Sociology Working Papers Department of Sociology

Sociology Working Papers Paper Number: 2008–04

Nominal comparability is not enough: Evaluating cross-national measures of educational attainment using ISEI scores

Silke L. Schneider

Department of Sociology University of Oxford Manor Road Oxford OX1 3UQ, UK http://www.sociology.ox.ac.uk/swp.html

Nominal comparability is not enough: Evaluating cross-national measures of educational attainment using ISEI scores∗ Silke L. Schneider 28th August 2008

Educational attainment is a core social background variable covered in each and every single social survey. Since educational qualifications are difficult to compare across countries, cross-national surveys pose a particular challenge to the measurement of educational attainment. In this paper, a number of cross-national measures of educational attainment—two versions of the International Standard Classification of Education (ISCED–97) and years of education—are evaluated using data from the European Social Survey. To begin with, the distributions of ISCED–97, simplified to the main levels of education, are examined for the different countries. Since this reveals a number of weaknesses of ISCED–97 as implemented in the ESS, an alternative way of simplifying ISCED–97 is proposed. In a next step, using linear regression models, it is shown how much explanatory power educational attainment loses when comparable variables are used, rather than country-specific categories. The outcome variable used for this validation is social status as measured by the International Socio-Economic Index. The results suggest that harmonisation always entails some loss of explanatory power. Moreover, the adequacy of years of education as well as the levels-only ISCED–97 differs strongly across countries. Of all measures tested, the proposed alternative simplification of ISCED–97 fares best: it shows the lowest relative loss of explanatory power and the lowest variation of losses across countries. Some recommendations on how to implement cross-national measures of educational attainment in international surveys are made. ∗

I would like to thank Anthony Heath, Tak Wing Chan (both Department of Sociology, University of Oxford) and Martina Brandt (Soziologisches Institut, University of Zürich) for their comments on earlier drafts of this paper. I would also like to acknowledge the receipt of financial support from the Economic and Social Research Council (ESRC), the Department of Sociology and Nuffield College, Oxford, UK as well as the Studienstiftung des Deutschen Volkes during the course of the research presented here.

2

1 Introduction

1 Introduction Over the last decades, more and more international data sets including more and more countries have become available to researchers. Examples are large academicallydriven public opinion surveys (cf. Heath, Fisher, & Smith, 2005) such as the European and World Values Studies (EVS and WVS, since 1981 and 1991 respectively; http://www.europeanvalues.nl and http://www.worldvaluessurvey.org), the International Social Survey Programme (ISSP, since 1983; http://www.issp.org), the European Social Survey (ESS, since 2002; http://www.europeansocialsurvey.org) and the Survey of Health, Ageing and Retirement in Europe (SHARE, since 2004; http: //www.share-project.org). There are moreover a number of public administrationdriven official surveys for monitoring the development of labour markets, incomes and living conditions co-ordinated by the Statistical Office of the European Communities (EUROSTAT), e. g. the European Union Labour Force Survey (EU-LFS, since 1983), the European Community Household Panel (ECHP, 1994-2001) and the European Union Statistics on Income and Living Conditions (EU-SILC, since 2004); and finally, OECD’s Programme of International Student Achievement (PISA, since 2000; http://www.pisa.oecd.org) and the International Adult Literacy Survey (IALS, 1994– 1998, OECD & Statistics Canada, 1995). This ongoing development gave rise to a previously unequalled popularity of crossnationally comparative social research. Cross-nationally comparable measurement is one of the most important challenges for quantitative (variable based) comparative social research (Przeworski & Teune, 1970), which intends to push comparability to the highest achievable level. The comparable measurement of social background variables is however a neglected area, and “. . . instruments allowing the compatible measurement of demographic and socio-economic variables are badly needed” (Wolf & Hoffmeyer-Zlotnik, 2003, p. 2). Educational attainment is one of the individuals’ socio-economic characteristics recorded in virtually all social surveys and is thus, like age and sex, a core social background variable. Measurement instruments for educational attainment usually consist of one or more questionnaire items covering the most common educational qualifications in a country in a set of response categories. In some surveys, information on ‘age when leaving continuous full-time education’ or ‘(full-time equivalent) years of education completed’ are included as well. Variables derived thereof in various ways are used in many individual-level empirical studies either as a main (predictor or outcome) or as a control variable. Examples of the former from different areas of cross-national research are studies of health inequalities (Eikemo, Huisman, Bambra, & Kunst, 2008; Mackenbach et al., 2008; von dem Knesebeck, Verde, & Dragano, 2006), a wide range of attitudes and public opinion (Kunovich & Slomczynski, 2007; Scheepers, Grotenhuis, & Slik, 2002; Weakliem, 2002), educational inequalities (Filmer, 2005; Shavit, Arum, & Gamoran, 2007; Treiman, Ganzeboom, & Rijken, 2003), educational homogamy (Domanski & Przybysz, 2007; Smits, Ultee, & Lammers, 1998; Ultee & Luijkx, 1990), school-to-work transitions (Müller, 2005; Saar, Unt, & Kogan, 2008; Shavit & Müller, 1998) and occupational attainment (Sieben &

