1

Curricular Choice: A Test of a Rational Choice Model

of Education

Forthcoming European Sociological Review

April 21, 2009

Limor Gabay-Egozi, Yossi Shavit

Tel Aviv University

and

Meir Yaish

University of Haifa

Word count: 10,820

Number of tables: 8

2

Abstract

Rational choice theories of education view student's educational decision

as a sequence of binary choices between options that entail long-term

utility and options that reduce short-term risk of failure. One of the best

articulated models of educational choice (Breen and Goldthorpe, 1997)

asserts that choice between alternative options is affected by students'

utility considerations, their expectations regarding the odds of success or

failure in alternative educational options, and their motivation to avoid

downward social mobility. We evaluated these propositions using data on

students' curricular choices in Tel Aviv-Jaffa high schools. We found that

educational choice was affected by subjective utility and failure

expectations, but not by class maintenance motivations. Just as

important, and contrary to the model's main assertion, educational

inequality between social strata was not mediated by any of these choice

mechanisms. Finally, and importantly, about a fifth of the students in Tel

Aviv-Jaffa did not choose between long-term utility and short-term risks,

but combined the two. These students, the hedgers, combined the riskier

scientific subjects that are expected to yield long-term utility with social

sciences and the humanities that reduce the risk of failure in the short

term, but are not expected to yield large long-term utilities. The hedgers,

moreover, were shown to be disproportionately female and drawn from

disadvantaged social strata. These results suggest that educational

systems that allow multiple rather than alternative choices may enhance

the attainment of working-class youth because they enable them to opt

for long term utility while providing a safety-net in the form of additional

safer subjects.

3

Theoretical Background

Introduction

In recent years policy makers have advocated choice in education as a means

to enhance equality of educational opportunity (Hayman et al., 1997; Plank

and Sykes, 2003). But as some argue, choice could be a stratifying rather

than an equalizing mechanism in the educational attainment process. For

example, Ayalon (2006), who studied students in Israeli high schools differing

in the degree of choice available to students, found that gender and

socioeconomic differences among them were more pronounced where choice

was prevalent. The implications of educational choice for educational

stratification therefore merit study.

One of the main debates in the literature on educational

stratification is encapsulated in the title of Gambetta’s (1987) book: Were

They Pushed or Did They Jump? Push factors refer to the various

structural and social constraints that determine students' educational

attainment, while pull factors refer to choice. These concepts echo

Boudon's (1974) distinction between primary and secondary effects. The

former are factors responsible for the association between social origins

and children’s scholastic ability and performance; the latter are factors

that account for educational choices controlling for class differences in

ability.

In recent years interest has grown in choice-related explanations of

educational stratification (e.g., Boudon, 1974; Gambetta, 1987;

Goldthorpe, 1996, 1998; Erikson and Jonsson, 1996; Breen and

4

Goldthorpe, 1997; Morgan, 1998, 2005). Rational-choice theory, for

example, assumes that individuals are conscious decision makers whose

choices are influenced by a costs-benefit calculus (Hedström and Stern,

2008). The growing interest in choice as an educational stratification

mechanism has led scholars to develop theoretical models to explain why

children of similar abilities but different class backgrounds are observed to

make different educational choices (Goldthorpe, 1996, 1998; Erikson and

Jonsson, 1996; Breen and Goldthorpe, 1997; Morgan, 1998, 2005). One

of the most influential of these models is Breen and Goldthorpe's

(henceforth BG) model of educational decision making. Although this

model was motivated by, and developed to explain persistent inequality in

class based educational opportunity (cf. Blossfeld and Shavit, 1993), it is

ultimately a model of educational choice (cf. Breen and Yaish, 2006;

Stocké, 2007; Van de Werfhorst and Hofstede, 2007), and we approach it

as such. Importantly, the BG model is theoretically well developed, and a

number of empirical studies have found support for its predictions

regarding the statistical association between class and educational

attainment patterns (Need and De Jong, 2000; Davies et al., 2002; Breen

and Yaish, 2006). Yet, the model's assumptions have rarely been tested

directly (cf. Stocké, 2007; Van de Werfhorst and Hofstede, 2007).

Choice and Educational Stratification

Most studies on the determinants of inequality of educational opportunity

focus on two main sets of variables: familial resources and characteristics

5

of the educational system. Among them are families' economic resources

(Duncan et al., 1998), cultural resources (De Graaf et al., 2000),

students’ track placement (Shavit, 1984, 1990) and the curriculum

offered in schools (Apple, 1990; Oakes, 1990; Ayalon, 1995, 2002;

Burkam et al., 1997). The privileged classes benefit from a variety of

material, cultural and cognitive assets which they mobilize to gain a persistent

edge in the competition for desired credentials, and the educational systems

are often structured in ways that benefit them further (e.g., Halsey et al.,

1980; Gamoran and Mare, 1989).

More recently, choice, rather than constraint, has attracted increasing

attention as a stratifying mechanism in the educational attainment process.

Some scholars have noted that despite the attenuation of structural

constraints on processes of educational stratification, inequality between

social strata in educational attainment persists. For example, Lucas argued

that despite the replacement of overt tracking by apparent choice in course

selection in America, de facto tracking persisted, as did class and race-based

inequality therein (Lucas, 1999). Scholars interested in pull factors argue

that inequality is also due to processes involving choice (e.g., Goldthorpe,

1996; Breen and Goldthorpe, 1997). Several studies have shown that

class differentials in educational attainment persist even after controlling

for push factors and scholastic performance (cf. Erikson and Jonsson,

1996; Breen and Jonsson, 2000; Need and De Jong, 2000; Davies et al.,

2002; Breen and Yaish, 2006). In England and Wales, for example, class

differentials in educational choice account for about 25 percent of inter-

6

class educational inequality (Erikson et al., 2005; Jackson et al., 2007).

The publication of BG's (1997) paper stimulated considerable

research on rational choice processes in education. Following Mare

(1981), BG view the educational attainment process as a sequence of

transition points at which students (and their families) decide whether to

drop out or to continue to the next level in any of the available

educational options. At each transition point students evaluate each available

option in terms of the expected utility to be gained by successfully completing

it, and in terms of the risk of failing to do so.1 Utility is defined as the

degree to which completion would enhance the student's future

occupational and economic attainment. Risk is defined as the odds of

failing to complete the prescribed course of study.

Importantly, the model implies that subjective utility and risk are

positively correlated. This is because educational trajectories that are

associated with high utility expectations are also associated with high

failure expectations. Thus, decision makers face an inherent dilemma

between maximizing the former and minimizing the latter.

The theoretical centerpiece of BG's model is the so called Relative

Risk Aversion mechanism (RRA). Accordingly, there exists a threshold

that determines a student’s minimum acceptable level of educational

attainment, which will guarantee entry to a class position at least as good

as that of their parents. In other words, students set their threshold at a

level that will minimize the risk of downward mobility (Breen and

Goldthorpe, 1997, p. 283-5). The model further assumes that students

7

from all social classes are equally motivated to maintain their parental

class position. The implications of the RRA mechanism for educational

choice, however, differ by class background. For upper class children it

implies choosing the risky option because this alone will lead them to the

upper class. Conversely, working-class children tend to choose less risky

options because they suffice for the attainment of working-class

occupations. When individuals reach an educational threshold which they

believe will gain them entry into the same social class position as their

parents', the costs of pursuing any further education (in terms of real

costs, earnings forgone, and the risk of failure to complete) outweigh the

utility of doing so (Breen and Yaish, 2006). Hence, net of scholastic

ability, the RRA mechanism raises the class differentials in educational

choice, and reproduces inequality of educational attainment between

classes.

