The central bank needs blockchain technology, as demonstrated by Bitcoin, to implement monetary policy

Translator's note: Why does the central bank issue digital currency based on blockchain? Because it can make the central bank's tools for implementing monetary policy effective. This view was mentioned by Andrew Haldane, chief economist of the Bank of England, in a public speech on September 18: The problem of monetary policy failure caused by the zero lower limit of interest rates can be alleviated in the short term through quantitative easing, but this will affect the independence and reputation of the central bank, and quantitative easing will eventually become ineffective. From some statistical data, the monetary interest rate is even a key factor in triggering economic and financial crises when it approaches the zero interest rate constraint, at least it has made a prediction. Breaking the zero interest rate lower line will prompt the central bank to issue digital cryptocurrency.

This article is a simplified version. The full English-Chinese translation is 30 pages in A4. Due to the 20,000-word limit on WeChat public accounts, it is difficult to send the full text. If you need it, please contact: wangliren123.

——————————

The global financial crisis has raised a host of questions about the future of central banking. These are the most fundamental challenges central banks will face in this generation, and perhaps longer. They include the role of money, macroprudential and microprudential policy, and optimal transparency and accountability. I will discuss two topics: the future of money and the future of monetary policy.

Both topics concern the core of the central bank’s special nature – its balance sheet. The central bank’s balance sheet is the basis for money and monetary policy making. The liabilities of the central bank define the amount of the so-called base money in circulation. The interest rate on central bank money defines monetary policy. In this sense, the money issued by the central bank and monetary policy are two sides of the same coin.

However, in practice, money and monetary policy are often separate issues. The liabilities of a central bank consist of two elements - money in society and bank deposits. In most central bank operating procedures, non-interest bearing money is supplied to the public on demand and flexibly, while the monetary policy stance consists of regulating the interest rate on bank reserves. In this way, the link between money and monetary policy has been severed.

The two themes have reconnected to some extent since the financial crisis. With interest rates falling to the floor in many developed countries, central banks have engaged in so-called quantitative easing (QE) - buying assets and lending to banks' accounts at the central bank. So far, these asset purchases have totaled $5 trillion, with more planned. The central bank's monetary policy arsenal has expanded.

These “unconventional” monetary policy actions reestablished some of the connection between central bank money issuance and monetary policy. But these unconventional policies were temporary, primarily measures related to the financial crisis. As quantitative easing becomes less effective over time, the connection between central bank money issuance and monetary policy will wither. The two sides of the same coin will be separated again, back to square one.

However, it is possible that monetary conditions will not fully return to their original state. If global real interest rates continue to fall, central banks may need to think about how to deal with the technical constraints imposed by the zero lower bound on interest rates on a more permanent basis. This may require a rethinking of the current practices of many central banks, which is a fundamental consideration.

I will discuss some medium-term options for dealing with this technological constraint. Since all of these options would represent significant shifts from the past, they would benefit from further study.

These are options for the future, and now is the time to conduct this research. But they are also relevant to monetary policy today, which I will reflect on at the end.

What is the zero lower bound?

After the global financial crisis, short-term interest rates in many countries fell sharply and have remained low ever since. Among the sample countries worldwide, 40% of short-term interest rates are below 1%, nearly two-thirds are below 3%, and 80% are below 5% (Figure 1). In some European countries, short-term interest rates have entered negative territory.

In large advanced economies, official interest rates are effectively zero. Japan has had official interest rates at zero for more than 20 years. In the United States, the eurozone and the United Kingdom, official interest rates have been zero for six years and counting. These countries have all enhanced their monetary policies, such as large-scale quantitative easing programs, injecting liquidity into the banking system and providing monetary policy guidance.

The need for unconventional measures stems from a technical constraint - money rates cannot be negative. You can set the interest rate on bank reserves negative - in fact, some countries have recently done so. But you can't do the same for currency, and whenever they turn negative, there is an incentive for deposit reserves to move into currency. It hinders the effectiveness of monetary policy and is known as the zero lower bound or zero interest rate lower bound problem.

