A Fallacy of Composition

Copyright, The New York Times Company

In 2009 the New York Yankees opened their new stadium on the north side of East 161st Street, replacing the historic stadium on the south side of the street. Not surprisingly, 2009 spending by consumers, news organizations and entertainment businesses, among others, on the north side of East 161st Street was a lot more than it had been in years past. It all started from the Yankees’ spending at the new location.

So a spending advocate might assert that this episode is proof that spending by one organization can stimulate spending by others, because the spending by the others on the north side of the street surged at exactly the same time that the Yankees started having their people work there.

Of course, such an analysis is flawed, because it ignores what happened on the other side of the street. Much of what happened north of East 161st Street was just a displacement of activity from the south side, rather than a creation of new activity. Even the construction workers building the stadium may well have been drawn from other tasks.

This pattern is not special to the Yankees’ move. A number of studies have shown that consumer spending associated with a sports team to a large degree displaces spending in other areas and displaces spending on other leisure activities; a family is unlikely to conclude that because there’s a new team in town or a new stadium, it should sharply increase its spending on entertainment.

Yet ignoring the displacement effects is exactly what Paul Krugman and Dean Baker have done in their praise of recent studies that use “cross-state variation in stimulus spending per capita to estimate the employment effects of the stimulus,” studies comparing states that received more stimulus to states that received less.

Spending from the American Recovery and Reinvestment Act (a.k.a., the “stimulus”) could be very much like the stadium spending — a locality that received more stimulus spending merely enjoyed a displacement of activity into its area from localities that received less spending, and that nationally little or no additional spending occurred as a result of the legislation.

If you want to know about the national effects of the stimulus, at least part of the analysis has to look at the nation as a whole. The same is true of the national effects of changes in labor supply. If one group suddenly becomes more willing to work, it is possible that the group solely takes jobs from the rest of the population, with no new jobs being created for the nation as a whole.

That is why, in addition to looking at the experiences of specific groups and specific regions, I have examined the effects of supply and demand on national employment. (Professor Krugman and Dr. Baker assert in so many words that I ignore national employment, though my papers on the subject look very much at national aggregates. For example, see Figure 3 and Figure 6 of this paper and Table 2, Table 3, Figure 2, Figure 3, Figure 4 or Figure 5 of this paper).

I found, for example, that national employment increases during the summer precisely because young people are more willing to work. Not surprisingly, the summer surge of young job seekers does seem to reduce employment of the rest of the population, but the net national effect is still almost a million more jobs in the summer.

For now, it appears that government spending reduces private spending, even while it may benefit specific regions or groups.

Exceptions to Keynesian Theory

Copyright, The New York Times Company

While taxpayers have been wondering if all of the extra government spending of the past couple of years has actually served to impede the recovery, Keynesian economists have been asking them to keep faith in the promise that government demand is the secret to economic recovery. Now Paul Krugman, an outspoken Keynesian stimulus advocate, admits that Keynesian theory has many exceptions.

It’s pretty easy to see how various types of government spending might reduce employment, rather than increase it: a number of government programs have been reducing the incentives for people to work, and reducing the incentives for business to hire.

Unemployment insurance is an example (among many) of how the work incentives of so-called stimulus programs operate. Unemployment insurance payments to individuals cease as soon as the individual starts to work again. I agree that such payments are compassionate, and may well be the right thing to do, but economists have long recognized that such compassion is not free: unemployment insurance reduces employment, rather than increasing it, because it penalizes beneficiaries for starting a new job.

Without offering any proof that incentives suddenly ceased mattering, stimulus advocates, and even the Congressional Budget Office, have recently ignored this effect. Many of them aim to prove the potency of unemployment insurance and other components of the stimulus law by insisting that the recession was caused by a lack of demand, and that any public policy that raises aggregate demand must be a big help.

Even if they’re right that the recession was a result of low demand, it does not follow that the way to recovery is to destroy supply, too. Before we turn away from one of the basic lessons of economics, we ought to have some evidence of the fundamental Keynesian proposition that “incentives to seek work are, for now, irrelevant.”

(Another tendency of Keynesians is to “prove” their supply claim by pointing to the existence of unemployment. Of course, unemployment exists in large numbers, but that does not tell us whether, and how much, incentives affect employment rates.)

Part of my research has been to examine episodes, from the current downturn, of changes in the willingness and availability of people to work. If, as Keynesians have been insisting, the incentives to work are in fact irrelevant in a recession, then none of these episodes would be associated with employment changes. (In their view, an increase in the number of people willing to work would just increase, one for one, the number of people who are unemployed.)

I looked at seasonal changes in labor supply. I looked at the increase in supply of workers to the nonresidential construction industry (workers who were leaving home building after the crash). I looked at the increase in the supply of elderly workers. I looked at the increase in supply of workers in Texas. In all of these recession-era episodes, more supply meant more jobs, and less supply meant fewer jobs.

(I also looked at some recession-era demand changes to see if they were at all constrained by supply, and they were — very much as they were before the recession.)

There is still no evidence to confirm the fundamental Keynesian proposition that supply doesn’t matter.

Rather than completely discard that proposition, Professor Krugman has recently formulated a theory of exceptions to the Keynesian theory, which he believes can help explain some of my findings:

Here’s the question: why do patterns of employment over time that are, in fact, normally supply-driven continue to be visible even during a demand-side slump? And here’s the answer: businesses make long-term decisions that influence hiring patterns over time, and those decisions continue to shape their behavior even when there is a surplus of labor.

