jusua DIPARTIMENTO DI SCIENZE SOCIALI E POLITICHE Education and social stratification I: Inequality of Educational Opportunities 1. Education and social stratification 2. Definition & measurement issues: What do we mean by IEO? 3. Modernization vs. social reproduction 4. Evidence I: IEO by parental class and education 5. Evidence II: IEO by family background over time 6. Explanations: why is there an OE association? 7. Trade off? A note on policies against inequality 1. Education and social stratification Social stratification in intergenerational perspective: the OED triangle- 3 hu The American patíonal Structure Peter M. BLau Otis Dudley Duncan íhTi reiki n Awď rd Wi n ra r Social stratification: the OED triangle (in detail) returns to education edu inequality of educational opportunities LM outcomes (employment, occupation, income....) DESO (direct effect of social origin) social origin (family background: parental oeeunation & education family background labour market Concerning the relation between education and stratification, we are interested in two transitions (OE and ED). Both transitions involve a selection process, but ideally, according to the meritocratic principle, we want them to be based on opposite mechanisms. In the first transition, a non-discriminatory principle works, as selection should be governed by equal opportunities. To ensure fairness, the OE association (inequality of educational opportunities) should be as low as possible. In the second transition, a discriminatory principle works, as selection should be governed by a suitability principle. To ensure fairness, the ED association should be as high as possible, to ensure efficiency, that is a good match between individual skills and requirements of the job. In a meritocratic context, both school achievement and job allocation should be based on merit, that is on the skills of individuals, not on any other ascribed feature of individuals (race, gender, political orientation, kin, friendship networks ecc.) 1. Definition & measurement issues: What do we mean by IEO? 10 Origin Destination w Education OE: association btw origins & education. INEQUALITY OF EDUCATIONAL OPPORTUNITIES We focus on family background, defined by parental occupation or education, and also look at gender, but many other origins might be considered, eg ethnicity, migratory experience, family structure and the like. Blau and Duncan (1967): O-E-D triangle 1. What is IEO? There are at least three major meanings to the concept of educational inequality (that is, three different definitions of E as a dependent variable in the OED triangle) 1. Inequality in the distribution of educational attainment 2. Inequality in the distribution of educational outcomes 3. Inequality in the distribution of educational opportunities 1. Inequality of educational attainment Inequality in the distribution of educational attainment is what was previously called the vertical stratification of schooling: at the micro level, some individuals go to school for longer, some for shorter spans of time; at the macro level, some countries have a more schooled population, some less. At the macro (country) level, it is measured by some index of dispersion or of variation of the amount of education achieved by the members of a given population. Given the fact that school systems have been constantly expanding over time, the study of the changing distribution of education over time (usually measured by year or cohort of birth) amounts to the study of educational expansion. Measuring inequality of educational attainment The amount of education of an individual, and thus the difference therein (dispersion, variation, inequality) can be measured in 2 ways: 1. his/her highest school title attained; 2. the number of years spent in school (the so-called pseudo-years, i.e. the no. of years required to get the highest title attained - otherwise school dropouts would get higher scores). Years are a metric variable, titles are a categorical variable: the same issues found for the case of the measurement of social position/status concerning socioeconomic scores and social classes apply here as well. From a theoretical and descriptive point of view, the issue is synthesis vs. detail, from a statistical pov it is linear regression models (OLS) vs. models for categorical variables (logit, probit). Measuring inequality of educational attainment European scholars prefer using the highest title, American ones prefer years. This relates to the different structure of the educational system and to the different legal status of titles. In Europe they are certified by the state («legal value» of the title, in Italian valore legale del titolo di studio), while in the US their value rests only on the awarding institution. So in the US what supposedly matters is having been in school, not having finished it. Moreover, this dichotomy also involves an argument concerning why education is useful for individuals and populations. According to human capital theory, education is useful since it makes people more productive and thus their wages become higher (market competition). According to credentials theory, education is useful since it provides tickets to access the better occupations. Measuring inequality of educational attainment The key issue is then whether there is a relation btw schooling and skills & productivity. HC theory assumes it exists. Credentialist theory takes on two positions. The argument, supported by evidence, is that productivity cannot be observed before employment, and that much of the job-related training takes place on the job. According to a weak version of credentialism, schooling is an indicator not of productivity directly, but of trainabilitv («training begets training))) and/or of general sociability (if someone has been through the school system for a number of years, it is likely that he or she will be able to comply with the behavioral requirements of the employing organization). In economics, this is called signaling theory (Spence 1973): school titles do not directly indicate productivity, but might be used as an indicator of it. Measuring inequality of educational attainment According to a strong version of credentialism (social reproduction theory), school certificates have nothing to do with productivity, are is just a signal of behavioral conformity. School titles do not relate to skills which make