting biases using longitudinal data (e.g. questionnaire changes, bias due to the day of the interview). Last, it is worth mentioning that individuals tend to use the scales of subjective well-being measures in a comparable since observed behaviours can be predicted by levels of happiness in cross-section (Clark 2001; O’Connor, 2020).

3 The Use of Subjective Well-being Data in Economics: Theoretical Considerations

Theoretical models in Economics are based on the concept of utility. Utility can be interpreted as a measure of satisfaction an individual derives from the consumption of commodities. It usually takes the following form: U = U(X) where U is the utility that depends on the vector of determinants X. While the use of utility as a theoretical tool is widespread, its empirical measurement is problematic. According to Marshall (1920) and Samuelson (1938), utility cannot be directly measured but it can be indirectly observed from individual choices. This is what economists call the revealed preferences approach.

The use of subjective well-being in Economics in the past forty years can be seen as an alternative way of measuring utility. According to Hirschauer et al. (2015), considering subjective well-being as a proxy for utility brings utility back to its utilitarian definition. Bentham (1789) stated that “Nature has placed mankind under the governance of two sovereign masters, pain and pleasure. It is for them alone to point out what we ought to do, […]. By the principle of utility is meant that principle which approves or disapproves of every action whatsoever according to the tendency it appears to have to augment or diminish the happiness of the party whose interest is in question: […]. I say of every action whatsoever, and therefore not only of every action of a private individual, but of every measure of government.” According to Bentham, the empirical measure of utility, or “felicific utility”, should incorporate different elements such as the intensity of pain and pleasures as well as their duration. Subjective well-being data arguably appear as a proxy for the measurement of the utility a la Bentham.

The standard approach in applied Economics is to use large databases and econometrics methods to estimate the following model:

$$SWB_{it} = \beta^{\prime}X_{it} +{\rm{\varepsilon }}_{it}$$

where \(SWB_{it}\), the subjective well-being of an individual (or a country) i at time t. The objective of this type of model is to estimate to what extent a given vector of characteristics, Xit, influences significantly the utility function. It might also be used to estimate empirically marginal rates of substitution and to test for the existence of concepts such as the interdependence of utilities.

4 The Easterlin Paradox and the Concept of Relative Utility

Richard Easterlin is one of the first researchers who brought measures of subjective well-being to academic research in the field of Economics. The so-called ‘Easterlin Paradox’ (1974, 1995) suggested that, despite substantial real income growth in Japan and Western countries over time, average happiness levels remained roughly constant. This paradox seems to contradict causal evidence and a parallel body of work that has shown three main facts: 1) within each country at a single point in time, richer individuals are more satisfied with their life (Blanchflower and Oswald 2004; Graham and Pettinato 2002). This correlation holds both for developed and developing countries. 2) Using panel data to control for individual fixed effects, Winkelmann and Winkelmann (1998), Ferrer-i-Carbonell (2005) and others concluded that not only the level of income matters, but also changes in income are positively correlated with changes in life satisfaction. Within this framework, Frijters et al. (2004) established  the causal impact of income changes exploiting quasi-natural experiments. 3) Life satisfaction is also positively correlated with macroeconomic variables such as GDP in very large cross-time cross-country samples (Di Tella et al. 2003; Alesina et al. 2004). One way to reconcile the above evidence on the positive relation between income and life satisfaction with the Easterlin Paradox is to consider the existence of a relative component in the utility function (Clark et al. 2008). This means that income is evaluated relative to others or to oneself in the past.

