What’s wrong with upside-down Keynesianism?

October 17th, 2018 at 5:00 pm



I’ve got a piece in today’s WaPo focusing on how the tax cuts have broken an important, fiscal linkage between budget deficits and the strength of the economy.

“When we close on full economic capacity, as is currently the case, tax revenues as a share of the economy should significantly rise, and deficits should fall. Instead, revenues have come way down, and deficits have climbed.

Why is the deficit 17 percent higher than last year, especially when the economy is growing faster, and unemployment is lower?

It’s primarily because the tax cuts have significantly reduced the amount of federal tax revenue the economy will spin off for any given growth rate. Increased spending also played a role but not as large a one as the tax cuts…

Consider these numbers. Using data back to the mid-1940s, I calculated the average deficit as a share of GDP over every year since the late 1940s that the unemployment rate was lower than or equal to 4.5 percent (it’s currently 3.7 percent). That average is -0.4 percent, as opposed to the -3.9 percent noted above for 2018. By the way, if I take this and last year’s deficit (-3.5 percent) out of that average, the result is a small surplus (0.1 percent).

This figure below shows this heretofore tight correlation between deficit and unemployment rate. To make the relationship easier to see, I’ve inverted the unemployment rate, so e.g., 4 becomes -4. When the economy tightened up, deficits used to come down. But the circled area at the end of the figure shows that under today’s fiscal policy that is no longer the case.”

You can see, in the circled part at the end of the figure, that the two lines were tracking each other as usual (meaning the economy was strengthening and the deficit was falling) but changed course in just the past few years. My WaPo piece gets into some details of how this relates to the tax cuts/spending increases and what should be done to rejoin the lines.

But this piece is on the economics and political economy of what’s going on in that circle; what one might call “upside-down Keynesianism” (UDK), or stimulating an economy that’s already closing in on full employment.

Before I get into the analysis, however, in case there’s anyone here unfamiliar with my rants of last year, I should clarify that this conversation abstracts from the fact that the Republican tax cut is a regressive, wasteful mess of tax complexity that is already exacerbating income and wealth inequality while opening up new loopholes that will promote tax avoidance and evasion until it is reversed. But this post isn’t about that. It’s about the macroeconomics of stimulating an already strong economy, not the composition of the stimulus.

What are the upsides and downsides of UDK?

The most obvious upside of UDK in the current economy is the extent to which the stimulus—deficit spending on tax cuts and spending programs—is pushing the economy closer to full capacity than would otherwise have occurred. In fact, there is both theoretical and empirical support for this upside.

First, “secular stagnation” argues that structural factors—inequality, aging demographics, persistent trade deficits—prevent the U.S. (and other advanced economies) from achieving truly full employment absent a push from “non-market” sources, such as fiscal stimulus or accommodative monetary policy.

The other theoretical support for UDK’s upsides is that economists must admit that we do not know our stars (u*, y*, r*)—the economy’s capacity indicators (the lowest unemployment rate consistent with stable prices, the level of potential GDP, the neutral interest rate)—within a policy-relevant confidence interval, and we’ve generally erred on the side of caution.

Therefore, stimulus at alleged full employment can help achieve actual full employment. In fact, the next figure shows that as actual unemployment has fallen well below the Fed’s estimates of u*, inflation is just now, after years of downside misses, hitting their 2 percent target (FWIW, I recently wrote up a related analysis which argues that their u* is about right; it’s just that inflation is really well anchored; as far as UDK is concerned, the upside is the same).

Empirically, even if we accept that standard estimates of u* (e.g.) are not too high, actual unemployment has been above the CBO’s u* for two-thirds of the quarters since 1980. In other words, the U.S. labor market has been slack far more often than not, a huge market failure, a significant factor in weakening worker bargaining power over these years, and a strong case for pushing beyond conventional measures of full employment.

