Swedsh Insttute for Socal Research (SOFI) Stockholm Unversty WORKING PAPER 7/2007 ARE WOMEN ASKING FOR LOW WAGES? GENDER DIFFERENCES IN WAGE BARGAINING STRATEGIES AND ENSUING BARGAINING SUCCESS by Jenny Säve-Söderbergh
Are Women Askng for Low Wages? Gender Dfferences n Wage Barganng Strateges and Ensung Barganng Success * Revsed May 2007 Jenny Säve-Söderbergh The Swedsh Insttute for Socal Research Unverstetsvägen 10 E Stockholm Unversty, Sweden. Emal: Jenny.Save-Soderbergh@sof.su.se Abstract Men and women s labor market outcomes dffer along pay, promoton and compettveness. Ths paper contrbutes by uncoverng results n a related unexplored feld usng unque data on ndvdual wage barganng. We fnd strkng gender dfferences. Women, lke men, also bargan, but they submt lower wage bds and are offered lower wages than men. The adjusted gender wage gap s lower wth postedwage jobs than wth ndvdual barganng, although less s ascrbable to the term assocated wth dscrmnaton. Both women and men use self-promotng, or compettve barganng strateges, but women self-promote at lower levels. Employers reward self-promoton but the larger the self-promoton, the larger s the gender gap n barganng success. Women therefore lack the ncentves to self-promote, whch helps to explan the gender dspartes. Keywords: Indvdual Wage Barganng, Compettveness, Barganng strateges, Self-promotng Barganng Strateges, Gender Wage Gap, and Dscrmnaton. JEL codes: M51, M52, J31, J16, * I would lke to thank Anders Björklund, Nabanta Datta Gupta, Lena Granqvst, Carolne Hoxby, Ethan Kaplan, Helena Persson, Mkael Prks, Åsa Rosén, Annka Sundén, Eskl Wadensjö and semnar partcpants at the IZA Summer School 2000, the Swedsh Insttute for Socal Research, the Department of Economcs at Stockholm Unversty, the Congress of the European Economc Assocaton n Stockholm and the 15 th annual EALE conference n Sevlle for valuable comments. I am also very grateful to Jusek for provdng the data. The usual dsclamer apples.
1 Introducton The labor market outcomes of men and women dffer along a varety of dmensons. Even after controllng for a broad range of demographc and background characterstcs, women earn sgnfcantly less than men do (see e.g. Altonj and Black 2000; Blau & Kahn, 2003, for an nternatonal comparson; Edn & Rchardson, 2002, and Albrecht et al, 2003, for Sweden 1 ). Women are also less lkely to have top-rankng postons (Bertrand & Hallock, 2001) and to have lower career moblty (Granqvst & Persson, 1999). Standard economc explanatons for such occupatonal dfferences nclude preferences, dscrmnaton and ablty. Common explanatons for women havng preferences dfferent from men, and thereby choosng to enter low-payng jobs, often emphasze women s role n the famly, whch affects ther human captal nvestment and career choces. Another common explanaton s dscrmnaton or antcpated dscrmnaton lmtng women s labor market opportuntes. Fnally, some argue that men and women dffer along unobservable dmensons of sklls, where these unobservable dfferences develop nto dsparate labor outcomes. In ths paper, several of these ssues are addressed by nvestgatng gender dfferences on the labor market from a new perspectve; that of ndvdual wage barganng. In partcular, we examne whether men and women dffer n the barganng strateges they use and whether the ensung barganng success, or the pay-off to a strategy, dffers by gender. Emprcal evdence on gender dfferences n real wage barganng s lmted. 2 The study therefore contrbutes by uncoverng results n an unexplored feld. 3 Nonetheless, ndvdual 1 As n most countres, the average raw gender wage gap n Sweden has narrowed snce the 1960s. In 1968 the average raw gender wage gap was almost 30 percent whle the equvalent for 1991 was 20 percent. However, snce 1991 the raw gender wage gap has ncreased slghtly (Albrecht et al, 2003). These developments are manly drven by correspondng developments n the overall wage dstrbuton (Edn & Rchardsson, 2002). 2 One mportant excepton s Babcock (2003) who provdes nterestng results and anecdotes. 3 Many studes, theoretcal as well as emprcal, have focused on the barganng power of unons. When t comes to post-employment wage barganng there s a vast amount of lterature, especally n the feld of personnel economcs and wth partcular emphass on the employers concerns. 2
barganng over wages s one of the man components of the nteracton n the labor market. If gender dfferences n barganng strateges exsts and the ensung outcomes dffer, t may suggest an explanaton to why women are less represented n compettve workng envronments, less represented n hgh payng jobs, or are less lkely to compete for promoton. The data explored s a unque dataset from two Swedsh surveys conducted n 1999 and 2000, whch nclude recent graduates wthn the socal scences. The data contans of a homogenous populaton where all are young graduates wth short labor market careers. Ths lmts ssues of selecton due to career nterruptons for famly concerns, job changes, or partcpaton selecton. In addton, the surveys ncorporate mportant control varables for dfferences n labor market opportuntes, outsde optons and ndvdual attrbutes. In the survey, respondents were asked to report whether they were asked to state an explct wage bd at the tme of applcaton for the ntal job they got wthn ther feld of major, and f so, the level of the wage bd. Moreover, they were asked to state the offered wage, or n ths case, the equvalent wage they accepted. Some respondents reported not beng asked about a wage bd, and hence are assumed to have appled for a job wth posted wages. The data thus allows for comparsons usng the posted wage group as a control group. In order to establsh whether or not there are gender dfferences n barganng and n barganng outcomes, we focus on four ssues. The frst ssue analyzed s whether there are gender dfferences n the propensty to choose to apply for a job wth wage barganng as a part of the applcaton process, compared to choosng a job wth a posted wage. Prevous studes have found that women are less lkely than men to ntate negotatons (Babcock & Laschever 2003; Babcock et al 2006). Recent expermental research has also shown that women choose compettve pay-offs to a lesser extent than men, (Datta Gupta et al, 2006; Nederle & Vesterlund, 2005). Women may addtonally not choose to bargan f they are more rsk averse 3
than men, as suggested by some research (e.g. Janakoplos & Bernasek, 1998; Powell & Ansc, 1997; Sundén & Surette, 1998). In Datta Gupta et al (2006) one explanaton to women not choosng a compettve pay-off s a hgher rsk-averson. The second ssue nvestgated s whether men and women submt dfferent wage bds. Women may submt lower wages than men do f they face dfferent labor markets, and/or f they have dfferent outsde optons. Women could face a monopsonstc labor market, a hgher level of famly responsbltes, or they may be restrcted n choosng a job due to a husband-and-wfe coordnaton of job locatons, all causng a downward pressure on female wage bds. Yet, consderng that the average age for women n Sweden at ther frst marrage was 29.9 n 1998 and 30.4 n 1999, and the equvalent for men was 32.4 and 32.9 (Statstcs Sweden, 2000), such consderatons should not be too predomnant for the ndvduals n the sample where women (men) are 27 (28) years on average. Lkewse, the average age for females havng ther frst chld n Sweden was approxmately 28.5 years n both 1999 and 2000 (Statstcs Sweden, 2000). Therefore, gven the relatvely hgh age, along wth Sweden havng generous publc parental chld-care benefts (wth a share earmarked for the father), the choce of jobs should not be too dfferent between men and women n the sample. Thrd, we explore the use of self-promotng barganng strateges. Assume that employers beleve the wage bd reveals nformaton on a personal attrbute of the applcant, whch s not observable. The extent of overbddng a smlar applcant, that s, submttng a hgher wage bd than an applcant wth dentcal observable attrbutes who bargan for the same type of job, or underbddng a smlar applcant, then reveals the ndvdual s evaluaton of these personal attrbutes. Overbddng could thus be consdered as a self-promotng strategy or compettve strategy though assumed to be upperly bounded by a loss n credblty, and thereby by the loss of the job opportunty. 4
