We report our examination of the properties of our three measures of delay in expected loss recognition in Table 4. In Panel A we show that the change in nonperforming loans reflects lagged and concurrent macro-economic conditions. Specifically, both the current and next period’s changes in nonperforming loans are positively correlated with current and lagged unemployment and negatively correlated with current and lagged industrial production. These correlations indicate that current economic conditions can be used to predict future and concurrent nonperforming loans. This supports the arguments made by Gambera (2000) that expected future nonperforming loans can be predicted using current macro- economic data. In Panel B we examine the association between our stock and market measures of delay in expected loss recognition and future nonperforming loans. There is generally a positive association between our beginning-of-the-period measures of shorter delay in loss recognition and the current and one period ahead change in nonperforming loans, suggesting that shorter delay banks put more weights on concurrent and fu ture nonperforming loans in determining provisions. This is consistent with these measures differentiating between smaller versus greater delay in expected loss recognition. In Fig. 1, we display the trend in the median adjusted R2 for both less delay (Eq. (4)) and more delay (Eq. (3)) models and the difference between the two models for each quarter. We find that our models explain a reasonable portion of the variation in provisions. For instance, the median-of-the-quarter adjusted R2 for the less delay model ranges from 30% to 70% during the sample period.28 In addition, following this trend, we show the descriptive statistics of the adjusted R2 on the bank-quarter basis. We find that there exists significant variation in the explanatory power for each model and, more importantly, in the difference in their adjusted R2. This latter finding supports our premise that there is significant variation in the application of the incurred loss model