The standard comparison effect states that individuals report on average lower levels of well-being when the income of their peers increases. Defining who the peers (or the reference group) are is crucial here. The standard approach in Economics is to consider that individuals with the same age, the same sex, the same education or living in the same region can be seen as peers. Using wage regressions to estimate the comparison income, Clark and Oswald (1996) show that a similar increase in own income and comparison income has no effect on well-being: the positive effect of the former is totally reduced by the negative effect of the latter. With a similar method but different datasets, Sloane and Williams (2000) and Levy-Garboua and Montmarquette (2004) confirmed that comparison income is negatively correlated with job satisfaction. The literature in Economics also extensively used cell averages to estimate the comparison income (McBride 2001; Blanchflower and Oswald 2004; Ferrer-i-Carbonell 2005; Luttmer 2005; Graham and Felton 2006). While these articles use data from different countries, different periods, and define the reference group in various ways, they all confirm that the lower the comparison income, the higher the own well-being. Card et al. (2012) address causality issues by exploiting a quasi-natural experiment setting. The state of California made the salary of any state employee public knowledge and a local newspaper set up a website to ease the access to this information. Card et al. (2012) informed a random subset of employees of three different campuses at the University of California about the existence of the website. Some days later, Card et al. (2012) surveyed all the employees from the three campuses and found that the informed employees whose wage was lower than their colleagues (defined as co-workers in the same occupation group and administrative unit) reported lower level of job satisfaction and a higher probability to quit. Note that they found no effect for employees at the top of the wage distribution.

A second bulk of the literature documents a positive relationship between one’s own well-being and the comparison income (Senik 2004; Clark et al. 2009). This phenomenon is referred as the information effect. Here researchers follow Hirshman’s (1973) interpretation and appeal to the information content that economic advances of others have: the presence of richer individuals signals that there is a possibility for oneself to get richer in the future, which increases own happiness even before any actual enrichment takes place. On this point see also D’Ambrosio and Frick (2012). These authors incorporate in the analysis the effects of time, that is of passing and being passed by others and, relying on panel data from Germany, they show that comparison and signal effects can coexist: the first is found with respect to those whose relative position did not change over time, while those who moved play an information role.

We also know that comparisons are not only interpersonal but also intra-personal. Using panel data, Clark (1999) and Grund and Sliwka (2007) show that past level of income has a negative effect on current well-being. Using German data, Di Tella et al. (2010) find that most of the effect on an increase in income vanishes after a year. Habituation refers to the evidence that people adapt to having more income, a phenomenon known as hedonic adaptation, or hedonic treadmill (see Lyubomirsky 2011, for an excellent survey).

5 The Main Individual Determinants of Well-being

5.1 Unemployment

Unemployment is probably the most widely studied of all personal characteristics, after income. The standard microeconomic model assumes that utility only depends on consumption and leisure, and labour affects utility only indirectly, i.e. it increases consumption but reduces leisure. In this model, unemployment is seen as voluntary. Consequently, we shall expect no effect of unemployment on well-being once controlling for income. The empirical literature, however, unanimously confirms that, keeping income constant, unemployment does reduce well-being. This is true both in cross-sections and panels (Clark and Oswald 1994; Winkelmann and Winkelmann 1998; Dolan et al. 2008; Frey and Stutzer 2010). Kassenboehmer and Haisken-DeNew (2009) used plant-closure as natural experiments and found similar results. Those findings are important because they suggest that unemployment is mostly involuntary.

Past unemployment also affects well-being today. This scarring effect has been extensively discussed (see Clark et al. 2001; Bell and Blanchflower 2011). In a recent work, Clark and Lepinteur (2019) find that the overall unemployment experience between the end of full-time education and age 30 significantly reduces life satisfaction at age 30.

The unemployment of other people matters too. Clark (2003) showed that the effect of unemployment on well-being is smaller for those with an unemployed partner or who live in higher-unemployment regions. He also showed that those whose well-being fell the most on entering unemployment leave unemployment faster, consistent with hysteresis in unemployment.

Having a job significantly increases well-being but job characteristics are of importance too. Job security is one of the most important job characteristics (Clark 2001). The more protected the better but, as income, relative job security matters (Clark and Postel-Vinay 2009; Georgieff and Lepinteur 2018). The effect of working time has been widely discussed as well. Correlational studies yield contradictory evidence but Loog and Collewet (2014) used an instrumental variable approach to estimate the causal impact of work-week length on life satisfaction. They conclude that working time has an inversed-U shaped effect and the optimal work-week length is just below 30 h. Lepinteur (2019a) confirms this finding by showing that mandatory reductions in working time in France (from 39 h per week to 35) and in Portugal (from 44 h per week to 40) increased the well-being of workers. Once again, relative working time also influences well-being (Booth and Van Ours 2008; Collewet et al. 2017, Lepinteur  2019b).

