INTRODUCTION
In this article I will discuss some promising new research lines in the study of social mobility in past societies, while at the same time placing the articles in this special issue within the perspective offered by those lines of research. I will do so under three broad headings: questions, data and methods. These are focused on, but not limited to, the study of intergenerational mobility.
ANSWERING OLD QUESTIONS AND CONCEIVING NEW ONES
In this field, new questions appear alongside older ones that have not yet been answered with any degree of precision. As the articles in this special issue have demonstrated, there remains too the classic question: ‘Has there been a long-term trend towards increased social mobility?’ Many historians and sociologists think there has been a trend away from traditional societies, where a person's social position was essentially inherited, to more open ones, where social position depends upon individual achievement. But if the question has long interested historians and sociologists, why has it not been conclusively answered before? How can it be answered with greater clarity and precision, while having regard to trends, differences between regions and the driving forces behind those forms of temporal and geographical variation? One obvious reason might well be that the driving forces behind changes in mobility have not been properly understood. That might indeed have been true in the past, and of course future research will be particularly illuminating, but the introductory article in this special issue has demonstrated that a great many driving forces, both individual and institutional, have already been identified. The bottleneck is most likely to be found elsewhere: in the difficulties of broadening the range of data and the set of methods needed to separate the wheat from the chaff.
There is another interesting point to note. In 1954 David Glass and his collaborators published a volume of essays entitled Social mobility in Britain.Footnote 1 The collection covered a wide range of topics, including the social grading of occupations, a comparative study of intergenerational social mobility in four countries, a study of social mobility between three generations, an essay on self-recruitment in the professions, a chapter on social mobility and marriage, one on the effect of the 1944 Education Act on mobility chances, and one on subjective aspects of social stratification. Some of these chapters cover ‘the usual suspects’ but others seem surprisingly modern, such as the one on mobility over three generations. That is an example of a topic that has more or less disappeared in the half century following the publication of Social mobility in Britain, only to resurface recently.Footnote 2 Over the years, while the quantitative study of social mobility has blossomed, the range of topics has narrowed – an observation already made by several scholars.Footnote 3 That might not have been entirely a coincidence. Rapid and large advances were made under the aegis of the international research group RC28 of the International Sociological Association, which organizes many meetings where like-minded quantitative sociologists present work in progress and comment on that of their peers. This research group functions as a nodal point in the study of social mobility. Rapid advances were made in large measure because of significant developments in methods of analysis, such as Blau and Duncan's status attainment models and log-linear models, including those of Goodman. Research has tended to become concentrated on topics where rapid advances could be made thanks to those methods of analysis. Possibly as a result, research in other directions has not been pursued with as much vigour, or success, resulting in a reduction in the range of topics studied. To the extent that quantitative social historians have studied social mobility, they too have profited from the rapid advances in answering certain questions, at the risk perhaps of shying away from other, possibly more entrenched, questions that could not be so easily answered by those new methods of analysis.
Furthermore, the field of comparative historical social mobility studies has been seriously hampered by the fact that no consensus existed on how to code occupational titles from various countries, languages or epochs to fit them into a common coding grid, let alone on using the same measures of class and rank. Those impediments are about to be removed, as will be discussed later in more detail: consensus on matters of coding and the classifying of historical occupational titles is slowly evolving; new methods of analysis are being developed, and as a result the range of questions that can be tackled is expanding.
