Due to the coronavirus, unemployment is high. It was fairly low before, and seemed poised to go even lower. Should American policymakers aim to go back - or to go further?
What even is full employment?
The labor market has three core indicators: unemployment, employment, and participation. Unemployment measures how many people don’t have jobs but are looking for one. Employment measures how many people have jobs. And participation measures how many people are employed or unemployed - i.e. how many people are actually in the labor market. We’d normally claim that full employment is when unemployment is at the lowest it can be: what’s known as the natural rate of unemployment, the “remainder” from the maximum amount of employment the economy can have. When unemployment is high, wages also grow slowly, because workers don’t have much of a choice with regard to wages and because they can’t afford to stay unemployed for long.
The key item here is monetary policy. Central banks normally have two mandates: full employment and low inflation. They’ve actually been fairly good at the latter… too good, in fact, because a lot of countries (including the US) haven’t been able to actually meet their inflation targets since the Great Recession. The Federal Reserve normally does an estimate of what the unemployment rate should be when the economy is strongest, and when they hit that rate they stop doing stimulus. Keeping unemployment as low as possible helps them achieve both their mandates because high wage growth means that prices also rise, both because of the higher cost of labor and also because of stronger demand.
Why would there still be unemployment even with a very strong labor market? An initial source of unemployment is frictional: people who recently quit their job, or whose company went under. These are people who are looking for a job and probably will find one, so even under a labor market that “covers” every worker, there would still be unemployment simply. It’s not easy to estimate it precisely, but it’s frequently assumed to be in the 2-4% range.
Secondly, there’s another relevant kind of unemployment: cyclical. The economy sometimes goes bad for various reasons, like a housing crisis or a pandemic, and companies might not be able to afford to retain all their workers or stay in business. If companies expect the economy to, overall, get “bad”. Normally, we’d expect monetary policy (and fiscal policy, to some extent) to prevent cyclical unemployment by smoothing out business cycles.
A third reason might be regulations. Companies are faced by many laws that force them to, for example, make their workers work for a certain set of hours, get certain amounts of days off, and especially be paid a specific wage. Because all of these have costs, then they might pile up and cost the economy jobs. The biggest regulation could be the minimum wage: because workers might be paid more than what the market demands, then perhaps this “price floor” hurts workers. Of course, there isn’t a lot of evidence for the minimum wage hurting employment all, and the increase in the minimum wage in 2009 (from 5.75 dollars to 7.25) wasn’t that big and was probably not that high.
There’s also the issue of costs of searching. For companies, hiring workers takes up a lot of time and training them costs a bunch of money - so they might not hire as much simply because of those costs. And to workers, looking for jobs costs time and money as well, so the pay they might receive gets “adjusted” by the costs. Since we assume that workers have a minimum wage they’d be willing to accept, then if market wages are too low after accounting for search costs, they might not take any jobs. And since companies know this, they compensate workers at above-market clearing levels, so the total number of people they hire is just a little bit smaller.
A final reason is quite puzzling: companies might be paying their workers too much. This sounds controversial, so let’s explain. Imagine that businesses don’t just need a given number of workers to earn a given amount of money - you also need them to put in all of their effort. This effort depends (let’s say, exclusively) on wages - and there’s a single wage level, for all workers, that gets them to put in that effort. Because you want workers to actually put in the work, you pay them that amount. And since it’s higher than the market wage, total employment is lower.
That’s called the efficiency wages theory, and you pay workers “too much” to be more efficient. It doesn’t really account for a weird fact: pay scales remain constant across industries and firms, which they maybe shouldn’t. What this means is that, for instance, a secretary in a big company makes roughly the same fraction of her boss’s salary as one in a smaller one, and the same happens between similarly sized companies in different sectors. But because workers are too different, this might not work. The reason is also efficiency wages - if workers also care about how much other people earn, they might feel like they’re being paid “unfairly” and put in less effort. If workers have similar preferences across industries, then this pattern should be roughly similar, and the wage level itself (not relative wages) is set by productivity, demand for the product, qualifications required, etc.
