Negative Interest Rates: Taking Stock of the Experience So Far

Luis Brandao-Marques; Marco Casiraghi; Gaston G Gelos; Gunes Kamber; Roland Meeks

doi:10.5089/9781513570082.087.A001

Jump to Content Jump to Main Navigation

Negative Interest Rates

Taking Stock of the Experience So Far

Author:

Mr. Luis Brandao-Marques

Mr. Luis Brandao-Marques
Search for other papers by Mr. Luis Brandao-Marques in
Current site
Google Scholar
PubMed

Marco Casiraghi

Marco Casiraghi
Search for other papers by Marco Casiraghi in
Current site
Google Scholar
PubMed

Mr. Gaston G Gelos

Mr. Gaston G Gelos
Search for other papers by Mr. Gaston G Gelos in
Current site
Google Scholar
PubMed

Gunes Kamber

Gunes Kamber
Search for other papers by Gunes Kamber in
Current site
Google Scholar
PubMed

, and

Mr. Roland Meeks

Mr. Roland Meeks
Search for other papers by Mr. Roland Meeks in
Current site
Google Scholar
PubMed

https://orcid.org/0000-0002-1321-3182

Publication Date:: 03 Mar 2021

ISBN:: 9781513570082

Language:: English

Keywords:: Negative interest rates; monetary policy; bank lending and profitability; nonbank financial institutions

Download PDF (1.1 MB)

This paper focuses on negative interest rate policies and covers a broad range of its effects, with a detailed discussion of findings in the academic literature and of broader country experiences.

Abstract

This paper focuses on negative interest rate policies and covers a broad range of its effects, with a detailed discussion of findings in the academic literature and of broader country experiences.

Introduction

Starting in 2012 several central banks introduced negative interest rate pol-icies (NIRP) that brought up many important questions. Central banks in Denmark, the euro area, Japan, Sweden, and Switzerland turned to NIRP as their economies faced cyclical headwinds and, in the cases of Denmark and Switzerland, strong upward pressure on their currencies even after short-term policy rates had been pushed to about zero. These moves were accompanied by many questions. To what extent would negative policy rates be transmitted to deposit and lending rates? Might the effects be counterproductive, with financial intermediaries reducing lending? Would banks, companies, and households switch massively to cash holdings? What would be the effects on the yield curve as a whole? Would the introduction of NIRP bring about disruptions in the functioning of money markets? Concerns were also raised about financial stability implications stemming from a potentially significant shift to risky assets by financial intermediaries. Beyond these concerns, NIRP was and remains politically controversial, partly since it is novel, counterintuitive, and often misunderstood. The economic effects of the COVID-19 crisis, and associated prospects of a protracted recovery in an environment where many central banks have exhausted conventional monetary policy space, have brought the issue back to the forefront.

This paper aims to take stock of the experience with NIRP so far. It summarizes the evidence accumulated since the discussion in IMF (2017). Several existing good surveys also cover NIRP as part of broader assessments. These either discuss the effects of unconventional monetary policies (UMP) more generally, including quantitative easing (QE) and forward guidance (Bhattarai and Neely 2016; Dell’Ariccia, Rabanal, and Sandri 2018; and CGFS 2019), cover low interest rates more broadly (CGFS 2018), or zoom in on one aspect of NIRP (for example, Brown [2020] on NIRP and bank lending). However, the specific aspects of NIRP deserve a deeper discussion, and an overall assessment of the effectiveness or desirability of NIRP should be based on its effects on the decisions of households, banks, and firms, in addition to asset prices and the overall economy. Hence, in contrast to other surveys, this paper focuses on NIRP, and covers a broad range of its effects. It comprises both a discussion of detailed findings in the academic literature and of broader country experiences, and points to open issues.

Overall, NIRP has likely supported growth and inflation. The evidence on the macroeconomic effects of NIRP remains sparse, partly because it is challenging to separate the effects of NIRP from those of other concurrent UMP measures. Still, for the euro area, NIRP seems to have had small but positive effects on inflation and growth (Rostagno and others 2019) and boosted corporate investment (Altavilla and others 2019b). In addition, in Japan, NIRP may have supported the economy through the exchange rate channel (Honda and Inoue 2019). Overall, the available evidence so far suggests that effects of NIRP on inflation and output may be comparable to those of conventional interest rate cuts or of other unconventional monetary policy.

The transmission of NIRP has been most visible in money market rates, but long-term yields have also responded. Across jurisdictions, money market rates have tracked policy rates closely as the latter moved below zero (Eisenschmidt and Smets 2019). NIRP also has contributed to the fall in longer-term yields, especially following the initial rounds of cuts pushing policy rates below zero, though some of the decline in long yields probably also reflects the coincident implementation of asset purchase programs (APPs) and enhanced forward guidance. No substantial disruption in money markets have been observed to date.

Deposit rates for non-retail customers and some lending rates have also fallen. In Denmark, the euro area and Sweden, where central banks cut interest rates multiple times into negative territory, these rates slowly dropped following each easing round. The decline in deposit rates has been more pronounced for corporate deposits, in line with the notion that, compared to retail depositors, it is costlier for companies to switch into cash (CGFS 2019). However, banks reacted by raising fees on retail deposits. There also is evidence that these cuts have helped to lower lending rates in the euro area and Switzerland, even if it is difficult to measure the effects because of many confounding factors (for example, the simultaneous announcement of targeted longer-term refinancing operations [TLTROs] in the euro area).

For the most part, bank lending volumes have also responded positively. However, for banking systems with a heavier reliance on deposit funding and larger holdings of very liquid assets, the impact of NIRP is weakened. This implies less stimulus to investment, consumption, and economic activity in general.

In general, bank profits have not significantly deteriorated. This is due to an increase in lending, the introduction of fees on deposit accounts, and the realization of capital gains. However, the evidence is not fully conclusive, since smaller and more specialized banks appear to have been adversely affected and some uncertainty about the effects of NIRP over the long term persists. That being said, the fact that NIRP has now been in place for five or more years without significant adverse effects on bank profitability is reassuring.

So far, there has not been a marked shift toward cash, although euro area banks have increased their cash-in-vault holdings. All in all, there is no evidence of a widespread increase in the use of cash in countries that have implemented NIRPs. Moreover, there is no clear indication that the intro-duction of NIRPs has been followed by an increase in vault cash in most countries. One exception, however, is the euro area, where vault cash held by monetary financial institutions (MFIs) remained stable during initial NIRPs announcement but started to rise significantly from 2016 when the European Central Bank (ECB) lowered a key policy interest rate to –40 basis points.¹

Overall, NIRP appears to have been effective, but it remains unclear how much further negative rates could go before inducing adverse effects. Esti-mates of how low policy rates can be before encouraging a wholesale switch to cash (typically referred to as the effective lower bound, ELB) have been revised downward across countries. In addition, the limit on NIRP after which financial intermediation by banks becomes seriously impaired (often referred to as the reversal rate) is still unknown. However, differences across countries in the structure of their financial systems play an important role in determining the degree to which NIRP can be deepened (The nreyro 2021). For example, the scope for reducing rates further in negative territory will be more limited in banking systems with more retail-oriented deposit funding, given that retail deposit rates, especially those held by households, appear to have a zero lower bound (ZLB). And politically, introducing or deepening NIRP in financial systems with an important presence of financial intermediaries that may stand to lose from nominal interest rates becoming negative (for example, money market funds) is likely to be more difficult.

Although a low-for-long environment creates significant financial stability concerns, NIRP per se does not appear to have compounded the problem. The accommodative monetary policy reaction to a lower equilibrium real rate (r*, which itself is driven by structural factors) tends to induce a search for yield and lower bank profitability. The evidence so far does not indicate that NIRP as such exacerbates these effects.

The evidence so far also underscores the notion that central banks should not rule out NIRP and should keep it as part of their toolkit. This is true even if they are unlikely to use it: when markets internalize that rates can be cut below zero, this is likely to translate into declines in longer horizon yields. This would suggest a greater loosening effect of NIRP for countries that currently have low but positive rates. Ultimately, given the low level of the neutral interest rate, many central banks may be forced to consider NIRP sooner or later, even if there are material adverse side effects.

The paper first discusses the conceptual underpinnings of NIRP and then, based on those, summarizes the existing empirical evidence and discusses specific implementation challenges. In Chapter 1, the paper attempts to derive a set of predictions about the effects of NIRP. Focusing on recent results, Chapter 2 reviews the fast-growing academic literature that attempts to quantify the effects of NIRP. Chapter 3 covers two operational challenges of NIRP: reserve tiering and communication. The concluding chapter identifies the shortcomings in the existing empirical evidence and suggests directions for future work. Details on country experiences are provided in the Annex 1.

Chapter 1 Conceptual Issues

Negative Interest Rates as a Policy Option

Why Seek Negative Policy Rates?

Central banks adopted NIRP against the backdrop of low neutral real interest rates, when the room for conventional policy easing had been exhausted. The neutral real rate of interest—the level of real rates at which demand equals potential output, and therefore there are no inflationary or deflationary pressures—has been declining globally, and in many advanced economies (AEs) is estimated to be close to zero (among others, see Del Negro and others 2019).¹ With inflation targets of about 2 percent, this has resulted in very low nominal interest rates as well. Whereas the median policy easing for AEs prior to the global financial crisis, after which neutral rates declined markedly was about 5 percentage points, in the following decade central banks implemented only small cuts when recessions struck (Figure 1). Achieving sufficient stimulus when neutral rates are low can require policy rates to be set below zero. Therefore, central banks turned to NIRP as part of a range of unconventional policy measures (Dell’Ariccia and others 2018).

