WPS8700 Policy Research Working Paper 8700 Long-Term Evolution of Inequality of Opportunity Maurizio Bussolo Daniele Checchi Vito Peragine Europe and Central Asia Region Office of the Chief Economist January 2019 Policy Research Working Paper 8700 Abstract The main goal of this paper is to document and analyze time, inequality of opportunity decreases with age: the the long-term evolution of inequality of opportunity effect of circumstances seems to weaken over the life cycle. and thus extend the recent empirical literature, which is This is a quite different age profile from that of inequality mainly concerned with its measurement at a specific point of outcomes (income or consumption), which generally in time. Using repeated cross-section surveys for five Euro- increases with age. A decomposition of the relative inequal- pean countries (France, Germany, Great Britain, Italy, and ity of opportunity allows highlighting some key drivers of Switzerland), the evolution of inequality of opportunity is its time evolution. In all the countries, there has been a clear measured for a period of about two decades for the whole enhancement of equality of educational opportunity (as populations, as well as for different birth cohorts. Relative captured by a downward trending intergenerational educa- inequality of opportunity represents an important portion tion persistence) and a reduction of the returns to education. of total income inequality, with values ranging from 30 However, for some countries, notably Italy, these trends to 50 percent according to the standard deviation of logs have failed to translate into decreasing inequality of oppor- (and reaching a lower share in case of mean log deviation) tunity in the income distribution because of the increasing and, for all the countries, it shows a stable or declining role of parental networking (an additional channel through time trend. When the birth cohorts are followed across which parental background affects the incomes of offspring). This paper is a product of the Office of the Chief Economist, Europe and Central Asia Region. It is part of a larger effort by the World Bank to provide open access to its research and make a contribution to development policy discussions around the world. Policy Research Working Papers are also posted on the Web at http://www.worldbank.org/research. The authors may be contacted at mbussolo@worldbank.org, daniele.checchi@gmail.com, peragine.vito@gmail.com. The Policy Research Working Paper Series disseminates the findings of work in progress to encourage the exchange of ideas about development issues. An objective of the series is to get the findings out quickly, even if the presentations are less than fully polished. The papers carry the names of the authors and should be cited accordingly. The findings, interpretations, and conclusions expressed in this paper are entirely those of the authors. They do not necessarily represent the views of the International Bank for Reconstruction and Development/World Bank and its affiliated organizations, or those of the Executive Directors of the World Bank or the governments they represent. Produced by the Research Support Team Long-Term Evolution of Inequality of Opportunity Maurizio Bussolo (World Bank), Daniele Checchi (University of Milan) and Vito Peragine (University of Bari) Keywords: Inequality of Opportunity, Decomposition methods, Education mobility, Returns to Education, Family Networking, Cohort Analysis. JEL codes: D31, D63, E24, I24, J62  This paper was started as background paper for the World Bank regional flagship report on “Towards a new social contract: Taking on distributional tensions in Europe and Central Asia”. We thank Jorg Neugschwender (Luxemburg Income project) and Teresa Randazzo (University of Bari) for technical assistance in building the dataset, and Tullio Jappelli (University of Naples, Italy) for extensive discussions. All errors are our own responsibility. 1. Introduction The recent empirical literature on equality of opportunity (EOp) has provided a significant body of evidence on the extent of inequality of opportunity in different countries. See Brunori et al. (2015) for a first assessment of the existing evidence and Ferreira and Peragine (2016), Ramos and Van de Gaer (2016) and Roemer and Trannoy (2015) for methodological and conceptual issues related to the measurement of EOp. A common feature of the existing literature is the static approach. Almost all empirical analyses use income distribution at a given point in time as the relevant distribution of individual advantages and are limited to computation of inequality of opportunity as a snapshot for a given country or set of countries.1 This paper instead is concerned with the evolution of inequality of opportunity, i.e. with a dynamic approach. In addition, the time variation of EOp allows to study its main determinants. By so doing, we move the research on EOp a step forward and propose and test a (simple) empirical model that can explain the generation of inequality of opportunity in a given economy. There are three different ways to analyze the evolution of inequality of opportunity, which correspond to three different concepts of inequality dynamics: (i) inequality measured across repeated snapshots of the population (repeated cross-sectional analysis); (ii) inequality measured along life courses (longitudinal analysis); (iii) inequality measured across generations (cohort analysis). While analysis (ii) requires the availability of a rich longitudinal data set containing information of individual incomes and circumstances over the entire life cycle of the individuals, the analyses (i) and (iii) can be potentially carried out by using repeated cross-section surveys, hence are much less data intensive. This is the reason why in the present paper we focus on analyses (i) and (iii). See Aaberge et al. (2011) for an analysis of long-term inequality of opportunity along the lines of concept (ii). 2. The model 2.1 Canonical models of inequality of opportunity The conceptual basis for the definition of inequality of opportunity is provided by the distinction, among the factors influencing the individual achievements, between individual efforts and pre- determined circumstances – defined as those which lie outside the realm of individual responsibility. The EOp approach considers that inequality due to the former is not ethically offensive, whereas it suggests that differences in individual outcomes due to the latter represent a violation of the principle of equality of opportunity and should be removed. In what follows we will follow the simple framework introduced by Checchi and Peragine (2010) to measure inequality of opportunity. Consider a distribution of income in a given population. Suppose that all determinants of , including the different forms of luck, can be classified into either a set of circumstances that lie beyond individual responsibility, belonging to a finite set Ω, or as responsibility characteristics, summarized by a variable , denoting effort,2 belonging to the set Θ. The outcome of interest is generated by a function : Ω Θ → ℝ such that: , (1) 1 Also the cross-country comparability is a relevant issue, given the potentially different definitions of outcome and circumstances involved in the analysis. 2 Effort could also be treated as a vector. However, we follow the literature and treat it as a scalar. 