The self-promotng strategy s assumed to reflect a self-assessed evaluaton of the applcant s unobservable productvty. A lower value than a smlar applcant, and to some extent a hgher value, would thus sgnal a lower, or hgher, unobservable productvty. Ths s supported by emprcal evdence suggestng that underbddng (and to some extent overbddng) market wages have been nterpreted to be sgnals of lower (hgher) productvty, (see Agell & Lundborg, 1995 and 2003, for survey evdence; and Fehr & Falk, 1999, for expermental evdence). Alternatvely, a self-promotng strategy could reveal preferences on compettveness or be taken to reveal preferences on rsk-takng such that hgh-rsk takers are those who also dare to overbd more. Followng socologcal research we may expect gender dfferences n employng such strateges. A well-documented tendency n socologcal research (see Smth & Powell, 1990, for further references) s that people often self-enhance ; they postvely dfferentate between ther own characterstcs or behavors relatve to the same characterstcs and behavors n other people. 4 Accordng to ths lne of research men tend to self-enhance to a greater extent than women do. Therefore men may have a hgher estmaton of ther personal ablty and thus would self-promote or overbd to a larger extent. 5 An alternatve explanaton s that men and women could be treated dfferently f they bargan. Bowles (2005) suggests that barganng poses a challenge for women because t calls for a type of domnatve masculne behavor (compettve asserton of one s self-nterest) that 4 Another explanaton could of course be that men and women have an objectve dfference n ablty, such that women have worse unversty performance and/or lower grades. As a matter of fact, the emprcal evdence suggests the opposte. In a report on all Swedsh unversty graduates for 1995/1996, t s found that women had a hgher performance score (number of credts acheved per semester) of 85 percent whle the equvalent for men were 80 percent. If only lookng at the socal scence majors, the gender dfference was 6 percent. Moreover, 18 percent of the women compared to 12 percent of the men were found n the upper end of the hgh-school grade dstrbuton, wth grade averages hgher than 4.0 (on a scale between 1.0 to 5.0), Statstcs Sweden and the Natonal Board for Hgher Educaton (1998). 5 Parallelng these fndngs regardng a dfference n subjectve ablty ratng, females appear to expect less n terms of wage gans due to possesson of a college degree (Brunello et al, 2004). Female students expect both sgnfcantly 5
contradcts the prescrptve norms of femnne behavor. The study also found that women who choose to bargan where evaluated as less nce and napproprately demandng (but not as less competent), whle ths was only true for males f barganng wth a male evaluator. Fourth, and fnally, we want to assess f men and women are equally successful gven the same self-promotng strategy, all observable characterstcs equal. Psychologcal lterature has found that women tend to be more cooperatve n barganng than men (see Walters et al, 1998, for a meta-analyss on the ssue), although the dfference s slght. If women are more cooperatve ether by nature or through socalzaton, or f they are perceved as beng more cooperatve, then they may not be equally rewarded, even for the same barganng strategy. Holm (2000) also fnds that both men and women tended to behave sgnfcantly more hawkshly (or non-cooperatvely) towards women when barganng, although ths knd of dscrmnaton aganst women worked as a coordnaton devce that boosted the earnngs of both sexes. 6 Agell & Bennmarker (2002) also found that frms wth a large share of female employees were less lkely to beleve that employees, who were unhappy about ther pay, would respond by reducng ther own efforts. Thus, even f women would enjoy a lower return from ther barganng, and would thus be less successful, employers may feel they wll not respond by reducng ther efforts to the same extent that men would. 7 lower startng salares after graduaton and beleve ther prospects to be worse, even f they rank themselves as hghly as the equvalent male rates hmself (see also Betts 1996). 6 Hultn & Szulkn (1999) also found that gender wage dfferentals are affected by the gender composton of an establshment s manageral staff. The negatve effect on women s wages due to a hgh male representaton among managers and supervsors was three tmes as strong wthn organzatons wth a hghly decentralzed wage-settng process than t was for females n general. The data n ths paper, however, reveals no nformaton on the sex of the employer. 7 In an effcency-wage experment, Schweren (2002) however, found that the average wage asked for by men and women was the equvalent but that women were offered lower wages after barganng. Moreover, men recprocated by makng more effort for hgh wage offers than women dd, but ths was manly a consequence of women not beng offered wages n the same range. 6
The emprcal analyss of the four ssues reveals strkng fndngs. We frst establsh that women actually choose a job where barganng s nvolved to the same extent as men do. Ths result thus dffer from prevous fndngs on women beng more prone to avod stuatons wth compettve pay-offs. Second, women consstently submt lower wage bds than men do, also when controllng for mportant job and ndvdual attrbutes. Thrd, as found by many prevous studes, women receve lower wages than men do. Interestngly, although wages are hgher for both men and women who have barganed, ex post, barganng for wages results n a hgher unadjusted gender wage gap by one percentage pont, wth the gap beng 0.95 for those applyng for posted wage and 0.94 for those who chose to bargan. Also after adjustng the wage gap to account for mportant job and ndvdual attrbutes, the dfference between the groups reman at one percentage pont. Yet, n a wage-decomposton (Blnder 1973; Oaxaca 1973), we fnd that the term commonly ascrbed to dscrmnaton explans less among those who barganed, 55 percent of the gender wage gap, whle for those who chose posted wages, the smlar term explan 68 percent of the wage gap. Thus, a falure to acknowledge ndvdual wage barganng may overestmate the gender wage gap typcally ascrbed to wage dscrmnaton. The use of self-promotng strateges also dffers by gender. Women do not necessarly use self-promotng strateges less often than smlar men do, but they do so at lower levels. Restrctng the sample nto those who strctly overbd a smlar canddate, women overbd by 9.4 percent, compared to men who overbd by 12.9 percent, on average. In general, self-promotng strateges are rewarded by employers. Overbddng has a postve effect on the barganng success, that s, the amount addtonally offered to the applcant relatve to a smlar applcant. Women, however, receve a lower pay-off relatve to men from the same self-promotng strategy. The results may parallel the fndngs on recprocal wage-settng 7
