Who compares to whom? The anatomy of income comparisons in Europe Andrew E. Clark* (Paris School of Economics and IZA) Claudia Senik** (Paris School of Economics, University Paris-Sorbonne, IZA and IUF) March 2009 Abstract. This paper provides unprecedented direct evidence on the intensity and direction of income comparisons. Income comparisons are considered to be important by a considerable majority of Europeans. They are associated with both lower levels of self-reported satisfaction and a greater demand for income redistribution. Colleagues are the most frequently-cited reference group, and the most innocuous one; comparisons to friends are both less widespread and much more toxic for satisfaction. This set of findings is consistent with information effects, as professional reference groups may represent a source of information about one’s future prospects rather than a simple benchmark for comparisons. Keywords: Income Comparisons, Happiness, Redistribution, European Social Survey JEL codes: D31, D63, I3, J31, Z13 * Andrew Clark, PSE, 48 Boulevard Jourdan, 75014 Paris, France. Tel.: +33-1-43-13-63-29. E-mail: Andrew.Clark@ens.fr. **Claudia Senik, PSE, 48 Boulevard Jourdan, 75014 Paris, France. Tel.: +33-1-4313-63-12. E-mail: senik@pse.ens.fr. The Norwegian Social Science Data Services (NSD) is the data archive and distributor of the ESS data. We are grateful to CEPREMAP for financial support. Who compares to whom? The anatomy of income comparisons in Europe 1. Introduction There is by now a very substantial literature on social interactions and relative utility. The key question in these literatures is: To whom do individuals compare, and with whom do they interact? Outside of the laboratory, where the reference group can be directly controlled (see Falk and Ichino, 2006, McBride, 2007 and Clark et al., 2009b), researchers have to make non-testable assumptions about the relevance of different reference groups for comparison purposes. Almost all empirical papers have got round this problem by simply imposing a certain reference group (e.g. Cappelli and Sherer, 1988, Clark and Oswald, 1996, MacBride, 2001, Bygren, 2004, Ferrer-i-Carbonnel, 2005, Luttmer, 2005, and Helliwell and Huang, 2009). Only rarely are different reference groups tested (as in Knight and Song, 2006). Even scarcer is direct survey evidence on the groups to which individuals compare. Knight and Song (2006) appeal to a survey of Chinese households and showed that two-thirds of respondents report that their main comparison group consists of individuals in their own village. In a very recent paper, Senik (2009) exploited a survey of 25 post-Transition countries (LITS), where people were explicitly asked to compare their current living standard with that of various groups of people that they used to know before the transition started (i.e. before 1989). She finds that the most important welfare impact comes from the deterioration of one’s living standard and from under-performing relative to one’s former schoolmates or colleagues, rather than from more general comparisons such as the individual’s self-ranking on the social ladder. In this paper we exploit new survey information from 22 European countries, which contains direct measures of the intensity and the direction of income comparisons. This information from Wave 3 of the European Social Survey (ESS) allows us to evaluate the extent to which individuals compare their income against that of relevant others; it also allows us to identify who these relevant others actually are. -1- With respect to how much income comparisons matter, empirical work almost always calculates an average effect for the whole sample, read off as the estimated coefficient on comparison income (y*) in a subjective well-being regression. Only rarely is heterogeneity in the relationship between well-being and income addressed (one exception is the latent class analysis in Clark et al., 2005). Here we can identify groups who compare their income more and groups who compare their income less. We can also observe whether the direction of comparisons (the relevant benchmarks) varies across those groups. The paper has three main aims in analysing this novel data: (i) Describe the extent of income comparisons across Europe (do we compare our incomes?); (ii) Describe the object of income comparisons across Europe (to whom do we compare?); and (iii) See how income comparisons affect subjective well-being and the demand for redistribution (does it matter?). One implication of a well-being function that includes others’ income (y*) in the form U=U(y, y-y*,…) is that greater comparisons (i.e. a higher value of U2) will increase or decrease well-being depending on whether the individual’s income is greater or less than y*. On the other hand, inequality aversion, as described in Fehr and Schmidt (1999), supposes that individuals don’t like earning less than y* (disadvantageous inequality), but nor do they necessarily like earning more (advantageous inequality): they are averse to all kinds of outcome inequality. In this case, a greater weight on comparisons will reduce well-being independently of the individual’s own position relative to others. The ESS dataset allows us to evaluate these kinds of predictions in a large-sample, nonexperimental way. However, we can also do somewhat more than that. While theory provides us with some ideas about the degree of income comparisons, it does not say much about whether the results are dependent on the object of such comparisons: comparing to one reference group may be far more noxious in satisfaction terms than comparing to another reference group. We are almost never able to draw this kind of distinction empirically within the same dataset. Here we have the possibility of doing so, and evaluating