In other words, Keynesian theory has exceptions that have to do with business’s long-term hiring decisions. For example, businesses have lived through enough seasonal cycles to know that they can normally make more money when their hiring patterns are responsive to the seasonal availability of people to work, so businesses continue to be responsive to the seasonal pattern of labor supply even during a deep recession when there are plenty of workers available throughout the year.

I don’t understand how Professor Krugman explains that the nonresidential construction industry took advantage of the plentiful supply of home builders after housing crashed (he also has no explanation for my minimum wage findings, Christmas seasonal findings or elderly employment findings). He also fails to explains why some business hiring patterns survive the recession intact, while other practices are completely different (e.g., businesses used to think they needed 138 million payroll employees, but by 2009 they got by with fewer than 130 million).

But even if Professor Krugman were correct that the ghost of labor supplies past haunts the recession through business’s long-term decisions, how can he be so sure that the labor-supply effects of government spending programs would not also have the same effects they did in the past?

For example, employers found that people were more difficult to hire and retain when a generous safety net was available. In this way, unemployment insurance would continue to reduce employment even after the recession began because employers have learned that the more generous the safety net, the more they must get by with fewer workers.

Would Keynesian stimulus spending work only when it came as a surprise? Or only when the spending was outside the range of prior business experience? Keynesian economists have not even begun to answer these questions. For now, Keynesian theory has so many exceptions that we might as well discard it.

Supply of Admissions

Professor Krugman is slowly converging to my view that supply matters even during recessions. I'll explain more tomorrow in a more lengthy blog post, but the basic idea has been in my seasonal supply paper (see the final para).

Labor Supply Always Matters

Copyright, The New York Times Company

As it has for the last three summers, the economy’s regular seasonal cycle will accumulate yet more evidence against Keynesian models of recession labor markets.

Two important seasons in the labor market are Christmas and summer. The Christmas season is an obvious time of high demand — people want to spend more in November and December. Basic economics says that Christmastime demand, while it lasts, raises wages, employment and hours, while it reduces unemployment.

Employment and work hours are also high during the summer, and as you are reading this, employment is likely to be surging well above what it was a month ago (the Census Bureau employment data for June is not due out until July 8, and the July data not until a month after that).

In the past, a few readers of this blog have given demand the credit for the summer job surge. In my view, demand contributes a little to the summer job surge (after all, agriculture, construction and other industries are expected to be more active when the weather is warmer), but supply is the primary reason that jobs are created during the summer.

One basis for my opinion is that people in vast numbers become available for work when school lets out for the summer, and about the same number are no longer available when school resumes. For example, about 20 million people 16 and older are attending school during the academic year, and very few of them are working full time. Nobody knows the students’ intentions for sure, but certainly millions of them would like to work during the summer.

I don’t know of any change in demand occurring over the summer that would number in the millions (even doubling the size of our military overnight would not create much more than a million jobs).

It’s easy to tell the summer supply stories and demand stories apart. The demand story is a lot like Christmas — customers demand, employers want to satisfy customers, so they hire more workers. If the demand story applied to summer, then we should see summer employment and work hours surge, and wages increase, too, while unemployment should dip.

If I’m right that the summer job surge comes from supply — the increased availability of workers — then summer wages and unemployment should follow patterns opposite to Christmas: wages should fall and unemployment surge.

The charts below display three labor market indicators — weekly hours worked, hourly pay for full-time jobs and unemployment — for the two seasons. (I use data from the Current Population Survey Merged Outgoing Rotation Group from January 2000 through December 2009).

The charts show seasonal spikes: the level of the indicator during the Christmas season (November and December) or the summer (June through August), relative to the indicator during the four months near the season. Wages and unemployment are represented as a proportional change from their off-season values, and spikes in hours are expressed as a proportion of a group’s average hours for the entire season and adjacent months.

(A person not at work counts as zero. To create this particular chart — other charts and tables are also available — I focus on people less than 35 years old, because their job turnover rates are greater and thus more visibly display the effects of short-term fluctuations like Christmas or summer. Please note that I have truncated the green hours bar and indicated its actual value with text, because the teenage summer hours spike is 0.295, which far exceeds the scale needed to display the other figures.)

Each group’s hours spike is positive on Christmas. During the summer, the hours spikes are positive only for the two younger age groups, which we expect because those are the groups attending school during the academic year and becoming suddenly available to work in June. All three Christmas wage spikes (middle panel) are positive, while all three summer wage spikes are negative.

It’s hard to believe that summer involves the kind of demand surge we see over Christmas; summer wages and unemployment go in exactly the opposite direction that they do during Christmas.

Thus, the summer job surge is nothing like Christmas. The economy creates jobs in the summer — even during the last several years, when our economy supposedly suffered from a lack of demand — because millions of people become willing and available to work. This is not to say that everything is working well in the labor market — employment is much lower than it should be — it’s just that greater labor supply remains one route to higher employment.

As noted by Greg Mankiw, the Harvard economist, and Gauti Eggertsson of the New York Fed, the fact that, even now, jobs are created when people are willing to work is a big challenge to the Keynesian economic model that assumes that labor supply is irrelevant during recessions, liquidity traps and other labor-market crises.

As Paul Krugman put it: “What’s limiting employment now is lack of demand for the things workers produce. Their incentives to seek work are, for now, irrelevant.”

That’s why Keynesians contend that expanding unemployment insurance can increase employment, even while they know that it erodes work incentives. Yes, unemployment is too high and employment is too low, and I applaud Keynesian economists for trying to understand that.

But they have gone too far and have ultimately given the wrong advice, in assuming without proof that labor supply is irrelevant. The labor market’s seasonal cycle shows pretty clearly that they’re wrong.

The Summer Job Surge: Supply or Demand?