NIRP and Other Unconventional Policies

An attractive characteristic of NIRP: it can reinforce the effects of forward guidance. Unconventional monetary policies such as forward guidance and APPs have frequently been deployed in concert with NIRP. One of the reasons is that these measures may often have mutually reinforcing effects, even if this was not always clear at first. The complementarity in the case of forward guidance is particularly direct, as both policies influence beliefs about the path of interest rates. By influencing beliefs about the lower bound of interest rates (Grisse, Krogstrup, and Schumacher 2017), NIRP reinforces the effect of forward guidance announcements that aim to lower long-term yields. By reducing the perceived asymmetry in possible future interest rate paths—through indicating that rates can go lower as well as higher—NIRP can lower the expected path of future rates and so long rates.²

NIRP can also be seen as a substitute for forward guidance when the credibility of the latter is imperfect. Forward guidance has been deployed when private sector expectations for rate hikes were running ahead of policymakers’ own expectations, with announcements inducing a fattening of the yield curve. The efficacy of these announcements hinges on their credibility, which is not observable. NIRP, however, is an observable action and could be more credible and, thus, more effective (Sims and Wu 2020). Furthermore, forward guidance is often viewed as a way to overcome the ZLB and NIRP, by allowing the central bank to track more closely the natural rate of interest, in part removes the need for such forward guidance (Rognlie 2016).

The interaction between NIRP and APPs is likely to be more complex. Just like NIRP, APPs also work through a “signaling channel” (Krishnamurthy and Vissing-Jorgensen 2011).³ Therefore, to the extent that the signaling channel of NIRP is material, negative rates may reinforce the effectiveness of APPs in the same manner that they reinforce that of forward guidance. In addition, since APP fattens the yield curve, NIRP may mitigate APP’s negative effect on bank profits from maturity transformation by removing the ZLB constraint on policy rates, as long as they remain above the reversal rate. However, there are circumstances under which APPs may detract from NIRP. This is because APPs increase the amount of bank reserves and, hence, the burden of NIRP on bank profits. Thus, for a sufficiently large central bank balance sheet, the contractionary bank lending channel may offset the expansionary signaling channel of NIRP (Sims and Wu, 2020).

What Are the Limits to Negative Policy Rates?

The Effective Lower Bound

The existence of cash means that zero is a special number for nominal, but not for real, returns. Real interest rates routinely fall below zero, often because inflation is higher than expected. Consequently, in general, zero has no special status when it comes to real rates of return on assets (Bernanke 2016). But a negative return in nominal terms is special thanks to the existence of physical cash, which is technologically constrained to maintain a constant face value.⁴ This constraint means that the return on cash dominates the return on assets that are close substitutes, such as central bank reserves and bank deposits, and pay negative nominal interest rates. Hicks (1937) used this observation to argue that the existence of cash placed a lower bound of zero on nominal rates, and the same basic assumption has been embedded in much subsequent analysis. However, storage and insurance costs mean that the nominal return to large amounts of cash can be negative.

Negative rates may induce banks to move reserve holdings to cash in vaults. Banks demand central bank reserves because they need it to settle transactions, and in some jurisdictions because of regulation.⁵ An overnight deposit with the central bank is an extremely close substitute for cash: both are liabilities of the central bank; are riskless in nominal terms; and are liquid, in the sense of being immediately available. Indeed, central banks typically guarantee the convertibility of central bank reserves into cash. This very close substitutability means that a policy rate below zero generates an incentive for banks to switch from central bank deposits into cash.⁶ However, the incentive for banks to switch out of reserves as nominal rates fall below zero is tempered by the practical difficulties of a wholesale move into cash. For example, shifting to cash settlement on the scale needed in large economies would be a vast logistical undertaking.

Negative rates would also make cash attractive for some households and firms. However, bank deposits offer obvious advantages over cash: deposits offer the convenience of electronic payments and holding large quantities of cash securely is costly. But for households and firms with smaller liquid asset balances, and less-frequent needs to make larger transactions, deposits may be flightier. Nonlinear effects may be conceivable, with massive cash withdrawals occurring if rates become sufficiently negative (IMF 2007).

Therefore, interest rates cannot be reduced below a technical minimum, known as the ELB, which may be below zero in many economies. The ELB is the interest rate below which there would be a move away from assets that carry nominal interest charges into cash, which is always redeemable at its nominal face value. Its primary determinant is generally taken to be the cost of storing and holding physical cash, which, given a zero nominal return, locates the ELB at or below zero interest rates in AEs,⁷ although it could be higher in emerging market and developing economies (EMDEs).⁸ So far no jurisdiction appears to have set a rate low enough to precipitate a material shift into cash.⁹

Estimates of the ELB vary and reflect different assumptions about storage, transportation, and insurance costs associated with holding large amounts of currency. The costs associated with holding and using cash are likely to vary across countries. For instance, storage costs depend on the largest denomination banknote available in each jurisdiction during the implementation of NIRP. Estimates for Canada, for example, imply an ELB between –25 and –75 basis points, with a midpoint estimate of –50 basis points.¹⁰ Similar estimates hold for the Czech Republic, euro area, and the United States and are somewhat lower for Denmark and Switzerland (Table 1).

^{Table 1.}

Estimates of the Effective Lower Bound

Source: IMF staff.

^{Table 1.}

Estimates of the Effective Lower Bound

Country	Estimate (percent)	Source
Canada	20.25 to 20.75	Witmer and Yang (2015)
Czech Republic	20.2 to 20.6	Kocunová and Havránek (2018)
Denmark	21.5	Rostagno and others (2016)
Euro area	20.7	Rostagno and others (2016)
Sweden	21.6	Rostagno and others (2016)
Switzerland	20.5	Rostagno and others (2016)
United States	20.35	Burke and others (2010)

Source: IMF staff.

^{Table 1.}

Estimates of the Effective Lower Bound

Country	Estimate (percent)	Source
Canada	20.25 to 20.75	Witmer and Yang (2015)
Czech Republic	20.2 to 20.6	Kocunová and Havránek (2018)
Denmark	21.5	Rostagno and others (2016)
Euro area	20.7	Rostagno and others (2016)
Sweden	21.6	Rostagno and others (2016)
Switzerland	20.5	Rostagno and others (2016)
United States	20.35	Burke and others (2010)

Source: IMF staff.

The Effective Lower Bound in Emerging Market and Developing Economies

In EMDEs, policy could be constrained by a positive ELB. There are a few reasons why the ELB in EMDEs may be positive. First, emerging market economies are typically more exposed to higher risks of currency substitution and capital fight, with investors demanding a risk premium on domestic currency assets (that is, in this context, the relevant concept of ELB is not the switch to banknotes but to foreign currencies). Second, financial inclusion is lower, cash usage more common, and households may more readily move away from bank deposits into domestic cash (that is, a higher ELB).¹¹ Therefore, despite low inflation and subdued economic activity, emerging markets’ central banks may be willing to keep a positive interest rate differential with AEs’ interest rates. This, in turn, could prevent emerging markets from adopting NIRP.¹²

Reversal Rates and Other Technical Minima

Even above the ELB, central banks may be unwilling to cut rates below a certain level for fear of adverse effects on financial intermediaries and credit dynamics. The interest rate below which these adverse effects could seriously impair or even reverse the pass-through of policy rates to lending and deposit rates is the “reversal rate.”¹³ The reversal rate may lie above, at, or below the ELB.¹⁴ Unlike the ELB, the reversal rate depends on the composition of financial intermediaries’ balance sheets and income, including their capitalization (Darracq Pariès, Kok, and Rottner 2020). Consequently, policy measures that affect the structure of intermediaries’ balance sheets (for example, micro- and macroprudential regulations) or the marginal returns to lending, such as TLTRO, as well as implementation details, such as tiering, contribute to determining the location of the reversal rate. For example, for jurisdictions wherein a material share of credit is provided by banks that rely heavily on retail deposits (which are thought to be sticky at zero), NIRP may have adverse effects on bank profitability, thus increasing the reversal rate. Such arguments were made for the UK case when interest rates there were cut to 50 basis points in 2009. However, the reversal rate may well be subject to change as the financial positions of key intermediaries change, as illustrated by the Bank of England’s subsequent reduction in rates to 10 basis points in the wake of the 2016 Brexit referendum.

The terms ELB and reversal rate are often used interchangeably, even in central bank communications. As discussed, the ELB and the reversal rate represent two different economic concepts. The confusion between the two hinders an informed debate regarding how low policy rates can go in negative territory. The term ELB is often used to denote a threshold below which the central bank deems it undesirable, rather than infeasible, to cut policy rates beyond, which is instead the reversal rate. However, central banks usually refer only to the ELB, without distinguishing between technical minima, and “reversal” rates.

The Expected Effects of Negative Interest Rates

NIRP is expected to provide substantial monetary accommodation and to boost aggregate demand. NIRP supports economic activity and inflation through the same channels as conventional interest rate cuts. However, NIRP may induce discontinuities in the behavior of households, firms, and financial intermediaries which may yield different effects compared to rate cuts above zero. One example is the increased demand for cash by the private sector, which was discussed above. Other potential discontinuities are discussed next.

