How much signal did the November MoM of the Average Hourly Earnings (AHE) contain? How does that compare to other indicators? In this (deliberately long) update, we analyze the recent developments in the Average Hourly Earnings (AHE) as well as other measures of wage and total compensation, and we provide our view of what the Fed staff might discuss with Powell and the FOMC.
Keep in mind
Research shows that the November MoM of the AHE is probably overstating the “true” pace of wage growth, which remains nevertheless robust. Evidence shows that wage (and possibly total compensation) growth remains above the level consistent with the Fed inflation target. There are some signs of total compensation disinflation, but overall, we remain skeptical of both wage disinflation and wage re-acceleration. In our view (and estimates), wage disinflation will take time. We are not there, not yet.
The November AHE figure: the CES survey response rate was low
The response rate of the establishment survey dropped significantly in November. Figure 1 shows the establishment survey response rates for first estimate by month in the last 7 years. The main takeaway is that we have not seen a comparable drop in recent years, not even during Covid. While the reason behind this drop is still unclear (the suspect is the timing of Thanksgiving but similar drops have been recorded in other surveys, including JOLTS in recent months), it raises serious concerns about the quality and reliability of the November data. For what we know, the data can be biased (one way or the other). Therefore, we should take the November data with great caution.
Figure 1. Establishment survey response rates for first estimate by month (%).
The November AHE figure: the “denominator effect”
The denominator does matter. AHE is a ratio between (average weekly) earnings and (average weekly) hours. Therefore, in any given month, the MoM of the AHE can increase either because the numerator (earnings) go up or because the denominator (hours) go down. The implications for consumers’ prices are, in principle, different. From an aggregate demand perspective, one should expect AHE to put upper pressure on consumers’ prices if the numerator increases (more earnings) but not necessarily if the denominator falls (in which case workers enjoy more leisure but do not have more spending power).
(Note: however, it remains true that from the point of view of the firms’ marginal cost, the fraction itself (that is, AHE) is more important. Having said so, the Compensation Per Hour (CPH) series is sending different signals than the AHE. We discuss CPH vs AHE below)
In the remaining of this note, we focus on the implications for aggregate spending, but we also discuss the signals from CPH and Unit Labor Cost (ULC).
How large was the “denominator effect” in November?
Table 1 shows the level of the AHE numerator (average weekly earnings), the level of the denominator (average weekly hours), the level of AHE and the MoM of AHE in recent months. In October, the denominator was unchanged with respect to September. Therefore, the AHE MoM in October (0.46%) was driven only by the numerator. In November, AHE grew 0.55% (rounding to 0.6%); the numerator and the denominator both moved: the numerator went up and the denominator went down.
Table 1 shows that if hours had remained constant between October and November (as they did between September and October), the AHE MoM in November would have been 0.26%, that is well below the published 0.55%.
Table 1. Average Hourly Earnings details.
Do hours move a lot and is it an issue?
Figure 2 shows the level of average weekly hours (the AHE denominator) in history. In the pre-Covid world, hours were roughly stable around 34.5, excluding during the GFC. However, since Covid hours have first increased significantly and then travelled back down. This implies that from an aggregate spending perspective, the signal from AHE was downwardly biased in the first phase of Covid and upwardly biased in recent months.
Figure 2. Average Weekly Hours (AWH), SA level.
Another way of looking at this issue is presented in Figure 3. The figure shows the correlation between the AHE MoM growth rate (y-axis) and the AHE denominator (AWH) MoM growth rate (x-axis). Each point in the chart shows the mean in the respective bin. If the changes of the denominator (hours) were irrelevant for the changes of the fraction (the AHE), one should observe no correlation in the chart. Instead, Figure 3 shows a visible negative correlation, especially for months in which hours drop significantly.
Figure 3. Correlation between MoM (% change) of AHE and AWH
The bottom line in our view is straightforward. From an aggregate spending perspective, months in which hours remain constant contain more signal than months in which hours move. For instance, in January 2022 AHE grew 0.57% MoM but in February 2022 AHE grew only 0.13% MoM. The “true” signal was probably in between the January and the February reading as hours dropped 0.57% in January but rose 0.29% in February (for the record, hours were flat in the 3 months before January 2022).
Needless to say, another simpler way of depurating from the denominator effect is to look at the numerator of AHE. In November, Average Weekly Earnings (AWE) increased 0.26% MoM, unsurprisingly close to the figure shown in Table 1 that we have calculated keeping the denominator constant between October and November.
The November AHE figure: where does AHE stand right now?
A simple way to mitigate the “denominator effect” is to look at AHE changes over longer horizons (so that the changes in the denominator are smaller or absent). Figure 4 shows the 3m/3m (a.r.), the 6m/6m (a.r.), the YoY, and the 2Yo2Y (a.r.) of AHE. In this moment, all metrics are remarkably close to 5%. This level is also in line with the signal from the Employment Cost Index (ECI), notoriously a better measure of total compensation. This suggests that, net of all biases, there are little signs of both, AHE disinflation and AHE acceleration. Put it differently, in our view the safest thing to say is that wage growth (which is persistent by nature) remains robust and that it will take time to disinflate it, unless labor markets cool off sharply.
