Educational Stratification in Russia During the Soviet Period

by Theodore P. Gerber, Michael Hout
Citation
Title:
Educational Stratification in Russia During the Soviet Period
Author:
Theodore P. Gerber, Michael Hout
Year: 
1995
Publication: 
The American Journal of Sociology
Volume: 
101
Issue: 
3
Start Page: 
611
End Page: 
660
Publisher: 
Language: 
English
URL: 
Select license: 
Select License
DOI: 
PMID: 
ISSN: 
Abstract:

Educational Stratification in Russia during the Soviet period1

Theodore P. Gerber

University of Oregon

Michael Hout

University of California, Berkeley

A national survey of educational stratification in Russia reveals substantial inequality of educational attainments throughout the Soviet period. Parents' education, main earner's occupation, and geographical origin contributed to these inequalities. Gender prefer- ences for men were removed, and for some transitions reversed. Although secondary education grew rapidly, higher education failed to keep pace. This disparity led to a university-level enrollment squeeze, and the resulting bottleneck hurt disadvantaged classes more than advantaged ones. In turn the effect of social origins on entering university increased after 1965. The upshot was no net change in the origin-based differences in the likelihood of attaining a VUZ degree across three postwar cohorts.

INTRODUCTION

The expansion of educational institutions throughout the 20th century has resulted in some striking similarities across Western industrialized societies (Blossfeld and Shavit 1993): (1) Enrollments (relative to cohort size) have increased dramatically, particularly at the lower secondary level (now compulsory in most industrial nations). (2) The association between social origins and years of schooling has fallen in many countries while the effects of origin on specific transitions have remained constant.

This is a revised version of a paper we first presented at the 13th World Congress of Sociology, Bielefeld, Germany, July 1994. We thank Michael Burawoy, Daniel Dohan, Martin Sanchez-Jankowski, and Yossi Shavit for useful discussions of the issues in this paper. Funding for this research has been provided by the National Science Foundation (SES 9209792), the American Council of Learned Societies, the University of California, Berkeley, Survey Research Center, and a Social Science Research Council Graduate Training Fellowship to the first author. Data collection was supported by grants from the John D. and Catherine T. MacArthur Foundation and the Spencer Foundation, with additional support from the National Science Foun- dation (SES 8822628). Address correspondence to Theodore P. Gerber, Department of Sociology, University of Oregon, 736 PLC, Eugene, Oregon 97403-1291.

O 1995 by The University of Chicago. All rights reserved.

0002-9602196110103-0003$01.50

AJS Volume 101 Number 3 (November 1995): 611-60 611 (3) Social origins affect completion of lower levels of the educational system (primary and lower secondary) more than they affect entry to or completion of higher levels. (4) Males' advantage over females in educa- tion has gradually disappeared; in some cases it has reversed. These patterns appear to transcend differences in socioeconomic and political systems, as they also hold for the formerly state socialist countries of Poland (Heyns and Bialecki 1993) and, with minor modifications, the former Czechoslovakia (Mateju 1993) and Hungary (Szelenyi and Aschaf- fenburg 1993).' Before closing the book on educational stratification un- der state socialism, however, we must consider the case of the Soviet Union. The Soviet Union was, after all, the leading Communist state. Furthermore, state socialism had a longer life there and was homegrown rather than imposed by a foreign occupying power. Thus, we ought to guard against generalizing to the Soviet Union itself the historical pat- terns found in its satellite states.

In the 1950s, Western sociologists took for granted the importance of education as a determinant of individuals' life chances in the Soviet Union. Inkeles and Bauer (1959) argued that the role of education in the Soviet stratification process was greater than in market societies because of the absence of self-employment as an alternative channel for upward mobility. More recently scholars have questioned the importance of edu- cation because the absence of "market discipline" eroded the returns to education (Connor 1991). Without economic returns why would young people invest in human capital by pursuing advanced education?

The Soviet regime for its part embraced an ideology of equal access, but pursued it only fitfully. A mix of strong and weak policies attempted to reduce inequalities in the educational attainments of the offspring of diverse classes. Some Soviet scholars have claimed a certain "equalizing and drawing together" (sblizhenie) of social classes with respect to educa- tional level (Vasil'eva et al. 1985). Other Soviet sociologists began in the 1960s to conduct highly localized studies of educational aspirations and attainments (e.g., Rutkevich and Fillipov 1978). They found substantial differences among groups of different social origin. Western commenta- tors familiar with such Soviet studies have concluded with certainty that some association between social origins and attainments, particularly with respect to attaining a college education, existed in Soviet society throughout the postwar era (Yanowitch 1977; Dobson 1977, 1980; Con- nor 1991). But the complete absence of educational stratification is too strict a test. No society could pass it (Blossfeld and Shavit 1993).

In Czechoslovakia the effect of father's education-but not father's occupation- first declined after the transition to Communism in 1948, but then returned to and exceeded its initial level. In Hungary the expansion of enrollments relative to cohort size at the low and high secondary levels was preceded by a contraction.

How unequal was access to higher education in the Soviet Union? The precise magnitudes of key parameters remain unclear, at least in part because studies are limited by their inadequate measures or by the limited scope of their samples. Too many studies address aspirations rather than achievements. None address all of the important educational transitions within an integrated framework. Change over time cannot be assessed; even the direction of change is in doubt. Soviet studies are also too localized-often confining themselves to single towns or even single en- terprises-to allow for cumulation of their separate results. Most employ overly broad social origin categories-such as the official tripartite Soviet class schema of "workers," "collective farmers," and "employees"- that may mask associations and make international comparisons impossi- ble. They present data in simple bivariate cross-tabulations, often with- out providing marginal distributions. Multivariate analysis and tests for statistical significance are virtually ~nknown.~

Cross-cohort comparisons with standardized indicators are available in only one study, and it is based on unreliable official statistics (Vasil'eva 1978).~

The Russian component of the Comparative Class Structure and Con- sciousness Project (Wright 1986; Hout, Wright, and Sanchez-Jankowski 1992) contains enough detail on education and social origin to estimate the effects pertaining to six Soviet cohorts. We follow the protocol of the studies collected by Blossfeld and Shavit (1993) and focus on several social origin variables and gender as the key indicators of equal access and opportunity. We estimate their effects on the probabilities of success- fully completing four successive educational transitions using a model proposed by Mare (1980) and now in wide use for comparative research. As background, we describe the Soviet education system and enrollment trends in Russia during the Soviet period, review previous findings on inequalities in Soviet educational attainments in more detail, and discuss various policies taken by Soviet leaders to address the historical inequali- ties and those that arose as a consequence of Soviet-era policies. Our results shed light on hitherto unexaminable aspects of educational strati- fication in Russia during the Soviet period and permit the addition of the Russian case to the comparative analysis of educational inequality.

There are some exceptions to this general depiction of Soviet studies of educational stratification. Rutkevich and Fillipov (1978) sample from six areas and regions. Dob- son (1977) cites two studies that provide correlation coefficients between parent's and children's educational level. However, these exceptions fail on other grounds to serve as adequate bases for specifying the magnitude or trend of origin effects on attain- ments.

The Harvard Project survey of Soviet citizens displaced by World War I1 (Inkeles and Bauer 1959) and the Soviet Interview Project survey of emigres in the late 1970s (Millar 1987) are not suitable due to the self-selected nature of the samples and the correlation between education and emigration.

Russia during the Soviet period is an interesting test case for the maxi- mally maintained inequality (MMI) framework proposed by Raftery and Hout (1993). The persistence of educational stratification across cohorts of Irish and British students suggested a generalization about inequality. Raftery and Hout inferred the general principle that privileged classes have a greater interest in the advancement of their own offspring than in blocking the advancement of students from other classes, this work proposes four specific hypotheses. The key proposition in MMI is that transition rates and odds ratios between social origin and educational transitions remain the same from cohort to cohort unless forced to change by increasing enrollments. More specifically:

  1. All else being equal, growth in the capacity of secondary and higher education will reflect the increased demand occasioned by population growth (if any) and the gradual upgrading of social origins over time (if any). In this case, origin-specific transition rates remain the same over time. This was the case for the two oldest Irish cohorts and also for all the English cohorts.
  2. If for some reason educational opportunity expands faster than the demand that is attributable to upgrading social origins rises, then transi- tion rates for all social origins increase, but the increase occurs in such a way as to preserve all the transition X class odds ratios. This happened between the two youngest Irish cohorts.
  3. If demand for a given level of education is saturated for the upper classes-that is, if some origin-specific transition rates approach or reach 100%-then the odds ratios decrease (i.e., the association between social origin and education is weakened). This diminished inequality of oppor- tunity occurs only if the expansion in enrollment cannot be accommo- dated in any other way. This was the case for entry to the second level between the second and third Irish cohorts.
  4. An equalization can be reversed. For example, if enrollment in academic secondary education had been more widespread among work- ing class youths in Ireland, the falling conditional probabilities of success- ful transition to higher education in the 1956 cohort might have led to an increase in the effect of origin on entry into higher education.

MMI might emerge from the rational choices of parents and students. Parents in advantaged circumstances and students from advantaged backgrounds are more interested in furthering their own goals and aspira- tions than they are in blocking the advance of lower-class students. In other words, MMI assumes that their interest is in the substantive goal of maximizing their own education, not in the derivative goal of main- taining class differentials.

As we shall show, Russian education modernized during the Soviet period. But that modernization was very uneven. Many reforms were tried, often with results that were not intended. Most important, the reforms at lower levels put pressures on upper levels of the educational system. Rapid expansion of secondary education initially lowered rates of completing academic secondary education (called "general" secondary in Russia). Then the expansion of general secondary education brought so many students to the doors of the universities that matriculation rates for student with academic diplomas decreased. According to MMI, we ought to expect educational stratification in Russia to increase during the Soviet period in response to these enrollment pressures. As the necessity to select intensifies, MMI predicts increased effects of social origin (the complement to the "saturation" argument).

Djilas (1965) and the "new class" theorists who have followed pre- dicted growing stratification in state socialist systems for different rea- sons. They did not anticipate enrollment crunches; they focused on the temptations and corruptions of power. In a totalitarian society, the moti- vation to aid one's offspring comes together with the means, and, ac- cording to Djilas, the temptation is irresistible. Hungarian education restratified in the second generation of Communism after an initial de- stratification (Simkus and Andorka 1982; Szelenyi and Aschaffenberg 1993). In Czechoslovakia, party membership was independent of occupa- tion, and the effects of occupation on education were mediated by politi- cal tests (Mateju 1993).

It is not possible to compare the first generation of Russian students to complete its schooling under Soviet power with previous generations because there are far too few of either group in our sample. We do, however, have enough information to compare six Soviet cohorts, begin- ning with those educated before the Second World War. We have speci- fied cohorts and a division of the educational process into transitions in such a way that allows us to separate the pressures for stratification that stem from the exercise of elite power from the pressures generated by institutional change.

THE SOVIET EDUCATIONAL SYSTEM

History and Structure

In the early years following the Russian Revolution in 1917, the Bolsheviks set about nationalizing all schooling. A campaign against illiteracy spurred rapid growth in enrollments. Experimentation in curriculum, pedagogy, and administration alternated with traditionalist backlashes (Matthews 1983) until the mid-1930s. Radical policies reached a head during Stalin's "cultural revolution" (roughly 1928-3 1). The two higher grades of second- ary education (then the eighth and ninth) were abolished in favor of univer- sal vocational training. A variety of "workers' faculties" and apprentice- ship schools earmarked to raise the educational qualifications of workers and peasants proliferated. Higher educational institutions, particularly those of a technical character, rapidly recruited large numbers of adult workers (Fitzpatrick 1979). The vacillating policies brought considerable disarray which, in turn, brought experimentation to a close. By 1934 tradi- tional educational philosophies were once again established. General sec- ondary education was reinstated. Universal textbooks were mandated. Wholly standardized, traditional pedagogical techniques, curricula, grades, and rigid examinations were reinstituted. University recruiters dis- continued policies that favored applicants of working-class origin.

