negative profit enter the regression, as might have been expected given the nature of the rules.
In column (3) we include in addition a measure of the utilization by the client of the limit granted to him in the previous year: the ratio of interest earned by the bank to the account limit. This is clearly of direct interest to the bank, since it loses money when funds are committed, but not used. This information is routinely collected on each client. Yet, this variable is uncorrelated with granted limit. We tried other measures of utilization of the limit (turnover on the account divided by granted limit, and maximum debt divided by granted limit), and none of these measures are significant.
In columns (4) and (5) we investigate the determinants of interest rates. Past interest rates seem to be the only significant determinant of today’s interest rates. Past loans, LTB and LWC do not enter the regression.
In sum, the actual policy followed by the bank seems to be characterized by systematic deviation from what the rules permit in the direction of inertia. To the extent that limits do change, what seems to matter is the size of the firm, as measured by its turnover and outlay, and not profitability or the utilization of the limit by the client.
It could be argued that inertia is actually rational and results from the fact that the past loan amount picks up all the information that the loan officer has accumulated about the firm that we do not observe. Prima facie, this explanation does not fit very well with the fact that the loan amount remains exactly the same—the past may be important but, as we already noted the firm’s needs are changing, if only because of inflation.
There is also a simple test of this view. The weight on past loans represents the banks experience with the firm: the fact that the weight is so high presumably reflects the fact that the past is very informative, suggesting a very stable environment. But a very stable environment necessarily implies that the bank knows a lot more about its old clients than it does about its newest clients. Therefore we should see the weight going up sharply with the age of the firm. Yet when we run the regressions predicting the loan amount separately for firms that have been the client of the bank for 5 years or more, and for those who have been clients for less than 5 years, we found that banks do not put less weight on the past loans for recent clients than for old clients. If anything, when we include today’s sales in the regression the bank seems to put