Bank insolvency risk and time-varying Z-score measures

We compare the different existing approaches to the construction of time-varying Z-score measures, plus an additional alternative one, using a panel of banks for the G20 group of countries covering the period 1992–2009. We examine which ways of estimating the moments used in these different approaches best fit the data, using a simple root mean squared error criterion. Our results are supportive of our alternative time-varying Z-score measure: it uses mean and standard deviation estimates of the return on assets calculated over full samples combined with current values of the capital-asset ratio, and is thus straightforward to implement.

► Empirically compare different ways to construct time-varying Z-score measures. ► RMSE criterion used to examine which is most consistent with the data. ► Mean/standard deviation of ROA for full sample with current capital-asset ratio preferred.