The most obvious downside of UDK is overheating. Though the inflation line in the previous figure shows little evidence of such pressures so far, these dynamics can be non-linear, as long-dormant correlations can reawaken at high capacity levels. In their new World Outlook chapter assessing current risks, the IMF worries that since “the US economy [is] already operating above potential, expansionary fiscal policy could lead to an inflation surprise, which may trigger a faster-than-currently anticipated rise in US interest rates, a tightening of global financial conditions, and further US dollar appreciation, with potentially negative spillovers for the global economy.”

The IMF are worrywarts about such developments, but economist Dean Baker, with whom I frequently collaborate on ways to gin up more demand, is clearly not. Yet, even he recently said that ”…we are likely getting close to full employment, so we probably don’t want too much larger of a deficit.”

But there’s a less obvious downside to UDK, one I raised in the WaPo piece and the one which concerns me most. Remember, Keynesian interventions are temporary injections of deficit spending to get over a negative, macro shock. But the tax cuts are intended to permanently damage our revenue-base, and as such, they are merely the latest installment of the Republicans’ longer-term agenda to never raise, but always cut, federal taxes. As I argued in the WaPo, the dangers to this low-revenue path strike me as acute and threatening in real time:

“Based on our aging demographics alone, we’ve long known that we will need more revenue over the next decade, not less. Add in geopolitical threats, climate change and the damage from increasingly intense storms (which is tied to the warmer climate), infrastructure, the need to push back on poverty and inequality, counter-cyclical fiscal policy that will be needed for the next downturn, and, it’s not hard to understand why a rising deficit at full economic capacity is so ill-advised. That is, unless you’re being paid not to understand these fiscal realities.”

In this regard, one of the biggest threats from the tax cuts that must be considered, even in the context of UDK, is the role it plays in emptying the Treasury’s coffers so that Republicans can point to all that debt as a rationale for cutting social insurance and safety net programs.

Keynes v. Laffer

If I’m correct about that threat, UDK may be a misnomer, as Keynesian stimulus is by definition temporary and, though they made some of the their tax cuts temporary for budget scoring purposes, the advocates of the cuts want them to be permanent. And yet, if you look at any credible economic projection, you find Keynesian dynamics in the forecasts: as fiscal stimulus fades in late 2019, as shown in the next figure, the forecasts expect GDP growth to slow and unemployment to rise. This expected reversal in fiscal impulse is behind the Fed’s latest forecast, which has real GDP up 3, 2.5, and 2 percent, 2018-20.

Source: GS Research

Advocates of the tax cuts, however, view this forecast as wrong, as it discounts Laffer effects, wherein the cuts allegedly pump up supply-side variables—capital investment, productivity, labor supply—such that GDP moves to a permanently higher growth path (roughly 3 percent as opposed to 2 percent).

In other words, barring another round of significant deficit spending, which, ftr, I would not rule out, in a few quarters we’ll have a real-time, cage-match between Keynes and Laffer.

If GDP slows from its current underlying, short-term trend of around 3 percent to its pre-tax-cut trend closer to 2 percent, as the forecasts predict, Laffer loses…again. To be clear, as the WaPo piece argues, he and his disciples have already lost on the assertion that tax cuts pay for themselves, but that was never even remotely believable.

There is, however, a wild card in play, one I’ve written about under the rubric of the FEPM: the full employment productivity multiplier. This idea, for which there’s suggestive evidence, is that in slack economies, firms can maintain profitability without being particularly efficient. However, at chock full employment, and especially with anchored inflation expectations, rising labor costs are a disciplining mechanism, enforcing the discovery of efficiency gains if firms are to maintain profit margins. These dynamics can also drive more capital investment that would not have been “necessary” in slack labor markets.

Thus far we haven’t seen much to suggest the “sugar-high” forecasts are wrong. Business investment is up, but no more than you’d expect at this point in the recovery. Productivity growth is still too low.

Moreover, even if there is a productivity multiplier that gets tapped as we close in on full employment, it will be hard to assign victory to either side. In theory, Keynes wins again, as the capital investments in the FEPM model are not a function of the lower, after-tax cost of capital as much as a Keynesian accelerator story, where full-employment-driven job and wage gains fuel stronger consumer demand. In response to higher demand and the desire to maintain margins, firms ramp up their investment. But empirically, it will be hard to tell one story from the other.

