The smallest loan size class (“c1”, with less than 250 millions lira), as discussed in detail in the
appendix, contains several elements of discontinuity in the time series and non-homogeneity in the
definition of the relevant variables. For this reason this class could not be considered in the estimates,
while, on the contrary, it has been included in the descriptive analysis of section 2. In any case, since this
specific size class mainly includes credit to households and very small individual firms (like small shops),
its exclusion does not affects the informational power of the second econometric analysis testing for the
“demand-determined” behaviour of the largest size-class loans. The definition of the remaining loan size
classes (c3, from 250 millions lira to 500 millions, c4, from 500 millions to 1 billion lira, c5, above 1
billion lira) have been “inherited” by our data set from the sampling criteria of the Bank of Italy and partly
reflect (as discussed in the appendix) the constraint imposed by the norms on confidentiality of banking
data.
3.1 The spread between lending rate and policy rate.
Since we cannot exclude simultaneity between some of the regressors and the dependent variables,
the estimates have been performed with the method of instrumental variables. The software employed is
DPD (Arellano-Bond, 1988 and further versions), the instrument employed for each variable (obviously
excluding the dummies) is the lagged variable itself.
TABLE 1
VARIABLES INCLUDED IN THE ESTIMATES
CONST = intercept
I2 = dummy for the average-sized banks.
I3 = dummy for the “small and minors” banks (according to the Bank of Italy classification).
c3 = dummy for the loan size class “c3” (i.e. from 250 millions lira to 500 millions lira).
c4 = dummy for the loan size class “c3” (i.e. from 500 millions lira to 1 billion lira).
c5_t = ratio between the bank overdrafts granted to the category “c5” and the total sum of bank
overdrafts granted to categories “c3”, “c4” and “c5”.
isco = expected demand for the next 3-4 months (ISCO data, based on entrepreneurs interviews). It is a
proxy for the firms’ forward-looking expectations and can assume positive or negative values, since it
expresses expectations on the possible increase or decrease in demand.
so_u = ratio between bad debts and used credit, for each geographic area, for each year, from each size
cathegory of banks to each size cathegory of loan facilities; because of the discontinuity in the criteria of
statistical sampling followed by the Bank of Italy, (briefly discussed in the appendix) the ratio does not
include the laon credit commitment. The variable has been used as a proxy for the “observation-specific”
risk.
ta_d = tat - tan
tan = interest rate on loans by the Bank of Italy, inclusive of any additional charge.
tat = interest rate on loan facilities , for each geographic area, foer each year, from each class size of
banks to each size class of loan facilities.
tdif = tma - tmi
21