The highly centralized, standardized, and regimented system crystal- lized by 1934 (Matthews 1972); it remained at least through the 1970s. The specific ministry in charge of education shifted from time to time, but cen- tral government organs retained direct responsibility for all aspects of its administration-especially staffing, curriculum details, and procedures for admitting students (Matthews 1983). The Communist Party wielded influence over educational policy from within these governmental organs and via declarations of policy goals by its Central Committee, often follow- ing Party Congresses (Prokof 'ev 1985). The basic course of general instruc- tion began at age 7, lasted 10 years,' and was divided into three stages:

Primary.-This initial stage comprises classes 1-4 until 1970, 1-3 thereafter.

Lower secondary.-This stage (nepol'noe srednee-literally "incomplete secondary") lasted through the seventh grade until 1958, the eighth grade thereafter.

Secondary.-This stage was composed of the remaining grades.

General educational schools came in part- and full-time varieties, some offering only primary classes, some primary and lower secondary, and some all three levels. In time, schools offering all three levels predomi- nated.

Parallel to and extending somewhat beyond the general secondary track was a vocational track, which was composed of two types of institu- tions. Vocational schools designed to teach manual skills to those with lower secondary education or less underwent a series of reorganizations and name changes before emerging as "professional-technical schools" (or PTUs by their Russian acronym) in 1959. Initially, these lower-grade vocational schools did not offer a general secondary curriculum and di-

Beginning in the early 1980s, increasing numbers of six-year-olds were enrolled in "grade zero" preparatory programs in general schools (Dunstan 1983). The standard 10-year general school program was supplemented by an additional eleventh year in some republics-notably the Baltics-where Russian language training increased the curriculum burden and required additional instruction time.

ploma, though tentatively in the mid-1950s and in increasing numbers during the 1970s SPTUs, schools combining vocational training with a secondary diploma, were established (Connor 1991). The other, higher- grade component of the vocational track was the "specialized secondary school" (also known as the tekhnikum), which trained lower-level profes- sionals such as technicians, nurses, librarians, and elementary school teachers. Some programs in such schools were available only to graduates of general secondary schools, others were designed for students entering with only lower secondary education.

Higher education took place in universities and more specialized "insti- tutes," all of which are lumped under the catch-all acronym "VUZ" (vysshee uchebnoe zavedenie, or "higher educational establishment"). It is important to note that following the rollback of the radical policies of the cultural revolution, admission to VUZs was highly competitive-based largely on entrance exams, though grades were also considered beginning in 1972 (Dobson 1977). In practice admission to a VUZ became possible only upon completion of the tenth grade of a general secondary school. In 1944 a secondary school diploma became a formal requirement for admis- sion (thus excluding all PTU graduates but the few who could win exemp- tions). Graduates of specialized secondary schools were required to work for at least three years before continuing their education, and graduates of SPTUs usually did not have adequate preparation for competitive entrance exams (Rutkevich and Fillipov 1978; Avis 1983). Thus, a crucial selection point in determining differential access to higher education occurred upon completion of lower secondary school, when students could proceed to gen- eral secondary school-which had no admission requirements other than completion of lower secondary (Yanowitch 1977)-or enter the vocational track or the workforce, either of which effectively precluded eventual VUZ entry. As one Soviet researcher concludes, noting that over 95% of VUZ entrants were general secondary graduates, the social composition of VUZ students is highly conditioned by the system of pre-VUZ education (Samoilova 1978, p. 113).~

Enrollments

Enrollments in all types of institutions grew rapidly during the Soviet period (fig. 1). Expansion at the primary level was greatest before the

In addition to the types of institutions discussed here, the system also included various types of special schools for the exceptionally gifted, mentally deficient, and for the ad- vanced education of certain elite Party members and military personnel (see Matthews 1982, chap. 6). Some such institutions-notably "special" high schools for the gifted- have been attacked on egalitarian grounds (Rutkevich and Fillipov 1978). However, since our data do not give any information on participation in such institutions-which in any event seem of marginal quantitative significance-we ignore them.

A. Primary

Grades 1-8

""

"Y

C) ;\I'

c
/ /

Q)

Op 15

3

Grades 9-10

B. Secondary and Higher

FIG.1.-Official enrollment data, Soviet Union, 1914-1985. Enrollment num- bers refer to academic year beginning in year listed. Pre-1939 numbers apply to pre-1939 Soviet borders (i.e., Baltic states are not included). Data for 1914 include all types of schools, not just day schools. Vocational school figures for 1932 and 1937 are from Fitzpatrick (1979, p. 238) and pertain, respectively, to 1931 and 1938. Other sources are TsSU (1977, pp. 7, 68, 145, 153, 213) and GKSPS (1989, pp. 48, 49, 148, 159, 193).

TABLE 1

YEAR^
1926-27 1939-40 1959-60 1970 1984-85
1. Total enrollments ............ 12.0 36.9 38.8 56.4 54.4
2. Population ..................... 147.0 194.1 208.8 241.7 276.3
3. Population 6-24 years          
old ............................ NA 72.9 70.6 84.1 84.0
4. Employed with at          
least lower secondary          
certificate (%) ............... NA 12.3 43.3 65.3 86.8

SOURCES.-Fitzpatrick (1979); GKSPS (1989); TsSU (1966, p. 8; 1977; 1985, p. 29); Gel'fand (1992, pp. 37, 94, 126, 168, 171).

* Statistics are in millions unless otherwise indicated.

t The first year in each pair pertains to rows 2 and 4, the second to rows 1 and 3. The figure for 1939-40 in row 3 applies to January 1, 1941.

Second World War; at higher levels it peaked following the war, espe- cially during the 1960s. This expansion kept ahead of the Soviet popula- tion (table 1). Between 1927 and 1985, enrollments in the five types of institutions listed in figure 1 grew by 453%, far outstripping the popula- tion growth of 188% that occurred at the same time. More important, the number of enrollments grew by 5.1% between 1940 and 1960 and then by an additional 45.4% by 1970, while the school-age population contracted by 3.2% during the first period, and grew by only 19.1% during the second. Thus, the somewhat low enrollment figure for 1960 reflects the tail end of the demographic "war dent" (the dip in population caused by the low number of births during World War 11; see Matthews

1982) as well as the impact of Khrushchev's education policies (discussed below). The proportion of the employed population entitled to claim at least a lower secondary education grew steadily following the Second World War. Precise data on coverage at various levels are scarce, but it would seem that universal primary education was declared a goal in 1928 and achieved by 1933 (Prokof'ev 1985, pp. 69-70). Compulsory lower secondary was officially declared in 1949 (Inkeles and Bauer 1959, p. 130), reaffirmed on the new eight-year basis in 1958, achieved in urban areas by 1962 (Matthews 1972, p. 260), and practically achieved in the Russian Federation by the mid-1970s, with 95% of those who entered the first grade in 1965 completing the eighth in 1974 (Rutkevich and Fillipov 1978, p. 74).

Trends in complete secondary school and VUZ coverage (fig. 2) are more complex and extremely important for our discussion. Beginning in

FIG.2.-Enrollment ratios based on official data and census counts, Soviet Union, 1940-85. Estimates (based on official fertility, mortality, and census data) of annual number of 17-year-olds on the first day of each year are provided by Gel'fand (1992). Annual enrollment figures are taken from TsSU (1977, pp. 93, 246; 1981, pp. 458, 468; 1983, p. 468; 1985, pp. 515, 525).

1945 enrollments in both grew rapidly, and the ratios of general second- ary school leavers to 17-year-olds and of VUZ entrants to 17-year-olds both rose; but from 1952 to 1958 the former rose at a more pitched rate, while the latter slightly declined.' This trend was briefly reversed starting in 1959, as the war dent's decline in 17-year-olds was not enough to override a decline in secondary school enrollments occasioned by Khru- shchev's reforms, while VUZ enrollments grew at a newly invigorated pace from 1957 to 1966. Khrushchev fell from power in 1964 and his policies were quickly abandoned; in the mid-1960s universal general sec- ondary education was proclaimed a goal (Yanowitch 1977, p. 73), eventu- ally enshrined in the Soviet constitution (Prokof'ev 1985, p. 82), and officially declared "achieved" in the late 1970s (Rutkevich and Fillipov 1978, p. 55; Vasil'eva et al. 1985, p. 158). As a result, general secondary graduation rates resumed their pitched growth. The policy of accommo- dating all students from primary through general secondary in one build- ing meant that the expansion of secondary enrollments could be accom- plished with a minimum of new construction. Universities could not expand as easily, nor was there the political mandate to do so. Conse- quently, VUZ openings could not keep pace with the size of the ensuing "baby boom" cohorts and higher secondary completion rates, causing

'The ratios displayed in fig. 2 are only rough indicators of the probability that a 17-year-old in a given cohort will complete general secondary school or enter a VUZ. Clearly, some accomplished either or both at ages other than 17, especially in the case of VUZ entry in the years following the war.

TABLE 2

% Planning to % That Actually Did
Parent's Occupation* Study Full-Time Study Full-Time
Urban intelligentsia ...................................    
Rural intelligentsia ...................................  
Workers: manufacturing, construction ...........  
Workers: communication, transportation .......  
Workers: agriculture ..................................  
Service workers .......................................  
Others ....................................................  
Total ................................. ............  

SOURCE.-Shubkin (1965, p. 65, table 3). Sample size is unclear.

* Based on father's occupation with substitution of mother's if father is not present

VUZ enrollment rates to fall off. These patterns are captured by the curve in figure 2 that represents the annual ratios of VUZ entrants to general secondary leavers, which, aside from the period 1959 to 1963, persistently declines-first abruptly, then more grad~ally.~

These enrollment trends imply that as coverage at the lower secondary level rapidly increased following the Second World War, the transition from lower to general secondary became more selective. From 1945 until 1953 annual VUZ intakes exceeded annual general secondary gradua- tions, meaning not much selection was occurring at this transition. How- ever, with the growth in general secondary enrollments, particularly from the mid-1960s onward, it was this transition-from general secondary to VUZ-that became the crucial selection point.

Origins and Attainments-Association, Mechanisms, Trends

From 1938 until the mid-1960s, no data on the social composition of VUZ students or the differential educational attainments of different ori- gin groups were published (Matthews 1972). Soviet studies revealing such origin-specific differences began appearing in the early 1960s, and came quickly to constitute a sizable literature that has been thoroughly summa-

'Part of the campaign for universal secondary education involved bolstering the number of SPTUs-vocational schools offering a secondary education along with vocational training. Enrollments in SPTUs-whose numbers increased from 180,000 in 1970 to 1,216,000 in 1975 to 2,168,000 in 1980 (TsSU 1982, p. 503)-are included in the vocational category in fig. 1. SPTU graduates are not included in the figures for general secondary leavers used to calculate the ratios in fig. 2.

TABLE 3

PARENT'S OCCUPATION* Ninth Grade Tekhnikum PTUIWork Nonmanual, requiring higher

education

.....................

........

........

86 11 3

Nonmanual, requiring specialized
secondary ...................................... 70 15 15
Skilled workers ................................... 52 2 7 2 1
Unskilled workers and nonmanual      
with no educational requirements .......... 25 25 50

SOURCE.-Vasil'eva (1973, p. 41, table 16). Sample size is unclear.

* Based on "higher" of two parents' occupations when they differ.

