Issue Briefs

Following are CBO’s responses to questions for the record following a House hearing into the CBO’s recent report comparing federal and private sector pay at which the AFGE union questioned the methods CBO used to reach its conclusion that overall federal employees earn 3 percent more, although with a wide variation by educational level.

Question. The American Federation of Government Employees (AFGE) has critiqued aspects of the Congressional Budget Office (CBO) study, saying it is not an accurate por¬trayal of the federal employee compensation differential. According to information provided to Committee staff, you will be engaging with AFGEin the coming weeks to discuss the methodology used in your report. After such meeting and pursuant to my request at the hearing, please respond to the criticisms leveled at your report by AFGE. Please discuss CBO’s treatment of demographic traits, including education, in the report. What was the reason for sorting employees by educational attainment? What are the benefits of doing so? Please respond to any additional criticisms of your report you discussed with AFGE.

Answer. On June 12, 2017, CBO discussed the concerns raised in the AFGEtestimony with Jacqueline Simon and another AFGEstaff member. Those concerns related to CBO’s adjustments for differences in demographic traits between the federal and private-sector workforces, its comparison of federal and private-sector compensation by workers’ level of education, and the approach it used to compare benefits between the sectors.

In its analysis, CBO adjusted for differences between federal and private-sector employees in various demographic traits (age, sex, race, ethnicity, marital status, immigration status, and citizenship) as well as education, years of work experience, occupation, size of employer, geographic location, and veteran status. Adjusting for differences in demographic traits is common practice when comparing compensation between two groups because some demo¬graphic traits are correlated with unobservable characteristics that can affect compensation. For example, if immigrants tend to be less proficient in English than native-born workers, then adjusting for immigration status will account for some of the differences in English proficiency between federal and private-sector workers even though English proficiency was not observable in the data. If adjustments for demographic traits had substantially changed the estimated differences in compensation between the sectors, then it would have been important to understand why that was the case, as such results might suggest that differences between the public and the private sector in average compensation reflected discrimination in one of those sectors on the basis of race, sex, immigration status, or other nonmerit factors. However, adjusting for demographic traits did not have much effect on CBO’s estimates of the differences between federal and private-sector compensation.

Reporting results by workers’ level of educational attainment reveals differences between fed¬eral and private-sector compensation that would not have been evident if CBO had averaged the results for workers at all levels of education. CBO categorized workers by educational attainment because education plays a particularly large role in determining compensation— in CBO’s analysis, it was the most important explanatory factor.

CBO estimated the cost of private-sector benefits using data from the Bureau of Labor Statistics’ National Compensation Survey (NCS), which has been used by economists at that agency to compare private-sector compensation with that of the employees of state and local governments. Because the NCS does not survey federal employees, CBO used data from the Office of Personnel Management to estimate the cost of federal employees’ bene¬fits. CBO then compared the costs of the benefits provided to federal and to private-sector employees, accounting for the same differences in workers’ characteristics that were used to analyze wages (education, years of work experience, occupation, size of employer, geo¬graphic location, veteran status, and demographic traits). To do that accurately, the agency adjusted the costs of defined benefit pension plans drawn from the NCS data so that they were comparable to the estimates of those costs for the federal government. Before 2006, many private-sector employers calculated the costs of their pension plans as though the risky assets that funded them would accrue returns at the rates they had in the past. The Pension Protection Act of 2006 generally requires private-sector employers to discount future annu¬ity payments at lower rates, but those rates are still higher than recent rates of return for Treasury bonds. For that reason, CBO increased the costs reported by the NCS so that they were consistent with the Treasury bond rates used for federal pension obligations.

Question. Were there any important cost drivers of federal compensation that could not be included in your study due to lack of data? If so, please state which benefits were excluded and explain the data barriers that prevent CBO from factoring such benefits into a federal employee’s total compensation package.