A few examples of the broadening of the research agenda must suffice here. Social mobility of women is a topic long neglected in historical research due to the relative paucity of sources and the problem of how to classify female occupational titles. Perhaps too, concern lingered about the problem of under-registration and interpretation of female occupational titles in census and vital registration data. Gendering patterns and the effects of social mobility make for an innovative approach, however, in that it documents and explains the social mobility of mothers and daughters over past centuries. That is interesting first of all in itself, because women in Europe were part of the labour market and experienced mobility, but also because many of the general hypotheses mentioned in the various articles in this volume on determinants have not previously been tested for women, and because gender roles are ascribed characteristics, and the effect of gender may of course have decreased over previous centuries. Studying the social mobility of men and women is both inherently interesting and possible to a greater extent than previously realized, though the relative scarcity of historical data and problems of interpretation still make it more difficult than studying only male occupational mobility. Several recent studies have begun to examine the social mobility of women.Footnote 4
Another interesting extension in recent research is the study of determinants of social mobility other than the occupations of fathers. Research into the twentieth century shows that about half the variance in occupational status was due to characteristics of the original family, of which again approximately half can be explained by measured variables such as parental occupation. Linked vital records – such as those used by Dribe and Lundh in their contribution to this special issue – allow us to investigate to what extent the total family effect changed during past centuries, and the degree to which it can be explained by parental occupations. So-called sibling models do that by comparing the similarity of the occupations of siblings with the similarity of the occupations of unrelated persons; data on regional contexts – see below – enable us to estimate multilevel sibling models.Footnote 5 Still other types of research look at the effects of wider kin networks, including grandparents.
Another expansion of the field concerns the broadening, beyond industrialization, of the study of determinants of social mobility. It should be noted that this is not a new idea, for in a sense Treiman's seminal article of 1970 is still a good and succinct survey of many other determinants,Footnote 6 but they have proved themselves to be rather difficult to study empirically. Even empirical study of the effects on social mobility of the historical process of industrialization has been very difficult, and to my knowledge the research by Zijdeman, notably his article in this special issue, is the first actually to operationalize industrialization and test its effects.Footnote 7 What is needed, and what is becoming available partly through the creation of large historical databases, are datasets with quantitative indicators of community characteristics, as will be discussed later in more detail. That would make it possible to test whether historical changes in the world of work had an effect on the likelihood of mobility, and if so what they were – and where and when changes in the educational system or in family values or management styles were more important, to name but a few of the potential determinants identified in the introduction to this volume. Although Zijdeman's approach is promising and important, the historical study of social mobility determinants is not limited to the use of a regional dataset with quantitative indicators of community characteristics. Other promising lines of research include the in-depth study, both quantitative and qualitative, of changes in the production process and the true divisions of work and systems of rewards in factories, companies, shops and the civil service over a long period of time. Notwithstanding Penn's impressive example, that type of research has not often been done.Footnote 8
DATA
Answering the older and the newer questions alike – only some of which have been mentioned above – requires historical data, and that is a field in which there have been important changes. This special issue is in a sense proof of this change, as the articles it contains rest on important new historical datasets created in recent years.
Vital registers still form an important source of information about social mobility. They consist of records of births, marriages and deaths, and occasionally include population registers. Apart from the occupations of married people, and those of their parents, marriage registers contain other useful information, such as age, marital status, date of marriage, place of birth and residence, address, sometimes even religion, and they reveal something about literacy too, at least whether or not an individual could sign his or her name.
These vital registers contain information about witnesses too, the sort of information that makes historical vital registration systems such an effective source of data from which the occupational and class structures of younger men and women can be compared with those of their parents. Related factors show up as well, such as whether an individual was a migrant (and if so of what type), literacy, geographical location and the social network as suggested by the identities of witnesses. It seems that the richness of those sources has not been fully realized nor tapped until now, not only with regard to their geographical coverage but also to the possibility they offer of going beyond the study of males to cover women too, and furthermore the width of the time horizon, allowing closer study of regional and temporal variations and ‘multiple measurement’.
