[Alan] Krueger argues that the patterns of wage changes and unemployment are roughly what one should expect. He focuses only on short-term employment (that is, employment less than 27 weeks), on the grounds that the long-term unemployed are more likely to be detached from the labor force and thus will exert less pressure on wages. Increases in real wages are measured with the Employment Cost Index data collected by the US Bureau of Labor Statistics, and then subtracting inflation as measured by the Personal Consumption Expenditures price index. In the figure below, the solid line shows the relationship between short-term unemployment and changes in real wages for the period from 1976-2008. (The dashed lines show the statistical confidence intervals on either side of this line.) The points labelled in blue are for the years since 2008. From 2009-2011, the points line up almost exactly on the relationship predicted from earlier data. For 2012-2014, the points are below the predicted relationship, although still comfortably within the range of past experience (as shown by the confidence intervals). For the first quarter of 2015, the point is above the historical prediction.
As an aside, note the particular selection of data series. I am not saying that Krueger is wrong for choosing short-term unemployment, the employment cost index, and the PCE deflator. In fact, I think he shows good taste here. But there are other choices available, and I can think of economists who have defiantly done so, cheered on by other prominent economists.
What I wish to point out is that the relationship as depicted is an anomaly with respect to textbook AS-AD, including both Keynesian economics and Sumnernomics. Timothy Taylor refers to the relationship as a Phillips Curve. However, the Phillips Curve relates nominal wages to unemployment, and the chart shows real wages and unemployment. Although in standard macro nominal wages may rise as the unemployment rate falls, real wages are supposed to move in the opposite direction. In standard macro, aggregate supply is derived from movement along the demand curve for labor. When real wages rise by less than productivity increases, demand for labor rises and output goes up. When real wages rise by more than productivity increases, demand for labor falls and output goes down.
Thus, rather than confirming conventional macroeconomic analysis, Krueger’s chart demonstrates an anomaly. In fact, this is hardly a new anomaly. The procyclical behavior of real wages was something that I had observed when I was in graduate school more than 40 years ago.
Of course, you can modify the Keynesian model to accommodate procyclical real wages. Or, you can find data that you believe demonstrate countercyclical real wages (I think that Sumner would try this latter approach). But that is because Keynesian economics is what I call an interpretive framework. How many anomalies you can tolerate before you discard an interpretive framework is a matter of choice. For me, the AS-AD paradigm has too many anomalies to live with.