The summer teen employment surge is largely a consequence of seasonality in supply, not demand. To see this, note that a pure summer demand surge would draw teens into the labor market with low teen summer unemployment, high summer real wages, and low summer unemployment among persons not enrolled in school during the academic year. The hypothetical demand surge would also have to be quite large – about as large as doubling the size of the nation’s military in a mere two months – because the end result is about a million new jobs for teens.

In fact, teen unemployment spikes in June as the labor market absorbs more than one million teens. Unemployment of persons aged 25 and older (not shown in the figures) is high throughout the summer, peaking in July at almost 700,000 persons above trend. Median nominal and real weekly wages for teens are often at their lowest of the year in the third quarter (July – September), and presumably hourly wages are even lower due to longer teen summer work weeks. These patterns reverse when the academic year ends.

Also consistent with the supply interpretation, Mulligan (2010b) shows how age groups with the largest summer log employment and log unemployment spikes are those with the greatest school enrollment rates during the academic year, and the summer log employment spike may even be negative for groups with near zero school enrollment.

Nor do many of the summer jobs for teens appear to be in industries that have a significant spike in labor demand, because 77% of those jobs are in industries that expand their employment of persons aged 25-34 less than two percent, if at all. Based on calculations from the May, July, and September 2005 Current Population survey, the top industry hiring teens in the summer was “arts, entertainment, and recreation” (accounting for 19 percent of the teen summer jobs), which had no change in the number of persons aged 25-34 employed. The second industry (also accounting for 19 percent) is “accommodation and food services,” which actually cut its employment of persons aged 25-34 by 4 percent during the summer.

These are all indicators of supply shifts during the summer that are large, and exceed the demand shifts.

Spending and Value Added over the Seasonal Cycle

I've heard some complaints about my seasonals article: that I should have looked at the impact of Christmas value-added rather than Christmas spending. If the complaints prove anything, they prove that Keynesian theory has not yet progressed beyond a statement of faith to a theory that might actually be tested.

First of all, this complaint is a great example of slippery Keynesian rhetoric. We have been told time and time again that the recession and sluggish recovery are to be blamed on too little consumer spending. We can and should help fix it -- create millions of jobs, they claim, with trillions of dollars of government spending. They even claim that unemployment insurance (UI) -- a transfer that is normally expected to reduce employment -- would, these days, increase employment because UI is spending.

In summary, the Keynesians say that spending really matters during a recession, and that employment, value-added, etc., would be the result of that sorely needed spending.

Now we're told that value-added is supposedly what's missing, not spending. So what happens to consumer spending in December is, all of the sudden, no big deal, because most of the December action is purportedly spending rather than value-added.

You might think that I'm being too harsh with the Keynesians because, after all, they are trying to explain economics to politicians and laymen who might be overwhelmed by a term like "value-added", so the Keynesians say "spending" when they really mean value-added.

But UI is an excellent example for sorting this out, because we all agree that UI is spending, but not value-added (UI actually pays people for not adding value!). So Christmas consumer spending should create as least as many jobs per dollar (probably more, for the incentive reasons I mentioned in my earlier post) than a UI program does.

Second of all, in claiming (without evidence) that the retail sales seasonal is very different from the value-added seasonal, econbrowser probably has the facts wrong. Professor Jeffrey Miron has written a book about the seasonal cycle, and reports that value-added falls sharply from the fourth quarter to the subsequent first quarter -- very much in line with the drop in retail sales. The value-added of Christmas is not all that different from the spending.

Third, I do not doubt that December retail sales are associated with some significant production in November, and October, and maybe even with some significant production in the 3rd quarter. But both of my recent posts on this topic repeatedly reference drops FROM DECEMBER TO JANUARY. At lot of the seasonal jobs lost at the end of the year include those that started in October and November, so in making December-January comparisons I do not have to assume that all of the employment and production associated with December spending occurred in December.

Keynesians are understandably nervous that the seasonal cycle appears to contradict the most fundamental tenets of their theory. Many of them really do think that government spending is as good as Christmas.

Does Supply Matter During Recessions? Evidence from the Holiday Seasonal

Copyright, The New York Times Company

There will be almost three million fewer jobs next month than in December, and this outcome reveals a lot about how the labor market works.

As I noted last week, employment normally falls sharply after Christmas, for the obvious reason that consumer demand is significantly less in January than in the preceding December. Next month should be no exception.

Three million is a lot of jobs – that happens to be the total number of jobs that the Obama administration contended that it had “created or saved” over the entire life of the stimulus law passed in 2009. Yet next month’s multimillion loss of employment will be largely ignored by news organizations, because it will not be much different than it has been any other January.

But it’s big news that the seasonal employment cycle continues to operate as normal. The cycle is, of course, the outcome of seasonal fluctuations in supply and demand, and Keynesian economists insist that supply and demand have not been operating normally since the recession began and that the economy has been caught in a “liquidity trap.”

Keynesian economists assert that labor supply – the willingness of people to work – doesn’t matter right now, which is why they advocate government stimuli that increase “demand” spending even while they erode work incentives. As Paul Krugman put it: “What’s limiting employment now is lack of demand for the things workers produce. Their incentives to seek work are, for now, irrelevant.”

In other words, a recession like the recent one is purported to be one of those rare times when demand is not constrained by supply.

My view is that lack of demand has not been the primary problem with our labor market, which is why the stimulus law has created no visible employment increase. Because of this disagreement among economists about the roles of supply and demand during recessions, it’s important to examine whether demand was operating normally during the latest recession.

If the Keynesians are right, the disappearance of Christmas demand in January would be even more devastating for total employment during a recession than it would be normally, when supply limits the effects of Christmas demand.