In sum, and abstracting from the awful regressivity, complexity, and aspirational permanence of the tax cuts, UDK has clear upsides. I don’t think the unemployment rate would be as low as it is right now without it. Its almost-50-year low is finally starting to generate wage gains that will reach those who have heretofore been left behind in this expansion, even in year nine.

UDK also invokes the risk of overheating. Yes, the Fed has lots of firepower to deal with that if need be, but the IMF could be right and we could end up with a hard versus a soft landing.

But at the end of the day, my biggest concern is less about UDK and more that we’re not really talking about temporary stimulus. We’re talking about a permanent reduction in revenues, sought by hard-right conservatives who have been gunning for social insurance programs forever, and thus view this current strategy as a twofer to both enrich their donors while starving the Treasury.

What does “full employment” mean in the era of anchored inflation expectations?

October 15th, 2018 at 6:00 am

There’s a new analysis out by a group of economists from the Goldman Sachs economic research team that raises the question of what the concept of full employment means in an era when the central bank expends significant and successful efforts to anchor inflationary expectations.

The paper (which lives behind a paywall) uses four distinct techniques to to derive different estimates of the natural rate of unemployment, aka u*, aka the lowest unemployment rate consistent with stable inflation. The results range from 4 to 4.8 percent, which, as their figure below shows, fit well within other commonly sourced versions, including CBO (4.6 percent) and the Fed (4.5 percent). Note also their forecast for the jobless rate to get to 3 percent (!) by the end of next year. As they say in the old country: “from their lips to Keynes’ ears.”

Source: GS Research

For those who are interested, I’ve pasted in their table at the end of this post which briefly describes their four methods and results, but anyone who still subscribes to Phillips Curve notions of full employment might well ask: “why, if the current unemployment rate is below all these estimates of the natural rate, is there so little acceleration in core inflation?” Here is their conclusion (my italics):

“The actual unemployment rate is already below all of these estimates, and we expect it to fall all the way to 3% in early 2020.  But…such a scenario is not as dramatic as it sounds: with well-anchored inflation expectations, a labor market overshoot is likely to result in above-target inflation, but not persistently accelerating inflation.  This is an important difference with the late 1960s, when labor market overheating led to runaway inflation.”

Now, one of my constant refrains these days is that the Fed’s 2 percent is an average target, not a ceiling. Given the many years of downside misses on the inflation target, we’re long overdue for a period of “above-target inflation.” So, if these analysts are correct, as I suspect they are, then this overshoot is a feature, not a bug.

If that’s true—if an unemployment rate significantly below the Fed’s natural rate estimate means we get something we very much want (super tight labor markets) as opposed to something we very much do not (spiraling inflation)—then, conditional on inflationary expectations remaining well-anchored, being above full employment is precisely where we should aspire to be.

This is an awfully different economic model. In this model, go ahead and bang out estimates of the natural rate if you must, but recognize that (u-u*)<0 (actual unemployment below your estimate) is not a signal to hit the growth brakes. In this model, your real job is to watch the indicators of inflation expectations and realizations. And if, as is currently the case, they seem well-contained, then there’s no reason to overreact.

One interesting aspect of this conclusion is that those of us who have criticized the Fed’s anchoring as being a drag on demand relative to more flexible inflation or level targeting (I’m talking to you, Beckworth) might consider that at least under this framework, solid anchoring enables stronger demand and lower unemployment than would otherwise prevail.

However, for those who benefit the most from such low unemployment to realize such gains, the members of the FOMC would have to recognize these dynamics. I actually think Chair Powell does, and he’s not alone, but there are others who look at that 3 percent at the end of the figure above and think, “not on my watch!”

Source: GS Research

Unemployment hits a 49 year low as jobs/wages stay on solid, hot-but-not-too-hot, trend.

