Inflation surprises and inflation expectations in the Euro area

ABSTRACT

In the second half of 2012, euro area inflation started declining and reached historical lows at the end of 2014. Market-based measures of inflation expectations also declined to unprecedented levels. During this disinflationary period, inflation releases have often surprised analysts on the downside. We provide evidence that inflation ‘surprises’ have significant effects on inflation expectations. The sensitivity of inflation expectations to the surprises, which has varied over time, disappeared after the introduction of the Asset Purchase Programme by the European Central Bank.

KEYWORDS:

• Inflation
• inflation expectations
• inflation swap rates
• monetary policy

JEL CLASSIFICATION:

• D84
• E37
• E31
• E58

Disclosure statement

No potential conflict of interest was reported by the authors.

Notes

¹ A lower value had only been reached in July 2009 (−0.6%), as the result of rapidly falling oil prices in late 2008.

² See, for instance, Stella and Stock (2015), Coibion and Gorodnichenko (2015) and Bobeica and Jarociński (2017).

³ The markets’ reaction to the April surprise was more muted since in that case the shock was less broad-based, being mainly due to unexpected seasonal changes in the prices of service in Germany.

⁴ The results are identical if we use the mean of the survey; the difference between the mean and the median is at most 0.1 percentage points.

⁵ Contracts are based on the flash release of the HICPxT, consistently with our inflation surprise measure.

⁶ To illustrate this point, consider for instance the case of a lower than expected value of inflation in month t: if this is interpreted by investors as determined by particular instances that are going to be reversed, i.e. the price level is expected to increase with a higher pace (compared with before the release) over the following months of the contract, then inflation swap rates should not be revised.

⁷ For instance, a forward inflation swap rate one-year one-year ahead in month t defines the expected inflation rate between t + 9 and t + 21.

⁸ The President of the ECB Draghi referred to these inflation expectations at the Jackson Hole Symposium in August 2014. Since then, they have attracted a lot of attention in particular during the period of decline between mid-2014 and June 2016; over this period, long-term inflation expectations declined by 0.8 p.p., to 1.3%, their historical minimum since mid-2004.

⁹ Data on oil prices in euro are taken from the ECB Statistical Data Warehouse (available at http://sdw.ecb.europa.eu/). The code is RTD.M.S0.N.P_OILBR.E. The €-coin indicator is available at https://eurocoin.cepr.org/.

¹⁰ We use the conventional 5 per cent confidence level to assess the statistical significance.

¹¹ The results are very similar if we use a three-year (36 months) window. In choosing the size of the window we need to strike a balance between allowing the estimation to detect changes in the coefficients, which would call for short windows, and the efficiency of the estimates, which would call for a longer window.

¹² Moreover, in our experiments with this methodology the number of break points is in most cases equal to the maximum number of breaks allowed for in the computation of the test. In light of these results, we prefer to assess the stability of the impact of the inflation surprises on inflation swap rates by estimating the baseline equation in (2) using a moving window of fixed size.

¹³ The coefficient on the inflation surprise for the one-year four-year ahead inflation expectations is statistically significant after mid-2013. However, we do not want to emphasize this result given that for all the other maturities the coefficients are not significant.

¹⁴ For example, for the one-year maturity, the impact of the surprise on swap rates is equal to 0.36 when a five-day window is used and to, respectively, 0.24 and 0.25 with a three- or two-day window. With a ten-day window the coefficient is 0.41. For the one-year one-year swap rates, a five-day window yields a coefficient of 0.17, whereas with a three- or two-day window the coefficient is, respectively, 0.12 and 0.08. In the case of a ten-day window the coefficient is equal to 0.15.

¹⁵ The index is computed daily by aggregating the difference between the release of macroeconomic variables and market analysts’ expectations – as polled by Bloomberg – for several macroeconomic variables, with the weight of each ‘surprise’ decaying over time.

¹⁶ The results are available upon requests.