1. Introduction
With this presentation, I would like to share the experience of the Bank of Albania (BoA) with the estimation of the lower policy rate bound, at a time, at the beginning of 2016, in which BoA was confronted with very low inflation rates, a challenging external environment, and a risk of disanchoring inflation expectations. The presentation is based on the framework developed and deployed by the Bank of Albania to estimate how far they could lower policy rates to counteract downside risks to price stability. The framework benefited from the IMF Technical Assistance (TA) funded by the Swiss Government. The reason we are presenting it here and have written a Working Paper is the belief of the authors that in a context of lower global real interest rates more frequently monetary authorities will be confronted with similar policy challenges and may benefit from the conceptual framework developed by the Bank of Albania.
In this presentation, I will first provide some information on the context in which the analysis underlying the Working Paper has been conducted. I will, then, present the framework used to conceptualize the problem of the lower policy rate bound. On the basis of this framework, the lower bound depends on three groups of factors: (1) those related to the effectiveness of the transmission mechanism; (2) those related to financial stability; and (3) some sundry extra factors. I will then present the framework developed by the BoA to monitor the dynamics of the variables in the above mentioned group. I will finally conclude.
2. The Context
At the beginning of 2016, the Bank of Albania, as evidenced by Figure 1, was faced with strong disinflationary forces deriving from disinflation from the euro area, declining commodity prices, and a one-off supply shock. The resolve and the timeliness with which BoA reacted, lowering the policy rate to the all-time low of 1.25 percent, anchored the inflation expectation and facilitated a gradual recovery of the inflation rate. At that time, however, when the extent and the length of disinflationary forces were unclear and the risk of dis-anchoring inflation expectations more material, key policy questions were: (1) how far policy rates can be lowered; (2) which factors should be taken into account to estimate the lower policy rate bound; (3) at which point non-standard measures could be considered relatively more effective in the pursuit of the price stability objective than conventional monetary policy; and (4) which measures would have been relatively more effective in Albania.
The paper and this presentation are based on the conceptual framework with which the Bank of Albania with IMF TA tried to address the first three questions. In our view, the same framework can be effectively used elsewhere.
3. Key Concepts
For the purpose of our analysis, we have introduced and defined three concepts: the zero lower bound, the technical lower bound, and the effective lower bound. Zero has long been considered synonymous with technical lower bound. Until a few years ago, the economic doctrine considered negative interest rates de facto impossible. According to the doctrine, since bank reserves and current account holdings can be converted into currency on demand and since currency has zero nominal interest rates, negative rates would have been impossible, as banks, companies, and individuals would have reacted to any attempt to lower the interest on deposits below zero by converting account holdings into currency. This assumption has never been fully accurate.
Already a century ago, Silvio Gesell, in 1916, proposed levying a stamp tax on currency to generate a negative interest rate. Keynes discussed this scheme, approvingly, in the General Theory. More recently, a number of modern-day variants of the stamp tax on currency have been proposed. However, even without recurring to these schemes, difficult to justify from a political perspective, the recent experiences of the Swiss National Bank, the Danmarks National Bank, and the ECB have shown that policy rates can be lowered below zero. Cash entails holding and transaction costs so that banks and corporates accept moderately negative account remuneration to avoid the cash holding and handling costs.
Hence, zero is not the theoretical minimum at which rates can be lowered without triggering cash hoarding and large-scale conversion of account holding into cash engendering a complete breakdown of the transmission mechanism, which we define as the technical lower bound, but also because there are several other considerations related to the effectiveness of the transmission mechanism and to financial stability which explain why a central bank may not wish to lower rates to the theoretical minimum possible. This defines an effective lower bound as the minimum level at which the main policy rate can be cut without engendering adverse consequences that may not be neutralized via additional policy tools and that may outweigh the expected benefits of the policy rate cut. Our proposed conceptual framework is visualized by Figure 2.