The issue of the zero lower bound is in some sense nothing new. It was discussed during the greatest and most devastating financial crisis of all time, the Great Depression. Keynes warned of the ineffectiveness of low interest rates in his book General Theory (Keynes (1936)). This led to his concept of the “liquidity trap”. Keynes also had a number of imaginative monetary and fiscal solutions to this problem, which I will discuss later.

For quite some time after the Great Depression, the zero lower bound issue disappeared from the policy horizon. It became an academic topic rather than a policy topic. Based on simulations conducted before the crisis, this neglect seemed benign. For example, Reifschneider and Williams (2000) found that with an inflation target of 2%, monetary policy was constrained by the zero lower bound only about 5% of the time.

More recent experience is instructive. The zero lower bound has become a real and persistent constraint for some central banks and a potential constraint for many more. Pre-crisis simulations of the possibility of a zero lower bound constraint were overstated by the low level of macroeconomic risks that characterized the Great Moderation. Recalibrating the experience of macroeconomic risks during the Great Recession has made the zero lower bound a clear and present danger.

However, the prevailing orthodoxy among academics and policymakers is that, while the zero lower bound has been more persistent than expected, the issue will eventually pass. As countries recover from the Great Recession, the zero lower bound constraint will relax, its policy relevance will weaken, and the debate on the zero lower bound will return to the academic stage, ultimately characterizing it as one of the lessons of the Great Depression.

However, this may be the wrong lesson. It is not just the crisis itself that caused the zero lower bound constraint: it in fact seems to have taken root long before the crisis. It is questionable whether the constraint will disappear once the global recovery is complete: the roots of the zero lower bound constraint may be structural and long-term. The “Great Recession”, in this sense, is different from the Great Depression.

Consider the recent behavior of global long-term real interest rates, see the graph.

Figure 2 plots the long-term real interest rates around the world over the past 35 years. Back in the 1990s, global real interest rates averaged around 4%. With an inflation target of 2%, this meant that nominal interest rates averaged around 6% over a typical cycle. At this level, monetary policy would have plenty of wiggle room around the zero lower bound—6%—to develop strategies to cushion the effects of business cycle troughs.

Yet over the past 30 years, real interest rates have fallen around the world (Broadbent (2014)). At the dawn of the crisis, they had halved to around 2%. Since then they have fallen further to zero and even negative values. A 2% inflation target would keep nominal interest rates at 2% on average over the cycle. This means that monetary policy has substantially less room for maneuver than it did a generation ago.

Too little headroom? One way to measure this is to look at past “typical” monetary policy easing cycles. Figure 3 plots interest rates in the UK, US, Japan and Germany since 1970, and Table 1 looks at easing cycles in these countries since 1970 and 1994 (Haldane (2015)). The “typical” easing cycle is between 3-5%. Either way, 2% headroom is usually not enough.

Another probability measure of the zero lower bound constraint is the probability of a recession. Table 2 looks at the cumulative recession probability measured over three time horizons (1, 5, and 10 years) and in three historical samples (1700, 1945 onwards for the UK, and 1870 onwards for a multi-country portfolio). Recessions occur roughly every 3 to 10 years. Over the course of a decade, they are extremely likely to occur.

So given these recession probabilities, what is the probability that interest rates will be at a level that will be sufficiently cut to cushion the effects of a typical recession? We can calculate the probability of a 3% interest rate using the market yield curve. These market-based probabilities for different horizons are shown in the last column of Table 2.

This suggests that the probability that policymakers will need 3 basis points of interest rate headroom is significantly greater than the probability that this headroom will be available. In other words, interest rates may need to go to zero at some point in the future. While not stated, these estimates suggest that the zero lower bound will exert a strong gravitational pull on interest rates for some time.

If the yield curve does not indicate the future path of global interest rates very well, this calculus will exaggerate the gravitational pull of the zero lower bound. For example, if global real interest rates quickly return on average to their historical levels of 2-3%, monetary policy space will be freed up again (Figure 4). While the Great Recession and the Great Depression look different today, at least in terms of long-term real interest rates, they may look different in the future.