If so, the holiday employment seasonal would be about twice as large during recessions – with January job losses in a recession of five million or six million, compared with the normal three million.

The chart below tests the Keynesian claims by comparing the holiday employment cycle across years. Each series measures the deviation of December employment from the October-January trend (Christmas shopping begins after October and is finished by January).

The green and red bars measure employment from the establishment survey and the household survey, respectively, both conducted by the Bureau of Labor Statistics.

The average holiday seasonal data previous to the recent recession is shown at the left, and the rest of the chart isolates the season for 2007, 2008 and 2009, some, if not all, of which years when our economy was purported to be in a supply-doesn’t-matter liquidity trap. (For further description of these and related calculations, see this paper)

Take, for example, the green bars from the establishment survey. On average from 1980 to 2006, December employment exceeded the October-January trend by 1.5 percent. During the years 2007, 2008 and 2009, the December employment seasonal turned out to be similar – about 1.3 percent, rather than the 3 percent or so that we would expect if the economics of supply and demand were really as different during those years as Keynesian economists say they are.

The red bar of the household survey is noticeably lower in 2009 but well within the range observed over the past several decades.

You might think that Christmas became smaller during the recession, and that change offset the purported extra impact of each dollar of Christmas spending. It is true that almost all kinds of spending are lower during a recession, but I adjusted for that by measuring the seasonal data in percentage terms.

The retail sales data show that, in percentage terms, the holiday spending surge was not much different from 2007 to 2009 than it was in previous years.

People can argue about whether demand effects are marginally bigger during a recession. But the seasonal cycle clearly shows a recession does not free demand – whether it be holiday demand or government demand – from the constraints of supply.

Does Labor Supply Matter During a Recession? Evidence from the Seasonal Cycle

During the recession of 2008-9, the federal government took a number of steps to help citizens and the economy, including expansion of food stamps and unemployment insurance, helping financially distressed homeowners refinance their mortgages, and offering tax credits to poor and middle class persons buying homes. The stimulus potential from these and other programs is said to derive from their redistribution of resources to persons with a high propensity to spend, but the same programs also implicitly raise marginal income tax rates because eligibility for them falls with the potential recipient’s income.

High marginal income tax rates by themselves “normally” reduce economic activity to some degree, rather than increase it, although there is plenty of room to debate the magnitude of incentive effects. For the same reason, social safety net programs are not expected to increase employment in the long run. But a number of economists believe that recessions are those rare instances when labor supply does not matter, and might even affect the aggregates in the opposite direction as usual (Eggertsson, 2010a). Thus, it is possible that government spending programs like unemployment insurance could stimulate economic activity during a recession, even while they eroded labor supply incentives, and even while those programs had very different effects in non-recession years.

The hypothesis that, as compared to non-recession years, demand matters more and supply matters less for determining aggregate employment and output at the margin in a recession is also the intellectual basis for Keynesian models of the business cycle (Eggertsson, 2010b, p. 2). Yet this hypothesis has not been the subject of much empirical testing, even though it is logically possible that supply matters at the margin just as much during times of severe labor market distortions as it does “normally.” The purpose of this paper is to examine the seasonals in the monthly U.S. data dating back to the 1940s to attempt to measure the degree to which labor supply and demand differentially affect employment and unemployment during recession periods than during non-recession periods.

The seasonal cycle has several analytical advantages. As Jeffrey Miron (1996, p. 17) explains, “The seasonal fluctuations are so large and regular that the timing of the peak or trough for any year is rarely affected by the phase of the business cycle in which that year happens to fall.” For example, Barksy and Miron (1989, Table 2) found that GNP falls 8 percent more than normal from Q4 to Q1. In a $14 trillion/year ($3.5 trillion/quarter) economy: that’s a sudden reduction of $280 billion, which is a larger change than even the largest year-to-year change in government spending created by the American Recovery and Reinvestment Act of 2009 (Congressional Budget Office, 2009, Table 2), and larger than other peacetime government spending shocks (Alesina and Ardagna, 2009; Auerbach and Gorodnichenko, 2010; Barro and Redlick, 2009).

Many economic fluctuations are not easily partitioned into “demand” or “supply,” but the seasonal cycle features an obvious demand change – Christmas – and an obvious supply change – the availability of teenagers for work during the summer. Moreover, these two seasonal impulses (measured as percentage changes from the previous and subsequent seasons – more on this below) react little to the business cycle, and thus provide the opportunity to measure different effects between recessions and non-recessions of a similar impulse. Finally, the seasonal cycles have occurred many times: there have been 12 summers and 12 Christmas’ during U.S. recessions since 1948. Even during the present recession – arguably different from many of the previous ones – Christmas and summer each occurred at least twice (depending on when the recession is deemed to have ended).

Previous work on the seasonal cycle has featured quarterly data, which had the advantage that the Bureau of Economic Analysis used to report seasonally unadjusted quarterly national accounts. However, unlike the labor market series used in this paper for which the raw data are seasonally unadjusted, much of the national accounts are built from seasonally adjusted inputs, and seasonally “unadjusted” national account series were obtained by attempting to remove the seasonally adjustments that had been implicitly introduced via the ingredients. More important, the supply and demand shifts of interest here do not coincide exactly with calendar quarters. The seasonal labor supply surge is seen already in June, which is part of the second quarter, and concludes in September, which is at the end of the third quarter. Obviously, Christmas is in December, and some of its activity spills into November, both of which are part of the fourth quarter, but the monthly data permit me to use October as a benchmark for Christmas, rather than the third quarter which would differ from the fourth not only in terms of Christmas demand but also in terms of summer labor supply.