Figure 4. Average Hourly Earnings metrics.
What other measures of wage and total compensation say at the moment?
The Atlanta Fed Wage Tracker (AFWT)
The AFTW calculates the growth rate of the individual worker wage. Therefore, it eliminates the hours and the aggregation biases of the AHE (reminder: the AFWT tracks the median of the individual workers’ wage changes, and not the change of the median workers’ wage). Figure 5 shows the 3m/3m (a.r.) and the 6m/6m (a.r.) implied by the published AFWT (which is published only as YoY), as calculated by Justin Bloesch (a post-Doc at Columbia). Figure 5 compares the AFWT metrics to the AHE metrics.
(For the record, Bloesch’s calculations and underlying assumptions can be found here).
Figure 5. AHE metrics vs (implied) AFWT metrics.
The bottom line from Figure 5 is that in recent months the AFWT has shown signs of plateauing, which is confirmed by the fact that the YoY is going sideways. The 6m/6m of the AFWT is estimated at 5.7% in October, while the 3m/3m is estimated at 3.4%. While Bloesch’s estimates are subject to some uncertainty, the overall impression is that the signals from the AFWT confirm the ones from the AHE.
What other measures of wage and total compensation say at the moment?
The Compensation Per Hour (CPH) and the Unit Labor Cost (ULC)
Two additional measures can shed light on the issues described above. The first one is the so-called “Compensation Per Hour” (CPH), the second is the Unit Labor Cost (ULC). Both indicators are published by the BLS in its “Productivity and Costs” report.
If one is interested in a measure of total compensation (as opposed to wage compensation only) accounting for movements in hours worked, then one should refer to CPH. The numerator of CPH includes all forms of monetary compensation (accrued wages and salaries, supplements, and employer contributions to employee benefit plans) net of taxes. On the other hand, ULC describes the relationship between CPH and labor productivity, or real output per hour, and can be used as an indicator of inflationary pressure on producers. Increases in hourly compensation increase unit labor costs; labor productivity increases offset compensation increases and lower unit labor costs. Put it differently (and with some simplifications), CPH is a measure of pressures on aggregate demand (controlling for hours worked), while ULC is a measure of pressures on firms’ prices.
Figure 6 shows recent developments in CPH and ULC.
Figure 6. Compensation Per Hour (CPH) and Unit Labor Cost (ULC), YoY.
In recent quarters we have observed moderation in CPH. The QoQ (a.r.) of CPH since the beginning of the year have been 2.1% (Q1), 2.3% (Q2), and 3.2% (Q3). These readings are in line with the pre-Covid period and they are all below the YoY shown above. Therefore, there is room for the light blue line in Figure 6 to move down going forward. On the other hand, the YoY of ULC remains well above pre Covid (around 5%), although in Q3 ULC grew 2.4% (a.r.). The reader should notice that 2022 is the only year in recent history in which the YoY of CPH is below ULC, the difference being driven by the recent drop in productivity. In this dimension, should productivity growth normalize going forward (as we expect), the red line in Figure 6 is expected to move back in line with pre-Covid readings.
Important note: in our view, the level of labor productivity has been mismeasured since the beginning of the pandemic but it is now in line with its pre-Covid trend. For this reason, in our view one can now trust the level of labor productivity but *not* its growth rate, which is probably downwardly biased. For this reason, the YoY of ULC in Figure 6 is in our view upwardly biased right now. The “true” ULC is probably already under the light blue line and the “true” pressure on firms’ prices is lower than in published statistics.
The bottom line is that the evidence from CPH is more benign than the AHE, as this measure (which is typically not discussed in the public debate) has already moderated. On the other hand, ULC is sending similar signals than AHE, unless one thinks (as we do) that the recent drop in productivity is driven by measurement issues.
Implications for the Fed staff
As mentioned in previous communication, in our experience the Fed staff relies on the ECI and CPH. The Fed staff uses the signals from AHE only to nowcast the wage component of the ECI. The staff regularly discusses (including with the Chair) measurement issues in the data. Therefore, in our view the Fed staff has discussed the details contained in this note with the DC members of the FOMC. As for the November AHE MoM, in our view the Fed staff message to Powell could be something like “we don’t really trust this 0.6% MoM and we think that the true signal is lower, unless otherwise proven in the next reports”.
As for our view, we continue to think that the signs of wage disinflation are limited. Wages are persistent by nature. Therefore, in our view (and models estimates) the baseline is that wage growth will moderate only with time. At the moment, wage growth continues to be well above the level consistent with the Fed target, with the exception of one measure of total compensation (CPH). To sum up: we are skeptical there are convincing signs of wage disinflation but we are also skeptical there are convincing signs of re-acceleration.