whether income comparisons matter more (for well-being) depending on the object of the comparisons (i.e. to whom?). -2- The broad distinction that we will draw here is between non-Hirschman and Hirschman (1973) types of comparisons. The first of these concerns the “standard” envy effect: we are left less well-off when those in our reference group do better. The second nuances this conclusion by noting that what happens to my reference group today may provide me with information about what will happen to me tomorrow. Here others’ good fortune likely continues to make me feel envious, but also has an element of good news about myself in the future. The net effect may then be negative, zero or positive, depending on the relative size of these status and signal phenomena. Of course, for such informational effects to take place, people in the reference group need to share some of the characteristics that will likely influence my own future trajectory. Senik (2004 and 2008) provides evidence of Hirschman-type informational effects in high mobility/uncertainty countries (Russia, the New Europe and the United States), showing that individual life satisfaction is positively correlated with the income of others with the same occupation. Clark et al. (2009a) appeal to linked employeremployee data to suggest that the net satisfaction effect of the income of other workers in the same firm is actually positive. These papers both consider professional reference groups. However, neither of them were able to evaluate the effect of these comparisons to colleagues relative to comparisons to friends or family members. We can carry out such calculations using the data available in the ESS. Equally, Senik (2009) evaluates the different well-being effects of various kinds of comparisons: to past living standards; former colleagues; former high-schoolmates; parents; and social rank. However, she has no information about how widespread these different types of comparisons are. The ESS data allows us to put all of these elements together, and consider the effect of income comparisons on individual well-being, depending on both their strength and the type of comparison under consideration. We believe that this is the first example of this type of analysis, especially in the context of a large-scale cross-country dataset. The paper is organised as follows. Section 2 describes the intensity of income comparisons across European countries, and Section 3 considers the incidence of the different reference groups (are income comparisons carried out relative to colleagues, -3- family members, friends or others?). Sections 4 and 5 then dissect the effect of the intensity and direction of income comparisons on well-being and the demand for redistribution respectively. Finally, Section 6 concludes. 2. What do people say about income comparisons? A set of novel and useful questions appeared in Wave 3 of the European Social Survey (the ESS: freely available from http://www.europeansocialsurvey.org). The ESS is a multi-country survey which has covered 30 different countries at various points over its first three rounds. Wave 3 of the ESS contains a special module on well-being, in which the key income comparison questions appeared. This wave, collected in 2006/2007, covers 23 different countries (although the comparison question was not asked in Cyprus). We restrict the sample to those of working age (16-65). On average, we have just under 1500 observations per country, for a total sample size of 34 379. A number of the questions in this well-being module deal with the broad questions of income comparisons to others. First, all respondents who are in employment are asked “How important is it for you to compare your income with other people’s incomes?”. They answer using a showcard, where 0 corresponds to “Not at all important”, and 6 is labelled “Very important”; none of the intervening values are labelled. The weighted distribution of answers to this income-comparison question across 22 European countries is shown in Table 1 below. Table1. “How important is it to you to compare your income with other people’s incomes?” Observations Weighted % Not at all important 1 2 3 4 5 Very important 5293 3783 3082 3768 3006 2095 1208 23.8 17.01 13.86 16.95 13.52 9.42 5.43 Total Note: Weighted statistics. 22234 100 -4- Table 1 shows that just under a quarter of Europeans say that it is not at all important to compare their incomes to others. In general there is something of a left-skew in the distribution of comparison intensity, but there are still substantial numbers of individuals towards the top end of the distribution. To illustrate the country distribution of the degree of income comparisons, we assign the cardinal numbers zero to six to the individual replies detailed above, and calculate the average intensity of income comparisons by country. As Table 1 indicates, countries with higher average scores therefore compare their incomes more. Figure 1 depicts these country averages, ordered from the least to the most comparisonsensitive. Although it is difficult to establish very precise country patterns, most Eastern European countries are found on the right-hand side of this figure, whereas Switzerland, Germany and Austria appear towards the left-hand side. In Figure 1, we also plot the level of Real GDP per capita, taken from the latest available Penn Table (Heston et al., 2006). Visual evidence does suggest the existence of a negative correlation between average income per head and the average intensity of comparisons across countries, suggesting that most comparisons are upward comparisons, and are made by the relatively poor. This impression can be formalised by calculating the correlation coefficient between the two series: this is -0.63, and significant at better than the one per cent level. As will be shown in the next paragraph, the same observation can be made within countries: people with lower incomes seem to compare more. -5- Figure 1. “How important is it to you to compare your income with other people’s incomes?”