Download Intergenerational Inequality: A Sociological Perspective

Journal of Economic Perspectives—Volume 16, Number 3—Summer 2002—Pages 31– 44
Intergenerational Inequality:
A Sociological Perspective
Robert Erikson and John H. Goldthorpe
W
hen economists are concerned with the inheritance of inequality, they
typically focus on the intergenerational transmission of income or
wealth. In contrast, sociologists are more likely to analyze intergenerational mobility between (and immobility in) different class positions.
One immediate consequence is that while economists usually work with intergenerational correlations of income or wealth treated as continuous variables, sociologists more often work with intergenerational patterns of association between class
positions that are treated categorically. The standard data array takes the form of
a contingency table in which class “origin” is crossed with class “destination.” The
former variable is usually indexed by class of father or other household “head” at
the time of a child’s—that is, the survey respondent’s—adolescence; the latter
variable, by the child’s (respondent’s) present class or class at time of inquiry. The
reliability with which father’s class can be established in survey interviews has been
subject to a good deal of investigation with reasonably satisfactory results (Hope,
Schwartz and Graham, 1986; Breen and Jonsson, 1997) and is in any event more
accessible than father’s income. The child’s, or respondent’s, class at time of
inquiry is not of course Ž xed, and signiŽ cant worklife or intragenerational mobility
does occur. But it is known that the frequency of such mobility falls off rather
sharply after around age 35, and intergenerational mobility tables are therefore
sometimes restricted to respondents over this age.
The difference in approach between sociologists and economists is not, however, absolute. Sociologists have studied intergenerational social mobility on the
basis of correlations of parents’ and children’s “socioeconomic status” scores (Blau
and Duncan, 1967; Featherman and Hauser, 1978). Following the pioneering work
y Robert Erikson is Professor of Sociology, Swedish Institute for Social Research, Stockholm,
Sweden. John H. Goldthorpe is OfŽcial Fellow, NufŽeld College, Oxford University, Oxford,
United Kingdom.
32
Journal of Economic Perspectives
of Atkinson, Maynard and Trinder (1983), several other economists have of late
adopted contingency table methods in analyzing intergenerational income mobility, using income quantile groups as their categories (for example, Dearden,
Machin and Reed, 1997; Hertz, 2001). Björklund and Jäntti (2000, p. 24) have in
fact recently called for further work of this kind, applying “more  exible measures
of association” in place of correlation (or regression) coefŽ cients, on the following
grounds: “There are, a priori, no good reasons to believe that the association
between fathers’ and sons’ incomes is the same throughout, over e.g. the income
range of fathers.” Atkinson, Maynard and Trinder (1983, p. 180) also make the
important point that a contingency table approach is able to bring out important
asymmetries in mobility patterns—for instance, long-range upward movements
being offset by more gradual “trickling down processes”—that correlation or regression coefŽ cients cannot capture.
Operationalizing Class
If the inheritance of inequality is treated in terms of class mobility, an obviously
crucial question that arises is that of how the concept of class is to be understood
and made operational. As an initial point here, we would distinguish the concept
of class from that of “socioeconomic status,” which has been widely used in
American social science—and sometimes as the basis for constructing occupational
categories rather than an interval-level scale.
We would regard class positions as being determined by employment relations
(for more detailed statements, see Erikson and Goldthorpe, 1992, chapter 2;
Goldthorpe, 2000, chapter 10). Thus, a primary division is that among employers,
self-employed workers and employees. However, employees, who make up the large
majority of the workforces of modern societies, require further differentiation,
which can, we believe, be provided in a theoretically consistent way by reference to
the mode of regulation of their employment. The problems employers face,
ultimately on account of the essential incompleteness of employment contracts and
more immediately in regard to work monitoring and human asset speciŽ city, lead
to contracts of signiŽ cantly differing form being offered to employees who are
engaged to carry out different kinds of work. These range from the “labor contract,” a simple recurrent spot contract for the purchase of a quantity of labor on
the basis of piece or time rates, via various modiŽ ed or mixed forms, through to the
“service relationship,” an exchange of a longer term and more diffuse kind in which
compensation for service to the employing organization involves important prospective elements, such as salary increments, expectations of continuity of employment or at least of employability and promotion prospects and career opportunities.
A class schema, using employment status and occupation as indicators of
employment relations, can then be drawn up on the general lines shown in
Table 1. Versions of this schema have in fact been widely applied in studies of
intergenerational mobility, and in other sociological research, since the 1980s, and
Robert Erikson and John H. Goldthorpe
33
Table 1
The Class Schema
Class
I.
II.
IIIa.
IIIb.
IVa.
IVb.
IVc.
V.
VI.
VIIa.
VIIb.
Occupational Grouping/Employment Status
Regulation of Employment
Professionals, administrators and managers, higher-grade
Professionals, administrators and managers, lower-grade;
technicians, higher-grade
Routine nonmanual employees, higher grade
Routine nonmanual employees, lower-grade
Small employers
Self-employed workers (nonprofessional)
Farmers
Technicians, lower grade; supervisors of manual workers
Skilled manual workers
Nonskilled manual workers (other than in agriculture)
Agricultural workers
service relationship
service relationship (modiŽ ed)
mixed
labor contract (modiŽ ed)
—
—
—
mixed
labor contract (modiŽ ed)
labor contract
labor contract
the schema is now attracting increasing interest from national and international
statistical agencies as a basis for ofŽ cial social classiŽ cations.1
It is important to note that since the schema aims to capture qualitative
differences in employment relations, the classes distinguished are not consistently
ordered according to some inherent hierarchical principle, such as, say, the “general desirability” of the positions they comprise. Their members may be relatively
advantaged or disadvantaged in different ways. Thus, routine nonmanual employees in Class IIIa may have lower average incomes than do small shopkeepers in Class
IVb or technicians and foremen in Class V, but more stable levels of income than
the former and better chances of promotion than the latter.