Effects on The Yield Curve, Bank Rates, and Exchange Rates

Central banks can affect the entire term structure of interest rates by chang-ing the expectations of future short-term interest rates or by influencing term premiums. Indeed, long-term rates reflect expectations about future short-term rates and the level of risk premiums, which amount to the com-pensation that investors demand for holding a nominal bond with an uncer-tain payoff. Changes in the central bank’s policy rate (a very short-term rate) will affect interest rates at longer maturities to the extent that they are interpreted to be persistent and, thus, affect market expectations about the future path for policy rates.¹⁵ Unconventional monetary policy, through forward guidance and other unconventional measures, can also affect term premiums.

If term premiums are positive, the presence of a lower bound on short-term nominal interest rates places a similar constraint on interest rates all along the term structure. But IMF estimates show that since 1998, across the major bond markets, term premiums, at their lowest point, have fallen to about –1 percent (IMF 2019, Figure 1.2). Therefore, longer-term bond yields would likely remain in positive territory if the long-term nominal neutral rate is above (positive) 2–3 percent. At the same time, with a natural rate below this threshold, long-term nominal rates may fall below zero even without policy rates going negative.

A one-time revision in investor beliefs about negative rates could produce an outsized effect on long-term yields. If an announcement of negative policy rates causes investors’ expectations of future rates to be revised down, longer-term yields will tend to fall too. The extent of the fall naturally depends on how long policy rates were expected to be at their lower bound. But where the removal (or reduction) of the bound is unexpected, and causes expectations of the path of rates to be materially lower over the medium term, the impact is likely to be stronger and likely be felt further along the term structure (Grisse, Krogstrup, and Schumacher 2017; de Groot and Haas 2020).

The degree of pass-through of negative policy interest rates to bank lending and deposit rates is likely to differ from country to country. In theory, pass-through should depend on a host of factors, including the relative importance of retail and wholesale deposits for bank funding, the competitive environment in which banks operate (for example, the degree of contest-ability of the markets for bank deposits and loans), the degree of adjustment costs, the prevalence of fixed-rate relative to variable rate loans, and the elasticities of demand for loans and deposits (Cottarelli and Kourelis 1994, Borio and Fritz 1995, van Leuvensteijn and others 2013). The speed and extent of pass-through will also tend to change over time as those factors vary (Hristov, Hulsewig, and Wollmershäuser 2014).

Generally, exchange rates are expected to respond to interest rate cuts below zero as they do when interest rates are positive. However, it is possible that the sensitivity of the exchange rate to interest rate differentials increases when rates become negative. One reason could be that the adoption of NIRP affects not only the level but also the distribution of the expected policy rates over the medium term.¹⁶ Another reason could be that “preferred habitat” effects (Vayanos and Vila 2009) cause exchange rates to behave differently once central banks adopt NIRP: suppose that only cross-border flows by mutual funds and other institutional investors are sensitive to interest rate differentials when rates are positive. However, when central banks adopt NIRP, it may be the case that a broader class of cross-border flows, including bank flows, also become sensitive to interest rate differentials. If these assumptions were true, they would imply an increased sensitivity of cross-border flows and exchange rates to interest rate differentials once rates become negative.

However, the literature seems to have largely overlooked the role of NIRP in the context of monetary regimes that target exchange rates. By contrast, a growing number of papers study whether central banks that hit the ELB can resort to foreign exchange intervention (FXI) to provide additional monetary stimulus (for example, Svensson 2003, Adrian and others 2020). These find-ings further support the view that FXI and NIRP could work as substitutes.¹⁷

How Might Negative Nominal Rates Affect Households’ Portfolio Choices?

The introduction of negative nominal rates may induce discontinuities in households’ behavior. Households may suffer from some degree of money illusion (confusing real and nominal quantities), or be extremely uncertain over the rate of inflation, which will make them dislike negative nominal rates more than negative real rates.¹⁸ Or they may see an explicit and visible “penalty,” in the form of negative rates, as unfair in a way that a more obscure “fee” or “charge” on their bank deposit or mutual fund is not. Such attitudes may combine with preferences that are not in line with standard expected utility theory. One example of such preferences is given by prospect theory (Kahneman and Tversky 1979) whereby the response to losses is more intense than the response to gains—in this case nominal losses and gains—a situation known as “loss aversion.” Another related example is aspiration-level theory whereby individuals want to reduce the probability of losses (that is, not meeting a level of aspiration)—a behavior called “loss probability aversion” (Diecidue and Van de Ven 2008). Under both theories, the introduction of negative rates could induce a sudden shift in household portfolios (for example, into more risky assets).¹⁹

How Might Negative Nominal Rates Affect Commercial Bank Profits and Lending?

Banks’ net interest margins (NIM) may suffer. If bank reserves pay a negative nominal interest rate, then bank income will decline if they cannot pass on the cost to their own depositors. Banks typically have some market power because retail customers value the safety and convenience of bank deposits, and because switching accounts to take advantage of better rates is seen as troublesome.²⁰ But if retail customers are strongly resistant to negative rates, for the reasons discussed above, banks that wish to maintain deposit funding must accept lower profits.²¹

On the other hand, negative rates may support banks’ net worth by boosting asset values and improving loan quality. The direct impact of NIRP on banks’ NIM may be offset by positive effects on other sources of income. If NIRP has the intended effect of easing economic conditions, the extent of provisioning charges declines along with borrowers’ improved ability to repay their loans. For tradeable assets, a similar revaluation may occur, and is reflected in mark-to-market gains. The equity value of the bank is potentially improved through both of these channels. But this benefit is transitory—capital gains are a one-off, and new loans will be priced to reflect better conditions.

When the negative net income effect outweighs the positive net worth effect, cuts in rates may hurt lending (Figure 2). A rich literature examines how incentive problems of various types can constrain the extent to which bank creditors are willing to fund their activities (see for example, Kashyap and Stein 1995, and Holmström and Tirole 1997). More recently, Brunnermeier and Koby (2019) have developed a simple model in which optimizing banks may respond to rate cuts with higher (rather than lower) loan rates, causing credit volumes to fall rather than to rise. The effect comes about because banks face constraints on their leverage and on their holdings of liquid assets. If negative rates cause bank net worth to decline too much, that leverage constraint becomes binding.²² To increase profits, banks then optimally choose to charge a higher rate than otherwise on the marginal loan. At the same time, a binding liquidity constraint leads them to choose a higher deposit rate than otherwise, reflecting the elevated shadow value of liquid assets.

However, other mechanisms may lead banks to lend more or make riskier loans in response to shrinking profitability and low policy rates. On the one hand, when banks have significant market power (the key ingredient of a “deposits channel of monetary policy”), they may respond to lower intermediation margins caused by a policy rate cut by increasing lending (Drechsler, Savov, and Schnabl 2017, 2018).²³ On the other hand, banks may increase risk-taking and lend to riskier borrowers if NIRP reduces banks’ net worth (see Dell’Ariccia, Laeven, and Marquez 2014).²⁴

Therefore, the net effect of negative rates on bank profits and lending is mostly an empirical question. The impact of NIRP on banks should be greater, the larger the retail deposits in overall liabilities, the greater the overall dependence on net interest income, the larger their share of liquid assets, and the lower the pricing power. On the other hand, banks should benefit from higher asset values and improved loan performance. The balance of these effects is uncertain, with more extended periods below zero tending to raise the likelihood that profitability could suffer.

What Could Be the Effects on Nonbank Financial Intermediaries?

Negative rates could potentially induce large outflows from money market funds (MMFs). Just like for banks, negative nominal interest rates are special for MMFs. This is because, in many jurisdictions, MMFs are in fact a form of narrow banks that issue short-term liquid liabilities and invest in liquid safe assets. Thus, like commercial banks, MMFs face the risk of redemptions when interest rates near the effective cost of holding cash. However, unlike commercial banks, MMFs work with very narrow interest margins because they have a limited ability to tilt the composition of their assets toward riskier, illiquid higher-yield assets.²⁵ A further complication arises for MMFs that offer constant net asset value or CNAV (for example, a constant 1 euro per share) with negative rates: such MMF models are either not sustainable, or forced to recur to share-cancelling mechanisms.²⁶ Finally, NIRP can encourage MMF exits because of the effect it may have on fund manager compensation, especially if management fees are a percentage of fund gross yields and these remain negative for a long time (Dwyer and others 2008).

Importantly, MMFs are likely to evolve and adapt their business model if interest rates become very negative and remain so for long. Faced with the risk of mass withdrawals, MMFs may start searching for yield to preserve their profitability and attractiveness to investors. Within the constraints set by regulation, MMFs will be able to adjust the composition of their portfo-lio and restructure the remuneration of their liabilities instead of exiting the market. This could in principle transform the industry, potentially making these intermediaries less liquid and riskier than they are now. These consid-erations are of primary importance in countries such as the United States, where MMFs represent a cornerstone of the financial system; a significant change in their balance sheet structure toward less-safe assets could threaten financial stability.²⁷ For instance, the functioning of short-term credit and Treasury markets could be affected if MMFs modified their business model.