In sum, BG's model postulates that students' educational choices are

affected by: their beliefs about the relative utility of available alternatives;

their beliefs about the relative odds of success or failure in each alternative,

and their motivation to avoid downward mobility. Although BG recognize class

differences in scholastic ability and economic resources (Breen and

Goldthorpe, 1997, p. 283), they further assume that net of ability, classes do

not differ in relative risk aversion motivation; future utility attributed to

educational options, and success expectations within different educational

trajectories. This study tests the motivational mechanisms underlying the

8

BG model in the context of curricular choice of secondary school students

in Tel Aviv-Jaffa in Israel.

BG's objective was theoretical and they did not test their

propositions empirically. In recent years others have attempted to do so

(Need and De Jong, 2000; Davies et al., 2002; Holm and Jager, 2005;

Breen and Yaish, 2006; Van de Werfhorst and Hofstede, 2007; Stocké,

2007), with mixed results. Most of these studies examine the model's

behavioral predictions rather than its motivational mechanisms directly.

For example, Need and De Jong (2000) analyzed data on Dutch

secondary school students. They evaluated the RRA proposition by testing

whether students wish to attain an educational level at least as high as

their parents'. They did not explore whether students' level of educational

aspiration was motivated by a fear of class demotion, as assumed by the

Breen and Goldthorpe model. Similarly, in their Danish study Davies et al.

(2002) noted that if RRA is correct, when students reach their parents'

level of education, they gain no additional utility from any further

education. They argue that this means that the impact of parental

education should be strongest on students' transitions up to the

educational level attained by their parents themselves, and would decline

thereafter. Again, this is an indirect derivative of RRA. Finally, Breen and

Yaish (2006) tested behavioral predictions of the BG model in England

and Wales but did not measure directly any of the motivational variables

suggested by the model.

9

Only two studies tested the actual motivational processes

underlying the model. Using survey data taken from secondary school

pupils in Amsterdam, Van de Werfhorst and Hofstede (2007) tested

whether, and to what extent, family cultural capital and relative risk

aversion explained the educational performance and ambitions of

secondary school students in Amsterdam. Unlike previous studies, they

measured RRA directly by asking students to indicate how closely they

agreed with a series of six statements such as "I find it important to

achieve a better job than my parents" (p. 399). The reliability of the

composite measure was 0.77. They found that the effects of social origin

on school performance were largely mediated by cultural capital rather

than by RRA, while RRA rather than cultural capital affected educational

ambitions. Hence, cultural reproduction theory provides an important

explanation for class inequalities in early school performance, whereas

ambitions are affected by concerns with mobility, as suggested by Breen

and Goldthorpe. By measuring RRA directly, Van de Werfhorst and

Hofstede advanced research on rational choice in education substantially,

but they did not measure the model's other motivational components.

Specifically they did not measure the subjective utility and the failure

expectations that students associate with the alternative options they

face.

Utilizing data on parents of fourth graders in Rhineland-Palatinate in

Germany, Stocké (1997), provided the most comprehensive test to date

of the motivational mechanisms proposed by the BG model: subjective

10

costs, success probabilities, and the status maintenance motivation

(RRA). He found that net of ability, parents' subjective beliefs about their

children's' odds of success, as well as RRA (but see below), strongly

affected their educational choices. However, contrary to BG's expectation,

he did not find that these mechanisms mediated class inequality in

secondary school choice.

Stocké measured RRA using two questionnaire items: (i) ‘For many

parents, the occupational future of their children is particularly important.

Would you please tell me how strongly it would bother you if your child

reached a less prestigious occupation than yourself?’ (ii) ‘Please think

about what your child will be able to reach in future with different

educational degrees. As how likely do you regard it that your child,

endowed with the different educational degrees, will be able to reach

occupationally at least what you reached?' (asked for all three degrees)

(verbatim, Stocké, 2007, p. 517 notes 4 and 5).

In our view, the first question measures RRA and is consistent with

the measure employed by Van de Werfhorst and Hofstede (2007). The

second question would appear to measure the labor market returns that

parents attribute to various educational credentials. In other words, it

measures utility expectations rather than RRA. We note in Stocké's

findings that while the measure of subjective utility affected educational

choices RRA did not.

The aim of our study is to provide a comprehensive and direct test

of the motivational mechanisms underlying the BG model. Like Stocké, we

11

simultaneously test whether subjective utility, failure expectations and

RRA affect educational choices. However, we employ Van de Werfhorst

and Hofstede's measure of RRA. We test the following hypotheses:

Hypothesis I: Students' educational choices are affected by their

subjective utility, namely their beliefs as to the likely returns on the

various educational options.

Hypothesis II: Students' educational choices are driven by their beliefs as

to their own odds of success or failure in the alternative educational

options.

Hypothesis III: Educational decisions are motivated by individuals'

apprehension of downward class mobility (i.e., the relative risk aversion).

Hypothesis IV: The three motivational mechanisms discussed above

mediate socio-economic inequality in educational choices.

Hypothesis V: Students of different socio-economic backgrounds (e.g.,

class) do not differ in any of the following: (i) their relative risk aversion;

(ii) their subjective future utility attributed to educational options; and

(iii) net of ability, their success expectations in the different educational

options.

The Setting, Data and Variables

The Setting

We tested these hypotheses with data on the choice of major subjects by

secondary school students in Tel Aviv-Jaffa in Israel. Our sample was

ninth and tenth graders who were about to choose a major for their

12

matriculation examinations. The choice of advanced courses and

examinations is an important junction in the socioeconomic life course of

young Israelis. At age twelve, after a year in pre-school and six years in

primary school, Israeli children enter middle schools where they spend

grades seven, eight and nine. This spell is followed by upper secondary

school in grades ten through twelve, where students are prepared for the

matriculation examinations leading to the award of the matriculation

certificate (bagrut), which is required for higher education.2

The secondary school curriculum is composed of compulsory and

elective subjects, such that students can choose from a range of

advanced academic courses in different subjects. The various subjects are

offered at different levels, ranging from one to five units, and students

must pass exams in subjects totaling at least 21 units to earn the

matriculation certificate.3

Seven subjects are compulsory (English, Mathematics, Civic

Studies, Bible Studies, History, Literature, and Hebrew language – in the

Hebrew school sector), and must be taken at either a basic (1-3 units) or

an advanced (4 or 5 units) level. Students must also take at least one

elective major at an advanced level. Advanced majors are usually chosen

during the spring term of tenth grade (although in one of the schools we

studied students made the choice as early as ninth grade).

The elective majors are not stratified formally, and they should in

principle all open meaningful educational opportunities. Nevertheless,

informal stratification of these majors does exist, such that students from

13

an advantaged social background tend to specialize in the sciences, while

those from a lesser social background tend to specialize in the humanities and

the social sciences (Ayalon, 1994, 1995; Ayalon and Yogev, 1997).

Admission into higher education predicates a matriculation

certificate, and the more prestigious departments and institutions of

higher education set several additional requirements: (1) advanced

English (at least 4 units); (2) a high matriculation grade point average;

(3) some technical and prestigious departments (e.g., engineering and

the sciences) and institutions (e.g., the Technion) specifically require

higher matriculation grades in advanced math and sciences. The

elaborate formula employed by universities also tends to favor applicants

who have passed more than one advanced level examination.

In recent cohorts, 54 percent of students obtained the matriculation

diploma (Ministry of Education, 2007). Of those who sat for matriculation

examinations, about 65 percent passed all the examinations necessary for

the diploma. Mathematics and English are generally regarded difficult

examinations, and unlike other difficult subjects (e.g., Physics, Chemistry

and Computer Science) they are compulsory. In an effort to raise

students' success rates in Mathematics schools evaluate their

performance early on and assign them to unit levels as early as ninth

grade. Once a student has been placed on a level he or she has little

further choice in the matter except to opt for a lower level.

