1) the data generation process is supplied by using the fitted value of the random effect estimate for
each of the three packages;
2) each package has used its own random number generator;
3) the total panel length has been kept fixed.
Taking into account these experimental conditions, observation of the graphs permits some
interesting considerations:
• TSP and STATA show about the same behavior for the normalized bias, which appears to
approach zero, and the Monte Carlo volatility, which approaches a constant value;
• LIMDEP shows a slightly different behavior: a non zero bias with a variability comparable to the
variability generated by TSP and STATA
6. Conclusions
This paper compares alternative econometric packages by analyzing the determinants of deposit
interest rates in the Italian banking system. The dependent variables are bank interest rates on total
deposits, current accounts, and certificates of deposit. We consider the influence on interest rates of the
Herfindahl index, the concentration ratio R3, the number of banks in each province, the growth rate of
deposits, the ratio between banking costs and total assets, and average staff costs per employee. With this
abundance of panel data, many different specifications have been estimated using the fixed-effects and
random-effects models. Our purpose was to find an answer to the caveats about numerical accuracy
raised by McCullogh and Vinod in the June 1999 issue of the Journal of Economic Literature. They
were very concerned about the little attention paid to numerical accuracy in the selection of econometric
packages.
The first conclusion of the paper is that the Herfindahl index does not influence deposit interest rates
in a panel regression for the years 1990-99; at the same time, the concentration ratio R3 has a negative
effect on deposit remuneration. With regard to current accounts, we found a negative effect of
concentration in the years 1990-96; cross section regressions show that this result derives mainly from the
years 1990-91, when branching was first liberalized. Concentration does not affect interest rates on
certificates of deposit and savings deposits. We also found a positive effect of the number of banks per
province on deposit rates and a negative effect of average staff costs per employee. Banks seem to take
into account their overall structure of costs when deciding the returns to pay on deposits.
17