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.
Part II Case Studies of amen Workers and In~rm~n bchno~gy
The Technological Transformation of White-ColIar Work: A Case Study of the Insurance Industry BARBARA BARAN Although white-colIar automation has received considerably less press than robots on the assembly-line, the introduction of computer-based technologies into the office has generated growing concern that industrialized working conditions and technological redundancy may be spreading to white-colIar settings. Because of the pervasiveness of occupational sex segregation within the office work force, it is also feared that women will bear the brunt of the restructuring process. Feldberg and Glenn (1983), for ex- ample, argue that whereas women's jobs are disproportionately disappearing and their opportunities for upward mobility declin- ing, men may actually benefit from the new technologies both because they will dominate the more highly skilled technical and professional jobs being created and because automation may cen- traTize control in the hands of (male) senior managers and systems analysts. The limited numbers of case studies that have been published on the impacts of office automation report conflicting findings. 25
26 THE INSURANCE INDUSTRY With regard to changes in the occupational structure, some re- searchers have found that job loss is concentrated among low- skilled clericals (Faunce et al., 1962; Roessner et al., 1985; Shep- herd, 1971), implying a general upgrading of labor. Other studies indicate, on the contrary, the elimination of skilled clerical ac- tivities, resulting in a polarization of the occupational structure (Feldberg and Glenn, 1977; Hoos, 1961; U.S. Bureau of Labor Statistics, 1965~. Similarly, whereas some analysts have reported less task fragmentation as the technology becomes more sophisti- cated (Shepherd, 1971; Matteis, 1979; Sirbu, 1982; Adler, 1983; Appelbaum, 1984), others suggest that job content is narrowed and worker autonomy reduced (Murphree, 1982; Greenbaum, 1979; Cummings, 1977; Feldberg and Glenn, 1983, 1977~. Finally, al- though in all cases women experienced the greatest job loss, in some reports it appeared that after automation women were rele- gated to lower-skilled activities (Feldberg and Glenn, 1977; Mur- phree, 1982), whereas in other accounts female clericals seemed to benefit from the new labor process (Matteis, 1979; Cummings, 1977~. The intent of my research on the insurance industry was to contribute to this nascent literature. The findings reported here are based on a two-year study of the impacts of automation on that industry. The first phase of this research was an in-depth case study of a major national property/casualty carrier, which included 26 interviews with employees in various parts of the company's operations (home office, branch office, data-processing center, commercial group, and personal lines centers) and dif- ferent levels of the occupational hierarchy, as well as analysis of extensive quantitative personnel data which the company made available. The second phase involved lengthy interviews with ex- ecutives, personnel managers, and systems analysts in 18 other companies, loosely stratified by size, product type, growth rate, distribution system, and so on; members of the industry's trade associations, agents' associations, and vendor companies were also interviewed. Third, a structures] telephone survey was conducted of 37 companies21 life firms and 16 property/casualty firms- again loosely stratified. All of the firms in both samples were among the top 100 companies in their industry segment. Together these 55 insurers account for approximately 55 percent of industry employment; they range from firms of over 50,000 employees to firms of less than 1,000. Finally, ~ have supplemented this field
BARBARA BARAN 27 work with secondary source material from government agencies, trade publications, and documents and survey data kindly pro- vided by the trade associations and consultants to the industry. Two kinds of conclusions emerged from this effort. The first, of course, are numbers of concrete observations which wiD be presented in summary form in this paper. In addition to these sectoraIly specific findings, however, analysis of the insurance in- dustry generated a set of more general hypotheses concerning the kinds of factors it is necessary to consider when attempting to as- sess the impacts of office automation technologies on a work force. Since the following discussion is not sufficiently comprehensive to cover all these issues, ~ want to discuss them briefly here before turning to the more detailed findings of the study. First, as the case of the insurance industry made clear, it is virtually impossible to separate technological innovation from other factors affecting the competitive dynamics of an industry. The competitive environment Is both a major determinant of the speed of diffusion of innovation and quite apart from technolog- ical changesignificantly affects the demand for labor. Second, and closely related, is that the impacts of the new technologies on the labor force are not limited to their effects on the organization of the work process. In insurance, changes in product offerings and in the structure of both the industry itself and the firms within it promise to be equally important influences on the kind and amount of labor employed. Third, it was clear from this study that "office automation" cannot be analyzed as a single phenomenon. Impacts vary on the basis of the specific kinds of technology being introduced (includ- ing, importantly, the generation of that technology) and the na- ture of the work process being automated (originating often in the unique characteristics of the product, market, and organizational structure of the industry). In terms of the first, ~ would underline the importance of periodizing the process of office automation; some of the disagreement in the literature can be attributed to did ferences in the generation of the technologies observed. In terms of the second, it ~ obvious that even in this one industry, the new labor processes will vary widely by product line. It is also important to point out that the insurance industry is very differ- ent from office-type settings in which word processing is the core application since systems development in the insurance industry
28 THE INSURANCE INDUSTRY has been driven by its data-processing needs. In fact, in some re- gards, more accurate parallels can be drawn between automation in the insurance industry and automation in manufacturing indus- tries. In both cases, production work is increasingly performed by machines while administrative support activity remains relatively labor-~ntensive (although, in insurance, services activities are also becoming extremely automated). Fourth, in assessing the impacts of automation on skill re- quirements and on the occupational structure of an industry, would stress the importance of analyzing the changes occurring in the entire labor process. In failing to do so, analysts often miss the forest for the trees. When the organization of work is being fundamentally restructured, it is not so useful to talk about how specific jobs are changing; instead, we need to begin to assess how particular job functions or activities are being reconstituted and recombined to produce new kinds of job categories. Although this observation may seem banal, most studies of the effects of office automation have focused almost exclusively on the clerical work force. As such, they miss one of the most important features of the current wave of applications, that is, the automation of professional functions and their transfer to less- skilled labor. As a result, these studies may be overly pessimistic both about the decline of clerical-type occupations and their likely skit levels; at the same time, they may be overly optimistic about the expansion of higher-level, challenging, well-paid work. Fifth, we have to be careful not to assume any necessary identity between skill levels and other job attributes, such as pay, satisfaction, or occupational mobility. Indeed, many of the new jobs emerging in the insurance industry may require- fairly high levels of skill and yet offer few rewards, material or otherwise. Finally, this last point is true in part because different cat- egories of workers will not only be differentially affected by the process of transformation, but the nature of the available labor force also shapes job design. In this case, the nature of the female labor market may be an import ant determinant of the emerging occupational structure. Some, although not all, of these themes will be explored in greater depth in the remainder of this paper. To give context to the discussion of the ways in which new technologies are having an impact on the labor force, the first section begins by describing the significant changes that have occurred in the last decade in terms
BARBARA BARAN 29 of the rate of diffusion and the kinds of computerized systems implemented in the insurance industry. THE PACE OF DIFFUSION AND THE NEW~IMPLEMENTATIONS Insurance companies, along with financial institutions and the government, were among the earliest users of electronic data- processing (EDP) computers. As early as 1959, the finance, in- surance, and real estate sector boasted the greatest number of computer installations per million employees (Phister, 19793; by 1970, the insurance industry employed a higher ratio of computer specialists than any except the high-technology manufacturing in- dustries (Bureau of Labor Statistics, 19Blb). Size of the organization was initially the primary factor influ- encing the implementation decision; companies attempted little, if any, formal cost-benefit justification. Applications were generally limited to structured accounting tasks, billing, and claims dis- bursements. With the exception of the move to direct billing- a transaction which had traditionally been handled at the agency level, the automation of these functions had little impact on the rest of the organization. For the first two decades, the pace of diffusion was extremely civilized and the industry proceeded much as it had for the last hundred years. The barriers to rapid adoption of more sophisti- cated office systems were both technological and organizational.t Beginning in the early 1970s, however, systems development took two unport ant new turns. First, the focus of automation shifted ~ The technological barriers included (a) equipment incompatibility, both among the various vendors and even within the product offerings of a partic- ular vendor, that prevented network extension and integrated system devel- opment; (b) substantial work station connection costs and communications costs; and (c) the difficulty of representing more complex white-collar activi- ties in computer algorithms. The attempt to impose a standardized logic on many procedures often generated too many exceptions to be cost-eEective and in some cases even diminished the efficiency of the organization. The organizational problems associated with the introduction of the new systems were perhaps even more substantial than these technical problems: these integrated systems required a fundamental reorganization of production, ser- vice, and distribution, since their effect was not limited to clerical labor in word-processing and data-processing departments; higher-level workers had their jobs redefined and in some cases even eliminated. Their resistance erected a powerful barrier to diffusion.
