Technicians (P.S.)* ..... medicine and technicians) and even have a higher propensity than men to hold a job match ...... Computer Technicians, Laboratory.
Higher Education: The More the Better? Gender Differences in Labour Market Returns to Tertiary Education in Israel Tally Katz-Gerro and Meir Yaish Paper Number 2000-05
Tally Katz-Gerro, Ben-Gurion University, Israel Meir Yaish, Nuffield College, Oxford
May 2000 Sociology Working Papers Editors: Diego Gambetta and Bridget Taylor Electronic Editor: Edmund Chattoe http://www.sociology.ox.ac.uk/swps/2000-05.html
Higher Education: The More the Better? Gender Differences in Labour Market Returns to Tertiary Education in Israel1
Abstract Although post-secondary credentials have become more important over the years, there is very little research on the transition from post-secondary school to work. Existing research refers mainly to differences between the academic and vocational tracks. In this paper we examine the transition from post-secondary education to work in Israel. We look at the transitions from different levels of higher degrees - vocational post-secondary, BA, MA and PhD - and from various fields of degree to the labour market. Three labour market outcomes are examined: employment status, occupational prestige attainment and job match. Data are drawn from a supplement to the 1983 Israeli census which includes a random sample of 10,914 individuals who represent Israel’s tertiary education (vocational or academic) degreeholders. The analyses employ a subset of Jews between the ages of 25-44 who were educated in Israel (N=5100). We find marked gender differences in occupational attainment among individuals with educational credentials in different fields of study. Several explanations for this finding are offered.
1
The authors are listed alphabetically. Send comments to Meir Yaish, Nuffield College, Oxford, OX1 1NF, UK,
Introduction In the past 50 years the influence of education on all realms of life has grown in importance and participation in the educational system has increased. The upgrading of the occupational structure as a result of industrialisation processes has created, inter alia, a demand for a skilled, sophisticated and highly-educated labour force. Concurrently, educational qualifications have become important for employment at the bottom of the occupational hierarchy (Collins 1971, p. 1003). As most individuals in modern societies attain some kind of secondary education, the proportion of occupations that require some kind of post-secondary education has grown and is projected to continue growing. This inflationary process may have undesired consequences. For one, as Collins (1971) argues, society at large (and labour markets in particular), may be faced with growing numbers of over-qualified individuals. If productivity is not associated with investments in these over-qualified individuals, as Collins suggests, part of an already scarce societal resource is simply wasted. It is also argued that the increase in society’s educational level is not coupled by a sufficient increase in demand for more qualified individuals in the labour market. As a result, unemployment rates may be relatively higher among the more (that is, over-) qualified graduates (cf. Hughes and O’Connel 1995). Although, as we have discussed above, post-secondary credentials become more important in modern society, there is very little research on the transition from different kinds of postsecondary education to work. Existing research refers mainly to differences between academic and vocational tracks (cf. Shavit and Muller 1998). In this paper we examine the transition from post-secondary education to work in Israel, with special emphasis on gender differences. We look at the transitions from different levels of higher education - for example, BA, MA and PhD - and from various fields of studies - for example, teaching and engineering (see Freeman 1980) to the labour market. We focus on gender differences because men and women tend to work in different occupations (LewinEpstain and Semyonov 1989) and attain different educational qualifications (Jacobs 1995). Further, the characteristics of male and female dominated occupations are distinct. Occupations dominated by women tend to have lower rates of pay, are less likely to offer fringe benefits and afford opportunities for training, promotion or the exercise of authority (Reskin 1993, p. 242). Three labour market outcomes are examined: employment status, occupational prestige attainment and job match. We aim to further our understanding of the transition from higher education to work - which has previously been conceptualised in terms of the labour market consequences of vocational vis-à-vis academic tracks in higher education - by including a more detailed differentiation of higher education tracks. More to the point, we employ models that include a specification of fields of study. We find marked gender differences in labour market outcomes among individuals with equal education levels but different fields of study. We conclude the paper with a discussion of these differences in an attempt to explain them. The Social Context Transitions between higher education and work are best understood in their specific national and historical contexts. In this paper, we present data on Israeli Jewish society which represents an interesting case study. Firstly, the higher education sector in Israel has expanded
dramatically over recent decades. As a result, a long-term growth in the enrolment of men and women in higher education is apparent in Israel. Secondly, and related to the above, Israel has a very high percentage of enrolment in post-secondary education. Table 1 shows how in the period 1961-1994, the proportion of those holding a post-secondary degree has increased from 2.7% to 11.7%, and the proportion of those holding an academic degree has increased from 6.4% to 20.6%. Note, moreover, that a large part of this increase occured in the 1970s and 1980s. (As we shall discuss below, our data are from 1984.) Table 1: Distribution of Tertiary Education Type by Year, Jews Aged 14+ Year 1961 1972 1980 1984 1990 1994
Vocational (%) 2.7 4.9 6.7 8.0 9.5 11.7
Academic (%) 6.4 9.1 12.1 13.3 16.1 20.6
Total (%) 9.1 14.0 18.8 21.3 25.6 32.3
Source: Israeli Statistical Abstracts, Number 46 (1995), table 22.3, p. 634.