NIRP could exacerbate the search for yield and associated risk-taking by other nonbank financial institutions. In particular, life insurance companies and other institutional investors may have an incentive to increase risk-taking because of guaranteed positive nominal returns to their policyholders and requirements to hold a certain fraction of their assets as liquid (potentially negative yielding) government securities. It is not clear, however, that this would imply a discontinuity at negative rates in the behavior of these investors, rather than being a problem with low interest rates in general. In addition, especially for liability-driven investors such as life insurance companies, low and negative yields may have the opposite effect and cause them to increase their demand for negative-yielding safe assets (even to the detriment of cash) as they try to hedge duration risk (Domanski, Shin, and Sushko 2017; Shin 2017).

Cross-Border Spillovers

As with other unconventional monetary policies, NIRPs could have cross-border spillover effects. Unconventional monetary policies in AEs support the world economy by boosting asset prices globally and benefit trade via increased economic activity. At the same time, as they create a low return environment for investors, they could also cause excessive capital flows into emerging markets, appreciating their currencies and easing financial conditions. These effects could, in turn, lower the competitiveness of recipient countries and, at the same time, threaten their financial stability by fueling an excessive buildup of leverage.

The nature and channels of international spillovers could be different when interest rates are negative. One potential reason why NIRPs could have a different spillover effect is the possibility that the behavior and the sensi-tivity of capital flows and exchange rates to interest rate differentials could be different than when interest rates are positive, as mentioned previously. Alternatively, the different nature and combination of shocks that lead to the adoption of NIRP could influence the nature of spillovers. Cross-border bank lending may also be impacted by NIRP, if, for example, it leads to a decline in banks’ lending activities, including cross-border bank lending. These effects could depend in part, however, on the currency denomination of international banking flows. For instance, although euro area and Japanese banks play an active role in extending credit to emerging markets, they mainly do so in US dollars (IMF 2019). Negative rates in jurisdictions such as the euro area or Japan could create an incentive to switch from US dollar-denominated foreign debt instruments issued in emerging markets to euro- or yen-denominated ones. This effect could be potentially stronger if NIRPs tended to increase the interest rate differential between US dollar assets and those denominated in other major currencies.

Summary

All in all, several frictions could cause the adoption of NIRP to have different effects from those observed when rates are cut in positive territory. As illustrated earlier, the existence of cash and financial constraints of various nature could induce agents in the economy to react differently when facing negative nominal interest rates. Bounded rationality may also imply that household change their behavior when nominal rates turn negative. In other words, the adoption of NIRP and further rate cuts below zero could have nonlinear effects on a vast set of financial and macroeconomic variables.

The potential adverse side effects of negative rates may intensify over time. It is possible that the erosion of bank profits is gradual because potential losses in net interest income are initially compensated by trading and fee income. In addition, firms and households may be slow in switching out of bank deposits into cash. Thus, the benefit of the policy may be outweighed by its cost as time passes by (see Balloch and Koby 2020).²⁸ Consequently, NIRP would likely be seen as temporary, and therefore less effective.

The empirical question is whether these effects have materialized in the economies where central banks have adopted NIRP. Box 1 reports a summary of key potential implementation problems of NIRP.

^{Box 1.}

Potential Implementation Problems

Interest rate cuts into (or in) negative territory may produce responses in output and inflation that are different to those in conventional (positive) territory. Differences in the transmission of NIRP may occur for the following main reasons.

1. Agents, including households and banks, may avoid paying negative interest rates by increasing their use of physical cash. This may happen at rates not far below zero, limiting the implementation of negative rates.
2. Banks may be unwilling to pass negative rates on to depositors or borrowers.
3. Bank interest margins may shrink, and balance sheets may weaken over time.
4. The response of households and firms to rate cuts may be weaker than with cuts in positive territory, particularly if bank lending is materially affected.
5. There may be a reversal rate at which policy rate cuts become contractionary for bank lending; the reversal rate may lie above, or below, the ELB.
6. A prolonged period of negative interest rates may amplify financial stability concerns (for example, excessive risk-taking by financial intermediaries or exit of MMFs).
7. The effect of the first rate cut below zero on the yield curve may be larger than the effect of subsequent cuts to the extent that the introduction of negative rates caused agents to revise down their beliefs about the policy path.
8. The effects on the exchange rate may be stronger than rate cuts in positive territory.

Chapter 2 Overview of Empirical and Quantitative Evidence

This chapter provides an overview of the quantitative evidence on NIRP in Denmark, the euro area, Japan, Sweden, and Switzerland.¹ It covers several academic studies that try to quantify the effects of NIRP on cash usage, financial variables, the behavior of households, nonfinancial firms, banks, and MMFs, as well as on output and inflation. It also draws on the descriptive analysis of raw data and on technical studies published in policy reports.

The quantification of the effects of NIRP on financial variables and the macroeconomy is difficult for a variety of reasons. First, it is difficult to disentangle the effect of NIRP from other UMP measures. This is because almost all NIRP announcements have been accompanied by other UMP measures. Second, there are not many instances of policy rate changes in or into negative territory, which reduces the power of empirical approaches. Third, identification of causal effects is difficult. Those studies that allow clean identification (for example, by examining the high-frequency response of asset prices or using micro data) only provide indirect evidence. By contrast, studies that try to measure aggregate effects directly face important identification challenges. Fourth, the analysis of NIRP suffers from a selection problem: if countries that expect high costs from NIRP do not implement it and only economies that expect low costs do, empirical studies may underestimate potential side effects. Finally, NIRP has only been adopted in high-income economies with deep financial markets, which limits the scope for exploiting cross-country heterogeneity to identify the role of structural factors in shaping the transmission of NIRP.

Cash Usage

There is no evidence of a widespread increase in the use of cash in countries that have implemented NIRPs. In some countries the use of cash has grown over the last two decades, but there is no indication that these increases coincide with the introduction of NIRPs (Figure 3). In the euro area and Japan, the ratios of banknotes in circulation to nominal GDP have been trending up over the last two decades.² Sweden, by contrast, has seen a significant downward trend driven by a shift to digital transactions.³ In Denmark, the ratio has been stable but started to decline recently, driven by the fall in the highest denomination banknotes. In Switzerland, there had been an increase in the use of cash starting 2008, but the trend has reversed recently. Trends in the usage of cash have not shown significant changes in countries that have not adopted NIRP.

In most countries, the introduction of NIRP has not been followed by an increase in the vault cash (Figure 4). However, in the euro area, vault cash held by MFIs remained stable during the initial NIRP announcement but increased significantly after 2016 when the ECB lowered interest rates to –0.40 percent. However, this was mostly driven by German MFIs (ECB 2018). Vault cash has been increasing in the United States as well, despite the fact that policy rates have remained in positive territory.

Financial Variables

Have the effects of monetary policy on the price of financial assets become different when central banks adopt NIRP? In the presence of a properly functioning transmission mechanism, changes in monetary policy rates are reflected in short-term money market rates and passed from there to the entire yield curve of risk-free rates. The term structure, in turn, works as a benchmark to price risky assets, stocks, and corporate bonds. Despite many confounding factors, the data can provide clues as to whether NIRP involves significant changes in monetary policy spillovers to financial assets. This is discussed next.

Money Market Rates

The impact of NIRP has been most visible in short-term money market rates. Across jurisdictions, they have tracked policy rates closely as the latter moved into negative territory (Figure 5 and Bech and Malkhozov 2016). NIRP has not weakened the pass-through of policy rates to money market rates in Denmark (Jensen and Spange 2015), the euro area (Eisenschmidt and Smets 2018), Japan, Sweden (Angrick and Nemoto 2017), and Switzerland (Grisse and Schumacher 2018).

Yield Curves

Yield curves have shifted downward after NIRP announcements (Figure 6). Government bond yields tend to exhibit an immediate and persistent negative response to the introduction of NIRP (Christensen 2019). Moreover, since the introduction of negative interest rates, a large amount of short- and medium-term government bond yields have turned negative, consistent with investors revising down their beliefs about the ELB. Similarly, Arteta and others (2016), in a cross-country event study analysis, find that both short- and long-term yields dropped significantly after policy announcements, with the maximum effects of NIRP occurring on two-year yields. The case of Japan, however, seems to have been somewhat different, with NIRP leading to a significant fattening of the yield curve, and explains the subsequent adoption of yield curve control (YCC) by the Bank of Japan (Westelius 2020).⁴

Once rates are negative, the impact of interest rate cuts on the yield curve appears to be similar to interest rate cuts in positive territory. The response of the yield curve in the euro area to changes in the policy rates before and after NIRP was qualitatively similar (Arteta and others 2016), especially in the short end of the yield curve (Wu and Xia 2020). Wu and Xia also suggest that successive cuts of already negative rates affect medium-term yields more strongly, possibly because of the use of forward guidance. Alternatively, the smaller effect of initial policy rate cuts on medium- and long-term yields when compared to later cuts could also reflect changing market expectations about the duration of NIRP.⁵

Exchange Rates

The empirical studies regarding the effects of NIRP on exchange rates find mixed evidence. Some argue that the impact of NIRP on exchange rates appears to have been short-lived because other domestic and international developments dominated the evolution of exchange rates (Arteta and others 2016; Hameed and Rose 2018; and Viñals, Gray, and Eckhold 2016). Others conclude that negative interest rates did not have any substantial impact on the behavior of exchange rates (Hameed and Rose 2018). However, Tornton and Vasilakis (2019) found NIRP to have contributed to weaker currencies and reduced exchange rate volatility. Moreover, the data on both the Danish krona and the Swiss franc suggest that NIRP contributed to depreciating the exchange rate, allowing central banks to reduce their reliance on FXI (Figure 7).⁶

NIRP may have increased the sensitivity of exchange rates to interest rate differentials by changing the investor base in currency markets. Some country-specific studies suggest a greater sensitivity of exchanges rates to interest rates under NIRP. In particular, Lane (2019) presents evidence that the sensitivity of the euro/dollar exchange rate to monetary policy expectations has risen since the introduction of NIRP in the euro area (see also references in Eisenschmidt and Smets 2019).⁷ In Denmark, NIRP may have led to higher banking outflows and depreciation pressures, as banks switch to holding more foreign assets to offset the costs of their reserves receiving negative interest rates (Khayat 2018). Moreover, the adoption of NIRP seems to have provided domestic banks in Switzerland with an incentive to raise their foreign currency exposure (Basten and Mariathasan 2018).

