econmis

The impact of female employment onmale salaries and careers: evidencefrom the English banking industry,

189019411By ANDREW SELTZER*

The late nineteenth- and early twentieth-century British labour market experiencedan influx of female clerical workers. Employers argued that female employmentincreased opportunities for men to advance; however, most male clerks regarded thisexpansion of the labour supply as a threat to their pay and status. This articleexamines the effects of female employment on male clerks using data from WilliamsDeacons Bank covering a period 25 years prior to and 25 years subsequent to theinitial employment of women. It is shown that, within position, women were substi-tutes for younger men, but not for senior men. In addition, the employment ofwomen in routine positions allowed the bank to expand its branch network, creatingnew higher-level positions, which were almost always filled by men.

The rapid growth in the female labour force participation rate was perhaps themost important change to the late nineteenth- and early twentieth-centuryBritish labour supply. Nowhere in the British economy were the effects of thisgreater than in the clerical sector. Mens greater physical strength gave them acomparative advantage in agriculture and manufacturing, which was reflected incorrespondingly higher wages.2 Clerical work, on the other hand, required littlephysical strength, and thus women had a comparative advantage in the sector.Clerical work was also seen to be more socially acceptable and more feminine thanother types of employment, particularly for middle-class women.3 Thus, when theclerical sector expanded rapidly in the late nineteenth and early twentieth centuryand barriers to female clerical employment were lifted, women rapidly moved intothe sector.Women comprised only 1 per cent of all clerks in 1871.This increasedto 11.1 per cent in 1901 and 44.8 per cent by 1921.4

* Author Affiliation: Royal Holloway, University of London; Centre for Economic History, Australian NationalUniversity; Institute for Compensation Studies, Cornell University; and IZA.

1 I wish to thank the staff of the Royal Bank of Scotland Group Archive in Edinburgh and London (DerekHammond, Laura Yeoman, Jenny Mountain, Ruth Reed, Alison Turton, Lucy Wright, and particularly PhilipWinterbottom) for their enormous help with the records used in this article. I also want to thank three anonymousreferees, Jeff Frank, Claudia Goldin, Elyce Rotella, participants at the Economic History Association meeting inBoston, and participants at seminars at London School of Economics, Lund University, and Royal Holloway,University of London for comments on an earlier draft. I acknowledge research funds from the Royal HollowayFaculty of History and Social Sciences. Remaining errors are mine alone.

2 Burnette, Gender, pp. 72135.3 Anderson, White blouse, pp. 910; Webb, Alleged difference, p. 653; Zimmeck, Jobs for the girls,

pp. 1568.4 Takahashi, Unrealised golden opportunities, p. 39.

bs_bs_banner

Economic History Review, 66, 4 (2013), pp. 10391062

Economic History Society 2013. Published by John Wiley & Sons Ltd, 9600 Garsington Road, Oxford OX4 2DQ, UK and 350 MainStreet, Malden, MA 02148, USA.

This expansion of female employment had several potentially important effectson clerical labour markets.There exists a sizable literature on clerical employmentand female labour market outcomes.5 However, the literature on the spill-overeffects to male workers is considerably smaller. Simple economic theory suggestsseveral potential effects of feminization. If employers used male and female clerksinterchangeably, female employment would reduce male wages through the basicprinciples of supply and demand. However, if there were complementaritiesbetween male and female clerks, increased female employment would lead tohigher male wages. Finally, the growth of female employment in the clerical sectorwas closely linked to the rise of dual labour markets.6 Men normally filled thehigh-pay, high-status positions and women were typically confined to routinepositions with low pay. The dual labour market hypothesis suggests that womenand men operated in different segments of the clerical labour market, and thus didnot directly compete with each other.

The debate among late nineteenth- and early twentieth-century clerks andemployers about the spill-over effects of feminization largely followed along thelines suggested by economic theory. Established male clerks typically saw womenas direct competition for jobs.7 Many male clerks feared that they would becrowded out by women, who typically earned lower salaries, particularly afterseveral years tenure.8 Debates within the clerical unions about female employ-ment inevitably turned to the detrimental effects that it had on male employmentand salaries.9 Subsequent scholars, such as Anderson, Klingender, Lewis,Wilson,and Zimmeck, have generally accepted that female employment resulted in lowermale salaries.10

Perhaps not surprisingly, contemporary employers and female clerks offered avery different view, arguing that female employment was necessitated by thegrowth of the sector and that male and female clerks were not close substitutes.They argued that increasing demand for clerical services was accompanied bymechanization of the sector, and this allowed for greater division of labour, which,alongside the standard workplace practice of requiring women to leave employ-ment upon marriage, created a system of dual labour markets.11 Male positionswere characterized by opportunities for promotion depending on ability andtenure, whereas female positions were low-level and routine, requiring little train-ing and offering few prospects for promotion.12 Dohrn argued that the growth of

5 See Seltzer, Female salaries, for a summary of this literature.6 See Reich, Gordon, and Edwards, Dual labor markets, on the dual labour market hypothesis.7 Rathbone, Remuneration, pp. 5660; Quicksilver, Should women be abolished?, Clerk (1920), pp. 45; A

junior, Another view of women in banks: a juniors grievance, Bank Officer, 4 (1922), p. 12; Bank OfficersGuild, Correspondence, Bank Officer (191921).

8 Webb, Alleged difference, pp. 64956; Goldin, Understanding the gender gap, pp. 10614; Seltzer, Femalesalaries, pp. 46671.

9 Anderson, Victorian clerks, p. 59; Bank Officers Guild, Correspondence (see above, n. 7); A junior,Another view (see above, n. 7).

10 Anderson, Victorian clerks, pp. 5265; Klingender, Condition, p. 22; Lewis, Women clerical workers, pp. 31,412; Wilson, Disillusionment, p. 249; Zimmeck, Jobs for the girls, p. 170.

11 Marriage bars were used by virtually all clerical employers, existing at the Civil Service, the Post Office,railways, libraries, schools, banks, and insurance companies. See Allman, Williams Deacons, p. 165; Anderson,White blouse, p. 10; Cohn, Process, pp. 97101; Liladhar and Kerslake, No more library classes, pp. 21920;Oram, Women teachers, pp. 4, 267; Parris, Staff relations, pp. 1456, 1512.

12 Bank Officers Guild, Correspondence (see above, n. 7); Cohn, Process, pp. 91116; Goldin, Understandingthe gender gap, pp. 11014; Jordan, Lady clerks; Seltzer, Female salaries, pp. 4624, 4713.

1040 ANDREW SELTZER

Economic History Society 2013 Economic History Review, 66, 4 (2013)

womens jobs benefited male clerks, stating that, Hiring an increasing number oflow paid women for menial work . . . kept the male clerks in a good mood becausethe more interesting and better paid work became their exclusive domain.13 Inaddition, female employment facilitated the expansion of the clerical sector, whichdisproportionately created positions above the entry level, and thus enhanced maleclerks prospects for advancement.

The debate over the effects of female employment on male careers has yet to besatisfactorily resolved. Most of the published views of contemporaries are fromtrade union members, who were disproportionately the sort of rank-and-file maleclerks that were most likely to be displaced by women. Upwardly-mobile youngmen and more senior staff who had already progressed through the ranks areconspicuously absent from these discussions. Subsequent scholars who haveaccepted that female employment reduced male salaries have relied heavily onthese published views. What is almost completely absent from both the contem-porary debates and later scholarship is quantitative analysis of the effects describedabove. This article provides such an analysis, focusing on the banking industry.Banking is a good case study for two reasons. First, banking (along with the CivilService, railroads, and insurance) was among the most elite of clerical industries,with higher pay and greater job security than the rest of the sector.14 Thus thebanking industry offers a broader picture of the effects of feminization on theclerical aristocracy. Second, the pattern of feminization in banking makes itpossible to use modern econometric techniques to estimate male earnings, and toseparate the effects of feminization from the effects of other factors occurringsimultaneously in the industry and the broader economy. Unlike the rest of theclerical sector, which feminized gradually from the 1870s, the feminization ofbanking occurred swiftly as a result of the labour market shock caused by theenlistment of over 40 per cent of male staff during the FirstWorldWar. Prior to thewar, banking had been almost entirely male, during the war it feminized veryrapidly, and afterwards most women remained at the banks.15 The discrete shockto banking labour markets caused by the war and variation in female employmentacross branches after the war makes it possible to use difference-in-differenceregression analysis to control for unobservable time-varying factors which affectedwages and estimate the effects of female employment on different types of malestaff.

