## Andreas Hadjar and Christiane Gross

Print publication date: 2016

Print ISBN-13: 9781447326106

Published to Policy Press Scholarship Online: January 2017

DOI: 10.1332/policypress/9781447326106.001.0001

Show Summary Details
Page of

PRINTED FROM POLICY PRESS SCHOLARSHIP ONLINE (www.policypress.universitypressscholarship.com). (c) Copyright Policy Press, 2021. All Rights Reserved. An individual user may print out a PDF of a single chapter of a monograph in PPSO for personal use. Subscriber: null; date: 28 July 2021

# Education systems and gender inequalities in educational returns

Chapter:
(p.259) Twelve Education systems and gender inequalities in educational returns
Source:
Education Systems and Inequalities
Publisher:
Policy Press
DOI:10.1332/policypress/9781447326106.003.0013

# Abstract and Keywords

Human capital accumulation has been identified in the economic literature, as one of the most relevant engines of economic growth. Changes in the skills premium affect income distribution and inequality in a country. This chapter concentrates on studying differences in outcomes in the labour market, in particular in terms of the level and dispersion of private returns on education by gender and related inequalities in earnings. A prime focus of the chapter is to introduce the definition and computation of the rate of return on education. After an overview, we analyse and compare wage profiles and returns on education inequalities across countries and across gender. We present estimates based on a microeconomic dataset for 12 countries with different institutions and education systems. Finally, the rules and impacts of several public policies on gender specific returns on education inequality is considered.

# Introduction

Although investment in human capital cannot be reduced only to investment in education, formal education does play a central role, and it is particularly interesting due to the crucial role of public intervention, which includes, among other things, the choice of the education system implemented in the country. The level of education has been identified as one of the most important engines of economic growth and has been seen as an important factor affecting many dimensions of social life, including the structure and dynamics of the family, and fertility patterns. Changes in the skill premium also affect income distribution and inequality in a country. From the viewpoint of policy, it is worth remembering the role attributed by the 2000 meeting of the European Council in Lisbon of the contribution of education and training towards the stated goal of making ‘Europe the most competitive and dynamic knowledge-based economy in the world, capable of sustainable economic growth with more and better jobs and greater social cohesion’ (European Commission, 2001: 3). The subsequent Report of the Education Council to the European Council, The Concrete Future Objective of Education and Training Systems (2001), states three main objectives: increasing the quality and effectiveness of education and training systems in the European Union (EU); facilitating the access of all to the education and training systems; and opening up education and training systems to the wider world. These are only some of the issues behind the active discussion of the levels and dynamics of investments in education, and their returns. The obvious motivation for the interest in a disaggregate analysis by gender is related to the disparities in earning profiles for men and women, confirmed in most of the countries. The wage profiles differ for many reasons. For instance, education attainments, length of active life, unemployment rates and (p.260) unemployment benefits vary by gender. Public policies can affect differently the incentives to invest in education and to participation in the labour market of men and women. All these policies, in particular the initiatives promoting the reconciliation of working and family life, differ greatly across countries, even within the EU, and emerge as a core concern of governments.

While theoretical analysis has been progressing, together with an increasing amount of empirical research, there are still many open questions. How does gender affect returns on education? How do public policies affect educational achievements and labour market participation? We believe that a comparative perspective, taking into account countries with different systems of education and policy interventions, is particularly useful in dealing with gender-related disparities in returns on education and labour market performance. According to the data, female dominance in higher education is a phenomenon of recent decades and differences across countries could be driven by the structure of the education systems. In particular, the effect of early versus late tracking could differently affect males and females. In comprehensive systems (late track), students attend the same schools throughout lower secondary school, until age 16. In selective systems (early track), students enter academic or vocational tracks prior to lower secondary school, at age 10. The different effects of late tracks by gender on educational attainment and choices about future studies create a gender difference around the time of puberty. In a late track system, females are more likely to choose academic tracks and to enter tertiary education. The positive impact on educational attainment is stronger for females, and the negative impact on males is higher in families with non-academic educational background (Pekkarinen, 2008). It would be expected that in countries with late track systems, males and females would have more similar education attainments than in countries with early track system; however, early tracking actually increases educational inequality (Hanushek and Wößmann, 2006). This was confirmed for the sample of countries we consider in this chapter: the higher the age when placed in tracks, the lower the gender gap in education. Given that individual education consists of a series of choices, the tracking decision can have long-term effects, influencing future income and labour market outcomes. The link between educational systems and educational achievements is discussed in other chapters. Here, we focus mostly on the consequences of different educational attainment, for males and females, on the returns on education. Cross-country gender differences do not depend only on educational systems; a multitude of factors are at play which may, (p.261) in the long run, adjust or change the initial impact. In this chapter, we contribute to the discussion by providing additional evidence. First, we present a model of individual choice to compute the rates of return on education. This approach allows us to take into account, and to assess the significance of, relevant variables: the wage premium, the structure of the income tax and of some public transfers and benefits, in particular maternity benefits, and the costs of investments in education, which also can vary by educational system. Second, we present estimations of the wage premium obtained using the EU Survey on Income and Living Conditions (EU-SILC), which improves the quality of the data previously available for comparative analysis. Finally, we consider 12 different countries spanning quite different situations in terms of labour market conditions, public policies and educational systems.