However, so far as overall economic status is concerned, individuals in Classes
I and II, representing the “service class” or “salariat,” could in fact be regarded as
generally advantaged over individuals in Classes IIIb, VI and VIIa and VIIb, representing the working class, in at least three ways that follow directly from the mode
of regulation of their employment and that, together, we would see as being of at
least comparable importance to current income alone. Members of the salariat are
advantaged over members of the working class in that they experience i) greater
long-term security of income through being less likely to lose their jobs and to
become unemployed; ii) less short-term (week-to-week or month-to-month)  uctu1
The schema has become known as either the EGP (Erikson, Goldthorpe and Portocarero) or CASMIN
(Comparative Analysis of Social Mobility in Industrial Nations) schema, the latter being the name of a
project directed from the University of Mannheim by Walter Müller and John Goldthorpe from
1984 –1990. Since 2000, a new instantiation of the schema has in fact been adopted as the ofŽ cial British
social classiŽ cation under the (somewhat unfortunate) name of the National Statistics Socio-Economic
ClassiŽ cation, and active consideration is presently been given to the use of the schema as the basis for
a general European Union social classiŽ cation. One valuable spin-off of such ofŽ cial interest is that
resources have been made available to test the validity of the schema: that is, the extent to which, as
implemented via information on employment status and occupation, it does in fact capture the kinds of
differences in employment relations that it is conceptually supposed to capture. The results of such tests
have been generally encouraging (for Britain, see Rose and O’Reilly, 1997, 1998).
34
Journal of Economic Perspectives
ation of income through being less dependent on piece rates, shift premiums,
overtime payments and less exposed to loss of pay on account of absence or illness;
and iii) better prospects of steadily increasing income over the life course—into
their 50s rather than their 30s—through having employment contracts that are
conducive to an upward-sloping age-earnings proŽ le (Lazear, 1995) with in turn
better prospects for the accumulation of wealth.2
One last point that needs to be made here is the following. Sociologists are
interested in class and class mobility not only as dependent, but also as independent, or explanatory, variables: ones that can be set in competition with other
variables, including income and income mobility, in their capacity to account for
variation in a wide range of life chances (say, health) and life choices (like political
partisanship). Empirically, class effects on such outcomes tend to persist even when
income is controlled. It is possible that class, as operationalized on the lines
indicated above, serves as a good proxy for permanent income. In addition,
though, we believe that its explanatory power stems from the fact that it is able to
capture important aspects of the social relations of economic life.
Analyzing Class Mobility
To measure the association between class origins and class destinations, sociologists most often use the odds ratio. For the simplest possible mobility table, that
with only two classes of origin and destination, say class 1 and class 2, the one
calculable odds ratio is given by
odds ratio 5
f11 /f12
,
f21 /f22
where f 1 1 is the frequency in cell (1,1), that referring to immobility within class 1,
f 1 2 is the frequency in cell (1,2), that referring to mobility from class 1 to class 2,
and so on. So in this case, the odds ratio gives the chances for an individual
originating in class 1 being found in class 1 rather than in class 2, relative to the
same chances for an individual originating in class 2. An odds ratio with the value
of 1 thus indicates the absence of association between origins and destinations (or
their statistical independence).
The odds ratio is attractive because it is a “margin insensitive” measure of
association (Bishop, Fienberg and Holland, 1975), which means that it is invariant
to the multiplication of any row or column of a contingency table by a (nonzero)
constant. In an intergenerational mobility table, what might be called the gross
association between origin class and destination class will be conditioned by differences in the overall distributions of these variables—the marginal distributions of
2
Moreover, even insofar as the classes cannot be perfectly ordered, we do not believe that this makes the
question of mobility between them irrelevant to issues of equality of opportunity and social justice. For
discussion of this point, see Marshall, Swift and Roberts (1997, appendix E).
Intergenerational Inequality: A Sociological Perspective
35
the table—that re ect changes in the proportions of individuals found in different
class positions across generations. For example, in the course of economic development, fewer children than fathers will become farmers, but more will become
managerial and professional employees. Thus, some intergenerational mobility will
of necessity occur in the form of out ow from the class of farmers and in ow to that
of managers and professionals. For many purposes, this mobility will be of interest
in itself.3 But odds ratios provide a measure of the association of origins and
destinations that is net of the effects of such class structural change and that can
therefore remain constant even when such change is extensive or, conversely, that
can alter even when such change is absent.4
In mobility tables with more than two categories, more than one odds ratio will
be calculable—in fact, one for every possible pair of origin categories considered in
relation to every pair of destination categories. Thus, the number of odds ratios
implicit in a square mobility table with k categories will be given by [(k 2 2 k) 2 ]/4,
although it can be shown that a “basic set” of (k 2 1) 2 odds ratios can be speciŽ ed
that will determine the remainder (Goodman, 1979).