Drowning in a bathtub
Normally, forecasts of unemployment reductions show that it goes down depending on how strong the economic recovery is - i.e. the more GDP growth, the faster the economy recovers. This is known as Okun’s Law; however, it doesn’t really seem to be true, because unemployment tends to go down at a more or less steady rate across multiple recessions. This counterexplanation is called the bathtub model: workers going in (the “faucet”) or workers going out (the “drain”). The model is mostly just basic math: when the economy does badly, more people flow out of employment than into it, so employment shrinks.
And more people flow into unemployment than out of it, so it grows… right? Not necessarily - since unemployment is a fraction of the participation in the labor force, participation going down might mean unemployment doesn’t go up. When the economy gets better: does unemployment decrease because people are flowing out of it, or because people flow into it at lower rates? This opens up the question: does participation go up or down during recoveries? And is it because people join the labor force, or because people stop leaving?
The reason people would leave the labor force is what’s known as the discouraged worker effect: people don’t think they’ll find jobs, so they don’t even bother looking for them. There’s also an additional worker effect, where participation increases because people need jobs more than ever, but it’s small outside of poorer countries. The Federal Reserve of San Francisco looked into it, and found something quite unintuitive: the problem is people leaving the workforce at higher/lower rates. The discouraged worker effect seems to completely dominate an “encouraged” worker effect, i.e. people enter the workforce at mostly steady rates (due to demographics, structural labor market conditions, etc.). This explanation means that exit rates are responsible for 96% of labor market participation changes, and entry rates for just 4%.
But here’s the thing: this result is mostly caused by bad data management. Let me elaborate. The distinction between people who don’t have jobs and are looking for them (unemployed), and people who don’t but aren’t (nonparticipants, or discouraged workers) isn’t very clear. Because it tends to rely on very arbitrary definitions on what counts of looking for work, people tend to sidestep in and out of the labor market for multiple consecutive months - known as NUNs, because they go from “Nonparticipant” to “Unemployed” to “Nonparticipant” again (the reverse would be an UNU).
Since it’s not especially probable that someone just pops into the labor market for tea and leaves inmediately, you’d have to adjust the data and consider that the medium letter is just wrong - so NUNs are just NNNs, and UNUs are just UUUs. Something similar would happen between people who are N one month and then U for several more - they were clearly an U the whole time. If you actually adjust for this, then the explanation gets flipped on its head - entry explains 97% of participation, and exit just 4%. Instead of entry being flat between 2015 and 2020 and exit gradually falling, you get the opposite - entry dropping during the Recession, and gradually rejoining while exit remains flat. If you look at annual figures (i.e. averages of the past 12 months), the same thing happens without needing wacky adjustments. And also in flows in and out of employment - you know, the one indicator without problems with the fractions.
Maximum employment
As previously said, unemployment and participation are kind of crappy measures. So let’s focus on employment - maximum employment is when the employment rate is the highest it can be. But how high is that?
Normally, the Federal Reserve looks at the composition of the labor force, looks at the highest employment rate for each demographic, and says “well, adjusting for demographic changes, then full employment is X%”. The Fed also only looks at recent figures - unemployment for some group might have been 1% in the 60s, but that was so long ago it barely matters. For all intents and purposes, the maximum employment rate is the one from 2005-2006, right before the Great Recession.
There’s two big problems with this. The first is that the early 2000s were called the “jobless recovery” from the dot com crash, because the labor market recovered so poorly. So employment was even higher just five years later - not a great start. You can see there was a big gap between the employment rate of 2000 and the rate during the mid 2000s. The Federal Reserve actually claimed to be at full employment in 2015 - halfway back to the pre Recession peak. Just assuming that lost growth (or lost jobs, or lost anything) can’t be gained back is madness - and it shows in estimates of GDP too.
And there’s even more: why would you assume that the maximum employment rate for each demographic is the same as for that demographic and not for the highest “achieving” one? Women, for example, are penalized in the labor market for having children, and most “NEETs” (young people not in education, employment, or training) are actually women who take care of family members. Racial minorities have the same problem too - why would you assume that, if minorities’ job prospects are grimmer, you wouldn’t be able to get past that? Ten percent of Black people were unemployed in 2005 - how is that full employment?