The primary focus of this article is on the salaries and career prospects of malestaff at Williams Deacons Bank, a mid-sized regional joint-stock branch bankbased in Manchester with a few offices in London.The data come from the banksextremely rich personnel records, which cover virtually all staff between 1890 and1936, and most through 1941.16 The records contain the number of male andfemale staff employed in each branch and the annual salary paid to each member

13 Dohrn, Pioneers, p. 50. Also see Heller, London clerical workers, pp. 1257.14 Wilson, Disillusionment, p. 15; Lockwood, Black-coated worker, pp. 44, 51.15 Takahashi, Unrealised golden opportunities, p. 39; Blackburn, Union character, p. 134.16 Royal Bank of Scotland Group Archive, Edinburgh (hereafter RBSG Archive), Williams Deacons Bank

Limited, staff registers, GB1502/WD/480/1 (covers 18901901), GB1502/WD/480/2 (covers 190110),GB1502/WD/481 (covers most northern branches, 190718), GB1502/WD/46/1 (covers London branches andMosley St, 191121), GB1502/WD/482/2 (covers most northern branches, 191927), GB1502/WD/482/3 andWD/482/4 (covers most northern branches, 192434), GB1502/WD/482/5 and GB1502/WD/482/6 (covers mostnorthern branches, 193141), GB1502/WD/46/2 (covers London branches and Mosley St, 192136).

IMPACT OF FEMALE EMPLOYMENT 1041


of staff.These records have been used to create a panel dataset covering the banksmale staff over extended periods both prior and subsequent to female employ-ment. These data are used to examine the extent to which the employment ofwomen from 1915 affected the salaries of male staff.

In addition to examining salaries, this article also examines the effects of femaleemployment on mens promotion prospects.This is a potentially important effect,as opportunities for promotion had long been viewed as one of the major attrac-tions of banking employment.17 Here there is less theoretical ambiguity than is thecase with salaries. Female employment had two separate effects, both of whichunambiguously increased mens promotion prospects. First, by making postsavailable to women, banks increased their pool of suitable applicants and thenumber of staff they could employ. Increased staff numbers were essential for theexpansion of branch networks following the First World War. The new brancheswere universally small, employing a manager and relatively few clerks.The expan-sion of the branch networks thus created new managerial positions, and, becausethe new branches employed relatively few clerks, resulted in an increase in theproportion of staff in managerial positions. Second, because promotion to branchmanager was not possible for women until well after the Second World War, whennew branches were opened, the newly created managerial positions always went tothe men.

The outline of the remainder of this article is as follows. Section I provides abrief historical background of the British banking industry, male careers, and theintroduction of female employment to the industry. Section II provides a briefhistorical background of Williams Deacons Bank and describes the rich dataset.Section III examines the impact of female employment on the probability thatmale staff would be promoted to manager or be assigned to clerical positions inlarge back offices. Section IV uses a difference-in-difference approach to estimatethe effect of female employment on mens salaries. Section V concludes.

I

In the mid-nineteenth century, most English banks were privately owned andoperated at most a small handful of offices.The late nineteenth and early twentiethcentury witnessed several important developments in the broader economy whichled to changes to the organization of the industry and its demand for labour. Thenumber of bank accounts grew dramatically as incomes increased.18 Mechaniza-tion lowered the costs of monitoring staff and communicating between branches.19

A series of education acts and the growth of commercial education increased thesupply of clerical labour.20 The industry responded to these changes with rapidconsolidation as larger joint-stock banks absorbed private banks and smallerjoint-stock banks. In addition, the larger banks opened numerous new branches.The total number of banking offices increased from 2,203 in 1890 to 10,082 in

17 Rae, Country banker, pp. 16570; Blackburn, Union character, pp. 756; Heller, London clerical workers,pp. 99102.

18 The number of accounts held by Williams Deacons increased by approximately 380% from 1891 to 1926;RBSG Archive, Williams Deacons Bank, GB1502/WD/377/1, particulars of branches (18901940).

19 Wardley, Commercial banking industry, pp. 7789.20 Heller, London clerical workers, pp. 15579.

1042 ANDREW SELTZER


1930.21 The average number of branches per bank increased from 7.1 in 1870 to14.3 in 1890, 31.5 in 1901, 67.5 in 1911, and 128.9 in 1920. By 1914, the industrywas dominated by a handful of very large banks which operated hundreds orthousands of branches. Although these changes preceded large-scale femaleemployment, they created the economies of scale in the back offices that laterproved crucial to the creation of womens jobs.

During the late nineteenth and early twentieth century, male bank clerks weregenerally viewed as being among the clerical aristocracy. Banking employmentwas long-term, secure, and characterized by internal labour markets. A typicalbanking career prior to the FirstWorldWar might be described as follows.22 Juniorswere hired shortly after completion of secondary schooling. For the next few years,they were effectively apprentices, and gradually learned the trade on the job. It wasfairly common for young men to leave banking during their first few years ofemployment. If a young man remained at the bank for three to five years andshowed promise, he would be moved from branch to branch and gradually bepromoted through the ranks, eventually reaching the level of manager after about1520 years. Pay was closely tied to tenure and was largely shielded from theexternal labour market.There were substantial differences in the responsibility andpay of managers based on branch size. The number of managers was strictlylimited by the number of branches, and only about 25 per cent of entrants werepromoted to the level of branch manager during their careers. However, all staffwho proved their loyalty and basic competence had secure jobs and higher paythan clerks in most other industries. Men who were not capable of advancing weretypically given back office positions at the larger branches. Pensions were unusuallygenerous and were typically based on end-of-career salaries.

The growth of the branch networks had important effects on the banks internallabour markets. In the mid-nineteenth century, a bank clerk was a generalist whotypically worked alongside the owner or general manager.This was not possible inthe much larger banks of the late nineteenth and early twentieth century, wheremost staff worked in either the ever-larger back offices of the main branches or inthe increasingly numerous smaller branches. The increase in the size of the banksand the need to monitor growing branch networks resulted in a dramatic increasein routine clerical work. Much of the work in the back offices became increasinglyspecialized, comprising tasks such as secretarial work, typing, coupons, and other[similar] posts.23 By contrast, work at the smaller branches remained focused oncustomer interface.

The transformation of banking offices had a dramatic effect on female employ-ment. In 1911 women comprised only 1.2 per cent of banking staff.24 Most privatebanks employed only men. The First World War provided the impetus for femaleemployment in the industry. Most younger male staff either volunteered or werecalled up into the Services. From 1915 it became necessary to bring in women as

21 Sheppard, Growth and role, pp. 11617.22 Blackburn, Union character, pp. 719; Heller, London clerical workers, pp. 99102; Seltzer and Frank, Pro-

motion tournaments, pp. i5670; Seltzer, Salaries and promotion, pp. 74852; idem, Did firms cut nominalwages?, pp. 11721; idem, Female salaries, pp. 46673.

23 RBSG Archive, Williams Deacons Bank Limited, GB1502/WD/49/8, letter from asst GM to Londonmanager, London managers staff papers (1928).

24 This compares to 24.5% of clerks in commerce, 8.8% in insurance, 20.7% in the civil service, 6.0% in law,and 1.3% at the railroads; Takahashi, Unrealised golden opportunities, p. 39.



temporary replacements. During the war, women performed the jobs of the men,including some high-level positions such as cashier or acting division head.25

Following the war, most female staff were given permanent positions, though anunusually large number left in 191920.26 Unlike the war years, almost all womenwere assigned routine clerical work during the interwar period, with little or noprospect for advancement.27 Traditionally these tasks had been done by juniorsand by older men who lacked the ability to progress through the ranks. However,after the war the banks increasingly saw these tasks as womens work. In part thiswas due to a widespread perception among employers that women were tempera-mentally better suited to routine and repetitive tasks.28 In addition, the diffusion ofnew technologies such as the typewriter and the adding machine meant that thesetasks required relatively little time to master, which made them ideally suited to afemale workforce that typically had short careers because of marriage bars.29

Perhaps most importantly, women were paid considerably less than men, particu-larly after several years of service.30

The real salaries of male bank clerks declined somewhat in the two decades priorto the First World War, and sharply during the war years.31 The bank clerksresponded to declining pay by unionizing. After a first abortive attempt in 1914,the Bank Officers Guild (BOG) was formed in 1917. Although discussions ofinflation, a fixed nominal pay scale, and declining real salaries dominated the earlyBOG meetings and the early issues of their trade journal, the Bank Officer, thefeminization of the industry and its impact on male staff quickly emerged as theBOGs second issue.32 Many male union members, particularly juniors andreturning war veterans, believed that they would be replaced or their wages wouldbe undercut by women and opposed all female employment in the industry. Oneclerk stated, Their competition cannot fail to depress the standard of living ofthose [men] compelled to earn in order to live.33 Another stated that women tendto keep down and probably further reduce the present [wage] scale of the variousbanks.34 Other BOG members supported female employment (or at leastaccepted it as inevitable), but also sought to establish safeguards to prevent the

25 Wardley, Women, pp. 46; RBSG Archive, Williams Deacons Bank Limited, GB1502/WD/49/1 andGB1502/WD/49/9, Female staff and Male staff, London managers staff papers (1921 and 1929).