an individual more productive (and profitable for employers), they only certify that the given person is apt for a given type of job. Measuring inequality of educational attainment Empirical analyses gave support to both theories. In the US some effect of the title has been found, in the 80s, by the so-called sheepskin models in the economics of education. The number of years (not pseudo-years) corresponding to time required for a title shows a stronger effect (there is a jump in wages or employment prob. corresponding to this number). This supports credentials theory. However, it has also been shown that even without the attainment of the title, some permanence in school has on average positive effects on occupational returns (wage, occupation etc.). For the Italian case, see Ballarino, Bison & Schadee (2011). This gives support to human capital theory. More on this when we will discuss returns to education. Measuring inequality of educational attainment The usual indices used in economics to measure income inequality are typically used (Gini, Theil etc.), although some complications arise when education is measured by means of categories (see Meschi & Scervini 2012, 2014). The pattern over time of the dispersion of education, measured by school years, is relatively well-known (Hout et al. 1993; Meschi & Scervini 2012, 2014): it is inverted U, similar to the Kuznets curve for income (educational Kuznets curve, see Milanovic for the Kuznets curve for income inequality). This depends on the expansion of education: when few people are educated, there is no variation and no inequality; when about half of the population is educated, the variation is at its maximum, and then it decreases as education becomes universal. Theoretical Kuznets curve (for any type of inequality) Gini Index A Greatest inequality Degree of development Measuring inequality of educational attainment However, recent research (Meschi & Scervini 2011) shows a third part of the curve, corresponding to a stage where tertiary education becomes important but not universal (by definition?) See the theoretical pattern in the next slide, and the empirical pattern of the SD of years of education by cohort for Italy in the following table. This is also similar to the pattern of Kuznets «waves» found by Milanovic (2016). Empirical educational Kuznets curve (Meschi & Scervini 2012) Second stage Third stage 15 Mean years of education Educational attainment (row %) and years of education (average and standard deviation), by birth cohort. Source: Ballarino and Panichella (2021) years of education birth cohort elementary junior high high schoo college total N average SD 30-34 73.6 15.5 7.9 3.1 100.0 7.198 7.53 4.20 35-39 66.6 20.2 9.8 3.4 100.0 8.213 7.87 4.39 40-44 54.3 26.6 14.1 5.0 100.0 8.615 8.57 4.77 45-49 40.8 32.9 18.4 8.0 100.0 10.144 10.466 9.60 5.08 50-54 24.1 41.1 24.9 9.9 100.0 10.52 4.78 55-59 12.8 45.7 31.3 10.1 100.0 11.485 11.07 4.46 60-64 7.0 50.6 33.3 9.0 100.0 12.272 11.078 11.13 4.18 65-69 5.0 48.3 36.2 10.6 100.0 11.53 4.28 70-74 3.5 41.3 39.7 15.5 100.0 6.072 12.42 4.52 75-79 3.4 34.3 44.3 18.1 100.0 2.781 13.09 4.58 total 28.6 37.3 25.4 8.7 100.0 88.324 10.32 4.77 2. Inequality of educational outcomes Educational outcomes refer to learning and competences (skills). Marks and grades assigned by teachers might also be used, but they are not reliable, since they often include biases. Besides individual biases, there are different grading standards over schools; subjects (humanities vs sciences); geographical areas (in Italy, N vs S) and teachers' cultures (egalitarians vs meritocrats), Skills aquired in school are measured by well-known learning tests collected by international random surveys like PISA, PIRLS, TIMMS, IALS-ALL-PIAAC. These tests have been elaborated in psychometry since the early XX century (IQ tests are actually tests of school learning), and appear to be a better predictor of later outcomes than the simple measures of attainment, which are not very detailed. International skills surveys PIRLS: Progress in International Reading Literacy Study: reading skills, 4° grade (aged about 8 yrs) TIMMS: Trends in International Mathematics and Science Study: math and science skills, 4° and 8° grade (aged about 9 and 13 yrs). PISA: Program for International Student Assessment: literacy, math and other skills, 10° grade (aged about 15 yrs). Started in the late 90s. IALS: International Adult Literacy Survey: literacy, numeracy and text comprehension, all adult population (also ALL: adult literacy and life survey). PIAAC: Programme for the International Assessment of Adult Competencies: skills used in work and in daily life, all adult population. International skills surveys In a number of countries such surveys are administered yearly to the whole student population, in order to have a third-party, independent assessment of student learning (output-based standardization). This custom is typical of de-centralized school systems as the Anglo-saxon ones, but is now spreading all over the world, because of the influence of international tests (PISA in particular) and policies of school decentralization. INVALSI: Italian yearly survey on the whole student population in grades 2, 8, 10 and 13. Managed by INVALSI, the national institute for school evaluation, created in 1999. Inequality of educational outcomes Survey skills measures are available only from the 90s on, so we cannot follow their pattern over time as in the case of measures of attainment. However, they are internationally comparable and the school-based surveys (PIRLS, TIMMS and PISA) include information relating to the school (typically collected from the school principal) and to the social background of the students, so they are extensively used by researchers in education, sociology and economics. At the aggregate level, they are very important for international policy-oriented comparison, and also, in big countries, for within-country comparisons (much work of this type has been done for the Italian case, eg Bratti, Checchi & Filippin 2008). Inequality of educational outcomes At the macro level, this type of data is very important for international policy-oriented comparison, and also, in big countries, for within-country comparisons. The richness of the data-sets allows to compare the effects of different institutional features of the school system, at all its levels, on the outcomes of the pupils, controlling for many confounders. (this is often done with «multilevel» models). For instance, it is possible to check whether IEO differs over different types of school systems (eg centralized vs. decentralized). At the individual level, they are widely used to study educational opportunities and their variation across social groups. But not over time. Moreover, being measure of ability they might be used to control for ability in the OED triangle (see below primary vs. secondary effects). 