3

2 Cross-national measures of educational attainment de Graaf, 2001; Treiman & Yip, 1989).1 Macro-sociological studies use aggregated data on education (e. g. Breen, 2005; Goesling & Baker, 2008; Jungbauer-Gans & Gross, 2007). Conceptually, educational attainment is conceived of in a variety of ways: as an indicator of general basic and/or instrumental skills, social position, qualification for labour market entry, and socialisation in terms of exposure to values prevalent in the educational system. Given the different ranges of countries and research areas involving education as a central variable, the actual concepts and measures of education used in the studies cited above differ greatly—it is in fact difficult to identify any two studies that operationalise educational attainment in the same way. Despite the wide utilisation of educational attainment measures, validation studies of different types of measures and, more specifically, different harmonised measures for cross-national research, are scarce (see section 2.2). The aim of this paper is thus to evaluate the consequences of using cross-nationally comparable education variables for substantive comparisons across European countries. This is achieved by comparing the predictive power of different measures of educational attainment using data from the ESS, namely country-specific categorical measures, two simplified versions of the International Standard Classification of Education 1997 (ISCED–97) and actual years of education. There are two central questions for assessing the validity of cross-national comparisons with these measures of educational attainment: 1) Which of the comparable measures displays the highest predictive power relative to the country-specific measure? and 2) Which of the comparable measures shows least variation across countries in terms of loss of relative explanatory power? The criterion variable used for this validation is social status as measured by the International Socio-Economic Index (ISEI; Ganzeboom, de Graaf, & Treiman, 1992; Ganzeboom & Treiman, 1996). Social status was chosen as the criterion because it is known to be closely associated with educational attainment. The association between educational attainment and social status can thus be expected to be very sensitive with respect to the quality of measurement of educational attainment. ISEI scores are a well-established instrument for measuring occupational status in cross-national research. Social status as measured by the ISEI is also closely related to many other socio-economic variables, like social class membership, occupational prestige and income (Ganzeboom et al., 1992).

2 Cross-national measures of educational attainment Different user communities have developed different ways of dealing with the crossnationally comparable measurement of education: Following Blau & Duncan (1967), social researchers often resorted to years of schooling (or measures derived thereof) as a proxy variable for educational attainment (e. g. Eikemo et al., 2008; Kunovich & Slomczynski, 2007; Scheepers et al., 2002; Treiman et al., 2003; Treiman & Yip, 1989). If such a variable was not covered in the data, it was derived from country-specific educa1

Since the introduction of PISA, many studies of pupil’s reading, math and science achievement have been published (e. g. Gorard & Smith, 2004; Marks, 2005; Park, 2008). The concept of academic achievement in terms competences actually acquired is not covered here.

