To understand why the zero-lower bound (ZLB) may pose a serious problem for central banks, we will discuss the operating principle of the respective tool. A reduction in the interest rate by the central bank has a two-pronged effect on returns on securities. First of all, reducing the central bank’s base rate means cheaper borrowing in the short-run. In addition, the current interest rate reduction affects the expectations of market participants regarding future trends in basic macroeconomic indicators and future monetary policy. If they perceive current interest rate reduction as a guidance on the central bank’s intent to pursue a looser policy in the future, both short-term and long-term interest rates will decline. Decreasing long-term borrowing costs stimulate investment and aggregate demand. Thus, a key component for an effective policy is strong correlation between short-term and long-term interest rates.

This correlation is threatened when the interest rate approaches zero (Ruge-Murcia, 2006). The problem here is that when zero is reached, the central bank’s ability to stimulate the economy through an interest rate policy becomes quite limited. Market participants realize that any further massive reduction in interest rates is impossible, and the central bank loses its ability to lower expected long-term rates. Two aspects must be taken into consideration in this situation: (1) the asymmetric nature of the effect of reducing and increasing short-term rates on the long-term end of the yield curve, and (2) the weak effect of short-term rates on long-term rates near the ZLB. Both aspects are attributable to the fact that market participants consider an interest rate policy to be ineffective near the ZLB, especially if this is a stimulating policy. These aspects of the policy effects were demonstrated based on Japanese data (Ruge-Murcia, 2006), while data on the U.S. stock market showed a reduction in the average impact of short-term rates on long-term rates (Grisse, 2015).

These aspects of the interest rate policy near the ZLB called for a revision in the role of interest rates as the main tool used by central banks in a liquidity trap. However, recent years have shown that the ZLB is often not rigid: in response to the crisis, the Riksbank, the National Bank of Denmark, the National Bank of Switzerland, the ECB, and the Bank of Japan have initiated negative interest rate policy (NIRP) by reducing rates on their basic deposit instruments to negative values. Thus, commercial banks keeping excessive reserves at the central bank were forced to put up with negative returns on these assets. The point of such rates is to make excessive reserves unprofitable for commercial banks. Theoretically, this may create incentives for massive lending or large-scale purchases of other financial assets by commercial banks which, in turn, should result in a growing aggregate demand and economic recovery.

The first experience with negative interest rates was by the Riksbank (the central bank of Sweden) which set a –0.25% interest rate on one-day deposit transactions in July 2009. At the same time, the rate on the basic instrument — one-week deposit and lending transactions (repurchase rate) — remained positive. The rate on deposit transactions remained negative for one year, after which monetary policy leveled out. Beginning in July 2014, the Riksbank was again forced to switch to negative rates (–0.5%) on deposit transactions. Moreover, within the two subsequent years, the bank made further rate reductions, which have been at –1.25% on one-day deposit transactions and –0.5% on repurchase transactions since February 2016.

The experience of the National Bank of Denmark is similar to the Riksbank in that the rate on one-week deposit certificates dropped below zero twice. The first bout of negative interest rates lasted from August 2012 to April 2014, followed by a short period of normal monetary policy and positive interest rates. However, the bank was again forced to switch to negative rates on deposit transactions in September 2014. Unlike Sweden, Denmark saw the beginning of a slight trend towards normal interest rates in January 2016: the interest rate on certificates of deposit was increased from –0.75% to –0.65%. Nevertheless, this process did not progress and Denmark still keeps these interest rates negative.

The ECB set the rate on deposit transactions at –0.1% in June 2014, which was followed by three reductions, and has remained at –0.4% since March 2016. At the same time, the lending interest rate was set at zero. The Bank of Japan was relatively late to emulate the ECB practices, setting a negative target rate on one-day deposit transactions (–0.1%) in June 2016.

The experience of the National Bank of Switzerland is largely similar to that of other central banks that established NIRP: in December 2014, the lower level of the target range for short-term deposit rates was set below zero, while beginning in January 2015, the entire target range for interest rates became negative. A distinctive feature of the interest rate policy in Switzerland was the sharp reduction in the target rate (from –0.25% to –0.75%) in January 2015, by contrast with the gradual reduction in rates by other central banks.