The full set of odds ratios implicit in a mobility table is taken to constitute the
“endogenous mobility regime” or, alternatively, the underlying “pattern of social
 uidity.” For testing substantive hypotheses about endogenous mobility regimes,
loglinear or logmultiplicative models, the parameters of which are odds ratios or
functions of odds ratios, are chie y used (Hout, 1983; Erikson and Goldthorpe,
1992). Such models can serve to represent particular hypotheses—for example,
that odds ratios are unchanged in a society over a period of time or are the same
across a number of societies, or that they change or differ in particular ways—and
the Ž t of selected models to the actual data of mobility tables can then be assessed
via standard statistical procedures.
When the main empirical features of endogenous mobility regimes have been
established and attention shifts to the actual processes of mobility that underlie
these regimes, loglinear models for the grouped data of mobility tables can be
rewritten as multinomial logistic regression models for individual-level data (Logan, 1983; Breen, 1994). Class position is the dependent variable, and class origin
Ž gures as one of a set of independent variables also including, for example,
measures of individual IQ, effort, educational attainment and other relevant variables. It thus becomes possible to examine how far the inclusion of such variables
in the analysis leads to the dependence of class of destination on class of origin
3
Sociologists do in fact analyze out ow and in ow rates in simple percentage terms: that is, by
considering the percentage distribution of all individuals of a given class of origin across all destination
classes or, conversely, the distribution of all individuals in a given class of destination across all origin
classes. But it is important to distinguish these “absolute” mobility rates from the “relative” rates
captured by odds ratios.
4
The motivation here could be thought of as somewhat similar to that behind the Galton measure of
intergenerational correlation in income, which normalizes for changes in the mean and standard
deviation of the income distribution over time. When intergenerational income mobility is studied via
a contingency table approach, using income quantile groups as the categories, the problem of controlling differing marginal distributions obviously does not arise.
36
Journal of Economic Perspectives
being reduced or, in other words, how far mobility regimes may be thought of as
meritocratic.
Results
Measuring intergenerational economic mobility through correlation or regression coefŽ cients, as economists most typically do, leads to results that can be very
concisely expressed. Sociologists’ results relating to the mobility regimes that
operate within class structures are more complex, since it is supposed that the
association between class origins and destinations may vary in strength across the
component cells of the mobility table—that is, from one intergenerational transition to another. This supposition turns out in fact to be fully warranted, so what is
lost in parsimony is gained in realism. With, then, some degree of simpliŽ cation,
the main Ž ndings from recent sociological research could be summarized as
follows.
First, in all modern societies, signiŽ cant associations between class of origin
and class of destination prevail. For men, at least, there is a broad similarity in
endogenous mobility regimes across societies.5 This represents an interesting parallel with the cross-national similarities in estimates of the extent of intergenerational income mobility noted by Björklund and Jännti (2000, p. 4, n. 4). Some
nationally speciŽ c variation in mobility regimes is also apparent; but, within this
variation, differences in the overall level of the origin-destination association, as
opposed to its pattern, is only one—in fact rather minor— element. Consequently,
no nation or nations stand out as showing decisively more social  uidity or
openness than the rest. The idea of American exceptionalism in this regard is a
myth (Erikson and Goldthorpe, 1985, 1992, chapter 9). Among economically
advanced societies, Sweden appears as the most open, but has still to be seen as
marking one end of a quite limited range of differentiation rather than being in
any sense sui generis.
Second, the main features of the cross-national commonality in mobility
regimes are the following. There is a general propensity for intergenerational class
immobility through the operation of what might be called class-speciŽ c inheritance
effects. These effects are relatively strong within Classes I and II, the salariat, and
Classes IVa and IVb, small employers and self-employed workers, and strongest of
all within Class IVc, that of farmers. In addition, there is a general propensity for
mobility to be reduced by “hierarchy” effects deriving from the overall advantages
and disadvantages associated with different class positions (as discussed above).
These effects especially operate between Classes I and II, on the one hand, and
Classes VIIa and VIIb, the nonskilled division of the working class, on the other. To
5
Although it has not so far been demonstrated, we would think it highly probable that such a result will
hold for women, also. The mobility regimes for men and women within particular nations have
repeatedly been shown to differ little—apart from the fact that odds ratios for women overall tend to be
slightly lower than for men.
Robert Erikson and John H. Goldthorpe
37
give some indication of the importance of class inheritance and hierarchy effects
together, the odds of a man originating in the salariat being himself found in the
salariat rather than in the nonskilled working class, relative to the same odds for a
man originating in the nonskilled working class, would, across modern societies, be
of the order of 15:1.6
Third, within particular societies, mobility regimes show a high degree of
constancy over time, and in some cases, such as Great Britain (Goldthorpe, Payne
and Llewellyn, 1987; Goldthorpe and Mills, forthcoming) or Japan (Ishida, 1995),
for periods extending back to the Ž rst half of the twentieth century. Loglinear
models that postulate no change in odds ratios reproduce the empirical data
remarkably well, usually misclassifying less than 5 percent of all individuals in the
mobility tables analyzed. In societies where trends in the overall level of  uidity can
be discerned, these are more often in the direction of increasing  uidity— odds
ratios moving generally closer to 1—than of decreasing  uidity. But such trends, as
well as being slight, would seem more often to be episodic as, say, in the United
States (Erikson and Goldthorpe, 1992, chapter 9) or Sweden ( Jonsson, forthcoming) than sustained, as in France (Vallet, forthcoming). The idea of a worldwide
and secular movement toward greater societal openness has been mooted (Ganzeboom, Luijkx and Treiman, 1989), but this is scarcely borne out by the evidence so
far accumulated. An alternative hypothesis is that the level of social  uidity will
inversely relate to the degree of economic inequality between classes (Erikson and
Goldthorpe, 1992, chapter 11; Goldthorpe, 2000, chapter 11). This hypothesis does
in fact Ž nd support from one of the very few studies of trends in intergenerational
income mobility so far carried out (Blanden et al., 2001).
