The Journal of Socio-Economics 39 (2010) 8188

Contents lists available at ScienceDirect

The Journal of Socio-Economics

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

Real household income and attitude toward immigrants: an empirical analysis

Leonardo Becchetti a, , Fiammetta Rossetti a , Stefano Castriota b
a b

University of Tor Vergata, Rome EURICSE, Trento, Italy

a r t i c l e

i n f o

a b s t r a c t
Previous studies have analyzed the (aggregate) effects of unemployment on attitudes towards immigrants and on right-wing crimes. In this paper, we investigate the effects of economic prosperity on attitudes towards immigrants, focusing not only on unemployment status but also on real household income. Using panel data from the German Socioeconomic Panel on around 33,000 individuals over the period 19922004 we nd a robust negative relationship between real personal household income and self-declared concern about immigrants, both in levels and rst differences. Both job loss and income reduction concerns about immigration. Our ndings document an interesting interaction between economic variables and social attitudes which does not depend on economic growth per se but on its capacity to generate higher economic wellbeing at individual level, not only for unemployed people but also for those in employment, who may face a fall in real income during economic downturns. 2009 Elsevier Inc. All rights reserved.

Article history: Received 15 December 2008 Received in revised form 25 June 2009 Accepted 22 July 2009 JEL classication: D64 I31 O15 Z13 Keywords: Social capital Morality Ethics Tolerance Immigration Household income

1. Introduction [I]t is in the progressive state, while the society is advancing to the further acquisition, rather than when it has acquired its full complement of riches, that the condition of the great body of the people, seems to be the happiest and the most comfortable. It is hard in the stationary, and miserable in the declining state. The progressive state is in reality the cheerful and the hearty state to all the different orders of society. The stationary is dull; the declining melancholy. Adam Smith, Wealth of Nations Evaluating the non-economic consequences of economic choices is extremely important for the correct denition of policy objectives. In fact, moral values, social attitudes and individual preferences are crucially oriented by the former. Many theoretical and empirical studies highlight that economic outcomes affect values, or that the same market structure and economic growth may produce alternative cultural values (such as antagonism and selfinterest) as side products, thereby interacting with civic virtues and helping to strengthen or deplete the moral fabric.

In this regard, two opposite theoretical schools of thought maintain that economic growth has respectively positive or negative consequences on moral values. A rst argument emphasizing the negative consequences of economic growth on moral values is the well-known trade-off between comfort and stimulation (Scitowsky, 1976), where afuence generated by economic growth increases individual comforts. This, in turn, dampens incentives to build those values and virtues which require effort to be maintained and strengthened. A second argument has been put forward by Hirsch (1976) who holds a well-known critical position on the moral consequences of market economy and economic growth. In his book entitled Social Limits to Growth, Hirsch identies three main negative effects of market economies on moral values: the tyranny of small decisions, the commercialization bias, and the depleting moral legacy. The rst is related to the classical coordination failure problem.1 The second refers to a typical Marxian argument which holds that in a free market economy everything, including moral values, becomes the object of exchange, generating

Tel.: +39 06 36300723; fax: +39 06 2020500. E-mail address: Becchetti@economia.uniroma2.it (L. Becchetti). 1053-5357/$ see front matter 2009 Elsevier Inc. All rights reserved. doi:10.1016/j.socec.2009.07.012

1 Individual choices, each made separately and thereby necessarily without taking account of the interaction between them, combine to have destructive social consequences. These consequences are destructive in the sense that they produce a worse result for the individual concerned than could have been obtained by coordination of individual choices with some method that took account of the mutual interaction. (Hirsch, 1976, p. 37).