Thus, rather sufficient observations have been made by this time to estimate the effects of NIRP. Certain tentative estimates challenge the notion that interest rates reduce their effectiveness as the main monetary policy tool when the zero level is reached. For example, an analysis of the experience of the central banks in Europe has shown that the transmission mechanism of moderately negative interest rates does not differ from the case of positive rates (Bech and Malkhozov, 2016). Based on data from Switzerland (Grisse, Schumacher, 2017) and 14 developed countries (Grisse et al., 2017), a detailed analysis has been performed on the correlation between short-term and long-term rates when approaching and then crossing the zero level. An asymmetric effect from reducing and increasing rates (predicted in Ruge-Murcia, 2006) was actually observed when rates approached zero. However, when central banks lowered the lower level of the target corridor of interest rates to a negative value, the effect of interest rate changes on the long-term end of the yield curve persisted. Moreover, after the zero level was crossed, the average effect of short-term rates on long-term rates not only failed to decrease, but even became stronger than when short-term rates were far from zero.

These results may be an indication that market players’ expectations have changed (Grisse and Schumacher, 2017). Before negative interest rate corridor boundary values had been introduced, market participants had considered the ZLB to be a serious problem for monetary policy. That is why regulators had failed to duly influence the expectations of market participants by reducing rates. Crossing the zero level demonstrated that the ZLB is, in effect, insignificant and that regulators can reduce interest rates further. Thus, the success of the interest rate policy at rates approaching zero is largely dependent on the expectations of market participants regarding the effective lower bound (EFB) on interest rates.

This conclusion is consistent with estimates of the EFB perceived by stock market participants’ interest in Europe (Lemke and Vladu, 2017). As actual rates in the eurozone approached zero, market participants perceived the EFB to be positive, albeit very low. In September 2014, EFB dropped sharply to –0.11%. The authors of the study attributed this to crossing of the zero level by the actual rates, which showed market participants the potential for the NIRP. Thus, the effectiveness of the interest rate policy is largely dependent on the regulator’s ability to convince market participants that the EFB lies lower.

One should hardly expect further interest rate reductions to have significant potential for stimulating the economy. The current negative interest rates can be considered moderate on the whole. However, their further reduction may cause rates for private non-bank agents to hit negative values. This, in turn, would force agents to forgo cashless payments and prefer cash, which earns no income but is not exposed to negative interest rates. Currently, certain costs associated with using cash (e.g. storage, transportation costs) prevent the massive abandonment of bank instruments. Nevertheless, concerns are being raised that further interest rate reductions or remaining below zero for a protracted period would lead to attempts by economic agents to cut the costs by using cash, which may ultimately take control over the money market from central banks (Bech and Malkhozov, 2016). Moreover, the NIRP has an adverse effect on the profitability of banks, which do not want to lose their clients by passing on costs of negative rates to them (Michail, 2019), which creates additional threats for the banking sector and further stimulation of the economy.

Central banks that have followed the NIRP are fewer in number than those that applied alternative unconventional policy measures, i.e. quantitative and qualitative easing, as described below. This is why one should not overestimate the effectiveness of the NIRP in countries still facing a need to stimulate the economy due to a liquidity trap.

Large-scale asset purchases by central banks, to influence commercial bank reserves and, through them, the macroeconomic equilibrium, are not a recent invention. As far back as 2001, the Bank of Japan launched the first official quantitative easing campaign, in an attempt to stimulate the economy. However, only after the global crisis did such programs achieve an unprecedented scale. As compared with pre-crisis levels, the value of central bank assets increased roughly two-fold in the eurozone and four-fold in the UK and the U.S. by 2013 (Joyce et al., 2012). The asset balance for the Bank of Japan increased comparatively moderately over the same period (Gambacorta et al., 2014).

Differences in timing between these programs are one of the important reasons for the differences in central bank balances during the first half of the post-crisis period. For example, the first quantitative easing program in the U.S., QE1, was announced in November 2008, whereas the Bank of England announced large-scale asset purchases in March 2009. The Bank of Japan and ECB’s first steps to buy assets were taken as early as summer 2008, but the main stages of the program were launched much later: in 2011 and 2012 by the ECB under pressure from the banking sector, and in 2013 by the Bank of Japan, forced by the need to fight the recession and due to the ineffectiveness of previous measures. The expansion of asset purchase programs led to a sharp increase in the central bank’s assets. By 2016, the Bank of Japan’s balance increased four-fold, while the ECB’s increased three-fold compared with pre-crisis levels (Borio and Zabai, 2016). Thus, the total scale of asset purchases was approximately the same for the central banks under review, although, as we see below, it had different economic effects.¹

The different results of asset purchases could be partly attributed to the specific tasks facing the central banks which led to differences in the designs of these programs. Both the Bank of England and the U.S. Fed aimed mostly at reducing long-term interest rates to stimulate the economy (Joyce et al., 2012). The main asset purchase targets for the Bank of England were government bonds, while the Fed combined purchases of government securities and mortgage instruments. At the same time, the ECB’s main task became preventing a banking crisis within the eurozone, which is why repurchases formed the basis for its unconventional transactions. The Bank of Japan faced the challenge of overcoming a protracted liquidity trap; therefore its programs combined lending to commercial banks and purchases of government and corporate securities.