found n Agell & Bennmarker (2002). But quantle regresson estmates reveal that gender dfferences do not exst n the 25 th quantle of barganng success, and that t s n partcular n the 75 th quantle where gender dfferences are largest. Consequently, n a barganng stuaton where the scope for mprovng the wage offer by a self-promotng strategy s hgh, women are less successful than men are from beng self-promotng. In summary, the lower pay-off to overbddng thus creates weaker ncentves for women to employ self-promotng strateges. Even though women choose to self-promote, ther strateges are ether consdered as less credble compared to smlar males strateges, or there s a glasscelng on the value of a self-promotng or compettve strategy for women. Hence the result may explan why women refran from enterng compettve workng envronments, or are less represented n hgh-payng jobs. The paper s organzed as follows. Secton 2 descrbes the data. Secton 3 presents the emprcal model and Secton 4 the results. Secton 5 offers some concludng remarks. 2 Data The data set s derved from two surveys conducted n 1999 and 2000 by Jusek, 8 a Swedsh trade unon for whte-collar workers. The ndvduals n the surveys were unversty graduates as at September 1997-1998 and September 1998-1999 who had receved at least three years unversty educaton. They should further have majored n one of fve felds: law, busness admnstraton and economcs, computer and systems scence, personnel management or socal scence. Only ndvduals who fulfll the above requrement and who were born later than 1964 are ncluded n 8 The Swedsh Assocaton of Graduates n Law, Busness Admnstraton and Economcs, Computer and Systems Scence, Personnel Management and Socal Scences. 8
the survey sample. 9 The data set ncludes 1911 ndvduals from the 1999 survey and 2054 from the 2000 survey. Two wage varables are used n the analyss. 10 The frst s the applcant s wage bd, that s, the wage asked for f employment were to occur (to the employer they accepted workng for). The second wage varable s the offered wage (or startng salary) at the tme when employment begns. Both these are gven as monthly gross wages. 11 For those who work part tme the survey explctly states that the wage should be converted nto a full-tme monthly gross wage. For the analyss of ndvduals who have stated an explct wage bd the number of observatons s reduced to 2112 for the pooled sample, wth 1022 from 1999 and 1090 from 2000. In the pooled sample there are 57 percent women and 43 percent men. Table 1 presents summary statstcs for the wage data. The frst noteworthy fndng s that, dvdng the sample nto those who chose jobs where wage barganng was a part of the applcaton process, we fnd that man and women equally lkely to choose such jobs. The share of women who choose jobs wth barganng s 54 percent and the equvalent for men s 53 percent wth any statstcally sgnfcant dfference rejected n a t test. Ths s partcularly nterestng as Babcock & Laschever (2003) fnd that only 7 percent of women compared to 57 percent of men tred to mprove ther ntal wage offer by negotaton. Moreover, n Gneezy et al 9 The response rate for the 1999 survey was 63 percent of the total of 4000 graduates between September 1997 and September 1998, and n the 2000 survey the response rate was 63 percent out of 4500 graduates between September 1998 and September 1999. 10 A thrd wage measure, the current wage, s also avalable n the survey. As ths measure was gven drectly after the questons on the wage bd and the offered wage, the error of reportng the offered wage as the current wage should be mnmzed. 11 The gross monthly wages are used rather than an hourly wage measure. Even though the surveys nclude nformaton on hours worked per week, the measure s ambguous due to the formulaton of the queston, whereby the respondents cannot mark up exactly 40 hours worked per week. Ether they must choose work less than 40 hours or 41-45 hours per week. The excluson of the alternatve exactly 40 hours may thus have meant that respondents msclassfed ther workng hours. As ths s the only measure of part-tme work avalable n the data, ndvduals who work part tme wll be treated as f they work full tme. In addton, ndvduals who have stated a wage bd below SEK 10 000 are excluded (3 ndvduals), snce the wage bd s more lkely to refer to part-tme than to full-tme employment. 9
(2003) & Datta Gupta et al (2006) t s found that women choose a compettve pay-off to a lower extent than women do. The second noteworthy fndng s that women on average submt lower wage bds. The average wage bd for the pooled sample n gross monthly wages for women s SEK 18 215 (appr US $ 2600), whle the equvalent for men s SEK 19 228 (appr US $ 2750). Thus the unadjusted relatve wage bd for women s approxmately 94.7 percent of the same bd for men, and the dfference s statstcally sgnfcant at the 99 percent level. Furthermore, men have a larger spread n ther wage bds. These gender dfferences n wage bddng also apply for both years separately. The thrd fndng s that offered wages are also lower for women on average. For men, the average offered wage s SEK 18 628 (appr US $ 2665), whle for women t s SEK 17 517 (appr US $ 2502), wth a statstcally sgnfcant dfference at the 99 percent level. The unadjusted relatve wage offer for women s thus approxmately 94.0 percent. Comparng these fndngs wth those who chose jobs wth posted wages, we fnd a smlar dfference n men s and women s wage offers wth a statstcally sgnfcant dfference. However, the gender wage gap for applcants to jobs wth posted wages s lower. Consequently the prelmnary results pont toward an ncrease n the gender wage gap followng ndvdual wage barganng. Nonetheless, wage offers are consderably hgher both for men and women who chose to ndvdually bargan over ther wages compared to those choosng posted wage offers. Between men who barganed and men who dd not, wage offers are sgnfcantly hgher, also statstcally, for the former group. The same result s found wthn the female group. Hence, although gender wage dfferentals are hgher wthn the group who barganed, as compared wth those who dd not, wages are consderably hgher n the former group. Thus smlar to Datta Gupta et al (2006) 10
we fnd that women are more lkely to choose a compettve pay when the economc ncentves are hgher. However, as the fndng refers to the unadjusted wage dfferentals, several other factors could explan the hgher wages subsequent to barganng and the ncreasng gender wage dfferentals. 3 Emprcal Model 3.1. Adjusted Wage Gaps To establsh whether dfferences n the gender wage gap preval f we adjust for crucal wagegeneratng characterstcs a standard Mncer wage model s estmated where: W OFF 0 1 2 3 4 β 5 = δ + β FEMALE + β MAJOR + β EXPERI + β AGE + SECTOR + + β 2000 + 6 CONTRACT + β 7TRAINEE + β 8METRO + β 9S υ [1]. MAJOR represents the dfferent felds of major to ndcate the hghest educatonal level completed by the applcant. Prevous job experence, EXPERI, s also ncluded as job experence may ncrease productvty. In both of the surveys subjects were able to ndcate whether they had had any employment between graduaton and the current poston, thus coverng postgraduate work only. In the 2000 survey respondents could also ndcate whether or not ther work was qualfed or unqualfed (wth no further nterpretaton gven n the survey). In the pooled sample only those who marked qualfed experence are ncluded n the category qualfed work experence, whle those wth any work experence are pooled wth those who marked unqualfed work experence. Note that snce the type of job experence cannot be dstngushed, qualfed work experence wll be underestmated. To account for any addtonal labor market experence, the age, AGE, of the respondent s ncluded. Ths measure s ntended to capture not any specfc job-market experence but 11
