Social and Economic Inequality in the Educational Career: Do the Effects of Social Background Characteristics Decline?

European Sociological Review, Vol. 13 No. 3, 305-321 305 Social and Economic Inequality in the Educational Career: Do the Effects of Social Background Characteristics Decline? Marianne Nondli Hansen This paper studies the influence of social and economic background on educational decisions during the complete career in the Norwegian educational system. The data consists of a sample of Norwegians between the age of 30 and 34 in 1990from the population censuses. Studies of the development of social inequality in educational attainment often use a sequential binary choice model. The model used here seeks to provide a more realistic representation of the structure of educational choice facing students, which involves multiple as well as binary choices. The results support those of previous studies in that social class was found to have the greatest impact in the first transition, from compulsory school to secondary education. However, the main impression is of continuing social inequality throughout the educational career. Those originating in the higher social classes are more likely than others to choose prestigious tracks of tertiary-level education. Parental income affet-tatherhnitv nftrarb and the leriel of theiinhwraityhi-gtw fnr thraie who rvtnmw aradrftmc ff^fhiefl with graduate and undergraduate degrees. Introduction The twentieth century has been a period of educational expansion at all levels, from primary to tertiary. This is true of Norway as well as other industrialised countries. The attainment of higher-level degrees among large proportions of the population may be expected to lead to inflation in the value of educational credentials and to increasing uncertainty about the returns from higher education (cf. Boudon, 1982; Collins, 1979). A tertiary-level degree per se will not ensure higher earnings, power, or privileges, but this is still true of some degrees, which have not become inflated to the same extent. In Norway, students in prestigious fields such as medicine or higher level business administration and engineering can look forward to secure jobs and high wages. Previous studies have indicated that social inequality in recruitment to such prestigious educational programmes is greater than it is in other tertiary education programmes (Erikson and Jonsson, 1993; Hansen, 1995). Because of these variations in the occupational outcomes of the different degrees, it is important to include tertiary education in analyses of social inequality in educational attainment. The main task in this paper is to assess the influence of social and economic background on educational decisions taken during the complete career in the Norwegian educational system. Does social inequality attenuate during the educational career, or do the effects of social and economic background characteristics persist? Following individuals up to the highest educational levels means that the sample population has to be of a relatively advanced age, because it is not unusual to graduate from university after the age of 30 (Berg, 1990: 26-27). The respondents in the O Oifbrd University Prcsj 1997

306 MARIANNE NORDU HANSEN sample used in this study were between the ages of 30 and 34 in 1990. The data are taken from the Norwegian population censuses of 1970 and 1990. The Model of Educational Choice Studies of the influence of social inequality on educational attainment often model the educational career as a sequence of binary choices. Such studies are concerned about the probability of reaching educational levely, given that one has reached the previous levely-1. Recent studies using this model indicate that social inequality is greatest early in the educational career. There seems to be a common pattern across countries of large inequalities in early transitions, and declining inequalities in the later transitions in the educational career (Mare, 1980; Blossfeld and Shavit, 1993). This means that social inequality in a specific cohort is greater in the transition from compulsory school to secondary-level education than in the transition from secondary-level to tertiary education among the members of the cohort who have survived through the secondary level. This conclusion is often based on educational classifications in which, in general, the highest level is university level. However, a study of nine European countries differentiating between higher and lower tertiary degrees indicates that inequality does not decline in a completely linear fashion in the sequence of educational transitions. Social inequality is greater in the transition from a lower to a higher tertiary degree than in the previous transition, from secondary to tertiary education (Miiller and Karle, 1993). For many educational choices the sequential binary choice model may be too simple. One discrepancy between this model and actual choice situations may be that the latter involve several rather than two possibilities. A second discrepancy may be that decisions about the level of education are taken all at once, at the same point in time, rather than in a sequence of continuation decisions. Those who choose a specific track thus do not later face the choice of continuing to a higher level. The discrepancy between the sequential binary choice model and the actual choice structure is perhaps most evident for the choice of tertiary level education. Students completing secondary school usually have the choice between multiple tracks of tertiary education. The level of education is often implied by the choice of track. Some courses of study are long term without involving further continuation decisions after the initial choice. Other choices lead only to undergraduate degrees, which cannot be used as a step towards a graduate degree. In that case, the binary choice model, which is concerned with the probability of reaching the next level given that one has reached the previous one, will be inappropriate. The model used here seeks to provide a more realistic representation of the structure of educational choice facing students. The first major decision made by Norwegian students is what to do after completing nine years of compulsory schooling at the age of 15-16 years. They have three alternatives. The first is to choose the academic track of secondary education, which qualifies them for tertiary education. 1 The second is to choose the vocational track, which ends in an occupational qualification, and the third possibility is to leave the educational system. Those who choose the vocational track later face the choice after one year of continuing their vocational education or leaving. The third possibility is to leave school altogether after compulsory schooling. The choice structure is shown in Figure 1. The academic track of secondary education takes three years. After completing their secondary examinations at the age of 18-19 years, the students can again been seen as facing a multiple choice. In this article I have simplified the options and reduced them to four alternatives. The first alternative is to choose a course of professional training, such as medicine, odontology, law, business administration, or engineering. These studies are estimated to take about 5-6.5 years. The second alternative is to embark on a university programme in the humanities, social sciences, or natural sciences. Students entering university later face the choice of whether to leave after obtaining an undergraduate degree or to continue and obtain a graduate degree. The total time estimated for obtaining a graduate degree is similar to a professional training about 5-7 years. The third alternative is to choose a short-cycle educational programme at a college of higher education, usually lasting three years, such as those offered