The other side of the transition from school to work is the labour market. Women’s participation rates in the Israeli paid economy are similar to those of women in other western societies (Ben-Porath and Gronau 1985). Forty five percent of Israeli women are active in the paid economy and a majority of them (more than 60%) are employed in full-time jobs (Stier 1998). As in most western societies, gender segregation in the Israeli labour force is high. Most Israeli women work in a few female dominated occupations (Cohen et al. 1987). In 1972, three quarters of working Israeli women were employed in female dominated occupations, and nearly one third of them were found in only seven (three digit) occupations (Izraeli 1979). In 1983, the proportion of women in predominantly female occupations was 77.2% (Cohen et al. 1987). Most Israeli women work in white-collar occupations and thus enjoy a relatively high level of occupational prestige (Semyonov and Kraus 1993). At the same time, women’s occupations pay less than men’s occupations (Izraeli and Gaier 1979). About half of all employed women work in the public sector, which is the sector where labour laws, especially those related to women’s work, are strictly observed (Kraus 1992). Research Question The question we address in this study is as follows: Is it beneficial, in terms of labour market returns, to acquire any higher education - quantity - or is it the field of study that matters quality? To this end, we examine three measures of labour market returns to different fields of study within tertiary education in Israel. We look at the association between higher education and employment status. We look at the returns to different fields of study and different degree types in terms of occupational prestige attainment. Finally, we look at the association between higher education and job match.
Data and Variables 1. The Sample Data are drawn from the 1984 Higher Education Survey in Israel. This survey is a supplement to the 1983 Israeli census which includes a random sample of 10,914 individuals who represent Israel’s tertiary education (vocational or academic) degree-holders. The analyses employ a subset of Jews aged 25-44 who were educated in Israel (N=5100).2 Concentrating on this relatively narrow age range ensures that we study individuals who were part of educational systems that were similarly situated in historical and social contexts.3 We exclude Arabs from the analysis since it is well established in the literature that Arabs and Jews in Israel do not share similar patterns of educational and occupational attainment (Shavit 1984). Furthermore, gender differences are more pronounced among Israeli Arabs since the majority of Arab women do not participate in the paid economy. We also restrict the analyses to younger cohorts (ages 25-44) in order to minimise the effect of career mobility. This allows us to examine the direct link between education and work. Put another way, we want to minimise the effects of experience and negotiation power on the transition into work, while focusing on individuals who were part of comparable educational systems. At the same time, we have to allow enough time for our respondents to complete military service, obtain a PhD and start working. Our post-secondary vocational degree-holders cannot accomplish this earlier than the age of 25, while our doctoral degree-holders can reasonably accomplish this by the age of 44. Finally, we carry out the analysis separately for men and women to examine whether schooling has different labour market consequences for the two genders. 2. Variables As our prime objective is to study the school-to-work transition, our dependent variables will represent the labour market side of this equation. For the purposes of the current study, it is important to focus on job held or entered after the completion of the last and highest level of tertiary education. At the same time, we do not wish to allow a long period of time to pass between graduation and entrance to the labour market. In other words, we wish to measure the direct association between school and work, and to control for any disturbances, mainly those connected with career mobility. Since the study is based on cross-sectional data, we cannot satisfy these criteria completely. We need, therefore, to organise our data in such a way that an acceptable linkage between school and work is achieved. First, as mentioned above, we are looking only at young cohorts (ages 25-44). Secondly, the respondents were asked about their last, that is their highest, educational qualification and the year they attained it. They were also asked to give information about their current labour force activities (that is, in 1984), their employment status (and occupation title) and the year in which they started to work. 2
