Skip to main content
The FRED® Blog

Posts tagged with: "AHETPI"

View this series on FRED

Has wage growth been slower than normal in the current business cycle?

You may have read in the popular press that wage growth seems much slower since the Great Recession compared with previous business cycles. Let’s see what FRED data can tell us. The graph above shows wage growth, defined as the annualized percentage change in the average hourly earnings of private production and nonsupervisory employees. To interpret the graph, note the gray bars, which indicate recessions since 1976, and the green vertical lines, which indicate the peaks of each business cycle. A generally U-shaped pattern occurs between the starts of consecutive recessions. At the start of a recession, the rate of wage growth falls for a number of months, then the trend is reversed as wages increase until the next recession, and the cycle repeats.

To better compare how wages behave across business cycles, we graph the wage behavior observed for each of the prior three business cycles and the current business cycle together. Each cycle is centered at zero, which denotes the month with the lowest wage growth for each business cycle. The current business cycle is identified by the purple line. This cycle started at a lower level of wage increases than the prior three cycles. More importantly, the wage increase from the low point has been following a lower trend: In prior cycles, wage increases exceeded 4%; the current cycle’s wage increases still have yet to reach 3%.

In a future blog post, we’ll look into possible reasons why the current business cycle’s wages have been increasing much more slowly.

How these graphs were created (plus some background): For the first graph, search for wages and select “Average Hourly Earnings of Production and Nonsupervisory Employees: Total Private.” From the “Edit Graph” panel, change the units to “Percent Change from a Year Ago.” The business cycles can be accented by adding green lines to the graph corresponding to each peak using the “Create user-defined line” option under the “Add Line” tab. For the second graph, change the units to “Index” and enter the date “1986-12-01.” This was the lowest point in wage growth for the associated business cycle, which had begun 65 months earlier and would last 43 months longer. To capture the entire business cycle with monthly data, check the “Display integer periods…” box and set the range from -65 to 43. If the units under the “Customize data” tab are changed to “Percent Change from a Year Ago,” the resulting graph shows the section of the first graph from July 1981 to July 1990. While this same result could have been achieved more easily by changing the date range of the original graph, an advantage of this approach is that it allows the same series to be plotted from multiple separate date ranges. Use the “Add Line” tab to add this same series to the graph four times. The options for each line will be the same as those for the first line, except that the custom index date and length of the date range will be different: A second low point occurred in September of 1992, 26 months into a cycle that would last 102 months longer, and the next in January 2004, 35 months into a cycle that would last 46 months longer. The present cycle had its low point 58 months in, during October 2004, and the end date of the cycle has yet to be determined. One way to resolve this problem is to set an unnecessarily high integer end date, like 200. FRED will then automatically fill in the latest available data.

Suggested by Ryan Mather and Don Schlagenhauf.

View on FRED, series used in this post: AHETPI

Wages with benefits

Nominal wages generally increase, but the picture is mixed for real wages. The green line in the top graph shows real wage growth, which is negative a fair amount of the time. Bursts in inflation can counteract the usually small increases in nominal wages. In fact, the strong growth of real wages at the end of the past recession is mostly due to a short episode of deflation.

But wages aren’t the whole story. A job usually also involves other types of compensation, such as the employer’s contribution to retirement pensions, health and life insurance, paid vacation and other leave, and any taxes the employer pays on these benefits. These benefits are now a substantial part of the cost of an employee, and they appear to be growing. The top graph shows that labor compensation growth is frequently higher than real wage growth. We can make this point more clearly by using index values: In the bottom graph, we set both series at 100 in 1970 and let them run. Real compensation growth is significantly higher: the 60% increase looks much better than the 3% increase for real wages.

How these graphs were created: Search for “real compensation” and click on the series shown. In the “Edit Graph” panel, add a new line by searching for “hourly earnings.” Then, within the same panel, add a series by searching for “CPI.” Apply formula a/b to the second line to make earnings real. For the first graph, set units for both lines to “Percent Change from Year Ago”; for the second line, you do this at the bottom of the panel. For the second graph, the selected units are “Index (scale value to 100 for chosen period)”; set the date as 1970-01-01.

Suggested by Christian Zimmermann.

View on FRED, series used in this post: AHETPI, CPIAUCSL, RCPHBS

Testing theory: marginal product and wages

Economic theory tells us that, in a perfectly competitive labor market, labor should be paid according to its “marginal product.” Now, without the jargon: The last workers to be hired by a business should receive pay that is equal to their contribution to the output of that business. So, let’s compare the data with the theory…

Unfortunately, we have no data on the marginal product. But fortunately, we have data on average product. Although it’s not a certainty, these two products should be correlated. So, the graph above shows real growth rates for average product and the average wage. But again, there’s a limitation to the data: We must use the wage of production workers only if we want a series that’s long enough to compare with average product.

Ultimately, it doesn’t look like these series are closely related. The two data limitations we have here could be undermining the relationship. Or the labor market could be less than perfectly competitive. Or the theory could be wrong. It’s difficult to say. But such is the life of an economist… For some more-rigorous research on this topic, take a look at this recent Economic Synopses essay.

How this graph was created: Search for “real output per hour” and select the series shown here. In the “Edit Graph” panel, add the next series by searching for “average hourly earnings” and taking the series with the longer duration. Then modify this series by adding the CPI data series and applying the formula a/b. Select “Percent Change from Year Ago” as the units.

Suggested by Christian Zimmermann.

View on FRED, series used in this post: AHETPI, CPIAUCSL, PRS84006091

Subscribe to the FRED newsletter

Follow us

Back to Top