82

L. Becchetti et al. / The Journal of Socio-Economics 39 (2010) 8188

corruption and venality and deteriorating the societys moral fabric. Thirdly, social morality is a legacy of the precapitalist and preindustrial past (Hirsch, 1976, p. 117) which is necessary if economic transactions have to work. Market economies, especially since industrialization, are characterized by negative values like individualism and avarice, and by negative social framework due to anonymity, mobility of workers, and the like, which deplete that legacy. An opposite school of thought challenges this view by emphasizing that economic growth has also benecial consequences in terms of higher tolerance and openness (Friedman, 2005). According to Friedman, Economic growth meaning a rising standard of living for a clear majority of citizens more often than not fosters greater opportunity, tolerance of diversity, social mobility, commitment to fairness and dedication to democracy. And conversely, when there is economic stagnation or decline the citizens moral character tends to decline accordingly, there being less tolerance, less openness, and less generosity to the poor and the disadvantaged. Intuitively, Friedmans contention has a sound economic rationale. A large part of public opinion believes that economic activity is a zero sum game in which the aggregate payoffs of losers and winners must necessarily be clear.2 In fact, from an algebraic point of view, this is true only when the GDP growth rate is non-positive. If the zero sum game is a crucial cause of worry, it is reasonable to assume that such worry is higher in stagnation periods, when economics is effectively a zero or a negative sum game in the aggregate,3 than it is in expansion periods, when the size of the cake grows and everyone can, in principle, have a larger slice of it.4 From an empirical viewpoint, a number of works have studied the determinants of tolerance towards immigrants, but to our knowledge those analyzing the relevance of economic variables have focused on the unemployment status, without considering the role played by real household income, which is the original contribution that this paper seeks to make. Peoples attitudes are the results of individual inherited traits, family background and education, age, the number of immigrants present in a certain region, the immigration policies adopted by the government, and the economic situation, which may make natives perceive immigrants as a threat to their employment situation or economic welfare. The relevance of mentality as the product of inherited traits and cultural background has been well described by Dustmann and Preston (2007), who explain the factors determining individuals opinions in regard to future immigration. They use three types of question on the perception of foreigners by native respondents in the BSAS: (i) questions related to race, (ii) questions related to the labor-market impact of foreigners, and (iii) questions related to the impact of foreigners on the economys welfare. Dustmann and Preston nd that racist attitudes constitute the main determinant

of opposition against future immigration, whereas labor market or welfare considerations are less decisive factors. The role of background was highlighted by one of the rst studies in this eld. The research was conducted in the 1940s in the United States by Adorno et al. (1950), who assumed that the characters structure is inuenced by the environment where a person has grown up: economic and social background, own and parents education, and so on, are considered the most inuential factors in the personalitys development. The importance of education for tolerance has been conrmed, among others, by Fertig and Schmidt (2002) with German data, and by Bauer et al. (2000) using ISSP data. Age is another important driver of attitudes towards foreigners: using data on Germany from the 1988 wave of the Eurobarometer, Gang and Rivera-Batiz (1994) found that students have the most positive attitudes and retired people the most negative ones. Tolerance is also inuenced by the concentration of ethnic minorities in certain areas, because it increases the hostility of native respondents towards these groups (Dustmann and Preston, 2000b; Gang et al., 2002), and by the immigration policies adopted by governments, since in countries with a more skill-based immigration policy (e.g. Canada) respondents tend to have a more positive attitude towards immigrants (Bauer et al., 2000). The role of economic variables is still a source of debate, given that studies on individual attitudes have not demonstrated a clear positive relationship between unemployment and resentment against immigrants, while studies on the role of real income are lacking (Falk and Zweimueller, 2005). Gang et al. (2002), using data from the Eurobarometer on several European countries over the period 19881997, nd that negative attitudes are affected by the degree of competition in the labor market with immigrants, although they do not nd any difference between employed and unemployed people. Negative effects of unemployment on tolerance have been found with German data by Gang and Rivera-Batiz (1994) and Bacher (2001a, 2001b). By contrast, no effect of unemployment on tolerance was found by Bauer et al. (2000), while the negative effects determined by Fertig and Schmidt (2002) become insignicant in the structural model. Our purpose in what follows is to shed more light on this muchdebated issue, our original contribution being the analysis of the effect of changes in real income on attitudes towards immigrants. During economic downturns, a signicant share of the population registers a fall in family income. Clearly, the most damaged in relative terms are those who become unemployed; but an analysis of the effect of diminishing purchasing power is necessary. In fact, Falk and Zweimueller (2005) show that unemployment is a driver of right-wing extremist crimes which are committed mainly by young males. However, real household income drops are experienced not only by those who lose their jobs, but also by many other individuals. These latter may express their frustration not through violent actions but rather through democratic votes in favor of less tolerant political parties, thereby depleting the societys moral fabric.

2 The idea of economics as a zero sum game nds an analogy in the second law of thermodynamics on the conservation of energy, which states that the total energy going into a system must equal the total energy coming out of it, and cannot be created or destroyed. 3 This is true at least from a static point of view when it is not considered that, even under constant GDP, a change in the share of investment may be crucial for raising the growth rate in the future. Furthermore, the nature itself of market transactions is such that, even in a zero growth economy, market trades increase individual wellbeing in proportion to the sum of consumer and producer surpluses. The zero sum game argument is therefore valid only in a static perspective and if the focus is exclusively on observable economic payoffs. 4 On translating our reasoning into terms of the happiness effects of material wellbeing, relative income and hedonic adaptation may make economic activity a zero sum game even in the presence of growth. In fact, if both per capita GDP and peoples expectations increase by the same amount, the overall populations wellbeing is unchanged. Therefore, the only way to increase ones satisfaction is by increasing ones relative income, which can happen only is somebody elses relative income diminishes.