The 1984 Higher Education Survey provides the most recent data on tertiary education in Israel. Nevertheless, some critics may still argue that these data are too ‘old’ for dealing with the questions we address here. Our reply to such a critique would be that the data we analyse are relevant for studying labour market consequences of the expansion of tertiary education through to the 1980s. What is more, we would argue that from a historical perspective, this study offers an examination of the initial expansion of tertiary education and its social consequences in Israel. 3 The Israeli educational system, like most others, underwent a major reform in the early 1970s. Much of this reform, moreover, was aimed at equalising educational opportunities at the secondary educational level. (For more on this reform, see Kraus et al. 1998 and the citations therein.) A major advantage of using the relatively ‘old’ 1984 Higher Education Survey is that most men and women in this survey entered secondary schools under
From these pieces of information we are able to classify the data in a more refined way and construct the dependent variables. Twenty two per cent of the respondents started their current job five years or more after they achieved their last educational qualification (see table 2). Since we do not wish to confuse the school-to-work transition with other factors, we are inclined to exclude this group from the analyses.4 Therefore, the analyses will concentrate on Israeli educated Jews, aged 25-44, who started their current job less than five years after graduation. Having established that, we can now explain the construction of the dependent variables. Table 2: Descriptive Statistics (Proportions, Means and Standard Deviations) by Gender, Jews Aged 25-44 Variable Father’s Education Primary or less Secondary Vocational Secondary Academic Post Secondary University Degree Ethnicity (Ashkenazim) Gender (Men) Religious High-School Secondary School Academic Vocational Other External Degree Post Secondary BA MA PhD Belong to 1984 Labour Force
4
All
Men
Women
36.5 14.2 23.9 5.9 19.5 89.1 57.8 17.2
38.0 13.8 23.4 6.1 18.6 88.2 14.0
34.5 14.8 24.5 5.6 20.7 90.3 21.5
70.8 24.1 5.0
59.8 34.0 6.2
85.9 10.6 3.5
35.9 45.7 15.2 3.3 88.3
28.5 50.0 17.0 4.4 91.6
46.0 39.6 12.3 1.7 83.8
Table A1 presents the results of a logistic regression estimating the effects of selected variables on the (log) odds of being in the ‘excluded’ group relative to the others. It shows that (i) in all fields of study alike women are less likely to be found in the excluded category, whereas men are more likely to be found there, and (ii) the older a person is and the later in time that person graduated, the lower the chances they will be found in the excluded category. The implications are that we exclude, ceteris paribus, men and young individuals. Women are more likely than men to enter the labour market immediately after graduation and tend to stay in the same job for a longer time. Young individuals approach their ‘occupational maturity’ in their early thirties (Goldthorpe
Variable Economic Sector Government Histadrut Private (Non-Profit) Private Educational Unknown
All
28.7 14.0 11.1 23.2 20.0 3.0 Occupation/Study Field Match 62.9 Beginning 1984 Job Relative to Graduation More Than Five Years Before 8.6 Four to One Year Before 12.4 Up to Five Years After 57.1 1 [More Than 6 Years After] [22.0] Occupational Prestige 75.84 (18.92) Age 33.36 (4.97) Year of Graduation 77.64 (4.75) N (minimum) 3409 1
Men
Women
34.1 14.9 9.3 31.0 7.2 3.4 63.3
20.9 12.7 13.6 12.0 38.4 2.5 62.2
9.5 13.5 57.9 [19.1] 77.17 (19.62) 33.90 (4.83) 77.90 (4.50) 2008
7.4 10.9 56.0 [25.7] 73.91 (17.70) 32.63 (5.06) 77.27 (5.06) 1401
Omitted from the analysis (see text).
