Athens, Ga. – The gap in wages between men and women has decreased sharply over the past 30 years, and a new University of Georgia study reveals that decline was even greater than previously recognized.
The research, published in the March edition of the journal Social Science Research, is one of the first to take into account a common bias that the authors argue has skewed previous results: People (especially men) tend to inflate their incomes when surveyed. Conversely, both men and women tend to underestimate income when reporting on someone else’s behalf. According to the UGA researchers, these biases, combined with changes over time in the proportion of men and women who report their own earnings, have obscured the decline in the gender wage gap.
The traditional estimation method, in which data from self-reports of income are combined with so-called proxy reports from others without any adjustment, indicates that the wage gap closed by 16 percent from 1979 to 2009. After removing gender-specific reporting effects, the UGA researchers find that the gap closed by 22 percent over the same period-a difference of nearly 50 percent.
“It appears that the gender wage gap has closed more quickly over time and by a greater amount than previous estimates suggest,” said study co-author Jeremy Reynolds, associate professor of sociology in the UGA Franklin College of Arts and Sciences.
Reynolds and study co-author Jeffrey Wenger, associate professor in the UGA School of Public and International Affairs, note that the wage gap was greatly underestimated in the 1980s, when most women self-reported and many men had spouses who reported on their behalf. In more recent years, when closer to half of men and women self-reported, estimates of the wage gap are likely to have been more accurate.
To be sure, the current gender wage gap-just over $6 per hour-is still significant, and the finding that the wage gap has shrunk more than once thought may be cold comfort for those experiencing today’s disparities. Still, Reynolds and Wenger’s research emphasizes that statistically controlling for how survey data is gathered is important for understanding earnings outcomes and other social behaviors. “We were surprised at how large these reporting effects were and that they were different for men and women,” Wenger said.
The study had its origins in the team’s previous work, which examined employment decisions made by married couples when their health status changed. As they were analyzing their data, they found that men and women often gave conflicting answers about their spouse’s health. With their interest piqued, they sought to explore whether similar discrepancies exist in the reporting of financial and employment information-and how those differences affect the conclusions that social scientists reach when examining long-term trends.
The researchers examined data from the Current Population Survey, a joint effort between the Bureau of Labor Statistics and the U.S. Census Bureau that is one of the most frequently used sources of labor market data. To help understand the roles that self and proxy reports play in reported wages, they tracked a subset of respondents over time to observe how changes in their reporting status were related to changes in wages. To account for other variables that might influence wages over time, they limited their analysis to respondents who were married, employed full time and in the same occupation and industry at both points in time.
The researchers emphasize their analysis is certainly not the final word on wage differences between men and women, since other statistical techniques could produce different estimates of how the wage gap has changed. They also emphasize they are not discounting the value of proxy reports, since previous studies suggest that survey responses from others are especially accurate for behaviors that people might be embarrassed about, such as dropping out of school or being incarcerated. What is needed, they said, is for researchers to be aware of potential reporting biases that may vary by topic, gender and time-and to correct accordingly.
“If you’re going to make comparisons between groups, it becomes pretty questionable to combine proxy and self-reports without thinking about those differences,” Reynolds said.