SOCIAL AND ECONOMIC INEQUALITY IN THE EDUCATIONAL CAREER 307 Professional tramng G«n*ral Unhmttyl Cod tg. LMvtng r\j University II ~ v\ Leaving Compulsofy school Vocrtona) Vocational D Luring Figure 1. The Norwegian educational system Leaving at teacher-training, nursing, business, or technical colleges. The final possibility is to leave the educational system after completing secondary school. The three tracks of tertiary-level education can only be ranked according to selectivity to some extent. The professional training programmes are the most selective. Admission is on the basis of competition, which has been severe throughout the whole post-war period. The main criterion for selection is secondary-school grades. Most of the other university faculties are open to all qualified applicants. For long periods the only entry requirement was a secondary-school certificate, while in other periods admission has been restricted and based on grades. 2 Graduate studies have always been restricted, with a system of admission based on the grades achieved for first degrees. Competition for admission to the short-cycle college courses has been far less severe than to the longterm professional training programmes. But due to the ceilings on the number of students, it has often been harder to be admitted to the most popular college programmes, such as social work, nursing, or journalism, than to the less restricted university programmes. Many students entering the shortcycle college courses actually have higher secondary-level grades than those entering university. Thus any ranking of these two tracks in terms of selectivity is inappropriate. Selection and Homogeneity in the Educational Career A long, but by no means uniform, tradition in sociology has regarded inequality in educational attainment as being primarily the result of cultural influences. According to one perspective, the different sets of values of the different social classes lead to inequality because education has a higher standing among the higher than among the lower classes (cf. for instance Hyman, 1953). Therefore those with lower-class backgrounds lack interest in school work, tend to receive low grades, and do not wish to obtain higher education. Another theory is that 'cultural deprivation' in lower-class environments means that children from such environments face difficulties at school because they do not bring with them from home the same knowledge and skills as the children from the higher classes (Coleman, 1966). Finally, inequality in educational attainment has been explained as the result of differences in cultural capital: Children from the higher classes, in contrast to those from the lower classes, have the linguistic and cultural competence required to succeed

308 MARIANNE NORDLI HANSEN at school, because it is the culture of the higher classes that is the basis of the knowledge taught at school (Bourdieu and Passeron, 1977). These explanations all mean that social inequality can be expected to be greatest early in the educational career. Those who have low motivation for school work or lack the ability to succeed - whether this is caused by cultural deprivation or by lack of cultural capital would have the highest probability of opting out early in their educational career. Therefore explanations of inequality in educational attainment that emphasize the association between class culture and school achievement fit well with the finding that the impact of social background shows a decline from the early to the later transitions. One explanation of declining inequality at the higher transition points, which may be designated the selection hypothesis, is based on differential selection (Mare, 1981; Blossfeld and Shavit, 1993). According to this hypothesis, students from different classes tend to make more similar choices at the higher than at the lower levels in the educational system because the effects of ability and motivation diminish at the higher levels. Given equal abilities, a smaller proportion of the lower-class students continue their educational career. Only those with the highest abilities attain educational qualifications above the lowest level. The result of this selection process is that the student body becomes more homogeneous with respect to abilities and motivation at every step up the educational ladder. The assumption that the abilities of students with different social backgrounds become more similar the higher the educational level seems reasonable in the light of previous research on the association between social class origin and grades (Boudon, 1974; Erikson and Jonsson, 1993). It has also been documented that there is little variation by class origin in the grades of university students in Norway, although it must be noted that the Norwegian research on this subject is scarce and often old (Aubert, 1963; Vangsnes, 1967; Aamodt, 1982; Hansen, 1986,1992). There is, however, reason to believe that social origin continues to influence educational choices at the higher educational levels. It has been argued that the costs and benefits of higher education vary according to social and economic background (cf. Boudon, 1974; Erikson and Jonsson, 1996). Therefore students with different social backgrounds tend to diverge with respect to educational choice, even if they have the same level of abilities or grades. Below I outline how two aspects of social background cultural characteristics and economic resources - may influence educational choice, independent of ability or achievement. Status Groups and Educational Choice One argument is that education serves as a basis for group cultures, or, to use Weber's concept, status groups (Weber, 1978 [1968]; Collins, 1971; Bourdieu, 1984, 1985). People in the status groups in which higher education is common will transfer their culture to their children. Therefore class culture may affect the ways in which the costs and benefits of educational choices are perceived. The benefits of choosing higher education will be greatest for those originating in the educational status groups. An important benefit of higher education is avoiding social degradation (cf. Boudon, 1974). Those originating in the educational status groups may have to aim at the highest educational levels to avoid a downgrading of their social position. Thus they would be especially eager to follow a professional training or undertake graduate studies. 3 Those originating in the lower classes do not risk social degradation if they abstain from the highest level of tertiary education, and stand to rise socially if they reach a medium level.therefore they are more likely than those originating in the higher class to be satisfied with the short-cycle, and therefore relatively inexpensive, college degrees. If the benefits of higher education are greatest among those originating in the educational status groups, they may be willing to spend more resources than others on reaching their educational goals. For instance, if their grades are too low to be accepted for the prestigious professional studies, they may be more willing than others to spend extra time on improving their grades and extra money on tuition. This view is supported by a recent study among medical students, a large majority of whom originate from higher classes, many having parents who themselves are doctors. As many as 75 per cent of these students had spent at least one year - and