5.2 Health

Layard et al. (2014) and Clark et al. (2019) developed a life-course model of well-being and demonstrate that the most important determinant of well-being is health. Dolan et al. (2008) show that both physical and mental health matters. Shields and Price (2005) find that the levels of well-being reported after a heart attack or a stroke are on average lower. Well-being is positively correlated with life expectancy (Danner et al. 2001) and negatively correlated with cardiovascular diseases (Steptoe et al. 2015).

Using longitudinal data, Oswald and Powdthavee (2008) show that individuals adapt only partially to negative health shocks, e.g. disability. While severe disability reduces well-being immediately, there is no full recovery in terms of well-being during the three subsequent years. Moreover, Clark et al. (2019) find that living with an ill partner reduces well-being.

5.3 Marital Status

Married individuals are found to be happier than the rest of the population in a variety of studies (see Stutzer and Frey 2006, for an excellent review). Their higher level of happiness, however, does not come from marriage per se. In Clark and Georgellis (2013), it appears that the positive effect of getting married fades away after the first 3 years. The observed premium in happiness among married individuals comes from the absence of adaptation to partnership.

In the same study, Clark and Georgellis (2013) show that divorce brings back individuals to the average level of happiness of the rest of the population. Widowhood has a negative impact on well-being that does not last.

5.4 Education

There is no consensus in the Economics literature about the effect of education on well-being (Blanchflower and Oswald 2004; Stutzer 2004). Using mandatory increases in school-leaving age as natural experiments to address endogeneity concerns, Oreopoulos and Salvanes (2011) report that education has a positive effect on well-being in the US while Clark and Jung (2017) found no effect in the UK. One way to explain such results is to consider that not only education increases earnings, but it also raises expectations (Clark et al. 2015). Using Australian data, Nikolaev (2018) showed that the reference group education is negatively correlated with well-being. This result is robust to a variety of definitions of the reference group and goes beyond the effect of relative income: this suggests that a higher education-level has the role of a desirable social status. Note that individuals with high levels of education are less prone to comparisons. Clark et al. (2019) replicated this analysis with British, German and American data and confirmed these results.

5.5 Age and Sex

One of the most stable findings in the literature is the U-shape curve between well-being and age (Gerdtham and Johannesson 2001; Hayo and Seifert 2003; Blanchflower and Oswald 2004) but convincing explanations are still missing today.

Women often report higher levels of cognitive well-being (Helliwell et al. 2016) but also more negative affect and stress (Nolen-Hoeksema and Rusting 1999; Kahneman and Deaton 2010). In a similar vein, Clark (1997) found that women are happier than men on the labour market and showed that the gender gap cannot be fully explained by factors such as observable characteristics and selection. He suggests that the gender gap might be explained by the fact that women have on average lower expectations on the labour market and evaluate the quality of their job more positively. Note that recent studies found that the gender gap in well-being decreases (Stevenson and Wolfers 2009) or has disappeared (Green et al. 2018).

6 Predicting the Behaviour of Happy People

Another strand of the literature in Economics considers subjective well-being as a factor that is likely to influence individual’s behaviour instead of treating it as an outcome. The standard model here is the following:

$$Y_{it + 1} = \gamma_{1} SWB_{it} + \gamma^{\prime}_{2} X_{it} +\epsilon_{it}$$

In this framework, \(Y_{it}\) is a future objective behaviour or decision that is plausibly influenced by current subjective well-being. It is important to mention here that \(\gamma_{1}\) is estimated keeping the vector of objective characteristics \(X_{it}\) constant as it allows isolating the contribution of subjective well-being and shows to what extent it incorporates a meaningful and independent information about individual behaviour. We here present a non-exhaustive list of papers that applied this approach.