4

2 Cross-national measures of educational attainment tion categories (‘virtual’ years of education). ‘Age at leaving full-time education’ (e. g. Weakliem, 2002) is a related measure which does not take the different starting ages of compulsory education in different countries into account. More rarely, scaling approaches are used in order to convert country-specific education categories to a common metric using some criterion variable (Smith & Garnier, 1987; Sorensen, 1983; Treiman & Terrell, 1975). Both approaches share the advantage that education can be parsimoniously included in statistical analyses as a linear variable. The most influential categorical measure in academic cross-national research is a scientifically motivated classification of educational credentials, the ‘CASMIN Educational Classification’ (Brauns, Scherer, & Steinmann, 2003; Brauns & Steinmann, 1999), which was used in a number of seminal cross-national studies on social stratification and mobility (e. g. Breen, 2004; Erikson & Goldthorpe, 1992; Shavit et al., 2007; Shavit & Müller, 1998). Another categorical scheme, the ISCED, was introduced by the United Nations Educational, Scientific and Cultural Organization (UNESCO) in 1976 (with a predecessor from 1958) and revised in 1997 (for an overview of the development of ISCED–97, see Smyth, 2008). ISCED–97 is currently implemented in a wide range of cross-national surveys, e. g. in the ESS, SHARE, EU-LFS, EU-SILC and the PISA studies. ISCED–76 was used e. g. in the ECHP and IALS. Among the studies using ISCED-related measures are Domanski & Przybysz (2007); Müller (2005); Saar et al. (2008); von dem Knesebeck et al. (2006) and Huisman, Kunst, & Mackenbach (2003). Since ISCED–97 is implemented in many surveys nowadays and ‘years of education’ in practice the most widely used measure of educational attainment, these two are the main focus of this study. In the following, ISCED–97 is described in some more detail.

2.1 The International Standard Classification of Education ISCED–97 (OECD, 1999; UNESCO, 1999, 2006 [1997]) is a multidimensional multipurpose classification for harmonising country-specific educational programmes into a cross-national framework for levels and fields of education (the latter will not be discussed here).2 It is mostly used for international statistical reporting by UNESCO, OECD and Eurostat, but it can be adapted for the measurement of individuals’ educational attainment in social surveys.3 ISCED–97 firstly distinguishes seven levels of education. The core criteria for the assignment of national educational programmes to ISCED–97 levels is programme content, proxied by minimum entrance requirements, typical starting age, certificates awarded, staff qualifications, and duration (among others). 2

The OECD (1999) provides the most detailed documentation on ISCED–97 for all OECD countries, including conversion tables mapping national educational programmes to the appropriate ISCED–97 categories. Similar tables are available from EUROSTAT (2005) for all countries of the European Union and some non-member countries. The UNESCO Institute for Statistics (2006) has additionally published such mappings for 19 developing countries. 3 Technical documentation specifying how to adapt ISCED–97 to the measurement of educational attainment and how to implement it in cross-national surveys is however not currently available. The link between the ISCED unit of classification, educational programmes, and individual’s educational attainment can however be established via the successful completion of an educational programme.

5

2 Cross-national measures of educational attainment ISCED level 0: Pre-primary education (excluding child care) ISCED level 1: Primary education (usually the first six years of formal schooling) ISCED level 2: Lower secondary education (the end of which usually coincides with the end of full-time compulsory schooling after around nine years of schooling) ISCED level 3: Upper secondary education (where maturity certificates and post-compulsory initial vocational qualifications are acquired) ISCED level 4: Post-secondary non-tertiary education (programmes between level 3 and 5, e. g. university entrance certificates for mature learners or non-tertiary vocational education following upper secondary education) ISCED level 5: First stage of tertiary education (university and vocational college education, exclusive of PhD/doctorate, and equivalent) ISCED level 6: Second stage of tertiary education (leading to an advanced research qualification, i. e. PhD/doctorate). ISCED–97 secondly uses four cross-cutting dimensions within some of these levels: Programme orientation: At levels 2, 3 and 4, there is a distinction between vocational or technical (V ), pre-vocational or pre-technical (P ) and general (G) programmes. Cumulative programme duration: At level 5, short (up to three years), medium (three to less than five years), long (five to less than six years) and very long (more than 6 years) programmes are differentiated. Position in the national degree and qualification structure: Again at level 5, first, second and third and further qualifications are distinguished. Programme destination: At levels 2, 3, 4 and 5 there is a distinction between A programmes leading to ever more advanced theoretically-based programmes that in the end give access to doctoral programmes (ISCED level 6); B programmes leading to more advanced vocational programmes up to vocational tertiary education (5B), and C programmes preparing for direct labour market entry or other programmes at the same ISCED level.4 Figure 1 shows the educational transition pattern as conceptualised in ISCED–97. This figure is already a simplification, as it does not show any sub-dimensions other than programme destination, and does not show all possible paths. In its entirety, ISCED–97 provides a large number of international categories, allowing for detailed categorical education variables for cross-national research. The first version of the ISCED, ISCED–76, differs from ISCED–97 in several respects: Firstly, in the earlier version, there were no complementary dimensions assigned to educational programmes. Secondly, level 4 in ISCED–97 (post-secondary non-tertiary education) was not available. Thirdly, ISCED–97 levels 5 and 6 version were distributed over three levels in ISCED–76 that do not overlap with the revised categories (4: tertiary qualifications without academic degree; 5: first university degree or equivalent; 6: post-graduate university degree or equivalent).