TABLE 4

SOVIETFINDINGS ATTAINMENT IN SIX REGIONS, EIGHTH

ON EDUCATIONAL AND TENTHGRADERS,1973

% Eighth Graders Planning % Tenth Graders Father's Education General Secondary Study Planning VUZ Study

Less than primary ................ 30.7 18.3
Primary .......................... 35.1 26.2
Lower secondary .................. 44.1 39.2
General secondary ................ 60.5 61.8
Specialized secondary ............ 50.9 55.8
Higher education ................. 78.7 76.9

SOURCE.-Rutkevich and Fillipov (1978, p. 94, table 21). N = 9,800 eighth graders; 9,278 tenth graders.

rized and analyzed in the West (see, e.g., Yanowitch 1977; Dobson 1977, 1980; Connor 1991). Here we present a sampling, in tables 2-5, to give an impression of the character and magnitude of inequalities discerned in Soviet research. These figures show that, in a variety of locations and at a variety of times, children whose parents had higher levels of educa- tion, higher status specialist jobs, or both were more likely than those whose parents had lower levels of education, low status jobs, or both to aspire to general secondary or VUZ education and to actually attain these levels. Aside from the educational and occupational dimensions of origins, the geographical dimension also receives significant attention in these studies, with urban-dwelling children having the advantage over

TABLE 5

% With Full General Father's Occupation Secondary Education

Agricultural manual ................................... 29.3
Unskilled manual ...................................... 32.1
Semiskilled manual .................................... 36.2
Skilled manual .......................................... 54.0

Unskilled nonmanual .......................

......

. 43.4

Nonmanual with higher education ................. 69.4

SOURCE.-Shubkin (1984, p. 65, table 15). N is unclear but the sample is described as one-fourth of employed 16-30-year-olds.

rural dwellers in aspirations and achievements (Samoilova 1978; Rut- kevich and Fillipov 1978; Vasil'eva 1978). Some evidence suggests ad- vantages accrue to those from higher income groups and those from dual-parent families (Rutkevich and Fillipov 1978). The educational aspirations and attainments of young men and women began to equalize by the early 1970s (Vasil'eva 1978; Avis 1983).

Given the absence (except for the period 1940-56) of entrance fees for general secondary and VUZ study and the meritocratic nature of VUZ selection, the selection mechanisms generating these origin-based inequalities can be reduced to related differences in aspirations and performance. The role of parental encouragement in shaping aspira- tions is often noted (Shubkin 1965; Samoilova 1978; Zaslavsky 1982), and students from lower social origins may be "discouraged by their poor academic performance, by their family's urgent need for addi- tional breadwinners, or by the trauma associated with the high proba- bility of failure in the intense competition" (Yanowitch 1977, p. 81). Regarding performance, apart from origin-specific differences in abil- ity-on which no data appears to exist-certain factors have been noted to enhance the performance of students from higher social origins: they are more likely to receive preschool preparation; their home environ- ment is characterized by a higher cultural level and is more conducive to study; their parents take greater interest in and more frequently moni- tor their work; they are less likely to be required to work to support the family; and they are more likely to have access to tutors and test-preparatory courses (Vasil'eva 1973; Dobson 1977, 1980; Rutkevich and Fillipov 1978 ). A key element in the urbanlrural distinction appears to be the inferior quality of primary and secondary training in rural areas (Zaslavsky 1982).

Some commentators optimistically equated access with equality. They claimed origin-based inequalities were declining and would continue to decline over time due to the spread of universal secondary education and the equalizing effects of economic growth (Rutkevich 1977; Rutkevich and Fillipov 1978; Dobson 1977). They assumed that economic growth would level the objective conditions of different social origin groups. Others, recognizing the increased selectivity at the point of VUZ entry due to the enrollment trends discussed above, claimed that such competi- tive conditions-which were only exacerbated in the mid-1960s by the postwar baby boom-would heighten the advantages of groups from higher social origins (Shubkin 1965; Avis 1983). This would imply in- creasing inheritance of high and low origins alike, because higher educa- tion is the main mechanism for the recruitment to high-status intelligen- tsia positions. In this vein, some Western commentators suggested that the class structure of Soviet society began to "crystallize" in the 1970s and predicted political instability due to the social malaise generated by frustrated mobility aspirations (Lapidus 1983; Connor 1991).

Claims about rising inequality and its supposed political consequences are based more on suppositions regarding the impact of the increased competition at the VUZ entry point than on actual data. It is possible that increased selectivity at that point was counterbalanced by decreased selection at the general secondary level, resulting in little or no change in the overall origin-specific probabilities of attaining higher education. Stratification might simply be delayed in parallel with the delay in selec- tion. Moreover, official data cited by Rutkevich and Fillipov (1978, p. 110) portray an increase in the representation of workers among VUZ students from 37.1% in 1969-70 to 45.5% in 1975-76 and a correspond- ing decrease in the representation of nonmanual employees from 54.4% to 47.0%. The reliability of such figures is highly questionable, however, given the inconsistent classification procedures of official VUZ statistics (students who held a job prior to VUZ entry are classified according to the nature of that job; the rest are classified according to parent's occupa- tion). All in all, it must be concluded that the true extent and direction of postwar trends in the association between social origins and educational attainments, as well as the precise magnitude of the net effects of different dimensions of origins, remain unclear.

Policies, Reforms

The rapid postwar growth of secondary education spawned hitherto un- heard of social problems-including youth unemployment and the spread of "overeducated, undertrained" (Connor 1991) graduates. The Khru- shchev regime (1956-64) undertook the first attempt since the early 1930s to reform secondary and higher education. Khrushchev's reforms sought to redress the disadvantages of students from lower social origins and to cope with the social problem. In 1956 tuition was abolished, polytechnical curricula were promoted in general secondary schools, and part-time VUZ study was assigned greater emphasis. Admission rules at VUZ were gradu- ally changed to encourage greater matriculation by students from lower social origins. To accomplish this, the Khrushchev reforms established generous quotas for applicants with two or more years of "production expe- rience" (Matthews 1982, p. 154). More sweeping measures were intro- duced in 1958. Eight rather than 10 years of schooling became the official norm for Soviet young people; the last two classes of general secondary school were extended to three, while the general curriculum was changed to include mandatory training in practical skills and work internships at enterprises or farms; quotas favoring production candidates for VUZ entry were increased (reaching 80% in some institutions); evening and correspon- dence courses were rapidly expanded, accounting for 79% of the 1.4 million increase in VUZ enrollments between 1958 and 1964; and (in 1959) state enterprises and collective farms were given the right to sponsor for VUZ admission individual employees, who were thus able to forego entrance exams (Matthews 1972, chaps. 9, 10; 1982).

Though hard evidence is not available, Khrushchev's reforms appar- ently failed on most counts, though they did temporarily reverse the relative growth trends of higher and secondary institutions (see fig. 2). More generally, they produced higher dropout rates, declining VUZ stan- dards, and much resentment, especially among faculty who had to work more evening hours to accommodate the part-time VUZ students. Fol- lowing Khrushchev's downfall in 1964, his reforms were reversed in short order, marking "a return towards more elitist trends" (Matthews 1972, p. 305). But, in a society expressly committed to overcoming the differences between social classes and creating greater social homogeneity (Rutkevich 1977), the persistence of educational inequalities-revealed by sociological research-was bound to remain a source of concern.

Though Soviet commentators could point to declining differences in the percentages of various social origin groups receiving lower secondary edu- cation (occasioned by the saturation of high-origin demand and the spread of universal lower secondary schooling) as evidence of a putative "equal- ization" of the educational levels of different social groups (e. g., Vasil'eva et al. 1985, p. 158), such claims were clearly undermined by public discus- sions of persistent inequalities at higher levels. Recognizing that inequali- ties were rooted in differences in actual ability, analysts justified them in functional terms-arguing that society had to make best use of the demon- strated abilities of its citizens in order to (in consummate dialectical fashion) progress sufficiently to overcome all differences in social situations-and tended to reject special policies to promote the underprivileged regardless of ability (Rutkevich 1977; Shubkin 1965, 1984).

Instead, the regime in the early 1970s adopted policies designed to equalize opportunities for the enhancement of ability, such as measures to improve the material base and quality of instruction in rural general schools (Rutkevich and Fillipov 1978). In 1969 special VUZ "preparatory departments" were established to provide a supplementary program of special preparation primarily to leading young workers, peasants, and demobilized soldiers (nominated by their workplace or unit), who were allowed direct entry to full-time VUZs without entrance exams upon successful completion of the program. The number of participants com- pleting such programs jumped from about 20,000 in 1970 (Rutkevich and Fillipov 1978, p. 63) to 97,000 in 1976 according to Avis (1983, p. 225), who argues that the expansion of such divisions accounted for a sizable portion of VUZ enrollment growth during the 1970s and had a significant impact on the social composition of VUZ students.

No reliable data has been presented demonstrating the actual impact, if any, of these reforms on origin-based educational inequality. Our re- sults allow us to determine if reform truly failed to reduce educational stratification in Soviet Russia as it did in all of the 13 countries studied in Shavit and Blossfeld (1993).~

DATA AND METHODS

Data were drawn from two surveys conducted in Russia with the same in- strument, which was designed in a collaborative effort of American and Russian research teams.'' A clustered sample representative of adults dwelling in the European part of the Russian Federation was interviewed under the auspices of the Institute of Sociology, at Moscow's Soviet Acad- emy of Sciences, in January-February 1991 (N = 2,177). In April 1991 a random sample of adults in two middle-sized cities in the Komi Autono- mous Republic, RSFSR, was interviewed under the direction of Michael Burawoy (N = 476).11 We combined the samples to increase our case base,

An additional sweeping set of reforms was introduced in 1984, primarily to address growing manpower problems by channeling more students into vocational studies (Rutkevich 1984). The reform was short-lived (Connor 1991, p. 99) and fell at the very end of our period of study, so we will disregard it.

lo It should be kept in mind that the reference population for our data-and thus our original empirical findings-is that of European Russia, while that for most of the figures and patterns described in previous sections is that of the entire Soviet Union. We are aware of the risks of generalizing from Russia to the USSR and regret this inconsistency of unit of analysis, but it is made unavoidable by the dearth of statistics and previous literature focusing on Russia.

" For details on design of the instrument, sampling and interview procedures, response rates, and quality checks, see Hout et al. (1992).

then removed cases with missing data and of those born after 1967 (who

may not have completed their education by 1991). The net is 2,141 cases.

To measure the effects of background on educational attainments, we employ a recent version (Raftery and Hout 1993; Blossfeld and Shavit 1993) of a model first pioneered by Mare (1980, 1981), which treats attain- ments as a series of transitions-not a scalar accumulation of years. Each transition has is own distinct magnitudes and patterns of effects. Let pi, represent the probability that student i at some level k -1 will success- fully complete the transition to the next level k. Let yik be the logistic transform of pi, and let Xij represent the set of J independent variables linearly related to yik by a set of regression coefficients pj:

Since the pj are not subscripted by either i or k, we include a variety of interaction terms among the Xij to allow for level- and cohort-specific effects, as well as to detect interactions among origin variables. Cohort- specific descriptive statistics on all of the independent variables are in table 6.

Our data enable us to model four transitions: (1) the completion of lower secondary school; (2) the completion of general secondary school, given the completion of lower secondary; (3) the entry to college, given completion of general secondary; and (4) completion of a college degree, given college entry.'' We coded respondents' highest level of education attained into five categories (see table 6), the first four made up of those who failed to make the correspondingly numbered transition, the last those who made all transitions. We assumed that all those with "some college" or higher had completed general secondary school, since this was by far the most common route to VUZ entry, as discussed above. We coded those indicating "lower vocational" education as having "lower secondary," since the lower vocational track was designed for lower- secondary leavers and is for most an educational cul-de-sac.13 Those

l2 We were not able to model the transition from completion of lower secondary to entry to general secondary since there was no "some general secondary" category available to respondents. We doubt, however, that the number of general secondary school dropouts is significant; these institutions have a two-year duration and various vocational alternatives are available to lower secondary leavers with less academic inclinations.

l3 We treat as "lower vocational" those who answered "PTU" to the question about high level of education completed. Some respondents giving this response may have attended SPTU, a form of vocational training institute not listed among the response categories. An SPTU is functionally the same as a PTU and inferior to the specialized technical education of the tekhnikum, despite its name.

r.

10 "aNw N 4

I ?t?? F"!"!':

4V, N N 0 m0-r.v)mNd

iD d N -N -N3

-m

,o mr-m-N
I

2 """?* 0 "??ZN NNdamt-3v)

b * --NN3

2

mmm-m m
I "!?"It "t""

9 * 00 mm*mNO-r.