Data

14

In the spring term of 2006 we collected data in four public secondary

schools in Tel Aviv-Jaffa. With a population of nearly 400,000, Tel Aviv-

Jaffa is Israel's largest city and is the core of the Tel Aviv Metropolitan

area with its population of about three million (Tel Aviv-Jaffa, 2007). Tel

Aviv-Jaffa has nine administrative districts, which are fairly homogeneous

socio-economically (Tel Aviv-Jaffa, 2006). Southern districts are typically

working class while northern ones are middle class.

We restricted our sample to Hebrew academic secular high schools

in Tel Aviv-Jaffa. We excluded vocational, religious, Arab and two magnet

schools because they differ in their curricular menu and emphasis.

Vocational schools offer a limited menu of academic subjects with an

abundance of technical ones, Jewish religious schools emphasize religious

studies and encourage their stronger students to take these rather than

the sciences, and Arab schools offer a somewhat different curriculum than

Hebrew schools. Specifically, they teach different languages (Hebrew and

English as foreign languages rather than English, Arabic and French), put

less emphasis on Bible studies, and teach some Koran instead. One of the

two magnet schools specializes in the arts and the other caters primarily

to Russian immigrant children and offers a mix of Russian and Hebrew

curricula. In the interest of simplification, we chose to restrict the sample

to schools which share a common curriculum.

Of the remaining seven schools in the city we sampled two in

working-class districts and two in middle-class districts. Fortunately, all

four school principals cooperated with the study and allowed us access to

15

their schools. The four schools offer students similar core curricula, as

well as the option to choose among the following advanced subjects:

Physics, Chemistry, Computer Science, Biology, Economics, Social

Sciences, History, Literature and Communication.

We distributed self-administered questionnaires among tenth

graders attending three of the schools and among both tenth and ninth

graders in the fourth (because students at that school where encouraged

to choose an advanced major as of ninth grade). In total, we collected

data in 28 classes and obtained 683 completed student questionnaires. Of

these, 18 percent were incomplete and were excluded from the analysis,

which accordingly covered 563 cases.4

Like Stocké's (2007) and Van de Werfhorst and Hofstede's (2007)

studies before us, our sample too is local rather than nationally

representative. However, our data include good measurements of the

central theoretical concepts under investigation, including relative risk

aversion, utility and failure expectations, as well as social origins and

other control variables (see below). While acknowledging the limited

generalizability of our findings beyond Tel Aviv-Jaffa, we believe that our

data are suitable to the task at hand. If the BG model holds as a general

theory, it should help explain socio-economic differences in educational

attainment in Tel Aviv and in other stratified communities. To anticipate,

the results of our study together with those of Stocké (2007) and Van de

Werfhorst and Hofstede (2007), add up to a rather consistent message

with respect to the BG model.

16

The Dependent Variables

We study the determinants of two types of curricular choices that

students make in anticipation of their matriculation examinations.

Respondents were presented with a list of electives that were available in

their school, and were asked to indicate whether or not they intend to

take each of them at an advanced level. The distribution of choices shown

in Table 1 indicates that boys tended to choose Physics, Computers,

Economics and other Social Sciences, while girls are more inclined to take

the Social Sciences, Communications, Economics and Biology.

Our first dependent variable distinguishes between those who chose

to take English at an advanced (4 or 5 units), rather than at a basic level

(3 units). Recall that admission into higher education predicates a

matriculation certificate with advanced English. Thus, the choice of

advanced English has very clear stratifying consequences. Given the

importance of advanced English for university enrolment, it is not

surprising to see in Table 1 that 93 Percent of our respondents intended

to take English at that level.5

Our second dependent variable concerns the choices that students

make among the elective subjects. Following common practice in the

literature on school curriculum (Apple, 1990; Kliebard, 1992; Kamens and

Benavot, 1992) we form the nine elective subjects into two groups: the

sciences (Physics, Chemistry, Computers, and Biology) and the

humanities and social sciences (Economics, Social Sciences, History,

17

Literature, and Communications). The sciences (hereafter hard subjects)

are generally regarded as more prestigious than the humanities and

social sciences (hereafter soft subjects) and are considered more

demanding (Ayalon and Yogev, 1997).

[TABLE 1 HERE]

In Table 2 we compare the three curricular choices in terms of

perceived utility and risk of failure. We measured perceived utility by

asking respondents the following question: "For each subject listed below,

in your opinion, if a student succeeds in this subject at the 5-unit level,

what are his or her chances of admission to a university?" [scale: 1 ('not

high at all') to 5 ('very good')]. Perception of risk is defined as students'

subjective belief that they would gain a low grade (64 or less on a 100-

point scale) in a subject were they to take it at an advanced level.

Students' perception of risk in the hard or the soft subjects is then refers

to the proportion of those who thought that their grade, in each subject

within the hard and the soft subjects, would be low. As seen in Table 2,

on average, (across subjects) students perceived the hard electives as

associated with both higher utility and a higher risk of failure than the soft

electives.6 The utility that students associate with advanced English is

higher than the mean utility associated with either the hard or the soft

electives. In addition, the risk of failure they associate with advanced

English is comparable to that associated with the soft subjects which are

perceived as relatively safe.

18

[TABLE 2 HERE]

As indicated earlier, although students are only required to take one

advanced subject to be eligible for a matriculation certificate, there are

strong incentives for taking more because students can then discount the

advanced subject in which they scored lowest. The great majority (96

percent) of students in our sample did indicate that they intended to take

two or more advanced electives. Although most students tended to pick

subjects within the two clusters of electives, some intended to take a mix

of both hard and soft subjects. Thus, our second dependent variable

consists of a three-category classification representing choice of advanced

subjects: (i) hard subjects (Physics, Chemistry, Computer Science, and

Biology); (ii) soft subjects (Economics, Social Sciences, History, Literature

and Communication); and, (iii) a mixture of both. As seen in Table 3, 36

percent of the students in our sample intended to take advanced subjects

in the sciences alone, 42 percent intended to take only soft subjects, and

22 percent intended to take a mix of the two.

[TABLE 3 HERE]

Thus, about a third of students who intended to take a hard elective

seemed to hedge their risk of failure by taking a soft elective too. This

strategy seems quite reasonable given that students are only required to

complete a single advanced elective course and can discount an extra

course in which they might do poorly.

19

Independent Variables

As noted, the rational choice model of education refers to three

motivational variables (i.e., secondary factors) affecting educational

choice: relative risk aversion, and, for each educational option, beliefs

about future utilities and beliefs about failure expectation.

We measured Relative Risk Aversion with the following questions:

"To what extent do you agree with each of the following statements?

[scale: 1 ('strongly disagree') to 5 ('strongly agree')]:

a. It is important for me to work in an occupation that is better than the occupations of my parents

b. It is important for me that my salary be at least equal to the level of my parents' salary.

c. My parents would not be satisfied if I were to work at a lower occupation than theirs.

d. I would like to attain a social class on a level at least equal to my parents'.

e. I am concerned that I might reach a social class that is lower than the social class of my parents

f. It is important for me to attain a higher level of education than the level of education of my parents"

This measure is based on Van de Werfhorst and Hofstede's six scale items

measurement of Relative Risk Aversion (2007). Items (b) to (e) represent

students' concern with downward mobility whereas items (a) and (f)

represent their attitude to upward mobility. Consistent with Van de

Werfhorst and Hofstede's measurement, factor analysis of these six items

found them to load on a single factor, which we interpret to represent the

class maintenance motivation (i.e., relative risk aversion). This factor

accounted for approximately 63 percent of the total variance in the test

20

items, and the factor loadings of five out of the six items were above .66

(item e=.47). The relatively low loading of item e is probably related to

the translation of the original statement, and was therefore removed from

the analysis due to translation issue.7 .The five item factor accounted for

approximately 72 percent of the total variance, and its reliability was

identical to that obtained by Werfhorst and Hofstede (.78).