. 3.1 2.9 Mean importance of income comparisons, left axis Real GDP per capita, right axis 2.7 40000 35000 30000 2.5 25000 2.3 20000 2.1 1.9 15000 10000 1.5 5000 Ne t he rla nd Fi s nl an Au d Sw stri itz a er lan Be d lg i Un Ge um r ite m d Ki any ng do De m nm ar Ir e k la Po nd rtu g No al rw a Bu y lg ar ia Fr an c Sw e ed en Ru ss Sl ia ov en ia Po lan d Sl ov ak Spa Re in pu bl ic 1.7 Notes: Average weighted income comparison by country (on a 0-6 scale). Real Gross Domestic Product per Capita in 2004, in current dollar prices, source: Heston, Summers and Aten, Penn World Table, 2006. To formalise the analysis of the intensity of income comparisons, we run a regression of the variable in Table 1 on a standard set of individual socio-demographic variables, as well as country dummies. Income is measured at the household level, and is banded into eleven categories. Income information is missing in three countries (Estonia, Hungary, and Ukraine). As the dependent variable is ordinal, rather than cardinal, we should ideally carry out ordered probit or logit analyses. It turns out that the results from cardinal analysis using OLS are extremely similar (as in Ferrer-i-Carbonell and Frijters, 2004), and for ease of interpretation we choose to present these in Table 2. -6- Table 2. The Importance of Income Comparisons. OLS estimates Important to compare income with other people's income Coefficient Standard errors Household income (omitted: Less than 150 Euros) 150 to 300 Euros 300 to 500 Euros 500 to 1000 Euros 1000 to 1500 Euros 1500 to 2000 Euros 2000 to 2500 Euros 2500 to 3000 Euros 3000 to 5000 Euros 5000 to 7500 Euros 7500 Euros or more -0.101 -0.178** -0.274*** -0.238** -0.288*** -0.314*** -0.332*** -0.338*** -0.323*** -0.338*** [0.148] [0.081] [0.094] [0.089] [0.092] [0.082] [0.089] [0.089] [0.099] [0.110] Occupation (omitted: Elementary professions) Armed Forces Legislators, senior officials and managers Professionals Technicians and associate professionals Clerks Service workers and shop and market sales workers Skilled agricultural and fishery workers Craft and related trades workers Plant and machine operators and assemblers Self-Employed (as opposed to employees) -0.617** -0.088 -0.019 -0.008 -0.045 -0.143 0.557*** 0.016 -0.027 -0.557*** [0.230] [0.121] [0.092] [0.085] [0.104] [0.110] [0.181] [0.104] [0.101] [0.044] Sector (omitted: Industry) Agriculture Services -0.297*** -0.017 [0.100] [0.035] Male -0.010 [0.037] Age (omitted: 16-25) 25-50 Over 50 0.035 -0.163* [0.087] [0.093] Education (omitted: Primary education) Did not complete primary education Lower secondary or second stage of basic Upper secondary Post-secondary, non-tertiary First stage of tertiary Second stage of tertiary -0.003 -0.161* -0.205** -0.109 0.070 0.146 [0.193] [0.091] [0.090] [0.098] [0.097] [0.147] Marital status (omitted: Never in couple) Married Separated Widowed Ever given birth to child -0.111** -0.206*** -0.214* 0.014 [0.047] [0.050] [0.112] [0.042] -7- Citizen 0.028 [0.069] Country dummies (omitted: Austria) Belgium Bulgaria Switzerland Germany Denmark Spain Finland France Great Britain Ireland Netherlands Norway Poland Portugal Russia Sweden Slovenia Slovakia -0.022 0.177 0.026 0.067** 0.190*** 0.594*** -0.073** 0.222*** 0.109*** 0.133*** -0.161*** 0.225*** 0.480*** 0.128** 0.306*** 0.332*** 0.496*** 0.818*** [0.032] [0.106] [0.039] [0.028] [0.039] [0.017] [0.032] [0.035] [0.027] [0.041] [0.024] [0.042] [0.066] [0.051] [0.076] [0.035] [0.030] [0.055] Constant 2.664 [0.152] Observations 15543 R-squared 0.037 Log likelihood -31131 Notes: *, ** and *** denote significance at the 10%, 5% and 1% levels respectively. Standard errors in parentheses are clustered by country. The results show that it is those living in households with lower levels of income who actually compare their incomes more. Those in the sixth to the tenth deciles of household income provide income comparison responses that are on average about onethird of a point lower on the one to six income-comparison scale. The self-employed (who constitute 11% of the sample) compare their incomes significantly less than do employees. In terms of industry, comparisons are less important in Agriculture than in Industry or Services. Those over 50 years old compare less, and the single compare more. There is no significant gender difference in the intensity of income comparisons. The country dummies are very significant. They show that, compared to Austria, respondents in the UK and Ireland, Southern Europe (Spain and Portugal) and most countries of Eastern Europe (Poland, Russia, Slovenia, and Slovakia) attach more weight to comparisons. This ranking of the estimated coefficients -8- on the country dummies is similar to that illustrated in Figure 1. In a separate regression where countries were grouped into more aggregated geographical zones, we find that Scandinavians (Denmark, Finland, Norway, and Sweden) compare less on average than do those living in central Continental Europe (Austria, Belgium, Switzerland, Germany, France and the Netherlands). 