Saying “everyone should have the same job prospects as white men” sounds a bit reductionist, but saying “if discrimination (or whatever you want to call it) is keeping your job prospects down, just suck it up” is even worse. In fact, getting every demographic to “white men” employment levels would permanently increase employment figures by some 20 million people.
The danger of complacency
There was a pretty nasty recession from 2007 to 2009, known as the Great Recession. Now, while the recession ended in 2009, its effects were longer-lasting: real GDP didn’t return to pre-recession levels until 2015, the output gap wasn’t closed until 2018, and the labor market didn’t recover to its 2006 levels until 2019. If you compare to the historical trend beforehand, this performance was even worse - real output per capita was nearly 10,000 dollars short per person before the pandemic than what 1946-2006 trends would have suggested. And if you look at actual Fed projections of employment (quite a contentious issue), the figure is actually much higher: 70,000 dollars in lifetime income losses.
Getting each demographic to its pre Recession, late 1990s peak (something achievable) would have an additional ten million jobs. Why was the recovery so slow? The US started on the right foot, with fiscal and monetary stimulus. The EU, in comparison, did barely anything of the sorts (particularly monetarily) and their unreasonably tight policies strangled the recoveries of basically all its member nations.
This was caused, primarily, by bad monetary policy. The most important thing to consider is that, well, people were very wrong about the economy, once again (like the 20s and 70s). At the heart of the story there is a political problem: Obama didn’t think monetary policy was very useful (as he said to CEA Chair Christina Romer, “it’s shot its wad”), and he didn’t have the political capacity for fiscal stimulus starting in 2011, since a very conservative GOP faction took over half of Congress. Plus, his apointees to the FOMC (the board that runs the Fed) weren’t very good, and their monetary policy decisions were bad - the Fed raised rates and stopped stimulating the economy in 2013, when unemployment was over 6%.
The Federal Reserve also relied on bad models of what the lowest possible rate of unemployment (the NAIRU, aka. the non accelerating rate of unemployment), which it located between 6.5% and 5%. That’s right - the Federal Reserve thought that the lowest unemployment could go was nearly 7%. While recovering from a recession
The Fed’s idea wasn’t just that it was doing all it could, but also that it had to stop doing anything lest inflation rise. This was because the Fed started tapering (i.e. cutting stimulus) when forecasts expected above-target inflation, not when there was much evidence for it. Their estimates of the natural rate weren’t good and ignored that labor participation was low- if the natural rate can be understood as being defined under full participation too, then having less participation than that over-estimates the NAIRU. Participation is how many people either have jobs or are looking for them, so ignoring it risks declaring full employment when people who would be unemployed don’t even bother to look for jobs because they don’t think they can find them.
The Federal Reserve triumphantly declared to be at the NAIRU at basically every point between 6% and 3.5% unemployment in the post-Obama era, and was flabbergasted when inflation was still never in the target range. Meanwhile, their less aggressive policies led to a really strong labor market- in fact, the strongest since before the Recession. In the words of macroeconomist Jon Steinsson:
Whether you look at the 1980s expansion, the 1990s expansion, or the 2010s expansion, the unemployment rate, if you just plot it, it just keeps falling. It keeps falling and falling and falling and it never levels out. Maybe at some point it would, but one view of that is that we’ve just never gotten to the point where we have true full employment.
I would like to point out something in the Fed’s defense: there wasn’t all that much they could do, at least not with their current toolkit. Their interest rates were super low, so low in fact that they hit what’s known as theZero Lower Bound: a point at which interest rates are so low they can’t go any lower. The Fed’s normal action is to buy government bonds and other high-quality assets to bring down interest rates, which they can’t do if they’re basically zero. But the Fed could have done plenty of things: the Bank of Japan became really aggressive in purchasing various securities and even regular bonds to stimulate the economy, andit worked. Michael Woodford, the guywho invented the whole concept of the ZLB in the first place,was not satisfied with the Fed’s work in 2012, from more aggressive commitments to even NGDP targeting. The Fed could have done what Japan did, could have heeded Sumner and Woodford’s advise, or could have even just printed out money and handed it out to regular people - known as helicopter money and invented by Milton Friedman.
Making full employment pay
Let’s circle back to the benefits of full employment. We’ve mentioned that wages grow faster when unemployment is low (and, logically, slower when it’s high). But whose wages, precisely?