Beliefs about future utilities were measured by the question cited

earlier concerning the subjective odds that a student (i.e., any student)

who succeeds in each subject will enter university. For each student we

computed the mean perception of success associated with the hard

subjects and the soft subjects. The ratio of the former to the latter was

then our measure of the relative utility that students attribute to the hard

subjects as against the soft subjects.

Relative Failure Expectation was measured as the difference

between the proportion of hard and soft subjects respectively, in which

the student expected to obtain a low grade (64 or less). High values

represent apprehension of the hard subjects.

Control Variables

Following well documented gender differences in curricular choices (e.g.,

Ayalon, 1995; Bradley, 2000), we include a dummy variable for gender.

Against the relatively constancy of class differentials in educational

attainment, in nearly all advanced societies gender differentials in levels

of educational attainment (favoring males over females) have declined

21

sharply and even reversed since the 1970s. BG attribute the decline to

perceptions of rising occupational and economic returns to women's

education (Breen and Goldthorpe, 1997, p. 296-7). In a footnote, they

also refer to gender differences in subject choice (Breen and Goldthorpe,

1997, p. 303, note 10) and suggest that since women expect their

careers to be interrupted by family obligations, they are prefer subjects

which lead to occupations that afford flexible working arrangements. To

anticipate, and consistent with other studies on gender differences in

curricular choices (e.g., Ayalon, 1995; Bradley, 2000), we find boys were

more likely than girls to take hard rather than soft electives.

We indicate social origin by two variables: parental education and

family economic resources. We began our analysis with class as an

indicator of social origin but soon realized that student reports of father's

occupation and of his employment status (which are the building blocks of

the class schema) are highly unreliable, therefore, we prefer to represent

social origins with parental education and their economic resources.

Parental education was measured as the highest qualification obtained by

either father or mother. It was measured by a six-category classification

which maintains a clear hierarchical order. In addition, Students were also

asked to provide information on the availability of a variety of durable

goods in the home (such as air-conditioning, computer, dishwasher, car,

etc.). Economic resources were measured as the sum of available items in

the parental household.8

22

Students' scholastic performance was indicated by their grades,

measured as the mean of self-reported grades in Hebrew, English and

Mathematics on the most recent report card. Grouping was the level at

which a student considered himself or herself in tenth-grade Mathematics.

It was coded as a dummy variable indicating advanced placement. As

noted earlier, group placement is largely determined by students’ prior

performance in Mathematics and once placement has been made it is an

evident push factor which affects subsequent curricular choice. Students

placed in low level Mathematics are unlikely to choose advanced hard

subjects.

Finally, we controlled for the number of subjects the student

intended to take at an advanced level, because the odds of taking a

mixture of both hard and soft subject should increase, the more subjects

students take. The inclusion of this variable in our models did not

substantially alter the effects of any of the other variables.

Empirical Findings

Descriptive Analysis

Table 4 presents descriptive statistics of the independent variables by

curricular choices: the choice of advanced and basic English, and the

choice among the three categories of advanced matriculation electives.

Consistent with previous studies in Israel (Ayalon and Yogev, 1997;

Ayalon, 2003; Katz-Gerro and Yaish, 2003) and elsewhere (Ma and

23

Willms, 1999; Bradley, 2000), we found that boys were more likely than

girls to take hard rather than soft subjects.

[TABLE 4 HERE]

As expected, we also see that students who chose advanced English

and the more demanding and rewarding sciences were disproportionately

drawn from a more advantaged social background than those who chose

the less demanding subjects (cf. Oakes et al., 1992; for Israel see Ayalon,

1994, 1995; Ayalon and Yogev, 1997). More importantly, on average,

students who chose a mix of hard and soft subjects (the hedgers) came

from lower strata than those who selected either pure category of

subjects (hard or soft ones). This is an interesting result that we shall

return to later.

Regarding scholastic performance and grouping, the entries in Table

4 reveal that students who chose hard subjects won higher grades than

those who selected the mixed or the soft subjects. Similarly, those who

chose hard subjects were more likely to have been placed in an advanced

group in math. The same apply for the differences between those who

choose advanced English compared with their counterparts who intended

to take the subject at a basic level.

Turning to the motivational factors, we see that those who selected

advanced English evinced more confidence in their ability to succeed in

the subject than those who chose it at a basic level. There are no

significant differences in Relative Risk Aversion and in utility expectations

24

between those who chose advanced and basic English. As for the three

clusters of electives, we see that students who chose mixed subjects

exhibited the highest degree of relative risk aversion, suggesting that

those who are concerned with class maintenance hedge their risk by

adopting a mixed strategy. The hard (and risky) subjects increase the

expected long-term utility of the mix and if one does poorly in these,

success in the soft subjects averts short-term failure. Thus, choosing a

mix of electives could minimize exposure to unwanted short-term risks,

while allowing the possibility of long-term success in higher education and

in the labor market.

Expectedly, those who selected hard subjects hold stronger beliefs

about their positive utility (i.e., admission to university) than those who

chose the other combinations, and those who selected soft subjects

showed the lowest confidence in their ability to gain academic success in

the hard subjects. Also expectedly, the more advanced subjects a student

chose, the more likely he or she was to choose a combination of mixed

subjects.

The Effects of Social Background on Performance and the

Motivational Factors

Before testing whether the three motivational variables explain

educational choices, we estimated, as shown in Table 5, the effect of

social background on performance, and that of social background and

performance on status maintenance motivation, utility considerations,

25

and failure expectations. Consistent with previous research, social

background exerted positive effects on scholastic performance. Further,

boys and girls did not differ in their performance but they did differ

significantly in the motivational factors.9

[TABLE 5 HERE]

The last three columns of Table 5 test Hypothesis IV, which

postulated that when scholastic performance is controlled, no

socioeconomic differences should appear in any of the three motivational

factors. The results indicate that students from advantaged social

background are less concerned with maintaining their parental

socioeconomic positions. Thus, with regard to the Relative Risk Aversion

mechanism, our finding, consistent with that of Stocké (2007), refutes

one of the core assumptions of the BG model, namely that all social strata

are equally concerned about status maintenance. A plausible explanation

for the negative association between relative risk aversion and social

background might be related to the degree of confidence that people of

different social backgrounds have in their social position. Students from

advantaged social strata may rely on their family to supply, if necessary,

the buffers against downward mobility, or many of them may even take

their position as given and are therefore less concerned with status

maintenance. By contrast, students from less privileged families may

suffer from status anxiety and fear of status loss. Thus, to the extent that

RRA has a positive effect on the choice of high utility subjects, it may

26

actually attenuate, rather than reproduce inequality between generations.

To anticipate, we shall soon show that RRA did not affect educational

choice and does not mediate the effects of origins on educational choice.

Contrary to BG's assumption, but consistent with Stocké's (2007) results,

we found significant socio-economic differences in beliefs about relative

utility. No such differences appeared in failure expectations, which is consist

with the BG model and Hypothesis IV.

Primary and Secondary Effects on Choice of Advanced Subjects

Reverting to the first three hypotheses outlined at the outset, we asked:

are students' educational choices affected by their apprehension of

downward class mobility (i.e., relative risk aversion), by their beliefs

about the likely future returns on the various choices, and by their beliefs

about their own odds of academic success or failure in the alternative

trajectories? We also asked whether these secondary factors mediated the

effects of social origins in educational choice, that is: is choice a

stratifying mechanism?

To answer these questions we first estimated a binary logit model

on the choice between advanced and basic levels in English. We then

examined, by applying a multinomial logit model, the determinants of

choice between the three curricular clusters identified earlier (i.e., hard,

soft and mixed subjects). These models included the following

independent variables: parental education, family's economic resources,

scholastic performance, grouping in math, relative risk aversion,

27

subjective utility (of advanced English or of the hard as against the soft

advanced subjects), and failure expectation (in English or in the hard as

against the soft subjects). We also controlled for sex in our models, and

in the multinomial model also for the number of chosen subjects. Table 6

presets the parameter estimates of the odds of choosing advanced

English. Tables 7a and 7b present the parameter estimates of the

contrasts between hard versus soft subjects (Table 7a) and between hard

subjects and a mix of hard and soft subjects (Table 7b).10 To facilitate

comparison of the effects of the continuous independent variables we

standardized them to a mean of zero and a unit standard deviation.