30 THE INSURANCE INDUSTRY from administrative functions such as accounting to automation of the production process itself, meaning primarily the underwriting, rating, and physical production of the insurance policy and the handling of claims disbursements. Second, in the place of single- task, batch-oriented machines, multitask, multimachine systems began to be introduced, often operating on-line. The process of systems integration was an evolutionary one. As the price/performance ratio of the hardware continued to im- prove and the software became more sophisticated, more and more functions in the operational areas were automated and systems became increasingly decentralized. Independent systems began to proliferate, often geared exclusively to a company's in-house data-/word-processing activities. Gradually, however, a counter- tendency developed as companies moved to link these systems together into what at first were fairly rudimentary networks. Ini- tially, in some of the local settings, applications were developed that went beyond the automation of a discrete task toward the performance of a wider set of connected operations. Automated linkages were then slowly constructed among systems, so that information could be electronically transferred and shared. On this basis, entry and processing functions were increas- ingly decentralized and distributed throughout the organization, but data bases were integrated and centralized. In the mid to late 1970s, what are often referred to as Voice automation" (OA) applications also began to be linked to these DP-based systems. A multitude of new office machines were introduced. Optical char- acter recognition devices were applied to premium billing and collection operations; computer output microfilm (COM) which permits the transfer of data directly from computer to microfiche- was used extensively by all insurers, and in conjunction with computer-assisted retrieval technology, increasingly replaced pa- per files; finally, more and more companies began to experiment with electronic mail, teleconferencing, electronic fund transfer, in- teractive data access via television, and other "office of the future" technologies. To an important extent, the introduction of these more so- phisticated systems in the 1970s was driven by a new competitive environment. Unprecedented levels of inflation and correspond- ingly high interest rates and important demographic and socio-
BARBARA BARAN 31 economic changes2 combined to shake up this once stodgy industry by increasing uncertainty in financial markets and, consequently, dramatically increasing competition in the financial services sec- tor. Deregulation fueled these competitive flames. Giants in the industry found themselves struggling for survival; and companies with a long, proud history of paternalism were suddenly forced to lay oh up to 1,500 employees virtually overnight. In many ways, In fact, the insurance industry was hit by a crisis not un- like the one faced by U.S. auto manufacturers; and in both cases, one important response on the part- of companies wan to turn to automationboth to increase the efficiency of their operations and to enhance their product offerings. Whereas the early applications were basically limited to the automation of discrete tasks (e.g., typing, calculating), the new implementations called for rationalization of an entire procedure (e.g., new business issuance, claims processing), and ultimately the restructuring and integration of all the procedures involved in a particular division, product line, or group of product lines. As such, in the categories of Bright's (1958) original analysis, there has been a dramatic leap in the span, level, and penetration of the automated systems and, therefore, in the impacts on the labor force. The changes that have occurred in the underwriting and claims support systems best illustrate this evolution, although will also touch briefly on three others: administrative/clerical, decision support, and agency systems. 2 On the demand side, inflationary pressures initiated the demand for alternatives to ordinary insurance that would provide adequate protection at lower cost or higher rates of return on savings. Demographic shifts fueled these trends. Dual income households and the more affluent dual career householdsgrew rapidly and consumer assets were increasingly concentrated here. The needs and tastes of this population differed significantly from the traditional insurance consumer with more emphasis on investment over mere savings, less concern with thrift and more with spending, and a demand for a fuller range of sophisticated financial services. On the supply side, competitors outside the insurance industry responded to the new market conditions by offering reasonable substitutes for many traditional insurance products and services. In addition, commercial banks began to lobby for the lifting of regulatory restrictions which bars them from the sale of insurance. Finally, insurance companies themselves moved to diversify their product offerings and successfully lobbied authorities to deregulate rates.
32 THE INSURANCE INDUSTRY UNDERWRITING AND CLAIMS S UPPORT SYSTEMS The underwriting function i' s a critical component of the in- surance policy production process; it entails an analysis of the risk involved and acceptance or rejection of that risk. As such, the heart of this operation has always been performed by profes- sional labor: insurance underwriters. However, today, in numbers of product lines, clerksaided by decision parameters embedded in computer software are responsible for the risks their compa- nies accept; and in the most standardized lines, computers are even performing this risk calculation task themselves. The un- derwriting systems now being used to produce property/casualty personal lines products (such as auto and homeowners' insurance) are a good example of these new applications. Traditionally, the production process for personal lines prod- ucts involved roughly the following: an agent gathered the client's policy information and forwarded this to an underwriting depart- ment, located either in the home or branch office; the underwriters established a file on the client, evaluated the risk, and determined the risk parameters, then sent the policy to the rating section. A rater (a skilled clerical employee) calculated the premium charges based on guidelines contained in numerous manuals; sometimes that information had to be communicated back to the agent so that the client could make a decision whether or not to use this particular company. In most cases, the policy went back to the underwriter who reviewed it and then sent it to a typing pool where policy typists prepared the various forms and documents. Finally, the policy was mailed back to the agent who forwarded it to the customer. In most large property/casualty carriers the first applications of computerized equipment to the underwriting process were im- plemented in the early 1970s and were aimed at speeding the underwriter's access to client files, shortening the time involved in rate malting, and of course Reproving the efficiency of policy production. A typical system of this vintage might simply have involved a stand-alone computerized rating system with most of the rating guidelines from the manuals built into the machine- and an automated policy issuance system, which performed the typing and assembly function. In many personal lines departments today, this kind of config-
BARBARA BARAN 33 Oration remains the current state of the art. Beginning in the late 1970s, however, some of the largest carriers moved to institute what they variously call "underwriting by exception," "pigeon- hole underwriting," or ~computer-assisted underwriting." The aim is very simply to have the computer itself perform the risk assessment function on as large a fraction of policies as possible, on the basis of underwriting decision rules which are built into the machine. In general, the production process associated with these kinds of systems looks something like the following: the agent sends the policy information to a personal lines department where a cleri- cal worker screens it and enters all routine risks directly into the machines; these policies are then often relayed in batch form to the carrier's national (or regional) computer center; the computer evaluates the risk, rates the policy, and produces it. Policies which fail to fit within the pre-established guidelines are kicked out by the machine and returned to an underwriter, who often now calls up the information on a terminal and works with it on the screen. Although these systems are extremely new, a few companies al- ready report that 50 percent to 90 percent of their personal lines policies are completely underwritten by the computer. Most personal lines systems fall between the two just described in their level of sophistication, but, in almost all cases, clerks are assuming greater, and sometimes almost exclusive, responsibility for underwriting. For example, in one major property/casualty carrier, the introduction of computer-assisted underwriting two years ago shifted the bulk of the underwriting function from the underwriting department to the operations department, a clerical operation which had formerly been confined to assembling and producing the physical policies. Although auto policies still go to the underwriters for an initial screening process, most property policies are sent directly to skilled clerks in the operations depart- ment, who themselves order all necessary inspection reports and issue the policies. In some rare instances, the underwriting function has even been integrated into the activity of a sales worker, entirely elim- inating it as a separate department. In these cases, a highly skilled clerical positioncustomer representative has been de- signed; this job includes answering inquiries from potential cus- tomers over the phone; accessing a terminal to produce on-the-spot
34 THE INSURANCE INDUSTRY rate quotes; negotiating the rates; and then if a sale is made, cap- turing additional information on-line. The computer underwrites and produces the policy that same night. Similar kinds of work process configurations have emerged in the claims departments of large insurance companies. In the life insurance division of the industry, for example, group health claims systems which were implemented early in the 1970s and are now probably the most sophisticated in the industry- have allowed skilled clerks to take over 75 percent or more of the claims function. ADMINISTRATIVE/CLERICAL, DEcIsIoN SUPPORT, AND AGENCY SYSTEMS In contrast to the underwriting and claims support systems, neither office automation applications nor decision support sys- tems have, as yet, made a significant change in the organization of work, although in the case of the former, companies have reported fourfold productivity increases when word-processing equipment is substituted for traditional typewriters and 70 to 85 percent rises in output per operator when CRTs replace MagIT equipment (Life Office Management Association, 1979~. By the late 1970s, word processors were widely diffused throughout the industry. Compa- nies concentrated on installing the text-processing systems first, since that is where large and immediate savings could be realized, and then gradually added more sophisticated technologies and applications electronic mail, voice mail, image processing, elec- tronic office support, and decision support. In most cases, word processing and data processing have been both organizationally and technically independent; companies have moved extremely slowly to merge the two through electronic communications capa- bilities. Management information or decision support systems are thus truly in their infancy. Although the integrated systems described earlier have made a wealth of information available on a timely- often dailybasis, few insurance managers have on-line access to these data bases. In the last few years, however, user acceptance has improved significantly due largely to the proliferation of per- sonal computers. One of the promises of these systems is that they can begin to bypass middle management and provide top
BARBARA BARAN 35 executives with ongoing access to the performance of their various departments. The story of the agency support systems is somewhat more complex. As in the case of the carriers, the greatest leaps in productivity have been attained in the production and servicing functions, but this is largely the result of extensive networking with the carriers and has therefore been highly dependent on the nature of the distribution system.3 Over the last 5 to 6 years, insurers with captive or salaried agency forces have moved rapidly to integrate sales electronically into the company's overall operations, allowing them to substantially rationalize the flow of work. For example, the keying function caI1 be moved to the field office, closer to the point of data generation; much of the servicing of the policies can be computerized and moved from the field to the home office; sales proposals and policies can often be printed on-site in the field offices, eliminating bulk ordering. The benefits of these integrated systems include improved speed and quality of quotations, policy production, billing and customer service; greater sophistication in marketing techniques; error reductions; and less duplication of clerical effort. EFFECTS OF AUTOMATION ON THE SIZE AND COMPOSITION OF THE WORK FORCE Taken together, the effects of these new systems on the insur- ance work force has already been dramatic and because integrated implementations did not begin in earnest until fairly recently, we can expect even more significant changes in the future. Work is being both eliminated and redesigned. According to a recent study funded by the U.S. Department of Labor (Drennan, 1983), productivity in the insurance industry as a whole grew rapidly between 1969 and 1979, at an average annual rate of 2.7 percent (in contract to a 1.1 average annual growth rate for all private industries); and these productivity gains accelerated 3 The four major marketing systems in the insurance industry are indepen- dent agents and brokers, exclusive or acaptiven agents, salaried employees (the Direct writers"), and direct mail. Both independent and exclusive agents are independent contractors, but the latter represent only one insurer. The great advantage of the direct writers and exclusive agency systems is the organizational integrity which has allowed them to make full use of integrated, computer-based systems.
