Many studies report that government-owned banks under-perform the private banks but no studies examine the reasons of this stylized fact empirically in emerging countries during 1993~2007. This study fulfills this gap. For simplicity, the under-performance of government banks is referred to as the GOB effect. Two matching theories, Nearest-Neighbor Matching (Nearest) and Mahalanobis Metric Matching (Mahala), are adopted to seek matched banks sharing similar characteristic variables located in the same countries. We propose three hypotheses, strong policy role hypothesis, weak policy role hypothesis and no policy role hypothesis, which classifies government-owned banks as having strong, weak and no policy roles, to explain the above phenomenon. Regarding to why GOB effects are commonly observed in emerging countries, this study find that government-owned banks, after being mandated to merge with a distressed or non-distressed bank, suffer adverse performance, supporting strong and weak policy role hypothesis. On the contrary, this study also finds government-owned banks undertaking no policy role perform equally as private-owned banks on average, supporting no policy role hypothesis. Next, by supporting the above three hypotheses, we suggest that political considerations indeed depress government bank performance and the GOB effect in emerging countries are coming from the policy roles influence.