Computer Chips and Paper Clips: Technology and Women's Employment, Volume II: Case Studies and Policy Perspectives (1987)

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.

Recent Trends in Clerical Employment: The Impact of Technological Change H. ALLAN HUNT and TIMOTHY L. HUNT THE GROWTH OF CLERICAL AND FEMALE EMPLOYMENT Clerical jobs are the largest single occupational group in the economy; they are also one of the most diverse. Generally people associate the traditional office occupations with the term "cleri- cal." Indeed, secretaries, typists, stenographers, file clerks, office machine operators, and receptionists do make up a large propor- tion of all clerical workers. But bookkeepers and bank tellers are also clerical workers according to the Bureau of the Census, as are bill collectors, insurance adjusters, postal carriers, factory expediters, and most enumerators and interviewers. The tremendous growth in the number of clerical workers in the United States is well known, but the true magnitude of this expansion cannot be appreciated without comparing it to the growth in total employment. Figure 1 shows that the proportion Facts and-observations presented in this document are the sole respon- sibility of the authors. The viewpoints do not necessarily represent positions of the W.E. Upjohn Institute for Employment Research. 223

224 0.20 0.18 0.16 O 0 14 _ . o o C£: J Cal J 0.12 0.10 0.08 0.06 0.04 0.02 o RECENT TRENDS IN CLERICAL EMPLOYMENT ; _ 1 940 1 950 1 960 1 970 1 980 YEAR FIGURE 1 Decadal growth in clerical employment as a proportion of total employment, 1940-1980. Source: Hunt and Hunt (1986~; based on 1940-1980 census data. of clerical workers in total employment has doubled in the last 40 years, from just under 1 employee in 10 in 1940 to 1 in 5 by 1980. However, the general recessionary conditions of the last 5 years combined with developments of the last 10 years or so in office technology raise one of the most puzzling questions about future employment: Will this trend continue? Those who are convinced that this trend cannot continue and may reverse itself base their predictions primarily on the introduc- tion to the office of microprocessor-based technologies. The incred- ible reductions in the cost of computing power, combined with the reductions in bulk made possible by microprocessor technology, may possibly constitute a revolutionary technical development. And there has been an apparent reduction in the rate of increase in the proportion of clerical workers. As Figure 1 indicates, while the clerical proportion rose almost linearly from 1940 to 1970, a slight reduction in the rate of increase occurred between 1970 and 1980. Is this the beginning of the end of clerical employment growth? If so, what will be the impact on women workers?

H. ALLAN HUNT AND TIMOTHYL. HUNT 225 Clerical jobs are also female jobs. Is it a coincidence that the expansion of clerical employment occurred simultaneously with the expansion of female labor force participation rates? To what extent have female job opportunities been linked to the expansion of the clerical work force? The overwhelming majority of clerical workers are in fact female, and this is even more true today than it was 30 years ago: from just over 60 percent female in 1950, the proportion grew to nearly 80 percent by 1980. Indeed, not only are clerical workers increasingly women but women are increasingly clerical workers. Between 1950 and 1980, the proportion of women workers who worked in clerical jobs grew from about 27 percent to over 35 percent. Thus, the sex segregation of clerical occupations appears to have been increasing, although there was very little increase in the proportion of females employed as clerical workers between 1970 and 1980.i The participation rate for men in clerical work was only 7.6 percent in 1980. Clearly, females are much more likely than males to work as clerical workers and clerical work has increased in importance as a source of employment for women. This paper reviews trends in clerical employment over the last 30 years and seeks evidence of the impact of changes in pro- cess technology on those trends. It also assesses prospects for the future of clerical employment. The next section reviews clerical employment trends from 1950 to 1980 by decade, as well as annual employment changes.2 Because this period encompasses the first computer revolution, the introduction of mainframe computers to the office, and the beginnings of the microcomputer age, the review can be interpreted as a search for the employment effects of tech- nological change. If changing office technologies displaced large numbers of clerical workers during the first computer revolution, the evidence should be in the employment record of the 1960s and 1970s. Similarly, if the current office technologies threaten clerical jobs, some evidence of this should be found in the employment figures of the early 1980s. The following section of the paper discusses the determinants of clerical employment in the broadest sense. The influence of industry occupational structure and industry employment trends ~ See Reskin and Hartmann (1985) for a discussion of sex segregation on the job. 2 A much more thorough review of existing data is provided by Hunt and Hunt (1986:Ch. 2 and 3~.

226 RECENT TRENDS IN CLERICAL EMPLOYMENT on clerical employment totals are examined. The aggregate change in clerical employment from 1972 to 1982 is decomposed into portions attributable to general economic growth, changes in the sectoral composition of the economy, and changes in occupational staking patterns. Evidence of the direct impact of technological change on office employment levels is sought for the finance and insurance industry, reputedly the most advanced user of office automation systems. This paper does not try to assess the influence of other impor- tant factors that will determine future labor market outcomes for clerical workers. In particular, there is no consideration of future supply issues. If female labor force participation rates continue to rise as they have in the past, the issue of job creation for women will be of even greater significance. On the other hand, if women increase their penetration of nontraditional female occupations, the number of females seeking clerical positions in the future may decline. Whether men are more likely to begin to look to cler- ical positions for career opportunities in the future presumably depends on labor market developments for clericals, as well as the job outlook in more traditional male occupations.3 Clearly these considerations are crucial to understanding whether the supply and demand of clerical workers will be in approximate balance in the labor market of the future, but these questions are beyond the scope of the present effort. We seek only to illuminate past trends in clerical employment and investigate the causes behind those trends. Throughout the analysis we strive to develop an under- standing of the employment implications of technological change for clerical workers. In a concluding section, we draw on these find- ings and on a review of existing forecasts of clerical employment to narrow the range of uncertainty about the probable future impact of technological change on the demand for clerical employment. A CLOSER LOOK AT CLERICAL EMPLOYMENT TRENDS EMPLOYMENT FROM 19 5 0 TO 1 9 8 0 DECENNIAL CENSUS DATA Table 1 reports the best derivable estimates of detailed cler- ical employment on a consistent basis across the 1950 to 1980 3 Many of these issues are addressed by Hartmann, Kraut, and Tilly (1986).

H. ALLAN HUNT AND TIMOTHYL. HUNT 227 time span. These figures are based on data from the decennial censuses, and although they are far from perfect, everything that can be done has been done to maximize the consistency of the estimates and minimize the distortions introduced by the mea- surement system and changes in it.4 Table 1 shows that there were just over 19 million clerical workers employed in 1980 in 42 separate clerical occupations ranging from secretary, the largest, to tabulating machine operator, the smallest. There were more than 4 million secretaries employed in 1980; they represented just over 4 Because these data have been adjusted rather extensively for consis- tency, the figures reported here do not correspond exactly with census figures from other sources. Comparisons of occupational data among decen- nial censuses are complicated, first, because the data came from a sample of all census respondents (though the numbers are very large by normal sampling standards), and second, because the measuring rod, the occupa- tional classification system, changes between censuses. In 1950, occupational employment was tabulated in 12 major groups and 469 detailed occupa- tional categories. In 1960, the same 12 major groups contained 494 detailed occupations; in 1970 there were only 417 detailed occupations but still ac- cumulated into 12 major occupational groups. The overall changes in the classification system can be regarded as relatively minor over this period, although with regard to individual occupations, major distortions can occur when an occupational category is added or deleted. In 1980, however, the magnitude of the differences in the occupational coding system are enor- mous: 503 detailed occupations, which have been reshuffled into 13 new major groups. For example, cashiers, who have previously been classified as clerical workers, are reclassified as sales workers; 1.65 million workers are thereby moved from one major occupational group to another. For the first time, there is a fundamental lack of consistency at the major occu- pational group level between adjacent census observations. To convert all occupational employment numbers to a consistent basis, the classification system of 1970 was chosen as the standard upon the advice of the Bureau of the Census. The comparison between 1960 and 1970 employment in terms of the 1970 classification was readily available from the Bureau (Bureau of the Census, 1972:Table 221, and Priebe et al., 1972~. The 1950 census employment could not be converted directly into 1970 categories, and were first reclassified into 1960 terms, using '' technical Paper No. 18; then those numbers were converted to the 1970' basis using Priebe et al. (1972~. The 1980 data were converted to the 1970 basis- using preliminary unpublished results from the Census Bureau. It requires a painstaking effort to bridge from one decennial census to the next in this manner, and the accuracy of the results is uncertain. All of the reclassification work is done on the- basis of sample results. The reclassified employment figures are thus subject both to the original sampling error in estimating occupational employment and the secondary sampling error involved ' in the reclassification study. These issues are discussed more fully in Hunt and Hunt (1986:Ch. 2~.

228 Go Go ·~ E o - :^ CC of Go o To US - o ._ As o rS ·O v ·_ >. m ~ ~ V o - 0Oo CO . - - o ._ Cal o 00 ~ O ~ ~ O ~ ~ ~ ~ ~ 00 CD ~ CD Cat ~ ~ us ~ ~ as co cot co oo oo 0 ~ u: ~ cs~ ~ cs oo co u: e~ CC ~ CD ~ ~ Ch ~ C~ u3 0 ~ C5) oo O ~ ~ o: C53 0 CO CO C9 e~ 00 CO ~ U: 00 0 ~ CD ~ 00 ~ 00 ~ ~ ~ CO ~ ~ ~ U: ~ ~ ~ CC ~ C53 ~ ~ _ C) ~ <9 ~ O _1 ~ ~ O) ~ O ~ ~ e~ ~ ~ c~ ~ 0 ~ u' u3 0 00 CC ~ ~ CD 00 C~ ~ 00 ~ O _ _ _ _ _ _ _ _ _ _ _ _ _ _ oo cs~ ~ c~ co ~ cs ~ e~ co 0 u: C~ co U: U~ ~ CO ~ oo O ~ 00 C~ O CO 00 ~ oo oo oo ~) ~ ~ 0 ~ co ~ e~ _ _ _ _ _ (D o: e~ ~ ~ {D ~ ~ O ~ ~ ~ CO u~ CD ~ ~ ~ O . a) e~ _~ ~ 0 t- 0 a~ e~ ~ o~ ~ ~ ~ o~ ~ ~ (D _ _ _ _ _ _ _ _ _ _ _ _ _ _ O CO 0 00 ~ ~ ~ Ut Ch—I co ~ ~ _I ~ ~ ~ oo cs e~ a~ t- e~ (D U~ Ut (D ~ ~ c~ ~ CO ~ CO _ _ _ _ ~ ~ ~ _t CD C9CDOOCDa,C~CO - CS}=U3~pOCO o ~ CD o Ut U~ CO CO 00 C9 00 C~ U~ ~ ~ O C~ CD ~ O ~ ~ C~ ~ a _ _ _ _ _ _ _ _ _ _ _ _ _ _ oo l~ ut l~ ~ e~ a: 0 ~ ~ cO CD N {D O 00 ~ U3 U~ CD t- t~ I a) _I oo 0 _I ~ e~ a~ _I _ _ _ _ C9 o O ,Y ~0 0 3 3 _ c ~ ._ ._ s" s" ~ a, _ ~ ~ ~ ~ 5 E 3 ~ ~ ~c4' ~ ~ — ~ C., ~ o CO oo .<s ·— ~ o o ~ ·— O ~=mv~ v · V. 0 _ v t0~o, a~ O > ~ o.= rY.. ~ ~ ~ ~ O b.— O q, O q, ~ C. ~ O O ~ U: ~ M m ~ v v co co 0 a~ co co u3 ~ ~ ~ . . . . . . . . · . c~ ~ ~ a, co ~ 0 co C~ ~ 00 <9 ~ ~ ~ ~4 ~ Co ~ O ~ ~ C~ ~ ~ ~ CO C9 00 e~ 0 ~ CS u: ~ ~ oO oo ~ 00 ~ C~ ~ ~ ~ ~ U: co co ~ co co co c~ ~ ~ ~ ~ oo cO ~ ~ ~ u, oo - oo o ~ ~ cO o: co ~ oo ~ ~ e~ ~ ~ co oo - - - - - - - - - - o ~ oo u, oo C~ o: (D C9 CD ~ ~ ~ e~ co ~ ~ e~ c. co 0 ~ ~ 0 e~ u~ e~ co ~ e~ 0 co ~ CD ~ ~ c~ ~ O O O 00 ~ ~ 00 ~ ~ ~ C5) - - - - - - - - - - c~ ~ cs ~ e~ c~ ~ co co e~ co u: ~ ~ ~ e~ o: ~ e~ a, ~ oo ~ ~ ~ e~ {D ) oo 0 co c~ e~ ~ ~ ~ oo - - - - - - - - - u~ ~ CO oo {D CO a~ ~ us Ct ~ ~ ~ O CD CD e~ cO 0 - L. ~ - ~ 0 0 tell O o L' O o~.Y ~ o ~ ~ ~ ~ ~ 9 ° ~V oq (V ~Y ~ ~ ~ ·O (~' `, 5 t} 00 ~ V ~ ~ ~ ~ ~ ~ ~