36 THE INSURANCE INDUSTRY so that by 1975-1979 the average annual percentage increase in productivity was 6.7 percent using a GNP deflator and 4.2 percent using an industry deflator.4 Also significant, registered productivity improvements varied within the industry. In the life insurance segment, where au- tomation of policy production proceeded most rapidly, purchases of policies increased by 49.2 percent (between 1970 and 1980) while the labor force expanded by only 9.8 percent. In the prop- erty/casualty segment, on the other hand where fewer lines of business were easily automated and, more important, insurers were expanding capacity to reap the benefits of high interest rates- employment kept pace with gains in output. From 1970 to 1978, average annual employment gains in property/casualty carriers were 3.2 percent, approximately one-fourth again higher than the industry-wide average and just slightly less than four times higher than the average gains of 0.85 percent annually in life carriers. By 1980, however, reckless expansion caught up with these companies. Between 1980 and 1982, sales plunged by 6.5 percent and employ- ment growth slowed considerably. Since then there may even have been some labor-shedding. Every property/casualty company we visited, with one exception, had experienced serious layoffs in the last four years, ranging between 5 percent and 15 percent of their work force. In the coming years, labor savings are expected to accelerate In all segments of the industry. With respect to future job Toss, three fairly comprehensive employment forecasts for the insurance industry have been re- cently published (Drennan, 1983; Leontief and Duchin, 1986; Roessner et al., 1985~; all three predict dramatic declines in in- dustry employment particularly clerical employmentbetween 4 Relatively crude measures of real productivity figures are presented here, because of the difficulty in calculating inputs from real estate, stock brokerage houses, and so on (estimated roughly at 6 percent of total industry employment). Unpublished productivity figures for the insurance industry calculated by the U.S. Department of Commerce show the dramatic improve- ment over the previous decade but suggest somewhat slower average annual gains during this period (2.0 percent). However, the Department of Labor figures were based solely on insurance carriers, whereas the Department of Commerce estimates include agents and brokers. The average annual em- ployment gain for agencies and brokers during these years was 4.5 percent, or almost double the industry-wide figure. This then explains the discrepancy and strengthens the hypothesis that the new technologies were indeed paying off in terms of labor savings at the insurance carriers.
BARBARA BARAN 37 1980 and the turn of the century. If, relative to output, requisite clerical labor input in the industry declines by even 10 percent between 1981 and 1991, approximately 100,000 fewer clerical jobs would be created. If, by the end of the century, the clerical la- bor coefficient were to fall to 70 percent of its present level the smallest decline predicted by these forecastersover 400,000 fewer clerical jobs would be created; and between 1990 and 2000, ap- proximately 100,000 clerks would actually lose their jobs through layoffs or attrition (Baran, 1985:225-234~. Clerical job loss may wed be even more serious. One of the studies is considerably more pessimistic; in addition, none of these studies take into account the effects of structural shifts likely to occur in the industry. For example, because agencies were responsible for 61 percent of all new clerical jobs in the last decade, a reorganization and consolida- tion of the distribution functio~which most industry observers expectshould have a major effect on clerical employment. Although layoffs in response to the overcapacity common in the 1980s (and the anticipated job loss) could affect the entire occupational structure, the impacts of technological change can be more discriminate. Over the last two decades, there have been visible shifts in the composition of employment and effects have varied with waves of implementation. The first impacts of computerization were, as we have seen, on the accounting functions of carriers; in addition, billing shifted from agencies and sales offices to the insurers' data-processing departments, and probably coincidentally, reporting requirements to state regulatory agencies increased (Appelbaum, 1984~. As a result of these various forces, the most significant change in the occupational structure of the industry during the 1960s was an extraordinary influx of technical personnel (see Table t). Whereas in 1960 professional and technical workers had constituted 3.2 percent of ad employment, by 1970 this share had grown to 5.8 percent (U.S. Bureau of I,abor Statistics, 1969, 19Blb).5 5 There are substantial problems involved in developing a longitudinal data series on employment in the insurance industry. There are three principal data bases: the first, the U.S. Census of Population, remained consistent between 1960 and 1970 and permits useful comparisons of oc- cupational change over that decade, although even for this time period the extent of disaggregation by occupation particularly insurance-specific occupations is woefully inadequate. By 1980, census data were truncated and occupational categories were significantly reorganized and redefined; this
38 THE INSURANCE INDUSTRY Over the next decade, data~processing technology continued to diffuse and text-processing equipment was introduced; competi- tion had heated up markedly and firms were rushing to reduce unit labor costs. By the end of this period, the impacts of technological change on clerical labor are strikingly clear (see Table 1~: clerical employment as a percentage of total employment in the industry fell from 50 percent to 45 percent; 73,000 clerical jobs were elim- inated. This dramatic drop in employment share contrasts with the experience of clerical labor in the economy as a whole and in other similar industries such as banking, securities, business services, and credit agencies. It occurred because the introduction and subsequent diffusion of new technologies occurred earlier for the insurance industry and was therefore more widespread by the late 1970s (Drennan, 1983~. Virtually all clerical categories experienced relative decline, but routine clerical occupations- such as keyboarding, filing, tra- ditional office machine operators, and so on were particularly makes comparison with earlier years difficult and, in some cases, useless. The second important source, the Bureau of Labor Statistics' Occupa- tional Employment Matrix, provides data reported for 1970, 1978, and 1980 which are estimates based on other surveys. In 1980, the Matrix shifted from basing itself primarily on census data to relying principally on data from the BLS Occupational Employment Survey (OES). However, the OES data fail to correspond even generally to the census data. To illustrate, the t\able below compares the occupational structure of the insurance industry as reported by the 1978 Matrix based upon the census and the 1980 Matrix based upon the OES: Census (1978) OES (1980) Professional/technical 6.79% 15.4% Managers/oflicials 12.08 13.8 Sales workers 34.54 14.6 Clerical workers 45.42 54.4 The third source of occupational data, the employer-based OES series, which appears to be the most reliable, has limited usefulness since it is available for the insurance industry only for the years 1978 and 1981. Despite this limitation, this report has relied on it and the OES-based Matrix and has used the census-based Matrix only when necessary for the earlier years. Although wide discrepancies in the data exist, at least the direction of charge is the same in most cases. Finally, we have used the BLS Employment arid Earnings series in comparing aggregate employment changes, since these data were the only longitudinal ones available disaggregated by major industry division. See Hunt and Hunt (1985) for a fuller discussion of methodological problems.