2 This can be seen as a reduced-form model in which income is exclusively determined by circumstances and effort, such that all individuals having the same circumstances and the same effort obtain the same income. Roughly speaking, the source of unfairness in this model is given by the effect that circumstance variables (which lie beyond individual responsibility) have on individual outcomes. A parametric implementation of the model above,3 which has been extensively used in the literature (see Bourguignon et al. 2007), considers estimating by OLS the following equation (2) and computes inequality of opportunity as the value of a given inequality measure ∙ applied to the distribution of the predicted values , where . Hence the value of absolute inequality of opportunity is given by while the value of relative inequality of opportunity is given by / . A dynamic version of the model can be obtained by introducing the time dimension in alternative ways. We could consider a first expression, in which income is assumed to vary with time, while circumstances are assumed to be time invariant: (3) Model (3) assumes that circumstances impact income in an identical way over the entire life. A variant of the same model considers the possibility of time-varying effects, possibly distinguishing between fixed and time-varying circumstances: (4) Both models (3) or (4) are highly demanding in terms of data, because their longitudinal structure requires repeated observations of the same individual, possibly under alternative sets of circumstances which are independent from her will. In addition, implementing models (3) or (4) would provide a picture of the evolution of EOp over the life cycle of the specific birth cohorts that are present at the start of the analysis. A less demanding approach in terms of data exploits the availability of repeated cross sections from the same population. If one is interested in understanding whether a society is experiencing changes in the EOp of its citizens, the relevant model considers (5) where is the income of individual sampled in survey . The data generating process is allowed to change over time among random draws from the (same country) population. The implicit assumption is the over-time stability of the population, such that changes in EOp can be attributed to changes in the relevant parameters and . Model (5) is specular to cross-country analysis, once is interpreted as a country indicator, but has the advantage of greater comparability of the underlying populations, originating from the same country. 3 In this paper we follow the ex ante approach. See Fleurbaey and Peragine (2013) for a comparison between the ex ante and ex post approaches to equality of opportunity. 3 If the number of cross-sections available for the same country is large enough, and their time span covers a sufficient number of years, one could interpret them as a pseudo-panel, in order to get as close as possible to model (3). In such a case the relevant model becomes (6) where is the income of individual born in year and sampled in survey . In such a case, EOp can be repeatedly measured along three dimensions: in a specific year of survey , repeated observations refer to different birth cohorts ’s; for a specific birth cohort , repeated observations refer to different dates of survey ’s; for a specific age cohort , repeated observations refer to different life cycles. Section 2.2 initially adopts the approach described by model (5). It uses repeated cross-section surveys of the population of a specific country and estimates, for each year, the relevant parameters of the model. An extension which uses the cohort structure of model (6) is also considered. Both these dynamic approaches provide interesting and distinct insights on the evolution of EOp. 2.2 Our empirical model This section presents a decomposition of measured inequality of opportunity into its constituting components in the same vein as what Solon (2004) did for intergenerational mobility of incomes. In the empirical literature (Ferreira and Peragine 2015), circumstances have included gender, age, ethnicity, region of birth, parental background (in terms of educational attainment and occupational status). For simplicity of exposition, let us consider circumstances as consisting of a single variable, parental education, indicated with where denote generations.4 We assume that parental background affects the income opportunity of the child through two main channels: educational investment and family networking.5 The first channel can be simply described by the intergenerational persistence of educational attainment (Black and Devereux 2011) (7) where is the education of the child, is the education of the parents, is a measure of intergenerational persistence and captures any unobservable component (like ability as well as effort). This intergenerational correlation can be justified on various grounds: cultural dependency (more educated parents value education more and press their children to follow in their footsteps), financial resources (more educated parents hold better jobs and earn higher salaries which allow larger resources to be invested in education); teaching practices (more educated parents are capable to support their children during their schooling career). Education is valued in the labor market. Following the Mincerian approach, we assume that individuals choose optimally the amount of schooling by balancing costs (foregone incomes) and benefits (higher wages expected in the future – see Heckman et al 2005). As a consequence, the earnings of people with different educational attainments will differ by an amount that will be proportional to the years of schooling, as in the following equation (where we abstract from the usual demographic information): log (8) 4 One could easily add additional circumstances (like gender, age and foreign citizenship, as we do in the empirical section) but the line of argument would remain unaffected. 5 Since parental background includes many other dimensions beyond education (like parental income, access to educational resources, family wealth, neighborhood), our model is observationally equivalent to many other models of intertemporal transmission of social status. See for example DeFraja (2002), 4 where is the income of the child, is the standard return to education and is a random error (capturing unobservable components – ability, effort – but also unpredictable components – luck). Besides helping providing education, parents may influence children’s outcomes by other means. To consider this additional influence, we adopt an extended mincerian equation as follows log (9) The inclusion of parental education can be justified as proxy for family networking in non- competitive labor markets, where connection referrals matter to obtain good jobs (Kramarz and Nordström 2014); it is also consistent with intergenerational transmission of financial assets through bequests. By replacing equation (7) into equation (9) we obtain: log (10) If we now denote with ∙ any inequality measure, we get (11) where we can notice that income inequality will be a function of the distribution of parental education (circumstances) and unobservable components (effort, ability and/or luck), as well as of the structural parameters of the income generating process. For consistency with most of the literature on earnings inequality, we have chosen the standard deviation of logs as our inequality indicator.6 In such a case 2 , (12) As previously mentioned, a relative measure of inequality of opportunity is given by the ratio between the inequality attributable to circumstances and total inequality. In the present case, the income attributable to circumstances is given by the predicted values ̂ , obtainable from the estimation of equations (7) and (9). The relative IOp is thus given by the following equation: 6 Analytic and empirical results are almost identical if we replace the standard deviation of logs with the mean log deviation. 5 , (13) , Equation (13) indicates that, other things constant, relative IOp declines when there is: 1) a reduction in the intergenerational persistence of education ̂ ; 2) a reduction in the (private) return to education ; 3) a reduction in the effect of family network in the labor market ; 4) an increase in the variance and covariance of the non-observable components and ̂; 5) a reduction in the variance of the educational attainment of the previous generation. We will focus mostly on the combination of parameters ̂ which summarizes the channels of intergenerational persistence. As it is intuitive, if the educational investment becomes irrelevant (because education yields insignificant returns in the labor market), then parents become unable to transmit privileges to their offspring, and inequality declines as a consequence. Similarly, if parents are unable to actively network on behalf of their children, the disadvantage due to circumstances will decline. The same approach can be used to assess the role of other circumstances. As a final example, consider the impact of gender: women are better achievers in schooling, but they are discriminated against in the labor market. Equations (7) and (9) are to be modified accordingly: (7)´ log (9)´ where now is a dummy variable for women, is the mean school gap achieved by women and is the gender wage gap. Since 1 , where is the fraction of women in the working population, then we get that relative inequality of opportunity now reads (13)´ , In this case, relative inequality of opportunity also depends on whether the schooling advantage for women exceeds (or falls short of) the labor market disadvantage , as well as from the gender composition of the labor force. 