229 ~ Cat . . en ~ ~ co ~ ~ 0 ~ en 0 oo ~ CO oo ~ on CD ~ O Cal ~ u, us en ~ ~ us 0 ~ ~ a: 0 ~ ~ ~ co en cot ~ ~ ~ co co en cat up en ~ co ~ en cat ~ O ~ ~ ~ en en oo O ~ (D en ~ en 0 ~ ~ oo ~ Cat CD ~ ~ ~ o oo U. ~ ~ ~ ~ oo oo ~ ~ ~ ~ of en 0 ~ co up ~ ~ ~ ~ ~ 0 Go ~ up ~ CO ~ 00 ~ ~ ~ ~ ~ ~ Cat ~ 03 CD oo 0 en ~ oo ~ ~ Is 0 ~ ~ ~ ~ CO CO O O ~ ~ CO 49 CO CD ~ 03 en on o ~ ~ Go ~ ~ ~ U3 0 ~ en Cal ~ Go O ~ Cry ~ CO —I CS, oo ~ ~ _ _ _ ~ _ _ _ _ en 0 ~ en ~ en ~ 00 Ut Up Go lo- ~—~ ~ ~ ~ O CD to ~ Go CD ~ en en co ~ ~ 0 ¢9 ~ on ~ {D co UP co ~ en en ~ ~ Go ~ 0 en co 0 0 en ~ 0 up Go en 0 0 Cal ~ cat ~ ~ ~ ~ ~ O CD 03 =~ ~ ~ ~ ~ ~ ~ ~ ~ Go a, O~ U. ~ CO ~ 0 ~ 00 CO 00 up DO 00 ~ ~ UP ~ up ~ CO ~ ~ . ~ 00 ~ CD {D UP ~ co ~ Go ~ ~ {D al on ~ en lo ~ ~ ~ co en en ~ UP on ~ ~ U: ~ ~ on Cot Go ~ ~ ~ 00 O C9 CD O 03 ~ Ed ~ ~ 0 ~ C53 up Go ~ ~ 0 Go en co UP 0 ~ ~ on {D ~ ~ CD a, a, ~ CD ED ~ up co ~ ~ 0 al co ~ co ED Go .a~ ~ en ~ ~4 ~ m 0 In O ~ ~ ~ ~ ~ O. 00 ~5 00 .= X X o) C ^ ~ al ~ ~ ~ `~i ~ ~ c, cs ~ ~ "c O A ~ A ~ > ~ ~ ~ ~ E a: sit o o ~ Go o o Cap ~ U. en ~ ~ en (D~= 0~ =~ - = CO o Ed U. ~ CO ~ ~ , ~ . an, A, ~ ~ ~ At_ An, tom 3mo~oE- :~ :^ GO ._ 0c o v al ._, al L. 3 a ._ a' _ a: 3 ~ °° V _ Cal ~ a 11 11 · ~ a . · . .. v 0 0 z u,

230 RECENT TRENDS IN CLERICAL EMPLO YMENT 4 percent of total employment and 21 percent of clerical employ- ment in that year. The second largest category was bookkeepers, with about 1.8 million employed, followed by cashiers, with 1.7 million. The only other clerical occupation that has approached 1 million employees is typists. Together, these "big four" clerical occupations accounted for 8.5 million jobs, or about 45 percent of all clerical employment in 1980. These same four occupations accounted for only 27 percent of clerical employment in 1950; all four of these occupations have grown substantially in employment during the last 30 years. On the other end of the scale in terms of size, there were only about 3,300 tabulating machine operators and about 7,600 telegraph op- erators employed in 1980. These occupations have been declining for some years, as have the next two smallest occupations, dupli- cating machine operators and calculating machine operators. Each of these occupations has been adversely impacted hv ~.h~.n~.c: in technology. ~ ¢ ~ ~ ~) ~^ ~~A b—= 111 When these same data for the various occupations are ex- · ~ · Omened In terms of their annual compound rates of change in employment between 1950 and 1980, computer and peripheral equipment operators far exceeded all other clerical occupations in their rate of increase. This occupation has grown from an employ- ment level of 868 persons in 1950 at the dawn of the computer age to nearly 400,000 persons in 1980, an annual rate of growth of over 22 percent. This is the labor market expression of the computer revolution, which began to substantially affect employment levels in computer-related occupations in the 1960s. It is interesting to note that the second fastest growing clerical occupation over the 1950 to 1980 period was teachers' aides: from high tech to high touch in one easy step! The number of teachers' aides increased from 6,000 to over 200,000 in this 30-year period, or about 12 percent per year. The third fastest growing clerical occupation was typists, even though there was actually a 23 percent decline in employment from 1970 to 1980. The phenomenal growth of typists in the 1950s and 1960s was sufficient to offset the recent decline, for an average annual rate of growth of 9 percent when the entire 30-year period is considered. Following in order of rate of growth are library attendants, clerical supervisors, bank tellers, receptionists, and cashiers. Clearly, there is not a high-tech oc- cupation among them, although they have all been impacted in

H. ALLAN HUNT AND TIMOTHY L. HUNT 231 one way or another by technological change as well as many other influences. There were also a few clerical occupations that showed abso- lute declines during this Midyear period. The most rapid declines were among stenographers and telegraph operators, declining in employment by about 5 percent annually. Both occupations have been impacted by technology, but not in an obvious way. The telegraph has been all but replaced by superior communications devices, and this has nearly eliminated the jobs of telegraph oper- ators. Improvements in dictation equipment and changing habits of users have spurred the decline in the stenographer occupation. In 1950, there were 2.3 secretaries per stenographer while by 1980 the ratio had risen to 44 to 1. Fairly rapid declines were also shown by tabulating machine operators and weighers. Actually, the tabulating machine opera- tors would have been the most rapidly retreating if 1960 had been taken as the base year. This occupation provides an excellent ex- ample of a technology-specific occupation that experiences rapid growth and then declines. Tabulating machines were very popu- lar in the 1950s for analyzing data on punched paper cards. The number of tabulating machine operators nearly tripled between 1950 and 1960. But data-processing technology moved rapidly beyond the capabilities of tabulating machines, and the number of employees in this occupation has fallen by nearly 90 percent since 1960. Rounding out the declining occupations are messengers and office helpers, calculating machine operators, and telephone op- erators. All appear to be office-technology-related declines, since the communications and computing capabilities of modern offices have rendered these jobs less essential than in the past. With the spectacular exception of the computer operator cat- egory, the rapid-growth jobs do not show any particular high- technology bent. On the other hand, the declining occupations do seem to offer a technological interpretation, at least in part. Whether this represents a general principle is not clear at this time. What is clear is that the bulk of clerical employment occurs in a few very large, very diffuse occupational titles, such as secre- tary and bookkeeper. This was more true in 1980 than in 1950. It is possible that one impact of office technology over the last 30 years has been to foster more generality in job title and perhaps in duties, but that cannot be conclusively demonstrated with the data that are currently available.

232 RECENT TRENDS IN CLERICAL EMPLOYMENT ANNUAL EMPLOYMENT CHANGES, CURRENT POPULATION SURVEY DATA The long-term decennial census data do not seem to demon- strate a widespread impact of technology on clerical occupations, but annual data from the Current Population Survey (CPS) may be more revealing. Figure 2 shows aggregate clerical employment as a proportion of total employment on an annual basis from 1958 to 1984. It clearly shows that the rate of increase of clerical work- ers relative to all employment was much slower in the 1970s than it was in the 1960s.5 Even more apparent is the stagnation in the proportion of clerical workers since 1980. Clerical workers did not fare as well in the last recessionary period as they did earlier. It is less certain what the downturn in the clerical proportion in 1984 means. Such a decline has been typical of recovery periods in the past (as in 1976-1977), when the number of production workers rises rapidly to restore the prerecession balance between produc- tion and nonproduction workers, including ciericals. Whether the trend of the early l980s is something different is not yet clear. The magnitude of the drop is unprecedented, but that does not prove that the cause is fundamentally different. A look at recent annual CPS data for detailed occupations may be instructive. Unfortunately, the only time period for which this can be done is the decade from 1972 to 1982. If the mi- croprocessor revolution is going to have catastrophic impacts on clerical employment, it should have become apparent by 1982 when the microcomputer population reached the 1 million unit level (Computer and Business Equipment Manufacturers Associ- ation, 1985:87~. While this period would seem to be adequate for analysm, it is complicated by the fact that the recession of 1981-1982 occurs right at the end of the period.6 In addition, because of the smaller sample used by the CPS, some reservation must be expressed about any particular annual observation. More 5 The ~r~n~r~nt: Harry ;- 1071 ~ ..1~ :~ ~ ~ n . .. ~.~ r I ,-~^ .~vu~~ Lo; 1~,llor1:u as lb resects one conversion to new census codes rather than any actual change in clerical employment levels. 6 It is frustrating to stop the analysis in 1982; however, the massive reorganization of the occupational classification system introduced to the CPS in 1983 (corresponding to the 1980 decennial census reclassification) prevents the development of consistent data for all occupations after 1982.

H. ALLAN HUNT AND TIMOTHYL. BUNT 0.19 0.18 o - CC O 0.17 o Or: CL 6 0.16 UJ J Cat 0.15 0.14 1960 1 964 YEAR 233 ,~ 1968 1972 1 976 1 980 1 984 FIGURE 2 Annual changes in clerical employment as a proportion of total employment, 1958-1984. Source: Hunt and Hunt (1986~; based on 1958-1984 CPS data. confidence can be put in trends that emerge over a period of 3 or 4 years. Table 2 shows the CPS clerical occupations sorted by the an- nual rate of change over the 1972-1982 decade. As in the decennial data for 1950 to 1980, computer and peripheral equipment oper- ators experienced the most rapid rate of increase of any clerical occupation, although it^was only about half the average annual rate shown for the 1950-1980 period. Bank tellers and insur- ance adjusters, examiners and investigators both edged ahead of teachers' aides in growth rates during the more recent decade. This reflects the falloff in the rate of growth in teachers' aides as employment growth in education as a whole faltered because of funding difficulties and a reduction in the student population. Other clerical occupations showing relatively rapid growth during the 1972 to 1982']ecade include cashiers, estimators and investigators, and receptionists. All three of these occupations involve direct customer contact and probably would fall into the "hard to automate" category. Messengers and office helpers emerge as a relatively rapidly growing clerical occupation in the