26 RBSG Archive, Williams Deacons Bank Limited, staff registers (see above, n. 16). In 1919 and 1920 anaverage of 78.5 women left the bank each year. Between 1921 and 1936, at most 37 left the bank in any year.

27 The Williams Deacons data suggest that very few women advanced to positions of responsibility. Higher-level positions normally came with a salary over the normal 250 maximum for women; RBSG Archive,WilliamsDeacons Bank Limited, GB1502/WD/49/1 and GB1502/WD/49/9, circular, London managers staff papers(1920). Only 11 (of 1,293) women who entered between 1915 and 1941 earned over 250 at any point in thesample period.

28 Webb, Alleged difference, p. 650; Zimmeck, Jobs for the girls, pp. 1589; Lewis, Women clerical workers,p. 37.

29 The marriage bar was universal in the banking industry throughout the sample period; Allman, WilliamsDeacons, p. 165; Heller, London clerical workers, pp. 11423.

30 Seltzer, Female salaries, pp. 460, 46970.31 Blackburn, Union character, pp. 1324; Lockwood, Black-coated worker, pp. 414; Seltzer, Salaries and

promotion, pp. 7427.32 Bank Officers Guild, Correspondence (see above, n. 7);Warwick Univ., Modern Records Centre (hereafter

MRC), Bank Officers Guild, MS 56, minutes of general meeting (1920). According to Feinstein, National income,tab. 65, prices increased by 24.7% between 1895 and 1914 and another 141.6% between 1914 and 1920.

33 Bank Officers Guild, Correspondence (see above, n. 7) (Sept. 1921), pp. 1415.34 A junior, Another view (see above, n. 7), p. 12.

1044 ANDREW SELTZER


utilization of women as cheap labour.35 Ultimately, equal pay for equal workemerged as the BOGs official position; however, because many women within theGuild feared that successful insistence on this point . . . would mean that verymany women would lose their posts, the BOG accepted that female salaries couldbe lower than male salaries so long as there were limit[s] put upon the type of worka woman can be called upon to perform.36

Employers and some female clerks believed that this view of the effects of femaleemployment was fundamentally flawed. They argued that because most womenwere limited to routine back office duties, they were not close substitutes for men.Instead female employment reduced the need for men to perform routine duties,which meant that more could be assigned to positions of responsibility. One femaleclerk wrote, If a junior is capable he will get promotion because the women are notofficially considered to be . . . making banking a life job, . . . [and] few of thewomen obtain promotion.37 Similarly, a 1918 Williams Deacons Bank internalmemo stated, [o]ur men may justly welcome [female clerks] retention as offeringto themthe menthe greater chances of responsible posts and less of thatroutine which to many active minds spells monotony.38

A second argument put forth by advocates of female employment was that itfacilitated the growth of the branch network, and thus the creation of new mana-gerial positions. At the end of the war, the banks were short-staffed and needed toretain women in order to continue to expand the networks. A 1918 WilliamsDeacons internal memo stated, we are still lamentably short of staff in spite of ourutmost endeavours to find and take suitable juniors.39 Another memo from laterin 1918 stated, Even if we get all these new-comers, as well as our demobilizedmen, we shall still need the help of many of our women clerks.40 Later scholarshave noted the existence of an age bulge whereby promotion opportunitiesincreased shortly after the war.41 For example, Lockwood argued that: Manyopportunities of promotion at quite early ages were provided by . . . the post-warexpansion in banking which persisted until the middle and late 1920s.42

II

Williams Deacons was the product of the 1890 merger between the London-basedWilliams, Deacon & Co. and the Manchester and Salford Bank.43 During theperiod of this study, the bank maintained a moderate size (57 branches in 1891and 206 in 1939) and a largely north-western focus. After running into troublesbecause of the decline of the Lancashire textile industry, it was absorbed by theRoyal Bank of Scotland in 1930, but continued to trade separately under its ownname until 1969.

35 MRC, Bank Officers Guild, MS 56, minutes of General Purposes Sub-Committee meeting (1926 and1928).

36 Ibid.37 E. A. Sulthorp, Correspondence, Bank Officer (June 1922), p. 16.38 RBSG Archive, Williams Deacons Bank Limited, GB1502/WD/HC221, memos (1918).39 Ibid.40 Ibid.41 Blackburn, Union character, p. 76; Lockwood, Black-coated worker, pp. 5767.42 Lockwood, Black-coated worker, p. 64.43 Allman, Williams Deacons, pp. 14150.



Similar to the rest of the industry, approximately 45 per cent of WilliamsDeacons male staff received Service Leave during the First World War. FromAugust 1915, the bank began to appoint women on a temporary basis. R. T.Hindley was appointed as the banks general manager in 1917, and he immediatelyset out plans for branch expansion after the end of the war. The expansion (from120 branches in 1918 to 200 in 1927) had important implications for the distri-bution of branch sizes, and thus positions at the bank. The new branches wereuniversally small, and the number of branches with five or fewer staff increasedfrom 44 in 1914 to 91 in 1928. Each of these branches had a manager, and thusthe growth of the branch network created new promotion opportunities. At thesame time, the increased volume of paperwork associated with the growth of thebranch networks required larger back offices at the main branches. Between 1914and 1928, the number of staff at the banks head office increased from 107 to 179.The expansion of the network required an increase in the size of the workforce, andwith the bank losing 46 men in the war and another 178 to resignation orretirement between 1915 and 1921, it became apparent that women would have tobe appointed on a permanent basis, a policy which was formalized in July 1920.44

The number of female staff dropped sharply after the war, due to resignations anda virtual cessation of female appointments until 1928.45 At the same time, thenumber of men appointed surged after the First World War and then declinedsharply from the late 1920s, when the bank stopped opening new branches. Thebank resumed regular appointment of women in the 1930s, and increased thenumber of female appointments after 1938, due to the mobilization of youngermale staff for the Second World War. Figure 1 shows the trends in the number ofstaff at the bank and in branches operated by the bank and the industry as a wholeover the sample period.

There are strong reasons to believe that Williams Deacons was fairly repre-sentative of personnel practices in banking and is thus a good case study of theindustry. Its pattern of female employment prior to, during, and subsequent to thewar was mirrored by the rest of the industry.46 Over the sample period, there is avery strong (0.992) and significant correlation between the number of branchesoperated by Williams Deacons and the number operated by the industry as awhole, and a weaker (0.16) but still significant correlation for the annual percent-age change in the number of branches.47 Seltzer has shown that other banks hadsimilar wage policies with regard to tenure and similar trends in real wages in theyears just before feminization.48 Moreover, the BOG normally referred to banks asa collective when discussing personnel issues, suggesting strong similarities in theirpractices.

The primary source of data is Williams Deacons personnel records, collectedfrom the Royal Bank of Scotland Group Archive in London and Edinburgh.49 Therecords, which are organized by branch, list virtually every employee at the bank

44 RBSG Archive, Williams Deacons Bank Limited, GB1502/WD/49/1 and GB1502/WD/49/9, circular,London managers staff papers (1920).

45 RBSG Archive, Williams Deacons Bank Limited, staff registers (see above, n. 16).46 Wardley, Women, pp. 45.47 The data are from Sheppard, Growth and role, pp. 11819, and RBSG Archive, Williams Deacons Bank

Limited, particulars of branches (see above, n. 16).48 Seltzer, Salaries and promotion.49 RBSG Archive, Williams Deacons Bank Limited, staff registers (see above, n. 16).

1046 ANDREW SELTZER


between 1890 and 1936, and at northern branches (except the head office)between 1890 and 1941.50 For each employee, the records contain dates of birth,entry to the bank, entry to the branch, and exit from the branch; reason for exit(resigned, died, retired, dismissed, transferred to another branch, and left); andcontinuous information about nominal wages. The wages have been recorded onan annual basis and deflated using Feinsteins price series.51 It is also possible touse the records to infer the number of staff employed at each branch and the staffmember who was the branch manager.52 The records have been used to constructa panel dataset which contains 34,977 annual observations of 2,117 male staff. Ofthese observations, 12,171 are from 18901914 and 19,381 are from 192041.The war years, which contain the remaining 3,425 observations, were atypical inseveral ways and in the analysis these years are controlled for using a dummyvariable or omitted altogether.53

50 The only missing observations are for staff who left between 1890 and 1895.51 Feinstein, National income, tab. 65. Wages and branch of employment were recorded for every employee

present for any part of a calendar year. This information was recorded as of 1 Oct. each year or, if the employeewas not present on 1 Oct., as of the last available date.