The core of the chapter is in the third and fourth sections. In the third section, extending the model according to de la Fuente (2003), we compute the rates of return on education by gender for 12 European countries. The de la Fuente model has become the theoretical reference for several studies. The main contribution here is to develop a new version of this theoretical model, aiming to capture gender-related features of the work experience and, on the basis of this model, to provide specific rates of return for men and women. The fourth section makes two main contributions. First, we discuss the data and present the estimates of the Mincerian coefficients, which measure the educational premium on wages and salaries. Calibrating the theoretical model, we then obtain the rates of return on education by gender and discuss the main results. Finally, in the fifth section, we analyse the impact of public policies on the returns. The sixth section presents conclusions.

# Theoretical background

The growth in per capita incomes of many countries during the 19th and 20th centuries are usually explained by stressing the role of the rapid increase of scientific and technical knowledge that raised the productivity of labour, and of the other inputs to production. It is clear that the exploitation of the benefits of technological improvements requires a parallel improvement in the skills of workers. Given that formal education is one of the most important factors affecting individual human capital and the ability to increase human capital,1 educational attainment plays an important role in affecting the growth outcomes of modern economies.

Two important distinctions must be made here. The theoretical literature on the role of schooling has developed along two different (p.262) lines of analysis: the human capital theory and the signalling approach. The theory of human capital, initiated by Becker (1964), considers how the allocation of time and resources to education affects the future productivity of workers. This approach assumes that schooling increases worker productivity and, consequently, their wages and salaries. It follows that the choice of schooling attainment is, at least partly, a factor affecting the productivity differences across workers. According to this approach, the monetary and non-monetary investment in education is, at least potentially, productive from both the individual and the social viewpoints, due to the increases in individual salaries and productivities that it prompts.

A second approach, built on Spence’s (1973) seminal contribution, emphasises the signalling role of education. Consider an environment where some privately, but not publicly, observable characteristic, such as ‘innate ability’, affects the probability of success in education, its cost and the productivity of the individuals on the job. The productivity, however, depends only upon the ‘innate ability’ and is not affected by education. In such a situation, education attainment can be used to screen individuals with high ability from those with low ability. Employers will then be able to use school credentials as an index of the productivity of individuals on the job. A wage premium for highly educated individuals is required to compensate them for the resources spent on education, and therefore, according to this approach, there is a private premium for education due to the higher wages of educated workers (exactly as in the human capital approach). The social gain from education is not due to an actual increase in the productivity of the educated individuals (as in the human capital model), however, but to the benefits related to the possibility of screening individuals of different abilities. These two approaches have quite different theoretical, empirical and policy implications. In the following, we will refer to the human capital approach.

In both approaches, an essential distinction is between private and social rates of return. Generally speaking, the private rate of return is the rate of discount equalising the expected private marginal costs and benefits of the investment in education.2 This rate is one of the main determinants of individual choice in educational attainment. It is affected by many public policies, including direct and indirect subsidies to education and income taxes. Instead, the social rate of return is the rate of discount equalising the expected marginal social costs and benefits of the investment in education. The difference between social and private returns is potentially large. For instance, the cost of public education systems enters the calculation of the social rate of return in (p.263) its entirety, while only the costs, monetary and non-monetary, paid by the individual student affect the private rate of return. Similarly, while after-tax incomes are the relevant variable with which to compute private returns, what matters in the computation of social returns are before-tax incomes. Finally, social rates of return should also take into account the effects of positive externalities of education,3 if any. In this chapter, we will only consider private returns; however, it is important to bear in mind the role of both private and social returns. For instance, the economic rationale for the large, generalised subsidies to education characterising most economically advanced countries rests on the presumed size of the social returns over the private ones. Note that in the following we are abstracting from the possible non-economic benefits of education that will be discussed in next chapter. We look at education attainments as the optimal outcomes of individual decisions. This approach does not explicitly consider the effects of education on consumption. This is because, in the model we present, it is possible to separate the individual choice about education from choices related to consumption. Once we limit the analysis of the education decision to its productive investment feature, the standard approach is to define the discounted sum of the future stream of income, net of private education costs as an objective function of the individual.