* N

m N -3 NN

-

C

0

6

Uv,

dt-mi mv, N
I ?"!"!* 1"??

v, N -N-NmmONN

z

m m N -4 -Nm

m

2

E z 8

aaaN v,b N

1 ?"!?" ?t??EN i 2 ZM*\D+OmN

--b

m

-

a

t;'s2

t-mmm -wet

2w ui

" "!""

-s E m-E N m -r. r.Nmt-Nd\Dr.

N * --*

e,

5 mr.0-v) M*

V]

???? t?"?

W m b N 0 NNmr.mt-Oa

m N -4 -NN

2 a

b

.... ............

........

Y .... .... .,..............

.... ............

.... ............

.... ............

il ................

................

................

................

................

................

................

................

................

...............

. . . . . . . . .......

.........,-. ......

....... .m : : : : : :

........

........ a,... .

........ --. . .

............ . ...
i . .-.

........ t-2SGi;z i

........

........ az3.N .-.

. . . . . . . .

........2zq:: iqj

........

Y ..,.

..: : :..8912;;.:1:

.A

fi:: : fi:: :g$EgGSi9;

"3:::.. .":*...:-E E.

...

2 ...

> 2::: g .:. gg,$gms

... :83.2fi c.s:2=

2: ..: 5 ij :mfima,aadAq

...

-... ...H0 m.g.g3: 2;

"7.:: -: : :%'zag1s ,z

ii c rd

EX jH c ' I,,* s 0

$2 :w $2 :wme '",

E-i5

'V]

~2 ;-2%

;;Z$#pn$$;.=

OE :> .?" hlrl

ofi .z fi-m

E;+a8 2,.4 2 2..gV]gi

zloV]esloXb$;,HH>>>>

E e

V]

Respondent's education (9%):
Less than lower secondary ............................... 12 47 2 6 10 2 0 0
Lower secondary ........................................... 40 27 4 1 40 44 44 41
Complete general secondary ............................. 14 9 7 15 18 15 17
Some college ............................................ 3 2 2 3 2 5 6
College degree or higher .................................. 30 15 2 5 33 34 36 36
Mother's education (%):              
Unknown ..................................................... 1 1 1 2 0 0 1
Less than lower secondary ............................... 52 89 80 65 45 32 2 1
Lower secondary ........................................... 17 3 8 11 2 1 27 24
Complete general secondary ............................. 2 1 6 9 18 2 3 2 8 36
Some college ............................................ 1 1 1 0 2 2 2
..................................College degree or higher 8 1 2 4 9 10 17
Father's education (9%):              
Unknown ..................................................... 5 3 4 5 5 4 5
...............................Less than lower secondary 46 77 67 5 0 39 3 2 24
Lower secondary ........................................... 17 9 10 16 2 1 20 2 3
Complete general secondary ............................. 18 6 9 2 0 2 0 27 2 3
Some college ........................................... 2 1 2 1 2 1 2
College degree or higher .................................. 13 4 7 8 13 16 2 3
N (listwise deletion) .......................................... 2. 141 270 371 228 534 358 380
NoTE.-Fo~ descriptions of variable codings. see text . * Urban = 1; rural = 0.              
t Female = 1; male = 0.              

with "specialized secondary" (tekhnikum), posed a dilemma, since the proportion of specialized secondary entrants with a general secondary diploma has fluctuated over time.14 Since 1975 marked a watershed after which such students constituted an increasing majority, we coded those specialized secondary graduates born in 1958 or later as having completed general secondary, the rest as having only lower secondary.''

We defined six cohorts, as shown in table 6, based on demographic and enrollment trends reflected in figure 2. The first cohort, which turned 17 years old between 1914 and 1945, is not a cohort in any meaningful sense, since it spans such a large number of years and was subject to wildly varying and contradictory policies, demographic trends, and events. There are too few cases, though, to divide it further, so it is included as a baseline. The second cohort (born 1929-38) turned 17 be- tween the end of the World War I1 and the onset of Khrushchev's reforms in 1956-a period marked by rebuilding and, toward the end, rapid expansion of general secondary school outstripping that of VUZs. The third cohort (born 1939-45) corresponds to the demographic "war dent" (births began declining in 1939; see Gel'fand 1992) and, roughly, the Khrushchev reform period when VUZ enrollments were increased rap- idly and students were redirected to vocational rather than general sec- ondary education. The fourth cohort (born 1946-54) marks the first wave of the postwar baby boom, the end of Khrushchev's reforms, and the

l4 It began in the 1950s at about 5%; climbed sharply toward the end of the decade, barely breaking the 50% mark in 1958; then it abruptly dropped to about 25% in 1963. In the mid-1970s it climbed again, passing 50% in 1975 and quickly reaching 6096, above where it remained through the end of the 1980s (TsSU 1977, p. 174; 1981, p. 468; GKSPS 1989, p. 176).

l5 Of course, this procedure implies that we underestimate the proportion of those with general secondary who were born before 1958 and overestimate the proportion of those with general secondary who were born between 1958 and 1967. Consequently, some of the changes in attainments-and in the effects shaping them-in the affected cohorts (the two youngest) may be artifactual. However, were we to code specialized secondary graduates the same for the entire sample, some of the stability across the relevant cohorts could have been artifactual. We adopted our procedure since it is likely to produce fewer measurement errors. We are encouraged by the fact that our cohort-specific measurements of the proportion of respondents who completed general secondary (see table 6 above) appear to correspond quite closely to the calculated (from fig. 2) cohort-specific averages of the percentage of 17-year-olds graduating from general secondary institutions, especially for the four youngest cohorts.

Data set . . .. .. . .... .. . .... . . .. .26 .34 .51 .54 .66 .79 Official . . . .. . . . . . . . . . . . . . . . . .. .05 .15 .45 .52 .68 .84

renewed expansion of general secondary, which exceeded the growth in annual contingents until 1966 and thereafter kept pace with it. The fifth cohort (born 1955-60) marks the second wave of the baby boom and the period when general secondary expansion outstripped both VUZ expan- sion and cohort growth. The sixth cohort (born 1961-67)' smaller than the previous two (Soviet births peaked in 1960, but a year later the war dent generation reached childbearing age; see Gel'fand 1992), experi- enced slower growth in general secondary leavers, and continuing caps on the numbers of VUZ entrants.

Cohort and level are demographic and institutional variables that must be controlled in order for us to understand how the main variables of interest-gender and social origin (which is composed of both parents' education, the main-earning parent's occupation, and urban origin) affect educational success. We used dummy variables for gender (1 = women; 0 = men) and urban origin (1 = those who lived in a town of 50,000 or more inhabitants when they were 16 years old; 0 = those from less populous environments). Mother's and father's education were coded as closely as possible to respondent's education, except we could not distin- guish between general and specialized secondary education for parents since they both provided full secondary diplomas.16

We defined origin classes based on the occupation of the main earner in the respondent's household when the respondent was growing up. First, we used procedures suggested by Erikson and Goldthorpe (1992, pp. 50-51) and adapted to fit state socialist circumstances to code the main earner's occupation and employment status into an eight-category version of Erikson and Goldthorpe's class schema. We then constructed a socioeconomic status index (SEI) to scale these class categories along one dimension, using a modified version of procedures employed by Dun- can (1961).17 Table 6 reports the SEI values for each class. Models with

l6 Preliminary analyses showed no difference between the effects of mother's and father's education on the four transition logits, so we pooled information by summing mother's and father's educations. Data for one or the other parent was missing for 111cases (5% of the valid sample). To avoid the bias that excluding them would have introduced, we incorporated a "family structure" effect as a dummy variable equal to "1" if one parent's education was missing and zero otherwise. When the family structure dummy proved to be insignificant in all four regressions, we dropped it and concluded that the loss of one parent was equivalent to losing the equivalent number of years of parental education. Our additive specification of parents' education implies that any disadvantage suffered by those with only one parent is due to the absence of the missing parent's contribution to educational origins. While this may be counter- intuitive, it is supported by the nonsignificance of a dummy variable for children missing one parent.

l7 The exact formula was SEI = .59(1,) f .55(E,), where I, refers to the percentage of class c with income in the top three deciles and E, to the percentage in class c

a dummy variable specification of main earner's occupation fit no better than ones with the linear socioeconomic scale (see the appendix). From this evidence we conclude that the departures from linearity are not significant.

Table 6 shows that our sample captured some long-term sociodemo- graphic trends in Soviet society that have been widely commented upon as representing its modernization (e.g., Lewin 1988): structural occupa- tional mobility depleted the ranks of agricultural workers and bolstered those of professionals and industrial workers, respondents' and parents' education increased monotonically over time, main earner's SEI in- creased almost monotonically, and urban origins became more common, especially following the war.

We give details of our variable selection and specification procedures in appendix A. By way of summary, we note that, in addition to the main effects of level, parents' education, main earner's occupation, ur- ban origin, and sex, we tested two-way interactions involving (1) level

x cohort; (2) each origin variable and sex x level and x cohort; (3) parents' education x occupation; and (4) parents' education, main earner's occupation, and urban origin x sex. We also tested three-way interactions involving (a) parents' education, main earner's occupation, urban origin, and sex x level x cohort; (b) parents' education x main earner's occupation x level; and (c) parents' education x sex x level-for a total of 24 effects, many of which were not significant. We smoothed the significant interaction effects by recoding the level and cohort vari- ables to linear specifications. l8

RESULTS

Changing Transition Rates

Our data on the cohort-specific success rates at each transition (fig. 3) reflect the trends displayed in figure 2 in a new light. Transition rates reveal squeezes on institutions that enrollment figures by themselves can-

with greater than general secondary education (including specialized secondary, some college, or more). Since data on income were not available for respondents' parents, the class-specific SEI values were computed using information on the respondents' earnings. The resulting SEI rankings reflect the Soviet particularities in occupational prestige rankings observed by Treiman (1977): manual work is upgraded and routine clerical work downgraded, relative to the rankings of such occupations in noncommunist countries. See Gradolph (1993) for a detailed discussion of this pattern of earnings differences among occupational groups in Russia.

'*Universal entry to secondary school eliminated variance in the first transition for the last three cohorts at level 1, so we treat the first transition as "missing" for the last three cohorts.

A. Incomplete secondary B. General secondary

30% ! I 30% 4 I Pre-1929-1939-1946 1955-1x1-Re-1929-1939-1946 1955-I%!1929 38 45 54 60 67 1929 38 45 54 60 67

Birth cohort Birth cohort

C. Enter WZ D. WZ degree

40%/ ;I]

30% i i j : , , i i j , Pre-1929-1939-1946 1955-1x1-Re-1929-1939-1946 1955-1%11929 38 45 54 60 67 1929 38 45 54 60 67

Birth cohort Birth cohort

FIG. 3.-Observed transition rates by level of education, European Russia,

1991.

not. When enrollment figures rise we cannot be sure that an individual's

chances for advancing from one level to the next higher level have gone

up or that there are simply more people "at risk" of moving up because

of changes at lower levels of the system.

The pressures on academic secondary and VUZs definitely rose during the Soviet period. Lower secondary education rose meteorically from just over half to nearly universal across the first four cohorts. Universal lower secondary was practically achieved by the time the 1946-54 cohort reached that level in the early 1960s. The transition rate from lower to general secondary rose more modestly until after the rise of incomplete secondary was over. The transition rate to general secondary dropped from 50% to 45% for the 1929-38 cohort due to the abrupt rise in lower secondary completion, then rose modestly to 55% for the small 1939-45 cohort. It stalled at 55% for the 1946-54 cohort due to the combination of Khrushchev's reforms and saturation at the lower secondary level for postwar "booming" cohorts," only to rise dramatically for the two youngest cohorts, reaching 80% by the late 1980s.