Fig. 1. The distribution of self-declared concern about foreigners.

L. Becchetti et al. / The Journal of Socio-Economics 39 (2010) 8188

83

Fig. 2. The distribution of changes in self-declared concern about foreigners.

The paper is divided into four sections (introduction and conclusions included). Section 2 describes our database and comments on the descriptive evidence. Section 3 presents and discusses the econometric ndings from different specications in which we regressed (changes of) self-declared concern about immigrants on (changes of) various measures of aggregate and individual socioeconomic conditions. Section 4 concludes. 2. Empirical analysis The data source for our empirical analysis is the German SocioEconomic Panel (GSOEP). This panel is unbalanced and covers the 19842004 period. However, we consider only 13 waves, from 1992 to 2004, because the tolerance question on which we focus was added to the survey only in 1992. The dataset in its original form was organised into separated blocks of variables for each survey year. The nal dataset was the result of large-scale selection and redenomination of variables in order to render them homogeneous across survey years. The dynamic panel was built by appending the 13 waves. The additional data le that the GSOEP offers to integrate base survey data with more accurate information on household income and education was used to improve the quality of our information. The dataset used for the analysis is an unbalanced panel with 32,880 different individuals for a total of 168,626 observations. Our dependent variable is based on the following question What is your attitude towards immigration to Germany. Are you concerned about it? for which the possible answers are: (i) Very Concerned, (ii) Somewhat Concerned and (iii) Not Concerned At All. We rescaled this measure in order to have 3 for Very Concerned and 1 for Not Concerned At All, giving the variable a more intuitive interpretation also in the estimation results. An important question is whether our variable is a good proxy for the concept of tolerance. In its Declaration on the Principles of Tolerance UNESCO denes tolerance as respect, acceptance and appreciation of the rich diversity of our worlds cultures, our forms of expression and ways of being human. Tolerance is harmony in difference.5 Although a single indirect question cannot capture the richness of the concept, we would point out that concern about immigrants is very likely to be negatively correlated with acceptance of cultural diversity, which forms the main part of the UNESCO denition. More generally, however, we can argue that concern about migration is denitely related to the Friedmans more general

Fig. 3. Dynamics of the share of respondents with the maximum level of concern about foreigners.

statement which relates economic decline to less tolerance, less openness, and less generosity to the poor and the disadvantaged. Fig. 1 shows the distribution of responses across the entire sample period. We observe that 23% of respondents declared themselves to be Not Concerned At All, 48% of them to be Somewhat Concerned, while 29% were Very Concerned. Fig. 2 illustrates the distribution of changes in concern about immigration in the sample period. Note that variation in concern is quite marked and involves more than 40% of sample respondents (around 37% onerung changes and around 4% two-rung changes). Fig. 3 shows the evolution across years of the share of individuals Very Concerned about foreigners for the overall sample and for German natives only: the aggregate level of concern appears to be quite stable over time. Figs. 4 and 5 repeat the exercise for, respectively, the individuals Somewhat Concerned and Not Concerned At All, with similar results. For a preliminary inspection of our dependent variable we calculated year-by-year transitions to different states of concern about

5 The Merriam Webster Dictionary denes tolerance, not very differently and more succinctly, as sympathy or indulgence for beliefs or practices differing from or conicting with ones own.

Fig. 4. Dynamics of the share of respondents with the average level of concern about foreigners.

84

L. Becchetti et al. / The Journal of Socio-Economics 39 (2010) 8188

Fig. 5. Dynamics of the share of respondents with the minimum level of concern about foreigners.

Table 1 Transition matrix of one-year changes in respondents self-declared concern about foreigners. Year t Year t 1 Not Concerned At All Somewhat Concerned Very Concerned Not Concerned At All 55.02 17.83 7.10 Somewhat Concerned 35.06 60.89 35.82 Very Concerned