SOCIAL AND ECONOMIC INEQUALITY IN THE EDUCATIONAL CAREER 309 often two or three improving their secondary grades (Wiers-Jcnssen eta/., 1995). Thus, even if abilities and achievements become more homogeneous at the upper educational levels, those with their origins in highly educated groups have the most compelling reasons for pursuing top-level educational programmes. Thus the impact of social class can be expected to persist up to the highest educational levels. This is the first hypothesis that will be tested out below. The Impact of Parental Income A second argument in favour of a continuing influence of social origin on educational choice is that the impact of economic resources will persist up to the highest educational levels. In an economic perspective, educational choices may be seen an investment in one's future human capital (cf. Becker, 1993 [1964]). Parental income is important for investment decisions, because richer families can more easily pay for their children's education, including the income loss when the children spend their time on education rather than at work. Within sociology the importance of parental income for educational choice has been emphasized especially by Boudon (1974; see also Gambetta, 1987). If parental income is an important factor in educational choice, there is no reason to expect this influence to attenuate at the higher educational levels (cf. Erikson and Jonsson, 1996). 4 On the contrary, the costs of education often increase with the age of the student. Students above the age of 18 19 often move away from the parental home, either because they want to be independent or because it is necessary in order to attend a higher education institution. Their living expenses will therefore be higher than if they live with their parents. They may have to pay tuition costs, at least if they attend private schools. Moreover, the foregone earnings will be greater the more the student would have been able to earn, and the longer the duration of their education. The earning ability of a 16 17-yearold, fbrinstance,issmallerthanthatofa22-23-year-old. If the costs of higher education become greater with advancing age, the impact of parental income on educational choice may well increase higher up in the educational system. This would at least be true if the parents had to fully finance their children's education. However, they are often not required to do this. Many governments finance colleges and universities, at which the tuition costs are fully or partly covered, and public or private loans or grants may be available. The more extensive such measures are, the more independent the children's educational choices will be of their parents' income. Thus, presumably because of such measures, parental income has been assumed to have a decreasing rather than an increasing impact on educational choice during the educational career. According to the life-course hypothesis ot educational choice, the older one becomes the more independent are one's decisions of the economic resources, as well as the preferences, of one's parents (Gambetta, 1987;Blossfed and Shavit, 1993). Adult students who have left home will be able to make their own decisions as to what they want to do. This is an alternative explanation to the selection hypothesis for the above-mentioned empirical finding that social inequality in continuation decisions tends to diminish during the educational career. Against the life-course hypothesis it may be argued that systems aimed at providing equal opportunities are often flawed, and thus unable to counteract the effect of parental income (cf. Becker, 1993 [1964]). 5 The impact of parental income on higher-level educational decisions may be expected to vary between countries, according to the efficiency of their measuresforequalizing opportunities. The Cost of Education in Norway In the Norwegian system, a large part of the tuition costs is paid by the state. Therefore most of the cost of higher education consists of living expenses and foregone earnings. This means that the total cost of an educational programme is determined by its length, and not, for instance, by the prestige of the educational institution. I mentioned above that the estimated time taken to obtain a higher-level degree in Norway is 5-7 years. This means that students should complete their education in their mid-20s. However, the mean age of graduates is higher than this. In 1987, when the members of the cohorts included in this study were between the ages of 27 and 31, the mean

310 MARIANNE NORDU HANSEN graduation age was 29 years for women and 28 years for men (Berg, 1990: Figure 5). 6 If students move away from home after taking their secondary-school certificate at the age of 18-19, for instance, they will on average need a further ten years to achieve a graduate degree or a professional qualification. This means that for Norwegian students the total cost of an education in terms of living expenses and loss of earnings may amount to a considerable sum. There is a government loan scheme that provides economic support for students while they are studying, which aims to eradicate the impact of parental income on educational choice. Under this scheme every student has the right to receive a loan. The interest rate on these loans is at present higher than the rates offered by commercial banks. 7 There is also a system of grants, but this accounts for a much smaller proportion of public funding. In contrast to many other countries, there are no special support schemes for students from low-income backgrounds or especially gifted students (cf. Johnstone, 1986; Jonssontf*/., 1996). 8 Having an official scheme for student loans is an advantage for families who, due to their economic situation, would be denied loans in commercial banks. But obviously, the terms of the loans will seem less arduous for students who can rely on their families in difficult situations. The future payments will seem easier for these students than those from families in a more difficult economic situation. The economic security of the latter will be lower, so they will have more reason to avoid making longlasting economic commitments than students who, for instance, can count on an inheritance. 9 Moreover, empirical studies have shown that students often receive economic support from their parents that covers at least some of their educational expenses (Hansen, 1992; Lowe, 1995; Wiers-Jenssen eta/., 1995). Affluent parents are those best able to provide the most substantial economic support. Thus, a system in which the living expenses for those who need support have to be repaid at high interest rates cannot be expected to eradicate the impact of parental income. Because of the variations in the costs of long- and short-term studies including the costs involved in improving secondary grades that are often incurred prior to starting professional training - it is likely that parental income will influence educational choices in tertiary-level education. This is the second hypothesis addressed below. Data and Methods Data To test the two hypotheses set out above - that the impact of social-class origin and parental income persists throughout the educational career - I use a representative sample of approximately 8 per cent of Norwegians aged 30 34 from the Norwegian population census of 1990. The sample consists of approximately 26,000 individuals. The data from the censuses are organized as a panel, so it is possible to follow individuals over a long period of time. The classification of social origin is based on information about the parents in the 1970 census (for information on the census data, see Central Bureau of Statistics, 1983; Furseth, 1990; Skogvoll,1992). Economic background is measured in terms of household earnings the sum of the parents' earnings. 10 The year of registration is 1970, when the members of the cohorts in the study were aged between 10 and 14. It might have been more informative to use data on household earnings ten years later, at the time when the decision about whether or not to continue education at the university level was taken, but this was not possible for technical reasons concerning the data. However, an association, although not a perfect one, can be expected between parental income in 1970 and in 1980. High-income families tend to continue to be so over a period of time, and vice versa forlow-income families. Moreover, having an affluent family provides long-term economic security as well as current economic support. The income measure is based on information from the public tax registers on the annual earnings of the parents, which corresponds roughly to gross salaries. This source of information minimizes individual differences in reporting and memory lapses. Household earnings are a good measure of the standard of living of the family, and thus of the possibility of their helping to finance the children's education, for employees. The measure is less suited to those who are self-employed. For this category the tax register shows the income after certain deductions, which means that their income is not comparable to that of an employee. Therefore those

