Reply: Is My Window Broken?
回应1984年关于货币增长预测的研究，指出基于简约原则的模型在预测货币增长上优于BPT模型，但两者在预测实际产出增长上无显著差异，并讨论时间序列估计的局限性。
by an informal notion of parsimony, reported in 1984. Although our 1984 specifications forecast money growth over this period better than the BPT specifications do, neither set of specifications clearly dominates the other in predicting real output growth. If we have a reservation about Poirier's method, it is that it focuses mainly on quantities estimated from the time series for each country. The ability to draw inferences from such estimates is quite limited, in substantial part because there are as yet so few postwar annual data but also because the underlying parameters of the time series processes may shift over time. Poirier's results bear out that estimates of these parameters are quite noisy in that the data typically do not map his diffuse priors over specification space into sharp posteriors. Our 1984 article sought to raise the signal-tonoise ratio by examining, using as many countries as possible, the cross-regime pattern of time series estimates. Admittedly, in computing these estimates we were guided by an informal notion of parsimony. We would definitely welcome a formalization thereof, especially one derived from priors about the residual vector.