3. The data Consistent estimates of the IOp described by equation (13) impose data requirements that are rather demanding: a) adequate information on circumstances (in addition to gender and age, some information on parental background and country of origin). b) a measure of disposable income that is comparable across surveys and across countries (if we intend to benchmark one country against the others). c) a sufficiently extended time coverage in order to capture meaningful dynamics and/or to apply birth/age cohort decomposition. 6 Existing sources of publicly available data are rather limited with respect to these three criteria. We resorted to the LIS Cross-National Data Center in Luxembourg (http://www.lisdatacenter.org/), which allowed us to process data from four countries (Italy, Germany, France and Switzerland), while a fifth country was obtained from accessing the original provider (United Kingdom – https://www.understandingsociety.ac.uk/). The surveys we have used are therefore the following: Italy: Survey on Household Incomes and Wealth (SHIW), collected by the Bank of Italy – 11 surveys, covering the period 1993-2014 (information on parental background is not available before the starting date – originally consisting of 112,690 individuals, which reduces to 107,846 when considering non-missing information. Germany: German Socio-economic Panel (SOEP) – 11 surveys, covering the period 1984-2013 – originally including 156,338 individuals, then reduced to 133,467 in case of non-missing information. France: Household Budget Survey (HBS), conducted by the Banque de France) – 6 surveys, covering the period 1978-2005 – originally consisting of 97,306 individuals, declining to 89,119 when missing information is excluded. Switzerland: Swiss Household Panel (SHP) – 6 surveys, covering the period 1999-2014 – originally consisting of 43,102 individuals, which then decline to 31,273 valid observations. United Kingdom: starts as British Household Panel (BHPS), replaced after 2009 by the Understanding Society-Household Longitudinal Survey (UKHLS) – considers 24 waves over the period 1991-2014 – originally consisting of 434,253 individuals, which then decline to 308,625 valid observations. Our selection rules include individuals aged 25-80 with a positive disposable income, harmonized according to the LIS procedure (variable DPI).7 Incomes are converted to constant prices using the national consumer price index. Parental education is typically a categorical variable recording the highest educational attainment in the parental couple. In order to estimate a unique coefficient associated to the intergenerational transmission of education, we have converted them into years of education.8 Descriptive statistics at survey/country disaggregation are reported in tables 1 to 5 in the Appendix. Using these data, total inequality, absolute inequality of opportunity (namely inequality computed over incomes predicted according to circumstances) and relative inequality of opportunity (see equation (13)) have been estimated for each country and for each survey/year. These measures are reported in tables 6 to 10, including two indicators of inequality (standard deviation of logs and mean 7 To avoid negative values associated to logs, we have excluded all individuals with yearly incomes below 10. Data for the United Kingdom were rather volatile with respect to top incomes: in order to avoid confounding factors associated to differences in sampling procedures, we have trimmed them excluding incomes exceeding the 99.5 centile. 8 In the Italian file, recoding implies the following conversion: [1] illiterate=0 years; [2] primary=5 years; [3] lower secondary=8 years; [4] upper secondary=13; [5] tertiary=18. In the German file, recoding implies the following conversion: [1] school not attended =0 years; [2] no school degree =4; [3] Secondary General School (Hauptschule)=9 years; [4] Intermediate School (Realschule)=10 years; [5] Technical High School (Fachoberschule)=12 years; [6] Upper Secondary School (Abitur)=13 years. In the Swiss file, recoding implies the following conversion: [1]1: Primary or first stage of basic education=6 years; [2] 2: Lower secondary or Second stage of basic education=9 years; [3] 3A&B Upper secondary education (preparation for tertiary education & voc.educ)=12 years; [4] 3C: Upper secondary education (entrance into the labor market)=11 years; [5] 4A: Post-secondary education non-tertiary (preparation for an institution for higher education)=13 years; [6] 5B: First stage of tertiary education (professional education)=15 years; [7] 5A/6 tertiary education (general education)=16 years. In the UK file recoding implies the following conversion: [1] no qualification =8 years; [2] some qualification=10 years; [3] post school qualification=12 years; [4] university degree=18 years. Eventually, in the case of France there is no information on parental education, but only on parental occupation. In order to retain the country, we have created a dummy variable corresponding to either [5] intermediate profession (foreman, nurse, etc.) or [6] executive, liberal profession. We interpret this variable as the (likely) completion of secondary or tertiary education. 7 log deviation), which behave in very similar ways. In addition, to apply the decomposition of relative inequality of opportunity as shown in equation (13), equations (7) (which captures intergenerational persistence in education) and (9) (Mincerian wage equation) have been estimated. Estimations were conducted at the country and year/survey level. For illustrative purposes, the results of these estimations at the country level and for the full sample (i.e. for all the surveys pooled together) are reported in table 11. One can notice that country samples are rather consistent, according to the impact exerted by the regressors. Education is adequately rewarded in all countries, with an estimated yearly return rate ranging between 5.4% in France and 13.2% in Great Britain. The intergenerational persistence in education is highest in Italy and Germany and lowest in Great Britain. There is also general evidence that parental education exerts an impact beyond favoring educational attainment of the next generation, as the coefficient in equation (9) is estimated positive and statistically significant in all countries (its magnitude being highest for continental countries). In all countries, women are on average penalized in terms of both schooling and incomes, while age exhibits an opposite trend: the younger age cohorts are better educated than the older ones, but incomes increase with age, the net effect being ambiguous. Finally, being born in less developed regions (South of Italy, East Germany) or holding a foreign citizenship is associated to lower incomes (but not necessarily lower schooling). To study the evolution of inequality of opportunity – the main objective of this paper – the estimation of the models reported in table 11 is performed for year survey/year and the results are graphically reported in figures 1, 3, 5, 7, and 9, and discussed in detail in section 4 below. Given that the sample sizes are large enough, it is possible to carry out the estimations at a more disaggregated level. By disaggregating the population in birth cohorts, we can estimate inequality of opportunity for each cohort and investigate whether the cohort-specific evolution