9.92 21.28 57.08

foreigners (see Table 1). As expected, persistence in a given level of self-declared concern is quite high (between 50 and 60% and higher in the intermediate response), but variability of the dependent variable for the same individual across years is not negligible, with a 35% probability of passing from an extreme state to the intermediate one. Much lower, and below 10%, are the probabilities of a sudden change from one extreme to the other extreme state. Overall, the transition probabilities conrm that there is substantial variation of self-declared concern, whose causes should be investigated. Table 2 reports the list of variables considered in our regressions with corresponding summary statistics. Among the possible macro variables we select the 3-year percent changes in the real per capita GDP at constant prices and the 3-year unemployment and ination rates, the underlying assumption being that people take time to react to changes in the macroeconomic situation. Data on GDP and unemployment are taken from Econstats and those on ination from the IMFs International Financial Statistics. The macroeconomic situation in our sample period is rather gloomy, and the average GDP growth rate is below 1.5%. A problem arises when using macroeconomic data like GDP growth rates, ination and unemployment because their variation in our sample is very limited (only 13 observations, 1 for each year)
Table 2 Descriptive statistics. Variable Concern about foreigners Yearly change in concern about foreigners German nationality Age Male Education (in years) Married Separated Single Divorced Widowed Unemployed Loss of job in the current year Real household income GDP growth (3-year moving average) Ination rate (3-year moving average) Unemployment rate (3-year moving average) Observed 168,626 132,834 91,153 168,623 168,626 164,268 160,509 160,509 160,509 160,509 160,509 167,983 145,047 161,760 168,626 168,626 168,626

against a total sample size of around 168,000 observations on the level of concern. This implies that statistical inference from the effects of these variables is poor and that their coefcients in econometric estimates may be unstable (see Frey and Stutzer, 2000). For this reason we decide to rely more on the one-year percent change in real household income to test our hypothesis on the effects of growth on tolerance, since this is measured at the individual level, and to use the macroeconomic variables mainly as a control in a subset of our specications. In our regressions we control for gender, age, nationality, education and marital status. With regard to employment status we create a dummy variable (Unemployed) which is equal to 1 if an individual is unemployed in a given year and 0 otherwise. Furthermore, we generate a dummy variable (Loss of job) to take account of also negative variations in the employment status: the variable takes value 1 if a person lost his/her job in the given year and 0 otherwise. Table 3 shows the correlation between the two dependent variables of interest, level of concern and change of level of concern, and the set of control variables that we use in our regressions. Inspection of the rst column shows that German nationality, age, being married, and unemployment status are positively correlated with the level of concern about immigration, while single status, higher education and income have a positive correlation with tolerance. Similar ndings apply for the change in the level of concern (except for unemployment status and education, which are no longer signicant). Note that both levels and changes in real household income are signicantly and negatively correlated with concerns about foreigners, while only the second variable is still negative and signicantly correlated with changes in such concern. Tables 4 and 5 treat the issue from a slightly different perspective because they show the distribution of selected variables across changes in self-declared concern about immigrants. In Table 4, we observe that the share of Germans, married persons, and individuals losing their jobs with the highest level of concern is markedly greater than that of the overall sample. The main difference in the marital status variables is in the share of those with the lowest level of self-declared concern. In this regard, singles and separated respondents exhibit a higher value than married ones (28.89 and 25.70, respectively, against 23.51). More relevant to the topic of this paper is that people with greater concern about foreigners declare lower real household incomes (the average monthly income for those with the highest level of concern is 2773 (1984) Dmarks, more than 200 Dmarks lower that that (2999 Dmarks) of those declaring the lowest level of concern). Repeating the calculations on changes in self-declared concern revealed (Table 5) that the average number of moves from not concerned at all to very concerned (highest to lowest level of tol-

Mean 2.06 0.00 0.72 45.31 0.48 11.55 0.62 0.02 0.24 0.06 0.06 0.07 0.03 2,900 1.21 1.51 8.53

S.D. 0.72 0.74 0.45 17.14 0.50 2.60 0.49 0.13 0.43 0.24 0.24 0.26 0.18 1,743 0.71 0.51 0.56

Min 1 2 0 17 0 7 0 0 0 0 0 0 0 0 0.10 0.99 6.03

Max 3 2 1 99 1 18 1 1 1 1 1 1 1 67,407 2.99 3.87 9.17

L. Becchetti et al. / The Journal of Socio-Economics 39 (2010) 8188 Table 3 Correlation of the level of concern and yearly changes in the concern about foreigners with selected variables. Variable Yearly change in concern about foreigners German nationality Age Male Education (in years) Married Separated Single Divorced Widowed Unemployed Loss of job in the current year Real household income % (real household income) GDP growth (3 year average) Ination rate (3 year average) Unemployment rate (3 year average)
* ** ***

85

Concern about foreigners 0.5175*** 0.0417*** 0.0522*** 0.0032 0.1509*** 0.0468*** 0.0023 0.0684*** 0.015*** 0.013*** 0.0299*** 0.0169*** 0.0477*** 0.0016 0.0118*** 0.0666*** 0.0053**

Yearly change in concern about foreigners 1 0.0163*** 0.0077*** 0.0043 0.003 0.005* 0.0032 0.0056** 0.0027 0.0014 0.0025 0.0042 0.0285*** 0.0214*** 0.0369*** 0.0216*** 0.0223***

p < 0.1. p < 0.05. p < 0.01.