The large amount of data means that it becomes feasible to break information down according to region and period. Regional and temporal variations can then be described with greater clarity than was possible before, so that even annual fluctuations in social mobility patterns can often be described, revealing, for example, whether years of warfare led to disturbances in mobility patterns, and if so for how long.Footnote 9 Regional variation is more clearly evident, and can be connected to regional variations in institutional contexts. That then opens up the possibility of testing for the effect of presumed determinants of social mobility, in a multilevel design. The large amount of data and its variety have another fortuitous effect on the scope for research. Marriage records contain data not only about the marriage of a particular bride and groom but about everyone mentioned in the marriage record, including parents and other relatives or friends who might have been acting as witnesses. In some datasets, marriage records have been linked either to other marriage records or to birth and death records and other vital registers, which allows for multiple measurement of the occupations of parents and children, and makes it possible to see how the effect of a parent's occupation on a child's occupation might change over the lifetime of the child.
The time horizon can be extended into earlier periods, which is useful when studying any slow process. For some countries vital registers with occupational information cover the eighteenth century as well, and some start even earlier.Footnote 10 That makes it equally possible to take into account a wider range of institutional contexts. The recent past is also revealed for study, and so historical results can be linked to those of the sociologists. The question then, incidentally, becomes one of the best way to make survey results comparable with those from historical vital registers?Footnote 11
The geographical coverage of the data may be pushed further in the direction of global comparisons. Most existing historical studies of social stratification and mobility deal with Europe and North America, and there is a clear need to extend the geographical coverage to other regions of the globe, including regions in Russia, Latin America and Asia. In deciding which countries to study, two considerations must be balanced: on the one hand choosing those institutional contexts thought to matter most while on the other ensuring the availability of both the historical data and qualified historians willing to engage in a collaborative enterprise. With the aim of increasing variation in the determinants of mobility, the geographical extension of the data could focus on regions with different cultural norms regarding parent–child relations and individual ambition as set against adaptation to the collective, and on regions which have experienced different political trajectories, such as from colonialism to post-colonialism. Other stratification mechanisms will be more clearly seen, notably those of race, for example due to slavery, as well as initial levels of social inequality, and economic growth.
Extending coverage to other parts of the globe is a worthwhile enterprise as it further increases the range of contexts. A geographical extension of the data is difficult though because of the absence of a research infrastructure and of collaborative projects between scholars in developed and developing countries.Footnote 12 Practically speaking, the existence of groups of scholars working on social mobility in Russia, Latin America and Asia (although not in Africa) favours the selection of regions with different institutional contexts in those areas. The registration system in Eastern Europe is similar to that in other parts of Europe.Footnote 13 The Russian and East European datasets are pertinent as they permit study of the effect on opportunities for social mobility of the rise and fall of communism, even if derived only from recent data.
For Latin America several options exist. Some historical studies have already been conducted for Brazil, Argentina and Uruguay.Footnote 14 Both Brazil and Argentina are countries with a stratification system based in part on race, even after the formal abolition of slavery. Of them, Brazil's economic growth is more recent while Argentina experienced success earlier.Footnote 15 In Asia, two groups of societies have records most favourable to the study of social mobility. In the Philippines and Macao, colonial powers introduced European-style systems of registration, while in Japan, Taiwan and Korea household registers exist, and some investigations have already been conducted using these records.Footnote 16
The range of data sources can be further extended. Alongside vital registers, other types of information can be used, notably autobiographical information, job advertisements in newspapers and data about education and wealth. Contemporary economic research tends to focus on income mobility rather than occupational mobility,Footnote 17 and while for historians relevant data are much harder to obtain, they are not impossible to find, as is illustrated by the use of poll tax registers in the article by Dribe and Lundh in this volume. The same goes for educational data. The classic status attainment model of Blau and Duncan, for that matter, places the education of parents in a central role similar to that of their occupation, but historical research using educational data is rare and generally limited to university students and their parents. That is because marriage registers generally do not contain information about education other than is implied by the ability to sign or not, while censuses sometimes do. It is sometimes possible to link vital registration data to school records.Footnote 18
Autobiographical information includes published information, whether single autobiographies, collections of autobiographies or interviews, as well as unpublished material including private interviews. Historical autobiographical material exists from the seventeenth century onwards and for all countries in Europe and North America.Footnote 19 The same exists outside Europe – in Brazil and the Philippines, for instance. The material will not be as uniform as vital registers with regard to dates, regions, and social classes covered, nor with regard to the precise nature of the information for each person. It has its own problems of interpretation in any case, including a certain bias towards elites and success stories. Nonetheless, collections of autobiographies by workers do exist, and historians have developed ways of interpreting them.