of inequality of opportunity differs from that of the full population. More in detail, we have partitioned birth years and ages in 5-year intervals, and we have retained only cells gathering at least 400 individuals. In each population subgroup, we have estimated inequality, inequality of opportunity and other structural parameters. This procedure is exemplified in table 12 and figure 11 for the Italian case. Despite having the population distributed over 66 cells (the potential number of cells depending on dates of initial and final surveys – top part of the table), only 53 satisfy a sufficient numerousness and are therefore retained for estimation of (relative) inequality of opportunity (bottom part of table 12). Once we have obtained these measures, if we ask ourselves what the time pattern of IOp is, we can plot these measures by birth cohort, as we have done in figure 11. Looking at the graph, one would be tempted to conclude that during the life course IOp exhibits an inverted U-shaped profile, at least in Italy. However, we would be confusing two different dimensions, namely age and cohort: some birth cohorts (for example the one born around the second world war) have experienced higher IOp at any age, compared to neighboring birth cohorts. Thus, we need a more rigorous method to summarize the information contained in the cells, possibly distinguishing between age and cohort effects. We have then followed Deaton (1997), and we have regressed the obtained measures onto age, cohort and survey dummies, imposing restrictions on the estimated coefficients for dummies. Results are reported in table 13, and then plotted using a smoothing procedure in figure 12 using the LOWESS command in Stata. Simple inspection of the coefficients indicates that the time profiles of the constrained and unconstrained estimates are rather similar, though the time trend may be different. The same procedure is also applied to the estimated structural parameters, weighting the observations by the inverse of their standard errors. 8 4. The results Having clarified our statistical procedure, it is now time to review our main results, which are fully summarized by figures 1 to 10. For each country, we report two sets of estimates: a) The first set contains the analysis by year/survey and reports the estimated values of four different variables: relative inequality of opportunity, return to education (), parental network () and the intergenerational persistence in education (̂ ) (see figures 1, 3, 5, 7, and 9); b) The second set contains the same variables calculated at different ages and for different birth cohorts. Hence, it reports respectively the age and the cohort profiles of each of the four estimates mentioned above (see figures 2, 4, 6, 8, and 10). Instead of reviewing the results twice, first by surveys and second by decomposing age and birth cohorts, we have preferred a thorough discussion by country. 4.1 Italy Starting with relative IOp, the analysis by survey shows a clear reduction in relative IOp at the beginning of the 2000s and then an increase at the beginning of the 2010s. In sum a rather constant time trend: the value of IOp is the same at the start and at the end of the period, also confirmed by the mean log deviation (MLD). As for the magnitude, it varies between 45% and 50% according to the standard deviation of logs and between 30% and 40% according to MLD (see figure 1). What is behind this high and rather constant time evolution of inequality of opportunity? The decomposition approach of this paper – and in particular considering the trends of intergenerational persistence of education, returns to education, and parental networking – can help answering this question. The intergenerational persistence of education shows a clear declining trend. This trend is well known and explained by the expansion in education that took place in Italy following the compulsory education reform at the beginning of the 1960s, with some signals of trend reversal in recent years. However, this declining trend has not translated into a declining inequality of opportunity in income. Furthermore, the return to education displays a downward trend, which should also help in reducing inequality of opportunity. Apparently, this reduction is not materializing because of the counterbalancing increasing trend of parental networking. Our suggested interpretation is that the increased equality of educational opportunity (associated to the decrease in intergenerational education persistence) and the reduced “value” of education in the labor market have failed to translate into a decrease of opportunity inequality in income because of the increasing role of parental networking. This interpretation is substantially confirmed when looking at both the age and the cohort analyses, which however shows some additional interesting facts (see figure 2). As for the age profile, the results show a clear declining pattern in relative inequality of opportunity, which is associated with a consistent declining trend in the return to education and a clear increasing trend in both intergenerational persistence and parental networking. The cohort profile follows a similar path in inequality of opportunity, return to education and parental network, while the intergenerational persistence shows a clear declining pattern, which is explained by the expansion in education level that took place in Italy during the last decades. Thus, the general declining pattern of intergenerational education observed in the analysis by survey seems to be mainly driven by the cohort effect. 4.2 Germany The analysis by survey shows a clear declining pattern in relative IOp, which takes values between 40% and 55% in the case of standard deviation of logs (between 20% and 50% in case of MLD). This is complemented by a fairly constant pattern of intergenerational education persistence and a weakly increasing trend of parental networking (which however is not statistically significant for most of the 9 sample period), while the return to education shows a declining trend in the 1980s and then a fairly stable pattern (see figure 3). As for the age profiles, the results show a clear declining pattern in the value of relative inequality of opportunity, which is associated with an inverted U-shaped trend of the return to education and a flat pattern of both intergenerational persistence of education and parental networking. The cohort profile follows a similar path in the values of inequality of opportunity, parental network, and intergenerational persistence of education, while the return to education is rather stationary across cohorts (see figure 4). 4.3 France The analysis by survey clearly shows a declining pattern in relative IOp, which takes values between 30% and 45% in the case of standard deviation of logs (between 20% and 30% in case of MLD). This is complemented by a decreasing trend in intergenerational education persistence. On the other hand, parental networking shows a pretty flat picture and the return to education a constant pattern with a decline in the last period (the first half of the 2000s). Hence, the declining trend of IOp might be mainly driven by the reduction in intergenerational educational persistence (see figure 5). As for the age profiles, our results show a clear declining pattern in the value of relative inequality of opportunity, which is associated with a consistent declining trend in the return to education and a clear increasing trend in both intergenerational persistence and parental networking. The cohort profile follows a similar path in the values of inequality of opportunity, although the pattern shows an increase in the very first period, and in the return to education and parental network, while the intergenerational persistence shows a clear declining pattern, which is explained by the expansion in education level that has taken place during the last decades (see figure 6). 