Other Assets

NIRP does not seem to have had a significant impact on equity prices in general, but bank equities may have suffered. Stock market indexes kept rising after policy rate cuts in negative territory (Figure 8). However, while reporting an overall positive reaction of stock prices to NIRP in Japan, Hong and Kandrac (2018) find the opposite for Japanese banks’ stocks. For the euro area, while announcements of UMP (including NIRP) on average benefited banks by increasing their stock prices and reducing their credit default swap (CDS) spreads (Altavilla and others 2018), there is some evidence that the effects of policy rate cuts on bank equity prices have turned negative since official rates went to or below zero (Ampudia and Van den Heuvel 2018; Heider, Saidi, and Schepens 2019; Balloch and Koby 2020; Bats, Giuliodori, and Houben 2020).⁸ Importantly, the more banks rely on deposit funding, the more their stock prices suffer from rate cuts in low or negative territory. This suggests that markets perceive the existence of a ZLB on deposits rates as harming bank profitability and provides indirect support for the reversal rate hypothesis.

The evidence on corporate bonds is mixed. In some countries, price indices on investment grade corporate bonds stopped rising or even started falling following the introduction of NIRP. Besides the caveats listed above, it is important to stress how these indices suffer from composition effects. If, within the investment grade category, the relative weight of firms with a low rating increases, the overall quality and thus the price of bonds in the index falls. This is likely to happen because one of the intermediate objectives of NIRP is to incentivize firm access to capital markets for external funding.

Household and Firm Behavior

Empirical studies of the response of household savings and portfolio choices to NIRP are largely absent. The authors are not aware of any empirical study of changes in investor behavior using microdata. At the macro level, Aizenman Cheung, and Ito (2019) find that the effect of low real interest rates on private savings could be negative, depending on economic and structural factors (for example, output volatility, old-age dependency ratio, and financial development) but their study is not specifically about NIRP.

Experimental evidence of the effects of NIRP on household saving and investment decisions is mixed. At least one study in which investors can choose between a risk-free and a risky asset, reductions in the risk-free rate can cause investors to increase the share of risky assets in their portfolios when rates turn negative, but a similar reduction does not occur when rates remain positive (Baars, Cordes, and Mohrschladt 2020). However, Bracha (2020) finds no change in risk-taking behavior in a similar experiment. Fur-thermore, experiments where investors can choose to withdraw (and spend) or keep a bank deposit with a negative nominal return suggest that investors seem tolerant of negative interest rates (Efendic and others 2019, Corneille and others 2020). Tolerance for negative rates seems to decline with the size of the deposit, the time horizon, and the size of negative rates. Tolerance for negative rates seems to be higher for regular savers or when savers know that interest rates will become positive later.

The evidence on how NIRP affects firm behavior is also mixed. On the one hand, in the euro area, firms with relationships with banks that pass through negative interest rates to corporate deposits increase their fixed investment (Altavilla and others 2019b). This effect comes mostly from firms with high cash holdings. Moreover, a survey by Commerzbank (2019) suggests that nonfinancial companies tend to change their (fixed and financial) investment strategies when interest rates turn negative: 37 percent of respondents said that they would switch to other types of assets, and 32 percent responded that they would increase (fixed) investment in their own company. Also, in Denmark, nonfinancial firms more exposed to negative deposit rates increase fixed investment and employment and reduce their leverage and bank deposit holdings (Abildgren and Kuchler 2020).⁹ On the other hand, micro-evidence for Japan suggests that nonfinancial firms curtail investment if they borrow from banks with greater exposure to negative rates (Inoue, Nakashima, and Takahashi 2019). In addition, a 2017 survey by the ifo Institute found that about 30 percent of firms affected by negative deposit interest rates reallocated their financial portfolios to other investments and repaid loans, but only 11 percent increased their fixed investment (Hainz, Marjenko, and Wildgruber 2017).

Bank Behavior

Overall, the evidence suggests that the effects of rate cuts below zero on bank lending and bank lending rates largely resemble those of cuts in positive territory. Although the magnitude of the effects depends on the intensity of NIRP, the responses of bank credit are qualitatively similar across banks. This notion applies in particular to volumes, while the response of lending rates is less clear cut. However, the impact of NIRP on bank funding costs appears different from that of conventional monetary policy: particularly for households, banks tend to raise fees and commissions rather than impose negative rates. However, there is some (but not conclusive) evidence that the impact of NIRP on lending is weakened for banking systems with a heavier reliance on deposit funding and larger holdings of very liquid assets. For such banking systems, NIRP would be less stimulative for investment, consumption, and economic activity in general.

Deposit and Lending Rates

Banks seem to respond to NIRP by increasing fees on retail deposits, while passing on negative rates partly to firms. For retail customers, banks try to overcome the ZLB on deposit rates and reduce their interest expenses by charging higher fees and commissions on retail depositors (Arce and others 2018, Bottero and others 2019 for the euro area; Basten and Mariathasan 2019 for Switzerland). In contrast, for corporate customers, negative rates are transmitted to rates on firm deposits (Altavilla and others 2019b; Deutsche Bundesbank 2020). For European banks, Klein (2020) finds no evidence for a nonlinear relation between policy and deposit rates at negative policy rates, or for a slower pass-through, contrary to what is expected in the presence of a ZLB for retail deposit rates.

The responsiveness of bank deposit rates to successive policy rate cuts after NIRP does not seem to have changed significantly. Deposit rates generally adjust slowly to lower policy rates. A substantial body of evidence shows that this was the case in the euro area prior to the 2007–09 financial crisis (Andries and Billon 2016). This may also be true for rate cuts below zero, as Figure 9 suggests (that is, successive rounds of rate cuts have produced smaller and slower reductions in deposit rates). There is little evidence that the short-term pass-through from policy to deposit rates slowed after the adoption of NIRP in Denmark and Sweden (Madaschi and Pablos-Nuevo 2017). The authors of this departmental paper do not find that the asso-ciation between policy rates and deposit rates has changed after the adop-tion of NIRP in any of the countries surveyed here, except perhaps for Denmark (Box 2).

NIRP seems to have lowered interest rates on new mortgages and corporate loans. Bank lending rates fell in Denmark after the NIRP, even if the immediate pass-through from policy rates may have declined (see Adolfsen and Spange 2020). Still, it is not clear whether the lower pass-through to lending rates in Denmark is due to NIRP or to the aftermath of the 2007–09 financial crisis, and Madaschi and Pablos-Nuevo (2017) do not corroborate such a decline in pass-through following NIRP for Denmark or Sweden. Floating rate mortgage rates fell substantially across the euro area and Swe-den after policy rates turned negative (Figure 10). Similarly, new corporate lending rates in the euro area dropped sharply after the ECB cut the deposit facility rate (DFR) into negative territory. The pass-through to corporate lending rates from the ECB’s first 10 basis point rate cut was more than 30 basis points after three months; however, the cut was accompanied by the announcement of TLTROs, which clouded the picture. A simple analysis does not support the notion that the association between policy rates and average bank lending rates in the countries covered in this survey has changed after the adoption of NIRP (Box 2).

There is substantial heterogeneity across banks in the effects of NIRP on lending rates. Bottero and others (2019) report that NIRP did lower loan rates amongst Italian banks and increased lending—particularly among banks holding larger shares of liquid assets. Similar results have been obtained for Switzerland by Basten and Mariathasan (2018), although they are at odds with Danthine (2018). In Denmark, there is no evidence that banks theoretically more exposed to NIRP (that is, with a higher reliance on deposit funding) responded differently than other banks (Adolfsen and Spange 2020). By contrast, Italian banks with a relatively high reliance on retail deposits tend to increase rates on loans to the nonfinancial private sector (Amzallag and others 2019), while Japanese banks that are more exposed to NIRP did not lower lending rates as much as other banks (Hong and Kandrac 2018).

Bank Profits

On average, bank profits have not significantly deteriorated, thanks to an increase in lending, the introduction of fees on deposit accounts, and the realization of capital gains.¹⁰ For banks in the European Union, Japan, and Switzerland, NIRP only had a small overall effect on profitability because losses in interest income were offset by gains in non‐interest income, such as fees, capital gains, and insurance income (Lopez, Rose, and Spiegel 2020), or because of lower loan-loss provisions (see Urbschat 2019 for evidence on German banks).¹¹ Larger were also likely to have made use of hedging strategies to protect margins (IMF 2020). Other studies find that overall bank profitability in the euro area has been largely unaffected by the introduction of NIRP once the total effects of this policy on asset quality are taken into account (Hong and Kandrac 2018, Altavilla and others 2019a, Stráský and Hwang 2019).

