Myths of the Chicago School

4. The quantitative tradition at Chicago

4.5 Burgess as the barometer of methodological tendencies

If Park, on balance, was hostile to statistics, Burgess certainly was not. Before coming to Chicago, Burgess had been involved with social surveys and statistics. He had worked with J.J. Sippy on the Belleville Survey (1913) and with F.W. Blackmar on the Lawrence Social Survey of 1916 (Burgess 1916). He encouraged students to make use of statistical data wherever appropriate. He was very much involved in developing census data for sociological use. 'Burgess, it may fairly be claimed, was the father of modern census tract statistics, both by example and as a co-ordinator of pressure on the Bureau of Census in drawing up their plans for 1930' (Bulmer, 1981a, p. 315).
 
Burgess, in conjunction with Newcomb, produced analyses of the 1920, and 1930 censuses (Burgess and Newcomb, 1931, 1933), and encouraged the analysis of the 1934 census (Newcomb and Lang, 1934). In conjunction with this Newcomb was awarded the 'University prize for originality in research ... for his contribution of a "Single Numerical Index of Age and Sex Dis­tribution of Population"' (Bulletin of the Society for Social Research, Dec., 1930).
 
Burgess' request for funding from the Social Science Research Council (Burgess, 1935a) for research that would lead to the presentation of materials on city growth, movement of population and formation of local communities by sector, prepared the ground for the later Fact Books of Chicago (Wirth and Furez, 1938; Wirth and Bernert, 1949; Hauser and Kitagawa, 1953; Kitagawa and Taeuber, 1963) [5]. His involvement with census analysis at a multi-disciplinary level goes back at least to 1924. As chairman of the Chicago Census Committee, Burgess circulated a paper (3/6/1924) to the Social Research Committee of Political Economy, Sociology and Political Science Departments on the development of the Chicago census. This committee was the forerunner of the Local Community Research Committee and Marshall, Merriam and Burgess were the executive committee.

Burgess was very instrumental in encouraging students who had been brought up to make case studies and to get qualitative data and that sort of thing to try and translate them into a more quantifiable form and to pay some attention to the statistics. He was much influenced by Ogburn, much encouraged I should say, he was already aware of the uses of quantified material, he was very much aware of the usefulness of the census data and that sort of thing and of course our delinquency area studies ... Yes, that preceded Ogburn ... Park was mainly grumbling about people who started on a problem with lets get some quantitative data, he wanted to start with 'lets find out about the way this person ticks' and get inside of him and inside of his community and find out what is going on. And if you have to use quantitative data [then] use it in the light of what you know, not start out with it and cover over your ignorance about it, as is what goes on with a lot of figures. So we were already using rather elementary quantitative material. And that was, I think, becoming more impressed on Faris, as Chairman, to get a statistician. (Cottrell, 1972)

Ogburn had a considerable impact on Burgess when he arrived, and Burgess supported Ogburn. He encouraged the development of multiple factor analysis by his students, and learned the technique from Lang and Cottrell, whom he had encouraged to attend Ogburn's class. 'Actually we were the first sociologists to use factor analysis on sociological data in the marriage prediction studies. It was not a central thing but a technique we tried out and got some help in the interpretation of the data (Cottrell, 1972).
 
Burgess had taken the opportunity, as a full professor, of attending Ogburn's quantitative courses and utilised the extensive quantitative expertise available at Chicago in his own research. This is clear from the development of his marriage adjustment research. In an enclosure to a letter to Wirth, Burgess (1935b) referred to the initial study of prediction of marriage adjustment he had undertaken in collaboration with Cottrell. The aim was to predict, at the time of marriage, the success or failure of the relationship. He referred to the fact that approximately one thousand schedules were secured and that a high degree of reliability was found 'as indicated by coefficients of correlation ranging from .86 to .96'. The construction of a 'marriage adjustment index out of eighteen items in the schedule' allowed for assignation of numerical values to adjustment in marriage. Thus they were able to identify the factors making for success in marriage. The further development of the research was weighted the most important factors and combined them 'for prediction by the method of partial and multiple correlation'.
 