SOCIAL AND ECONOMIC INEQUALITY IN THE EDUCATIONAL CAREER 311 with self-employed fathers, such as farmers and shopkeepers, are omitted from the analysis of educational choice. Social-class origin is assessed by the father's occupation, if the father was present in the household in 1970, and if not, by the mother's occupation. The classification of social-class origin is fairly detailed, and intended to describe horizontal differences, between class factions, as well as vertical stratification (cf. Bourdieu, 1984). The idea is to distinguish between groups that differ with respect to cultural characteristics that may influence educational choices. Such differences are likely to be greatest at the highest levels, therefore a horizontal distinction is made only at this level. The first category consists of managers and business executives. 11 These are differentiated from top-level groups with more cultural capital. The latter are expected to have especially close links with the educational system, and therefore be particularly likely to have children who choose higher education. This group consists of higher-grade professionals and higher-grade teachers and engineers, public administrators, and various cultural-sector occupations. The third category consists of medium-level employees, the fourth category of lower-level employees and workers in service occupations, the fifth of skilled workers, and the sixth of unskilled workers. Approximately 12 per cent of the total sample have missing social-class origins. This is a heterogeneous category. Some parents failed to answer the census questionnaire. Others were living abroad at the time, or have immigrated to Norway since 1970. Finally, some were single mothers without any occupation, or parents who were disabled, sick, or unemployed. When those with missing class origins and the self-employed are excluded, this leaves us with a sample of approximately 17,100 respondents. Table 1 shows the mean income in the various class categories and the distribution by social-class origin of the sample. As expected, the mean parental income is highest among managers and business executives and lowest among unskilled workers. Method In the analysis of educational choice, I first address the question of social selection in the four transi- Table 1. Class origin andpartntaliruomc Social class origin 1. Managers, executives 2. Higher-grade professionals, teachers, engineers, administrators 3. Medium-level employees 4. Lower-level employees, service 5. Skilled workers 6. Unskilled workers Total Mean income (NOK 000s) 66.970 66.210 51.580 44.090 39.580 35.260 43.270 N 695 1019 3565 1430 2931 7486 17129 % 4.1 5.9 20.8 8.4 17.1 43.7 100.0 tions shown in Figure 1. The conditional proportion choosing a track by social class origin is assessed at every transition i.e. the proportion choosing a specific track given that they have reached the previous educational level. The next step of the analysis is to assess the impact of class origin on continuation probabilities independent of parental income, and vice versa. This is done by specifying logistic regression models for each of the four educational transitions. The models vary in complexity. Two are binary, and represent a choice of whether to continue the career or leave. This applies to those who chose secondary level vocational education, and those who chose academic university studies. In the binary logistic regression model the probability of moving to a higher level, given that you have reached the previous level, can be expressed as where e denotes exponentiation based on the constant e, and gfc) is the sum of the effects of the variables assumed to affect the probability of moving to a higher level, including a constant term PQ. The model of the choice after completing compulsory school has three alternatives, and those who complete secondary general studies are seen as having the choice between four alternatives. A general expression of the conditional probability of the three-category model is

312 MARIANNE NORDLI HANSEN P(Y=j\x) = e*(») (cf. Hosmer and Lemeshow, 1989: ch.8). Each of the models includes the effects of class origin, parental income, and gender. Social class origin is the categorical variable with six levels as described above. To obtain an appropriate representation of the impact of income, the effect of parental income squared is tested as well as parental income. Previous studies have indicated that women differ greatly from men in their choice of tertiary-level education, and that women originating in the lower classes are less likely to choose male-dominated educational fields, such as engineering or business administration, than women originating in the higher classes (Hansen, 1993).To test for differences between genders with regard to the impact of class, the interactions between class origin and gender are included in addition to the main effect of gender. Results Social-Class Origin Table 2 shows the observed transition rates by socialclass origin in the four educational transitions. In the first transition, from compulsory school to secondary-level education, there are large differences by class origin: 77 per cent of the sons and daughters of higher-grade professionals and teachers, engineers and administrators, etc. choose general studies. In contrast, this is true of only 20 per cent of those with their origins among unskilled workers, who tend to prefer vocational education; the largest proportion of school leavers are found in this class. The differences are smaller in the second transition concerning those who have chosen the vocational track in transition 1. But those originating among higher-grade professionals and teachers, engineers and administrators, in addition to the sons and daughters of managers and business executives, still have the highest continuation rates. Approximately 22 per cent of those who completed general studies at secondary school choose to leave after this stage. This means that about 27 per cent of the total cohort reaches the level of tertiary education, disregarding those who were omitted from the study. 12 This large proportion supports the view that analyses of social inequality in educational attainment should include tertiary education. We again see that the category of sons and daughters of higher-grade professionals, teachers, engineers, and administrators, etc. is distinguished by having the highest percentage going on to professional training and other university studies. The highest proportion entering the short-cycle college programmes is found among the sons and daughters of medium-level employees. The sons and daughters of skilled workers have the highest proportion of school leavers. Finally, we also find class differences in the transition from an undergraduate to a graduate degree. In sum, we see that those originating in the category of higher-grade professionals, teachers, engineers, and administrators distinguish themselves by having the highest transition rates to general studies at the secondary level, to professional and academic studies at university level, and to graduate studies among those who choose academic university studies. I have argued that those originating in this class are likely to have especially close links with the educational system, and indeed they do have higher transition rates to professional and academic tertiary education than the sons and daughters of the managers and executives, which is the category with the highest mean earnings (cf. Table 1). Thus it is not the category with the highest mean earnings that has the highest transition rates to professional and academic tertiary education. This supports the view that motivational differences based on status group cultures may exist. Economic Background and Social Origin Table 3 shows the results of logistic regression models estimating the effects of social origin, parental income, and gender. In part A, showing the coefficients pertaining to the first two transitions, we see a substantial effect of social class and income in the first transition. That the effect of income squared is negative indicates that the impact of income decreases at the higher income levels. This is not the case in the second transition, in which the effect of income squared is not significant, and therefore

SOCIAL AND ECONOMIC INEQUALITY IN THE EDUCATIONAL CAREER 313 Table 2. Educational transitions by social dossorigin (%) (A) Secondary level education Social origin Transition 1 General Vocational Leaving Transition 2 \focational II 1. Managers, executives 2. Higher grade professionals, teachers, engineers, administrators 3. Medium-level employees 4. Lower-level employees, service 5. Skilled workers 6. Unskilled workers Total N (B) Tertiary level education Social origin 1. Managers, executives 2. Higher-grade professionals, teachers, engineers, administrators 3. Medium-level employees 4. Lower-level employees, service 5. Skilled workers 6. Unskilled workers Total N Transition 3 Professional 16.3 21.1 12.3 9.2 6.1 5.8 11.0 655 63.8 76.6 52.5 37.7 27.8 20.1 34.8 5990 Academic 20.8 253 18.6. 12.0 13.9 13.0 17.0 1014 32.2 21.5 40.4 51.5 57.4 58.4 50.7 8725 College 44.1 40.8 49.7 553 50.4 53.3 49.6 2961 4.0 2.0 7.0 10.7 14.7 21.5 14.4 2504 Leaving 18.8 12.8 19.4 23.5 29.6 27.9 22.4 1337 573 573 50.9 48.9 46.1 39.8 44.5 3886 Transition 4 Graduate degree Note: Transition 2 applies to those who choose the vocational track in Transition 1. Transition 4 applies to those who choose the academic track in Transition 3 Those who only took the introductory course (Examcn philosophicum) arc omitted in this transition (f=213) The highest proportion in every column is presented in bold type. not reported. There are also significant, though small, effects of gender in the first two transitions. The interaction effects between class and gender were not significant in these two first transitions. Part B shows the coefficients for the third transition, from secondary school to tertiary-level education, for each of the three tracks relative to school leaving. The first three columns show the estimates for the main effects of the independent variables. The next three columns include the interaction effects between class and gender. Income squared was not significant and therefore is not reported. The largest coefficients in the first part of the table are found in the first column, for professional 32.4 36.4 29.3 24.5 28.6 26.0 30.1 236 studies versus leaving. The positive coefficients show that the conditional probability of choosing professional studies rather than leaving is smallest for the reference category, comprising the sons and daughters of unskilled workers. The coefficient pertaining to the class category of higher-grade professionals and teachers, engineers, and administrators, etc. is considerably larger than those pertaining to all the other class categories. Its size indicates that there are substantial differences in recruitment patterns: the odds ratio is approximately 5, indicating that the odds of choosing professional education versus leaving is 5 times as high among the sons and daughters of higher-grade professionals and teachers, engineers, and administrators

