In the last 15 years, monetary policy in the euro area, as elsewhere, has developed to include a variety of instruments beyond the steering of short-term interest rates. This multi-instrument monetary policy has allowed central banks to control the position of the yield curve as well as its slope and to influence general financing conditions. New instruments designed by the ECB include forward guidance, asset purchase programmes, and targeted loans to banks.
When discussing exit from monetary easing today, the ECB has to decide not only on the timing and speed of exit, but also on the sequence. Indeed, it has been announced that a withdrawal from asset purchases will go first and interest rate increases will follow, which will have the effect of steepening the safe yield curve.
Although this strategy has the desirable consequence of favouring the profitability of banks and insurance companies, it will require a relatively rapid exit from the asset purchase programme in order to avoid a prolonged – and in current circumstances – unjustified situation in which deposit rates are negative. A fast exit, on the other hand, carries the danger of putting upward pressure on the yields of government bonds issued by more indebted countries, creating tensions in those markets. These tensions could jeopardise the effectiveness of interest rate policy by causing a fragmentation of financial markets and impairing the transmission mechanism of interest rate policy.
In this column, we report selected results from our recent paper (Reichlin et al. 2021b) and show that the desired effects of monetary policy can be obtained in the euro area only if changes in policy are not coupled with changes in sovereign risk premia.
We obtain impulse response functions from unanticipated changes at the short end of the yield curve from vector autoregression (VAR) analysis and compare different experiments.
In all experiments, we identify unexpected changes in monetary policy by using the external instrumental variables (IV) method of Stock and Watson (2008). As instruments for conventional monetary policy, we use the target and the timing surprises constructed by Altavilla et al. (2019).
Monetary policy surprises
The surprises are constructed from high-frequency event studies that record price changes in interest rates before and after the announcement of a monetary decision by the central bank. Assuming that markets incorporate information about economic fundamentals efficiently prior to the announcement, only unanticipated changes in policy should affect prices. We utilise seven overnight index swap (OIS) rates at varying maturities: 1 month, 3 months, 6 months, 1 year, 2 years, 5 years and 10 years. The reason for choosing the OIS is that it is free of any credit risk as the contract does not involve the exchange of principal and does not entail any credit risk. Hence, the swap rate reflects only expectations about the policy rate at the maturity of the forward agreement.
Altavilla et al. (2019b) compute four factors that summarise high-frequency surprises on the entire maturity curve: a target factor, a timing factor, a forward guidance (FWG) factor, and a quantitative easing (QE) factor. Here we will focus on the first two – target and timing – since they capture unexpected changes in the short rate. The target factor loads predominantly on the short-term rate and is extracted from a narrow window around the ECB press releases. The other three factors are extracted from price revisions to the yield curve during the ECB press conference. The timing factor has higher loadings at the shorter maturities; the forward guidance factor has higher loadings on 1- to 2-year maturity rates; the QE factor mostly captures variation at the long end of the yield curve (10-year Treasury rates). The factors are orthogonal.
Figure 1 reports the sum of target and timing surprises along with dates of key policy decisions.
Figure 1 Time series plot of short-term surprises
The VAR model is estimated on several monthly real, nominal, and financial variables for the sample 2001–2019. The output measure is proxied by an interpolated series of real GDP, while the GDP deflator is the indicator of prices. We also include the nominal effective exchange rate (NEER), the trade balance, the stock market index STOXX 50, corporate spreads for financial and nonfinancial corporations, the German one-year rate, oil prices, the VIX, and an index of financial stress.
The VAR is estimated using Bayesian methods. In particular, we use a normal inverse Wishart prior and optimise the tightness parameter following the approach of Giannone et al. (2015).
Results from three experiments
We report here results of median impulse responses from an interest rate tightening of 100 basis points of the 1-year German rate and for three experiments. These are all illustrated in Figure 2.
Figure 2 Impact of a positive surprise on short-term interest rates
Experiment 1: Impulse response functions of target + timing surprises
This is the baseline exercise using target plus timing surprises as instruments. The shock is interpreted as an unexpected change in the short interest rate. The impulses are depicted in Figure 2 by the dashed line and labelled “original instrument”.
The effect of this shock is counter-intuitive, with both output and inflation increasing as a result of the tightening. The effect on stock prices, which is large and positive, is also counter-intuitive.
This suggests that the instrument is not appropriate. We move then on the second experiment.
Experiment 2: Impulse response functions using informationally robust instruments
As first pointed out by Romer and Romer (2000), interest rate surprises may convey not only information on monetary policy, but also information on economic development. If this is the case, they cannot be used as valid instruments since the exogeneity assumption is violated. To cope with this problem, we use the approach suggested by Miranda-Agrippino and Ricco (2020) for creating informationally robust instruments. This consists of running a regression of the target and timing factors on information about economic fundamentals that is either known to the markets or is disseminated by the central bank.
In the reported results, we use as information on economic fundamentals the commercial forecasts from Reuters’ polls, which are monthly surveys of forecasters for the variables included in the ECB projections. The polls also include quarterly forecasts on the MRO rate at a horizon of up to three quarters ahead.
The impulses are depicted in Figure 2 by the blue line and labelled “IR target+timing (Rpolls)”.
With the correction, the output effect becomes negative for the first few months and then remains persistently positive, although the effect is small and insignificant. The effect on inflation is negative but also small. The counter-intuitive positive effect on output and inflation of experiment 1 is now corrected. This is an indication that a tightening surprise was also read by markets as a positive surprise regarding the state of the economy. Once that effect is cleaned away, the response of output and inflation changes. The puzzle on the response of the stock market, however, remains.1
We move then to our third experiment.