Higher education institutions could accommodate modestly growing enrollments but could not keep up with the rapidly growing numbers of students who were eligible for VUZ entry. As we saw in figure 2, VUZ entrants exceeded general secondary leavers in the immediate postwar years because returning soldiers enrolled to resume the educations that military service interrupted. After that disruption, VUZ entries grew slowly while the number of general secondary leavers exploded. Conse- quently, the general secondary-to-VUZ transition rate declined steadily from roughly three-fourths of the 1929-38 cohort to barely one-half of the 1961-67 cohort. Decline on this scale is unprecedented. Italy and Ireland experienced a squeeze on higher education entry after rapid expansion of secondary education (Cobalti 1990; Shavit and Westerbeek 1995; Raftery and Hout 1993), but nothing on this scale-fewer cohorts were affected and the decline was more modest.

Information on VUZ dropout rates is not generally available. Ac- cording to official criticisms they grew as a result of Khrushchev's re- forms, but hard data have been lacking. Our data show virtually no change in the transition from VUZ entry to a VUZ degree. The VUZ success rate drops slightly for the third cohort (they were in university during Krushchev's reforms) and again in the last two cohorts. The de-

l9 The decline in general secondary coverage occasioned by Khrushchev's reforms affected people born in the 1940s, i.e., the latter half of cohort 3 and the first few years of cohort 4, so it is not visible in the chart. We broke the cohorts where we did so that we could focus on Krushchev's effects on VUZ entry (for which our 1939-45 cohort is well situated).

clines in the cohorts born since 1955 are more significant than this figure

makes them appear because they are exclusively among men.

Some of the change over time is the consequence of previous changes; that is, once educated, parents have more resources and higher aspira- tions for their children than their parents had for them. This puts upward pressure on educational institutions. How much of the trend toward more education in Russia was the second-generation consequence of the pre- ceding generation's advance, which was reflected as higher origins for our respondents from later cohorts? With the model of transition rates (described in the appendix) we can compare cohorts while holding ori- gin constant. Figure 4 shows each cohort's expected transition rates for urban-origin men and women whose main earner was a skilled worker and whose parents had complete secondary education. It also shows each cohort's overall transition rates for comparison.

The upgraded origins of recent cohorts are sufficient to sustain univer- sal incomplete secondary education. Even 50 years ago, virtually no par- ents who themselves had completed secondary school would allow their children to drop out before completing some secondary education.

Social composition also plays a role in sustaining high levels of general secondary education in recent cohorts. While general secondary educa- tion for the parents is "average" for recent cohorts, the early cohorts would have had higher rates of general secondary education if their par- ents had that much education. The improving composition masks some pressures for reversal in Soviet policy toward secondary education. Although the most recent cohort has the highest secondary completion rates, the 1939-45 cohort stands significantly above the adjacent ones. The first cohort to be in secondary school after World War I1 (born 1929-38) actually achieved less than the preceding cohort. The 1939-45 cohort-attending secondary school in the late 1950s-achieved far more. The Krushchev reforms harmed the 1946-54 cohort. For men whose parents achieved secondary education, there is no net trend in their own transition rates; women's chances improved by almost 20 percentage points.

The drop in the transition to higher education is all the more precipi- tous for those from educated backgrounds. Nearly all of the offspring of people with general secondary education born before World War I1 could expect to enroll in VUZ; that expectation dropped by 40 percentage points over the six cohorts we consider (about 50 years).

As noted above, the steady state in graduation rates for those who enter VUZ masks a drop in men's graduation rates in recent cohorts (fig. 4, part D). The changing social origins do not affect the trend because origins only weakly affect earning a degree once the student has entered a VUZ.

A. Incomplete secondary B. General secondary

30% 4       4 30% !         I
  Re 1929 1939 1946 1955 1961   Re 1929 1939 1946 1955 1961  
  1929 38 45 54 60 67   1929 38 45 54 60 67  
    Birth cohort         Birth cohort      
    C. Enter WZ         D. WZ degree      

AWomen

(all)

\ \ \ \ \ \ \

\%

Men

30% 4 i 30% 4 I Pre-1929-1939-1946 1955-1961-Re-1929-1939-1946 1955-19611929 38 45 54 60 67 1929 38 45 54 60 67

Birth cohort Birth cohort

FIG.4.-Expected transition rates by level of education and gender, European Russia, 1991 (main earner is a skilled worker, SEI =36, and both parents have general secondary, sum = 6).

Changing Educational Stratification

Social origins and gender significantly affect the chances of educational success at each transition for each cohort (see table 7)." The pattern of effects is complicated, with effects varying in magnitude across transi- tions and cohorts. The constants capture the combined effects of cohort size and institutional change. Parents' education interacts with main earner's occupation and sex (although these interactions are the same for all cohorts and transitions). This means that the effects of parents' educa- tion are greater for high status families and that the effect of class is greater for families in which the parents have academic educations. The return to each unit of parents' education is in all cases .07 higher for women than for men.

Geographical origins strongly affected entry to secondary education in the first three cohorts (urban origins were advantageous). Urban origins' ability to aid the completion of general secondary school was statistically significant but modest, except for its strong effect on the 1939-45 coh~rt.~'

The urban-rural gap at VUZ entry was appreciable for all cohorts. This cross-level pattern suggests that performance-based selection-clearly embodied in the VUZ entrance exam-was more favorable to urban youths than were the pure self-selection mechanisms (aspirations, analysis of chances, etc.) operative at the general secondary level. It supports the argument that the disadvantage of rural youths was rooted in the inferior quality of primary and secondary training outside the cities

"The coefficients in table 7 show the effects of social origin and gender on the chances of educational success at each transition for each cohort and serve as the basis for the remaining analyses in this article. The coefficients were calculated from our preferred statistical model, which we describe in the appendix. In our initial models the fluctua- tions of coefficients across cohorts and transitions were considerable. In order to produce results we could interpret, we constrained many of the interactions to be linear functions of cohort and/or level (see the appendix below). The linear constraints dampened some, but not all, of the large fluctuations, and eliminated those which were not statistically significant. It is important to note that the amplitude of the fluctuations that remain is less than that of the corresponding unconstrained coeffi- cients. To take one example, freeing each cohort-specific effect of gender on comple- tion of a VUZ degree produces the following six coefficients (in order of cohort):

-1.79, -0.63, 3.94, -1.04, 0.26, and 1.05. These compare with -0.81, -0.81, 1.44, -0.81, 0.32, and 1.44, according to the preferred model. Freeing this particular constraint resulted in a likelihood ratio of .40 (5 df). Similar experiments with some of the other highly constrained results likewise confirmed that (a) lack of fit and (b) odd patterns like this are both features of the data, not artifacts introduced by our procedures.

The sharp and brief rise in the urban origin effect on general secondary completion for the 1929-45 cohort is consistent with the observation of many (e.g., Medvedev and Medvedev 1978) that Khrushchev's policies generally wrought great harm in the countryside.

TABLE 7
LOGISTIC REGRESSION BY LEVELAND COHORT: RUSSIA,1991
COEFFICIENTS EUROPEAN
INDEPENDENTVARIABLE Pre-1929 a
Incomplete secondary:
Parents' education .................................................... .70 Main earner's SEI (X 100) ............................................ -.03 Urban origin .............................................................. .89 Woman ..................................................................... -.38 Parents' education X main earners' SEI ( X 100) ............... .2 1
1929-38 .70 -.03 .89 -.38 .21 1939-45 .03 6.62 .89 -.38 .2 1 1946-54 ... ... ... ... ... 1955-60 ... ... 1961-67
...
...
...
...
...

Parents' education x woman

.........................

..............

Constant ...................................................................

General secondary:
Parents' education .......................................................
Main earner's SEI (X 100) ............................................
Urban origin ..............................................................
Woman ..................... ...............................................
Parents' education X main earner's SEI (X 100) ...............
Parents' education x woman .......................................
Constant ...................................................................

(Zaslavsky 1982). Among university students, those of rural origins were slightly more likely to graduate. To overcome their disadvantage and gain entry to a VUZ, rural youths would have had to display exceptional drive and ability, which would serve them well once they got to the university (see Mare [1980, 19931 for a more general form of this "unob- servable heterogeneity" argument).

Women's educational disadvantage-more pronounced in less edu- cated families of originz2-was wiped out during the Soviet period. In this regard, Russia is consistent with all 13 cases examined in Blossfeld and Shavit (1993).

Due to the interaction between parents' education and class we con- sider them together (table 8). The spread between the most and least favorable parents' education and class (using high and low extreme val- ues for each variable) gives an indication of the overall combined effect of the two variables for a given transition and cohort. Research on changing educational stratification in other countries leads us to expect little varia- tion among cohorts and weaker effects at higher levels of the educational system (Blossfeld and Shavit 1993).~~

With some notable exceptions, Rus- sia conforms to the general pattern established elsewhere. The combined effect is greatest at the first transition, equal for the two subsequent ones, and lowest for the final one. There are important exceptions occurring in the pre-1929 and 1939-45 cohorts. For the oldest cohort, the spread is greatest at VUZ entry-reflecting the high magnitudes of both the main education and class effects-but opposite in sign. The VUZ recruitment policies in effect at various times during the 1920s and 1930s apparently were effective and reversed advantage to lower class groups.24 For the 1939-45 cohort, the education effect dropped as the first transition neared saturation. At VUZ entry, a heightened class effect this time favors those with higher origins, indicating that Khrushchev's egalitarian reforms not only failed, but had the opposite result to that which was intended. As for the final transition, we have no ready explanation for the rather

22 For parents' education = 8 (both parents have some college), add .56 to the gender effect. For parents' education = 6 (both have high school degrees), add .42 to the gender effect, etc.

23 AS Mare (1993) has recently shown, there are a number of possible ways unobserved determinants of educational success might influence measured origin effects at a given transition. Unfortunately, our data do not contain the information on siblings neces- sary to correct for this in the manner he suggests.

24 Recall that the oldest cohort spans 30 years of fluctuating and occasionally contra- dictory education policies. The negative sign of the class effect, which must be viewed in conjunction with the interaction between parental education and class, testifies to the effectiveness of those policies, however short-lived, which were designed to favor working-class students, especially at the VUZ entry level.

TABLE 8

COMBINEDEFFECTSOF PARENTS'EDUCATION OCCUPATIONON

AND MAIN EARNER'S EDUCATIONAL BY LEVEL OF EDUCATION AND COHORT:

SUCCESS MEN, EUROPEANRUSSIA, 1991

Incomplete secondary: Class I, education 8 ....... Class I, education 2 ....... Class VIIb, education 8 ....... Class VIIb, education 2 ....... Spread ................ General secondary: Class I, education 8 ....... Class I, education 2 ....... Class VIIb, education 8 ....... Class VIIb, education 2 ....... Spread ................ Enter VUZ: Class I, education 8 ....... Class I, education 2 ....... Class VIIb, education 8 ....... Class VIIb, education 2 ....... Spread ................ VUZ degree: Class I, education 8 ....... Class I, education 2 ....... Class VIIb, education 8 ....... Class VIIb, education 2 ....... Spread ................

NOTE.-"Spread" row gives maximum difference in combined effect magnitude for relevant cohort.

dramatic intercohort fluctuations, as they include several changes in sign for each effect. The high negative magnitude of the class effect for the war cohort does suggest that Khrushchev's reforms had the effect of increasing the dropout rate among VUZ students of higher origins, and the subsequent reversal casts doubt on the effectiveness of the "prepara- tory departments" at increasing the retention of worker-origin VUZ stu- dents.