3. Inequality of educational opportunities 0 5 a x 1 2 s o _ p-{ 3 n ID q q-1 ~ u 5 q « This is the major interest of sociologists who study education with an interest in social stratification and inequality. Very generally, in this field we study how the distribution of educational attainment or outcomes differs across different social groups, and how this difference changes over time (intergenerational transmission of education). Groups of interest include: genders; social groups variously defined by parental occupation (strata or classes), income (in economics), education (intergenerational reproduction of educational inequality); ethnic groups; migrants vs natives; as well as the full set of the intersections among such groups. In the following 1 will concentrate on social groups defined by family background, looking in particular at parental occupation and education. Why do we care for IEO? There at least three good reasons to study the association between social origins and educational achievement (OE association). •a descriptive one: as education is one of the main predictors of occupation and social position, its determinants are of major sociological importance (see next slide). •an ethical one: the OE association can go against the legitimacy criteria of contemporary social and political systems (see previous slides on the EBM and the EGE arguments). •an analytical one: despite the major progresses of the field, scholars are still divided concerning both the pattern over time of the OE association and the mechanisms who drive it. A scientist wants to know how things really work. And this is what makes research on El - IEO an amazing field. Why do we care? As education is the major determinant of the occupational status of an individual, and thus of her income, the OE association is one of the major structural parameters (in the sense of Blau 1977) of a society. "education is the main factor in both upward mobility and the reproduction of status from generation to generation" (Hout and DiPrete 2006: 6) Research shows education to be related to a set of other individual and societal outcomes, such as civicness (Dee 2003); crime (Lochner and Moretti 2004); health and happiness (Hartog and Oosterbeek 1998); social cohesion (Green et al. 2006). General reviews are provided by Hout (2012) in sociology and by Oreopoulos & Salvanes (2011) in economics. "Many good things come from education" (Hout 2012) Inequality of educational opportunities The first major question is then whether inequality of educational opportunities (IEO) exists: Have the offspring of different social groups access to the same school opportunities? is there an association between social origins (family background) and school attainment or outcomes? Given that in most of the cases this association exists, the second question is whether it changes over time. The time-span was originally defined by the availability of survey data: survey started to get fielded in the mid-20th century, but now in many countries historical data sets are being made available, thus extending our observation window well into the 19th century (not for Italy, unfortunately). A third question concerns the mechanisms explaining it. Why an OE association exists? 3. Modernization vs. social reproduction Modernization theory The importance of education as a factor structuring micro social stratification and macro societal outcomes is well-present in the founding fathers of sociology and in positivist social research, before WW2. For instance, in his comparative work Max Weber compares different types of education as structuring different types of societal arrangements and power relations. Also Comte and Spencer (but not Marx & Engels) underline the importance of education in the development of modern societies. Durkheim's first professorship was in Science Sociale et Pedagogie. Corrado Gini, the Italian statistician who invented the well-known index of inequality, in the 30s collected and analyzed data on IEO in a set of countries. Modernization theory After WW2, modernization theory (MT) systematized much of this work. According to MT, educational inequalities should diminish and finally disappear because of the way modern societies work. In modern societies, both firms and public institutions are constrained to hire their personnel on the basis of productivity (because of market competition in the case of firms and political competition both internal and international in the case of institutions). The organization who would not select personnel on this basis would lose out to competition. ED gets stronger. Aware of this, families and individuals try to get as much school as they can, thus producing school expansion. The school system develops to fulfill these requirements, in order to produce individuals with vast knowledge and well socialized to bureaucracy, and providing them with a certification (school title) of what they have learned, and how (final grades). Modernization theory The internal working of the school system has to be meritocratic and inspired by universalistic values, otherwise it would not produce the skills required by the modern socioeconomic system. OE should disappear, provided ability/intelligence are independent from social origin. However, during the 60s and 70s a number of studies cast doubt on this optimistic picture. In particular, three empirical works changed the way we look at the relation between schooling and inequality: the Coleman Report (1966), The American Occupational Structure itself (Blau and Duncan 1967) and Schooling in Capitalist America (Bowles and Gintis 1976). The critical 60s and 70s First, the results of the so-called Coleman report (Coleman et al. 1966), analyzing data of the first great school survey commissioned by the US government to study IEO with a strong policy mission to reduce it, showed the optimism not to be