Unemployment and Wage Inflation: Recent Findings Using State Data

A 2015 Economic Synopses essay, written by Maximiliano Dvorkin and Hannah Shell, explored the relationship between labor market conditions and wage growth in the aftermath of the Great Recession. Given that the U.S. economy has been characterized by relatively high inflation and low unemployment in recent years, we decided to reexamine the relationship between unemployment and wage growth using both state- and individual-level data.

Background on the Phillips Curve

Before we get into the data, let us discuss the theory behind the empirical relationship between unemployment and wages. This relationship has been extensively studied in economics. The first economist to explore the connection between wage inflation and the unemployment rate was A.W. Phillips.

Unemployment is one way to summarize how much competition there is in the labor market. Phillips greatly influenced the economics field with his much-debated 1958 article, in which he argued that when unemployment is high, nominal wages grow more slowly, and conversely, when unemployment is low, nominal wages grow more quickly. In other words, when the labor market is tight (loose), wages grow faster (slower). This relationship is often referred to as “the Phillips curve.” Although the Phillips curve should be interpreted with caution, because correlation does not imply causation, this is a useful framework for viewing how labor market conditions and wage growth possibly interact.

Phillips Curve at the State Level

The Phillips curve is typically studied at the national level. However, it can also be applied to U.S. states, where there is more heterogeneity in unemployment rates and wage inflation. For example, during the Great Recession (December 2007 to June 2009), the national unemployment rate reached at 9.5%. At the state level, however, the highest unemployment rate during that time ranged from 4% (North Dakota) to 14% (Michigan). Given this variation among states, it makes sense that wage pressures from unemployment would also vary. So, does the Phillips curve hold at the state level?

To investigate the relationship between unemployment and wage growth at the state level, we used wage data from the Quarterly Workforce Indicators from the Census Bureau and state unemployment dataSince this wage data is available only at a quarterly frequency, we average the monthly unemployment rate to a quarterly frequency. from the Bureau of Labor Statistics. We limited our sample to workers in the private sector from 2009 to 2019 to study the relationship before the COVID-19 pandemic. Then, we took averages of the wages and unemployment rate for each state and year.

The figure plots a data point for each state-year combination, with the year-over-year percent change in earnings on the y-axis and the unemployment rate on the x-axis. The red line is the regression line. The slope of the line is -0.24, highlighting the negative relationship between unemployment and the year-over-year percent change in wages. This suggests that the Phillips curve holds at the state level, albeit not perfectly.

Relationship between Unemployment and Earnings Growth at the State Level

SOURCES: Census Bureau, Bureau of Labor Statistics and authors’ calculations.

NOTES: The red line is the regression line from 2009 to 2019. The data are for workers in the private sector.

Phillips Curve at the Individual Level

To further explore the relationship between unemployment and wage growth, we used microdata from Homebase. Homebase is a private payroll company that collects wage, hours and location data of individual workers. Using Homebase data, we can track an individual’s wages across time. We limited our sample to U.S. workers that have been at the same company for at least one year and have an hourly wage of at least $5, which leaves out some tipped employees whose tip income isn’t included in the Homebase data. Our sample’s data are from January 2018 to September 2023. Then, we took a monthly average of each worker’s hourly wage rate. Next, we matched workers with their reported state’s unemployment rate for each month to gauge labor market conditions at the time of employment.

We ran regressions for four different factors linked to wage inflation; these were our dependent variables. The first two regressions capture the probability that a worker experienced a wage increase or a change (a pay raise or a pay cut), respectively. We define a wage change to be at least 10 cents over a month. In our sample, on average, 10.7% of workers saw a wage increase in any given month and 12.5% of workers saw a wage change in either direction.

The last two regressions capture the magnitude of the wage change for all workers and the wage change for those receiving a pay raise. In any given month, the workers in our sample saw an average wage increase of 0.44%; when looking only at those receiving a raise, the average increase was 6%.

For each regression, we used the state unemployment rate and dummy variables for years 2020 through 2023 as the independent variables (i.e., the explanatory variables) to capture the average wage change in each year for the whole economy.

We also included industry-fixed effects to control for persistent differences in wage changes across industries. For example, wage inflation may be greater, on average, in the construction sector than in the food services sector. Thus, including industry-fixed effects in the regression allows us to estimate the relationship between unemployment and wage growth that is not influenced by persistent industry differences.

The table shows the regression results. In all four regressions, the coefficient for the state unemployment rate is negative. This indicates that there is a negative relationship between the state unemployment rate and individual wage growth, suggesting that the Phillips curve holds also when using individual wage change data. In particular, in times when unemployment is high in the state, relative to average, the probability that a worker experiences a wage increase (column 1) or a wage change (column 2) is lower. Moreover, in times when unemployment is high in the state, wages decline mildly (column 3). In addition, the increase in wages for workers receiving a pay raise is lower when unemployment is high (column 4).

Implications

Both the state data and individual-level data suggest that periods with low unemployment rates are also periods with more rapid wage growth. Because wages are an important component of firms’ costs, a booming labor market may signal higher inflationary pressures. Recently, labor markets have shown some indications of cooling, which would be correlated with lower levels of wage increases. In fact, the dummy variables capturing average effects of different years show that 2023 has seen lower levels of wage increases and lower probability of wage changes than those in 2022.

Note

1. Since this wage data is available only at a quarterly frequency, we average the monthly unemployment rate to a quarterly frequency.