experence n other job-related aspects, dmensons that are thought to ncrease productvty. As there s no nformaton on pre-graduaton work experence, the varable age may capture some of ths effect. Job- and employer-related characterstcs are also ncluded to control for dfferences n market opportuntes. Frst, wages may dffer between job sectors, SECTOR, and hence dummy varables are ncluded to control for applyng to the prvate, governmental or muncpal sector. 12 Furthermore, ndvduals wth temporary employment contracts (tral perod, project bass or substtute contracts), CONTRACT, are separated from ndvduals wth permanent employment contracts. As some ndvduals have appled for tranee postons, a dummy varable has been added, TRAINEE. Because a job locaton n a metropoltan area could represent a more flexble labor market and thereby ncrease the barganng power of the applcant we dstngush between a job locaton n a metropoltan and non-metropoltan area, survey effects a survey dummy, to fulfll the usual crtera of ndependence. METRO. To account for any year or S 2000 s ncluded. Fnally υ s the error component, assumed To analyze the adjusted wage gap n wage bds, a smlar Mncer model to equaton (1) for the wage bds s estmated, wth an addtonal proxy for the applcant s outsde opton. The estmated model s then: W BID 0 1 2 3 4 β 5 = δ + β FEMALE + β MAJOR + β EXPERI + β AGE + SECTOR + + 10 β 6CONTRACT + β 7TRAINEE + β 8METRO + β 9S 2000 + β SEARCH TIME + ε [2] 12 Some ndvduals have reported workng n a sector denoted other and are dropped (45 ndvduals) snce the survey lacks any nformaton as to what s defned as other. Indvduals workng outsde Sweden are also dropped (25 workng n Nordc countres, 37 n other parts of Europe and 12 outsde Europe) along wth those beng selfemployed (14 ndvduals), or who partcpate n labor market programs (67 ndvduals). 12
One proxy for outsde opton s the tme spent on searchng for the job, SEARCHTIME. Assumng unobservable productvty s postvely correlated wth the probablty of gettng a job, wage bds would declne wth the tme spent searchng for a job. 13 Another measure of the outsde opton conssts of the number of jobs appled for pror to acceptng the current employment, APPLIEDJOB S. Assume job utlty s hgher for the more able applcants, and therefore abler applcants would need to search for fewer jobs than an equvalent applcant wth lower unobserved ablty would have to. Ths varable s however only avalable for the 1999 survey. Table 2 dsplays the summary statstcs for the control varables. There are some gender dfferences worth notng. Men and women have majored n dfferent felds, wth more women among those majorng n socal scences and personnel management, whereas more men major n the feld of computer and system scence. 14 The same dfference s found for those who chose a posted wage job. Further, more men are employed n the prvate sector and they are older on average than the women. 15 Men are also more strongly represented n the group havng no search tme 16, whle women are more often found n the groups who searched for more than sx months before fndng (or acceptng) ther current jobs. 13 However, tme spent n searchng for the job may also have a contrary effect. Assumng that ndvduals wth more tme to search wll also learn more about how to fnd a good job-match, then we could fnd a postve effect on wage bds dependng on the tme spent n job search. 14 Women have noted that they work less than 40 hours per week to a greater extent than men, see footnote 16. As regards those who chose to bargan (posted wages), 21 (18) percent of the women work part tme whle the equvalent for men s 14 (12) percent. Yet, gven the age of the ndvduals ncluded n the sample, t s reasonable to assume few appled for part-tme work as the average age for females havng ther frst chld n Sweden s approxmately 28 years n both 1999 and 2000 (Statstcs Sweden, 2000). A t-test, moreover, reveals that the share of women over 28 years s smaller than the share of over 28 men n the group whch has noted work less than 40 hours. The dfference s also statstcally sgnfcant at the 5 percent level. 15 The hgher average age among men may be explaned by the compulsory mltary servce for young men n Sweden. 16 Note that the group no search tme ncludes () ndvduals who got the job before graduaton, () those who returned to a prevous employer from the tme before ther unversty studes, or () those who found ther jobs wthn a month of graduaton. 13
There are also some dfferences between the groups who appled for a job requrng wage barganng and those who appled for jobs wth posted wages. Gender dfferences between these groups, however, are small. Ths last could then be taken as evdence that there are no partcular selecton effects along gender lnes. 3.2. Self-promotng Barganng Strateges and Barganng Success To assess the extent to whch applcants overbd, or underbd, smlar applcants a measure whch relates the wage bd to a reference wage bd the employer may have expected gven the applcants characterstcs s developed. The reference wage s derved from regressng the wage bd accordng to equaton [2] and from ths then predct a market wage bd for every applcant who chose to bargan over ther wages, PREDMARKET BID. The measure of overbddng s then gven by: BID OVERBID = [3]. PREDMARKET BID When the measure, OVERBID s above 1, the applcant has bd a wage hgher than a smlar applcant would, gven the same characterstcs. When t s below 1 the applcant has bd a wage lower than a smlar applcant would, n general, gven the same characterstcs. Smlarly, we develop a measure whch captures the barganng outcome or barganng success, relatve to what the applcant may have expected gven the job and applcant characterstcs. Lkewse, the actual wage offer s compared to a calculated reference wage offer derved by estmatng equaton [1] for the whole sample and from ths predct a market wage offer for each applcant who chose to bargan over ther wage, The measure of barganng success s then gven by: PREDMARKET OFFER. 14
OFFER BARGSUCCES S = [4]. PREDMARKET OFFER When the measure s above 1 (below 1) the applcant receved an offer hgher (lower) than that predcted by hs or her wage- generatng characterstcs or by the job attrbutes. We allow for two dfferent specfcatons of the reference wage offer and wage bds. In a frst specfcaton a gender dummy s not ncluded when estmatng the reference wages. The reference wages are thereby set accordng to what s expected of a smlar canddate regardless of gender. In a second specfcaton the gender dummy s ncluded to take nto account that men and women may be dfferently valued n the market (the commonly found negatve wage premum for women) and therefore do not serve as substtutes for a smlar canddate. Note that wth the frst specfcaton, women s overbddng s undervalued f men and women do not appear as substtutes to the employer. Fnally we want to elct the dfferences n the pay-off from a self-promotng barganng strategy. To measure the pay-off we regress the level of overbddng on the barganng outcome, BARGSUCCES S, such that: BARGSUCCES S = α + β FEM + β OVERBID + β FEM * OVERBID + ε [5]. F ob FEMob The β ob measures the effect of overbddng on the barganng success,.e. the pay-off at dfferent levels of overbddng, relatve to a smlar applcant. Notce that β ob can capture the pay-off to unobservable ablty, preferences for rsk-takng or compettveness. The female dummy, FEM, and nteracton effects, FEM * OVERBID are ncluded to compare barganng outcome dfferences between men and women. To allow for a better characterzaton of the condtonal dstrbuton of the barganng success a quantle regresson method s used. Snce f overbddng nfluence the condtonal dstrbuton of the barganng success other than at the mean, a quantle regresson model 15
