A simple enquiry on heterogeneous lending rates and lending behaviour

conservative choice to test for “dmand determined” behaviour of the dependent variable, since “tma”
reflects the incentive of the banks rather than borrowers. By looking at the final “parsimonious”
specification reported below, we see that the own interest rate with three lags “tat(-3)” is largely
significant and has a negative coefficient, as expected. By looking at the lag structure of the variable
“so_u” and at the values of its coefficient one could say that the share of loans to class c5 is roughly
positively correlate with an increase of the bad debts over the last 2-3 quarters. However the second and
third lag are significant at the level of confidence of 88% and 83% respectively, while the 4th lag is not
very significant (although in preliminary a variable deletion test the zero-restriction on its coefficient was
rejected). Nevertheless, by looking at the (very close in absolute value) coefficients of “so_u(-2)” and
“so_u(-3)”, what seems to be significant is rather the “increase” in the bad debts: this again is consistent
with an accomodative behaviour of banks and, therefore, a demand determined share of “c5”. Also the
positive (and significant at the level of confidence of 83%) if the variable “isco” is consistent with a
“demand determined” behaviour. None of the other variables are significant and it is interesting to remark
in this case that no dummy variable turned out to be significant.

TABLE 3

FINAL “PARSIMONIOUS” SPECIFICATION

EQUATION FOR THE RATIO BETWEEN THE LARGEST SIZE CLASS OF LOAN CREDITS
AND THE TOTAL CREDIT GRANTED FOR ALL SIZE CLASSES

The data employed for the estimates cover the period 1990 QI - 1998 QIV

++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++

D.P.D. RESULTS

LEVELS IV

Number of firms:     18       Sample period is 1991 QII to 1998 QI

Observations:       504       Degrees of freedom:       497

Dependent variable is:    c5_t

Instruments used are:

CONST c5_t(-2)  tat(-4) so_u(-3) so_u(-4) so_u(-5) isco(-1)

ONE-STEP ESTIMATES WITH ROBUST TEST STATISTICS

25

<<   22   23   24   25   26   27   28   >>

More intriguing information