4

These subcategories A, B and C are however defined somewhat differently at different ISCED levels.

6

2 Cross-national measures of educational attainment

0 1

2A 3A

2B

2C

3B

3A,3B

3C

LM LM

4A

4B LM

5A 6

5B LM

LM

Figure 1: ISCED–97 transition pattern according to UNESCO (2006 [1997], p. 18). Note: LM=Labour Market.

2.2 Previous evaluations of cross-national measures of educational attainment Since data using the revised ISCED were not available until recently, ISCED–97 has not yet been thoroughly evaluated. Kerckhoff and colleagues published two articles concerning the evaluation of ISCED–76 and the CASMIN educational scheme for a small number of countries, namely Great Britain, Germany, the Netherlands, Sweden and the United States on the one hand (Kerckhoff & Dylan, 1999) and Great Britain, the Netherlands and the United States on the other (Kerckhoff, Ezell, & Brown, 2002). Braun and Müller (1997) performed a similar evaluation for the United States, Germany, Poland and Hungary, covering CASMIN, years of schooling and a very basic three-category scheme. The methodological rationale of the analyses performed in these evaluations was that there are two sources of variation in the association between educational attainment and any other variable (e. g. occupational status) over countries: Firstly, the ‘true’ association between educational attainment and the related variable may differ over countries.

7

2 Cross-national measures of educational attainment Secondly, different cross-national measures of educational attainment reflect the ‘true’ level of educational attainment to different degrees. Therefore the strength of the association between educational attainment and the other variable will also depend on which specific measure of educational attainment is used. This latter point refers to the validity of different measures of educational attainment. It can be evaluated by comparing, within countries, the association of different measures of educational attainment with a given criterion. For the evaluation of ISCED–76, Kerckhoff and Dylan (1999) used data from the IALS for Great Britain and the United States. They compared the R2 s resulting from the regression of occupational prestige scores (Treiman, 1977) and cognitive test scores on education measured in three ways: Using ISCED–76 as implemented by the national researchers, using ISCED–76 as recommended by the OECD, and using the countryspecific educational attainment classification. The national classification was used as a benchmark. The authors summarised the results as follows (Kerckhoff & Dylan, 1999, p. 770): 1. “Relatively minor changes in the classification of the indigenous credentials into standard categories (in this case the ISCED categories) can alter the associations between educational attainment and both occupational prestige and cognitive skill. 2. Both constructions of standard categories underestimate the associations between educational attainment and two important outcomes (occupational prestige and cognitive skill). 3. Underestimation is much greater in one country (Great Britain) than in another (the United States). 4. The different underestimates lead to misleading impressions of the relative importance of educational attainment in the two countries by making the differences look larger than they actually are”. The comparison of CASMIN and ISCED–76 in Kerckhoff et al. (2002) show that in Great Britain and the Netherlands, CASMIN performs better than ISCED–76. In the US in turn, ISCED–76 works better. Thus the reclassifications imposed by either international framework are not equally adequate in all countries. In sum, “the important conclusion to be reached is that the way in which the standard categories are constructed from the indigenous categories can make a major difference in the kinds of results that are produced in comparative research” (Kerckhoff & Dylan, 1999, p. 769, italics original). It is also noted that neither CASMIN nor ISCED–76 offer enough categories to adequately represent all important distinctions in educational credentials. Braun and Müller (1997) use a wider set of criterion variables (including e. g. gender attitudes and income) to evaluate CASMIN, ‘years of schooling’ and a basic threecategory scheme, again comparing the explanatory power of the cross-national and the ‘indigenous’ education dummies. In all countries but the US, ‘years of schooling’ had (sometimes considerably) less explanatory power than the country-specific education categories, regardless of the criterion variable chosen. Using CASMIN resulted in less—but still sometimes undesirable—loss of predictive power. The minimal three-category scheme shows the weakest performance of all measures.

