Intergenerational class mobility and the convergence thesis: England, France and Sweden¹

Compared rates of intergenerational class mobility among the adult male populations of England, France and Sweden have been taken as a basis for evaluating current arguments concerning mobility patterns within western industrial societies and for testing two different versions of the thesis which claims that these patterns display an essential similarity.

Abstract

In two papers previously published, we have compared rates of intergenerational class mobility among the adult male populations of England, France and Sweden (Erikson, Goldthorpe and Portocarero 1979, 1982). The empirical results reported in these papers have been taken as a basis for evaluating current arguments concerning mobility patterns within western industrial societies and, in particular, for testing two different versions of the thesis which claims that these patterns display an essential similarity. The earliest and simplest version of this thesis is that due to Lipset and Zetterberg, which holds that the actually observed – or, as we would wish to say, the absolute – rates of mobility between broadly defined classes tend to be 'much the same' from one western industrial society to another (Lipset and Zetterberg 1959). The data presented in the first of our two papers stood in some opposition to this claim. While our results could lend support to the idea of there being a 'family resemblance' among the class mobility patterns of England, France and Sweden, each of these countries was at the same time found to have a fairly distinctive 'mobility profile' when intergenerational movements in class position were examined on the basis of a ninefold class schema. Inflow rates, or patterns of class recruitment, showed especially marked cross-national variation. A major factor creating such variation was evidently that of historically-determined differences in the class structures of the three societies, most notably ones associated with the relative sizes of their agricultural sectors and with differing rates of contraction of employment in agriculture in the course of economic development. It is, however, awareness of precisely this possibility of structurally induced variations in absolute mobility rates which distinguishes the subsequent reformulation of the Lipset-Zetterberg thesis undertaken by Featherman, Lancaster Jones and Hauser (Featherman, Lancaster Jones and Hauser 1975: 339–41). What these latter authors propose is that an essential similarity in mobility patterns across western industrial societies should be sought not at the 'phenotypical' level of absolute rates, where the effects of structural differences will be apparent, but rather at the 'genotypical' level of the relative rates which underlie the absolute ones: or, in other words, at the level of mobility rates as assessed net of structural effects. In the second of our two papers we have been able to show that evidence for England, France and Sweden is certainly more favourable to this revised version of the thesis of cross-national similarity than it is to the original version. A detailed model was developed of the common pattern of relative mobility rates – or of what we would term social fluidity – in England, France and Sweden, and this proved to have a reasonable degree of fit to our empirical data. It was at all events confirmed that the cross-national differences that were found in absolute rates were overwhelmingly attributable to structural effects rather than to differences existing at the level of relative rates also. At the same time, though, the model did not achieve a perfect fit and, more importantly, one systematic, if small, national deviation from it was detected. In the Swedish case, generally lower tendencies towards class immobility could be observed than the model would predict: a finding which, as we noted, is of obvious interest in the context of the current debate over how far social-democratic rule within capitalist societies can be effective in reducing class-based inequalities in social power and advantage, and thus in producing greater equality of life-chances and a more 'open' form of society. The present paper follows in direct continuation of our two earlier ones, and aims to overcome one major limitation of the analyses which they presented: namely, that these entirely took the form of static comparisons. While a basis was thus provided for assessing how closely at one particular point in time – specifically, the early 1970s – the mobility patterns of the three societies studied resembled each other, no evidence could be adduced bearing directly on the question of whether over time these patterns were tending to increase or decrease in their degree of similarity. It would seem clear that for Lipset and Zetterberg the similarity in mobility patterns that they claim among the industrial societies of the western world represents the outcome of an increasing similarity in their occupational and class structures. Thus, the possibility arises, and calls for investigation, that the Lipset-Zetterberg thesis is not so much mistaken as premature: that the cross-national differences in absolute mobility rates which can be displayed on the basis of data for the early 1970s may also be shown to be ones that are steadily diminishing. In their reformulation of the Lipset-Zetterberg thesis in terms of relative rates, Featherman, Lancaster Jones and Hauser do not themselves contend that essentially similar patterns of relative rates will likewise result from wider trends of change that western industrial societies increasingly show in common. However, an argument to this effect might be thought to be implicit in their position and has, in any event, been made out clearly enough by other authors. Thus, for example, Blau and Duncan and subsequently Treiman have argued that the functioning of a modem industrial society entails a secular movement away from 'ascription' and towards 'achievement' as the leading principle of selection for