Heterogeneity between stressed and non-stressed countries 32

Table 6 below illustrates the results from the baseline regressions adding the dummy variable GIIPS to examine whether the banks in non-stressed countries behaved differently compared to those in stressed (GIIPS) countries.

TABLE 6. Results when controlling for the difference in results for the part of GIIPS countries (using the NIM as the business model variable)

Dependent variable:

∆ DFRt × NIM t-1 0.845^* -0.563 -2.169^*

NOTES: ***, **, * indicate significance at the 1%, 5%, and 10% risk levels, respectively.

Dummy variable GIIPS takes value 1 if the observations are from Greece, Italy, Ireland, Portugal, or Spain; otherwise, its value is 0.

The results in the first column are based on the random-effects model, while the results in the other columns are based on the fixed-effects model. The choices of the model are supported by Hausman Test. Time effects are included in the fixed-effects model but not reported. The constant term of the random model is not reported either.

The regression results in column (1) are not able to identify statistically significant effects for the part of GIIPS countries on the pass-through of the DFR to reserve ratio and liquid assets. Therefore, banks in the sound countries and stressed countries seem to adjust their reserves and liquid assets in a similar way against the DFR change. Nevertheless, it should be noticed that the banks with higher NIM in stressed countries probably raise their liquid assets in the asset portfolio as the interaction of NIM and GIIPS in column (2) is statistically significant at the 10% level.

On the other hand, the new country factor creates a statistically significant heterogeneity in bank lending responses to a change in DFR. In particular, banks in GIIPS countries build up about 2.875% more loans relative to assets than the banks in non-GIIPS when DFR drops by one percentage point.

However, when recognizing that if the banks in GIIPS countries have 1% NIM higher than average, their increases in loans are just 1.17% (=2.875-1.708) higher than the banks’ in non-GIIPS countries. This is evidenced by the statistically significant coefficients of GIIPS in interaction with the previous NIM and DFR change, indicating that the location of the bank also has an effect on the pass-through of DFR conditional on NIM. For the other factors, based on these results there is no major difference from the baseline results reported in section 4.1.2.

A possible explanation for the above findings may lie in the liquidity shortage of banks in stressed countries during the crises and the low-for-long period. As shown in Figure 3, the GIIPS banks already have had lower reserve ratios than the non-GIIPS banks and they actually seem to have kept their reserve ratio at the minimum required level in 2013 and 2014. It implies that the banks in stressed countries have held less funding for lending activities, but the reduction of policy rates over the sample period probably has succeeded in helping the banks in the stressed countries to improve their funding possibilities, and thus, ability to increase lending. But if banks in the GIIPS countries previously have had higher NIM which stands for the interest sensitivity in my model, they may be more cautious to expose further risk in the later period due to their current high interest-rate risk. This may be the reason why they increased their lending less aggressively than the banks with lower NIM. Besides, NIM also measures the bank profitability specifically related to the interest-bearing activities, so banks may prefer preserving their previous profitability by decreasing their risk exposure, especially under the high real economic uncertainty observed during the low-for-long era and NIRP.

This finding partly supports the findings in Altavilla, Canova, et al. (2019) and Buchholz et al. (2020). Although Altavilla, Canova, et al. (2019) agree about the irrelevance of operating location in the bank heterogeneities in response to the conventional monetary policy, they find that the banks located in GIIPS have been more affected by the unconventional policies. Nevertheless, they believe the quantitative easing and credit easing policies contribute to such heterogeneities, while the effect of DFR is quite limited. Buchholz et al. (2020) also find that banks in GIIPS countries behave differently from those in non-GIIPS countries although their conclusions related to the DFR effects on banks’

asset reallocation contradict mine, as discussed in the baseline results (see section 4.1.2).

4.1.4 Effect of negative interest rate policy

In this section, the dummy variable GIIPS in the previous section is replaced by After2014 to investigate the differences in the results before and after the implementation of negative interest rates in the euro area. The results are shown in Table 7.

TABLE 7. Results when controlling for the effects of negative interest rate era (using the NIM as the business model variable)

NIMt-1 -0.065 -0.249 0.204

NOTES: ***, **, * indicate significance at the 1%, 5%, and 10% risk levels, respectively.

Dummy variable After2014 takes value 1 with the observations from 2015 afterward;

otherwise, its value is 0.

The results in the first column are based on the random-effects model, while the results in the other columns are based on the fixed-effects model. The choices of the model are supported by Hausman Test. Time effects are included in the fixed-effects model but not reported. The constant term of the random model is not reported either.

Different from GIIPS, the effect of the negative interest regime is significant and negative on the relationship between DFR and banks’ reserves.

During the negative rate period (after 2014), when DFR drops 100 basis points, banks further increase their reserve ratio by 6.16% in addition to individual

DFR change effect. Furthermore, banks with different NIM level seem to have similar reactions to the NIRP.

For the part of lending activities, the highly positive coefficient of the dummy variable After2014 (28.996) shows a considerable stimulus on lending after 2014. Obviously, some parts of this stimulus may come from the other (unconventional) monetary policy actions implemented simultaneously with the NIRP. Moreover, the introduction of the NIRP seems to change the interaction of DFR and NIM significantly. Banks with higher interest sensitivity (proxied by 1% higher-than-average NIM values) increase their loan ratio by 4.95% more than on average when the DFR decreases (the relationship between DFR and loan change is negative). However, for the period since 2014, the coefficient on the triple effect term points out that the loans are likely to decrease 0.82% (= 5.769 – 4.950) rather than increase as before (the relationship turns to positive). Therefore, the NIRP seems to contract (rather than encourage) bank lending, and the final effect of a reduction in DFR under the NIRP is simply a rise in excess reserves. Indeed, the higher interest-sensitive banks are more affected, and this is complemented to the change in NIM effect on lending during the pre- and post-NIRP discussed in part 4.1.2.

As is well known and discussed also by Heider et al. (2018), the ECB decided to apply the NIRP due to the deteriorating economic activity under which lending opportunities were more limited and riskier than before. Given the high uncertainties in the business environment due to the abnormal negative era and depressed real investment opportunities, banks seem to have been interested to save more liquidity as the buffer for the unexpected events stemming from the aggregate economy. The results in this section seem to be consistent with Heider et al. (2018) that the banks generally want to withhold more liquidity in response to a decline in DFR. However, the result from the estimates of loan changes argues for the effect of other unconventional policies that have created a huge stimulus on the bank lending, but such stimulus is partly contracted by the NIRP if the banks have relatively high interest-rate sensitivity (proxied by NIM). This finding gets partial support from Altavilla, Canova, et al. (2019) who also argue for the superior effects of the quantitative easing and credit easing policies over the NIRP in pushing bank lending.