314 MARIANNE NORDLI HANSEN Table 3. Tbe effects of gender, social class origin, and parental income in five educational transitions: logistic regression models (A) From compulsory to secondary school, and from vocational level I to vocational level II Variable* Transition I General Vocational "Transition 2 Vocational II Male Social class origin: 1 " 1. Managers, executives 2. Higher-grade professionals, teachers, engineers, administrators 3. Medium-level employees 4. Lower-level employees, service 5. Skilled workers Income (10,000s) Income squared Intercept Number of cases -0.176* (0.051) 2.126* (0.209) 2.948* (0.235) 1.595* (0.081) 1.057* (0.100) 0.566* (0.071) 0.504* (0.055) -0.019* (0.006) -1.506 17129 * In each case the variable is compared with the Moving education* status. Reference category=unskilled workers. * Coefficient* significant at the (105 level. (B) The transition to tertiary education Variable Male Social class origin: 1. Managers, executive! 2. Professional, higherlevel teachers, engineers, administrators 3. Medium level employees 4. Lower level employees, service 5. Skilled workers Income (10,000s) Male*Class 1 Male'Class 2 Male*ClaH 3 Male*Class4 Male*Class 5 Intercept Number of cases Transition 3 without interaction effects Professional Academic College 1.601* (0.115) 0.862* (0.225) 1.602* (0.195) 0.804* (0.154) 0.439* (0.209) -0.110 (0.198) 0.189* (0.028) -3.329 5940 0.230* (0.085) 0.4O9* (0.194) 0.999* (0.169) 0.470* (0.122) -0.051 (0.177) -0.062 (0.144) 0.147* (0.025) -1.425-0.177* (0.067) 0.079 (0.159) 0.355* (0.143) 0.211* (0.093) 0.159 (0.124) -0.137 (0.102) 0.050* (0.020) 0.544-0.030(0.046) 0.897'(0.212) 1.152'(0.241) 0.573* (0.079) 0.472* (0.094) 0.306* (0.062) 0.234* (0.051) -0.012* (0.006) 0.378 0.109(0.045) 0.576* (0.148) 0.518* (0.148) 0.316* (0.068) 0.331* (0.084) 0.228* (0.060) 0.086* (0.017) - -1.303 8674 Transition 3 with interaction effects Professional Academic College 2.365* (0.338) 1.548* (0.445) 2.434* (0.383) 1.364* (0.357) 1.303* (0.452) 0.357 (0.463) 0.184* (0.028) -1.086* (0.501) -1.339* (0.431) -0.884* (0.394) -1.289* (0.513) -0.649 (0.516) -3.830 0.790* (0.177) 0.861* (0.265) 1.444* (0.230) 0.935* (0.172) 0.546* (0.250) -0.045 (0.218) 0.145* (0.026) -0.916* (0.356) -0.909* (0.304) -0.938* (0.232) -1.197* (0.355) -0.092 (0.292) -1.692 0.229 (0.123) 0.447* (0.209) 0.778* (0.189) 0.481* (0.123) 0.473* (0.174) -0.010 (0.135) 0.048 (0.022) -0.808* (0.293) -0.961* (0.263) -0.587* (0.175) -0.666* (0.246) -0.296 (0.204) 0.375 " In each case the variable is compared with the leaving education'status. b Reference category=unski!ied workers. * Coefficients significant at the (105 level.