different positions within the social division of labour; and consequently, it is held, the degree of association that exists between the positions held by parents and by their offspring will tend to decline – independently of any effects in this same direction that may result from structural changes (Blau and Duncan 1967; Treiman 1970). Thus, it may be expected that patterns of relative mobility chances within industrial societies will become more alike in becoming more equal – in responding, it might be said, to a common functional need for increased social fluidity or openness. If, then, we are to be in a position to examine arguments such as the foregoing – which are obviously ones of central relevance to larger issues of the extent to which the overall development of industrial societies is set on 'convergent' lines – it is clear that analyses are required that are both comparative and diachronic: that is, which are concerned with similarities and differences in cross-national mobility trends. It is a major aim of the present paper to provide such analyses in the case of intergenerational class mobility in England, France and Sweden. We shall first consider trends in observed or absolute mobility, and try to establish how far, over recent decades, these trends have been of a convergent kind. Then, second, we shall seek to assess how far trends in absolute rates, whatever their nature, have been accompanied, and in part determined, by shifts in relative mobility rates: that is, by changes in mobility chances which cannot be (directly) attributed to structural influences. We shall for the most part proceed, as previously, through the secondary analysis of data on the mobility experience of the adult male populations of England, France and Sweden which are available from national sample surveys undertaken in these three countries in the early 1970s. Details of these inquiries are given in our earlier publications. In thus working from single surveys we shall not have the possibility of observing trends directly: rather, we shall be forced to infer them from the examination of the mobility experience of survey respondents of differing birth dates. However, for each of the three countries we shall be able to make some check on the results of such analyses, at least in regard to relative rates, by setting them against those obtained from other studies of mobility trends which draw on data from inquiries undertaken at two different points in time – although, unfortunately, data so categorised that they cannot be rendered strictly comparable with those on which the present work rests. Our previous comparative analyses of intergenerational mobility rates were made on the basis of a ninefold class schema, the rubric of which is given in the first column of Table I or on the basis of the sevenfold collapse of this schema which is shown in the second column of the table. However, in attempts to infer mobility trends by comparing the mobility experience of successive birth cohorts defined within our national samples, the continued use of such rather elaborate categorisations tends to create difficulties in that cell values become in some instances unduly low from the standpoint of reliability. We have, therefore, in the present paper resorted to further collapses of our class schema – in fact, to the fivefold and threefold versions shown in the third and fourth columns of Table I. Our method of inferring trends also gives rise to a more fundamental methodological problem, namely, that the mobility experience of men in a particular birth cohort will reflect not only 'period' effects – which are what concern us – but also 'age' effects and effects associated with that cohort's relationship to others within the total population. This problem is not one that can ever be fully resolved on the basis of the data available to us; and, because of this, our assessment of mobility trends cannot be simply a matter of 'reading off' statistical results but must embody an important element of interpretation. However, as regards age effects at least, we can in this respect follow certain useful guidelines that are empirically informed. We may, to begin with, regard the older men in our samples, say, those aged 35 and over, as having reached a stage of 'occupational maturity', at which further major changes in their class position are relatively unlikely: that is to say, we may take results for our older cohorts as giving a reasonably reliable indication of the 'completed' pattern of the collective mobility experience of their members. On the other hand, then, we may suppose that results for the younger men in our samples will reflect important 'age' effects, in that many of these men who are still in the early stages of their worklives may be expected to experience significant mobility in the future. Further, though, we can reckon that such worklife mobility will show certain rather well-defined tendencies: specifically, that it will result, through a predominance of upward 'career' mobility, in there being a net increase in the proportion of these men eventually found in Classes I and II of our schema, and also in the self-employed categories comprised by Classes IVa, IVb and IVc; and, conversely, that it will produce a net decrease in the proportion found in the routine nonmanual positions of Class III and the manual wage-earning ones of Classes V/VI, and VIIa and VIIb. We may, thus, interpret the data actually recorded for our younger respondents with these tendencies in mind.2 Finally in this respect it is also relevant to note that the problems generally associated with birth cohort analysis are in fact greatly reduced in so far as we are here concerned specifically with investigating the degree of cross-national convergence in mobility trends. This is so since if mobility rates and their determinants are becoming standardized within western industrial societies, there is no obvious reason why this tendency should not show up in more similar patterns of mobility over working life as well as in