234 ~ be ~ V en Go o Be. C~4 - be Cat TIC v ¢ - - C~ Go o en - oo o ._ As o - ct v .c o Cat m al c - lo. en 00 00 CO en 4, - - o . C: o ~ 00 ~ O ~ . . . . . ~ CS) CD CD u' O c~ 00 ~ ~ ~ ~ O ~ ~ {D . . . . . . . . . . . U~ ~ C~ CO CC C~ X ~ O C~ CO O C~ U: ~ ~ _ C~ O ~ ~ <9 CO CD O ~ 00 ~ ~ ~ 00 00 ~ ~ ~ ~ ~ CD C~ 00 0 u, ~ a: ut ~ ~ O {D C5) oo co ~ ~ ~ ~ ~ u: u: u: u: CD ~ ~ ~ ~ ~ ~ 0: 00 u: C~ ~ _ C~ u: ~ oo ~ ~ C5) ~ u, oo ~ eo CO ~ ~ ~ C~ ~ ~ ~ 00 CD ~ 00 ~ C~ U: ~ _ CO U: U: CO _ ~ ~ ~ ~ ~ _ ~ _ ~ ~ ~ ~ ~ ~ ~ ~ ~ 0 oo _ _ ~ C~ ~ CO O ~ u: _ O ~ ~ ~ ~ _ CD ~4 ~ (5) u: u: ~ co ~ ~ CD _ oo c~ c~ ~ ~ _ CO 00 CO 00 ~ O O ~ O ~ u~ ~ ~ 00 CD 0 U, ~ ~ CO {9 0 oo oo ~ 0 U: 0 ~ ~ 0 CO ~ ~ ~ ~ ~ ~ ~ C9 cs e~ CS) ~ ~ e~ ~ o~ C5) ~ u~ a~ C5) 0 e~ ~ ~ ~ C~ U: ~ ~ ~ ~ ~ oo U, C~ CO ~ _ o' CO u, oo a) ut 0 a, ~ o' CD O ~ ~ ~ O ~ 00 c5~ ~ u~ e~ ~ oo co ~ ~ ~ co ~ e~ oo ~ oo ~ a~ c~ ~ ~ ~ ~ c~ F ~ _ ~ _ _ ~ ~ CO e~ co a~ e~ 0 ~ oo a~ a~ ut u~ ~ oo ~ _ ou) u: O0 oom~r~~` ~c~ ~ooe~ oo ~ ~ c~ e~ co ~ ~ o~ c~ o: oo U: ~ ~ ~ oo ~ ~ U, ~ ~ 00 oo o, ~ ~ o u~ ~ u~ c~ ~ co ~ co ~ oo ~ a~ 0 O ~ e~ e~ a e~ 0 ~ 0 ~ a, 0' CD a, ~ ~ ~o ~ oo C~ O CO ~ _~ Co _ _ C' 00 ~ O oo {D e~ 0 CD _ _ CO ~ a~ 0 ~ oo ~ O ~ ~ ~ ~ a) a, 0 co c~ ~ 0 0 ~ u~ a~ ~ co e~ e~ o: 0 a~ co c~ c~ ~ c~ ~ co ~ ~ o, e~ ~ ~ a, ~ C~ L' o _ bO C~ ._ s" ·l a~ O _ ~ ~ ~ aS $4.— X s~ s.. ~V ~ ~ °q Pe p, a, ~ ~ ~0 _ ~ L. o bO ._ CO ._ a, 4, ~ O ~ 0e C o tt,.~ ~ .o ~ ~ ~ 3 ~ ~ ~ ~, ~ ·c ~ o ~ ~ - ,= U 0 ~ ~ ~ 00 0 E ~ ~=, ~ x ~ ~ ~ .cE c ~, u O :, ° c ~ x ° v m ~ ~ vm ~v>~m ~v

235 u~ ~ ~ ~ ~ r~ c~ 0 oo co ~ ~ co ~ oo co oo c~ . . . . . . . . . . . . . . . . . . C~ ~ C. ~ ~ ~ ~ ~ o o o o o o o C~ ~ CO ~ , ~ ~ ~ ~ 0 oo u~ ~ ~ C~ ~ O ~ oo ~ ~ ~ e~ <9 CO ~ ~ CO C~ U ~ Ci~ U: U: ~ eD ~ C~ ~ oo ~ (9 oo ~ ~ co e~ ~ co ~ ~ ~ e~ c~ ~ ~ ~ c~ ~4 oo Ut ~ o ~ oo o ~ C~ CO U: C~ ~ ~ C9 ~ CO ~ CD C~ ~ u~ ~ ~ C~ CD CO O c~ ~ ~ co e~ ~ co u, ~ ~ o: ~ u: C~ O CO U: C~ CD ~ ~ 00 u~ ~ u: C~ ~ ~ ~ CO C~ 03 ~ ut ~ ~ CO 0O ~ ~ CS~ ~ u: c~ ~ ~ co e~ ~ o: u, ~ ~ co c~ ~ e~ 0 co cs' C5) ~ ~ ~ oo ~ co oo ~ e~ co ~ ~F ~ CO o ~ o ~ ~ ~ ~ ~ CO ~ u3 CO <9 C~ 0O ut c~ ~ ~ ~ ~ a~ ~ ~ —4 c~ c~ u~ c~ O co CO ~ ~ O ~ 00 CD e~ oo ~ CO oO CD ~ ~ ~ u: ~ ~ C9 oo cr) c~ ~ a~ ~ ~ ~ c~ e~ u~ e~ 0 o ~ ~ "' CO _' N ~ ~ ~ ~— O ~ U, _l O t_ CO ~ oo a) u~ co a~ oo ~ t~ ~ u~ oo ~ 0 t- e~ ~ u~ oo N t— CO O~ _~ CO ~ _1 —1 N N u~ O~ O CO C53 ~ N O~ —I ~ ~ C9 @) O ~ ~ O} ~ ~ C~) ~ _I t~ ~ ~ _~ N U:~ ~ ~ ~ t— ~ ~ CS) ~ ~ ~ O N t— 03 N _~ CO ~ ~ _ N N ~ N C~ CO Cr~ C/~ ~ ~ N 00 _I CO {D U~ 00 ~ ~ a) U:) 00 0 _I ~ ~ 0 co 0 co o~ o~ ~ ~ eo u: r~ Cb co ~F {D O N ~— 03 N ~ 03 ~ _I _1 N N ~ N O a) _I _4 N CO 00 C~ 0~ 0 ~ U~ 00 O~ 00 03 U~ C~ 03 ~ ~ lO ~ 0 e~ 0 N 1O ~ @) u:~ ~— ~) ~ <~ ~ ~ U:} O N t— 0O N ~ 03 ~ _I _t O~ N ~ N O a) u, oo o' ~ O ~ e~ ~ o' ~ ~ ~ oo o' O O ~ ~ U) GO t— O O _I U~ CO N ~ 00 CD ~ O ~F ~ 1O 0 O~ ~ O~ N _~ CO ~ _I r4 e~ O~ ~ a~ O ~ _1 ~I C~ N _' le' O —1 @) oo ~ ~ _ I 03 N 1O ~ ~ 1O oo oo O) 0 oo @) o~ IO ·) ~ ~ t oo N a~ ~ N N It) @~ _1 —1 CO ~ _I _I N N 1&~ N O CO ~1 _1 ~ 4, ·o ,Y ~ co ~ a 4, ~ ~ ~ 4, b r O ~ ·, 5 0 1! w ,_ w u A a' ~ 5 _ ~ '. ~ ~ w ~ ~ ~ ~ C O .= :~:~mcn~vcn~m~E-m ~n ~ :~ ._ ~ ._ 3 4~ a, 4~ A4 11 ~ V . . · V o o Z; ~

236 H. ALLAN HUNT AND TIMOTHYL. HUNT 1970s, which is in contrast with their declining employment from 1950 to 1970. The number of bill collectors increased at 3.6 percent annually during the decade, and non-Post Office mail handlers in- creased at 3.5 percent. No particular pattern emerges from the listing of clerical occupations that grew more rapidly than average during this recent decade. At the other end of the distribution, the declining occupations, stenographers and telephone operators, are joined by bookkeeping and billing machine operators in rather rapid decline for the 1972 to 1982 period. Small annual declines were registered for typists, postal clerks, mail carriers, and stock clerks and storekeepers. Bookkeeping and billing machine operators are another clerical occupation that may be impacted by the microprocessor revolu- tion. As microcomputers have become more widely distributed, increasing attention has been paid to creating accounting software that will run on the micros. This has undoubtedly impacted the number of bookkeeping machine operators. What is not clear is whether it has reduced the number of people doing bookkeeping work. Since they are not doing it on a special purpose device, it would no longer be appropriate to call them bookkeeping machine operators, however, and the job titles are very likely changed. Without more careful study of both inputs and outputs of of- fice production, definitive conclusions about the impacts of office technology cannot be drawn. This analysis leaves the impression that office technology cannot be regarded as an important deter- minant of clerical employment for more than a few occupations. Nonetheless, it is also the case that clerical employment growth has slowed in the last decade. DETERMINANTS OF CLERICAL EMPLOYMENT Analysis of the trends in occupational employment indicated that some clerical jobs were growing while others were declining. On an aggregate basis, it was seen that clerical jobs as a whole were becoming relatively more important as a proportion of total jobs in the economy, although that growth slowed in the 1970s. It also appeared that in the recession of 198~1982 the proportion of clerical jobs did not increase significantly as it has in past recessions. What might be causing those movements?

H. ALlAN HUNT AND TIMOTHYL. HUNT 237 In the broadest terms possible, employment depends upon the demand for output and the productivity of the workers who pro- duce that output. National output is generally measured by gross national product (GNP), the value of all final goods and services produced In the economy in a year. This simple relationship, al- though devoid of occupational and industrial content, emphasizes two important points. First, if one accepts the notion that productivity is more or less fixed in the short run by the technological structure of production, then it should be clear that changes in GNP aggregate demand in the economy—drive any changes in employment. Many factors affect both the level and rate of growth of GNP, including unfore- seen shocks to the economy, such as the energy crises of the 1970s, which at least temporarily disrupt the national economic system, and business cycles, which may vary tremendously in terms of their length and severity. All occupations and industries are adversely affected by the failure of GNP to grow sufficiently; likewise, all oc- cupations and industries tend to benefit from reasonable economic growth. The second factor that influences employment is productivity. The concern, of course, is that productivity growth will outpace the growth of GNP. During recessionary periods it is not unusual for the lack of jobs to be blamed on labor-saving technology, and office automation is a case in point. However, productivity growth and GNP growth are actually intertwined: productivity gains allow the possibility of economic growth. Historically, technological change has not created permanent unemployment for millions of workers; instead it has raised living standards. To be sure, there have been winners and losers in this process, both among firms and individuals, but the net result has been real economic growth. No one can guarantee that history will repeat itself with office automation today, but some appear to be too easily persuaded that history will not repeat itself, that office automation and other labor-saving technologies will wipe out millions of jobs. The demand for labor is a derived demand based upon the de- mand for the goods or services which that labor produces. The rise and fall of occupations is related to the rise and fall of products and services which are produced in particular industries. Prospects for clerical workers, then, are linked to the prospects of the industries in which they work.

