- Email: [email protected]
The gender earnings gap among pharmacists
Research in Social and Administrative Pharmacy 8 (2012) 285–297
Original Research
The gender earnings gap among pharmacists Manuel J. Carvajal, Ph.D.a,*, Graciela M. Armayor, Pharm.D., M.S.a, Lisa Deziel, Pharm.D., Ph.D.b a
Department of Sociobehavioral and Administrative Pharmacy, Nova Southeastern University College of Pharmacy, 3200 S. University Drive, Fort Lauderdale, FL 33328, USA b Department of Pharmacy Practice, Nova Southeastern University College of Pharmacy, Fort Lauderdale, FL 33328, USA
Abstract Background: A gender earnings gap exists across professions. Compared with men, women earn consistently lower income levels. The determinants of wages and salaries should be explored to assess whether a gender earnings gap exists in the pharmacy profession. Objectives: The objectives of this study were to (1) compare the responses of male and female pharmacists’ earnings with human-capital stock, workers’ preferences, and opinion variables and (2) assess whether the earnings determination models for male and female pharmacists yielded similar results in estimating the wage-and-salary gap through earnings projections, the influence of each explanatory variable, and gender differences in statistical significance. Methods: Data were collected through the use of a 37-question survey mailed to registered pharmacists in South Florida, United States. Earnings functions were formulated and tested separately for male and female pharmacists using unlogged and semilog equation forms. Number of hours worked, human-capital stock, job preferences, and opinion variables were hypothesized to explain wage-and-salary differentials. Results: The empirical evidence led to 3 major conclusions: (1) men’s and women’s earnings sometimes were influenced by different stimuli, and when they responded to the same variables, the effect often was different; (2) although the influence of some explanatory variables on earnings differed in the unlogged and semilog equations, the earnings projections derived from both equation forms for male and female pharmacists were remarkably similar and yielded nearly identical male-female earnings ratios; and (3) controlling for number of hours worked, human-capital stock, job preferences, and opinion variables reduced the initial unadjusted male-female earnings ratios only slightly, which pointed toward the presence of gender bias. Conclusion: After controlling for human-capital stock, job-related characteristics, and opinion variables, male pharmacists continued to earn higher income levels than female pharmacists. Ó 2012 Elsevier Inc. All rights reserved. Keywords: Wages; Gender; Earnings gap; Pharmacist workforce
* Corresponding author. Tel.: þ1 954 262 1322; fax: þ1 954 262 2278. E-mail address: [email protected] (M.J. Carvajal). 1551-7411/$ - see front matter Ó 2012 Elsevier Inc. All rights reserved. doi:10.1016/j.sapharm.2011.06.003
286
Carvajal et al. / Research in Social and Administrative Pharmacy 8 (2012) 285–297
Introduction Beyond the obvious effect of number of hours worked, wages and salaries are determined by the interaction of 3 sets of forces: human-capital stock, workers’ preferences, and employers’ characteristics. All 3 sets, as well as the amount of labor supplied, are largely influenced by gender. When one is confronted with the abundant evidence that, compared with men, women earn consistently lower income levels,1-7 one would expect that one or more of these 3 sets of forces provide the underlying reasons for the differential. Most studies comparing earnings and other labor outcomes of men and women do so across occupations, which introduces unwanted stochastic disturbances (ie, greater statistical error terms diminishing goodness of fit) brought about by differences in training, employment structure, wage rate distributions, etc, that taint the effect of explanatory variables on wages and salaries.8 Even studies focusing on a single profession attempt to analyze, more often than not, data sets that extend nationwide and sometimes internationally; consequently, unmeasured variations in regional income, prices, tax structures, cultural practices and customs, and so on may wrongly attribute gender earnings differentials to variables specified in a model when, in fact, they originate in uncontrolled variables. A far more methodologically rigorous approach to probing potentially different responses by men and women to labor market conditions, in terms of human-capital stock, individual preferences, and gender bias, calls for narrowing the comparison scope to persons with the same training performing the same range of activities in the same place at the same time.9 This article sought to pursue such analytical approach by formulating and estimating, using ordinary least squares, male and female pharmacists’ earnings functions from a culturally diverse group of practitioners living in a relatively small area. Two sets of parameters were estimated for each genderda set explaining variations in wage-and-salary earnings and a set explaining variations in the natural logarithm of wage-and-salary earnings. Normally, income studies use the semilog version to reduce the right-side skew (ie, presence of outliers) innate to most wage-and-salary distributions.10-12 This procedure forces analysts to interpret the effects of the explanatory variables on the wage gap by observing relative differences in earnings rather than absolute amounts, although
absolute amounts may be obtained via antilog conversion. The data set developed for this study had an approximately normal earnings distribution, so a second objective was to estimate and compare earnings functions for each gender alternatively using unlogged and logged values of earnings as the dependent variable. The question raised here was whether the 2 formulations yielded similar results estimating the wage-and-salary gap through earnings projections, the influence of each explanatory variable, and gender differences in statistical significance. Determinants of wage-and-salary earnings Variation in human-capital stock is the most frequently used line of reasoning explaining gender differences in pay. Skills acquired through formal education, professional experience, and other forms of investment generate a stock of capital that increases workers’ productivity and becomes appealing to employers. Because men and women play different roles dictated by society, they often end up with heterogeneous levels of commitment in the career and home spheres of their lives, whereby women assume the primary child care and household responsibilities. Compared with men, they work fewer hours in the marketplace, are more likely to work part time, interrupt their careers more often, and invest less in themselves in the areas of on-the-job training and acquisition of knowledge beyond formal education.13-17 Less human-capital stock often translates into lower wages and salaries, so the observed long-term gender earnings gap may be at least partly attributed to the relatively lower marginal costs of humancapital production experienced by men.18 Women’s socially defined disproportionate involvement with child care and household responsibilities also leads them to develop different sets of tastes and preferences for job characteristics than those exhibited by men.19,20 Men usually seek jobs in which pecuniary factors such as pay, overtime hours, and advancement opportunities are emphasized, whereas women tend to prefer jobs with flexible schedules, high levels of job satisfaction, and other nonpecuniary advantages that act as compensating differentials.12,21,22 Thus, a portion of the commonly observed long-term gender disparities in pay may be attributed to compensating differentials for working conditions and jobs generally preferred by women. The third set of forces configuring gender wage-and-salary earnings differentials has to do
Carvajal et al. / Research in Social and Administrative Pharmacy 8 (2012) 285–297
with employers’ characteristics and identifies constraints and obstacles that impede access to, or advancement at, work by women. It is based on the notion that workers’ earnings largely depend on institutional factors rather than market forces. These factors may include seniority, collective bargaining, and others such as gender bias over which the worker has no control. Although there are some exceptions, the literature is replete with studies pointing to the effects of gender bias and discrimination.23-30 Standard articles on the topic systematically attempt to identify or control for variables measuring differences in human capital and preferences between men and women. Whatever gender gap remains after such adjustments usually is interpreted as evidence of employers’ characteristics, including gender bias.31 Methods Data To reduce unwanted stochastic disturbances, the study was confined to South Florida, an area covering 10,400 square miles, comprising the 8 contiguous counties in the southern tip of the peninsula: Broward, Martin, Miami-Dade, and Palm Beach on the Atlantic coast and Collier, Hendry, Lee, and Monroe on the Gulf of Mexico, or inland. In September 2006 there were 5846 licensed pharmacists, according to a list provided by the Florida Department of Health. A survey questionnaire explicitly designed to obtain data for this purpose was mailed to every pharmacist on the list in October 2006, with a reminder sent in January 2007. Of the 5846 questionnaires mailed, 140 were returned as undeliverable because of various reasonsdthe pharmacist moved without leaving a forwarding address, passed away, was no longer working, etc. A total of 1478 pharmacists returned their questionnaires for a response rate of 25.9%. The number of observations and the rate of response compared favorably with those reported by similar studies.32-39 Of the 1478 responses, 626 men and 590 women provided data for every variable included in the earnings determination models developed below. Identification and characteristics of variables in the model The wage-and-salary earnings functions formulated here and tested using ordinary least squares contained the same explanatory variables in order
287
to compare responses from men and women and measure the gender earnings gap under different conditions.10,40 An alternative pooled model, in which the gender effect was identified by a dummy variable, was discarded because it assumed implicitly, and probably incorrectly, that the responses to covariates were equal for both genders.31 In addition, the estimation of separate earnings functions for male and female pharmacists, rather than the use of a gender dummy variable, allowed for disaggregating disparities in outcome (ie, the gender earnings gap, if it existed) into components attributable to the explanatory variables identified in the model and the process by virtue of which these explanatory variables affected wageand-salary earnings. The comparison is particularly relevant when one considers differences by not only gender but also age because they relate to work and the changing profile of the health care professions, with more women than men graduating from pharmacy and other academic programs.41-45 The model presented here interpreted, for each gender, annual wage-and-salary earnings as a function of average workweek, human-capital stock, job preferences, and opinion variables, as follows: Eij ¼ ai þHijk bki þXijk gki þYijk dki þZijk qki þuij and ln Eij ¼ ui þHijk lki þ Xijk xki þ Yijk tki þ Zijk jki þvij where Eij is a vector of values of annual wage-andsalary earnings, in dollars, reported by the jth pharmacist of the ith gender; ln Eij is a vector of the natural logarithm values of annual wage-and-salary earnings, in dollars, reported by the jth pharmacist of the ith gender; Hijk is a matrix of values (k ¼ 2) of the linear and quadratic terms of average number of hours worked per week by the jth pharmacist of the ith gender; Xijk is a matrix of values (k ¼ 4) of human capital characteristics including type of pharmacy degree, professional experience (linear and quadratic terms), and number of children reported by the jth pharmacist of the ith gender; Yijk is a matrix of values (k ¼ 2) of job-related characteristics including pharmacy setting and main role as a practitioner reported by the jth pharmacist of the ith gender; Zijk is a matrix of values (k ¼ 3) of opinion variables including job satisfaction, availability of advancement opportunities, and job security expressed by the jth pharmacist of the ith gender;
288
Carvajal et al. / Research in Social and Administrative Pharmacy 8 (2012) 285–297
uij and vij are vectors of normally and independently distributed stochastic disturbance terms in the unlogged and semilog equations, respectively, with mean zero and variance sui2 and svi,2 respectively, pertaining to the jth pharmacist of the ith gender; ai and ui are the constant terms estimated for the ith gender in the unlogged and semilog equations, respectively, using ordinary least squares; bki, gki, dki, and qki are vectors of k parameters, 1 parameter per explanatory variable within their respective group, being estimated in the unlogged equations for the ith gender using ordinary least squares; lki, xki, tki, and jki are vectors of k parameters, 1 parameter per explanatory variable within their respective group, being estimated in the semilog equations for the ith gender using ordinary least squares; and where i ¼ 1 for male pharmacists and i ¼ 2 for female pharmacists; j ¼ 1, ., ni; and ni is the number of pharmacists corresponding to the ith gender (n1 ¼ 626 and n2 ¼ 590). The least squares coefficients of the 2 equation forms were not directly comparable. In the unlogged equation, the coefficients represented the estimated dollar change in earnings brought about by a 1-unit change in the corresponding
explanatory variable, whereas in the semilog equation the coefficients denoted exponential values, that is, the percentage by which earnings were estimated to change for each 1-unit change in the corresponding explanatory variable. At any level of semilog earnings, an addition of 0.693147 units to the value of the dependent variable doubled earnings; a subtraction of 0.693147 units cut earnings in half. In the empirical evidence (next) section, results from both equation forms were made comparable by expressing unlogged estimated changes in percentage terms and by calculating elasticities at the 1% increase in the means of the explanatory variables.