more similar mobility rates overall. In order to amplify this point, let us suppose that we are comparing just two countries on the basis of mobility data which, in each case, relate to three cohorts of men, born in 1910, 1920 and 1930 respectively. Consider now the three sets of results displayed in the three panels of Figure I, which have been constructed specifically to illustrate the main possibilities of interest to us. As can be seen, the horizontal axes of the graphs refer to years, and the vertical axes to some measure of the degree of similarity in mobility rates. The lines then plot the similarity that exists between the mobility rates of each of the three corresponding pairs of cohorts from the year in which their members became 20 up to 1970. Points on the lines marked 'x' indicate the extent of the similarity that would be observed in 1970, while the points marked 'o' indicate the extent of the similarity that would be observed if the cohorts were in each case compared when their members were aged 40. Graphs representing three possible relationships between convergence in intergenerational mobility rates, across three birth cohorts in two countries, and differences in patterns of worklife mobility Panel A illustrates a rather simple case of convergence: the similarity in mobility rates increases as one moves from the 1910 to the 1930 cohorts. But – what might be thought empirically rather implausible – the course of worklife mobility that men typically follow is already cross-nationally identical from the earliest cohort, so that the degree of similarity shown up in the mobility rates of the pairs of cohorts does not alter over the lifetimes of their members – i.e. all three lines on the graph are horizontal. Thus, a comparison made at one point in time (for example, 1970) of the mobility of cohorts of differing age would not, under these conditions, be misleading: it would give quite accurate information about the change (increase) in similarity that has occurred. In Panel B we have, in contrast, a case in which the typical course of worklife mobility differs greatly between the two countries, with the result that the similarity in the mobility experience of corresponding cohorts becomes much greater towards the end of their working lives than it was earlier – i.e. all the lines slope upwards. In such a case, then, it could be that, as is shown, a comparison made at one point in time (1970) would lead to a judgment of divergent trends – the mobility of the 1910 cohorts appears more similar than that of the 1930 cohorts – whereas if men at the same age (40) are compared, a convergent trend is apparent. It is therefore in such circumstances as these that a comparison made at one point in time might be regarded as misleading in that it would fail to detect a convergent movement. However, it would still have to be recognized that the convergence that is involved here is of a very peculiar kind: some increased similarity in overall mobility goes together with persistingly dissimilar – that is, non-converging – patterns of mobility over working life. And it could scarcely be held that this state of affairs is envisaged in, or indeed evidently consistent with, any version of the arguments that were reviewed in the previous section. It is in fact in Panel C that we have illustrated the case which these arguments may best be taken as proposing: that is, one in which cross-national mobility experience becomes more similar across successive cohorts and over the course of working life – i.e. all three lines slope upwards, but that for the youngest cohort is the closest to the horizontal. And what is then important to note is that while the differences in the degree of similarity in mobility that are shown up in the comparison made in 1970 are less than those revealed when men are compared at age 40, there is still, as in the case of Panel A, no danger of the former comparison failing to reveal the convergent trend that exists. Perhaps the most direct way in which to test a thesis of convergent mobility trends, at least in regard to absolute rates, would be to compare percentage outflow rates across successive birth-year cohorts. While we cannot do this directly, because of the small numbers of cases that would be observed for each cohort, we may attempt to circumvent this obstacle by using moving averages of these rates. If, for each of our three countries, one were to plot such averages for the mobility of men of a given class of origin to a given class of destination, then an underlying pattern of convergence should be displayed as three convergent lines on the graph. A practical difficulty with this method is that it tends to generate a large amount of information: a graph is required for every cell of any mobility table that one may construct. Also, we found that where cell numbers became small, even moving averages calculated for each birth year of our respondents could show large fluctuations, although the span of years covered by each average was quite wide. For these reasons, then, we decided to apply the method in question only to data at a high level of aggregation: that is, to our basic intergenerational mobility data as organized on the basis of the threefold collapse of our class schema shown in Table I. It may however be noted that this collapse into nonmanual, manual and farm classes has the advantage of being, nominally at least, the same as that utilized by Lipset and Zetterberg (although in fact they excluded the farm class from most of their analyses). Taking this version of the class schema, it is thus possible to derive from our data birth-year averages of percentage outflow rates for men in each of our three countries and for each of nine possible origin-to-destination transitions. However, even with this degree of aggregation, the numbers involved in two transitions – those from nonmanual to farm and from manual to farm – are too small to permit any very reliable conclusions to be drawn and these transitions are therefore left out of consideration. Graphs for five of the remaining seven transitions for men age 35 to 64 are shown in II-VI, the two omitted graphs being essentially the obverses of ones presented.3 In these graphs the moving average for a particular year is based on the percentage outflow value for men born in that year and the corresponding values for the seven previous years and the seven following years – i.e. fifteen values are used in all. However, weighting formulae are applied in order to give greater weight to values closer to the mid-point one.4 Moving averages of outflow rates: nonmanual origins to manual destinations.