BARBARA BARAN TABLE 1 Percentage Distribution of Insurance Employment by Occupation, 1960,1970, 1978 Occupation Percent of Total Employment Percent Change from 1960 1970 1978 1970 to 1978 Managers/officers 13.3 11.8 12.1 21.4 Professional/ technical workers 3.2 5.8 6.0 24.0 Clerical workers 47.4 50.0 45.4 8.0 Computer operators 0.7 1.S 119.0 Keypunch operators .1.8 1.2 -21.9 Statistical clerks 2.8 2.5 6.4 Bookkeepers 3.6 2.8 -7.4 Adjustera/examiners 7.3 9.8 58.8 File clerks 2.5 1.7 -19.7 Mail handlers 0.7 0.5 -10.7 Secretaries 13.2 12.0 8.2 Typists 6.9 5.1 -11.7 Sales SS.6 30.0 S`.5 37.2 NOTES: Percents for aubeategories under the category "Clerical workers" do not total the percents for the category as a whole due to other clerical occupations not listed. Columns do not add to 100.0 due to other occupations not listed. SOURCE: 1960 data, U.S. Bureau of Labor Statistics (1969); for all other years, U.S. Bureau of Labor Statistics (1981b:Vol. I). 39 seriously affected. In absolute terms, the number of bookkeeping machine operators fell by 60 percent, calculating machine opera- tors and duplicating machine operators by 41 percent, tabulating machine operators by 76 percent, keypunch operators by 22 per- cent, file clerks by 20 percent, and typists by 12 percent (Bureau of Labor Statistics, 1981b). Secretarial employment in the insurance industry continued to grow at about ~ percent a year reflecting the relatively small impact of automation on traditional administrative support tasks, yet this growth rate was less than one-third that of secretarial labor in the economy as a whole and less than half the rate of employment growth in the insurance industry. Even in this occu- pational area, then, the introduction of word-processing equipment and rationalizing moves such as the pooling of secretaries took a toll on employment opportunities. Professional, managerial, and especially sales jobs, on the other hand, grew at a fairly rapid clip. Although there are no current data available to reflect the re- cent implementations of integrated on-line systems, it is clear that, especially in the carriers, clerical jobs are continuing to disappear (see Table 2~. Between 1978 and 1981, of the clerical occupational titles reported in the BI`S Occupational Employment Survey of In- surance, approximately two-thirds grew more slowly than overall employment and close to one-half experienced absolute job loss.
40 THE INSURANCE INDUSTRY TABLE 2 Percentage Change in Employment in Insurance Carriers by Selected Occupation, from 1978 to 1981 Occupation Percent Change in Employment Total employment Managers and officers Professional and technical Actuary Systems analyst Account ant / auditor Claims examiner, property/casualty Underwriter Computer programmer Clerical Computer operator Bookkeeper, hand Claims adjuster Correspondence clerk File clerk General clerk Rater Secretary Stenographer Typist Sales S.6 10.9 9.5 21.7 42.5 -51.1 10.8 3.0 36.2 3.5 0.4 -15.2 -7.7 -1.5 -15.9 -5.9 -1.6 6.6 -31.4 -2.9 6.6 SOURCE: Unpublished data from the U.S. Bureau of Labor Statistics (1978, 1981a). However, clerical categories such as "claims examiner, life, acci- dent and health," "insurance clerk," and epoxy change clerk" continued to experience steady growth, probably reflecting the growing responsibility of clerks for the claims and underwriting processes. Significantly, from 1978 onward the impacts of automation on professional work have become evident in aggregate data. Between 1978 and 1981, underwriters lost employment share and the num- ber of accountants and auditors plunged dramatically. Although the professional and technical category still grew faster than total employment, most of this expansion was directly attributable to the continued addition of computer professionals (systems analysts and programmers); excluding these two categories, professional oc- cupations grew at about half the rate of total employment. Growth in the number of sales personnel also began to slow down (to about
BARBARA BARAN 41 half the rate of the previous period)probably reflecting declining profitability and the attempts on the part of insurers to improve the efficiency of their distribution systems; but managerial ranks continued to swell. Three conclusions can be drawn from these data: First, at least in relative terms, insurance is employing fewer people; and in increasing numbers of occupational categories, the relative decline in employment share has escalated into absolute job loss. Sec- ond, the weight of the occupational structure has shifted upward, tilting toward the higher-skilled categories of labormanagerial, professional, and sales. Third, however, even within the clerical category, routine occupations are being eluninated most rapidly. In contradiction, then, to observers who feared that widespread deskilling would accompany automation of this industry, the new technologies seem to be raising aggregate skill levels both across occupations and within occupations. When the declining and growing clerical occupations in the life insurance segment of the industry are ranked on the basis of current wage levels (as a proxy for complexity of the job), this upward shift in clerical skills is confirmed. Almost without exception, all occupations at the bot- tom of the pay scale are shrinking more rapidly than those at the top. Similarly, when the average skill level of declining and growing occupations in both the 197~1978 and 1978-1981 peri- ods are computed on the basis of the U.S. Department of Labor's (1977) Dict;io nary of Occupational Titles job evaluation scheme, growing occupations in both time periods ranked higher in the areas of mathematical skills, language requirements, and specific vocational training (Baran, 1985:115-118~. Although these aggregate data are suggestive, they fail to capture the logic of the new implementations and therefore are not very useful either in predicting the future or in describing the kinds of new jobs actually emerging. For that kind of detail, it Is necessary to go to the level of the shop floor. CHANGES IN THE NATURE OF WORK: THE VIEW FROM THE SHOP FLOOR Conceptually, what is perhaps most notable about the emerg- ing organization of work in the insurance industry is that it rep- resents an electronic transcendence of the assembly-line, or more
42 THE INSURANCE INDUSTRY precisely of the detailed division of labor which has come to domi- nate ad modern industry, both blue and white collar. Despite the fact that the office automation literature, particularly the critical literature (see, for example, Braverman, 1974; Cummings, 1977; DeKadt, 1979; Driscoll, 1979, 1980; Feldberg and Glenn, 1977, 1983; Greenbaum, 1979; Hoos, 1961; Nussbaum and Gregory, 1980), has been dominated by the fear that industrialized labor processes would accompany the rising capital intensity of white- colIar production, ironically it was the pre-automated rather than the newer highly automated labor processes in the insurance in- dustry that most closely resembled an industrial assembly-line. Traditionally, in huge open offices, white-colIar insurance workers sat grouped by function underwriters, raters, typists, file clerks, and so on; the paper flowed manually from one station to another as each worker completed his or her portion of the production task. In line with established principles of industrial engineering and scientific management over the last two decades, this work was increasingly fragmented into its component parts and simpler functions were turned over to less-skilled labor. Just the check issuance procedure in a preautomated chains office illustrates the extraordinary degree to which work was divided: If a check had to be issued, first it was typed by a typist; then another clerical verified the amount; a third person audited the claim to insure that the doctor charged appropriately for the service provided; a fourth person actually ~burst" the check (took apart the carbons); and a fifth then put the check through the signing machine. Including supervisorial oversight, between six and seven people were involved in this one procedure alone (personal interview). Task fragmentation was not limited to processing functions. In many property/casualty companies, for example, underwriting wan also rationalized; at the bottom end, routine underwriting tasks were turned over to a newly created clerical position, "un- derwriting technical assistant," and, at the top end, the number of specialty underwriters increased. The early applications of data-processing and text-processing equipment tended to follow and intensify the Taylorist logic un- derlying this form of work organization, producing ever more highly ~industrial~zed" work settings. In the case of data pro- cessing, the early mainframes were large, noisy, and often required special rooms. Because of these technical requirements and the
BARBARA BARAN 43 tendency to automate in conformity with the rationalized bureau- cratic structure, separate data-processing departments were estab- lished in all insurance companies. These eventually generated their own hierarchy of systems analysts, programmers, computer oper- ators, data-entry operators, and so on. With the improvements in telecommunications in the last decade, the entry function was often separated sharply from the rest of the operation and fre- quently spatially isolated in suburbs or small towns where land and labor were cheaper. These were the "electronic sweatshops." Work within the centers was machine-linked, machine-paced, and often machine-monitored as well; correspondingly, turnover was extremely high. Many of the early applications of word process- ing had a similar character in that key entry was often separated from other clerical activities and centralized in a word-processing center. Beginning in the late 1970s, however, as companies began to introduce integrated systems, they also embarked on a serious re- design of the labor process, which promises to reverse many of these earlier trends. Whereas the organizing logic of the last sev- eral decades produced an amazing proliferation of "detail" work- ers, today the white-colIar assembly-lines are being internalized within the machines themselves; an electronic reintegration of the work process is manifesting itself in at least three important new trends. First, multifunction jobs are replacing the extremely narrowed occupational categories of the past. On the basis of the new inte- grated systems, insurers have been able to consolidate all policy holder data into central master records stored in the company's main computer installation. In the past, these records were du- plicated in up to a dozen functional units; now users in remote sites can access and alter relevant policies and the changes can be integrated into the master record. One individual can therefore handle multiple service transactions and functional units can be consolidated since the individual master record for each policy is a complete data base. Second, "mentally and "manuals labor are (slowly) being rein- tegrated as entry functions are slipped unobtrusively into the ac- tivity of higher-level workers, such as underwriters and claims adjustors. The apocryphal example is the insurance salesman, equipped with a portable terminal, who enters all necessary pol- icy data directly from the customer's home; through a regular
