Estimating The State of Labor Markets (Unemployment Statistics vs. Economic Realities)

Here is my elaboration on the work I mentioned here.

Let’s begin by downloading historical monthly official unemployment rates according to the St Louis Fed website. We’ll be using the series UNRATE, M0892AUSM156SNBR, M0892BUSM156SNBR, and M0892CUSM156NNBR. The first thing we want to do is turn these into a single, monthly series. To begin with, we will take M0892AUSM156SNBR for granted in the early part of the data. Subsequently, when it overlaps with M0892BUSM156SNBR, we will favor that latter series. During the first month of the latter, the two series agree, but they diverge thereafter. It’s not clear why, but M0892BUSM156SNBR seems more realistic. Finally, we have the issue that no seasonally adjusted series covers the year 1947. M0892CUSM156NNBR covers that year, but is full of seasonal noise. Fortunately, M0892CUSM156NNBR overlaps with UNRATE, so it is possible to estimate a seasonal cycle in M0892CUSM156NNBR and remove it. By removing the estimated average monthly differences, I can bridge the gap between the series. Now, using the series GPDIA and GDP data, I can find out what portion of spending in each year, back to 1929, was investment. This is significant because during normal times, there seems to be an inverse relationship between changes in the unemployment rate, and changes in the “investment rate.” As it turns out, this works fairly well through the entire period, except for one glaring discrepancy:


Four years, from 1942-1945 inclusive, fall off the general pattern of this relationship. Why? Simple. During World War II, the forcible removal by conscription of ten million able bodied young men (I assume they would have been young men?) from the labor force into the armed forces, the volunteering of many more who wished to avoid the high likelihood of assignment to infantry associated with the draft, the movement of people into draft exempt war production industries, all naturally led to a reduction in the official unemployment rate. To quote Robert Higgs:

Between 1940 and 1944 unemployment fell by either 7.45 million (official measure) or 4.62 million (Darby measure), while the armed forces increased by 10.87 million. Even if one views eliminating civilian unemployment as tantamount to producing prosperity, one must recognize that placing either 146 or 235 persons (depending on the unemployment concept used) in the armed forces to gain a reduction of 100 persons in civilian unemployment was a grotesque way to achieve prosperity, even if a job were a job.
So, clearly, the unemployment rate during those years does not accurately reflect the actual state of the economy and the ability of people to get a job-it was easier for those not fighting in the war to get jobs only because the supply of labor was artificially contracted. Based on the above relationship, the level of private investment at the time would have supported an unemployment rate of about 20%. But that relationship is a little noisy, I mainly mention it because it fairly starkly illustrates the unreliability of the unemployment rate during that period. I used such relationships with investment and the the private economy’s deviation from long term trend, to estimate, with a multiplicative adjustment, and an additive adjustment, what the unemployment rate would have looked like had there not been so many people simply removed from the labor force. With respect to the investment rate, I appear to have made a conservative (in the sense of err on the low side for an upward adjustment) estimate. I only menti0n this because I cannot, unfortunately, remember exactly how I arrived at the adjustment? I realize this is the sort of comment to invite a great deal of skepticism. I can only comment that I recall that I had some method at the time, and there are defensible methods whereby I know I could arrive at an even stronger answer. I was reconstructing work I did a couple of years ago in the course of writing this post. If you’re disappointed that I couldn’t reconstruct the origin of this adjustment, your disappointment cannot match my own. I think it was based on the private output gap? Either way, you can trust me or not on my final numbers for this period, I’ll leave that up to you. Now, I end up with a monthly record like this:
But…Something kind of extraordinary is going on at the end there. During the alleged recovery from the recent recession, the labor force and the working age population, between the ages of 16 and 64, have diverged in a noticeable way:
We see that, in spite of the growth of the working age population, the labor force began to stagnate in 2008. If we assume an alternate scenario, in which the labor force, but not the number of people actually employed, had continued to grow from 2008 on at the 2000-2007 linear trend, the following happens to the unemployment rate:
I can go further, though, and estimate annual unemployment rates for the US back to 1790. Using my estimates of the private economy and it’s deviation from it’s long term trend, accounting for mismeasured inflation during WWII, based on the relationship between the two:
I chose a quadratic curve to prevent the formula from outputting negative values, and the point at which the relationship would output values going in the “wrong” direction, never occurred in the historical private output gap data. This allows me to extend unemployment estimates back to 1790:
It is on this basis that I find that the unemployment rate was lower, significantly so, during the Classical Gold Standard, than it has been recently, under the Federal Reserve.
Ceterum censeo Subsidium Foederati esse delendam.

Comments Off on Estimating The State of Labor Markets (Unemployment Statistics vs. Economic Realities)

Filed under Uncategorized

Comments are closed.