(a) Moving averages of outflow rates: manual origins to nonmanual occupations.(a) Moving averages of outflow rates: farm origins to non-manual destinations Moving averages of outflow rates: farm origins to manual destinations Moving averages of outflow rates: farm origins to farm destinations From Figure II a clear idea can be gained of how, using our method, convergent trends in mobility rates may be shown up. It is at once evident from the graph that mobility rates indicative of intergenerational mobility from nonmanual origins to manual positions have tended to become more similar within our three countries.5 As the outflow from nonmanual origins to agricultural occupations has been negligible in all countries and for all cohorts, it follows that a similar convergence has taken place in the extent of intergenerational stability within the nonmanual classes. It can moreover be seen that it is England and France which have, so to speak, converged on the Swedish rate: that is, through an increase in nonmanual to manual movement in the French case – implying a corresponding decline in nonmanual immobility – and through the converse trends in the English. However, so far as the other four graphs of III-VI are concerned, no similarly clear-cut instances of convergence spring to the eye. A general tendency might still be claimed for the English and French rates to come closer together; but the important point is that in these cases this movement does not at the same time bring them closer to the Swedish rates. On the contrary, the latter tend to diverge from the growing similarity evident between those for the other two countries, and in a rather systematic way: that is, in showing less intergenerational stability within both the manual and the farm classes and, correspondingly, greater outflow from both manual and farm origins to nonmanual positions and also from farm origins to manual positions. In other words, it could be said that in these respects Sweden would seem to become more distinctive in displaying a generally higher level of mobility. In sum, then, the data presented in III-VI, when considered overall, cannot be regarded as providing the thesis of convergence in (absolute) mobility rates with any very large measure of empirical support. Instances of convergence can clearly be seen, but so too can instances of divergence which theoretical arguments of the kind earlier reviewed could not readily provide for. An alternative way of examining trends in absolute mobility rates is through the use of dissimilarity indices (DIs). In an earlier paper (Erikson, Goldthorpe and Portocarero 1979) we used DIs to assess the extent of cross-national differences in percentage outflow distributions for men originating in each of the nine classes distinguished in our schema. Here, in Table II, we make the same kind of comparison, but on the basis of the sevenfold version of the schema and with each national sample being divided into three broad – 15 year – birth cohorts. In this case, then, the focus of interest will be not so much on the actual size of the DI for any class of origin and pair of countries, but rather on whether over the three cohorts DIs are tending to fall, thus suggesting that some convergence in rates is in train. In Tables III, IV and V we also present the actual outflow distributions from which the DIs of Table II are derived, so that it can be seen how any shift in DIs has actually come about. It will however be apparent from these tables that we are here sometimes working with disturbingly small numbers, and that it may well thus be dangerous to attach too much significance to any one figure. Our attention should rather be concentrated on the overall pattern of results that emerges, and on the question of whether this pattern, while based on a different and more detailed version of our class schema, is consistent with that produced by our previous analysis. As regards, to begin with, outflow from Classes I and II, the data of Table II would indicate that in this case the question of convergence scarcely arises. For the DIs reported for all cohorts, in each of the three pairwise comparisons made, have remarkably low values. In general, we would think it reasonable to interpret DIs of around 10 or less as indicating that in fact only very slight, if not indeed negligible, differences exist between the distributions to which they relate. We have earlier noted that if Classes I and II are taken together, their outflow pattern does display, in line with Lipset and Zetterberg's original thesis, a rather striking cross-national similarity – the salient feature of which is that in each country just over 60 per cent of the sons of Class I and II fathers are themselves found in Class I and II positions. From Table III it can then be seen that this close similarity is already evident in the oldest of our three cohorts – that of men born entirely in the first quarter of the twentieth century – and is maintained in the next cohort as, in each country alike, the proportion of Class I and II sons who appear as intergenerationally stable increases somewhat. Furthermore, recalling the guidelines earlier mentioned for interpreting results for our youngest cohort, we may reasonably suppose that in this cohort there is, at all events, little likelihood in any of the three countries of a decline in the proportion of stable Class I and II sons – as, in the course of their future working lives, some