8

3 Educational attainment in the ESS These results are worrying and instructive at the same time. However, the results reached by Kerckhoff and colleagues using ISCED–76 might differ from those that would be achieved with ISCED–97 today: the major revision of the ISCED undertaken in 1997 may have changed the performance of the measure. Therefore, these studies need updating. Moreover, as the IALS data only included the major occupational categories of ISCO–88, the measures of occupational status and prestige used in both studies are very crude, which might have influenced the results, probably by underestimating the degree of underestimation resulting from the harmonisation of educational attainment. Furthermore, although ‘years of education’ is often used in cross-national research (see section 2), only the study by Braun and Müller (1997) has evaluated the predictive power of ‘years of education’. Finally, we still lack knowledge about the effects of using comparable education measures in a wider range of countries. It would therefore be useful to check the performance of ‘years of education’ and ISCED–97 in more recent cross-national surveys.

3 Educational attainment in the ESS The ESS (for the most recent technical report, see Jowell & the Central Co-ordinating Team, 2008) is a biennial repeated cross-sectional survey carried out in around thirty mostly European countries.5 Its main focus is change in peoples’ attitudes and underlying values over time. It also contains a number of variables capturing the social background of the respondents as well as their partners and parents, which in principle makes it attractive for comparative social stratification and mobility research. For this study, data from the first three rounds (collected in 2002/2003, 2004/2005 and 2006/2007) are analysed. The ESS includes three measures of educational attainment. Firstly, respondents were asked about how many (full-time equivalent) years they spent in education.6 Secondly, respondents were asked for their highest level of education completed using country-specific questionnaire items and response categories. Thirdly, the country-specific variables were reclassified into a simplified version of ISCED–97, which was agreed upon centrally before fielding the survey. This means that an ex-ante output harmonisation approach (Ehling, 2003) was chosen for the ESS. The simplified ISCED variable edulvl only distinguishes the main ISCED levels of education, but none of the sub-dimensions ‘programme orientation’, ‘destination’ or ‘duration’.7 The country-specific variables often do not distinguish 5

These are Austria (AT), Belgium (BE), Bulgaria (BG), the Czech Republic (CZ), Cyprus (CY), Denmark (DK), Estonia (EE), Finland (FI), France (FR), Germany (DE), Greece (GR), Hungary (HU), Iceland (IS), Ireland (IE), Israel (IL), Italy (IT), Latvia (LV), Luxembourg (LU), the Netherlands (NL), Norway (NO), Poland (PL), Portugal (PT), Romania (RO), Russia (RU), Slovakia (SK), Slovenia (SI), Spain (ES), Sweden (SE), Switzerland (CH), Turkey (TR), the Ukraine (UA) and the United Kingdom (UK). Not all countries participated in every round of the survey. 6 Questionnaire item F7, round 3: About how many years of education have you completed, whether full-time or part-time? Please report these in full-time equivalents and include compulsory years of schooling. 7 This implementation thus does not follow the recommendations by the scientific advisors to the ESS (see Erikson & Jonsson, n.d.).