Table 4 Percent distribution of selected variables across levels of self-declared concern about foreigners. Control variable Concern about foreigners in Germany 1 Male German nationality Married Separated Single Divorced Loss of job in the current year Real household income (average) Total sample 24.06 18.82 21.69 25.70 28.89 22.56 21.16 2999 23.51 2 45.81 49.73 48.79 44.70 45.79 45.53 44.20 2929 47.80 3 30.13 31.45 29.52 29.60 25.32 31.91 34.65 2773 28.69

report on average an increase of 1.83%. Surprisingly, negative family shocks seem to be associated with a reduction in concern about foreigners. An example of this is the share of separated respondents who become more tolerant, which is 20.1% and above the sample average. The ndings just described are obviously non-conclusive because composition effects, inherited traits, and reverse causality may signicantly blur the picture. The econometric estimates presented in the following section sought to solve these problems and to verify the signicance of the correlations observed. 3. Econometric ndings The discrete qualitative nature of the dependent variable requires that we estimate the model with a (random effects) panel ordered probit approach. We estimate the model both in levels and rst differences. The model in levels includes the full set of regressors and a time trend, while the model in rst differences comprises only a subset of the variables (which are expected to affect also rst differences) and no time trend. The use of rst differences helps us disentangle the effects of life events from those of psychological traits inherited from childhood. Concern about immigrants is affected by economic or family shocks, but it can also be determined by xed characteristics which are generally modeled as time-invariant individual intercepts in panel data (xed effects). By rst differencing our specication we eliminate the xed effects from the model (for a similar approach see, among others, Ravallion and Lokshin, 2002, and Ferrer-i-Carbonell and Frijters, 2004). The independent variables we select included various socio-economic individual controls (gender, age, nationality, education, marital status, employment status, loss of employment), the real household income and some aggregate macroeconomic variables (GDP growth, ination and the unemployment rate). Tables 4 and 5 present some evidence on the effect of selected variables on tolerance towards foreigners. The regressions in Table 6 seem to conrm the descriptive ndings. Being married, divorced or unemployed increases the level of concern, while higher income and education levels have the opposite effect. Age and male gender have a positive effect on concern about foreigners, although the effect becomes statistically not signicant when only native Germans are considered. With regard to macro variables, ination and unemployment rates are both signicant and have the expected sign. The positive sign is consistent with the hypothesis that higher levels of ination correspond to a reduction in pur-

Note: All numbers refer to the percent distribution of selected variables, apart from the Real Household Income, for which we report the averages conditional to the level of concern.

erance) was very low (on average 2.3%) but became higher for those respondents who had lost their jobs (3.27%). Furthermore, there is an evident negative correlation between the change in real household income and the change in the level of concern about foreigners. Those moving from the lowest to the highest level of concern register on average a real annual reduction of 2.28% in their monthly household income, while those moving in the opposite direction (from the highest to the lowest level of concern)

Table 5 Percent distribution of selected variables across changes in self-declared concern about foreigners. Control variable Concern about foreigners in Germany 2 Male German nationality Married Separated Single Divorced Loss of job in the last year % (real household income) Total sample 2.09 1.71 1.86 2.14 2.17 2.22 2.52 1.83 2.01 1 18.61 18.30 18.63 20.10 19.21 18.45 18.52 1.38 18.80 0 58.35 59.66 58.83 57.43 57.96 58.22 56.62 2.92 58.42 1 18.42 17.97 18.46 17.81 18.27 18.32 19.07 3.81 18.46 2 2.53 2.36 2.23 2.52 2.39 2.79 3.27 2.28 2.31

Note: All numbers refer to the percent distribution of selected variables, apart from the % (real household income), for which we report the average change.

86

L. Becchetti et al. / The Journal of Socio-Economics 39 (2010) 8188

Table 6 Determinants of levels of self-declared concern about foreigners. L1 German nationality Age Male Education Married Separated Divorced Widowed Unemployed ln (real household income) GDP growth (3 year average) Ination rate (3 year average) Unemployment rate (3 year average) cut1 cut2 Obs. LR 2 65.37229 (19.42) 63.62875 (18.90) 149,424 1,910 55.92872 (13.59) 54.22733 (13.18) 80,451 846 0.0037404 (8.10) 0.0357562 (2.75) 0.0834658 (32.90) 0.1033214 (6.45) 0.0201817 (0.61) 0.1207357 (4.83) 0.0692768 (2.25) 0.0467198 (3.08) 0.106838 (11.57) L2 0.418345 (19.58) 0.0004708 (0.76) 0.0079964 (0.46) 0.0626314 (17.53) 0.0957589 (4.41) 0.0003522 (0.01) 0.0616793 (1.85) 0.062852 (1.62) 0.0379108 (1.98) 0.1075701 (8.52) L3 L4 0.41219 (19.23) 0.0004813 (0.77) 0.0079654 (0.46) 0.0627287 (17.49) 0.0971186 (4.46) 0.0032189 (0.07) 0.0624045 (1.87) 0.0625348 (1.60) 0.0342072 (1.78) 0.1058542 (7.53) 0.2543266 (10.13) 0.5038958 (9.44) 0.3864928 (15.77) 46.16019 (4.15) 47.86692 (4.31) 80,451 1,205