The association between Ogburn and Burgess persisted until their retirement in 1952. In 1950, for example, Ogburn recorded that

one of Burgess's students, Strauss, did some research once on the reasons for choosing a mate for marriage. He investigated three reasons.... I rang Burgess yesterday and suggested that he have another research done on these same couples to see which class of choice yielded more stability and which showed more separations and divorces. (Ogburn journal, 11th November 1950)

Throughout, however, Burgess was concerned to ensure the mutual development of case study and statistics. In the development of a book on the 'Family in the Urban Community', for example, Burgess (1935c) proposed that the volume 'makes use of both statistics and case study material'. Over a thirty-year period Burgess consistently argued for a methodological eclecticism and against monism. In many respects Burgess's methodological comments are a barometer of prevailing direction of methodological concerns in sociology in America from 1920 to 1950.
 
For Burgess development of prediction studies was of the highest 'theoretical and practical importance. Increase in the efficiency, precision, and scope of prediction is a chief aim of the social sciences as it is of all science' (Burgess, 1941 p. 55). This involvement in prediction studies and the development of quantitative research alongside case study was also reflected in his teaching. From 1919 to 1941 Burgess had taught on the 'Field studies' course which gradually became a student practical, in that it was 'designed to provide direction and suggestion for either special research or a community survey' (Burgess, 1952). In 1940 he began teaching 'Methods of sociological research' which was described as:

Methods of historical research, field observation, mapping, interviewing, evaluation of human documents, and case study as used in sociology -- especially in human ecology and social psychology -- and the relationships of these methods to statistical procedures. (Burgess, 1952)

This course continued until 1944 when Burgess taught 'Introduction to statistical sociology', until 1947, which involved

Practical methods of analysing sociological data -- the questionnaire, graphical presentation, interpretation of statistics, the nature of statistical evidence, statistical fallacies. (Burgess, 1952)

Meanwhile he began a course on the 'Problems and methods of prediction' (1941), which was concerned with the theory of prediction and forecasting and included a comparison of 'statistical prediction and forecasts from case studies'. Burgess was proud of his role in pioneering prediction studies (Burgess, 1941) and either directly, or through his students, was involved in such studies in various areas including school and college success, occupational adjustment, criminal recidivism, mental breakdown, adjustment in army camps, juvenile probation, adult probation, cadet camps, and selection of occupation. The first marriage prediction study ever undertaken in 1931 was published in a report by L.S. Cottrell and Burgess in 1939 entitled 'Predicting Success or Failure in Marriage'. A progress report on this study to the Social Science Research Committee at Chicago in the Winter Quarter of 1931 noted of the study that

its objective is to work out a method for predicting the statistical probabilities both (a) of the continuance of the marriage status and (b) of happiness in marriage....
The schedule of eight pages is placed in the hands of persons who have been married from two to five years....
The revised schedule includes the shorter personality test by Professor Thurstone which it is believed will be a valuable addition to the study.
In addition to securing these schedules, a small number of case studies have been secured partly for the purpose of checking the schedule. (Burgess, 1931)

Burgess' prediction studies had a wide impact, for example the work of Lewis Terman ('Psychological Factors in Marital Happiness') and E. Lowell Kelly ('Study of Engaged Couples') used the Burgess-Cottrell scale of marital adjustment, while Clifford Kilpatrick (University of Minnesota) and Harvey Locke (Indiana University) undertook similar marriage study work (Burgess, 1941).
 
These studies gradually became a test of the case study as a means of objective data gathering, as Burgess turned more and more towards a factor analytic, correlational study of prediction. He noted that the prediction work was related to other work going on at the University of Chicago that

has been most helpful in the development of methods of research. W.F. Ogburn, who has carried on considerable work in the prediction of political behavior has given valuable advice in the initial development of both the parole and marriage studies. L.L. Thurstone and his students have been of great assistance particularly in the application of methods of factor analysis (in our first study) and of matrix algebra (in the present study) to problems of prediction. S.A. Stouffer, who has developed a significant mathematical formula for the prediction of intermigration between cities has been increasingly helpful in giving advice and suggestions upon problems of statistical and case-study prediction. His interest in the possibilities of interrelating statistical and case-study techniques goes back to his doctoral dissertation [1930]. He has had a long interest in the problems of prediction as indicated by certain of his published papers. Dr. David Slight, who has given valuable advice on psychiatric aspects of the project, has shown a strong interest in which I equally share in the possibilities of working out a joint project on psychiatric and sociological phases of behavior important for prediction. (Burgess, 1941, p. 58–59) [6]

