Behavioral and Social Science: 50 Years of Discovery (1986)

Below is the uncorrected machine-read text of this chapter, intended to provide our own search engines and external engines with highly rich, chapter-representative searchable text of each book. Because it is UNCORRECTED material, please consider the following text as a useful but insufficient proxy for the authoritative book pages.

Measuring Social Change ALBERT J. REISS, JR. INTRODUCTION Surely among the most influential models of social change was that developed by William Fielding Ogburn (1922b). Ogburn described a process of invention followed by cultural change, followed by social disorganiza- tion, and finally social adjustment (Ogburn & Nimkoff, 1940:8771. Ogburn concluded that public policies and interventions meant to guide modern social change would depend heavily upon the development of a unified national statistical system to collect and process information about social trends (Ogburn, 1929:9581. Although Ogburn's vision of a unified statistical system has not been realized, he may well have regarded this as but a lag in adjustment to which all inventions give rise. This essay does not attempt to assess systematically Ogburn's (1922b) theory of social change, his contributions to our understanding of social trends (1928-1935, 1942), or the development of statistical systems (Og- burn, 1919; President's Research Committee on Social Trends, 19331. But it draws heavily upon that vital heritage. Three major questions are addressed: (1) How do inventions, especially those of the behavioral and social sci- ences, affect social changes and adaptations? (2) How do social changes affect measurement? And, (3) How do contemporary behavioral and social science models, concepts, and methods affect our understanding of society and how it changes? SOCIAL INVENTIONS In Ogburn's view, inventions, particularly mechanical ones, are the source of all cultural growth and evolution. Inventions also cause disruptions in 36

MEASURING SOCIAL CHANGE 37 related parts of culture and in social organization, necessitating adaptations and adjustments. But these adjustments take time, and Ogburn therefore called them cultural lags, noting that "Over the long course of social evolution, measured in thousands of years, cultural lags are invisible. At any particular moment, however, they may be numerous and acute" (Og- burn in Duncan, 1964:301. Although Ogburn emphasized that social inventions can cause social change (1934: 162), his theory and his own work gave priority to mechanical inventions (1922b:76-77; Ogburn & Nimkoff, 1940:809-8101.~ This be- nign neglect of social inventions is coupled with Ogburn's firm conviction that the behavioral and social sciences can shorten cultural lags. Nowhere did he summarize this belief better than in his chapter on invention in Recent Social Trends (President's Research Committee on Social Trends, 1933:166~: Society will hardly decide to discourage science and invention, for these have added knowledge and have brought material welfare. And as to the difficulties and problems they create, the solution would seem to lie not so much in discouraging natural science as in encouraging social science. The problem of the better adaptation of society to its large and changing material culture and the problem of lessening the delay in this adjustment are cardinal problems for social science. Ogburn concluded an essay on trends in social science with these obser- vations (1934:2621: The greatest obstacles to the development of science in the social field are complexity of the factors and the distorting influence of bias. These are formidable, but certainly the trends of the present century are most encouraging, and we may look forward, because of social science, to a greater control by man of his social environment. The relatively lesser emphasis that Ogburn placed on Me role of social as compared win material technology persists to this day. Even social and behavioral scientists tend to overlook their role in processes of social change. In fact, it is quite plausible that social inventions, especially those of the behavioral and social sciences, are a major cause of change, as well as key elements in society's adaptation to change. The selective perception that limits recognition of the role of behavioral and social science inventions may indeed count as a cultural lag. ~Ogburn's interest in social inventions, their effects, and lags in adapting to them preceded the writing and publication of his classic study, Social Change (1922b). His doctoral dissertation (1912) was on child-labor legislation. While teaching at Reed College in Oregon, he became interested in the initiative and referendum as methods of direct legislation (1914, 1915). Still later, he was interested in the consequences of women's suffrage (Ogburn & Goltra, 1919). As Duncan concludes, however, this early interest in social inventions arose, in part, from political sympathies with social problems and reforms.

38 ALBERT J. REISS, JR. Underlying the major themes for this first section is a speculation that the relative contributions of the respective sciences and technologies to social change are altering substantially. Modern societies have come to depend heavily on the behavioral and social sciences and their technologies and cannot run without them. As material technology replaces labor, non- material technology may come to dominate social change, if it has not already done so. Major Social Inventions and Their Consequences Ogburn was fascinated by the effect of what he distinguished as major technological inventions such as the ship, the airplane, the internal com- bustion engine, and the elevator. He also devised lists of significant social inventions (1934:162), such as the minimum wage law, the juvenile court, Esperanto, installment selling, and group insurance. Yet he apparently never attempted to differentiate between social and behavioral inventions with potentially major versus those with more limited or minor effects. Some social and behavioral science inventions, nonetheless, have had such sig- nificant and widespread impact that one cannot imagine modern democratic societies operating without them. Two such inventions, noted in the first report of the Committee on Basic Research in the Behavioral and Social Sciences (Adams et al., 1982) are singled out here: human testing and sample surveys. Human Testing Ogburn (1950) generally attributed invention to three fundamental causes: mental ability, social demand, and the accumulation of cultural elements from which inventions are fashioned. To pinpoint the origins of a particular invention is not a simple task, given the multiplicity of able minds, the variation in the sources of demand, and the different patterns that elements of the cultural base may take. The invention of human testing is usually attributed to a nineteenth- century scientific interest in the study of individual differences. The history of tests of distinctly mental abilities is better documented than other major forms of human testing (Wigdor & Garner, 19821. Tests of mental abilities derived from psychologists' attempts to understand differences in intelli- gence among individuals. Gallon (1869) first devised a series of sensory discrimination tests to shed light on individual differences, followed by Cattell (1890) and others who developed batteries to test sensory and motor abilities. But it was a demand within the French Minister of Education, to distinguish subnormal from normal children in Paris schools, that led Binet, in collaboration with Simon (1905), to introduce the concept of mental age and scales to measure it.

MEASURING SOCIAL CHANGE 39 Ogburn often noted that inventions diffuse more readily where there is a demand for them; the Binet-Simon scale diffused quickly. The test was translated into English by Goddard in the United States in 1908, into Italian by Ferrari in 1908, and into German by Bobetrag in 1912 (Klineberg, 1933:323~. Translation was followed by revision, such as the Stanford- Binet test published by Terman and his collaborators in 1916 (Klineberg, 1933:3241. Although testing has been important to the conduct of research and was a product of psychological laboratories, its development and invention have been highly responsive to social demands arising outside the laboratory, initially by the public schools to sort children and somewhat later by the U.S. Army to screen World War I draftees. Testing is now at least as consequential for the major operating organizations in industrial societies as for the conduct of research. The testing industry is integral to four major organizational tasks: (1) selection of persons as employees or clients; (2) classification of employees or clients according to organizational tasks, (3) assessment of human performance within organizations; and (4) assessment of the "human output" of organizations. Ogburn distingushed primary from derivative effects of invention. Since societies and their organizations do not systematically collect and process information about such effects, even less so for social than mechanical inventions, it is far easier to identify qualitatively than to document the quantitative impact of the invention of human testing. The primary effects are clearly on employment and the management of organizations. Testing occupations generate substantial employment in the U.S. Civil Service, the Armed Forces, public and private school systems, and in large private industrial firms, most of which employ testing extensively in at least one of the four organizational tasks mentioned above, as well as in the devel- opment, production, and marketing of tests themselves. Public controversy and litigation may surround the use of testing in organizational management. Because many organizations base selection and promotion on testing, test information can be influential in legal proceed- ings. The testing industry has been challenged to produce different kinds of tests as a consequence of such litigation. The courts have played a substantial role, for example, in structuring tests for selecting and promoting women and minorities in police and fire departments. Derivative effects of behavioral and social inventions include the spur they often provide to mechanical inventions. The first high-speed printer (essential for modern computers) was developed for a scoring machine by the educational tester Lindquist. In the highly competitive educational achievement testing industry, the rapid scoring and delivery of test results to schools was critical to market shares. As this example illustrates, social

40 ALBERT J. REISS, JR. invention and mechanical invention are seldom independent of one another. The design of modern control systems necessarily involves both human performance measures and technological components. The displacement of humans by computerized robots is also a replacement of some human skills by other human skills. The machine's displacement of manual or mechanical labor moves the labor force toward the cognitive skills that are most dis- tinctively human. It seems no exaggeration to estimate that the average person in an in- dustrial society encounters the products of the testing industry virtually every year for the first two decades of life and in many cases for much of his or her career. Even where not subject to standardized tests, occupational life is controlled by elementary concepts of ability and achievement de- veloped in testing. Increasingly, testing concepts enter the debate over major issues in society, such as the recent controversy over merit pay for teachers especially whether merit can be based on testing teacher performance. Aside from the considerable effect on every other sector of society, the invention of testing precipitated many new inventions in statistics and other behavioral and social sciences. These inventions have significantly affected the conduct of research, and the results of that research have in turn affected society. The early testing of intelligence and mental abilities led to Spear- man's attention to the reliability of measures and his positing of the G factor in intelligence (Spearman, 1904~; this development gave rise to factor analysis, especially with Holzinger's (1930, 1931) development of the bi- factor method (through a study with K. Pearson and collaboration with Spearman, 19251. A variety of statistical factoring methods were soon invented as the concept of intelligence changed with empirical testing, including multiple-factor methods (Thurstone, 1931, 1935) and principle component methods (Kelley, 1928, 1935; and Hotelling, 1933~. As factor analysis was extended to other human traits and characteristics, e.g., human emotions (Burt, 1915, 1939), attitudes, and opinions, awareness of its limitations led to statistical inventions for discerning latent structures (Gutt- man, 1950; Lazarsfeld, l9SO, 1954, 1967; Rasch, 1968, 1980) and statis- tical interactions (Goodman, 1970~.2 These analytical innovations have shaped theory and hypothesis testing in behavioral and social sciences and, 2The history of social science inventions should become an important part of any sociology of knowledge as well as being integral to the study of social change. The ways that demand shapes intellectual agendas is not well understood. Consider the fact that Lazarsfeld undertook his work on latent structure analysis and Guttman on scale analysis in connection with research for the Research Branch of the Information and Education Division of the U.S. War Department in World War II.