There is a technical lower bound, as the theoretical minimum at which policy rates can be lowered. It may be above or below zero based on several considerations related to the relative convenience of cash versus electronic money, the regulatory premium attached to retail deposits, and the perceived soundness of the banking system. However, in light of considerations related to the effectiveness of the transmission mechanism, and to financial stability considerations, a central bank may refrain from lowering the rates to that technical minimum and consider a different and higher level as the minimum at which rates can be effectively lowered. If the availability of alternative, nonconventional monetary policy tools is taken into consideration, which at a certain point can be considered relatively more effective, the effective lower bound can be set at a higher level. On the other hand, the availability of risk-mitigating policy tools to address some of the negative consequences of low policy rates (for example tiered remuneration of excess reserves) may push the effective lower bound lower. We will now look more closely at the factors from which the effective lower bound depends.
4. Impairment in the Transmission Mechanism
The first group of factors concerns the possible impairment in the transmission mechanism at low policy rates.
The impairment in the transmission mechanism is one of the main reasons central banks may refrain from lowering policy rates further. The lower effectiveness of the traditional monetary policy channel can derive from rigidities and non-linearities, but may also take in euroized economies a peculiar form, which is a function of the exchange rate. At already low levels, lower policy rates may not be transmitted to lending rates. Lower policy rates are in general transmitted to the real economy primarily via lower lending rates that should ultimately stimulate credit demand, credit growth, consumption, investment, and aggregate demand. They are also transmitted via lower deposit rates that, ceteris paribus, should encourage higher present consumption and lower savings. However, if financial intermediation is impaired, banks may not be able to lower deposit rates to prevent deposit outflows. Furthermore, if funding costs are rigid at the zero bound, banks may not react to lower policy rates with lower lending rates as they attempt to maintain their lending margin unchanged with a view to preserving profitability.
In case a negative deposit rate, banks may try to compensate for the increase in the cost of retail deposits by increasing their lending rates. This effect reportedly took place in a few countries in which policy rates were lowered below zero. It may, however, also take place when retail deposit rates are approaching the zero bound. In these cases, further monetary policy easing may either result in compression in interest rate margin or is hardly transmitted to lower lending rates. This is well documented.
However, in partially euroized economies, like Albania, the impairment in the transmission mechanism can take a peculiar form, as evidenced in Figure 3, providing a stylized description of the bank lending and exchange rate channels of monetary transmission.
The bank lending rate channel is activated through the impact of a monetary policy decision on the cost of funding for the banking system. Changes in the cost of funding are then transmitted to bank lending rates, which, in turn, will affect the spending decisions of households and businesses, and finally affect domestic demand, the labor market, and inflation. Bank lending rates are also driven by a variety of risk premia, among which one of the most prominent is that related to the probability that the borrowers default totally or partially on their obligations. Such default probability is usually influenced by cyclical conditions and tends to grow during recessions and decrease during expansions; increases in the probability of default tends to translate to the fraction of nonperforming loans (NPL) in total lending, which can be used as a proxy for the probability of defaulting.
The exchange rate channel is activated through the change that a policy decision implies in the relative returns of domestic and foreign denominated assets. This implies movements in the exchange rate. The exchange rate channel usually impacts on inflation both directly and indirectly. The direct impact stems from the effect of the exchange rate movement on the price of imported goods; the indirect impact is associated to gains/losses in foreign competitiveness, which affect real exports and real GDP.
In ideal conditions, the transmission of monetary policy decisions through these channels has a predictable and consistent impact.
In a euroized economy, the two channels of transmission would tend to interact negatively. An exchange rate movement would affect the value of loans denominated in euro and have an impact on the ability of the borrowers to repay their euro-denominated debts, which would, in turn, imply a change in their probability of default, and in the fraction of NPLs.
With the transmission mechanism depicted in Figure 3, the impact of a monetary policy decision becomes difficult to predict even in qualitative terms. The exchange rate channel, while still involving its direct and indirect effects on real GDP and inflation, would now have an additional effect on NPLs, and hence on bank lending rates, which would tend to counterbalance the normal impact of the monetary policy decision. In an extreme case, and assuming that the exchange rate depreciates following a monetary policy easing, NPLs could be so adversely affected that the resulting transmission to bank lending rates could totally impair the functioning of the bank lending rate channel via higher rates and lower credit provision, and possibly offset the positive effects of the depreciation on inflation, stemming from higher competitiveness and higher import prices.