238 RECENT TRENDS IN CLERICAL EMPLOYMENT TABLE 3 Employment by Industry, 1982 Percent of Total Percent of All Clerical Industry Percent of Workers in Clerical Workers Employment Total Industry in a Employment Employed Industry "thousands) Employment Clerical Jobs- tthoUBandS) in Industry Agriculture 3,401 3.4 2.4 83 0.4 Mining 1,028 1.0 12.5 128 0.7 Construction 5,756 5.8 7.8 451 2.4 Durables 11,968 12.0 12.6 1,513 8.2 Nondurables 8,318 8.4 12.9 1,074 5.8 Utilities 6,552 6.6 22.3 1,463 7.9 Wholesale trade 4,120 4.1 20.5 844 4.6 Retail trade 16,638 16.7 17.1 2,840 15.4 Finance 6,270 6.3 43.9 2,750 14.9 Services 30,259 30.4 18.1 5,473 29.7 Public administration 5,218 5.2 35.0 1,827 9.9 Total 99,528 100.0 18.5 18,466 100.0 aThis percentage is also known as the clerical staffing ratio. SOURCE: Calculated by the authors from CPS data. CLERICAL EMPLOYMENT BY INDUSTRY The relative importance of the various industries in the na- tional employment picture is presented in Table 3, based on CPS data for 1982. By far the most important of the individual one- digit industries is the service sector. It accounts for a little over 30 percent of all employment, almost double the size of the next biggest sector, retail trade. Table 3 also shows the clerical staffing ratio, which indicates how important the clerical jobs are in each of these industries. A staffing ratio for an occupation measures the relative importance of that occupation compared to all others in an industry. The ratio is obtained by dividing employment in that occupation in an industry by total industry employment. (The staffing ratios of all occupations within an industry must sum to one; shown as percents, they must sum to 100.) According to Table 3, the finance industry shows the highest percentage of clerical workers; nearly 45 percent of all employees in this industry are clerical workers. While not shown in the table, twice as many clerical workers work in finance as any other type of worker. Public administration is also a heavy employer of clerical workers; 35 percent of all jobs in this industry are clerical. It is

H. ALLAN HUNT AND TIMOTHYL. HUNT 239 followed by utilities and wholesale trade, which also utilize heavy proportions (more than 20 percent) of clerical workers to produce their output. However, in neither of these industries are clerical workers as dominant as in finance or public administration. The service industry and retail trade both show between 15 and 20 per- cent of their total employment in clerical occupations (although their needs for other occupations do not Took similar at all). The durable and nondurable manufacturing industries show between 10 and 15 percent of their total employment in clerical occupations (both are the home base of operatives, of course). Last is the con- struction industry, which employs relatively few clerical workers (but is the dominant user of skilled craft workers in the economy). Clearly, different industries use very different mixes of occupa- tions to produce their final output. The occupational staffing ra- tios are relatively unique to each type of production. It is this vari- ation in the staffing ratios between industries that makes trends in industry employment an important influence on the distribution of occupations throughout the economy. The absolute number of clerical jobs in each of the major industries is also presented in Table 3. About 5.5 million clerical workers can be found in the service industry. Just under 3 million clerical jobs are located in each of two sectors, retail trade and finance. These three sectors combined—services, ret se} trade, and finance account for more than 11 million clerical jobs, almost 60 percent of total clerical employment. Clerical workers may be dispersed broadly throughout the national economy, but these three sectors are especially important to total clerical employment. The 20 most important industrial employers of clerical workers in 1982 are presented in Table 4. These data are from the Bureau of Labor Statistics (BI,S) Occupational Employment Statistics (OES) program, which currently collects the most detailed data on occupational employment by industry.7 The entries in the table are ranked by the number of clerical employees in each industry. The clerical staffing ratios and total employment in each industry are also included. Finally, the percent of total clerical jobs accounted 7 These data provide the historical basis for the staffing ratios in the Bureau of Labor Statistics (BLS) industry-occupation matrix, which is sub- sequently used as the starting point for the BLS projections of future occupational employment. See BLS (1981~.

240 RECENT TRENDS IN CLERICAL EMPLOYMENT TABLE 4 Clerical Employment by Industry, 1982 Total Clerical Cumulative Industry Clerical Staffing Percentage of Percentage of Employment Employment Ratio Total Clerical Total Clerical Industry (thousands) (thousands) (percent) Employment Employment State and local government and educational services 13,068 2,512 19.2 13.4 13.4 Miscellaneous retail trade 10,476 2,496 23.8 13.3 26.8 Wholesale trade 5,294 1,531 28.9 8.2 34.9 Banking 1,650 1,180 71.5 6.3 41.2 Federal government 2,739 1,138 41.5 6.1 47.3 Insurance 1,700 911 53.6 4.9 52.2 Miscellaneous business services 3,139 896 28.5 4.8 57.0 Hospitals 4,166 666 16.0 3.6 60.5 Social services, museums, and membership organizations 2,755 587 21.3 3.1 63.7 Credit agencies, security and commodity brokers 1,015 577 56.9 3.1 66.8 Legal and miscellaneous services 1,628 560 34.4 3.0 69.7 Telephone and other communication 1,174 529 45.1 2.8 72.6 Physician and dental offices 1,309 394 30.1 2.1 74.7 Construction 3,913 324 8.S 1.7 76.4 Eating and drinking places 4,781 224 d.7 1.2 77.6 Electric services and gas distribution 792 207 26.2 1.1 78.7 Trucking and warehousing 1,206 199 16.5 1.1 79.8 Miscellaneous printing and publishing 846 192 22.8 1.0 80.8 Real estate 986 188 19.1 1.0 81.8 Miscellaneous personal services 1,219 186 15.3 1.0 82.8 SOURCE: Calculated by the authors, based on data tape from the 1982-1995 OES/BLS occupational employment projections. for by each of the 20 industries as well as the cumulative total is also reported. The top 10 industries in terms of clerical employment account for about two-thirds of all clerical employment. The top 20 in- dustries account for over 80 percent of all clerical jobs. Federal, state, and local government sectors are clearly important to cler- ical employment; jointly they account for more than 3.6 million clerical jobs or almost 20 percent of the total. We can also see the importance of banking and insurance, the two largest sectors within finance in terms of clerical employment. Finally, clerical jobs are important in a variety of service sector industries, from business services to personal services. None of the top 10 clerical employment industries is from the goods-producing sector.

H. ALLAN HUNT AND TIMOTHYL. HUNT 220 200 11 Go 45, 1 80 - X UJ me 1 60 140 120 100 241 Top 10 Industries 0~0 All 105 Industries _~ _ _; f ~ I, . . . 1960 1965 1970 YEAR 1975 1 980 FIGURE 3 Total employment in the top 10 clerical-employing industries, 1958-1984. Source: Hunt and Hunt (1986~; [based on 1958-1984 BLS data. Since industry employment is so crucial to occupational em- ployment levels, the trends over the last 27 years in industry employment are presented in Figure 3 and Table 5. Figure 3 ag- gregates the employment in the top 10 industries, while Table 5 presents the employment trends for each of the 10 industries. The numbers are reported in index number form to make it easier to compare the growth trends. The average growth in employment for all industries is also reported to facilitate comparisons between the particular industry and the average for all industries. Figure 3 demonstrates a number of important features of the top 10 clerical employment industries. First, these industries have been much less susceptible to the vagaries of the business cycle than all industries. The growth rate of the sum of these 10 sectors has remained positive through two of the three recessions during the period. It was only in 1982, during the worst recession since World War IT, that the composite growth rate of these 10 sectors turned negative—and then, barely so. Second, the average growth rate of these 10 industries has clearly outdistanced the all-industry average for the entire 27-year period, but this is mostly because

242 A :^ o - v o so U3 ~4 C. ._ DO o C) V] o Be: o :^ o - - ~s to ~ Go hi me En ~ 0 ~ 0 _ ~ .O _ O I 0 o. ~ 0 ED _ ~ 0 0 ~ a) _ {4 . ·C} ~ ~ ~ 0 4, C ~ ~ U) ~ a 0 ~ . - ~ V v ~ ~ ~ m 1 — 0 O . :~: ~ ~ ~ 0 _ ~ au 0 0 0 ~ ·— (~~ ~ 4, ~ amen 0 _ > ~ q,, 0 4, REV ~ Hi: m ~ ._ ~ 4,, O ~ TIC A ~ 0 ~ 0 — ~ U ~ ~= 0 0 ·— TV 4 ~ ~15 a ~ ~ 7 ~ ~ ~ q, O ~ us Us ~ GO ~ OF ~ en us ~ ~ 0 ~ ~ ~ a) en Go us a) 0 ~ ~ 0 ~ O 0 0 0 0 0 ~ _ ~ e4 C`4 e~ e~ o: o~ co ~ o' ~F ~ u~ u, a, cc u, co CD —4 _~ _ ~ _I _~ _I _~ _~ _1 _~ _~ ~ _~ _~ _~ _1 ~ _I ~ —4 ~ ~ ~ —4 _ 0 ~ ~ ~ a) co 0 c~ F O ~P ao o~ ~ ~ oo o~ 0 C53 co 0 c~ e~ ~ e~ O 0 0 0 ~ ~ e~ e4 o~ ~ ~ ~ u~ ~o co c9 r~ ~ oo ~ C53 0 ~ ~ ~ ~ c ~ _~ ~ _~ ~ _1 _~ _~ _ ~ ~ ~ ~ _I _~ e~ ~ ~ e4 c~ c~ O 00 C~ OD ~ 0O ~F a) 0 co oo ~ a, co ~ ~ ~ 0 e~ ~ o~ a, 0 0 0 co 0 ~ c~ e~ ~ o~ o~ o: ~ ~ ~ ~ u~ ~ u~ ~o tD CO ~ ~ r~ oo ~ ~ ~ C~ a _~ _I _1 ~ _1 _1 ~ _~ ~ ~ ~ ~ _4 _1 _~ _1 ~ _~ ~ _~ _~ ~ _ O C53 cO ~F a~ e~ ~ C~) ~ ~ ~ ~ ~ e~ ~ e~ 0 e~ a~ co oo ~ ~ ~ o~ 0 0 ~ e~ c~ o~ a~ ~ D ~ 00 00 o0 ~ O O O ~ e~ o~ u~ CD ~ ~ —~ ~ ~ ~ ~ ~ ~ ~ ~ e~ ~ c~ c~ ~ c~ ~4 oo oo 0 co oo 0 c~ ~ 03 ~ co ~ c~ ~ u~ co oo co ~ 0 0 ~ 0 ~ co 0 C4 ~ a O O ~ e~ c4 ~ ~P ~ hD ~ 00 ~ O ~ ~ e~ a~ u~ CC~ ~ ao a~ 0 c~ o~ o: c~4 ~ ~ _1 ~ ~ _1 _I _ _I ~ ~ ~ C~ c~ c~ ~q co a~ co o~ o: O O CD a, u: (g ~ o, e~ ~ ~ oo c~ ~ e~ oo <9 co o, cS, C53 c~ ~ O — O O ~ ~ C~ C1) ~ U: ~ ~ ~ e~ ~ co ~ oo 0 c~ co ut oO C~ ~ O ~ u~ oo U~ _ ~ _4 ~ c~ c~ co co co o: oo ~ ~ ~ ut u: ~ co O ~ CO ~ CD ~ ~ C`4 ~ a~ C`4 u~ ~ ~ o: u~ ~ O e~ oo ~ O ~F ~ ao 0 0 0 0 0 0 ~ ~ ~ ~ c~ eo co ca co ~ ~F ~F u~ c9 c9 CD (D CS) ~ ~ ~ ~ ~ - 4 ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ - 4 0 e~ ~ ~P ~ oo ~ ~ ~ ~ u, co ~ oo co ~ ~ ut u, ~ C~ ~ ~ ~ u, C~ 0 0 0 0 0 0 0 0 ~ e4 c~ c~ c~ e4 c~ ~ c~ e~ co e~ 4 c ~ ~ _1 _~ _I _~ ~ ~ _~ ~ ~ ~ ~ _~ ~ _1 ~ ~ _~ _~ ~ _~ _1 ~ ~ ~ _ 0 ~ ~ e~ co 0 ~ os ~ ~ 00 O O O ~ ~ e~ o: ~ ~ ~ _1 ~ ~ ~ ~ _I ~ ~ _' ~ ~ ~ (D e~ O ~ a~— ~F ~ oa oo ~ O O ~ c~ 0O ~ ~ ~ C9 c~ e~ ~ e~ o co u~ u~ ~ ~ C~ {D ~ ~ ~ ~ ~ ~F a~ ~ ~ oo o~ oo ~ ~ ~ 0 ~ CO u 0 0 0 0 0 0 ~ ~ e~ O4 ~ co co o~ a~ ~ ~P ~F Ut ~0 C~~ ~ ~ 00 _~ _~ ~ ~ _~ ~ r4 _~ _~ _1 ~ ~ ~ ~ ~ ~ _~ _ _I ~ ~ _~ ~ —4 _~ —4 _ 0 a~ ~ r~ ~ e~ co ~ ~ oo ~ ~o ~ ~ C5~ ~ oo co oo u~ oo ~ oo 0 0 a} 0 0 0 ~ ~ e~ co co o: ~ ~ ~P ~ u~ u~ co co ~ ~ co c9 ~ ~ _1 ~ ~ _~ ~ ~ _~ _4 _4 ~ _~ ~ ~ ~ _~ _I ~ _~ _1 o ~ oo ~ 'o O O O ~ ~ e~ c~ - l ~ ~ ~ ~ ~ ~ oo c~ 0 ~ e~ c~ ~ u' IO CD (D C9 <9 CO a) ~ ~ ~ ~ <5) ~ ~ - 4 ~ ~ ~4 ~ - ~ 'o ~ ~ ~ ~ o a, ~ ~ co ~ ~4 u' O ~ ~ (9 ;) ~ ~ - 4 ~ - i - ~ ~ C9 0O ~ Ut O oO c~ cO ~ u~ ~ ~ co ;) ~ ~ ~ - ~ ~ ~ - ~ ~ o~ ~ u~ e~ e~ co ~ oo CO a~ oo oo c~ ~ e~ C~ O ~ e~ o' ~ co ~ oo oo a, C53~=C53 ~ ~ ~4 ~4 ~4 0 . - 0 . - o l . - u) m ~C o o 0 ~5 0 o ~C a, 43 ~C O o 0 ~ a5 C74 C) ~ — oo V U~ . . o z . . v o U)