Section I of this paper takes for granted that recessions are appropriately characterized as times of severe labor market malfunctions, and briefly shows that a couple of familiar theories predict that labor demand matters significantly more at the margin, and labor supply matters significantly less, during recessions than during non-recession years. However, other theories of labor market distortions predict that the incidence of supply and demand shifts would be no different during recessions than they would be during non-recessions, so these incidence questions must ultimately be answered with empirical evidence.

I find that the summer seasonals for teen employment, teen unemployment and total employment are large and in the direction to be expected if labor supply had shifted significantly more than labor demand. However, the seasonal cycles for recessions and non-recessions are not significantly different from each other.

The Christmas seasonals seem to be essentially the same in recession years as in non-recession years.

Even the 2008 and 2009 summers and Christmas’ looked a lot like summers and Christmas’ in nonrecession years.

These findings contradict the view – which is the basis for much fiscal policy and business cycle analysis – that labor supply shifts have little (or even perverse) effects on aggregate employment during a recession, and contradict the view that demand shifts encounter significantly fewer supply constraints during a recession than they normally would. Admittedly, recessions are times when the labor market does not function well, but nonetheless labor supply and demand seem to operate on the margin during recessions in much the same way that they do during non-recession years.

The Christmas cycle is at least as large as the high frequency peacetime government spending changes that have been observed in U.S. history, so my results might imply that fiscal demand shocks would have much the same employment effects in a recession as they would in non-recession years. Of course, the seasonal results by themselves do not rule out the possibility that a fiscal demand increase significantly increases employment regardless of whether or not it were a recession (although see Alesina and Ardagna, 2009 and Barro and Redlick, 2009, on this point).

It is possible that the labor market has different mechanisms to adapt to various supply and demand shifts, and that certain types of fiscal policy might be different from Christmas in this regard. The seasonal cycle is also easily anticipated.

Either case raises the question of how, exactly, fiscal policy might be different from Christmas, why government spending might encounter fewer supply constraints than Christmas does, and how that information can be used to better design fiscal policy during recessions.

The Seasonal Job Surge: Supply or Demand?

Copyright, The New York Times Company

Keynesian economists designed a fiscal stimulus law that targets aggregate demand and runs roughshod over supply incentives. Evidence on the seasonal labor market shows why their stimulus was ultimately self-defeating.

Keynesian economists contend that labor supply — the willingness of people to work — doesn’t matter right now (or even that a more willing work force would reduce employment). If they are right, then a stimulus law that increased spending might work even if the law were taking incentives away from people to work and employers to hire.

Under the assumption that the cycle for teenage employment is a supply phenomenon — that is, the seasonal surge is more a result of teenagers’ becoming available for work than it is a change in employer demands — I showed last week that the Keynesians are wrong. Supply matters a lot, even in the depths of this recession, which is why the stimulus law was doomed to failure.

A few readers asserted that the teenage employment surge is a result of labor demand rather than supply. Fortunately, economics offers several tests of whether demand or supply is the dominant factor in the labor market.

First, the supply side of the problem obviously is related to school enrollment: population segments with more school enrollment during the academic year have a greater supply shift when summer arrives. So if supply were an important factor, we should see a larger summer job surge for groups with higher school enrollment rates.

The chart below confirms this hypothesis, using data from 1980-2007: the summer employment spike is a greater percentage for the younger teens, and those are exactly the age groups for whom school enrollment is highest. (The summer employment spike is measured by the log deviation of July’s value from the average of May and September values.)

In fact, there is essentially no employment spike for the 25-to-34 age group, and that group has hardly anyone in school. If summer labor demand were as great as my critics claim, summer ought to bring some extra employment for the 25-to-34 group too.

Summer does involve some demand shifts, but the unemployment data readily show that the summer supply increase dwarfs demand in the labor market for young workers. Figure 2 below displays unemployment spikes for the same age groups.

When school lets out, students storm into the job market and jobs are created for most of them. A few students spend some of the summer unemployed because students, and not their employers, are the ones who suddenly decided that summer is the time when they are available to work. That’s why the summer unemployment spike is positive — unemployment per capita for these groups is greater during the summer than during the academic year, especially for the groups with more school enrollment.

Not surprisingly, summer unemployment is higher than normal even for nonstudents, because many young people out of school compete in the labor market with students who are set free from school each summer.

The Keynesian claim that supply doesn’t matter in recession is quite radical and contradicted by the labor market’s seasonal variation. The sooner we can undo the supply-harming provisions in Obama-era legislation, the sooner our labor market can have a real recovery.

Summer: Is it Demand or Supply?

A few readers have wondered whether the summer employment surge, especially for teens, reflects demand or supply.

Of course, to some degree the demand for teenage workers shifts during the summer. But the demand shift is less than the supply shift. To see this, you could look at the seasonal for wages or, as I did in my paper, look at the seasonal for teenage unemployment.



If demand were shifting more than supply, then teen unemployment would be low in the summer -- summer would be the easiest time for an eager teenager to find a job. If supply shifted more than demand, then teen unemployment would be high in the summer, especially early in the summer before the market had much time to absorb all of those students storming out of school, because each teen job searcher has so many other teens with whom to compete.

Odd Niche Employment

This article at Atlantic.com acknowledges that labor supply matters -- that people with weak work incentives will be "pickier" about job offers.

At the same time it quibbles with my work on teenagers' seasonal, claiming that summer jobs for teens are "odd niche employment."

It doing so, it has begged the question! From where do the "odd niches" come? Outer space? Is it just a coincidence that teens enjoy their odd niche at exactly the time they are available to work, rather than, say, March-May? Is it just a coincidence that the summertime niche is for teens, and not for, say, elderly people?