well-founded. The report showed a strong persistence of both race and family background effects on educational outcomes; the effect of the family on the outcomes were much stronger than that of the school, and other social and socio-psychological variables, such as the ethnic composition of schools and individual locus of control, to be not very relevant. Second, The American Occupational Structure by Blau and Duncan (1967) showed the ED association to be slowly increasing over time, but it also found the OE association to be stable over time, and also a direct effect of family background on occupational destination (OD), controlling for education. The critical 60s and 70s Third, Schooling in Capitalist America, by Marxist economists Bowles and Gintis (1976), argued that schools mostly socialize individuals to authority, and selects them on the basis of compliance to it. Correspondence principle: the school system reproduces capitalist society and prepares individuals to it. 1. The hierarchical structure of school reproduces social hierarchy. 2. Compliance and obedience are rewarded in students more than actual knowledge (as shown by research relating personality traits of students and teachers' marks). 3. Reward is external to activity, as it is in society. As workers do work because of the wage, not because they like it, similarly students do study because of the title, not because they like it. Alienation (see Marx 1844). The critical 60s and 70s 4. Knowledge in schools is fragmented into disciplines, as society is fragmented into social classes and groups because of the social division of labour. 5. Schools are heterogeneous so to prepare students to different social positions. Lower grades and vocational tracks prepare workers, with a teaching & learning process based on passive reception and compliance, while good high schools and colleges prepare managers, professionals and entrepreneurs, based on autonomy, creativity and personal re-elaboration of trasmitted notions. Schooling (teaching & learning) differs according to the social position kids are preparing to. This work converged with critical pedagogy (Bernstein 1971) and with sociological theories of cultural reproduction (Bourdieu & Passeron 1970) and educational credentialism (Collins 1979) in criticizing the link between schooling and productivity. The criticism of schooling From the 70s on, the optimism of modernization theory gave way to a counter-movement, whose theoretical expectations were often influenced by Marxism, which during the 70s underwent a revival: "Neo-marxism" eg. the "Frankfurt school" (Marcuse, Adorno); the French structuralists and post-structuralists (Althusser; Foucault; Bourdieu); "analytical marxism" in the Anglo-Saxon world (Wright; Roemer; Elster). Neo-marxism (NM) had a strong political impact on the student movement of 1968. Indeed, one of its key features was a strong critique of schooling, which was seen as a means by which the ruling class holds on to its social and economic power. According to NM, the proletarian revolution, predicted by Marx and Engels, did not come about in wealthy capitalist countries ("late capitalism") because of ideological consent to the system guaranteed by collaborative trade unions, consumption and -more importantly - culture transmitted in schools. Neo-marxism and schooling Marx & Engels, themselves well-schooled intellectuals, were not fully aware of the importance of schooling as a part of the modern state and society, since their work preceded the diffusion of postprimary schooling. Wrt schooling, classical Marxism was not different from other modernist and positivist theories (M & E were admirers of Darwin). The socialist movement favoured schooling as a means of emancipation of the working class, and strongly contributed with its political strength to the expansion of schooling. The cultural origins of the criticism of schooling are to be found in the Romantic movement, who developed a criticism of economic and political modernity, accused of breaking the "organic" original tissue of life and society (eg Rousseau). Neo-marxism and schooling Romantic criticism of modernity included both reactionary criticism and the actual observation of the downsides of social modernization and marketization. The concept of "alienation" is one of the more important in this approach. It was developed by the young Marx (1844, although doubts have been raised on the authenticity of those pages - see Rojahn 1983) in order to denounce the fact that workers, as dependent of an employer who owns the object of their work, are dispossessed of a part of their life, so they are "made other" - the literal meaning of the term. In this sense, only the abolition of private property and of market relations might reverse this situation. More widely, by "alienation" Marxists and critical sociologists and activists complain the lack of control on their lives on the part of individuals. Neo-marxism and schooling Workers should become their own masters, in order to eliminate alienation. Many perspectives have been proposed to this end, but no one among them managed to be fully satisfying from the point of view of effectiveness. No other property arrangement creates the incentives related to private property. The division of labour cannot be reversed, as it appears. Underlying the concept of alienation, moreover, there is the romantic idea of a full, organic relation btw individuals, society and their destiny. An emphatic and hardly empirical idea of happiness is also related to this idea. Sometimes, a distant past is idealized, when this fullness did actually exist. Often this past is described as communist ("primitive communism" in the old Engels), and counterposed to the dire present. Of course, as we have seen, this idea does not make any historical sense. Neo-marxism and schooling The Soviet union, after some brief flirting with romantic anti-school stances, built a school system whose functioning was not really different from those of capitalist countries, strongly selective and stratified, with a focus on vocational training (Matthews 1982). On the contrary, Neo-marxism, developed from the 60s on, blended Marx & Engels work with romantic and anti-positivist philosophy and pedagogy (and criticism of quantitative research). Wrt schooling, it had a different stance fro classical Marxism, criticizing modern school as a key