Fourth, educational attainment is a major—probably the major—mediating
factor in class mobility (Ishida, Müller and Ridge, 1995; Marshall, Swift and Roberts, 1997), although this is more apparent when education is measured by highest
level of qualiŽ cation achieved (academic or vocational) rather than by number of
years of education completed, as is the usual American practice.7 In the British
case, the tradition of birth cohort studies provides data sets that allow for the effects
of IQ and of effort (in the sense at least of academic motivation as measured on
standard psychological scales) to be reliably compared with that of education. The
latter proves to be clearly stronger, and further, the effects of IQ and effort appear
to operate largely via educational attainment, at all events so far as the mediation
of early life mobility (up to around age 30) is concerned (Breen and Goldthorpe,
1999).
Fifth, modern societies are not meritocracies in the sense that, once educational qualiŽ cations and other “merit” variables are controlled, class of destination
6
Sector effects, operating between the classes of farmers and agricultural workers and the rest, reduce
propensities for mobility still more strongly than do hierarchy effects and were indeed a major feature
of the mobility regimes of many modern societies even up to the middle decades of the twentieth
century, although they are by now of much reduced importance overall.
7
The standardized measure that is chie y used here is the CASMIN educational classiŽ cation (König,
Lüttinger and Müller, 1988; Brauns and Steinmann, 1999).
38
Journal of Economic Perspectives
is no longer dependent on class of origin. To the contrary, a signiŽ cant and often
substantial dependence remains (Marshall, Swift and Roberts, 1997; Breen and
Goldthorpe, 1999, 2001). In some cases, for example, Sweden, the persisting effect
of class origins has been shown to extend to income, also (Erikson and Jonsson,
1998). Thus, as Breen and Goldthorpe have put it (1999, p. 21): “Children of
disadvantaged class origins have to display far more merit [as indicated by educational
attainment or by IQ and effort] than do children of more advantaged origins in
order to attain similar class positions.”
Sixth, the mediating role of education varies signiŽ cantly in its importance
from one type of intergenerational transition to another. Thus, educational qualiŽ cations have been shown to be of no importance at all in mediating intergenerational immobility (for which there is a high propensity) within any of the subdivisions of Class IV: that is, among small employers, self-employed workers or farmers
(Ishida, Müller and Ridge, 1995). What appears crucial here is the direct intergenerational transmission of “going concerns” or of economic capital in other
forms—a factor that Bowles and Gintis in this issue also Ž nd important for the
intergenerational income correlation. Further, several studies now in progress
suggest that educational qualiŽ cations are of greater importance in “long-range”
upward mobility—as, say, from working-class origins into the salariat—than they are
in intergenerational immobility within the salariat (for example, Guzzo, 2002).
Here in particular, the advantages of a disaggregated, contingency table approach
can be seen. Effects that bear on mobility from speciŽ c origins to speciŽ c destinations can show up in a way that would not be possible if the same regression rules
were simply assumed to apply across the board.
Problems and Prospects
There are at least two outstanding problems, in part related, that we would
recognize, along with Bowles and Gintis.
The Ž rst is what Bowles and Gintis call the “black box” problem. More than
half of their preferred coefŽ cient for the intergenerational transmission of
income remains unaccounted for even when all conventional explanatory variables are included in the analysis. We in fact believe that the situation is less
clear—and possibly worse—than Bowles and Gintis suppose, since we would
question whether their efforts to disentangle the relative importance of heredity
and environment and also of direct transfers of assets warrant such precise
estimates as they venture.
For example, a key assumption underlying their estimates is that the
difference in the intergenerational correlations of earnings or income between
identical and fraternal twins will be indicative of the importance of genetic
effects, since twins of both kinds share the same environmental conditions.
Bowles and Gintis do acknowledge that environments may in fact be somewhat
more similar for identical than for fraternal twins; but, we would argue, there is
Intergenerational Inequality: A Sociological Perspective
39
more to it than that. The often symbiotic relation between identical twins will
lead them to strive to keep together, while fraternal twins may actively seek to
distinguish themselves from each other. Thus, identical twins may follow the
same education, not because their genetic make-up leads them both to choose
it, but simply because they do not want to be separated; and the same wish may
likewise in uence their family life, place of residence and other factors. 8 If one
identical twin gets a job offer, he or she may not take it in order to avoid
separation, geographic and/or social, from the other twin. Now the more equal
conditions, including in income, that are thus likely to result among identical
twins are certainly linked to the fact that they have the same genes. But how can
the effect in question be generalized to the population at large?9
However, we would not wish to claim that sociologists are any better placed
than economists as regards to the black box problem—that is, as it arises in their
understanding of the association between class of origin and class of destination.