4.4 United Kingdom The analysis by survey (see figure 7) shows a declining pattern in relative IOp, which takes values between 30% and 50% in the case of standard deviation of log incomes (between 10% and 35% in case of MLD). On the other hand, it is observed a stable pattern in parental networking and a weakly declining trend in both intergenerational education persistence and return to education. Hence the declining trend of IOp might be mainly driven by the reduction in intergenerational educational persistence. As for the age profiles, the results show a clear declining pattern in the value of relative inequality of opportunity, which is associated with a declining pattern in the return to education. On the other hand, both parental network and intergenerational persistence of education show an increasing trend. The cohort profile follows a similar path, except for the intergenerational persistence of education, which is more stable, while the return to education shows a more stable path (see figure 8). 4.5 Switzerland The analysis by survey shows a clear declining pattern in relative IOp, which takes values between 30% and 40% in the case of standard deviation of logs (between 15% and 25% in case of MLD). This is complemented by a fairly increasing pattern of both intergenerational education persistence and parental networking, while the return to education shows a decreasing trend (see figure 9). As for the age profiles, the results show a clear declining pattern in the value of relative inequality of opportunity, which is associated with an inverted U-shape of the return to education, a fairly stable trend of parental networking and an increasing pattern of intergenerational persistence of education. The cohort profile follows a fairly similar path, except for the return to education that, after an increase for the first cohorts, then remains stable (see figure 10). 10 4.6 Summing up In general, our empirical results are consistent with theoretical expectations. More precisely, the relationships between the trends of inequality of opportunity in the income space, intergenerational persistence in education, return to education and parental networking are consistent with the conjectures based on equation (13). In addition, it is possible to highlight the following stylized facts: i) in all the countries and the period considered, inequality of opportunity represents an important portion of total income inequality, with values ranging from 30% to 50% according to standard deviation of logs (and reaching a lower share in the case of mean log deviation); ii) in general, inequality of opportunity shows a stable or declining pattern over the period considered in all countries; iii) on the other hand, in all countries considered, there has been a clear enhancement of equality of educational opportunity (as captured by the intergenerational education persistence); iv) in some countries the egalitarian process taking place in the education system has failed to translate into decreasing opportunity inequality in the space of income because of the increasing role of parental networking and the reduced “value” of education in the labor market. This mechanism seems to be at work notably in Italy; v) in some other countries (France, Germany and Great Britain), where both returns to education and the family networking followed a more constant pattern, inequality of opportunity seems to decrease both in the education and in the income space. The decomposing of inequality of opportunity trends according to the age and cohort effects allow to identify the following additional facts: vi) in all the countries considered, inequality of opportunity decreases with age: the effect of the circumstances at birth seems to weaken over the life cycle. This pattern is quite different from the age profile of the inequality of outcomes (income or consumption), which generally increases with age; vii) the decreasing pattern of relative inequality of opportunity in France and Italy is associated with a consistent declining trend in the return to education and a clear increasing trend in both intergenerational persistence and parental networking. Great Britain shows an increase in intergenerational education persistence, while Germany is characterized by a stable trend of intergenerational education persistence; viii) the cohort effect, on the other hand, shows a more mixed picture: while for Great Britain and Germany the data show a declining path in the values of inequality of opportunity, with the younger generation experiencing lower IOp levels, both Italy and France are characterized by an inverted U- shape pattern; ix) these trends are associated, in Germany and Great Britain, with a stable or weakly increasing trend of the intergenerational educational persistence, while in Italy and France with a clear declining trend in the intergenerational persistence of education, which is explained by the expansion in education level that has taken place during the last decades. 5. Concluding remarks This paper contributes to the analysis of inequality of opportunity in three respects. First, by using extended samples, it is capable to detect time trends, showing that the role of circumstances (parental background, gender, age, and place of birth) in shaping income distribution has declined over the last two decades in all the countries considered in the present analysis. Depending on the inequality index we choose, inequality of opportunity accounts for between one-third (MLD) and half (standard deviation of logs) of total inequality in personal disposable incomes, at least for the four largest economies in the European Union. 11 Second, we exploit the large sample sizes to obtain inequality measures by age group and birth cohorts, thus being able to decompose observed trends in age profiles and birth cohort changes. For the five countries under analysis, the observed inequality of opportunity exhibits an inverted U- shaped pattern over the life cycle. Moreover, the most recent age cohorts have experienced a lower IOp, thus appearing as the main beneficiaries of the overall decline in inequality. Third, the paper proposes a theoretical framework identifying the variables that affect (positively or negatively) inequality of opportunity. The framework is then estimated, and the data confirm the predicted signs. The analysis has focused on the role of three variables: the intergenerational persistence in educational attainment, the return of education, and the networking activity of parents. While the first two variables exhibit a declining trend, which other things constant should produce a decline in IOp, the third one appears to be rising in many countries, thus counteracting the effects of the first two. Consequently, the fair optimism that descriptive statistics suggest with respect to income inequality should be mitigated by paying attention to educational persistence and labor market segmentation. References Aaberge, R., M. Mogstad, & Peragine, V. (2011). Measuring Long-term Inequality of Opportunity, Journal of Public Economics, 95(3-4), 193-204. Black, S., and P. Devereux. (2011). “Recent Developments in Intergenerational Mobility.” Handbook of Labor Economics , 4(B), Ch-16, pages 1487- 1541 Bourguignon F., Ferreira F.H.G., Menéndez, M. (2007). Inequality Of Opportunity In Brazil, Review of Income and Wealth, 53(4), 585-618. Checchi, D., Peragine, V. (2010). Inequality of opportunity in Italy, Journal of Economic Inequality, 8(4), 429-450. Deaton, A. (1997). “The analysis of household survey. A microeconometric approach to development policy” Johns Hopkins University Press (Baltimore and London) DeFraja, G. (2002). "The Design of Optimal Education Policies," Review of Economic Studies, vol. 69(2), pages 437-466. Ferreira, F.H.G., Peragine, V. (2016). Individual responsibility and equality of opportunity, in M. Adler and M. Fleurbaey (eds.), Handbook of Well Being and Public Policy, Oxford: Oxford University Press. Fleurbaey M., Peragine, V. (2011). Ex ante versus ex post equality of opportunity, Economica, vol. 80(317), pages 118-130. Heckman, James J., Lochner, Lance J. and Todd, Petra E., 2006. “Earnings Functions, Rates of Return and Treatment Effects: The Mincer Equation and Beyond” in Erik Hanushek and F. Welch (ed.), Handbook of the Economics of Education. Elsevier, vol.1 Kramarz, Francis and Oskar Nordström Skans, 2014. "When Strong Ties are Strong: Networks and Youth Labour Market Entry," Review of Economic Studies, Oxford University Press, vol. 81(3), pages 1164-1200. Ramos, X., Van de Gaer, D. (2016). Approaches to inequality of opportunity: Principles, measures and evidence, Journal of Economic Surveys, 30(5), 855-883. Roemer, J.E., A. Trannoy (2015). Equality of opportunity, in A.B. Atkinson and F. Bourguignon (eds). Handbook of Income Distribution, vol.2B, Amsterdam: North Holland. 