The great variety of asset purchase programs pursued by major central banks led to a certain difficulty in their classification and some confusion around definitions. The literature most often cites quantitative, qualitative, and credit easing programs. Quantitative easing programs include those that significantly increase the money base, including asset purchases and credit transactions by the central bank (Fawley and Neely, 2013). Credit easing programs include those aimed at improving particular lending terms by reducing particular rates or by restoring a particular segment of the financial market. Thus, credit easing programs, which increase the central bank’s asset balance, are also considered to be quantitative easing measures. According to another classification (Joyce et al., 2012), quantitative easing programs include government security purchases, while credit easing involves the purchase of commercial securities. According to this classification, the sets of credit and quantitative easing measures do not overlap, while the repurchase transactions actively pursued by the ECB do not fit any of these definitions.

Qualitative easing covers programs aimed at reducing the risks associated with assets owned by private agents in the market (Farmer, 2012). For example, such a program may include the purchases of private high-risk securities by the central bank with the simultaneous sale of risk-free government bonds. Such transactions effectively transfer risk from private agent balance sheets to the central bank. Notably, the official policy pursued by the Bank of Japan since 2013 is called “quantitative and qualitative easing” (QQE). The word “qualitative” was included in the title to distance it from Japan’s failed quantitative easing between 2001 and 2006, and from the policy pursued by the Fed (Ryuzo and Tatsuyoshi, 2017). In fact, the purpose of QQE was to double the money base, to which end the central bank conducted large-scale purchases of government bonds. As we can see, there is no single classification for these various asset purchase methods at this point. However, most of the programs actually carried out by central banks successfully combine various easing options, making the discussion on the definitions of various measures sound somewhat spurious.

All asset purchase programs are ultimately aimed at reducing long-term interest rates which, if successful, would stimulate investment and aggregate demand. Moreover, asset purchases from commercial banks improve their liquidity, resolve possible balance issues, and produce funds for investment. The literature on the effects of such unconventional monetary policy measures examines whether long-term interest rates are reduced by large-scale asset purchases. There is a consensus at present that asset purchases reduce long-term rates, but estimates of the absolute values of those reductions vary (see, e.g., Gagnon et al., 2011; Joyce et al., 2011; Krishnamurthy and Vissing-Jorgensen, 2011; Hamilton and Wu, 2012; D’Amico and King, 2013; Borio and Zabai, 2016; Altavilla and Giannone, 2017). Most often these estimates are based on an event analysis methodology which uses high-frequency financial data to determine the instantaneous effect of new asset purchase announcements, free from all other factors. Unfortunately, high-frequency data cannot be used to estimate the macroeconomic effects of unconventional policies as inflation, output, and unemployment statistics are published quite rarely. This is why researchers trying to estimate the effect of asset purchases on output and inflation are forced to tackle the complicated task of purging data from other influencing factors. The resulting estimates diverge so greatly that it makes any definite conclusions regarding the effectiveness of these programs quite complicated.² Moreover, inflation and economic growth rates are still considerably lower than before the crisis, leaving open the question whether asset purchase programs are effective (Borio and Zabai, 2016).

Moreover, the effectiveness of an unconventional policy depends on the stage of the crisis: the maximum effectiveness can be achieved if the central bank has managed to take measures during the acute stage (Bech et al., 2014; Jannsen et al., 2015). The effectiveness of programs declines during the recovery stage of a financial crisis. This might be an explanation for the effect discovered in an analysis of the impact of asset purchase programs by the Fed and the Bank of England (Hesse et al., 2018): the initial programs yielded quite strong effects, whereas those of latter programs had virtually no effect. A comparison of macroeconomic trends between the United States and the eurozone shows that the obvious differences between the results of asset purchase programs are at least partly attributable to the fact that the measures were taken during different periods. In the U.S., the first QE programs were launched practically at the peak of the crisis, whereas in the eurozone the active implementation of programs was protracted, which is already one reason why their impact was weaker.

Making asset purchases over a long period of time as the main method to counter the aftermath of a crisis poses threats for the economy. For example, as such programs lose their effectiveness, the scale of asset purchases must increasingly be expanded (Jannsen et al., 2015). This increases the chance of bubbles in financial markets and financial crises due to increasing risk-taking. In addition, asset purchase opportunities are limited by the market size, which hardly makes asset purchases a panacea against crises. Consequently, to increase the effectiveness of the above programs, they should be augmented with measures that influence the expectations of market participants.

