Academic journal article
By Levanon, Gad
Business Economics , Vol. 45, No. 1
This article evaluates which economic indicators are the most useful for signaling recessions. The article uses a modified Markov switching method to compare the timing of recession signals across many indicators. In its present form, it is difficult to use the Markov switching methods for comparing recession signals across indicators. First, the regimes in the Markov switching method do not necessarily align with recession periods. Second, the definitions of the two regimes are likely to be different across indicators. However, if some modifications are made to the Markov switching method, the method can he helpful for comparing recession signals across indicators. This article shows that by converting Markov switching probabilities into percentiles, the Markov switching method can be useful in comparing the qualify of recession signals across indicators. Using the method, hundreds of indicators are ranked based on their leading ability during different sample periods. Finally, the performance of the indicators during the current recession is evaluated.
Business Economics (2010) 45, 16 27.
Keywords: leading indicators, recessions, forecasting, business cycles, Markov switching
The forecasting and detection of turning points in the economy is one of the most studied and practiced areas in macroeconomics. Leading and coincident indicators play an important role in signaling the different phases of the business cycle. The eXisting literature provides several methods for evaluating business cycle indicators. Forecasting important measures of economic activity like GDP or forecasting recessions is one method [Rudebusch and Williams 2007]. Determining turning points in economic indicators and comparing them to the official National Bureau of Economic Research (NBER) recession dates is another method [Bry and Boschan 1971].
In this article, I suggest a new method for evaluating business cycle indicators and comparing their leading abilities. The method uses the Markov switching methodology, which is one of the most commonly used methods for estimating recession probabilities from economic indicators. In its present form, the Markov switching method is not helpful for comparing recession signals across indicators for two principal reasons. First, the regime probabilities associated with each indicator in the Markov method do not necessarily align with recession periods. Second, the definitions of regimes across different indicators are likely to be different.
To mitigate these problems, in this article the Markov switching probabilities from each indicator are first converted into percentiles, which will allow for comparisons across indicators. In order to decide what percentile qualifies as a recession signal, a threshold will then be created that is proportional to the percent of recession months out of all the months in the sample. The result is a new method for eXtracting recession signals out of indicators.
The second main contribution of this article is to use the new method to systematically compare hundreds of economic indicators and rank them by their performance as leading or coincident indicators. To the best of my knowledge, this is the first article that aims at doing that.
This article proceeds as follows: Section 1 describes the Markov switching method. Section 2 analyzes the performance of indicators during the 1959-2009 period, a larger set of indicators during the 1989-2009 period, and the indicators during the current recession. Section 3 provides some concluding comments.
1. The Method
Recessions are the most negative periods of economic activity. An optimal leading indicator provides a recession signal only before or during a recession and should not signal a recession during a period when there is not a recession, or in other words, provide a false signal. In addition, they should not fail to provide a signal when a recession does happen--a missed signal. …