GB545 Unit 8 Pay Equity Analysis Example

From 9.5 Percent to 2.2: A Unit 8 Pay Equity Analysis of 520 Engineers

[Student Name]

Purdue University Global

GB545: Strategic Total Rewards and Financial Impact

Unit 8 Pay Equity Analysis

[Instructor Name]

[Date]

The company, its engineers and the dataset are composite; the figures were produced by script from a composite dataset, the statutes and sources cited are real, and the analysis is educational, not legal advice.

Purpose and Legal Context

The analysis asks whether women and men in comparable engineering jobs at Tellwater are paid the same, after accounting for legitimate factors that affect pay. The Equal Pay Act bars paying one sex less than the other for substantially equal jobs (29 U.S.C. § 206, 2024), and prohibits discrimination in compensation (42 U.S.C. § 2000e-2, 2024). Studies like this one are often commissioned through counsel so that their findings can be reviewed before action is taken; this analysis notes that practice without advising on it. A parallel model examining race and ethnicity was run separately and is not reported here.

Population

The population is all 520 engineers employed on the analysis date: 126 women and 394 men. Mean base pay is $123,782 for women and $136,748 for men, a raw difference of 9.5 percent.

Distribution by level. Engineers are placed in five levels. Women: 28 at level one, 34 at level two, 32 at level three, 22 at level four and 10 at level five. Men: 70, 88, 112, 82 and 42. So 62 of the 126 women, 49.2 percent, sit in the two lowest levels, against 158 of 394 men, 40.1 percent.

Method

Each model regresses the natural logarithm of annual base salary on an indicator for female and a set of controls. The logarithm is used because pay differences tend to be proportional, so a single coefficient can describe a percentage difference across junior and senior engineers alike. A coefficient on female is converted to a percentage by taking its exponential and subtracting one. Wooldridge (2020) describes this specification and its interpretation, and the models were estimated by ordinary least squares in a short script so that anyone with the data could reproduce them.

Model one includes only the female indicator.

Model two adds indicators for the five engineering levels.

Model three adds years of tenure and its square, indicators for the three work locations and indicators for hardware, firmware or software function.

Results

Model one. The coefficient on female is negative 0.0997, which converts to a gap of 9.5 percent, reproducing the raw difference. R-squared is 0.031, meaning sex alone explains very little of the variation in pay.

Model two. Adding level reduces the coefficient to negative 0.0294, a gap of 2.9 percent, and raises R-squared to 0.912. Level explains most of the variation in engineers' pay, and it absorbs about two-thirds of the raw gap.

Model three. Adding tenure, location and function reduces the coefficient to negative 0.0223, with a standard error of 0.0048 and a t-statistic of negative 4.63. R-squared is 0.959. The coefficient converts to a gap of 2.2 percent, about $2,740 a year at the women's mean salary. A t-statistic of that size means the estimate would almost never appear through random variation alone if women and men with the same level, tenure, location and function were actually paid alike.

Even after comparing women and men at the same level, with the same tenure, location and function, women are paid about 2.2 percent less, and that difference is unlikely to be chance.

What the Controls Absorbed

Most of the narrowing from 9.5 to 2.2 percent comes from level. That is the correct comparison for the question of equal pay for comparable work, because a level five engineer and a level one engineer do not do comparable jobs. But it does not mean level is innocent. If women are placed at lower levels when hired, or promoted more slowly, then controlling for level hides part of the gap rather than explaining it. Blau and Kahn (2017) show that in national data much of the pay difference between women and men traces to the occupations and industries they work in, and that placement within employers matters as well.

The distribution table is the second finding. Women are 24 percent of Tellwater's engineers but are overrepresented at the two lowest levels. The analysis cannot say why from pay data alone. It could reflect when women were hired, differences in prior experience or how hiring managers set starting levels and promotion decisions.

Remedy

For each woman, the model three equation, with the female coefficient set to zero, gives a predicted salary: what an engineer with her level, tenure, location and function would be paid if sex had no effect. Ninety-one of the 126 women are paid below that predicted salary. Raising each to her predicted pay costs $478,989 in base salary, an average of about $5,264 for those adjusted, or $538,384 once loaded for payroll taxes and benefits at the company's factor of 1.124. Women already at or above predicted pay are not adjusted, and men below their predicted pay are reviewed separately under the company's normal process.

Checking the Remedy

Before the remedy was priced, the model three predictions were checked in two ways. First, the gaps between each engineer's actual and predicted salary were examined by level and location to make sure no group was systematically poorly predicted, which would make its adjustments unreliable. None was. Second, the 91 adjustments were reviewed by the compensation team against each engineer's file, to catch cases where a documented factor outside the model, such as a recent transfer from a lower-cost location under a written agreement, explained the shortfall. Four such cases were found, and the model's adjustment was kept for each after counsel's review because the documented factor did not account for the full difference. After the adjustments, rerunning model three gives a female coefficient near zero and not statistically significant, which confirms that the remedy closes the adjusted gap it was designed to close.

Recommendations

Adjust pay now. Make the $478,989 in adjustments in the next payroll cycle rather than waiting for the merit cycle, so the gap does not compound with percentage increases.

Review level placement and promotion. Examine starting levels for engineers hired in the past five years and promotion rates by sex, since the distribution suggests the larger part of the gap may lie in level itself.

Repeat the analysis each year. Run the same three models annually before merit decisions are finalized, so any new gap is caught early.

Limits

The models account only for the factors in the data. Performance ratings were excluded because they may themselves reflect bias, and prior experience before Tellwater was not available. Groups at the highest levels are small, so estimates for them alone would be unstable.

References

Blau, F. D., & Kahn, L. M. (2017). The gender wage gap: Extent, trends, and explanations. Journal of Economic Literature, 55(3), 789-865. https://doi.org/10.1257/jel.20160995

29 U.S.C. § 206 (2024). https://www.law.cornell.edu/uscode/text/29/206

42 U.S.C. § 2000e-2 (2024). https://www.law.cornell.edu/uscode/text/42/2000e-2

Wooldridge, J. M. (2020). Introductory econometrics: A modern approach (7th ed.). Cengage.