University of Amsterdam
Amsterdam School of Economics
Inequality and social mobility: is tertiary education a
stratifying or equalizing force?
Mila Quacquarelli
11777591
BSc Economics and Business Economics
Thesis supervisor: Kees Haasnoot
ii
Statement of Originality
This document is written by Student Mila Quacquarelli who declares to take full responsibility for the contents of this document.
I declare that the text and the work presented in this document are original and that no sources other than those mentioned in the text and its references have been used in creating it.
The Faculty of Economics and Business is responsible solely for the supervision of completion of the work, not for the contents.
Abstract:
Accessing top-performing universities is an important factor for achieving or, rather, maintaining a certain status-quo. A high correlation between family background and attendance of prestigious universities can be seen as a symptom of an educational channel that reinforces inequalities, where intergenerational immobility is supported by selective and expensive institutions. This research explores the link between the degree of social mobility and the nature of the tertiary education sector on a country level. By performing a cross-sectional analysis on a sample of 82 countries, the paper concludes that an elitist higher educational environment negatively affects the impact of tertiary education on social mobility.
Keywords: Education and Inequality, Equality of Opportunity, Social Stratification,
Higher Education, Human Capital Investment.
iv
Table of Contents
Introduction ... 1
Theoretical Framework ... 3
i. Inequality and economic growth ... 4
ii. Inequality and social mobility, educational channel ... 6
iii. Recent trends in accumulation of income/wealth, social mobility and tertiary education ... 8 Hypotheses ... 11 Methodology ... 12 Models ... 12 Model 1 ... 12 Model 2 ... 12 Data ... 13 Results ... 17 Main findings ... 17
Ordinary least squares model ... 17
Ordered logistic model ... 19
Robustness analyses ... 20 Alternative measurements ... 20 Restricted sample ... 23 Discussion ... 24 Conclusion ... 26 References ... 27 Appendix A ... 29 Appendix B ... 31 Appendix C ... 32 Appendix D ... 34
Introduction
The recently published “World Inequality Report 2018” (WIR) provides evidence for a trend in income inequality that resembles very closely that of persistent inequality. Specifically, the top 1% earners of the US national income have seen their share of income increase by 20%, whereas the bottom 50th percentile income earners have
witnessed their share of national income shrink more and more between 1980 and 2016. Additionally, the report presents us with a depiction of the relationship between college attendance rate and parental income rank as shown in Figure 1: the almost 45° line is very telling of the role that education might play in allowing for intergenerational inequalities to become persistent in nature. The potential for education to serve as one of the channels for which inequalities become persistent can be perceived as a symptom of a badly functioning system.
Figure 1. Relationship between parental income rank and share of children who attended college.
Educational access might turn to be a tool that, indeed, enforces a certain degree of social stratification rather than allowing for more mobility. If there is a mechanism for which those who are already wealthy enter the educational channels that will drive them to the best-paying jobs, and the same situation does not hold for
2
less affluent individuals, persistent inequality is generated. The enduring nature that characterizes inequalities at their core levels will be the relevant reading key provided in this research. As it will be argued in the upcoming analysis, the persistency of a highly unequal income and wealth distributions are harmful to the economy because they lead to an underinvestment in human capital: education not only is one channel through which inequalities are perpetuated, but also a medium through which more innovation and knowledge can be provided to our society.
From a historical perspective, Beller and Hout (2006) point out that social movements towards more educational opportunities in the 1960s improved the lives of disadvantaged people thereafter. In their opinions, these achievements loosened the ties between occupational and income origins and destinations among college graduates. Nowadays, however, with skyrocketing costs where, for instance, the yearly tuition fees at Columbia University are of the same magnitude as the US GDP per capita, the role of education as a social lifter has become questionable. Furthermore, we observe a very heterogenous distribution of entry requirements, where top colleges have become substantially more selective compared to the 1960s and average colleges have not increased their selectivity, as Hoxby (2009) recognizes. One aspect of these criteria is often disregarded: the economic background of the prospective student’s family is a key player in the pre-enrolment educational performance and in the likelihood of meeting certain entry requirements. The educational possibilities given to individuals coming from a wealthier family are different than those achievable by the ones from a more troubled economic background. Recent empirical analyses (Kinsler & Pavan, 2011; Belley & Lochner 2007) found a positive and significant effect on the quality of college attended and family income by looking at the National Longitudinal Survey of Youth cohorts, both in 1979 and 1997.
Figure 2a presents the conceptual model that is behind this research and
provides context for the upcoming data analysis. The paper itself will focus only on empirically assessing the extent to which education serves as an equalizer or rather a stratifying force of the socio-economic conditions of its participants (relationship drawn in red). Put differently, the goal of this analysis is to evaluate whether the nature of tertiary education in a given country, i.e. whether it is elitist or not, has an impact on
the degree of intergenerational social mobility that is observed as a result of increased investment in the tertiary education sector. This enquiry will be explored by performing a cross-sectional analysis using both an ordinary least squares (OLS) model as well as an ordered logistic model. The other relationships depicted in Figure 2a are briefly discussed in the theoretical framework to provide the reader with more context on where this research is positioned with respects to the academic debate.
Figure 2a. Conceptual framework
The rest of this paper is structured as follows: firstly, the reader will be presented with theoretical concepts and empirical observations that provide context for the conceptual framework behind this research. Subsequently, the models being tested will be presented and described. Next, more information on the collection of data and the construction of the employed indices will be provided. Upon having presented the results and their robustness analysis, a discussion on the main limitations of this research – and possible ideas to overcome them in future analyses – will appear. Lastly, this paper will be concluded with a number of final remarks.
Theoretical Framework
The aim of this section is to provide the reader with the necessary theoretical underpinnings and empirical figures to firmly grasp the conceptual framework briefly introduced above. As outlined in Figure 2b, popular views on the relationship between inequality and growth will be presented first. Second, the link between inequality and educational access will be explored, together with the concept of social mobility. Finally, a number of empirical findings will be presented such to inform the reader of the recent trends in what appears as persistent inequality and its relationship to mobility.
4 Figure 2b. Conceptual framework – order of analyzed subtopics
i. Inequality and economic growth
The prevailing literature on the impact of inequality on economic growth reveals the perspective that is often taken when studying inequality. Although focusing on different channels and transmission mechanisms, it appears that inequality has been regarded as detrimental for growth not necessarily because of its very nature, but rather because of the consequent voting behaviors it entails. Specifically, models employed in the past aimed at showing that inequality slows economic growth via the preferences over taxation that the actors of the economy have. Assuming a democratic electoral system, the representative median voter coming from a more unequal society will prefer higher levels of taxation on capital, resulting in slowing economic growth by diminishing the speed and efficiency at which capital produces returns. Put differently, the redistributive policies arising to tackle inequality introduce distortions, damaging growth (Alesina & Rodrik, 1994; Persson & Tabellini, 1991).