Experiment 3: Impulse responses using robust instruments plus conditioning for movements in risk premia
In this experiment, we apply an additional correction by cleaning the monetary policy surprise from movements in sovereign spreads, implying a change in the periphery’s yield in the opposite direction to the change in monetary policy. A feature of monetary union is that, in bad times, there is a flight to safety towards German bonds and away from the periphery countries' government bond markets. In times of financial stress, a tightening may be associated with an increase in sovereign spreads and, as a consequence, in geographical differences in financial conditions. This can impair the transmission mechanism of monetary policy to output, inflation, and the stock market. Eliminating episodes in which spread surprises co-move negatively with the target and timing surprises means we should expect a stronger effect on output, inflation, and the stock market.
The spread correction is done by eliminating those months from the instrument where the spread surprise co-moves negatively with the informationally robust instrument and the spread surprise is more than one standard deviation away from the mean. The impulses are depicted by the red line and labelled “IR target+timing (Rpolls)+spread correct”.
Figure 2 shows that the effect of the correction is sizeable for output, inflation (which now have the expected sign), and, in particular, stock prices, whose response becomes insignificant rather than counter-intuitively positive. The responses of the VIX and the index of financial stress are also small and insignificant. This is to be expected since we eliminated episodes of sovereign tensions from the sample. The movements in corporate spreads are now less pronounced or insignificant.
Note that, in all experiments, an interest rate tightening results in a strong appreciation of the euro.
Implications for monetary policy
Monetary policy in all jurisdictions is about steering the yield curve via a variety of tools. In the euro area, the ECB faces an extra dimension to monetary policy since the policies which affect the ‘common’ risk-free yield curve (typically proxied by the OIS curve) also affect risk premia associated with country-specific yield curves (countries face their own default risks).
During periods of macroeconomic and financial stress, the monetary union experiences the double phenomenon of a flight to safety (i.e. a flight to the German bund, in particular, by foreign investors) and home bias in sovereign purchases causing financial market fragmentation along geographical lines.
To ensure the smooth transmission of monetary policy throughout the Union and to control the two dimensions of monetary policy, the ECB needs to calibrate interest rate policy, forward guidance, and asset purchases so as to steer the common ‘risk-free’ yield curve while at the same time preventing spikes in interest rates in some jurisdictions (see Reichlin et al 2021a and Reichlin 2021 for a discussion).
This is controversial since it can lead to moral hazard. For this reason, we do not support the idea that the ECB should always and unconditionally choose policies aimed at compressing cross-country spreads. Indeed, to the extent that spreads may in part reflect financial frictions and self-fulfilling dynamics, they also reflect differences in bond default probabilities. The liquidity and solvency attributes, however, are difficult to separate, which creates a problem in defining a policy target for the ECB based on the ‘risk-free’ yield curve only.
These considerations have inspired the response of the ECB to the Covid crisis. In these circumstances, the ECB implemented the PEPP programme, which allowed purchases in different proportions to the capital keys.
In designing tightening tools today, the ECB needs to maintain that flexibility and allow targeted purchases if necessary. For this, it needs to design a new instrument following the model of the PEPP. Having that instrument will make it possible to exit the asset purchase programme and then increase short-term interest rates without too much disruption.
Without this, as our empirical exercise shows, risk premia will impair the effectiveness of tightening on inflation, output, and the stock market.
Altavilla, C, L Brugnolini, R S Gürkaynak, R Motto and G Ragusa (2019a), “The euro area monetary policy event-study database”, VoxEU.org, 3 October.
Altavilla, C, L Brugnolini, R S Gürkaynak, R Motto and G Ragusa (2019b), “Measuring euro area monetary policy”, Journal of Monetary Economics 108: 162-179.
Giannone, D, M Lenza and G E Primiceri (2015), “Prior selection for vector autoregressions”, Review of Economics and Statistics 97(2): 436-451.
Jarociński, M and P Karadi (2020a), “The transmission of policy and economic news in the announcements of the US Federal Reserve”, VoxEU.org, 3 October.
Jarociński, M and P Karadi (2020b), “Deconstructing monetary policy surprises—the role of information shocks”, American Economic Journal: Macroeconomics 12(2): 1-43.
Miranda-Agrippino, S and G Ricco (2021), “The transmission of monetary policy shocks”, American Economic Journal: Macroeconomics 13(3): 74-107.
Reichlin, L (2021a), “Non-standard monetary policy instruments: effectiveness and risks”, in Central banks in a shifting world, conference proceedings of the ECB Forum on Central Banking, Sintra, 11-12 November.
Reichlin, L, G Ricco and A Tuteja (2021b) “Monetary Policy Signals and Shocks in the Euro Area”, mimeo, London Business School.
Reichlin, L, K Adam, W J McKibbin, M McMahon, R Reis, G Ricco and B Weder di Mauro (2021), The ECB Strategy: The 2021 Review and its Future, CEPR Press.
Romer, C D and D H Romer (2000), “Federal Reserve information and the behavior of interest rates”, American Economic Review 90(3): 429-457.
Stock, J H and M W Watson (2008), What’s new in econometrics: Time series, Lecture 7: Structural vars, National Institute for Economic Research.
Data sources and descriptions
1 Jarocinski and Karadi (2020) remove the stock market puzzle by using an information correction which differs from ours. Our conjecture is that their information correction removes both information and risk premium shocks. By examining the two corrections separately, we can assess the importance of the risk premia effect (see also Jarocinski and Karadi 2018 for the US case).