The odd niche for teens comes from their willingness to work! If other groups were willing to work -- willing to give up their spare time and effort to help a business produce more, and accept pay that was commensurate with their production -- then emplopyment niches would appear for them too!

[NOTE: none of this says that the labor market is functioning well -- I have written elsewhere that is is not -- just that labor supply still matters in the usual way, despite the presence of large labor market distortions.]

Recession Supply Debate

I have shown that there is a close theoretical relationship between crowding out (of government spending) and the quantity effects of supply. To recap,

• the crowding out view of government spending says that supply is scarce in the aggregate: if you want more employment and output, factor rents have to adjust to induce suppliers to be willing to do the work that it takes to have a job and be productive. The higher factor rents created by public sector demand will reduce private sector factor demand.
• by the same logic, a significant increase in labor supply should significantly increase employment and output, precisely because supply is scarce
• Thus, if you think crowding out is unimportant, you must conclude that supply is not an important determinant of the aggregate quantities of employment and output.

My recent paper reviews this theory, and offers evidence from the 2008-9 recession that aggregate employment was impacted by supply. Yesterday, I updated part of that work: my work on the aggregate effects of seasonal supply shifts (thanks to NGM for the link and cliff notes!).

In doing this work, I have benefitted from discussions with Dr. Gauti Eggertsson at the FRB of NY, who believes that fiscal policy has no crowding out effect at times like this, when the economy is supposedly in a liquidity trap.

So you may be interested in our exchange to date (NOTE: our exchange is ongoing, and I expect that Dr. Eggertsson has some new points yet to make, or will help me better understand points he has already made).

• Summer 2009: Dr. Eggertsson was working on his theory paper claiming that, in a liquidity trap, the economy would paradoxically react to MORE labor supply by having LESS employment. He told me about this in September when I visited FRBNY.
• Summer 2009: Although not yet aware of Eggertsson's paradox, I was skeptical of even the more moderate view that "supply matters less during a recession." As soon as the July 2009 employment data was released, I ran my first test on the seasonally unadjusted employment series. It showed that labor supply does affect aggregate employment, and to about the same degree that it did before the recession.
• Fall 2009: I learned about Dr. Eggertsson's Paradox of Toil paper, as well as Professor Woodford's paper claiming that crowding out effects might (in theory) disappear in a liquidity trap, so I began to collect the various Great Recession labor market events that might confirm or contradict their claims, and emailed some of my results to them (my paper was later distributed more widely by NBER).
• May 2010: Dr. Eggertsson kindly wrote back, and pointed me to his new reply paper. In his view, empirical work that properly tests the aggregate effects of supply shifts in a liquidity-trapped economy could not only confirm or reject his particular model, but "pretty much any model that features nominal rigidities" (p. 2, emphasis in the original). So a lot is at stake here!
• May 2010: With regard to my seasonal fluctuations test, Dr. Eggertsson said that a seasonal labor supply increase would not REDUCE employment (even while other sorts of labor supply shifts would), but would leave it UNCHANGED: compare Figures 1 and 2 in his reply paper, respectively.
• July 2010: I write back to Dr. Eggertsson that his Figures 1 and 2 are hardly different in the sense that they BOTH say that the aggregate effects of supply are fundamentally different in a liquidity trap than they "normally" are. Figure 1 says that more labor supply reduces aggregate employment, Figure 2 says aggregate employment is unchanged; both of those are fundamentally different than my view that more labor supply INCREASES aggregate employment regardless of the state of the liquidity trap. The seasonal data support my view, and contradict both his Figure 1 and his Figure 2.
• July 2010: I also wrote requesting guidance as to, in theory, which labor supply shifts, and which fiscal policy shifts, should be analyzed with his Figure 1, and which should be analyzed with his Figure 2. In any case, both versions of his theory are contradicted by the seasonal employment data
• ... more to come.
  

The Seasonal Job Surge: 2010 Edition

Copyright, The New York Times Company

National employment has increased by three million since January, and the change tells us a lot about how the labor market operates.

The headline from Friday’s employment report was that employment was lower in July than in the previous month and has hardly recovered since the beginning of the year. That headline report does not refer to actual estimates of employed persons, but rather to seasonally adjusted estimates.

The fact is that national employment was two million to three million higher in July than it was six months before (employment estimates vary somewhat between the business surveys and the household survey).

If previous seasonal patterns continue, much of that increase is expected to reverse itself by January. So, at first glance, there is no reason to get excited about the actual two million to three million increase.

But it’s big news that the seasonal employment cycle continues to operate as normal. The cycle is, of course, the outcome of seasonal fluctuations in supply and demand, and Keynesian economists insist that supply and demand are not operating normally since the recession began and that the economy has been caught in a “liquidity trap.”

Keynesian economists claim that labor supply — the willingness of people to work — doesn’t matter right now (or even that a more willing work force would reduce employment), which is why they advocate government stimuli that increase spending even while they erode work incentives. As Paul Krugman put it: “What’s limiting employment now is lack of demand for the things workers produce. Their incentives to seek work are, for now, irrelevant.”

My view is that the stimulus law is partly responsible for today’s low employment, precisely because so many of its components have eroded work incentives. That’s why it’s so important to know whether supply operates as normal.

If the Keynesians are right, then the fact that teenagers become more willing and available to work when school lets out would not, during the recession, increase employment (and might even decrease it — Keynesians are not yet clear whether they think that the seasonal supply cycle would have a paradoxical effect, or merely no effect).

The chart below tests the Keynesian claims by comparing the seasonal employment cycle across years. Each series is the ratio of teenage employment to the employment of people ages 20 to 24. The teenage group is the most affected by the academic year, because they have so many more members in school. The 20-24 group is used as a benchmark to reflect the general changes in employment, and its possibly specific effects on the types of jobs that would disproportionately employ people under age 25.

