The Falling Participation Rate – Lack Of Jobs, Not Retirement

The Labour Force Participation Rate  in the United States has shifted from being an economic statistic that most investors had never heard of to becoming a relatively heated point of debate. In this article, I summarise two recent pieces of research with conflicting conclusions. My conclusions are somewhat tentative, but I agree with the more conventional interpretation of the data  – the fall in the Participation Rate partly reflects demographics, but it also reflects a lack of jobs.

Background: Why The Participation Rate Matters

The debate over the causes of the fall of the Participation Rate appears fairly arcane, but it matters for the outlook over central bank policy. The reason is that this affects out interpretation of the Unemployment Rate, which is one of best measures of spare capacity in the economy.

On paper, an “output gap” - the difference between the actual level of real GDP and “potential” GDP – is the best summary of spare capacity in the economy. Unfortunately, we only have a so-so idea of what real GDP is doing right now, and an even worse idea of what “potential GDP” is. What we see is the GDP-based output gap does a very good job of explaining what happened 10 years ago, which is when most of the revisions to GDP and the output gap are finished. The Unemployment Rate, on the other hand, is not revised (not counting the updates to the seasonal adjustment factors). It is also associated with the most important market in the economy – the labour market. Other measures of spare capacity, such as manufacturing capacity utilisation, have become less meaningful as manufacturing activity shifts overseas. 

Returning to the Participation Rate, it colors the interpretation of the fall in the Unemployment Rate. Since the percentage of the population that is working is not rising, the Unemployment Rate is only falling because an increasing percentage of the population is leaving the labour force. This is because people who are outside the labour force, such as retirees or students, are not counted as employed or unemployed. If people have only left the labour force because they cannot find a job (or find something else to do, as I discussed here), an improving labour market may bring them back. 

To provide a simplified version of the debate between the doves and the hawks:

• Hawks: The fall in the Participation Rate is mainly due to demographics (older people retire) and thus the fall in the unemployment rate is telling us that spare capacity is diminishing.
• Doves: Demographics cannot explain all of the fall of the Participation Rate. People will return to the labour market as jobs are created, and so those re-entrants will allow the economy to grow a lot longer before capacity constraints are hit.

Since the timing of Fed rate hikes will be determined by when capacity constraints are hit, this debate matters for the bond market.

From “Battling Straight Lines” To “Battling Fed Researchers”

The first paper I want to note is this paper – “On the Causes of Declines in the Labor Force Participation Rate” - by Shigeru Fujita, of the Philadelphia Fed. In it, he used a usually ignored subset of the Current Population Statistics (CPS, usually referred to as the Household Survey) – the micro data (raw survey results). Within that dataset, reasons are given for leaving the labour force:

1. Retirement,
2. Health issues (disabled),
3. “Other”.

If people quit looking for work because they see no chance of finding a job, it would fall under “Other”. As I noted in the article I linked to earlier, there may be an increasing number of students as a result of lack of jobs. (Additionally, I have seen arguments that there have been a rising number of out-of-work people claiming disability benefits. I have not looked at this, but some of the rise in the number of disabled people would be an expected side effect of an aging population.) 

The paper has conclusions about the causes of the fall of the Participation Rate over a longer horizon, but the most interesting conclusion is the following:

The decline in the participation rate since the first quarter of 2012 is entirely accounted for by increases in nonparticipation due to retirement.

The implication drawn from this is that since those workers are retired, they are thus not expected to return to the labour force. This implies that since the first quarter of 2012, the fall in the Participation rate is entirely the result of demographics, and thus the fall in the Unemployment Rate over the past two years is legitimate. Note however, that the Participation Rate was depressed due to non-demographic factors before 2012, so even on this interpretation, the Unemployment Rate is “too low”.

The difficulty with looking at this claim in detail is that it would take a very considerable effort to replicate the analysis; it is a big, unfriendly, data set to work with. I will instead switch to look at another paper, and draw a comparison to its conclusions.