H. ALLAN HUNT AND TIMOTHYL. HUNT 243 employment in these sectors does not ordinarily retreat during recessionary periods, but continues to expand. As Table 5 shows, growth patterns among the top 10 indus- tries have been diverse. Their average growth over the 27-year period was 122 percent; all but one grew faster than the economy- wide rate of 67 percent. The most. robust growth has clearly oc- curred in miscellaneous business services; hospitals; banking; and credit agencies and security and commodity brokers. The growth in employment in miscellaneous business services is particularly striking, with more than six times as many workers in this sector in 1984 as there were in 1958. This sector provides myriad ser- vices to business firms, from accounting to customized computer software to consulting advice. Also striking is the growth rate of employment in hospitals; employment tripled over the period 1958 to 1984. Some of the causes of this growth, such as the aging of the population and the increasing availability of medical insurance for retirees and the indigent through government programs, are well known. In any event, the growth of this sector has not been affected by business cycles and may have even accelerated during the last recessionary period. Surprisingly, however, hospital em- ployment growth slowed in 1983 and actually turned negative in 1984. The recent emphasis on cost containment and the shortening of hospital stays may be having an impact on employment. It is also clear that the finance sector especially banking; credit agencies and security and commodity brokers; and insur- ancc contributed significantly to clerical job growth during these years. All three of these sectors have staffing ratios for clerical workers in excess of 50 percent, the highest of all industries. In- surance employment grew at about the economywide average until about 1974, then began to accelerate, and outdistanced the na- tional economy in job growth thereafter (except for 1984~. The growth of employment in banking was consistently higher than that for insurance, nearly tripling from 1958 to 1984. The slowest growth among the 10 industries with large clerical employment occurred in the federal government. The federal gov- ernment has not been a source of significant employment growth in the last 20 years; its growth is well below average. State and local government employment, while generally above average in growth, actually declined absolutely during the 198~1982 reces- sion. By the end of 1984, employment had still not exceeded its

244 RECENT TRENDS IN CLERICAL EMPLOYMENT peak employment level, achieved in 1980. This is significant be- cause it is the first such decline and sluggish recovery in recent history for the largest single employer of clerical workers among the 105 industries in this analysis. Of course, the gnawing question is: Will these industries show rapid employment growth in the future? That question cannot be answered at this point. However, the nation is still experiencing a long-run shift from a good~producing economy to a service-producing economy. Historically, clerical workers have benefited from this shift since service industries employ much higher proportions of clerical workers. Thus, even if staffing ratios begin to fall for clerical workers (because of office automation or other factors), it is still possible for them to grow at or above the average rate for all jobs because they are concentrated in the non-goods-producing sectors. Clerical workers have so far had a fortunate industry mix in their employment pattern. TECHNOLOGICAL CHANGE AND CLERICAL EMPLOYMENT GROWTH Because one of the goals of office automation is to improve labor productivity, actual gains in labor productivity may be the best measure of the degree to which this goal is realized. The major problems in attempting to estimate the true gains from office au- tomation are twofold. First, it is impossible to glean from current data any information whatsoever about the relative importance of office automation spending by firm or industry. Investment data are subdivided only into the two broad subcomponents of machin- ery and equipment and structures. Second, complete data about clerical jobs are not available over time, and no information links workers to the use of equipment. Thus, it is impossible to estimate the productivity gains specifically attributable to clerical workers utilizing various types of electronic office technology.8 One simple approach to examining the productivity gains from office automation is to study those sectors that are significant em- ployers of clerical workers and that are also believed to be leaders in office automation. The broad industrial sector of finance and insurance is the recognized leader in the field of office automation, data. ~ See Hunt and Hunt (1985) for a more detailed discussion of the available

H. ALLAN HUNT AND TIMOTHYL. HUNT 245 and more than one-half of the workers in this sector are clerical workers. Of the 105 detailed industries analyzed in this paper, the finance and insurance sector is composed of three of these in- dustries: banking, insurance, and credit agencies and security and commodity brokers. Thus, if office automation improves produc- tivity, these three industries are logical candidates to demonstrate the effects of such gains. Figure 4 reports the productivity gains for banking, insurance, and credit agencies and security and commodity brokers for the period 1958-1983.9 The data are reported in index number form to better depict the percentage changes in productivity from year to year. The productivity trend for all private nonfarm employment is reported as well to facilitate a comparison of these industries with the aggregate. Surprisingly, Figure 4 reveals no discernible trend that can be attributed to office automation. The productiv- ity gains in banking, insurance, and credit agencies and security and commodity brokers have all tended historically to lag behind the average for private nonfarm employment. In fact, productivity for credit agencies and security and commodity brokers was very slightly Tower in 1983 than in 1958, and productivity deteriorated absolutely in insurance after 1977. Since 1981, banking produc- tivity has improved relative to all private nonfarm productivity, but the improvement is hardly revolutionary, especially given that banking productivity declined from 1979 to 1981. It should be emphasized that these are not measures of the productivity gains for clerical workers nor can these gains be at- tributed to office automation. They are industry-wide measures for output gains due to all improvements across all employees. All that can be fairly concluded is that there is nothing in the aggre- gate industry data to support the notion that office automation has engendered significant overall productivity gains in these three industries. Have these industries been investing in office automation? Al- though office automation expenditures are not reported separately, Figure 5 reports, in index number form, real investment spending per employee for the financial industries. The data are for finance 9 The gross output in constant dollar terms and employment measures are defined in Bureau of Labor Statistics, 1979; the data utilized here are from an unpublished update (April 1985~.

246 RECENT TRENDS IN CLERICAL EMPLOYMENT 150 _ 140 _ 130 120 CO A) 0 - A Ad 110 100 90 .\ 80 70 60 _, _— , ~ , . <~-:r~~ t!1 _.1 ~ TOTAL '' / INSURi RANKS \ \ {.~ \ \ /! ,~ - ~ — CREDIT AGEN/ \' 50 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 -at 1960 1965 1970 YEAR 1975 1980 FIGURE 4 Productivity gains in terms of output per hour for banking, insurance, and credit agencies, 1958-1983. Source: Based on unpublished data provided by BLS, April 8, 1985. and insurance firms, the smallest level of industrial detail avail- able for such investment data (Seskin and Sullivan, 1985~. Once again, the totals for private nonfarm employment are also shown for comparison. In contrast to the lack of any "takeoff" evident in the productivity data for finance and insurance, the investment data in Figure 5 clearly indicate much higher than average invest- ment trends in finance and insurance after 1966-1967. In fact, investment per employee virtually exploded, growing a little more than five times the average for all private nonfarm employment after 1966-1967.~° Although there is no doubt that the finance and insurance industries are investing heavily in new capital equipment, it is less certain that they are investing in office automation. What can be concluded is that the dramatic growth in investment in finance and insurance has not resulted in measurable labor productivity gains to date. One explanation of this puzzling situation is that the aggregate industry data may be flawed, but given such pronounced \ id Historically, absolute investment per employee in finance and insurance has tended to be much less than the average for all nonfarm private industries.

H. ALLANHUNT AND TIMOTHYL. HUNT 550 500 - g 450 11 co - X he - z oh 400 300 250 200 150 100 50 247 / Finance ~ insurance `~' / 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 ___ ~ ~— ~ - `_—J - Tot Pnvate Nonfann 1 1 1 1 1 1 1 1 1 1 1960 1965 1970 1975 1980 YEAR FIGURE 5 Real investment spending per employee for finance and insur- ance firms, 1958-1983. Source: Based on unpublished data provided by BLS, April 8, 1985. investment growth in finance and insurance over such a Tong period of time, it is difficult to believe that the productivity data are so flawed that they would not register at least the beginning of a new trend, if one were actually occurring.' ~ The lack of productivity increase shown in the aggregate data is borne out to some extent by other sources. Although formal case studies of the economic impacts of office automation are gen- erally lacking, there is fragmentary information that casts doubt on the most wildly optimistic productivity claims of advocates of it There are serious concerns about the quality of the aggregate data (see Hunt and Hunt, 1985~. BLS has constructed special productivity indexes in banking, measuring output as services rendered and labor input as actual hours worked. The overall result, 1967-1980, is that productivity growth in banking remains very slightly below the national average for all nonfarm businesses (Brand and Duke, 1982~. i2 Salerno (1985) and Strassman (1985) concur in the paucity of studies. For a review of a considerable literature about the sociological impacts of office automation, see Attewell and Rule (1984~. From the economists' perspective, this literature is lacking in a systematic treatment of output, capital input, prices of outputs and inputs, and other economic variables.

248 RECENT TRENDS IN CLERICAL EMPLOYMENT office automation. First, a number of recent books (Bailey et al., 1985; Diebold, 1985; and Katzan, 1982) designed to be guides for managers interested in improving productivity through office automation contain surprisingly few references to the actual ex- periences of firms or to the productivity gains that managers can reasonably hope to achieve with office automation. For instance, Katzan includes an entire chapter on word processing, but provides no hint about the likely potential productivity gains. Second, Paul Strassman, an executive and office automation specialist with Xerox, in a recent assessment (1985) of the technol- ogy with which he has been associated for over 20 years, eschews the current focus on hardware. Although he is optimistic about the potential productivity gains from computers and information technology generally, he stresses that the hardware is less relevant than the people using it. In fact, he suggests (1985:159) that the growth rates in office automation investment of the early 1980s are unsustainable unless that investment produces demonstrable returns. Strassman finds little evidence of such returns currently: The preliminary findings of my research raise doubts about the as- sumptions which managements in the businesses I have sampled so far must have made when they increased their computer-technology budgets in pursuit of improved productivity. Strassman thinks the payoff will come when management focuses on strategic goals and the people who will accomplish those goals, rather than on the methods to achieve those goals. Third, a recent study points to the hidden costs of adop- tion of information technology, costs so large that productivity gains do not occur. Researchers King and Kraemer (1981) con- tend that, while the cost of hardware is falling, the total cost of electronic computing is rising rapidly (1981:102~: "Software procurement, software maintenance, and data management and computing management, are all becoming increasingly expensive." Furthermore, because many of the non-hardware costs tend to be hidden from normal accounting procedures used to justify imple- mentation, these costs do not necessarily affect the implementation decision itself- although they would adversely impact the firm's actual operating results. New positions and even departments are springing up in firms to evaluate software, perform system main- tenance, coordinate among different users, etc. Firms often find that "off-the-shelf" software is unsatisfactory for their computing