Indeed, in one respect, the problem appears even more embarrassing for sociologists. Of late, a number of studies have indicated that the part that is played in
mediating intergenerational class mobility by educational attainment— our most
important conventional explanatory variable—is if anything, declining (for Sweden,
see Jonsson, 1992; for Britain, Breen and Goldthorpe, 2001; and Goldthorpe and
Mills, forthcoming; for France, Vallet, forthcoming; and for Ireland, Whelan and
Layte, 2002).1 0
We are, then, led strongly to agree with Bowles and Gintis that if the black box
problem is to be overcome, we will need to examine the possible importance in the
inheritance of inequality of a wider range of individual attributes than has so far
been considered and, in particular, to be less exclusively concerned with cognitive
abilities and with skills, at least as usually understood. We Ž nd much of interest in
their suggestions regarding “group membership” effects on economic success and
what they elsewhere elaborate as the effects of “incentive enhancing preferences”
(Bowles and Gintis, 2001). However, we would also want to take further than Bowles
and Gintis a more direct, demand-side approach: that is, asking just what are the
attributes of potential employees that employers are looking for and how these
desiderata might be changing.
In modern economies, there would appear to be, at virtually all levels of the
8
Ashenfelter and Krueger (1994, p. 1159), on whose work Bowles and Gintis chie y rely, do in fact note
that identical twins far more often studied together and were also somewhat less often married than
fraternal twins.
9
It could further be argued that models such as that used by Bowles and Gintis, in which the effects of
heredity and environment are treated as simply additive, are misspeciŽ ed. Evidence is now available
(Maccoby, 2000) of interaction effects between hereditary factors and parental childbearing regimes,
analogous to those that have for long been demonstrated between genes and environment in the case
of various plants, fruit  ies and so on.
10
When such results are reported, economists usually react with some surprise or even skepticism on the
grounds that earnings returns to education are tending to increase. However, apart from the fact that
there is no necessary inconsistency here, for at least some of the countries referred to in the text, the
evidence of such increasing returns is not all that compelling. For example, for France, see Baudelot and
Glaude (1989); and for Britain, see Chevalier and Walker (2001).
40
Journal of Economic Perspectives
class structure, an increasing range of what have been referred to as “peopleprocessing” or, somewhat more cynically, “high-touch” occupations— obvious
growth areas being in the leisure, entertainment and hospitality industries, in
public relations and the media and in the personalized selling of high-value goods
and services. In occupations of this kind, employers may well view cognitive abilities
(above some threshold level) and conventional skills as being of less relevance than
such attributes as physical appearance, dress sense, accent, self-presentation, lifestyle and savoir faire, along with related “social” or “interpersonal” skills.1 1 For
Britain at least, there is some amount of supporting evidence for this speculation
from preliminary results from both the content analysis of job advertisements and
organizational case studies ( Jackson, 2001; Jackson, Goldthorpe and Mills, 2002;
Warhurst and Nickson, 2001).
Insofar as such a shift in the pattern of employers’ requirements is in train, an
important implication follows. Increasing economic value now attaches to individual attributes of a kind less likely to be achieved through the educational system
than ascribed through processes of socialization within generally more advantaged
families and communities. In turn, a possible explanation is indicated for both the
(apparently widening) gap that arises if we seek to account for the patterning of
mobility regimes solely in terms of education and for the fact that education plays
a greater part in accounting for upward mobility into more advantaged classes than
for intergenerational immobility within these classes.12 Men and women with
advantaged class backgrounds acquire, more or less as a matter of course, attributes
that help them maintain their position even if their educational attainments are
only modest.
The second problem that we share with Bowles and Gintis is raised by the
rather limited success of public policy aimed at reducing intergenerational inequality and, more speciŽ cally, that of explaining why, even if education plays a lesser
role in mediating such inequality than is often supposed, the massive expansion
and often radical reform of educational institutions in modern societies has not
had a more evident egalitarian outcome.
In addressing this problem, we would start from two further sets of empirical
results and a conceptual distinction. The Ž rst set of results are those that show that
in most societies, class differentials in educational attainment have in fact proved
highly resistant to change, even across decades of educational expansion and
reform, if these differentials are understood and modeled in terms of the relative
odds of children of different class origins making or not making the successive
“transitions” by which educational careers are deŽ ned (Blossfeld and Shavit, 1993;
Mare, 1980, 1981; Breen and Jonsson, 2000).
11
On this issue, we would add that the repeated references of Bowles and Gintes to “good looks” as
genetically transmitted seems to miss the point that “good looks” are, at least in some large part, a social
construct, modeled on the example of superior classes or status groups.
12
Other alternative or complementary explanations for this development could of course be proposed:
for example, that, as a consequence of the expansion of educational provision and of the growing
numbers of individuals who have qualiŽ cations of some kind at all levels, the value of education in both
signalling by potential employees and in screening by employers is reduced.
Robert Erikson and John H. Goldthorpe
41
The conceptual distinction, due to Boudon (1974), is between the “primary”
and the “secondary” effects that are at work in this regard. Primary effects are those,
whether genetic or cultural, that create class differentials in “demonstrated ability”
early in children’s educational careers and in this way condition the options
subsequently open to them. Secondary effects are those that later operate through
the choices that children, together perhaps with their parents, actually make
among the options they have available.