Contrary to Alesina and Rodrik (1994), Persson and Tabellini (1991) allow for an intergenerational transfer of wealth. In this way, the average capital stock accumulated by the previous generation has a positive externality effect on the initial endowment of tangible resources passed on to the next generation. Another major difference lays on the emphasis given by the researchers on the endowment of skills (whereas Alesina and Rodrik (1994) only distinguish between unskilled workers and capital owners). Both studies reach the same above-mentioned conclusion, but Persson and Tabellini (1991) make a crucial assumption: the level of capital ownership is strictly related to the level of skills individuals are endowed with. Here, the link between possession of specific skills and the possibility to achieve higher returns on personal investments appears to be problematic. This is because skill levels are perceived as the only vehicle differentiating people’s attainment of better investments. However, this link is highly challenged when higher-return investments are possible
based on the educational training that individuals receive. From this perspective, two individuals endowed with similar skills are likely not to reach the same outcome if there are substantial differences in their financial capabilities. Especially once intergenerational transfers of wealth are allowed for, it becomes apparent to see how these two similarly skilled individuals might have very different career paths.
Perotti (1992) emphasizes another important characteristic of the model presented by Persson and Tabellini (1991): the taxation policy is also employed for financing public education, which is considered by the author as a factor that enhances growth. This positive effect of public education on growth is also pointed out by Saint-Paul and Verdier (1993). Interestingly, they mention how the relative cost of education decreases as we move forward in the global income distribution: in a very poor economy, only the upper income classes will be able to invest in human capital, realizing the potential skills which people are born with. This situation is not ideal, as the contribution of human capital is valuable for economic growth and, therefore, to be encouraged as much as possible.
The importance of education for economic growth is further analyzed by Saint-Paul and Verdier (1993), presenting us with the following scenario: higher inequality, together with very well responding political and educational systems, will lead to a transitory higher economic growth that results from an investment in public education. The intuition goes as follows. In a more unequal society where voting rights are extended to everyone, people in the lower income bracket will need to benefit from public education to achieve a certain degree of investment in human capital. Because of this, and given that redistributive taxation is employed to finance public schools, the lower earners will demand higher taxation of the higher earners’ returns. The subsequent investment in public education is not only seen as beneficial for the low-income voter who advocated for such policy, but also for individuals with no voting rights due to their recent arrival in the country under scrutiny. In sum, the authors provide us with an overview of generational benefit from government-financed educational institutions. This scenario reaches an opposite conclusion to what presented above: higher inequality leads to a temporary increase in growth rather than its slowdown due to voting behavior. This perspective is particularly beneficial as it emphasizes other drivers of growth, rather than solely attributing it to capital
6
accumulation. Specifically, this idea allows us to reflect on the importance of education to achieve a better level of economic well-being and higher rates of growth.
In addition to that, Galor and Zeira (1993) show that there is a substantial problem in the credit market for which borrowing is constrained. Because of it, the initial wealth endowment is a crucial determinant of whether higher education can be accessed or not. Given this constraint, intergenerational transfers of wealth appear to strongly influence the factors determining economic growth via the channel of education (Perotti, 1996). Finally, this analysis reflects a structural issue that is found in the context of a highly unequal distribution of resources: the inability to materialize potential skills is not only detrimental for the individuals unable to access higher education but also for the economy as a whole.
ii. Inequality and social mobility, educational channel
Together with a number of other economists, Becker and Tomes (1979) argue that bright minds will somehow find the way to finance their high-quality education regardless of the possible initial adverse financial conditions they may face. Since innate abilities do not depend upon redistribution, the authors claim that investment in public education will not do any good in enhancing the human capital investment channel for fueling growth. In more recent years, however, Piketty (2000) has shown to condemn this reasoning. He claims that this phenomenon would only be realistic if the endowment of abilities was extraordinarily heterogenous, which is not what is observed. In reality, he supports the already outlined idea that innate talents and skills are important, but they are of a potential nature only. Their realization and subsequent translation into human capital investment fundamentally depends on the educational possibilities given to the individuals in question. As similarly argued by Brezis and Hellier (2018), human capital materialization depends on randomly distributed inherent talents but also on family background and the field of study that is chosen by the given individual in a given education system.
In addition to that, Piketty (2000) presents a crucial point for our upcoming analysis: dynasties with little initial wealth face limited investment opportunities and remain poor. This is not only detrimental for the dynasties’ themselves, but it also appears detrimental on a country-level because of the loss in potential human capital
investment that is hindered by the unfavorable initial wealth endowments. His point allows us to see persistent inequality and intergenerational immobility as the cause and effect of the presence of inequalities. Differently from the empirical researches outlined above, the detrimental effects of inequality are analyzed for their very nature rather than focusing on the unfavorable redistributive policies that result from them. This is a very important conclusive remark for the upcoming analysis.
Finally, Figure 3 provides further insights into the relationship between intergenerational mobility and inequality. Following the intuition outlined above, more unequal countries are presented as being less socially mobile too. Of course, the direction of causality in this relation is vague.
Figure 3. Relationship between Social Mobility and Gini coefficient for the analyzed countries in this research.
8
iii. Recent trends in accumulation of income/wealth, social mobility and tertiary education
After having observed how inequality affects growth via the human capital channel, we now turn to another strictly connected relationship, that of education on inequality itself. So, not only does education play a fundamental role to impact growth via the human capital transmission mechanism, but it also represents an important channel of itself for the extent to which inequality is perpetuated over generations. As mentioned by Piketty (2000) and Brezis and Hellier (2018), educational access and investment opportunities are very much connected to initial resource endowments, rather than skill endowments alone.
In the WIR (2018), we are presented with figures on the percentage change in incomes occurred in the past couple of decades. The report extends the time frame of the original author, Branko Milanovic, and provides us with recent evidence on the “elephant curve”, named after its peculiar shape (see Figure 4). This depiction is particularly important for our analysis as it empirically tells us the global trend of income growth depending on the global distribution of income. Data show that the upper middle class has seen very little relative increase in their incomes over the time period under scrutiny, whereas the very top global income earners have seen a very large relative increase in their incomes. In sum, the very rich became even richer in relative terms, phenomenon that can also be seen in Figure 5.
Figure 4. Elephant Curve.
10
Milanovic (2016) observes that this phenomenon has been largely motivated by little access to top performing educational institutions by the non-top earning individuals. If the wealthiest accrue even more resources thanks to the ability to attend the most prestigious schools, then intergenerational immobility can be attributed to a large extent to educational possibilities. Because of this, he stresses the need for allowing more access to education as a crucial means to address the large gap between the bottom 90th percentile of global earners (mainly found in Western Europe
and the US) and the very elite.