3. Who Compares to Whom? The general intensity of income comparisons, described above, is difficult to interpret precisely without knowing to whom people actually compare. If comparison benchmarks are endogenous, as suggested in the literature, is likely that the reference group will depend on the respondent’s age, position in their life cycle, labour market status, and so on. The third wave of the ESS survey includes the following question: “Whose income would you be most likely to compare your own with? Please choose one of the groups on this card: Work colleagues/ Family members/ Friends/ Others/ Don’t compare/ Not applicable/ Don’t know.” The distribution of answers to this question is shown in Table 3. Table 3. “Whose income would you be most likely to compare your own with?” Observations Work colleagues 6 159 Family members 929 Friends 2 382 Others 1 192 Don't compare 5 185 Note: Weighted statistics. Weighted % 38.93 6.03 14.94 7.39 32.72 Of those who identify a comparison group, most people say that they compare to their work colleagues or (to a far lesser extent) their friends. However, 33% of respondents say that they do not compare, which is consistent with the results in Table 1, where 24% of respondents answered that comparisons are not at all important, and 17% chose 1 on the 7-step scale for the importance of income comparisons. This is also consistent with the partial correlation between comparison benchmarks and the general importance of comparisons: those who chose “Don’t compare” in Table 3 also declare that they attach -9- significantly less importance to comparisons: controlling for the same right-hand side variables as in Table 2, the partial correlation coefficient between the two is significantly negative (not shown). Partial correlations also suggest that the greater the intensity of income comparisons, the higher is the probability that the respondent designate colleagues as the comparison group. Regarding the determinants of social comparisons, it appears that, as suggested by the literature, reference groups are correlated with the respondent’s socio-economic situation. Table 4 presents the results of multinomial logit regressions for the probability of reference-group choice. The coefficients represent the probability of choosing the group in question relative to the omitted group (“work colleagues”). Table 4. “Whose income would you be most likely to compare your own with?” Multinomial logit regression. Omitted category: “Work colleagues”. Family members Friends Others Coefficient Std. Err Coefficient Std. Err Coefficient Std. Err Household income (omitted: Less than 150 Euros) 150 to 300 Euros 0.11 0.29 0.06 0.12 0.19 -0.36 300 to 500 Euros 0.20 0.23 0.10 0.00 0.27 0.26 500 to 1000 Euros 0.20 0.14 -0.14 0.29 0.40 0.26 1000 to 1500 Euros 0.43 0.24 0.26 0.17 -0.19 0.32 1500 to 2000 Euros 0.14 0.28 0.18 0.15 -0.25 0.31 2000 to 2500 Euros 0.28 0.23 0.09 0.16 -0.58 0.34 2500 to 3000 Euros 0.39 0.25 0.13 0.17 -0.34 0.33 3000 to 5000 Euros 0.37 0.25 -0.03 0.19 -0.42 0.33 5000 to 7500 Euros 0.27 0.33 0.03 0.20 -0.58 0.32 7500 Euros or more -0.08 0.44 0.06 0.23 -0.51 0.33 Occupation (omitted: Elementary professions) Armed Forces -1.70 1.07 0.30 0.40 0.08 0.58 Legislators, senior officials and managers 0.23 0.04 0.13 0.12 -0.79 0.41 Professionals 0.18 0.11 0.03 0.18 -0.92 -0.30 Technicians and associate professionals 0.16 0.10 -0.05 0.17 -0.83 -0.23 Clerks 0.13 -0.14 0.14 0.21 -0.66 -0.39 Service workers and shop and market sales 0.14 -0.01 0.12 0.20 0.14 -0.41 workers Skilled agricultural and fishery workers -0.26 0.33 0.18 0.30 0.41 0.36 Craft and related trades workers 0.17 0.11 -0.12 0.15 -0.47 -0.24 Plant and machine operators and assemblers 0.20 0.08 0.14 -0.56 -0.21 -0.34 Self-Employed (as opposed to employees) 0.54 0.14 0.93 0.12 0.86 - 10 - 0.13 Sector (omitted: Industry) Agriculture Services Male Age (omitted: 16-25) 25-50 Over 50 Education (omitted: Primary education) Not completed primary education Lower secondary or second stage of basic Upper secondary Post-secondary, non-tertiary First stage of tertiary Second stage of tertiary Marital status (omitted: Never in couple) Married Separated Widowed Ever given birth to child Citizen Country (omitted: Austria) Belgium Bulgaria Switzerland Germany Denmark Spain Finland France Great Britain Ireland Netherlands Norway Poland Portugal Russia Sweden Slovenia Slovakia 0.36 0.08 0.30 0.13 -0.17 0.15 0.15 0.07 0.15 0.03 0.22 0.08 -0.48 0.09 0.07 0.07 0.10 0.09 -0.57 -0.66 0.15 0.15 -0.78 -1.25 0.10 0.15 0.15 0.22 0.14 0.15 0.25 -0.15 -0.17 -0.23 -0.06 -0.01 0.27 0.16 0.18 0.21 0.21 0.31 -0.08 -0.29 -0.27 -0.33 -0.26 -0.04 0.33 0.10 0.08 0.11 0.10 0.16 -0.37 0.13 -0.09 0.05 0.11 0.30 0.25 0.20 0.20 0.17 0.21 0.28 0.31 -0.12 -0.20 0.09 0.18 0.34 -0.32 -0.27 0.09 0.08 0.08 0.23 0.01 -0.01 0.07 0.11 0.13 0.31 0.24 0.11 -0.12 0.09 0.20 0.09 -0.26 0.16 0.21 0.11 0.17 0.21 -0.17 -1.36 -0.35 -0.56 -0.46 0.32 0.36 -0.02 0.10 0.22 -0.36 -1.08 0.47 -0.38 -0.43 -0.73 -0.40 -0.90 -0.61 0.08 0.19 0.09 0.08 0.09 0.05 0.07 0.07 0.07 0.07 0.04 0.11 0.14 0.07 0.17 0.07 0.06 0.15 0.31 -0.39 0.12 0.08 -0.23 -0.37 0.10 -0.12 -0.72 0.11 -0.10 -0.36 -0.43 0.67 -1.41 0.16 -1.12 -0.60 -0.90 0.25 0.07 0.12 0.09 0.06 0.08 0.04 0.06 0.06 0.07 0.08 0.05 0.09 0.10 0.04 0.09 0.06 0.04 0.08 0.22 0.56 1.26 0.07 0.64 0.12 0.59 0.69 0.25 0.31 0.16 0.35 0.02 0.77 -0.33 -0.67 -0.43 -0.26 0.09 -2.20 0.06 0.19 0.07 0.06 0.06 0.03 0.06 0.07 0.05 0.07 0.04 0.07 0.13 0.07 0.17 0.06 0.06 0.14 0.38 - 11 - Observations 10374 Pseudo R2 0.06 Log Likelihood -10783 Notes: Multinomial logit regressions of the answer to the question “Whose income would you be most likely to compare your own with?” The possible answers are Work colleagues (the omitted category), Family members, Friends, Others, Don’t compare, Not applicable, and Don’t know. The regression omits those who say that they “Don’t compare”. Standard errors in parentheses are clustered by country. Bold coefficients are significant at the five per cent level. Table 2 suggested that the self-employed compare less in general. Table 4 nuances this result: it shows that they are significantly less likely to compare to their colleagues, but more likely to compare to family members, friends and “others”. Men