Seasonal Adjustment of Israeli Consumer Price Inflation
The present work applies the structural method for seasonal analysis, presented by Harvey (1993) among others, to estimate seasonal factors of monthly inflation in Israel during the years 1992-2010. One advantage of this method-compared with the X-12 method as it is employed by Israel's CBS (Central Bureau of Statistics)-is that it enables us to integrate external information (such as exchange rate devaluation and world price inflation) in the estimation; this helps to extract the time varying seasonal factors. Another advantage is that the definition of seasonality under the structural approach is more clearly formulated and therefore contributes to a better identification of the seasonal factors. The seasonal factors estimated by the structural approach are less volatile than those of the CBS, and in recent years the two approaches have yielded significantly different estimates. As a result, the development of the seasonally-adjusted inflation rate is often very different between the two methods, and under the structural method it is more correlated with the actual inflation. Thus, for example, from September 2008 to February 2009, actual inflation was -2.4 percent (in annual terms). The seasonally-adjusted inflation rate based on the structural method was -0.6 percent, whereas that based on the CBS was positive, 1.4 percent - a gap of 2 percentage points, as wide as the range of the inflation target range of 1-3 percent. The above results highlight the importance of having reliable estimates for the seasonal factors for monitoring the development of monthly inflation. To compare the two methods we formulated two kinds of tests. The first one is a "forecast accuracy test", which compares the forecast performance of the inflation rate using the seasonal factors estimated by both methods. It appears that the structural method, even without using any additional information, does much better than the X-12 method in forecasting inflation. Also, employing external information in the estimation phase improves the forecasting performance of the structural method. The second test is the "correlation and bias test". It tests the possible existence of a correlation between the estimated seasonal-factors and external variables that are supposed to affect trend inflation, such as the exchange rate depreciation rate and the inflation of imported-goods prices (seasonally adjusted). We interpret such a correlation, which was found in the case of the CBS estimators, as a failure of the method to achieve a good decomposition of inflation into its seasonal and non-seasonal components. Such a result also means (as will be explained later) that the seasonally adjusted inflation rate, estimated by using the CBS estimates of seasonality, is a biased estimate of trend inflation. An interesting by-product of the structural method is the estimation of a time varying pass-through from exchange rate changes to inflation. Such estimation also enables decomposing inflation into two components: an external one, which captures the direct effects of exchange rate devaluations and global price inflation, and a domestic one. The results of such decomposition, into external and domestic components, suggest that during the period 2000-2010, although the CPI had no definite trend, the domestic component increased since 2003. Despite the above mentioned benefits of the structural method, it is important to emphasize that this work does not criticize the X-12 method, nor its implementation by the CBS. On the contrary, we think that it is important that the CBS maintains its present approach of not involving subjective considerations in the process of seasonal adjustments. In the present research we focus on a specific variable (inflation) during a specific period of time (1992-2010), for which we are equipped with information which has accumulated over time and we find worth using. The contribution of this work is not to replace the estimates of the CBS, but to offer additional, complementary estimates of the seasonal components of monthly inflation. By so doing we hope to supply an additional point of view for monitoring and assessing the evolution of monthly inflation, which is an important element for monetary-policy formulation.