Comment on Harding and Pagan 'The econometric analysis of some constructed binary time series'

This comment discusses Harding and Pagan's (2007) article that advocates modeling the NBER business cycle chronology as the outcome of the two-quarter rule. The comment shows that the two-quarter rule does not fare well as a description of the decision-making of the NBER with real-time data available at the time the NBER declared the turning points. In addition, it is not clear how generally one could posit tractable rules-such as the two-quarter rule-for other constructed binary time series, such as stock market booms and busts. As an alternative to modeling the NBER chronology per se, this comment suggests a modified Qual VAR that includes autoregressive dynamics in the latent business cycle index. Out-of-sample forecast results from this model look promising with real-time data without the econometric shortcomings highlighted in Harding and Pagan's critique of the literature to date.