Results Gender comparisons Table 1 shows means and standard deviations of wage-and-salary earnings reported by male and female pharmacists, as well as variables postulated to affect them. On average, women earned 7.9% less than men, and the gap was statistically significant. The disparity, however, was not necessarily indicative of gender bias; it merely might reflect the influence of gender-selective investments in human capital or voluntary trade-offs of income for nonpecuniary variables. A more meaningful gender comparison of the earnings
Table 1 Means (and standard deviations) of variables related to male and female pharmacists’ wage-and-salary earnings Variable
Means (and standard deviations) Men
Women
626 102,550a (36,456) 39.6b (11.3)
590 95,054a (25,780) 38.5b (10.0)
Human-capital variables (X) Pharm.D. degree (%) Professional experience (years) Children in the household (number)
0.256a (0.437) 27.8a (15.3) 1.7a (1.4)
0.498a (0.500) 14.7a (10.5) 1.2a (1.2)
Job preference variables (Y) Work in retail chain pharmacy (%) Main role administrative (%)
0.412 (0.493) 0.109c (0.311)
0.400 (0.490) 0.085c (0.279)
7.2 (2.2) 4.2a (3.1) 7.7a (2.3)
7.2 (2.2) 5.4a (3.0) 8.1a (2.0)
Number of observations Annual wage-and-salary earnings (dollars) (E) Average workweek (hours)
Opinion variables (Z) Job satisfaction (0-10 scale) Advancement opportunities (0-10 scale) Job security (0-10 scale) a b c
Male and female values significantly different from each other (P ! .01). Male and female values significantly different from each other (P ! .10). Male and female values significantly different from each other (P ! .20).
Carvajal et al. / Research in Social and Administrative Pharmacy 8 (2012) 285–297
determinants in the model must be undertaken throughout this study to ascertain the source(s) of the disparity. Number of hours worked was perhaps the most obvious of these determinants. If women worked in the marketplace fewer hours than their male counterparts because they took on relatively more work at home, one would expect women’s earnings to be lower than men’s earnings, even if both were paid the same wage rate. The data revealed that indeed male pharmacists worked relatively longer hours, and the difference was statistically significant. Consequently, at least a portion of the observed pay gap seemed to be attributable to differences in amount of work. This variable appeared in the model with linear and quadratic terms to measure rate of change. The model contained 3 human-capital variables: type of academic degree, professional experience, and presence of children in the household. Normally, workers are paid according to the quantity and quality of skills they bring to the labor market; the greater and better their educational attainment, the higher their incomes are likely to be. Because pharmacists receive homogeneous training and have to pass standardized board exams, variation in years of schooling was not as relevant in this study as the type of academic degree attained. In response to recent clinical demands on practitioners, the traditional bachelor of science degree was being phased out at the turn of the century; the Pharm.D. is the new required degree for pharmacists entering the profession in the United States and calls for at least 1 more year of school work. On average, throughout the United States, pharmacists with a Pharm.D. degree are estimated to earn approximately 3% to 6% higher incomes than their colleagues with a baccalaureate degree, although the differential does not control for other variables.46,47 The effect of type of academic degree was measured here with a dummy variable receiving a value of 1 if the jth pharmacist of the ith gender held a Pharm.D. degree, a value of 0 otherwise. The data from this study revealed that almost twice as many women as men possessed a Pharm.D. degree, which was consistent with the volume of women’s entry into the profession in the last 20 years. Professional experience is another earnings determinant related to productivity. Pharmacists are expected to be rewarded for their experience because the longer they have practiced their profession, the greater has been their investment in acquiring skills leading them to do a better job.
289
Linear and quadratic terms were specified for this variable to measure the direction of the effect and the rate of change. Men reported almost twice as many years of professional experience as did women, which might reflect women’s greater propensity to interrupt their careers for the sake of child care and housework,48-50 as well as their relatively recent entry into the pharmacy workforce.41 The other variable related to human capital recorded the presence of children in the household. Women generally devote more time than men to child-rearing activities such as being primarily responsible for children’s physical care, playing with them, and helping with homework plus performing additional housework brought about by children (ie, cleaning, cooking, etc).51,52 Compared with men, the presence of children was expected to influence women’s labor supply, as well as their ability and willingness to invest in their humancapital stock, and consequently their earnings, to a greater extent. On average, female pharmacists analyzed here had fewer children than their male counterparts, which evidences the contention that women are forced to make choices between having a family and pursuing a career that men usually are not forced to make. The second set of variables explored in this earnings determination model focused on jobrelated characteristics reflecting pharmacists’ preferences. The characteristics included were type of primary practice site and main role as practitioner. Both indicators were captured as dummy variables. The practice site variable received a value of 1 if the pharmacist reported that he/ she worked primarily in a retail chain establishment and a value of 0 otherwise; pharmacists also were assigned a value of 1 if their primary work activity was administrative in nature and a value of 0 otherwise. The percentages of practitioners working in retail chain were very similar for both genders, and the percentage of men who reported working as administrators was only marginally greater than the percentage of women. The third set of variables in the model consisted of respondents’ opinions regarding their jobs. These opinions reflected perceptions and attitudes conditioned by the various rewards, or lack thereof, that different groups of individuals experienced in their work environment.53 They measured how much pharmacists liked specific aspects of their professional activities, thus serving as proxies for compensating differentials. The broad hypothesis tested here was that interpretation of what one does influences the wages one demands in exchange for his/her
290
Carvajal et al. / Research in Social and Administrative Pharmacy 8 (2012) 285–297
work, and the interpretation might be different for men and women. The opinion variablesdoverall job satisfaction rating, perception of advancement opportunities, and perception of job securityd were measured along intensity scales of 0 to 10, with 10 showing the greatest intensity. This procedure had the advantage of response homogeneity insofar because pharmacists were able to express their perception of various facets of their professional life following a common measurement standard. Because there was no room for outliers, the mean provided an adequate measure of central tendency. According to their responses, pharmacists in the data set reported middle to moderate levels of all 3 opinion variables. There was no difference between genders in the scores for overall job satisfaction, but women experienced greater levels of perceived advancement opportunities and job security than men. Judging by the size of the standard deviations, responses within gender seemed to be homogeneous. Empirical evidence The estimated least squares coefficients of the explanatory variables, their standard errors, and levels of significance are presented in Table 2 for the unlogged equations and in Table 3 for the semilog equations. All coefficients were
statistically significant for at least one gender in 1 of the 2 versions, and the F ratios were highly significant. The adjusted R2 values were very high for cross-sectional data and explained more of the semilog than the unlogged earnings variation. The positive linear coefficients and negative quadratic coefficients of the workweek variable, all significant, revealed that more hours worked led both men and women to greater wage-andsalary earnings at a decreasing rate. According to the coefficients of the unlogged equations, at the means of the workweek variable, 1 more hour of work yielded an increase in wages and salaries of 1.9% for men and 1.3% for women. The semilog coefficients yielded greater increments in wages and salaries out of 1 more hour of work per week: 2.8% for men and 2.3% for women. The influence of work hours on earnings should not be compared for both genders using solely the values of the least squares coefficients because earnings, as well as workweek distributions, differed for men and women. A full assessment of the magnitude of such influence required the estimation of labor input elasticities of earnings, which measured the ratio of a percentage change in wages and salaries brought about by a percentage change in the number of
Table 2 Values of least squares coefficients, (standard errors), and levels of significance of the unlogged earnings determination model estimated for male and female pharmacists Variable
Estimated least squares coefficients, (standard errors), and levels of significance Term
Constant term Workweek (linear term) Workweek (quadratic term) Pharm.D. degree Experience (linear term) Experience (quadratic term) Number of children Work in retail chain pharmacy Main role administrative Job satisfaction Advancement opportunities Job security F statistic Adjusted R2 a b c
Statistically significant (P ! .01). Statistically significant (P ! .10). Statistically significant (P ! .20).
a b1 b2 g1 g2 g3 g4 d1 d2 q1 q2 q3
Gender Men (i ¼ 1)
Women (i ¼ 2)
23,775.8 3173.9a (314.4) 16.2a (3.7) 10,752.1a (2896.9) 866.1a (279.2) 17.3a (4.8) 2513.7a (780.4) 65.3 (2163.0) 3638.2 (3383.5) 915.1b (501.5) 612.7b (358.8) 642.0c (468.4)
35,213.5 4494.5a (240.2) 37.8a (3.2) 5491.6a (1608.1) 1003.4a (196.4) 22.5a (4.5) 662.1 (603.9) 5392.5a (1490.7) 7361.8a (2532.5) 20.3 (331.2) 328.7c (246.1) 162.3 (358.3)
66.16a 0.534
85.63a 0.612
291
Carvajal et al. / Research in Social and Administrative Pharmacy 8 (2012) 285–297
Table 3 Values of least squares coefficients, their standard errors, and levels of significance of the semilog earnings determination model estimated for male and female pharmacists Variable
Estimated least squares coefficients, (standard errors), and levels of significance Term
Constant term Workweek (linear term) Workweek (quadratic term) Pharm.D. degree Experience (linear term) Experience (quadratic term) Number of children Work in retail chain pharmacy Main role administrative Job satisfaction Advancement opportunities Job security
Gender
u l1 l2 x1 x2 x3 x4 t1
b c
9.229905 0.083438a (0.003590) 0.000782a (0.000048) 0.058198b (0.024034) 0.012114a (0.002935) 0.000298a (0.000067) 0.014222c (0.009026) 0.083002a (0.022281)
(0.033327) (0.004940) (0.003534) (0.00461)
169.21a 0.748
0.090546b 0.001716 0.001757 0.003991
(0.037852) (0.004950) (0.003678) (0.005355)
98.64a 0.646
Statistically significant (P ! .01). Statistically significant (P ! .05). Statistically significant (P ! .20).
hours worked. These elasticities (Table 4), computed at the means of the variables, showed, for example, that in the unlogged equations, a 1% rise in weekly hours worked was expected to increase men’s earnings by 0.73% and women’s earnings by 0.64%. The semilog estimates were more elasticdthe same 1% rise in labor input was expected to increase men’s earnings by 1.10% and women’s earnings by 0.89%. In both equation forms, male pharmacists’ earnings appeared to be more responsive than female pharmacists’ earnings to hours of labor input. Table 4 Earnings elasticities for male and female pharmacists derived at the means of the variables from the estimated coefficients of the unlogged and semilog equations Variables in the equations
Earnings elasticities Men (i ¼ 1) Women (i ¼ 2)
Unlogged equation Workweek Professional experience Presence of children
0.73 0.03 0.04
0.64 0.05
Semilog equation Workweek Professional experience Presence of children
1.10 0.04 0.03
0.89 0.05 0.02
a
Women (i ¼ 2)
9.174006 0.077695a (0.003096) 0.000625a (0.000037) 0.083213a (0.028534) 0.010324a (0.002750) 0.000213a (0.000047) 0.020268a (0.007687) 0.030900c (0.021305) 0.001725 0.005947 0.004857c 0.007350c
t2 j1 j2 j3
F statistic Adjusted R2 a
Men (i ¼ 1)
Square term not statistically significant.