This point is also emphasized by Brezis and Hellier (2018), who observe that being admitted to elite colleges has become increasingly more difficult for people born in middle class families. Specifically, they show that in the US, SAT scores, which are fundamental for admission in American colleges, are highly correlated with family education and wealth (Brezis & Temin, 2008). Similarly, Albouy and Wanecq (2003) observe that since the end of WWII, the difference in likelihood of accessing the French Grandes écoles conditional on socio-economic background has followed a U-shaped curve. This finding is particularly striking, especially when combined with a generational admission rate to such institutions of 4% per academic year, as found by Brezis (2018).
Furthermore, Brezis (2018) observes that we have recently witnessed a phenomenon of “massification” in the intake of students in tertiary educational systems. Brezis and Hellier (2018) observe the same circumstance of world-wide democratization of higher education, reporting that the generational share of individuals pursuing tertiary education has shifted from about 10% to 60% in the present period. This increase in enrolment rates can be regarded as beneficial because of the consequent investments in human capital that stem from it. However, Su et al. (2012) observe that non-elitist public post-secondary colleges have increased their enrolment rate by 525% against 250% in elite colleges, between 1959 and 2008. Combining these results suggests that the overall ratio of tertiary educated students increased but that, conditional on the quality and prestige of universities, this increase has been unequal. In addition to that, Kinsler and Pavan (2011) stress that the costs of completing college in the US have increased to a particularly dramatic extent within
the top of the institutions’ quality distribution. Still in the US, Desrochers and Wellman (2011) report that Ivy League colleges devote three times more spending per student (i.e. proxy for educational quality) than other universities.
Summarizing the main points observed until now, it is possible to see a correspondence of quality and costs for which bottom income earners can hardly achieve the better quality institutions, whereas top income earners have the resources to potentially access better-performing universities. Rationale confirmed by, for instance, Chetty et al. (2017): they observe that Stony Brook University (New York) – taken as representative of the non-top colleges - has 16.4% of students coming from the bottom quintile, whereas Ivy-League colleges have 3.8% of students from the same income bracket.
Hypotheses
The theories covered in the previous section suggest that the materialization of human capital investments through the channel of education is beneficial for the individuals’ well-being and their careers. This being said, it has also been mentioned a recent phenomenon of higher enrolment rates in the tertiary sector, when compared to past rates. This process alone is expected to yield positive effects on intergenerational mobility, which leads to the first outcome predicted by this research:
Hypothesis 1: there is a positive relationship between investment in tertiary education and social mobility.
However, the above-mentioned theories and figures also suggest that, when considering the differentiated quality and accessibility of the given educational institutions, devoting more resources towards a sector that is highly concentrated with selective and expensive universities will not be as beneficial for those unable to access them. This leads to the second outcome predicted by this study:
Hypothesis 2: investment in tertiary education has a less positive effect on social mobility if the educational system is more elitist.
12
Methodology
As mentioned above, the only relationship within the conceptual framework that is empirically assessed in this paper is that between tertiary education and social mobility. First, the models that follow from the hypotheses outlined above will be tested using an ordinary least squares model. On top of the high interpretability of the resulting coefficients, OLS was opted for due to the properties entailed by it being the best linear unbiased estimator (BLUE). Second, the analysis will be performed using an ordered logistic model too. This choice was made in order to operate with a model that takes into account the bounded nature of the dependent variable. Further information on how the variables are composed, together with data collection processes are discussed in the “Data” section below.
Models
Model 1. Only (tertiary) education is taken into account, together with controls:
!"#! = &"+ &#∙ )*+! + &%∙ ,"--!+ &&∙ .! + /! (1)
In line with hypothesis 1, a positive value for coefficient β1 is expected.
Model 2. The interaction between (tertiary) education and elitism is also considered,
together with controls:
!"#! = &"+ &#∙ )*+!× )12!+ &%∙ )*+! + &&∙ )12! + &' ∙ ,"--! + &(∙ .! + 3! (2)
The above-mentioned hypotheses entail an expected negative β1,as well as a
positive β2.
The chosen depended variable, Mob, refers to intergenerational social mobility. The main goal of this variable is to tie intergenerational income elasticity, the popular way to look at mobility in economic terms, with other factors that pertain to the social mobility sphere (such as fair wages, work opportunities, health and, among others, education). In this context, social mobility is understood as the ability to discern an individual’s career prospects from its socio-economic background.
Key variable for this analysis, the elitism index (Eli) provides information on the nature of the tertiary education system for a given country, and heavily relies on the assessment of university performances on a global scale. For the scope of this analysis, a country has a more elitist educational environment when there is a larger distance between the top performing universities and the standard or bottom performing ones, as it will further be discussed below. The Edu variable measures the priority that a specific government gives to tertiary education relative to other levels of education (i.e. primary and secondary), expenditure-wise.
Finally, corruption (Corr) and unemployment (U) are included to control for phenomena outside of the narrow scope of this research but which, nonetheless, tackle important transmission channels related to education. Specifically, a corrupt institutional environment would negatively contribute to the achievement of a higher degree of mobility because of the lack of transparency in the process of materializing the human capital upon completion of tertiary level education. Similarly, a heavier burden carried by higher levels of unemployment among individuals with advanced education would signal a certain level of saturation in the job market, thereby negatively influencing social mobility via the reduced number of job opportunities available.
Data
Most of the data has been retrieved from the World Bank dataset, with the exception of figures on social mobility and elitism. Due to the lack of alternative access, the data relative to social mobility has been manually transcribed from the Global Social Mobility Report (World Economic Forum, 2020). Data on university performance, which has been crucial for the construction of the elitism index, was retrieved from the webpage on world university rankings of the academic year 2019-2020, formed by the Center for World University Rankings (CWUR).
Given the cross-sectional nature of this analysis, the dataset only comprises countries for which information on social mobility was available. Specifically, this research is carried within a sample of 82 countries and it is not restricted to OECD territories only. For each indicator, if data were available for a multitude of years, the most recent observation was taken into account only. This decision was made in order
14
to best capture the social and economic structures that are behind the figures on social mobility, which are extremely recent. Most observations fall into the timespan between 2013 and 2020, fulfilling the goal of having as up-to-date figures as possible, conditional on their availability. Only three1 observations refer to data made available
in, respectively, 2003, 2010 and 2011.
An important remark is that both the social mobility index and the university rankings have a dual characterization: a score and a consequent ranking of their performance on a global scale. In this research, only the former is used in the analysis to avoid assuming equal distances between countries’ performances, which would be implied by the use of rankings. For interpretational purposes, however, taking into account the position of countries in relative terms might help towards a better understanding of the results’ implications from a socio-institutional perspective. Because of that, ranked positions will appear in the results section.