H. ALLAN HUNT AND TIMOTHYL. HUNT 249 needs, necessitating significant investment in software program- ming. As electronic computing becomes more widespread in firms through the adoption of personal computers, King and Kraemer (1981:101) believe it will become increasingly difficult for man- agement to track these costs. Users at all levels dedicate some portion of their time to routine maintenance tasks. Worse, some may even develop a personal interest in the technology that diverts them from other work. According to King and Kraemer manage- ment seldom knows the ongoing costs of training, normal system maintenance, or unplanned downtime that are in fact incurred because of the firm's utilization of information technologies. They cite a variety of other studies and fragmentary data, which appear to indicate that the annual costs for system maintenance run at least 20 percent of the cost of the development of the system it- self, and may even be much higher. They think the costs due to breakdowns may be particularly significant in highly integrated systems. According to King and Kraemer (1981:107), when systems become integrated and units become more interde- pendent in a real-time sense, problems in one system or unit can literally stop progress in others simply by disruption of the process of interaction. As integration increases, interdependency increases. Together, these two phenomena result in increased costs. Fourth, International Data Corporation (IDC), which may be the information industry's largest market research and consulting firm, has repeatedly stressed that the labor productivity gains from office automation fall far short of justifying the purchase of the equipment. According to IDC (1984, 1983, 1982), the direct labor savings attributable to an office automation project over a 5-year period usually amount to no more than one-half the cost of implementation of the system. IDC states that this rule of thumb does not include the training costs of implementing office automation. However, it also does not include any improvements in the quality of the output of offices. IDC concludes that the quality improvements justify the adoption of office automation. Even some computer vendors are not emphasizing cost savings per se in their attempts to sell office automation. A booklet about cost justification that Wang Laboratories makes available to potential customers stresses the complexity of the cost justification process for office automation. One of the premises of the booklet (Wang Laboratories, Inc., 1985:3) is that information technology systems

250 RECENT TRENDS IN CLERICAL EMPLOYMENT are fundamentally "different from other kinds of capital equipment investments and should be treated differently with regard to cost justification." In the six examples of firms that have successfully costjustified their systems included in the booklet, the emphasis in all cases is on improvements in quality rather than on direct cost savings. In general, these sources indicate little improvement in pro- ductivity overall (Katzan, Strassman), higher costs of installation and operation (King and Kraemer), or quality rather than quan- tity improvements (IDC and Wang). Still other reasons why office automation may not have a significant impact on productivity include its limited spread to date, weaknesses in its technical ca- pabilities, and its role in making production more information · . , Intensive. First, one of the most obvious reasons that office automation may not have created measurable productivity gains industry-wide is that the diffusion of the technology may not have proceeded nearly as far as implied by the popular media. According to a national random survey by Honeywell Inc. (1983) of 1,264 general office secretaries employed in information-intensive establishments with 100 or more employees, office automation equipment is not yet in widespread use in most offices. Fewer than one-half of the secretaries reported having access to an electronic memory typewriter/word processor/personal computer in the general office area in which they work; less than one-fourth possessed any of this equipment at their individual workstation (1983:lIl-S). Almost none of the secretaries reported having direct access to electronic mail, computerized scheduling, or computerized filing, while about 15 percent said that such equipment was located somewhere in the office area (1983:~-5~. The limited use of automation found is especially surprising because the survey was limited to large establishments in information-intensive industries, exactly where one avouch expect to find once automation in place. Second, once automation may not be producing the promised productivity gains, because technical constraints inherent in the current technology reduce its electiveness. Word processors may not be that much different from their earlier nonelectronic prede- cessors; both are tools used to accomplish standard office func- tions. More advanced devices and appropriate methods may be necessary to bring about more significant changes in production methods. There also appear to be severe hardware and software

H. ALLAN HUNT AND TIMOTHYL. HUNT 251 compatibility problems across computer systems, which limit ap- plications. Electronic mail, for example, is currently limited by system incompatibilities. Even when direct communications sys- tems are installed, for example, in the form of a local area network, it will likely still be a relatively primitive system. It may not be possible to use the local area network to access the large data bases on the firm's mainframe computer. It may not be possible to transmit graphics via the network. While it may be possible to access a user who is not on the local area network, the pro- cedure may be too tedious and cumbersome to be truly useful in the transmission of serious business messages. In short, the allow- able traffic on the local area network may be very structured and severely limited by the available hardware and software. The office with instantaneous access to any data base around the world and total communications flexibility still lies somewhere in the future. Many writers have compared this stage in the evolution of computers with that of automobiles in the 1920s. The technology for automobiles had already been firmly established by that date. What was needed, however, were the highways which would make it possible to effectively utilize the technology. According to this analogy, computers now need "pathways" to effectively communi- cate across dissimilar hardware and software systems before it will be possible to realize their full potential. The third reason that office automation may not be having an impact on productivity is that much of what we term office automation is not being purchased as labor-saving process tech- nology at all. A deepening of capital may be occurring as products and services become more information intensive and new products and services are developed. Office automation may represent sim- ply additional capital support for office workers rather than capital substitution for labor. In part, this is a restatement of the quality argument pre- sented earlier. Office automation may be seen as a means to en- sure a firm's competitive survival by providing improved access to data and feedback. Aside from the question of whether electronic office technology saves labor time directly, there is no doubt that it permits more adequate analytical support for decision making, timely answers to customer inquiries, more rapid tracking of sales data allowing better inventory control, and many other gains in quality. Those who have utilized electronic spreadsheet software know that it results in a whole new world of opportunities for

252 RECENT TRENDS IN CLERICAL EMPLOYMENT tabular and graphical analyses. The availability of personal com- puters taps hidden computing needs that executives always had but that could not be satisfied previously because of insufficient staff or time available to do so on the firm's mainframe computer. The diffusion of the newer and smaller micros is eliminating this roadblock. Managers are simply taking advantage of the lower marginal cost of computing by utilizing it in new and different ways. This result is compounded by the difficulty of adequately measuring output from an office; it is extremely difficult to know how much the new technologies have added to the bottom line of the firm. In particular, changes in quality are especially difficult to measure. In summary, this review of the technological influences on clerical employment has been relatively unsatisfying. There are no general time series data about office automation spending by industry or about the application of devices by individual occu- pations. The analysis of overall productivity gains in finance and insurance did not provide any evidence that office automation is producing significant productivity gains in that sector despite the fact that real investment spending in finance and insurance has increased rapidly since the late 1960s. Similarly, other sources, citing a variety of explanations, support the notion of an apparent lack of productivity gains from office automation to date. Conse- quently, it seems likely that productivity gains are not having a major impact on clerical employment trends. DECOMPOSITION OF CLERICAL EMPLOYMENT CHANGES Employment change in an occupation can be understood as resulting from three factors: (1) general economic growth, (2) differential rates of industry growth (because each industry uses a different mix of occupations), and (3) the relative importance of each occupation within each industry, which can be thought of as the intensity of use of a particular occupation (or the occupational staffing ratio introduced above). The above discussion focused on industry growth rates and the impact of office automation on the intensity of use of clerical-workers. If the net effect of office automation is the displacement of clerical jobs, over time clerical staffing ratios will fall. In the following discussion, an analytical tool is used to summarize the effects of all three factors. Using the mathematical technique of decomposition, occupational

H. ALLAN HUNT AND TIMOTHY L. HUNT 253 employment changes can be artificially separated into their three components. The decomposition methodology provides another, more sys- tematic opportunity to assess the technological influence of office automation on clerical jobs. In general, changing staffing ratios are probably the most visible manifestation of the specific ejects of technological change on occupational employment. For exam- ple, the staffing ratios for computer-related occupations have risen in many industries over time because of the dramatic increases in the use of computers. On the other hand, the staffing ratios for stenographers have been falling over a Tong period of time because of the adoption of dictation equipment, a technological change that reduces the need for stenographers.~3 The mathematical decomposition technique holds some of the factors constant over time while allowing others to vary. For exam- ple, the effects of a change in staffing ratios on clerical employment during a particular time period can be determined by comparing current clerical employment with simulated employment levels ob- tained by holding staffing ratios constant at their original levels but using current industry employment as the multiplier. In other words the simulated clerical employment level uses the "correct" industry levels actual current employment in these industries- but the "wrong" staffing ratios those that existed at the start of the period. Thus, the difference between actual current clerical employment and simulated employment indicates the extent to which the change can be attributed to staffing ratio changes.~4 While one may presume that these changes are due to techno- logical change, the mathematical decomposition of occupational employment growth is not an explanation of cause and effect; many complex economic and noneconomic forces lie behind the numbers revealed by the technique. i3 It should be emphasized that staffing ratios may change for reasons other than the use of innovations, such as organizational change, job title change with no change in job content, or others. Any time an individual occupational staffing ratio changes, all of the remaining staffing ratios in that industry will change as well. (This occurs because the sum of the staking ratios in an industry must equal one.) For example, if a particular industry was very successful in automating production worker jobs, perhaps by using robots, then the relative importance of other jobs such as clericals, professionals, etc. will increase. id A more technical description of the decomposition is provided in Hunt and Hunt (1986:Ch. 4~.

254 RECENT TRENDS IN CLERICAL EMPLOYMENT TABLE 6 Clerical Employment Growth by Industry, 1972-1982 Industry Clerical Emoloyment Changes Change in Change in 1972 1982 Employ- Employ- Employ- Employ- ment 1972- ment 1972- ment ment 1982 1982 t000'n) (000's) (000's) (%) Agriculture 48 83 35 72.9 Mining 59 128 69 116.9 Construction 362 451 89 24.6 Durables 1,352 1,513 161 11.9 Nondurables 1,040 1,074 34 3.3 Utilities 1,307 1,463 156 11.9 Wholesale trade 684 844 160 23.4 Retail trade 2,099 2,840 741 35.3 Finance 2,007 2,750 743 37.0 Services 3,691 5,473 1,782 48.3 Public administration 1,678 1,827 149 8.9 Total 14,326 18,446 4,120 28.8 Decomposition of Clerical Employment Changes, 1972-1982 Absolute Changes (in thousands) Percent of Employment Due to: Changes Due to: Differential Differential Aggregate Rates of Staffing Aggregate Rates of Staffing Economic Industry Ratio Econorn~c Industry Ratio Industry Growth Growth Changes Growth Growth Changes Agriculture 10 -11 36 21.1 -22.9 75.0 Mining 12 29 28 21.1 49.1 47.5 Construction 77 -44 56 21.1 -12.1 15.S Durables 286 -244 119 . 21.1 -18.0 8.8 Nondurables 220 -222 36 21.1 -21.3 3.5 Utilities 276 -23 -97 21.1 -1.8 -7.4 Wholesale trade 145 86 -71 21.1 12.6 -10.4 Retail trade 444 45 252 21.1 2.1 12.0 Finance 424 457 -138 21.1 22.8 -6.9 Services 781 605 396 21.1 16.4 10.7 Public admin. 355 -53 -153 21.1 -3.2 -9.1 Total 3,029 625 466 21.1 4.4 3.2 NOTES: Totals and percentages may not add exactly due to rounding. Percentages based on 1972 employment. SOURCE: Calculated by the authors, based on data from the Current Population Sunrey. The results of the decomposition for the major occupational group of clerical workers are presented in Table 6. One-digit industries and occupations are used in this analysis, based on CPS data, because the sample is far too small to provide both industrial and occupational detail below that level. The time period for the analysis, 1972-1982, is selected because that is the only recent time span for which consistent data are available. From 1972 to 1982