means of reproduction of the capitalist exploitations of workers. Neo-marxism and schooling According to hrench Marxist "structuralist" philosopher L. Althussef (1976), the school system is a "state ideological apparatus" which produces individuals fit for the capitalist socio-economic system. Capitalist relations of production (and exploitation) are the structure, individuals depend on it. According to this anti-individualist, strongly Durkheimian position, individuals are just produced by social "practices", social activities aimed at reproducing the existing power relations and social hierarchy (eg the economy, the polity, culture). The reproduction of society takes place at a level which is unattainable by individuals. This is a type of conspiracy theory: what we see is not reality, but is somehow pre-ordained in order for everybody to be fooled. Moreover, as Durkheimian sociology in general, this theory hypostatizes capitalism (or power, or the state) by personifying something that is just the macro organization and outcome of micro-level behaviour. "Society" does not exist per se. Neo-marxism and schooling Pierre Bourdieu, a hrench sociologist and philosopher, took forward this position by underscoring the role of intellectuals and of highbrow culture in the reproduction of capitalist society. He critized Althusser's theory for its macro nature and for its "conspiracy" traits, and maintained that social reproduction is not determined at some macro-level, but it happens by the convergence of a number of individual actions. However, it is not clear how and why individuals act to fulfill the system's needs. Social structure is divided into "fields" (similar to Althusser's "practices") and inviduals' behave according to what B. calls "habitus". I am a professor and behave according to my professorial habitus, otherwise other people would not get from mw what they expect. This is similar to classical role theory in social psychology, and does not explain why often people do not behave according to their habitus. 4. Evidence I: lEO by family background Evidence Let us now look at the evidence concerning IEO by family background in Italy. We will also look at it in comparative perspective, and take a brief look at gender inequality. The ETM (educational transitions model) A major innovation in stratification research concerning schooling processes was provided by Robert Mare (1980; 1981), an American sociologist and demographer (1952-2021) who developed the so-called educational transition model (ETM), also called "Mare model". In the ETM, school attainment is measured by the highest educational level a person has attained. From and individual point of view, educational attainment is a process of completing, or not completing, each one of a set of sequential transitions. The model builds on the cumulative-sequential character of the modern school system (see Ballarino &Schadee 2010) The ETM («Mare model») College / Grade 12 \ Grade 11 Drop Out / Grade 10 r Drop Out \ Drop Out Fig. 1.—Perspective of traditional education transitions analyses The ETM («Mare model») Students in any given grade either continue on to the next grade or level of schooling or end their formal education. Educational attainment is thus analyzed as the cumulation of a sequence of binary choices, btw stopping and continuing school, for each school grade (or, more in the European way, each school level). Technically, the ETM produces a set of regressions of the probability of making each transition (completing/not completing each grade or level - 0/1 binary variable). Models might be estimated as linear probability, logit or probit models, and the "population at risk" might be all the population (unconditional models) or just those who graduated from the previous level (conditional models). Independent variables might be any 0 in the OED triangle: parental class or education, gender, ethnicity, migration or whichever ascriptive characteristic of interest. IEO in Italy Let us look, now, at IEO in Italy according to the ETM. These are our own analyses (courtesy of dr. Cantalini), on the data from the 2009 Multipurpose survey by Istat (random sample of Italian adult population). We first estimate Educ = a + b*Class of Origin Then we estimate Educ = a + b*Parental Education Then we estimate Educ = a + b*Class of Origin + c*Parental o Education di di m 'iche These analyses give us a cross-sectional, static picture. In order to — ■-< 5 o check how IEO changes over time, we will also add cohort of birth (j q~ 1 ~ u 5 q « to the picture (see below). IEO in Italy: education by class of origin Educ = a + b*Class of Origin First, we look at the total OE pattern, by regressing the probability of making each of the three school transitions we consider (E: to low sec, upper sec. and tertiary title) on class of origin, for the whole Italian population of 2009 (first slide, models control for gender and geographical area. The fourth model is an Ologit model of education by social class of origin, which is explained below). Conditioned logistic regressions (Mare model) ^ogtetkf regression Primary - Low sec - Upper sec -low sec_upper sec_Tertiary_ Social class of origin (ref. Service) WhC UPB APB UWC AWC -0.02** (-0.03 - -0.00) -0.07*** (-0.09 - -0.06) q *** (-0.23 - -0.20) -0.15*** (-0.17--0.14) (-0.29 - -0.25) -0.06*** (-0.08 - -0.04) 0 22*** (-0.25 - -0.20) -0.36*** (-0.39 - -0.33) -0.33*** (-0.35 - -0.30) -0.47*** (-0.50 - -0.44) (-0.20--0.14) (-0.30 - -0.24) 0 32*** (-0.36 - -0.28) -0.35*** (-0.38 - -0.32) -0.37*** (-0.41 - -0.33) -0.67*** (-0.78 - -0.55) -1.47*** (-1.59--1.35) -2.50*** (-2.62 - -2.37) -2.05*** (-2.16--1.94) -2.95*** (-3.08 - -2.82) Observations 21,435 17,033 10,540 22,291 Parental social class (occupation) vs parental education A typical finding in IEO research is that when 0 is measured by parental education, the OE association is stronger and more stable over cohorts than when 0 is measured by parental social class - that is, occupation (for Italy, see Ballarino and Schadee 2008). The interpretation is that parental education indicates parental immaterial resources in terms of skills, abilities, motivation and expertise concerning the school system, while parental social class (which is based on occupation) indicates material parental resources in terms of money and wealth. Most would agree that the former are more important than the latter for