Table 1. Descriptive Statistics
Variable Obs Mean Std.Dev. Min Max
Social Mobility 82 62.209 13.675 34.5 85.2 Elitism 61 3.479 1.637 .071 7.349 UniScore 2000 71.636 5.069 65.8 100 Education 71 88.380 10.042 57.325 100 Corruption 80 .372 .999 -1.150 2.212 Unemployment 79 6.529 5.685 1.185 30.280
As shown in Table 1, the social mobility index can take values between 0 and 100, the latter extreme signaling the highest degree of social mobility attainable. Its construction (as performed by the World Economic Forum) consists of an aggregation of scores (1-100) matching the 10 pillars assigned to the indicator, namely: health, education access, education quality and equity, lifelong learning, technology access, work opportunities, fair wage distribution, working condition, social protection and
inclusive institutions. Each pillar has sub-measurements that can be explored in Appendix A. Finally, the aggregate measurement is the arithmetic mean of the scores of each pillar per given country.
The university rankings (as constructed by the CWUR) used in this research consist of an aggregate of the following characteristics: quality of education, alumni employment, quality of faculty and research performance. These measurements combine verifiable data and robust indicators2, further information on its construction
can be found in Appendix B. The analyzed university performance scores take values between 65.8 and 100, which are respectively the lowest and highest educative quality achievements of the top 2000 performing universities on a global scale. Overall, there are 99 countries where these universities are located, 71 of which are paired with the available data for the social mobility index. For these matched observations, 10 countries had only one university entering the first 2000th position worldwide.
Moreover, the elitism index has been constructed as shown in Equation 3, where n refers to the number of universities in a given country within the top 2000 global university performances. For that given country, the standard deviation of the institutions’ performances has been employed as the indicator of elitism. The larger the variability of the UniScore figures, the more elitist the tertiary educational system has been assessed to be. This is because a greater gap between best and worst performing universities is associated with a higher difference in education quality being offered to students. As argued in the theoretical framework, top universities have become increasingly more selective and more expensive, whereas standard ones have not changed their selection criteria as intensely. Combining these observations together, a good proxy for the degree of elitism is found in the abovementioned figures on performance variability.
)124256 = 7)*## ∑) 9.:2;<"-=! − .:2;<"-=?%
!+# (3)
16
The corruption variable captures the perception of the degree in which public power is devoted to private interests, and it is reported in units of a standard normal distribution (hence ranging between negative and positive figures, as depicted in Table
1), where a higher value signals a better control of said phenomenon. Furthermore,
the unemployment figures reflect the percentage of individuals in the labor force with advanced education (i.e. holding a bachelor’s, master’s or doctoral degree or equivalent according to the International Standard Classification of Education 2011) and no employment.
Finally, measurements on education and elitism have been centered around the mean for interpretational purposes. Because of that, coefficients β2 and β3 for Model 2will illustrate the contribution of the given regressor conditional on the other interacted indicator being associated with its mean value (rather than zero).
Results
In what follows, the reader is presented with the main results for Model 1 and
Model 2 using an OLS model and an ordered logistic model respectively. Next, the
robustness analyses performed using OLS models are shown. Main findings
Ordinary least squares model
The two-tailed OLS estimates3 are reported in Table 2.
Table 2. Main outcomes for Model 1 and Model 2, OLS.
(1) (2) (3) (4) (5) (6)
Variables Mob Mob Mob Mob Mob Mob
Edu * Eli -0.116 -0.194** (0.0780) (0.0738) Education 0.474*** 0.132* 0.319** 0.165 0.211 -0.0927 (0.134) (0.0747) (0.119) (0.118) (0.138) (0.151) Elitism 4.180*** 0.416 4.404*** 0.797 (0.732) (0.903) (0.712) (0.824) Control of Corru. 10.15*** 8.520*** 8.978*** (0.784) (1.472) (1.406) Unemployment -0.715*** -0.610* -0.564* (0.262) (0.306) (0.308) Constant 62.94*** 62.97*** 66.49*** 64.38*** 67.19*** 65.19*** (1.543) (1.638) (1.200) (1.864) (1.233) (1.911) Observations 71 69 51 49 51 49 R-squared 0.120 0.783 0.500 0.776 0.512 0.801
Robust standard errors in parentheses *** p<0.01, ** p<0.05, * p<0.1
Starting from the specification of Model 1 - columns (1) and (2) - a positive effect of education is found on the social mobility score. To better grasp the magnitude of the outcome found in column (1), for instance, suppose that the public spending devoted to tertiary education in Italy increased such to reach the expenditure implemented by the Finnish government (which is roughly 3,7 percentage points higher). In that case, Italy’s social mobility index would increase by 1.77 units. This would lead the Mediterranean peninsula to reach the 30th rank in social mobility
3 Note: robust estimation techniques have been employed, upon having performed a test to check for
18
performance globally, moving from the 34th position that is currently held. A similar
reasoning, but with a smaller magnitude is found in column (2). The intuition behind the abovementioned results is that improving the quality of tertiary institutions by increasing their availability of public funds will positively affect the country’s social mobility. This positive impact of education on mobility has been shortly outlined in the theoretical framework and does not appear as a surprise. However, despite the positive and significant results, Italy would still be far from the 3rd place held by Finland,
suggesting that education alone does little to improve the Italian social mobility as a whole.
For the sake of completeness, columns (3) and (4) are displayed. However, the reader should keep in mind that the understanding of elitism is strictly related to its interaction with education rather than having a valuable contribution for itself.
Finally, columns (5) and (6) present us with the outcomes of Model 2. In particular, the coefficient of the interaction term in column (6) tells us that an increase of 1 percentage point in Edu would endure a decrease in its effectiveness on social mobility of 0.194 units for each additional unit of Eli. Interestingly, we can observe that the main effect of the (mean centered) education variable becomes insignificant once the interaction term and the control variables are introduced in the model. Hence, for countries having a mean elitism level, the impact of education on social mobility becomes insignificant, which is associated with a resulting null change in the dependent variable. However, for a positive deviation of elitism4 from its mean level,
for example of 0.16 in the case of Italian universities, there would be a decrease in the effect of education on the social mobility index by roughly 0.03 units per each percentage point increase in the government expenditure devoted to tertiary education. Considering once again the comparison between Finland and Italy and allowing for differences in elitism levels of the educational institutions: an increase in Italian government expenditure matching the Finnish expenditure level, given Italian universities’ elitism, would result in a 0.11 units decrease in social mobility. This would
lead to no consequent change in the ranked performance position of Italy, which would remain in the 34th place.
In terms of the controls, the results are in line with the above-presented expectations: a better control of corruption is positively associated with social mobility, and a higher level of unemployment amongst highly educated individuals is negatively reflected by Mob.
Ordered logistic model
In Table 3, the reader can find the results stemming from the use of an ordered logistic model. For the purpose of this analysis, four categories5 within the variable Mob have been constructed and consequently considered. In Appendix D, the reader
is provided with further details on the category pertaining to each country, as well as the chosen cut-off points. Finally, the homoskedasticity of errors and the assumption of parallel regressors have been successfully verified prior to the analysis.
Table 3. Main outcomes for Model 1 and Model 2, ordered logistic.
