The Aruoba-Diebold-Scotti Business Conditions Index

Last weekend I attended an excellent conference on business cycles hosted by UC Riverside (program details here). Among the many interesting presentations was an update from University of Maryland Professor Boragan Aruoba on the index of current business conditions that he developed with Professor Frank Diebold of the University of Pennsylvania and Federal Reserve economist Chiara Scotti.

The idea behind the Aruoba-Diebold-Scotti index is to use data as it arrives from various sources to update an impression of the overall level of U.S. economic activity. I earlier described a similar effort for interpreting European data by Gabriel Perez-Quiros and Maximo Camacho. Though both approaches use a state-space framework, the details are a bit different, with ADS taking the perspective that there is an underlying “daily” level of economic activity whose status evolves slowly over time. The basic idea behind either specification is that, given the previous assessment of where the economy seemed to be as of our last available data reading, and given those slowly evolving dynamics, one would have had a prediction for what the next available indicator would show. If that new datum, when reported, is more favorable than expected, the inferred level of economic activity is revised slightly upward, while if less favorable, the inferred index is revised slightly downward. The ADS index makes use of quarterly data for real GDP growth, monthly data for payroll employment, industrial production, personal income less transfer payments, and manufacturing and trade sales, and the weekly initial unemployment claims that I highlighted last week.

Here’s what the output of that inference looks like when applied to data since 1960. Values of the index below zero indicate slower than normal growth, whereas values above zero correspond to above average growth. Shaded regions on the diagram represent the dates of U.S. recessions, as determined by the NBER Business Cycle Dating Committee.


One of the things that I learned from Boragan at the UCR conference was that the Federal Reserve Bank of Philadelphia is now updating the ADS index with each new data release and is reporting those updates on a useful webpage. Thus for example, the Federal Reserve reported today that the index of industrial production fell by 1.5% between February and March, giving it a cumulative first-quarter decline of 20.0% at an annual rate.


The industrial production numbers were released at 9:30 a.m. EDT, and by 10:30, the Philadelphia Fed website had the new implied economic assessments. The new data led the model to a slightly more pessimistic assessment of where the economy stood in mid February, for example, causing a decline in the February 16 value from a previous estimate of -3.24 to a new assessment of -3.32. The downward revisions to the most recent inferences were barely noticeable.


Does the fact that the index is on its way up from the low point give us reason to believe that the worst of the downturn is now behind us? The answer is, not necessarily. A key aspect of any model like this is the presumed underlying dynamics that govern where you expect the economy to be before you receive a new observation. Those dynamics imply a significant degree of mean reversion, i.e., imply that a -5% annual growth rate for real GDP is not going to persist. That’s not so much an assumption on the part of the researchers as an implication of the data. Any process you fit to the observed data will have some degree of that mean reversion property, because the fact is, historically -5% real GDP growth rates just weren’t something we saw sustained for a long episode, at least for the sample period to which this model was fit. Given that the index is now so far below the mean of zero, the unconditional forecasts are always going to be headed back up. Perhaps a more meaningful signal of a real improvement would come when the index gets back up above -1% or so.

Although I wouldn’t want to read too much into it, I find it helpful to have this convenient quantitative summary of how each day’s economic news might alter an objective assessment of where the economy currently stands.

Originally published at Econbrowser and reproduced here with the author’s permission.