what children's educational attainment is concerned. In general, btw the two measures there is a strong correlation, as shown in the following slide. Primary or less Low sec Upper sec Tertiary Total Service 20.2 18.5 27.4 33.9 1,627 WhC 28.9 35.3 29.1 6.7 3,759 UPB 62.5 29.0 7.7 0.8 3,229 APB 91.4 7.2 1.1 0.3 2,732 UWC 78.1 19.6 2.0 0.3 7,673 AWC 94.2 5.0 0.7 0.1 2,823 Total 66.8 20.2 9.1 4.0 21,843 In this data the rank correlation btw parental social class and parental education is .48, quite high. This depends on the ED association: those with better education are mostly found in the service class or in the WC, while the primary educated are mostly working in agriculture or in the UWC. Parental education Let us substitute parental education to class of origin as a measure of social origin: Educ = a + b*Parental Education W substitute parental education to parental social class in a set of ETM models of the probability of making each of three transition, showing the average probability for all individuals in the data set, by parental education (next slide). Parental education The empirical patterns are not really different from what seen above for education by social class. Indeed, this depends on the high correlation btw the two variables, which in turn depends on the ED association. In the regression models there is a small technical difference from the models for social class: in the case of social class, the reference category, set to 0, was the service class, so the distances were negative (disadvantages). Here the reference category are those with primary educated parents, so the distances are positive (advantages). Take this into account when reading the tables and graphs. Conditioned logistic regressions (Mare model) Primary -low sec Low sec- Upper sec upper sec_Tertiary Ordered logistic regression Parental education (ref. Primary or less) Low sec Upper sec Tertiary 0 20*** (0.19-0.21) 0 23*** (0.22 - 0.24) 0.25*** (0.24 - 0.26) 0 23*** (0.22 - 0.25) 0.41*** (0.39 - 0.42) 0.46*** (0.44 - 0.48) 0.10*** (0.08-0.11) 0.25*** (0.22 - 0.27) 0.50*** (0.47 - 0.53) \ 34*** (1.28-1.41) 2 34*** (2.25 - 2.44) 3.45*** (3.30 - 3.60) Observations 22,084 17,469 10,777 23,000 Parental social class (occupation) vs parental education Now we look at the probabilities of making each transition as predicted by a set of ETM models including both parental social class and education as indicators for social origin (0). Educ = a + b*Class of Origin + c*Parental Education We look at the effect of each of the two indicator of parental resources while controlling for the other one. Multivariate regression: you estimate the association btw an independent variable (here, parental class or education) and a dependent variable (here, own educational achievement), net of other factors included in the model (here, parental education or class). This model simulates for each coefficient a situation where all other independent variables are equal over individuals included in the analysis. Models control also for geographical area and gender. Parental social class (occupation) vs parental education It is clear that the impact of parental education is stronger than the one of parental social class. Btw the service class and the WC, for instance, there is no difference in the lower and intermediate transition. To the contrary, those with parents with tertiary education have an advantage wrt to all other parental education groups over all transitions, and the advantage is stronger for the transition to university. Moreover, if we compare the parameters of this model with those of the previous two models, for class of origin and parental education only, it appears that the parameters for class here are much weaker, while those for parental education decreased, but only to some extent. Conditioned logistic regressions (Mare model) Primary -low sec Low sec -upper sec Upper sec Tertiary Ordered logistic regression Social class of origin (ref. Service) WhC UPB APB UWC AWC Parental education (ref. Primary or less) Low sec Upper sec Tertiary 0.01 (-0.02 - 0.03) -0.00 (-0.03 - 0.02) Q ^ *** -0.01 (-0.04 - 0.03) -0.08*** -0.06*** (-0.09 - -0.03) -0.08*** (-0.12--0.05) (-0.12--0.05) -0.17" -0.11 (-0.14 - -0.09) (-0.21 - -0.13) (-0.15 - -0.06) -0.06" -0.14" -0.15" (-0.08 - -0.03) (-0.18 - -0.11) (-0.18 - -0.11) -0.15" -0.26" -0.16" (-0.18--0.13) (-0.30 --0.22) (-0.21 --0.11) 0.16" 0.18" 0.07" (0.15-0.17) (0.16-0.19) (0.05-0.09) 0.19" 0.32" 0.18" (0.17 - 0.20) (0.30 - 0.35) (0.15 - 0.21) 0.22" 0.39" 0.40" (0.20 - 0.23) (0.36 - 0.42) (0.35 - 0.44) Q /| -J*** (-0.28--0.05) -0.37*** (-0.50--0.24) -1.48*** (-1.61 - -1.34) 0 72*** (-0.84--0.60) /| g2*** (-1.96 - -1.68) 1.33*** (1.26 - 1.40) 2 01 *** (1.91-2.12) 2 QQ*** (2.73-3.06) Observations 21,037 16,788 10,448 21,843 UNIVERSITA DEGLl STUDÍ DI MILÁNO DTPARTIMENTO DI SCIENZE SOC1ALI E POLITICHE IEO in Italy The previous analyses gave us a cross-sectional, static picture. However, we are interested in change over time: according to modernization theory, the OE association should decrease over time, while according to social reproduction theory it should not change over time. In order to check for this, we add to our models an interaction term between parental class (or education) and cohort of birth. This amounts to check whether and how the association btw parental class (or education) changes by cohort of birth. The equation is: Educ = a + b*Class of Origin + c*Class of Origin*Cohort of Birth Technically, it results as a set of identical models (regression equations) for each cohort of birth. The persistent inequality paradigm The first great comparative international project on IEO was carried out in the second half of the 80s, directed by Y. Shavit and H.-P. Blossfeld (1993), and is one of the main achievements of the third generation of stratification research. The book was titled Persistent Inequality and is still important. The teams involved in the project studied 13 countries with a similar design, using categorical measures of family background (social class) and education (highest educational level achieved). Analyses were based on the ETM approach. The results showed the OE association to have been stable over time in 11 of the 13 countries studied, the exceptions being Sweden and the Netherlands, where IEO was decreasing. The persistent inequality paradigm Indeed, the empirical findings were actually more mixed (Treiman & Ganzeboom 1998). At the lower educational levels, the OE parameters were decreasing over time in most of the countries, as a consequence of educational expansion. This did not happen at the upper levels, for which conditional models were estimated. 