SOCIAL AND ECONOMIC INEQUALITY IN THE EDUCATIONAL CAREER 315 Table 3 (Continued) (Q The transition to graduate studies triable Male Social class origin: 1 1. Managers, executives 2. Higher-grade professionals, teachers, engineers, administrators 3. Medium-level employees 4. Lower-level employees, service 5. Skilled workers Income (10,000s) Intercept Number of cases 1 Reference category = unskilled workers. Coefficients significant it the 0.05 level. Transition 4 Graduate vs. undergraduate degree 1.019* (0.168) 0.006 (0.366) 0.218 (0.292) 0.004 (0.252) -0.106 (0.396) 0.038 (0.306) 0.132* (0.041) -2.199 766 as among the sons and daughters of unskilled workers. 13 However, the odds ratio of choosing general education versus leaving in the first transition is as high as 19, displaying a higher impact of social class. Moreover, after controlling for the other variables, men are more likely than women to choose professional training rather than leaving. The same is true of those from more affluent as opposed to poorer families. A similar pattern, though less pronounced, is found in the second column, for academic university studies versus leaving. There are four significant effects in the third column. These indicate that there are significant differences first, between men and women; second, between the categories of higher grade-professionals and teachers, engineers and administrators, and medium-level employees on the one hand, and on the other, the remaining classes; andfinally,that income has apositive effect on choosing short-cycle college studies rather than leaving. A number of interactions between class and gender in the three final columns are significant, and negative. These coefficients indicate that the gender-based differences in the probability of choosing professional or academic studies are greatest among those with their origins arrtong unskilled workers. Part C shows the results for the final transition, made by those who choose academic university studies. Both parental income and gender affects the transition probability to graduate studies. Controlled for class and gender, those from more affluent families are more likely to continue than those from poorer families. The effect of income squared was not significant. The coefficients pertaining to class are small and not significant, which means that class origin does not affect the probability of continuing to graduate studies when gender and parental income are controlled. Gender has a large effect on the probability of continuing to graduate studies. This effect may at least to some extent be the result of the previously noted age difference between men and women in graduation age (cf. Berg, 1990). A greater proportion of women than men may not yet have finished their degree when educational level is measured at the age of 30-35. Some of the findings in Table 3 are illustrated in Figures 2 5. In these figures the expected probabilities of continuing or choosing a specific educational track are calculated for the sons and daughters of the higher-level cultural-sector class category of higher-grade professionals, teachers, engineers, and administrators, on the one hand, and unskilled workers on the other. These tend to be polarized with respect to educational choice. The figures show the variations in the expected probabilities on the basis of the models in Table 3 for men and women by level of parental income. 14 Figure 2 shows the expected probabilities of choosing general studies at secondary school by class, gender, and parental income. The curves illustrate class differences based on the assumption of equal parental income. Class is expected to have a strong impact on educational choice, even with similar levels of parental income. Moreover, income has a strong impact on the expected probability within each class of choosing general studies. 15 Of course, the assumption of equal income is unrealistic. As can be seen in Table 1, the mean income of higher-grade professionals, teachers, engineers, and administrators is almost twice as large as that of unskilled workers. However, there is some overlap in the income distributions of the two classes. In addition to the high-earning professionals, the higher-level cultural-sector class consists of groups that may have fairly low earnings,

316 MARIANNE NORDU HANSEN 2 0 3 0 4 0 5 0 6 0 70 8090 100 Parental income (NOK thousand*) Men higher level Men working class Women higher level - Women working class Figure 2. General secondary studies: transition by doss and genderfigure 4. Academic university studies: transition by class and such as various cultural-sector occupations. Moreover, if the father is an unskilled worker, the family may still have relatively high earnings if the mother also works outside the home. The differences in the expected probabilities would be greater if we took differences in the income distribution into account. One would then have to add an income effect to the class effect illustrated in Figure 2. Such an income effect might be expressed in terms of the relationship between the estimates of the expected probabilities for the mean incomes of 20 30 40 50 60 70 80 90 100 Parental income (NOK thousands) Men higher level Women higher level Men working class - Women working class Figure 3. Professional training: transition by class and gender 0.6 -r- 0.55 0.5 - >>0.45 - I 0.4- I 0.35 0.3 :i 0.25-0.2 - - H 0.15-0.1-0.05 gender 0 20 30 40 50 60 70 80 90 100 Parental income (NOK thousands) Men higher level Men working class Women higher level - Women working class 30 40 50 60 70 80 90 100 Parental income (NOK thousands) Men working class Women working class FigureS. Graduate degree: transition by gender the two classes (NOK 35,000 and 66,000).The difference in the expected probabilities among men with their background in the higher-level cultural sector with the mean parental income of their class and that for men with their background among unskilled workers is approximately 0.14. The ratio between the expected probabilities of these men is nearly 1.8. This means that if an unskilled worker increases his income from the mean income of his class to that of, for instance, a university professor, his son's probability of choosing general studies would be almost twice as great.

SOCIAL AND ECONOMIC INEQUALITY IN THE EDUCATIONAL CAREER 317 In Figure 3 the pattern of expected probabilities of choosing professional training is illustrated in the same way as in Figure 2. Again there are large differencesbetween the classes, given equal levels of parental income, and parental income has a strong impact on the expected probability within each category. For instance, the probability of choosing professional training among the men from the higher-level cultural class increases from approximately 0.25 to 0.35, if their parental income increases from the mean income of category 6 to the mean income of their own class. 16 This is an increase of 40 per cent. The interaction coefficients between class and gender in Table 3 indicate that men and women from the higher-level cultural class are more similar than those with their background among unskilled workers. We see that the probability of choosing professional training among women with an unskilled worker background is very low, between 0.01 and 0.03. The ratio between the probabilities of men and women from this class is therefore therefore larger than for the high-level cultural class, where the women display a fairly similar pattern to the men. Figure 4 shows the expected probabilities of choosing academic university studies. We see that women from the higher-level cultural class display greater expected probabilities of making this choice than men with an unskilled worker background. In the upper income levels the women from the higherlevel cultural class have a higher expected probability of choosing academic university studies that the men from the same class. From Figure 3 we know that men with this level of parental income prefer professional training. Table 3 showed that there were no significant class differences in the probability of continuing to graduate studies among those who chose academic university studies in the fourth transition. Therefore only the expected continuation probabilities for men and women in the category of unskilled workers are shown in Figure 5. Again, we see that the model indicates that parental income has a strong impact on educational choice. Conclusion and Discussion My task in this paper was toassess the impact of socialclass origin and parental income on educational choice in the entire course of a student's educational career. I argued that this is necessary because of the increase in the proportion of the population reaching tertiary education and the subsequent inflation in the value of educational credentials. The value of graduate degrees and degrees from the most selective tracks - which are still scarce - are not inflated to the same extent as the undergraduate degrees. Thus any study of the association between social-class origin and access to jobs with high earnings, power, and privilege, should differentiate between degrees and types of tertiary education. The findings of the present study support those of previous studies in that they confirm that in the first transition from compulsory to secondary education a substantial social selection process takes place. Approximately 35 per cent of the students in this study chose the general track of secondary schooling, opening up the possibility of continuing to tertiary-level education. Both social-class origin and parental income have a strong impact on the choice of secondary education. However, the main impression isthatofcontinuing social inequality throughout the educational career. Social origin and parental income continue to affect the choice of track at the tertiary level. Those with their origins in the educated upper classes highergrade professionals, teachers, administrators, engineers, etc. - and those with affluent parents are most likely to choose professional and academic studies. Among those who choose academic university studies, the probability of obtaining a graduate degree increases with the income level of the parents. Those with lower-class parents, and whose families are in the lower range of the income distribution, are most likely to choose short-cycle college studies or to leave after secondary school. This supports the two hypotheses that both social-class origin and family income affect educational choice at the higher levels of the educational system. Why do we find these patterns? I emphasized above that the costs and benefits of higher education vary. The impact of the economic cost will be greater the lower the income of the family. The results indicate that parental income has a considerable effect on educational choice at all levels. This supports the assumption that economic resources are essential for educational choice. This is also true in the Norwegian education system, in which students can