The second set of empirical results then serves to show that secondary
effects do in fact play a signiŽ cant role in the persistence of class differentials in educational attainment—and in turn represent an obvious focus for
policy interventions. These results, from research undertaken within a variety
of national educational systems (for full references, see Goldthorpe, 1996),
reveal that even when level of demonstrated ability is held constant, children
of more advantaged class origins take more ambitious educational options—
for instance, stay on rather than leave or choose academic rather than vocational courses—than do children of less advantaged origins. For example, in
Sweden in the early 1990s, among children with average grades in primary
school, about twice as many from Class I backgrounds as from Class VI and VII
backgrounds entered academic tracks in secondary school (Erikson and
Jonsson, 1996, p. 77).
We are therefore entirely sympathetic to the argument put forward by Bowles
and Gintis that the intergenerational persistence of differences in educational
attainment results, at least in some important part, from actions taken by parents
and offspring under the in uence of a range of subjective dispositions or traits,
such as attitudes to risk, orientations to the future and sense of personal efŽ cacy,
that themselves tend to be intergenerationally transmitted. From such a position, a
number of sociologists have in fact developed models of educational choice (in
some cases to a formal level) and of mobility strategies, in which ideas of risk
aversion, the discounting of future rewards and belief in returns to effort and
probabilities of success all Ž gure (Erikson and Jonsson, 1996; Goldthorpe, 1996;
Breen and Goldthorpe, 1997; Breen, 1999, 2001; Goldthorpe, 2000, chapter 11;
Jonsson and Erikson, 2000).13
However, our class structural approach allows us, we believe, to give a rather
more detailed and differentiated account than Bowles and Gintis offer in this issue
of the social grounding of these dispositions that help preserve inequality across
generations. Bowles and Gintis observe that “less well-off people may be more likely
to be risk averse, to discount the future and have a low sense of efŽ cacy” (p. 18)
than the better-off. However, children from less advantaged origins may have good
13
The model advanced by Breen and Goldthorpe (1997) has been subjected to various attempts at
empirical testing, the most sophisticated of which, setting hypotheses derived from this model against
ones derived from the “linear social distance” approach of Akerlof (1997), is by Davies, Heinesen and
Holm (forthcoming). The development of this model by Breen (2001) is also of interest in suggesting
a behavioral basis for the models of educational transitions proposed by Mare (1980, 1981), in response
to the critique of Cameron and Heckman (1998).
42
Journal of Economic Perspectives
reasons to avoid high-risk alternatives, even if risk aversion is equally distributed
across classes ( Jonsson and Erikson, 2000 p. 364). We would emphasize in this
respect not only differences in current levels of income but, further, the differences
in economic security, stability and prospects that we previously identiŽ ed as the
main sources of class advantage and disadvantage. It is notable, for example, that
children of Class IIIa (routine nonmanual) families are regularly found to have
levels of educational attainment more similar to those of Class I and II families than
to those of Class VI (skilled working class) families, even though in average income,
they are much closer to the latter.
In addition, we would see the speciŽ c educational and occupational goals that
young people pursue as being likewise best understood in relation to their class
origins, following the “structural” theory of aspirations initially set out by Keller and
Zavalloni (1964) and developed by Boudon (1974). From this standpoint, levels of
aspiration are to be assessed not in absolute terms but relative to the positions of
those holding them. Instead of the emphasis being on class differences in aspirations, it comes rather to be placed on the shared priority within all classes that
children should achieve educational levels and class positions not less desirable than
those of their parents or, in other words, should avoid downward mobility.14
However, it may then be the case with children of less advantaged class backgrounds that the safest strategy to this end—for example, opting for a vocational
rather than an academic course that carries higher risks of failure—is not best
suited to achieving upward mobility.
In sum, intergenerational inequality has important self-maintaining properties. It creates conditions under which individuals in less advantaged positions
choose and act in ways that can in themselves be understood as adaptively quite
rational (rather than, say, being the expression of “dysfunctional” subcultures) yet
which, in aggregate, serve to perpetuate the status quo. Educational expansion and
reform alone should not therefore be expected to serve as very effective instruments of public policy aimed at creating greater equality of opportunity in the sense
of “a more level playing Ž eld.” Complementary efforts to reduce inequalities of
condition, and especially class inequalities in economic security, stability and
prospects, will also be required.
y We are indebted to Tony Atkinson, Anders Björklund, Richard Breen, Adam Swift and the
editors for very helpful comments on an earlier draft of this paper.
14
One could alternatively talk, as in Breen and Goldthorpe (1997), of equal relative risk aversion across
classes, and there is an obvious afŽ nity with the “prospect theory” of Kahneman and Tversky (1979),
according to which the slope of utility curves is greater in the domain of losses than in the domain of
gains.
Intergenerational Inequality: A Sociological Perspective
43
References
Akerlof, George A. 1997. “Social Distance and
Social Decisions.” Econometrica. 65:5, pp. 1005– 027.
Ashenfelter, Orley and Alan Krueger. 1994.
“Estimates of the Economic Return to Schooling
from a New Sample of Twins.” American Economic
Review. December, 84:5, pp. 1157–173.
Atkinson, A. B., A. K. Maynard and C. G.
Trinder. 1983. Parents and Children: Incomes in
Two Generations. London: Heinemann.
Baudelot, C. and M. Glaude, 1989. “Les
diplômes se dévaluent-ils en se multipliant?”
Economie et Statistique. October, 225, pp. 3–16.