However, the evidence is not conclusive, and it could be capturing only the short-term effects of NIRP, which may be reversed over time. For positive interest rates, Alessandri and Nelson (2015) and English, van den Heuvel, and Zakrajšek (2018) find evidence that rate cuts initially increase bank NIM and profits, but after some time the effect is reversed, consistent with loan pricing frictions. In line with this finding, at least one study shows that NIRP squeezed bank profits through a significant contraction in the NIM, which more than offset capital gains on security holdings (Molyneux, Reghezza, and Xie 2019). Also, Klein (2020) finds that a policy rate cut in negative territory implies a larger drop in NIM for European banks than an equivalent cut above zero. Finally, the expectation of large adverse medium- to long-term effects on bank profitability, potentially ofsetting any temporary increase in profits, could explain the evidence discussed earlier on bank stock prices falling after NIRP.

Still, smaller and more specialized banks appear to have been adversely affected.¹² In particular, banks that are small, are not engaged in cross-border lending, face significant competition, are real estate and mortgage specialists, or operate in countries where floating loan rates predominate, see the biggest declines in profits and NIM after the introduction of NIRP (Molyneux, Reghezza, and Xie 2019).¹³ However, Coleman and Stebunovs (2019) find that NIRP adversely affected the profitability of all euro area banks, regardless of their business models, but this policy seems to have accounted for only a small fraction of the difference in profitability between US and European banks.

Lending Volumes and Asset Quality

According to some studies, banks with more liquid assets and greater access to wholesale funding are able to increase lending more after NIRP. Studies that use different cross-sectional characteristics to measure the exposure to NIRP find a stronger increase in lending by banks with a larger share of liquid assets (Bottero and others 2019) and more excess reserves with the central bank (Basten and Mariathasan 2019). Moreover, banks with a lower share of deposit funding increase their supply of credit more (Heider, Saidi, and Schepens 2019; Lopez, Rose, and Spiegel 2020) or as much as (Bottero and others 2019) other banks.¹⁴ In addition, Inoue, Nakashima, and Takahashi (2019) and Eggertson and others (2019) find that in Japan and Sweden, respectively, a larger share of retail deposits is actually associated with lower lending.^15, ¹⁶ The finding that banks that rely more on wholesale funding increase lending more than those that depend more on deposits is in line with the bank lending channel.

Some other studies, however, find that banks that rely more on deposits increase their lending as much, and often more so, than their peers with lower shares of deposit funding. For example, Tan (2019) and Schelling and Towbin (2020) find that banks increase lending, but the effect is stronger for banks with high deposit ratios and which rely more on retail deposits. One explanation for this finding is that banks try to compensate the decline in interest income by increasing lending volumes (Klein 2020),¹⁷ which would be consistent with Drechsler, Savov, and Schnabl’s (2017) deposits channel of monetary policy.

Banks take on more (but not necessarily excessive) risk following the adop-tion of NIRP. This result holds in particular for loans (Basten and Mariathasan 2019; Bottero and others 2019; and Heider, Saidi, and Schepens 2019), with some evidence pointing to banks terming out loans (IMF 2020), but also for securities (Bubeck, Maddaloni, and Peydró 2020). Furthermore, smaller banks that are more reliant on deposits for funding seem to become riskier (Nucera and others 2017; Heider, Saidi, and Schepens 2019; Schelling and Towbin 2020), as do those banks with lower capital ratios (Inoue, Nakashima, and Takahashi 2019)¹⁸ or with stocks that have experienced larger drops in prices following the adoption of NIRP (Hong and Kandrac 2018). These findings are consistent with Dell’Ariccia, Laeven, and Marquez’s (2014) risk-taking channel of monetary policy. However, the increase in ex ante risk-taking does not translate into higher nonperforming loans (ex post risk) and is probably not excessive. This is consistent with additional lending by banks to financially constrained firms which lack access to credit but are otherwise profitable (Bottero and others 2019), but it can also be consistent with NIRP improving the ex post creditworthiness of borrowers, or simply with nonperforming loans being a lagged indicator of credit quality.

Bank Funding Structure

Countries differ sharply in terms of the importance of retail deposits as a source of bank funding. The country-level heterogeneity in the funding structure of banking sectors matters because the degree of reliance on retail deposits plays a key role in determining the effects of NIRP. Figure 11 shows the evolution of the share in bank deposits held by the nonfinancial private sector in total bank liabilities over the last two decades for the five economies where NIRP was implemented. The figure also reports the dynamics of household and firm deposits separately, because these two types of agents may respond differently to NIRP in terms of portfolio decisions. At the time of the introduction of NIRP, the share of deposits held by the nonfinancial sector ranged from 15 percent in Denmark to nearly 70 percent in Japan. Tus, the effects and transmission of NIRP in Scandinavian and euro area countries are more likely to resemble the impact of conventional cuts than in Japan or Switzerland.

NIRP does not seem to have affected bank reliance on retail deposits. Specifically, the share of household or nonfinancial corporation deposits over total liabilities has not fallen in any of the relevant economies following the adoption of NIRP. In some cases, for instance the euro area, the reliance on this source of funding has even risen since the policy rate moved into negative territory (Eisenschmidt and Smets 2019, Deutsche Bundesbank 2019). This is also found by Lopez, Rose, and Spiegel (2020) for banks in economies that have adopted NIRP.

However, many confounding factors may be at work. Among other factors, the evolution of deposits may reflect the adoption of unconventional monetary policy measures, such as QE. When the central bank purchases assets directly from households or firms, this mechanically causes an increase in bank deposits held by these sectors. However, this descriptive evidence suggests that neither households nor nonfinancial firms have significantly rebalanced their portfolios away from bank deposits.

Money Market Funds

Overall, MMFs have weathered NIRP relatively well. Although MMFs saw an increase in MMF redemptions following the introduction of NIRP, at least for the euro area, assets under management (AUM) by MMF recovered quickly (ECB 2015). And, because the assets held by MMFs were generally safe and liquid, redemption-induced liquidations were not disruptive to funding markets. Moreover, the empirical evidence suggests that the level of policy rates is not the key driver of MMF performance. In fact, the profitability of MMFs mainly depends on the difference between the rate on the central bank deposit facility and the yield on short-term debt securities, which typically represent a large share of MMF holdings (see Bua, Dunne, and Sorbo 2019 for evidence on Irish funds). These results point to the importance of the policy mix, since the gap between the relevant policy rates and short-term government bond yields is influenced by other unconventional monetary policy measures, such as QE.

There were, however, some changes in behavior. As interest rates and yields turned negative in the euro area, several MMF managers waived managements fees and the extended maturities of their investments.¹⁹ Initially, CNAV funds in Europe also used the so-called Reverse Distribution Mechanism—whereby shares are cancelled and then distributed across remaining shares—to maintain a unit share price when interest rates are negative. This practice has now been halted by the European Securities and Markets Authority.²⁰ However, these changes in behavior are not exclusive to NIRP. In fact, in response to policies that kept money market rates close to or at zero for a long period of time, MMFs in the United States have also reached for yield, reduced fees, or exited the market (Di Maggio and Kacperczyk 2017).

However, if policy rates were allowed to go below the ELB, the ensuing fight for cash would be detrimental not just to banks but also to MMFs. The evidence so far on the absence of important redemption pressures or many MMF closures under NIRP may be the outcome of policy rates remaining only mildly negative. The same is true for bank profits’ observed resilience to NIRP being specific to the current environment of moderately negative interest rates. It is possible, therefore, that both industries are significantly disrupted if policy rates become very negative.

Other Nonbank Financial Institutions

There are very few studies of how life insurance companies and pension funds have fared after the introduction of NIRP. In Denmark, they have seen their profit margins fall since 2012 (Danmarks Nationalbank 2018). This could be because NIRP depresses current interest income,²¹ or because legacy policies with significantly positive guaranteed returns have dragged down profitability in a low-return environment.²² However, in Switzerland, where a majority of pension funds pay negative interest on their bank deposits, the negative effect of NIRP on current interest income has been outweighed by the positive effects on capital gains in fixed income, equity, and real estate (Bauer, Bee, and Weisser 2019). And, in Japan, following the introduction of NIRP investment in foreign securities by insurers and pension funds experienced a large increase (Honda and Inoue 2019).

Impact on Inflation and Output

Judging by its effect on long-term yields, there is some indication that NIRP stimulated economic activity and inflation. For example, Rostagno and others (2019) find a large impact of negative interest rates on macroeconomic outcomes in the euro area. Those authors separately identify the effects of NIRP, forward guidance, and asset purchase programs and find that the cumulative impact of negative interest rates on the yield curve has been almost one to one across the maturity structure. This is substantially more than the estimates of the effect of conventional policy rate cuts (that is, in positive territory) on long-term yields for the United States (–42 basis points for a 100 basis point policy rate cut; Hanson and Stein 2015), the euro area (–17 to –45 basis points; Brand, Buncic, and Turunen 2010), and Japan (–40 basis points; Braun and Shioji 2006). Moreover, it compares well with the effect on 5-year yields of QE between 2008 and 2012: –130 basis points for the United States, –90 for the United Kingdom, –25 for Japan, and –13 for the euro area (Hausken and Ncube 2013).²³

Still, direct evidence on the overall effects of NIRP on inflation and output is scarce. On the one hand, Rostagno and others (2019) report that uncon-ventional monetary policies have had a substantial combined effect on output and inflation over the period 2015–18 (about 2.7 percent on GDP and one-third of a percentage point on inflation). Further, their decomposition of the effects of various unconventional policies assigns about a fifth of the overall impact to NIRP. This suggests that a staggered 50 basis point rate cut yields about 0.5 percent increase in GDP and 0.07 percent point increase in inflation. Boucinha and Burlon (2020) corroborate these findings. For Japan, Honda and Inoue (2019) present evidence that NIRP may have supported the economy through the exchange rate channel. On the other hand, Michail (2019) uses a counterfactual estimation technique and finds that the impact of NIRP did not have a significant effect on inflation in in Denmark, Sweden, and Switzerland.

