23 We examine the sensitivity of our results to the number of observations required and find that requiring 8 or 10 observations does not change the results. 24 Following Nicholas et al. (2009), we include the contemporaneous change in nonperforming assets in the smaller delay model (Eq. (4)). However, we also try including this term in the backward looking model. We continue to find the capital crunch and pro-cyclicality results only in greater delay banks, although the difference between smaller and greater delay banks is reduced.