Monetary policy can contribute to lower interest rates and risk spreads, which stimulates risk-taking in financial markets. This is an important mechanism in the portfolio rebalancing channel of quantitative easing (QE) (Joyce et al. 2012). Since asset purchases by the central bank change the composition of investment portfolios, preferred-habitat investors will react by buying assets that are a close substitute to the assets sold. This process will raise the price of assets not purchased by the central bank as well. Figure 1 shows that in the UK, the US, and the Eurozone, stock prices rose strongly after announcements of new QE programmes.
Figure 1 Stock prices and QE for the UK, the US and the Eurozone, 2007-2016
By encouraging risk-taking, QE may – as an unintended side-effect – contribute to asset price bubbles when asset price developments are out of sync with economic fundamentals. In the longer term, the bursting of these bubbles may even make the disinflation that central banks are trying to address worse. Hence, for monetary policy it is important to know whether asset price developments indicate future (upward and downward) risks to price stability, and at what lag. Previous research has documented thoroughly the channels through which asset prices affect inflation (for an overview, see Papademos and Stark 2010 ).
In a new paper, we focus on the tail risks of inflation, and not on inflation itself (De Haan and Van den End 2017). We investigate the information content of financial variables, such as stock prices, private credit and interest rates, as signalling devices of two possible inflationary regimes: low or even negative inflation, and high inflation. These are defined as inflation below, or above, more than one standard deviation from its mean. We employ a non-parametric signalling approach, the receiver operating characteristic (ROC) curve. This method has recently become more popular among economists, because it fully maps all possible trade-offs between Type 1 errors (missed signals) and Type 2 errors (false alarms). The area under the ROC curve (AUROC) captures an indicator’s forecasting performance in a single measure (Drehmann and Juselius 2014). Our sample starts in the first quarter of 1985 and ended with the fourth quarter of 2014.
Figure 2 shows the maximum signalling value found among lead times from 8 to 0 quarters, for each variable and each of the 11 countries in our sample, if statistically significant (AUROC > 0.5). Panel 2A plots the signalling value for high inflation, panel 2B for low inflation or deflation. A comparison of the significant peak signals for high versus low inflation suggests that high credit and asset prices and low interest rates gave a stronger signal for high inflation than for low inflation. The maximum significant signals for high inflation in panel A exceeded the signals for low inflation in panel B in most cases. Nonetheless, in some countries, high asset prices were a significant indicator of low inflation as well. In the UK, Japan and the Netherlands, high credit, equity and house prices were signalling episodes of both high inflation and low or negative inflation.
Figure 2 Maximum signalling value (areas under the ROC curve) for lead time 8 to 0 quarters, by country, 1985-2014
A. Maximum signal for high inflation
B. Maximum signal for low inflation/deflation
Note: Lead times have been chosen that give maximum signals (on vertical axis). Significant signalling values extracted from high (above trend) levels of credit, house and equity prices (lower confidence bound AUROC > 0.5). Significant signalling values extracted from low (vis-à-vis mean) levels of sovereign and corporate bond yields (lower confidence bound [1 – AUROC] > 0.5). Insignificant AUROC values on vertical axis below 0.5 are not shown.
Figure 3 plots the lead times corresponding to the significant peak signals shown in Figure 2. The lead times indicate that the transmission of high credit and asset prices to episodes of low or negative inflation can be long. The credit, house price, and government bond yield signals for low inflation have a longer lead time than for high inflation. The significant leads of credit and house price signals for high inflation are mostly limited to a range of 0 to 3 quarters (panel 3A), but the leads of the significant credit and house price signals for low inflation range from 0 to 8 quarters (panel 3B).
Figure 3 Lead of maximum signal, by country, 1985-2014
A. Lead of maximum signal for high inflation
B. Lead of maximum signal for low inflation
Note: Lead times (on vertical axis) have been chosen that gives maximum signals (in several cases the lead time is zero, implying that no bar is shown). Significant signalling values extracted from high (above trend) levels of credit, house and equity prices (lower confidence bound AUROC > 0.5). Significant signalling values extracted from low levels of sovereign and corporate bond yields (lower confidence bound [1 – AUROC] > 0.
Overall, the results indicate that the signalling value of asset prices for inflation is unstable. Higher asset prices do not always signal higher inflation, but can also precede a low inflation regime. It is likely that the relatively long lead time for low inflation is related to the unwinding process of high asset prices in a burst phase.
This research suggests that monetary policy that works through the asset price channel may not contribute to price stability. Stimulating effects on asset prices may lead to booming asset markets that feed high inflation, but in the longer term may create low or negative inflation if an asset price bubble bursts. Given that QE in particular affects asset prices, it implies that this instrument can compromise the inflation objective if it leads to excessive asset price developments. The unstable relationship between asset prices and price stability does not imply that asset price developments do not contain useful information for central banks. Even when the effects of asset price increases on inflation are unclear ex ante, a central bank should be mindful of asset price movements for reasons of financial stability.
Authors’ note: The views expressed are those of the authors and do not necessarily reflect official positions of DNB.
De Haan, L and J W van den End (2017), “The signalling content of asset prices for inflation: Implications for Quantitative Easing”, Economic Systems, forthcoming. Working paper version available here.
Drehmann, M and M Juselius (2014), “Evaluating early warning indicators of banking crises: Satisfying policy requirements”, International Journal of Forecasting 30: 759-780.
Joyce, M, D Miles, A Scott and D Vayanos (2012), “Quantitative easing and other unconventional monetary policy: an introduction”, The Economic Journal 122: 271-288.
Papademos, L D and J Stark (2010), Enhancing monetary analysis, European Central Bank.