Results

6.1 Diffs-in-diffs and fixed effects

The first column of Table 2 [ PDF 161.4KB | 7 pages ] gives the results of a simple difference in loan balances before and after the interest rate increase in Tikkapara and Kalyanpur. The coefficient is positive, reflecting the upward trend in overall loan balances (as seen in Figure 4 [ PDF 161.4KB | 7 pages ]). The basic differences-in-differences estimator (column 2) uses Geneva to estimate and subtract off this trend; the net impact of the interest rate increase is a 136 taka reduction in loan demand. The impact is small (the implied elasticity is -0.25) but, given the large sample size, statistically significant.

The interest rate elasticity falls (to -0.29) once customer age and their time with the bank are included as controls in the third column. The fourth column increases the flexibility of the specification by allowing for a full set of month-year dummy variables. The ability to better control for underlying trends lowers the interest rate elasticity further, to -0.39.

Few controls for heterogeneity in customer tastes and constraints are available in the data, but the long time series dimension of the panel allows precise estimation of account fixed effects. Adding controls for account fixed effects in column 5 absorbs a significant amount of variation (the R-squared increases from 0.13 to 0.69) and reduces the interest rate elasticity to -0.73. The next two columns show that the elasticities are in a similar range when estimated for Tikkapara (-0.86) and Kalyanpur (-0.70) separately. The eighth column of Table 2 allows the estimated impacts to vary by month; we see that the immediate responsiveness is small (an elasticity of -0.18), but increases substantially over time (to -1.18 a year after the change). The ninth column allows trends in Tikkapara and Kalyanpur to differ from Geneva’s base trend both before and after the interest rate change (we retain the differential intercepts of the standard diffs-indiffs model, in addition to account fixed effects). The treatment effect in this model is both the shift in the intercept and the differential trend associated with being in Tikkapara or Kalyanpur in the months after the interest rate increase. The average elasticity in this specification is -0.72, which is in the same range as the previous results.

The panel used above is not balanced: customers enter the program at different points and some exit before January 2001. One concern is that the changing mix of customers over time affects the results. So, in column 10, we restrict attention to a balanced panel made up of customers who are in the panel for at least six months prior to the February 2000 interest rate increase. These estimates from the balanced panel yield similar results to the base specification in column 5; the estimated interest rate elasticity is -0.79.

Finally, it is worth noting that our elasticity calculations are based on the change in loan balances relative to loans in the initial period. Given the upward trend in loan balances this could overstate the responsiveness of loans to changes in the interest rate. When we recompute our baseline elasticity (-0.73 in column 5) using an arc elasticity (averaged between March 2000 and January 2001), the estimated elasticity decreases only slightly, to -0.68, which does not qualitatively affect our results.

6.2 Heterogeneous effects

In this subsection we examine the heterogeneity of our main result along three dimensions: estimation window, lower wealth versus higher wealth, and borrowing capacity.

6.2.1 Estimation window

We begin by examining how the choice of estimation window affects our results. In our main results, we include the period 12 months before and after the change in interest rates. The concern is that with a longer estimation window trends in the data could be driving the results. Thus in Table 3 [ PDF 161.4KB | 7 pages ], columns 1 and 2, we narrow the window to nine months and to three months before and after the policy change. Dropping the last three months of our sample is useful because of the apparent increase in the growth of loans seen in Tikkapara and Kalyanpur in Figure 4 [ PDF 161.4KB | 7 pages ] at the end of the sample period. For the narrower window, the estimated elasticity is -0.7, similar to our baseline estimate. In column 2, for the narrowest window, the estimated responsiveness is smaller, though still negative and significant at standard levels. The elasticity in this specification is -0.37.

6.2.2 Lower-wealth versus higher-wealth borrowers

In the next three columns, we investigate how wealth affects the impact of interest rates on loan balances. Without a comprehensive measure of wealth, we turn to data on average saving balances. In column 3, the sample is restricted to households who did not save at least 100 taka during one of the months between June 1999 and August 1999,¹⁴ and the estimation window is restricted to October 1999 to January 2001. Column 4 considers households who saved over 100 taka during any month in the June to August period. The estimates show that the “low-saving” group is more responsive to the interest rate than the “high-saving” group, with an elasticity of -0.86 compared to - 0.26. We conclude that the composition of SafeSave’s loan portfolio shifted toward (relatively) wealthier clients compared to the composition before the increase.