a
Practitioners who held a Pharm.D. degree earned substantially higher incomes than their counterparts without it, a greater differential than those surveyed earlier in the literature: $10,752, or 10.5% more for men, and $5492, or 5.8% more for women. These estimates, however, were similar to the estimates generated by the semilog equations, namely, men earning 8.6% more and women earning 6.0% more than pharmacists practicing without a Pharm.D. degree. From a human-capital perspective, the return to training should be invariant to gender; thus, the differences in the coefficients of the male and female equations suggested the presence of gender bias. Judging by the signs of the coefficients, it might seem that more professional experience led to greater earnings at a decreasing rate for both genders in both equation forms. Yet a more thorough analysis revealed that male pharmacists experienced an age compression problem, whereby beyond a certain point, reported income declined as age increased. In the unlogged equations, at the means of the explanatory variable 1 more year of experience translated into a loss of wages and salaries of 1.1% for men and a gain of 3.2% for women. These estimates indicated that earnings peaked at 25.0 years of experience for men and 22.3 years of experience for women. Because male pharmacists exhibited almost twice as many years of
292
Carvajal et al. / Research in Social and Administrative Pharmacy 8 (2012) 285–297
professional experience as did women, their mean was beyond the peak, whereas the female pharmacists’ mean was not, largely because of their more recent entry into the profession. Perhaps this phenomenon might be attributed in part to the tendency by many older pharmacists, primarily men, to remain in the workforce on a part-time basis,54 that is, earning less because of fewer hours worked, with more years of experience. The differential gender effect of professional experience on earnings was corroborated by the results of the semilog equations although with less intensity. According to the semilog equations, at the means of the explanatory variable 1 more year of experience represented a loss of earnings of 0.2% for men and a gain of 0.3% for women. The experienced elasticities of earnings, computed at the means of the variables, were consistent with these findings. The values of the elasticities, presented in Table 4, were low and similar in both equation forms, thus suggesting that although the coefficients were statistically significant, the effect of professional experience on earnings was small for both genders. The presence of children in the household did not behave as expected. Having children affected male pharmacists’ earnings more than it affected the earnings of female pharmacists. The male pharmacists’ coefficients were highly significant in both equation forms, whereas the female pharmacists’ coefficient lacked significance in the unlogged equation and was barely significant in the semilog equation. On average, in the unlogged equations, men earned 2.5% higher wages for each additional child, whereas in the semilog equation men earned 2.0% and women earned 1.4% higher wages for each additional child. The positive female earnings coefficient for children contradicted a long list of empirical findings reporting a negative effect, often attributed to lower labor supplied on account of children.55-62 Perhaps the presence of children did not affect female pharmacists’ wages and salaries so decisively in this study because some women had left the labor force, even temporarily, on account of children, and there was no access to their earnings. In any event, the earnings of both male and female pharmacists appeared to be children inelastic. Working in a retail environment, the first job preference variable in the equations exerted a definitive positive influence on pay for female pharmacists and a less clearly defined influence for male pharmacists. This might be a compensating differential insofar as role conflict and job stress
commonly are reported in the literature to be relatively high in retail chain settings.63,64 According to the unlogged estimates, women who worked in retail chain pharmacies earned, on average, about $5400 or 5.7% more than women who worked in other settings. This estimate was highly significant, unlike the male pharmacists’ coefficient, which lacked statistical significance. In the semilog equations, the female coefficient, also highly significant, indicated that female pharmacists who work in the retail chain setting got an even greater compensating differential, in the order of 8.7%, whereas the male coefficient, only marginally significant, indicated that men working in retail chain earned 3.1% more than men who practiced in other settings. Working primarily in an administrative capacity, the other job preference variable clearly affected women but not men. The male coefficients lacked statistical significance. Female pharmacists received higher wages and salary for doing administrative work: 7.7% in the unlogged equation and 9.5% in the semilog equation. Collectively, the 3 opinion variables identified in the model did not explain much of the variation in earnings. Five of the 6 female pharmacists’ coefficients, in both equation forms, lacked statistical significance, and the significant coefficients did not behave as compensating differentials. The first of the opinion variables, job satisfaction, was designed to approximate workers’ comprehensive evaluation in a given position and a specific setting. The job satisfaction coefficient for men was positive, suggesting that as male pharmacists’ satisfaction with their jobs increased, so did their earnings; a more likely explanation would be that pharmacists who earned higher wages and salaries tended to be more satisfied with their jobs. Whatever the cause-effect relationship, the impact was substantial. The estimate of the unlogged equation indicated that men scoring a maximum of 10 in the job satisfaction scale earned, on average, $9151 or approximately 8.9% more than men scoring a minimum of 0. Availability of advancement opportunities was the only opinion variable with a statistically significant coefficient for female pharmacists, and this occurred in the unlogged equation. The signs for both genders were positive, meaning that more opportunities led to higher wages and salaries, which accorded with the findings by Carvajal and Hardigan.11 The difference between pharmacists scoring 10 and 0 in this variable represented 6.0% higher earnings for men and 3.5% higher earnings
293
Carvajal et al. / Research in Social and Administrative Pharmacy 8 (2012) 285–297
for women in the unlogged equations, compared with 5.0% higher earnings for men in the semilog equation. Job security, the last opinion variable, also showed positive least squares coefficients for male pharmacists, the opposite of what one might expect if a trade-off existed between wages, a relatively short-term concern, and job security, a longer term consideration. What seemed to emerge here was evidence that men’s better paying jobs were perceived as being more stable, too. On average, compared with male pharmacists scoring a minimum of 0, those scoring a maximum of 10 in this variable earned 6.3% and 7.6% more wages and salaries, according to the unlogged and semilog equations, respectively. Projections of earnings One of the main advantages of estimating separate earnings functions for men and women is that disparities in outcome (ie, wage-and-salary earnings) can be disaggregated into a portion attributable to gender characteristics (ie, average workweek, human-capital stock, job preferences, and perceptions and opinions) and another portion pertaining to how these characteristics affect the outcome. The procedure was originally proposed by Blinder65 and Oaxaca66 and has been used more recently, among others, by Broyles,8 Hersch,67 and Kim.68 This decomposition technique permits the projection of wage-and-salary earnings, on average, by male pharmacists if they exhibited the labor market characteristics typical of alternatively male and female pharmacists and by the same token, the projection of average earnings by female pharmacists if they had the characteristics typical of both genders. According to Table 1, the typical male pharmacist in the sample may be configured as follows: He worked an average of 39.6 hours per week, possessed 27.8 years of professional experience, and had an average of 1.7 children. He had
a 25.6% probability of holding a Pharm.D. degree, a 41.2% probability of working in a retail chain pharmacy, and a 10.9% probability of doing primarily administrative work. His opinion variable scores were 7.2 for job satisfaction, 4.2 for availability of advancement opportunities, and 7.7 for perception of job security. The typical female pharmacist worked 38.5 hours per week, possessed 14.7 years of experience, and had an average of 1.2 children. She had a 49.8% probability of holding a Pharm.D. degree, a 40.0% probability of working in a retail chain pharmacy, and an 8.5% probability of doing primarily administrative work. Her opinion variable scores were 7.2 for job satisfaction, 5.4 for availability of advancement opportunities, and 8.1 for perception of job security. The earnings projection using the least squares coefficients estimated for male and female pharmacists in the unlogged and semilog equations, under the typical labor market conditions specified in the previous paragraph, are presented in Table 5. Within gender, the projections of the unlogged and semilog equations were very similar. The initial unadjusted male-female ratio of 1.079 was only slightly reduced when men’s as well as women’s labor market characteristics were projected with the coefficients estimated for both genders. The projected male-female earnings ratios were quite consistent, thus suggesting that observed gender differences in wages and salaries might not be substantially attributed to differences in number of hours worked, humancapital stock, workers’ preferences, or perceptions and opinions. Rather, they seem to arise from employers’ characteristics, including gender bias. Discussion The probe into wages and salaries earned by male and female pharmacists conducted here
Table 5 Projected wage-and-salary earnings of male and female pharmacists under different scenarios Typical labor market characteristics
Projected wage-and-salary earnings (dollars) Male-female Male pharmacists’ coefficients Female pharmacists’ coefficients earnings ratio
Male pharmacists’ characteristics Unlogged equation Semilog equation
108,769 107,059
102,330 101,024
1.063 1.060
Female pharmacists’ characteristics Unlogged equation Semilog equation
107,204 107,691
101,232 100,031
1.059 1.077
294
Carvajal et al. / Research in Social and Administrative Pharmacy 8 (2012) 285–297
yielded 3 major generalizations. First, men’s and women’s earnings sometimes were influenced by different factors, and when they responded to the same variables, the effect often was different. Men’s (but not women’s) earnings generally were affected by the opinion variables, whereas women’s (but not men’s) earnings were influenced by engaging primarily in an administrative role. Working more hours, holding a Pharm.D. degree, and having more children increased men’s earnings more than women’s earnings, but the opposite was true for possessing greater professional experience and working in retail chain pharmacy. Consequently, the observed heterogeneity in the earnings functions of male and female pharmacists lent methodological validity to the separate equation estimation procedure. The second major generalization of this article is that the effect of some explanatory variables on earnings differed in the unlogged and semilog equations. The male coefficient for working in a retail chain pharmacy and the female coefficient for number of children in the household lacked statistical significance in the unlogged equation but not in the semilog equation, whereas the male coefficient for job satisfaction and the female coefficient for perception of advancement opportunities lacked significance in the semilog equation but not in the unlogged equation. Differences also were apparent in the magnitude of the impact. The estimated semilog elasticities of workweek were greater for both genders than were the estimated unlogged elasticities, as was the impact of professional experience. In addition, the estimated earnings increase for male pharmacists brought about by holding a Pharm.D. degree, and scoring higher in the advancement opportunities index was greater for the unlogged than the semilog form; however, the estimated compensating differentials for women working in retail chain pharmacy and doing administrative work, as well as the effect on earnings of perceived job security by men, were greater in the semilog than the unlogged equations. Notwithstanding these differences in individual explanatory variables, the wage-and-salary earnings projections derived from both equation forms for male and female pharmacists were remarkably similar and yielded nearly identical male-female earnings ratios. Thus, although disparities of interpretation of gender differences in statistical significance and the influence of explanatory variables on earnings might arise from the 2 equation forms, both procedures led to the same conclusion. Compared with the
unlogged equations, the semilog equations were characterized by greater F and adjusted R2 values, especially for men. The third major generalization is that controlling for human-capital stock, job preferences, and opinion variables reduced the initial unadjusted male-female earnings ratio only slightly, even after entering typically male labor market characteristics into the female equations and female characteristics into the male equations. The empirical evidence here points toward the presence of gender bias. This bias, approximately 6%, was relatively small but firm and indicated that the pharmacy profession did not convert individual characteristics into earnings outcomes in the same manner for men and women in South Florida. In other words, after controlling for the effect of the explanatory variables, male pharmacists continued to earn higher income levels than female pharmacists. In interpreting these results, one must take into consideration that the analytical scope of the study was rather narrow. It focused on one profession, with homogeneous training and licensure requirements, and was circumscribed to a small region. These 3 conditions posed advantages and disadvantages. The main advantage was that frequently unaccounted sources of stochastic disturbances such as occupational differences in training and relative scarcity, variation in prices and geographical settings, sociocultural patterns, etc, which usually distort the effect of explanatory variables on earnings, were eliminated, thus producing more accurate estimates and better goodness of fit than would be otherwise. The main disadvantage, and the first limitation of the study, was that the empirical findings might not be fully generalizable. Other occupations and/or the overall labor force may be characterized by relationships that are different from the ones ruling the pharmacy profession. The estimates reported here were largely configured by unique conditions prevailing in South Florida, an area unusually rich in gender, ethnic, and age mix of both population and practitioners. The study also was limited by the absence of concrete data concerning the working status of pharmacists in the sample. Although all respondents provided job-related information, this information might have referred to a previous job if specific pharmacists were not working at the time of response. Furthermore, if a substantial portion of practitioners chose not to respond to the survey because they were not working, either voluntarily
Carvajal et al. / Research in Social and Administrative Pharmacy 8 (2012) 285–297
(eg, retired, raising children, etc) or not (eg, unemployed), the indicators reported here might not be unbiased, although given the sample size, they probably were consistent. Another limitation was related to potential biases arising from the omission of explanatory variables. Although many of the relevant wageand-salary earnings determinants commonly identified in the literature were considered here, others might have been omitted. The standard consideration of critics of gender studies is that if the effect of omitted variables were taken into account, observed gender differences in earnings would disappear. A fourth limitation was that the study rested on a survey administered only once. Therefore, it was inadequate to ascertain whether gender differences in earnings fluctuate over time, especially in the aftermath of the recession, or the extent to which such fluctuations are affected by the proliferation of pharmacy schools in Florida as well as the rest of the United States. Future research ought to include longitudinal data to understand better the evolving role of earnings determinants over time.