number of those found at the time of inquiry in Class III, V/VI and VIIa positions especially will gain access to Classes I and II. The expansion of what we have termed the 'service class' of professional, higher technical, administrative and managerial employees can be regarded as a fairly general feature of the recent history of the advanced societies of the west, and as one which in itself makes for a high level of intergenerational 'succession' within this class (cf. Goldthorpe 1982). In so far, then, as there is any possibility of convergence in outflow patterns from Classes I and II, this must lie in the distribution of men of Class I and II origins who apparently have not succeeded in maintaining their fathers' positions. In fact, the only instance that is perhaps worthy of note here occurs in the French case in the decline over the two older cohorts in the proportion of such men who are found in Class IVa and b positions – or, that is, within the ranks of the petty bourgeoisie – and the corresponding increase in the proportion found in Class V/VI positions. It is primarily on account of this shift that the French outflow pattern overall becomes in the second cohort yet closer to the Swedish and English patterns than it was in the first. We may now move on to consider outflow patterns from what we may think of as the 'intermediate' classes of our schema – that is, Classes III, IVa and b and IVc. In the case of Class III, that of routine nonmanual employees, the small numbers of men originating in this class in the Swedish sample means that the only comparison that we can usefully make is that between England and France. The relevant DIs in Table II do in fact fall across the three cohorts distinguished, indicating thus that a convergence in outflow patterns has occurred. From examination of the actual distributions, as shown in Table IV, it can be seen that this increased similarity has not come about in any straightforward manner, but its main source can perhaps best be revealed if we consider outflow from Class III origins to Classes V/VI and VIIa taken together, as representing industrial working-class occupations. It could then be said that while in the first, or oldest, French cohort the proportion of sons of Class III fathers who were found in such working-class positions is less than a third but then rises markedly to over two-fifths in the second, in the English case the proportion falls to around two-fifths in the second cohort from being nearer to a half in the first. In other words, we find here the same pattern as was revealed for nonmanual to manual outflows generally by our analyses based on moving averages. Turning next to mobility from Class IVa and b – that is, petty-bourgeois – origins, we may note, first of all, that here again the England/France comparison in Table II shows clear evidence of convergence. Further, as Table IV reveals, this is again the result, to an important extent, of a growing similarity in outflows to working-class (Class V/VI and VIIa) positions. Over the three cohorts, the sons of the French petty bourgeoisie appear increasingly likely to be found within the industrial working class, but among their English counterparts no such trend is evident. Thus, while in the oldest cohort only around a quarter of the former as against some two-fifths of the latter have apparently been 'proletarianized', in the youngest cohort one might expect the eventual proportion who could be so regarded to lie somewhat between a third and two-fifths in each country alike. In addition, one may also recognize as, so to speak, the complementary aspect of this process of convergence, the decline in the extent of intergenerational stability within Classes IVa and b in France, which is again a trend not matched in the English case. Here too, then, a rather large dissimilarity in the first cohort would seem to be more or less eliminated by the third. We have previously remarked that the French petty bourgeoisie would appear to be distinctive in the degree to which its sons 'succeeded' their fathers and avoided entry into the working class (Erikson, Goldthorpe and Portocarero 1979: 435–6). Our cohort analyses now enable us to qualify this observation by recognizing further that, in the comparison with England, the differences evident in these respects are very largely ones which arise from the experience of men who were born in, say, the first third of the century. We may thus regard the foregoing as a fairly well defined example of convergence in mobility patterns. However, when we move on to the two further comparisons that we can make in regard to Classes IVa and b – that is, between England and Sweden and between France and Sweden – convergent trends are less readily discerned. It is true that in both instances the DIs shown in Table II fall over the first two of our cohorts, but this fall is not maintained into the third. Examination of the relevant data in Table IV would then suggest that this is chiefly because in the Swedish case a decline in the intergenerational stability of the petty bourgeoisie has occurred as in the French, but of a still more sudden and rapid kind. This conclusion would seem justified by the finding for the third Swedish cohort that less than 10 per cent of men of Class IVa and b origins occupied positions similar to their fathers – even if one were to accept the possibility of 'delayed' counter-mobility which might lead some number of these men back from working-class or other employee positions into self-employment. Although, then, one could say that in no cohort are cross-national differences in outflow from Classes IVa and b exceptionally large, the important point for present purposes is that in two out of the three comparisons that we can display, there is no particular reason for supposing that, over time, these differences are tending to narrow rather than to widen. Finally, as regards the intermediate classes of our schema, we must consider mobility patterns associated with Class IVc, that of self-employed workers,