12 Solon, G. (2004). “A Model of Intergenerational Mobility Variation over Time and Place.” Chapter 2 in Miles Corak (ed.), Generational Income Mobility in North America and Europe Cambridge, UK: Cambridge University Press 13 Figure 1 – Italy, by survey Italy Relative inequality of opportunity Return to education .09 .5 .4 .49 .38 .08 st.deviation logs mean log dev. .48 .36 .07 .47 .34 .46 .32 .06 .45 .3 1995 2000 2005 2010 2015 .05 year of survey 1995 2000 2005 2010 2015 st.deviation logs mean log dev. year of survey Parental networking Intergenerational persistence in education .04 .5 .03 .45 .02 .01 .4 1995 2000 2005 2010 2015 1995 2000 2005 2010 2015 year of survey year of survey Regressors include gender, age, age², born in South Italy and foreign citizenship Figure 2 – Italy, age-cohort decomposition Italy Inequality of opportunity - Standard deviation of logs personal incomes Return to education .08 .1 .06 .08 beta1 beta1 .04 .06 .02 .04 0 30 40 50 60 70 80 30 40 50 60 70 80 Age Age . . .15 .05 .04 .1 .02 .03 beta1 beta1 .05 .01 0 0 1920 1940 1960 1980 1920 1940 1960 1980 Born in... Born in... . . Parental networking Intergenerational persistence in education -.04 -.02 0 -.01 -.08 -.06 beta1 beta1 -.02 -.1 -.03 -.12 30 40 50 60 70 80 30 40 50 60 70 80 Age Age . . .3 0 .2 -.02 beta1 beta1 .1 -.04 0 -.06 -.1 1920 1940 1960 1980 1920 1940 1960 1980 Born in... Born in... . . Regressors include gender, age, age², born in South Italy and foreign citizenship - Age and cohort effects obtained from Deaton's decomposition 14 Figure 3 – Germany, by survey Germany Relative inequality of opportunity Return to education .55 .09 .5 .5 .4 st.deviation logs mean log dev. .08 .45 .3 .07 .4 .2 1980 1990 2000 2010 2020 .06 year of survey 1980 1990 2000 2010 2020 st.deviation logs mean log dev. year of survey Parental networking Intergenerational persistence in education .03 .75 .02 .7 .01 0 .65 -.01 -.02 .6 1980 1990 2000 2010 2020 1980 1990 2000 2010 2020 year of survey year of survey Regressors include gender, age, age², born in East Germany and foreign citizenship Figure 4 – Germany, age-cohort decomposition Germany Inequality of opportunity - Standard deviation of logs personal incomes Return to education .4 .04 .3 .03 beta1 beta1 .02 .2 .01 .1 0 0 30 40 50 60 70 80 30 40 50 60 70 80 Age Age . . .5 .03 .02 .4 .01 beta1 beta1 .3 0 -.02 -.01 .2 .1 1920 1940 1960 1980 2000 1920 1940 1960 1980 2000 Born in... Born in... . . Parental networking Intergenerational persistence in education .02 0 0 -.05 beta1 beta1 -.02 -.1 -.04 -.15 -.06 -.2 30 40 50 60 70 80 30 40 50 60 70 80 Age Age . . .05 .2 0 0 beta1 beta1 -.2 -.05 -.4 -.1 1920 1940 1960 1980 2000 1920 1940 1960 1980 2000 Born in... Born in... . . Regressors include gender, age, age², born in East Germany and foreign citizenship - Age and cohort effects obtained from Deaton's decomposition 15 Figure 5 – France, by survey France Relative inequality of opportunity Return to education .065 .46 .3 .44 .28 .06 st.deviation logs mean log dev. .42 .26 .055 .24 .4 .38 .22 .05 .36 .2 1980 1985 1990 1995 2000 2005 .045 year of survey 1980 1985 1990 1995 2000 2005 st.deviation logs mean log dev. year of survey Parental networking Intergenerational persistence in education 5 .2 4.5 .15 .1 4 .05 3.5 1980 1985 1990 1995 2000 2005 1980 1985 1990 1995 2000 2005 year of survey year of survey Regressors include gender, age, age² and foreign citizenship Parental education is absent and is replaced by dummy indicating middle-high parental occupations Figure 6 – France, age-cohort decomposition France Inequality of opportunity - Standard deviation of logs personal incomes Return to education .04 .1 .03 .05 beta1 beta1 .02 0 .01 -.05 0 30 40 50 60 70 80 30 40 50 60 70 80 Age Age . . .06 .2 .04 .15 beta1 beta1 .02 .1 0 .05 -.02 1920 1940 1960 1980 2000 1920 1940 1960 1980 2000 Born in... Born in... . . Parental networking Intergenerational persistence in education -.2 .1 .05 -.4 -.05 0 beta1 beta1 -.6 -.8 -.1 -.15 -1 30 40 50 60 70 80 30 40 50 60 70 80 Age Age . . .2 2 1.5 .1 beta1 beta1 1 0 .5 -.1 0 -.2 1920 1940 1960 1980 2000 1920 1940 1960 1980 2000 Born in... Born in... . . Regressors include gender, age, age² and foreign citizenship - Age and cohort effects obtained from Deaton's decomposition Parental education is absent and is replaced by dummy indicating middle-high parental occupations 16 Figure 7 – Great Britain, by survey GreatBritain Relative inequality of opportunity Return to education .35 .25 .5 .3 .45 st.deviation logs mean log dev. .2 .25 .4 .2 .35 .15 .15 .3 .1 1990 1995 2000 2005 2010 2015 .1 survey 1990 1995 2000 2005 2010 2015 st.deviation logs mean log dev. survey Parental networking Intergenerational persistence in education .15 .025 .02 .14 .015 .13 .01 .12 .005 .11 0 1990 1995 2000 2005 2010 2015 1990 1995 2000 2005 2010 2015 survey survey Regressors include gender, age, age², born in England and foreign citizenship Data trimmed at 99.5th centile - vertical dashed line indicates change of survey Figure 8 – Great Britain, age-cohort decomposition GreatBritain Inequality of opportunity - Standard deviation of logs personal incomes Return to education .4 .15 .3 .1 beta1 beta1 .2 .05 .1 0 0 30 40 50 60 70 80 30 40 50 60 70 80 Age Age . . .25 .8 .6 .2 beta1 beta1 .15 .4 .2 .1 .05 0 1920 1940 1960 1980 2000 1920 1940 1960 1980 2000 Born in... Born in... . . Parental networking Intergenerational persistence in education .005 .03 0 .02 beta1 beta1 -.005 .01 -.01 -.015 0 30 40 50 60 70 80 30 40 50 60 70 80 Age Age . . .15 0 -.03 -.02 -.01 .1 beta1 beta1 .05 0 -.04 -.05 -.05 1920 1940 1960 1980 2000 1920 1940 1960 1980 2000 Born in... Born in... . . Regressors include gender, age, age², born in England and foreign citizenship - Age and cohort effects obtained from Deaton's decomposition 17 Figure 9 – Switzerland, by survey Switzerland Relative inequality of opportunity Return to education .24 .12 .38 .22 .36 .11 st.deviation logs mean log dev. .2 .34 .1 .18 .32 .09 .16 .3 .14 .08 2000 2005 2010 2015 year of survey 2000 2005 2010 2015 st.deviation logs mean log dev. year of survey Parental networking Intergenerational persistence in education .05 .38 .36 .04 .34 .03 .32 .02 .3 .01 .28 2000 2005 2010 2015 2000 2005 2010 2015 year of survey year of survey Regressors include gender, age, age² and foreign citizenship Figure 10 – Switzerland, age-cohort decomposition Switzerland Inequality of opportunity - Standard deviation of logs personal incomes Return to education .05 .3 .04 .2 .02 .03 beta1 beta1 .1 0 .01 -.1 0 30 40 50 60 70 80 30 40 50 60 70 80 Age Age . . .25 .04 .02 .2 .15 0 beta1 beta1 -.02 .1 -.04 .05 -.06 0 1920 1930 1940 1950 1960 1920 1930 1940 1950 1960 Born in... Born in... . . Parental networking Intergenerational persistence in education .02 0 -.1 0 beta1 beta1 -.02 -.2 -.04 -.3 30 40 50 60 70 80 30 40 50 60 70 80 Age Age . . .02 .05 0 0 beta1 beta1 -.05 -.02 -.1 -.04 -.15 1920 1930 1940 1950 1960 1920 1930 1940 1950 1960 Born in... Born in... . . Regressors include gender, age, age² and foreign citizenship - Age and cohort effects obtained from Deaton's decomposition 18 Appendix Table 1 – Descriptive statistics - Italy highest st.deviation highest years personal personal respondent respondent years of logs of education in fraction of survey disposable disposable years of years of education in fraction of observations personal the parental born year income income education education the parental women disposable couple abroad (mean) (median) (mean) (st.deviation) couple incomes (sd.deviation) (mean) Italy 1993 12851 17491.9 15335.0 1.21 7.90 4.32 4.52 4.17 0.52 0.00 1995 12875 17103.5 15019.8 1.21 8.16 4.38 4.55 4.14 0.52 0.00 1998 11275 18497.0 16457.8 1.21 8.95 4.30 5.20 4.21 