Is It One Break or Ongoing Permanent Shocks

7 The relative importance of permanent (trend) versus cyclical shocks 8 to GDP has been a central issue in macroeconomics since the work 9 of Nelson and Plosser (1982). Morley et al. (2003) find large trend 10 shocks. In contrast, Perron and Wada (2009) argue that allowing for 11 a onetime change in the mean growth rate at 1973:1 nearly eliminates 12 evidence for any trend variance. We re-estimate the Perron and Wada 13 model conditional on a trend break having occurred at any one quar14 ter. We then average the conditional estimates of the trend variance 15 over the probability that the break occurred in a specified quarter. We 16 do this with a Bayesian model average which incorporates break date 17 uncertainty into a trend-cycle decomposition of U.S. real GDP. We 18 find a break date around 2006:1. The median estimate of the size of 19 the drop in the mean growth rate is even larger than the decrease esti20 mated by Perron and Wada. Allowing for a break significantly reduces 21 estimates of trend variance, as Perron and Wada suggest. However, 22 enough spread remains in the posterior distribution of the trend vari23 ance to indicate that available data does not definitively settle the 24 question. Although a model with a fixed break date is preferred over 25 an uncertain break date by the Bayes factor, assuming a fixed break 26 date at 2006:1 makes only a modest difference in the posterior for the 27 trend variance. 28 JEL-Classification: C11, C22, E32 29