Low Interest Rates and Banks’ Interest Margins: Does Deposit Market Concentration Matter?

We focus on the impact of low interest rates on banks’ interest margins and not on banks’ return on assets (ROA). The effects of low interest rates on bank profitability are less obvious, even if interest margins were to decline. Banks may benefit (at least in the short run) from low interest rates through valuation gains on fixed-income securities they hold. Furthermore, lower interest rates may also spur the economy, increasing the profitability from new lending and the provision of other financial services. Banks may also offset the decline in the NIM by increasing non-interest income, such as from fees and other charges.

Some recent papers focus on the impact of negative interest rates on banks’ interest margins; see, for example, Molyneux et al. (2019) and Lopez et al. (2020). Both studies report that negative interest rates reduced banks’ net interest margins. Nonetheless, Lopez et al. (2020) conclude that bank profitability as a whole has, thus far, been unaffected by negative nominal interest rates, because banks were able to compensate for declining net interest margins.

In his seminal paper, English (2002) studied the link between interest rate risk and bank interest margins in ten industrialized countries reporting that a steep yield curve raises interest margins. Similarly, Alessandri and Nelson (2015) established a positive relationship between the level and the slope of the yield curve and bank profitability in the United Kingdom. There is an extensive literature on other determinants of bank profitability and interest margins (see, for example, Kasman et al. 2010; López-Espinosa et al. 2011; Entrop et al. 2015; Andrievskaya and Semenova 2016; Birchwood et al. 2017).

Other recent studies examine the influence of low and negative rates on bank lending supply and risk-taking (see, for example, Borio and Gambacorta 2017; Heider et al. 2019; Basten and Mariathasan 2020; Boungou 2020; Molyneux et al. 2020).

We also consider national thresholds based on the 25^th percentile of the distribution of short-term local interest rates to identify low interest rate periods. The advantage of using a distribution-based threshold versus a standard threshold for all countries is that it allows accounting for cross country differences in interest rate levels. However, the main reasons why low interest rates may affect banks’ NIMs differently apply if interest rates are close to the effective lower bound and not just historically low. As will be explained in more detail in Section 3, using a uniform threshold implies that some countries are never in a low interest rate environment and this will help identification.

Earlier studies reporting that more concentrated markets exhibit greater price rigidity include Hannan and Berger (1991) and Neumark and Sharpe (1992).

In contrast, Bikker and Vervliet (2018) find that low interest rates compress banks’ net interest margins using data of 3,582 US banks from 2001 to 2015.

There is related research investigating the impact of bank concentration on bank performance. For example, studies have examined how bank concentration affects the connection between monetary policy and bank lending (Olivero et al. 2011; Adams and Amel 2011; Drechsler et al. 2017), bank earnings volatility (de Haan and Poghosyan 2012), bank profitability (Mirzaei et al. 2013) and financial stability (Beck et al. 2006; Jiménez et al. 2013; IJtsma et al. 2017).

We do not include US banks for two reasons. First, they would dominate our sample. Second, the conditioning impact of deposit concentration in the relationship between (low) interest rates and banks’ NIMs has been investigated already by Drechsler et al. (2021). However, we perform a sensitivity analysis including large US banks (see Section 5).

A related paper is by Wang et al. (2020a, b) who show that bank market power interacts with capital regulation to reverse the effect of US monetary policy when the federal funds rate is very low. Specifically, they estimate that, when the federal funds rate is below 0.9%, further cuts in the policy rate can be contractionary.

Note that interest rates used are annual averages, while all bank-level data that relate to flows, are measured at the end of the year.

As suggested by one of the reviewers, we have re-estimated all the results without a lagged dependent variable. Our results overall seemed robust and, in some cases, even stronger, although in a few cases in the robustness analyses results are weaker. Results for the main specification without the lag are shown in Table A.10 in the online Appendix. Results for the other tables are available on request.

Mirzaei et al. (2013: 2925) motivate the inclusion of inflation as follows: “we include the inflation rate and real GDP growth as proxies for business cycle fluctuations. Demirgüç-Kunt et al. (2004) have shown that banks in inflationary environments have wider margins and greater returns. Other studies (e.g. Bourke 1989; Molyneux and Thornton 1992; Demirgüç-Kunt and Huizinga 1999) have also demonstrated a positive relationship between nominal inflation rates and profitability. It is argued that the impact of inflation on profitability depends on whether future inflation is perfectly predicted or not. If bank managers fully anticipate inflation, then they increase lending rates more than deposit rates, maintaining the level of inflation-indexed real profits.”.

Several recent papers using bank panel data to investigate bank performance use time FE (see, for example, Boungou 2020; Elekdag et al. 2020; and Molyneux et al. 2020). Time FE capture structural changes and changes in overall economic conditions. With the financial crisis that started in 2008, changes in global regulation standards, technological changes, and perhaps changes in interest rates expectation worldwide, it seems critical to include these. Although short-term rates across countries are correlated there are also significant differences between the countries in our sample as can be seen in Fig. 1.

Table A.1 in the online Appendix shows the results of the estimation without time fixed effects. As in Claessens et al. (2018), the coefficient of the spread is positive when time fixed effects are excluded.

The correlation is close to zero for the period between 1995 and 2009, and negative at about -0.2 between 2010 and 2018.

Wang et al. (2020a, b) provide a good summary of the advantage of the Lerner Index over other non-structural indicators of banks market power such as the Boon indicator and Panzar-Rosse H-statistic.

Davide Mare kindly sent us the code that was used in Clark et al. (2018) for estimating separately a deposit side and a loan side Lerner index following the Forssbæck and Shehzad (2015) approach.

We start the sample in 2000, the second full year of the eurozone, because some of the controls are lagged one period and include only small countries that were in the eurozone from that period forward. We do not include Greece as that would imply that we could start our sample later. However, including Greece and using fewer years gives similar results (available on request). Apart from Greece, notably Ireland and Portugal were severely impacted by the euro sovereign debt crisis over 2010–12.