0.52 0.00 2000 11280 18827.7 16973.7 1.19 8.94 4.25 5.04 4.13 0.51 0.00 2002 10161 18797.5 16839.8 1.21 8.94 4.17 5.21 4.13 0.52 0.00 2004 9983 19741.8 17396.7 1.17 9.18 4.15 5.25 4.24 0.52 0.00 2006 9734 20611.4 18504.9 1.15 9.55 4.01 5.53 4.11 0.52 0.02 2008 6239 22629.3 19974.7 0.92 9.70 4.05 5.58 4.16 0.36 0.04 2010 6127 22123.2 19667.8 0.95 10.11 4.02 5.89 4.20 0.43 0.04 2012 6179 20435.3 18239.1 0.94 10.22 4.02 5.96 4.26 0.43 0.07 2014 11142 17817.8 16666.9 1.11 9.99 3.99 5.78 4.08 0.53 0.07 Total 107846 19065.8 17129.5 1.15 9.09 4.24 5.23 4.19 0.50 0.02 Table 2 – Descriptive statistics – Germany highest st.deviation highest years personal personal respondent respondent years of logs of education in fraction of survey disposable disposable years of years of education in fraction of observations personal the parental born year income income education education the parental women disposable couple abroad (mean) (median) (mean) (st.deviation) couple incomes (sd.deviation) (mean) Germany 1984 7034 15832.1 14558.9 1.57 10.38 3.16 8.50 2.68 0.51 0.24 1987 6833 17040.5 15627.8 1.50 10.45 3.17 8.54 2.65 0.51 0.24 1991 9270 23964.3 19590.6 1.23 11.18 3.47 8.82 2.31 0.52 0.17 1992 9118 24713.8 21100.3 1.21 11.21 3.46 8.86 2.28 0.52 0.17 1995 9343 25353.1 21669.0 1.17 11.37 3.46 8.89 2.26 0.52 0.18 1998 10002 26218.4 22023.8 1.09 11.49 3.48 9.03 2.14 0.53 0.15 2001 17188 32599.4 23837.3 1.11 12.08 3.57 9.34 1.94 0.52 0.12 2004 15349 31976.3 23460.1 1.09 12.20 3.60 9.42 1.91 0.52 0.11 2007 14611 31331.3 22767.6 1.05 12.33 3.62 9.52 1.85 0.52 0.09 2010 16010 29897.0 22305.6 1.03 12.32 3.62 9.61 1.78 0.53 0.09 2013 18709 30436.0 23221.5 0.98 12.49 3.65 9.78 1.80 0.55 0.09 Total 133467 27957.3 21313.8 1.18 11.82 3.59 9.25 2.11 0.53 0.13 Table 3 – Descriptive statistics – France st.deviation fraction of personal personal respondent respondent fraction of logs parents in fraction of survey disposable disposable years of years of parents in top fraction of observations personal top born year income income education education occupations women disposable occupations abroad (mean) (median) (mean) (st.deviation) (st.dev) incomes (mean) France 1978 13617 22298.4 18697.3 1.22 6.99 5.28 0.13 0.34 0.47 0.05 1984 15921 18460.3 16610.8 1.10 6.71 5.01 0.14 0.35 0.50 0.04 1989 12411 18854.2 16599.4 1.02 7.19 5.07 0.16 0.37 0.50 0.04 1994 16275 20397.3 17392.7 1.12 8.31 5.00 0.19 0.39 0.52 0.08 2000 15623 20749.7 17747.5 1.02 8.74 5.02 0.21 0.41 0.53 0.10 2005 15272 21892.6 18936.3 0.98 9.37 5.05 0.24 0.42 0.53 0.12 Total 89119 20444.9 17646.2 1.08 7.92 5.16 0.18 0.38 0.51 0.07 19 Table 4 – Descriptive statistics – Great Britain highest st.deviation highest years personal personal respondent respondent years of logs of education in fraction of survey disposable disposable years of years of education in fraction of observations personal the parental born year income income education education the parental women disposable couple abroad (mean) (median) (mean) (st.deviation) couple incomes (sd.deviation) (mean) Great Britain 1991 4250 9628.8 7793.0 1.05 10.80 1.33 9.86 2.55 0.56 0.06 1992 4344 10175.4 8418.7 1.02 10.83 1.32 9.90 2.58 0.56 0.06 1993 4444 10487.5 8582.7 1.01 10.85 1.31 9.94 2.61 0.56 0.06 1994 4599 10748.2 8651.2 1.01 10.87 1.31 9.99 2.62 0.56 0.05 1995 4752 11356.6 9149.7 1.00 10.89 1.31 10.04 2.66 0.55 0.05 1996 4988 11775.5 9684.9 0.98 10.92 1.31 10.07 2.66 0.55 0.05 1997 5125 12343.4 10279.9 0.99 10.93 1.30 10.11 2.68 0.55 0.05 1998 5276 12673.5 10487.1 0.98 10.95 1.29 10.14 2.68 0.55 0.05 1999 7974 12660.5 10461.3 0.97 10.94 1.27 10.11 2.67 0.55 0.05 2000 8382 13478.0 11081.8 0.95 10.95 1.26 10.13 2.67 0.55 0.05 2001 10457 13865.6 11349.4 0.91 10.97 1.28 10.03 2.64 0.55 0.05 2002 10629 14628.7 11920.2 0.94 10.99 1.27 10.07 2.67 0.55 0.05 2003 11149 15243.9 12451.8 0.92 11.02 1.27 10.11 2.68 0.54 0.05 2004 10339 15838.2 13100.0 0.89 11.04 1.26 10.14 2.71 0.55 0.04 2005 9950 16374.9 13511.4 0.90 11.05 1.25 10.16 2.71 0.55 0.05 2006 9540 17001.2 13916.2 0.87 11.06 1.25 10.17 2.71 0.55 0.04 2007 9000 17734.9 14355.5 0.88 11.08 1.24 10.19 2.73 0.55 0.04 2008 8553 18462.5 15011.6 0.87 11.10 1.22 10.21 2.74 0.55 0.04 2009 28934 19932.8 15814.4 0.99 11.26 1.28 10.62 3.05 0.56 0.16 2010 35477 20650.6 16680.0 0.92 11.26 1.26 10.59 3.02 0.56 0.14 2011 30910 21255.4 17324.6 0.92 11.28 1.25 10.62 3.02 0.56 0.13 2012 28631 21792.4 17696.6 0.92 11.31 1.24 10.68 3.05 0.56 0.13 2013 26803 22235.6 18004.2 0.91 11.33 1.23 10.72 3.07 0.56 0.13 2014 24119 23403.6 18828.8 0.94 11.35 1.23 10.76 3.09 0.56 0.13 Total 308625 18357.2 14641.7 0.97 11.16 1.27 10.42 2.91 0.56 0.10 Table 5 – Descriptive statistics – Switzerland highest st.deviation highest years personal personal respondent respondent years of logs of education in fraction of survey disposable disposable years of years of education in fraction of observations personal the parental born year income income education education the parental women disposable couple abroad (mean) (median) (mean) (st.deviation) couple incomes (sd.deviation) (mean) Switzerland 1999 4327 63707.1 57579.3 1.19 12.81 2.08 11.76 2.30 0.52 0.00 2002 3737 62533.1 54500.3 1.22 12.93 2.10 11.82 2.30 0.54 0.00 2005 5006 64389.9 54462.5 1.22 13.09 2.11 11.93 2.31 0.55 0.15 2008 5373 64798.3 55044.9 1.24 13.17 2.13 11.93 2.31 0.56 0.15 2011 5341 70051.9 58400.3 1.13 13.24 2.13 11.96 2.31 0.55 0.15 2014 7489 72643.8 60558.3 1.15 13.40 2.18 11.98 2.48 0.53 0.16 Total 31273 67087.3 57076.7 1.19 13.15 2.14 11.91 2.35 0.54 0.12 20 Table 6 – Inequality and inequality of opportunity - Italy 1 2 3 4 5 6 mean log st.dev.log relative deviation relative predicted mean log st.dev.log inequality of predicted inequality of survey incomes deviation incomes opportunity incomes opportunity (absolute incomes (2/1) (absolute (5/4) IOp) IOp) Italy 1993 1.206 0.580 0.481 0.448 0.166 0.370 1995 1.206 0.562 0.466 0.440 0.158 0.358 1998 1.214 0.587 0.483 0.458 0.170 0.371 2000 1.190 0.592 0.497 0.425 0.174 0.409 2002 1.207 0.588 0.487 0.418 0.171 0.408 2004 1.171 0.580 0.496 0.414 0.166 0.402 2006 1.145 0.542 0.473 0.384 0.144 0.375 2008 0.921 0.415 0.450 0.267 0.084 0.314 2010 0.946 0.441 0.466 0.298 0.095 0.320 2012 0.941 0.423 0.450 0.294 0.088 0.300 2014 1.108 0.523 0.471 0.363 0.137 0.377 Total 1.140 0.545 0.477 0.397 0.148 0.370 Table 7 – Inequality and inequality of opportunity - Germany 1 2 3 4 5 6 mean log st.dev.log relative deviation relative predicted mean log st.dev.log inequality of predicted inequality of survey incomes deviation incomes opportunity incomes opportunity (absolute incomes (2/1) (absolute (5/4) IOp) IOp) Germany 1984 1.569 0.841 0.536 0.669 0.325 0.486 1987 1.495 0.762 0.510 0.619 0.271 0.438 1991 1.232 0.619 0.502 0.469 0.185 0.394 1992 1.216 0.613 0.504 0.456 0.181 0.397 1995 1.177 0.547 0.465 0.435 0.145 0.334 1998 1.099 0.488 0.444 0.400 0.116 0.291 2001 1.112 0.484 0.435 0.467 0.114 0.244 2004 1.090 0.457 0.419 0.449 0.102 0.227 2007 1.048 0.454 0.433 0.433 0.100 0.231 2010 1.032 0.431 0.418 0.407 0.091 0.224 2013 0.980 0.403 0.411 0.387 0.080 0.206 Total 1.136 0.515 0.449 0.453 0.134 0.286 Table 8 – Inequality and inequality of opportunity – France 1 2 3 4 5 6 mean log st.dev.log relative deviation relative predicted mean log st.dev.log inequality of predicted inequality of survey incomes deviation incomes opportunity incomes opportunity (absolute incomes (2/1) (absolute (5/4) IOp) IOp) France 1978 1.22 0.558 0.457 0.505 0.148 0.293 1984 1.099 0.471 0.429 0.399 0.107 0.269 1989 1.02 0.428 0.419 0.363 0.09 0.247 1994 1.121 0.444 0.396 0.398 0.098 0.245 2000 1.019 0.406 0.399 0.347 0.082 0.238 2005 0.981 0.363 0.37 0.32 0.066 0.206 Total 1.076 0.444 0.411 0.387 0.098 0.249 21 Table 9 – Inequality and inequality of opportunity – Great Britain 1 2 3 4 5 6 mean log st.dev.log relative deviation relative predicted mean log st.dev.log inequality of predicted inequality of survey incomes deviation incomes opportunity incomes opportunity (absolute incomes (2/1) (absolute (5/4) IOp) IOp) Great Britain 1991 1.011 0.510 0.505 0.391 0.129 0.329 1992 0.994 0.473 0.476 0.378 0.111 0.294 1993 0.983 0.467 0.475 0.369 0.108 0.293 1994 0.989 0.456 0.461 0.369 0.103 0.278 1995 0.985 0.445 0.451 