Apart from their use for purposes of illustration and readability, autobiographies deepen our understanding of the processes of social mobility by identifying how men and women obtain work – for example with the help of parents, family and neighbours, or by sitting formal exams – and which of those methods is perceived as the most important. In the British case, for example, it appears that time-honoured resources such as parents, neighbours and friends were more important than written examinations or private or official agencies well into the twentieth century.Footnote 20 The French and German autobiographies analysed by Maynes reveal, for instance, which of several siblings were allowed more schooling than others and why, and what the perceived effect of schooling was on occupational achievement.Footnote 21 Interesting geographical patterns emerged when she mapped the places where autobiographers recalled teachers who inspired them compared with places where teachers were recalled as being tyrants:
far from being random, the variation in school accounts follows the contours of gender and political geography. Memoirs by Central European authors, especially male authors, show that the classroom tyrant was a fixture of popular culture. Beloved teachers appear only occasionally in German tales, but they play dominant roles in autobiographies from throughout France. In their patterned variety, these school stories provide a privileged inroad into the intellectual and political meaning of schooling for Europe's popular classes.Footnote 22
Job adverts in newspapers could also be used to study mechanisms of stratification and mobility in the past, notably by studying the process of matching new jobs to new arrivals on the labour market. For what sort of jobs are newspaper advertisements placed, and for which are they not, and what sort of things – in terms of educational qualifications and personal characteristics – are being asked for? Newspaper advertisements allow us to identify such characteristics, whether formal qualifications or personal attributes, and which jobs they are wanted for, and they suggest how the matter of gender is approached.Footnote 23 Long-running newspapers carrying job advertisements also exist outside Europe and North America.
Those are only some of the examples illustrating the potential for broadening the range of data.Footnote 24
METHODS
The large amount of data means that it becomes feasible to break down information so that regional variations, for instance, can be described with greater clarity than was possible before, and be connected to regional variations in institutional contexts. That opens up the possibility of testing for the effect of presumed determinants of social mobility in a multilevel design. For the first time in the history of stratification and mobility research, data on changes in institutional characteristics per region are now available for various regions over past centuries. It is now possible to collect information for a large number of regions on factors such as education (including the number and type of schools, for example), communication, transport, the dominant religion, family types and inheritance laws and patterns. We are now in a good position to test any institutional effects on social mobility patterns, and thereby establish a clear hierarchy of importance: that is to demonstrate which determinants had a strong overall influence on social mobility, which were effective only for a limited period or a specific region, and which had no effect at all. By linking mobility patterns to the variety of personal and institutional determinants, our understanding of the theoretical forces that drive social mobility increases greatly. The linking is done using a multilevel regression design.Footnote 25 Multilevel regression techniques allow reliable statistical estimates if data are clustered regionally, and allow us to estimate the effect of regional characteristics such as the number and type of schools.
Apart from the multilevel models, such as those used by Zijdeman, other state-of-the-art statistical models from the social sciences will continue to be used, including the logistic regression models used by Van de Putte, Oris and Matthijs and the log-linear models employed in this special issue.Footnote 26 They will allow new estimates of relative mobility to be produced, to help us decide whether relative mobility is stable or the data suggest increasing openness in the long run or more complex patterns. As the articles in this issue have demonstrated, log-linear models allow us to dissect what is in essence a bivariate association into substantively interesting parts, focusing on general and specific barriers to mobility, such as between sectors of the economy or overall and class-specific over-inheritance.