As usual, making the optimal choice requires that the marginal benefit (here measured by the expected discounted value of the increase in future income caused by a marginal increase in education attainment) equals marginal costs (here measured by the net, monetary and opportunity, costs of the increase in education).4 Benefits are usually computed to the age of retirement.5 Bear in mind that when we are looking at the choice of the length of schooling, S, as the optimal choice of an individual, the discount rate is, from the individual viewpoint, an exogenous parameter: individuals will invest in education as long as r > i,6 where r is the rate of returns on education and i denotes the interest rate prevailing in the economy.7 Evidently, to proceed to an empirical test of the theory, more structure needs to be imposed. As already mentioned, the human capital approach postulates that schooling increases wages by increasing the workers’ productivity and, therefore, that the choice of schooling causes at least part of the productivity differences among workers. The most common empirical approximation of the human capital framework is the earnings equation derived by Mincer (1974). In this chapter, we will follow this approach. A recent critical discussion of the Mincerian approach is provided by Heckman et al (2003; see also Harmon et al, 2003). Here, we will only present the intuition behind the approach.

(p.264) The reduced, and empirically testable, Mincerian equation postulates that the logarithm (log) of the observable earnings at time t are (approximately) given by the sum of four components: a constant term, α‎, which is a linear function of the amount of formal schooling S and a quadratic function of the work experience x. This translates into the canonical Mincer specification used in most empirical studies: given a sample of individuals, denoted by the subscript i, observed at time t, we usually proceed to estimate the econometric model:

(1)
$Display mathematics$

where wi is an earning measure for individual i with schooling Si and work experience xi, while ui is a disturbance term representing other relevant factors not explicitly measured.8 In equation (1), as in most applications in the literature, all the parameters are identical across individuals, so that they do not depend on any individual specific feature (such as innate ability): the effect of these sort of individual features is ‘hidden’ in the error term ui. Notwithstanding its limitations, the Mincerian model is the key reference in most empirical studies on education. We present the empirical strategy, a discussion of some of the econometric issues involved in its actual estimation late in the fourth section. Here, it suffices to say that there is a large body of literature estimating the Mincerian equation and variations. The estimated values of θ‎ vary significantly across the different studies depending upon the data and sample used, and the econometric techniques adopted. For instance, the exact specification of the model varies somewhat across studies (introducing demographic or other control variables assumed to affect the wage premium, and so on). The methodologies adopted to estimate the model also vary. In this chapter, we present estimates of the Mincerian equation obtained using the Heckman’s selection model to control for potential selection bias into employment.9 More generally, in addition to the basic ordinary least squares method, techniques based on instrumental variables and proxy variables are also widely adopted in the literature (for an overview of the literature see Mendolicchio and Rhein, 2014).

While the formal derivation of equation (1) is in line with the standard applications of the Mincerian approach, the interpretation of the results is slightly different: the estimated Mincerian coefficient θ‎ is treated as a measure of the education premium embedded in observable earnings; however, this coefficient is then used as an input for the computation of the internal rate of return of investments in (p.265) education, by gender, together with several other parameters which also affect the returns (that is, for calibrating the theoretical model).10

# The approach of the analysis

As already discussed, many studies analyse the rate of returns on education, embedding the wage premia. Most of them provide gender-free estimates. Conversely, we compute separately the returns on education of men and women entering the job market at the end of their formal education and exiting the job market at the age of retirement. To describe the decision on schooling, we present an extension of the model proposed by de la Fuente (2003). The structure of the model has the advantage of considering the costs of the investment in education, taxes, probabilities of employment during school and after school, and unemployment benefits, and therefore it allows us to compare the way these variables affect individuals in several countries (covered in the sample). To study all the possible factors driving the gender gap in returns, however, we take the de la Fuente model one step further by also considering parameters related to maternity issues. Since the actual female work experience may be affected by maternity episodes, we introduce maternity leave benefits and maternity related monetary benefits and the interaction between fertility rates and education. Looking at international data, it turns out that there is a negative correlation between education and fertility: the higher the level of education, the lower the fertility rate. To disentangle the impact of the fertility path on education, we must account for the position of the individual woman in the labour market. In all the European countries, in order to reconcile women’s family life and work, the law establishes for a working woman the right to leave her job for a period of time for maternity and childcare. A fraction of this period is paid by the firms or by the public insurance system. For all women having a child, independent of their position in the labour market, the government usually pays a cash benefit. The child benefit programmes differ dramatically across countries, and thus it is natural to ask how (if) these policies affect the investment in education. The first question is how to define and measure the returns of the investment in education by gender? To answer these questions, we first study an individual decision model accounting for these public policies and, calibrating the model, we obtain the private rate of returns on education, which depends upon the marginal costs and benefits of the investment, where the costs are the sum of direct and opportunity costs, and the benefits are the sum of the change in the wage profile, the (p.266) probability of employment and differences in unemployment benefits due to higher levels of education. For women, they will also depend on the change in the profile of benefits related to maternity. Finally, we discuss the elasticities of the returns on education with respect to the policy parameters and evaluate them numerically11 (see fifth section). In the following, we present the theoretical approach behind the model. The numerical values are presented in the next sections of the chapter.