The transmission of monetary policy is further complicated by difficult-to-predict non-linearities. A reduction in policy rates may engender an exchange rate depreciation only when the interest rate differential drops below a certain threshold; likewise, only large exchange rate depreciations may have a sizeable effect on NPLs; a spike in NPLs may impair the banking lending channel only when they exceed certain levels eroding the banks’ capital ratios.
5. Financial Stability
Regardless of any financial stability mandate of the central bank and although the extent in which monetary policy should take into account financial stability considerations in its reaction function is an issue not entirely settled yet, it is broadly agreed that (1) financial stability risks – when they materialize – engender a significant and long-lasting output loss; (2) financial stability risks can be heightened when monetary policy enters the relatively uncharted waters of low policy rates; and (3) monetary policy is a relatively blunt tool to deal with financial stability risks. There are extra tools that may in principle better address and mitigate these risks. In this context, we have identified five risks engendered by low policy rates in the Albanian context. I will focus in my presentation on just a few of them.
5.1 Policy rates, exchange rate, and NPLs
We have discussed already the effect that any exchange rate depreciation may have on NPLs. This is confirmed by the Albanian experience with the exchange rate depreciation in 2009, as evidenced in Figure 4.
Allowing for the lags and non-linearities present, this depreciation was one of the major drivers in the subsequent increase in the risk premia, as evidenced by the clearly higher incidence of default in the portfolio of FX loans vis-à-vis the portfolio of lek loans. Stress test analysis at the Bank of Albania indicates that a 10 percent depreciation adds 3.1 percentage points to NPLs for foreign currency loans in the first year of the exercise and 2.1 percentage points in the second year. Higher NPLs do not matter only for the potential impairment of the transmission mechanism, but also from a financial stability perspective.
5.2 Euroization
Low lek interest could also encourage more euroization, everything else being equal. Demand of euro deposits is in part driven by optimal portfolio choice of risk-averse investors interested in the risk-adjusted return of their portfolios and in euro deposits as hedge against adverse foreign exchange shocks. Lower domestic interest rates by reducing ceteris paribus the relative attractiveness of domestic deposits vis-à-vis foreign exchange deposits, increase euroization, and accentuate in the long-term the financial vulnerabilities arising from higher levels of euroization. Hysteresis effects from past periods of instability mean that foreign currency deposits are considered a better safeguard against adverse economic shocks. Domestic deposits should therefore provide a yield premium to make depositors indifferent between the two.
Analysis of past data shows a significant correlation between interest rate differentials and the share of lek deposits. Except during brief crisis-related episodes, the interest rate differential remains an important explanatory variable of the currency composition of new deposit flows. This is evidenced by Figure 5, showing a strong correlation between interest rate differential and the currency composition of new deposits.
Rigidities, non-linearities at the zero lower bound accentuate the problem due to negative ECB policy rates. If foreign currency deposit rates cannot fall below zero, the BoA may consider it undesirable to lower policy rates beyond the point at which the bank rates and the domestic assets nominal yields fall to a level no longer attractive compared with the yields on foreign currency deposits, which will unlikely fall below zero regardless of the foreign policy rates. If leaning against and reducing euroization is a policy objective, the preservation of the interest rate differential necessary to prevent any large-scale shift becomes a more compelling policy constraint.
5.3 Funding composition
A deteriorating funding structure would also be expected to impact the bank risk-taking behavior. Another side effect of the low interest environment is a “thinning of the middle” in the maturity structure of the banks’ funding structure. As interest rates in “traditional” savings instruments (bank deposits with a maturity between one and 12 months) go down, savers have moved either into current accounts and demand deposits, in search of liquidity, or into longer tenured deposits, in search of yields. This shift is depicted in Panel A of Figure 6. This movement has not induced though: (1) a shift into more marketable or liquid instruments on the asset side of the banking system, such as government papers (Figure Panel B); (2) a shift into shorter-term government papers (Figure Panel C), or (3) a shift into credit of lower duration (Figure Panel D).