318 MARIANNE NORDLIHANSEN obtain loans at high interest rates irrespective of family income or school achievement. An important benefit ftom choosing higher education for those originating in the educated upper classes is that they avoid social degradation. If avoiding social degradation is a motive behind educational choice, those with their origins among higher-grade professionals, teachers, engineers, administrators, etc., would be more likely to pursue higher education than those with their origins among managers or business executives. Those originating in the latter category may be expected to have alternative avenues to success apart from the educational system. They can leave school at a relatively young age and perhaps pursue a business career that is not necessarily dependent on educational credentials. Those who wish to follow in the footsteps of a father who is, for instance, a physician or university professor, can only do so by acquiring a top-level education. The above findings support these assumptions. The educated upper classes are far more likely to choose the academic track at secondary school and professional or academic tertiary education than any of the other classes. If it is especially important for those from the educated upper classes to reach the highest educational levels, why do we not find significant class differences in continuation to graduate degrees when income is controlled? One possibility is that academic ability is the main factor behind selection for these degrees. It is easy to enter university, but it is also easy to fail. Therefore those with the lowest abilities will tend to drop out of their studies, and they will not continue with graduate studies, which require the minimum grade of an undergraduate degree. If there is no association between academic ability and class among those who enter university, and if grades are decisive for continuation, we should find no association between continuation rates and social-class origin, once parental income is controlled. Another explanation of the class differences in the choice of tertiary education track is that they are caused by differences in achievement. Unfortunately, there are no measures of school achievement in my data. However, I have assumed that class differences with regard to grades would decrease at the highest educational levels, an assumption that is supported by previous empirical studies. It is still possible that class differences with regard to grades, though smaller than at the lower levels, lead to class differences in the choice of higher education (cf. Erikson and Jonsson, 1996). The professional fields - which those in the educated upper classes tend to choose, particularly the men are more selective than the two other tracks, requiring especially high grades from secondary school. It is possible that the high attendance of those originating in the educated upper classes in thesefieldsisprimarily evidence of theirgood school grades. However, the higher educational ambitions among those originating in these categories may be an important reason for their scholarly success, in addition to ability. If they already tend aspire to the restricted professional studies at the level of secondary school, their level of ambition is likely to influence their results. An additional spur may be the wish to avoid social degradation. In addition, those who are admitted to professional studies often spend a good deal of time improving their secondary grades after leaving school. As mentioned above, this is true of 75 per cent of the current group of medical students in Norway. This investment of time and money is made without any guarantee of success. Moreover, the future grade requirements in these studies are never entirely certain, because they depend on the number of applicants and their grade levels at any one time. Clearly, the decision whether or not to invest so much time and money with no guarantee of success will depend on the resources in the family. Class differences with regard to grades is an unlikely explanation for the impact of class and income on the choice of academic university studies and short-cycle college studies. Students originating in the working classes are especially likely to choose these college courses, but the grade requirements for such courses are often higher than for the less restricted university system. Therefore there must be other explanations for the inequality in recruitment to university and college studies than differences in achievement. Notes 1. In today's system it also is possible to be qualified for tertiary education on the basis of occupational

SOCIAL AND ECONOMIC INEQUALITY IN THE EDUCATIONAL CAREER 319 training combined with work experience, and examinations in certain compulsory subjects (cf. NOU, 1988, 34). However, admission on this basis is so exceptional that it is reasonable to see vocational education as a path leading away from tertiary education. 2. Academic university studies had a relatively unrestricted entry for the cohorts involved in this study. The trend in recent years has been that most university faculties practise some form of restricted entry. At the end of the 1980s the number of students increased dramatically, and all the faculties at the University of Oslo, with the exception of the faculty of Theology, set a ceiling on the number of students it would admit. 3. Another benefit may be that children from the higher classes feel more at ease in the educational system. Because they like being at school better, the consumption value of higher education will be greater for them than for lower class students (cf. Erikson andjonsson, 1996, 22-27). 4. Erikson and Jonsson conclude that the impact of parental income is low in early educational transitions in Sweden. Less is known about the transition to university level education, they argue, but find reasons to believe that economic background is important at this level (Erikson andjonsson, 1993,1996). 5. Becker criticizes the US federal student loan program for having too low maximum loans, for misplaced and high subsidies and for high default rates (Becker, 1993, 22 [1964]). 6. The mean age varies considerably from faculty to faculty. In the humanities, the mean graduation age was 35, in the social sciences and in medicine 30, in law and natural sciences 29. There are two main reasons for the high graduation age. First, many students spend much time in paid employment (cf. Berg, 1990). Secondly, many have spent time doing other things before commencing higher level education, often improving their grades from secondary school. 7. One differences between the educational loans and the commercial bank loans, however, is that no interest is paid on the educational loans until the studies are completed.the normal term of repayment is 20 years, so that people have to spend a large part of their occupational life repaying loans from their years of education. 8. The Norwegian system can be contrasted to the British. Jonsson ttal. (1996) argue that the scholarships and/or free places given to talented British lower class children have significantly contributed to reducing class inequalities in Britain. 9. This assumption is sustained by a study of the distribution of the resources of the State Educational Loan Fund showing that the mean size of public support for students increases with the income of their parents (Hansen and Rogg, 1991). This pattern was more evident among students in the mid 1980s than among students ten years previously. 10. Unfortunately there is no information about the number of siblings in my data set, so it is not possible to construct an income measure adjusted for family size. 11. For a presentation of the ideas behind the scheme and classification procedures, see Hansen (1995). 12. Approximately 30 per cent of the present cohorts enter universities or colleges in their early twenties (Educational Statistics, Universities and Colleges, 1993: Figure 4) 13. The odds ratio is estimated by exponentiation of the coefficient. 14. For the calculation of probabilities in polytomous logistic regression models, see Hosmer and Lemeshow (1989, ch. 8). 15. The impact of parental income on educational choice within classes is also tested out by introducing interactions between class and income into the choice models, and by performing separate analyses for each class. These analyses support the pattern shown in Figure 2 and the subsequent figures. Parental income also tends to affect educational choice in the analyses of separate classes. The effects of the interaction terms between class and income tend to be non-significant. However, the results indicate that the effect of parental income is strongest in the category medium level employees. 16. Note that these estimates arc based on the income distribution of the total sample. This is different from the actual distribution of those who chose tertiary education, due to the selection by income that earlier had previously occurred in the transition from compulsory to secondary education. Acknowledgements I am grateful to Nils Asbiornsen, Arne Maastekaasa and anonymous reviewers for helpful comments on an earlier version of this paper. References Aamodt, P. O. (1982) Utdanning og sosial bakgrunn (Education and social background). Samjunnsthmomiskestuditr, 51 (Oslo -Kongsvinger, Central Bureau of Statistics).