compare less to family members than do women. Comparisons to colleagues (and “others”) increase after the age of 25, whereas the opposite is true of comparisons to family members and friends. Higher levels of education are positively correlated with comparisons to colleagues, but reduce comparisons to friends. The married compare more to colleagues and family members but less to friends. Respondents with children compare more to family members. These results are consistent with the choice of comparison benchmark being closely related to the type of regular social interactions that are experienced. In terms of geography, those in central Continental Europe compare much more to their colleagues than do respondents elsewhere. Spaniards and Irishmen compare significantly more to family members, as do Scandinavians. Those in Eastern Europe compare significantly less to their family than elsewhere. Esping-Andersen (1999) has proposed a classification of European countries in terms of their system of social protection. Southern European countries appear to rely much more heavily on family solidarity than other groups of countries. The primary importance of comparisons with family members in this group of countries is consistent with this classification. To go one step further, we also used another question in the survey: “How often do you meet socially with friends, relatives or work colleagues?”. The distribution of answers to this question is shown in the Appendix. The more often people declare that they meet socially, the more likely they are to declare that they compare to friends, and the less likely they are to compare to colleagues or family members (and to “others”). Income comparisons are also less important for those who meet socially more often. - 12 - In summary, when talking about “income comparisons”, what most people have in mind is comparisons to colleagues, rather than to other non-professional groups. But beyond this general finding, comparison benchmarks are also endogenous: people compare to those that they meet the most often. We now ask whether all of the comparison groups are identical in terms of their effects of subjective well-being. 4. The Effect of Comparisons on Well-Being Comparisons are important, and comparison benchmarks vary according to the individual’s situation in the labour market and society. But what is the well-being effect of comparisons? To investigate, we analyse the relationship between life satisfaction, on the one hand, and the intensity and direction of comparisons on the other. Respondents answered the question “How happy are you?” by ticking one response on an 11-step ladder, where 0 was labelled “extremely unhappy” and 10 “extremely happy”. The overall distribution of responses is shown in Table 5 below. Table 5. “How happy are you?” Extremely unhappy 1 2 3 4 5 6 7 8 9 Extremely happy Total Observations 115 142 280 545 642 2142 1893 4395 6836 4200 2306 Weighted % 0.49 0.6 1.19 2.32 2.73 9.12 8.06 18.7 29.09 17.88 9.81 23496 100 Note: Weighted statistics. We first check that the life satisfaction function is “well-behaved” in terms of the usual correlates of subjective well-being (see Clark et al., 2008b). Self-declared happiness - 13 - was indeed higher for richer, younger people, the married, women (although not significantly so in Southern and Eastern Europe) and for more-skilled workers. The ranking of regions in terms of happiness is, in descending order: Scandinavia, the British Isles, Continental Europe, Southern Europe, and finally, Eastern Europe. In order to make the results more directly interpretable, we sometimes appeal to an alternative specification of the happiness estimate: we dichotomize the happiness variable, coding those who declare that they stand at over 7 on the 11-step happiness ladder as 1 (56% of the sample), and the rest as 0. We call this new dichotomized variable “high happiness”. 4.1 Those who compare more are less happy Table 6 illustrates the significantly negative impact of comparison intensity on subjective happiness. Comparisons here are measured cardinally from 0 to 6; using a dummy variable for comparisons being important produces the same qualitative results. The regression estimates using the dependent variable “high happiness” and a dummy for comparisons being “very important” (i.e. over 4 on the 7-step ladder) show that those saying that comparisons are “very important” are 7% less likely to be very happy. We can check this negative relationship between the intensity of comparisons and happiness by using a number of different notions of well-being present in the survey, namely “satisfaction with life as a whole”, “satisfaction with how life turned out so far”, subjective general health, “feeling of depression during the past week”, “feeling always optimistic about my future”, as well as more economic notions of satisfaction such as “satisfaction with one’s standard of living”, “difficulty of living with present income”, “satisfaction with job” and feeling that one “gets paid appropriately, considering efforts and achievements” (see the Appendix for the descriptive statistics regarding these variables). As shown in Table 6, the greater is the importance that people attach to comparisons, the lower they rank on these different satisfaction scales. - 14 - Table 6. The Welfare Effect of Comparison Intensity. OLS Estimates of Various Subjective Satisfaction Variables How happy are you Important to compare Observations R-squared -0.095*** [0.013] 15502 0.178 How satisfied with life as a whole -0.122*** [0.015] 15520 0.234 Satisfied with how life turned out so far Health Feeling optimistic -0.074*** [0.011] 15523 0.176 -0.012*** [0.004] 15534 0.156 -0.039*** [0.004] 15510 0.058 Felt depressed, how often past week Satisfied with How satisfied with