To assess the likelihood of real interest rates reverting to the mean, we first need to understand why they are falling. Colleagues at the Bank have recently attempted to identify the drivers of global real interest rates (Rachel and Smith (forthcoming)). The factors they identify can be considered to account for most of the 450 basis point fall since the 1980s. These factors include low trend growth (100 basis points); worsening demographic trends (90 basis points); low investment rates due to falling input prices (50 basis points); rising inequality (45 basis points); and excess savings in emerging markets (25 basis points).

These factors are not illusory and are not likely to reverse quickly. So low levels of global real interest rates are likely to persist, well below their long-run averages. This time may indeed be different, not only from the recent past (1980s and 1990s) but also from the distant past (1930s). If so, central banks may find themselves running up against the zero lower bound on interest rates on a recurring basis.

Ignoring the zero lower bound is no longer a benign proposition, so even if policymakers do not know how long it will constrain, risk management based on thinking about policy options that could relax the zero lower bound on a lasting basis is beneficial (see Bitter and Panigirtzoglou (2003), Bitter (2004), Bitter (2009), Kembauer (2015)). From a brief list, I discuss three such options, each of which represents a significant departure from current central bank practice and would therefore benefit from further research and reflection.

Review of monetary policy provisions

Over the past few decades, inflation targets in advanced economies have fallen steadily, typically clustering around 2% (Figure 5). In emerging economies, they have fallen faster still, currently around 4%. The decline in average inflation targets has reflected a decline in inflation itself (Figure 6). In fact, inflation is currently not meeting the above targets, running slightly below them on average by about 1.5 percentage points.

After the inflation panics of the 1970s and early 1980s, the deflationary trajectory has been very beneficial. Nonetheless, the steady decline in inflation and the decline in global real interest rates have come at a cost. Suppressing inflation through tight monetary policy has output costs, albeit temporary ones. More importantly, by reducing the steady-state level of nominal interest rates, a lower inflation target will increase the probability of binding at the zero lower bound.

So, one option to relax this constraint is to raise the inflation target. For example, raising the inflation target from 2% to 4% would provide 2 percentage points of room for interest rate wiggle. About an order of magnitude of researchers think this is desirable (Ball (2014) and Blanchard et al (2010)). Simulations show that a 4% inflation target gives enough monetary policy room to cushion almost the largest recession in history (Leigh (2009)).

In other words, the optimal inflation target may be different, depending on the government and the likelihood of being bound by the zero lower bound. This likelihood in turn depends on the equilibrium level of the real interest rate. If the equilibrium real interest rate shifts, the optimal inflation target should also shift (Reifschneider and Williams (2000)). That is, in theory, this would also support a revision of monetary regulations.

What are the costs of doing so? The welfare costs of inflation rise through a variety of channels and are well documented in the literature. But there is no evidence that these costs increase when steady-state inflation rises from 2% to 4%. Cross-national studies find evidence of a negative relationship between inflation and economic growth, but this negative effect is generally observed only when current inflation is above the inflation target.

Now let me add some important caveats: This evidence comes almost entirely from periods of high and falling inflation. The dynamics of inflation expectations in periods of high and steadily falling inflation and periods of low and abrupt increases in inflation are likely to be fundamentally different. There is a risk of excesses and asymmetric responses in inflation expectations at a turning point.

The pattern of UK inflation expectations over time suggests a stable sequence of deflationary ratchets, changes in monetary policy regimes, and a gradual build-up of credibility (Figure 7). Inflation expectations have been anchored and increasingly tightly fixed around the inflation target. Once anchored and fixed, they rise and fall with the waves. Even through the storm of the Great Recession, long-term measures of UK inflation expectations have been remarkably stable.

Reputations, in all walks of life, are hard-earned and easily lost. The same is true of inflation reputation. Therefore, deliberately not fixing the inflation rate, but promising to fix it at a certain level, takes unknown monetary risk. Once the inflation rate is not fixed, inflation expectations are more difficult to explore. Uncertainty about the direction and level may undermine macroeconomic stability.

