【正文】 evised form 20 February 2020。 Bank concentration。 Survival analysis The effects of concentration and petitiveness in banking on the industry39。 risksensitive, incentivepatible deposit insurance cannot be implemented in a petitive, deregulated environment (Chan, Greenbaum, amp。 franchise values and further encourage gambling (Hellmann, 18 Murdock, amp。 concentration may be socially preferable to perfect petition (Allen amp。s (1972) results can be indirectly related to the bank concentration–crisis relationship. To the extent bank concentration will motivate excessive risktaking and bring about the possibility of bankruptcy, it will motivate the bank to abandon incentivepatible behavior. Additionally, increased bank concentration may create banks that are toobigtofail, which provide wealth effects for the ―big‖ banks and increase their profitability, while nonincluded banks are adversely affected (O39。 Shaw, 1990). Other studies argue that the relationship between bank petition, excessive risk taking, and bank crisis depends on the type of the deposit insurance scheme. For example, Matutes and Vives (2020) examine the link between bank petition and risktaking incentives, whose effects depend on the extent of petition and the deposit insurance regime (none, flat premium, riskbased premium). Cordella and Yeyati (2020) also study the impact of petition on banks39。 riskiness, a riskbased deposit insurance or, alternatively, the public disclosure of financial 19 information, are likely to mitigate this effect (Cordella amp。 Reinhart, 1999). Considering the relevance of this issue, one would want to improve upon the aforementioned contradictory results. However, until very recently, the appropriate data were not available. In fact, a mon shorting of previous empirical studies is their countryspecific nature. In addition to De Nicol243。231。s estimation technique: (i) The discriminant analysis in the form of a binary logistic regression is often used to identify the variables that influence the oute. Even though this type of analysis is useful in identifying a few variables or binations of variables in a large set of variables, it provides little insight into the way the explanatory variables affect survival, if in fact the binary dependent variable has such an implication. (ii) When the potential censoring times are related to the explanatory variables, discriminant analysis will provide biased results. One of the most important differences between the oute variables modeled through a logistic regression and survival analysis is the fact that we may observe the survival time only partially. In the context of this paper, for a country experiencing a banking crisis, time is the oute variable of interest, the actual survival time. However, for countries that are not experiencing a crisis at the end of the study, time indicates the length of ―followup,‖ which is a 20 partial or inplete observation of survival time. These inplete observations are referred to as being censored. Therefore, using survival analysis, one can consider only two possibilities for a country: in a banking crisis or not known to be in a banking crisis. The pervasiveness of censoring problem seems to be relevant in the bank crisis data. The number of years remaining in the data following the last bank crisis until the end of the sample period has the mean of years and the mode of 8 years, where the minimum and maximum are 2 and 12 years, respectively. Therefore, there is a substantial amount of observations during which countries are not known to be involved in a bank crisis. (iii) The inclusion of the actual failure time as an explanatory variable in a discriminant analysis would be a serious error, as the failure time is part of the response, not part of the factors influencing the response. Therefore, the above points imply that survival analysis places tougher restrictions on the data and on the affect of the covariates on failure (banking crisis) than the logistic regression. (iv) Finally, as it is demonstrated in Section 3, survival analysis provides a superior description of the survivaltime data, produces survival functions for different groups of countries, and offers insights into the dynamics in hazard rates associated with different country groups. Second, this paper places a stronger emphasis on the differences between developed and developing countries than the BDL study. Previous studies (Matutes amp。 Claessens amp。 BDL) emphasize the fact that the plex relationship between concentration, petition, and bank failures has not been adequately investigated. Therefore, this paper addresses the contradictory results regarding the concentration–crisis as well as the regulation– crisis relationship. The data suggest that developing countries have higher bank concentration as well as stricter bank regulations. However, as the overall results of the BDL study and those of this paper suggest, higher bank concentration and stricter bank regulations decrease the hazard of bank failure. This paper takes the analysis a step further and provides an answer to the abovementioned controversial results by examining the determinants of bank concentration. The empirical results of this paper largely confirm those of the BDL study with 21 respect to the effects of bank concentration, regulations, and macroeconomic policies on the hazard of bank failure. Higher concentration in the banking sector, tougher banking restrictions, higher degrees of banking freedom, higher growth rates of real GDP, improving terms of trade, larger depreciation of the home currency, economic freedom, accountability, and, generally speaking, being a developed country decrease the hazard of bank failure. However, a generous deposit insurance scheme, higher real interest rates, inflation rates, dom