MEASURING SOCIAL CHANGE 41 as Holzinger noted in 1941, have had major applications in physics, med- icine, and business forecasting (1941:51. Sample Surveys Modern sample surveys rest on early inventions. The principles of random selection, objective probability, and stratified random sampling are well over a thousand years old (Duncan, 1984:iv). Survey modes of data collection also have been around for a considerable time. But the coalescence and systematization of these inventions into the modern stratified probability survey of a population are a product of modern be- havioral and social science, coming mostly within the last 50 years. As in the case of testing, there is a dearth of data to assess the effects of this invention, particularly its role in social change. Yet, we can plausibly argue that, except for institutional data collected as a by-product of orga- nizational routines, the sample survey has become the major mode for linking action to intelligence in modern democratic societies. Even news organizations do not any longer claim to speak for the aggregate except in a metaphorical sense; but the opinion poll is accepted as doing so. It is difficult to trace all of the ways that the sample survey has come to dominate organizational and individual decisions and operations. A few examples are offered simply to illustrate how pervasive it has become and how instrumental it is in changing behavior. Perhaps nowhere has the invention of sample surveys altered the pattern of activity as much as in American electoral politics. Despite an abundance of skepticism about candidate and opinion polls, no candidate runs for major political office without a private polling operation. Media coverage of elections compares candidates in terms of their poll status; legislative and executive action is responsive to poll information; and political issue and candidate polls are a substantial American industry. A second major area where surveys dominate is in providing intelligence for government decisionmaking. Much of the information for operating the government comes from sample surveys. The IRS, for example, has used sample surveys in its Audit Control Programs since 1948, and as an estab- lished part of its Taxpayer Compliance Measurement Program (TCMP) since 1962 (Long, 1980:551. These surveys of tax returns and filing com- pliance in the general population have become a principal means for the IRS to set its enforcement strategy. Major short-term policy indicators on unemployment and the cost of living are based wholly or in part upon sample surveys. The Survey Division of the Bureau of the Census has become one of its largest, quite apart from many other divisions within the bureau also operating sample surveys or collecting information through them. The Current Population Survey annually reaches about 1 in 1,000

42 ALBERT J. REISS, JR. households. No organization of any size remains unsurveyed by some gov- ernment organization (though not always by sample surveys). A third major area for sample surveys is marketing. Market research may be the dominant sector in sample surveying, surpassing the resources al- located to surveys by governments though data for precise comparisons are lacking. There are several kinds of market research. Sample surveys affect product development and sales strategies. They locate territories or populations for marketing a particular good or service. Surveys estimate the demand for new products or satisfaction with existing ones. The mass media, which rely on sample surveys for news, rely even more heavily on them for market information. No industry is more sensitive to the sample survey than tele- vision, in which ratings of network programs determine advertising revenues and the fate of writers, producers, and stars. As a fourth major consequence, the sample survey has become the major means of developing social indicators in postindustr~al society. Sample survey information is aggregated into indicators in two different, albeit related, ways. Surveys are used cross-sectionally at a point in time-to evaluate relative performances or outputs, as in the Nielsen ratings of television programs, or to compare electoral candidate strengths. Social indicators are also used to forecast, monitor, control, or respond to the course of change over time. For example, the monthly Current Population Survey estimates unemployment, residential tenure, and vacancy rates; the semiannual National Crime Survey examines victimization rates; the Annual Housing Survey reports characteristics of housing units; and the National Health Survey examines illness, use of health care services, and health- related expenditures. Sample surveys are also important in applied social science research, especially by nonacademic organizations. Not only has evaluation research become a substantial private industry, but major organizations such as the Armed Forces have developed a considerable in-house capability for sample surveys; it has been said that the most surveyed population in the world is the Armed Forces of the United States; certainly the American soldier in World War II served the most surveyed military in history (Stouffer et al., 1950). Finally, the sample survey is one of the major methodological foundations of the modern behavioral and social sciences. Despite widespread use in government and by profit and nonprofit organizations, major innovations and inventions in sample surveying continue to stem mainly from the ac- ademic social science community. Exceptions occur, primarily in the de- velopment of efficient means of surveying, such as computer-assisted telephone interviewing (CATI); yet even when such innovations occur outside the

MEASURING SOCIAL CHANGE 43 academic community, assessment of their utility and continuing innovation generally moves within it. This brief review of the pervasive effects of two major behavioral and social science inventions human testing and sample surveys illustrates their major impact on patterns of life in modern societies and draws attention to the possibility that the relatively lower scientific prestige of the behavioral and social sciences rests in part on their not studying the social impact of their inventions. Were there systematic investigations of such inventions and their effects, we might discover that in postindustrial society behavioral and social science inventions are more consequential for social change than material inven- tions. Ogburn developed his theory of cultural evolution by focusing on the material inventions and advances in physical science and mathematics that contributed to the Industrial Revolution. That view scanted the great social inventions of earlier societies, such as bureaucratic administration and empires (Eisenstadt, 1963) and antedated most of modern behavioral and social science.3 The role of economics in setting government policies and in the social control of economies has grown considerably since the work in Recent Social Trends. Although a president had sought the advice of academic social science in the "President's Research Committee on Social Trends," the committee seemed not to have imagined the significant role that behavioral and social science inventions would come to play in corporate organizational life and government in America. Ogburn believed that the cultural base of social invention accumulated less rapidly in modern times Man Tat of mechanical invention (Ogburn & Nimkoff, 1940:7921.4 This slower grown, in turn, slows the rate of new social invention. Yet there appears to be greater accumulation in the behavioral and social sciences Tan Ogburn expected. Rapid expansion of the knowledge base has been especially evident in cognitive psychology and linguistics. A final word may be in order here on the reluctance to examine the impact of behavioral and social science inventions on society and especially on social change. Lags in adaptation due to such inventions may be intrin 30gburn observes en passant: "The fact that technology is at present so powerful a cause of cultural lags, and consequent social disorganization, does not deny that other variables such as social inventions or population changes are creating lags also . . . the lag of social changes behind technological progress is simply a special case of the general phenomenon of unequal rates of change of the correlated parts of culture" (Ogburn & Nimkoff, 1940:893). 4The matter is empirical. It is not clear that the cultural base of social inventions cumulates any less rapidly in the modern world. Boulding (1978) argues that the homogenization of societies throughout the world may lead to less diversity in the cultural base and thus in the long run threaten the survival of culture.

44 sically shorter than for material inventions. But also, the dominant social theories have conceptualized societies as relatively stable structures, with an emphasis on the ways that such stable structures are maintained.S Models of social structural change seem less well developed, less often tested, and more focused on radical or revolutionary change than on ordered but ac- celerated change.6 The literature on organizations, for example, emphasizes the resistance that organizations display to deliberately contrived interven- tions. This strategy of theory construction and testing downplays the im- portant ways that inventions occur and are diffused in society most often other than by deliberate intervention and promotes the false premise that invention and intervention are ordinarily successful in producing change, except where organizational resistance is powerful enough. The contrary seems to be the case. Most experiments and inventions fail, or succeed in producing entirely unintended effects. We may learn more about how to produce intended effects through social invention by looking to the unin- tended consequences of purposive social action (Merton, 19361. ALBERT J. RElSS, JR. Reduction of Cultural Lags Although Ogburn subordinated the role of behavioral and social science inventions in causing cultural charge, 7 he assigned to these sciences a special role in facilitating Me agitation of society to changing material culture (1934:1661. Ogburn believed Mat Me failure of ~nshtudons to adapt to ad- vanc~ng technology produced nearly all social maladjustment and disorgaru sOgburn (1957b:8-9) concluded that the study of social trends carries two major messages: "The first general message that knowledge of social trends brings to us is that there is much stability in society, even though there be a period of great and rapid social change.... The second lesson we learn from a knowledge of social trends is that there is a sort of inevitability about social trends.... It is difficult to buck a social trend. It may be slowed up a bit, but generally a social trend continues its course.... Success is more likely to come to those who work for and with a social trend than to those who work against it." 6Antipathy toward military institutions, for example, may account for a general neglect of how organizations may change quite rapidly and as a consequence of social inventions. In the history of race relations in the United States, for example, little attention is given to how the U.S. military organizations became egalitarian and at an accelerated rate compared with any other sector of American society (and that religious organizations are among the most recalcitrant to change and racially segregated at the local level). 7In Part VII, "Social Change," of Sociology, Ogburn recognized that assigning a priority to mechanical invention is partly a function of the precision with which an invention can be dated. He also recognized the problem of an infinite regress of causation that complicates assignment of priority in social change. He concluded with a mechanical analogy: "When all the interconnected parts of a culture are in motion, and each part exerts a force on some other part, the origin of the motion cannot be located" (Ogburn & Nimkoff, 1940:866-867).