Conclusion Beyond its limitations and methodological concerns, the study was successful in comparing the responses of male and female pharmacists’ earnings to human-capital stock, workers’ preferences, and opinion variables; exploring the impact of each explanatory variable; establishing gender differences in statistical significance; and estimating the wage-and-salary gap through earnings projections. The empirical results raise a fundamental policy issue, namely, how to eliminate, or at least reduce, the gender bias appearing in the empirical results. Although the most obvious factor contributing to the gender gap, according to many studies, is direct discrimination against women in the workforce, the literature identifies other possible contributing factors including lack of informal mentoring provided to women, absence of networking with colleagues, less effective salary negotiating behaviors, and failure by employers to comply with equal employment opportunity regulations.69-71 Future research in the area of gender earnings gap should build on the results of the present study by assessing the effects of these variables on male and female pharmacists’ wages and salaries.
295
References 1. Blau FD, Kahn LM. Changes in the labor supply behavior of married women: 1980-2000. Forschungsinstitut zur Zukunft der Arbeit Institute for the Study of Labor Discussion Paper Series. No.13583. 2006. 2. Carvajal MJ. Economic grounds for affirmative action: the evidence on architects and engineers in South Florida. Rev Soc Econ 2006;64:515–538. 3. Graham JW, Smith SA. Gender differences in employment and earnings in science and engineering in the US. Econ Educ Rev 2005;24:341–354. 4. Srinivas S. Social attitudes and the gender pay gap in the USA in recent years. Int J Soc Econ 2007;34: 268–275. 5. Weinberger DH. Earnings by gender: evidence from Census 2000. Mon Labor Rev 2007;130:26–34. 6. Green CT, Jegadeesh N, Tang Y. Gender and job performance: evidence from Wall Street. National Bureau of Economic Research. working paper 12897. 2007. Accessed November 10, 2010. 7. Bradbury KL. Education and wages in the 1980s and 1990s: are all groups moving up together? New Engl Econ Rev 2002;First Quarter:19–46. 8. Broyles P. The gender gap of STEM professions in the United States. Int J Sociol Soc Policy 2009;29: 214–226. 9. Robinson D. Differences in occupational earnings by sex. Int Lab Rev 1998;137:3–31. 10. Bhave DP, Glomb TM. Emotional labour demands, wages and gender: a within-person, between-jobs study. J Occup Organ Psychol 2009;82:683–707. 11. Carvajal MJ, Hardigan PC. Gender and wage earnings differences among pharmacists. J Pharm Finance Econ Policy 2004;13:5–32. 12. Noonan MC, Corcoran ME, Courant PN. Pay differences among the highly trained: cohort differences in the sex gap in lawyers’ earnings. Soc Forces 2005;84: 853–872. 13. Chichilnisky G. The gender gap. Rev Dev Econ 2008; 12:828–844. 14. Cunningham M. Influences of gender ideology and housework allocation on women’s employment over the life course. Soc Sci Res 2008;37:254–267. 15. Fogli A, Veldkamp L. Nature or nurture? Learning and female labor force dynamics. Federal Reserve Bank of Minneapolis, Research Department Staff Report. No. 386, 2007. 16. Poeschl G, Pinto I, Mu´rias C, Silva A, Ribeiro R. Representations of family practices, belief in sex differences, and sexism. Sex Roles 2006;55: 111–121. 17. Quesenberry JL, Trauth EM, Morgan AJ. Understanding the “mommy tracks”: a framework for analyzing work-family balance in the IT workforce. Inform Resour Manag J 2006;19:37–53. 18. Wu H. Can the human capital approach explain lifecycle wage differentials between races and sexes? Econ Inq 2007;45:24–39.
296
Carvajal et al. / Research in Social and Administrative Pharmacy 8 (2012) 285–297
19. Bender KA, Donohue SM, Heywood JS. Job satisfaction and gender segregation. Oxf Econ Pap 2005; 57:479–496. 20. Kim S. Gender differences in the job satisfaction of public employees: a study of Seoul Metropolitan Government, Korea. Sex Roles 2005;52:667–681. 21. Carvajal MJ, Hardigan PC. Pharmacists’ intergender differences in behavior and opinions: is work input an important mediator? Internet J Allied Health Sci Pract 2008;6. 22. Theodossiou I, Vasileiou E. Making the risk of job loss a way of life: does it affect job satisfaction? Res Econ 2007;61:71–83. 23. Allen MW, Armstrong DJ, Riemenschneider CK, Reid MF. Making sense of the barriers women face in the information technology work force: standpoint theory, self-disclosure, and causal maps. Sex Roles 2006;54:831–844. 24. Blau FD, DeVaro J. New evidence on gender differences in promotion rates: an empirical analysis of a sample of new hires. Ind Relat 2007;46: 511–550. 25. Kee HJ. Glass ceiling or sticky floor? Exploring the Australian gender pay gap. Econ Rec 2006;82: 408–427. 26. Leggon CB. Women in science: racial and ethnic differences and the differences they make. J Tech Tran 2006;31:325–333. 27. Leung D. The male/female earning gap and female self-employment. J Soc Econ 2006;35:759–779. 28. Levy H. Health insurance and the wage gap. National Bureau of Economic Research. working paper 11975. January 2006. Accessed 10.11.10. 29. Long A. Happily ever after? A study of job satisfaction in Australia. Econ Rec 2005;81:303–321. 30. Luce S, Brenner M. Women and class: what has happened in forty years? Mon Rev 2006;58:80–93. 31. Black DA, Haviland AM, Sanders SG, Taylor LJ. Gender wage disparities among the highest educated. J Hum Resour 2008;43:630–659. 32. Desselle SP. Survey of certified pharmacy technicians in the United States: a quality-of-worklife study. J Am Pharm Assoc (2003) 2005;45:458–465. 33. Jacobson CC, Resneck JS Jr, Kimball AB. Generational differences in practice patterns of dermatologists in the United States. Arch Dermatol 2004;140: 1477–1482. 34. Kahaleh A, Gaither CA. Effects of empowerment on pharmacists’ organizational behaviors. J Am Pharm Assoc (2003) 2005;45:700–709. 35. Lapane KL, Hughes CM. Job satisfaction and stress among pharmacists in the long-term care sector. Consult Pharm 2006;21:287–292. 36. Lin BY, Yeh Y-C, Lin W-H. The influence of job characteristics on job outcomes of pharmacists in hospital, clinic, and community pharmacies. J Med Syst 2007;31:224–229. 37. LoBuono C. Exclusive R.Ph. salary survey: cashing in. Drug Top 2003;147:46–48. 54–55.
38. Weeks WB, Wallace AE, MacKenzie TA. Gender differences in anesthesiologists’ annual incomes. Anesthesiology 2007;106:806–811. 39. Yutzie JD, Shellito JL, Helmer SD, Chang FC. Gender differences in general surgical careers: results of a post-residency survey. Am J Surg 2005;190:955– 959. 40. van Klaveren C, van Praag B, van den Brink HM. A public good version of the collective household model: an empirical approach with an application to British household data. Rev Econ Household 2008;6:169–191. 41. Carvajal MJ, Armayor GM, Hardigan PC. In: Socioeconomic Profile of Pharmacists in South Florida, 1. Fort Lauderdale, FL: Nova Southeastern University, College of Pharmacy Monograph Series; 2010. 42. Jovic E, Wallace JE, Lemaire J. The generation and gender shifts in medicine: an exploratory survey of internal medicine physicians. BMC Health Serv Res 2006;6:55. 43. Kennedy MM. Managing different generations requires new skills, insightful leadership. Physician Exec 2003;29:20–23. 44. Shields MC, Shields MT. Working with generation X physicians. Physician Exec 2003;29:14–18. 45. Southard G, Lewis J. Building a workplace that recognizes generational diversity. Public Manag 2004; 86:8–12. 46. Levy S. Take this job and love it? Drug Top 2008;152: 34. 36–38;40;42. 47. Ukens C. Pay dirt: pharmacist salaries are still climbing, but at a slower rate. Our exclusive survey shows who gets top dollar. Drug Top 2007;151:38–42. 48. Aanerud R, Morrison E, Homer L, Rudd E, Nerad M, Cerny J. Widening the lens on gender and tenure: looking beyond the academic labor market. NWSA J 2007;19:105–123. 49. Blau FD, Ferber MA, Winkler AE. The Economics of Women, Men, and Work. Upper Saddle River, NJ: Prentice Hall; 2002. 50. Keaveny TJ, Inderrieden EJ, Toumanoff PG. Gender differences in pay of young management professionals in the United States: a comprehensive view. J Labor Res 2007;28:327–346. 51. Blank RM, Shierholz H. Exploring gender differences in employment and wage trends among less-skilled workers. National Bureau of Economic Research working paper. 12494. 2006. Accessed 10.11.10. 52. McElwain AK, Korabik K, Rosin HM. An examination of gender differences in work-family conflict. Can J Behav Sci 2005;37:283–298. 53. Malka A, Chatman JA. Intrinsic and extrinsic work orientations as moderators of the effect of annual income on subjective well-being: a longitudinal study. Pers Soc Psychol Bull 2003;29:737–746. 54. Knapp KK, Cultice JM. New pharmacist supply projections: lower separation rates and increased graduates boost supply estimates. J Am Pharm Assoc (2003) 2007;47:463–470.
Carvajal et al. / Research in Social and Administrative Pharmacy 8 (2012) 285–297 55. Blasius JJ, Pae E-K. Work-pattern differences between male and female orthodontists. Am J Orthod Dentofacial Orthop 2005;128:283–290. 56. Cohany SR, Sok E. Trends in labor force participation of married mothers of infants. Mon Labor Rev 2007;130:9–16. 57. Corrigall EA, Konrad AM. Gender role attitudes and careers: a longitudinal study. Sex Roles 2007; 56:847–855. 58. Ginther DK, Kahn S. Does science promote women? Evidence from academia 1973-2001. National Bureau of Economic Research working paper Series. No. 12,691. 2006. 59. Mosisa A, Hipple S. Trends in labor force participation in the United States. Mon Labor Rev 2006;129: 35–57. 60. Napari S. The early career gender wage gap. The London School of Economics and Political Science Center for Economic Performance Discussion Paper. No.738. 2006. 61. Pettit B, Hook J. The structure of women’s employment in comparative perspective. Soc Forces 2005; 84:779–801. 62. Stephan PE, Levin SG. Leaving careers in IT: gender differences in retention. J Tech Tran 2005;30:383– 396. 63. Mott DA, Doucette WR, Gabriel PE, Pendersen CA, Schommer JC. Pharmacists attitudes toward worklife:
64. 65.
66. 67. 68. 69.
70.
71.
297
results from a national survey of pharmacists. J Am Pharm Assoc (2003) 2004;44:326–336. Ukens C. Pharmacists stressed but satisfied with jobs, study finds. Drug Top 2004;148:42–43. Blinder AS. Wage discrimination: reduced form and structural estimates. J Hum Resour 1973;8: 436–455. Oaxaca R. Male-female wage differentials in urban labor markets. Int Econ Rev 1973;14:693–709. Hersch J. Sex discrimination in the labor market. Found Trends Microecon 2006;2:281–361. Kim M. Race and gender differences in the earnings of black workers. Ind Relat 2009;48:466–488. Kim J. Gender inequality in the U.S. labor market: evidence from the 2000 census. In: Karsten M, ed. Gender, Race, and Ethnicity in the Workplace [Three Volumes]: Issues and Challenges for Today’s Organizations. Vol ABC-CLIO eBook. Santa Barbara, CA: Praeger; 2006:269–290. Accessed 10.03.11. Gerhart B, Rynes S. Determinants and consequences of salary negotiations by male and female MBA graduates. J Appl Psychol 1991;76:256–262. Barnes M. Glass ceiling. In: Brzuzy S, Lind A, eds. Battleground: Women, Gender, and Sexuality [Two Volumes]. ABC-CLIO eBook Collection. Santa Barbara, CA: Greenwood; 2007:233–236. Accessed 10.03.11.
