Trend Lines

For decades, one of the most salient features of women’s status in the labor market was their tendency to work in a fairly small number of relatively low-paying, predominantly female jobs. That has changed and as it has differences in aggregate average wages between men and women (the most often cited data on the “wage gap”) have narrowed. Still, the aggregate average wages of women are  almost 20 percent lower than the aggregate average for men.  But that data say almost nothing about whether employers compensate men and women in similar positions differently.  Only a thin veneer of empiricism covers what seems to me to be fundamentally ideological arguments in the public debate on what, if anything, government should do to address the gender wage gap issue.  Proponents of “paycheck equity” measures don’t often acknowledge that research shows that differences in labor force participation, occupation, experience, etc. account for the majority of the difference between the average wages of men and women. While those who believe markets (including labor markets) operate perfectly, and who reject almost all government interjections into markets, fail to note that even accounting for differences in occupations etc., there is still a significant portion of the pay differences between men and women that is unaccounted for or “unexplained.”  The unexplained portion of the wage gap is probably 8 percent or less and whether this difference in wages is the result of biases or discrimination that can be remedied by legislation is the fundamental public policy at issue.

Living with four, smart, talented, women with strong opinions makes me approach this subject with trepidation.  It is possible that one or more of them may talk to me even less than they do now  (I am comforted knowing that it gets incrementally more difficult for this to occur as the amount approaches zero) if my interpretation of the issue is in error in their eyes.  Still, as the father of daughters, I want to know if there are any biases in the labor market that may constrain their earnings and limit their ability to care for me in my dotage.

To truly know whether biases or discrimination (intentional or not) contribute to the wage gap you have to do the impossible, you have to compare the wages of men and women with identical ages, education, qualifications, experience, motivations, work habits etc., in identical jobs, working for identical companies, and you have to do it for the entire male and female workforce (across all occupations).  Nobody can do that so in the absence of a Boston Celtics, Boston Bruins, or UNH hockey game on television this past Saturday evening I undertook a smaller task to gain the kind of first-hand insight into the issue that only analyzing data can provide.

My goal was to compare the wages and salaries of full-time working men and women who are as similar in education, age, experience, occupation and other factors (marriage, children etc.) as possible, to see if wage and salary difference exist when as many key characteristics of individuals are as similar as possible.  The 2009 March Supplement of the U.S. Census Bureau’s “Current Population Survey” contains information on the college majors of those who have earned degrees, in addition to the usual labor force, earnings, and demographic information it provides about survey respondents.  I extracted microdata for over 5,000 survey respondents nationally whose highest educational attainment was a bachelor’s degree in accounting and who were employed full-time as  accountants or auditors.  Focusing on just one occupation and one level of educational attainment limits the ability to generalize my results (that is, results and conclusions may not apply to other occupations or to other levels of educational attainment) but choosing just one occupation that has a relatively equal employment distribution between men and women, and just one education level is likely to provide the cleanest evidence or lack thereof for the wage gap.

Many factors contribute to gender wage differences between men and women in the same occupation.  In the end, like other researches, my analysis showed that under 10 percent of the difference between the wages and salaries of men and women (in accounting) cannot be explained by difference in hours worked, the presence of children etc.  This little exercise cannot say whether those “unexplained” or unaccounted for factors mean discrimination or bias or less pernicious forces that are not captured or measured with available variables.  Regardless, the results were fascinating (to me at least) for a number of economic and sociological reasons.

At all age levels, women in accounting worked (on average) fewer hours weekly than did men (Figure below).  This was true for women with and without children and whether they were married or not.  Regression models show that each additional hour worked per week actually added more to the annual wage and salary of women than to men ($641 to $568).  This suggest to me that women are rewarded equally for additional labor and that some of the wage gap in accounting is simply attributable to the longer (on average) hours worked by men.

Child birth and care is associated with lower labor force participation among women but on average, women in accounting worked fewer hours than did men whether or not they had children.  This is not an epiphany because child care responsibilities still disproportionately fall on women.  The presence of children, however, does have a much more negative impact on hours worked for women than it does for men. The effect on hours worked diminishes by age for women – presumably because children are more likely to be older and in need of less care.  Interestingly, among younger men, the presence of children reduces hours worked (not as much as it does for women) but among older men, the presence of children is associated with small increases in hours worked.  I hope this implies larger roles in child care among younger men that will contribute to a further narrowing of the wage gap.

Wages grow with age and experience (typically for most individuals into their late fifties) but among accountants with bachelor’s degrees, they grow less, on average, annually for women than for men.  The cumulative effects of hours worked, breaks in labor force participation among women for child care, and other factors can account for some of this.  But key factors that can determine how much wages grow over time (getting promotions, asking for raises, motivation etc.) cannot be discerned from the data.  If women aren’t promoted as readily or don’t seek raises as often or as large,  their annual wage growth would be expected to be lower.

The figure below shows some of the wage and salary impacts for men and women of different variables.  Interestingly, marriage seems to have a much more positive impact on the earnings of men  than it does for women,  adding $3,415 to earnings of men but just $920 for women when controlling for other variables (i.e. age and hours worked, children,  etc.).

Apparently some stereotypes are based in reality and it makes me wonder what aggregate economic and societal implications this may be having as more young men seem to resist or delay marriage or long-term relationships.  The big take away for me was the impact that marriage as well as the “unexplained” difference attributable to gender had on wages and salaries when earnings were not “constrained,” meaning that the data included the earnings from the highest earners, the “outliers” or top 15 percent who earn much higher salaries than typically earned in the occupation.  For most of my analyses I selected data for individuals whose wages and salaries were between the 15^th and 85^th percentile of all earners among accountants.  But I also ran the same analyses using data that was not constrained at the top of the earnings scale.  This would likely include business owners and individuals in higher level positions.   When data was not constrained based on wages and salaries, the impact of marriage on the wages and salaries of women was significantly negative in contrast to the wage and salary constrained data where it was positive (albeit a much smaller relationship with earnings than for males).  The unexplained difference attributable to gender was also disproportionately large compared to the constrained data set.  What this says to me is that the largest evidence of a gender wage gap occurs at the upper end of wages and salaries and may be more attributable to lower percentages of women in higher positions (a “glass ceiling” effect) than a paycheck fairness effect.  Does this mean that women are at a disadvantage in becoming high earners if they are married, and if so, is it the result of personal choices or does the labor market systemically punish them relative to men?  Does being married necessarily mean that a woman will be less likely to rise to the top of their occupation?   These results do suggest a glass ceiling that contributes to the wage gap and this should trouble all fathers of daughters, brothers,  husbands and even friends.  Unfortunately, it doesn’t say much about what we can do about it.

We do have laws that prevent wage and salary discrimination based on race and gender and yet the wage gap persists (even as it has shrunk).  As more women populate higher levels of occupations and organizations (it will happen as I suggest here and here and in several other posts), it is likely that whatever unexplained gender-based differences in wages and salaries (that are not based on occupation, education, experience etc.) will decline rapidly.  To me, addressing why more women don’t reach higher positions in organizations  is likely to be a much more effective prescription for whatever gender gap continues to exist than are mandates that seek to compensate for factors that we aren’t even sure contribute to wage and salary differentials between genders.