Numbers Don’t Lie? 

By Floyd Norris 

The labor market is booming.  The average unemployment rate in 2023 was 3.64%, down a smidgen from the previous year and the lowest for any total year since 1969. 

But is that just because fewer people are looking for work, and therefore are not counted as unemployed?   

A way to seek that answer is to look at the Labor Force Participation Ratio (LFPR) – the proportion of working age people who are either employed or looking for work.   

The LFPR does seem to indicate there is a problem.  The 2023 average was 62.6%, better than during the Pandemic but otherwise the lowest such figure since 1977.  It was five percentage points higher in 1997. 

That statistic makes it appear that for whatever reason,  whether a lack of available jobs or a lack of people willing to work, Americans just aren’t working like they used to.  The low unemployment rate, it would seem, came because a lot of people dropped out of the labor force, not because they found jobs. 

But appearances can be misleading.  In reality, people are now more likely than ever to be working. 

How can that be true?  To a statistician, it is an example of what is called “Simpson’s Paradox,” named for Edward H. Simpson, who became interested in mathematical statistics as a young man during World War II when he worked as a cryptanalyst in Bletchley Park, where the British broke German codes. The whole can be very different from the sum of the parts – if the size of the parts changes. 

The Bureau of Labor Statistics, which reports the job and unemployment figures, counts everyone who is at least 16 years old as being of working age.  That is not the way most of us view it. We don’t expect most young people to get jobs – or at least not full-time ones – until they finish their education.  And we don’t expect most people of retirement age to continue working. 

The Bureau publishes data by age and gender, mostly for five-year age intervals. It does not adjust many of those numbers for seasonal changes, so the numbers discussed here are annual averages of the 12 unadjusted monthly figures. 

I compared the 2023 figures to those for 2006 – a year chosen because its unemployment rate was relatively low just before the Great Financial Crisis sent jobless rates up sharply.  (It was still higher than the current level.  The average unemployment rate for 2006 was 4.6%, compared to 3.6% in 2023.) 

In 2023, a greater proportion of women in all but one of the 13 age groups was in the labor force than were women in the same age groups 17 years before.  The sole exception was teenagers.  But with more and more baby boomers hitting retirement age, overall the proportion of women who wanted to work had declined by more than five percentage points. 

In 2006, the oldest baby boomers – those born in 1946 – celebrated their 60th birthdays.  In 2023, those same people turned 77.  In 2006, 17% of working age women were over 65.  Last year the figure was 23%. 

The results for men are not as stark. There was a long-term trend in which prime-age (25-54) male labor market participation gradually eroded, and that shows in these figures. There is some indication that decline may have halted in recent years, but it is not conclusive. 

The declines in male LFPR were concentrated in the groups under 45.  Those between 45 and 54 came close to breaking even, while all groups over 55 saw more participation.  

Incidents of Simpson’s Paradox are easy to miss in analyzing data, and that can lead to very misleading conclusions.  In his book “Probably Overthinking It,” published last year by the University of Chicago Press, Allen B. Downey, a professor emeritus of Computer Science at Olin College, points to one statistic that gained some unfortunate attention early in the Covid-19 Pandemic. 

A British government report in 2021 showed, in the accurate report by one journalist, that “vaccinated English adults under 60 are dying at twice the rate of unvaccinated people the same age.”  

“Among skeptics,” Downey writes, “it seemed like proof that vaccines were not just ineffective, but harmful.” 

But, Downey says, if you cut through the numbers, it turned out that was a matter of age.  At that point, virtually all vaccinated people were relatively old, and unvaccinated ones were generally young.   Older people are always more likely to die than younger ones, Pandemic or no Pandemic. 

“The excess mortality among vaccinated people in 2021 is entirely explained by Simpson’s Paradox,” Downey wrote. “We can conclude that the vaccine saved many lives.” 

There was considerable economic stimulus in this country as the Pandemic raged, including cash grants and more generous unemployment benefits. And the reported LFPR declined at the same time.  Did those benefits make people lazier, and do we therefore need to cut benefits to get people back to work? 

Simpson’s Paradox helps us to understand that was not the case.  The majority of age groups are more likely to work now than their predecessors were in years past – something you might expect in a strong economy with rising wages.  The declines in the overall labor market participation rates reflect the aging population – a trend that will continue. 

One way to understand the underlying reality of the changing labor market is to calculate what the total LFPR would have been over time if every group – men 16-19, men 20-24 and so on — had the same LFPR as it actually did in each year, but that the population share of each group had been the same in every year as it was in 2023.  

If population shares had been consistent, the total LFPR would have set a record high in 2023, not been near the level of the late 1970s.  Americans aren’t lazier than they once were. But they are older. 

To get more workers in coming years, it probably will be necessary to get more people of actual working age.  Since no new working age people are being born – you have to be at least 16 to be counted as working age – getting a larger working age population in the United States would require increased immigration. 

The economy has made many people more eager to work than were their predecessors, and more confident they can find jobs.  It has helped to push up wages of some of the occupations historically at the bottom of the ladder, to the consternation of employers who see higher wages for their workers as inflationary – not to mention as a risk to profit margins. 

The increase in government benefits during the Pandemic did not persuade people to stay home and live on the dole.  It encouraged many to get jobs, which often paid better than before. And those jobs presumably attracted more immigrants, wanted or not. 

To the surprise of many economists, the economy has been able to add many more jobs than seemed likely a year or two ago without creating an inflationary spiral.  But surely there are limits to that growth. 

Immigration deserves to be a major political issue this year.  But not in the way it has become one.