## The model

The model that we present in this chapter can best be seen as an extension of the one by de la Fuente. We consider the after-tax earnings of an individual in full-time employment as an increasing function of schooling. To capture differences in the progressivity of the tax system across countries, we include both the average and the marginal rate of income tax. We take into account that, if unemployed, individuals obtain unemployment benefits that may or may not be related to their previous earnings and to their average earnings.12 We also consider the possible changes in the probability of being employed. This probability depends upon the unemployment rate, which is per se a function of the education level. It can be seen from the data that for all countries in the sample, independent of gender, the higher the education level, the lower the unemployment rate, and the higher the probability of being employed. We also consider that, while in school, individuals devote a fraction of their time to studying and attending school, and therefore the potential labour supply of students and their probability of being employed are lower than that of a full-time worker.13 We explicitly introduce maternity and parental leave and child benefits for women as follows: we compute the fraction of her working life that the representative woman can spend on maternity leave. This will depend upon the number of children, if any, and upon the length of maternity leave allowed by law. During this fraction of her active life, a female member of the labour force can, legally, be on maternity leave.14 Finally, schooling also implies direct private costs. Using these variables, we can redefine and compute the present value of expected net lifetime earnings, following the approach previously discussed. Denote with g the rate of productivity growth, then, the private rate of return on education (RRE) is the value r = R+g, such that the average level of education is the optimal solution to the problem of maximisng the present value of the expected net lifetime earnings, for the representative agent, man or woman, who studies for S years and retires at time U. A straightforward computation based on the (p.267) solution to the optimisation problem shows that R is obtained solving the following equation:

(2)
$Display mathematics$

Equation (2) may be easily interpreted: the denominator can be seen as the sum of the marginal opportunity, Δ‎OC, and direct costs of education, Δ‎DC. Similarly, the numerator gives the marginal effect of education on earnings. For men, this effect can be decomposed into two components: one related to the wage profile, Δ‎W, which is driven by the Mincerian parameter, θ‎, and a second one related to the effect of education on the probability of employment and unemployment benefits on income, Δ‎E. The tax system is extremely important in this kind of analysis, because of its effect on Δ‎W: the more progressive the tax system, the lower the impact of the wage premia on the rates of return. In the case of women, there is a third component, Δ‎F, due to the effect of education on fertility and the benefits that a woman obtains if she is out of work because she is on maternity leave.

The last component Δ‎F can be interpreted as the marginal (percentage) increase of income due to the change of the fertility rate caused by an increase in the level of education. Δ‎E measures the marginal (percentage) effect of the increase in education on income due to the change in the probability of employment. Similarly, Δ‎W measures the effect on after-tax incomes due to the effects that an increase in education has on the earning function. Since the left-hand side of equation (2) is strictly increasing in RRE, the larger the value of the right-hand side, the larger the value of the private returns on education.15

# Estimations, calibrations and main results

One of the key building blocks of the analysis presented in this chapter is the estimate of the wage premia θ‎. In a multi-country analysis, the main difficulty is in comparability of the data. The EU-SILC dataset gives us the opportunity to use recent and comparable micro-data. Since 2004, the EU-SILC data have been collected annually by the national statistical offices for the purpose of providing comparable information on income and the poverty situation in EU member countries. The EU-SILC data have replaced the European Community Household Panel (ECHP) previously used in many studies. The dataset contains cross-sectional information about household financial behaviour and fundamental individual socio-demographic characteristics such as age, gender, highest completed degree, parent backgrounds, family (p.268) composition, working status and so on. On the basis of the availability and comparability of the data, we selected 12 countries with different welfare regimes, institutions and educational systems.16 We restricted the sample to the working age population (men and women aged 25–64). A methodological problem arises when estimating the schooling coefficients due to the possibility of non-random selection of the sample from the workforce (Heckman, 1979, 1980). A priori, the relevance of this problem might vary across genders. Given the aim of this study, it is particularly important to take this possible bias into account, and thus we estimate the wage equation, the Mincerian equation, using the selection model by Heckman to control for potential selection bias into employment.17 The Heckman selection equation is:

(3)
$Display mathematics$

where YChi is the number of young children, aged 0–5, while OChi is the number of older children, aged 6–17, in the household. FInci is a measure of the income of the other members of the family.18 Finally, ei is a zero mean error term. Then, given a sample of individuals, denoted by the subscript i, observed at time t, we proceed to estimate the corrected wage equation:

(4)
$Display mathematics$

where we control for schooling, defined as the number of years of education, work experience19 measured as the real experience of the individual20 and, including several dummy variables, for marital status, public vs. private sectors, native-born vs. foreign-born individuals, part-time vs. full-time jobs, and parent’s educational background.21 As usual,

$λ ^ i$

is the inverse of the Mills ratio,22 estimated from the first stage, and ui is a disturbance term representing other explanatory variables.