(1) (2) (3) (4)
Variables Mob Mob Mob Mob
Edu * Eli 0.971 0.934** (0.0253) (0.0318) Education 1.068*** 1.036 1.055 0.957 (0.0247) (0.0307) (0.0409) (0.0586) Elitism 2.372*** 1.423 (0.543) (0.445) Control of Corru. 19.86*** 15.44*** (10.67) (10.58) Unemployment 0.889** 0.879 (0.0517) (0.0754) /cut1 0.239*** 0.0655*** 0.0272*** 0.0132*** (0.0744) (0.0411) (0.0214) (0.0144) /cut2 0.842 0.657 0.357*** 0.302 (0.209) (0.337) (0.136) (0.220) /cut3 6.155*** 123.5*** 7.420*** 57.89*** (2.059) (123.9) (3.513) (64.52) Observations 71 69 51 49
Standard errors in parentheses *** p<0.01, ** p<0.05, * p<0.1
20
It should be noted that Table 3 displays the odds ratios resulting from the ordered logistic models. Columns (1) and (2) test for Model 1, whereas columns (3) and (4) refer to Model 2. The intuition for interpreting the results goes as follows. For instance, column (1) suggests that a 1 percentage point increase in the expenditure on tertiary education would lead to the odds from category to category increasing by 1.068. In other words, the odds of being in category “Very high mobility” over the “High mobility” category would be greater, given an increase in Edu; the same would hold for “High mobility” over “Low mobility”, and “Low mobility” over “Very low mobility”, in line with the parallel regressors assumption. Moreover, we can observe that the main effect of education becomes non-significant once elitism is introduced in the model. At the same time, the coefficient of the interaction term in column (4) being smaller than 1 is suggestive of an adverse effect of education on social mobility once elitism is taken into account. Although quantitatively assessing the results is not as straightforward as in Table 2, they appear to be of a similar qualitative kind for both
Model 1 and Model 2.
Robustness analyses
Alternative measurements
Given the numerous possible combinations stemming from interacting two terms that can be measured in alternative ways, such robustness analysis has only been performed with regards to Model 2. In addition to that, the analysis has been performed by employing only OLS estimations because of the ease in interpretation of the resulting coefficients. Furthermore, the assumption of parallel regressors for the use of ordered logistic models were violated by the employment of different measurements for education and elitism. Therefore, only the OLS estimates for the robustness check are shown and discussed.
Firstly, Table 4 reports the OLS estimations in which Edu has been measured in alternative ways but keeping all the other variables measured as in the original model. In columns (1) and (2), education has been regarded as the spending on tertiary education over the total public expenditure. Columns (3) and (4) take into account the % of GDP spent on tertiary education. Finally, expenditure on tertiary education other than staff compensation is taken into account in columns (5) and (6).
Table 4. Robustness analysis, alternative measurements – elitism as described in Equation 3.
(1) (2) (3) (4) (5) (6)
Variables Mob Mob Mob Mob Mob Mob
Edu*Eli 0.494 -0.305 0.321 -1.347 -0.0785 -0.0619 (0.793) (0.494) (1.300) (0.909) (0.0668) (0.0428) Education -1.507 -2.203** 8.260** 5.231** -0.110 -0.0906 (0.997) (0.986) (3.309) (2.143) (0.117) (0.0695) Elitism 4.959*** 0.931 3.903*** 0.631 4.670*** 1.096 (0.877) (0.875) (0.752) (0.850) (0.787) (0.891) Control of Corru. 9.234*** 7.212*** 7.829*** (1.433) (1.445) (1.409) Unemployment -0.589* -0.927*** -1.076** (0.346) (0.254) (0.494) Constant 65.98*** 63.52*** 65.46*** 66.56*** 67.10*** 67.13*** (1.338) (2.293) (1.355) (1.817) (1.322) (2.397) Observations 53 51 54 52 46 44 R-squared 0.454 0.794 0.506 0.786 0.450 0.815
Education variable Expenditure on tertiary education as a % of total public expenditure Expenditure on tertiary education as a % of GDP Expenditure on tertiary education
other than staff compensation Robust standard errors in parentheses
*** p<0.01, ** p<0.05, * p<0.1
Secondly, Table 5 proceeds in reporting OLS estimates for different measurements of Edu but, contrary to what discussed above, the elitism index is now constructed as follows:
)124256 = !@A(.:2;<"-=) − !2:(.:2;<"-=) (4)
Columns (1) and (2) show education as measured in the main model, together with the alternative construction of the elitism index. The rest of the table shows the same sequence of alternative education measurements as presented in Table 4.
22
Table 5. Robustness analysis, alternative measurements – elitism as described in Equation 4.
(1) (2) (3) (4) (5) (6) (7) (8)
Variables Mob Mob Mob Mob Mob Mob Mob Mob
Edu*Eli‡‡‡ -0.0188 -0.0370** 0.252* -0.0417 0.0608 -0.200 -0.00698 -0.00111 (0.0279) (0.0159) (0.148) (0.0833) (0.404) (0.232) (0.0153) (0.00755) Education 0.271* -0.0927 -0.353 -2.227*** 10.23*** 2.436 -0.0762 -0.0591 (0.140) (0.134) (0.885) (0.645) (3.203) (2.062) (0.107) (0.0513) Elitism‡ 0.853*** 0.190 0.865*** 0.0538 0.662*** 0.0684 0.881*** 0.104 (0.206) (0.133) (0.210) (0.120) (0.201) (0.121) (0.229) (0.138) Control of Corru. 9.219*** 10.14*** 8.571*** 9.150*** (0.993) (1.107) (1.125) (0.942) Unemployment -0.809*** -0.707** -0.937*** -1.133*** (0.299) (0.291) (0.279) (0.319) Constant 65.92*** 65.41*** 65.11*** 63.12*** 64.55*** 65.44*** 65.87*** 66.24*** (1.224) (1.966) (1.406) (2.021) (1.465) (1.910) (1.501) (1.864) Observations 61 59 63 61 64 62 56 54 R-squared 0.354 0.805 0.303 0.816 0.371 0.793 0.276 0.828
Education variable Expenditure on tertiary
education as a % of total public expenditure
Expenditure on tertiary
education as a % of GDP education other than staff Expenditure on tertiary compensation Robust standard errors in parentheses*** p<0.01, ** p<0.05, * p<0.1
Restricted sample
Furthermore, the analyzed sample contained a number of countries for which only a limited number of universities entered the first 2000th rankings on a global scale.
To see whether these observations heavily influenced the outcomes of this research, the original specification of Model 2 has also been tested for a subset of countries in which there were at least 4 educational institutions performing within the top 2000 ones. Table 6 illustrates the results of the two-tailed OLS estimates:
Table 6. Robustness analysis, restricted sample.