Dummy variables retained their original metric.

In Model I in these tables, we regressed educational choice on social

background alone. In models II and III we added performance and

grouping to each model respectively. These two models were meant to

assess the extent to which performance and grouping mediate the effects

of social background. In model IV we regressed educational choice on the

three motivational factors alone, and model V included all variables

simultaneously.11

[TABLE 6 HERE]

Advanced Versus Basic English

Model I in Table 6 indicates that students with educated parents were

more likely to opt for advanced English. When we added performance to

this model (Model II) the effect of parental education was reduced, and

28

adding grouping in Math (Model III) rendered the effect of parental

education even weaker and statistically insignificant. High achieving

students and those who had been assigned to an advanced grouping in

math were more likely to choose English at advanced level. Importantly,

the choice between advanced or basic level English was largely

determined by performance and grouping, and no effects of social

background remained to be explained by secondary factors.

In Model IV the analysis focused on the gross effects of the three

secondary factors. The effects of relative risk aversion and utility

consideration were very small and statistically insignificant. By contrast,

students who hold low failure expectations in advanced English were more

likely to select advanced English. When, in Model V, we also controlled for

social background, performance and grouping in Math, the net effects of

failure expectation was slightly reduced but maintained its statistical

power (compared with Model IV).

In sum, the main determinants of choice between advanced and

basic English were scholastic performance, grouping in math, and failure

expectations in English, but not relative risk aversion or utility

perceptions. Moreover, although the odds of choosing advanced English

are affected by students' subjective failure expectations, they are largely

determined by previous performance and no effects of social background

remained to be explained by secondary factors.

Hard versus Soft Subjects

29

The analysis of choice between hard and soft electives yields similar

results to those obtained above. Model I in Table 7a indicates that males,

and students with educated parents, were more likely to opt for hard than

for soft subjects. When we added performance to this model (Model II)

the effect of parental education was reduced, and adding grouping (Model

III) rendered the effect of parental education even weaker and

statistically insignificant. Strong students and those who had been

assigned to an advanced grouping in math were more likely to choose

hard rather than soft subjects. Importantly, the choice between the hard

and soft clusters of subjects was largely determined by performance and

grouping, and no effects of social background remained to be explained

by secondary factors.

[TABLE 7a HERE]

In Model IV the analysis focused on the gross effects of the three

secondary factors. The relative risk aversion mechanism was very small

and statistically insignificant. By contrast, the effects of both relative

utility and relative failure expectation were significant, and in the

expected directions: students who expected relatively higher utility from

the hard subjects and hold low failure expectations in them were more

likely to select hard than soft subjects. When, in Model V, we also

controlled for social background, performance and grouping the net

effects of relative utility and relative failure expectation were slightly

reduced but maintained their statistical power (compared with Model IV).

30

Again, the coefficient for relative risk aversion in Model V was not

significant statistically.

Finally, we found that the odds of choosing hard rather than soft

subjects declined as the number of subject selections increased.

Evidently, students who choose the hard sciences are more likely than

those who choose soft subjects to concentrate on the relatively difficult

task at hand and avoid spreading their attention over several subjects.

Hard versus a Mix of Subjects

In Table 7b we present the effects of the independent variables on the log

odds of choosing hard rather than mixed subjects. These results, too, are

quite similar to those reported in Tables 6 and 7a. Boys were more likely

to choose hard subject than a mix of hard and soft ones; the

socioeconomic differences in these odds were fully accounted for by

performance and grouping; the effects of RRA were statistically

insignificant; those of perceived relative utility and relative risk of failure

were statistically significant and in the expected directions, and the

secondary factors did not mediate socio-economic differences in choice.

Finally, the odds of taking a mix of subjects increased substantially as the

number of selections increased.

[TABLE 7b HERE]

In sum, the main determinants of choice among hard subjects, soft

subjects and a mix of the two were scholastic performance, grouping in

31

math, relative failure expectations, and subjective utility, but not relative

risk aversion. Just as important, the three motivational factors posited by

BG's Rational Action Theory did not seem to mediate the effects of social

background.

Thus far, then, our results lead to the same conclusion; secondary

effects are absent with respect to educational inequality between social

strata in curricula choices within the Israeli secondary system.

Summary and Discussion

It is generally recognized that students from privileged social backgrounds

progress farther in the educational system than those from less advantageous

origins. Rational choice models of education, in particular the BG model,

attribute these inequalities, in part, to differences in educational choice made

by individuals from different social strata. According to the model, the

educational process consists of a sequence of branching points at which

students are faced with binary choices. On the one hand, the high road is

demanding and risky in the short term, but can lead to rewarding life-course

outcomes in the long term. On the other hand, the low road is less

demanding and safer in the short term, but can offer only limited long-term

rewards. That is, long-term utility and short-term risks are positively

associated. It is further assumed that on reaching educational decision points

students consider their success probability in each alternative, its utility,

defined as the probability that it will lead to desirable occupational outcomes,

and the extent to which it will protect them from downward social mobility.

32

The desire to avoid downward mobility gives rise to class differentials in

educational choice. Students from the upper strata are impelled to choose the

high road, which is necessary if they are to maintain their social position,

while those of lower strata choose the low road, which is sufficient to maintain

their social position. Finally, the BG model postulates that social reproduction

is mediated, in part, by differential educational choice.

In this paper we put these propositions to an empirical test.

Specifically, we examined the role of utility considerations, of failure

expectations in education, and of the class maintenance motivation in shaping

educational choice. We examined the validity of these assumptions by

analyzing recently collected data from a purpose-designed survey of Tel Aviv-

Jaffa high school students who in 2006 were about to select school subjects

at an advanced level for their matriculation examinations.

In the main, our results cast doubt on some of the main assumptions of

the BG model. Summarizing our results concerning each assumption in turn,

first we found that educational choice was affected by the relative utility

that students attribute to the alternative choice options. It was also

affected by students' subjective failure expectations on the alternative

educational routes. However, it was not affected by the so-called Relative

Risk Aversion motivation (i.e., the class maintenance motivation). These

results are consistent with those reported by Stocké (2007) for German

parents choosing secondary educational tracks for their children. Hence,

the role of mobility concerns at this stage of the educational career is

questionable.

33

Further, although affected by social origin (and two of the three also

hold effect on educational choice), secondary factors did not mediate

class inequalities in educational choice, which were largely mediated by

performance and prior grouping in Math. In addition, relative risk aversion

was negatively related to social origins: students from the lower strata

seemed more concerned about downward social mobility that those from

the more privileged strata. Although BG's predictions about the

determinants of educational choices were validated in part. We agree with

Stocké (2007) and with Nash (2003, 2006) that rational choice theory in

education focuses too heavily on secondary effects. Our findings indicate

that secondary effects on educational attainment are weaker compared

with primary effects. Research on the relative importance of secondary

effects on educational attainment indicates that the relative importance of

secondary effects depend, in part, on the specific educational transition in

question and on the societal context in question.12

Moreover, and importantly, a large proportion of our respondents took

neither the high road nor the low road, preferring a mixture of the two. We

suggest that students who choose both high-utility and low-risk options

hedge short-term risks with long-term utility. We find that hedgers came

disproportionately from less affluent and less educated families, had lower

scholastic achievements, and were more concerned with class

maintenance, than those who chose either pure hard subjects or pure soft

subjects. The possibility to choose a mix of subjects is not anticipated by

rational choice models of education.