There is a further point to remember here, which is that society's preferences are different from what economists say. The costs of inflation calculated by economists may be different from what the general public perceives. International surveys show that the public is much more concerned about inflation than the academic literature suggests (Shiller (1997)).

In the UK, the Bank of England regularly conducts surveys of the public’s attitudes towards inflation and the inflation target. This reveals society’s inflation preference. The public’s attitude towards the Bank is highly correlated with their perception of inflation (Figure 8). This highlights the importance of low inflation to the reputation of the central bank. It also shows that the public prefers inflation below the current target rather than above it.

Consistent with the above, when the public is asked directly for their opinion on inflation targets, they generally think that the inflation target is probably a bit too high (Figure 9). Taken together, this evidence suggests that an inflation target above current levels would not only put central banks’ reputations at risk. More importantly, it would be out of sync with public preferences.

The choice of the UK inflation target is rightly ultimately a decision for the government, not the Bank of England. But Friedrich Hayek once likened controlling inflation to holding a tiger by the tail (Hayek (1979)). It seems to me that at a time when we seem to be taming the tiger, twisting its tail would be a courageous move.

The second policy option is to accept the zero lower bound constraint and allow currently “unconventional” monetary measures to become “conventional.” This could mean including quantitative easing as a monetary tool in the monetary policy arsenal throughout the year (both in normal times and in times of crisis).
This approach has its appeal. Unlike raising the inflation target, it does not lead to unknown monetary conditions. Quantitative easing has been implemented by many central banks in advanced economies in recent years, providing a rich evidence base for assessing the effectiveness of this policy approach. Moreover, this evidence base suggests that the quantitative easing policies adopted so far have achieved the desired effects.

One way to assess the effects of quantitative easing is to evaluate how asset prices react after the announcement of the intervention. Event studies have found that most quantitative easing interventions have statistically significant effects on asset prices, such as short-term and long-term interest rates, corporate bond yields, exchange rates, and financial market uncertainty.

Figure 10 is a simple eyeballing of the responses of financial market variables to a series of quantitative easing interventions by the Bank of England, the Federal Reserve, the European Central Bank, and the Bank of Japan over the past few years. Each intervention is measured as a percentage of the country's GDP. These charts are not meant to predict quantitative easing, but in general the responses are largely correctly predicted. In many of these cases, they are also statistically significant.

The second strand of this literature is to look beyond the immediate effects of QE on asset markets and focus instead on medium-term demand and inflation. Typically, these studies use validated VAR time series techniques (e.g., Baumeister and Benati (2012)). An example is the work of my colleagues at the Bank, Martin Weir and Thomas Wieladek (Weir and Wieladek (2015)).

Figure 11 is from Weale and Wieladek and shows the impact of quantitative easing policies equivalent to 1% of annual GDP on real GDP and inflation in the UK and the US using one of the four identification schemes proposed in their study. The impact is correctly predicted and is statistically, quantitatively and persistently significant, with an average peak impact of 0.3% of GDP in the following 12 months.

On the surface, these simple math sheets make it seem plausible that QE has a strong, immediate impact on stimulating demand and inflation, with a shock coefficient that is no different from conventional interest rate policy. On the surface, this suggests that QE may be a practical and proven way to keep the monetary policy engine running and the economy moving slowly, and in the future, finds interest rates stuck at the zero lower bound.

The case against QE becoming a permanent monetary policy tool, in my view, is based on three considerations. First, the effects of QE as a monetary instrument are likely to be highly state-dependent and therefore uncertain, at least relative to interest rates. This uncertainty is not simply a result of the evidence being more limited about QE than about interest rates. Rather, it is an inherent property of the transmission mechanism of QE.

The effectiveness of all monetary interventions depends on market imperfections. The non-neutrality of interest rates depends on imperfections in goods and labor markets. Commodity price and wage stickiness—for example, due to cross-wage and price contracts (Taylor (1979), Fischer (1977), Calvo (1983), Yun (1996), Mankiw and Reis (2002))—allows changes in nominal interest rates to affect real activity. The effectiveness of QE also depends on these frictions in goods and labor markets. But it also depends on imperfections in asset markets. These are necessary for portfolio adjustments induced by QE to have an impact on risk premia and asset prices (Joyce et al (2014)).