The estimated effect of interest rate cuts on output under NIRP is com-parable to those of conventional policy rate cuts and quantitative easing, but the effects on inflation may have been modest. The estimates of the effect of a 100 basis point policy rate cut based on pre-GFC data (that is, away from the ZLB) suggest an increase in output of 0.5–1 percent in the United States (Ramey 2016) and the United Kingdom (Cloyne and Hurt-gen 2016), 0.3–0.7 percent in the euro area (van Els and others 2003), and 0.7–0.8 percent in Japan (Miyao 2002).²⁴ In addition, the effect of a 100 basis point drop in long-term rates caused by QE is estimated to have been a rise of 1.1–1.4 percent in GDP and of 0.9–1.5 percent in CPI in the United States, a 2.5–3 and 1.5–4.2 percent rise in GDP and CPI, respectively, in the United Kingdom, and a 0.75–1 percent increase in GDP and 0.35–0.6 rise in CPI in select euro area countries (Kapetanios and others 2012; Baumeister and Benati 2013; Altavilla, Giannone, and Lenza 2016; and Weale and Wieladek 2016).²⁵ Overall, these results suggest that all countries that have implemented NIRP have yet to reach the reversal rate, which may be at or below –1 percent (see Darracq Pariès, Kok, and Rottner 2020).

Cross-Border Spillovers

Few studies have examined cross-border spillovers of NIRP, and no com-prehensive, systematic assessment exists. For example, the authors were not able to find any empirical study on the effect of NIRP on cross-border flows to and from non-NIRP economies. Still, there is some evidence that NIRP in Japan had positive spillovers to equity markets in other Asian countries (Fukuda 2018). In addition, the adoption of NIRP by the ECB had significant positive spillovers to financial markets in the Czech Republic, Poland, Romania, and Sweden (Varghese and Zhang 2018) and to bond yields in Asia (Fieldkircher, Huber, and Punzi 2020). However, Arteta and others (2016) find that, although the adoption of NIRP by central banks in AEs had positive spillovers to asset prices in EMDEs, they were in line with those of other expansionary monetary policy announcements.

^{Box 2.}

Pass-Through of Policy Rate Changes to Bank Rates

The authors tested the hypothesis that the association between monetary policy rate changes and changes in bank deposit or lending rates has changed after the adoption of NIRP. For this purpose, they ran a regression of changes in the policy rate on changes in the bank rate as follows:

{Δ i}_{t}^{b} = β_{0} + β_{1} {Δ i}_{t}^{p} + β_{2} {Δ i}_{t - 1}^{p} + β_{3} {Δ i}_{t - 2}^{p} + ϵ_{t},

in which i^b is either a bank deposit or lending rate, i^p is the policy rate (premeasured with a very short-term money market rate, which is the operational target of the central bank), and t denotes the month. The authors included lagged values of the explanatory variable as there may be some inertia in the transmission of changes in policy rates to deposit and lending rates. The authors then tested whether the coefficients on changes in policy rates are stable over the sample using Ploberger and Krämer’s (1992) CUSUM test. The table here shows the dates for which the test rejects the null of no structural break.

Source: IMF staff. Note: “..” means the null of hypothesis of no structural break is never rejected at the 5 percent significance level. The test statistic is that of Ploberger and Krämer’s (1992) CUSUM test based on OLS residuals. Deposit rates for Japan and Switzerland are not available by sector of the borrower. See Annex 2 for data sources.

	Deposit Rates		Lending Rates
	Nonfinancial Firms	Households	Nonfinancial Firms	Households
Denmark	..	Oct/07-Nov/08, May/09-Aug/09, Jul/12-Feb/13	..	..
Euro area		Oct-08	Dec-11	..
Japan	..
Sweden		..	..
Switzerland	..	..	..	..

	Deposit Rates		Lending Rates
	Nonfinancial Firms	Households	Nonfinancial Firms	Households
Denmark	..	Oct/07-Nov/08, May/09-Aug/09, Jul/12-Feb/13	..	..
Euro area		Oct-08	Dec-11	..
Japan	..
Sweden		..	..
Switzerland	..	..	..	..

The results point to the existence of structural breaks in the pass-through of policy rates to bank rates only in Denmark and the euro area. Among these, only the break that goes from July 2012 to February 2013 in Denmark occurred after the adoption of NIRP (July 2012).

All in all, the estimates do not support the notion that NIRP is associated with a structural break in the transmission of policy rates to bank rates. However, the results must be taken as suggestive for two reasons. First, they do not control for potential confounding factors, including the fact that rate cuts, especially after the adoption of NIRP, were often taken in conjunction with UMP measures. Second, they do not imply causality and simply measure association.

Chapter 3 Implementation Issues

Tiering Reserve Regimes

Several central banks have introduced a tiering reserve regime to mitigate the impact of NIRP on bank profitability.¹ This system consists of remunerating a share of reserves at a rate higher than the marginal policy rate. In practice, central banks have implemented tiering in slightly different ways (see Annex 1). The size of the exemption is usually based on banking-system characteristics. For instance, the exemption is sometimes proportional to reserve requirements or reflects the activity on money markets. In terms of remuneration, the interest rate on exempt reserves can be zero (as in the euro area and Switzerland). As of today, Japan is the only case of a 3-tier system, in which central bank reserves are divided in three different categories with three different remunerations (a positive rate on the basic balance, a zero rate for the macro add-on balance, and a negative rate for the policy-rate balance).

When choosing the level of exemption, central banks face a trade-of between shielding bank profitability and ensuring a full transmission of monetary policy. On the one hand, the share of exempt reserves needs to be large enough to significantly lower the average cost of holding them. On the other hand, the exemption cannot be too large—otherwise the relevant money market rates will drift away from the (negative) marginal policy rate toward the rate at which the exempt reserves are remunerated (among others, see Whitesell 2006).²

All in all, the experience with tiering reserve regimes seems positive. Exempting a share of reserves has not hindered the monetary transmission mechanism, while it has provided support to the bank interest rate margin, potentially reducing the reversal rate. Counterfactual evidence for Japan suggests a small but beneficial effect on bank lending (Balloch and Koby 2020). At the same time, the trade-of between preserving the transmission mechanism and sustaining bank profitability may become more difficult to manage as excess reserves keep growing as a result of unconventional mea-sures, such as QE.

Communications

The adoption of NIRP involves significant communication challenges. As for other UMP measures, central banks need to clearly justify the use of negative rates and explain their expected effects and the channel through which they work. In other words, central bank communication should establish a direct and sound link between NIRP, the macroeconomic outlook, and the central bank’s objectives. For UMP in general, and for NIRP in particular, this communication is made more difficult because of the uncertainty surrounding its effects, the potential implications for financial stability, and the public perceptions about its distributional effects.³

Still, at some level, the communication of NIRP may be easier than for other UMP because the transmission channels are better understood. All in all, communication may benefit from the fact that the theory and imple-mentation of NIRP are well established, while suffering from a high degree of uncertainty on its empirical effects.⁴ Indeed, negative rates are the most conventional among UMPs in several respects—in principle rate cuts below zero are not so different than cuts above zero, and its limits—the existence of the ZLB and reversal rates, as discussed earlier—are also well understood. Central banks can monitor the key variables that proxy for the intensity of negative side effects, for instance the use of cash or banks’ reliance on retail deposits. Nevertheless, the levels of the reversal rate and, to a lesser extent, of the ELB are difficult to estimate with precision, also because they depend on characteristics of the banking sector that may vary over time and can be influenced by regulation.

Central bank communication has emphasized different aspects of NIRP to support the decision to adopt or reject it. For example, the ECB has reiter-ated that negative rates should not be viewed as extraordinary, highlighted its positive effects especially when adopted as part of a broad package of UMPs, and shown confidence in its ability to counteract drawbacks should they become quantitively significant. These arguments have typically been backed by analytical work. By contrast, some central banks have been reluctant to disclose the analysis underpinning their assessments of NIRP where this has rested on market-sensitive information, such as confidential supervisory and regulatory data. For instance, the Federal Reserve and the Bank of England refrained from introducing negative rates during the Great Recession on the grounds of potential damage to the proper functioning of financial markets with associated heightened risks to financial stability.