52 See Seltzer, Female salaries, on the identification of managers.53 The dummy variable covers the period 191519 rather than the war years per se (191418) because men did

not begin enlisting until late in 1914 and because demobilization and the temporary employment of womencontinued through 1919.

1,400 800

700

Women

Men

Total

BranchesWilliams Deacon's Bank

BranchesEngland average 600

500

400

300

200

100

0

1,200

1,000

800

600

Sta

ff

Bra

nche

s

400

200

0

1890

1892

1894

1896

1898

1900

1902

1904

1906

1908

1910

1912

1914

1916

1918

1920

1922

1924

1926

1928

1930

1932

1934

1936

Figure 1. Number of branches and staff, 18901936Note: The series BranchesEngland average shows the average number of branches per bank across all trading banks inEngland.Sources: Sheppard, Growth and role, pp. 11617; RBSG Archive, Williams Deacons Bank Limited, staff registers and particularsof branches (see above, n. 16).



In addition to the wage records, the Royal Bank of Scotland Archive containsrecords on the branches, which provide information about the age of the branch,the value of advances and deposits, and the number of accounts and staff.54 Thearchive also contains numerous memos, minutes, job descriptions, and so on, thatprovide rich institutional information about the bank and its policies with regardto staff.55

III

As noted in previous sections, the banks, female clerks, and some subsequentscholars have argued that men were more likely to advance to positions of respon-sibility as a result of the employment of women.There are two theoretical reasonsfor this. First, because women were rarely promoted to positions of responsibilityand never to manager, their presence reduced the competition for these positions.Second, the lower labour costs associated with female employment allowed thebanks to expand. This section examines whether men at Williams Deacons were,in fact, more likely to advance to positions of responsibility and less likely to beassigned routine back office positions after 1915.56

To examine the likelihood of men being assigned to positions of responsibility,it is necessary to consider the percentage of men with sufficient service to beconsidered for promotion to branch manager who actually held managerial posi-tions. Figure 2 shows this percentage, which is measured using three differentdenominators: the total number of male staff, the total number of staff (includingwomen), and the total number of staff at branches opened before 1915 (includingwomen).The numbers in both the numerator and denominator include only staffwith at least 10 years service, about the minimum needed to be considered forpromotion to branch manager.The analysis has been restricted to staff with at least10 years of service, as this provides a reasonable proxy for how a contemporarymale bank clerk would have viewed his promotion prospects.57 The male staffseries can be thought of as the actual percentage of men holding managerialpositions. The all staff and pre-1915 series can be thought of as showingcounterfactual percentages in managerial positions under different assumptionsabout the banks expansion.The all staff series shows the percentage of men whowould have been in managerial positions had the bank expanded in exactly the waythat it actually did, with the exception of hiring only men after 1915. The pre-1915 series shows the percentage of men who would have been in managerialpositions had the bank not opened any new branches after 1914, but expandedexisting branches as it actually did and continued to employ only men. The gapbetween these series can be thought of as a lower bound estimate of the actualeffect of female employment on mens promotion prospects, as the counterfactual

54 RBSG Archive, Williams Deacons Bank Limited, particulars of branches (see above, n. 16).55 RBSG Archive, Williams Deacons Bank Limited, memos, circular, male staff, female staff, and letter from

asst GM (see above, nn. 23, 25, 27, 38).56 The analysis in this section focuses solely on incumbent employees. It is likely that female employment also

meant that some young men who would otherwise have filled the routine positions in banking were forced to takelower-paid and lower-status positions in other industries.

57 I have also estimated the series in fig. 2 restricting the sample to staff with 12 and 15 years of service, as thesewere more typical periods of time until initial promotion to manager. The results are substantively the same asthose shown in fig. 2.

1048 ANDREW SELTZER


contains an implicit assumption that male clerks hired after 1915 would have quitat the same rate observed for women. In fact, men had a much higher probabilityof surviving to 10 years tenure (73.8 per cent for men versus 24.2 for womenamong entrants between 1915 and 1926), and thus it is likely that had the bankonly hired men after 1915, from 1925 there would have been more staff withsufficient seniority to be considered for promotion than was actually the case.

It can be seen from figure 2 that men were more likely to be in managerialpositions after 1915. The age bulge described by Blackburn and Lockwood isapparent from the sharp increase in the percentage in managerial positions from1915 and the subsequent decline and return to prewar levels by 1936. By con-struction, the male staff and the pre-1915 series are the same until 1915 and themale staff and all staff series are the same until 1925, 10 years after the bankbegan employing women. A comparison of the male staff with the all staff, andof the male staff with the pre-1915 series, shows that the increase in promotionopportunities was due to both the expansion of the branch network and theemployment of women. The gap between the male staff and all staff series canbe thought of as the effect of the exclusion of women from managerial positions.From 1925 (the first year that any women would have had sufficient seniority tohave been considered for promotion), the absence of women in managerial posi-tions increased mens promotion opportunities. The larger gap between the allstaff and pre-1915 series can be thought of as the effect of branch expansion. Itcan be seen that managerial opportunities in the existing branches were declining

35

30

25

20

15

10

5

0

1895

1897

1899

1901

1903

1905

1907

1909

1911

1913

1915

1917

1919

1921

1923

1925

1927

1929

1931

1933

1935

Managers as a % of staff inbranches opened before 1915

Managers as a %of all staff

Managers as a %of male staff

Figure 2. Staff in managerial positions, 18951936Note: Figures computed as a percentage of staff with at least 10 years tenure.Source: RBSG Archive, Williams Deacons Bank Limited, staff registers (see above, n. 16).



after 1915, as the number of managers in existing branches remained constant andthe number of men of promotable age in these branches continued to grow.

Finally, it appears from figure 2 that increased promotion opportunities for menlasted for approximately one generation, as by 1936 the percentage of male staff inmanagerial positions had returned to its prewar level. However, much of thedecline in promotion opportunities in the 1930s was likely a phenomenon that wasspecific to Williams Deacons. The decline in the percentage in managerial posi-tions from the mid-1920s shown in the all staff series can be attributed to twofactors: an increase in the number of men of promotable tenure (due to thematuring of the cohort hired after the war) and the end of the expansion of thebranch network in the late 1920s. All of the major banks expanded after the FirstWorld War, and thus probably had substantial increases in the number of men ofpromotable age from the late 1920s. However, it can be seen in figure 1 thatWilliams Deacons was atypical in terms of branch expansion; the decline of theLancashire textile industry led it to stop opening new branches in the mid-1920,whereas branch growth continued at other banks through the 1930s.

Table 1 provides a more detailed analysis of the changes in promotions tobranch manager over time.The percentage of men of promotable seniority (witha tenure of between 10 and 30 years) receiving promotion increased sharply andthe average age of first promotion to manager decreased slightly after 1919. Thechange in average age understates the actual increase in promotion opportunitiesbecause it reflects two opposing trends. First, promotion before the age of 30 wasextremely rare before 1915 (1.6 per cent of promotions), but fairly commonplaceafter 1919 (18.9 per cent of promotions). Second, there was a slight increase in thepromotion rate for older men (from 14.7 per cent of promotions between 1890and 1914 to 26.1 per cent between 1920 and 1936).Thus improvements in careerprospects occurred at two different margins: the most talented men were promotedthrough the ranks faster after 1919 and some marginally talented men, who wouldnot have been promoted in the earlier period, received late-career promotions inthe later period.

In addition to arguing that female employment increased promotion opportu-nities, senior bank managers also claimed that female employment alleviated the

Table 1. Age and tenure at the time of first promotion to branch manager

YearPromotions

(% of eligible)Age at firstpromotion

% less thanage 30

% greater thanage 45

18909 13 (6.4%) 36.83 0.00 0.0019004 13 (6.2%) 37.17 7.69 15.3819059 20 (6.8%) 40.74 0.00 20.00191014 21 (7.0%) 39.46 0.00 19.05191519 32 (10.0%) 38.78 3.13 25.0019204 55 (18.1%) 37.31 20.00 21.8219259 43 (14.0%) 34.15 32.56 11.6319306 51 (8.1%) 37.80 11.76 35.2918901914 67 (6.7%) 38.83 1.60 14.72192036 149 (12.0%) 36.78 18.92 26.13

Note: Col. 2 shows the number of staff promoted to branch manager and the % of male staff in non-managerial positions withtenure between 10 and 30 years who were promoted to branch manager.Sources: RBSG Archive, Williams Deacons Bank Limited, staff registers; Seltzer, Salaries and promotion, p. 683.