0.0038015 (8.19) 0.035822 (2.75) 0.0831164 (32.47) 0.1068075 (6.64) 0.0199797 (0.60) 0.1210891 (4.83) 0.0677047 (2.20) 0.0412602 (2.71) 0.108765 (10.71) 0.2728341 (16.15) 0.5063628 (14.04) 0.4142591 (24.73) 35.46754 (4.69) 37.21807 (4.92) 149,424 2,726

Legend: the dependent variable is the self-declared level of concern about foreigners, the value 3 being the maximum and 1 the minimum. Regressions were run by using random effects ordered probit with time trend.

chasing power and therefore have a positive effect on worry about foreigners, while higher levels of unemployment increase the perception of competition in society.6 A puzzling result relates to GDP growth, which has a positive effect on the level of concern. Consider, however, that, as said before, this variable takes only 13 different values in a sample containing 149,424 (Regr. L3) and 80,451 (Regr. L4) observations. Moreover, the GDP growth is extremely at in the period considered. Table 7 presents the results of the rst difference regressions. The dependent variable is the change of declared level of concern while the independent variables are age, separation status, change in real household income, loss of job, and the three macro variables used in the previous tables. While regressions D1 and D3 include the full sample, D2 and D4 restrict the sample to native Germans. Age and loss of job are positively associated with an increase in the level of concern, while a rise in real income has a signicant and negative effect on it. The results on ination and unemployment rates are in line with those in Table 6, while the coefcient of the GDP growth is no longer statistically signicant, possibly solving the puzzle. 4. Interpretations and robustness checks The above-described econometric ndings support the hypothesis of a signicant association, and in the expected direction,

between real household income and concern about foreigners. Can we conclude that economic growth has positive moral consequences from this point of view? The answer seems to be afrmative when prosperity translates into signicant changes in individual economic conditions (not only growth but also the distribution of growth dividends matters). Our ndings are inevitably weaker when we look at the relationship between the dependent variable and the aggregate measure of GDP variation, given the structure of our data and the availability of only 13 different values for GDP. For this reason, we perform some robustness checks by running separate regressions by gender, nationality and age subgroups, with and without macro variables, for regressions in levels (Table 8a) and in rst differences (Table 8b). Table 8a and b report for these regressions only the magnitude and the t-statistics of the coefcient of the (change of) concern, documenting that all the coefcients are signicant and negative.7 When performing our analysis a problem of reverse causality arose because concern about foreigners and growth inuence each other. We identify a signicant nexus between the two variables of interest, net of the xed effects, but cannot directly test whether this is determined by a direct causality link from changes in household income to changes in self-declared concern. However, the characteristics of our variables and the rst difference estimates induce us to conclude that the direct causality is the most likely interpretation of our ndings. While it is plausible to assume from

6 Several empirical papers document that it is not just the personal unemployment status but also the aggregate level of unemployment which negatively affects individual happiness (Di Tella and MacCulloch, 2003; Becchetti and Santoro, 2007). The interpretation of this nding is that higher unemployment increases the fear of being red and also entails costs for workers in terms of higher taxes or increased support to unemployed family members.

7 A nal robustness check is performed by replacing our income measure with equivalised real household income using the OECD standard correction, which divides actual income by the weighted sum of family members (the rst adult is given a unit weight, additional adult members weight 0.5 and children weight 0.3). The results are substantially unchanged. They are omitted for reasons of space and are available upon request.