The estimated coefficient of schooling in a Mincer wage equation, θ‎, can be conveniently interpreted as the wage premium. It gives the average percentage increase in wage due to an increase in schooling, in our case an additional year of school. We then embed the estimated (p.269) values of θ‎ as parameters affecting the individual decision problem, together with several other parameters that try to capture the relevant characteristics of the labour markets and public policies, as explained in the previous section. Finally, to obtain the returns on education by gender, we calibrate – separately for men and women and by country – the theoretical model presented in the previous section.23 We present the estimates of the Mincerian coefficients24,25 and the rates of returns, by country and by gender in Table 12.1.

Table 12.1: Mincerian coefficients and RRE by country, by gender

Country

Heckit

RRE (%)

Austria

θ‎M

θ‎W

rM

rW

0.0495***(0.0028)

0.0614***(0.0037)

5.00

5.81

Belgium

0.0443***(0.0024)

0.0531***(0.0034)

4.64

5.47

Denmark

0.0446***(0.0026)

0.0508***(0.0025)

6.10

6.88

France

0.0462***(0.0022)

0.0484***(0.0024)

5.75

6.36

Germany

0.0461***(0.0018)

0.0449***(0.0023)

5.32

4.82

Ireland

0.0569***(0.0036)

0.0859***(0.0043)

6.81

9.92

Italy

0.0363***(0.0010)

0.0406*** (0.0016)

3.88

4.48

Luxembourg

0.0827***(0.0023)

0.0842*** (0.0037)

8.63

8.68

Netherlands

0.0404***(0.0022)

0.0286***(0.0028)

4.57

2.72

Portugal

0.0727***(0.0034)

0.0940***(0.0034)

7.90

9.67

Spain

0.0577***(0.0023)

0.0705***(0.0039)

6.41

7.97

Sweden

0.0543***(0.0038)

0.0342***(0.0042)

5.23

2.76

Notes θ‎M and θ‎W own estimations using the EU-SILC data (2007). Standard error in parenthesis: significant at

(***) 1%,

**5%,

*10% level. rM and rW own calculations based on the calibration of the model.

(p.270) To interpret the results, remember that θ‎M and θ‎W measure the average percentage increase in future earnings due to an increase in schooling, while rM and rW measure the internal rate of returns from the investment.26 Let us first focus on men. In most countries, private returns for men range between 4% and 6%, with an average of 5.7%. The minimum value, 3.9%, is in Italy, and the maximum is 8.6% in Luxembourg.

The private returns of women vary greatly across countries, with an average of 6%. They are much lower than the average in Sweden (2.8%) and in the Netherlands (2.7%). For Ireland, Luxembourg and Portugal, the rates are much higher than average: 9.9%. 9.7% and 8.7%, respectively.

Our results show that private returns on education for females are higher than those for males in most of the countries in the sample. The only exceptions are Sweden, the Netherlands and Germany. For example, in Germany, returns on education are equal to 5.32% for males and 4.82% for females; more important, the difference is not statistically significant. This can be explained taking into consideration some peculiar features of their labour markets.

In Sweden, the public sector is a relevant component of gross domestic product (GDP) and there is a higher percentage of skilled women, compared to skilled men, working in this sector. Our results show that the wage level is typically lower in the public sector. Differences by gender in the skill composition of workers in public employment can therefore explain the gender gap in the returns on education.

In the Netherlands, the female job market is characterised by a high proportion of women in the labour force who work part-time. This proportion is even higher among low-skilled women and working mothers. In a separate analysis, not reported in the chapter, we combine educational levels and part-time experience. The wage premium decreases with education levels: this suggests that the high incidence of part-time jobs plays a role in explaining the lower rates of return for females.

For Germany, the values are in line with the results of previous studies. Occupational segregation by gender, in particular in the low-wage sectors, is a well-known and widely discussed phenomenon in Germany. Segregation makes it difficult for highly skilled women to obtain jobs in the upper part of the occupational hierarchy. There is also evidence that, if they do obtain such jobs, they no longer suffer wage discrimination. Clearly, this phenomenon has an impact on the returns on education for females.