On a more mundane level, there are already developments under way with regard to the coding and classifying of occupational data which will make historical comparative studies of social mobility easier. Occupations from historical vital registers are now often coded uniformly in HISCO. HISCO is an occupational classification system which is both international and historical, and it has been applied to data originating from fifteen European countries as well as Brazil, Canada, the Philippines and Russia.Footnote 27 HISCO has been implemented in major European and foreign databases, and may now be considered the standard for classifying historical occupations.
Based on HISCO, comparative historical measures of social class and status have been developed – the so-called HISCLASS scheme, with twelve social classes – and measures of social status – the so-called continuous HISCAM status scale of occupations.Footnote 28 They make it possible not only to classify occupations comparably across time and place – as is demonstrated by the articles in this special issue – but to investigate empirically what variations existed in the social status of occupations between generations, countries, or time periods, and between men and women. The development of HISCO and related measures is intended to help overcome some problems and document others, but certainly not to obscure unresolved problems. Testing for problems with the common coding schemes, class schemes, and status scales, as has been done for instance in the contribution by Fonseca and Guimaraes in this special issue, remains very valuable. If problems cannot be resolved then they can at least be documented and their magnitude or significance assessed. Detailed knowledge about the social and cultural context of each society leads to a position where it is feasible to evaluate comparability across samples.
CONCLUSION
Some twenty-five years ago Conzen reviewed American quantitative historical studies dealing with social mobility. She lamented the lack of context:Footnote 29 ‘too often there simply has not remained sufficient time or energy to consult the literary sources that would have added depth and local texture’, which leads to a certain ‘arid quality’ according to critics.Footnote 30 She noted that ‘much of the work has been additive rather than cumulative’,Footnote 31 while hoping that ‘greater sensitivity to limitations in evidence, and greater statistical and theoretical sophistication may yet redeem its initial promise’.Footnote 32 Conzen concluded that ‘the answer … is not less quantification, but … better quantification, more awareness of the limitations of the data, greater willingness to move beyond descriptive statistics to multivariate explanatory models where appropriate, and greater attention to explicit theory in posing questions and interpreting findings’.Footnote 33
The new wave of historical social mobility studies of which this collection is a part promises to provide better answers to key questions in history, economics, sociology and demography concerning trends and determinants of social mobility (broadly understood to include social homogamy, the history of the career and intergenerational social mobility), and to approach those issues by examining a considerably longer period than has been done before, able as we now are to compare different parts of the globe and integrate and test ideas from the three disciplines. The combination of ideas, methods and data with a large disciplinary, temporal and regional coverage will allow both new and better answers to old questions and fresh answers to new questions, for example about the effects of fathers and mothers on the chances of mobility for their children, or about the nature and causes of female mobility. In the end, most of all it will serve to historicize social mobility research, by actually measuring temporal change rather than assuming it, and by looking at the effects on social change of different historical institutional circumstances. The causes of social change between generations can be pinpointed to specific institutional settings, rather than being assumed to be invariant. The combined testing of the effect on social mobility of the relevant personal and institutional characteristics is a necessary step towards resolving the conundrum of the theoretical factors said to have had a bearing on trends in mobility patterns in past centuries. Without cutting away the undergrowth, so to speak, no elegant and persuasive historical theory for the historical process of intergenerational mobility will ever be formulated. Removing the undergrowth will, I believe, reveal a ‘big picture’ of what was important in determining social destinies in the past. The coming decade might be the time to redeem the initial promise.
ACKNOWLEDGEMENTS
This article has been written under the Advanced Investigator Grant of the European Research Council, ‘Towards Open Societies? Trends, Variations and Driving Forces of Intergenerational Social Mobility in Europe over the Past Three Centuries’ ERCAdG, n230279 (2009–2014). I would like to thank Ineke Maas and my other colleagues in the Department of Sociology at Utrecht University and at the International Institute of Social History for their comments on earlier versions.