Attempts by regulators to influence the expectations of private agents produced a system of forward guidance on future monetary policy changes, which is a vital component of any central bank’s information policy. The key role of such guidance is that, in a liquidity trap, possibly the only opportunity for macroeconomic policy to stimulate the economy is the central bank’s promise that even after overcoming the crisis, interest rates will remain low for a long time (see, e.g., Jung et al., 2005). If private agents trust these promises by the authorities, this will motivate them to increase spending and will stimulate business activity even during the current period of zero interest rates. In fact, all leading central banks have used this tool to varying extents, promising that rates would remain low over a lengthy period. Notably, the effect of such guidance on long-term interest rates was, on the whole, positive in the U.S. and Japan, although some authors doubt the effectiveness of this tool in the eurozone and in the UK.³

In order to boost confidence, the central bank needs to develop an effective guidance system that would be seen by the population and key macroeconomic agents as a binding commitment by the central bank (Eggertsson, Woodford, 2003). The way to build a successful forward guidance system is for the central bank to take consistent and timely measures. A key role in this strategy is given to forward guidance on future changes in the monetary policy, which would be perceived by private agents as the central bank’s binding commitment (Bodenstein et al., 2012). However, they will only trust guidance on interest rate changes (increasing inflation expectations in a liquidity trap) if the central bank has a solid reputation (Nakata, 2015).

There is a discussion on changing the official targets established by central banks to overcome the deflation bias and build a reputation as a less conservative agent. Two policy modification options are usually considered: adding alternative goals to the inflation goal and changing the inflation target rate. For example, concerns have been raised that the target inflation rate of 2% that is accepted in many developed countries may turn out to be an insufficient buffer to prevent future liquidity traps if the reduction in real interest rates takes on a more permanent nature (Williams, 2009). Central banks have been recommended to prefer a 4% target inflation rate instead of 2%, as this would not lead to considerable adverse effects for the economy and would keep the economy from falling into a liquidity trap (Ball, 2013). At the same time, increasing the target inflation rate means a commitment to pursue a more stimulative policy in the future as compared to the current target inflation rate.

According to some estimates, the optimal inflation rate is positively correlated with the probability of falling into a liquidity trap and its average duration (Carreras et al., 2016). Therefore, the calibrations of the New Keynesian model, consistent with data from the most recent liquidity trap episodes, admit an optimal inflation rate exceeding 2%. Moreover, in a model that takes menu costs into account, and consistent with micro data on the price behavior of firms, the optimal inflation rate was placed at 5% (Blanco, 2016). Still there is an evidence that the optimal long-term inflation rate depends on the discretion of monetary policy (Billi, 2011). If the central bank has access to the binding commitment technology, the optimal inflation rate in the model does not exceed 1% in the U.S.; if the technology is unavailable and the central bank implements a discretionary policy, the optimal inflation rate exceeds 10%.⁶ The Japanese experience demonstrates that increasing the target inflation rate may encourage economic recovery (Michelis and Iacoviello, 2016).

However, there are concerns that this kind of policy shift may strongly affect the population’s confidence in the central bank (Walsh, 2011; Blinder et al., 2017). It is noted that any modification of the target inflation rate should be gradual and accompanied by informational support from the central bank, so that market participants understand the reasons for the policy changes (Blanchard et al., 2010). This would enable them to adjust their contracts to the new conditions and to avoid a redistribution of wealth from lenders to debtors. In addition, a full explanation of the reforms being undertaken would help to maintain the confidence of the economic agents in the central bank which would pave the way for a successful future policy.

Adding new target indicators to the central bank’s loss function, or changing their relative weights, can also help in fighting the liquidity trap. For example, increasing the weight of output in the standard Taylor rule will reduce the adverse impact of negative shocks on output when the target inflation rate is reached (Williams, 2009). However, the events of recent years bring the question to light regarding the potential transition from inflation targeting to price level targeting (Walsh, 2011). This policy presumes that the economic recession increases the gap between the real price level and the target trajectory of their level. As a consequence, in the future the central bank would be forced to pursue a more stimulative policy to narrow this gap. Thus, this policy entails a higher inflation rate in the future (exceeding the target in the case of inflation targeting), lower long-term interest rates, and higher recovery rates for aggregate demand. Consequently, targeting the price trajectory may solve the time inconsistency problem for an optimal policy. It has been demonstrated that switching to price level targeting may improve the effectiveness of guidance on future changes in monetary policy (Cole, 2018). Nevertheless, switching to this policy during a crisis may considerably undermine confidence in the central bank, thus calling for further discussion (Walsh, 2011).