In a note “Labor Force Participation Rates Revisted”, on the Atlanta Fed Macroblog, Timothy Dunne and Ellie Terry look at the explanation of the falling Participation Rate based upon the “macro” (summary) data of the same CPS survey. The article gives a graph of the contribution due to falling participation rates within age cohorts (e.g., 35-44 years), as well as the effect of changing weights of the cohorts (since older cohorts have a lower Participation Rate, an increased weight of oldsters lowers the aggregate Participation Rate) over (roughly) the same time interval – 2012 to the third quarter of 2013. The youngest age cohort (16-24 years) has bounced back, and is adding to the aggregate participation rate. However, the cohorts from 25-54 are subtracting from the Participation Rate since their Participation Rates are falling (and this is larger than the effect of the 16- to 24-year old cohort).

Unfortunately, I did not have time to load the whole data series so that I could generate graphs. But using the BLS web data retrieval form, we see the following:

Series Id:           LNS11300089
Seasonally Adjusted
Series title:        (Seas) Labor Force Participation Rate - 25-34 yrs.
Labor force status:  Civilian labor force participation rate
Type of data:        Percent or rate
Age:                 25 to 34 years

Year	Jan	Feb	Mar	Apr	May	Jun	Jul	Aug	Sep	Oct	Nov	Dec	Annual
2012	81.7	81.6	81.9	81.9	81.7	81.6	81.2	81.3	81.5	82.1	81.6	81.8
2013	81.7	81.7	81.4	81.3	81.3	81.4	81.1	81.2	81.0	80.6	81.2	81.0

Series Id:           LNS11300091
Seasonally Adjusted
Series title:        (Seas) Labor Force Participation Rate - 35-44 yrs.
Labor force status:  Civilian labor force participation rate
Type of data:        Percent or rate
Age:                 35 to 44 years

Year	Jan	Feb	Mar	Apr	May	Jun	Jul	Aug	Sep	Oct	Nov	Dec	Annual
2012	82.8	82.9	82.5	82.4	82.6	82.7	82.5	82.6	82.5	82.7	82.1	82.5
2013	82.2	82.1	82.2	82.1	82.6	82.1	82.5	82.4	82.3	81.9	82.1	82.1

In other words, for these two cohorts (that are well away from the traditional retirement age), the Participation Rate fell by 0.7% over the time interval isolated by Shigeru Fujita. Although I have not attempted to replicate the full results of Dunne and Terry, this snapshot tells me that their results are at least directional correct – about half of the fall in the Participation Rate in the past two years is due to those under 55 years dropping out of work faster than than was typical for their age.

As far as I can tell, the only way to reconcile the conclusions drawn by the different Fed researchers is that there has been a lot of retirements within the sub-54 age group. Although early retirement blogs are increasingly popular, a sudden upsurge in financially independent young people seems hard to reconcile with the survey evidence on the state of personal finances for most households.

Even if early retirement is the actual explanation, I question this implication drawn by Shigeru Fujita:

The likelihood of those who left the labor force due to retirement or disability rejoining the labor force is small and has been largely insensitive to business cycle conditions in the past 

Even at age 54, there is considerable scope for retirees to return to at least part-time employment, especially given the shift towards part-time jobs within the economy.

I also wonder about the reliability of the subset of the "micro" data that was used to draw the conclusions about retirement. I do not think that this part of the data set has been heavily analysed in the past. Do discouraged workers have a tendency to answer that they are “retired” if they have been stuck in long-term unemployment? If they do, this makes this part of the data set inapplicable for the question at hand. Since the U.S. job market has not been this weak since the Depression, the behaviour of the data will be different now when compared to previous cycles.

Conclusions

To my mind, the fall in the Participation Rate amongst younger cohorts tells us that the employment market is in an abnormally weak state. Since these rates are still falling, I find it hard to see why the environment would be radically different over the past two years.

Although I cannot produce a quantitative statement like  – “[insert number]% of the fall in the Participation Rate since [insert date here] was due to an ageing population” – I question the usefulness of attempting to precisely answer that question. If the percentage of the population that is employed has stopped rising, I think that it could take up to a decade for spare capacity in the U.S. labour market to disappear (even assuming that a recession does not occur during that interval). Therefore, the primary question is whether job growth will accelerate, and it is only if this acceleration occurs that the secondary question of the Participation Rate will begin to matter.

(c) Brian Romanchuk 2014