consttutes a superor way of descrbng the whole dstrbuton compared to OLS, see Koenker and Hallok (2001). Note that standard errors are obtaned by bootstrap methods. OLS estmates are also ncluded for comparson. 4 Results 4.1 The Mncer Model for the Wage Bds and Wage Offers The results for the Mncer wage model for the wage bd model are reported n Table III, column 1 and 2. What the estmated model frst reveals s that women submt lower wage bds than men do, also after controllng for a broad range of factors. The dfference s approxmately 3 percent, or between 2.9-3.5 percent dependng upon the number of controls. The control varables all have the expected sgns, and are apart from beng a tranee, or not, statstcally sgnfcant. 17 The results for the offered wage are reported n Table III, column 3-5. The thrd column refers to fndngs based on the total sample whle the fourth and ffth refers to a dvson of the sample nto those choosng a job nvolvng barganng and those choosng a job wth posted wage offers. As n many prevous studes on the gender wage gap we fnd that women are offered lower wages, even when educaton, experence, age and other work-specfc characterstcs are controlled for. The gender dummy reveals an average effect of 2.7 percent lower offered wages for women. Dvdng the sample, we obtan a strkng fndng that the gender dfference n offered wages s hgher for those who chose to bargan relatve to those wth posted-wage jobs, 3.2 percent versus 2.6 percent, and that the dfference s hgher than that for the wage bds, 2.9 17 There are two coeffcents worth notng. Frst, only f the applcant has searched for more than nne months s there a postve effect on wage bds, suggestng that a long search tme could have an nformatonal value on the jobmatch qualty. One possble explanaton could however be that respondents msclassfed ther search tme. Wthn ths group, 27 percent have had a job for more than sx moths pror to the current one, n comparson to 11 percent for the whole sample. However, the correlaton between postgraduate work experence and searchng for more than
percent. Thus there s a gender dfference not only n barganng but also an even stronger one n the post-barganng outcome. To further assess the extent to whch the adjusted wage gap s explaned by genderspecfc dfferences n characterstcs, or to whch these characterstcs are dfferently prced, the wdely employed method of decomposton suggested by Blnder (1973) and Oaxaca (1973) s used (for a thorough explanaton see Appendx A). In Appendx Table I we present the OLS estmates of equaton [2] separated by gender and barganng. The results of the wage decomposton are dsplayed n Table 4. To what extent would the gap then decrease, f women s characterstcs were the same as men s? The second term () ndcate that the gap would decrease by 2.7 percentage ponts for those choosng a job wth wage barganng and by 1.6 percentage ponts for those choosng a posted wage-job, f women had the same characterstcs as men. Ths resembles results found n other papers usng Swedsh data (see e.g. Edn & Rchardson, 2001). The term commonly assocated wth dscrmnaton then accounts for 55 percent of the wage gap for those who barganed and 68 percent for those wth posted wages. Hence, even though the gap s larger for women n the former group, less of t can be ascrbed to dscrmnaton. What ths fndng ndcates s that earler studes whch measure gender wage dfferentals on a bass of offered wages and wthout knowng whether the offered wage s a result of ndvdual wage barganng behavor or not, may actually have overestmated the unexplaned share or the part commonly ascrbed to dscrmnaton. 4.2. Self-promotng Barganng Strateges and the Barganng Success The summary statstcs of the measure of overbddng are gven n Table 5. The frst fndng s that women overbd the reference wage less often than men do and f they do, they overbd by a 9 months s no greater than 0.19 for unqualfed work and 0.28 for qualfed work experence. Second, the number of 17
smaller amount. In partcular, on average, a woman who strctly overbds a smlar canddate, man or woman, overbds wth 9 percent, whle men do so by 13 percent on average. If we nstead compare the wage bd to a reference wage whch takes the gender nto account, then women overbd a smlar canddate to a larger extent than men do and by a hgher amount, on average. Ths comes from the fact that when the reference wage takes the gender nto account, then the reference wage for a female applcant s sgnfcantly lower than for a male applcant (due to the sgnfcant negatve gender dummy n the references wage regresson). Therefore more women have a rato above 1. But when we restrct the sample nto those who strctly overbd a smlar canddate, women overbd by a smaller amount, 9.4 percent, compared to men who overbd by 12.9 percent on average. Hence also when takng the negatve gender premum nto account, women use self-promotng strateges to a lesser extent. Lookng at the level of underbddng, we fnd the opposte pattern. If men underbd a smlar applcant, then he underbds by a hgher level than women do, although the dfference s not so large. Fgure 1 llustrates our fndngs well. In Fgure 1 the condtonal dstrbutons of overbddng usng gender adjusted reference wages, for men and women s depcted. We can then see that men s dstrbuton s layng to the rght of the female dstrbuton suggestng that they have a hgher level of overbddng smlar applcants, especally for hgh levels of overbddng. Table 6 reports the results from the estmatons of equaton [5], when not takng the female dummy nto account. A comparson of the OLS estmates and the medan, or 50 th, quantle estmates provdes very smlar results suggestng that women obtan lower barganng pay-off when self-promotng or overbddng a smlar canddate. For example, assume an applcant decdes to overbd a smlar applcant by 20 percent nstead of 10 percent. If the jobs appled for has a clearly negatve effect on the wage bd, also when the length of search s controlled for. 18
applcant s a woman ths strategy s rewarded by 8.9 percentage ponts hgher offered wage, whle the equvalent for a male s 9.4 percentage ponts..8.6.4.2 0.5 1 1.5 2 2.5 OVERBID Male Female OVERBID OVERBID Fgure 1 The Condtonal Dstrbuton of Overbddng for men (dashed lne) and women (flled lne). Comparng the effect of overbddng at dfferent quantles of the barganng success, we frst note that the effect s much larger n the 75 th quantle compared to, n partcular, the 25 th quantle. Assumng a smlar example to the one above, a woman who s n the 75 th quantle of the barganng success receves as a counter-offer whch s 9.6 percentage ponts hgher whle the equvalent for a male s 10.5 percentage ponts. In Fgure 2 predcted dfferences of overbddng n the hghest quantle compared to the lowest quantle of barganng success s shown. Note that these dfferences are all sgnfcant at the 99 percent level. In sum, the estmated relatonshp frst suggests that women fare better from self-promotng, or overbddng, n the groups where the 19