44 THE INSURANCE INDUSTRY telephone link, that information is received by the company's mainframe which then underwrites, rates, and produces the pol- icy. Although the current state of the art is perhaps a decade away from this futurist scenario (and many procedures simply do not lend themselves to such extensive automation), companies are beginning to install systems that require professionals to perform some, if not all, of the requisite data entry. Finally, apart from this kind of electronic integration of tasks, in several procedures and product lines, single-activity units are being eliminated in favor of multiactivity teams. Whereas formerly typists, raters, and underwriters were divided into separate units, each with its own supervisor, now a small team consisting of one or two of each kind of worker will service some subset (often geographical) of the company's customers. Although these overall trends are visible throughout the en- tire industry, a variety of actual job configurations are emerging both within companies and across companies largely reflecting differences among product lines. In the case of high-volume, stan- dardized lines, two principal types of work organization seem to be developing. The first of these is found in purest form in the highly automated personal-lines underwriting departments of many large property/casualty carriers. Here both rating and risk assessment have been largel~r assumed by the computer; as a result, func- tions formerly divided among entry clerks, raters, underwriting assistantsas well as additional low-level underwriting tasks have been consolidated and turned over to a highly skilled clerical worker. Although closely circumscribed in their decision making, these clerks are a long way from the typing pool. Judgment is required, and often they are required to interact directly with the agents, a level of responsibility formerly reserved for professional under- writers. Most important, these workers (limited by the decision parameters built into the machines) are ahnost solely responsible for the soundness and accuracy of the millions of routine risks their companies write. Underwriters have also had their jobs redefined as clerks have taken over their lower-level functions. First, they have become "exceptions handlers, responsible only for the policies that fad! to fit into the "pigeon-holes~ of the computerized system; thus, their work has become more complex. Second, there has been a
BARBARA BARAN 45 reorientation of the job function away from churning out policies and toward planning and marketing. A similar job configuration has emerged in the claims oper- ation of many group life and health insurers. On the basis of highly integrated computer systems, clerks have taken over most of the work of professional claims examiners; the small number of remaining examiners handle only the problem cases. The emergence of this kind of work organization, where rou- tine clerical work is eliminated, professional jobs are reduced in number while being enlarged in depth and scope of responsibility, and skilled clerks become the bulk of the work force, seems to de- pend on several factors. First, the operation is high volume, mak- ing automation cost-effective; second, the product is standardized, making it relatively easy to translate the decision parameters into algorithmic form; third, these operations still involve semiskilled functions that cannot be automated, often because interpersonal interaction is required. Alternatively, where products are standardized and volume is high but semiskilled functions can be assumed almost entirely by the computer system, the work force is highly bifurcated between a large number of routine data-entry clerks and a very small number of skilled professionals. At the clerical level, there is little real decision making and no interaction with agents or clients. Finally, in product lines that are too low in volume, unique, or complex to lend themselves to this level of standardization, automation is less extensive and more confined to discrete tasks. Clerical work ~ routinized and narrow; professional work is com- puter-assisted to the extent technically feasible, but profession- als are not being replaced either by clerks or by machines. In these situations, there are serious constraints on the automation of professional functions, either because low volume makes such sophisticated systems unjustifiably costly or because the work it- self is too complex or unstructured. In the longer run, it is possible that some professional functions will be turned over to less-skilled labor as the software improves; but it is also likely that clerical labor will be bypassed entirely as data entry is increasingly folded into the activity of higher-level originators. The nature of the product may be the primary determinant of which of these job configurations emerges. The second seems to be considerably more prevalent in life insurance companies (about
~6 THE INSURANCE INDUSTRY 75 percent of our sample, as opposed to 50 percent for prop- erty/casualty firms); and even within property/casualty personal lines, the production of homeowners policies often afforded cler- icals greater discretion and wider responsibility than auto policy production. The crucial parameter in both cases seemed to be the difficulty of interpreting the necessary supporting reports (e.g., medical records, auto history, property evaluation). However, there were important differences in apparently identical product situations to emphasize the importance of a managerial decision level in the design of the labor process. Although each of these job configurations has different rami- fications for the employed work force, there are certain important similarities among all three that distinguish them from the earlier labor processes, beginning with their impacts on skill require- ments. First, even in the case of routine entry operators, clericals have an increasingly important responsibility for the quality of the data. In the context of integrated data bases, errors are more significant, more costly, and more difficult to detect and correct, because data entered anywhere throughout the system is immedi- ately recorded in all relevant files. There are multiple reasons for the new organizational and occupational designs just described, but in all cases an important motivation was the reduction of error; in each instance greater accountability was assigned to first-line workers. Single-source entry and team systems, for example, both encourage and demand that an individual or small group- often aided by feedback from the machines assume responsibility for the accuracy of their work. The move to decentralize and de- specialize entry was encouraged to an important degree because centralized work settings generated intolerably high error rates. To the extent then that responsibility is an essential component of skill, skill requirements have been rising. Similarly, there is evidence that even routine data-entry posi- tions involve greater levels of mental concentration than traditional text entry, to the extent that the work requires the operator to know and use a multiplicity of codes (Adler, 1983~. In the old- fashioned insurance typing pools it was possible, after sufficient experience on the job, to semi-automatically pull out the appro- priate forms from the appropriate stack and fill in the requisite information virtually without thinking, much like driving a car down a familiar route; the work was boring and demanded little in the way of mental effort.
BARBARA BARAN 47 As Adler notes when reviewing recent research, this kind of comfortable familiarity is never achieved when the language of the work is algorithmic. In addition, with increased computerization the span of functions for which any individual worker is respon- sible has widened; each transaction involves the entry of a wide- variety of codes and discrete pieces of information, each step of which must be performed quickly, accurately, and in the correct sequence. The new systems then demand a kind of continual men- tal alertness which might also appropriately be understood as a new kind of skill. Finally, the fact that the systems are continually changingboth in response to technological improvements and in the interest of product innovationmeans that the work force must be unusually flexible and adaptable and sufficiently polyva- lent to easily learn new routines, new codes, and new procedures. On the other hand, as Adler noted in the case of routine cleri- cal activity in banking, what may be most notable about many of the new jobs is precisely the extent to which they demand much of the worker with little reward (AdIer, 1983~: "If the abstrac- tion of means increases training requirements and mental effort, the abstraction of ends leads to a very different resultwork is often experienced as boring." There are also other roots to this perceived disaffection. Although objectively work has been reinte- grated through the expansive reach of the computerized system, for any individual worker, task variety may have diminished. For many insurance clerks, the previous work process involved a mul- tiplicity of activities, which in combination relieved the boredom of the day. Now all these functions are of the sane quality, that is, they center around the machine and the manipulation of its abstract symbols. Worse, the sustained concentration that is re- quired introduces a new element of stress. Finally, the sociability of a job declines dramatically in the transition to a computerized work setting. Evaluating the new skilled clerical categories is even more difficult. On the one hand, there is no doubt that they are an improvement over most clerical jobs along a series of dimensions: task variety has widened rather than narrowed, the span of decision making is broader, the responsibility for production is greater, and general training and educational requirements have risen. How- ever, pay scales remain low and in some cases work is closely monitored and even paced by the system, heightening occupa- tional stress. A recent survey by the 9 to 5, National Association
48 THE INSURANCE INDUSTRY of Working Women (1984) found that respondents whose work Is subject to computerized monitoring are much more likely to rate their jobs as "very stressfully than other office automation users. Similarly, those working under production quotas report signifi- cantly higher levels of stress than other office workers.6 Equally important, although skill levels may be rising, opportunities for occupational mobility may be diminishing. As lower-level pros fessional categories are eliminated, there is, In the words of one manager, increasingly a "quantum leap" between the computer- linked clerical positions and the next step up the occupational ladder. In this sense, career paths may be structurally truncated. Even for the professional and managerial labor forces the reor- ganization of work is only a mixed blessing. Although their work is likely to become more challenging, many jobs may be eliminated.7 WOMEN AND MINORITY WORKERS Because of the prevalence of occupational stratification by gender and race, this reshaping of work and of the occupational structure has particularly affected female and minority workers. 6 The survey presented a picture of two prototypical skilled, computer- linked clerical occupations: customer service representative and claims ex- aminer in which white claims examiners were more likely than any of the other clerical categories reported to find their work always ainteresting and challenging" (50 percent of claims examiners as compared to 30 percent of white legal secretaries, the second-ranked occupation). They were almost twice as likely as other clerks to work under production quotas and 44 percent described their jobs as Always very stressful (as compared to 29 percent of general clerks in automated settings). Customer service representatives were even more likely to have set production quotas (72 percent) and to experience their jobs as very stressful (60 percent). 7 In the case of sales workers, competitive pressures have probably been a more important source of change than automation although the two are clearly integrally connected. The primary impacts of the new technologies on the agency employees have been: first, a decline in their role as service providers and an increased emphasis on their sales function; second, a much closer connection on all levels with the rest of the organization. Skill requirements have clearly risen in some product lines particularly financial services planningas evidenced by new training and licensing requirements. However, in more standardized product lines, both the role and necessary competencies of the sales worker are being eroded; the direct marketing clerk described earlier is a good example of this kind of change.