(1) (2)
Variables Mob Mob
Edu * Eli -0.143 -0.140* (0.112) (0.0727) Education 0.165 -0.0808 (0.168) (0.156) Elitism 4.389*** -0.0970 (1.021) (0.745) Control of Corru. 9.918*** (1.476) Unemployment -0.556* (0.322) Constant 68.25*** 64.48*** (1.447) (2.050) Observations 42 40 R-squared 0.429 0.797
Robust standard errors in parentheses *** p<0.01, ** p<0.05, * p<0.1
By looking at the results for the coefficient of the interaction term, it is possible to conclude that those observations did not qualitatively impact the conclusions reached within the original sample. However, they did have an influence over its magnitude. In addition to that, the effect of education alone resembles the original specification in terms of providing a non-significant contribution towards social mobility. Finally, only Model 2 has been tested for the restricted sample as Model 1 is not concerned with measurements deriving from university performances, hence it is expected not to be influenced by such change in sample definition.
24
Discussion
To begin with, the results presented in the previous section suggest that
Hypothesis 1 holds for the original specification of the model. In addition to that, the
outcomes were mostly in line with Hypothesis 2; the education variable, however, appeared not to be fully consistent. Overall, the idea of higher investment in the tertiary education sector being associated with an increased social mobility can be confirmed. At the same time, this positive effect tends to be adversely impacted by the level of elitism found in the given tertiary sector of the analyzed country.
From the robustness analyses it is possible to conclude that, given the use of the control variables, employing an alternative measurement for elitism as described in Equation 4 does not impact the main conclusions of the original specification7.
Although presenting an impact on social mobility of smaller magnitude, the interaction term is still characterized by a negative sign.
However, the same does not hold for alternative measurements of education: no matter how elitism is defined, this research is not robust for measurements of education differing from the original model. Despite not being an exhaustive list, the following intuitions may be possible explanations for these findings. For example, looking at the spending on tertiary education as a percentage of total government expenditure on all sectors (so, including health and social services) might not be a good indicator given the nature of the dependent variable. Specifically, expenditure on other resources that contribute to the publicly provided goods might have a stronger link to social mobility than the government spending on tertiary education alone. Similarly, as described in the theoretical framework, looking at the expenditure on tertiary education as a % of GDP might not be a good idea because the level of GDP itself might largely influence the fraction that will be devoted towards investments in higher education. Because of this, the given alternative measurement might capture forces that are outside of the scope of this research.
An important limitation that needs to be addressed is the lack of a time horizon in the performed analysis. The choice of a cross-sectional study derives from the way in which social mobility has been measured. Specifically, the global indicator was available for a limited duration. Since societal transformations are very slow and require a large observed timespan to be effectively captured, this approach was opted for. The alternative approach of handling panel data would have not been successful for this research as it would have entailed the use of fixed effects, not advisable when dealing with slow-changing variables (specifically, its addition would have meant capturing most of the indeed slow variation). Furthermore, avoiding the implementation of fixed effects would have still been problematic because of the serial correlation in observations.
Another limitation worth mentioning pertains the construction of the elitism index. Although the university performance scores were available for a satisfying amount of institutions, a number of countries only had one school entering the first 2000th global ranking. This could have been problematic for the interpretation of the
elitism index as a unique observation for that given country would have misleadingly pointed to a small distance between the maximum observation and the minimum one. In the understanding of this paper, in fact, a smaller distance has been associated with a milder degree of elitism. This, however, can clearly not be successfully applied to the case of a unique observation. Luckily, using Equation 3 as the main definition of elitism avoided falling in this inaccuracy as the standard deviation of a unique observation has not been computed. However, using Equation 4 to check the robustness of the results to alternative measurements of education implied taking it into account (i.e. having zero-distance observations that cannot be directly translated into an absence of elitism).
Nonetheless, the robustness check presented in Table 6 successfully showed that the overall conclusions drawn from the model were not qualitatively driven by countries with less than four universities performing within the top 2000 ones. The smaller magnitude of the interaction term, however, needs to be addressed. For the original sample, it can be argued that having a low number of universities could have induced a higher probability of them being further apart from each other, hence pointing at a higher level of elitism, given the way in which such variable has been
26
measured in this research. Because of that, the recorded effect on social mobility might have been overestimated in the original sample definition.
Lastly, if education is looked at as a black box transforming inputs into outcomes, this research has only taken into account data on outcomes for the formation of the elitism index. It would have been a good practice to also take into account the rigidity of the entry requirements to such institutions, together with data on their tuition fees. The availability of these pieces of information would have provided a more realistic idea of the level of elitism present in a given country. For example, this research shows high figures for elitism in Denmark, Sweden and Norway, whereas researches such as the one performed by Brezis (2018) – who takes into account the mentioned inputs - points at the same countries as being an example of the lower bound of elitism. However, this paper and the above-mentioned one converge in seeing countries such as France, the United Kingdom and the United States as instances of the opposite bound, with a rather high level of elitism.
Conclusion
Upon having extensively explored both the theoretical standpoints and empirical observations connecting inequality, economic growth and social mobility, this research has shown that the nature of the tertiary educational system for a given country negatively contributes to the mobility experienced by the inhabitants of said territory via the educational channel. Overall, it has been observed that tertiary education alone positively impacts intergenerational mobility. However, once higher education is considered together with its level of elitism, the educational channel appears to become a mild stratifying force rather than an equalizing one. Because of this, the above-mentioned policy recommendation suggested by Milanovic (2016) resonates with these findings: enabling a less elitist higher educational environment appears as a positive feature for the achievement of a higher degree of social mobility. Finally, as the magnitude of the results and the robustness analysis pointed out, such recommendation appears to have a rather small impact. Therefore, it is expected that policy implications within the educational sector ought to be coordinated with actions pertaining other social services too.
References
Albouy, V., & Wanecq, T., (2003). Les inégalités sociales d’accès aux grandes écoles [Social inequalities in the access to the Grandes écoles]. Economie et
Statistiques, 361, 27–52.
Alvaredo, F., Chancel, L., Piketty, T., Saez, E., & Zucman, G. (Eds.). (2018). World
inequality report 2018. Belknap Press.
Alesina, A., & Rodrik, D. (1994). Distributive politics and economic growth. The
quarterly journal of economics, 109(2), 465-490.
Becker, G. S., & Tomes, N. (1979). An equilibrium theory of the distribution of
income and intergenerational mobility. Journal of political Economy, 87(6), 1153-1189.
Belley, P., & Lochner, L. (2007). The changing role of family income and ability in determining educational achievement. Journal of Human capital, 1(1), 37-89. Beller, E., & Hout, M. (2006). Intergenerational social mobility: The United States in
comparative perspective. The future of children, 19-36.
Brezis, E. S. (2018). Elitism in higher education and inequality: Why are the Nordic countries so special?. Intereconomics, 53(4), 201-208.
Brezis, E. S., & Hellier, J. (2018). Social mobility at the top and the higher education system. European Journal of Political Economy, 52, 36-54.
Brezis Elise, S., & Temin, P. (2008). Elites and Economic Outcomes. New Palgrave
Dictionary of Economics.