Paid Living standard of job appropriately comfortably living Important to compare 0.021*** -0.113*** -0.124*** -0.126*** -0.039*** [0.003] [0.010] [0.013] [0.008] [0.003] Observations 15496 15530 15523 15487 15524 R-squared 0.086 0.302 0.085 0.134 0.377 Notes: Each column corresponds to a separate regression. Controls: income, occupation, self-employed, sector, age, education level, marital status, children, citizen dummy, and country dummies. *, ** and *** denote significance at the 10%, 5% and 1% levels respectively. Standard errors in parentheses are clustered by country. It is of interest to note that four of the groups who compare their income the most in Table 2 (those with lower income, employees, those aged under 50, and singles) are often found to report lower levels of well-being in empirical analyses. This is entirely consistent with Table 6’s suggestion that such income comparisons render individuals unhappy on average. It is thus possible that the degree of comparison intensity be an omitted variable that underlies some of the common correlations in the well-being literature. 4.2 Which comparisons hurt the most? With reference to those who declare that they do not compare, all comparisons are welfare-reducing, but not equally so: the results in Table 7 show that comparisons to friends and “other” are twice as painful as comparisons to colleagues and family. The same results hold after dropping those who do not compare: those who declare that they compare to colleagues are significantly happier than those who choose friends (or “others”) as their benchmark. Those who declare that they compare to their colleagues are 4% more likely to declare that they are very happy than those who choose their friends as a benchmark (Table 7). This actually nuances the above finding that increased social activity reduces the importance of income comparisons. While this is true, the - 15 - remaining comparisons are more likely to be friends, and this is the most painful type of comparison for well-being. - 16 - Table 7. Happiness and comparison benchmarks. OLS estimates of self-declared happiness. Whose income would you be most likely to compare your own with? Work colleagues 0.138*** [0.040] 0.143* [0.081] Observations R-squared log likelihood -0.169*** [0.034] -0.162* [0.088] -0.313*** [0.048] -0.406*** [0.088] 15421 0.172 -29050 Omitted category Don’t compare Friends Family members Friends Others -0.104* [0.051] 10348 0.179 -19254 Notes: Column 1: regression over the entire sample. Omitted category: “Don’t compare”. Column 2 drops those who reply that they “don’t compare”. Omitted category: “Friends”. Controls: income, occupation, self-employment, sector, age, education level, marital status, children, citizen dummy, and country dummies. Standard errors in parentheses are clustered by country. *, ** and *** denote significance at the 10%, 5% and 1% levels respectively. 4.3. Heterogeneity Are these effects the same for everybody, or is there some heterogeneity in the welfare effect of comparisons? To elucidate, we interacted the answer to the comparison questions with dummy variables indicating the socio-professional status of respondents. As Ai and Norton (2003) have shown that the coefficients in interactions in ordered logit or probit models are not easily interpretable, we ran a simple OLS estimate of happiness on the interacted terms, controlling for the usual socio-demographic variables and main effects. Regarding the impact of the intensity of comparisons, the resulting estimates show that comparisons are less painful for the youngest group (aged up to 25), those in higher professions (senior managers), and service workers and machine operators (as compared to elementary occupations). They are also less painful in the UK and Ireland. None of the other interactions (with sex, income and employment status - self-employed versus employee) were significant. - 17 - Turning to the impact of different directions of comparisons, interaction effects suggest that comparisons to colleagues are particularly painful for the older (the coefficient on the interaction with the 25-50 age category is -0.152 [0.086] and that with the 51-65 category is -0.164 [0.082]) and the poorer (the coefficient on the interaction with those who earn less that the median wage is -0.194 [0.052]), but less so in higher occupations. Income comparisons to family members are more painful for Southern Europeans and older people (over 50 years). Income comparisons to friends are significantly more painful in Great Britain than in Continental Europe; they are significantly less painful for those aged over 50 (which contrasts with comparisons to colleagues), and for certain occupation categories such as professionals and service workers. The interaction terms were never significant for comparisons to “others”. This set of findings rules out the idea that people endogenously choose their comparison benchmark in order to maximize their well-being. For instance, we found that Southern Europeans compare more to family members even though this is the most painful type of comparison for them. People seem to compare to those they interact with often, rather than those to whom they enjoy comparing their own outcomes. These results concerning the welfare impact of the intensity and the direction of comparisons also naturally raise the question whether those for whom comparisons are very important are less happy when they compare to a certain group rather to another. Experiments along theses lines did not reveal any interesting results: comparison intensity reduces well-being pretty much independently of the reference group. 