0.368 0.098 0.267 1996 0.966 0.418 0.433 0.353 0.087 0.246 1997 0.954 0.441 0.462 0.346 0.096 0.277 1998 0.947 0.437 0.462 0.343 0.094 0.275 1999 0.947 0.416 0.440 0.337 0.086 0.254 2000 0.925 0.415 0.448 0.325 0.085 0.260 2001 0.904 0.425 0.470 0.318 0.089 0.279 2002 0.936 0.416 0.444 0.332 0.084 0.254 2003 0.911 0.406 0.446 0.322 0.080 0.250 2004 0.886 0.394 0.445 0.303 0.076 0.251 2005 0.899 0.390 0.434 0.306 0.075 0.244 2006 0.874 0.353 0.404 0.295 0.062 0.208 2007 0.878 0.354 0.403 0.304 0.062 0.203 2008 0.857 0.358 0.417 0.291 0.063 0.216 2009 0.991 0.329 0.332 0.360 0.053 0.146 2010 0.926 0.301 0.325 0.324 0.045 0.138 2011 0.924 0.290 0.314 0.317 0.042 0.132 2012 0.925 0.288 0.311 0.315 0.041 0.130 2013 0.920 0.282 0.307 0.311 0.040 0.127 2014 0.933 0.290 0.311 0.317 0.042 0.133 Total 0.933 0.350 0.375 0.327 0.063 0.190 Table 10 – Inequality and inequality of opportunity – Switzerland 1 2 3 4 5 6 mean log st.dev.log relative relative deviation predicted inequality mean log inequality st.dev.log predicted survey incomes of deviation of incomes incomes (absolute opportunity incomes opportunity (absolute IOp) (2/1) (5/4) IOp) Switzerland 1999 1.194 0.456 0.382 0.428 0.102 0.237 2002 1.223 0.448 0.366 0.449 0.100 0.222 2005 1.225 0.386 0.315 0.496 0.075 0.150 2008 1.240 0.370 0.298 0.491 0.069 0.140 2011 1.132 0.381 0.337 0.454 0.073 0.160 2014 1.149 0.369 0.322 0.447 0.068 0.151 Total 1.189 0.396 0.333 0.461 0.078 0.171 22 Table 11 – Estimation of relevant equations (7)-(9)-(10), by country full sample Italy Germany France Great Britain Switzerland 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 log log log log log log log log log log years of personal personal years of personal personal years of personal personal years of personal personal years of personal personal dep.variable education disposable disposable education disposable disposable education disposable disposable education disposable disposable education disposable disposable income income income income income income income income income income female -0.664*** -0.785*** -0.834*** -0.860*** -0.928*** -0.989*** -0.509*** -0.779*** -0.807*** -0.042*** -0.537*** -0.542*** -0.930*** -0.650*** -0.738*** [0.027] [0.008] [0.008] [0.022] [0.007] [0.008] [0.033] [0.007] [0.007] [0.005] [0.004] [0.004] [0.028] [0.015] [0.015] age -0.089*** 0.029*** 0.034*** -0.019*** 0.012*** 0.015*** -0.103*** 0.023*** 0.020*** -0.022*** 0.021*** 0.027*** -0.020*** 0.024*** 0.026*** [0.001] [0.002] [0.002] [0.001] [0.002] [0.002] [0.001] [0.002] [0.002] [0.000] [0.001] [0.001] [0.001] [0.004] [0.004] age² -0.000*** -0.000*** -0.000*** -0.000*** -0.000*** -0.000*** -0.000*** -0.000*** -0.000*** -0.000*** [0.000] [0.000] [0.000] [0.000] [0.000] [0.000] [0.000] [0.000] [0.000] [0.000] years of education 0.078*** 0.072*** 0.054*** 0.132*** 0.095*** [0.001] [0.001] [0.001] [0.002] [0.004] parental education (yrs) 0.460*** 0.022*** 0.058*** 0.667*** 0.005** 0.054*** 3.953*** 0.113*** 0.328*** 0.114*** 0.018*** 0.033*** 0.325*** 0.023*** 0.054*** [0.003] [0.001] [0.001] [0.008] [0.002] [0.002] [0.042] [0.009] [0.009] [0.001] [0.001] [0.001] [0.007] [0.004] [0.004] born in a specific regions -0.602*** -0.378*** -0.426*** 0.666*** -0.184*** -0.136*** -0.026*** 0.005 0.001 [0.028] [0.009] [0.009] [0.029] [0.007] [0.008] [0.006] [0.004] [0.005] born abroad -0.685*** -0.475*** -0.524*** 0.375*** -0.253*** -0.227*** -2.199*** -0.105*** -0.225*** 0.376*** -0.130*** -0.080*** -0.013 -0.147*** -0.149*** [0.100] [0.032] [0.031] [0.043] [0.015] [0.015] [0.073] [0.013] [0.013] [0.013] [0.008] [0.008] [0.051] [0.026] [0.027] constant 10.901*** 8.052*** 8.591*** 6.063*** 8.574*** 8.897*** 11.077*** 8.922*** 9.458*** 10.678*** 7.157*** 8.352*** 10.380*** 8.874*** 9.759*** [0.075] [0.067] [0.068] [0.092] [0.055] [0.056] [0.070] [0.039] [0.040] [0.023] [0.033] [0.029] [0.103] [0.110] [0.103] Observations 107846 107846 107846 133253 133253 133253 89119 89119 89119 259608 259608 259608 30984 30984 30984 R² 0.439 0.285 0.239 0.162 0.277 0.244 0.241 0.229 0.175 0.209 0.222 0.199 0.211 0.144 0.119 Robust standard errors in brackets - sample weights - survey dummies included - statistical significance *** p<0.01, ** p<0.05, * p<0.1 Specific regions include South for Italy, East for Germany, England for Great Britain.- parental education for France correspond to high occupations 23 Table 12 – Estimation by age-cohort subgroups – Italy number of observations age groups birth cohorts 25-29 30-34 35-39 40-44 45-49 50-54 55-59 60-64 65-69 70-74 75-80 Total (1910-1914) 0 0 0 0 0 0 0 0 0 0 199 199 (1915-1919) 0 0 0 0 0 0 0 0 0 126 772 898 (1920-1924) 0 0 0 0 0 0 0 0 205 1638 1433 3276 (1925-1929) 0 0 0 0 0 0 0 240 2076 1708 1724 5748 (1930-1934) 0 0 0 0 0 0 266 2267 2165 2026 1607 8331 (1935-1939) 0 0 0 0 0 322 2512 2535 2656 1676 1568 11269 (1940-1944) 0 0 0 0 285 2616 2677 2643 1643 1705 0 11569 (1945-1949) 0 0 0 286 2896 3047 3017 1947 1956 0 0 13149 (1950-1954) 0 0 270 2482 3052 3112 1866 2018 0 0 0 12800 (1955-1959) 0 259 2395 2830 2914 1855 1847 0 0 0 0 12100 (1960-1964) 194 2068 2663 2921 1895 2028 0 0 0 0 0 11769 (1965-1969) 1047 1868 2386 1732 1730 0 0 0 0 0 0 8763 (1970-1974) 787 1479 1157 1351 0 0 0 0 0 0 0 4774 (1975-1979) 593 681 872 0 0 0 0 0 0 0 0 2146 (1980-1984) 343 508 0 0 0 0 0 0 0 0 0 851 (1985-1989) 204 0 0 0 0 0 0 0 0 0 0 204 Total 3168 6863 9743 11602 12772 12980 12185 11650 10701 8879 7303 107846 (relative) inequality of opportunity age groups birth cohorts 25-29 30-34 35-39 40-44 45-49 50-54 55-59 60-64 65-69 70-74 75-80 (1915-1919) 0.401 (1920-1924) 0.381 0.402 (1925-1929) 0.43 0.483 0.442 (1930-1934) 0.489 0.482 0.450 0.371 (1935-1939) 0.500 0.495 0.456 0.470 0.402 (1940-1944) 0.524 0.501 0.530 0.526 0.508 (1945-1949) 0.466 0.542 0.526 0.440 0.475 (1950-1954) 0.476 0.506 0.489 0.472 0.449 (1955-1959) 0.505 0.509 0.530 0.505 0.455 (1960-1964) 0.503 0.508 0.483 0.463 0.505 (1965-1969) 0.465 0.502 0.462 0.477 0.494 (1970-1974) 0.454 0.476 0.404 0.481 (1975-1979) 0.431 0.406 0.438 (1980-1984) 0.417 24 Table 13 – Deaton’s decomposition by age-cohort subgroups – Italy - OLS 1 2 unconstrained constrained dep.variable IOp st.dev.log IOp st.dev.log age=27 0.007 0.089** [0.034] [0.033] age=32 0.032 0.101*** [0.026] [0.031] age=37 0.022 0.084*** [0.028] [0.029] age=42 0.04 0.094*** [0.026] [0.028] age=47 0.043* 0.090*** [0.024] [0.026] age=52 0.060** 0.099*** [0.022] [0.025] age=57 0.042* 0.072*** [0.021] [0.023] age=62 0.040** 0.063*** [0.019] [0.021] age=67 0.037** 0.052** [0.018] [0.020] age=72 0.039** 0.046** [0.017] [0.019] birth=1917 0.093* [0.050] birth=1922 -0.039 0.05 [0.031] [0.044] birth=1927 0.017 0.104** [0.029] [0.042] birth=1932 0.018 0.092** [0.027] [0.040] birth=1937 0.035 0.102** [0.025] [0.039] birth=1942 0.076*** 0.136*** [0.025] [0.038] birth=1947 0.048* 0.099** [0.025] [0.037] birth=1952 0.036 0.079** [0.025] [0.036] birth=1957 0.062** 0.098*** [0.025] [0.035] birth=1962 0.056** 0.083** [0.025] [0.033] birth=1967 0.054** 0.073** [0.026] [0.033] birth=1972 0.032 0.044 [0.028] [0.033] birth=1977 0.017 0.02 [0.030] [0.034] survey=1994 0.016 -0.007* [0.014] [0.004] survey=1999 0.035** 0.012* [0.014] [0.007] survey=2004 0.02 -0.005* [0.012] [0.003] survey=2009 -0.011 [0.012] Constant 0.385*** 0.315*** [0.021] [0.041] Observations 53 53 R-squared 0.81 Standard errors in brackets - statistical significance *** p<0.01, ** p<0.05, * p<0.1 Constraints: ( 1) - survey1 - survey2 - survey3 - omitted.survey4 - omitted.survey5 = 0 ( 2) - survey1 - 5*survey2 - 10*survey3 - 15*omitted.survey4 - 20*oomitted.survey5 = 0 25 Figure 11 – Age profiles for inequality of opportunity, by birth cohorts - Italy Inequality of opportunity - standard deviation of logs - Italy .55 .5 .45 .4 .35 25-29 30-34 35-39 40-44 45-49 50-54 55-59 60-64 65-69 70-74 75-80 age birth1920-24 birth1925-29 birth1930-34 birth1935-39 birth1940-44 birth1945-49 birth1950-54 birth1955-59 birth1960-64 birth1965-69 birth1970-74 birth1975-79 Figure 12 – Profiles for inequality of opportunity, by birth and cohorts – Italy Inequality of opportunity - Standard deviation of logs personal incomes .1 .08 beta1 .06 .04 30 40 50 60 70 80 Age . .15 .1 beta1 .05 0 1920 1940 1960 1980 Born in... . 26