BARBARA BARAN TABLE 3 Number of New Jobs Created in the Insurance Industry by Gender, 1960-1982 49 Year Number of New Jobs (thousands) Women Men Percentage of New Jobs Held by Women Insurance carriers 1960-1970 198.0 109.1 88.9 55.1 1970-1978 143.9 144.7 -(~.8 101.0 1978-1982 63.2 58.6 4.6 92.7 Agents/brokers 1960-1970 70.7 46.6 24.1 65.9 1970-1978 115.2 86.6 28.6 75.2 1978-1982 71.0 55.3 15.7 77.9 Total 1970-1982 393.3 345.2 48.1 87.7 SOURCE: U.S. Bureau of Labor Statistics (1983~. Therefore, it is impossible to evaluate the effects of the new tech- nologies on the work force without considering these more selective impacts. EFFECTS OF AUTOMATION ON THE FEMALE WORK FORCE Despite the fact that femaTe-dominated jobs are disappearing in great numbers, empirically the proportion of women in the insurance industry has grown dramatically in the last decade. By 1982, women comprised 61 percent of the industry's entire work force, up from 54 percent in 1970 (Bureau of Labor Statistics, 1983). Between 1970 and 1978 women claimed approximately 231,000 of the 259,000 new jobs created; and over the next 4 years women gained 113,900 jobs, whereas men gained only 20,300 (see Table 3~. In total during this period, female employment rose by 49 percent compared to a meager 8 percent rise in male employment. Since during this same decade clerical workers declined as a percentage of the work force, the increase in female employ- ment cannot be explained by the disproportionate growth of tra- ditionally female-typed jobs. On the contrary, what seems to be
50 THE INSURANCE INDUSTRY occurring is a major movement of women into traditional male occupations professional, managerial, technical, and even cleri- cal. Between 1970 and 1979, the proportion of female managers and officers grew from I] percent to 24 percent of the insurance work force; professionals grew from 17 percent to 38 percent; and technicians, from 38 percent to 65 percent. The percentage of women insurance examiners and investigators (formerly male clerical occupations) grew from 9 percent in 1962 to 26 percent in 1971 and to 58 percent in 1981 (Appelbaum, 1984~. Overall in the insurance industry, the ratio of women to men in professional and technical occupations rose by 19 and 27 per- centage points in 1971 and 1981, respectively, as opposed to a 4.3 percentage point gain for women throughout the economy in pro- fessional and technical occupations combined. Although the disag- gregated comparative statistics are not extremely reliable, women in the insurance m~ustry seem to have increased their share of professional employment more rapidly than in any other major sector of the economy (see computations in Baran, 1985:144~. There are a number of plausible explanations for this rapid transition from male to female labor. The first is simply affirma- tive action victories. Because of EEOC oversight and successful affirmative action suits waged against numbers of insurers, com- panies throughout the industry have developed more egalitarian hiring and promotion policies.8 At the same time, however, there is evidence that women are being hired in preference to men for the new computer-linked jobs, whether those are presently designated professional, technical, or clerical. For example, the professional stabs in the new highly automated personal-lines centers of one company are so overwhelmingly female that an administrator of one joked that they are under pressure to develop affirmative ac- tion goals for men. She explained that the reason the company chose to hire women was that they are more flexible than men in adjusting to the computer-mediated labor process. In another company, the introduction of computer-assisted underwriting has shifted the bulk of policy processing from a ~ For example, in response to affirmative action litigation, between 1977 and 1983, one company we studied increased its percentage of female man- agers from 4 percent to 33 percent; professionals, from 27 percent to 46 percent; and technicians from 29 percent to 67 percent.
BARBARA BARAN 51 department that is over 60 percent male to one which is entirely female. In still another, the change to a computerized claims process was accompanied by an increase in the percentage of female employees from approximately 25 percent of the claims force to over 60 percent. In line with this hypothesis also, the percentage of female professional and technical workers in the industry varies widely by product line. In the more highly automated life and medi- cal/health segments, women's share of employment in these occu- pational categories is considerably higher.9 The differences in the proportions of female labor are not solely attributable to automation; other characteristics of the job militate for or against the employment of women. For example, in all firms women hold a greater percentage of professional and man- agerial positions in the personal lines of business where external contacts are individual rather than corporate; similarly, women have been preferred for office rather than field operations (such as claims adjustment or sales). Nevertheless, the jobs which have been more accessible to women also lend themselves more easily to automation. Thus, although the congruence between female presence in an activity acid its level of mechanization to some ex- tent reflects an older discriminatory structure of employment, the growing dominance of women in these occupations and product divisions threatens to strengthen and perpetuate the inequality. In a small number of companies, there seems to be some concern over this prospect. A personnel manager in one major property/casualty carrier said that she worries that numbers of occupational categories are segregating females and may be de- valued as a result. Despite the rapid movement of women out of the clerical ghettos in this company, 79 percent remained in posi- tions that were 80 percent or more female and 71 percent were in occupations that were 90 to 100 percent female. Approximately 20 percent of nonclerical jobs were overwhelmingly occupied by women. 9 In 1979, 40 percent of all professionals in life carriers and 43 percent of professionals in medical/health carriers were women, as opposed to 35 percent in property/casualty companies. Similarly, 68 percent of all life insurance technicians were female and 81 percent of all medical/health technicians, whereas women held only 50 percent of all technical jobs in the property/casualty segment.
52 THE INSURANCE INDUSTRY Not surprisingly, then, although women are moving up in the occupational hierarchy, female wage rates in the industry re- mam extremely low. In our case study company, for example, female professionals earn only 16 percent more than male clericals (whereas male professionals earn 50 percent more); and the ma- jority of white female managers (57 percent) earn on the average $4 less per week than white male clericals. Overall in the in- surance industry, real average hourly earnings for nonsupervisory personnel fell by 4 percent between 1968 and 1978, in contrast to a slight rise in real wages during this same period across all industries. Disaggregat~ng by industry segment, the drop in wages corresponds directly to the increase in the proportion of female labor (Bureau of Labor Statistics, 1983~. Automation and feminization are thus proceeding as twin and highly interrelated processes and we should expect this trend to continue. Automation of professional functions is creating pre- cisely the kinds of jobs women have traditionally held in offices: the work is fairly routine, semi-skilled, and responsible, but offers Tow monetary reward and little opportunity for occupational mo- bility. As Oppenheimer (1968) has argued, employment of men instead of women in such occupations would mean a rise in the price of labor, a decline in its quality, or both. In today's environment, however, insurance companies are having an increasingly difficult time finding low-cost, high-quality female labor. As a result, such labor may be becoming a significant locational determinant. Nelson's (1982) study of the locational determinants of automated office activities (including insurance) concluded that, holding land costs constant, companies have cho- sen to site such operations in areas with a disproportionately high percentage of suburban married women. Compared with women in the central cities, suburban wives are often less career-oriented and may therefore be more willing to accept jobs with lirn~ted occupational mobility; because they are much more likely to be supplementing rather than solely providing the household income, they may be more content with modest pay scales.~° t° An executive of the Fantus Co., a subsidiary of Dun and Bradstreet, which specializes in corporate location, made an argument similar to Nelson's regarding insurance company relocations in particular (Best'& Rechew, 1979~; see also Kroll (1984~.
BARBARA BARAN 53 Ross's (1985) study of insurance company relocations and our own case work and survey tended to corroborate this hypothesis. Of the 75 percent of our sample who indicated that major spatial changes had occurred in their company within the last 5 years, 71 percent had moved a greater percentage of their operations to suburban locations. OverwheIrningly, they cited labor quality, labor costs, or both as the primary locational criteria. EFFECTS OF AUToMAT~oN ON MINoRITY WORKERS Although the relocations and, more generally, the changes occurring in the labor process may favor some categories of female labor, they threaten to have a negative effect on other categories. The situation of minority workers is particularly problematical. In general over the last 15 years, paralleling the progress of women, the position of minority labor in the insurance indus- try has improved significantly. Whereas in 1970, the percentage of blacks in insurance was seriously below the economy-wide av- erage in all occupational categories, by 1980 the EEOC reports indicated higher than average minority employment in many oc- cupational categories including higher percentages of black em- ployment. Disaggregating these gains by industry segment, it is clear that minority workers also fared best In highly automated sectors. Given the occupational structure of the industry, the majority of these minority employees are women. Approximately one-fifth of the female work force is composed of minorities as opposed to one-tenth of the mate work force. Overwhelmingly these women are concentrated in clerical positions (74 percent). Minority men also hold a disproportionate share of clerical jobs; 32 percent of all minority men in the industry are clericals compared to 7 percent of all white men. ii In the medical/health segment, in 1980, minority workers held 24 percent of all jobs; although their greatest share of employment was in clerical work (almost 30 percent), minority workers held unusually high percentages of jobs in all occupational categories. Minority employment was lowest in property/casualty firms (14.1 percent) and insurance agencies (12.4 percent), although even here in both cases minority workers held a disproportionate share of clerical jobs (U.S. Equal Employment Opportunity Commission, 1980).