The wages of low-income workers, specifically. Every one point drop in unemployment results in annual growth rate for the bottom 10% of workers that is 0.5% faster. For workers around the median wage, it’s 0.4%; and for the top 5%, it’s just 0.3%. These results are symmetric, by the way, so a 1% increase slows wage growth by the same amount, and increases in the employment rate have similar outcomes in term of wage growth.
This relationship actually became weaker after the Great Recession, because of what’s known as "downward nominal wage rigidity”. That’s quite a mouthful, so let’s break it down: rigidity is when something can’t be adjusted - for example, companies can’t change their prices because it would be too much of a hassle to customers. And this rigidity affects nominal wages, i.e. the amount of money in a non-inflation adjusted paycheck. Finally, it’s downward because it means wages can’t be cut down - normally you’d expect some workers to take pay cuts, but they generally don’t.
Because of the demographics involved, i.e. that minorities and women make lower wages and are unemployed at higher rates than other groups, then a tighter labor market would benefit them more than other groups. So reducing unemployment, and increasing employment, wouldn’t just increase wages - it would increase them most at the bottom and close earnings gaps a bit. In fact, that is what was seen in 2018 and 2019, when the labor market started tightening significantly. That a tighter labor market means lower wage inequality and higher upward mobility isn’t new - it’s been known since the 1970s (by Arthur Okun, of Okun’s Law fame).
And back then it wasn’t even the tightest it could be - unemployment was much lower than the Fed thought it might be, and their preferred measure of inflation was still below target. The normal measure of tightness is the ratio of vacancies to unemployed people, and before the COVID pandemic it was at its highest levels ever. However, since a lot of job searchers are actually employed, and the participation issues measured above, the actual level of tightness was lower than expected - although still at early 2000s levels.
A final issue is that, for a variety of reasons that I don’t want to get into (you can read more on my post about the topic), the labor market has become less competitive over time. Because employers hold many cards, then sluggish growth lets them keep wages too low for employees. The same logic as a monopoly charging too much for its products applies here - so it’s known as monopsony, a monopoly of supply. In fact, if wages were fully set by productivity, the median worker would earn 17% more - meaning that concentration has real costs. If, though, employers have to compete for workers (and not the other way around), then you can run around monopsony.
Forcing companies to replace man with machine
Now, you might argue, keeping wages too high might actually backfire, by making labor too expensive and incentivizing companies to replace workers with machines. This isn’t just a valid complaint - it’s actually something you want to incentivize.
Why? That sounds like a terrible idea - dooming everyone to mass unemployment because of technical advances. But the kind of technical advances aren’t the kind of thing that lays off millions of people, just that makes labor of a certain type of skill more productive. And, as I’ve mentioned before, US productivity has been in the dumps for a decade or so - so more machines would be a good thing, because it would get productivity up.
But how does full employment cause more productivity, and and how does technology benefit workers? Let me explain. Higher wages means labor is more expensive - so it’s incentivized to replace it with machines, holding productivity constant. This might have been why the Industrial Revolution happened in England and not elsewhere: it had really high wages (compared to China) and really low energy costs (compared to, say, the Netherlands), so replacing labor of equal productivity with energy was a solid winner. Now, the “equal productivity” part is super dicey, as well as the fact that maybe China needed a lot more labor per unit of output for various non-technology reasons. This theory is also shaky because of a bunch of other differences, like the widespread use of (very cheap) child labor, the role of market size, and the fact that China might not have had strong enough institutions for the Revolution t even occur. Something similar might have happened between Japan and the United States in the 1880 - 1960 period, based on the nature of innovations in agriculture that “freed up” labor for manufacturing.
Okay, we have the wages-productivity relationship. But what about the inverse? Everyone’s talking about productivity growing and wages falling behind, so it can’t be good, right?. Plus, even if workers actually got the full gains from productivity, how would we not have them just be for the ones who actually became more productive?
Well, productivity increasing and wages not keeping track isn’t particularly true. The key assumption here is that faster productivity means divergence - but that’s just not true. In fact, if you look into it, median total compensation (i.e. wages and other benefits, like bonuses or childcare) has tracked productivity very closely - an increase of productivity of 1% has led to increases in median wages of 0.7%, and of production/non-supervisory wages (i.e. the ones lower income workers make) by 0.4% to 0.7% between 1973 and 2016 - a time of significant labor market slack.