L. Becchetti et al. / The Journal of Socio-Economics 39 (2010) 8188 Table 7 Determinants of the changes in the level of self-declared concern about foreigners. D1 Age Separated % (real household income) Loss of job in the current year GDP growth (3 year) Ination rate (3 year) Unemployment rate (3 year) cut1 cut2 cut3 cut4 Obs. LR 2 2.007969 (165.94) 0.7549865 (78.26) 0.8886012 (91.69) 2.078249 (174.34) 119,749 94.32 2.070311 (156.36) 0.7980824 (76.73) 0.8711034 (83.52) 2.063497 (160.81) 104,682 61.1 0.0005871 (3.13) 0.0315256 (1.28) 0.0640617 (7.88) 0.033971 (1.92) D2 0.0004486 (2.25) 0.0370762 (1.39) 0.0609623 (6.88) 0.0380809 (1.95) D3 0.000496 (2.64) 0.0375394 (1.53) 0.0770799 (9.10) 0.0263016 (1.48) 0.7413222 (1.11) 0.1519849 (7.40) 0.1222583 (13.17) 0.7545282 (7.34) 0.5032755 (4.91) 2.155982 (21.02) 3.350964 (32.53) 119,749 396.01 D4 0.0002665 (1.33) 0.0423656 (1.59) 0.0880567 (9.45) 0.0303214 (1.55) 0.8754212 (1.19) 0.1322912 (5.45) 0.1628867 (16.24) 0.4706481 (4.16) 0.8059779 (7.14) 2.485061 (21.98) 3.684308 (32.44) 104,682 371.97

87

Legend: the dependent variable is the one-year change in self-declared level of concern about foreigners. Regressions were run by using random effects ordered probit.

Table 8 Robustness checks on the relationship between real household income and selfdeclared concern sensitivity analysis. (a) Concern about foreigners in Germany Coefcient of the log of real household income Male Female German Not German Age 43 Age > 43 Macro variables 0.088207 (6.58) 0.1258096 (9.87) 0.0539165 (3.67) 0.2603877 (10.40) 0.110712 (8.28) 0.1143805 (8.77) No 0.102644 (6.96) 0.1158054 (8.26) 0.1220635 (7.37) 0.0805345 (3.00) 0.1036904 (7.07) 0.1336803 (9.27) Yes

an aggregate point of view that less concern creates an environment more favourable to economic prosperity, it is more difcult to imagine that short-term increases in self-declared concern about foreigners among the interviewees can generate immediate negative changes in their respective household earning capacities. Finally, reverse causality must denitely be excluded when we compare changes in aggregate economic indicators and changes in respondents self-declared concern about foreigners. The results on ination and unemployment go in the right direction in rst difference estimates. It is difcult to imagine that a reduction in worry about foreigners (or omitted variables related to it) may be the cause of the improved performance of the above-mentioned macroeconomic indicators.8 5. Conclusions A large part of public opinion tends to believe that economic activity is a zero sum game in which the gain of a given group of individuals is necessarily offset by the loss of another group. In periods of stagnation in which real GDP growth is non-positive, the economy is, by denition, a zero (or a negative) sum game, at least from the algebraic point of view (see footnotes 3 and 4 for qualications and distinctions). For this reason, low growth and high unemployment and ination rates (which create the conditions for economic stagnation) may contribute to reinforcing the belief that

(b) Concern about foreigners in Germany Coefcient of the one year log difference in real household income Male Female German Not German Age 43 Age > 43 Macro variables 0.0505615 (4.37) 0.077682 (6.80) 0.065591 (4.91) 0.0716945 (3.52) 0.0424949 (3.68) 0.0859836 (7.51) No 0.0673059 (5.60) 0.0867719 (7.27) 0.0984528 (7.06) 0.0324494 (1.57) 0.0526402 (4.39) 0.1004478 (8.37) Yes

Legend: The tables report coefcients and t-statistics of the level of concern (a) and change of level of concern (b) from regressions run by gender, nationality and age subsamples. The reference specications for (a) are regression L1 (without macro variables) and L3 (with macro variables) in Table 6 and those for (b) are regressions D1 (without macro variables) and D3 (with macro variables) in Table 7.

8 As a nal robustness check on the causality nexus between the two variables we estimate a two-equation panel VAR model where, in the rst (second) equation, self-declared tolerance (real household income) is regressed on its past values and on past values of real household income (self-declared tolerance). The advantage of the panel VAR model is that it jointly tests for the existence of the two possible causality directions between the selected variables (Holtz-Eakin et al., 1988). A Granger causality test conrms the negative and signicant effect from real household income to self-declared tolerance. The problem with the panel VAR model is that we are forced to assume that the discrete dichotomous structure of self-declared tolerance may be approximated by a continuous variable. Frey and Stutzer (2006) have done something similar on the 010 scale self-satisfaction variable, nding that the results of OLS and orderer probit models are not very different. However, since our variable is on a 13 scale, we believe that this assumption is stronger in our case (even though not necessarily implausible) and therefore prefer to omit the results of the estimates (available upon request) from the paper.

