Last May, the Atlanta Fed temporarily paused updates to the Wage Growth Tracker (WGT), a decision made in response to a change in the underlying data used to construct the WGT. The WGT uses data from the Current Population Survey (CPS), which is jointly administered by the US Census Bureau and Bureau of Labor Statistics, to depict the median growth rate in individual hourly earnings. To generate the WGT, the Atlanta Fed computes the percent change in hourly-equivalent earnings for an individual in the CPS between the current month and the same month a year before and derives the median of individual wage growth on a monthly basis. The published WGT is the three-month moving average of this monthly series. Recently, the US Census Bureau changed its rule about how the CPS reports data on earnings over a certain amount as part of an effort to improve data quality. The data under the new rule was used for the first time in April for the WGT calculations. The Atlanta Fed decided to halt updates to the WGT to investigate the potential impacts associated with the change to the underlying data on the WGT. Based on our investigation, we are confident of the WGT's future performance. However, users should be aware that this change in the data gives rise to a different sample used in the calculation of the WGT, a change that leads to modest differences in the WGT calculation—as we discuss below.
A change to the topcoding rule—and why it could affect the WGT
Prior to this recent change, all workers earning more than a certain income threshold were assigned the same fixed value in the CPS data. This procedure, called topcoding, was implemented to prevent a survey respondent with extremely high earnings from being identified by the publicly available information in the CPS data. Starting in 2003, the topcoding cap kicked in at $2,885 per week (or about $150,000 per year). With wage growth over time, this fixed earnings cap meant that more and more workers' earnings data were being truncated, which reduced the sample size for analysis. To improve the data quality, Census changed the fixed earnings cap to a dynamic topcoding rule, which instead caps the top 3 percent of earnings every month.
When implementing this dynamic rule, however, Census capped the top 3 percent of earnings separately for earnings reported by hour and by week (the CPS reports earnings on either a per-hour or per-week basis). Given that the actual data for the capped earnings are not available in the publicly released CPS micro dataset, the WGT's calculation doesn't include these earnings. In other words, the recent change in topcoding rules impacts the sample used to calculate the WGT and thus could affect the median level of wage growth used to construct the WGT.
To better understand the rule change's impact on the sample included in the WGT calculation, consider the following. Under the old rule, the capped earnings always resided at the top of the earnings distribution because weekly and hourly reported earnings were treated symmetrically by capping weekly earnings at $2,885 and hourly earnings at the equivalent level after being converted to weekly earnings using hours worked per week. Hence, under the old rule, the equivalent cap for hourly earnings depended on hours worked per week and was $72 per hour for a 40-hour work week. Under the new dynamic rule, the cap for earnings reported by hour is no longer equivalent to the cap for earnings reported by week because the 3 percent rule is separately applied to these two categories. The implied cap under the new rule for earnings reported by week is much higher than $2,885 per week, and the implied cap for earnings reported by hour is much lower than the equivalent levels under the old rule. It is important to reiterate that the dynamic nature of the new topcoding rule implies that the topcoding caps will change every month. Using April 2024 (subject to the new topcoding rule) as an example, the highest reported noncapped weekly earnings exceeded $4,000 per week, and the highest reported noncapped hourly earnings were just over $55 per hour. This means that under the new rule, earnings between $2,885 and $4,000 were available (observable) in April 2024 for workers reporting their earnings on a weekly basis, but under the old rule, data about these earnings would not have been available. On the other hand, for workers reporting their earnings by hour and usually working 40 hours a week, hourly earnings between $55 and $72 were capped and not available in April 2024 but would have been available under the old rule. So although the implementation of the new topcoding rule increases the number of noncapped weekly earnings, it reduces the number of noncapped hourly earnings.
Implications for the WGT
Since actual data about capped earnings are unavailable, we can't definitively answer the seemingly straightforward question, "How does the recent change in the topcoding rule affect the WGT?" But to shed light on the potential impact of the rule change, we created an earnings sample that is consistent across time regarding topcoding, thus neutralizing the rule changes' impact. We created this sample by applying the most restrictive components of both the old and new rules. Specifically, we apply the old rule of topcoding for all weekly earnings above $2,885 for the data after the rule change, and we also apply the new rule of topcoding to the top 3 percent of workers reporting hourly earnings for all the data before the rule change. An alternative WGT can then be constructed from this newly created sample, which can be compared with the Atlanta Fed WGT that uses noncapped earnings whenever available.
Figure 1 compares the Atlanta Fed WGT (blue line) to the alternative WGT (orange dashed line) over the last two years. As you can see, these series are closely aligned both before and after the recent rule change, an observation that makes us confident that the WGT will remain a reliable measure of wage growth.
Figure 1 makes it clear that the WGT rose sharply in April 2024, which also happens to be the first month that the rule change affected the WGT. Without actual data on capped earnings, we can't comprehensively assess the contribution that the rule change might have made to the April jump shown by the WGT. However, the fact that the alternative WGT based on the sample that neutralizes differences in the topcoding procedures (the orange dashed line in figure 1) has a jump of similar size in April suggests that the change in the topcoding rule is not the main cause of the jump. Since the WGT is a three-month moving average of the median rate of growth in wages, the April 2024 jump also affects the WGT in May and June 2024. But notice that after we remove the effect of the April 2024 jump (in July 2024), the July 2024 WGT reading drops below the March 2024 reading.
We are closely watching the incoming data and working with the US Census Bureau and the Bureau of Labor Statistics to continue evaluating the impact of the change in the topcoding rules on the WGT. We have now resumed the publication of the WGT and will continue to adhere to our past schedule of updates. The WGT is available on the WGT web page and in the Atlanta Fed's EconomyNow app. We feel it's important to remind the WGT users, and Macroblog readers, that the change in the topcoding rules in the data means a change in the underlying sample of earnings used in the WGT calculation. The impact on the WGT is likely to be modest based on our study. Stay tuned as we continue to improve this data tool!