(p.271) Quantitatively, the returns on education depend crucially on the wage profile (wage premium and labour income taxes). Looking at the composition of the numerator of equation (2), we see that the main component of the benefits depends on the coefficient of the Mincerian equation, while the effects of education on the probability of employment and the fertility effect vary greatly across countries and are of a smaller order of magnitude.27 We can conclude that the gender gap in returns on education can be explained mainly by the Mincerian coefficients which more than compensate for the negative effects on female returns triggered by higher unemployment rates and maternity related benefits. Finally, looking at the denominator of equation (2), we have found that the key components of costs are opportunity costs. In only three countries (Austria, Portugal and Spain), do direct costs exceed 5% of earnings. On the other hand, opportunity costs are (at the margin) always above 75% of earnings.

# Elasticities and public policies

Relevant public policies vary substantially across countries, and therefore questions arise. What is the effect of changes in the policy parameters on rates of returns by gender? Do policies affect gender-specific returns differently? What is the effect of maternity on women’s investments in education? For instance, increases in maternity leave and childcare benefits have a direct effect on the RRE because they decrease the opportunity cost of maternity. There is also an indirect effect because the changes in these policies also affect the fertility rate and may influence the values of the probability of being employed. By direct computation,28 we find that, for the sample of countries considered here, both elasticities have a negative sign. As one would expect, increases in childcare benefits increase the opportunity cost of schooling and the return decreases. It turns out that the effect of an increase in maternity leaves is also always negative. This is somewhat counterintuitive because an increase in the value of maternity leave increases the expected future income, however, it also increases the opportunity costs of schooling. The impact of a change in opportunity costs dominates all the others. Both results are in line with the empirical evidence. Following the same approach, we estimate the elasticities of returns on education with respect to unemployment benefits, marginal and average tax rates. The numerical values are presented in Table 12.2.

The first two columns of Table 12.2 report the elasticities of the returns on education for men and women with respect to the replacement rates. An increase in unemployment benefits has a negative (p.272)

Table 12.2: Elasticities of RRE by country, by gender

Country

Unempl. benefits

Marginal tax rate

Average tax rate

Mat. leave

Childcare

Male

Female

Male

Female

Male

Female

Female

Female

Austria

-0.68

-0.61

-5.93

-6.34

1.96

-5.54

-0.01

-0.09

Belgium

-1.07

-1.24

-7.33

-7.91

2.68

3.34

-0.03

-0.03

Denmark

-0.31

-0.51

-5.42

-5.88

5.08

10.01

-0.12

-0.05

France

-0.64

-0.62

-3.07

-3.15

1.99

0.19

-0.24

-0.13

Germany

-1.91

-2.13

-5.33

-5.22

2.30

-3.80

-0.05

-0.20

Ireland

-0.63

-0.92

-2.78

-5.98

0.00

0.93

-0.07

-0.16

Italy

-0.35

-0.33

-3.96

-3.95

0.92

-0.22

-0.10

-

Luxembourg

-0.28

-0.28

-6.20

-6.16

2.66

2.01

-0.09

-0.18

Netherlands

-0.21

-0.31

-5.01

-5.01

3.65

-3.70

-0.07

-0.01

Portugal

-0.23

-0.18

-5.54

-9.18

1.64

-1.95

-0.09

-

Spain

-0.33

-0.28

-3.68

-4.80

0.68

-1.08

-0.07

-

Sweden

-0.32

-0.11

-6.98

-6.65

2.72

9.84

-0.32

-0.11

Note The elasticities with regard to maternity leave and childcare benefits are computed for the female population.

impact on the returns on education for both genders. The magnitude of these effects is, however, fairly small except for Germany, where an increase of 1% in unemployment benefits causes a decrease of 1.9% in returns on education for men. The negative impact is even higher for women, 2.1%. What is much more relevant is the impact of a change in the tax system. The sign of the elasticity with respect to the marginal tax rate is negative for both genders, for all the countries. To give an example, an increase of 1% in the marginal tax rate decreases by about 5% the returns on education in Germany, for both genders. On the other hand, the impact of an increase in the average tax rate can vary according to the position of the individual in the wage distribution. For men, the impact is always negative. For women, the sign changes considerably across countries. It may be positive or negative. It is positive and significant in countries such as Denmark and Sweden, where the tax system is more progressive and the relative earnings distribution of females is fairly close to that for men. In Germany, for example, an increase of 1% in the average tax rate implies a decrease of about 3% in female returns on education. While an increase of 1% in childcare benefits decreases the returns of women of 0.2%, the elasticity with respect to maternity leaves is even smaller with a value (p.273) equal to -0.05. The impact of both policies on the returns is, usually, small. To interpret the results properly, bear in mind that the focus of the analysis is on only one of the possible channels for the effectiveness of public policies. In particular, they could possibly have a quantitative impact on the choice of whether to participate in the labour force. The empirical approach (see previous section) allows us to correct only partially for this endogenous decision.