88

L. Becchetti et al. / The Journal of Socio-Economics 39 (2010) 8188 Bacher, J., 2001a. Macht Arbeitlosigkeit rechtsextrem? In: Zempel, J., Bacher, J., Moser, K. (Eds.), Erwerbslosigkeit. Ursaken, Auswirkungen und Interventionen. Leske and Budrich, Opladen, pp. 171186. Bacher, J., 2001b. In welchen Lebensbereichen lernen Judendliche Auslanderfeindlichkeit? Ergebnisse einer Befrangung bei BerufsschulerInnen. Kolner Zeitschrift fur Soziologie und Sozialpsychologie 53, 334349. Bauer, T.K., Lofstrom, M., Zimmermann, K.F., 2000. Immigration policy, assimilation of immigrants and natives sentiments towards immigrants: evidence from 12 OECD-countries. Swedish Economic Policy Review 7, 1153. Becchetti, L., Santoro, M., 2007. The wealth-unhappiness paradox: a relational goods/Baumol disease explanation. In: Bruni, L., La Porta, L. (Eds.), Handbook on the Economics of Happiness. Elgar, pp. 239262. Di Tella, R., MacCulloch, R., 2003. The macroeconomics of happiness. Review of Economics and Statistics 85 (4), 809827. Dustmann, C., Preston, I., 2007. Racial and economic factors in attitudes to immigration. The B.E Journal of Economic Analysis and Policy 7 (1), 16551685. Dustmann, C., Preston, I., 2000b. Attitudes to ethnic minorities, ethnic context and location decisions. Economic Journal 111, 353373. Falk, A. and Zweimueller, J. (2005), Unemployment and right-wing extremist crime. CEPR Discussion Paper No. 4997. Ferrer-i-Carbonell, A., Frijters, P., 2004. How important is methodology for the estimates of the determinants of happiness? Economic Journal 114, 641659. Fertig, M., Schmidt, C.M. (2002), The perception of foreigners and Jews in Germanya structural analysis of a large opinion survey. IZA Discussion Paper No. 431. Frey, B.S., Stutzer, A., 2000. Happiness, economy and institutions. Economic Journal (110), 918938. Frey, B.S., Stutzer, A., 2006. Does marriage make people happy, or do happy people get married? Journal of Socio-Economics 35 (2), 326347. Friedman, B.M., 2005. The Moral Consequences of Economic Growth. Alfred A. Knopf, New York. Gang, I.N., Rivera-Batiz, F.L., 1994. Unemployment and attitudes towards foreigners in Germany. In: Steinmann, Gunter, Ulrich, Ralf E. (Eds.), The Economic Consequences of Immigration to Germany. Physica-Verlag, Heidelberg, pp. 121154. Gang, I.N., Rivera-Batiz, F.L., Yun, M.S. (2002), Economic strain, ethnic concentration and attitudes towards foreigners in the European Union. IZA Discussion Paper No. 578. Hirsch, F., 1976. Social Limits to Growth. Harvard University Press, Cambridge, MA. Holtz-Eakin, D., Newey, W., Rosen, H.S., 1988. Estimating vector autoregressions with panel data. Econometrica (56), 13711395. Ravallion, M., Lokshin, M., 2002. Self-rated economic welfare in Russia. European Economic Review 46, 14531473. Scitowsky, T., 1976. The Joyless Economy: An Enquiry into Human Satisfaction and Consumer Dissatisfaction. Oxford University Press, New York.

economic wellbeing is a cake of xed size whose slices must be competitively acquired at the expense of others. This perception is expected to reduce tolerance toward other participants in the competition, all the more so if they are foreigners or belong to a cultural minority. This reasonable argument implies that economic prosperity may also have positive moral consequences by abating peoples fear of becoming losers in the competitive race and their related concerns about competitors and foreign workers. In this paper, we take advantage of the existence of a large panel dataset relating individual declarations of concern about foreigners with various socio-demographic variables. Our descriptive and econometric ndings document the existence of a robust negative nexus between real household income and self-declared concern about foreigners, net of the impact of many control variables and of the effects of time-invariant idiosyncratic inherited traits of individual personality. Although we have observed a similar nexus when considering ination and unemployment rates, our ndings are more fragile in this case because we had to rely on only thirteen annual observations for the period 19922004. In this latter case, results from statistical inference are less stable and do not allow us to draw denite conclusions. Overall, we cannot reject the hypothesis that economic growth exerts a positive inuence on moral values and social attitudes, even though our results qualify the assumption by showing that the effect crucially depends on the transmission of positive changes in prosperity from the aggregate to the individual level. This in turn implies that, for determination of this effect, the distributional consequences of economic growth need to be taken carefully into account. References
Adorno, T.W., Frenkel-Brunswick, E., Levinson, D.J., Sanford, R.N., 1950. The Authoritarian Personality. Harper, New York.