Original Research
The gender earnings gap among pharmacists Manuel J. Carvajal, Ph.D.a,*, Graciela M. Armayor, Pharm.D., M.S.a, Lisa Deziel, Pharm.D., Ph.D.b a
Department of Sociobehavioral and Administrative Pharmacy, Nova Southeastern University College of Pharmacy, 3200 S. University Drive, Fort Lauderdale, FL 33328, USA b Department of Pharmacy Practice, Nova Southeastern University College of Pharmacy, Fort Lauderdale, FL 33328, USA
Abstract Background: A gender earnings gap exists across professions. Compared with men, women earn consistently lower income levels. The determinants of wages and salaries should be explored to assess whether a gender earnings gap exists in the pharmacy profession. Objectives: The objectives of this study were to (1) compare the responses of male and female pharmacists’ earnings with human-capital stock, workers’ preferences, and opinion variables and (2) assess whether the earnings determination models for male and female pharmacists yielded similar results in estimating the wage-and-salary gap through earnings projections, the influence of each explanatory variable, and gender differences in statistical significance. Methods: Data were collected through the use of a 37-question survey mailed to registered pharmacists in South Florida, United States. Earnings functions were formulated and tested separately for male and female pharmacists using unlogged and semilog equation forms. Number of hours worked, human-capital stock, job preferences, and opinion variables were hypothesized to explain wage-and-salary differentials. Results: The empirical evidence led to 3 major conclusions: (1) men’s and women’s earnings sometimes were influenced by different stimuli, and when they responded to the same variables, the effect often was different; (2) although the influence of some explanatory variables on earnings differed in the unlogged and semilog equations, the earnings projections derived from both equation forms for male and female pharmacists were remarkably similar and yielded nearly identical male-female earnings ratios; and (3) controlling for number of hours worked, human-capital stock, job preferences, and opinion variables reduced the initial unadjusted male-female earnings ratios only slightly, which pointed toward the presence of gender bias. Conclusion: After controlling for human-capital stock, job-related characteristics, and opinion variables, male pharmacists continued to earn higher income levels than female pharmacists. Ó 2012 Elsevier Inc. All rights reserved. Keywords: Wages; Gender; Earnings gap; Pharmacist workforce
* Corresponding author. Tel.: þ1 954 262 1322; fax: þ1 954 262 2278. E-mail address: [email protected] (M.J. Carvajal). 1551-7411/$ - see front matter Ó 2012 Elsevier Inc. All rights reserved. doi:10.1016/j.sapharm.2011.06.003
286
Carvajal et al. / Research in Social and Administrative Pharmacy 8 (2012) 285–297
Introduction Beyond the obvious effect of number of hours worked, wages and salaries are determined by the interaction of 3 sets of forces: human-capital stock, workers’ preferences, and employers’ characteristics. All 3 sets, as well as the amount of labor supplied, are largely influenced by gender. When one is confronted with the abundant evidence that, compared with men, women earn consistently lower income levels,1-7 one would expect that one or more of these 3 sets of forces provide the underlying reasons for the differential. Most studies comparing earnings and other labor outcomes of men and women do so across occupations, which introduces unwanted stochastic disturbances (ie, greater statistical error terms diminishing goodness of fit) brought about by differences in training, employment structure, wage rate distributions, etc, that taint the effect of explanatory variables on wages and salaries.8 Even studies focusing on a single profession attempt to analyze, more often than not, data sets that extend nationwide and sometimes internationally; consequently, unmeasured variations in regional income, prices, tax structures, cultural practices and customs, and so on may wrongly attribute gender earnings differentials to variables specified in a model when, in fact, they originate in uncontrolled variables. A far more methodologically rigorous approach to probing potentially different responses by men and women to labor market conditions, in terms of human-capital stock, individual preferences, and gender bias, calls for narrowing the comparison scope to persons with the same training performing the same range of activities in the same place at the same time.9 This article sought to pursue such analytical approach by formulating and estimating, using ordinary least squares, male and female pharmacists’ earnings functions from a culturally diverse group of practitioners living in a relatively small area. Two sets of parameters were estimated for each genderda set explaining variations in wage-and-salary earnings and a set explaining variations in the natural logarithm of wage-and-salary earnings. Normally, income studies use the semilog version to reduce the right-side skew (ie, presence of outliers) innate to most wage-and-salary distributions.10-12 This procedure forces analysts to interpret the effects of the explanatory variables on the wage gap by observing relative differences in earnings rather than absolute amounts, although
absolute amounts may be obtained via antilog conversion. The data set developed for this study had an approximately normal earnings distribution, so a second objective was to estimate and compare earnings functions for each gender alternatively using unlogged and logged values of earnings as the dependent variable. The question raised here was whether the 2 formulations yielded similar results estimating the wage-and-salary gap through earnings projections, the influence of each explanatory variable, and gender differences in statistical significance. Determinants of wage-and-salary earnings Variation in human-capital stock is the most frequently used line of reasoning explaining gender differences in pay. Skills acquired through formal education, professional experience, and other forms of investment generate a stock of capital that increases workers’ productivity and becomes appealing to employers. Because men and women play different roles dictated by society, they often end up with heterogeneous levels of commitment in the career and home spheres of their lives, whereby women assume the primary child care and household responsibilities. Compared with men, they work fewer hours in the marketplace, are more likely to work part time, interrupt their careers more often, and invest less in themselves in the areas of on-the-job training and acquisition of knowledge beyond formal education.13-17 Less human-capital stock often translates into lower wages and salaries, so the observed long-term gender earnings gap may be at least partly attributed to the relatively lower marginal costs of humancapital production experienced by men.18 Women’s socially defined disproportionate involvement with child care and household responsibilities also leads them to develop different sets of tastes and preferences for job characteristics than those exhibited by men.19,20 Men usually seek jobs in which pecuniary factors such as pay, overtime hours, and advancement opportunities are emphasized, whereas women tend to prefer jobs with flexible schedules, high levels of job satisfaction, and other nonpecuniary advantages that act as compensating differentials.12,21,22 Thus, a portion of the commonly observed long-term gender disparities in pay may be attributed to compensating differentials for working conditions and jobs generally preferred by women. The third set of forces configuring gender wage-and-salary earnings differentials has to do
Carvajal et al. / Research in Social and Administrative Pharmacy 8 (2012) 285–297
with employers’ characteristics and identifies constraints and obstacles that impede access to, or advancement at, work by women. It is based on the notion that workers’ earnings largely depend on institutional factors rather than market forces. These factors may include seniority, collective bargaining, and others such as gender bias over which the worker has no control. Although there are some exceptions, the literature is replete with studies pointing to the effects of gender bias and discrimination.23-30 Standard articles on the topic systematically attempt to identify or control for variables measuring differences in human capital and preferences between men and women. Whatever gender gap remains after such adjustments usually is interpreted as evidence of employers’ characteristics, including gender bias.31 Methods Data To reduce unwanted stochastic disturbances, the study was confined to South Florida, an area covering 10,400 square miles, comprising the 8 contiguous counties in the southern tip of the peninsula: Broward, Martin, Miami-Dade, and Palm Beach on the Atlantic coast and Collier, Hendry, Lee, and Monroe on the Gulf of Mexico, or inland. In September 2006 there were 5846 licensed pharmacists, according to a list provided by the Florida Department of Health. A survey questionnaire explicitly designed to obtain data for this purpose was mailed to every pharmacist on the list in October 2006, with a reminder sent in January 2007. Of the 5846 questionnaires mailed, 140 were returned as undeliverable because of various reasonsdthe pharmacist moved without leaving a forwarding address, passed away, was no longer working, etc. A total of 1478 pharmacists returned their questionnaires for a response rate of 25.9%. The number of observations and the rate of response compared favorably with those reported by similar studies.32-39 Of the 1478 responses, 626 men and 590 women provided data for every variable included in the earnings determination models developed below. Identification and characteristics of variables in the model The wage-and-salary earnings functions formulated here and tested using ordinary least squares contained the same explanatory variables in order
287
to compare responses from men and women and measure the gender earnings gap under different conditions.10,40 An alternative pooled model, in which the gender effect was identified by a dummy variable, was discarded because it assumed implicitly, and probably incorrectly, that the responses to covariates were equal for both genders.31 In addition, the estimation of separate earnings functions for male and female pharmacists, rather than the use of a gender dummy variable, allowed for disaggregating disparities in outcome (ie, the gender earnings gap, if it existed) into components attributable to the explanatory variables identified in the model and the process by virtue of which these explanatory variables affected wageand-salary earnings. The comparison is particularly relevant when one considers differences by not only gender but also age because they relate to work and the changing profile of the health care professions, with more women than men graduating from pharmacy and other academic programs.41-45 The model presented here interpreted, for each gender, annual wage-and-salary earnings as a function of average workweek, human-capital stock, job preferences, and opinion variables, as follows: Eij ¼ ai þHijk bki þXijk gki þYijk dki þZijk qki þuij and ln Eij ¼ ui þHijk lki þ Xijk xki þ Yijk tki þ Zijk jki þvij where Eij is a vector of values of annual wage-andsalary earnings, in dollars, reported by the jth pharmacist of the ith gender; ln Eij is a vector of the natural logarithm values of annual wage-and-salary earnings, in dollars, reported by the jth pharmacist of the ith gender; Hijk is a matrix of values (k ¼ 2) of the linear and quadratic terms of average number of hours worked per week by the jth pharmacist of the ith gender; Xijk is a matrix of values (k ¼ 4) of human capital characteristics including type of pharmacy degree, professional experience (linear and quadratic terms), and number of children reported by the jth pharmacist of the ith gender; Yijk is a matrix of values (k ¼ 2) of job-related characteristics including pharmacy setting and main role as a practitioner reported by the jth pharmacist of the ith gender; Zijk is a matrix of values (k ¼ 3) of opinion variables including job satisfaction, availability of advancement opportunities, and job security expressed by the jth pharmacist of the ith gender;
288
Carvajal et al. / Research in Social and Administrative Pharmacy 8 (2012) 285–297
uij and vij are vectors of normally and independently distributed stochastic disturbance terms in the unlogged and semilog equations, respectively, with mean zero and variance sui2 and svi,2 respectively, pertaining to the jth pharmacist of the ith gender; ai and ui are the constant terms estimated for the ith gender in the unlogged and semilog equations, respectively, using ordinary least squares; bki, gki, dki, and qki are vectors of k parameters, 1 parameter per explanatory variable within their respective group, being estimated in the unlogged equations for the ith gender using ordinary least squares; lki, xki, tki, and jki are vectors of k parameters, 1 parameter per explanatory variable within their respective group, being estimated in the semilog equations for the ith gender using ordinary least squares; and where i ¼ 1 for male pharmacists and i ¼ 2 for female pharmacists; j ¼ 1, ., ni; and ni is the number of pharmacists corresponding to the ith gender (n1 ¼ 626 and n2 ¼ 590). The least squares coefficients of the 2 equation forms were not directly comparable. In the unlogged equation, the coefficients represented the estimated dollar change in earnings brought about by a 1-unit change in the corresponding
explanatory variable, whereas in the semilog equation the coefficients denoted exponential values, that is, the percentage by which earnings were estimated to change for each 1-unit change in the corresponding explanatory variable. At any level of semilog earnings, an addition of 0.693147 units to the value of the dependent variable doubled earnings; a subtraction of 0.693147 units cut earnings in half. In the empirical evidence (next) section, results from both equation forms were made comparable by expressing unlogged estimated changes in percentage terms and by calculating elasticities at the 1% increase in the means of the explanatory variables.