Center for World University Rankings (CWUR). (2020) World University Rankings
2019-20. Retrieved from https://cwur.org/2019-20.php .
Chetty, R., Friedman, J. N., Saez, E., Turner, N., & Yagan, D. (2017). Mobility report
cards: The role of colleges in intergenerational mobility (No. w23618). national
bureau of economic research.
Desrochers, D. M., & Wellman, J. V. (2011). Trends in College Spending,
1999-2009: Where Does the Money Come From? Where Does it Go? What Does it Buy?. Delta Cost Project.
Galor, O., & Zeira, J. (1993). Income distribution and macroeconomics. The review
of economic studies, 60(1), 35-52.
Hoxby, C. M. (2009). The changing selectivity of American colleges. Journal of
28
Kinsler, J., & Pavan, R. (2011). Family income and higher education choices: The importance of accounting for college quality. Journal of human capital, 5(4), 453-477.
Mahassen, N. (2014). A quantitative approach to world university rankings. Center
for World University Rankings.
Milanovic, B. (2000). The median-voter hypothesis, income inequality, and income redistribution: an empirical test with the required data. European Journal of
Political Economy, 16(3), 367-410.
Milanovic, B. (2016). Global inequality: A new approach for the age of globalization. Harvard University Press.
Perotti, R. (1992). Income distribution, politics, and growth. The American Economic
Review, 82(2), 311-316.
Perotti, R. (1996). Growth, income distribution, and democracy: What the data say. Journal of Economic growth, 1(2), 149-187.
Persson, T., & Tabellini, G. (1991). Is inequality harmful for growth? Theory and
evidence (No. w3599). National Bureau of Economic Research.
Piketty, T. (2000). Theories of persistent inequality and integenerational mobility. Handbooks in Economics, 16, 429-476.
Saint-Paul, G., & Verdier, T. (1993). Education, democracy and growth. Journal of
development Economics, 42(2), 399-407.
Su, X., Kaganovič, M. I., & Schiopu, I. (2012). College expansion and curriculum
choice. University of Alberta, Department of Economics.
World Bank, Education Statistics - All Indicators. (2020). [Data file]. Retrieved from
https://databank.worldbank.org/source/education-statistics-%5E-all-indicators/
World Bank, World Development Indicators. (2020). [Data file]. Retrieved from
https://databank.worldbank.org/source/world-development-indicators/
World Bank, World Governance Indicators. (2020). [Data file]. Retrieved from
https://databank.worldbank.org/source/worldwide-governance-indicators/
World Economic Forum (2020). Global Social Mobility Report. Retrived from
https://www.weforum.org/reports/global-social-mobility-index-2020-why-economies-benefit-from-fixing-inequality
Appendix A
Construction of the Global Social Mobility index as performed by the World
Economic Forum. Each indicator is given equal importance (i.e. weight) within the pillar, and they are described as follows:
Pillar I Health (0-100 best) 10%
1.1 Adolescent birth rate per 1000 women 1.2 Prevalence of malnourishment (% of 5-19 year olds)
1.3 Health Access and Quality Index (0-100 best) 1.4 Inequality-adjusted healthy life expectancy index (0-100 best)
Pillar II Education Access (0-100 best) 10%
2.1 Pre-primary enrolment (%)
2.2 Quality of vocational training (1-7 best)
2.3 NEET ratio (% of 15-24 year olds)
2.4 Out-of-school children (%)
2.5 Inequality-adjusted education index (0-100 best)
Pillar III Education Quality and Equity (0-100 best) 10%
3.1 Children below minimum proficiency (%) 3.2 Pupils per teacher in pre-primary education (%)
3.3 Pupils per teacher in primary education (%) 3.4 Pupils per teacher in secondary education (%)
3.5 Harmonized learning outcomes (score)
3.6 Social diversity in schools (score)
3.7 Lack of education material among disadvantaged children (%)
Pillar IV Lifelong Learning (0-100 best) 10%
4.1 Extent of staff training (1-7 best)
4.2 Active labour market policies (1-7 best)
4.3 Access to basic services through ICTs (1-7 best) 4.4 Percentage of firms offering formal training
4.5 Digital skills among active population (1-7 best)
Pillar V Technology Access (0-100 best) 10%
5.1 Internet users (%)
5.2 Fixed-broadband internet subscriptions per 100 pop.
5.3 Mobile-broadband subscription per 100 pop. 5.4 Population covered by at least 3G mobile network (%)
5.5 Rural Population with electricity access (%) 5.6 Internet access in schools (1-7 best)
Pillar VI Work Opportunities (0-100 best) 10%
6.1 Unemployment among labour force with basic education (%) 6.2 Unemployment among labour force with intermediate education (%)
30 6.4 Unemployment in rural areas (%)
6.5 Ratio of female to male labour force participation rate 6.6 Workers in vulnerable employment (%)
Pillar VII Fair Wage Distribution (0-100 best) 10%
7.1 Low pay incidence (% of workers)
7.2 Ratio of bottom 40% to top 10% labour income share (%)
7.3 Ratio of bottom 50% to top 50% labour income share (%) 7.4 Mean income of bottom 40% (% of national mean income)
7.5 Adjusted labour income share (%)
Pillar VIII Working Conditions (0-100 best) 10%
8.1 Workers' Rights Index (0-100 best)
8.2 Cooperation in labour-employer relations (1-7 best)
8.3 Meritocracy at work (1-7 best)
8.4 Employees working more than 48 hours per week (%)
8.5 Collective bargaining coverage ratio (%)
Pillar IX Social Protection (0-100 best) 10%
9.1 Guaranteed min. income benefits (% of median income) 9.2 Social protection coverage (% of population)
9.3 Social protection spending (% of GDP) 9.4 Social safety net protection (1-7 best)
Pillar X Inclusive Institutions (0-100 best) 10%
10.1 Corruption Perceptions Index (0-100 best) 10.2 Government and public services efficiency (score)
10.3 Inclusiveness of institutions (score) 10.4 Political stability and protection from violence (score)
Source:
Appendix B
Components of the university performance scores (and subsequent rankings)
as constructed by the CWUR:
1. Quality of Education, measured by the number of a university's alumni who have won major academic distinctions relative to the university's size (25%) 2. Alumni Employment, measured by the number of a university's alumni who
have held top executive positions at the world's largest companies relative to the university's size (25%)
3. Quality of Faculty, measured by the number of faculty members who have won major academic distinctions (10%)
4. Research Performance:
i) Research Output, measured by the the total number of research papers (10%)
ii) High-Quality Publications, measured by the number of research papers appearing in top-tier journals (10%)
iii) Influence, measured by the number of research papers appearing in highly-influential journals (10%)
iv) Citations, measured by the number of highly-cited research papers (10%) Source: https://cwur.org/methodology/world-university-rankings.php
32
Appendix C
Elitism index for the sample analyzed in this research:
Country From Equation 3 From Equation 4 Mean centered (equation 3) Mean centered (equation 4)
Argentina 3,086893 10,3 -0,39212 0,802817 Armenia 0 -9,49718 Australia 4,81129 16,8 1,332273 7,302816 Austria 3,375822 11,5 -0,10319 2,002817 Bangladesh 0 -9,49718 Belgium 3,972461 11,5 0,493443 2,002817 Brazil 3,126146 14,8 -0,35287 5,302817 Bulgaria 0,989949 1,4 -2,48907 -8,09718 Cameroon 0 -9,49718 Canada 5,818331 21,4 2,339314 11,90282 Chile 2,948236 9,6 -0,53078 0,102818 China 3,802102 18,2 0,323084 8,702818 Colombia 2,106961 5,4 -1,37206 -4,09718 Costa Rica 0 -9,49718 Croatia 3,550587 8 0,07157 -1,49718 Cyprus 3,464823 4,9 -0,01419 -4,59718 Czech Republic 3,645229 11,7 0,166211 2,202817 Denmark 6,431951 18,3 2,952934 8,802816 Egypt, Arab Rep. 2,154339 7,8 -1,32468 -1,69718 Estonia 3,534119 6,7 0,055102 -2,79718 Finland 3,52231 11,7 0,043293 2,202817 France 4,867267 20,6 1,38825 11,10282 Georgia 1,697056 2,4 -1,78196 -7,09718 Germany 4,447816 17,7 0,968799 8,202818 Ghana 0,070711 0,1 -3,40831 -9,39718 Greece 3,483115 10,9 0,004097 1,402817 Hungary 3,014852 7,3 -0,46417 -2,19718 Iceland 4,737615 6,7 1,258598 -2,79718 India 2,496857 9,1 -0,98216 -0,39718 Indonesia 0 -9,49718 Ireland 3,662687 11 0,18367 1,502817 Israel 6,626839 17,5 3,147822 8,002817 Italy 3,634761 14,3 0,155744 4,802817 Japan 4,231831 23,8 0,752813 14,30282 Korea, Rep. 4,058442 20,4 0,579425 10,90282 Latvia 0 -9,49718 Lithuania 2,658947 6 -0,82007 -3,49718 Luxembourg 0 -9,49718 Malaysia 2,90566 8,4 -0,57336 -1,09718 Malta 0 -9,49718 Mexico 2,919017 11,1 -0,56 1,602818 Morocco 0,707107 1,6 -2,77191 -7,89718 Netherlands 4,592762 17,1 1,113745 7,602818 New Zealand 3,184533 9,3 -0,29448 -0,19718 Norway 5,866517 17,9 2,3875 8,402817 Pakistan 2,098094 5,6 -1,38092 -3,89718
Peru 1,697056 2,4 -1,78196 -7,09718 Philippines 0,707107 1 -2,77191 -8,49718 Poland 2,366007 9,9 -1,11301 0,402817 Portugal 4,152773 12 0,673756 2,502817 Romania 1,845924 5,2 -1,63309 -4,29718 Russian Federation 3,104458 11,9 -0,37456 2,402817 Saudi Arabia 4,705923 10,9 1,226906 1,402817 Senegal 0 -9,49718 Serbia 3,726035 7,9 0,247017 -1,59718 Singapore 7,348639 13,9 3,869622 4,402817 Slovak Republic 2,832402 6,8 -0,64662 -2,69718 Slovenia 4,355074 9,7 0,876056 0,202817 South Africa 3,872007 11,5 0,392989 2,002817 Spain 3,243124 14,2 -0,23589 4,702817 Sri Lanka 1,626346 2,3 -1,85267 -7,19718 Sweden 5,995403 19,2 2,516386 9,702818 Switzerland 6,397277 20,6 2,91826 11,10282 Thailand 2,934063 8,4 -0,54495 -1,09718 Tunisia 1,502221 3,4 -1,9768 -6,09718 Turkey 2,080838 7,6 -1,39818 -1,89718 Ukraine 1,767767 2,5 -1,71125 -6,99718 United Kingdom 6,068923 28,2 2,589906 18,70282 United States 6,72496 34,2 3,245943 24,70282 Uruguay 0 -9,49718 Vietnam 0,861684 1,8 -2,61733 -7,69718
34
Appendix D
Social mobility categories
Country Social Mobillity Category Country Social Mobillity Category
Argentina Low social mobility South Africa Very low social mobility Australia High social mobility Spain High social mobility Austria Very high social mobility Sweden Very high social mobility Belgium Very high social mobility Switzerland Very high social mobility Brazil Low social mobility Thailand Low social mobility Canada High social mobility Tunisia Low social mobility Chile Low social mobility Turkey Very low social mobility China Low social mobility United Kingdom High social mobility Colombia Very low social mobility United States High social mobility Croatia High social mobility Vietnam Low social mobility Czech Republic High social mobility Albania Low social mobility Denmark Very high social mobility Armenia Low social mobility Egypt, Arab Rep. Very low social mobility Bangladesh Very low social mobility Finland Very high social mobility Bulgaria High social mobility France High social mobility Cameroon Very low social mobility Germany Very high social mobility Costa Rica Low social mobility Greece Low social mobility Cote d'Ivoire Very low social mobility Hungary High social mobility Cyprus High social mobility India Very low social mobility Ecuador Low social mobility Ireland High social mobility El Salvador Very low social mobility Israel High social mobility Estonia High social mobility Italy High social mobility Georgia Low social mobility Japan High social mobility Ghana Very low social mobility Korea, Rep. High social mobility Guatemala Very low social mobility Lithuania High social mobility Honduras Very low social mobility Malaysia Low social mobility Iceland Very high social mobility Mexico Low social mobility Indonesia Very low social mobility Morocco Very low social mobility Kazakhstan High social mobility Netherlands Very high social mobility Lao PDR Very low social mobility New Zealand High social mobility Latvia High social mobility Norway Very high social mobility Luxembourg Very high social mobility Pakistan Very low social mobility Malta High social mobility Poland High social mobility Moldova Low social mobility Portugal High social mobility Panama Low social mobility Romania High social mobility Paraguay Very low social mobility Russian Federation High social mobility Peru Very low social mobility Saudi Arabia Low social mobility Philippines Low social mobility Serbia High social mobility Senegal Very low social mobility Singapore High social mobility Sri Lanka Low social mobility Slovak Republic High social mobility Ukraine Low social mobility Slovenia High social mobility Uruguay High social mobility
The four categories have been constructed taking into account the absence of observations for the boundaries of the social mobility index. Moreover, the mean of such variable has been taken as a pivotal figure for the formation of said categories, as it can be observed below:
Category Lower bound Upper Bound
Very low social mobility 0 51.32
Low social mobility 51.32 62.20
High social mobility 62.20 78.67
Very high social mobility 78.67 100