34

This is an important finding, indicating that the model ignores

systemic differences in the structure of available choices. The availability

of multiple rather than binary options can affect educational choice and

class-based educational inequality by allowing students to hedge long-

term utility with short-term risk. Comparative research across educational

systems that offer students a variety of alternatives to choose from is in

order.

Our finding that the less privileged in society adopted a hedging

strategy also calls into question the argument that curricular choice

masks transparent tracks visible only to the more affluent parents (Lucas,

1999). We find that students from less affluent and less educated families

seemed to be aware of the risks and utilities associated with the different

subjects available to them, and they mixed them to their apparent

advantage. The availability of options need not work to the detriment of

greater equality of educational opportunity, as suggested by Lucas (also by

Ayalon, 2006).

Research is also called for on the relationship between students'

stated intentions and the courses they actually take. While actual course

taking is affected by students' preferences, it is also affected by the

availability of courses at schools, by teachers' recommendations, by schools'

policies of student selection, and by a host of other institutional constraints.

In a future study we hope to evaluate the role of subjective choices and of

hedging, as well as of push factors, on the actual courses students take.

35

Notes

1. When evaluating their available educational option, students (and their

parents) are also assumed to take into account the direct and indirect

costs of education. Although perceived cost is an important variable in

rational-choice theory it is not discussed or measured in this study

because high-school alternatives (i.e., choice of major) in Israel do not

differ in their expected costs.

2. Upper secondary education consists of two main tracks: academic

and vocational. Until recently the vocational track trained most

students for a trade and prepared them for the world of work rather

than for further study. In recent years, however, most vocational-

track students have been able to sit for matriculation examinations,

although their success rates are lower than those of their academic

counterparts (e.g., Ayalon and Shavit, 2004).

3. The number of units refers to the time devoted to the subject. A

unit equals one hour a week for three years, or three hours a week

for one year. The number of units at which a subject is taken

corresponds to its level and degree of difficulty (Ayalon, 1994).

4. Students' sampling probabilities were computed as the product of

the school's sampling probability (0.29 and 0.66 in the middle class

and working class neighborhoods, respectively) and the students'

sampling probability within the school, which was computed as the

ratio of the number of respondents and the number of tenth graders

(ninth and tenth graders in one of the schools). In the analysis, the

36

cases were weighted inversely to their sampling probabilities. In

unreported analyses we also controlled for sampling districts (north,

south) and obtained substantively identical results.

5. Some schools track students in math early-on in secondary school

and leave them little choice in the matter. In unreported analysis

we found that when compared to known distributions of students by

level of study in math, their stated choices are highly unrealistic.

Therefore, we do not analyze the choice of advanced mathematics.

6. The results presented in Table 2 for the hard and soft clusters are

similar to those obtained by inspecting the utility and failure

expectation for each of their composite subject separately, with one

exception: both the perceived utility and failure expectations for

Economics are similar to those for Biology. However, since the great

majority (96%) of students who intended to take economics also

chose other social science subjects its allocation to the social

science category seems justified.

7. Specifically, the wording in Van de Werfhorst and Hofstede's item e

was: 'I am concerned that I might reach a social class that is lower

than the social class of my parents', whereas in our Hebrew version

concerned was worded as afraid.

8. In unreported analyses we also controlled for number of siblings,

and students' reading habits, but, net of parental education and

economic resources, most of their effects did not reach statistical

37

significance. Dropping them from the analysis did not alter the

substance of our results in any way.

9. In a separate study that is currently in progress, we attempt to

account for gender differences in the motivational factors.

10. The effects on a third possible contrast (soft vs. mix) are not

reported and can be computed as the differences between the

respective effects in Tables 7a and 7b. Of all effects on the third

contrast, only that of relative failure expectation is significant and in

the expected direction. It does not mediate the effects of social

origins.

11. When models IV and V were fitted with each of the choice

mechanisms (RRA, RU and RFE) only, each effect was similar to the

effect reported here.

12. See, for example, the sharp difference in the estimated relative

importance of secondary effects in England and Wales (Jackson et

al., 2007) compared with Germany (Reimer and Schindler, 2007).

38

Acknowledgements

This study was supported by an Eshkol Fellowship from the Israeli Ministry of

Science and Technology, and by a fellowship from the Horowitz Institute at

Tel Aviv University to the first author, and by the Weinberg Chair of Social

Stratification and Inequality at Tel Aviv University. An earlier version was

presented in 2007 at the Annual meeting of the American Sociological

Association and at the Montreal meeting of RC28. We are thankful to those

who contributed important comments, especially to John Goldthorpe, Richard

Breen and Moshe Semyonov. Finally, we would also like to thank the two

anonymous referees for their useful suggestions and comments on previous

version of this manuscript.

39

REFERENCES

Apple, M. W. (1990). Ideology and Curriculum (2nd edition). New York:

Routledge.

Ayalon, H. (1994). Monopolizing Knowledge - the Ethnic-Composition and

Curriculum of Israeli High-Schools. Sociology of Education, 67, 264-

278.

Ayalon, H. (1995). Math as a Gatekeeper: Ethnic and Gender Inequality in

Course Taking of the Sciences in Israel. American Journal of

Education, 104, 34-56.

Ayalon, H. (2002). Mathematics and Sciences Course Taking among Arab

Students in Israel: A Case of Unexpected Gender Equality.

Educational Evaluation and Policy Analysis, 24, 63-80.

Ayalon, H. (2003). Women and Men Go to University: Mathematical

Background and Gender Differences in Choice of Field in Higher

Education. Sex Roles, 48, 277-290.

Ayalon, H. (2006). Nonhierarchical Curriculum Differentiation and

Inequality in Achievement: A Different Story or More of the Same?.

Teachers College Record, 108, 1186-1213.

Ayalon, H. and Shavit, Y. (2004). Educational Reforms and Inequality in

Israel: The MMI Hypothesis Revisited. Sociology of Education, 77,

103-120.

40

Ayalon, H. and Yogev, A. (1997). Students, Schools, and Enrollment in

Science and Humanity Courses in Israeli Secondary Education.

Educational Evaluation and Policy Analysis, 19, 339-353.

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 Inequalities: a Comparative Study

of Educational Attainment in Thirteen Countries. Boulder Colorado,

Westview Press, pp. 1-23.

Boudon, R. (1974). Education, Opportunity, and Social Inequality:

Changing Prospects in Western Society. New York: Wiley.

Bradley, K. (2000). The Incorporation of Women into Higher Education:

Paradoxical Outcomes?. Sociology of Education, 73, 1-18.

Breen, R.and Goldthorpe, J. H. (1997). Explaining Educational

Differentials - Towards a Formal Rational Action Theory. Rationality

and Society, 9, 275-305.

Breen, R. and Jonsson, J. O. (2000). Analyzing Educational Careers: A

Multinomial Transition Model. American Sociological Review, 65,

754-772.

Breen, R. and Yaish, M. (2006). Testing the Breen-Goldthorpe Model of

Educational Decision Making. In Stephen, L., Grusky, D. B. and

Fields, G. S (Eds), Mobility and Inequality: Frontiers of Research

from Sociology and Economics. Stanford: Stanford University Press,

pp. 323-358.

41

Burkam, D. T., Lee, V. E. and Smerdon, B. A. (1997). Gender and Science

Learning Early in High School: Subject Matter and Laboratory

Experience. American Educational Research Journal, 34, 297–331.

Davies, R., Heinesen, E. and Holm, H. (2002). The Relative Risk Aversion

Hypothesis. Journal of Population Economics, 15, 683-713.

De Graaf, N. D., De Graaf, P. M. and Kraaykamp, G. (2000). Parental

Cultural Capital and Educational Attainment in the Netherlands: A

Refinement of the Cultural Capital Perspective. Sociology of

Education,73, 92-11.