9

3 Educational attainment in the ESS PhDs from other university degrees as required by ISCED–97. The intended 7-level ISCED variable is thus only available for a limited number of countries. In order to keep the other countries in the analyses, a six-category version of the ISCED, aggregating ISCED levels 5 and 6, is used in addition to the 7-level version. In order to get a first idea about how data coded with the simplified ISCED look like in the individual ESS countries, the distribution of educational attainment according to the levels-only ISCED–97 is shown in table 1.8 The distributions are presented for the three rounds of the ESS separately for three reasons: firstly, a number of countries changed their country-specific variables in round two or three. Secondly, not all countries were included in all three rounds. Lastly, changes over rounds are informative with respect to the reliability of the measure: Since the survey rounds are only two years apart, true changes in the distributions (due to cohort replacement) can be assumed to be minimal, and should be continuous. Categories that are significantly larger in one round than in another are marked by superscripts A, B and C (meaning that the marked category is larger than in round 1, 2 and 3 respectively in the respective country). Three things stand out here: Firstly, and unsurprisingly, there are large differences between countries in the distribution of educational attainment, with all but ISCED level 6 being substantial in at least some countries. Secondly, there are statistically significant (p ≤ 0.05) changes in distributions of ISCED–97 within countries over the three ESS rounds. The most substantial changes (sometimes exceeding 10 percentage points) can be observed in a number of countries that changed their country-specific variables in round 2 or 3 (marked by an asterisk in the column ‘ESS round’).9 Even in those countries where the country-specific variable was not changed in-between different ESS rounds, there are statistically significant changes in the ISCED distributions, namely in the Czech Republic, Germany, Finland, Luxembourg, the Netherlands, Slovenia and the UK. Actually the only countries with no significant changes in the harmonised education distribution over ESS-rounds are Belgium, Portugal and Slovakia. Thirdly, there are many instances where large proportions (up to nearly 80 %) of the sample fall into a single ISCED–97 level. This is clearly most problematic at ISCED level 3, and particularly in Central and Eastern European Countries: In Bulgaria, the Czech Republic, Cyprus, Hungary, Poland, Slovakia, Slovenia and Switzerland, more than 50 % of the population are consistently classified as ISCED level 3. In Germany, Estonia, France and Sweden, the proportion of respondents classified at ISCED level 3 is consistently above 40 %. Quite diverse qualifications are however awarded at this level of education, namely general and vocational ones, as well as qualifications giving access to 8

The levels-only ISCED–97 reported here uses the OECD (1999) recommendations for mapping ISCED to country-specific education categories. It is thus not identical with the variable ‘edulvl’ included in the ESS data sets, which included many classification errors (see the first evaluations of the education measurement in the ESS by Kolsrud & Skjåk 2005 and Schneider 2007). 9 The changes in distributions can be attributed to actual changes in the country-specific response categories in the case of Denmark, Switzerland and ISCED levels 0 to 2 in Spain; and to a change in the data collection procedures in the case of Norway in round 2. With respect to Estonia, France, Greece, Hungary, Italy, Poland, Sweden and ISCED levels 3 to 6 in Spain as well as changes between round 2 and 3 in Norway, differences in distributions cannot be attributed to changes in countryspecific variables.

10

3 Educational attainment in the ESS Table 1: Distributions of the ISCED–97 main levels of education in the first three rounds of the European Social Survey

country

ESS round

BE BE BE BG CH CH CH CY CZ CZ DE DE DE DK DK DK EE EE ES ES ES FI FI FI FR FR FR GR GR HU HU HU IL IS IT IT LU LU

1 2 3 3 1 2 3* 3 1 2 1 2 3 1 2* 3 2 3* 1 2* 3 1 2 3 1 2 3* 1 2* 1 2* 3 1 2 1 2* 1 2

Distribution of the main ISCED–97 levels, in % 0

1

2

3

1.3 0.8 0.4 1.6 0.2 0.1 1.0ab 0.6 0.1 0.2 0.1 0.2 0.1 0.1 0.3 0.4 0.1

8.4 8.1 6.9 2.7 0.6 1.5 3.5ab 10.3 0.2 0.2 0.9 1.2 0.9 0.3 0.5 0.9 0.7 0.3 15.1b 9.5 15.5b 12.2c 9.9 9.0 12.4 9.8 11.0 26.4 27.1 2.4b 1.0 1.2 0.9 2.1 15.4b 11.9 24.2 23.4

