Whereas microeconomic studies point to pronounced downward rigidity of nominal wages in the US economy, the standard Phillips curve neglects such a feature. Using a stochastic frontier model we find macroeconomic evidence of a strictly nonnegative error in an otherwise standard Phillips curve in post-war data on the US nonfinancial corporate sector. This error depends on growth in the profit ratio, output, and trend productivity, which should all determine the flexibility of wage adjustments. As the error usually surges during an economic downturn, the empirical model suggests that the downward pressure on inflation arising from higher unemployment in a standard Phillips curve framework is significantly cushioned. This might help to understand the robustness of inflation especially in the most recent past. In general, the cyclical dynamics of inflation appear to be more complex than captured by a conventional Phillips curve.