Evaluating MMI

The MMI hypothesis predicts that rapid rises in educational success at one level will-if they lead to saturation of upper-class demand-result in a decrease in educational stratification at that level and an increase in educational stratification at the next level, unless they are offset by corresponding increases. At the lower secondary level, middle-class de- mand is already practically saturated among the urban population for the oldest cohort (fig. 4 above), so the gains in overall transition rate at this level led to pronounced destratification across the first three cohorts (results not shown), as less privileged groups gained from the expansion of coverage. As noted above, the overall general secondary completion rate actually declined for the 1929-38 cohort and, to a lesser extent, the 1946-54 cohort, largely as a result of this expansion. Also, the overall VUZ entry rate falls across the four youngest cohorts. According to MMI, drops in the conditional transition rates at both of these levels should lead to increased stratification. Though general secondary education coverage grows rapidly for the 1939-45 group and the two youngest cohorts, upper-class demand never reaches full saturation: the expected completion rate for men with the highest main earner's SEI (71) and parents' education (10) values peaks at 95% for the 1939-45 cohort, drops to 88% for the subsequent 1946-54 cohort, then rises monotonically to reach 94% in the youngest (1961-67) cohort. Thus, according to the strong version of MMI, we expect no change in the amount of stratifica- tion at the general secondary level, rapid growth notwithstanding. To evaluate these predictions, we examine changes in the effects of parents' education and main earner's occupation on success at the general second- ary and VUZ entry levels. In the following figures we use probabilities expected under the model in table 7 rather than observed probabilities (some of which would be based on very few cases-N = 0, 1, or 2 in some combinations of parental education and cohort).

Among urban Russians, there is some evidence that increased stratifi- cation did materialize at the general secondary level for the 1929-38 and 1946-54 cohorts (fig. 5). The slopes of these lines are steeper than those

A. Women B. Women

loo% T

:

0 20   40 60 80 0 2 4 6 8 10
      Main earner's SEI       Parents' educations (sum)  
      C. Men       D. Men  
40j j i , ; i j j ,
30% 0 20   40 Main earner's SEI 60 80 0 2 4 6 8 Parents' educations (sum) 10

FIG. 5.-Expected probability of completing general secondary by social ori- gin, cohort, and gender, European Russia, 1991. For A and C, both parents, general secondary education (sum = 6); for B and D, main earner, skilled worker (SEI = 36).

of the others in each case, though the difference appears to be minimal among men." To verify that this graphic form is not obscuring a trend in stratifica- tion, we calculated the ratio of the form

where i indexes parents' education, j indexes main earner's occupation, k indexes cohort, and 1 indexes level. For a given gender, and for geo- graphic origin, Rijk,will be less than one because the top combination of parents' education and main earner's occupation-that is, both parents with college degree (i = lo), Class I (j= 1)-will have a higher expected probability than any other social origin combination. If stratification is increasing, then the gap between the top and other combinations will widen, and the Rijk,will decrease away from one over time.

The ratios of expected probabilities of finishing general secondary edu- cation change appreciably over time (as the selected Rjk,values in fig. 6 show), particularly for persons with disadvantaged origins. Strikingly, the intercohort pattern of changes mirrors precisely the pattern of changes in the overall transition rate (fig. 6, part B). This indicates that not only did decreases in the overall rate indeed lead to increased stratification as predicted by MMI, but also, in contradiction to MMI, increases in the overall rate led to decreases in stratification, despite the fact that satura- tion of upper-class demand was not attained in the course of these in- creases. Over the course of the Soviet period, there is practically no net trend among men. Among women, however, a positive trend is dis- cernable over time, indicating less stratification in recent cohorts than in earlier ones.

Stratification increased at the point of entering higher education. The effect of main earner's occupation on the probability of entering VUZ was much greater for the 1939-45 cohort than for any other, particularly among women (fig. 7, parts A and C). Though the effect of class on the log odds of entering VUZ does not change, class differences in the

25 The expected probabilities in fig. 5-particularly in part B, which shows the rela- tionship between success and parents' education for urban women-vary over a wider range at lower-origin than higher-origin parents' education despite the absence of a significant interaction between parents' education and cohort at level 2. The differ- ences are due to the inherent nonlinearity of the logistic regression model.The effect of continuous variable X on the probability of success p is given by the partial deriva- tive of p with respect to X: dyldX = P,p (1 -p), where P, is the logistic regression coefficient for the effect of X on y. For fixed P,, dy1dX increases from zero to 0.25P, as p increases from zero to 0.5, and then decreases to zero again as p continues to increase from 0.5 to 1. See Petersen (1985) for details.

80% --70% -- \ 'wO 0-     80% --70% -+   0-' observed)
60%         60% --      
40% Urban men Urban women ---Rural men --Rural women=     50% 40% ----      
20% J he1929 192938 19391946 45 54 Birth cohort 195560 I 196167 20% -1 he1929 192938 1939194645 54 Birth cohort 195560 I 196167
20% J         i 20%           1
  he 1929 1939 1946 1955 1961 Re 1929 1939 1946 1955 1%1  
  1929 38 45 54 60 67 1929 38 45 54 60 67  
      Birth cohort         Birth cohort      

FIG. 6.-Ratios of expected complete general secondary rates by gender and geographic origin, European Russia, 1991 (main earner's class and parents' edu- cation controlled; baseline is main earners' SEI = 71). For A, class I (SEI = 71); general education (sum = 6); for B, class IIIa (SEI = 56), general education (sum = 6); for C, class VIIa (SEI = 29), general education (sum = 6); for D, class VIIa (SEI = 29), primary education (sum = 2).

A. Women B. Women

30% 30% -1 I

0 20 40 60 80 0 2 4 6 8 10

Main earner's SEI Parents' educations (sum)

C. Men D. Men

30%

0 20 40 60 80 0 2 4 6 8 10

Main earner's SEI Parents' educations (sum)

FIG.7.-Expected probability of entering VUZ by social origin, cohort, and gender, European Russia, 1991. For A and C, both parents, general secondary education (sum = 6); for B and D, main earner, skilled worker (SEI = 36).

probabilities of VUZ entry in the three youngest compared with the two oldest cohorts increase because the constants decrease. Thus the unchanging effect of class has a larger impact on the probability of VUZ entry.26 This is our evidence that the failure of the university system to keep up with growing numbers of persons eligible to enter lowered the probability of entry to the point where even the highest educational cate- gories fell short of 100% entry. The effect of parents' education is by far the greatest in the pre-1929 cohort (fig. 7, parts B and D). The profound curve in the line for the pre-1929 cohort is attributable to the ceiling effect built into the logistic regression model." For women the probability of VUZ entry is distributed over a wider range in the last two cohorts than in the other post-1929 cohorts; among men it is the last three cohorts that have a wider range of expected probabilities. The underlying param- eter did not change, but the logjam in higher education reduced every- one's chances enough that even the people in those cohorts from high educational backgrounds do not approach 100% enrollment.

The ratio Rij,,

of each social origin combination's probability of VUZ entry to that of the comparison group declines over time for all but the lowest educational category (fig. 8). This is our clearest evidence of increased stratification. If stratification were not increasing at the univer- sity level, Rij,,

would show no net trend. The fact that it declines means that each class's chances of VUZ entry-relative to the chances of the top class-fall over time.

The increased stratification at the university level canceled the egalitar- ian effects of increased access to general secondary education. This is most clearly seen in the cumulative probability of attaining a VUZ degree for each social origin and gender (fig. 9). We multiplied the expected probability of each transition to obtain an expected cumulative probabil- ity of a degree. The key cohorts are highlighted with darker, wider lines in figure 9. Parts A and C show a slightly higher effect of main earner's occupation among women and men born 1939-45 and 1961-67 and among men born 1955-60. Parts B and D show that only the oldest cohort experienced substantially sharper stratification by parents' edu- cation.

Figure 9 is useful for descriptive purposes, too. It shows that even middle-class young peoples' prospects of earning a university degree dete- riorated throughout the Soviet period. Young people from working-class and peasant origins whose parents had academic educations also experi-

26 This is an intrinsic feature of the logistic regression model, which implies that
dpldX = P,p(l -p).
''When p = .5, dpldX = P,p(l -p) is greater than when p = .8 by a factor of 251
16 (more than a 50% increase in dp/dX).

Urban men
Urban women
---Rural men
30% --Rural women
20°h 4     1 20% 4         I
  Re 1929 1939 1946. 1955 1x1   Re 1929 1939 1946 1955 1961
  1929 38 45 54 60 67   1929 38 45 54 60 67
  Birth cohort           Birth cohort    
20% iI
Re 1929 1939 1946 1955 1x1 Re 1929 1939 1946 1955 I%]
1929 38 45 54 60 67 1929 38 45 54 60 67
    Birth cohort         Birth cohort    

FIG. 8.-Ratios of expected VUZ entry rates by gender and geographic origin, European Russia, 1991 (main earner's class and parents' education controlled; baseline is main earners' SEI = 71; parents' education sum = 10). For A, class I (SEI = 71), general education (sum = 6); for B, class IIIa (SEI = 56), general education (sum = 6); for C, class VIIa (SEI = 29), general education (sum = 6); for D, class VIIa (SEI = 29), primary education (sum = 2).

A. Women B. Women

oO/, i

0%-I

0 20 40 60 SO 0246810

Main earner's SEI Parents' educations (sum)

C. Men D. Men

0% /0% /

0 20 40 60 SO 0246810

Main earner's SEI Parents' educations (sum)

FIG. 9.-Expected cumulative probability of a VUZ degree by social origin, cohort, and gender, European Russia, 1991. For A and C, both parents, general secondary education (sum = 6); for B and D, main earner, skilled worker (SEI = 36).

enced a steady deterioration of their prospects for earning a university degree. The most disadvantaged young people-those whose parents were working class and had achieved less than secondary education- saw an initial rise in their prospects during the early part of the Soviet period, but they fell as did everyone else's after the 1960s.

CONCLUSIONS

Our 1991 national survey of social structure and consciousness in Russia

reveals substantial inequality of educational attainments throughout the

Soviet period. Parents' education, main earner's occupation, and geo-

graphical origin remained instrumental to these inequalities. Gender dif-

ferences-initially favorable to men-were removed during the Soviet

period, and for some transitions were reversed.

Secondary education grew rapidly in Russia throughout this century, much as it has done in other industrialized nations. Russian higher educa- tion, though, failed to keep pace with the rapid growth of secondary education. This disparity led to an enrollment squeeze at the university level. Far more students were prepared for university education than the higher educational system (VUZ) could accommodate. The result was a dramatic fall in the proportion of secondary school graduates who went on to higher education between 1962 and 1980. In the cohorts born after 1945, all social origin groups experienced declining prospects for making the transition from academic secondary to university education. For men of any social origin, this helped produce a drop in the overall odds of attaining a VUZ degree across the three postwar cohorts, while the gains by women relative to men at each educational transition offset the impact of the bottleneck at VUZ entry. As a result women's odds of attaining a VUZ degree over this same period neither rose or fell.

The rapid expansion of general secondary education led to greater equality of educational opportunity. Although all origin groups benefited to some extent, Russians from worker or peasant origins whose parents had minimal education gained the most from the expansion. The re- sulting bottleneck at the VUZ entry point hurt these same disadvantaged groups more than the rest, however, so the effect of social origins on the conditional probabilities of making this crucial transition actually increased during the later decades of the Soviet period. The overall effect of more equality of opportunity at the secondary level and less at the university level was little net change across the three postwar cohorts in the social-origin-based differences in the likelihood of attaining a VUZ degree. While this finding lends support to scholars who claim that the Soviet class structure crystallized during the 1970s and 1980s (e.g., Lapi- dus 1983), the image of crystallization fails to capture the related trend of all origin groups-even the most advantaged-facing declining proba- bilities of attaining a VUZ degree. Furthermore, the contradictions that produced "crystallizationn-rising opportunity at the secondary level canceled out at the university level-are both complicated and subtle enough that we doubt that they could have had the destabilizing political consequences Lapidus and others envisioned.

Reform fared no better in the Soviet Union than in the 13 countries studied by Blossfeld and Shavit (1993). While there was less inequality in chances of attaining a VUZ degree for those who reached college age immediately following the war than for those who did so earlier (1929-38 vs. pre-1929 cohorts), there was greater class inequality for the cohort born during the war (except among urban men) despite the favorable conditions associated with its small size. This is decisive proof of the adverse results of Khrushchev's efforts to make access to higher educa- tion more egalitarian. Attempts in the 1970s to bolster the representation and performance of worker- and peasant-origin youths at the VUZ level through the creation of preparatory departments succeeded, at best, only in preventing these groups from losing further ground.