18.6 17.2 17.3 21.1 9.8c 8.8 6.6 11.1 7.1 9.3 9.0 8.2 9.3 15.8 13.4 12.9 9.1 11.4 25.7 28.0 25.9 16.2bc 12.5 10.5 11.5 13.0c 9.2 19.1 17.2 23.2b 14.7 23.8b 13.4 22.0 35.5 32.5 7.8 10.7a

38.1 39.4 39.1 50.4 57.9 58.6 55.6 50.7 77.5 78.1 47.3 46.0 50.7b 49.1bc 37.8 35.7 46.3 44.2 21.0b 13.2 18.0b 37.6 38.8 39.1 42.7 43.6 47.5 31.9 31.6 56.6 56.5 57.3 35.5 37.8 38.1 39.7 40.3 37.4

9.5c 14.4ac 5.1 0.3 0.2 1.2ab 3.0 2.9 1.6 5.0b 2.8 1.3 0.4 0.6 1.8 1.3 0.7 0.7 2.7a

Continued on the next page

11

4

6.3

4.4

5.8 5.4 6.7

8.7c 4.8 8.4 8.6 11.2

0.9 2.4 4.4a

valid

5&6

6

cases

33.6 34.5 36.2 17.9 31.5 31.0 28.8 27.3 15.1b 12.3 36.9c 39.1c 32.4 34.7 48.1a 50.1a 35.2 39.2 20.3 26.3a 24.3 33.6 38.7a 40.2a 30.5 30.7 30.8 17.6 21.3a 16.6 27.5ac 17.1 48.4 38.1 9.8 14.4a 24.7 21.4

1.4 1.1 0.6

1186 1161 1142 898 1460 1502 1250 683 824 2012 1961 1870 1902 1050 1032 1043 1190 900 1047 1064 1240 1305 1314 1216 996 1215 1412 1555 1481 1099 1033 959 1547 341 855 1022 910 1116

1.8

1.2 1.1 1.1 0.9 1.1 1.2 0.3 0.6 0.8 0.8 0.9 1.1 2.1 2.1

0.3 0.5 0.8 0.9

0.9 0.9

3 Educational attainment in the ESS Table 1: Distributions of the ISCED–97 main levels of education in the first three rounds of the European Social Survey, % (continued) country

ESS round

NL NL NL NO NO NO PL PL PL PT PT PT RU SE SE SE SI SI SI SK SK TR UA UA UK UK min max average

1 2 3 1 2** 3 1 2* 3 1 2 3* 3 1 2 3* 1 2 3 2 3 2 2 3 1 3

Distribution across ISCED–97 levels, %

valid

0

1

2

3

4

5&6

6

cases

0.5 0.4 1.0 0.1 0.2 0.3 0.7 0.1 0.2 5.9 4.6 4.4

4.4 5.6 4.8 0.1

24.5 27.6 28.8a 33.9 42.6a 44.4a 14.2 14.0 14.8 11.4 11.1 12.3 60.0 28.4 29.2 34.1ab 20.2 21.0 25.4a 13.4 14.9 7.2 62.5 61.0 39.1 48.2a

0.3 0.7 0.2 0.6 1.0 1.1

20.5 0.2 0.1 0.1

30.6 29.4 31.1 53.8bc 31.9 33.1 61.2 65.0 63.9 13.6 15.4 16.7 32.0 49.3 48.4 46.2 57.5 60.8 56.7 72.1 70.3 15.2 31.0 30.3 39.5c 33.5

6.4 6.9 9.2a 2.9 8.9a 11.1a 5.2 5.9 4.6

1.5 48.5 1.7 1.4 0.5 0.2

33.6c 30.0c 25.2 9.3 16.4ac 11.1 18.7b 14.7 16.3 15.4 13.3 14.5 7.1 15.1 15.8 14.9 18.6 16.0 16.2 10.2 9.6 8.5 4.6 7.1b 20.8 18.0

