Peer behaviors in bank lending decisions using panel-data and convolutional neural network method

Economic growth is strongly depend on financial institutions and many economic activities are in need of lending from banking system, so one of the challenges of managers in banking is making proper lending decisions. this study investigates the effect of peer-banks behavior on lending decisions. We consider both characteristics (size, liquidity, profit, growth, credit risk) and lending behaviors on bank`s loan lending decisions. Data from 19 banks and financial institutions accepted in Tehran Stock Exchange during 2015 to 2020 has been applied with both panel-data method and convolutional neural network(CNN) to find out if there is any convergent behavior. According to results, the bank`s characteristic(size, profitability, liquidity and risk) have a significant impact on loan lending decisions; also the average lending rate of other banks and financial institutions has a negative impact on bank loan lending behavior. So according to panel-data model lending decisions in banks are not convergent, and banks do not imitate their counterparts in making their lending decisions; but average liquidity in the industry and the average credit risk of rivals have a positive and significant impact on loan lending decisions. Examining hypotheses by the convolutional neural network method also showed a divergent relationship between lending decisions of similar banks by considering all the features. In fact, banks do not imitate lending decisions in the same way, but they consider the information of industry and similar banks in their decisions. Therefore, banks should be aware of the psychological impact of competitors' decisions and the characteristics of similar.