There are different behaviours and outcomes that are influenced by the current level of subjective well-being. Individuals with higher levels of well-being have better health outcomes. One famous contribution here is the “Nun Study,” in which nuns who wrote more positive short descriptions of their life in their late teens and early twenties were significantly more likely to still be alive 60 years later (Danner et al. 2001). Using larger samples and more sophisticated statistical methods, Diener and Chan (2011) and Banks et al. (2012) find that life satisfaction is significantly associated with better future health outcomes.

Subjective well-being predicts a variety of outcomes and behaviours on the labour market. Conditional on current employment, higher levels of life satisfaction today reduce the probability of being unemployed in the future (O’Connor2020). Similarly, keeping the objective job characteristics constant, Clark (2001) showed that a low level of job satisfaction predicts a higher probability of job quit.

Finally, Liberini et al. (2017) showed that measures of subjective well-being can be used to predict individual voting behaviours. Using British data and an instrumental variable approach, they demonstrate that the higher the well-being scores, the higher the probability to support the incumbent. Using worldwide data, Ward (2020) finds similar results. At the macroeconomic level, national happiness was a better predictor of the incumbent government vote share than the GDP growth rate, unemployment rate and inflation rate.

7 “Objective” Well-being in Economics: Unidimensional and Multidimensional Poverty

Poor individuals have an objectively low level of well-being since they are not able to satisfy their basic needs, that may, or may not, depend on the society they live in. When this is the case, poverty is said to be relative, when this is not the case, poverty is absolute. These considerations are captured by the choice of the level of income (or consumption) that determines who is poor, the so-called, poverty line. An absolute poverty line is a number that is fixed according to some criterion and does not change constantly over time, such as, for example, the international poverty line set by the World Bank of $1.90 a day. A relative poverty line is a number that depends on the distribution of resources of the country of residence of the individual under analysis, such as the poverty lines set by the EU member states equal to 60% of the median of the distribution of equivalent household disposable income of the year under analysis. Any change in the median income will be automatically reflected in an update of the poverty line.

The fundamental contribution to the measurement of unidimensional poverty, that is of poverty that looks only at one dimension of well-being such as income or consumption, is based on Sen (1976). Sen (1976) viewed poverty measurement as involving two exercises: (i) the identification problem: the identification of the poor, and (ii) the aggregation problem: the aggregation of the characteristics of the poor into an overall indicator of poverty. The identification problem was discussed above and requires the specification of a poverty line—a demarcation line separating poor individuals from non-poor persons in the population. Once the poor persons are identified, the next step is to aggregate the information on the poor into an indexthat will quantify the extent of poverty.

Many indices of poverty have been proposed in the literature. The most popular include headcount, poverty gap and squared poverty gap indices. Headcount measures the proportion of the population that is poor, that is, it consists of a simple count of the poor as a fraction of the total number of individuals in society. This index, of the incidence of poverty, is very simple and easy to understand but it lacks any consideration of the depth of poverty, that is, of how poor the individual is. The poverty gap index was proposed to overcome this shortcoming and to capture the intensity of poverty, that is the amount of money the poor needs to cross the poverty line. This difference is known as the individual poverty gap. The poverty gap index is equal to the average, over the population, of the individual’s poverty gaps as a proportion of the poverty line. The third index of poverty, the squared poverty gap, averages the poverty gap squared relative to the poverty line. It is a measure that considers inequality among the poor since it gives much more weight to people with larger poverty gaps. These three indices look at the three I’s of poverty: incidence, intensity and inequality.

Researchers are in agreement that well-being is multidimensional; to assess all aspects of poverty, more than one dimension needs to be considered, such as, for example, income, health and education of the individual. The measurement of multidimensional poverty is much more complicated that what we summarized above for the unidimensional case. Multiple poverty lines should be set, one for each dimension of poverty, and the aggregation stage is more complex given the possible associations existing between the dimensions that need to be included (going back to the example above, a positive correlation exists between income, health and education). An additional difficulty is given by the type of variables relevant for the measurement of poverty that are very often ordinal or categorical, adding an additional layer of difficulty to the measurement of association between dimensions and the aggregation step. We refer the reader interested to know more to the handbook edited by D’Ambrosio (2018).