1050 ANDREW SELTZER


need for men to be in routine and tedious back office positions.58 The WilliamDeacons data do not contain sufficiently detailed information about position todirectly identify the number of staff in routine positions. Because these positionswere disproportionately in the larger back offices, the percentage in clerical posi-tions at the larger branches (20+ staff) is considered as a proxy for the percentagedoing routine duties.This is shown in figure 3, which is constructed the same wayas figure 2, with the exception that the denominator is not restricted to staff withat least 10 years tenure (as this was not a prerequisite for assignment to a largerback office). By construction, the male staff, all staff, and pre-1915 series arethe same until 1915. It is evident from figure 3 that men were less likely to be inclerical positions in the large back offices after 1915. As with the changes in thepercentage in managerial positions, this was due to both the expansion of thebranch network and the employment of women. From 1915, the prevalence ofwomen in the larger back offices meant that fewer men were assigned thesepositions. Much of the employment growth in the existing branches was in thelarge back offices, and, but for the opening of new branches, the proportion of staffin these positions would have increased relative to prewar levels.

58 RBSG Archive, Williams Deacons Bank Limited, GB1502/WD/HC221, memos (1918).

65

60

55

50

45

40

35

30

25

20

1895

1897

1899

1901

1903

1905

1907

1909

1911

1913

1915

1917

1919

1921

1923

1925

1927

1929

1931

1933

1935

Male clerks in branches with20+ staff as a % of male staff

Clerks in branches with20+ staff as a % of all staff

Clerks in branches with 20+ staff as a %of all staff in branches opened before 1915

Figure 3. Staff in clerical positions at large branchesNote: The series Clerks in branches with 20+ staff as a % of all staff in branches opened before 1915 includes both men andwomen in the numerator (clerks in branches with 20+ staff) and in the denominator (all staff in branches opened before 1915).The series Male clerks in branches with 20+ staff as a % of all staff includes only men in the numerator and men and womenin the denominator. The series Male clerks in branches with 20+ staff as a % of male staff includes only men in both thenumerator and the denominator.Source: As for fig. 2.



The effects shown in figures 2 and 3 are fairly large. For example, a comparisonof the actual percentage of men in managerial and clerical positions in the largeback offices (the male staff series) to the counterfactual percentages (the all staffand pre-1915 series) for 1928 shows the following.The actual percentages of men(of suitable tenure) in managerial positions and clerical positions in the large backoffices were 30.8 and 36.2, respectively. The percentages of all staff in thesepositions were 24.5 and 40.8, respectively.The percentages of all staff in the olderbranches in these positions were 16.9 and 49.9, respectively. In other words, malestaff would have been about 20 per cent less likely to be in managerial positionsand 11 per cent more likely to be in large back office clerical positions if the bankhad expanded as it did but only employed men than was actually the case. Inaddition, male staff would have been nearly 50 per cent less likely to be inmanagerial positions and 40 per cent more likely to be in large back office clericalpositions if the bank had expanded the existing branches as it did, had not openedany new branches, and had continued employing only men, than was actually thecase.

IV

In order to examine the effects of female employment on male salaries, this articleborrows an empirical approach from the literature on migration.59 In the stylizedmigration model, unskilled migrants are substitutes in production for the unskillednative-born and complements to the skilled native-born. Cities in the destinationcountry are treated as islands with separate labour markets. Migrants are morelikely to go to cities that are geographically close to their home country. Thisimplies that border cities will have larger labour supply effects from immigrationthan interior cities. Importantly, the geographic attraction of border cities isexogenous and does not depend on differences in local labour market conditionsin potential destination cities. Thus the effect of migration on the wages of thenative-born can be estimated by comparing wage changes of different types ofworkers across cities of varying distance to the source countries following changesin the overall immigration rate.60

This approach can be easily modified to the case of female labour supply at bankbranches.The islands analogy is perhaps more appropriate for bank branches thanfor cities.This is because bank staff were assigned positions by their employers anddid not have the right to move between branches in response to salary differencesacross branches. Female employment created a shock to the banks labour supply.The extent of this shock was not distributed evenly across branches. A branchwhich was relatively specialized in routine back office work (which was regarded aswomens work) would have received a larger supply shock than one that was morespecialized in the customer interface (which remained all-male throughout theinterwar period). The testable implication of this model is that after 1915, thesalaries of junior (but not senior) men at the branches with a high percentage offemale staff should have declined relative to those at branches with a low percent-age of female staff.

59 Card, Immigrant inflows, pp. 3641; Borjas, Economic analysis, pp. 1697704.60 One caveat to this approach is that there may be a general equilibrium effect whereby natives migrate

between cities in response to immigration, reducing the extent of relative earnings differences; Borjas, Economicanalysis.

1052 ANDREW SELTZER


The model can be estimated using data on male salaries at Williams Deaconsprior to and subsequent to 1915. The underlying methodological approach is toestimate a difference-in-difference within the well-known framework of the Mincerwage regression.61 The determinants of real male salaries are examined usingregressions of the following form:

Ln real wage a X b POST b AVGFEMb POST

i t i t i i t, , ,( ) = + + ++

1 23

19141914419144

i i t

i i t i t i t

AVGFEMb POST AVGFEM TENURE

+ +,

, , ,(1)

where i,t denotes individual i and time period t, X is a vector of control variables,POST1914 is a dummy which equals 1 if the observation is from 191541,AVGFEM is the average percentage of female staff at the branch between 1919and 1936, and POST1914*AVGFEM and POST1914*AVGFEM*TENURE areinteractions.

The control variables included in the regression are a fairly standard set ofpersonal and workplace characteristics.62 These are: TENURE (and TENURE2,TENURE3, and TENURE4); ENTRY AGE (and its square); HEAD OFFICE, adummy for whether the individual was employed at the banks head office;WWI,a dummy for the years 191519; INFLATION, the national inflation rate;MANAGER, a dummy for whether the individual was the branch manager;STAFF, the number of staff at the branch; the interaction of MANAGER andSTAFF; and a vector of location dummy variables (London, Wales, Cheshire,Derbyshire, Lancashire, Merseyside, and Yorkshirethe omitted category isGreater Manchester).

The effect of female employment on different types of male clerks is identifiedby the difference-in-difference, which is given by POST1914, AVGFEM, and thetwo interaction terms. The logic of the approach is as follows. The variableAVGFEM acts as a proxy for the nature of work at the different branches. Theunderlying assumption is that the nature of work at individual branches remainedessentially unchanged over time, and thus each branch had a fairly constantcomposition of positions throughout the period of this study. Because women wereassigned exclusively to back office positions, a constant composition of positionswould imply that the extent to which individual branches were feminizable wasalso essentially constant over time. Appendix I examines this assumption in detail.The time dummy, POST1914, identifies all time-specific effects, including thoseunrelated to female employment, such as the unionization of the banking labourforce, improvements in technology, changes in the male labour supply due to theFirst World War, the decline of the Lancashire textile industry, and broaderchanges in the British economy. Because the time dummy potentially captures theeffect of several different factors, there are no prior expectations for the sign of itscoefficient, b2. In addition, its coefficient, b2, is not interpreted as an effect of

61 Mincer, Investments, pp. 2912; idem, Schooling, pp. 723, 8396.62 The data do not contain any information on education or previous employment. However, this is much less

of a concern for this article than would normally be the case with Mincer-type wage regressions because newentrants to the banking sector were remarkably homogeneous with respect to their educational background.Moreover, to the extent that there were differences in the level and quality of education or prior experience, thesewould be captured in the regressions by the individual fixed effects.



feminization; however, any global labour supply effects of feminization will besubsumed within this coefficient.

The difference-in-difference is given by the two interaction variables.The inter-action AVGFEM*POST1914 identifies an effect of female employment that is thesame across all male staff at a given type of branch.The coefficient on this variablecan be thought of as the effect of being in a female-intensive branch after 1915,controlling for both period and the inherent feminizability of the branch. Theinteraction AVGFEM*POST1914*TENURE identifies the extent to which thiseffect differed across men with different levels of tenure. As noted previously,younger men often held the same positions as women and thus it is likely thatfemale employment reduced their earnings through the standard labour supplyeffect.Thus one would expect a negative value for the coefficient b3. On the otherhand, because women faced glass ceilings, they did not compete directly withsenior men for positions of responsibility. Consequently, one would expect that theeffect of feminization would be less for senior men than junior men, and thus apositive value for the coefficient b4.63

Table 2 shows summary statistics for the regression variables over the periods18901914, 191519, 192041, and 18901941. The expected signs on thecontrol variables can be drawn from the basic model of human capital. Pay in theclerical sector was strongly attached to tenure and one would expect salaries toincrease at a decreasing rate with tenure (that is, that the net effect of TENURE,TENURE2,TENURE3, andTENURE4 is positive but decreases with tenure).Thehead office and larger branches had more staff in routine clerical positions than thesmaller branches, but also more staff in positions of responsibility such as divisionheads, and thus the signs on the coefficients for HEAD OFFICE and STAFF aretheoretically ambiguous. In line with tournament theory, one would expect a

63 There are two types of long-tenured male clerks in the data: those with insufficient talent to be promoted andthose in positions of responsibility, such as heads of divisions.Women were substitutes for the first type, but notthe second. The proportion of male clerks in positions of responsibility increased with tenure, and hence thecoefficient b4 is expected to be positive. However, the mix of the two types is not observable in the data and henceit is theoretically ambiguous whether the net effect of b3 and b4 will be positive or negative at any given tenure.

