setting their coefficient equal to zero, in spite of their low level of significance shown by the t-statistics.
Therefore, their low level of significance might be due to the contemporaneous presence of other
dummies (which are also meant to “capture” other relevant effects associated to the specific behaviour of
lenders in their approach to risk or investment strategies and the specific risk of some geographic areas).
By looking at the coefficients of tdif(-3) and tdif(-4) one could interpret the dependent variable “ta_d” as
negatively affected by the variable “tdif(-3)” (with a coefficient close to -0.20) and also negatively
affected by the difference “tdif(-3)-tdif(-4)” (with a coefficient of close to -0.75), although these two
variables are not very significant. The variable referred to the size dummies of the bank loans are not
significant at all, although they were significant in the general unrestricted model with a level of
confidence of 76% and 85%. However, since for the sake of our comments all the dummies have been
considered jointly, the loan size dummies appear in the estimates. The results suggest that the persistently
higher interest rate set by the banks on the smaller sized overdrafts might not be simply associated to the
size of the credit contract, but could reflect the impact of other more conventional variables, associated to
the riskiness of the borrower, the competitive context (captured by the variable “tdif”), the size of the
lender, whose dummies turn out to be significant at a level of confidence of 77% and 79%, and, partly, to
the geographic area (specially for what concerns the region Lazio, whose dummy is significant with the
level of confidence of 97%, but also for the areas 3, 5 and 6, whose dummies are significant at the levels
of confidence of 82%, 94% and 80% respectively). Finally, the final parsimonious specification
marginally shows some marginal first-order residual correlation.
3.2 Are largest loans demand-determined?
As explained before, our conjecture is that the share of largest loans over the total amount of loans
granted to the sum of the size classes “c3”, “c4”, “c5” is “demand determined” if it is negatively
correlated with its own interest rate, positively correlated with the demand expectations, and reflect an
accomodative behaviour from the banks. On the basis of the previous considerations, the following
general unrestricted model has been estimated (see the appendix for the details and the variable deletion
tests).
4 4 4 4 4
c5^ijt=CONST+Σηt-kC5^ijt-k+Σφt-ktatt-k+Σλt-ktani-jt+Σθt-kSθ^+Σξt-ktma+
k=1 k=0 k=0 k=0 k=0
+ dummies for banks size +
+ dummies for the geographic area +
+ white noise (2)
As an alternative lending rate “tma” and not “tmi” has been employed: this constitute a
24