Answer. CBO’s analysis included wages and the costs of benefits such as retirement income, health insurance, paid leave, and other, legally mandated, benefits. In the agency’s judg¬ment, those are the most important drivers of the cost of federal compensation. The analysis excluded certain benefits some workers receive—for example, the above-market rate of return the federal government offers its employees through the G Fund (one of the investment options in their retirement plan) and the stock options that some private-sector businesses provide to their employees. Because such benefits are less costly to provide or less commonly available than those included in the analysis, CBO expects that their effect on the cost of important to understand why that was the case, as such results might suggest that differences between the public and the private sector in average compensation reflected discrimination in one of those sectors on the basis of race, sex, immigration status, or other nonmerit factors. However, adjusting for demographic traits did not have much effect on CBO’s estimates of the differences between federal and private-sector compensation.

Reporting results by workers’ level of educational attainment reveals differences between fed¬eral and private-sector compensation that would not have been evident if CBO had averaged the results for workers at all levels of education. CBO categorized workers by educational attainment because education plays a particularly large role in determining compensation— in CBO’s analysis, it was the most important explanatory factor.
CBO estimated the cost of private-sector benefits using data from the Bureau of Labor Statistics’ National Compensation Survey (NCS), which has been used by economists at that agency to compare private-sector compensation with that of the employees of state and local governments.2 Because the NCS does not survey federal employees, CBO used data from the Office of Personnel Management to estimate the cost of federal employees’ bene¬fits. CBO then compared the costs of the benefits provided to federal and to private-sector employees, accounting for the same differences in workers’ characteristics that were used to analyze wages (education, years of work experience, occupation, size of employer, geo¬graphic location, veteran status, and demographic traits). To do that accurately, the agency adjusted the costs of defined benefit pension plans drawn from the NCS data so that they were comparable to the estimates of those costs for the federal government. Before 2006, many private-sector employers calculated the costs of their pension plans as though the risky assets that funded them would accrue returns at the rates they had in the past. The Pension Protection Act of 2006 generally requires private-sector employers to discount future annu¬ity payments at lower rates, but those rates are still higher than recent rates of return for Treasury bonds. For that reason, CBO increased the costs reported by the NCS so that they were consistent with the Treasury bond rates used for federal pension obligations.

Question. Were there any important cost drivers of federal compensation that could not be included in your study due to lack of data? If so, please state which benefits were excluded and explain the data barriers that prevent CBO from factoring such benefits into a federal employee’s total compensation package.

Answer. CBO’s analysis included wages and the costs of benefits such as retirement income, health insurance, paid leave, and other, legally mandated, benefits. In the agency’s judg¬ment, those are the most important drivers of the cost of federal compensation. The analysis excluded certain benefits some workers receive—for example, the above-market rate of return the federal government offers its employees through the G Fund (one of the investment options in their retirement plan) and the stock options that some private-sector businesses provide to their employees. Because such benefits are less costly to provide or less commonly available than those included in the analysis, CBO expects that their effect on the cost of federal compensation and on the difference between federal and private-sector compensation would be limited.

Additional factors can affect the cost of recruiting and retaining a highly qualified federal workforce. For example, compared with the private sector, the government may be able to attract and retain highly qualified employees at lower levels of pay because federal jobs offer greater job security. Conversely, because federal employees’ total compensation includes a greater share of retirement benefits, which workers might find less valuable than an equiv¬alent amount of cash received today, total compensation might need to be higher in the government than in the private sector. However, quantifying the extent to which such factors affect the amount of compensation the government needs to provide to workers was beyond the scope of CBO’s analysis, and the agency is not aware of data sets that could be used to accurately quantify such effects.

Question. CBO found the dispersion of federal wages is constrained relative to those in the private sector. What may be causing this? What are the practical implications of these constraints?

Answer. The difference between the wages of the highest- and the lowest-paid employees was smaller in the federal government than in the private sector, even after accounting for employees’ education and other observable traits. The narrower dispersion of wages among federal employees may reflect the constraints of federal pay systems, which make it harder for managers to reward the best performers or to limit the pay of poor performers. Such constraints probably both limit the government’s ability to attract and retain highly qualified workers and cause the government to pay more than necessary to retain workers who are less productive.