Duncan, G. J., Yeung, J. W., Brooks-Gunn, J. and Smith, J. R. (1998).

How Much Does Childhood Poverty Affect the Life Chances of

Children?. American Sociological Review, 63, 406-423.

Erikson, R. E., Goldthorpe, J. H., Jackson, M., Yaish, M. and Cox, D. R.

(2005). On Class Differentials in Educational Attainment.

Proceedings of the National Academy of Sciences of the United

States of America, 102, 9730-9733.

Erikson, R. E. and Jonsson, J. O. (1996). Can Education Be Equalized? The

Swedish Case in Comparative Perspective. Boulder, CO: Westview.

Gambetta, D. (1987). Were They Pushed or Did They Jump? Individual

Decision Mechanisms in Education. Cambridge: Cambridge

University Press.

42

Gamoran, A. and Mare, R. D. (1989). Secondary School Tracking and

Educational Inequality: Compensation, Reinforcement, or

Neutrality?. American Journal of Sociology, 94, 1146-1183.

Goldthorpe, J. H. (1996). Class Analysis and the Reorientation of Class

Theory: The Case of Persisting Differential in Educational

Attainment. British Journal of Sociology, 47, 481-505.

Goldthorpe, J. H. (1998). Rational Action Theory for Sociology. British

Journal of Sociology, 49, 167-192.

Halsey, A. H., Heath, A. F. and Ridge, J. M. (1980). Origins and

Destinations: Family, Class and Education in Modern Britain.

Oxford: Clarendon Press.

Hayman, F., Golan, H. and Shapira, R. (1997). School Autonomy and

Parental Choice: Steps in Local Educational Planning. In Shapira, R.

and Cookson, P.W. (Eds), Autonomy and Change in Context: An

International Perspective. Oxford: Pergamon, pp. 77-108.

Hedström, P. and Stern, C. (2008). Rational Choice and Sociology. In

Durlauf, S. and Blume, L., The New Palgrave Dictionary of

Economics (2nd edition). New York: Palgrave Macmillan.

Holm, A. and Jaeger, M. M. (2005). Relative Risk Aversion and Social

Reproduction in Intergenerational Educational Attainment:

Application of a Dynamic Discrete Choice Model. Paper presented at

the summer meeting of ISA-RC28 (Social Stratification and

Mobility), Los Angeles, August.

43

Jackson, M., Erikson, R. E., Goldthorpe, J. H. and Yaish, M. (2007).

Primary and Secondary Effects in Class Differentials in Educational

Attainment: the Transition to A-Level Courses in England and

Wales. Acta Sociologica, 50, 211-229.

Kamens, D. H. and Benavot, A. (1992). A Comparative and Historical

Analysis of Mathematics and Science Curricula, 1800-1986. In

Meyer, J.W., Kamens, D.H. and Benavot, A. (Eds), School

Knowledge for the Masses : World Models and National Primary

Curricular Categories in the Twentieth Century. Washington, DC:

Falmer Press, pp. 101-123.

Katz-Gerro, T. and Yaish, M. (2003). Higher Education: Is More Better?

Gender Differences in Labour Market Returns to Tertiary Education

in Israel. Oxford Review of Education, 30, 571-592.

Kliebard, H. M. (1992). The Decline of Humanistic Studies in the American

School Curriculum. In Kliebard, H.M. (Ed), Forging the American

Curriculum: Essays in Curriculum History and Theory. New York:

Routledge, pp. 3-26.

Lucas, S. R. (1999). Tracking Inequality: Stratification and Mobility in

American High Schools. New York: Teachers College Press.

Ma, X. and Willms, D. J. (1999). Dropping Out of Advanced Mathematics:

How Much Do Students and School Contribute to the Problem?.

Educational Evaluation and Policy Analysis, 21, 365-383.

44

Mare, R. D. (1981). Change and Stability in Educational Stratification.

American Sociological Review, 46, 72-87.

Ministry of Education, http://www.education.gov.il/netuney_bchinot/,

accessed on 26 December 2007.

Morgan, S. L. (1998). Adolescent Educational Expectations: Rationalized,

Fantasized, or Both?. Rationality and Society, 10, 131-62.

Morgan, S. L. (2005). On the Edge of Commitment: Educational

Attainment and Race in the United States. Stanford: Stanford

University Press.

Nash, R. (2003). Inequality Difference in Education: Is a Real Explanation

of Primary and Secondary Effects Possible?. British Journal of

Sociology, 54, 433-451.

Nash, R. (2006). Controlling for 'Ability': A Conceptual and Empirical

Study of Primary and Secondary Effects. British Journal of Sociology

of Education, 27, 157-172.

Need, A. and De Jong, U. (2000). Educational Differentials in the

Netherlands - Testing Rational Action Theory. Rationality and

Society, 13, 71-98.

Oakes, J. A. (1990).Opportunity, Achievement and Choice: Women and

Minority Students in Science and Mathematics. Review of Research

in Education, 16, 153–222.

45

Oakes, J. A., Gamoran, A. and Page, R.N. (1992). Curriculum

Differentiation: Opportunities, Outcomes and Meanings. In Jackson,

P.W. (Ed), Handbook of Research on Curriculum. New York:

Macmillan, pp. 570-608.

Plank, D. N. and Sykes,G. (2003). Choosing Choice: School Choice in

International Perspective. New York: Teachers College Press.

Reimer, D. and Schindler, S. (2007). Primary and Secondary Effects in

Access to Higher Education Germany. Conference of the European

Research Network EQUALSOC Network (Economic Change, Unequal

Life-Chances and Quality of Life), Dijon, France.

Shavit, Y. (1984). Tracking and Ethnicity in Israeli Secondary Education.

American Sociological Review, 49, 210-220.

Shavit, Y. (1990). Segregation, Tracking and the Educational Attainment

of Minorities: Arabs and Oriental Jews in Israel. American

Sociological Review, 55, 115-126.

Stocké, V. (2007). Explaining Educational Decision and Effects of Families’

Social Class Position: An Empirical Test of the Breen–Goldthorpe

Model of Educational Attainment. European Sociological Review, 23,

505-519.

Van de Werfhorst, H. G. and Hofstede, S. (2007). Cultural Capital or

Relative Risk Aversion? Two Mechanisms for Educational Inequality

Compared. The British Journal of Sociology, 58, 391-415.

46

Tel Aviv-Jaffa, Municipality. http://www.tel-aviv.gov.il/Hebrew/Education/

schools/Index.htm, accessed March 2006.

Tel Aviv-Yafo. 2007. Statistical Yearbook No. 46. Tel Aviv-Jaffa: Tel Aviv-

Yafo Municipality.

47

Authors' Address

Limor Gabay-Egozi (to whom correspondence should be addressed),

Department of Sociology and Anthropology, Tel Aviv University, Ramat

Aviv, Tel Aviv 69978, Israel. Email: gabaylim@post.tau.ac.il.

Yossi Shavit, Department of Sociology and Anthropology, Tel Aviv

University, Ramat Aviv, Tel Aviv 69978, Israel. Email: ys@post.tau.ac.il.

Meir Yaish, Department of Sociology and Anthropology, Haifa University,

Haifa 31905, Israel. Email: myaish@univ.haifa.ac.il.

48

Table 1 Percent distributions of subject choice by gender

Advanced Subjects Boys Girls Total

English 93 92 93

Physics 33 12 23

Chemistry 22 17 19

Computers 28 13 20

Biology 19 20 20

Economics 30 25 27

Social Science 30 52 41

History 16 15 16

Literature 12 14 13

Communications 14 24 19

Table 2 Means (and standard deviations) of perceived utility and perceived risk in advanced English, and in hard and soft advanced Electives

Advanced English

Hard Subjects

Soft Subjects

Perceived utility in university admission 4.44 4.09 3.50* (0.79) (0.78) (0.80) Perceived risk of failure 0.03 0.15 0.06* (0.16) (0.26) (0.15)

NOTE: Asterisks indicate significant (p<0.05) differences between the means of the hard and the soft clusters.