And why do we keep the gains from being pooled in just one sector? A foe turned friend: Baumol’s Cost Disease. The Cost Disease, discovered by William Baumol in 1966, consists of answering the question: why did wages for musicians increase over time if musicians aren’t more productive? The reason is opportunity cost: musicians would leave music to be industrial workers if their wages didn’t grow, so they “mooched off” the productivity gains of other sectors. And it’s called a “disease” because it affects the cost of things like education and healthcare, which get progressively more and more expensive because of high labor costs and tepid productivity gains of their own.
But one’s man disease is another’s man health: because productivity gains have to “spill over” into other sectors, then maybe a more competitive labor market boosts wages even further through this phenomenon. And because wages keep going up, firms innovate even harder - maybe even in services, which have had an abyssmal record productivity-wise so far.
Conclusion
Full employment, aka low unemploment, high participation, and high employment, is good. Not only because it’s more of a good thing and less of a bad thing, but also because it helps wages grow faster, making wage inequality less bad. It can also offset labor market concentration (which should be offset anyways) and might even help the US’s sluggish productivity back on track.
Now, it should be pointed out that “returning to mid 90s levels” might not even be full employment. The US had higher employment rates that similarly developed nations back then, but it stopped a long time ago - and European nations, which recovered much worse, actually surpassed it in terms of employment rates long ago. As the Bitcoin maniacs say “to the Moon!”.
Sources
Full employment
Bivens (2018), “The Fed’s current path might be leaving lots of money on the table unnecessarily”
Zipperer (2016), “Employment fell because of the Great Recession, not the minimum wage”
Lutz Hendricks, “Mortenson Pissarides Model”, Class slides, 2017
David Autor, “Lecture Note: Efficiency wages, the Shapiro-Stiglitz Model”, 2003
Akerlof & Yellen, “The Fair Wage-Effort Hypothesis and Unemployment”, 1984
Participation and unemployment
McCarthy, Potter, & Cee Ng, “Okun’s Law and Long Expansions”, New York Fed, 2012
Sahin & Patterson, “The Bathtub Model of Unemployment”, New York Fed, 2012
Ernie Tedeschi, “Participation and the Hot Labor Market”, Employ America, 2019
Maximum employment
Matthew Klein, “The Case for 38 Million More U.S. Jobs by 2031”, 2021
Joseph Politano, “The Fault in R-Stars: The Problems in Policymaking Based on "Natural" Rates”, 2021
Claudia Sahm, “Racism skews our beliefs about what's possible”, 2021
Alex Williams, “Potential Output: Little Explanation for a Big Number”, Employ America, 2021
The Great Recession
Matthew Klein, “Let's Overshoot”, 2021
Joseph Politano, “The Great European Undershoot”, 2021
Matthew Yglesias, “Obama's biggest economic policy mistake”, 2014, Vox
Dylan Matthews, “The Fed’s bad predictions are hurting us”, Vox, 2019
Mike Konczal, “How low can unemployment go? Economists keep getting the answer wrong.”, Vox, 2018
The benefits of full employment
Okun (1973), “Upward Mobility in a High-Pressure Economy”
Bivens & Zipperer (2018), “The importance of locking in full employment for the long haul”
Bivens (2021), “The promise and limits of high-pressure labor markets for narrowing racial gaps”
Abraham, Haltiwanger, & Rendell (2020), “How tight is the US labor market?”
Productivity
Noah Smith, “A virtuous cycle of worker power and technology?”, 2021
Matt Clancy, “Innovation gets (mostly) harder”, 2021
Brad DeLong, “Robert Allen: The British Industrial Revolution in Global Perspective”, 2012
Pseudoerasmus, “Random thoughts on critiques of Allen’s theory of the Industrial Revolution”, 2016
Summers & Stansbury (2018), “Productivity and Pay: Is the Link Broken?”
The Fed´s pursuit of Nominal Stability would do wonders for finding "maximum employment".
https://marcusnunes.substack.com/p/macroeconomic-patterns-and-stories