Inflation and relative price variability in the euro area: evidence from a panel threshold model

Abstract

The impact of inflation on Relative Price Variability (RPV) generates an important channel for real effects of inflation. This article provides first evidence on the empirical relation between inflation and RPV in the euro area. Stirred by the widespread use of inflation caps or target bands in monetary policy practice, we are particularly interested in threshold effects of inflation. In line with the predictions of monetary search models, our results indicate that expected inflation significantly increases RPV only if inflation is either very low (below 0.95% per annum (p.a.)) or very high (above 4.96% p.a.).

Acknowledgements

We thank Alexander Bick, Bruce Hansen and Uwe Hassler for helpful comments and suggestions. The research for this article was partly conducted while Juliane Scharff was visiting the Economic Research Centre of the Deutsche Bundesbank. Financial support by the Monetary Stability Foundation is gratefully acknowledged. This research was supported by the Deutsche Forschungsgemeinschaft through the CRC 649 ‘Economic Risk’.

Notes

¹ New Keynesian stochastic dynamic general equilibrium models support price stability as an outcome of optimal monetary policy only because inflation increases RPV beyond its efficient level, see, e.g. Woodford (2003). As Green (2005, p. 132) says, price dispersion is ‘the root of all evil’ caused by inflation.

² Similarly, Caraballo et al. (2006) show in a multi-country study that the relation between inflation and relative prices depends on the inflationary context.

³ Threshold models nest the linear case, such that they can be viewed as a first, natural step to generalize the standard inflation–RPV equations. Alternatively, Fielding and Mizen (2008) employ nonparametric methods to analyse the functional form of the inflation–RPV linkage in the US.

⁴ In 2007, these are Austria, Belgium, Finland, France, Germany, Greece, Ireland, Italy, Luxembourg, the Netherlands, Portugal, Slovenia and Spain. The HICP subcategories are food and nonalcoholic beverages; alcoholic beverages, tobacco and narcotics; clothing and footwear; housing, water, electricity, gas and other fuels; furnishings, household equipment and routine maintenance of the house; health; transport; communication; recreation and culture; education; restaurants and hotels; miscellaneous goods and services. The data are seasonally adjusted by the Census X11-method.

⁵ The Davidson–MacKinnon test computes a test of exogeneity for a fixed-effect regression estimated via instrumental variables, see Davidson and MacKinnon (1993). A rejection of the null hypothesis indicates that endogenous regressors’ effects on the estimates are meaningful, and instrumental variables techniques are required.

⁶ The country-specific inflation forecasts are shown in the Appendix. Since we found no evidence on time varying inflation uncertainty in our sample, inflation uncertainty does not appear in the RPV equation as an additional regressor, compare Neumann and von Hagen (1991).

⁷ Hansen (1999) groups the regression residuals by individual and takes the sample with size N as the empirical distribution. Since N is limited but T is large in our empirical analysis, we treat the sample as the empirical distribution to be used for bootstrapping. For the bootstrap procedure, the variable x_it and the threshold variable q_it are given, i.e. their values are fixed in repeated bootstrap samples. We take with replacement a sample of size NT from the empirical distribution and create a bootstrap sample under the null hypothesis of no threshold. This bootstrap sample is used to estimate the model under H ₀ and H ₁ and to calculate the bootstrap value of the likelihood ratio statistic. This procedure is frequently repeated – 1000 bootstrap replications in our application – and the bootstrap estimate of the asymptotic p-value under H ₀ is the percentage of draws for which the simulated likelihood ratio statistic exceeds the actual statistic. The null hypothesis of no threshold effect is rejected if the p-value is smaller than the desired significance level.

⁸ The GAUSS program underlying this analysis is based on the GAUSS code by Bruce Hansen, see http://www.ssc.wisc.edu/~bhansen/.

⁹ Inflation thresholds can be confirmed in a comprehensive sensitivity analysis for alternative specifications, including subsets of countries, alternative threshold variables and inflation measures. For brevity, these results are not presented but are available upon request.

¹⁰ In the same vein, Caglayan et al. (2008) find that intra-market price dispersion in Turkey is initially decreasing in expected inflation but is eventually increasing.