The summers of 2009 and 2010 are the depth of this recession; the recession was still pretty mild in the summer of 2008. The years 2006 and 2007 are before the recession.

All five of the years show a teenage employment spike in July. In fact, the July 2009 and July 2010 spikes are about the same size — 15 to 17 percent (relative to the previous May) — as the 18 percent average of the previous three years.

People can argue about whether supply effects are marginally smaller during a recession. But the seasonal cycle clearly shows that labor supply remains important and cannot be approximated as a nonfactor, or even a sharply diminished factor, in today’s employment fluctuations.

For the Moment, Men Regain the Majority

Copyright, The New York Times Company

By one measure, women were a majority of the nation’s work force for the five months ended March 2010. Employment trends suggest they will regain this majority next winter.

The chart below shows payroll employees by gender since January 2009. According to the Bureau of Labor Statistics, the majority of the nation’s payroll employees were female in February and March 2009. Women regained their majority status in November 2009 and held it until April 2009, when men held 50.04 percent of the nation’s payroll jobs.



Employment follows a seasonal pattern, and a somewhat different seasonal pattern for men than for women. Thus, the summer work force is more male than the winter work force, and it is no accident that women outnumbered men in colder months. So I expect that men will retain a narrow majority of payroll jobs until next winter.

The chart shows that the female majority was larger this winter than it was in the previous one, in large part because almost a million construction jobs were lost in 2009, and the construction industry employs many more men than women. The construction industry will not shed so many jobs this year, but it will take more than a year to recreate the construction and other male-dominated jobs that were lost in the previous year.

For this reason, I expect that women will again hold a majority of payroll jobs for several months next winter.

A 'Paradox of Toil'?

Copyright, The New York Times Company

Some economists have been recently discussing a “paradox of toil,” meaning that an increased willingness to work actually depresses the economy. But evidence from this recession clearly shows that the paradox of toil is of little practical importance.

Paul Krugman explained Monday on his blog:

when you’re in a liquidity trap … If you cut taxes on labor income, this expands labor supply — which puts downward pressure on wages and leads to expectations of deflation, which increases the real interest rate, which leads to lower output and employment. [emphasis added]

This paradox of toil is of interest because it turns standard economics on its head, and helps rationalize the view that active fiscal policy can boost the economy even while it reduces incentives to work. Professor Krugman and other fiscal stimulus advocates tell us that the paradox of toil describes our economy during this recession.

Fortunately, the relevance of the paradox does not have to be taken on faith, but can be examined with data from 2008 and 2009.

For example, if the paradox described this recession, then the expansion of labor supply that occurs at the beginning of every summer as students become available to work would lead to lower employment. Or on the other hand, an increase in the minimum wage would increase employment as it raises prices and costs and leads to inflation.

In fact, the 2009 labor market reacted to these events in the conventional way, with no paradox.

When school let out for the summer of 2009, teenage employment increased by over a million, and total employment increased by about 700,000 (these May-July comparisons are necessarily seasonally unadjusted, because the school year is part of the seasonal cycle). The expansion in labor supply did not reduce total employment, as Professor Krugman would have us believe would have happened in a “paradoxical” year like 2009. Total employment expanded seasonally as it did in previous years.

The federal minimum hourly wage was increased from $6.55 to $7.25 at the end of July 2009. It did not have the stimulating effect that Professor Krugman assumes.

Teenage employment fell 1.5 million after that increase, as compared to the 1 million that teenage employment typically falls at the end of summers that do not have minimum-wage increases. Total employment also fell more than the usual seasonal patterns would suggest.

The evidence shows that the laws of economics remain in full force, despite the present “liquidity trap.”

More on the Summer Jobs Surge

Copyright, The New York Times Company

Last week, I wrote about the seasonal employment surge: Almost 700,000 more Americans reported in July that they were employed than did in May. Since then, data releases, reader comments and additional space permit further interpretation of this fact.

The Census Bureau’s household survey found that, without seasonal adjustments, 141,055,000 Americans were working in July 2009, as compared to 140,363,000 who were working in May 2009. The two-month increase of 692,000 is significant, even when compared with some of the heaviest monthly job losses experienced during this recession.

Most of the media attention during this recession has been rightly focused on the seasonally adjusted employment series, because those remove the annual seasonal cycle and are therefore the best series for understanding whether the recession might be over. However, the seasonal employment and spending patterns themselves are normally the outcome of supply and demand, so they are a good laboratory for detecting a sudden change in the role of supply in determining employment outcomes.

Last week I explained how the supply of labor is an important factor every summer, and the summer of 2009 has been no exception: The end of the academic year makes more students willing and available to work, and the market finds work for them to do.

A lot of commentary has been quick to link employment to spending in the economy, and has given little attention to the supply of labor. However, late last week we learned that the volume of retail sales (not adjusted for seasonality) was actually lower in July than in May, and that the dollar amount of retail sales was essentially constant.

One reader inquired whether labor supply, while still relevant even in the depths of this recession, might be less important than in the past. I expect that there is something to this reader’s idea, because the minimum wage has increased three years in a row and is expected to create an excess supply of workers (especially among teenagers, and especially this year when the real minimum wage had already gotten pretty high even before July’s increase went into effect).

The chart below investigates this idea by comparing the seasonal employment cycle across years. Each series is the ratio of teenage employment to the employment of those 20 to 24. The teenage group is the most affected by the academic year because they have so many more members in school. The 20-to-24 age group is used as a benchmark to reflect the general decline in employment, and its possibly specific effects on the types of jobs that would disproportionately employ young people.