Results Gender comparisons Table 1 shows means and standard deviations of wage-and-salary earnings reported by male and female pharmacists, as well as variables postulated to affect them. On average, women earned 7.9% less than men, and the gap was statistically significant. The disparity, however, was not necessarily indicative of gender bias; it merely might reflect the influence of gender-selective investments in human capital or voluntary trade-offs of income for nonpecuniary variables. A more meaningful gender comparison of the earnings
Table 1 Means (and standard deviations) of variables related to male and female pharmacists’ wage-and-salary earnings Variable
Means (and standard deviations) Men
Women
626 102,550a (36,456) 39.6b (11.3)
590 95,054a (25,780) 38.5b (10.0)
Human-capital variables (X) Pharm.D. degree (%) Professional experience (years) Children in the household (number)
0.256a (0.437) 27.8a (15.3) 1.7a (1.4)
0.498a (0.500) 14.7a (10.5) 1.2a (1.2)
Job preference variables (Y) Work in retail chain pharmacy (%) Main role administrative (%)
0.412 (0.493) 0.109c (0.311)
0.400 (0.490) 0.085c (0.279)
7.2 (2.2) 4.2a (3.1) 7.7a (2.3)
7.2 (2.2) 5.4a (3.0) 8.1a (2.0)
Number of observations Annual wage-and-salary earnings (dollars) (E) Average workweek (hours)
Opinion variables (Z) Job satisfaction (0-10 scale) Advancement opportunities (0-10 scale) Job security (0-10 scale) a b c
Male and female values significantly different from each other (P ! .01). Male and female values significantly different from each other (P ! .10). Male and female values significantly different from each other (P ! .20).
Carvajal et al. / Research in Social and Administrative Pharmacy 8 (2012) 285–297
determinants in the model must be undertaken throughout this study to ascertain the source(s) of the disparity. Number of hours worked was perhaps the most obvious of these determinants. If women worked in the marketplace fewer hours than their male counterparts because they took on relatively more work at home, one would expect women’s earnings to be lower than men’s earnings, even if both were paid the same wage rate. The data revealed that indeed male pharmacists worked relatively longer hours, and the difference was statistically significant. Consequently, at least a portion of the observed pay gap seemed to be attributable to differences in amount of work. This variable appeared in the model with linear and quadratic terms to measure rate of change. The model contained 3 human-capital variables: type of academic degree, professional experience, and presence of children in the household. Normally, workers are paid according to the quantity and quality of skills they bring to the labor market; the greater and better their educational attainment, the higher their incomes are likely to be. Because pharmacists receive homogeneous training and have to pass standardized board exams, variation in years of schooling was not as relevant in this study as the type of academic degree attained. In response to recent clinical demands on practitioners, the traditional bachelor of science degree was being phased out at the turn of the century; the Pharm.D. is the new required degree for pharmacists entering the profession in the United States and calls for at least 1 more year of school work. On average, throughout the United States, pharmacists with a Pharm.D. degree are estimated to earn approximately 3% to 6% higher incomes than their colleagues with a baccalaureate degree, although the differential does not control for other variables.46,47 The effect of type of academic degree was measured here with a dummy variable receiving a value of 1 if the jth pharmacist of the ith gender held a Pharm.D. degree, a value of 0 otherwise. The data from this study revealed that almost twice as many women as men possessed a Pharm.D. degree, which was consistent with the volume of women’s entry into the profession in the last 20 years. Professional experience is another earnings determinant related to productivity. Pharmacists are expected to be rewarded for their experience because the longer they have practiced their profession, the greater has been their investment in acquiring skills leading them to do a better job.
289
Linear and quadratic terms were specified for this variable to measure the direction of the effect and the rate of change. Men reported almost twice as many years of professional experience as did women, which might reflect women’s greater propensity to interrupt their careers for the sake of child care and housework,48-50 as well as their relatively recent entry into the pharmacy workforce.41 The other variable related to human capital recorded the presence of children in the household. Women generally devote more time than men to child-rearing activities such as being primarily responsible for children’s physical care, playing with them, and helping with homework plus performing additional housework brought about by children (ie, cleaning, cooking, etc).51,52 Compared with men, the presence of children was expected to influence women’s labor supply, as well as their ability and willingness to invest in their humancapital stock, and consequently their earnings, to a greater extent. On average, female pharmacists analyzed here had fewer children than their male counterparts, which evidences the contention that women are forced to make choices between having a family and pursuing a career that men usually are not forced to make. The second set of variables explored in this earnings determination model focused on jobrelated characteristics reflecting pharmacists’ preferences. The characteristics included were type of primary practice site and main role as practitioner. Both indicators were captured as dummy variables. The practice site variable received a value of 1 if the pharmacist reported that he/ she worked primarily in a retail chain establishment and a value of 0 otherwise; pharmacists also were assigned a value of 1 if their primary work activity was administrative in nature and a value of 0 otherwise. The percentages of practitioners working in retail chain were very similar for both genders, and the percentage of men who reported working as administrators was only marginally greater than the percentage of women. The third set of variables in the model consisted of respondents’ opinions regarding their jobs. These opinions reflected perceptions and attitudes conditioned by the various rewards, or lack thereof, that different groups of individuals experienced in their work environment.53 They measured how much pharmacists liked specific aspects of their professional activities, thus serving as proxies for compensating differentials. The broad hypothesis tested here was that interpretation of what one does influences the wages one demands in exchange for his/her
290
Carvajal et al. / Research in Social and Administrative Pharmacy 8 (2012) 285–297
work, and the interpretation might be different for men and women. The opinion variablesdoverall job satisfaction rating, perception of advancement opportunities, and perception of job securityd were measured along intensity scales of 0 to 10, with 10 showing the greatest intensity. This procedure had the advantage of response homogeneity insofar because pharmacists were able to express their perception of various facets of their professional life following a common measurement standard. Because there was no room for outliers, the mean provided an adequate measure of central tendency. According to their responses, pharmacists in the data set reported middle to moderate levels of all 3 opinion variables. There was no difference between genders in the scores for overall job satisfaction, but women experienced greater levels of perceived advancement opportunities and job security than men. Judging by the size of the standard deviations, responses within gender seemed to be homogeneous. Empirical evidence The estimated least squares coefficients of the explanatory variables, their standard errors, and levels of significance are presented in Table 2 for the unlogged equations and in Table 3 for the semilog equations. All coefficients were
statistically significant for at least one gender in 1 of the 2 versions, and the F ratios were highly significant. The adjusted R2 values were very high for cross-sectional data and explained more of the semilog than the unlogged earnings variation. The positive linear coefficients and negative quadratic coefficients of the workweek variable, all significant, revealed that more hours worked led both men and women to greater wage-andsalary earnings at a decreasing rate. According to the coefficients of the unlogged equations, at the means of the workweek variable, 1 more hour of work yielded an increase in wages and salaries of 1.9% for men and 1.3% for women. The semilog coefficients yielded greater increments in wages and salaries out of 1 more hour of work per week: 2.8% for men and 2.3% for women. The influence of work hours on earnings should not be compared for both genders using solely the values of the least squares coefficients because earnings, as well as workweek distributions, differed for men and women. A full assessment of the magnitude of such influence required the estimation of labor input elasticities of earnings, which measured the ratio of a percentage change in wages and salaries brought about by a percentage change in the number of
Table 2 Values of least squares coefficients, (standard errors), and levels of significance of the unlogged earnings determination model estimated for male and female pharmacists Variable
Estimated least squares coefficients, (standard errors), and levels of significance Term
Constant term Workweek (linear term) Workweek (quadratic term) Pharm.D. degree Experience (linear term) Experience (quadratic term) Number of children Work in retail chain pharmacy Main role administrative Job satisfaction Advancement opportunities Job security F statistic Adjusted R2 a b c
Statistically significant (P ! .01). Statistically significant (P ! .10). Statistically significant (P ! .20).
a b1 b2 g1 g2 g3 g4 d1 d2 q1 q2 q3
Gender Men (i ¼ 1)
Women (i ¼ 2)
23,775.8 3173.9a (314.4) 16.2a (3.7) 10,752.1a (2896.9) 866.1a (279.2) 17.3a (4.8) 2513.7a (780.4) 65.3 (2163.0) 3638.2 (3383.5) 915.1b (501.5) 612.7b (358.8) 642.0c (468.4)
35,213.5 4494.5a (240.2) 37.8a (3.2) 5491.6a (1608.1) 1003.4a (196.4) 22.5a (4.5) 662.1 (603.9) 5392.5a (1490.7) 7361.8a (2532.5) 20.3 (331.2) 328.7c (246.1) 162.3 (358.3)
66.16a 0.534
85.63a 0.612
291
Carvajal et al. / Research in Social and Administrative Pharmacy 8 (2012) 285–297
Table 3 Values of least squares coefficients, their standard errors, and levels of significance of the semilog earnings determination model estimated for male and female pharmacists Variable
Estimated least squares coefficients, (standard errors), and levels of significance Term
Constant term Workweek (linear term) Workweek (quadratic term) Pharm.D. degree Experience (linear term) Experience (quadratic term) Number of children Work in retail chain pharmacy Main role administrative Job satisfaction Advancement opportunities Job security
Gender
u l1 l2 x1 x2 x3 x4 t1
b c
9.229905 0.083438a (0.003590) 0.000782a (0.000048) 0.058198b (0.024034) 0.012114a (0.002935) 0.000298a (0.000067) 0.014222c (0.009026) 0.083002a (0.022281)
(0.033327) (0.004940) (0.003534) (0.00461)
169.21a 0.748
0.090546b 0.001716 0.001757 0.003991
(0.037852) (0.004950) (0.003678) (0.005355)
98.64a 0.646
Statistically significant (P ! .01). Statistically significant (P ! .05). Statistically significant (P ! .20).
hours worked. These elasticities (Table 4), computed at the means of the variables, showed, for example, that in the unlogged equations, a 1% rise in weekly hours worked was expected to increase men’s earnings by 0.73% and women’s earnings by 0.64%. The semilog estimates were more elasticdthe same 1% rise in labor input was expected to increase men’s earnings by 1.10% and women’s earnings by 0.89%. In both equation forms, male pharmacists’ earnings appeared to be more responsive than female pharmacists’ earnings to hours of labor input. Table 4 Earnings elasticities for male and female pharmacists derived at the means of the variables from the estimated coefficients of the unlogged and semilog equations Variables in the equations
Earnings elasticities Men (i ¼ 1) Women (i ¼ 2)
Unlogged equation Workweek Professional experience Presence of children
0.73 0.03 0.04
0.64 0.05
Semilog equation Workweek Professional experience Presence of children
1.10 0.04 0.03
0.89 0.05 0.02
a
Women (i ¼ 2)
9.174006 0.077695a (0.003096) 0.000625a (0.000037) 0.083213a (0.028534) 0.010324a (0.002750) 0.000213a (0.000047) 0.020268a (0.007687) 0.030900c (0.021305) 0.001725 0.005947 0.004857c 0.007350c
t2 j1 j2 j3
F statistic Adjusted R2 a
Men (i ¼ 1)
Square term not statistically significant.