0 5 a x 1 2 s o In the following graphs, estimates for Italy (Multiscope 2009 data) are reported. 1_ p, 3 M ID u - J Q P< e w (J 3 Q « ETM for Italy: predicted probabilities of making 3 educational transitions, by class of origin and cohort of birth Ser 00 - co - cn _^e — --.c t-1-1-1-r 1930 1940 1950 1960 1970 WhC 00 - cd - cn - T-1-1-1-r 1930 1940 1950 1960 1970 UPB 00 _ cd - cn - ~\-1-1-1-r 1930 1940 1950 1960 1970 APB 00 _ cd cn - t-1-1-1-r 930 1940 1950 1960 1970 uwc -|-1-1-1-r 1930 1940 1950 1960 1970 oo _ cd - cn - AWC ^-i .....» ♦ --1 1930 1940 1950 1960 1970 Lower secondary Upper secondary —*— Tertiary ETM for Italy: class differences in the predicted probabilities of making 3 educational transitions, by cohort of birth WhC T UPB APB cd J£----1 :i ---1 \7 -- 1 1930 1940 1950 1960 1970 1930 1940 1950 1960 1970 1930 1940 1950 1960 1970 uwc t-1-1-1-r 1930 1940 1950 1960 1970 cd AWC I/-''' - ----pi 1930 1940 1950 1960 1970 Lower secondary - Upper secondary —*— Tertiary IEO in Italy: interpretation So the general picture provided by empirical evidence is: IEO decreases in the transition to lower secondary, for all classes and particularly for the agricultural classes, but also in the transition to upper secondary, for some classes. In the transition to tertiary, however, IEO is persistent. What matters more? It is a matter of interpretation. In terms of numbers, the lower transitions involve more people, so this is more important (see Shavit, Arum and Gamoran 2006). However, one could also emphasize persistence at the higher levels, as Mare, Blossfeld & Shavit did, and the fact that equalization at the lower level (the changes in the distribution of education) takes place because of the ceiling effect, while the persistence of IEO at the tertiary transition (the allocation of education) is what matters for social inequality. The persistent inequality paradigm The ETM, by definition, does not produce a single OE parameter whose pattern over time can be taken as a measure of increasing/decreasing selection on competences vs on heritage, so interpretation is important: one has to choice on which transition to focus. Shavit & Blossfeld (1993) gave more weight to persistent inequality in the higher transitions (to upper sec. and university) than to decreasing inequality in the lower ones (to elementary and low. sec). But in terms of scale the latter was much stronger than the former, involving more people. S & B's interpretation was based on Mare's distinction btw distribution and allocation (see Ballarino & Schadee 2010 for a more detailed discussion). The persistent inequality paradigm Distribution is the distribution of education over social classes (or any other group) and its change over time due to educational expansion, while allocation are the relative probabilities of different classes to make each educational transition. The distinction is similar to the one between absolute and relative mobility proposed by Erikson and Goldthorpe (1992), as discussed concerning the mobility table. Shavit & Blossfeld concluded that the decrease if IEO at the lower school levels was just a function of educational expansion, thus just a matter of distribution (absolute mobility), while allocative inequality persisted because of stability at the higher levels. This paradigm, however, was challenged in the 2000s. The cumulative logit model (ordered logit) The ETM, in fact, provides detailed evidence, but this comes at the expense of synthesis. Moreover, interpretations can always be discussed, depending on the weight one gives to the different measures of OE at the different levels. Much following work has tried to overcome this limit. The preferred solution has been the substitution of the ETM with models who constrain the estimates of the transition-specific probabilities into a single parameter: the most widely used is the ordered logit model (ologit and gologit in Stata), also called cumulative or ordinal logit. These models estimate the probability to make the transition to university taking also the previous transitions into account. As educational levels can be taken as ordered categories (differently from social classes), this can be done. It has also been done in economics, where the ordered probit is preferred (eg Cameron and Heckman 1998). The cumulative logit model (ordered logit) These "new" (in fact they had been around in psychology since the 40s) models typically show the OE parameters to have been declining over time, contrary to the persistent inequality paradigm. The main comparative papers were done by Breen, Luijkx, Mueller & Pollack (2009; 2010), who showed inequality to be decreasing over cohorts in all countries they studied (including Italy, although to a lesser extent than other countries). Systematic empirical comparisons between the ordered logit and the ETM, indeed, have shown results to be wholly consistent across models (Ballarino & Schadee 2010). The ordered logit gives more weight to the decrease of IEO at the lower levels, while the ETM keeps it separated from stability of IEO at the upper levels. It is then possible to give more interpretative weight to the latter. The cumulative logit model (ordered logit) In the next slide, a cumulative logit model for education by social class (remember our measure of IEO is the distance in education btw classes), by cohort, is shown. It is estimated (with Ologit) on the 2009 Multiscopo data, and is thus fully comparable to the previous ETMs shown before. These are not predicted probabilities (although it is possible to calculate them), but, for each class & cohort, the "log odds", that is the logarithm of the ratio btw the % of individuals who did the transition and the % who did not. The reference category are those with an origin in the service class, so we are measuring an advantage, which clearly decreases over time for the agricultural classes and for the UWC (not for the WC and for the UPB). So the general picture is not really different from the one provided by the ETM, but it is more synthetic and less detailed. UNIVERSITA DEGLl STUDÍ DI MILÁNO dtpartimento di scienze soc1ali e politiche