Behind the CT DOL Town Economic Indexes
Updated: Jan 10
It is commendable that the Connecticut Department of Labor folks develop and regularly publish an economic performance metric for each of the 169 towns in Connecticut. They call it the Connecticut Town Economic Index.
The Indexes are published online and via the print edition of Connecticut DOL's Economic Digest. The latest Town Economic Index series was published in October 2019 and contains annual data for the 2005-2018 period. The raw data is also available and can be downloaded.
But the project is flawed - and examining the reasons behind the shortcomings of their approach is what brings us here today.
I am going make two claims and a conclusion here; first the claims: (i) the CTEI is nothing but the unemployment rate gussied up into a wan composite index; and, (ii) the CTEI reflects very little of the real economy although it claims to channel town's economic health. The conclusion is at the end of this essay.
Admittedly, having information specific to a town - if the DOL had gotten it right - would be highly useful. It would allow an assessment of how a town is doing relative to its neighbors or relative to the State at large; it would allow fruitful insight as to how a town's economy is performing relative to itself.
What goes into the CT Town Economic Indexes and what does it presume to do? The authors of the State Economic Index claim the CTEI measures "each town or city's overall economic health." (Joo & Placzek, Economic Digest, Oct. 2019, at 1).
Alas, the CT Town Economic Index methodology does not deliver. Let's take a look behind the numbers.
The CTEI is an equally weighted average of the following four economic series:
1. total covered business establishments,
2. total covered employment,
3. inflation-adjusted covered annual average wages,
4. and the unemployment rate.
Take a look at the CTEI in action. The graph below shows the Town Economic Indexes for the 3 towns in my stomping grounds here in the New Haven region - and the comparable series for Connecticut.
It tells us - for instance - that the towns in the New Haven region appear to be on the up and up but my hometown of Hamden has faltered relatively speaking. As an aside - New Haven's performance is noteworthy; it went from being the regional laggard to the star performer.
But wait a minute; looking at the graph gives me a strange vibe - like I have seen this before. You may have seen this series too.
And, in fact, you have. And it is the first clue as to a serious problem. Here is a graph for the unemployment rate for the same towns and the state (with unemployment reversed to match what the CT DOL folks did).
Uncanny? Not really. As it turns out the Town Indexes are just the run o' the mill unemployment rates wrapped in fluff. This next visual shows the correlations via the size of the circles and the correlation coefficient. Look at the correlation coefficient between unemployment and the CTEI (-0.98) - and the big fat circle is a warning.
In fact, the correlations in the matrix above also tell us that it is probably redundant to include the employment series in addition to the unemployment rate in assembling the index; employment and unemployment are highly correlated (the associated correlation coefficient is - 0.91). This means that they contribute practically the same information to the Index.
Its not hard to see the problem. The 169 towns don’t vary much by total covered business establishment, employment, and real wages.
Sure, employment in North Stonington (2018 = 150.1) has fared better than in Barkhamsted (2018 = 53.5). And we could probably find some notable differences in business establishments. But by far the biggest differences between towns, from among these four measures, are in unemployment. Unemployment in Hartford (7 percent) was almost three times as great as Union (2.6 percent) in 2018; in fact, the difference between these two is more than two standard deviations (st dev = 1.8); in statistical distance, this is downright YUUGGEE!
To check out still another relationship between unemployment and the CTEI I mapped 2018 data,the cross-section if you will:
Which displays the linearity and a correlation of 0.661.
And here is the scatter between the two variables over the entire time period.
The correlation between these two series - by the way is -0.91.
Last, let's take a different look by examining the contribution to variance; that is to say - how much does the individual series contribute to the realized variation in the CTEI? Here is the visual:
This is using the raw data. The variation of unemployment accounts for practically half of the variation in the CTEI. Unemployment is driving the CTEI.
What appears to be happening here is that the folks who put together this index are just needlessly gussying up the unemployment series and giving it a fancy new name. Let's take a look, here are the two series over time.
My second point turns on appraising the DOL's claim that the CTEI tracks economic health. Short answer: it doesn't.
For this we take a look at the statewide measure of the same variables; the CTEI for the state at large. One would think that a measure of economic health would track real GDP closely; so we take a look at the historical relationship between the CTEI Index for the State of Connecticut and Real GDP for the State of Connecticut; and for grins we also take a look at CTEI vs nominal GDP and CTEI vs GDP per capita, for good measure. Here are all the series.
Alas, the CTEI is off on its own.
Here is a closer look at the CTEI and Real GDP. Data is from FRED (CTRGSP); I indexed the series to equal 100 in 2010 to match the CTEI indexing.
As noted above, the CTEI appears to be completely detached from the real economy.
And here as mentioned is the mapping to Nominal GNP.
Seemingly more accommodating but still appears to be so only by coincidence because the only monetary constituent series in the CTEI is real wages, not nominal wages.
I think i can summarize now: the claims as to what the CT DOL Town Economic Indexes project is providing is a tad too ambitious. The indexes are nothing else but the unemployment rate dressed up for the holidays. And perhaps more distressing it is that they are an economic measure untethered from the real economy. It contributes nothing new to our insight of the economy of the towns. In other words, it has practically zero information content.
By the way, the observations here are equally applicable to the State version of these indexes which DOL also publishes. They are applicable because the State Indexes use the same methodology. In other words the State Indexes are just a rewrite of the rankings that would result if State rankings were based on State unemployment rates alone.