MEASURING SOCIAL CHANGE 45 zation (Ogburn & Nimkoff, 1940:8901. In a 1957 addendum to the theory of cultural lags, Ogburn (1957a: 172) reasoned that lags accumulated more rapidly in modern society because of the volume and accelerated rate of technological change. Although acknowledging that lags might be reduced by retarding the development of the natural sciences or following Stamp's (1937) suggestion for a moratorium on mechanical invention, he did not take these suggestions senously, believing that such courses of action required too high a degree of planning and control (Ogburn & Nimkoff, 1940:8901. Although the accu- mulation of lags was thus inevitable, it could still be reduced. For example, wars and revolutions reduce accumulated lags in a society (Ogburn, 1957a:1721. Another less radical way to reduce lags is Trough the technology of the behavioral and social sciences (President's Research Committee on Social Trends, 1933: 1661. But just how to achieve this Ogburn failed to make clear. The answer would have to lie in the production of knowledge-based innovation and invention designed to increase adaptation to cultural changes or to reduce the effects of their accumulation. Below I will illustrate two different ways in which social science- both basic and applied-can function in restructuring societies in consequence of changes in culture. Statistics and Quality Control The invention and diffusion of statistical quality control illustrates how social inventions can cope with the cultural dislocations caused by material and nonmaterial inventions. The coalescence of mechanical inventions into the modern mass production assembly-line factory produced the problem of assuring uniformity and high precision. Departures from strict production standards have consequences ranging from mechanical failure to increased transaction costs; these can be very signif- icant in competitive markets or under other conditions where the tolerance for failure is small. Statistical quality control is Me statistical surveillance of repetitive processes. It is used primarily for two purposes: process control to evaluate future per- formance and acceptance inspection to evaluate past performance (Walks & Roberts, 1956:4951. In either type of control, samples are drawn to make decisions about a population. For process control, We population is an infinite number of expected results from repetitions of Me same process; for acceptance inspection, it is Me quality of a finite set of existing items. The basic invention of statistical quality control was developed in the 1920s by an industrial statesman, Shewhart,8 who invented the statistical quality control chart (1925, 1926a, 1926b, 1927, 1930, 19311. Its wide ~Shewhart dates the invention of the statistical quality control chart as 1924 (1939:4).

46 ALBERT J. REISS, JR. spread dissemination came in the 1940s and resulted from the demands of the War Production Board, which deemed quality production of military goods essential to winning the Second World War, especially in light of the high quality of the German industrial complex (Walks & Roberts, 1956:495, 5121. Wald's method of sequential analysis (1945), although developed initially for use in scientific research, proved so useful for ac- ceptance inspection that an estimated 6,000 U.S. plants used it within two years of its development in 1943 (Walks & Roberts, 1956:5181. Other organizational innovation accompanied this rapid diffusion. Inten- sive training courses in quality control were developed at Stanford Uni- versity and given in most major industrial centers during the war. Among the many consequences of diffusion was the founding of the American Society for Quality Control, made up largely of applied statisticians working in industrial applications.9 Ogburn concluded from his studies that the acceptance of inventions and their integration into cultures other than the one of origin depended upon the similarity of the cultures involved (Ogburn & Nimkoff, 1940:8291. He was also disinclined to assign causal roles to individuals either in invention or diffusion (Ogburn, 19261. For Ogburn, the existence of independent invention demonstrated that the cultural base predominates over individual ability or uniqueness. Ogburn's view may be correct in the long run, but in the short-run case of quality control, there were key individual disseminators. One of these was W. Edwards Deming, a government statistician originally in the De- partment of Agriculture and later at the Bureau of the Census and on independent government assignment. The introduction and rapid diffusion of statistical quality control in Japan seems largely due to the efforts of Deming. Since 1951, the Union of Japanese Scientists and Engineers has recognized his importance to Japanese industry by creating a major award, the Deming Prize, for contributions to statistical quality control in industry (American Statistical Association, 1983:11.~° Some believe that the com- petitive margin of Japanese over U.S. products is attributable to a higher integration of statistical quality control in Japanese industry. 9Although statistical quality control was initially developed and applied in industry, the invention has wide applications since it is applicable to any kind of repetitive process, e.g., communicable diseases, medical experiments with human subjects, and accounting processes. Esthete is no Deming Prize in the U.S., although he was honored in 1983 by the American Statistical Association for his contributions to "statistical quality control at home and abroad" with the Samuel S. Wilks Medal Award. Deming also has been decorated for his work in the name of the Emperor of Japan with the Second Order Medal of the Sacred Treasure. Nearing age 83, the peripatetic Deming was absent from the award ceremony, unable to fit it into his schedule without a few months' notice! (American Statistical Association, 1983:1).

MEASURING SOCIAL CHANGE 47 Cohort Analysis A second example of how behavioral and social sci- ences permit adaptation to social change is the use of cohort analysis. A cohort is an aggregate of individuals of similar age who are exposed to or experience certain events during the same period of time. Cohort analysis is a quantitative description and analysis of occurrences from the time a cohort is exposed to these events (Ryder, 1968:5461. The continued entry of new cohorts provides a continuing opportunity to modify society. Cohorts consequently are central to the study of social change. But there also may be effects associated with age or aging per se, and changes brought about by external influences or events that affect all people alive at the time. These three sources of change in a population are referred to as cohort, age, and period effects. A cohort analysis, as Ryder (1968:550) points out, differs from a lon- gitudinal or panel analysis in that the latter examine changes in the individual members of a population or sample over time, while cohort analysis ex- amines the changing characteristics of an aggregate through time: it is macro- rather than microlongitudinal. The value of a cohort analysis to our understanding of social change can be illustrated by the studies of changing attitudes toward racial integration in the United States (Taylor et al., 1978:481. Opinion polls between the 1950s and 1980 showed considerable shift in white attitudes favoring racial integra- tion. Underlying that shift, however, were different cohort trends. Although all age groups showed some shift with aging, this factor accounted for only about 10 percent of the total attitude change. Almost half of all change was due to the succession of cohorts in the population, with older, less favorable cohorts being replaced by new, more favorable ones. Almost half of the change in favorableness by 1980 is due simply to Rose younger cohorts comprising an ever greater portion of Me population. By simple extrapolation we would forecast Rat within a matter of decades We vast majority of We population will favor racial integration. This type of cohort analysis shows Rat lag re- ductions often occur Trough We mechanism of population replacement. ~ ~ But cohort analysis does not substitute for theoretical models of what causes particular changes. In the example, we still need to explain why the younger cohorts are most favorable. Is it due, for example, to indoctrination, to greater contact with unlike persons in environments such as schools, to involvement in social movements that support certain racial attitudes, or to some combination of these and other explanatory variables? While cohort analysis can aid us in under- standing changes at the population level, it does not provide a substantive theoretical explanation of how such changes occur at the macrolevel of individual members of that population or at the microinstitutional and organizational level of changes. The failure to develop explanatory micro- and macromodels of social change severely limits our understanding of it. For a more extended discussion and set of examples of uses of cohort analysis, see Reiss (1982b).

48 ALBERT J. REISS, JR. Cohort analysis has become one of the more powerful instruments of social planning, especially when linked to social and demographic fore- casting. It has figured prominently, for example, in calculating changes in the Social Security system, in planning health care prison and prisoner legislation. Cohort analysis does not necessarily depend in the short run on forecasting the birthrate. One does not, for example-, have to forecast the birthrate to estimate in 1980 the size of the 1990 prison population, although forecasts or assumptions about rates of death, immigration, offending, and sanc- tioning are required. But given relatively constant rates of offending or transitional probabilities, one can forecast, for example, how much of the expected annual change in crime rates can be attributed to changes in the age composition of a population. Blumstein, Cohen, and Miller (1980), for example, demonstrated that the prison population in Pennsylvania is es- pecially sensitive to changes in age and race composition and that due to these characteristics of living cohorts, arrests would peak in 1980, prison commitments in 1985, and the prison population by 1990. Blumstein also apprised the Pennsylvania legislature that sentencing legislation it was con- sidering could have a considerable immediate impact on the size of the population in Pennsylvania prisons, exacerbating overcrowding. The in- crease in mandatory sentence length for selected offenses originally con- templated by the Pennsylvania legislature in 1976 would have increased the prison population by 50 percent at peak. Recent work by Greenwood (1982) has shown that changes in the selection of inmates, assuming the appropriate procedures can be developed, have the potential to reduce the size of prison populations. Building prisons that take a considerable period of time to complete, do little to solve overcrowding during the building period, and result in overcapacity soon after their completion would be ill- advised. Such cohort analyses appeared to dampen the expansion of prison capacity, though clearly decisions to increase prison capacity are subject to other contingencies (Blumstein, 19831. Demographic and economic forecasting is a part of behavioral and social science analyses that has become an institutionalized element in managing economies. Forecasting is not only the province of the President's Council of Economic Advisers and the Congressional Budget Office but of large businesses, many of which produce their own commercial forecasts or adapt forecasting models to their ends. As Thompson (1967:21-22) concluded, when organizations cannot deal with environmental fluctuations by buff- ering their technical cores or smoothing input and output transactions, they will seek to anticipate environmental changes by forecasting them. Such forecasting permits planned adaptation to anticipated changes. Many pro- ducers now make medium- to long-term investment and production deci and in considering

MEASURING SOCIAL CHANGE 49 signs by estimating how cohort succession can be expected to affect the market for old and new products. CONSEQUENCES OF SOCIAL CHANGE FOR MEASURES AND MEASUREMENT Concepts and Measures as Products of Social Life The social process itself is the source of most basic concepts and pro- cedures of social measurement (Duncan, 1984:ii). All ways of knowing are socially organized (Biderman & Reiss, 1967), and all methods of inquiry into social life are subject to its substantive laws even as they attempt to discover and test those laws (Reiss, 1980~. Duncan's recent work on social measurement (1984) draws attention to the fact that many of our basic concepts and procedures of social mea- surement such as voting, counting people, money, social rank, rewards and punishments, randomization, and sampling did not originate in the pursuit of scientific knowledge but rather as the consequence of practical problem solving. Not only do we depend upon social processes to invent many of our concepts and measures, but the development and maintenance of a social science depend in the long run, as Duncan notes, upon "what the society wants or allows to be measured and is able and willing to pay for. How it will be measured or, in any event, the socially tolerable limits on concepts and measurement-is also socially determined." (Of course, phy- sicists and astronomers who struggle for appropriations to finance massive particle accelerators or space science vehicles face similar constraints.) Not the least of purposes in acquiring information for statistical indicators is to gain support for particular courses of action. Florence Nightingale's (1858) work on the collecting and processing of statistics to monitor sanitary conditions in British army life illustrates how both the origin and institu- tionalization of statistical indicators depend upon social purposes and pro- cesses. Nightingale collected information on deaths due to preventable causes in British army hospitals at home and in the Crimean War and invented statistical graphics to depict that information as part of her cam- paign to improve the sanitary condition of army hospitals. She even authored an anonymous publication, Mortality of the British Army, to mobilize public and parliamentary attention to her diagrams. In private correspondence she referred to these unique graphics as coxcombs because of their shape and the red and blue colors she used (Cook, 1914:376; Funkhouser, 1937:3441. Nightingale discoursed on the "inaccuracy of hospital statistics" and "the want in General Military Hospitals of a Statistical Department engaged in registration" (1858:288-332) and succeeded in securing such departments.