H. ALLAN HUNT AND TIMOTHY L. HUNT 255 the number of clerical jobs increased by just over 4 million for a 28.8 percent gain over 1972 employment levels. During that some time span total employment increased by 21.1 percent. Clerical jobs grew faster than the average for all jobs, which also means that clerical jobs were becoming relatively more important in the national economy. The bottom row in Table 6 indicates that the bulk of all new clerical jobs, a little more than 3 million, were added as a con- sequence of the overall growth of the economy; another 625,000 clerical jobs were added because clerical workers were more preva- lent in industries that were growing faster than the average for all industries (the factor labeled differential rates of industry growth in the table); finally, 466,000 clerical jobs were added because of increasing staffing ratios for clerical jobs; that amounts to 3.3 percent of the 1972 employment level for clerical workers. This does not mean that staffing ratios in all industries were increasing for clerical occupations, but rather that the net effect across all industries was positive. Neither changing staffing ratios nor differ- ential rates of industry growth, however, were major contributors to clerical employment growth in the 10 years from 1972 to 1982, although both factors were modestly positive during the period. An examination of the data for each industry shows that staffing ratios for clerical jobs are actually falling in a number of industries. Most interesting are the results for the finance sector, as noted above probably the biggest user of office automation to date. The finance sector has been a rapidly growing sector as indicated by the 37 percent overall growth rate of clerical jobs in that sector versus the 28.8 percent growth rate for all clerical jobs. Thus, the elects of falling staffing ratios, which would have reduced jobs in this sector by 6.9 percent from 1972 employment levels, were more than recovered by the fast growth of the industry itself. However, if the industry had not expanded so rapidly, there would have been actual reductions in employment of clerical workers in the finance sector. Staffing ratios for clerical jobs have also been falling in three other important industries utilities, wholesale trade, and public administration. The decline in public administration may not be due primarily to office automation. Except in the postal service, which has automated the mail-sorting operation, government has not been in the forefront in adopting automation. Government was one of the sIowest-growing sectors during this time period. Perhaps

256 RECENT TRENDS IN CLERICAL EMPLOYMENT government administrators, when faced with tight budgets and rising demands for services, economized more on clerical jobs than other positions. It is not easy to provide an adequate explanation of the fall In staffing ratios for public administration or the other industries. It is possible that they are linked to office automation. Clearly, more study of these trends is called for. It is possible that office automation is raising the productivity of clerical workers, contributing to the falling staffing ratios in those sectors, and thereby negatively impacting clerical jobs. It is puzzling that the aggregate productivity data for finance and insurance showed below-average productivity growth for the sec- tor as a whole, yet the decomposition analysis showed declining staging ratios for clerical jobs within finance and insurance. Many unanswered questions remain about the causes of employment trends for clericals in sectors such as finance and insurance. The changing composition of industries has tended to favor clerical jobs, but the influence of industry mix was only moderately positive during the 1970s, and some sectors that are heavy em- ployers of clericals have recently experienced much slower growth or even absolute declines in total employment. This is particularly true for hospitals, and state and local government, the latter of which is the largest single employer of clericab. Even though other sectors, notably services, will likely continue to be fast growing, there is no strong reason to think that industry mix will play a more significant role in the future employment outlook for clerical workers; it may well play a less-significant role. Most importantly, this analysis reveals that economic growth is by far the most important factor in determining clerical employ- ment. Economic growth accounted for three-fourths of the clerical jobs created. Yet this factor may be in the process of becoming less favorable to clerical employment. In the past, clerical employment continued to grow when the economy as a whole experienced a re- cession. In the last recession, however, clerical jobs failed to grow. They may be becoming more like other jobs in their sensitivity to general economic conditions. CONCLUSION Clerical employment has grown rapidly in the last 40 years or so, but many factors suggest that the growth of clerical jobs

H. ALLAN HUNT AND TIMOTHY L. HUNT 257 has slowed significantly. Clerical workers have had a favorable in- dustry mix in their employment pattern, benefiting from the shift toward finance and other service-related industries because those industries employ much higher proportions of clerical workers. Furthermore, the relative importance of clerical jobs has tended -to rise within industries. Thus, in the past all the factors have tended to be positive and the result has been spectacular growth in clerical employment. Given the rapid growth in clerical jobs over the last 30 years or so, it appears reasonable to conclude that many goods and services have been becoming more and more in- formation intensive per unit of output over time. This has tended to boost clerical employment. However, it is clear that the rate of growth of clerical jobs slowed during the last decade. Clericals did not benefit from the last recession as they have in earlier recessions, nor are some of the sectors that are important employers of clericals growing as fast as they once were. Finally, although office automation may not be producing a revolution, it should at least contribute to the slowing of employment growth in these occupations in the future. Changing staffing ratios, probably the most visible manifesta- tion of the specific effects of technological change on occupational employment, had a moderately positive effect on the employment growth of clerical workers from 1972 to 1982, creating about 450,000 new clerical jobs (compared to about 3 million created by aggregate economic growth and 600,000 by the concentration of clerical workers in rapidly growing industries). Although the net effect of changing staffing ratios on clerical employment was modestly positive across all industries, there were a few sectors, notably finance, where the effect was negative. This is taken as evidence of the adverse impact of technological change on cleri- cal employment. Investment in this sector has been dramatically higher than the historical average for that sector for the last 15 years, but measured industry-wide productivity gains in finance and insurance do not support the thesis that office automation is having a significant impact. This lack of measured productivity results remains a puzzle. Nevertheless, future employment may well be impacted by the capital buildup in this and other sectors that have traditionally been large employers of clerical workers.

258 RECENT TRENDSINCLERICALEMPLOYMENT A LOOK TOWARD THE FUTURE It is, of course, risky to attempt to predict what might happen to clerical employment in the future, but the changes that lie behind the trends noted above provide some clues, as do recent forecasts made by the Bureau of Labor Statistics (BES) and other researchers. These will not be fully reviewed here, but their major conclusions are summarized as part of our assessment of the likely future for clerical employment. (For a more complete discussion see Hunt and Hunt, 1986.) The BES occupational projections represent the most impor- tant effort of the federal government to anticipate future needs for specific occupations. The BES methodology is based on a mod- eling framework that accounts for many economic variables. The resulting occupational projections are not necessarily superior to others, but they do have the advantage of being produced in a comprehensive and reasonably consistent manner. Other forecasts of clerical employment growth are not nearly as comprehensive as that of the BES. For example, Leontief and Duchin (1986) of New York University analyze the impacts of computer automation on employment from 1963 to 2000. Their research is limited to cer- tain specified computer technologies and does not consider other productivity-enhancing technologies or any other source of produc- tivity growth. Roessner et al. (1985) of the Georgia Institute of Technology, examines clerical jobs in two industries, banking and insurance. Finally, Drennan (1983) of Columbia University looks at clerical employment in six industries. His projection methodol- ogy utilizes extrapolation of historical trends after accounting for the effects of the 1980-1982 recession. The BES anticipates near average growth for clerical jobs. Between 1982 and 1995, BES anticipates an average growth in employment of 28.1 percent. Growth in employment of clerical occupations is anticipated to be 26.5 percent. The effect of differ- ential rates of industry growth is expected to be slightly positive at 1.6 percent (smaller than their contribution between 1972 and 1982~. The effect of changing staffing ratios is expected to be negative at minus 3.1 percent. In fact, all the existing forecasts of employment in clerical occupations are unanimous in predicting that staffing ratios for clerical jobs will decline in the years ahead, presumably because of office automation. The fall in staffing ratios anticipated by BI`S is modest compared with other predictions.

H. ALLAN HUNT AND TIMOTHYL. HUNT 259 Still, it is significant that the only turnaround from historical trends anticipated by BES at the major occupational group level due to changing staffing ratios is that for clerical workers. At least through 1982, the decomposition analysis discussed earlier showed that the staffing ratios for clerical jobs were rising, whereas the BES forecast (base year 1982) and other forecasts expect that this trend wiB be reversed in the years ahead. Growth rates forecast by BES for 95 specific clerical occu- pations range from a positive 76.1 percent to rn~nus 20 percent. The range in the portion due to clerical staffing ratio changes is from plus 38.4 percent to minus 55.6 percent. Clearly, though the effect of BES's staffing ratio for clerical workers is negative overall, BES expects many positive staffing effects for clerical workers as well. The fastest-growing clerical jobs are expecter] to be com- puter operators, claims adjusters, insurance checkers, peripheral EDP equipment operators, telephone ad takers, claims clerks, and credit authorizers. All are expected to have staffing ratio im- pacts equivalent to increases in employment levels of 20 percent or more. Besides the obvious technological impacts of comput- ers on this list, it may be important to note that many of these occupations require the worker to interact in some way with the customer who is being served. That may provide a clue as to why BES thinks secretaries will not decline in importance, or perhaps why cashiers are the 10th fastest-growing occupation, with a 48.2 percent growth rate. Again a world of both high tech and high touch is anticipated. Roessner et al., Drennan, and Leontief and Duchin all con- clude that office automation will have a much greater negative im- pact on clerical jobs than the BES predicts, with Leontief-Duchin and Roessner et al. predicting absolute declines in clerical em- ployment within the next decade. Roessner et al. is particularly pointed about his concerns regarding the BES methodology and forecasts, while Drennan's projections appear to be nearer the BES position. We find the studies of Leontief and Duchin and Roessner et al. to be seriously flawed for serving policy needs; we think that Leontie£Duchin and Roessner et al. are unduly pessimistic about the outlook for clerical jobs. There are a variety of reasons that support our contention. First, although Leontief-Duchin use the BES forecast for demand for goods and services, they predict revolutionary change in the

260 RECENT TRENDS IN CLERICAL EMPLOYMENT process technology by which those goods and services will be pro- duced. The revolution in office automation is assumed to leave the demand side of the marketplace unchanged. But that is not the way a complex, dynamic market economy operates. If office automation were adopted rapidly, it would change the relative costs of production for those goods and services that are inten- sive users of office automation. Those lower production costs will generate lower prices. Since office work is concentrated in the ser- vice sector, where demand growth has been above average, there is every reason to think that both the lower prices and income growth over time will generate additional demand. This scenario is even more plausible when one realizes that the product markets themselves are not static. So the new electronic office technolo- gies may provide the impetus for the development of entirely new goods and services. Industry interrelationships may change or scale economies may be so significant that they fuel the devel- opment of a mass market that heretofore was undreamed of. In our opinion it is inappropriate to fix demand or the growth of demand and then assume a revolutionary change on the supply side of the market. Obviously, such a partial analysis can create false impressions about the true impacts of office automation. Second, it appears that none of these other studies account for the tendency of output to become more information intensive over time. Yet this is a process that has been occurring for some time. The production recipes for many different goods and services today require more information processing than yesterday. This is not simply a function of the changing composition of demand but relates to the ingredients for a standard unit of output. To the extent that this trend continues in the future, it implies that office automation may have less impact on clerical employment levels than anticipated by some researchers. Third, it should be mentioned that office automation is likely to lower the marginal cost of some new types of work so much that the required labor needed to produce that work rises by more than the labor-saving impact of the new techniques themselves. The common example is redrafts of documents with word processing. The probability that this will occur may be enhanced by our inability to measure output from offices in the first place. This type of new work or rework is explicitly rejected by I`eontief- Duchin, and perhaps implicitly by Roessner et al.

H. ALLAN HUNT AND TIMOTHYL. HUNT 261 Fourth, these studies do not account for the fact that the new technologies must be cost-effective and relatively reliable for widespread application. The technologies may appear to be cost- less, producing quantum leaps in productivity for the users. Yet there are purchase costs, installation costs, and ongoing costs that must be accounted for. The ongoing costs include system main- tenance, software development, and training, among others. The cost of unscheduled downtime may become increasingly significant with more integrated systems. Finally, Leontie£Duchin and Roessner et al. appear to us to be truly overoptimistic technologically, both in terms of what office automation equipment can do and in the speed of diffusion of that equipment. Leontief-Duchin assume that word processors alone wit] produce productivity gains for typists and secretaries of 500 percent. Yet this assumption is based upon a short, anonymous trade journal article that is five times more optimistic than the other articles which Leontief-Duchin reference. Roessner et al., on the other hand, emphasizes the potential for two breakthrough technologies, voice input and artificial intelligence. He assumes that innovations will occur in these technologies in the next few years, that they will be successfully marketed, and that they will dramatically reduce clerical employment in banking and insurance during the 1990s. Our major complaint with the technological assumptions of both Leontie£Duchin and Roessner et al. is not that they may be technically wrong, although there is ample reason to question them, but that the level of uncertainty about the technical fore- casts is so great that no one should seriously want to base policy decisions on them. Artificial intelligence, for example, is a technol- ogy that has been touted since the 1950s as a major breakthrough. Perhaps we will always be overoptimistic about new technologies; it stems to be part of the human condition. But that is no justi- fication to shape public policy based on our dreams of the future. Suffice it to say that we are unconvinced that technology will evolve as far or as fast as Leontief-Duchin and Roessner et al. predict. This is the kind of analysis that leads to the fear that we will experience massive technological unemployment at some point in the future. Various analysts have been predicting such an event at least since the dawn of the industrial age. Somehow the employment apocalypse is always just ahead, yet fortunately we never quite reach it.

