The deposit growth rate and provincial per-capita income do not influence deposit rates. It seems
that only banking variables exert any influence.
Cross-sections (Table 5) for the period 1990-99 confirm that the Herfindahl index does not influence
interest rates for most of the years examined. We also split our sample, considering the five macro-regions
(NorthWest, NorthEast, Center, South, Sicily and Sardinia). In this respect Table 6 shows that the
Herfindahl index negatively influences deposit rates only in the provinces in the NorthWest and the Center,
while the concentration ratio R3 is statistically significant only in the NorthEast and Center. The statistical
significance of concentration indicators is accepted only at the 10 per cent level. These regressions also
seem to confirm a weak effect of concentration on deposit rates.
Many American studies distinguish the analysis of current account rates from the examination of time
or saving deposit rates. This paper also tries to distinguish between the different technical forms of deposits.
In the years 1990-96, concentration negatively affects current account rates (Table 7), but not the returns
on CDs (Table 8) and deposits redeemable at notice (in the latter case we do not report the results). With
reference to certificates of deposit, the result may depend on their fast reaction to variations in money
market yields.
We further investigated the determinants of current account rates, running regressions on cross
sections for the years 1990-96 (Table 9). The regressions show that the Herfindahl index negatively
influences current account deposit rates only in 1990-91;16 in the following years, perhaps as a
consequence of the 1990 liberalization of branching in Italy, concentration does not affect deposit rates
negatively.17 The other control variables maintain their statistical significance.
Coming back to the American debate, it is possible to conclude that the evidence for Italy appears to
be nearer to those studies which criticize the Berger and Hannan papers. First, concentration does not
influence deposit rates in the period 1990-99. Second, the relationship between concentration and rates
seems to vary over time.
In the next section we turn to the analysis of different econometric software for the estimation of
panel data.
16 The concentration ratio R3 is statistically significant only in 1990.
17 On the contrary, the second part of table 9 shows that if the Herfindahl index and the number of banks are
used as regressors, the two variables have both a positive and statistically significant coefficient in the years
1993-96. The result corresponds to the efficiency market view, according to which a higher concentration
may imply a larger deposits’ remuneration, because of the intense competition between the few oligopolists.