dstrbuton of offered wages s narrower among smlar applcants,.e. n the 25 th quantle. Second, women do relatvely worse when the dstrbuton of offered wages among smlar applcants s wder, hence where there s more scope for ndvdual wage barganng. Predcted dff quant75-25.4.3.2.1 0.5 1 1.5 2 2.5 maledff7525 femaledff7525 Fgure 2 The predcted dfference from overbddng n the 75 th quantle relatve to overbddng n the 25 th quantle for men (dashed lne) and women (sold lne) separately. In Table 7 we report the equvalent estmates controllng for the fact that men and women are not evaluated as perfect substtutes on average by employers. The results suggest a smlar relatonshp, also when ths effect s taken nto account. Fnally n Table 8, we report separate estmates for men and women. Agan we fnd the pay-off dfference between men and women to be accentuated the larger the scope for ndvdual wage barganng. Ths gender dfference n pay-off n wage barganng may frst reflect the gender dfferences n motvaton and pay noted n Agell & Bennmarker (2002). Although women enjoy 20
a lower return from ther barganng, and are thus less successful, employers may feel they wll not respond by reducng ther efforts to the same extent that men would. Alternatvely, the result could suggest that women come up aganst a glass celng when the scope for barganng, and the use of self-promotng barganng strateges s large despte havng the same wage-generatng characterstcs needed. Ths result could suggest that employers are more dscrmnatory wth a more wde spread wage dstrbuton. Assumng that n hgher-payng jobs or n hgh-profle jobs, employers allow for larger wage dstrbutons then the results may also then parallel the results found n Albrecht et al (2003), where n Swedsh wage data the gender log wage gap ncreases throughout the wage dstrbuton, and actually accelerates at the upper end of the dstrbuton. Lkewse, f the level of the wage bd s seen as a sgnal of unobservable productvty, the fndng may suggest that employers fnd women s sgnals to be less credble. Consequently, f women are aware of the lower margnal gan from ncreasng ther relatve wage bd, the ncentve to overbd n barganng declnes relatve to men s. Thus we would fnd that women do not ask for hgher wages relatve to smlar applcants, n partcular the larger the scope for barganng, resultng n them fndng themselves n a vcous crcle, such that f they do not ask for hgher wages they wll not be gven them. Hence ncentves for selfpromotng barganng strateges could be an explanaton for the fndng that female wage bds are lower. 5 Concludng Remarks Men and women s choces and outcomes n the labor market dffer along many dmensons. Dfferences n barganng strateges may capture many of these dspartes. For example, the use 21
of self-promotng strateges may be necessary to nfluence the probablty to be promoted. Alternatvely they may be necessary when applyng for a better pad job. The general concluson to be drawn from ths paper s that gender dfferences do exst n both wage barganng behavor and barganng outcomes. It s found that women both ask for lower wages and receve lower counter-offers from employers. These results are robust to the ncluson of controls for ndvdual-specfc and employer-specfc characterstcs. Another fndng here s that f wage barganng s acknowledged when explanng the wage offer gender gap, the unexplaned part of the gap s reduced. Ths suggests that f gender dfferences are neglected n wage barganng, the term commonly attrbuted to dscrmnaton n wage decompostons may be overestmated. Instead, t may be the gender dfferences n wage barganng that reflect dscrmnaton. Gender dfferences were found n the use of self-promotng strateges. Women selfpromoted less. Frst, f ths arses due to self-perceved ablty dfferences or self-enhancement, then gender wage dfferentals may be exacerbated by ndvdual wage barganng more than f the employer uses wage-offer postng. Second, f as suggested by ths study the economc ncentves are weaker for women then women may decde to choose occupatons where selfpromoton or competton s less lkely to be mportant. As ths study s unque n ts knd, the results regardng ndvdual wage barganng dfferences are strkng but yet prmary. Therefore, n vew of the fndngs n ths paper, a stronger focus on barganng behavor even when assessng the dfferent outcomes for men and women n the labor market, s a useful drecton for future research. 22
References Agell, J., Lundborg, P. (1995) Theores of Pay and Unemployment: Survey Evdence from Swedsh Manufacturng Frms, Scandnavan Journal of Economcs, Vol. 97, 295-307. Agell, J., Lundborg, P. (2003) Survey Evdence on Wage Rgdty and Unemployment: Sweden n the 1990s, Scandnavan Journal of Economcs, Vol. 105, 15-30. Agell, J., Bennmarker, H. (2002) Wage polcy and Endogenous Wage Rgdty: A Representatve Vew from Wthn, IFAU Workng Paper 2002:12, Uppsala unversty. Akerlof, G. (1982) Labor Contracts as Partal Gft Exchange, Quarterly Journal of Economcs, Vol. 97, 543-69. Akerlof, G., Yellen, J. (1990) The Far Wage-Effort Hypothess and Unemployment, Quarterly Journal of Economcs, Vol. 105, 255-283. Albrecht, J., Bjorklund, A., Vroman, S. (2003) Is there a Glass Celng n Sweden?, Journal of Labor Economcs, Vol. 21, 145-177. Babcock, L., Laschever, S. (2003) Women don t Ask, Prnceton, NJ: Prnceton Unversty Press. Babcock, L., Gelfand, M., Small, D., Stayn, H. (2006) Gender Dfferences n the Propensty to Inttate Negotaton, n Socal Psychology and Economcs, ed. De Cremer, D., Murnghan, J. K., Zeelenberg, M., Lawrence Erlbaum Assocates, New Jersey. Becker, G. (1971) The Economcs of Dscrmnaton, 2 nd edton, Chcago, The Unversty of Chcago Press. Bewley, T. (1995) A Depressed Labor Market as Explaned by Partcpants, Amercan Economc Revew Papers and Proceedngs, Vol. 85, 250-254. Betts, J. R. (1996) What do students know about wages: Evdence from a Survey of Undergraduates, Journal of Human Resources, Vol. 31, 27-57. 23
Blau, F. D., Kahn, L. M. (2003) Understandng Internatonal Dfferences n the Gender Pay Gap, Journal of Labor Economcs, Vol. 21, 106-144. Blnder, A. S. (1973) Wage Dscrmnaton: Reduced Form and Structural Estmates, Journal of Human Resources, Vol. 8, 436-455. Bowles, H., R. (2005) It Depends Who Is Askng and Who You Ask: Socal Incentves for Sex Dfferences n the Propensty to Intate Negotaton, KSG Faculty Research Workng Paper Seres RWP05-045. Brunello, G., Lucfora, C., Wnter-Ebmer, R. (2004) The Wage Expectatons of European College Students, Journal of Human Resources, Vol.39, 1116-1142. Campbell, C., Kamlan, K. (1997) The Reason for Wage Rgdty: Evdence from Survey of Frms, Quarterly Journal of Economcs, Vol. 112, 759-89. Edn, P. A., Rchardson, K. (2002) Swmmng wth the Tde: Soldarty Wage Polcy and the Gender Earnngs Gap, Scandnavan Journal of Economcs, Vol. 104, 49-67. Fehr, E., Falk, A. (1999) Wage Rgdty n a Compettve Incomplete Contract Market, Journal of Poltcal Economy, Vol. 107, 106-134. Gneezy, U., Nederle, M., Rustchn, A. (2003) Performance n Compettve Envronments: Gender Dfferences, Quarterly Journal of Economcs, Vol. 118, 1049-1074. Goldn, C., Polachek S. (1987) Resdual Dfferences by Sex: Perspectves on the Gender Gap n Earnngs, Amercan Economc Revew, Vol. 77, 143-151. Granqvst, L., Persson, H. (1999) Gender dfferences n career opportuntes n the wholesale and retal trade ndustry, n Persson H., Essays on Labour Demand and Career Moblty, Dssertaton Seres, No. 40, The Swedsh Insttute for Socal Research, Stockholm Unversty. 24
