What motivates mergers in banking? The data show that merger activity is concentrated among very large banks. A large literature on the banking structure has studied this question by estimating cost functions and has provided mixed evidence. A crucial assumption is the exogeneity of input prices.If this assumption fails, result may be biased. This paper adopts the production function method proposed by Levinsohn and Petrin (2003) to separate the impact of productivity from scale economies in banking. To avoid this bias, I use recovery rates of non-performing loans, charge off rates, and cash holdings as proxies for productivity. The proxy method illustrates that the industry operates with significant diseconomies of scale, while the OLS method generates opposite results. Therefore, this finding supports the view that improvements in productivity cause mergers, which is also consistent with data. Finally, I introduce the Quantile Proxy Method to capture the impacts of both input endogeneity and size heterogeneity. This method reveals that medium size banks have largest diseconomies of scale, while top 5% experience somewhat extensive economies of scale. This result sheds light on the fact why many mergers occur among large banks: large parties involved in a consolidation benefit from both productivity improvements and scale economies. / text