a
Practitioners who held a Pharm.D. degree earned substantially higher incomes than their counterparts without it, a greater differential than those surveyed earlier in the literature: $10,752, or 10.5% more for men, and $5492, or 5.8% more for women. These estimates, however, were similar to the estimates generated by the semilog equations, namely, men earning 8.6% more and women earning 6.0% more than pharmacists practicing without a Pharm.D. degree. From a human-capital perspective, the return to training should be invariant to gender; thus, the differences in the coefficients of the male and female equations suggested the presence of gender bias. Judging by the signs of the coefficients, it might seem that more professional experience led to greater earnings at a decreasing rate for both genders in both equation forms. Yet a more thorough analysis revealed that male pharmacists experienced an age compression problem, whereby beyond a certain point, reported income declined as age increased. In the unlogged equations, at the means of the explanatory variable 1 more year of experience translated into a loss of wages and salaries of 1.1% for men and a gain of 3.2% for women. These estimates indicated that earnings peaked at 25.0 years of experience for men and 22.3 years of experience for women. Because male pharmacists exhibited almost twice as many years of
292
Carvajal et al. / Research in Social and Administrative Pharmacy 8 (2012) 285–297
professional experience as did women, their mean was beyond the peak, whereas the female pharmacists’ mean was not, largely because of their more recent entry into the profession. Perhaps this phenomenon might be attributed in part to the tendency by many older pharmacists, primarily men, to remain in the workforce on a part-time basis,54 that is, earning less because of fewer hours worked, with more years of experience. The differential gender effect of professional experience on earnings was corroborated by the results of the semilog equations although with less intensity. According to the semilog equations, at the means of the explanatory variable 1 more year of experience represented a loss of earnings of 0.2% for men and a gain of 0.3% for women. The experienced elasticities of earnings, computed at the means of the variables, were consistent with these findings. The values of the elasticities, presented in Table 4, were low and similar in both equation forms, thus suggesting that although the coefficients were statistically significant, the effect of professional experience on earnings was small for both genders. The presence of children in the household did not behave as expected. Having children affected male pharmacists’ earnings more than it affected the earnings of female pharmacists. The male pharmacists’ coefficients were highly significant in both equation forms, whereas the female pharmacists’ coefficient lacked significance in the unlogged equation and was barely significant in the semilog equation. On average, in the unlogged equations, men earned 2.5% higher wages for each additional child, whereas in the semilog equation men earned 2.0% and women earned 1.4% higher wages for each additional child. The positive female earnings coefficient for children contradicted a long list of empirical findings reporting a negative effect, often attributed to lower labor supplied on account of children.55-62 Perhaps the presence of children did not affect female pharmacists’ wages and salaries so decisively in this study because some women had left the labor force, even temporarily, on account of children, and there was no access to their earnings. In any event, the earnings of both male and female pharmacists appeared to be children inelastic. Working in a retail environment, the first job preference variable in the equations exerted a definitive positive influence on pay for female pharmacists and a less clearly defined influence for male pharmacists. This might be a compensating differential insofar as role conflict and job stress
commonly are reported in the literature to be relatively high in retail chain settings.63,64 According to the unlogged estimates, women who worked in retail chain pharmacies earned, on average, about $5400 or 5.7% more than women who worked in other settings. This estimate was highly significant, unlike the male pharmacists’ coefficient, which lacked statistical significance. In the semilog equations, the female coefficient, also highly significant, indicated that female pharmacists who work in the retail chain setting got an even greater compensating differential, in the order of 8.7%, whereas the male coefficient, only marginally significant, indicated that men working in retail chain earned 3.1% more than men who practiced in other settings. Working primarily in an administrative capacity, the other job preference variable clearly affected women but not men. The male coefficients lacked statistical significance. Female pharmacists received higher wages and salary for doing administrative work: 7.7% in the unlogged equation and 9.5% in the semilog equation. Collectively, the 3 opinion variables identified in the model did not explain much of the variation in earnings. Five of the 6 female pharmacists’ coefficients, in both equation forms, lacked statistical significance, and the significant coefficients did not behave as compensating differentials. The first of the opinion variables, job satisfaction, was designed to approximate workers’ comprehensive evaluation in a given position and a specific setting. The job satisfaction coefficient for men was positive, suggesting that as male pharmacists’ satisfaction with their jobs increased, so did their earnings; a more likely explanation would be that pharmacists who earned higher wages and salaries tended to be more satisfied with their jobs. Whatever the cause-effect relationship, the impact was substantial. The estimate of the unlogged equation indicated that men scoring a maximum of 10 in the job satisfaction scale earned, on average, $9151 or approximately 8.9% more than men scoring a minimum of 0. Availability of advancement opportunities was the only opinion variable with a statistically significant coefficient for female pharmacists, and this occurred in the unlogged equation. The signs for both genders were positive, meaning that more opportunities led to higher wages and salaries, which accorded with the findings by Carvajal and Hardigan.11 The difference between pharmacists scoring 10 and 0 in this variable represented 6.0% higher earnings for men and 3.5% higher earnings
293
Carvajal et al. / Research in Social and Administrative Pharmacy 8 (2012) 285–297
for women in the unlogged equations, compared with 5.0% higher earnings for men in the semilog equation. Job security, the last opinion variable, also showed positive least squares coefficients for male pharmacists, the opposite of what one might expect if a trade-off existed between wages, a relatively short-term concern, and job security, a longer term consideration. What seemed to emerge here was evidence that men’s better paying jobs were perceived as being more stable, too. On average, compared with male pharmacists scoring a minimum of 0, those scoring a maximum of 10 in this variable earned 6.3% and 7.6% more wages and salaries, according to the unlogged and semilog equations, respectively. Projections of earnings One of the main advantages of estimating separate earnings functions for men and women is that disparities in outcome (ie, wage-and-salary earnings) can be disaggregated into a portion attributable to gender characteristics (ie, average workweek, human-capital stock, job preferences, and perceptions and opinions) and another portion pertaining to how these characteristics affect the outcome. The procedure was originally proposed by Blinder65 and Oaxaca66 and has been used more recently, among others, by Broyles,8 Hersch,67 and Kim.68 This decomposition technique permits the projection of wage-and-salary earnings, on average, by male pharmacists if they exhibited the labor market characteristics typical of alternatively male and female pharmacists and by the same token, the projection of average earnings by female pharmacists if they had the characteristics typical of both genders. According to Table 1, the typical male pharmacist in the sample may be configured as follows: He worked an average of 39.6 hours per week, possessed 27.8 years of professional experience, and had an average of 1.7 children. He had
a 25.6% probability of holding a Pharm.D. degree, a 41.2% probability of working in a retail chain pharmacy, and a 10.9% probability of doing primarily administrative work. His opinion variable scores were 7.2 for job satisfaction, 4.2 for availability of advancement opportunities, and 7.7 for perception of job security. The typical female pharmacist worked 38.5 hours per week, possessed 14.7 years of experience, and had an average of 1.2 children. She had a 49.8% probability of holding a Pharm.D. degree, a 40.0% probability of working in a retail chain pharmacy, and an 8.5% probability of doing primarily administrative work. Her opinion variable scores were 7.2 for job satisfaction, 5.4 for availability of advancement opportunities, and 8.1 for perception of job security. The earnings projection using the least squares coefficients estimated for male and female pharmacists in the unlogged and semilog equations, under the typical labor market conditions specified in the previous paragraph, are presented in Table 5. Within gender, the projections of the unlogged and semilog equations were very similar. The initial unadjusted male-female ratio of 1.079 was only slightly reduced when men’s as well as women’s labor market characteristics were projected with the coefficients estimated for both genders. The projected male-female earnings ratios were quite consistent, thus suggesting that observed gender differences in wages and salaries might not be substantially attributed to differences in number of hours worked, humancapital stock, workers’ preferences, or perceptions and opinions. Rather, they seem to arise from employers’ characteristics, including gender bias. Discussion The probe into wages and salaries earned by male and female pharmacists conducted here
Table 5 Projected wage-and-salary earnings of male and female pharmacists under different scenarios Typical labor market characteristics
Projected wage-and-salary earnings (dollars) Male-female Male pharmacists’ coefficients Female pharmacists’ coefficients earnings ratio
Male pharmacists’ characteristics Unlogged equation Semilog equation
108,769 107,059
102,330 101,024
1.063 1.060
Female pharmacists’ characteristics Unlogged equation Semilog equation
107,204 107,691
101,232 100,031
1.059 1.077
294
Carvajal et al. / Research in Social and Administrative Pharmacy 8 (2012) 285–297
yielded 3 major generalizations. First, men’s and women’s earnings sometimes were influenced by different factors, and when they responded to the same variables, the effect often was different. Men’s (but not women’s) earnings generally were affected by the opinion variables, whereas women’s (but not men’s) earnings were influenced by engaging primarily in an administrative role. Working more hours, holding a Pharm.D. degree, and having more children increased men’s earnings more than women’s earnings, but the opposite was true for possessing greater professional experience and working in retail chain pharmacy. Consequently, the observed heterogeneity in the earnings functions of male and female pharmacists lent methodological validity to the separate equation estimation procedure. The second major generalization of this article is that the effect of some explanatory variables on earnings differed in the unlogged and semilog equations. The male coefficient for working in a retail chain pharmacy and the female coefficient for number of children in the household lacked statistical significance in the unlogged equation but not in the semilog equation, whereas the male coefficient for job satisfaction and the female coefficient for perception of advancement opportunities lacked significance in the semilog equation but not in the unlogged equation. Differences also were apparent in the magnitude of the impact. The estimated semilog elasticities of workweek were greater for both genders than were the estimated unlogged elasticities, as was the impact of professional experience. In addition, the estimated earnings increase for male pharmacists brought about by holding a Pharm.D. degree, and scoring higher in the advancement opportunities index was greater for the unlogged than the semilog form; however, the estimated compensating differentials for women working in retail chain pharmacy and doing administrative work, as well as the effect on earnings of perceived job security by men, were greater in the semilog than the unlogged equations. Notwithstanding these differences in individual explanatory variables, the wage-and-salary earnings projections derived from both equation forms for male and female pharmacists were remarkably similar and yielded nearly identical male-female earnings ratios. Thus, although disparities of interpretation of gender differences in statistical significance and the influence of explanatory variables on earnings might arise from the 2 equation forms, both procedures led to the same conclusion. Compared with the
unlogged equations, the semilog equations were characterized by greater F and adjusted R2 values, especially for men. The third major generalization is that controlling for human-capital stock, job preferences, and opinion variables reduced the initial unadjusted male-female earnings ratio only slightly, even after entering typically male labor market characteristics into the female equations and female characteristics into the male equations. The empirical evidence here points toward the presence of gender bias. This bias, approximately 6%, was relatively small but firm and indicated that the pharmacy profession did not convert individual characteristics into earnings outcomes in the same manner for men and women in South Florida. In other words, after controlling for the effect of the explanatory variables, male pharmacists continued to earn higher income levels than female pharmacists. In interpreting these results, one must take into consideration that the analytical scope of the study was rather narrow. It focused on one profession, with homogeneous training and licensure requirements, and was circumscribed to a small region. These 3 conditions posed advantages and disadvantages. The main advantage was that frequently unaccounted sources of stochastic disturbances such as occupational differences in training and relative scarcity, variation in prices and geographical settings, sociocultural patterns, etc, which usually distort the effect of explanatory variables on earnings, were eliminated, thus producing more accurate estimates and better goodness of fit than would be otherwise. The main disadvantage, and the first limitation of the study, was that the empirical findings might not be fully generalizable. Other occupations and/or the overall labor force may be characterized by relationships that are different from the ones ruling the pharmacy profession. The estimates reported here were largely configured by unique conditions prevailing in South Florida, an area unusually rich in gender, ethnic, and age mix of both population and practitioners. The study also was limited by the absence of concrete data concerning the working status of pharmacists in the sample. Although all respondents provided job-related information, this information might have referred to a previous job if specific pharmacists were not working at the time of response. Furthermore, if a substantial portion of practitioners chose not to respond to the survey because they were not working, either voluntarily
Carvajal et al. / Research in Social and Administrative Pharmacy 8 (2012) 285–297
(eg, retired, raising children, etc) or not (eg, unemployed), the indicators reported here might not be unbiased, although given the sample size, they probably were consistent. Another limitation was related to potential biases arising from the omission of explanatory variables. Although many of the relevant wageand-salary earnings determinants commonly identified in the literature were considered here, others might have been omitted. The standard consideration of critics of gender studies is that if the effect of omitted variables were taken into account, observed gender differences in earnings would disappear. A fourth limitation was that the study rested on a survey administered only once. Therefore, it was inadequate to ascertain whether gender differences in earnings fluctuate over time, especially in the aftermath of the recession, or the extent to which such fluctuations are affected by the proliferation of pharmacy schools in Florida as well as the rest of the United States. Future research ought to include longitudinal data to understand better the evolving role of earnings determinants over time.