Granqvst, L., Regner, H. (2004) Den nya lönebldnngen: En forsknngsöverskt och analys av lönebldnngen för akademker, report for the Swedsh Confederaton of Professonal Assocatons. Gupta, N. D., Poulsen, A., Vlleval, M-C. (2005) Male and Female Compettve Behavor: Expermental Evdence, IZA Dscusson Paper 1833, IZA, Bonn. Holm, H. J. (2000) Gender-Based Focal Ponts, Games & Economc Behavor, Vol. 32, 292-314. Hultn, M., Sulkn, R. (1999) Wages and Unequal Access to Organzatonal Power: An Emprcal Test of Gender Dscrmnaton, Admnstratve Scence Quarterly, Vol. 44, 453-472. Janakoplos, N., Bernasek A. (1998) Are Women More Rsk Averse?, Economc Inqury, Vol. 36, 620-630. Jusek, (2000) Lönestatstk 99, JUSEK. Koenker, R., Hallock, H. (2001) Quantle Regresson, Journal of Economc Perspectves, Vol. 15, 143-156. Nederle, M., Vesterlund, L (2007) Do Women Shy away from Competton? Do Men Compete too Much?, forthcomng, Quarterly Journal of Economcs, August 2007. Oaxaca, R. (1973) Male-Female Dfferentals n Urban Labor Markets, Internatonal Economc Revew, Vol. 14, 693-709. Powell, M., Ansc, D. (1997) Gender Dfferences n Rsk Behavor n Fnancal Decson- makng: An Expermental Analyss, Journal of Economc Psychology, Vol. 18, 605-628. Schweren, C. (2002) An Erroneous Belef: Payng Women Lower Wages Does Not Pay, mmeo, Faculty of Economcs and Busness Admnstraton, Unversty of Maastrcht. 25
Smth, H. L., Powell, B. (1990) Great Expectatons: Varatons n Income Expectatons Among College Senors, Socology of Educaton, Vol. 63, 194-207. Statstcs Sweden, Natonal Agency for Hgher Educaton (1998) Kvnnor och män högskolan från gymnasum tll forskarutbldnng 1986/87 1995/96, Natonal Agency for Hgher Educaton Report 1998:44 R, Stockholm. Sundén, A. E., Surette, B. J. (1998) Gender Dfferences n the Allocaton of Assets n Retrement Savngs Plans, Amercan Economc Revew, Vol. 88, 207-211. Walters, A. E., Stuhlmacher, A. F., Meyer, L. L. (1998) Gender and Negotator Compettveness: A Meta-Analyss, Organzatonal Behavor and Human Decson Processes, Vol. 76, 1-29. 26
Tables TABLE 1 SUMMARY STATISTICS ON WAGE BIDS AND WAGE OFFERS BMALE BFEM Raw- Wage gap PWMALE PWFEM Raw- Wage gap WAGE BID (SEK) 19 312*** 18 196 0.942 (3 288.5) (2 663.9) ln. WAGE BID 9.85*** 9.80 WAGE OFFER (SEK) 18 628*** a 17 517 a 0.938 16 925*** 16 047 0.948 (3 311.1) (2560.2) (2 964.1) (2 337.6) ln. WAGE OFFER 9.82*** 9.76 9.72*** 9.67 No of Obs 901 1222 812 1030 Note: Numbers n parentheses are standard devatons. B refers to those choosng a job nvolvng ndvdual wage barganng and PW refers to those choosng a job wth a posted wage. ***/**/* denote statstcal gender dfferences at the 1/5/10 percent levels respectvely n a t-test of equal varance. a/b/c denote statstcal dfference between barganers and non-barganers at the 1/5/10 percent levels respectvely. 27
TABLE 2 SUMMARY STATISTICS ON SELECTED CONTROL VARIABLES BMALE BFEM PWMALE PWFEM Busness Adm. & Economcs 47 % ** a 42 % a 40 %** 35 % Law 9 %*** a 13 % a 29 %** 33 % Computer and System sc. 27 %*** b 14 % a 17 %*** 10 % Personnel Management 4 %*** a 10 % a 2 %** 7 % Socal Scence 13 %*** 21 % a 11 %*** 16 % Age 28.4*** 27.5 28.0*** 27.2 No work experence 58 %*** a 47 % a 68 %*** 59 % Unqualfed work experence 25 %*** a 35 % a 20 %*** 27 % Qualfed work experence 17 % a 17 % b 12 %* 14 % Temporary 22 %*** a 35 % a 40 %*** 48 % Tranee 10 %*** a 7 % a 16 % 15 % Metropoltan 69 % 66 % a 67 %*** 61 % Search < 1 month 45 %*** 40 % 52 %*** 47 % Search 1-3 months 16 % 15 % 19 % 21 % Search 3-6 months 14 % 14 % 13 % 13 % Search 6-9 months 9 %*** 14 % 8 % 9 % Search > 9 months Appled 1-5 jobs 15 %* 53 %* 17 % 48 % 8 %** 63 % 10 % 60 % Appled 6-20 jobs 29 % 27 % 24 % 26 % Appled 21-50 jobs 12 %** 17 % 9 % 9 % Appled >50 jobs No of obs 6 % 901 8 % 1222 4 % 812 5 % 1030 Note: Numbers n parentheses are standard devatons. B refers to those choosng a job nvolvng ndvdual wage barganng and PW refers to those choosng a job wth a posted wage. ***/**/* denote statstcal gender dfferences at the 1/5/10 percent levels respectvely n a t-test of equal varance. a/b/c denote statstcal dfference between barganers and non-barganers at the 1/5/10 percent levels respectvely. Employment contracts are permanent or temporary (tral perod, project bass or substtute contract.).. Metropoltan areas: Stockholm, Gothenburg and Malmo. Search <1 month ncludes ndvduals who got job pror to graduaton, who returned to same employer as before studyng, and those who searched for less than a month. Only avalable for 1999. 28
TABLE 3 REGRESSION RESULTS OF THE STANDARD MINCER MODEL FOR THE WAGE BID AND WAGE OFFERS, SEPARATED BY GENDER Wage Bd Wage Offer Sample B B 1999 ALL B PW Female -0.029*** -0.035*** -0.027*** -0.032*** -0.026*** (0.006) (0.008) (0.004) (0.006) (0.006) Law -0.012-0.013-0.038*** -0.012-0.035*** (0.009) (0.013) (0.006) (0.009) (0.009) Computer & Sys. Scen 0.051*** 0.033*** 0.072*** 0.063*** 0.079*** (0.007) (0.011) (0.006) (0.008) (0.009) Personnel Man. -0.028*** -0.039** -0.022** -0.021* -0.037** (0.011) (0.018) (0.009) (0.011) (0.015) Socal Scence -0.038*** -0.031*** -0.030*** -0.038*** -0.024** (0.008) (0.011) (0.006) (0.008) (0.010) Unqual. Work Exper. 0.024*** 0.042*** 0.036*** 0.037*** 0.013* (0.007) (0.009) (0.005) (0.007) (0.007) Qualf. Work Exper. 0.087*** 0.104*** 0.113*** 0.075*** (0.009) (0.006) (0.008) (0.009) Age 0.014*** 0.013*** 0.011*** 0.013*** 0.006*** (0.001) (0.001) (0.001) (0.001) (0.001) Governmental Sector -0.039*** -0.035*** -0.047*** -0.040*** -0.036*** (0.007) (0.010) (0.005) (0.007) (0.008) Muncpal Sector -0.031*** -0.037*** -0.019** -0.024** -0.033** (0.010) (0.014) (0.009) (0.011) (0.016) Temporary contr. -0.023*** -0.030*** -0.038*** -0.027*** -0.034*** (0.006) (0.009) (0.005) (0.006) (0.007) Tranee 0.008 0.019-0.008 0.007-0.006 (0.009) (0.012) (0.006) (0.010) (0.008) Metropoltan Area 0.060*** 0.062*** 0.065*** 0.070*** 0.059*** (0.006) (0.008) (0.004) (0.006) (0.006) Survey 2000 0.042*** 0.045*** 0.033*** 0.055*** (0.006) (0.005) (0.006) (0.007) Search 1-3 months -0.017** -0.016 (0.008) (0.011) Search 3-6 months -0.010-0.003 (0.009) (0.013) Search 6-9 months -0.003 0.011 (0.010) (0.014) Search >9 months 0.036*** 0.046*** (0.009) (0.013) Appled 6-20 Jobs -0.039*** (0.009) Appled 21-50 Jobs -0.042*** (0.011) Appled >50 Jobs -0.060*** (0.015) Constant 10.741*** 10.704*** 10.497*** 10.657*** 10.071*** (0.069) (0.098) (0.056) (0.072) (0.084) R-squared 0.313 0.306 0.323 0.309 0.307 Adjusted R-squared 0.307 0.293 0.320 0.304 0.302 F-value 53.207 23.307 134.416 67.182 57.828 No of obs. 2123 1022 3965 2123 1842 Note: Standard errors are n talcs. All values are n SEK. ***/**/* denote statstcal sgnfcance at the 1/5/10 percent levels respectvely. B refers to those choosng a job nvolvng ndvdual wage barganng and PW refers to those choosng a job wth a posted wage. For a varable descrpton see Table II. 29
TABLE 4 DECOMPOSING THE GENDER WAGE GAP IN OFFERED WAGES FOR THOSE WITH A JOB REQUIRING WAGE BARGAINING AND THOSE WITH A POSTED-WAGE JOB THE MALE-FEMALE WAGE GAP EXPLAINED BY DIFFERENCES IN: (I) CHARACTERISTICS (II) THE RETURN TO CHARACTERISTICS lnw ln β X X ) M W F M ( M F [ β ( X X ) M ] THE MINCER MODEL FOR POSTED-WAGE JOBS 0.050 0.016 [-0.025] THE MINCER MODEL FOR 0.060 0.027 BARGAINING JOBS [-0.028] Estmates follow from estmatons reported n Appendx Table I, columns 1 and 2. Estmates follow from estmatons reported n Appendx Table I, columns 3 and 4. F F X F ( M F β β ) [ X ( β β M )] M F 0.034 [-0.018] 0.033 [-0.030] 30