so ALBERT J. RElSS, JR. Because behavioral and social concepts and measures originate in social processes and require support if they are to be maintained, a national statistical system to measure and monitor social changes can survive only if it is folly institutionalized. Those familiar with the polling industry know how difficult it is to find a record of the same attitude or opinion over time, since repeating of questions depends upon perceived public interest and the salability of information. Bases of support for individual indicators often are eroded by social change. Even the quinquennial U.S. Census of Manufactures was abandoned during World War II. Those fa- miliar with attempts to institutionalize the collection of certain basic indi- cators of attitudes and opinion know also how much the National Opinion Research Center's General Social Survey depends on continuing renewal of support from the National Science Foundation. The only annual time series to span U.S. history are those required by the Constitution for pur- poses of apportioning democratic representation and for taxation. Over-time measures not only depend upon social processes for their origin and maintenance but may change subtly as a consequence of those processes. This is apparent in the development of time series about law enforcement and legal regulation. The behavior of such series depends upon diverse processes such as legislation, changes in executive authority, bureaucratic administration, the changing technology of data collection, and reporting practices. Major fluctuations in the enforcement actions of the Food and Drug Administration from 1919 to 1977, for example, can be traced to changing legislative mandates for enforcement, the rise and fall in the budget for inspectors, and changes in statistical reporting practices (Reiss & Bid- erman, 1980:229-2411.~2 The major lesson here is that our understanding of the behavior of any statistical indicator series depends heavily upon monitoring- and also mea- suring the social processes that affect the concepts and measures used in that series: We must model and measure the behavioral processes underlying the series. This enterprise does not consist simply of modeling error in the measures, as it is conventionally viewed. For hidden within such "error" is change itself. The seeming increase in litigiousness of the American public might, for example, be attributable almost entirely to the growth of laws, regulations, legal rules, and the scope of judicial opinion. People None of the few recommendations of the President's Crime Commission on Law Enforcement and the Administration of Justice ever implemented was the institutionalization of the National Crime Survey of Victimization by Crime developed by social scientists. Some even regard it as a precarious institutionalization since it has had to survive frequent threats of elimination. Annual estimates have been made from 1973 to 1983.

MEASURING SOCIAL CHANGE 51 may be no more litigious perhaps even less so in matters that could always have been litigated. Stated more technically, measures of validity and reliability are themselves grounded in postulates that compromise the measurement of change. Social Change and the Organization of Ways of Knowing The mea- surement of social life and changes in it depends upon socially organized ways of knowing. The concept of a "real" or "true" rate of crime, for example, independent of organized efforts to detect and measure crime, is illusory (Biderman & Reiss, 19671. There are no rates without an organized intelligence system to demand, collect, and process people's accounts, whether the people are scientific observers, police officers, victims, or jurists. More generally, there are only socially organized ways of knowing because all criteria (and measures) for knowing, defining, and processing social facts lie in social organization (Biderman & Reiss, 1967:91. Ogburn conceptualized the organization of ways of knowing as a social invention, and indeed our understanding is enhanced if we see knowledge as the product of invention. Many small inventions, for example, went into what we think of as the modern sample survey: not only inventions such as statistical probability, sampling, and analytical techniques, but organi- zational procedures to train and supervise people in acquiring information in interviews and to link these people together in a sequential process. These observations have a number of implications for the study of social change. One is that we must understand how these socially organized ways of knowing change over time as a consequence of material and nonmaterial technology and the effect that such changes have on our concepts and measures. When one moves from in-person to phone interviewing and from an interview schedule to a computer-assisted recording, one cannot simply assume that these will produce the same results, for each is enmeshed in a somewhat different social modality. The second implication is that more attention should be paid to how changes in organization affect the definitions of measures and to under- standing how information reflects the character of its collection, processing, and reporting. We have come to realize that processes such as illegal immigration affect organized ways of knowing such as censuring. Due to the unequal residential distribution of illegal immigrants, the resulting "un- dercount" of the population can affect state representation on the House side of Congress. The magnitude of the undercount can be reduced by other socially organized ways of knowing, including multiple record systems and new statistical methods of estimation (Fienberg, 1972; El-Koratzy et al., 1977; Ericksen & Kadane, 1983; Levine et al., 19851. But few scientists or others appear to realize that the very concept of an undercount is rooted

52 ALBERT J. RElSS, JR. in socially organized and epistemologically dubious assumptions the no- tion, for example, that a method exists whereby a large, mobile, dynamic population can be exactly counted or characterized at a fixed point in time by enumerating everyone who resides at a fixed location at that point in time. The Paradox of Method All methods for inquiry into nature are gov- erned by the substantive laws of nature at the same time that these methods are the means for discovering and testing those laws (Reiss, 1980:11. This paradox can be resolved only by approximation: improvements in methods will advance the formulation and testing of substantive theories, and re- ciprocally, improvements in substantive theory become available for ad- vances in method. The paradox of method can be regarded as parallel to Kaplan's (1964:53-54) paradox of conceptualization: "The proper concepts are needed to formulate a good theory, but we need a good theory to arrive at the proper concepts." This too can be dealt with only by successive approximation. In exploring the study of social change, it is important to grasp fully one of the important implications of the paradox of method: Since tests of theories and the development of knowledge depend upon methods, the development of substantive knowledge underlying methods is essential. Put another way, research on methods is of strategic importance because the development of all knowledge depends upon it. The knowledge most central to We development of any science, then, is the substantive theory and testing germane to its methods and measures. Some behavioral and social sciences depend rather heavily upon indirect modes of observational such as asking subjects to provide information about past behavior. These methods rest on theories about remembering. The methods are designed to retrieve information from long-term memory by short-term recall probes, but the theory of how subjects get information from the past into the present and report it is far from adequately tested. We readily recognize that other processes may intervene, such as forgetting and deception (including self-deception). We cannot understand the methods that rest on these postulates without further understanding of memory and motivation. That understanding in turn rests, at least in part, upon testing the theory using these methods; hence an instance of the paradox. Note that this observation differs from i3This predilection for indirect over direct methods of observation is atypical among the sciences; whenever possible, the traditional sciences opt for direct observation. The predilection is not easily explained, though some social processes impede direct measurement (Reiss, 1968, 1971, 1976).

MEASURING SOCIAL CHANGE 53 that underlying the use of multiple methods of measurement. There the argument is simply that any particular method taps multiple processes and sources of variance (Webb et al., 1966:4~. Here the point is that any method is inextricably woven with substantive theories about the behavior under- lying the method. Even so, the assumption of multiple confirmation in the multimethod approach (Campbell ~ Fiske, 1959; Webb et al., 1966:5) depends upon substantive theories and knowledge: it rests on the assumption that components can be weighted according to their known extraneous variation and in combination by their independence from the same sources of bias. Indeed, one problem of a multimethod approach is to determine why methods are not substantially of equal weight.14 There seems to be no escape, then, from the paradoxes of theory and method. We must be prepared, consequently, to devote considerable effort to understanding the substantive theories that underlie our methods. Indeed, it can be argued that the most critical theory for the social sciences is that germane to its methods (Reiss, 1980~. We must therefore draw two more implications of the paradox of method for the study of social change. First, concepts and measures used to in- vestigate social change are vulnerable to secular changes. Second, the methods of measuring and analyzing social change are vulnerable to secular changes. The problem for theorists and empirical investigators is how to measure social change when both the measures and that which is being measured are changing. Or correlatively, how to measure social change when that which is being measured is changing while the concepts and measures are too limited or rigid to detect it. The task seems inordinately complex, but must be faced if we are to scientifically understand social change. Consequences of Institutionalizing Measures of Changes The difficulties social scientists encounter in measuring social change may preclude the kind of precision we commonly associate with the physical sciences. The past may never be kept in such a way that we can tie it to the future, once we discover or invent new ways of measuring change. Moreover, any evolutionary or dynamic theory of change providing for emergence or forecasting faces problems of how to tie the present to the past and the future. Tithe law of evidence and of torts similarly recognizes that all methods are not to be weighted equally (Prosser, 1964), though it resolves the matter by legal rather than scientific criteria of evidence.

54 ALBERT J. RElSS, JR. Most indicators of change emerge from social processes, and their future depends upon such processes. Consequently, social indicators are subject to at least two major types of changes. On the one hand, the broad processes of change affect the concept and its measures, a matter we shall return to below. On the other hand, the organization responsible for data collection will, from time to time in response to discontinuity between the concept, its operational measures, and changed conditions redefine the concept, substantively and operationally, or introduce new methods or techniques of data collection and measurement. The changes may be perceived as occumng within the phenomena under continuous observation or they may be accretions to that class of phenomena. Examples of how concepts are redefined in accord with changes in the observed phenomena are the repeated modifications of the U.S. Bureau of the Census in the definition of a dwelling unit and a household. These changes respond largely to the way that housing and living arrangements change from one decade to the next. The changing definition of "head of household" by the bureau is a response to public representations that the concept was biased toward older persons and males, and biased in opera- tional procedures and conceptualization. Other sex-linked census concepts include "secondary family workers" and "housewife in the labor force." Yet other seemingly simple census concepts vulnerable to secular change include "ethnic status" (vulnerable to changing patterns of marnage), "na- tive language," "county of ongin," and "ethnic status of parents." Examples of increments to a class occur quite commonly for legislated concepts. Congress, for example, mandated the collection of information on arson as a violent index crime in Uniform Cnme Reporting. A somewhat extended example may best serve to illustrate how new concepts and new measures emerge in response to changed conditions in society, and how old concepts may no longer measure the same conditions as a consequence of changes The example is drawn from the history of measuring employment. For much of our industrialized history, a major way of describing the economy was to measure the economically active, employed population. This example was used in an earlier discussion of this problem in measurement (Reiss, 1982b). Robert Parke assisted in developing that example. Esthete are equally fascinating questions of measuring employment in preindustrial history and its relationship to industrial employment. How does one conceptualize and measure employment, for example, in household-based economies or cottage industries? Is the notion of employment so closely tied to a form of social organization involving an employer and an employee who derives income from that work that it becomes inapplicable to earlier periods of history? In what sense can persons in earlier periods be described as "self-employed"?