Clearly, then, the Chicagoans, not only made extensive use of statistics in the inter-war period but also developed the rigorous hypothesis testing and cumulative theory development principles that later came to be identified so closely with the post 1945 'Columbia School'.
 
In his account of prediction studies, Burgess specified the following stages. Stage one is to locate the best criteria of adjustment. The second stage is to isolate the best predictive factors on three fronts: the past experience of subject; the present personality traits; and contingency conditions. The third stage is the combination and weighting of predictive items, using 'factor analysis and matrix algebra' (Burgess, 1941 p. 53). The fourth stage is the assessment of the feasibility of prediction from intensive case studies to discover dynamic factors difficult to arrive at by schedules and to reduce to statistical formulations either (i) by case study analysis alone or (ii) in conjunction with statistical techniques (Burgess, 1941 p. 54)
 
Chicago sociologists generally aimed at explanatory accounts in the Thomasian mould. Burgess (1944a) provided a clear designation of the explanatory process in quantitatively oriented nomothetic sociology when he outlined the five steps in any factor analytic causal attributable study. These are
 
1. selecting criteria (of success) [Y]
 
2. selecting determining factors [Xs]
 
3. establishing temporal precedence
 
4. correlating each X with Y
 
5. combining (useful) Xs into a prediction score for use on others not in the study.
 
This outline closely parallels the Columbia model of operationalised non-spurious time prioritised correlation (Labowitz and Hagedorn, 1971; Hirschi and Selvin, 1972; Boudon, 1974).
 
The Chicagoans were also quite aware of the potential or possibilities of the large-scale scheduled interview or questionnaire survey. Indeed, the earliest development of this form of data gathering for social scientific rather than official purposes was the work done through the Local Community Research Committee in the politics department at Chicago (see Appendix 2). The close interdepartmental ties, meant that the large-scale survey was an accepted and developed tool of social science at Chicago (Bulmer, 1981a). Burgess was very interested in developing the technique and had been since his early involvement in social surveys (Burgess, 1916). By the early 1920s he was considering the likely effectiveness of a large-scale interview-based sociological survey (Burns, 1924).      
 
Burgess's views on technique tended to be very flexible, his main concern was that society be conceived of as an organism rather than as an aggregate of atomistic units. He was prepared, then, to countenance any method that may be of use in developing sociological theory within this overall perspective. Thus he embraced all methodological developments, including advanced statistical techniques, provided that they did not swamp his underlying methodological concerns. Thus, he noted that factor analysis assumes

that factors operate in the individual case as in the average of all cases in the sample of the population upon which predictions are based. This assumption does violence to the clinically-minded person who perceives in each case a unique configuration of dynamic factors.(Burgess, 1944a, p. 30).

In a paper read to the Annual Meeting of the Iowa Association of Economists and Sociologists in 1927 Burgess had asserted the need for empirical work in sociology and his fundamental view of the organic nature of society. He addressed himself directly to the role of statistics and argued that they had tremendous potential and utility if used in conjunction with other techniques and were located within the theoretical perspective he advanced. To adopt an atomistic statistical monism was, for Burgess, to ignore the relationships that lay beneath the surface of ostensive appearances. In this, he reflected Park's Jamesian concern about the 'reality behind the mask' and Thomas's distinction between attitude and value. He argued that statistics alone were inadequate because it was important to

recognise that quantitative methods deal in the main with the cruder, more external aspects of human behavior, and that some other more sympathetic and discerning method is necessary to probe beneath the surface and to depict and analyse the inner life of the person. (Burgess, 1927, p. 112)