The postwar increase in educational stratification at the university level in Russia is highly unusual. Blossfeld and Shavit (1993) identify only two industrialized nations (Sweden and the Netherlands) where so- cial class barriers actually fell. But they find none where social class barriers became higher. In Hungary and Czechoslovakia, Soviet-dominated regimes were implementing quota schemes that fostered a destratification of education in the first socialist generation (Szelenyi and Aschaffenberg 1993; Mateju 1993). The Soviets also tried quotas but quickly abandoned them (Matthews 1973). 28

The MMI model anticipated this accentuation of class differences in higher education in a system manifestly committed to eliminating such differences. The Russian case implies a modification of MMI, however. We encountered instances of decreases in inequality corresponding to expansions at the general secondary level, even without ceiling effects on privileged social origin groups. It is possible that 100% expected transi- tion rates for the upper classes at a given level is simply an unrealistic definition of saturation of their demand. Especially at higher levels, it is likely that some youths of any social origin group would opt not to con- tinue their education. While the top social origin groups expected transi- tion rates for the cohorts in question did not reach loo%, it did come relatively close.

Overall, the concept underlying MMI-that the level of inequality in

The Soviets did maintain ethnic quotas in the non-Russian republics until at least 1991.

attainment at a given educational level is inversely associated with the growth (or decline) of enrollments relative to the eligible population size-receives strong support from our results. Increasing transition rates at one level might produce greater equality at that level but will produce greater inequality at a subsequent level unless the pressure is relieved by proportionate expansion there as well. This occurred twice during the Soviet period. First, a sudden expansion of lower secondary education for the 1929-38 cohort brought about a drop in the conditional probability of completing general secondary education, to the detriment of the lower- origin groups. Then the postwar expansion of general secondary school created an enrollment squeeze at the VUZ level which again dispropor- tionately hurt the chances of lower-origin classes.

The irony of inequality induced by the pressures of expansion at lower levels in Russian education is that it occurred despite an intention to reduce inequalities. The Soviets saw centralization as a collective control mechanism that would efficiently allocate human capital and simulta- neously counteract the tendency of individual choices to perpetuate in- equalities across generations. We cannot evaluate their allocation of human capital. Our assessment of the intergenerational persistence of educational inequality shows that, despite some early successes in reduc- ing the importance of social origins for educational achievement, educa- tional stratification was greater at the end of the Soviet period than at the beginning. The various reform efforts undertaken during the Soviet period not only failed to counteract the stratifying pressures created by individual differences in resources but also produced stratifying pressures of its own. By reforming only parts of the system at a time (starting at the lowest levels), educational policymakers put pressure on the levels above. The planners could not anticipate that opening access at one level would increase stratification at the next level, but that outcome is now clear.

APPENDIX

Constructing Linear Cohort Contrasts

We employ the same approach as that used by Raftery and Hout (1993) and Hout, Raftery, and Bell (1993), analyzing the sample of transitions, not of individuals. We pooled data for the four transitions (at each of which the ineligible population is censored) and the dependent variable is a dummy variable, success, which equals one if the transition is made and zero if it is not.

To arrive at a preferred model, we followed a set modeling procedure (table Al). The first step (A) was to settle upon a specification of the

main earner's class variable and the main effects to be tested. We deter-

mined that the scalar specification of the main earning parent's class

based on the calculated SEI scores was preferable to a dummy-variable

specification (models 1 and 2). The scalar specification uses six fewer

degrees of freedom at the cost of only 11 chi-square points, and is prefera-

ble from a modeling standpoint since it makes it easier to test for interac-

tions. We also determined that variables for family intactness and farm-

worker status did not improve fit and were not significant, so they were

discarded for the remainder of the analysis (models 3 and 4).

The next step (B)involved entering dummy-variable specifications of the all the two-way interactions involving level. Cohort x level interac- tions were included as 20 dummies corresponding to five of the six cohort categories for each level. This specification has the consequence of freeing the constants at each cohort-l.evel combination. A priori constraints on the constants can lead to biased estimates of other effects; with this speci- fication we avoid that danger. The interactions of level with parents' education, parent's class, urban origin, and gender (female) are entered as dummies to establish level-specific values for each of these background effects, which will serve as baselines in the search for level-specific inter- actions between cohort and the effect (i.e., level X cohort X background effect interactions). In step C we enter the two-way interactions among background variables (model S), remove the nonsignificant ones (model 6), and determine that no interactions between these significant effects and level are significant (not shown).

We then identify three-way interactions between each background variable, level, and cohort, and we constrain those that are significant to be level-specific ordinal recodings of the cohort variable; these recod- ings serve as multipliers of the main effect of the background variable at the relevant level. We proceed one level at a time (steps D-G), identi- fying then constraining these interactions in the three stages illustrated for step D (the two latter stages are collapsed into one for steps E-G). First, dummy-variable specifications of the level x cohort interactions at a given level are multiplied by each of the four background variables (model 8). This uses 4(C, -1)degrees of freedom, where C, is the number of cohorts for which parameters are estimated at level k (three in the case of level 1, otherwise six). In a substep, we adjusted the dummy specifications so that within each respective set of interactions the omitted category is either the highest or lowest. This allows us to evaluate the statistical significance of the interaction-if none of the "dummy interac- tion" terms differ significantly from the (level-specific) baseline magni- tude of the effect, there is no significant change in the effect across co- horts. In the next stage, we remove insignificant interactions (model 9).

This leaves a series of dummy interaction terms, some or all of which

TABLE A1
Step and Model Chi-square df Comments
A. Pick class specification, evaluate intactness and kolkhoz effects: 1 L(d) + E + W + U + P(d) ...............................................    
2L(d)+E+W+U+P ...................................................    
o\ P (n 3 [2] + oneparent(d) ...........................................................
4 [2] + farm(d) ...................................................................
   
  B. Enter level x cohort and background X level two ways (as dummies): 5 [Z] + LC(d) + EL(d) + PL(d) + UL(d) + WL(d) ....................    
  C. Enter four two ways among background, remove nonsignificant ones: 6 [5] + PE + EW + PW + UW ........................................... 4,580  
  7 [6] -PW -UW .............................................................. 4,582 No PEL(d) or EWL(d) are significant
  D. Identify and constrain background by level X cohort: 8 [7] + ELlC(d) + PLlC(d) + ULlC(d) + WLlC(d) ................. 4,574 No ULlC or WLlC are significant
  9 [8] -ULlC(d) -WLlC(d) ................................................. 4,578  
  10 [9] with 1 df specification of ELlC and PLlC .......................... 4,580 ELlC (3 = 1) (else = 0)
      PL1 C (3 = 1) (else = 0)
  E. Identify and constrain background X level 2 X cohort: 11 [lo] + ELZC(d) + PLZC(d) + ULZC(d) + WL2C(d) ............. 4,560 No EL2C or PL2C are significant
  12 [ll] -EL2 C(d) -PL2 C(d); 1 df specification of UL2 C and of    
    4.578 UL2C (3 = 1) (else = 0)
      WL2C (1-3 = 1) (4 = 2) (5 = 3) (6 = 4)
F. Identify and constrain background X level 3 X cohort: 13 1121 + EL3C(d) + PL3C(d) + UL3C(d) + WL3C(d) ............... 4,749 4,558 No UL3C are significant
14 [13] UL3C(d); 1 df specification of EL3C, PLSC, and WL3C ... 4,755 4,575 EL3C (1 = 1) (else = 0)
      PL3C(3 = 1)(2,4,5,6 = 2)(1 = 3)
        WL3C (2, 3 = 1) (1 = 2) (6, 5 = 3) (4 = 4)
G. Identify and constrain background x level 4 x cohort:            
15 [14] + EL4C(d) + PL4C(d) + UL4C(d) + WL4C(d) ............... 4,729 4.555 No UL4C are significant
16 [IS] -UL4C(d); with 1 df specification of EL4C, PL4C, and      

WL4C

.............................................. ...

........................

4,736

H. Constrain background x level 2-ways:
17 1161 with 1 df specifications for EL, PL, UL, WL ..................... 4,743

Q\

m

(n I. Constrain level x cohort:
18 1171 with 1 df specifications for LlC, LZC, L3C, L4C ................ 4,747

J. Constrain main level effect:
19 [18] with 1 df specification for level ....................................... 4,747

NOTE.-L = level; E = parents' education (sum); W = woman (male = 0; female = 1); U = urban origin (rural = 0; urban = 1); P = main earning parent's class; one parent = dummy indicating one missing parent; farm = collective farm origin dummy; C = birth cohort; (d) = indicates dummy variable specification of the variable in question; otherwise specification is ordinal. For P, ordinal specification is the class-specific SEI score. For L and C, and interaction terms including them, ordinal specifications are recodes explained under "comments." L1, L2, L3, and L4 refer to dummy specifications of level where the respective level equals 1. Thus, e.g., "PL3C (3 = 1) (2, 4, 5, 6 = 2) (1 = 3)" means: parent's SEI X level recoded as (3 = 1) (else = 0) x cohort recoded as (3 = 1) (2, 4, 5, 6 = 2) (1 = 3).

TABLE A2

LOGISTICREGRESSIONCOEFFICIENTS,PREFERREDMODEL

Variable* b SE P

Level (1 = 1) (3 = 2) (4 = 3) (2 = 4) .....................
Level 1 x cohort (1) (2) (3) ..................... ..............
Level 2 X cohort (6) (1, 3, 5) (2, 4) .........................
Level 3 X cohort (6) (1, 3, 4, 5) (2 = 5) ..................
Level 4 x cohort (6) (5) (1, 2, 3) (4) ........................

........ . . . .. . . .. . . . . . . . . Parents' education (sum)

Parents' SEI ..................................................... Urban 16 ........................................................ Woman . . . . . . . . . . . . . . . . . . . . . .. . . .. . . .... . .. . . .. . . .. . . . . . . . . . . . . . . .. . Parents' SEI x parents' education .........................

........ . . .

Parents' education X woman

...... .... . . ...................

Parents' education X level (2, 3, 4 = 1) .... . . . . . . . . . .....

Parent's SEI X level (1, 2) (4) (3) ................... ......
Urban 16 X level (1) (3) (2) (4) ...............................
Woman by level (3, 4) (2) (1) ................................
Parents' education X cohort (3 = 1)for L1 ....... .......
Parents' SEI x cohort (3 = 1) for L1 .......... . ... . . . . . . .
Urban 16 X cohort (3 = 1)for L2 ..........................
Woman X cohort (1-3) (4) (5) (6) for L2 ..................
Parents' education x cohort (1 = 1) for L3 ..............
Parents' SEI x cohort (3) (2, 4, 5, 6) (1) for L3 .........
Woman X cohort (2, 3) (1) (6, 5) (4) for L3 ...............
Parents' education X cohort (1, 4-6 = 1)for L4 .......
Parents' SEI X cohort (6) (1, 5) (2, 4) (3) for L4 ........
Woman x cohort (1, 2, 4) (5) (3, 6) for L4 ...............
Constant ... . .. . . . . . . ... . . . . .... . . . . . . . . . . . . . . . . . . . . . .. . ... . .. . . . .

NOTE-xi(-2 LL) = 4,747; df = 4,595.

