This paper examines non-performing loans (NPLs) using the Turkish banking sector data by a panel vector autoregression (VAR) approach for the period between 2002Q4 and 2017Q4. The panel VAR analysis is used to test the existence of interdependencies among major bank-level variables and macroeconomic indicators. Given earlier evidence on changing dynamics of NPLs in Turkey after the global crisis, the analysis is repeated by sub-periods covering the pre-crisis and the post-crisis periods as a

more »

... ustness check. Overall, the panel VAR confirms the existence of strong feedback effects among the selected variables. Also, the robustness check supports the presence of feedback effects and also verifies the changing dynamics of NPLs by producing more significant responses compared to the benchmark model. Furthermore, the robustness analysis shows that NPLs are more responsive to macroeconomic conditions in the post-crisis period. Despite the evidence for changing dynamics, capital adequacy, profitability and efficiency still feature out as important bank-specific variables, the impulses of which produce significant and plausible responses in NPLs in all cases. Likewise, NPLs respond reasonably and strongly to shocks in fundamental macroeconomic indicators like inflation, GDP growth, unemployment and debt stock. In this regard, the analysis of impulse-response functions reveals that a positive shock to growth leads to the reduction of NPLs while higher inflation, unemployment and debt stock lead to higher NPLs as expected. Future studies may focus on the underlying structural forces driving the NPLs, which, however, is beyond the scope of this paper.