Bishop, Y. M., S. E. Fienberg, and P. W. Holland. 1975. Discrete Multivariate Analysis. Cambridge, Mass.: MIT Press.
Björklund, Anders and Markus Jännti. 2000.
“Intergenerational Mobility of Socio-economic
Status in Comparative Perspective.” Nordic Journal of Political Economy. 26:1, pp. 3–32.
Blanden, J. et al. 2001. “Changes in Intergenerational Mobility in Britain.” Discussion Paper
517, Centre for Economic Performance, London.
Blau, P. M. and O.D. Duncan. 1967. The American Occupational Structure. New York: Wiley.
Blossfeld, Hans-Peter and Yossi Shavit, eds.
1993. Persistent Inequality: Changing Educational
Attainment in Thirteen Countries. Boulder, Colo.:
Westview Press.
Boudon, Raymond. 1974. Education, Opportunity and Social Inequality. New York: Wiley.
Bowles, Samuel and Herbert Gintis. 2001.
“The Inheritance of Economic Status: Education, Class and Genetics,” in International Encyclopedia of the Social and Behavioral Sciences: Genetics,
Behavior and Society, Volume 6. Marcus Feldman
and Paul Baltes eds. New York: Oxford University Press and Elsevier, pp. 4132–141.
Brauns, H. and S. Steinmann. 1999. “Educational Reform in France, West Germany and the
United Kingdom: Updating the CASMIN Educational ClassiŽ cation.” ZUMA-Nachrichtung. May,
44, pp. 7– 44.
Breen, Richard. 1994. “Individual Level Models for Mobility Tables and other Cross-ClassiŽ cations.” Sociological Methods and Research. November, 23:2, pp. 147–73.
Breen, Richard. 1999. “Beliefs, Rational
Choice and Bayesian Learning.” Rationality and
Society. 11:4, pp. 463–79.
Breen, Richard. 2001. “A Rational Choice
Model of Educational Inequality.” Instituto Juan
March, Madrid, Working Paper 2001/166.
Breen, Richard and John H. Goldthorpe.
1997. “Explaining Educational Differentials: Towards a Formal Rational Action Theory.” Rationality and Society. 9:3, pp. 275–305.
Breen, Richard and John H. Goldthorpe.
1999. “Class Inequality and Meritocracy: a Critique of Saunders and an Alternative Analysis.”
British Journal of Sociology, 50:1, pp. 1–27.
Breen, Richard and John H. Goldthorpe.
2001. “Class, Mobility and Merit: The Experience of Two British Birth Cohorts.” European
Sociological Review. 17:2, pp. 81–101.
Breen, R. and J. O. Jonsson. 1997. “How Reliable are Studies of Social Mobility?” Research in
Social StratiŽcation and Mobility. 15, pp. 91–112.
Breen, Richard and J. O. Jonsson. 2000. “Analyzing Educational Careers: A Multinomial
Transition Model.” American Sociological Review.
65:5, pp. 754 –72.
Cameron, Stephen V. and James J. Heckman.
1998. “Life Cycle Schooling and Dynamic Selection Bias: Models and Evidence for Five Cohorts
of American Males.” Journal of Political Economy.
April, 106:2, pp. 262–333.
Chevalier, A. and I. Walker. 2001. “Further
Results on the Returns to Education in the
U.K.,” in Education and Earnings in Europe: A
Cross-Country Analysis of Returns to Education. I.
Walker, N. Westergaard-Nielsen and C. Harmon,
eds. London: Edward Elgar, pp. 302–30.
Davies, Richard, Eskil Heinesen and Anders
Holm. Forthcoming. “The Relative Risk Aversion Hypothesis of Educational Choice.” Journal
of Population Economics.
Dearden, Lorraine S., Steve Machin and Howard
Reed. 1997. “Intergenerational Mobility in Britain.” Economic Journal. January, 107:440, pp. 47– 66.
Erikson, Robert and John H. Goldthorpe.
1985. “Are American Rates of Social Mobility
Exceptionally High? New Evidence on an Old
Issue.” European Sociological Review. 1:1, pp. 1–22.
Erikson, Robert and John H. Goldthorpe.
1992. The Constant Flux: A Study of Class Mobility
in Industrial Societies. Oxford: Clarendon Press.
Erikson, Robert and Jan O. Jonsson, eds.
1996. Can Education be Equalized? The Swedish Case
in Comparative Perspective. Boulder, Colo.: Westview Press.
Erikson, Robert and Jan O. Jonsson. 1998.
“Social Origin as an Interest-bearing Asset: Family Background and Labour Market Rewards
among Employees in Sweden.” Acta Sociologica.
41:1, pp. 19 –36.
Featherman, D. L. and R. M. Hauser. 1978.
Opportunity and Change. New York: Academic Press.
Ganzeboom, Harry G. B., Ruud Luijkx and
Donald J. Treiman. 1989. “Intergenerational Class
Mobility in Comparative Perspective.” Research in
Social StratiŽcation and Mobility. 8, pp. 3–55.
Goldthorpe, John H. 1996. “Class Analysis and
44
Journal of Economic Perspectives
the Reorientation of Class Theory: The Case of
Persisting Differentials in Educational Attainment.” British Journal of Sociology. September,
47:3, pp. 481–505.
Goldthorpe, John H. 2000. On Sociology: Numbers, Narratives and the Integration of Research and
Theory. Oxford: Oxford University Press.