The summer of 2009 is the depth of this recession, whereas the recession was still pretty mild in the summer of 2008. The years 2006 and 2007 are before the recession, and the year 2006 before any minimum wage increase.

All four of the years – including 2009 – show a teenage employment spike in July. In fact, the July 2009 spike is about the same size – 15 percent (relative to the previous May) – as the 18 percent average of the prior three years.

It is true that a spike of 15 is less than a spike of 18, and this is consistent with the claim that the role of supply is slightly less important than it was before the recession. But the seasonal cycle clearly shows that labor supply remains important, and cannot be written off when looking at today’s job market fluctuations.

Unfortunately, demand-oriented public policy has done much during this recession to discourage the supply of labor, and little to encourage it.

Supply Obviously Matters: The Case of the Seasonal Cycle

Copyright, The New York Times Company

On Friday, the Bureau of Labor Statistics released a report of yet another month of job losses. But underneath the national, seasonally adjusted totals is a seasonal employment rise that contradicts much of what’s been claimed about our labor market.

To my amazement, the laws of economics have been suspended in economic commentary during this recession, especially among advocates of active government intervention. In particular, it has been repeatedly claimed that the supply of labor — the willingness of workers to work — is completely irrelevant for determining the total amount of employment in today’s economy. All that matters, so they say, is the demand for labor, and that demand is scarce.

The New York Times quoted a labor economist at Harvard as saying: “Traditionally, many economists have been leery of prolonged unemployment benefits because they can reduce the incentive to seek work. But that should not be a concern now because jobs remain so scarce.”

Paul Krugman wrote on his blog this week how he thought it ludicrous that any part of this recession’s employment decline derived from people’s willingness (or unwillingness) to work. Economist.com agreed.

This viewpoint is empirically testable, because at least one thing proceeded as normal this recession: School let out for the summer. As usual, students are more willing and available to work during their summer break than they are during the school year.

If it were true that willingness and availability for work were irrelevant for the number of people actually employed, then we should see little or no effect of summer break on the amount of employment (as measured by the number of people who tell the government they are working). Students on summer vacation and newly minted graduates should be entering the work force in the hopes of getting a job, but since the job market is so tough (especially on young people), employment across America should not actually go up.

The government data released on Friday contradict this view.

Obviously, the 16-to-19 age group has more members in school during the academic year than does wider population of adults. Thus, a decomposition of total employment by age helps us see what happened to employment as a result of the ending of the academic year.

The chart below displays seasonally unadjusted monthly employment for those 16 to 19, and all ages 16 and over, minus the number employed in the group in May 2009. When school let out, employment surged among the school-age groups: More than a million persons 16 to 19 said they were employed in July than employed in May.



Source: Analysis of Bureau of Labor Statistics by Casey B. Mulligan

Most of the media attention since Friday has been rightly focused on the seasonally adjusted employment series, because those remove the annual seasonal cycle and are therefore the best series for understanding whether the recession might be over.

However, the seasonal employment patterns themselves are normally the outcome of supply and demand, so they are a good laboratory for detecting a sudden change in the role of supply in determining employment outcomes.

If demand were the only thing that mattered, it would still be possible for employment to surge among those of school age, if the eager students were just “taking jobs” from older people. But the chart shows that the school-age employment surge was not fully offset by an employment drop by others, because total employment also increased as school let out.

Students became available to work, and the market found work for them to do.

Some may argue that summer itself is entirely a demand phenomenon: That spending surged this summer, akin to surges in summers past, and that created new jobs.

But so far the data show no summer 2009 surge in spending, or in wages. Seasonally unadjusted retail sales and hourly wages both fell more than 1 percent between May and June 2009 (the latest month available).

Thus, the national employment rate does depend on how many people are willing and available to work, and not solely on employer demand. Employment is higher when more people are willing and available to work even in the depths of recession.

The basic economic result that both supply and demand determine market outcomes has tremendous policy relevance. It is part of the case against protectionism. It helps explain why the federal government needs to roll back some of the terrible incentives it has created, such as the marginal income tax rates in excess of 100 percent embedded in President Bush’s and President Obama’s means-tested mortgage modification programs (that is, tests that evaluate things like a family’s income, net worth and debts to determine if the family is eligible to have its mortgage modified).

To be clear, much of the unemployment rate spike and the employment collapse do not result from bad policies initiated during this recession. Demand is relevant too. For example, the housing collapse would put a lot of construction employees out of work regardless of what the government did.

But public policy has made those bad employment numbers even worse, because supply also matters, and cannot be ignored.

Has Employment Been Rising?!

Maybe, maybe not.

The BLS offers four measures of national employment, and one of them has been rising for a while now:

1. establishment survey, seasonally adjusted (-247,000)
2. establishment survey, not seasonally adjusted (-333,000)
3. household survey, seasonally adjusted (-155,000)
4. household survey, not seasonally adjusted (+229,000). It has gone up THREE OF THE LAST FOUR MONTHS

For good reasons, the establishment survey is better for measuring monthly fluctuations, and we generally prefer seasonal adjustments to none. But that preference is based on THEORIES so, unless you are 100% sure (as opposed to 99% or less sure) about the details of those theories, you should give some weight to the other series.

Here's the graph of series 4 from www.bls.gov:



This is BLS series id LNU02000000 "(Unadj) Employment Level, in thousands".
Note that this series' employment changes from each Dec to Jan are affected by "changes in the population controls".

[Added Aug 9: If your purpose is to understand what is driving our labor market (as opposed to making a declaration as to whether employment went up or down), the seasonally unadjusted household survey may be the most informative. Tune in this wednesday morning when I explain more....]