54 THE INSURANCE INDUSTRY The explanation for the influx of minority workers is identical in most respects to the argument ~ made with regard to women: affirmative action gains and the automation of the labor process. At the same time, there was an unportant difference in the dynamic of entry that may be extremely important in the longer run. Long excluded from office-type employment, at least in the private sector, minority women especially black womenbegan entering the clerical work force in significant numbers for the first tune in the late 1960s. To some extent these women sought and were granted new job opportunities as a result of the Civil Rights movement and its demands for greater equality in the job market; to some extent, they were pushed to enter the clerical labor force as opportunities for employment as domestic service workers declined. Nevertheless, they entered the office after the division of labor in administrative support activity had become excessively detailedand they entered at the bottom, in what are often referred to as ~back-office" jobs. This is still where minority clericals are overwhelmingly con- centrated in the data-processing centers, typing pools, and filing departments. As of 1980, almost 90 percent of all secretaries and 84 percent of all receptionists were white, whereas minority workers represented 25 percent of ah typists; 27 percent of all file clerks, office machine operators (other than computer operators), and messengers; and 28 percent of ad mad! clerks. Undoubtedly, then, the high proportion of minority clericals in the insurance industry is a direct indication that much of its office activity is production-oriented, rather than more traditional administrative support. In addition, the kinds of industrialized work settings created by the first waves of mechanization in the industry favored minority employment. Today the concern is that these jobs are disappearing, through mechanization, relocation, and perhaps also the new labor process configurations. First, as Table 4 suggests, of the eleven clerical occupations with the greatest minority representation, all but two are declining; in contrast, virtually all of the growing occupations are dominated by whites. Second, the movement of automated insurance activities out of the central cities to suburbs and white t2 These data are for the economy as a whole (U.S. Equal Employment Opportunity Commission, 1980~.
BARBARA BARAN TABLE 4 Occupations in the Insurance Industry, by Race 55 Occupation Percent White Percent Black All occupations 82.0 10.0 Professional/technical/sales Underwriter 88.95 6.27 Computer systems analyst 89.05 4.64 Operations and systems research 88.73 6.02 Actuary 94.16 1.79 Statistician 84.86 7.88 Computer programmer 86.75 . 5.58 Insurance sales occupations 90.81 5.16 Clerical: above average blacka G eneral office supervisor 83.12 10.17 Computer operator 80.13 11.64 Peripheral equipment operatorb 79.44 12.67 Typistb'C 75.30 15.78 Correspondence clerks 81.08 13.43 Order clerk 81.08 11.69 File clerkb'C 73.13 17.52 Billing, posting, calculating machine operatorb 77.67 12.50 Duplicating machine operatorb'C 73.25 17.48 Office machine operator, not elsewhere classifiedb'C 73.17 Telephone operatorb 79.05 Mail clerkb 72.28 Messengerb'C 73.03 Clerical: below average blacka 17.12 14.38 18.72 17.59 Computer equipment supervisor 86.78 7.30 Financial record-processing supervisor 90.16 4.67 Secretary 89.00 5.73
56 TABLE 4 (continued) THE INSURANCE INDUSTRY Occupation Percent White Percent Black Stenographe b,c 84.74 9.10 Receptionistb 84.24 8.15 Bookkeeper/accountant/audit clerk 90.01 4.33 Payroll/timekeeping clerk 85.53 8.15 Billing clerk 85.34 7.95 Cost and rate clerks (including raters)C 85.29 7.88 NOTE: Employment share percentages are for the entire economy, not the insurance industry, for 1980. aClerical occupations are grouped according to whether the proportion of blacks is higher or lower than the proportion of blacks for "All occupations" 610.0 percent). Absolute decline 1970-1978 in insurance industry employment (Occupational Employment Matrix, Bureau of Labor Statistics, 1981b). CAbsolute decline 1978-1981 in insurance industry employment (Occupational Employment Surrey, Bureau of Labor Statistics, 1978, 1981a). SOURCE: U.S. Equal Employment Opportunity Commission (1980~. towns also threatens minority employment. Again using the ex- ample of our case study company in the absence of reliable ag- gregate data, the three new, highly automated "personal-lines" centers described earlier were located in towns where minorities represented 3.1 percent, 3.3 percent, and 14.3 percent of the popu- lation (whereas their representation in the population as a whole is closer to 20 percent). This company's centralized data-processing center (soon to be closed) and home office also relocated from a central city with a minority population of over 40 percent to sub- urban locations which are 90 percent white. Third, and somewhat more speculatively, the move to team-type work configurations, which involve close working relations among high-level and lower- leve! employees, may well favor the hiring of "socially compatible" white women (Nelson, 1982; Storper, 1981~. The historically low level of rn~nority representation in secretarial and receptionist jobs
BARBARA BARAN 57 lends circurr~stantial evidence to this hypothesis. As the bulk- processing centers are closed and key entry functions are moved to decentralized settings, minority clericals are in real danger of being displaced. And the centers are closing. A vice president in one company we interviewed predicted that within 2 years their six processing centers located across the country, which now em- ploy approximately 1,500 to 1,800 women, will have been closed. Although word processing is still somewhat centralized in many companies, over 70 percent of our sample in the survey responded that the trend in their company is toward decentralization of this function. Finally, and perhaps most important, are the prior disadvan- tages minorities face in the educational arena. If it is true, as believe, that the lesser-skilled jobs in the industry will continue to disappear, workers without adequate mathematics and literacy skills will be disadvantaged. To the extent that these occupa- tions increasingly involve interpersonal communication, workers for whom English is their second language may also be passed over. CONCLUSION To conclude, it seems appropriate to return briefly to the primary concerns raised in the office automation literature and to summarize the perspectives presented here on the most contentious questions: Will automation deskill white-collar work? Will office automation result in widespread technological redundancy? Will women workers bear the burden of the restructuring process? In terms of the deskilling debate, ~ have argued that three kinds of processes seem to be raising skill requirements in the in- surance industry. First, the rapid automation and elimination of much low-skilled work as computers assume responsibility for the most-structured functions is making the occupational structure increasingly top heavy. Conservatively, we can expect that during this coming decade roughly two-thirds of all new jobs in the indus- try will be nonclerical; over the next decade, net job gain should all occurinnonclerica~jobcategories(Baran,1985:234-238~. Second, the transfer of higher functions to less-skilled laboras decision parameters are embedded within computer softwareis creating new categories of skilled clerks, biasing the clerical hierarchy up- ward as well. For just this reason, in fact, there is some evidence
58 THE INSURANCE INDUSTRY that clerical skills are higher in mas~produced product lines where clerks are able to assume greater responsibility for the entire pro- duction process than in specialty lines where they remain adjuncts to professional labor. Finally, especially within the context of inte- grated data bases, computer-mediated work seems to demand new skills of the clerical work force. In all these ways, therefore, what we are witnessing ~ a rolling process of deskilling and reskilling which, in aggregate, should increase the industry's demand for skilled labor. At the same time, however, as ~ suggested at length earlier, opportunities for occupational mobility may decline, the quality of work life may deteriorate, and salary scales may remain low. Turning to the question of job redundancy, an important corol- lary to this argument about skills, ~ have argued that clerical job loss may indeed reach serious proportions over the next two decades. In the last 10 years women have moved rapidly into profes- sional, technical, and managerial occupations. However, since ap- proximately 70 percent of the almost ~ million female employees in the insurance industry are clericals, the predicted drop in clerical jobs would have important implications for employment opportu- nities for women. Assuming that women gain even 50 percent of ad new nonclerical jobs and continue to constitute 94 percent of all clerical employment gained or lost, under conditions of moder- ate declines in clerical employment women will gain slightly less than half of all new jobs created in the industry over the next two decades; in contrast, between 1970 and 1980, women gained closer to 88 percent of all new jobs. If the decline in clerical employment is more substantial as two of the studies cited predict (Leontief and Duchin, 1986; Roessner et al., 1985) female employment in the industry would actually fall by 10 percent or more, unless, of course, women claim a much larger share of nonclerical employ- ment (Baran, 1985:238-239~. While women are moving up the occupational hierarchy in significant numbers, at the same time women at the bottom may be losing their jobs. In this sense, the effects of automation on the female work force will vary importantly by class and race. For minority clericals and less-educated white women, especially in the central cities, the threat of redundancy is serious. For skilled clerks, particularly in suburbs and small towns, there will prob- ably be jobs but not opportunities, unless new kinds of training
BARBARA BARAN 59 programs are developed. For college-educated women, there may be new opportunities but there is also the danger, discussed ear- lier, that numbers of professional, technical, and lower managerial positions may resegregate females and be implicitly or explicitly reclassified downward as a result. The truth is that in a real sense the future is both uncer- tain and open-ended. Ultimately the question of whether the new technologies are used to create more good jobs than bad ones will depend on decisions made by the leadership of the firm; these decisions will be nonetheless shaped in significant ways by public policy. There is, for example, historical evidence that skill avail- abilities and shortages impact directly on job design (Levitan et al., 1981~. Public programs that absorb the costs of training- including both general education and the kind of ongoing retrain- ing that a period of rapid technological change requiresmake "working smarter" strategies both more attractive to companies and, in fact, possible. Policy wiD also play an important role in determining which workers bear the burden of the transition. Without a greater pub- lic emphasis on education and training, for example, it is likely that many of the workers presently employed in great numbers in the insurance industry will not only be expelled, but may also find themselves unable to secure comparable work. Similarly, aggres- sive affirmative action policies Knight act as a countertendency to the gender bias of job loss. One of the greatest dangers of studies such as this, which at- tempt to analyze the impacts of technological change on the work force, is their underlying assumption of technological determinism. In fact, the findings of these studies are ambiguous and indeter- minate not only because the processes we are analyzing are in a state of flux but, more important, because collectively we have considerable control over the eventual outcome of these processes. Nevertheless, with these caveats firmly in mind, ~ think that it is safe to predict that we need to begin to prepare our work force particularly our female work forcefor a labor market in which there will be many fewer routine clerical jobs.