262 RECENT TRENDS IN CLERICAL EMPLOYMENT Because of the uncertainties about future demand and the capabilities of future technologies, we would encourage a focus on shorter-range occupational forecasting, exactly the opposite approach being suggested by Leontief-Duchin and Roessner et al. Roessner et al. argues that public policy makers need a longer time period for planning. But, if technological change is occurring faster today, then it is becoming even more impossible to develop long-run employment forecasts. Surely it is folly to think that we can peer 15 to 20 years into the future and see the detailed occupational and industrial structure of this nation. In fact, we think that the current :BES efforts, which produce about a Midyear planning horizon, tax existing forecasting abilities to the limit.iS What has this review shown for the future of clerical jobs? First, we think the pessimists who claim that these jobs will either stop growing absolutely or actually decline are wrong. The forces of economic growth, the shift toward services, and the current limitations of office automation technologies all argue strongly against this scenario. However, it is clear that the historical rate of growth of clerical jobs has slowed. Clericals did not benefit from the last recession as they have in earlier recessions, nor are some of the sectors that are important employers of clericals growing as fast as they once were. Finally, office automation is likely to at least contribute to the slowing of employment growth in these occupations in the future. We think that the overall growth of clerical jobs in the future will be average to slightly below average. There is broad agreement among forecasts that clerical jobs will not continue their rapid growth of the past few decades. The recent slowdown in the growth of clerical jobs is very likely permanent. We find no persuasive evidence, however, that there will be a significant decline in the absolute number of clerical jobs. The forecasts of declining clerical employment are based on overopti- mistic expectations of technological improvements or exaggerated productivity claims on behalf of existing technology. In our opin- ion, current office technology offers significant improvements in product quality and modest improvements in productivity. There 15 We also think BLS should be more open about their handling of technological change in their forecasts. Like the other forecasters discussed here, BLS should reveal the basis for its judgments concerning technological change.

H. ALLAN BUNT AND TIMO THY L. HUNT 263 is as yet no empirical evidence of an office productivity revolu- tion that will displace significant numbers of clerical workers. The growth of computer-related clerical positions will continue to be strong. Office automation is not sufficiently advanced at this point to slow the growth of these jobs. Many factors will contribute to the continued, if slower, job growth of clericals in the future. Chief among these is the simple fact that clericals are so diffused in the national economy. More- over, to the extent that clerical jobs are concentrated in particular industries, it has been in industries growing faster than average. Even if the growth rate of some of these slows, as it has in some financial industries for example, others, such as services, are likely to continue to grow at above average rates. Therefore, even al- lowing for negative employment impacts from office automation, the growth of this large, diverse, and diffused major occupational group should not be much below the average growth for all occu- pations for the next decade. Many commentators believe that back-office clerical jobs will disapppear. We do not think this is likely. An analogy to man- ufacturing may be useful. Automation has not caused the total elimination of production workers in manufacturing, but these jobs have not been increasing in absolute terms for 40 years either. IJike most of the forecasts discussed, we think the back-office jobs are more threatened by automation than other positions. They share with production workers a routinization of tasks, which tends to support automation. This will not necessarily lead to their demise but their growth will be well below average. A good example of the limits of technological change is pro- vided by bank tellers. The growth of this occupation has slowed in recent years. The future growth prospects for bank tellers appear to be directly tied to the public's acceptance of automatic teller machines. But these machines today are used mostly for with- drawals and cannot handle nonroutine transactions. They cannot be thought of as a substitute for a fully staffed bank. Furthermore, it is difficult to know if and when the public will be willing to break the human link in making banking transactions. Secretaries fall somewhere between the back-office jobs and those positions that involve considerable customer interface. Therefore, secretarial employment growth may slow but will not stop. We think that the growth of secretaries will be average to below average, but the absolute number of these jobs will definitely

264 RECENT TRENDS IN CLERICAL EMPLOYMENT increase. Secretarial positions require a variety of skills and many are generalist in nature; they are more difficult to eliminate with automation. A variety of skills helps to ensure that the automa- tion of any one of these skills leaves the job intact. It seems clear that secretaries of the future will require an even greater variety of skills and will utilize much more capital equipment. Computer technology is still not ready to tackle the unstruc- tured situations where humans excel. Clerical positions that in- volve direct interface with customers or co-workers are likely to experience at least average growth. The office of the future will require both "high tech" and "high touch" occupations. REFERENCES Attewell, Paul, and James Rule 1984 Computing and organizations: what we know and what we don't know. Communicahon~ of the A CM (December) : 1184-1192. Bailey, Andrew D., James H. Gerlach, and Andrew B. Winston 1985 Office Systems, Technology and Organizations. Reston, Va.: Reston Publishing Company. Brand, Horst, and John Duke 1982 Productivity in banking: computers spur the advance. Monthly Labor Review (December) :19-27. Bureau of the Census 1972 1970 Census of Populah on, Detailed Characteristics, United Stated Sum' mark. PC(1~-D(1~. Washington, D.C.: U.S. Department of Com- merce. Bureau of Labor Statistics 1981 The National Indwtry-Occupahon Matriz, 1970, 1978, and Projected 1990. Bulletin 2~)86. Washington, D.C.: U.S. Department of Labor. 1979 Time Series Data for Input-Output Industries: Output;, Price, and Em- ployment. Bulletin 2018. Washington, D.C.: U.S. Department of Labor. March. Computer and Business Equipment Manufacturers Association 1985 The Computer and Brained Equipment Industry Marketing Data Book. Washington, D.C.: Computer and Business Equipment Manufac- turers Association. Diebold, John 1985 Managing Information: The Challenge and the Opportunity. New York: AMACOM. Drenn an, Matthew P. 1983 Implications of Computer and Communications Technology for Less Skilled Service Employment Opportunities. Final report pre- pared for the Employment and Training Administration. Wash- ington, D.C.: U.S. Department of Labor. Hartmann, Heidi I., Robert E. Kraut, and Louise A. Tilly, eds. 1986 Computer Chips and Paper Clips: Tecknolo~y and Women's Employment. Vol. I. Washington, D.C.: National Academy Press.

H. ALLAN HUNT AND TIMOTHYL. HUNT 265 Honeywell Inc. 1983 Office Automation and the Workplace: A National Survey. Minneapolis, Minn.: Honeywell Inc. Hunt, H. Allan, and Timothy L. Hunt 1986 Clerical Employment and Technological Change. Kalamazoo, Mich.: W.E. Upjohn Institute for Employment Research. Hunt, Timothy L., and H. Allan Hunt 1985 An Assessment of Data Sources to Study the Employment Effects of Technological Change. Pp. 1-116 in Technological Employment Effcet`: Into rim Report. Panel on Technology and Women's Employ- ment, Committee on Women's Employment and Related Social Issues. National Research Council. Washington, D.C.: National Academy Press. International Data Corporation 1982 Information systems for tomorrow's office. Fortune (October 18~: 17-81. 1983 Office systems for the eighties: automation and the bottom line. Fortune (October 3~:89-162. 1984 Information systems for tomorrow's office. Fortune (October 15~: 99-138. Katsan, Harry, Jr. 1982 O.~cc Automations A Mar~ager's Guide. New York: American Man- agement Associations. King, John L., and Kenneth L. Kraemer 1981 Cost as a social impact of information technology. Pp. 93-129 in Mitchell L. Moss, ea., Telecommunications and Productivity. Reading, Mass.: Addison-Wesley. Leontief, Wassily, and Faye Duchin 1986 The E`uture Impact of Automation on Workers. New York: Oxford University Press. Priebe, John A., Joan Heinkel, and Stanley Greene 1972 1970 Occupation and Industry Classification Systems in Terms of their 1960 Occupation and Industry Elements. Technical Paper No. 26. Washington, D.C.: Bureau of the Census, U.S. Department of Commerce. July. Reskin, Barbara F., and Heidi I. Hartmann, eds. 1985 Womcn!s Work, Millie Work. Scz Segregation on the Job. Washington, D.C.: National Academy Press. Roessner, J. David, Robert M. Mason, Alan L. Porter, Frederick A. Rossini, A. Perry Schwartz, and Keith R. Nelms 1985 The Impact of Office Automation on Clerical Employment, 1985- 2000: Forecasting Techniques and Plausible Futures in Banking and Insurance. Westport, Conn.: Quorum Books. Salerno, Lynn M. 1985 What happened to the computer revolution? Harvard Business Review (Novembe - December):129-138. Se~kin, Eugene P., and David P. Sullivan 1985 Revised estimates of new plant and equipment expenditures in the United States, 1947-1983. Survey of Current Bueinc~` (February): 16 - 27.

266 RECENT TRENDS IN CLERICAl EMPLOYMENT Strassman, Paul A. 1985 Ir~formation Payoff: The lEan~forma:tion of Work in the Electronic Age. New York: Free Press. Wang Laboratories, Inc. 1985 Issues in Information Processing: Cost Justification Lowell, Mass.: Wang Laboratories, Inc. BIBLIOGRAPHY Administrative Management 1978 The many cases for WP. Adnuni~trative Management (April):70-71. Bureau of the Census 1982 1980 Census of Population" Classic ed Index of Indu~!nes and Occupations. Washington, D.C.: U.S. Department of Commerce. Business Week 1985 The computer slump. Burners Week (June 24~:74-81. Downing, Hazel 1980 Word Processors and the Oppression of Women. Pp. 275-288 in Tom Forester, ea., The Microelectronics Revolution. Cambridge, Mass.: MIT' Press. Employment and Training Administration 1977 Dictionary of Occupational Titles. 4th ed. Washington, D.C.: U.S. Department of Labor. Gold, Bela 1981 Robotics, programmable automation and increasing competitive- ness. Pp. 91-117 in Exploratory Workshop on the Social Impacts of Robotics: Summary and Issues. Office of Technology Assessment, Congress of the United States. Washington, D.C.: U.S. Govern- ment Printing Office. Karan, Mary A. 1982 Word processing: when it doesn't work. Computer World (March): 16. Klein, Deborah P. 1984 Occupational employment statistics for 1972-82. Employment and Earnings (January):13. Kutscher, Ronald E. 1982 New economic projections through 1990 an overview. Pp. 1-9 in Econorn~c Projections to t990. Washington, D.C.: U.S. Department of Labor. Miller, Ann, Donald J. Treiman, Pamela S. Cain, and Patricia A. Roos, eds. 1980 Work, Jobs, and Occupations: A Critical Remew of the Dictionary of Occupational Titles." Washington, D.C.: National Academy Press. Murphree, Mary 1981 Rationalization and Satisfaction in Clerical Work: A Case Study of Wall Street Legal Secretaries. Ph.D. dissertation. Columbia University, New York. Office of Economic Growth and Employment Projections 1981 Projected Occupational Staffing Patterns of Industries. OES Tech- nical Paper Number 2. Washington, D.C.: U.S. Department of Labor.

H. ALLAN HUNT AND TIMOTHY L. HUNT 267 Office of Technology Assessment 1984 Effects of Ir~formation Technology on Financial Services Sy~ter~w. Wash- ington, D.C.: Congress of the United States. Silvestri, George T., John M. Lukasiewicz, and Marcus E. Einstein 1983 Occupational employment projections through 1995. Monthly Labor Review (November):37-49. Stanford Research Institute 1984 Drop in business demand for personal computers. The SRI Journal (May):4-5. Winston, Patrick H. 1985 The AI business: a perspective. Manufacturing Engineering (March).