320 MARIANNE NORDU HANSEN Aubcrt, V. (1963) Eksamcnslcaraktercrog sosial bakgrunn (Grades and Social Background). Tidsskrifl for samfunnsforskning, 4, 189-214. Berg, L. (1990) Studielepet. Vorprosjtktrapport {The Student Career), Report 11. NAVF's utredningsinstitutt, Oslo. Becker, G. S. (1993 [1964]) Human Capital: A Theoretical and Empirical Analysis with Special Reference to Education, 3rd edn. University of Chicago Press, Chicago. Blossfeld, H.-P. and Shavit, Y. (1993) Persisting barriers: changes in educational opportunities in thirteen countries. In Shavit, Y. and Blossfeld, H.-P. (eds.) Persistent Inequality: Changing Educational Attainment in Thirteen Countries. Westview Press, Boulder, CO., pp. 1-24. Boudon, R. (1974) Education, Opportunity and Social Inequality. Wiley, New York. Boudon, R. (1982) The Unintended Consequences of Social A ction. St. Martin, London. Bourdieu, P. (1984) Distinction. Harvard University Press, Cambridge, Mass. Bourdieu, P. (1985) Social space and the genesis of groups. Theory and Society, 14, 723-744. Bourdieu, P. and Passeron, J. C. (1977) Reproduction. In Education, Society and Culture. Sage, Beverly Hills, Calif. Central Bureau of Statistics (1983) Folke- og Boligtelling 1970: Do/tumentasjon. (The Population and Housingcensus Hout, M., Raftery, A. E., and Bell, E. O. (1993) Making 1970: Documentation), Internal Notes 12. Central the grade: educational stratification in the United Bureau of Statistics, Oslo. States, 1925-1989. In Shavit, Y and Blossfeld, H.-P. Coleman, J. S. (1966) Equality of EducationalOpportunity. (eds.), Persistent Inequality: Changing Educational Washington, U. S. Department of Health, Education A ttainment in Thirteen Countries. Westview Press, and Welfare. Boulder, CO., pp. 25-49. Collins, R. (1979) The Credential Society: An Historical Hosmer, D. W. and Lemeshow, S. (\989) Applied logistic Sociology of Education and Stratification. Academic Regression. Wiley, New York. Press, New York. Hyman, H. (1953) The value systems of different classes. In Educational Statistics (1983) Universities and colleges, 1 Bcndix, R. and Upset, S. M. (eds.), Class, Status and Norges OffisielU Statistikk. (NOS; October). Central Power. Free Press, Glencoe, 111. Bureau of Statistics, Oslo. Johnstone, D. B. (1986) Sharing the Costs of Higher Education: Student Financial A ssistance in the United Kingdom, Educational Statistics (1993) Universities and colleges, 1 Norges Offisielle Statistikk (NOS; October). Central The Federal Republic of Germany, France, Sweden and the Bureau of Statistics, Oslo. United States. College Entrance Examination Board, Erikson, R. and Jonsson, J. O. (1993) Ursprung ocb New York. Utbildning: SocialSnedrekrytering til HSgre Studier. (Ori- Jonsson, J. O. (1993) Persisting inequalities in Sweden. In gin and Education: Social Inequality in Recruitment to Higher Education). Stockholm, Utbildningsdepartementet, Statens offendige utredninger, 85. Erikson, R. and Jonsson, J. O. (1996) Introduction: explaining class inequality in education: the Swedish test case. In Erikson, R. and Jonsson, J. O. (eds.), Can Education be Equalised: The Swedish Case in Comparative Perspective. Westview Press, Boulder, CO., pp. 1-63. Furseth, J. (1990) Folke- og boligtelling 1990. (The Population and Housing Census 1990). Central Bureau of Statistics, Kongsvinger. Gambetta, D. (1987) Did Tbejjump or Were They Pushed? Cambridge University Press, Cambridge. Hansen, M. N. (1986) Social inequality in educational attainment in Norway. Research in the Sociology of Education and Socialisation, 6, 103-132. Hansen, M. N. (1992) Sosial bakgrunn blant begynnerstudentenc: betydningen av kulturelle og okonomiske forhold. (Social background among first-year students: the impact of cultural and economic conditions). In Berg, L. (ed.) Begynnerstudenlen. NAVF's utredningsinstitutt, Oslo. Hansen, M. N. (1993) Kjonnssegregering i hoyere utdanning: Betydningen av foreldrenes fagutdanning og sosiale bakgrunn for studenters valg av fag (Sex segregation in higher education). Tidsskrift for samfunnsforskning, 34, 3-29. Hansen, M. N. (1995) Class and Inequality in Norway: The Impact of Social Class Origin on Education, Occupational Success, Marriage and Divorce in the Post-War Generation, Report 95/15. Institute for social Research, Oslo. Hansen, M. N. and Rogg, E. (1991) Hoyere utdanning I Norge: Rekruttering, finansiering og omfordeling. Tidsskriftfor Samfunnsforskning, 32, 387-416. Shavit, Y. and Blossfeld, H.-P. (eds.), Persistent Inequality: Changing Educational Attainment in Thirteen Countries. Westview Press, Boulder, CO., pp. 101 132. Jonsson, J. O., Mills, C. and Muller, W. (1996) A half century of increasing educational openness? social class, gender and educational attainment. In Erikson, R. and Jonsson, J. O. (eds.) Can Education be Equal-