60 THE INSURANCE INDUSTRY REFERENCES Adler, Paul 1983 Rethinking the Skill Requirements of the New Technologies. Work- ing Paper. Boston: Harvard Business School. Appelbaum, Eileen 1984 The Impact of Technology on Skill Requirements and Occupational Structure in the Insurance Industry, 1960-1990. Unpublished pa- per. Temple University, Philadelphia. Baran, Barbara 1985 Technological Innovation and Deregulation: The Transformation of the Labor Process in the Insurance Industry. Final Report for the U.S. Congress Office of Technology Assessment under Contract No. 433-3610.0. University of California, Berkeley. Bcst's Review 1979 Insurance office locations in the 1980s. But' Review (August):62- 63. Braverman, Harry 1974 Labor and Monopoly Capital. New York: Monthly Review Press. Bright, James R. 1958 Does automation raise skill requirements? Harvard Brained Review (July-August) :85-98. Bureau of Labor Statistics, U.S. Department of Labor 1965 Impact of Office Automation on the Insuranec Industry. Bulletin 1468. 1969 Tomorrow's Manpower Needs. Vol. II. National Rend and Outlook: Industry Employment and Occupational Structure. Bulletin 1606. 1978 Occupational Employment Survey: Insurance. Unpublished data. 1981a Occupational Employment Survey: Insurance. Unpublished data. 1981b The National Indwtry-Ocetlpation Employment Matriz, 1970, 1978, and Projected 1990. Bulletin 2086, Vols. I and II. 1983 Employment and Earnings 30(January). Cummings, Laird 1977 The Rationalization and Automation of Clerical Work. Unpub- lished Master's thesis. Brooklyn College, New York. DeKadt, Maarten 1979 Insurance: a clerical work factory. Pp. 242-256 in Andrew Zim- balist, ea., Case Studies in the Labor Process. New York: Monthly Review Press. Drennan, Matthew P. 1983 Implications of Computer and Communications Necrology for Less Skilled Service Employmcut Opportunities. Final Report to the U.S. Depart- ment of Labor under Grant No. USDL 21-36-80-31. New York: Columbia University. Driscoll, James W. 1980 Office Automation: The Dynamics of a Technological Boondoggle. Presented at the International Office Automation Symposium, Stanford University. 1979 People and the automated office. Datamation (November):106-112. Faunce, William, Einar Hardin, and Eugene H. Jacobson 1962 Automation and the employee. Annals of the American Academy of Political and Social Science (340~:60-68.
BARBARA BARAN 61 Feldberg, Roslyn, and Evelyn Nakano Glenn 1983 Technology and work degradation: effects of automation on women clerical workers. Chap. 4 in J. Rothschild, ea., Machina En Dea: Feminist Per~pectiocs on Technology. Elmsford, N.Y.: Pergamon. 1977 Degraded and deskilled: the proletarianization of clerical work. Social Problems 25:52-64. Greenbaum, Joan M. 1979 In the Name of Efficiency. Philadelphia: Temple University Press. Hoos, Ida 1961 Automation in the Office. Washington, D.C.: Public Affairs Press. Hunt, H. Allan, and Timothy L. Hunt 1985 An assessment of data sources to study the employment effects of technological change. Pp. 1-116 in Technology and Employment Effete. Interim Report of the Panel on Technology and Women's Employment, Committee on Women's Employment and Related Social Issues, National Research Council. Washington, D.C.: Na- tional Academy Press. Kroll, Cynthia 1984 Employment Growth and Office Space Along the 680 Corridor: Booming Supply and Potential Demand in a Suburban Area. Working Paper 84-75, Center for Real Estate and Urban Eco- nomics. Berkeley: University of California. Leontief, Wassily, and Faye Duchin 1986 The F~urc Intact of Automation on Workers. New York: Oxford University Press. Levitan, Sar, Garth L. Mangum, and Ray Marshall 1981 Human Repoured and Labor Maraca: Employment and Training in the American Economy. New York: Harper and Row. Life Office Management Association 1979 Word Processing Survey. Atlanta, Gal: Life Office Management Association. Matteis, Richard J. 1979 The new back office focuses on customer service. Harvard Bwinc~` Review 5 7(March-April) :146-159. Murphree, Mary 1982 Impact of Office Automation on Secretaries and Word Processing Operators. Presented at the International Conference on Office Work and New Technology, Boston. Nelson, Kristin 1982 Labor Supply Characteristics and Trends in the Location of Rou- tine Offices in the San Francisco Bay Area. Paper presented at the 78th annual meeting of the Association of American Geographers, San Antonio, Tex. 9 to 5, National Association of Working Women 1984 The To-5 National Survey on Women and Stress. Cleveland, Ohio: National Association of Working Women. Nussbaum, Karen, and Judith Gregory 1980 Race Against Time: Automation of the Office. Cleveland, Ohio: Working Women's Education P`und. Oppenheimer, Valerie 1968 The sex labeling of jobs. Indwirial Relations 7,3:219-234.
62 THE INSURANCE INDUSTRY Phister, Montgomery, Jr. 1979 Data Processing Technology and Economics. Santa Monica, Calif.: Santa Monica Publishing Company. Roessner, J. David, Robert M. Mason, Alan L. Porter, Frederick A. Rossini, A. Perry Schwartz, and Keith R. Nelms. 1985 The Impact of Ofi.cc Automation on Clerical Employment, 1985-2000: Forecasting Tcchr~iq~ce and Plausible Futures in Barging arid In~uranec. Westport, Conn.: Quorum Books. Ross, Jean 1985 Technology and the Relocation of Employment in the Insurance Industry. Unpublished master's thesis. Department of City and Regional Planning, University of (California, Berkeley. Shepherd, J. 1971 Automation and Alienation: A Study of Office and Factory Workers. Cambridge, Mass.: MIT Press. Sirbu, Marvin A. 1982 Understanding the Social and Economic Impacts of Office Au- tomation. Unpublished paper. Cambridge, Mass.: Massachusetts Institute of Technology. Storper, Michael 1981 Toward a structural theory of industrial location. In J. Rees, G. Hewing, and H.A. Stafford, eds., Industrial Location and Regional Syatenu. New York: J.F. Bergen Publishers, Inc. U.S. Department of Labor 1977 Dictionary of Occupational Fitly, 4th edition. 1981 Selected Characteristics of Occupations Defined in the Dictionary of Oc- cupational Titlce. U.S. Equal Employment Opportunity Commission 1980 Minoritice and Women in Pupate Industry. ZuboR, Shoshanah 1982 Problems of symbolic toil. Dissent 29(Winter):51~1.