Conclusion Beyond its limitations and methodological concerns, the study was successful in comparing the responses of male and female pharmacists’ earnings to human-capital stock, workers’ preferences, and opinion variables; exploring the impact of each explanatory variable; establishing gender differences in statistical significance; and estimating the wage-and-salary gap through earnings projections. The empirical results raise a fundamental policy issue, namely, how to eliminate, or at least reduce, the gender bias appearing in the empirical results. Although the most obvious factor contributing to the gender gap, according to many studies, is direct discrimination against women in the workforce, the literature identifies other possible contributing factors including lack of informal mentoring provided to women, absence of networking with colleagues, less effective salary negotiating behaviors, and failure by employers to comply with equal employment opportunity regulations.69-71 Future research in the area of gender earnings gap should build on the results of the present study by assessing the effects of these variables on male and female pharmacists’ wages and salaries.
295
References 1. Blau FD, Kahn LM. Changes in the labor supply behavior of married women: 1980-2000. Forschungsinstitut zur Zukunft der Arbeit Institute for the Study of Labor Discussion Paper Series. No.13583. 2006. 2. Carvajal MJ. Economic grounds for affirmative action: the evidence on architects and engineers in South Florida. Rev Soc Econ 2006;64:515–538. 3. Graham JW, Smith SA. Gender differences in employment and earnings in science and engineering in the US. Econ Educ Rev 2005;24:341–354. 4. Srinivas S. Social attitudes and the gender pay gap in the USA in recent years. Int J Soc Econ 2007;34: 268–275. 5. Weinberger DH. Earnings by gender: evidence from Census 2000. Mon Labor Rev 2007;130:26–34. 6. Green CT, Jegadeesh N, Tang Y. Gender and job performance: evidence from Wall Street. National Bureau of Economic Research. working paper 12897. 2007. Accessed November 10, 2010. 7. Bradbury KL. Education and wages in the 1980s and 1990s: are all groups moving up together? New Engl Econ Rev 2002;First Quarter:19–46. 8. Broyles P. The gender gap of STEM professions in the United States. Int J Sociol Soc Policy 2009;29: 214–226. 9. Robinson D. Differences in occupational earnings by sex. Int Lab Rev 1998;137:3–31. 10. Bhave DP, Glomb TM. Emotional labour demands, wages and gender: a within-person, between-jobs study. J Occup Organ Psychol 2009;82:683–707. 11. Carvajal MJ, Hardigan PC. Gender and wage earnings differences among pharmacists. J Pharm Finance Econ Policy 2004;13:5–32. 12. Noonan MC, Corcoran ME, Courant PN. Pay differences among the highly trained: cohort differences in the sex gap in lawyers’ earnings. Soc Forces 2005;84: 853–872. 13. Chichilnisky G. The gender gap. Rev Dev Econ 2008; 12:828–844. 14. Cunningham M. Influences of gender ideology and housework allocation on women’s employment over the life course. Soc Sci Res 2008;37:254–267. 15. Fogli A, Veldkamp L. Nature or nurture? Learning and female labor force dynamics. Federal Reserve Bank of Minneapolis, Research Department Staff Report. No. 386, 2007. 16. Poeschl G, Pinto I, Mu´rias C, Silva A, Ribeiro R. Representations of family practices, belief in sex differences, and sexism. Sex Roles 2006;55: 111–121. 17. Quesenberry JL, Trauth EM, Morgan AJ. Understanding the “mommy tracks”: a framework for analyzing work-family balance in the IT workforce. Inform Resour Manag J 2006;19:37–53. 18. Wu H. Can the human capital approach explain lifecycle wage differentials between races and sexes? Econ Inq 2007;45:24–39.
296
Carvajal et al. / Research in Social and Administrative Pharmacy 8 (2012) 285–297
19. Bender KA, Donohue SM, Heywood JS. Job satisfaction and gender segregation. Oxf Econ Pap 2005; 57:479–496. 20. Kim S. Gender differences in the job satisfaction of public employees: a study of Seoul Metropolitan Government, Korea. Sex Roles 2005;52:667–681. 21. Carvajal MJ, Hardigan PC. Pharmacists’ intergender differences in behavior and opinions: is work input an important mediator? Internet J Allied Health Sci Pract 2008;6. 22. Theodossiou I, Vasileiou E. Making the risk of job loss a way of life: does it affect job satisfaction? Res Econ 2007;61:71–83. 23. Allen MW, Armstrong DJ, Riemenschneider CK, Reid MF. Making sense of the barriers women face in the information technology work force: standpoint theory, self-disclosure, and causal maps. Sex Roles 2006;54:831–844. 24. Blau FD, DeVaro J. New evidence on gender differences in promotion rates: an empirical analysis of a sample of new hires. Ind Relat 2007;46: 511–550. 25. Kee HJ. Glass ceiling or sticky floor? Exploring the Australian gender pay gap. Econ Rec 2006;82: 408–427. 26. Leggon CB. Women in science: racial and ethnic differences and the differences they make. J Tech Tran 2006;31:325–333. 27. Leung D. The male/female earning gap and female self-employment. J Soc Econ 2006;35:759–779. 28. Levy H. Health insurance and the wage gap. National Bureau of Economic Research. working paper 11975. January 2006. Accessed 10.11.10. 29. Long A. Happily ever after? A study of job satisfaction in Australia. Econ Rec 2005;81:303–321. 30. Luce S, Brenner M. Women and class: what has happened in forty years? Mon Rev 2006;58:80–93. 31. Black DA, Haviland AM, Sanders SG, Taylor LJ. Gender wage disparities among the highest educated. J Hum Resour 2008;43:630–659. 32. Desselle SP. Survey of certified pharmacy technicians in the United States: a quality-of-worklife study. J Am Pharm Assoc (2003) 2005;45:458–465. 33. Jacobson CC, Resneck JS Jr, Kimball AB. Generational differences in practice patterns of dermatologists in the United States. Arch Dermatol 2004;140: 1477–1482. 34. Kahaleh A, Gaither CA. Effects of empowerment on pharmacists’ organizational behaviors. J Am Pharm Assoc (2003) 2005;45:700–709. 35. Lapane KL, Hughes CM. Job satisfaction and stress among pharmacists in the long-term care sector. Consult Pharm 2006;21:287–292. 36. Lin BY, Yeh Y-C, Lin W-H. The influence of job characteristics on job outcomes of pharmacists in hospital, clinic, and community pharmacies. J Med Syst 2007;31:224–229. 37. LoBuono C. Exclusive R.Ph. salary survey: cashing in. Drug Top 2003;147:46–48. 54–55.
38. Weeks WB, Wallace AE, MacKenzie TA. Gender differences in anesthesiologists’ annual incomes. Anesthesiology 2007;106:806–811. 39. Yutzie JD, Shellito JL, Helmer SD, Chang FC. Gender differences in general surgical careers: results of a post-residency survey. Am J Surg 2005;190:955– 959. 40. van Klaveren C, van Praag B, van den Brink HM. A public good version of the collective household model: an empirical approach with an application to British household data. Rev Econ Household 2008;6:169–191. 41. Carvajal MJ, Armayor GM, Hardigan PC. In: Socioeconomic Profile of Pharmacists in South Florida, 1. Fort Lauderdale, FL: Nova Southeastern University, College of Pharmacy Monograph Series; 2010. 42. Jovic E, Wallace JE, Lemaire J. The generation and gender shifts in medicine: an exploratory survey of internal medicine physicians. BMC Health Serv Res 2006;6:55. 43. Kennedy MM. Managing different generations requires new skills, insightful leadership. Physician Exec 2003;29:20–23. 44. Shields MC, Shields MT. Working with generation X physicians. Physician Exec 2003;29:14–18. 45. Southard G, Lewis J. Building a workplace that recognizes generational diversity. Public Manag 2004; 86:8–12. 46. Levy S. Take this job and love it? Drug Top 2008;152: 34. 36–38;40;42. 47. Ukens C. Pay dirt: pharmacist salaries are still climbing, but at a slower rate. Our exclusive survey shows who gets top dollar. Drug Top 2007;151:38–42. 48. Aanerud R, Morrison E, Homer L, Rudd E, Nerad M, Cerny J. Widening the lens on gender and tenure: looking beyond the academic labor market. NWSA J 2007;19:105–123. 49. Blau FD, Ferber MA, Winkler AE. The Economics of Women, Men, and Work. Upper Saddle River, NJ: Prentice Hall; 2002. 50. Keaveny TJ, Inderrieden EJ, Toumanoff PG. Gender differences in pay of young management professionals in the United States: a comprehensive view. J Labor Res 2007;28:327–346. 51. Blank RM, Shierholz H. Exploring gender differences in employment and wage trends among less-skilled workers. National Bureau of Economic Research working paper. 12494. 2006. Accessed 10.11.10. 52. McElwain AK, Korabik K, Rosin HM. An examination of gender differences in work-family conflict. Can J Behav Sci 2005;37:283–298. 53. Malka A, Chatman JA. Intrinsic and extrinsic work orientations as moderators of the effect of annual income on subjective well-being: a longitudinal study. Pers Soc Psychol Bull 2003;29:737–746. 54. Knapp KK, Cultice JM. New pharmacist supply projections: lower separation rates and increased graduates boost supply estimates. J Am Pharm Assoc (2003) 2007;47:463–470.
Carvajal et al. / Research in Social and Administrative Pharmacy 8 (2012) 285–297 55. Blasius JJ, Pae E-K. Work-pattern differences between male and female orthodontists. Am J Orthod Dentofacial Orthop 2005;128:283–290. 56. Cohany SR, Sok E. Trends in labor force participation of married mothers of infants. Mon Labor Rev 2007;130:9–16. 57. Corrigall EA, Konrad AM. Gender role attitudes and careers: a longitudinal study. Sex Roles 2007; 56:847–855. 58. Ginther DK, Kahn S. Does science promote women? Evidence from academia 1973-2001. National Bureau of Economic Research working paper Series. No. 12,691. 2006. 59. Mosisa A, Hipple S. Trends in labor force participation in the United States. Mon Labor Rev 2006;129: 35–57. 60. Napari S. The early career gender wage gap. The London School of Economics and Political Science Center for Economic Performance Discussion Paper. No.738. 2006. 61. Pettit B, Hook J. The structure of women’s employment in comparative perspective. Soc Forces 2005; 84:779–801. 62. Stephan PE, Levin SG. Leaving careers in IT: gender differences in retention. J Tech Tran 2005;30:383– 396. 63. Mott DA, Doucette WR, Gabriel PE, Pendersen CA, Schommer JC. Pharmacists attitudes toward worklife:
64. 65.
66. 67. 68. 69.
70.
71.
297
results from a national survey of pharmacists. J Am Pharm Assoc (2003) 2004;44:326–336. Ukens C. Pharmacists stressed but satisfied with jobs, study finds. Drug Top 2004;148:42–43. Blinder AS. Wage discrimination: reduced form and structural estimates. J Hum Resour 1973;8: 436–455. Oaxaca R. Male-female wage differentials in urban labor markets. Int Econ Rev 1973;14:693–709. Hersch J. Sex discrimination in the labor market. Found Trends Microecon 2006;2:281–361. Kim M. Race and gender differences in the earnings of black workers. Ind Relat 2009;48:466–488. Kim J. Gender inequality in the U.S. labor market: evidence from the 2000 census. In: Karsten M, ed. Gender, Race, and Ethnicity in the Workplace [Three Volumes]: Issues and Challenges for Today’s Organizations. Vol ABC-CLIO eBook. Santa Barbara, CA: Praeger; 2006:269–290. Accessed 10.03.11. Gerhart B, Rynes S. Determinants and consequences of salary negotiations by male and female MBA graduates. J Appl Psychol 1991;76:256–262. Barnes M. Glass ceiling. In: Brzuzy S, Lind A, eds. Battleground: Women, Gender, and Sexuality [Two Volumes]. ABC-CLIO eBook Collection. Santa Barbara, CA: Greenwood; 2007:233–236. Accessed 10.03.11.