MEASURING SOCIAL CHANGE 55 This was conceptualized in terms of a workforce comprised of all gainfully occupied workers in the nation. Moreover, each worker was considered to have a usual occupation, which was asked about as a means of providing information on labor resources (Jaffe & Stewart, 1951; Shiskin, 19761. Labor resources were regarded as important to jobs, yet only occupational status, not employment status, was usually measured. The 1930 Census, for example, reported statistics on the gainfully employed work force by their usual occupation. There were no national statistics on unemployment, because this usual occupational status concept of the work force did not include current employment status. A conception of unemployment linked to the business cycle had, of course, existed for quite some time. But prior to 1930, interest in unem- ployment was largely confined to labor unions. Unions had few resources for systematic documentation. Unemployment during the episodic panics and depressions of the nineteenth and early twentieth centuries led to oc- casional estimates of unemployment and, beginning in the 1920s, a growing concern with the effects of changing technology led to studies of shifts in occupational composition, employment, and unemployment in industrial sectors undergoing rapid technological change. There were also estimates of seasonal unemployment in agriculture. But all of these estimates were based on a presumption that unemployment was a temporary dislocation. Consequently, there was surprisingly little statistical information on un- employment in the chapter on labor in Recent Social Trends (President's Research Committee on Social Trends, 1933:xvi). The national statistical system was unprepared to measure it. As Jaffe and Stewart (1951:7) con- cluded, although gainfully occupied statistics may have been useful for social policy in the nineteenth century, they were of little value during the Great Depression. Being collected only once every 10 years, they had especially little value in illuminating the immediate problem of mass un- employment. Although unreported in Recent Social Trends perhaps be- cause it was only a single observation in time- the 1930 Census was the fast attempt to measure national unemployment by asking all persons re- porting a gainful occupation whether they were at work the preceding day. That estimate, limited as it was, was soon considered far too low. Although sample surveys were just emerging, government agencies attempted to measure unemployment by surveys in 18 cities during the early 1930s and nationwide in 1937 in the so-called Biggers survey Qaffe & Stewart, 1951:9~. These early attempts to measure unemployment showed the limitation of the concept of a work force made up of gainful workers in usual occupations. Under this concept, the only people who could be counted as unemployed were those "established workers" who did not have jobs. Excluded from the unemployed were those without previous employment, such as young

56 ALBERT J. REISS, JR. people entering the work force for the first time. Housewives, occasional workers, and those who earned only supplemental income were also gen- erally omitted. Not regarded as gainful workers, they would not be counted as unemployed when they sought but could not find work. The work force of the gainfully employed was subject to attrition only through retirement, injury, death, and emigration; it was augmented only through the maturation of young people and immigration (Bancroft, 19791. Of course, the idea of a usual occupation or a customary status such as that of housewife was consistent with a stable society based on fixed statuses, not a rapidly chang- ing urban economy with mobile factory employment. Customary statuses are unlikely to change in the short run; hence, measurement could occur at long intervals such as the decennial census. The rapid growth of unemployment during the Great Depression and public interest in short-run unemployment changes emphasized a need for frequent estimates of unemployment and people who wanted indeed needed-jobs. This required a different concept of the relationship of in- dividuals to the labor market. The gainful worker concept had emphasized occupational status and experience as resources available to fill jobs. The emerging concept was a labor force made up of all persons at work, looking for work in the preceding month, or temporarily laid off with an expectation of being called back to work. This concept emphasized the current activity of individuals, and what emerged was monthly measurement. Under work force concepts, an increase in the gainfully occupied necessarily meant a decrease in the unemployed. Under the labor force concept, the unemployed are measured somewhat separately from the employed. The numbers of employed and unemployed can both rise (or both fall) if there are persons entering the labor force from among those not in the labor force, such as first entrants, housewives, students, retired persons, and those discharged from military service.~7 Much, then, has changed: the concept of participation in a labor market, He ways that employment status and activity are measured, and the fre- quency with which measures are taken. Yet the matter does not end there. In 1961, President Kennedy appointed a Committee to Appraise Em- ployment and Unemployment Statistics, partly based on the awareness that He notion of unemployment had both conceptual and operational difficul ~7The consequences of this quasi-independent variation can be a political paradox. In the 1976 Ford-Carter debate, Carter correctly claimed that the number of unemployed had risen in the most recent month and President Ford correctly responded that the number of employed had risen. Four years later, the same claims were correctly made in the Carter-Reagan debates with Carter's role reversed. (This example is not offered to suggest there invariably is a presidential statistic and a candidate statistic on the labor market!)

MEASURING SOCIAL CHANGE 57 ties. The line between a person looking for work and not looking was vague, and the reasons for not looking for work were many. In the end, the committee decided to exclude "discouraged" workers, persons who had given up looking for work because they saw little chance of getting a job. In 1967, the Bureau of Labor Statistics implemented that recommen- dation and excluded a substantial segment of the unemployed by restricting the concept of looking for work to those searching for work during the preceding four weeks. At the same time it began to record persons as "discouraged" if they said they were not looking for work but would work were there a job for them. As recently as 1979, Finnegan advised the National Commission on Employment and Unemployment Statistics (1979:215-217) that while the practice of reporting discouraged worker statistics- measured every quarter rather than monthly should be continued, this group should not be counted as unemployed. But the matter did not end there either. Discouraged workers are disproportionally distributed among persons under 20 years and over 60 years of age and among blacks and women. Should such persons not be regarded as in some way unemployed? Clearly, short-run as well as long-run changes in the society are at the heart of such conceptual and political debates. Our example of unemployment highlights a central problem in the mea- surement of social change that the concepts, operational definitions, and measures used to chart this change are themselves changing over time as a result of social processes. What seems called for is far more inquiry into how one adjusts concepts and their measures over time. Splicing of mea- sures, synthetic estimation, and multiple measures all provide some means for coping with this issue. Yet there is all too seldom provision for adjusting statistical series to secular changes. Some Consequences of the Organization of Statistical Indicators No society could afford the effort to acquire and organize all of the statistical information that might usefully enter into decisionmaking. But societies vary considerably in the extent to which they institutionalize sta- tistical indicators and the uses to which they are put. Japan, for example, has a national statistics day; the United States does not. The classical bureaucracies of Japan and some other countries such as France may have institutionalized a respect for statistics that is less well developed in the United States. There is no national statistical system in the United States, in the sense that no one really knows what statistics are collected annually. Many private organizations collect and compile annual series. Any major university has probably several hundred such series. Even individual mem

58 ALBERT J. REISS, JR. hers of a population keep their own statistics. Indeed, were an inventory taken of statistical series, government series would probably contribute the minority. Our concern in this section is with the consequences of the current organization of statistical indicators for monitoring and investigating social change. Perhaps the major consequence of the current organization of our statistical system is that we cannot readily compile them into meaningful aggregates beyond the level at which they are collected. Judicial statistics are gathered for every court in the land, but it is difficult to combine more than a small part of them into state (much less national) statistical series. Conversely, most national statistics are collected in such a way that we cannot make local estimates from them. This is especially true for series collected by the survey method, but it is so for many other modes of data collection and compilation as well. A second major consequence is that statistics gathered by private orga- nizations generally are inaccessible for aggregation or analysis unless re- porting is centrally coordinated and controlled. Private organizational data are seldom compiled unless there is legislation creating a voluntary or mandatory system of reporting. As a consequence, we cannot measure very much change using that vast resource. A third consequence is that the major developmental and analytical re- sources are concentrated in federal statistical systems designed to meet particular needs of federal legislative, executive, and judicial agencies. The resulting lack of attention to local variation may have the consequence that matters requiring collective attention as well as social changes come to be defined for a national aggregate rather Can in terns of their local variability. To the degree that statistical information is important in reaching decisions, this imbalance may bias toward federal rather than state and local adaptation. Countries with central statistical systems such as France were historically organized to gather information for each department and unit thereof. It might be useful to learn more about the role such regionalized statistical systems play in social change in contrast with the United States. I would draw attention to the problem of developing concepts and measures of social change that meet agreed-upon requirements of all levels of government. Volunteer sample surveys, rotating panels, and synthetic estimates are some of the ways of doing so. This is an area for social invention. THE CONCEPTUALIZATION AND MEASUREMENT OF SOCIAL CHANGE Most current theories and models of social change are deficient in ex- planatory precision and predictive power. This section considers some of

MEASURING SOCIAL CHANGE 59 the ways that contemporary models and concepts could be improved to better our understanding of social change. Individualistic Biases in Studying Social Change The dominant theories of social life in the United States postulate indi- viduals as basic units and especially presume individual actors who make rational decisions. References to collective choice and organizational actors are often translucent, magnifying postulates about the behavior of individ- uals to collective actions. Durkheim's view of society as a reality sui generis is honored more in the breach, as interpretative commentary. Ogburn was well aware of the domination of individualism in explaining social change, formulating the problem as the role of the "great man" versus "social forces" (Ogburn, 1926~. His own well-known view was Hat great men are but a medium of social change (1926:2311. Historical theories, such as those of Sorokin (1937-41, 1943, 1947), likewise assign a key role to social forces, which dominate He behavior of individual actors. But the con- ha~y emphasis upon individual actors and individual welfare still dominates much contemporary theory and research, biasing He Reagent of social change. Earlier I used unemployment as an example of a major social indicator. Unemployment is a characteristic attached to persons. So is the concept of the discouraged worker, based as it is on motivation of individuals in a labor market. Even though it is apparent to labor economists that employment attaches to jobs, we lack indicators of job vacancies per se; Here is no national indicator of jobs comparable to that on unemployed individuals. In his ground-breaking studies of vacancy-chaining, Harrison White (1970) noted that social scientists had focused on social mobility as individual move- ment through jobs, neglecting the fact that a job, as a position in an organi- zation, must open up or become vacant to constitute an opportunity for mobility. White studied movements of vacancies Trough bureaucratic hierarchies-how vacancies are filled and how the filling of positions relates to organizational structure and process. Positions and opportunities tend to be linked in chains, and these constitute the relative openness to upward mobility in a bureau- cracy.~8 Although recent studies of social mobility have examined how or- ganizational structures affect occupational mobility, defining these as opportunity structures (Rosenbaum, 1976, 1984), the study of mobility has not shifted to focus on changes in opportunity structures themselves. To the extent that theories dictate what is problematic, they also dictate Parenthetically we note that the trickle-down market allocation theory of housing in economics could be tested in a vacancy-chaining model.

