Identifying Changes in Mean, Seasonality, Persistence and Volatility for G7 and Euro Area Inflation*

Abstract

We propose an iterative decomposition that tests and accounts for multiple structural breaks in the mean, seasonality, dynamics and conditional volatility, while also accounting for outliers. Considering each component separately within each iteration leads to greater flexibility compared with joint procedures. Monte Carlo analysis shows the procedure performs well. Applied to monthly CPI inflation in G7 countries and the Euro area, we uncover mean and seasonality breaks for all countries and, allowing for these, changes in persistence are generally also indicated. Further, volatility reductions are widespread in the early to mid 1980s, with some countries exhibiting increases from 1999 onwards.

What is the Globalisation of Inflation?

“…In line with results for international analyses that employ univariate inflation models (including Benati, 2008, andBataa et al, 2014), our results measured through the domestic-foreign inflation model in equation (6) indicate that aggregate inflation persistence has generally declined over this period. Note, however, that our measure of persistence is given by Moreover, close inspection of Figure 7 b, indicates that all inflation targeters (except Finland) also experience a fall in core inflation persistence during the early 1990s.…”

“…Broadly similar numbers of volatility as coefficient breaks are uncovered, with three again the maximum number detected (for aggregate and energy inflation in Japan). Our finding of fewer breaks overall than univariate studies that examine the mean level of inflation (such as Benati, 2008, andBataa et al, 2014) may be attributed to our methodology 16 that explicitly includes dynamics in the breaks analysis in order to avoid the oversizing of HAC methods, as discussed in Section 3 above.…”

“…However, Bai and Perron (2006) show that inference is improved when the dynamics are explicitly modelled, as in our approach. Secondly, our procedure simplifies the iterations in respect of volatility breaks, since the results of Bataa et al (2013, Unlike Bataa et al (2014), seasonality is not of interest in the present study and (as discussed in Section 2) all series are seasonally adjusted prior to the application of the iterative procedure. This is justified by the robustness analysis in Bataa et al (2014), which indicates that the detection of structural breaks in the mean and dynamics is not substantively affected by the method of accounting for seasonality.…”

“…The iterative methodology proposed by Bataa et al (2014) employs structural break tests in conjunction with the outlier detection and removal procedure of Stock and Watson (2003) to examine structural breaks in each of the seasonal, mean, dynamic and volatility components of univariate inflation. While based on this approach, our procedure differs in three important respects.…”

“…The vertical axis measures the estimated coefficient value, while the horizontal axis represents time. Therefore, a detected coefficient break date (given in Table A (2011) and Bataa et al (2014). It should be noted that this phenomenon apparently applies only in the very shortrun, since the summed coefficients of Figure 5 panel a indicate that shifts to lagged foreign inflation generally play a smaller, rather than greater, role over time.…”