The Connecticut State Economic Index
Connecticut State Economic Index Outpaces Nation in 2023
The above is the headline in the October edition of the Connecticut Economic Digest. You can find the issue here.
##The State Economic Index
The graph of the SEI index below is built with the data provided in the Digest.
The graph shows the 2022 and 2023 State Economic Index (SEI) levels as well as the year-to-year difference, for all states. Connecticut scored 100.3 in 2022 and 104 in 2023, a YTY growth rate of 3.7 percent.
And here is a clearer view of Connecticut’s performance in terms of 23-24 annual growth.
Connecticut’s performance is commendable. And worth celebrating.
But it is here where some folks start to wonder - what is the SEI and what is driving this fortuitous result?
The SEI is a composite index. This means that is a weighted sum of several economic metrics. Specifically, it is an equally weighted average of (i) the number of establishments; (ii) Employment; (iii) Real Wages; and, (iv) Unemployment Rate.
The choice of the particular variables assembled and the weighting proportions applied are the index builder’s conceit.
The intent of all composite metrics is to represent or map the performance of some latent, hidden feature of the economy that we could all understand as performance. So whether the composite index achieves its intended object is unknowable. That is why there are so many of these “performance indexes.” Because they can.
The famous Andrew Gelman takedown of the State Human Development Index comes to mind. In that piece which you can find here, Gelman takes apart the US States version of the Human Development Index - and finds that it is highly correlated with State income. As Gelman puts it: “you’re pretty much just mapping state income and giving it a fancy transformation and a fancy new name.”
With that in mind, lets take a closer look at the SEI. First up is the realization that the four variable selected are apples and oranges differing in at least one key element: variability.
The table below displays a measure of the variability of each variable. By far, unemployment is hugely variable - and not surprisingly, the biggest difference between the states is in terms of unemployment.
Another measure, the correlation between the SEI and the variables suggests that it is unemployment that may be driving these results; echoes of Gelman.
Lets dive right in, this time with a different tool, a random forest feature importance. The results are in the figure below. Not surprisingly, it shows unemployment as the key variable in the resulting rankings.
In the figure below we visually directly show the 2023 relationship between the SEI and the unemployment rate (index) used in the construction of the SEI. Most of the states fall on the 45-degree line.
Concluding Comment
It appears to be clear. to paraphrase Gelman: you’re pretty much just mapping state unemployment rate and giving it a fancy transformation and a fancy new name.
Which is a bit disappointing from the CT DOL folks. One expects a bit more in their analysis.
A.E. Rodriguez, PhD Professor Economics & Business Analytics Pompea College of Business University of New Haven
arodriguez@newhaven.edu