5. Income comparisons and the demand for redistribution Do painful comparisons make people more favourable to compensating measures, such as income redistribution? Here we use the following question: Using this card, please say to what extent you agree or disagree with each of the following statement: “The government should take measures to reduce differences in income levels”. - 18 - Table 8. “The government should take measures to reduce differences in income levels” Frequency Strongly agree Agree Neither agree nor disagree Disagree Strongly disagree Weighted % 6433 9709 3655 2892 602 27,62 41,69 15,69 12,42 2,59 Note: Weighted statistics. Sample: age between 16 and 66. A preliminary simple estimate of the demand for income redistribution by the State showed that Scandinavians are less in favour of income distribution, as compared to those in Continental Europe, while Southern and Eastern Europeans are more in favour. The richer categories, those in higher occupations, the self-employed, men, the younger, the married, the more educated and those who have never had children are less in favour of income distribution. Table 8 shows that the more people consider income comparisons to be important, the more they are in favour of income redistribution by the State. However, excluding those who declare that they “don’t compare”, Table 9 also shows that it is comparisons to family members and “others”, much more than to colleagues, that prompt the demand for income redistribution. - 19 - Table 9. Comparisons and the demand for income redistribution by the State. OLS estimates of the demand for redistribution. Important to compare income with other people's income 0.027*** [0.008] Whose income would you be most likely to compare your own with? Omitted category: “others” Work colleagues -0.074** [0.033] Family members 0.019 [0.047] Friends -0.074 [0.046] Observations 15411 Observations 10307 R-squared 0.150 R-squared 0.158 Log likelihood -21689 Log likelihood -14329 Notes: Column 1 corresponds to a regression over the entire sample; Column 2 drops those who reply that they “don’t compare”. Omitted category: “others”. Dependent variable: Agree that “The government should take measures to reduce differences in income levels”. Other controls: income, occupation, self-employed, sector, age, education level, marital status, children, citizen dummy, and country dummies. Standard errors in parentheses are clustered by country. *, ** and *** denote significance at the 10%, 5% and 1% levels respectively. The demand for redistribution therefore seems to follow the same patterns as the welfare effects of comparisons: the intensity of comparisons reduces subjective happiness and increases the demand for redistribution. Comparisons to colleagues are the least painful and also produce the weakest correlations with the demand for income redistribution. 6. Conclusions This paper has analysed unprecedented direct evidence on the intensity and direction of income comparisons. Income comparisons are acknowledged as important by the majority of Europeans. They are associated with lower levels of self-declared happiness, and this negative impact increases with the age of the respondent but falls with their income. Colleagues are the most frequently-cited reference group. They are also the - 20 - most innocuous, i.e. that which produces the smallest fall in happiness; comparisons to friends are much more damaging for happiness. Cultural differences are apparent: the negative welfare effect of comparisons is larger in Continental countries and smaller in the British Isles, which is also the region in which comparisons most frequently concern colleagues. Southern Europeans seem to be more family-oriented: they compare more to family members and suffer more from this type of comparisons than do inhabitants in other parts of Europe. Income comparisons are also associated with a greater demand for income redistribution. It is comparisons to family members and “others”, much more than comparisons to colleagues, that prompt the demand for income redistribution. Finally, the choice of comparison benchmarks seems to be endogenous, being closely related to the type of regular social interactions that respondents experience. For example, those who “socialize” more compare less to their colleagues. The above empirical findings are redolent of an information effect à la Hirschman (1977). Senik (2004, 2008) and Clark et al. (2009a) have produced evidence consistent with professional reference groups acting as a source of information regarding one’s future prospects, as well as being a simple benchmark for comparisons. This reflects that individuals in these groups share the same productive characteristics as the respondent, and hence the same expected earnings profile. This is obviously less true of other non-professional groups, such as family members or friends. This could explain why people mostly compare to their colleagues (in order to acquire information about their professional future) and why these comparisons are not as damaging to welfare as comparisons to family members, friends or “other” groups. - 21 - References Ai, C. and Norton, E. (2003). "Interaction Terms in Logit and Probit Models". Economics Letters, 80, 123-129. Bygren, M. (2004). "Pay reference standards and pay satisfaction: what do workers evaluate their pay against?". Social Science Research, 33, 206-224. Cappelli, P. and Sherer, P. (1988), “Satisfaction, Market Wages and Labour Relations: An Airline Study”, Industrial Relations, 27, 56-73. Clark A.E., Diener E., Georgellis T. and Lucas R. (2008b), “Lags and Leads in Life Satisfaction: A Test of the Baseline Hypothesis”, Economic Journal, 118, F222-243. Clark A.E., Kristensen N. and Westergård-Nielsen N. (2009a), “Job Satisfaction and Co-worker Wages: Status or Signal?”, Economic Journal, 119, 430-447. Clark, A.E. and Oswald, A (1996), “Satisfaction and comparison income”, Journal of Public Economics, 61, 359-81. Clark, A.E., Etilé, F., Postel-Vinay, F., Senik, C., and van der Straeten, K. (2005). "Heterogeneity in Reported Well-Being: Evidence from Twelve European Countries". Economic Journal, 115, C118-C132. Clark, A.E., Frijters, P. and Shields, M. (2008a). "Relative Income, Happiness and Utility: An Explanation for the Easterlin Paradox and Other Puzzles". Journal of Economic Literature, 46, 95-144. Clark, A.E., Masclet, D. and Villeval, M.