The State Coincidence Index (SCI) is an economic recession indicator. The SCI is a combined index score published by the Federal Reserve Bank of Philadelphia to approximate the health of each US state. By simply tracking how many of the 50 states are experiencing a month-over-month decrease in their SCI, we can make a model to approximate the likelihood that the aggregate US economy is also shrinking and heading into recession. Since these index scores are usually available several months prior to NBER's GDP data (and therefore ahead of NBER's declaration of an official recession), this could be useful signal to have, given typical market performance during recessions.
The below chart shows the number of US states with a month-over-month reduction in these SCI scores. The average over the dataset shows that any given month there are about 9 states with declining SCI's. The average number on the first month of a recession is about 26 states. As of September 1, 2023 there are currently 19 states with a decreasing SCI, indicating a Normal risk of upcoming recession. The chart below shows these values over time, as well as the average and +/- standard deviation bands.
This model was originally developed in a paper by Kevin Kliesen and Cassandra Marks at the Federal Reserve Bank of St. Louis.
Looking at Standard Deviations
In line with other market models covered on this site, it's interesting to look at the current data in terms of the historical norm, and standard deviations from that norm.
It should be noted that this methodology is illustrative only. Statistically, this data won't fall into a standard distribution and is bounded by the 0-50 range (as it's simply a count of US states). Which is to say, there is a zero percent chance that 51 states will ever have a negative month-over-month SCI growth rate -- at least until Puerto Rico is recognized as a state.
On average there are 8.93 states with a negative coincidence index. The standard deviation is 11.91. Again, since it's impossible for fewer than 0 states to have shrinking SCI values, the standard deviation analysis here is a bit nonsensical. (Which is to say it's impossible for a datapoint to be more than one standard deviation below the mean, since that would require a negative number of states.) That said, for modeling purposes we consider a value of 0 to represent "low" risk of upcoming recession.
Theory & Data
The Federal Reserve Bank of Philadelphia publishes State Coincident Indexes for each US state every month. These indexes are a summary statistic the represents four main variables:
- Nonfarm payroll employment
- Average hours worked in manufacturing by production workers
- Unemployment rate
- Inflation-adjusted wage and salary disbursements
The link above shares additional data on the history, methodology, and data sources used to construct the index scores. Ultimately, the SCI number for each state is representative of current economic conditions. A rising SCI is an expanding state economy, and a shrinking SCI is a shrinking state economy. Since the aggregate US economy is just the sum of all 50 state economies, this data should be indicative of overall national economic conditions.
The model charted above is very straightforward: by counting the number of states each month with a shrinking economy (i.e., with a month-over-month declining SCI value), we can forecast an upcoming (or present) recession. Indeed, as can be clearly seen in the chart, each national recession since SCI data became available in 1980 was preceded by a spike in the number of states with declining SCIs. There are very few false positives over the dataset. Only in January 2003 did over half the states have declining SCIs and the US managed to avoid a recession.
US economic recessions are officially declared by the National Bureau of Economic Research (NBER). It is commonly said that a recession is defined as two successive quarters of declining GDP, but in actuality there is no strict rule to define a recession. NBER defines recession as:
NBER's traditional definition of a recession is that it is a significant decline in economic activity that is spread across the economy and that lasts more than a few months. The committee's view is that while each of the three criteria—depth, diffusion, and duration—needs to be met individually to some degree, extreme conditions revealed by one criterion may partially offset weaker indications from another.
Additionally, NBER doesn't declare the beginning or end of a recession until 3-6 months after it has occurred, making the official declaration essentially useless as a tool to understand the current state of the economy. It is far more helpful to have other, more real-time indicators to understand how things are going.
Predictive Value of the Model
The SCI model is great at predicting national recessions - that is clearly evident from looking at the chart above. Every single US recession sees a spike in number of states with shrinking SCIs, with an average of 26 states contracting on the month the recession begins.
The real question is whether or not this information is useful in predicting future stock market returns, and the answer is no, not really. Even though we know that the stock market tends to drop during recessions, it doesn't do so in a predictable enough way to create a trading strategy. The first month of recessions do tend to have negative stock market returns, but there isn't enough data here to rigorously create a trading strategy--which makes sense, since if there were it would have long ago been exploited and arbitraged away.
The below table cites all data and sources used in constructing the charts, or otherwise referred to, on this page.
|Original Model||Are State Economic Conditions a Harbinger of a National Recession?
By Kevin L. Kliesen, Cassandra Marks. Published by The Federal Reserve Bank of St. Louis, December 28, 2022
|State Coincident Index Data||State Coincident Index Data, published by The Federal Reserve Bank of Philadelphia.
James Stock and Mark Watson developed the basic model for constructing a coincident index for the USA. Theodore Crone and Alan Clayton-Matthews adapted the basic model for the states.