49

Table 3 Percent distribution of elective choice clusters

Choice categories %

Hard subjects only: 36 Physics, Chemistry, Computers and Biology

Soft subjects only: 42 Social Science, Economics, History, Literature and Communications

Mix of hard and soft subjects 22

Total % (N)

100 (563)

50

Table 4 Means (and standard errors) of key variables by choice categories

Variables

All (1)

English advanced

(2)

English basic (3)

Hard Subjects

(4)

Mixed

(5)

Soft Subjects

(6)

Social Background Sex (boys) 0.51 0.51 0.48 0.63 0.55* 0.39* (0.50) (0.50) (0.51) (0.49) (0.50) (0.49)

Parental Education 4.67 4.73* 3.86 5.10* 4.17* 4.57* (1.56) (1.53) (1.65) (1.31) (1.67) (1.60)

Economic Resources 6.17 6.24* 5.39 6.59* 5.39* 6.23 (2.21) (2.18) (2.23) (1.86) (2.23) (2.36)

Scholastic Performance & Grouping Scholastic Performance 82.81 83.44* 74.75 87.40* 78.91 80.90* (11.12) (10.95) (10.10) (9.35) (11.82) (10.82)

Grouping in Math (advanced level) 0.78 0.80* 0.58 0.94* 0.74 0.68* (0.41) (0.40) (0.50) (0.24) (0.44) (0.47)

Motivational Factors Relative Risk Aversion -0.10 -0.10 -0.11 -0.19* 0.18* -0.18 (1.00) (1.00) (1.03) (1.05) (0.95) (0.96)

English Utility 4.43 4.44 4.28 -- -- -- (0.79) (0.79) (0.79)

Relative Utility 1.20 -- -- 1.35* 1.14 1.11* (hard vs. Soft subjects) (0.28) (0.31) (0.23) (0.22)

English Failure Expectation 0.05 0.03* 0.34 -- -- -- (0.22) (0.16) (0.48)

Relative Failure Expectation 0.09 -- -- 0.01 0.07* 0.17* (hard vs. soft subjects) (0.23) (0.14) (0.21) (0.27)

Number of Chosen Subjects 1.94 -- -- 1.37* 2.95* 1.88* (0.99) (0.62) (1.09) (0.74)

Number of Cases 563 517 46 189 144 230 NOTE: Asterisks in column 2 indicate statistically significant (p<0.05) differences between the means

shown in columns 2 and 3. Asterisks in column 4 indicate statistically significant (p<0.05) differences between the means shown in columns 4 and 5. Asterisks in column 5 indicate statistically significant (p<0.05) differences between the means shown in columns 5 and 6. Asterisks in column 6 indicate statistically significant (p<0.05) differences between the means shown in columns 4 and 6.

51

Table 5 OLS regression coefficients (and standard errors) on performance and motivational factors

Independent Variables

Scholastic Performance

Relative

Risk Aversion

Relative Utility of Sciences

Relative Failure

Expectations in Sciences

Sex (Boys) -0.77 0.29* 0.07* -0.07* (0.83) (0.08) (0.02) (0.02)

Social Background Parental Education 1.68* -0.16* 0.02* -0.01 (0.30) (0.03) (0.01) (0.01)

Economic Resources 1.05* -0.08* 0.01 0.01 (0.21) (0.02) (0.01) (0.01)

Scholastic Performance Scholastic Performance 0.00 0.004* -0.003* (0.00) (0.00) (0.00)

Constant 68.89* 0.97* 0.66* 0.38*

(1.54) (0.29) (0.08) (0.07)

Adjusted R2 0.14 0.14 0.09 0.05

Significance level: *p<0.05

52

Table 6 The effects (and standard errors) of primary and motivational factors on the odds of choosing English at an advanced rather than basic level

Independent Variables

I II III IV V

Sex (boys) 0.14 0.23 0.17 0.18 0.15 (0.32) (0.32) (0.33) (0.34) (0.35)

Social Background

Parental education 0.38* 0.28~ 0.22 0.12 (0.16) (0.17) (0.17) (0.20)

Economic resources 0.25 0.10 0.07 0.08 (0.18) (0.19) (0.19) (0.20)

Scholastic Performance & Grouping

Scholastic Performance 0.54* 0.51* 0.32~ (0.16) (0.16) (0.17)

Grouping in Math (advanced level) 0.62~ 0.58 (0.35) (0.38)

Motivational Factors

Relative Risk Aversion 0.09 0.23 (0.17) (0.19)

Utility in English 0.04 0.07 (0.16) (0.16)

Failure Expectation in English -2.90* -2.22* (0.42) (0.48)

Constant 2.61* 2.58* 2.16* 2.87* 2.47*

(0.23) (0.23) (0.32) (0.25) (0.36)

Cox and Snell R2 0.022 0.041 0.045 0.069 0.083

Significance levels: *p<0.05 ~ 0.05<p<0.09

53

Table 7a The effects (and standard errors) of primary and motivational factors on the odds of choosing hard rather than soft clusters of electives

Independent Variables

I II III IV V

Sex (boys) 1.23* 1.34* 1.31* 1.02* 1.14* (0.21) (0.22) (0.23) (0.24) (0.26)

Social Background

Parental education 0.39* 0.24* 0.09 0.12 (0.12) (0.13) (0.13) (0.15)

Economic resources -0.07 -0.20 -0.19 -0.22 (0.12) (0.13) (0.13) (0.15)

Scholastic Performance & Grouping

Scholastic Performance 0.82* 0.76* 0.55* (0.14) (0.15) (0.16)

Grouping in Math (advanced level) 1.91* 1.55* (0.34) (0.40)

Motivational Factors

Relative Risk Aversion 0.02 0.14 (0.12) (0.13)

Relative Utility 1.19* 1.05* (hard vs. soft subjects) (0.16) (0.16)

Relative Failure Expectation -1.23* -1.06* (hard vs. soft subjects) (0.18) (0.19)

Number of Subjects Chosen -1.41* -1.41* -1.55* -1.70* -1.84* (0.19) (0.19) (0.21) (0.22) (0.24)

Constant -1.39* -1.55* -3.15* -1.59* -3.07* (0.18) (0.19) (0.37) (0.22) (0.43)

Cox and Snell R2 0.386 0.425 0.460 0.507 0.540

Significance level: *p<0.05

54

Table 7b The effects (and standard errors) of primary and motivational factors on the odds of choosing hard rather than mixed clusters of electives

Independent Variables

I II III IV V

Sex (boys) 0.81* 0.92* 0.92* 0.74* 0.87* (0.28) (0.29) (0.30) (0.30) (0.31)

Social Background

Parental education 0.35* 0.21 0.13 0.17 (0.16) (0.16) (0.16) (0.18)

Economic resources 0.10 -0.03 0.01 -0.05 (0.16) (0.16) (0.17) (0.18)

Scholastic Performance & Grouping

Scholastic Performance 0.81* 0.77* 0.60* (0.17) (0.17) (0.19)

Grouping in Math (advanced level) 1.31* 1.11* (0.43) (0.47)

Motivational Factors

Relative Risk Aversion -0.14 0.03 (0.15) (0.17)

Relative Utility 0.92* 0.77* (hard vs. soft subjects) (0.18) (0.18)

Relative Failure Expectation -0.84* -0.73* (hard vs. soft subjects) (0.20) (0.22)

Number of Subjects Chosen -2.63* -2.62* -2.76* -2.93* -3.03* (0.23) (0.23) (0.24) (0.25) (0.27) Constant -0.18* -0.35 -1.51* -0.55* -1.68* (0.23) (0.24) (0.45) (0.26) (0.50)

Significance level: *p<0.05