-C. (2009b). "Effort and Comparison Income". Industrial and Labor Relations Review, forthcoming. Esping-Andersen G., 1999 (first edition 1990), Les trois mondes de l’État-providence, PUF. Falk, A. and Knell, M. (2004), “Choosing the Joneses: Endogenous goals and reference standards”, Scandinavian Journal of Economics, 106, 417-435. - 22 - Falk, A., and Ichino, A. (2006). "Clean Evidence on Peer Pressure". Journal of Labor Economics, 24, 39-57. Fehr, E., and Schmidt, K. (1999). "A Theory of Fairness, Competition and Cooperation". Quarterly Journal of Economics, 114, 817-868. Ferrer-i-Carbonell, A. (2005), “Income and well-being: An empirical analysis of the comparison income effect”, Journal of Public Economics, 89, 997-1019. Ferrer-i-Carbonell A. and Frijters P. (2004), “How Important is Methodology for the Estimates of the Determinants of Happiness?”, Economic Journal, 114, 641-659. Helliwell, J.F., and Huang, H. (2009). "How’s the Job? Well-Being and Social Capital in the Workplace". Industrial and Labor Relations Review, forthcoming. Heston A., Summers R. and Aten B., “Penn World Table Version 6.2”, Center for International Comparisons of Production, Income and Prices at the University of Pennsylvania, September 2006. Hirschman A., with Rothschild M. (1973), “The Changing Tolerance for Income Inequality in the Course of Economic Development”, Quarterly Journal of Economics, 87, 544-566. Knight, J. and Song, L., 2006, “Subjective well-being and its determinants in rural China”, University of Nottingham, mimeo. Luttmer E. (2005), “Neighbors as Negatives: Relative Earnings and Well-Being”, Quarterly Journal of Economics, 120, 963-1002. MacBride M. (2001). “Relative-Income Effects on Subjective Well-being in the Crosssection”, Journal of Economic Behavior and Organization , 45, 251-278. McBride, M. (2006). "An Experimental Study of Happiness and Aspiration Formation". University of California-Irvine, Mimeo. Senik C. (2004), "When Information Dominates Comparison. Learning from Russian Subjective Panel Data”, Journal of Public Economics, 88, 2099-2133. - 23 - Senik C. (2008), “Ambition and Jealousy. Income Interactions in the “Old Europe” versus the “New Europe” and the United States”, Economica, 75, 495-513. Senik C. (2009), “Direct Evidence on Income Comparisons and their Welfare Effects”, Journal of Economic Behavior and Organization, forthcoming. - 24 - Appendix. Weighted Descriptive Statistics Sample: respondents aged 16-65 years old. Observations Percentage Household income Less than 150 Euros per month 150 to 300 Euros 300 to 500 Euros 500 to 1000 Euros 1000 to 1500 Euros 1500 to 2000 Euros 2000 to 2500 Euros 2500 to 3000 Euros 3000 to 5000 Euros 5000 to 7500 Euros 7500 Euros or more 952 1221 1765 2459 2375 2419 2407 2478 4387 2028 1101 4.04 5.17 7.48 10.42 10.07 10.25 10.20 10.50 18.60 8.59 4.67 Occupation Armed forces Legislators senior officials and managers Professionals Technicians and associate professionals Clerks Service workers and shop and market sales workers Skilled agricultural and fishery workers Craft and related trades workers Plant and machine operators and assemblers Elementary occupations 71 1869 3245 3853 2338 3427 625 2709 1704 2050 0.33 8.54 14.82 17.60 10.68 15.65 2.85 12.38 7.78 9.36 Sector (NACE) Agriculture Industry Services 871 5815 14948 4.03 26.88 69.10 Marital status Married Separated Widowed Never in couple 13901 2244 579 6768 59.17 9.55 2.46 28.81 Had child Citizen 23294 23321 68.25 96.14 Father’s employment status when resp. was 14 Employee Self-employed Not working Father dead/ absent Note: Weighted statistics. 16149 4471 791 1354 70,94 19,64 3,47 5,95 - 25 - Weighted number of observations per country/regions European regions : Continental British Austria Belgium Switzerland Germany France Netherlands United Kingdom Ireland Portugal Spain Norway Denmark Finland Sweden Bulgaria Poland Russia Slovenia Slovakia Note: Weighted statistics. 1276 1272 1159 1700 1487 1391 Southern Scandinavian Eastern 1491 1081 847 915 1449 1082 1374 1481 878 1185 1670 959 895 Subjective Variables Obs. Mean Std. Dev. Min Max Happy Comparisons are important I compare to colleagues I compare to family I compare to friends I compare to others I compare to other people Redistribute income Often meet socially with friends. relatives or colleagues Satisfied with life as a whole Satisfied with how life turned out so far Health Feel optimistic about future How often felt depressed past week Satisfied with standard of living Satisfied with job Paid appropriately considering efforts and achievements Living comfortably on present income? Note: Weighted statistics. 23235 16045 15959 15959 15959 15959 15959 23027 23257 23274 23275 23305 23241 23202 23284 16182 16133 23276 7.38 2.30 0.39 0.06 0.15 0.07 0.33 3.79 5.12 7.04 7.06 3.90 3.71 1.45 6.73 7.16 3.08 3.06 1.89 1.83 0.49 0.24 0.36 0.26 0.47 1.06 1.50 2.23 1.93 0.85 0.90 0.67 2.25 2.09 1.14 0.86 0 0 0 0 0 0 0 1 1 0 0 1 1 1 0 0 1 1 10 6 1 1 1 1 1 5 7 10 10 5 5 4 10 10 5 4 - 26 - “How often do you meet socially with friends, relatives or work colleagues?” Never Less than once a month Once a month Several times a month Once a week Several times a week Every day Note: Weighted statistics. Observations Weighted Percentage 109 831 1 188 2 998 3 052 5 028 2 618 0.68 5.41 7.60 18.96 19.35 31.42 16.58 - 27 -