1.0 1.8

1721 1338 1342 1470 1257 1212 1311 1116 1084 938 1223 1364 1473 1354 1285 1279 988 881 946 951 1135 1244 1201 1220 1364 1542

0.1 20.5 1.8

0.1 55.6 9.0

4.6 35.5 15.5

13.2 78.1 43.0

0.9 11.2 6.1

7.2 62.5 29.1

0.2 2.1 1.0

341 2012 1219

0.3 0.9 0.8 0.6 0.5 0.3 0.8

0.4 0.1 53.8 55.6 52.2 0.9 6.9c 5.7 4.0 3.0 1.7 1.5

3.5 3.7

0.6 0.7 0.7 1.0

2.0

1.1 0.8

Source: European Social Survey 2002-2007, own calculations. Respondents aged 25-65. Assignment to ISCED levels according to OECD (1999). Note: Asterisks in column ‘ESS round’: *=country-specific variable changed in this round; **=Data source changed in this round (from register to survey data). Marking for statistically significant (p ≤ 0.05) changes in distributions across ESS rounds with superscripts: a: category significantly larger than in round 1 in the respective country. b: . . . than in round 2 . . . c: . . . than in round 3 . . .

12

4 A European Survey Version of ISCED–97 higher education and qualifications not giving access to higher education—information that remains hidden when using the levels-only version of ISCED–97. In Denmark, Finland and Norway (2004 and 2006) as well as in Israel, Russia, the Ukraine and the UK, close to or more than 40 % of the sample are classified as ISCED level 5 and 6. Again, we know that this category is rather heterogeneous since there are short vocational programmes (classified as ISCED 5B) as well as undergraduate and postgraduate university degrees (classified ISCED 5A) as well as PhDs (ISCED 6) included in this one category. These important distinctions are not reflected by the levels-only ISCED–97. The size of categories might also be problematic at ISCED level 1 in Portugal and Turkey, although the level of heterogeneity is probably rather low at this level of education. Finally, in most European countries, the differentiation between ISCED 0 and 1 is negligible, as is ISCED level 6. ISCED 4 does not exist in all countries and is marginal in some others, which can often be explained by the fact that country-specific measures do not cover the respective programmes very well (Schneider & Kogan, 2008, p. 32f.). Therefore, the bulk of people are found at ISCED 2, 3 and 5 in the large majority of countries, which are by no means homogeneous categories. To summarise, the seven main ISCED–97 levels on the one hand do not well reflect the diversity of country-specific credentials at the most common levels of education. On the other hand, the levels-only ISCED–97 distinguishes between levels which are not common in the European context nowadays. Prima facie, the levels-only ISCED–97 thus looks neither efficient nor sufficiently differentiated.

4 A European Survey Version of ISCED–97 A detailed version of ISCED–97 is currently not used in any actual survey. Possible reasons are lacking documentation on how to implement ISCED–97 in social surveys and the complexity of the classification framework, which makes it difficult to implement and practically useless for any actual statistical survey analyses. Since the levels-only ISCED–97 does not look like an optimal way of simplifying ISCED–97, a “differently simplified” version of ISCED–97 for use in European and possibly other developed countries is proposed here: The European Survey Version of ISCED–97 (ES–ISCED). The ES–ISCED builds on the concepts and mappings of country-specific to international categories provided by ISCED–97, but incorporates the crucial idea behind the CASMIN scheme (see section 2), namely the differentiation of types of qualifications within levels of education. The ES–ISCED offers a more efficient simplification of ISCED–97 by collapsing certain (sub-)categories of the classification, but not others, so that important distinctions are maintained, whereas less relevant ones are dropped. Table 2 shows an overview of ES–ISCED. In ES–ISCED, ISCED levels 0 and 1 are summarised in one category. The various sub-categories within ISCED level 3 are simplified to a dichotomy, and ISCED level 4 is absorbed into the adjacent levels. ISCED level 6 finally is so small that it is not efficient to reserve an extra category for it (unless there is an explicit interest in respondents

13

Table 2: The European Survey version of the International Standard Classification of Education 1997 (ES–ISCED)

ES–ISCED

ISCED–97

I

0,1

II

2, 3C