His suggestion had been to proffer a mixture of case study and statistics and he argued that the two were not incompatible techniques. He reminded his audience that Le Play (commonly regarded, at the time, as the 'father' of statistical analysis in social science) introduced case study as the 'hand-maiden of statistics' and that Healy's well known statistical analyses into the causes of delinquency were elaborated by the use of case study material, which acted to provide Healy with the insights his statistical analyses alone were unable to provide. Burgess was not simply advocating uncritically or naively the use of case studies. He argued, for example, that they provided a way of testing hypotheses, and suggested that the first systematic use of case studies in this respect had been the Polish Peasant study (Thomas and Znaniecki, 1918). Burgess also pointed to the suggestions of Karl Pearson that scientific method involved classifying facts, noting mutual relations and describing their sequences, and argued that case study provided just such a facility.
 
Burgess did not assume that case study was an established technique, however.

The assumption sometimes made that case studies and statistics were opposed to each other, or that statistics succeeded case studies in the 1920s, does not correspond to reality. Case studies and statistics developed side by side and supplemented each other. Students might use both approaches in their dissertations, as I did in my dissertation on Suicide. (Cavan, 1983, p. 414)

The debates about the relative merits of statistics and case study that peaked at the end of the 1920s and the mid-1930s were not debates about an old established method (case study) being superseded by a new growing method (statistics). As Burgess argued, the systematic use of case study as a method was in its infancy in 1927 and, like the rapidly expanding field of statistics, needed to be developed and nurtured. He advocated quite specifically that the two techniques be embraced jointly.

It is probably sufficient to point out that the methods of statistics and of case study are not in conflict with each other: they are in fact mutually complementary. Statistical comparisons and correlations may often suggest leads for research by the case-study method, and documentary materials as they reveal social processes will inevitably point the way to more adequate statistical indices. (Burgess, 1927, p. 120)

This is a position that Burgess addressed himself to consistently for the next 20 years. Throughout he attempted to develop research that adopted the complementary approach, his emphasis shifting, as he became more and more involved with prediction studies, to evaluating the efficacy of each method for predictive purposes. Thus, as part of his research outline in 1944, he pointed to how his results will provide an assessment of whether 'prediction by case-study analysis of dynamic factors is superior or inferior to prediction by statistical techniques' (Burgess 1944a, p. 54).
 
Even at this stage, having made enormous use of multivariate analysis, factor analysis and other statistical techniques, Burgess is reluctant to concede ground to quantification and treads the delicate line that divided the two sides of the Polish Peasant debate (Social Science Research Council, 1939).

The main research method relied upon by students of personality, social organization and disorganization and collective behavior has been the personal document. Through letters, interviews and life histories data have been obtained by which the processes of personal and social interaction might be analyzed. Once a process is identified and defined then it is possible but often difficult to develop statistical methods for more precise and accurate measurement as in the case of ideal type concept already discussed. (Burgess, 1944a, p. 21)

Burgess persisted in his advocacy of the synthesis of case study and statistics and pointed to how such interrelation may be achieved. He maintained that case study is a useful adjunct to statistics at an exploratory level and that it is useful for interpreting statistical findings that would otherwise be unintelligible. Of central importance for Burgess in the further use of case study was whether or not the problems of explanatory generalisation from case study data may be resolved. Burgess admitted that the strongest supporters of case study are those interested in idiographic rather than nomothetic study. Burgess himself, reflecting the long tradition at Chicago, is interested in nomothetic study and conceded that

so far the data available in comparison of statistical and case study predictions appear to indicate the superiority of quantitative methods. This seems in large part due to the difficulty of controlling the personal equation of the clinical investigator who tends to judge cases on the basis of his training or personal experience. (Burgess, 1944a, p. 31)

In this respect Burgess tended to underplay the shift in the debate, fostered by Blumer, which centred on the conceptualisation process. However, he referred to the 'conceptualisation' debate when addressing the penchant for scaling. He warned against the vogue of scale construction lest it proceed without 'due care for careful development of prior conceptualisation'. It was the critique of conceptualisation, especially of the problems and aims of operationalisation, which lay at the heart of the differences between quantifiers and methodological sceptics such as Blumer and Lerner (Social Science Research Council, 1939).
 