The first dependent variable is Employment Status: whether or not respondents were employed at the time of the survey (1984). Here we want to assess whether the (log) odds of being employed decrease or increase depending on education levels - so we can test the overqualification argument (de la Fuente and Smith 1995) - and on a specific field of study (controlling for education levels). In Table 2 we can see that the level of employment in our sub-sample is high (about 90%) and that women’s employment rate is somewhat lower than that of men. The second dependent variable is Occupational Prestige. We use the Kraus prestige scale (Kraus 1976) which is based on the 1972 three-digit standard occupational codes in Israel and ranges from 0-100. We chose to study status and not income attainment since we believe that prestige is a more adequate measure that reflects structural elements in the labour market, while income attainment is associated with the individual: experience, negotiating skills and so on. Thus, we study occupational prestige attainment for different degrees and fields of study in Israel to assess whether “more is better” or whether it is a question of studying for the “right” degree. Table 2 shows that men attain, on average, a higher level of occupational prestige than women. It must be noted that the average level of occupational prestige in the national population, for the same age groups, is considerably lower for both men and women (about 43 points, see Kraus et al. 1998). This is due to the fact that our sample population includes only highly-educated Israelis, and education is a key determinant of occupational prestige. The third dependent variable is a Job Match variable that we constructed. We assign each respondent to one of 11 academic or 5 post-secondary (mainly vocational) fields of study, based on information about the respondent’s last (representative) field of study. (See table A5 in the appendix.) We then construct occupational groups based on the three-digit occupational codes. We start with 13 occupational groups that are similar to the fields of study we have in our file (see table A6). These are our matched jobs. The remaining occupations, which could
not be matched, are assigned to a modified version of the 11 major occupational groups in Israel (cf. Kraus and Hodge 1990, Appendix C, pp. 185-188).5 Thus, altogether we created 13 occupational groups that we consider to be qualification related occupational groups (matched jobs) and 6 occupational groups which could not be matched. Based on these we construct a dummy variable (1=matched, 0=otherwise). Table 2 shows that nearly two-thirds of the Israeli respondents have a matched job occupation with very little gender difference. The labour market outcomes mentioned above (that is, the dependent variables) are determined by three main factors: social origins, educational history and labour force activities. We represent these factors by a series of independent variables. Social origin measures are represented by three variables. Father’s Education is defined by five categories: primary education or less, secondary vocational education, secondary academic education, post-secondary education and university degree. We construct five dummy variables from these categories and contrast the first four with the last one. Father’s education is also taken as a proxy for father’s occupational prestige since the latter is not available. Ethnicity is measured on the basis of the respondents’ and their fathers’ country of birth: European-American origin and Israeli origin (Ashkenazi) contrasted with Asian-African origin (Sephardi). We can see in table 2 that the majority of the respondents are of Ashkenazi origin (90%). It is well documented that in Israeli society, Sephardic Jews attain less education relative to Ashkenazi Jews (Kraus and Hodge 1990). Religiosity is measured on the basis of attendance at a religious high school as opposed to a non-religious high school. Respondent’s educational history was measured by three variables. Secondary School is a measure with three categories: academic secondary school, vocational secondary school and external secondary school. The latter refers to individuals who obtained their secondary certificate (matriculation diploma) in alternative systems. Table 2 shows that men are represented over three times more than women in the vocational track. These vocational postsecondary studies, however, lead to the matriculation diploma and are not to be confused with vocational studies that prevent one from obtaining a post-secondary degree (cf. Shavit 1984). Degree is a measure with four categories, representing the highest degree obtained: Post Secondary (mostly vocational), BA, MA and PhD. Here we can also see a disparity between the genders; women are over-represented in post-secondary degrees and under-represented in academic degrees, whereas for men the opposite is true. Field of Study is a measure with nearly 80 specific fields of study. From this information we construct 16 categories (see table A5). Figure 1 presents the distribution of fields of study by gender. It shows that women are over-represented in the humanities, arts, social sciences, teaching and nursing school, while men are over represented in economics, law, medical school and engineering. Similar results have been reported in the US (cf. Jacobs 1995, table 2, p. 88).
5
Firstly, we combine sales with proprietors, and then skilled, semi-skilled and unskilled workers. Secondly, we combine professional with scientific professional, since most of them are assigned to the matched occupations. Finally, clerical workers (which are part of the 11 major occupational groups in Israel), are assigned to the
Figure 1: Distribution of Gender by Field of Study, Jews Aged 25-44
30
20
Per cent
10
Women
0
men Hum Teacher
Art
Soc Econ
Medicin Mat Law ParaM ed Nuater
Eng Teachpesr
Tech
Celrcial Nuesr
Ohter
The labour force measures are represented by three variables. The Job Match variable was mentioned above and we made use of it both as a dependent and independent variable. Economic Sector is a measure with two broad categories: public sector and private sector. The former was separated into three categories: government, histadrut (Israeli union) and educational sectors. Private sector was divided into two categories: non-profit private sector and the remaining private sectors. We construct five dummy variables from these categories and contrast the first four with the last one. Table 2 shows that women are over-represented in the educational sector and under-represented in the government and private sectors. We introduce this variable in the analyses to control for differences in outcomes that may be related to the structure of the Israeli labour market and not to the individual. Timespan Between Graduation and 1984 Job is a measure with four categories. Those who began to work before graduation were assigned to the first two categories: the first category represents individuals who began their job more than four years before graduation; the second category represents individuals who began their job 1-4 years before graduation. The third and fourth categories consist of those who started to work after they graduated. To remind the reader, however, we excluded from the analyses those who started to work more than five years after graduation (that is, the fourth category). We believe that the second category consists of individuals who work in occupations that reflect their field of study - they started to work soon after they began to study. The third category is likely to consist of individuals in their first job after graduation. To control for these differences, the three categories were included in the analysis as dummy variables, contrasting the first and the second categories with the third. In addition, to control for changes in the opportunity structure of the Israeli labour market, we employ two variables measuring time: Age at the time of interview and Year of Graduation.