TABLE 5 THE DISTRIBUTION OF OVERBIDDING OR SELF-PROMOMTING BARGAINING STRATEGY ALL MALE FEMALE OVERBID 1.008 1.024*** 0.996*** (0.131) (0.146) (0.116) OVERBID < 1 0.924 0.925 0.924 (0.056) (0.058) (0.056) share 0.54 0.51 0.57 OVERBID > 1 1.108 1.130*** 1.090*** (0.123) (0.137) (0.107) share 0.46 0.49*** 0.43*** OVERBID F 1.008 1.011*** 1.027 (0.130) (0.144) (0.122) OVERBID F < 1 0.925 0.920** 0.928** (0.057) (0.059) (0.055) Share 0.55 0.56 0.53 OVERBID F > 1 1.108 1.129*** 1.094*** (0.121) (0.137) (0.108) share 0.45 0.44* 0.47* Note: Numbers n parenthess are standard devatons. ***/**/* denote a statstcally sgnfcant gender dfference at the 1/5/10 percent levels respectvely. OVERBID s defned as the rato of the wage bd and a reference wage bd predcted for each ndvdual usng the same controls as n Table III column 1, except for the gender dummy. OVERBID F s defned as the rato of the wage bd and a reference wage bd predcted for each ndvdual usng the same controls as n Table III column 1 ncludng the gender dummy 31
TABLE 6 REGRESSION RESULTS ON THE SUCCESS OF THE WAGE BARGAINING BARG SUCCESS BARG SUCCESS BARG SUCCESS BARG SUCCESS 25 TH QUANTILE 50 TH QUANTILE 75 TH QUANTILE OLS OVERBID 0.808*** 0.943*** 1.047*** A 0.936*** (0.017) (0.017) (0.017) (0.017) FEM*OVERBID 0.035-0.052** -0.083*** A -0.050** (0.025) (0.024) (0.022) (0.025) FEMALE -0.038 0.043* 0.082*** A 0.044* (0.025) (0.025) (0.022) (0.025) Constant 0.181*** 0.096*** 0.022 A 0.094*** (0.018) (0.017) (0.017) (0.017) Pseudo R-sq 0.3884 0.4628 0.5455 R-squared 0.727 Adj-R-squared 0.726 F-value 1879.51 p-value 0.000 No of Obs. 2123 2123 2123 2123 Note: Standard errors are n talcs. ***/**/* denote statstcal sgnfcance at the 1/5/10 percent levels respectvely. A ndcates a statstcally sgnfcant dfference between the 75 th and 25 th quantle at the 1 percent level. The dependent varable s the barganng success defned as the rato of the offered wage and a predcted reference wage offer, usng the same control varables as n Table III, column 3 except for the gender dummy. OVERBID s defned as the rato of the wage bd and a reference wage bd predcted for each ndvdual usng the same controls as n Table III column 1, except for the gender dummy. 32
TABLE 7 REGRESSION RESULTS ON THE SUCCESS OF THE WAGE BARGAINING BARG SUCCESS F BARG SUCCESS F BARG SUCCESS F BARG SUCCESS F 25 TH QUANTILE 50 TH QUANTILE 75 TH QUANTILE OLS OVERBID F 0.806*** 0.946*** 1.044*** A 0.937*** (0.017) (0.018) (0.017) (0.017) FEM* OVERBID F 0.045* -0.061** -0.079*** A -0.053** (0.025) (0.026) (0.022) (0.025) FEMALE -0.047* 0.052** 0.077*** A 0.048* (0.025) (0.027) (0.023) (0.025) Constant 0.183*** 0.093*** 0.026 A 0.093*** (0.018) (0.019) (0.017) (0.017) Pseudo R-sq 0.3861 0.4603 0.5395 R-squared 0.723 Adj-R-squared 0.722 F-value 1840.715 p-value 0.000 No of Obs. 2123 2123 2123 2123 Note: Standard errors are n talcs. ***/**/* denote statstcal sgnfcance at the 1/5/10 percent levels respectvely. A ndcates a statstcally sgnfcant dfference between the 75 th and 25 th quantle at the 1 percent level. The dependent varable s the barganng success defned as the rato of the offered wage and a predcted reference wage offer, usng the same control varables as n Table III, column 3 ncludng the gender dummy. OVERBID F s defned as the rato of the wage bd and a reference wage bd predcted for each ndvdual usng the same controls as n Table III column 1, ncludng the gender dummy. 33
TABLE 8 REGRESSION RESULTS ON THE SUCCESS OF THE WAGE BARGAINING BS F BS F BS F BS F BS F BS F BS F BS F 25 TH QUANT 50 TH QUANT 75 TH QUANT. OLS 25 TH QUANT 50 TH QUANT 75 TH QUANT OLS WOMEN WOMEN WOMEN WOMEN MEN MEN MEN MEN OVERBID F 0.851*** 0.885*** 0.965*** A 0.884*** 0.806*** 0.946*** 1.044*** A 0.937*** (0.021) (0.017) (0.013) (0.017) (0.023) (0.017) (0.022) (0.018) Constant 0.136*** 0.145*** 0.103*** 0.141*** 0.183*** 0.093*** 0.026 0.093*** (0.021) (0.018) (0.014) (0.017) (0.024) (0.017) (0.023) (0.019) Ps. R-sq 0.3856 0.4389 0.5056 0.3854 0.4840 0.5733 R-sq 0.697 0.744 Adj Rsq 0.696 0.744 F-value 2802.526 2618.993 p-value 0.000 0 No of Obs. 1222 1222 1222 1222 901 901 901 901 Note: Standard errors are n talcs.***/**/* denote statstcal sgnfcance at the 1/5/10 percent levels respectvely. A ndcates a statstcally sgnfcant dfference between the 75 th and 25 th quantle at the 1 percent level. OVERBID F s defned as the rato of the wage bd and a reference wage bd predcted for each ndvdual usng the same controls as n Table III column 1, ncludng the gender dummy. 34
APPENDIX TABLE I REGRESSION RESULTS OF THE STANDARD MINCER MODEL FOR THE WAGE OFFERS USED IN THE OAXACA-BLINDER DECOMPOSITION Barganng Posted Wage Sample Male Female Male Female Law -0.004-0.013-0.010-0.048*** (0.017) (0.011) (0.015) (0.011) Computer & Sys. Scen 0.045*** 0.084*** 0.075*** 0.088*** (0.011) (0.011) (0.013) (0.013) Personnel Man. -0.063*** -0.007 0.020-0.056*** (0.024) (0.012) (0.034) (0.016) Socal Scence -0.045*** -0.028*** 0.005-0.035*** (0.016) (0.009) (0.016) (0.012) Unqual. Work Exper. 0.056*** 0.026*** 0.033*** -0.000 (0.012) (0.008) (0.012) (0.009) Qualf. Work Exper. 0.148*** 0.086*** 0.102*** 0.051*** (0.013) (0.010) (0.015) (0.012) Age 0.012*** 0.013*** 0.002 0.008*** (0.002) (0.001) (0.002) (0.001) Governmental Sector -0.052*** -0.030*** -0.068*** -0.017* (0.013) (0.009) (0.013) (0.010) Muncpal Sector -0.024-0.022* -0.029-0.028 (0.021) (0.012) (0.032) (0.018) Temporary contr. -0.035*** -0.021*** -0.041*** -0.026*** (0.011) (0.007) (0.012) (0.009) Tranee -0.025* 0.038*** -0.020 0.007 (0.015) (0.013) (0.013) (0.010) Metropoltan Area 0.077*** 0.068*** 0.061*** 0.056*** (0.010) (0.007) (0.010) (0.008) Survey 2000 0.018 0.044*** 0.055*** 0.055*** (0.011) (0.008) (0.010) (0.008) Constant 10.599*** 10.640*** 9.837*** 10.208*** (0.126) (0.086) (0.138) (0.104) R-squared 0.292 0.299 0.313 0.288 Adjusted R-squared 0.282 0.291 0.302 0.279 F-value 28.185 39.640 27.952 31.611 No of obs. 901 1222 812 1030 Note: Standard errors are n talcs. All values are n SEK. For varable defnton see Table II. ***/**/* denote statstcal sgnfcance at the 1/5/10 percent levels respectvely. 35
Appendx A: The Oaxaca Blnder Wage Decomposton Let ln WM be the natural logarthm monthly mean wage for men and ln WF be the equvalent for women. Further, let X M and X F denote the mean values of the characterstcs used as explanatory varables and let β M and β F denote the slope coeffcents of the male and female wage equatons respectvely. The wage gap can then be wrtten as ln WM lnwf = β M ( X M X F ) + X F ( β M β F ), [2] () () () where frst term () equals the male-female wage dfferental n log wages. The second term () s the share of the wage gap whch can be credted to dfferences n the mean values of the explanatory varables. The fnal and thrd term () s the share of the wage gap that s credted to dfferent returns connected wth those characterstcs or varables (.e. measurable and seemngly dentcal characterstcs). It s ths fnal term that s assumed to be ascrbable to dscrmnaton. 18 18 There are two lmtatons to ths decomposton and the extent to whch the thrd term can be assumed to reflect dscrmnaton. Frst, a crucal element for dvdng the decomposton nto one non-dscrmnatory (explaned) and one dscrmnatory (unexplaned) part, s that all productvty-related aspects that affect wages have to be ncluded n the wage equaton. Second, the slope coeffcents used n the decomposton ought to have a sgnfcant mpact on the wages. Both of the above condtons are hard to fulfll, whch means that the results should be analyzed wth ths shortcomng n mnd. Moreover, two other ponts should be made. If dentcal characterstcs do not yeld the same return, ths cannot always be ascrbed to dscrmnaton (Goldn and Polachek, 1987). Moreover, dfferences n characterstcs may n ther turn be due to dscrmnaton. 36