# Conclusions

In this chapter, we discussed gender inequalities in returns on education in countries implementing different policies and educational systems, and we analysed the role and the impact of institutional variables in the decision of the individual to invest in education. Towards this aim, we embedded the Mincerian coefficients in an individual decision problem, together with several parameters capturing the characteristics of the labour market and tax system, the costs of the investment in education and public policies which may affect the incentive to invest. To compute the gender-specific returns on education, we have also explicitly considered policy variables related to maternity episodes. The results show that the returns on education of females are higher than those of males in all countries in the sample except Germany, the Netherlands and Sweden. The gender gap in the returns on education can be explained mainly by the Mincerian coefficients, typically larger for women, which more than compensate for the negative effects of women’s returns triggered by higher unemployment rates and maternity related benefits. Finally, the effects of the returns on education of the policy parameters can be evaluated in two ways: computing the elasticities of the returns and, as a robustness check, using several counterfactual experiments.

Evaluating the elasticities, we can conclude that an increase in unemployment benefits, by increasing the opportunity cost to be employed, always has a negative, but weak, impact on the returns on education, for both men and women. Women seem to be, on average, more sensitive to this policy.

An increase in marginal tax rates always has a strong and negative impact for men. For women, the elasticities are, in absolute value, even larger. This is due to the relative position of females in the earnings distribution. An increase in the average tax rates always has a positive impact on men’s returns, while it can have a negative or positive impact on the returns on education of women, depending upon the progressivity of the tax system and their relative position in (p.274) the earnings distribution. The more progressive the tax system, the greater the negative impact on the rates of return.

Finally, in each country, the elasticities with respect to maternity and childcare benefits are negative and not high, that is, an increase in maternity and childcare benefits always implies a weak decrease in women’s returns on education.

To interpret these results properly, note that the analysis considers one dimension of maternity related policies: the effect on the rates of return on education and on their differences across gender. These policies may have aims which are beyond the scope of this study, for instance to promote an increase in fertility. From this viewpoint, the small values of the elasticities presented are reassuring, in that they suggest that they can be implemented at a fairly small cost in terms of returns on education, and hence of investment in human capital.

As a robustness check, to assess the impact of public policies, it is also possible to run several counterfactual experiments simulating the theoretical model assuming different, hypothetical, policy scenarios. Since all the results of the counterfactual experiments confirm the impact of the various policies implicit in the elasticities, we preferred to report the values of the elasticities in the chapter, which are somewhat more intuitive. Nevertheless, we would like to present an interesting result from one simulation where we evaluated the impact of public financing of education, by comparing the actual returns – presented in the fourth section – with those that would prevail in a scenario where individuals had to bear the total cost of education. The provision of education services has a very high impact on the rates of return. If individuals had to directly bear the full cost of their education, the rates of return would substantially decrease for both genders and, in particular: returns on education would decrease, on average, by 1.7% for men and 2.8% for women. As a final exercise, we have computed the rates of return as if there were no public intervention at all. We call this simulation the ‘basic scenario’. Comparing the returns of the basic scenario with the actual ones, we note that the latter are, in most countries, higher, suggesting that the positive effects of education spending are more important than the negative effects of taxes and unemployment benefits.

References

Bibliography references:

Becker, G. (1964). Human Capital and the Personal Distribution of Income. Ann Arbor: University of Michigan Press.

de la Fuente, A. (2003). Human Capital in a Global and Knowledge-based Economy. part II: Assessment at the EU Country Level. Report for the European Commission. Brussels: DG for Employment and Social Affairs.

(p.277) European Commission (2001). The Concrete Future Objectives of Education Systems, COM (59 final). Brussels: European Comission.

Hanushek, E.A. and Wößmann, L. (2006). Does educational tracking affect performance and inequality? Differences-in-differences evidence across countries. The Economic Journal, 116(510), C63–C76.

Harmon, C., Oosterbeek, H. and Walker, I. (2003). The return to education: microeconomics. Journal of Economic Survey, 17(2), 115–155.

Heckman, J. (1979). Sample selection bias as a specification error. Econometrica, 47(1), 153–161.

Heckman, J. (1980). Sample selection bias as a specification error with an application to the estimation of labour supply functions. In J.P. Smith (ed.), Female Labour Supply: Theory and Estimation. Princeton, NJ: Princeton University Press, 206–248.

Heckman, J., Lochner, L.J. and Todd, P.E. (2003). Fifty Years of Mincer Earnings Regressions. Technical Report 9732, National Bureau of Economic Research.

Mendolicchio, C. and Rhein, T. (2014). The gender gap of returns on education across West European countries. International Journal of Manpower, 35(3), 219–249.

Mincer, J. (1974). Schooling, Experience and Earnings. New York: National Bureau of Economic Research.

Pekkarinen, T. (2008). Gender differences in educational attainment: Evidence on the role of tracking from a Finnish quasi-experiment. Scandinavian Journal of Economics, 110(4), 807–825.