5.4 Other factors
In addition to financial stability and the effectiveness of the transmission mechanism, central banks may want to consider additional factors in their estimation of the effective lower bound. Such considerations would derive from other objectives the central banks are legally mandated to pursue. Alternatively, these considerations might be derived from risk factors that would indirectly and over long-term horizons affect economic or financial stability, or might impede on the ability of the central banks to pursue their primary objectives. We identify, here, two possible risk factors, such as the effect on savings, which in developing economies, starting from a lower capital stock, can be important to preserve the incentive for gradual fixed-capital formation, and any effect on credit allocation.
6. The Monetary Policy Barometer
In order to monitor in a comprehensive manner the effect of the factors affecting the estimation of the effective lower policy rate bound, the BoA developed the “barometer,” a monitoring tool, in 2016. The tool shown in Figure 7 represents a way to operationalize the monitoring of the dimensions described and how they affect the effective lower policy rate bound.
The barometer is used as a signaling tool for when the transmission of lowering interest rates is weakening and/or is producing undesired effects with regard to financial stability concerns. Several indicators are observed. Their performance is compared against a benchmark measured as their respective average level and/or standard deviation over the past two years. Deviations of the current value from the mean are observed to obtain warning signals. The values of each indicator are normalized in a scale of 0 to 1, where these numbers correspond to the lowest and highest values of the indicator over the past two years. The mean of the series indicates the midpoint of this interval. Hence, current values approaching the upper limit (1) signal worsening of the performance of that indicator. The opposite holds if current values are approaching the lower limit of (0). In order to have a simple visual interpretation, values falling in the red area represent a signal for warning, while those falling in the green area represent safe conditions. The barometer represents an effective tool to monitor the different relevant indicators comprehensively. It signals the extent to which undesirable effects are materializing and the relative strength of these effects. It facilitates and guides policy judgment, but it does not replace it. The indicators to use in the barometer, their relative importance, their number, and the scale to use for each indicator can obviously be adjusted.
7. Conclusions
The technical lower bound can be quantitatively estimated on the basis of the currency carry costs relative to the account holding costs. The effective lower bound does not coincide with the technical lower bound. There are several reasons central banks may not wish to lower rates to the theoretical minimum.
The effective lower bound depends not only on the structural features of the economy and of the financial system but also on how central banks interpret the trade-off between the effects of different available options and the relative merits of different policy tools.
The effective lower bound may change over time, not only because economic agents adapt, but also because policy options vary and the case for low policy rates may become more compelling after alternative non-conventional tools have been used.
The effective lower bound varies from country to country, and it requires policy judgment aided by a well-established monitoring framework. In the case of Albania, from a theoretical perspective, the single, most significant risk to which the Albanian economy and the financial system are exposed is a significant, abrupt depreciation of the exchange rate, impairing the debt servicing capability and the creditworthiness of unhedged borrowers. Such depreciation may result from a lower interest rate differential between foreign currency and domestic assets altering their relative attractiveness. Since the BoA entered the easing cycle, not only has such depreciation not taken place, but the exchange rate has appreciated. This would indicate that there is room for further policy easing as needed.
Despite the exchange rate stability, the BoA monitoring system provided initial evidences of disintermediation, weakening in the funding profile of banks with a shift toward shorter-dated funding, higher demand of large-denomination lek banknotes as store of value, and lower, though still healthy, banks’ profitability. At the same time, the decreased interest differential increased the risk of fueling large-scale portfolio shifts and subsequent large-scale depreciation, while the relatively persisting higher euroization levels increased awareness of the need to lean against euroization. At the time in which the persistent downside risks to baseline scenario of the gradual recovery in inflation rate could not exclude the need of further accommodative measures, the analysis facilitated by the BoA’s barometer led to the conclusion that the continuous effectiveness of the transmission mechanism and the overall stability of the exchange rate provided room for additional policy rate cuts, in case they had been needed. However, the indicators of stress and the lack of some of the traditional transmission channels of very low policy rates – such as higher risk -taking toward alternative investment instruments – indicated that such room was overall limited, that the room should have been used with caution, and that the possible use of non-conventional policy tools could have become a more compelling option once the limited room had been used.
Guido Della Valle, Erald Themeli, Romain Veyrune, Ezequiel Cabezon, and Shaoyu Guo, 2018, “The effective lower bound for the policy rate in euroized economies: an application to the case of Albania,” International Monetary Fund, Working Paper, forthcoming..