n Most recent analyses The following analyses use a pooled data set including all of the recent data set including the information needed in order to answer to our questions on IEO. Data set include: Istat Multiscopo (1998; 2003; 2009); ILFI (1997-2005); IMS (1985); ESS (2002-2018); Eu-Silc (2005; 2011); Sharelife (2008/9). The first set of models is an ETM and the results refer to parental education, with tertiary education set to 0 (reference category), controlling for parental class. The second model is an ologit UNIVERSITA DEGLl STUDÍ DI MILÁNO dtpartimento di scienze soc1ali e politiche -.6 -.4 -.2 0 -.6 -.4 -.2 0 -6 -4 -.2 0 Relative probability of getting a school degree by parental education, by cohort. Italy (Baiiarino&Panicheiia202i) Licenza media Diploma Laurea <£T <§- $r & rft «9> vS> & r$ >fl> Licenza media o meno Diploma Cumulative probability of getting school degrees by parental social class, by cohort. Italy (Baiiarino & Panicheiia 2023) 1 -0- -1 - UNIVERSITA DEGLl STUDÍ Dl MILANO DIPARTIMENTO Dl SCIENZE SOCIALI E POLITICHE -2--3- i i i i i i i i i i i i i & A* ^ j>" a? J?1 jr at a 4 / / / / f / / / B ill" cohort Birth cohort —♦-Parental education: Lower secondary -----Parental education. Upper secondary —►— Parental education: Tertiary -♦- Paren-.al ^due-alien Lower secondary -----Parental education: Upper secondary —t— Parental education: Tertiary US: Transition lower secondary ■ upper secondary education (women) US: Transition upper secondary - tertiary education (women) I \ f f f f f 4 Birth cohort Birth cohort -4— Parental education: Lower secondary -—— Parental education: Upper secondary — Parental education: Tertiary -♦- Parental education: Lower secondary ------Parental education: Upper secondary —*— Parental education: Tertiary USA (Blossfeld Blossfeld a Blossfeld 2016) Fig.2 Predicted transition probabilities (with 95 *X confidence intervals) to upper secondary and tertiary education for successive cohorts in the United States (US) (Source: Authors' calculation) I - J * n^TranaMor bw Of mgdhJ ■ly - uppor vscccoary oducrton(rnor __1-----i---- -------_| _x__2-- -*-» A-*--1-^ J R. Traneiton upper socnndary - laroary oducation |mori) i ■ ■ _____--------------HE- "'" 1 . —*-i-i—i—i-í / / / / f é B ríf" rcnci / / / / / / / Briti acixrf! —»— Panrati sduoÉcn Lowor —-— Parents, odjcatcrt Upcor mxrorry ■ ;:xi u ~*i srtjratnr T citary • Pntt artnttnrt umwt wccrory -----PwiM «Ii ■> li Lrjpsr secondary —■•—■ Parents, aducafon Tartary PL Transition lower secondary • upper secondary educator (women) PL Transition upper secondary - tartary educate r (woman! / / / / / / / / / / / / / / Birth cohort Birth cohort • Pinrci educator —~— Parana artjcatort --Pirora educator Tartary Poland (Blossfeld Blossfeld a Blossfeld 2016) Fig. 3 Predicted transition probabilities (with 95 1 confidence intervals) to upper secondary and tertiary education for successive cohorts in Poland iPIj (Source: Authors' calculation! i kmor socmanry - upper socondnry adiulian {merfi KR TranUlcr upper socondnry ■ lartlsry odjcatcr imoni --. - to rniilun Uffwf nnilii; ---- - South Korea (Blossfeld Blossfeld a Blossfeld 2016) Fig.4 Predicted transition probabilities ■with 95 ri confidence intervals) to upper secondary and tertiary education for successive cohorts in the Republic of South Korea i KR i (Source: Authors' calculation) IEO in comparative perspective. Ologit models Breen, Luijkx, Müller and Pollack (2009 for M, 2010 for F) estimate Ordered Logit models for IEO in a number of wealthy countries, using the best data available for each country. They find substantial reduction in the OE association (also using parental education as 0) in all countries observed. (A t7j "d o o 0 1 - 2 - 3 - 4 - 0- 1 - 2 3 4 - Germany France Great Britain Sweden ~l-TT-1-~l-r & if J* 1*5 n*> Italy Poland Ireland Netherlands i-1-1 j-n-1-3-r nr y t> ^ av °r ^ ^ ^ nj 00 O ro "i* c-. Nj bo u) o J_I_I_I_I_I_I C O o -> bo o ro — nj j> 30-34 - 35-39 - 40-44 - 45-49 - 50-54 - 55-59 - 60-64 - 65-69 - 70-74 - 75-79 - 80-84 - i —' nj bo o nj nj ho nj 0» bo O J_I_I_L O 3 O 120 "O Q) 3 n Sľ ro o ro o O Policies to reduce inequality and cumulative advantage If it is so, policies oriented towards reducing inequality face strong issues, since they should intervene over a number of mechanisms and contexts. In particular, they should displace the workings of market competition and of the family, who tend to produce and reproduce over generations, respectively, inequality. In both cases, policies face serious difficulties in actually intervening on all mechanisms involved in creating inequality. It might be a good idea to intervene as soon as possible. According to Nobel laureate economist J. Heckman, returns to investment in human capital decrease rapidly over time (see slide). Optimal Investment Levels Birth Age 7 Age 18 Source: Heckman & Carneiro (2003) Human Capital Policy ^1 Policies to reduce inequality and cumulative advantage The Heckman curve is a stylized fact (in economics, SFs are statements based on empirical evidence raised from different sources, and not corresponding to a single measurement). It is based on 4 points: 1. The neurological structure of the brain and the formation of skills depend on an interaction btw genoma and the individual's experience of the environment. 2. The mastering of skills shows a hierarchical structure, so that more advanced skills are built on the more basice ones, in a HI cumulative structure ("learning begets learning") PE|| 3. Different types of skills (cognitive, social, emotional etc.) are WUm interdependent, and are built starting from childhood. ^1 Policies to reduce inequality and cumulative advantage 4. This is a lifetime process, but the creation of skills is stronger and faster in the childhood years, during which neurological circuits and the behaviour they are associated with are more plastic and then more to be influenced by the environment. This is of course a general, average pattern. It does not mean that investment in learning at later ages does not have any return. There might be, indeed, different curves for different types of skills. Some authors have tried to empirically estimate the curve, by directly measuring costs and benefits of policy interventions oriented to learning. It appears that the curve is actually flatter than the way it is represented by Heckman (Rea and Burton 2020). The debate is ongoing.