Burgess was both alert to, and interested in, developments in quantification in the social sciences outside of Chicago, notably the mathematical orientation espoused by Lundberg (1929, 1936, 1942) and attempted by Dodd (1940) and the sociometry of Moreno. Burgess saw sociometry as predated by Bogardus' 'Social Distance Scale' but as being a more systematic development of it. Burgess argued that sociometry contributed to the bridging of the gaps between social analysis and statistics by concentrating on group interactive processes rather than the conventional approach of statistical analyses, which was to deal with aggregates of atomistic items. Thus he saw great potential in sociometry as a methodological tool in his advocacy of an eclectic approach.
 
However, this did not deter him from his main concern and contention that statistics and case study are mutually complementary. He argued that case studies suggest problems for quantitative analysis, which need case studies to interpret results fully, which leads to more problems, and so on. In other words, the interactive use of case study and statistics is compatible with a cumulative development of knowledge model. Such a model underpinned American sociological research for half a century and became embodied in middle-range theorising (Merton, 1948). This is further discussed in Chapter five.
 
The examination of the role of the Chicagoans in empirical social research and the development of methodology has indicated that Chicago sociology did not exhibit tendencies towards a single methodic approach. Rather, it tended to be eclectic, and certainly changed throughout the first half of the century. The Chicagoans tended to integrate both the subjective and the objective aspects of the social world and in doing so were not particularly unique (see Social Science Research Council, 1939). They were very much aware of methodological debates and this is reflected in the discussions in the Society for Social Research at Chicago. In its regular meetings and annual institutes, both with a large proportion of visiting speakers, current debates were vigorously enjoined.

Notes

[ 5] Burgess (1935a, p. 1) began

The purpose of this study is to assemble, present in tabular and graphic form, and interpret the materials now available through the census and other sources on the local communities and sectors of the city of Chicago. At the present time these data are available in raw and unanalysed form which makes then difficult to use for any specific practical purpose.'
These other sources included social history data on local communities collected by V. Palmer. The Social Science Research Committee at Chicago agreed to underwrite publication costs, and the research was seen as part of a programme of work involving members of the Committee on the history, demographic and socio-economic analysis of Chicago. Return

[ 6] This quote summarises Burgess' major research interest from 1930 to 1940 and shows how it relates to other research interests at Chicago and these are primarily quantitative. It reflects a wide base of interest, psychological and psychiatric as well as sociological. However, one must read this critically, in the sense that it is an application for funding, that an experienced proposal writer like Burgess will aim to include those elements he thinks will be well received and therefore may be including an 'overloading' of quantification background in order to secure funds in a climate which is more hospitable to quantified research. On the other hand, this represents a genuine interest for Burgess and reflects a concern with relating case study and statistics, which has been central to Chicago research for twenty years. It reflects the interests of the Social Science Research Council as evident in the 1939 debate on the Polish Peasant and guidelines issued in the late 1930s and early 1940s, and had been a subject of debate in, for example, the American Sociological Review since 1936. Again, one might see Burgess as reflecting external concerns in his proposal but these concerns were promoted by, among others, the Chicago sociologists (albeit from different sides). The 'canny' side of the proposal is perhaps reflected in Burgess' next paragraph that reflects the national and institutional interest in the area.

At the present time under the auspices of the Social Science Research Council a sub-committee on Predictive Methods in Social Adjustment is making a comparative study of methods of prediction now in use in the field of school success, vocational adjustment, marriage adjustment and criminal recidivism. The members of the sub-committee are Mr. Stouffer, chairman; L.S. Cottrell, Cornell University; E. Lowell Kelly, Purdue University and E.E. Richardson, U.S. Civil Service Commission with Paul Horst, psychologist, Proctor and Gamble, making the report. The report together with its recommendations will, in my judgement, give a great impetus to the improvement of present methods and to the rise of the level of research in this field.' (Burgess, 1941, p. 59)

This latter point, novel method, is an area of major concern for the Social Science Research Council as shown in its circular of 1945. Return

Next 4.6 Methodological debates in the Society for Social Research