Baltensperger (1972a, b) proposes the risk-cost hypothesis that banks decid
e the number of loans by considering the costs arising from diversifiable p
ortfolio risk. Thus, the banks do not minimize operation costs, but total c
osts including risk costs. This paper examines empirically whether the risk
-cost hypothesis is valid, using financial panel data from Japanese banks f
rom 1981 to 1994. Estimating the first-order condition of total cost minimi
zation together with an operation cost function, we find that the hypothesi
s is supported. Dividing the sample into different types of banks, it is fo
und that the hypothesis is valid for city and regional banks, but not for s
econd regional banks. (C) 1999 Elsevier Science B.V.