Market frictions are not all created equal. Goods and labor market frictions arise from contractual arrangements or rules of thumb between employers (buyers) and employees (sellers), labor or buyers and sellers of goods (Blanchard and Fisher (1989)). Because changing these arrangements is costly, they tend to be static. Thus, goods and labor market frictions are relatively fixed, or at least not variables that change over time.

This is not the case with asset market frictions. Capital market frictions are shaped by the constraints of investors’ portfolios and risk preferences (Vayanos and Vera (2009)). Both of these factors are likely to vary over time and are highly state-dependent (Baker and Wurgler (2007) and Groom, Sapienza, and Zingales (2013)). Portfolios themselves change rapidly and at high frequencies, as do risk premia. Changes in asset risk premia are therefore large and unpredictable (Bollerslev and Todorov (2011)). Asset prices tend to exhibit excess sensitivity. Asset market returns are therefore likely to be volatile, fat-tailed, and highly state-dependent (Haldane and Nelson (2012)).
All of this has direct implications for the transmission mechanism of QE. If asset frictions are highly state-dependent and unstable, the effects of QE will be similar. Estimates of the effects of QE during periods of high risk premia and turbulent financial conditions are likely to be very different from those during periods of stability and low risk premia.

In line with this, existing empirical studies point out that the margin error of the simple quantitative easing calculation table (Table 3). Quantitative easing policies of different stages and contents have completely different effects. Figure 12 compares the impact of Japan's first round of quantitative easing (2003-2008) and the second round (2008-2015). The calculation tables are not only different in scale, but also in sign (positive or negative?). The key point of microeconomics is that the uncertainty brought by QE is essentially greater than that brought by interest rates. This has a great impact on regulating its use.

Second, implementing large-scale quantitative easing or making it permanent could blur the line between monetary and fiscal policy, although the process is subtle. Consider the mechanism of quantitative easing, which usually involves the central bank buying government or private sector assets. If this is done permanently and on a large enough scale, it is a quasi-fiscal action.

If the central bank engages in quantitative easing by purchasing government bonds, this may have an impact on the cost of servicing debt – indeed, this is one channel through which quantitative easing operates. If the purchases are permanent, this also has implications for the amount of debt the government needs to issue given the fiscal stance. Either way, there are direct implications for the government’s budget constraint (Kirby and Meaning (2015)).

If QE is implemented through the purchase of private assets, although this may not have an immediate fiscal impact, it has an indirect effect on the government's intertemporal budget constraint. For example, if the risks of purchasing these private assets are made clear, the liability will eventually fall on the government's finances. Whether through public or private purchases, QE can lead to serious fiscal consequences.

This fuzzy fiscal/monetary boundary may be limited if asset purchases are modest and temporary. The quantitative easing currently being pursued in advanced economies probably meets these conditions. The fiscal/monetary boundary can also be drawn in a variety of ways to reduce the risk of fuzziness. For example, in the UK there are clear protocols to protect banks from financial losses caused by quantitative easing.

Nonetheless, if QE grows in size and persistence, the boundaries begin to blur. QE becomes both a fiscal and monetary policy risk, a victim of so-called “fiscal dominance” (Woodford (2001), Cochrane (2011), BIS (2012), Roubini (2014)). This would erode another monetary prize that has been hard-won over decades – central bank independence. In short, as QE becomes permanent, the credibility of monetary policy would decline dramatically.

Third, another channel through which quantitative easing works is through the exchange rate. Traditional interest rate policy is implemented through the exchange rate channel. But because quantitative easing directly affects reserve assets held by the private sector, the potential for spillover effects in asset markets, including the foreign exchange market, may be greater.

Signs of QE intervention coincide with a significant exchange rate reaction in many cases (Figure 13). The domestic currency depreciates after the announcement of domestic QE. However, especially for small open economies, spillovers from international QE can be equally important. Figure 14 shows the impact of UK and US QE on UK output and inflation, expressed as a percentage of gross domestic product (GDP). US QE appears to have had a greater impact on the UK economy than UK QE.

