Wages and prices of foreign goods in the inflationary process in Iceland

In this paper we discuss the relationships between the CPI in Iceland, the unit labour cost, and the price of foreign goods, and their role in equations for forecasting inflation. We find that the logs of these variables are cointegrated, the cointegrating vectors are stable over many different data periods, and the coefficients satisfy the homogeneity condition. On the other hand, the coefficients in regressions of log difference of the CPI on log differences of the other variables, and a constant, are unstable, and for data for the last two decades, the homogeneity condition is always rejected. The coefficient for changes in unit labour cost, the price of the most important cost item, is often insignificant, while the constant, which shouldn‘t be in the equation, is frequently highly significant. It is shown that the estimates of coefficients in the equation in log differences of the variables depend on the coefficients of correlations between the variables, and their standard deviations, which have diverged very much since the turn of the century. Large standard deviations of changes in unit labour cost, and especially of changes in the price of foreign goods, compared to standard deviations of changes in the CPI, contribute to lower coefficient estimates, and to the significance of the constant. In the paper we discuss how the long-run, cointegrated, relationship between the logs of the variables can be used to obtain valuable information for forecasting the rate of inflation. We also present estimation of an equation for the log difference in CPI where the error-correction term is the estimated error of an AR-equation for the errors from the equation in logs.