Panel (b) of Figure 2 shows the estimated
recession probabilities averaged across two-regime models for each of the three variables that arguably receive the most attention in both academic and popular discussions of short-run fluctuations—real GDP growth, employment growth, and the change in the unemployment rate.4 The models using either employment or the unemployment rate yield the same two key results as the model using GDP: the two regimes look like recessions and expansions, and the periods the model assigns high recession probabilities to overlap closely with NBER recessions. Indeed, the correspondence to the NBER recession dates from averaging the three sets of estimates is even closer than that using GDP alone. Of course there are differences (an issue we discuss below), but the overall fit is remarkable.