* Recodings of level and cohort for interaction terms are presented in abbreviated fashion: For recod- ings in dummy form (e.g., parents' education X level) original values set equal to "1" are given in parentheses, remaining values are set at zero. For recodings in ordinal form (e.g., parent's SEI X level) initial categories recoded with the same value are regrouped within parentheses, which are arranged in ascending order of the new values, beginning with "1." See table A1 and appendix text for more detail.

may be significantly different from the maximum or minimum baseline effect. But this specification of three-way interactions is not parsimonious and leaves no way of ascertaining the significance of the differences among nonbaseline categories without a series of two-way t-tests that greatly complicate the analysis. To correct these flaws, we undertook a third stage, where we respecified each significant level x cohort interac- tion as a single, ordinal (sometimes dummy) multiplier of the relevant background effect (model 10). These "linearized" interaction terms are level-specific recodings of the ordinal cohort variable based on the ob- served pattern among the dummy interaction terms. The recodings are

given in the "Comments" column. For example, inspection of the dummy interaction terms for parents' education x cohort at level 1 suggested a significant difference in the size of the effect between the third cohort and the first two, but not between the first two. The interaction is therefore respecified as a single variable (model lo), which equals par- ents' education for level 1, cohort 3, zero for all other levels and cohorts. Similarly, the dummy interaction terms for sex x cohort at level 2 (model 11) suggested that, while for each cohort the effect differed significantly from the baseline value (maximum at cohort 6), there was also a pattern of differences across some, but not all, of the remaining cohorts. A linear specification of this pattern was significant (model 12). We were able to fit statistically significant "constrained" specifications (either ordinal or dummy in form) for each three-way interaction that was significant in the dummy interaction specification. In every case, the one degree of freedom specification resulted in little or no loss of fit (cf. models 9 and 10). This multistaged "identify and constrain" procedure thus provides not only a rigorous test for changes in the magnitudes of background effects across cohorts at specific transitions; it also provides the most parsimonious specifications of these changes.

In the final three steps we constrain in a similar fashion the two-way background x level interactions (H),the two-way level x cohort inter- action (I)-which is respecified as a set of four ordinal recodings of the cohort variable, one corresponding to each level-and the main effect of level (J).The result is our preferred model, presented in table A2, which we used to calculate the magnitudes of the background effects and con- stants for each level and cohort combination (table 7).

REFERENCES

Avis, George. 1983. "Access to Higher Education in the Soviet Union." Pp. 199-239 in Soviet Education in the 1980s, edited by Janusz Tomiak. New York: St. Martin's.

Blossfeld, Hans-Peter, and Yossi Shavit. 1993. "Persisting Barriers: Changes in Edu- cational Opportunity in Thirteen Countries." Pp. 1-23 in Persistent Inequality: Educational Attainment in Thirteen Countries, edited by Yossi Shavit and Hans- Peter Blossfeld. Boulder, Colo.: Westview.

Cobalti, Antonio. 1990. "Schooling Inequalities in Italy: Trends over Time." European Sociological Review 6:199-214. Connor, Walter D. 1991. The Accidental Proletariat. Princeton, N.J.: Princeton Uni-

versity Press. Djilas, Milovan. 1965. The New Class. New York: Praeger. Dobson, Richard B. 1977. "Social Status and Inequality of Access to Higher Educa-

tion in the USSR." Pp. 254-74 in Power and Ideology in Education, edited by Jerome Karabel and A. H. Halsey. New York: Oxford University Press. . 1980. "Education and Opportunity." Pp. 115-37 in Contemporary Soviet

Society: Sociological Perspectives, edited by Jerry G. Pankhurst and Michael Paul Sacks. New York: Praeger.

Duncan, Otis Dudley. 1961. "A Socio-Economic Index for All Occupations." Pp. 109-38 in Occupation and Social Status, edited by A. J. Reiss. New York: Free Press.

Dunstan, John. 1983. "Now They Are Six." Pp. 62-85 in Soviet Education in the 1980s, edited by Janusz Tomiak. New York: St. Martin's. Erikson, Robert, and John H. Goldthorpe. 1992. The Constant Flux. Oxford: Clarendon Press. Fitzpatrick, Sheila. 1979. Education and Social Mobility in the Soviet Union, 1921- 1934. Cambridge: Cambridge University Press. Gel'fand, V. S. 1992. Naselenie SSSR za 50 Let (1941-1990) (The population of the USSR in the course of 50 years). Perm: Permskogo Universiteta.

Gosudarstvennyi Komitet SSSR po Statistike (GKSPS). 1989. Narodnoe Obrazovanie i Kul'tura v SSSR: Statisticheskii Sbornik (Public education and culture in the USSR: Statistical collection). Moscow: Financy i Statistika.

Gradolph, Rebecca S. 1993. "Stratification in a Non-Market Setting: Determinants of Standard of Living in Russia, 1991." Paper presented at the annual meeting of the American Sociological Association, Miami, Florida, August.

Heyns, Barbara, and Ireneusz Bialecki. 1993. "Educational Inequalities in Postwar Poland." Pp. 303-35 in Persistent Inequality: Educational Attainment in Thirteen Countries, edited by Yossi Shavit and Hans-Peter Blossfeld. Boulder, Colo.: Westview.

Hout, Michael, Adrian E. Raftery, and Eleanor 0. Bell. 1993. "Making the Grade: Educational Stratification in the United States, 1925-1989." Pp. 25-49 in Persistent Inequality: Educational Attainment in Thirteen Countries, edited by Yossi Shavit and Hans-Peter Blossfeld. Boulder, Colo.: Westview.

Hout, Michael, Erik Olin Wright, and Martin SBnchez-Jankowski. 1992. Class Struc- ture and Consciousness in Russia(MRDF). Berkeley, Calif.: Survey Research Center. Inkeles, Alex, and Raymond A. Bauer. 1959. The Soviet Citizen: Daily Life in a

Totalitarian Society. Cambridge, Mass.: Harvard University Press. Lapidus, Gail Warshofsky. 1983. "Social Trends." Pp. 186-249 in After Brezhnev, edited by Robert F. Byrnes. Bloomington: Indiana University Press. Lewin, Moshe. 1988. The Gorbachev Phenomenon. Berkeley and Los Angeles: Uni- versity of California Press. Mare, Robert D. 1980. "Social Background and Educational Continuation Deci- sions." Journal of the American Statistical Association 75:295-305. . 1981. "Change and Stability in Educational Stratificiation." American Socio- logical Review 46:72-87.

,1993. "Educational Stratification on Observed and Unobserved Components of Family Background." Pp. 351-76 in Persistent Inequality: Educational Attain- ment in Thirteen Countries, edited by Yossi Shavit and Hans-Peter Blossfeld. Boul- der, Colo.: Westview.

Mateju, Petr. 1993. "Who Won and Who Lost in a Socialist Redistribution in Czecho- slovakia?" Pp. 251-72 in Persistent Inequality: Educational Attainment in Thir- teen Countries, edited by Yossi Shavit and Hans-Peter Blossfeld. Boulder, Colo.: Westview.

Matthews, Mervyn. 1972. Class and Society in Soviet Russia. New York: Walker & Company. . 1982. Education in the Soviet Union. London: George Allen & Unwin. . 1983. "Long-Term Trends in Soviet Education." Pp. 1-23 in Soviet Educa- tion in the 1980s, edited by Janusz Tomiak. New York: St. Martin's.

Medvedev, Roy A., and Zhores A. Medvedev. 1978. Khrushchev: The Years in Power. New York: Norton.

Millar, James. 1987. "History, Methods, and the Problem of Bias." Pp. 3-30 in

Politics, Work, and Daily Life in the USSR: A Survey of Fomzer Soviet Citizens,

edited by James Millar. Cambridge: Cambridge University Press. Petersen, Trond. 1985. "A Comment on Presenting Results from Logit and Probit Models." American Sociological Review 50: 130-3 1. Prokof'ev, M. A,, ed. 1985. Narodnoe Obrazovanie v SSSR (Public education in the USSR). Moscow: Pedagogika.

Raftery, Adrian E., and Michael Hout. 1993. "Maximally Maintained Inequality: Expansion, Reform, and Opportunity in Irish Education, 1921-75." Sociology of Education 66:41-62.

Rutkevich, M. N. 1977. lntelligentsiia v Razvitom Sotsialisticheskom Obshchestve (The intelligentsia in developed socialist society). Moscow: Izdatel'stvo Politicheskoi Literatury.

. 1984. "Reforma Obrazovaniia, Potrebnosti Obshchestva, Molodezh' " (The reform of education, the demands of society, youth). Sotsiologicheskie Issledovaniia 4119-28.

Rutkevich, M. N., and F. R. Fillipov, eds. 1978. Vysshaia Shkola kak Faktor Izmene- niia Sotsial'noi Struktury Razvitogo Sotsialisticheskogo Obshchestva (Higher edu- cation as a factor of change of the social structure of developed socialist society). Moscow: Nauka.

Samoilova, E. E. 1978. Naselenie i Obrazovanie (The population and education). Moscow: Statistika. Shubkin, V. N. 1965. "Molodezh' Vstupaet v Zhizn'" (Youth sets out in life). Voprosy Filosoji 19:57-70. , ed. 1984. Trudiashchaiasia Molodezh': Obrazovanie, Professiia, Mobili'nost' (Working youth: Education, profession, mobility). Moscow: Nauka.

Shavit, Yossi, and Karin Westerbeek. 1995. "Educational Reforms and Educational Stratification in Italy." Paper presented at the annual meeting of the International Sociological Association Research Committee on Stratification and Mobility, Zu- rich, May 25-27.

Simkus, Albert, and Rudolf Andarka. 1982. "Inequalities in Educational Attainment in Hungary, 1923-1973." American Sociological Review 47:740-51.

Szelenyi, Szonja, and Karen Aschaffenburg. 1993. "Inequalities in Educational Op- portunity in Hungary." Pp. 273-302 in Persistent Inequality: Educational Attain- ment in Thirteen Countries, edited by Yossi Shavit and Hans-Peter Blossfeld. Boul- der, Colo.: Westview.

Treiman, Donald J. 1977. Occupational Prestige in Comparative Perspective. New York: Academic Press.

Tsentral'noe Statisticheskoe Upravlenie (TsSU). 1966. Narodnoe Khoziaistvo SSSR v 1965 Godu: Statisticheskii Ezhegodnik (The economy of the USSR in 1965: Statis- tical annual). Moscow: Statistika. . 1977. Narodnoe Obrazovanie, Nauka, i Kul'tura v SSSR: Statisticheskii Sbornik (Public education, science, and culture in the USSR: Statistical collection). Moscow: Statistika. . 1981. Narodnoe Khoziaistvo SSSR v 1980 Godu: Statisticheskii Ezhegodnik (The economy of the USSR in 1980: Statistical annual). Moscow: Finansy i Statistika. . 1982. Narodnoe Khoziaistvo SSSR 1922-1982: Iubileinyi Statisticheskii Ez- hegodnik (Anniversary statistical annual). Moscow: Finansy i Statistika. , 1983. Narodnoe Khoziaistvo SSSR v 1982 Godu: Statisticheskii Ezhegodnik (The economy of the USSR in 1982: Statistical annual). Moscow: Finansy i Sta- tistika. . 1985. Narodnoe Khoziaistvo SSSR v 1984 Godu: Statisticheskii Ezhegodnik (The economy of the USSR in 1980: Statistical annual). Moscow: Finansy i Sta- tistika.

Vasil'eva, E. K. 1973. Sotsial'no-Professional'nyi Uroven' Gorodskoi Molokezhi. (The social-professional level of urban youth). Leningrad: Izdatel'stvo Leningradskogo Universiteta. . 1978. Sotsial'no-Ekonomicheskaia Struktura Naseleniia SSSR (The social- economic structure of the population of the USSR). Moscow: Statistika.

Vasil'eva, E. K., Eliseeva, I. I., Kashina, 0. N., and Laptev, V. I. 1985. Dinamika Naseleniia SSSR 1960-1980 gg (The dynamic of the population of the USSR, 1960- 80). Moscow: Finansy i Statistika.

Wright, Erik Olin. 1986. "The Comparative Project on Class Structure and Class Consciousness: An Overview." Acta Sociologica 32:3-22.

Yanowitch, Murray. 1977. Social and Economic Inequality in the Soviet Union: Six Studies. White Plains, N.Y.: M. E. Sharpe.

Zaslavsky, Victor. 1982. The Neo-Stalinist State. Armonk, N.Y.: M. E. Sharpe.

Comments
  • Recommend Us