Spence, M. (1973). Job market signalling. Quarterly Journal of Economics, 87(3), 355–373. (p.278)

## Notes:

(1) Via, for instance, learning-by-doing or other informal processes.

(2) As with any other investment, investment in education entails a comparison between the current outlay and the expectation of a future return. The individual is therefore comparing values at different points in time: the costs they have to pay today to obtain, (p.275) for example, a higher degree, and the future higher wage on the labour market. The procedure used to compare gains and losses at different points in time is discounting: for instance, X euros obtained t years from today are equivalent to X/(1+r)t euro today, where r is the appropriate discount rate. The internal private RRE is the discount rate such that the cost of a small increase in education will be fully compensated by the increase in the discounted value of the gains it induces (where both costs and benefits are computed at the level of schooling optimally chosen by the individual).

(3) An example of externalities are peer effects. To the extent that schooling and learning on the job are group activities, the level of other people’s activity may have a direct effect on the learning level of an individual. This is also related to the educational system implemented in the country.

(4) In this way we can convert a risky future cash flow into today’s monetary equivalent. In our specific case, one additional year in education will (with some probability) increase the future (yearly) income. This expected flow of income is discounted today to measure the total marginal benefit of the investment in education. The same approach is used for the total marginal costs. More details are given on p. 268.

(5) This implicitly assumes that the amount of post-retirement benefits are the capitalised values of individual contributions to pension funds (an assumption which, strictly speaking, is warranted only for fully funded contributive pension systems).

(6) Under the standard assumption of perfect capital markets.

(7) As we will see, in the empirical implementation of the model, r is computed ex-post, as the value which ‘rationalises’ the observed average level of education attainment as the optimal choice of a representative individual.

(8) These error terms are assumed to be independent of the other explanatory variables.

(9) There are few estimations which adopt this approach. Usually, they focus on single-country analysis.

(10) This approach also gives us the opportunity to analyse the role and the impact of Mincerian coefficients in the decision of the individual to invest in education.

(11) Elasticity measures the percentage change of an economic variable, in our case the RRE, to a percentage change in another variable; in this case we will consider, one by one, several policy variables. For each policy, the elasticities, at country level, are computed and evaluated separately for men and women. A full computation of all elasticities is available upon request.

(12) To compare the different unemployment systems implemented in the countries, the unemployment benefits are computed as the sum of two components. One captures the benefits related to the previous net earnings, while the second captures benefits that are related to the average net earnings. Whether one of the components is different to zero will depend on the unemployment system of the country.

(13) Taxes are corrected accordingly.

(14) During this period, she can be either employed or unemployed, with some probability. If employed, she will receive a fraction of her previous earnings, plus other benefits related to childcare and typically independent of the personal income and depending instead on the average income of the country. If unemployed, her income will be determined by the unemployment benefits plus the maternity related benefits which are, however, independent of employment.

(p.276) (15) For the full discussion and derivation of the model, and all formal definitions of the components see Mendolicchio and Rhein (2014), Appendix 1.

(16) The European countries we include in the sample can be divided according to educational system into three groups: early tracking countries (Austria – age 10, Germany – age 10), medium tracking countries (Belgium – age 12, France – age13, Italy – age 14, Luxembourg – age 12, the Netherlands – age 12, Portugal – age 15), late tracking countries (Denmark – age 16, Spain – age 16, Sweden – age16).

(17) Not for all observations is a positive outcome reported. In our case, we can observe the wage only for individuals who work. Since people who work are selected non-randomly from the population, estimating the determinants of wages from this subpopulation may introduce bias. The Heckman model is a two-step statistical approach which allows us to correct for selection bias. In the first stage, it estimates the probability of working using a probit model, see equation (3). In the second stage, it corrects for self-selection by incorporating a transformation of these predicted individual probabilities as an additional explanatory variable,

$λ ^$

see equation (4).

(18) In our case it is the total household income minus own labour income.

(19) Experience is included as a quadratic term to capture the concavity of the earning profile.

(20) If this information is not available, potential experience or age might be used.

(21) The parent’s educational background is measured by the higher number of years of schooling of mother or father.

(22) The inverse of the Mills ratio is the ratio of the probability density function to the cumulative distribution function.

(23) More details are available upon request.

(24) When convenient, we will use the subscripts W and M to denote the values of the parameters for women and men, respectively.

(25) In a multi-country analysis the main question is: are these measures comparable across countries? Here, given that the model is fully specified, the effects are comparable across countries and across genders.

(26) As already explained, it is the discount rate such that the cost of a small increase in education will be compensated by the increase in the discounted value of the gains it induces.

(27) Data for the decomposition is available upon request.

(28) For the computation of the elasticities by gender, see Mendolicchio and Rhein (2014), Appendix 2.