The paper presents the use of the difference GMM, the system GMM and the Panel VAR for the purpose of determining the critical determinants of non-performing loans. The aim of the paper is to point out the factors that explain the volatility of NPLs in a time of crisis. The study focused on a sample of 18 Tunisian banks observed during the period 2008-2018. The paper seeks to identify the impact of crucial macro, microeconomic and governance variables on the NPLs. The results suggest that the deterioration in asset quality can be attributed to both macroeconomic and bank-specific factors. The liquidity risk has a positive and significant correlation with the NPLs of Tunisian banks. The variable "Revolution" presents a positive though not significant relationship with these. Also, the results emphasize the strength of macrofinancial feedback loops in Tunisia. As for the effect of the positive shock of the revolution on the NPL level, we note that it is significant and negative. The decomposition of the sample into two sub-samples: pre-revolution period and post-revolution period allowed showing that the ROA and the ownership structure affect negatively and significantly the NPLs of the banks in the two periods, while the capital affects them positively. It appears that bank-specific factors explain well the volatility of NPLs, especially in the post-revolution period. Finally, by a descriptive study, we have shown that the COVID-19 crisis explains the volatility of the NPLs of Tunisian banks.