Table 2. Summary statistics

Mean(18901941)

Mean(18901914)

Mean(191519)

Mean(192041)

Ln(REAL WAGE) 4.81 (0.84) 4.76 (0.88) 4.28 (0.84) 4.93 (0.77)TENURE 15.18 (11.94) 13.86 (11.47) 16.23 (12.57) 15.83 (12.04)ENTRY AGE 17.68 (2.86) 18.23 (3.44) 17.48 (3.09) 17.37 (2.31)HEAD OFFICE 0.13 (0.34) 0.17 (0.38) 0.15 (0.36) 0.10 (0.30)LONDON 0.17 (0.38) 0.22 (0.42) 0.22 (0.41) 0.13 (0.34)INFLATION 2.05 (8.00) 0.80 (2.08) 16.10 (5.33) 0.36 (8.32)MANAGER 0.14 (0.34) 0.11 (0.31) 0.12 (0.32) 0.16 (0.36)STAFF 34.38 (39.78) 40.51 (38.68) 43.91 (46.41) 28.84 (38.29)MANAGER*STAFF 1.05 (6.09) 1.05 (6.62) 1.01 (6.56) 1.06 (5.64)POST1914 0.65 (0.48) 0.0 (0.00) 1.0 (0.00) 1.0 (0.00)AVGFEM 0.24 (0.10) 0.27 (0.08) 0.26 (0.08) 0.22 (0.11)POST1914*AVGFEM 0.15 (0.14) 0.0 (0.00) 0.26 (0.08) 0.22 (0.11)POST1914*AVGFEM *TENURE 2.46 (3.50) 0.0 (0.00) 4.37 (3.87) 3.66 (3.69)Sample size 34,976 12,171 3,425 19,380

Sources: RBSG Archive, Williams Deacons Bank Limited, staff registers; Feinstein, National income, tab. 65.

1054 ANDREW SELTZER


managerial premium that was increasing in branch size, and thus positive coeffi-cients on MANAGER and MANAGER*STAFF.64 Finally, the regional dummiesare included as controls for local labour market conditions, and there are no priorexpectations on their signs.

The regression results are shown in table 3. As the data contain multipleobservations of the same individuals, each of the regressions includes individualfixed effects. In order to provide sensitivity analysis, several regression specifica-tions are reported in the different columns. Columns 1 and 2 report the resultsfrom regressions covering the full sample period and excluding the 191519observations, respectively. Columns 3 and 4 show the results of regressions split bytime period (18901914 and 19151941).The control variables are the same as inthe first two specifications, with the obvious exception of omitting POST1914(and correspondingly changing the interaction variables). These regressions serveas a check for the robustness of the first set of regression results. Column 5 showsthe results of a regression on the 18901914 data which includes a time trend andan interaction between the time trend and AVGFEM. This regression is includedto test for the existence of pre-trends, in other words whether the difference-in-difference is measuring something other than the effects of feminization.

Each of the regressions is strongly significant overall, the signs of thecoefficients for the control variables are generally in line with the predictions of

64 See Rosen, Prizes and incentives, pp. 7027, and Lazear and Rosen, Rank-order tournaments, pp. 8447,on tournament models.

Table 3. Determinants of log real salary

SampleALL

18901941ExWWI

18901941ALL

18901914ALL

191541ALL

18901914

TENURE 0.22* (178.1) 0.22* (177.9) 0.24* (128.3) 0.20* (157.7) 0.16* (21.3)TENURE2*100 -0.94* (81.8) -0.96* (82.8) -1.24* (70.8) -0.78* (66.1) -1.22* (70.8)TENURE3*1000 0.20* (52.7) 0.21* (53.8) 0.29* (48.2) 0.17* (42.1) 0.29* (48.0)TENURE4*10000 -0.017* (39.5) -0.017* (40.6) -0.025* (37.0) -0.014* (31.5) -0.25* (36.8)HEAD OFFICE 0.019 (1.9) 0.029* (2.8) 0.032 (2.2) -0.10* (8.9) 0.041* (2.7)INFLATION -0.014* (81.2) -0.014* (76.8) -0.006* (7.5) -0.012* (79.3) -0.006* (7.7)MANAGER 0.20* (33.1) 0.20* (33.9) 0.08* (7.4) 0.13* (19.1) 0.08* (7.5)STAFF 0.0007* (8.4) 0.0007* (7.0) 0.0005* (3.5) 0.0014* (14.7) 0.0005* (3.2)MANAGER*STAFF 0.009* (31.2) 0.009* (32.9) 0.007* (16.3) 0.004* (12.1) 0.007* (16.4)POST1914 -0.19* (17.0) -0.18* (15.6)AVGFEM -0.05 (1.5) -0.03 (0.9) -0.061 (1.1) -0.079* (3.2) -0.23* (3.4)POST1914*AVGFEM -0.15* (3.5) -0.22* (5.0)POST1914*AVGFEM

*TENURE0.019* (19.3) 0.024* (23.2)

AVGFEM*TENURE 0.009* (3.4) 0.010* (7.7)WWI -0.25* (50.1) -0.29* (52.8)TIMETREND 0.07* (9.3)TIMETREND*

AVGFEM0.02* (5.7)

Constant 3.36* (292.4) 3.37* (286.0) 3.34* (190.6) 3.26* (369.2) 3.40* (163.7)R2 0.900 0.903 0.857 0.928 0.858F 13,993.1* 13,850.5* 3,899.7* 14,378.6* 3,730.0*Sample size 34,976 31,541 12,171 22,805 12,171

Notes: All regressions include dummy variables for location (London,Wales, Cheshire, Derbyshire, Lancashire, Merseyside, andYorkshire). The coefficients on these variables are not reported due to space constraints, but are available from the author uponrequest. All regressions include individual fixed effects. t-statistics in parentheses. * Significant at a 1% level.Sources: As for tab. 2.



the human capital model, and most of the coefficients are statistically signifi-cant and robust to specification. The most important results for the purposesof this article concern the two interaction terms POST1914*AVGFEM andPOST1914*AVGFEM*TENURE from the first two regression specifications.65

The coefficients on both of these variables have the expected signs and are stronglysignificant.The coefficients from the first column imply that, relative to the earlierperiod, between 1915 and 1941 a newly hired clerk at a branch with no womenearned approximately 4.5 per cent more than an otherwise similar clerk at a branchwhich was 30 per cent female (about the 66th percentile of the distribution).66

However, the regressions also imply that the earnings of clerks with nine years ofservice would have been about the same at the two branches and clerks with 20years of service at a branch that was 30 per cent female would have earnedapproximately 6 per cent more than their counterparts at an all-male branch.These values are relatively large and show that women were substitutes for juniormen but were complements to senior men. As noted previously, these estimatesonly capture the relative effect of feminization across different types of branches.There could also be a global effect for all men independent of branch size that issubsumed within the estimated coefficient on the time period dummy. While it isnot possible to determine the actual size of this effect (as the time period dummycaptures all time-specific factors influencing salaries, including those unrelated tofemale employment), as a benchmark it has been calculated that, assuming that40 per cent of the total period effect was due to feminization and the remainder toother factors, then female employment would have reduced the salary of a newmale entrant by about 11 per cent and had approximately zero net effect for a maleclerk who spent a 40-year career in branches that were 30 per cent female.67 Thesefigures are broadly consistent with qualitative evidence from contemporary maleclerks. The sizable estimated effect of female employment on junior salaries isconsistent with the views expressed by juniors and returning veterans that theywere being crowding out by women.68 However, as the number of junior hiresbegan to decline and the returning veterans became more established, the issue ofwomen undercutting male salaries began to disappear from the union journals anddiscussions. By the late 1920s, letters from even modestly senior clerks aboutfemale employment ceased to appear in the Bank Officer and female employmentbecame a much less prominent discussion topic at the BOG meetings.

Another important regression result concerns MANAGER andMANAGER*STAFF.The coefficients from the first column imply that a managerof a branch with two staff earned about 21.8 per cent more than an otherwise

65 To check the robustness of the results to the measure of feminizability, I ran regressions using STAFF as aproxy for feminizability instead of AVGFEM. The logic of this proxy is that women were typically assigned toroutine duties in the back offices of the larger branches, and thus STAFF and feminizability are likely to bestrongly positively correlated. These regressions exclude AVGFEM and its interactions, but include interactionsbetween STAFF and POST1914 and between STAFF, POST1914, and TENURE.The regressions also includethe same control variables as in tab. 3.The results of these regressions are remarkably similar to those presentedin the article, reinforcing the interpretation of the results presented. A full set of these regressions is available uponrequest from the author.

