years (e.g., avgrowth_{it }

=

P

t−1 t−3

(growth_{it}). Bank environment growth is a weighted average of

the growth rates in the states in which a bank operates:

bkgrowth_{bit }

=

X

ω_{bi}avgrowth

bit

i∈states

where the weights ω_{bi }

are the percentage of bank branches bank b had in state i in 1980:

N . Data on branch locations are from the Directory of Commercial Bank Offices b i s∈states ω_{bi }= P in India.^{37 }Result N_{bs }s: We test this hypothesis with two pairs of linear regressions. First, we consider

using the location of a bank’s headquarters as an indicator of the growth environment in which a bank operates. Because the regulatory environment in India changes significantly beginning in the 1991-1992, we estimate our equations for the entire time period, 1985-2000, as well as the “post-reform period” of 1992-2000.

[TABLE 10 ABOUT HERE] The results are reported in Table 10. Columns 1 and 2 report the results using the growth environment of the state in which a bank is headquartered, with the first column representing the results for the entire period and the second, results for the post-reform era. Specifically, we estimate

ln(CD_{bit})

= α + β ∗ a v g r o w t h b i t + γ + ( S p r e a d t ∗ a v g r o w t h b i t ) ∗ I S p r e a d t > 0

(8)

+ γ − ( S p r e a d t ∗ a v g r o w t h b i t ) ∗ I S p r e a d t < 0 + θ i + δ t + ε b i t

^{37 }Branch data are from Reserve Bank of India (2001). We have NSDP for all of the states in which bank head- quarters are located. However, in constructing the index, NSDP for the following were not available: Jharkhand, Uttaranchal, Chandigrah, Dadra and Nagar Haveli, Chattisgarh, and Lakhsadeep. Rather than drop any bank that had a branch in one of these states, the ω_{bi }weights are constructed using only the set of branches for which GSP data are available. A second problem is that the growth data are not available for a few states for 1998 and 1999 (Nagaland, Sikkim, Andaman and Nicobar) or 1999 (Goa, Jammu & Kashmir). The two most logical ways of constructing indices in the absence of this data, namely (i) not using those states when constructing state

weights, and thus not using the growth information from 1985-1997, which includes these states, and a second set ω

1985-1997, and (ii) using one set of weights ω

1 bi

from

2 bi

which exclude these states in 1998 and 1999 produce

essentially identical results.

## We choose the latter, because we feel Jammu and Kashmir and Goa warrant inclusion

throughout the 1985-1998 period.

32