Several lessons can be drawn on how to design communication in prepara-tion and during the implementation of NIRP. First, central banks should clearly explain the expected benefits of adopting NIRP and the likelihood and size of potential side effects of modestly negative rates. Central banks should push back against the notion that there exists a “red line” at zero, which is too risky to cross.⁵ Second, central banks need to stress that they will continuously and closely monitor the pass-through of policy rates to deposit and lending rates, as well as potential side effects and act to alleviate them with adequate instruments (for example, the adoption of a tiering reserve regime). Finally, communication should highlight the difference between nominal and real interest rates, reiterating the prominence of the latter in affecting macroeconomic conditions. To this end, central banks should put additional effort into explaining the rationale underlying negative rates— including the use of “tiered” or “targeted” communications geared toward stakeholders with differing levels of economic literacy.

Chapter 4 Conclusions

Tis paper sheds light on the cross-country experience with NIRP so far. To conclude, the authors briefly summarize what they have learned, what remains to be better understood, and why, given the evidence so far, central banks have been somewhat hesitant in adopting or deepening NIRP.

What Do We Know About the Effects of NIRP?

Overall, most of the theoretical negative side effects associated with NIRP have failed to materialize or have turned out to be less relevant than expected. Economists and policymakers have identified a number of potential drawbacks of NIRP, but none of them have emerged with such an intensity as to tilt the cost-benefit analysis in favor of removing this instrument from the central bank toolbox. For instance, the transmission mechanism of monetary policy does not appear to change significantly when official rates become negative. Moreover, overall, bank profitability has not significantly suffered so far. In other words, the reversal rate remains a theoretical concept which has not been empirically validated and, most likely, not yet breached (Arce and others 2020; The nreyro 2021), and banks do not appear to have engaged in excessive risk-taking. Of course, these side effects may still arise if NIRP remains in place for a long time or policy rates go even more negative, approaching the reversal rate.

What Do We Not Yet Know About the Effects of NIRP?

The literature so far has largely overlooked the impact of negative interest rates on financial intermediaries other than banks. Although pension funds and insurance companies do not typically offer overnight deposits and thus the constraint on lowering the corresponding rates below zero is not an issue, other non-linearities may arise when market rates become negative. Among others, legal or behavioral constraints to offering negative nominal returns could affect the profitability of nonbanks. Given the importance of these institutions for the financial system, the absence of empirical evidence on the impact of negative rates on their behavior is surprising.

The role of bank competition in shaping outcomes is not yet well under-stood. As highlighted in IMF (2017), absent competition from other inter-mediaries or capital markets, the transmission of negative policy rates to bank lending rates will be weaker, as banks would try and preserve their intermediation margin. To the best of our knowledge, no study has tested this hypothesis, despite the availability of relevant data.

Another interesting direction for future research is to further study the determinants of the corporate channel identified by Altavilla and others (2019b). According to this channel, cash-rich firms with relationships with banks that charge negative rates on deposits are more likely to use their liquidity to increase investment. What drives this channel is still unclear. For instance, the role of multiple bank relationships could be investigated. If cash-rich firms can easily move their liquidity across financial institutions (including nonbanks), then negative rates on corporate deposit may simply lead these firms to reallocate their liquidity across intermediaries, without any significant impact on investment. By contrast, frictions that prevent firms from easily establish new bank relationships, and thus move their funds around, could induce a reallocation from corporate deposits to other less liquid assets, such as fixed capital.

Why Have Central Banks Not Resorted to NIRP More Often?

NIRP may not be a first-choice policy option for many central banks because they have other unconventional policy options.¹ The need for monetary policy accommodation when central banks’ main policy rates are close to zero can be met with in a variety of ways. For example, quantitative easing, credit easing, funding-for-lending schemes, and YCC have all been actively used by AE central banks, but not many have ever used NIRP. A reasonable inference is that those central banks judge that the least costly means of meeting their objectives does not involve NIRP.² For example, NIRP may be perceived as costly in jurisdictions that allow CNAV MMFs, at least in the short term.

Specific institutional constraints are likely to have also played a role. In some jurisdictions, central banks may lack the authority to set negative rates (Swoboda 2018). In others, they may be unable to easily enact other forms of unconventional monetary policy. One example is the potential constraints on the ECB’s ability to undertake purchases of government securities arising from the prohibitions on monetary finance written into European treaties. Such constraints may make the commitment to undertake unlimited purchases that underpin YCC harder to support, although recent announcements make clear that the ECB can conduct very sizeable operations if needed (ECB 2020).

The structure of funding markets may also discourage some central banks from pursuing NIRP. This may be the case, for instance, in economies where banks rely extensively on retail deposits. It could also apply to economies where MMFs intermediate a significant fraction of savings, especially if most of those funds have constant or stable NAV. However, the experiences and evidence surveyed in this paper should dampen these concerns, at least for moderately negative rates.

However, central banks should not rule out NIRP and keep as part of its toolkit, even if they are unlikely to use it. Allowing for the possibility of temporary cuts into negative territory by not ruling out NIRP could still shift expectations and be effective in lowering long-term rates. If markets internalize the concept that rates can be cut below zero or below what was previously seen as their effective minimum, the shift in market expectations is likely to produce declines in longer horizon yields—even if the move is initially seen as temporary. This would suggest a greater loosening effect of NIRP for countries that currently have low but positive rates.

Finally, there may still exist some room for central banks to cut rates further. The absence of a fight to cash at moderately negative interest rates could mean that the ELB is deeply negative, probably below the estimates of about –0.5 to –0.75 percent from many older studies (Table 1) and perhaps as low as –2 percent (Lilley and Rogoff 2019).³ Deeply negative rates could be effective as a temporary tool in extreme situations, and may be required given that AE central banks, on average, cut policy rates by 500 basis points in typical recessions. Still, there is considerable uncertainty about how negative rates can become, before significant adverse effects on bank lending become apparent. And prolonged periods of negative rates could elicit a strong pushback from households and certain segments of the financial sector (for example, money market funds).

Save
Cite
Email this content

Share Link

Copy this link, or click below to email it to a friend
Email this content
or copy the link directly:

https://www.elibrary.imf.org/view/journals/087/2021/003/article-A001-en.xml
The link was not copied. Your current browser may not support copying via this button.

Link copied successfully

Collapse
Expand

Negative Interest Rates: Taking Stock of the Experience So Far

Author:

Mr. Luis Brandao-Marques

Marco Casiraghi

Mr. Gaston G Gelos

Gunes Kamber

, and

Mr. Roland Meeks

Volume/Issue:: Volume 2021: Issue 003

Publisher:: International Monetary Fund

ISBN:: 9781513570082

ISSN:: 2616-5333

Pages:: 85

DOI:: https://doi.org/10.5089/9781513570082.087

Keywords
Introduction
Chapter 1 Conceptual Issues
Chapter 2 Overview of Empirical and Quantitative Evidence
Chapter 3 Implementation Issues
- Tiering Reserve Regimes
- Communications
Chapter 4 Conclusions

View raw image

^{Figure 1.}

Median Policy Easing during AE Recessions, 1990–2019

(Percentage points)

View raw image

^{Figure 2.}

Potential Effects on Bank Net Worth

View raw image

^{Figure 3.}

Cash-to-Consumption Ratio

View raw image

^{Figure 4.}

Vault Cash Held by Monetary and Financial Institutions (MFIs)

View raw image

^{Figure 5.}

Money Market Rates

View raw image

^{Figure 6.}

Government Bond Yields

View raw image

^{Figure 7.}

Exchange Rates

View raw image

^{Figure 8.}

Corporate Bond and Equity Prices

View raw image

^{Figure 9.}

Bank Deposit Rates

View raw image

^{Figure 10.}

Bank Lending Rates

View raw image

^{Figure 11.}

Bank Funding

Foresee

	Deposit Rates		Lending Rates
	Nonfinancial Firms	Households	Nonfinancial Firms	Households
Denmark	..	Oct/07-Nov/08, May/09-Aug/09, Jul/12-Feb/13	..	..
Euro area		Oct-08	Dec-11	..
Japan	..
Sweden		..	..
Switzerland	..	..	..	..

Topics

Countries

Series

About

Resources

Abstract

Introduction

Chapter 1 Conceptual Issues

Negative Interest Rates as a Policy Option

Why Seek Negative Policy Rates?

NIRP and Other Unconventional Policies

What Are the Limits to Negative Policy Rates?

The Effective Lower Bound

The Effective Lower Bound in Emerging Market and Developing Economies

Reversal Rates and Other Technical Minima

The Expected Effects of Negative Interest Rates

Effects on The Yield Curve, Bank Rates, and Exchange Rates

How Might Negative Nominal Rates Affect Households’ Portfolio Choices?

How Might Negative Nominal Rates Affect Commercial Bank Profits and Lending?

What Could Be the Effects on Nonbank Financial Intermediaries?

Cross-Border Spillovers

Summary

Chapter 2 Overview of Empirical and Quantitative Evidence

Cash Usage

Financial Variables

Money Market Rates

Yield Curves

Exchange Rates

Other Assets

Household and Firm Behavior

Bank Behavior

Deposit and Lending Rates

Bank Profits

Lending Volumes and Asset Quality

Bank Funding Structure

Money Market Funds

Other Nonbank Financial Institutions

Impact on Inflation and Output

Cross-Border Spillovers

Chapter 3 Implementation Issues

Tiering Reserve Regimes

Communications

Chapter 4 Conclusions

What Do We Know About the Effects of NIRP?

What Do We Not Yet Know About the Effects of NIRP?

Why Have Central Banks Not Resorted to NIRP More Often?

Share Link

Table of Contents

Headings

Figures

Metrics

Foresee