Goldthorpe, John H. and Colin Mills. Forthcoming. “Trends in Intergenerational Class Mobility in Britain in the Late Twentieth Century,”
in National Patterns of Social Mobility: Convergence
or Divergence. Richard Breen, ed. Oxford: Oxford
University Press.
Goldthorpe, John H., Clive Payne and Catriona Llewellyn. 1987. Social Mobility and Class
Structure in Modern Britain, Second Edition. Oxford: Clarendon Press.
Goodman, Leo A. 1979. “Simple Models for the
Analysis of Association in Cross-ClassiŽ cations
Having Ordered Categories.” Journal of the American Statistical Association. 74:367, pp. 537–52.
Guzzo, Sal. 2002. “Getting in Through the
Back Door: Equal Educational QualiŽ cations,
Unequal Occupational Outcomes.” ISA Research Committee on Social StratiŽ cation and
Mobility, Oxford, April.
Hertz, Thomas. 2001. “Intergenerational
Transition Probabilities in the PSID.” Center for
Health and Well-Being, Princeton University.
Hope, Keith, Joseph Schwartz and Sara Graham. 1986. “Uncovering the Pattern of Social
StratiŽ cation: A Two-Year Test-Retest Inquiry.” British Journal of Sociology. September, 37, pp. 397– 430.
Hout, Michael. 1983. Mobility Tables. Beverly
Hills: Sage.
Ishida, Hiroshi. 1995. “Intergenerational Class
Mobility and Reproduction,” in Social StratiŽcation and Mobility: Basic Analysis and Cross-National
Comparison. H. Ishida, ed. Tokyo: SSM Research
Series, pp. 145–97.
Ishida, Hiroshi, Walter Müller and John M.
Ridge. 1995. “Class Origin, Class Destination,
and Education: A Cross-National Study of Ten
Industrial Nations.” American Journal of Sociology.
July, 101, pp. 145–93.
Jackson, Michelle. 2001. “Meritocracy, Education
and Occupational Attainment: What do Employers Really See as Merit?” Sociology Working Paper,
Department of Sociology, University of Oxford.
Jackson, Michelle, John H. Goldthorpe and
Colin Mills. 2002. “Education, Employers and
Class Mobility.” ISA Research Committee on
Social StratiŽ cation and Mobility, Oxford, April.
Jonsson, Jan O. 1992. Towards a Merit-Selective
Society? Stockholm: Swedish Institute for Social
Research.
Jonsson, Jan O. Forthcoming. “Social Mobility
in Sweden, 1976 –1999,” in National Patterns of
Social Mobility: Convergence or Divergence. Richard
Breen, ed. Oxford: Oxford University Press.
Jonsson, Jan O. and Robert Erikson. 2000.
“Understanding Educational Inequality: The
Swedish Experience.” L’année sociologique. 50:2,
pp. 345– 82.
Kahneman, D. and A. Tversky. 1979. “Prospect
Theory: An Analysis of Decision under Risk.”
Econometrica. 47:2, pp. 263–91.
Keller, S. and M. Zavalloni. 1964. “Ambition
and Social Class: A RespeciŽ cation.” Social Forces.
43:1, pp. 58 –70.
König, Wolfgang, Paul Lüttinger and Walter
Müller. 1988. “A Comparative Analysis of the Development and Structure of Educational Systems:
Methodological Foundations and the Construction of a Comparative Educational Scale.” CASMIN
Working Paper 12, University of Mannheim.
Lazear, Edward P. 1995. Personnel Economics.
Cambridge, Mass.: MIT Press.
Logan, J. A. 1983. “A Multivariate Model for
Mobility Tables.” American Journal of Sociology.
September, 89, pp. 324 – 49.
Maccoby, Eleanor E. 2000. “Parenting and its
Effects on Children: On Reading and Misreading Behavior Genetics.” Annual Review of Psychology. 51, pp. 1–27.
Mare, Robert D. 1980. “Social Background
and School Continuation Decisions.” Journal of the
American Statistical Association. 75:1, pp. 295–305.
Mare, Robert D. 1981. “Change and Stability
in Educational StratiŽ cation.” American Sociological Review. 46:1, pp. 72– 87.
Marshall, Gordon, Adam Swift and Stephen
Roberts. 1997. Against the Odds? Social Class and
Social Justice in Industrial Societies. Oxford: Clarendon Press.
Rose, David and Karen O’Reilly, eds. 1997.
Constructing Classes: Towards a New Social ClassiŽcation for the U.K. London: Economic and Social
Research Council and OfŽ ce for National Statistics.
Rose, David and Karen O’Reilly. 1998. The
ESRC Review of Government Social ClassiŽcations.
London: OfŽ ce for National Statistics and Economic and Social Research Council.
Vallet, Louis-André. Forthcoming. “Change
in Intergenerational Class Mobility in France
Analysed According to the CASMIN Perspective
from the 1970s to the 1990s,” in National Patterns
of Social Mobility: Convergence or Divergence. Richard
Breen, ed. Oxford: Oxford University Press.
Warhurst, Chris and Dennis Nickson. 2001.
Looking Good, Sounding Right: Style Counselling in
the New Economy. London: The Industrial Society.
Whelan, Christopher T. and Richard Layte.
2002. “Late Industrialization and the Increased
Merit Selection Hypothesis: Ireland as a Test
Case.” European Sociological Review. 18:1, pp. 35–50.