influences. As a proxy for the first element, we employ the trading amount of government
bonds (BTRADE). We adopt the ratio of time deposits with unregulated interest rates to the
total time deposits (FREEDEP) as the data of the second element. The variable is
calculated separately for city banks and regional banks.17 As for the third element, we use
a year dummy (SEGDUM), which takes the value of zero until 1993, and unity thereafter.
The degree of competition may be affected by various other factors, which should be
added as explanatory variables in the regression equation. The first candidate for these
factors is the effect of the business cycle. We use the composite index (CIX) to represent
this factor. The real GDP growth rate (GDP) and inflation rates (INFL) are also examined
as alternatives to CIX. The second element is the market structure of the loan market.
The Structure-Conduct-Performance hypothesis predicts that the less concentrated the
market, the more competitive it becomes. The efficiency structure hypothesis proposed by
Demsetz (1973), however, predicts the opposite relationship. We use the Herfindahl index
of city banks and regional banks (HI) as a measure of market concentration.
In view of these arguments, our regression equation is
θk = α + α1BTRADE + α2FREEDEPk + α SEGDUM
k,t 0 1 t 2 k ,t 3 t
(10)
+ α4 CIX (or GDP and INFL ) + α5HIk + α6REGDUM k,
4 t t t 5 k,t 6 k
where k = city or regional banks, and REGDUM is the dummy variable
representing regional banks.
17 For regional banks, only the data for the first regional banks are used, since the second regional
banks were undergoing conversion from mutual banks, and the data was not available during
1989-1991.
18