In search of tea leaves.

AE Rodriguez , Michelle Brandao

The talk of a forthcoming recession is deafening. Everyone is trying to read the tea leaves. The search for a timely gauge of economic activity never wavers. Ideally, we would like a current measure that correlates nicely with economic activity. A proxy in-hand would enable us to forecast the economy. This applies to the nation as a whole as well as to the State of Connecticut.

A proxy is necessary and useful ‘cause the official measures of economic activity are dated. For instance – we write this in mid-August and the latest measure of CT Gross Domestic Product is from the first quarter of 2019.

The timelier Purchasing Managers Index (PMI) has been used to approximate economic activity at large for the US. Strictly speaking the PMI measures the growth of economic activity in the manufacturing sector. How well does the PMI correlate with the growth of GDP rate? For the US – it does fairly well. Figure 1 is a scatterplot showing a strong positive correlation (a correlation coefficient of around 0.71).

The PMI is a composite index grouping five diffusion indexes –with equal weights. It groups New Orders, Production, Employment, Supplier Deliveries, and Inventories. A PMI of greater (less) than 50 percent indicates that more (less) than half of survey respondents reported better (worse) economic conditions than they reported in the previous index. As can be seen in Figure 1, for the United States GDP Growth was more positive than negative when the PMI was above 50 percent. What does this reveal? The latest PMI we have is for July – and it remains above 50 where it has been for awhile although it has declined in the last four periods.

Can the PMI provide similar a function for Connecticut? Unfortunately no. Figure 2 is a scatterplot that displays the correlation between PMI and Growth in Connecticut GDP. Alas, the correlation ceases to be significant; the correlation coefficient is a measly 0.21.

Note that Connecticut actually publishes its own Diffusion Index. It is published by the CT Department of Labor and is known as the CT Business Sector Scorecard.

Figure 3 displays the correlation between CT GDP Growth and the Business Scorecard Diffusion Index. The coefficient is 0.23. Not a very good indicator.

Its latest datapoint is for June – and it registered a 0 – the equivalent of a 50 in terms of the PMI scale. The CT Scorecard reflects a big fat tie between those sectors moving UP and those sectors moving DOWN; the June Index had only 4 of the 6 sectors reporting. The CT index tracks Housing Permits, Exports, Gaming Slots Revenues, Air Passenger Count, the Connecticut Manufacturing Production Index, and Average Manufacturing Weekly Hours. The Scorecard used to track Major Attraction Visitors – but DOL discontinued publishing it last year. Ironically, an analysis of important predictors of CT GDP Growth shows Major Attraction Visitors as one of the most important ones. More on this on a later blog.

It is clear that one uses the Business Scorecard at one’s peril. Can we salvage something? It is possible to pick the best predictors chosen by the CT DOL and adjust the Diffusion Index to reflect the influence of only the top three most important (predictor) variables: Major Attraction Visitors, Air Passenger Count, and Average Manufacturing Weekly Hours – which are the ones that account for most of the (modest) predictive oomph present in the Scorecard. And here is the result. Figure 4 displays the relationship between Connecticut Economic Growth and the “adjusted” Scorecard. It remains a weak-ish correlation – but it’s the best we got, given this lot.

The correlation coefficient is now 0.37. Small wrinkle: the adjusted scorecard was built with data on Major Attraction Visitors. The CT DOL does not publish this anymore. Maybe we can get the DOL to continue reporting it. Otherwise – as to predicting the downturn in CT – no joy.

NOTES: Rodriguez is Professor and Chair of Economics & Business Analytics at the University of New Haven (). Brandao is a graduate student in the Sport Management – Analytics program at the University of New Haven ().

The PMI was aggregated to a quarterly statistic by averaging across quarters. We used Real Gross Domestic Product Percent Change from Preceding Period (FRED key: A191RL1Q225SBEA) and Real GDP for CT, Percent Change,Seasonally Adjusted Annual Rate (FRED key: CTRQGSP). Data period examined was first quarter of 2005 through the first quarter of 2019. The monthly series in the CT Business Scorecard were similarly aggregated to quarterly. The Business Scorecard Index was recoded using the PMI methodology. That is to say, the series that were positive from period to period were divided by those that were positive and negative combined – and multiplied by 100. The selection of the most important predictors was via random forests.


For attribution, please cite this work as

Rodriguez & Brandao (2019, Aug. 16). The Least Explanation: PMI AND GDP? THE DOL BUSINESS SCORECARD AND GDP? DO THEY CORRELATE FOR CONNECTICUT? . Retrieved from

BibTeX citation

  author = {Rodriguez, AE and Brandao, Michelle},
  url = {},
  year = {2019}