60 ALBERT I. REISS, dR. the choice of measures. There is a predilection for developing instruments to secure information from and about individuals rather than from and about organizations, and our concepts and measures more often refer to individual statuses than organizational positions. The individualistic bias certainly predominates in methods development. The sample survey is built primarily around sampling individuals and is poorly developed to secure information about organizations. Indeed, most methods for collecting information on organizations actually collect infor- mation about individuals in organizations or rely on individual surrogates for the organization. Far too little attention is paid to measures other than surveys (Sinaiko & Broedling, 19761. Even when organizations are the object of study, the line of discussion is affected by the individualistic bias. For example, Kish (1965) noted that around the time of the first Sputnik, about half of U.S. high schools offered no physics; a quarter, no chemistry; and a quarter, no geometry. He then noted that this did not tell us how many high school students could take courses in chemistry, physics, or geometry, for the schools offering no such courses, though large in number, were small in size, accounting for only 2 percent of all high school students. It is clear that average school characteristics would give a misleading description of conditions for the average student. But Kish failed to point out the relevance of the original organizational statistic for deploying col- lective resources. For some purposes the distribution of schools is critical, e.g., if a government sought to equalize educational opportunities for all students. To do that, the government would have to merge schools or divide resources among existing schools; and in either case, a large number of schools would be involved. Fifty percent of all high schools, for example, might need a new physics laboratory and an instructor trained to offer physics. As a consequence, the market for physics teachers might change drastically and organizational consequences on teacher recruitment and training would be considerable. One can readily imagine a whole train of organi- zational, structural, and individual consequences stemming from how one reads these statistics and decides to act on them. We routinely conceptualize and measure the size and composition of populations of individuals but have only recently come to think seriously about doing so for populations of organizations.~9 Yet the size of the or- ganizational population in the United States is greater than that of individ ~9Neeworks are even more complex. Consider thee apart from unmarried siblings, the kinship network is never the same for any two individuals. For further discussion of research on organi- zations, see Hannan, in this volume.

MEASURING SOCIAL CHANGE 61 uals. Consider that a household is a form of organization, that it is not co- terminous with the family as a form of organization, and that these may be regarded as two distinct populations of organizations. Alexis de Tocque- ville sensed the multiplicity of organizations when he characterized America as a society of joiners; yet he focused primarily on the individual charac- teristics of the joiners and less so on the fact that American society was creating an enormous number of organizations for individuals to join, many such organizations persisting well beyond the involvement or the life of those who founded them. The suggestion here is that organizations may play as great, if not a greater, role in social change than individuals, and that the bias toward individualism fails to take into account how populations of organizations are both causes and effects of such change. This, indeed, is not the only consequence of the individual bias; other units are neglected as well, such as units of culture. It has perhaps seemed simpler to count individuals when faced with the difficulties of devising and counting organizational popu lations or cultural products. Ogburn held that change in material culture invention is fundamental to social change. One test of his theory required counting the numbers of inventions so that one could calculate the rate of growth of the technological base. Although he used patents to count the growth of invention, he rec- ognized the limitations of this indicator. It perhaps is unfortunate that he did so little to try to count social inventions. Ogburn's neglect of how one conceptualized and counted social inven- tions should be seen in historical perspective. Societal intelligence on the growth of science and technology is little advanced over Ogburn's day. At the core of counting inventions are conceptual problems of what constitutes invention and of how one measures the growth of knowledge in the sciences. Some 30 years ago Lazarsfeld and Barton (1951) wrote a piece on mea- surement in the social sciences for a volume edited by Lerner and Lasswell on the policy sciences. They drew attention to the fact that while the individual was a primary unit of observation and measurement, there are also units that are not based upon the primary characteristics of individual members precisely because no individual data correspond to them. A com- munity can have a speed law; individuals cannot. A community can be characterized in terms of the proportion of its members who violate those laws. A primary characteristic of a unit, they noted, had to be distinguished from its analytic characteristics, which refer to component elements. Or- ganizations may be seen in terms of their individual members or in terms of properties that cannot attach to individuals. Although individual analytic characteristics may help explain social change, as, for example, the proportion of a society's manpower that is employed

62 ALBERT J. REISS, JR. in science, individual scientists do not and cannot have a technological base. The neglect of the latter for the former data may account for our being in a disadvantageous position to theorize on and measure social change. Even in our use of cohorts consider the examples used earlier in this paper we usually look at cohorts of individuals, rarely at cohorts of organizations. Bankruptcy is generally expressed in annual rates rather than survival rates in a birth cohort of organizations. The study of social change should focus, then, much more on the primary characteristics of organizations, which should be regarded more in terms of functional subunits than participating individuals. We can readily see that resources (such as laboratories) and relational properties (such as hi- erarchy) are primarily characteristics of organizations, not individuals. We must systematically collect better information on characteristics of orga- nizations and units of material and nonmaterial culture, to use Ogburn's terms, if we are to understand cultural and social change. Individual versus Collective Welfare There is a bias in welfare models of human behavior toward optimizing or maximizing individual welfare rather than the welfare of collectivities such as organizations.20 Trade-offs commonly are seen in terms of individual rather than collective costs and benefits. The quality of life is measured in terms of individual rather than collective units: Is this community a good one for scientists rather than for science? Is the housing stock fit for individual habitation rather than what kind of collective life is possible, given the housing stock? How one asks the question can make a difference. We look at the crime rates of com- munities in terms of victimizations in a population of individuals, neglecting the high rate of victimization of organizations and collective property parks, schools, playgrounds. We look at individual careers in crime rather than at community careers in crime (Reiss, 1982a, 19831; yet the latter may explain much of the former. Concepts such as justice, social cohesion, and social integration are not reducible to the lives of a society's individual members, nor can they be measured simply by summing observations for individuals. Changes in particular social indicators can have collective and individual effects. A change in the divorce rate, for example, is both a change in the status of 20Measures of social welfare as conventionally defined in political economy should not be confused with measures of collective welfare. Measures of social welfare typically are based on the concept of a collective consensus based on individual preference scores. Although such pref- erence measures may be technically infeasible, they presume that consensus measures on welfare preferences optimize or maximize collective welfare, which is an empirical matter.

MEASURING SOCIAL CHANGE 63 individuals and a change in social relationships and organizational structure of society. Most divorces increase the number of single-person households and decrease the number of two-or-more-person households. Divorces alter the relationships of husband to wife, children to parents, insurers to insured, and the taxable income and legal status of the parties, to mention only a few consequences of changes in the divorce rate. Such changes may produce chain reactions. The divorce rate can have a substantial effect on the size and occupancy rate of the housing stock, which may affect the burglary rate (a crime against housing units rather than individuals per se). Understanding social change would thus seem to require understanding of collective as well as individual welfare, and how changes in collective welfare are consequential for individual welfare. We may need to think more about the well-being of science in society, less about the consequences of science for the quality of individual life. Controversies over the risks of science must be viewed not only in terms of the risk to individuals, such as by gene splicing, but also of how the failure to do gene splicing research may affect the state of science in a competitive order of societies. Lags in Measuring Social Change Ogburn edited an annual series of the May issue of The American Journal of Sociology called Social Changes in [Year] from 1928 to 1935 and in May 1934, one entitled "Social Change and the New Deal." In the early volumes, Ogburn made clear that his purpose was not that of editing a conventional yearbook but rather "scientific analyses of social change . . . " (19291. The Great Depression, with its marked social changes, had con- sequences for the publication of his annual series. In his introduction to Social Change in 1932 (1933:823-824) he observed: The American Journal of Sociology has itself been influenced by these economic changes, and a policy of retrenchment in the interests of economy has affected the size of this special issue. We have had to reduce the number of the articles, as it did not seem possible to reduce the length of the articles further and have them of any scientific merit.... In order to do this, some of the topics covered regularly In the annual "Social Change" issue have been omitted.... In some cases the omission of certain topics is not a particularly serious loss because extensive data are not always collected every year in sufficient volume to note significant changes, and a two year interval will show the changes most clearly. This is true, for instance, in the case of social legislation. Most of our state legislatures meet only once in two years. The effect that changes in society can have upon its intelligence system is disclosed here rather dramatically. A second issue is also evident here. With what frequency shall we collect