66 Formally, this is calculated as Y X= . The gap above is given by -0.15(0.30) + 0.019(0*.30*0) = -0.45.67 Formally, this is done by summing the estimated effects from the regression for each level of tenure between

0 and 40.68 A junior, Another view (see above, n. 7); One of Byngs Boys, Correspondence, Bank Officer, 14 (1920),

p. 10.

1056 ANDREW SELTZER


similar clerk, whereas a manager at a branch with 10 staff earned about 29 per centmore than an otherwise similar clerk. Approximately one-quarter of all male staffreached the level of branch manager, and, as shown in the previous section, atalented man was likely to reach the level of manager perhaps six to eight yearsfaster as a result of female employment. The magnitude of these coefficientsimplies that, for talented junior staff, the salary boost associated with fasterpromotion would have greatly outweighed the negative consequences of directcompetition from women early in the career.

As a check for the robustness of the results in columns 1 and 2, columns 3 and4 show the results of regressions which are split by time period (18901914 and19151941). The control variables are the same as in the first two specifica-tions, with the obvious exception of omitting POST1914 (and correspondinglychanging the interaction variables). The coefficients on AVGFEM andAVGFEM*TENURE are somewhat smaller than for the first two regressionspecifications, but are nonetheless broadly consistent with the other results. Thecoefficient on AVGFEM is statistically insignificant for the earlier period, butsignificant at a 1 per cent level for the later period. Both coefficients are slightlylarger in absolute value for the post-1914 period than for the earlier period.Thesecoefficients imply that juniors were paid relatively more in the more feminizablebranches in the earlier period, whereas senior staff were paid relatively more in themore feminizable branches in the later period. However, the estimated effects aresmall. The estimated relative salary decline for a newly hired clerk in a branchwhich was 30 per cent female (compared to a similar clerk in a branch with nowomen) was about 0.5 per cent from 18901914 to 191541. By contrast, therelative salary of a clerk with 30 years of service in a branch that was 30 per centfemale would have increased by about 0.5 per cent from 18901914 to 191541.

One caveat of the difference-in-difference technique is that it is possible that theestimated outcome in the post-1914 period was simply a continuation of pre-existing trends from the earlier period. The existence of a pre-trend would implythat the correlation between wages and feminization actually measures the effectsof something else, such as mechanization of the more routine tasks, which beganbefore the First World War and continued through the 1920s. In the context of thecurrent analysis, there is a fundamental assumption that there was not a greaterrelative decline in wages at the more feminizable branches prior to the actualemployment of women. To test this possibility, column 5 in table 3 shows theresults of a regression using the 18901914 data. The control variables are thesame as in the other regressions, and the effects of a possible pre-trend arecaptured by the inclusion of a time trend (PRETREND) and the interaction ofPRETREND and AVGFEM.The coefficient on the interaction variable is positiveand statistically significant at a 1 per cent level.Thus relative wages at the female-intensive branches were increasing prior to the actual employment of women,suggesting that mechanization prior to feminization led to increasing relativesalaries at the more feminizable branches, and thus the regression results may, infact, underestimate the negative impact of female employment on male salaries.

V

The expansion of female employment along with immigration created the largestsupply shock to the late nineteenth- and early twentieth-century British labour



market. Contemporaries were divided on the impact of female employment in theclerical sector on male salaries and careers. Rank-and-file male clerks perceivedwomen to be a considerable threat to their own positions and used their tradeunions to lobby against female employment and later for equality of wages. Femalestaff and employers argued that women typically did very different jobs than men,and thus posed little direct competition to them, particularly for high-pay, high-prestige positions.

Using an extremely rich micro-data set from Williams Deacons Bank coveringvirtually all staff at the bank between 1890 and 1941, this article finds somesubstance in both of these seemingly contradictory arguments. Female employ-ment had different effects across different types of male staff and thus had impor-tant consequences for the distribution of mens income in the clerical sector.Women typically held the same jobs as junior men, and the evidence shows thatafter 1915 the salaries of junior men in the more female-intensive branchesdeclined by about 5 per cent relative to otherwise similar junior men at exclusivelymale branches. However, women were complementary inputs to older male staff.The salary effect was approximately zero for staff with about nine years tenure andwas positive for older staff. In addition, female employment also increased pro-motion opportunities for male clerks. The employment of women was necessaryfor the expansion of the branch network, which created new managerial openingsand increased the proportion of men in managerial positions. Male careers becamemore dependent on individual ability as a result. The most talented men werepromoted earlier because managerial positions opened more regularly as the bankgrew larger. Moreover, older, marginally talented men also had more opportunitiesto advance.

Date submitted 23 February 2012Revised version submitted 22 June 2012Accepted 17 August 2012

DOI: 10.1111/j.1468-0289.2012.00678.x

Footnote referencesAllman, H. A., Williams Deacons, 17711970 (Manchester, 1971).Anderson, G., Victorian clerks (Manchester, 1976).Anderson, G., The white blouse revolution, in G. Anderson, ed., The white blouse revolution: female office workers

since 1870 (Manchester, 1988), pp. 126.Blackburn, R. M., Union character and social class: a study of white-collar unionism (1967).Borjas, G. J., The economic analysis of immigration, in O. C. Ashenfelter and D. Card, eds., Handbook of labor

economics (Amsterdam, 1999), pp. 1697760.Burnette, J., Gender, work and wages in industrial revolution Britain (New York, 2008).Card, D., Immigrant inflows, native outflows, and the local labor market impacts of higher immigration, Journal

of Labor Economics, 19 (2001), pp. 2264.Cohn, S., The process of occupational sex-typing: the feminization of clerical labor in Great Britain (Philadelphia, Pa.,

1985).Dohrn, S., Pioneers in a dead-end profession: the first women clerks in banks and insurance companies, in G.

Anderson, ed., The white blouse revolution: female office workers since 1870 (Manchester, 1988), pp. 4866.Feinstein, C. H., National income, expenditure and output of the United Kingdom, 18551965 (Cambridge, 1972).Goldin, C. D., Understanding the gender gap: an economic history of American women (New York, 1990).Heller, M., London clerical workers, 18801914 (2011).Jordan, E., The lady clerks at the Prudential: the beginning of vertical segregation by sex in clerical work in

nineteenth-century Britain, Gender and History, 8 (1996), pp. 6581.Klingender, F. D., The condition of clerical labour in Britain (1935).

1058 ANDREW SELTZER


Lazear, E. P. and Rosen, S., Rank-order tournaments as optimum labor contracts, Journal of Political Economy,89 (1981), pp. 84164.

Lewis, J. E., Women clerical workers in the late nineteenth and early twentieth centuries, in G. Anderson, ed.,The white blouse revolution: female office workers since 1870 (Manchester, 1988), pp. 2747.

Liladhar, J. and Kerslake, E., No more library classes for Catherine: marital status, career progression and libraryemployment in 1950s England, Womens Studies International Forum, 22 (1999), pp. 21524.

Lockwood, F. D., The black-coated worker: a study in class consciousness (1958).Mincer, J. A., Investments in human capital and personal income distribution, Journal of Political Economy, LXVI

(1958), pp. 281302.Mincer, J. A., Schooling, experience, and earnings (Cambridge, Mass., 1974).Oram, A., Women teachers and feminist politics, 190039 (Manchester, 1996).Parris, H., Staff relations in the civil service: fifty years ofWhitleyism (1973).Rae, G., The country banker: his clients, cares, and work from an experience of forty years (7th edn. 1930).Rathbone, E. F., The remuneration of womens services, Economic Journal, 27 (1917), pp. 5568.Reich, M., Gordon, D. M., and Edwards, R. C., Dual labor markets: a theory of labor market segmentation,

American Economic Review, 63 (1973), pp. 35965.Rosen, S., Prizes and incentives in elimination tournaments, American Economic Review, 76 (1986), pp. 70115.Seltzer, A. J., Did firms cut nominal wages in a deflationary environment?: micro-level evidence from the late

19th and early 20th century banking industry, Explorations in Economic History, 47 (2010), pp. 11225.Seltzer, A. J., Salaries and promotion opportunities in the English banking industry, 18901936, Business

History, 52 (2010), pp. 73759.Seltzer, A. J., Female salaries and careers in British banking, 191541, Explorations in Economic History, 48

(2011), pp. 46177.Seltzer, A. J. and Frank, J., Promotion tournaments and white collar careers: evidence from Williams Deacons

Bank, 18901941, Oxford Economic Papers, 59, S1 (2007), pp. i4972.Sheppard, D. K., The growth and role of UK financial institutions, 18801962 (1971).Takahashi, A., Unrealised golden opportunities: the development of female clerical work, 18701939 (unpub.