We examine this effect directly in column 5 where we estimate the effect of the interest rate change on size of loans taken by poor borrowers. We use a tripledifferences estimator: we are comparing the growth in the amount loaned to the poor relative to the rich in Tikkapara and Kalyanpur, before and after the interest rate increase, subtracting out the same difference from Geneva to control for the time trend. We find that there was a 250 taka decrease in the typical size of loan taken by poor borrowers because of the interest rate change, a decrease of 12 percent relative to the mean. Note that the decrease in amount loaned is relative to Geneva. In absolute terms, the amount loaned to the poor relative to the rich decreased in Tikkapara and Kalyanpur by 624 taka, compared to a decrease of 373 in Geneva over the same period.

We also explored possible differences in responsiveness by gender and age (in specifications not reported here), but do not find substantial differences in estimated elasticities.

6.2.3 Borrowing capacity

Table 2 [ PDF 161.4KB | 7 pages ] presented a range of estimates of the change in loan balances in response to changes in the interest rate. Though these estimates account for time trends, time effects more flexibly, and observed and unobserved individual characteristics, the estimates do not account for variation in borrowing capacity. In particular, individuals with low borrowing capacity are less able to respond to changes in interest rates than individuals with higher capacity (this is most transparent for individuals with zero borrowing capacity). In columns 6 and 7 we address this by taking advantage of our knowledge of the rules used by SafeSave to determine the maximum loan capacity of borrowers (the rules are detailed in the appendix).

Column 6 introduces capacity as a control in our main specification. The estimated interest-rate responsiveness and elasticity are somewhat greater than our baseline result, -0.88 compared to -0.73. The coefficient on capacity is 0.2, which suggests that households increase their borrowing by only 20 percent of an increase in borrowing capacity.

However, measurement error, simultaneity, and omitted variable bias with respect to the capacity measure are serious concerns. Though in principle we measure borrowing capacity precisely, there is always scope for some variation in a branch’s loan decision. There is also a serious concern of simultaneity. A common shock could drive both savings (which is the most important component of capacity) and borrowing. In particular, the presumption is that a negative shock would decrease savings and increase the demand for loans, potentially biasing our results downward. Finally, borrowing capacity is determined mostly by savings, which could affect the demand for loans for reasons other than borrowing capacity.

We address all of these concerns by instrumenting for loan capacity using the length of time the individual has been in the program. As discussed in Section 5, time in program plausibly satisfies the key requirements for a valid instrumental variable. Results are presented in column 7. We note that the F-statistic of the instrument in the first stage is reasonably high (10.75). The estimated effect of interest rates on borrowing increases in absolute value. The implied interest rate elasticity is now -1.04, the lowest elasticity that we find in any specification.

6.3 Mechanisms

Several mechanisms could account for our results: reductions in the probability of taking loans, reductions in the size of those loans, increased speed in repaying loans, or some combination of these. In Table 4 [ PDF 161.4KB | 7 pages ] we examine each of these outcomes using our base specification.

In column 1 we begin by considering a linear probability model where the dependent variable is a dummy variable for whether the borrower takes a loan in a given month. We find a five percentage point increase in the probability of taking a loan. If borrowers are taking more loans, but average loan balances are decreasing, it would suggest either that the size of loans is decreasing or that repayment rates are accelerating. Columns 2, 3, and 4 show that both of these are the case. We find that the amount borrowed decreases by about 17 percent relative to the typical loan size. For the amount repaid, we find an increase of approximately 100 taka, or 60 percent relative to the typical monthly repayment. At the same time, the time between loans fell by about one month, suggesting that the interest rate increase induced borrowers to take more frequent, smaller loans and to repay them more quickly than before.

The fifth column shows that, as expected, withdrawals from savings accounts rise, to compensate for the decrease in borrowing. Deposits also fall, but the coefficient is small and not statistically significant.

The views expressed in this paper are the views of the authors and do not necessarily reflect the views or policies of the Asian Development Bank Institute (ADBI), the Asian Development Bank (ADB), its Board of Directors, or the governments they represent. ADBI does not guarantee the accuracy of the data included in this paper and accepts no responsibility for any consequences of their use. Terminology used may not necessarily be consistent with ADB official terms..

Comment(s)

There are [0] comment(s) for this entry. Post a comment.