On reading a recent post by Ed Dolan at Economonitor with some evidence of the lack of a strong Phillips relationship for consumer-price inflation in US data, it occurred to me to try a measure of total compensation per hour with recent data. The wage relationship estimated over all available quarters, using averaged monthly observations for the civilian unemployment rate, is shown above, with a scatter plot and an estimated regression line. Like the relationship estimated by Dolan, the regression line above suffers from a rather loose fit (constant: 6.87; slope coefficient: -.29; R-squared = .02). A complete explanation of inflation is complicated and of course also involves other costs, including raw materials such as fuel. The latter costs are subject of course to “cost-push”-type inflation at times, as are wages. Exchange rates of course affect these costs.
A time series graph below displays both series over the entire sample period, 1948q1 to 2014q1. As some have observed, the exceedingly high unemployment rates of the post-financial-crisis era (blue line) have resulted in very weak or negative compensation growth rates (red line). The latter are not adjusted for inflation in the figures, since we are focusing on nominal data in this post. The downward trend in nominal wage growth in the right side of the figure (red line) helps to explain recent declines in the so-called wage share, which measures the fraction of national income going to labor costs. (However, see this New York Times article for some evidence that falling unemployment is beginning to bring some inflation-adjusted wage growth to parts of the US.)
By the way, if inflation were to become a large problem (and it seems well-contained now), non-recessionary methods exist to try to alleviate it. Even where the Phillips-curve relationship is strong, the human costs of using it to combat inflation are usually very high, given the existence of alternative policies that could perhaps be given a try in the US.