64 ALBERT J. REISS, JR. measures of social change? Ogburn calls our attention to the fact that frequency of measurement is in part tied to the social processes themselves. Changes of some kinds especially those he would have characterized as adaptive are institutionalized, such as in the periodic meeting of legis- latures. The response to change will determine in part the scale and fre- quency of measurement and thus the capacity of science to detect and measure such changes. A third point was also mentioned briefly in Ogburn's introduction-the problem of lags in our intelligence on social change. He notes in particular the lag between an event, its measurement, and analytical understanding of it (1933:823-8241: There are few aspects of our social life that have not been markedly affected by this most severe economic depression of modern times. The papers in this volume indicate many of these changes and their effects. The extremely dramatic events, which began in the latter part of February and reached a climax in the most extensive closing of the banks ever known, have particularly significant effects. These, however, are not re- corded in this volume, which is restricted to 1932. Some time has to elapse after an event for the data to be collected and recorded so that it is possible to submit them to scientific analyses. News events are almost simultaneous, but there must be a lag before the scientific analyses can occur. Here we see a major and continuing issue in conceptualizing, measuring, and monitoring social change that of how our intelligence systems can be developed to collect information on events as they take place and how we can reduce the lag between collection of information and scientific analysis. We often fail to collect information rapidly that is essential to scientific analysis while, at the same time, far more information lies in our collection systems than we can process. How to resolve these problems is not altogether clear. A good theory helps, but data collection also depends upon social processes. Forecasting and testing likewise depend upon these processes. Since Ogburn's day, great strides have been made in time-series analysis by developing forecasting models and identifying causal and "leading indi- cator" models. Despite the inevitability of some lags in analysis and model testing, more attention still needs to be given to short-run forecasting as well as long-run theories of social change. Our capacity to measure and monitor social change depends upon using and encouraging all processes that represent investments in knowing about and understanding social change. Economics in particular has both a macro theory of change and methods of short-run forecasting. The theory develops episodically and partly as a consequence of social change; such concepts as stagflation emerge to cope with the inadequacies of the theory and to fit it more closely to a world "out there." Economics may have progressed rapidly precisely because it

MEASURING SOCIAL CHANGE 65 forecasts. Forecasts that fail are crucial steps in learning about theories and discovering where their weaknesses lie. But that means we must build models of social life so that it can be forecast. Both demography and economics have done so and learned much by failed forecasts; other be- havioral and social sciences might well take note. At this juncture of theory building on social change, social change itself is the way to theory building. Thus, theory testing and failed forecasts may be the best paths to scientific understanding of social change. Need for a National Statistical System Although one can easily demonstrate that our national indicators of eco- nomic and social change are more highly developed in some areas than was true on publication of Recent Social Trends, in many areas there has been little improvement. For example, we still have few national indicators of legal change, and we rely almost exclusively on ad hoc surveys to monitor changes in values and value practices such as religious belief and obser- vance. This discontinuity and variability in indicator development, collec- tion, and reporting amounts to a failure to develop the national statistical system Ogburn envisioned. Some of this is due to benign neglect by the social sciences since World War II of macro social change in advanced societies. One of the requirements for developing and testing theories of social change is a set of concepts and their indicators measured over time, within the domain of a national statistical system. I note three salient conclusions about requirements for a national statistical system: First, we require research devoted to building explanatory models of social change in order to structure a national statistical system that can usefully measure and monitor this change. Second, because of the limits of present models of social change and underinvestment in their development and testing, we generally lack data on potential explanatory variables for He trends that are monitored and 2iWorld War II appears to have been a historical dividing point in the study of social change. In the postwar period, the fashion in studying social change shifted to the "third world," the "developing nations," and "economic and cultural development." Modeling efforts shifted to how one might simulate the growth of economies and, increasingly for the noneconomic sciences, to the effects of rapid social change on traditional cultures. Although this latter interest fell within the domain of Ogburn's lag formulation, its shortcomings (see Smelser, in this volume) failed to generate interest in revising the model.

66 ALBERT J. RElSS, JR. measured; the major exception occurs for selected models of the economy.22 Generally, when explanatory variables are collected, their analysis is ad- ministratively segregated from analysis of the trends as such, substantially reducing the ability to test theoretical models. There are substantial problems in linking and analyzing extant information for like units (Tuma & HaDnan, 1979a, 1979b) and myriad problems stemming from lack of standardization. As a result, census variables such as age or years of school completed have to become surrogates for almost every conceivable explanatory concept. Third, we lack an adequate system of indicators of science and tech- nology. The annual report of the National Science Board, Science Indicators (1983), has almost no major indicators to monitor substantive changes in science and technology, much less a set of explanatory variables related to such changes. Most glaring is the absence of indicators for the behavioral and social sciences and technologies based on them. The failure to collect information on the content of science and technology, especially on inven- tions, has two important consequences. One is that we do not measure changes in the rate of behavioral and social science inventions and tech- nology and their contribution to contemporary society. The other, and more important, is that we are unable to measure relative contributions to social change, especially the contributions of behavioral and social science and technology compared to that of physical science and technology. A SUMMING UP The decades since the publication of Recent Social Trends have been a period largely of benign neglect by the behavioral and social sciences in modeling and measuring social change, economics being the major excep- tion. This neglect may owe in part to the reticence of theorists, save for economists, to address matters of social change. But scientific knowledge shapes and is shaped by such change; it becomes practically meaningful in the context of what kind of change is, or seems, possible; and it is tested against these consequences. It may be well to remember that Ogburn, as Duncan (1964:vii) calls to our attention, "saw science primarily as an accumulation of knowledge, but an accumulation whose structure is subject to continual change as new relationships among its parts are perceived or as discoveries shed new light on supposed relationships." Perhaps the period of benign neglect is drawing to a close, in which 22 In acknowledging this, attention is also drawn to the dissatisfaction with the precision of measures of variables estimated in structural equation models of the economy or of leading in- dicators (see Klein, in this volume).

MEASURING SOCIAL CHANGE 67 event it is essential to attend to the kinds of problems touched upon in this essay. We must understand better the role of behavioral and social science knowledge and inventions in social change. We must examine the effects of social change on theory, concepts, and measures, including their capacity to record and render social change intelligible. Time has favored Ogburn's conviction that statistical intelligence systems have a critical role to play in the processes of science and in society as a whole. * * * I wish to thank Otis Dudley Duncan and Barbara Laslett for their helpful substantive comments. REFERENCES Adams, R., Smelser, N., and Treiman, D., eds. 1982 Behavioral and Social Science Research: A National Resource. Washington, D.C.. National Academy Press. American Statistical Association 1983 AMSTAT News, Number 100. Bancroft, G. 1979 Some problems of concept and measurement. In National Commission on Employment and Unemployment Statistics, Counting the Labor Force: Readings in Labor Force Statistics, Appendix Vol. IlI. Washington, D.C.: U.S. Government Printing Office. Biderman, A. D., and Reiss, A. J., Jr. 1967 On exploring the "dark figure" of crime. The Annals 374:1-15. Binet, A., and Simon, T. 1905 Methodes nouvelles pour le diagnostic du niveau intellectual des anormaux. L'Anne~e Psychologique X1:191-244. Blumstein, A. 1983 Prisons: population, capacity, and alternatives. Pp. 229-250 in Wilson, J. Q., ea., Crime and Public Policy. San Francisco: ICS Press. Blumstein, A., Cohen, J., and Miller, H. D. 1980 Demographical disaggregated projections of prison populations. Journal of Criminal Justice 8:1-26. Boulding, K. E. 1978 Ecodynamics: A New Theory of Societal Evolution. Beverly Hills, Calif.: Sage. Burt, C. 1915 General and specific factors underlying the primary emotions. Report of the British Association for the Advancement of Science 85:694-696. 1939 The factorial analysis of emotional traits. Character and Personality 7:238-254, 285 299. Campbell, D. T., and Fiske, D. W. 1959 Convergent and discriminant validation by the multitrait-multimethod matrix. Psy- chological Bulletin 56:81-105. Cattell, J. McK. 1890 Mental tests and measurements. Mind XV:373-380. Cook, E. 1914 The Life of Florence Nightingale . London.

68 ALBERT J. REISS, JR. Duncan, O. D. 1964 Introduction. William F. Ogburn On Culture and Social Change: Selected Papers. Chicago: University of Chicago Press. 1984 Notes on Social Measurement: Historical and Critical. New York: Russell Sage Foun- dation-Basic Books. Eisenstadt, S. N. 1963 The Political Systems of Empires: The Rise and Fall of Historical Bureaucratic So- cieties. Glencoe: The Free Press. El-Korazty, M. N., Imrey, P. B., Koch, G. and Wells, H. B. 1977 Estimating the total number of events with data from multiple-record systems: a review of methodological strategies. International Statistical Review 45:129-157. Ericksen, E. P., and Kadane, J. B. 1983 Estimating the Population in a Census Year: 1980 and Beyond. Carnegie-Mellon University Technical Report #260, Pittsburgh, Pennsylvania. Fienberg, S. E. 1972 The multiple recapture census for closed populations and incomplete 2k contingency tables. Biometrika 59:591-603. Funkhouser, H. G. 1937 Graphical representation of statistical data. Osiris 3:319-344. Gallon, F. 1931 1869 Hereditary Genius: An Inquiry into Its Laws and Consequences. New York: Horizon Press (1952). Goodman, L. A. 1970 The multivariate analysis of qualitative data: interactions among multiple classification. Journal of the American Statistical Association 65:226-256. Greenwood, P. 1982 Selective Incapacitation. Santa Monica, Calif.: The Rand Corporation. Gullman, L. 1950 The problem of attitude and opinion measurement; Basis for scalogram analysis; and The principle components of scale analysis. Pp. 46-90, 271-272, and 312-361 in Stouffer, S. A., et al., eds., Measurement and Prediction, Vol. 4 of Studies in Social Psychology in World War II. Princeton, N.J.: Princeton University Press. Holzinger, K. J. 1930 Statistical Resume of the Spearrnan Two-Factor Theory. Chicago: University of Chi cago Press. On factor theory. Conference on Individual Differences in Special and General Abil ities. Washington, D.C.: National Research Council. 1941 Factor Analysis. Chicago: University of Chicago Press. Hotelling, H. 1933 Analysis of a complex of statistical variables into principle components. Journal of Educational Psychology 24:417-441, 498-520. Jaffe, A. J., and Stewart, C. D. 1951 Manpower Resources and Utilization: Principles of Working Force Analysis. New York: John Wiley. Kaplan, A. 1964 The Conduct of Inquiry. San Francisco, Calif.: Chandler.  Kelley, T. L. 1928 Crossroads in the Mind of Man: A Study of Differential Mental Abilities. Stanford, Calif.: Stanford University Press.

