One big question regarding the US economy is whether wage growth is accelerating. You might think that this is a pretty straightforward question to answer, but it’s not. There are many measures of wage growth, and they don’t all point in the same direction. For example, we can compare the year-over-year wage growth in recent years (Chart 1)according to four different measures, three from the US Bureau of Labor Statistics: the Employment Cost Index (wages and salaries), average hourly earnings (Establishment Survey), median weekly earnings (Current Population Survey), and the new Atlanta Fed Wage Growth Tracker. Some of these measures show a significant pickup, but some show no acceleration at all. What should we make of this?
Chart 1. Different measures point in different directions
Sources: Bureau of Labor Statistics, Atlanta Federal Reserve Bank; calculations by The Conference Board
Each wage indicator is somewhat different, but they all measure nominal wages (that is, they are not adjusted for inflation). Each measure is determined by “true wage growth,” but also by composition effects, inaccurate coverage, and other noise in the data. Theoretically, true wage growth reacts to wage growth drivers such as labor market tightness and inflation. Thus, indicators that are more correlated with these wage growth drivers probably measure true wage growth more accurately.
In the following exercise we try to determine which wage measures correlate best with labor market tightness and inflation. We do this by running sequential regressions of the different measures of compensation growth (four-quarter growth rate) on three variables:
- The unemployment gap, which subtracts the natural rate of unemployment from the current rate of unemployment,
- The labor market differential (from The Conference Board Consumer Confidence Survey), which is the percentage of respondents who say jobs are plentiful less those who say jobs are hard to get, and
- The core consumer price index (inflation).
The sample period for this exercise is 1986Q1 to 2015Q1. Table 1 shows the adjusted R-squared measure (a measure of how well the independent variables explain the variation in wage growth—that is, the “goodness” of fit) for each wage measure.
Table 1. Comparing wage measures, 1986Q1 – 2015Q1
|Wage measure (1986Q1-2015Q1)||R-squared|
|1.||Employment Cost Index: Wages and salaries||0.891|
|2.||Raised compensation in the last 3 months? (NFIB)||0.690|
|3.||Average hourly earnings (CES)||0.641|
|4.||Median weekly earnings (CPS)||0.390|
|5.||Business compensation per hour (BLS Productivity)||0.343|
Source: Calculations by The Conference Board
The results: wages and salaries from the Employment Cost Index (ECI) are by far the most correlated with wage growth drivers. Surprisingly, the next best correlated in this group is a measure from the National Federation of Independent Business (NFIB) showing the percentage of companies that have raised compensation in the past three months. The widely watched average hourly earnings from the BLS Establishment Survey is ranked just third, and the bottom two measures—CPS median weekly earnings and compensation per hour in the business sector from the BLS Productivity dataset—both have very low R-squared measures.
In recent months, the Atlanta Fed started producing a Wage Growth Tracker that measures median wage growth over a 12-month period for the same individuals (you can read here about their methodology). This series starts in 1997, so we repeated our earlier exercise starting from the first quarter of 1997, which, in addition to allowing us to include the Atlanta Fed measure, serves as a nice check of robustness for the rest of the results. The results in Table 2 show that changing the sample period did not significantly alter the results. It also shows that the Atlanta Fed Wage Growth Tracker is the wage measure most correlated with wage growth drivers, perhaps because it is the least impacted by changes in composition, as it tracks wage growth for the same individuals over time.
Table 2. Comparing wage measures, 1997Q1 – 2015Q1
|Wage measure (1997Q1-2015Q1)||R-squared|
|1.||Wage Growth Tracker (Atlanta Fed)||0.946|
|2.||Employment Cost Index: Wages and salaries||0.902|
|3.||Raised compensation in the last 3 months? (NFIB)||0.712|
|4.||Average hourly earnings (CES)||0.655|
|5.||Median weekly earnings (CPS)||0.545|
|6.||Business compensation per hour (BLS Productivity)||0.442|
Source: Calculations by The Conference Board
So what do these results tell us about whether or not wage growth is accelerating right now? According to our analysis, the three top measures—the Atlanta Fed Wage Growth Tracker, the ECI, and the NFIB—all point to significant acceleration in wage growth in recent quarters, which suggests that true wage growth has been accelerating. Moreover, since the results suggest a very strong relationship between market forces and the Wage Growth Tracker and the Employment Cost Index, a future trajectory that includes further tightening of the labor market strongly suggests that wage growth is likely to accelerate significantly in the near future.
Further supporting this conclusion is a recent paper by the Dallas Fed. The authors find that as long as the unemployment rate is above its long-run average, tightening of the labor market does not have a big impact on wage growth. However, when the unemployment rate starts to drop below its long-term average, the impact on wage growth becomes significantly stronger.
My own takeaway is to monitor developments especially in the Atlanta Fed Wage Growth Tracker and the Employment Cost Index, for which the second quarter results will be published later this month.
 It turns out that this measure is much better than the unemployment gap in predicting wage growth.
 When you add benefits to wages from the ECI, the correlation deteriorates significantly, which suggests that the growth rate in benefits has very low correlation with market forces.