Another way to solve the deflation bias in a liquidity trap may be an issuance of non-indexed public debt in order to influence future macroeconomic policy and shape inflation expectations. A deflation bias could be regarded as a problem with the population’s lack of confidence under liquidity trap (Eggertsson, 2006). In this situation, increasing public debt, whose real value depends upon inflation, creates future incentives to keep the interest rate low to reduce debt service cost on the national debt and to raise the inflation to devalue the nominal public debt. These measures may have a positive effect on the population’s confidence level and assist in developing an effective macroeconomic policy in a liquidity trap.

This idea was developed further in a review of quantitative easing policy (Bhattarai et al., 2015) As part of effective quantitative easing, the population receives guidance that by purchasing private securities, and thereby increasing the nominal debt, the state shows interest in maintaining low interest rates not only at the present time but also in the future, when near-zero rates cease to be a limitation for implementing effective macroeconomic policy.

However, accumulating nominal public debt should not be regarded as an obvious solution for improving the effectiveness of macroeconomic policy in a liquidity trap. Increasing nominal debt will inevitably affect output, exchange rates, financial stability, and, accordingly, the economic situation in general. Therefore, certain caution is required in using nominal public debt as a tool to influence the expectations of private agents in a liquidity trap.

Admitting that promises to keep the interest rate lower for longer are themselves associated with the significant difficulties in improving the population’s confidence, there are suggestions to shape expectations for a lower national currency value in the future. The logic of this idea is as follows: expectations for a lower national currency value mean expectations that the competitive power of domestic products in the global market and aggregate demand will grow. Also, a cheaper national currency requires maintaining lower real interest rates in the future. Thus, promises to devalue the currency in the future are, in a large part, a substitute for a policy of guidance on future rate reductions, but may cause the problem of a time-inconsistency of optimal policy.

For example, a scheme has been suggested for creating binding commitments for central banks (Svensson, 2000). First of all, an optimal policy should include a temporary transition to price level targeting, while the target price level should grow over time. At the same time, the central bank must devalue the national currency and switch to a crawling peg, so that the devaluation rate is pegged to the gap between the domestic inflation rate and the average global inflation rate. As soon as the ultimate price target is reached, the central bank can give up foreign exchange rate control and switch either to standard inflation targeting or to price level targeting.

It is easier for the central bank to shape expectations with respect to its commitment to foreign exchange rate control than to reducing interest rates in the future, as currency devaluation for the central bank is associated with purchases of foreign reserves and increasing the money supply. Unlike a revaluation, this process may last infinitely, as the central bank is not limited in increasing money supply. Ideally, the recognition of this policy (unlike the disputable future interest rate reduction) by the population boosts their confidence. By demonstrating its commitment to its promises regarding the exchange rate, the central bank may ultimately convince the population that the current policy will be successful, which is just what is needed to solve the time inconsistency problem.

The central bank’s concerns regarding the size of its net worth and the fear of losing independence also create a necessary binding commitment mechanism to maintain a low national currency value in the future (Jeanne and Svensson, 2007). Most central banks try to maintain a rather large net worth, since a considerable reduction in net worth may force the central bank to turn to the government for help, which would mean a lower degree of independence from the government. Thus, accumulating foreign reserves, combined with the current devaluation of the national currency, creates a binding commitment for the central bank to avoid currency appreciation, since a sharp appreciation in the currency would reduce the value of foreign reserves in the national currency, thereby reducing net worth. Thus, the central bank finds incentives to avoid currency appreciation in the future and, accordingly, to maintain lower interest rates to solve the time inconsistency problem.

Regardless of the central bank’s choice of method to stimulate the economy in a liquidity trap, its effectiveness will be contingent upon whether the bank has a solid reputation, and whether the announcements by central bank representatives are perceived as the regulator’s binding commitments. The policy of maintaining low interest rates following severe crises is most likely to enjoy confidence if the private agents expect periodic future shocks that hinder economic activity (Nakata, 2018). If such shocks threaten, the central bank will be reluctant to raise the nominal rate immediately after overcoming the crisis, as it may otherwise provoke a significant economic recession and thereby severely undermine the population’s confidence in its policy. Thus, a potential loss of reputation given comparatively frequent shocks will force the central bank to honor its promises and maintain a low nominal interest rate even during an economic recovery. In this case, due to the potential threat of shocks, the long-term motives for maintaining a low nominal rate will be more important for the central bank than short-term motives for fighting inflation. Private agents will recognize the motives behind the central bank’s behavior in the current circumstances, making it possible to solve or at least mitigate the time inconsistency problem and resulting in an effective macroeconomic policy in a liquidity trap.