MEASURING SOCIAL CHANGE 69 1935 Essential Traits of Mental Life. Harvard Studies in Education 26. Cambridge, Mass.: Harvard University Press. Kish, L. 1965 Sampling organizations and groups of unequal sizes. American Sociological Review 30:564-572. Klineberg, O. 1933 Mental tests. Encyclopaedia of the Social Sciences X:323-329. New York: Macmillan. Lazarsfeld, P. F. 1950 The logical and mathematical foundation of latent structure analysis; The interpretation and computation of some latent structures. Pp. 362-472 in Stouffer, S. A., et al., eds., Measurement and Prediction, Vol. 4 of The American Soldier: Studies in Social Psychology in World War II. Princeton, N.J.: Princeton University Press. 1954 A conceptual approach to latent structure analysis. Pp. 349-387 in Lazarsfeld, P. F., ea., Mathematical Thinking in the Social Sciences. Glencoe, Ill.: The Free Press. 1967 The Uses of Sociology. New York: Basic Books. Lazarsfeld, P. F., and Barton, A. H. 1951 Qualitative measurement in the social sciences: classification, topologies, and indices. In Lerner, D., and Laswell, H., eds., The Policy Sciences. Stanford, Calif.: Stanford University Press. Levine, D. B., Hill, K., and Warren, R., eds. 1985 Long, S. B. 1980 Immigration Statistics, A Story of Neglect. Panel on Immigration Statistics, Committee on National Statistics, National Research Council. Washington, D.C.: National Acad- emy Press. The internal Revenue Service: Measuring Tax Offenses and Enforcement Response. Washington, D.C.: National Institute of Justice (September). Merton, R.K. 1936 The unanticipated consequences of purposive social action. American Sociological Review 1:894-904. National Commission on Employment and Unemployment Statistics 1979 Counting the Labor Force. Washington, D.C.: U.S. Government Printing Office. National Science Board 1983 Science Indicators 1982: An Analysis of the State of U.S. Science, Engineering, and Technology. Washington, D.C.: National Science Board and National Science Foun- dation. Nightingale, F. 1858 Notes of Matters Affecting the Health, Eff ciency, and Hospital Administration of the British Army. London: Harrison and Sons. Ogburn, W. F. 1917 1914 1915 1919 Progress and Uniformity in Child-Labor Legislation: A Study in Statistical Measure ment. Studies in History, Economics, and Public Law 48(2). New York: Columbia University Press. Methods of direct legislation in Oregon. Quarterly Publication of the American Sta tistical Association XIV. Initiative and referendum tested in hard times. Survey 23:693-694. Measurement of the cost of living and wages. Annals of the American Academy of Political and Social Science vm:23s-242. 1922a Bias, psychoanalysis, and the subjective in relation to the social sciences. Publications of the American Sociological Society XVII:62-74.

70 ALBERT J. RElSS, JR. 1922b Social Change: With Respect to Culture and Original Nature. New York: Viking Press. The great man versus social forces. Social Forces V:225-231. Social Changes in . . . (1928-1935). Annual series of the May issue of The Amer 1926 1928 1935 ican Journal of Sociology. Reprinted annually by University of Chicago. 1929 Social Changes in 1928. Chicago University of Chicago Press. 1933 Social changes in 1932: introduction. The American Journal of Sociology 38:823 824. 1934 Trends in Social Science. Science LXXIX:257-262. 1942 Man and his machines: teaching American youth how invention changes the modern world. Problems in Modern Life, no. 3. Washington, D.C.: National Council for Social Studies, National Association of Secondary School Principals, Departments of the National Education Association. 1950 Social Change. New York: Viking Press. 1957a Cultural lag as theory. Sociology and Social Research XLI:167-173. 1957b Social trends. Sociology and Social Research XLII:3-9. Ogburn, W. P., and Goltra, I. 1919 How women vote. Political Science Quarterly 34:175-183. Ogburn, W. F., and Nimkoff, M. F. 1940 Sociology. New York: Houghton Mifflin. President's Research Committee on Social Trends 1933 Recent Social Trends in the United States. New York: McGraw-Hill. Prosser, W. L. 1964 Handbook of the Law of Torts. 3rd ed. St. Paul: West. Rasch, G. 1968 An individualistic approach to item analysis. In Lazarsfeld, P. F., and Henry, N. W., eds., Readings in Mathematical Social Science. Cambridge, Mass.: MIT Press. 1980 Probabilistic Models for Some intelligence and Attainment Tests (with a new foreword and afterword by Benjamin D. Wright); expanded edition, University of Chicago Press. Reiss, A. J., Jr. 1968 Stuff and nonsense about social surveys and observation. Pp. 351-367 in Becker, H., Geer, B., Reisman, D., and Weiss, R., eds., Institutions and the Person. Chicago: Aldine. Systematic observation of natural social phenomena. Pp. 3-33 in Costner, H., ea., Sociological Methodology. Systematic observation surveys of natural social phenomena. Pp. 123-141 in Sinaiko, W., and Broedling, L. A., eds., Perspectives on Attitude Assessment: Surveys and Their Alternatives. Champaign, Ill.: Pendleton Publications. Exploring the central paradox of method in social science inquiry. Yale University (Copyright 1980), New Haven, Connecticut. 1982a How serious is serious cnme? Vanderbilt Law Review 35:541-585. 1982b Statistical measurement of social change. The 5-Year Outlook on Science and Tech nology 1981 2:649-667. Washington, D.C.: The National Science Foundation, Source Matenals. 1983 Cnme control and the quality of life. American Behavioral Scientist 27:43-58. Reiss, A. J., Jr., and Biderman, A. D. 1980 Data Sources on White-Collar Law-Breaking. Washington, D.C.: National Institute of Justice (September). Rosenbaum, James 1976 Making Inequality: The Hidden Curriculum of High School Tracking. New York: Wiley. 1971 1976 1980

MEASURING SOCIAL CHANGE 7 1984 Career Mobility in a Corporate Hierarchy. Orlando: Academic Press. Ryder, N. B. 1968 Cohort analysis. International Encyclopedia of the Social Sciences 2:546-550. New York: Macmillan & The Free Press. Shewhart, 1925 W. A. The application of statistics as an aid in maintaining quality of a manufactured product. Journal of the American Statistical Association XX (December), New series no. 152: 546-548. 1926a Correction of data for errors of measurement. The Bell System Technical Journal 5:11 - 26. 1926b Quality control charts. The Bell System Technical Journal 5:593-603. 1927 Quality control. The Bell System Technical Journal 6:722-735. 1930 Economic quality control of manufactured product. The Bell System Technical Journal 10:364-389. 1931 Economic Control of Quality of Manufactured Product. New York: D. Van Nostrand. 1939 Statistical Method from the Viewpoint of Quality Control (with the editorial assistance of W. Edwards Deming). Washington, D.C.: The Graduate School, U.S. Department of Agriculture. Shiskin, J. 1976 Employment and unemployment: the doughnut or the hole? Monthly Labor Review 99:3- 10. Sinaiko, A. W., and Broedling, L. A. 1976 Perspectives on Attitude Assessment: Surveys and Their Alternatives. Champaign, Ill.: Pendleton Publications. Sorokin, P. A. 1937-41 Social and Cultural Dynamics. 4 vole. New York: American Book Co. 1943 Sociocultural Causality, Space, Time. Durham: Duke University Press. 1947 Society, Culture and Personality: Their Structure and Dynamics. New York: Harper & Bros. Spearman, C. 1904 General intelligence, objectivity determined and measured. American Journal of Psy- chology XV:201-293. 1927 The Abilities of Man. New York: Macmillan. Spearman, C., and Holzinger, K. J. 1925 Note on the sampling error of tetrad differences. British Journal of Psychology XVI: 86-89. Stamp, J. 1937 The Science of Social Adjustment. London: The Macmillan Company. Stouffer, S. A., et al. 1950 The American Soldier, Studies in Social Psychology in World War II, Vol. I, Ad- justment During Army Life, and Vol. II, Combat and Its Aftermath. Princeton, N.J.: Princeton University Press. Taylor, D. G., Sheatsley, P. B., and Greeley, A. M. 1978 Attitudes toward racial integration. Scientific American 238:42-49. Thompson, J. D. 1967 Organizations in Action. New York: McGraw-Hill. Thurstone, L. L. 1931 Multiple-factor analysis. Psychological Review 38:406-417. 1935 The Vectors of Mind. Chicago: University of Chicago Press. Tuma, N. B., and Hannan, M. T. 1979a Approaches to the censoring problem in analysis of event histories. Pp. 209-240 in

72 ALBERT J. REISS, JR. Schuessler, K., ea., Sociological Methodology. San Francisco: Jossey-Bass. 1979b Dynamic analysis of event histories. American Journal of Sociology 84(January):820 854. Wald, A. 1945 Sequential Analysis of Statistical Data: Applications. Statistical Research Group, Co lumbia University. New York: Columbia University Press. Wallis, W. A., and Roberts, H. V. 1956 Statistics: A New Approach. New York: The Free Press. Webb, E. J., Campbell, D. T., Schwartz, R. D., and Sechrest, L. 1966 Unobtrusive Measures: Nonreactive Research in the Social Sciences. Chicago: Rand McNally. White, H. 1970 Chains of Opportunity. Cambridge, Mass.: Harvard University Press. Wigdor, A. K., and Garner, W. R., eds. 1982 Ability Testing: Uses, Consequences, and Controversies. 2 vole. Committee on Ability Testing, National Research Council. Washington, D.C.: National Academy Press.