Academic journal article Economic Inquiry

The Level and Uncertainty of Inflation: Results from Oecd Forecasts

Academic journal article Economic Inquiry

The Level and Uncertainty of Inflation: Results from Oecd Forecasts

Article excerpt

BRYCE E. KANAGO [*]

There is considerable evidence that inflation variability and the level of inflation are positively related across countries. Evidence of a within-county relation is mixed. Evidence for a significant positive relation comes mostly from studies using some survey measure; contrary evidence comes mostly from studies using regression errors. Our measure of uncertainty is the squared forecast-error from Organization for Economic Cooperation and Development inflation forecasts. Most countries do not exhibit a positive and significant relation. The greatest number of positive coefficients is for relative uncertainty regressed on contemporaneous inflation. (JEL E310, E370, 1342, 1340, 1323)

I. INTRODUCTION

Although empirical studies have not reached a unanimous conclusion, most economists believe one detriment of high inflation is the greater uncertainty it generates about future inflation. Affirmative evidence of a positive relation between inflation and uncertainty is most frequently found in cross-country studies that regress some measure of variability or uncertainty for each country on their average inflation rate. Time-series regressions for individual countries provide mixed results. Those time-series studies that derive inflation uncertainty from a survey of forecasters conclude that higher inflation causes greater uncertainty. Time-series studies that derive uncertainty from ex-post regressions constructed by researchers provide most of the negative evidence. [1]

It is tempting to dismiss this negative evidence as the consequence of poorly or opportunistically constructed ex-post forecasts. In this paper we use annual inflation forecasts for Organization for Economic Cooperation and Development (OECD) countries that have been published in the OECD's Economic Outlook for about 30 years. These data circumvent the problems of constructing forecasts after the fact. For each date and each country, a single forecast is provided. By contrast, survey data, such as the Livingston data, average the forecasts of a number of respondents. To the extent that the survey sample is representative of the population, the surveys may provide a better approximation to the public's expectations. However, survey respondents have nothing to lose if their forecasts are wildly inaccurate, whereas professional forecasters put their reputation on the line. A distinct advantage of the published forecasts is that the OECD provides forecasts for 24 countries, whereas surveys are available for only a few countries. Of course, ex-post forecasts could be constructed for the same set of countries. While these ex-post forecasts could utilize modern econometrics techniques such as Autoregressive Conditional Heteroscedasticity (ARCH) and time-varying parameters, they cannot replicate the specialized human-capital, the judgment, or the particular perspective of the times that were used to construct ex-ante forecasts.

Davis and Kanago [1998] regressed the squared-error from ex-ante inflation forecasts produced and published by Business International on inflation and found no significant relation for most countries. This result makes it less tempting to dismiss the time-series evidence of ex-post inflation forecasts. We extend the results of Davis and Kanago [1998] by computing the same regression but using squared errors from ex-ante forecasts made by the OECD and published in Economic Outlook. The 24 countries in the OECD data include Canada, Iceland, Luxembourg, New Zealand, the United States, and Turkey that are not included in the Business International data. For some countries the OECD data includes a few more years than the Business International data. The results from the OECD and Business International data are broadly similar; few countries show a significant relation between inflation and uncertainty.

We further test the robustness of these results by using relative uncertainty as the dependent variable. …

Search by... Author
Show... All Results Primary Sources Peer-reviewed

Oops!

An unknown error has occurred. Please click the button below to reload the page. If the problem persists, please try again in a little while.