Given that the business cycles of the UK and the US are closely synchronized, cross-border spillovers are likely to have a positive impact on both countries. However, if the business cycles are biased, the situation is different, and the spillover effects of international quantitative easing policies may complicate the formulation of domestic monetary policy. In fact, the cross-border impact of quantitative easing can be seen as a potential externality to the international monetary system.

If quantitative easing is modest and temporary, as it has been in recent years, this systemic externality is small. But making quantitative easing permanent or operating on an industrial scale would amplify this systemic externality. In my view, this would have adverse implications for the long-term stability of the global financial system and lead to risks.

For these three reasons, I am skeptical that full-year QE is the ideal steady-state monetary solution to the zero lower bound problem. While I do so, I am unbiased about whether QE has a further temporary role to play in stimulating aggregate demand, or whether it is necessary.

Negative interest rates

This brings me to the third option, perhaps the most radical and enduring. It brings together money issuance and monetary policy and involves finding the technical means to impose negative interest rates or breaking the constraints that set such rates on physical money.

These options are not new. A century ago, Silvio Gesell proposed a stamp tax on currency to generate a negative interest rate (Gesell (1916)). Keynes discussed this proposal approvingly in his General Theory. More recently, a number of modern variants of currency stamp taxes have been proposed—for example, randomly invalidating the value of banknotes through serial numbers.

A more radical proposal is still to remove the zero lower bound constraint entirely by abolishing paper money. This proposal has recently found its supporters (e.g., Rogoff (2014)). In addition to solving the zero lower bound problem, it has the added benefit of taxing illegal activities such as drug trafficking that use paper money.

A third option is to set a clear exchange rate between paper money and electronic money (or banknotes). Assuming that electronic money is a publicly accepted unit of account rather than currency, a steady depreciation of paper money against digital money could effectively generate a negative interest rate, which is not a new idea.

In principle, all of these options could solve the zero lower bound problem. In practice, they all face a significant behavioral constraint. Government-backed money is a social convention, a unit of account, and a slightly more narrow medium of exchange. These social conventions are not easy to change, whether through taxation, conversion, or abolition. This is why, despite the lack of appeal of paper money, its demand continues to rise, at a rate higher than the GDP of many countries.

An interesting solution would be to maintain the principle of government-backed money, but issue it in electronic form rather than paper. This would preserve the social conventions of legal units of account and mediums of exchange, while the money is held in digital form rather than in a physical wallet. But it would make it easier and quicker to levy negative interest rates on the money, thus easing the zero lower bound constraint.

Is such a currency technology feasible? In a sense, there is nothing new about state-issued digital currencies, as is the case with commercial banks’ deposits with central banks. However, the technology that underpins digital cryptocurrencies has changed rapidly in recent years, and the reason for this change is simple: Bitcoin.

In its short life, Bitcoin has become a monetary enigma. It has sparked unprecedented divergence of opinion (e.g., Yermack (2013), Shin (2015)). Some countries have banned its use, while others have encouraged it. Some economists have accused it of being a panacea for the evils of fiat money. Others have declared it a panacea for the evils of fiat money.

I think it is now fairly clear that the distributed payment technology presented by Bitcoin has real potential. On the surface, it solves a deep problem in monetary economics: the “substance of money”, how to establish trust in a distributed network. Bitcoin’s “blockchain/ledger chain” technology seems to offer an imaginative solution to the distributed trust problem.

Whether a variation of this technology could support a central bank-issued digital form of currency is an open question. So too would the public accept it as a replacement for paper money. Central bank-issued electronic money also raises big logistical and behavioral questions. What kind of use would it actually have? What security and privacy risks would it raise? How would publicly and privately issued money interact with one another?

There are no easy answers to these questions. This is why work on central bank digital currencies is a core part of the Bank’s current research agenda (Bank of England (2015)). While the barriers to implementation are high, the potential gains from easing the zero lower bound are also high. Perhaps central bank money is mature enough for its own major technological leap, triggered by the pressing need for a zero lower bound.