M.A. thesis, Royal Holloway, Univ. of London, 1994).Wardley, P., The commercial banking industry and its part in the emergence and consolidation of the corporate

economy in Britain before 1940, Journal of Industrial History, 3, 2 (2000), pp. 7197.Wardley, P., Women, mechanization and cost-savings in twentieth century British banks and other finan-

cial institutions, paper presented at the International Economic History Association XIV Conference,Helsinki (2006) http://www.google.co.uk/url?sa=t&rct=j&q=&esrc=s&source=web&cd=2&ved=0CD0QFjAB&url=http%3A%2F%2Fwww2.uwe.ac.uk%2Ffaculties%2FCAHE%2FHPP%2Fstaff%2Fstafflist%2Fstaff_pwardley_banking.pdf&ei=K9ECUefXKKq40QXHhICwDw&usg=AFQjCNEggFkkN1pzG0w2FoE1a6HE84VCsQ&bvm=bv.41524429,d.d2k (accessed on 5 Feb. 2013),

Webb, S. J., The alleged difference in the wages paid to men and women for similar work, Economic Journal, 1(1891), pp. 63562.

Wilson, R. G., Disillusionment or new opportunities? The changing nature of work in offices, Glasgow 18801914(Aldershot, 1998).

Zimmeck, M., Jobs for the girls: the expansion of clerical work for women, 18501914, in A.V. John, ed., Unequalopportunities: womens employment in England, 18001918 (Oxford, 1986), pp. 15277.

APPENDIX I: MEASURING FEMINIZABILITY

The identification strategy used in the econometric analysis in this article rests heavily onthe assumption that the feminizability of the individual branches was essentially constantover time and is reasonably measured by the variable AVGFEM. This appendix examinesthese assumptions.

The basis for the assumption of essentially constant branch-level feminizability is thatmost women were restricted to routine clerical tasks and the distribution of the differenttypes of tasks was fairly constant at the branch level over time. It thus follows that womenwould be more likely to be assigned to branches which had large back offices and a largevolume of clerical work.This definition of feminizability does not rely on the employmentof women at a given point in time; rather it is essentially a characteristic of the nature ofwork at individual branches. If the mix of tasks was constant over time at a particularbranch, then so too was the inherent feminizability of the branch.

The data do not provide any direct evidence about the mix of work being done at theindividual branches, so it is necessary to measure feminizability in a different way. Themeasure used is the average percentage of female staff at each branch between 1919 and



1936.The period 191518 is not used in constructing the variable because women duringthis period were employed as replacements for men in the services, rather than as clerks forroutine tasks. It is likely that the percentage of female staff at a given branch during theseyears was more a characteristic of the age distribution of the male staff (as younger menwere more likely to enlist) than the nature of the work at the branch. In addition, the period193741 is not used in constructing the variable, as there are no data on the Londonbranches or the Head Office for these years.

The use of AVGFEM to measure feminizability in the regressions can only be justifiedif, first, the annual proportion of women at the branch level (PERFEM) was fairly constantover the period 191936 and, second, the variation of PERFEM across branches wasdriven by branch-level characteristics that were similar before and after 1915.69 To addressthe assumption that it was fairly constant over time at the branch level, table A1 shows amatrix of correlation coefficients for PERFEM over the period 191936.The average of the153 correlation coefficients is 0.664 and every correlation is positive and significant at a 1per cent level. The correlations decline somewhat as the time between observationsincreases; however, even comparing 1919 to 1936 there is a strong positive correlation. Aregression of PERFEM has been run on a series of dummy variables for branch. Theadjusted R2 for the regression is 0.526, confirming that branch alone explains much of thevariation of PERFEM. This evidence strongly suggests that PERFEM is, at least in part,picking up branch-level characteristics that changed relatively little over time.

To test the assumption that variation in feminizability across branches was largely drivenby time-invariant branch characteristics, regressions of PERFEM were run on a vector ofbranch characteristics using annual data from the period 191536. The independentvariables are the available branch characteristics: total employment at the branch and itssquare, age of the branch, a dummy for London, and the number of accounts. Table A2shows the regression results for a specification with just these variables, a specification withthese variables and year dummies, and a specification including year and branch dummies(and excluding the London dummy). The results show that much of the variation infeminization can be explained by a few observable branch characteristics. In addition, thestrong significance of the year dummies suggests that much of the variation occurring overtime within individual branches can be explained by changes in the banks overall policiesconcerning female employment, rather than year-to-year changes at the branch level.Theseregressions show that feminization during the period 191536 was driven by branchcharacteristics, but not whether these characteristics were similar prior and subsequent to1915. To test whether these characteristics were relatively constant over the entire periodof this study, I calculated the predicted values from the second regression specification forall years between 1891 and 1936. I then calculated the correlation coefficients of thepredicted values from each year from 1891 to 1914 with each year from 1915 to 1936.Thecorrelations ranged from 0.49 to 0.99 and averaged 0.94, strongly implying that feminiz-ability was fairly constant over time within individual branches and that it is unlikely thatthere was much change in the feminizability of the branches over time.

69 Note that AVGFEM is simply the average of PERFEM over the period 191936. There was considerablevariation in feminization across branches over the period in which the bank employed women. Between 1915 and1941, about 43% of annual observations of branches have no women. At the other end of the scale, during thewar years women comprised as much as 80% of some branches.

1060 ANDREW SELTZER


Tab

leA

1.C

orre

latio

nm

atri

xfo

rP

ER

FE

M

Year

1919

1920

1921

1922

1923

1924

1925

1926

1927

1928

1929

1930

1931

1932

1933

1934

1935

1936

1919

1.00

1920

0.76

1.00

1921

0.71

0.88

1.00

1922

0.66

0.82

0.93

1.00

1923

0.68

0.77

0.87

0.92

1.00

1924

0.63

0.72

0.80

0.86

0.93

1.00

1925

0.62

0.70

0.75

0.79

0.87

0.92

1.00

1926

0.57

0.63

0.66

0.71

0.78

0.82

0.91

1.00

1927

0.45

0.54

0.57

0.62

0.72

0.77

0.86

0.93

1.00

1928

0.44

0.57

0.56

0.63

0.70

0.74

0.82

0.90

0.96

1.00

1929

0.43

0.49

0.50

0.56

0.63

0.63

0.70

0.79

0.85

0.89

1.00

1930

0.44

0.50

0.51

0.56

0.64

0.63

0.69

0.76

0.81

0.85

0.93

1.00

1931

0.46

0.52

0.51

0.57

0.64

0.61

0.70

0.75

0.78

0.80

0.91

0.96

1.00

1932

0.44

0.45

0.44

0.51

0.55

0.56

0.62

0.67

0.67

0.69

0.78

0.84

0.87

1.00

1933

0.48

0.50

0.50

0.56

0.62

0.63

0.69

0.71

0.71

0.70

0.75

0.79

0.81

0.92

1.00

1934

0.43

0.46

0.47

0.53

0.58

0.60

0.68

0.69

0.68

0.67

0.70

0.77

0.79

0.88

0.92

1.00

1935

0.37

0.35

0.37

0.45

0.48

0.49

0.55

0.55

0.53

0.55

0.59

0.63

0.65

0.76

0.78

0.85

1.00

1936

0.38

0.36

0.37

0.45

0.48

0.50

0.59

0.58

0.54

0.57

0.58

0.64

0.66

0.73

0.75

0.80

0.92

1.00

Sou

rce:

RB

SG

Arc

hive

,Will

iam

sD

eaco

ns

Ban

kL

imit

ed,

staf

fre

gist

ers.



Table A2. The determinants of PERFEM

1 2 3

STAFF 0.012* (9.54) 0.004* (3.45) 0.016* (9.33)STAFF2 -0.00006* (9.22) -0.00002* (3.87) -0.00006* (5.79)BRANCH AGE 0.0018* (9.76) 0.0022* (13.28) -0.0004 (0.48)LONDON 0.104* (5.96) 0.107* (7.07)ACCOUNTS -0.00004* (5.33) -0.000005 (0.64) -.000008 (0.63)Constant 0.072* (14.52) 0.057* (4.61) -0.030 (0.43)Year dummies NO YES YESBranch dummies NO NO YESF 126.74* 64.86* 35.56*Adjusted R2 0.195 0.390 0.692N 2,598 2,598 2,598

Notes: t-statistics in parentheses. * Significant at a 1% level.Sources: RBSG Archive, Williams Deacons Bank Limited, staff registers and particulars of branches.

1062 ANDREW SELTZER


Documents

econmis