Results 1. Labour Force Participation We begin our substantive examination with the following question: what is the association between higher education and employment status? We are particularly interested in two types of effects which education may have on employment status. The first is a ‘quantitative’ effect; does more education increase or decrease the likelihood of being employed? The second is a ‘qualitative’ effect; do different fields of study have different labour market consequences (controlling for levels of education)? We are interested to see, moreover, if these effects have different labour market consequences across genders. The detailed analysis is presented in table A2, and we discuss and present the most relevant findings. Turning to our first point of interest - the quantitative issue - we present in table 3 the ‘net effect’ of each level of degree (relative to a BA degree) on the likelihood of being out of employment in 1984, for men and women separately. Apart from the ordinary b coefficients of these contrasts (which are extracted from table A2), we present in table 3 the quasivariance of each b coefficient.6 The advantage of the quasi-variances is great when interpreting the effects of dummy variables. The standard interpretation of regression coefficients referring to a set of dummy variables is of a pairwise comparison. The researcher sets one of the coefficients (commonly known as the omitted category) to zero, and interprets the other coefficient as a measure of deviation from this benchmark. However, it is not possible to infer from this analysis the existence of (statistically significant) differences between the effects of pairwise comparisons that do not include this omitted category. In order to know if there are some statistically significant differences of that sort, one has to change the omitted category and present all possible pairwise comparisons. The use of quasi-variances allows the researcher to calculate all possible pairwise differences and to test whether or not these differences are statistically significant without re-analysing the data. Consider, for example, the entries in the fifth column of table A2. The (logged) ‘net effect’ (b coefficient) of a vocational post-secondary degree (relative to a BA degree) on employment status is -0.322, with standard error 0.218. In table 3 we present this coefficient and its quasivariance. We calculate its quasi-standard error as follows: Q − s.e. = (Q − varB. A. + Q − varVOCATION ) , where Q-s.e. is the standard error for the comparison between a vocational post-secondary degree and a BA degree, Q-varB.A. is the quasi-variance of the coefficient for a BA degree and Q-varVOCATION is the quasi-variance of the coefficient for a vocational post-secondary degree. The Q-s.e. for this comparison (0.218) and the other two comparisons with BA are then presented in the third column of table 3. It is possible to calculate the Q-s.e. of any contrast (with the corresponding b coefficient difference) in this way.7
6
The quasi-variances reported in this paper were generated by a web based QV calculator . For full documentation and explanation of this methodology, see Firth (1999). 7 For example, if one is interested in calculating the contrast MA/PhD the following steps are required. First, the difference of the b coefficients between MA and PhD is calculated: (-0.504)-(-1.964)=1.46. This is equivalent to setting the MA coefficient to zero, and the new b coefficient (1.46) means that men with an MA degree are over 1.46
It is clear from table 3 that marked gender differences exist in the effect of education on employment status. For women, the effect of educational level on employment status is not statistically significant in the three comparisons presented (for the contrasts with BA degree).8 For men, on the other hand, an MA degree and a PhD degree (compared to a BA degree) increase the likelihood of being employed in 1984.9 The general pattern we find for men suggests that the higher the degree the more likely men are to be employed. These findings, in turn, would appear to run contrary to the over-qualification hypothesis. That is, the more qualified graduates, compared to the less qualified graduates, do not have a higher risk of being out of employment. Furthermore, in the case of men the evidence presented above suggests the opposite. Table 3: Net Effect (Log Odds) of Educational Levels on Being Out of Employment (From Model II in table A2) for Israeli Jews, Aged 25-44 in 1984, by Gender.
Educational Level Vocational Degree (P.S.) BA MA PhD
b -0.322
Men Q.Var.a 0.042
Q-s.e.b 0.218
b 0.252
Women Q.Var. 0.016
Q-s.e. 0.159
Set to 0 -0.504* -1.964*
0.005 0.054 1.030
0.244 1.017
Set to 0 -0.336 -5.161
0.010 0.059 55.076c
0.262 7.422
Notes: * p