October 5th, 2018 at 9:41 am

The nation’s payrolls grew by 134,000 jobs last month and the unemployment rate fell to a 49-year low of 3.7 percent. While the jobs number came in well below expectations, that should not be considered bad news. First, Hurricane Florence may have slightly dampened monthly payrolls (see below discussion of the impact of hurricanes on the jobs data). But more importantly, upward revisions for the prior two months’ data reveal a trend over the past three months of 190,000 jobs per month, a solid pace of job gains for this stage of the labor market recovery. Hourly pay was up 2.8 percent, just slightly off last month’s cyclical high of 2.9 percent.

The decline in unemployment is “real,” meaning it occurred through fewer unemployed persons as opposed to people leaving the labor market. The black unemployment rate, at 6 percent, is only slightly higher than its all-time low of 5.9 percent earlier this year. At its peak in the depths of the last recession, black joblessness was almost 17 percent, meaning it is now down about 11 percentage points, close to twice the decline of the overall rate (about 6 points) and a critical reminder of the disproportionate benefits of full employment to less-advantaged workers.

To get a cleaner take on the underlying pace of job gains, our monthly smoother takes averages of the number of jobs added over 3, 6, and 12-month periods. In all three cases, the bars are hovering around 200,000, a very solid pace of job gains for year nine of the economic expansion.

In September, as noted, hourly wages for all private-sector workers grew 2.8 percent on a year-over-year basis, and 2.7 percent for lower-paid workers (the 80 percent of the workforce that’s in blue-collar or non-managerial jobs). The figures below show that the trends for both groups have gained some speed as unemployment has come down. Such acceleration, which, it must be underscored, is occurring at a smooth at not at all unusually quick pace (compare it to the speed of wage growth decline in the downturn), has been long-awaited and should be welcomed, not feared as inflationary. As a figure below shows, there is no evidence that wage growth is bleeding into higher (core) inflation.

Turning to the sectoral jobs data, there’s perhaps some Florence effects in the 17,000 loss of leisure/hospitality jobs (which include hotels, waitpersons, food prep) and the 20,000 jobs lost in retail trade (“brick and mortar” stores). But this a relatively small losses in noisy, monthly data, and the BLS series on those who missed work due to weather last month showed a relatively small jump, especially relative to much larger storms in the past.

One data point in the report that is less positive is the flat, recent trend in the employment rate for prime age workers (25-54), a measure of labor demand and a source of “room-to-run” claims made about the current labor market. That is, as the figure shows, the prime-age employment rate, while still not quite back to its prior peak before the recession, has been steadily climbing back, signal that the strong labor market was pulling more workers in from the sidelines, and suggesting greater labor market capacity than most economists presumed. However, if the recent flattening (hard to see in the figure, but the rate has wiggled around 79.3 percent for about seven months) persists, it will suggest this source of labor supply could be tapped out. Moreover, as I have written previously, and as you can kind of see in the figure if you squint at it for awhile, this indicator tends to flatten before a recessions. However, whether this is a pause in its upward trajectory or a stopping point is yet to be seen.

The last figure is in here to underscore the point made above about how core inflation remains “well-behaved,” even as the unemployment rate is close to a full point below the Fed’s “natural rate” (its estimate of the lowest jobless rate consistent with stable inflation). Even as wage growth slowly trends up (the yellow line in the figure), core inflation is merely at the Fed’s 2 percent target. Importantly, since the target is an average, not a ceiling, and given how long inflation has missed the target to the downside, this figure suggests the Fed chair Powell’s view of a strong but not overheating labor market/economy is correct.

Hurricanes and the jobs data.

The jobs data information comes from two surveys. Payroll and wage data are drawn from the Establishment Survey, while unemployment and much other utilization information comes from the Household Survey. In both cases, the surveys’ “reference period” is the week of the month that includes the 12th, meaning Florence occurred during the survey week.

This can affect the data in various ways. First, regarding payrolls, even if someone can’t get to work, if they are paid for work, they are considered to be on the payroll and counted (of course, if a hurricane prevents the submission of data from firms to the Bureau, this could interrupt this process). However, if, due to the storm, someone misses work and is unpaid, they are not counted that week, and this dynamic often suppresses employment, especially in severe cases.

Since low-wage workers are more likely to fall out of the survey in this manner, hurricanes can artificially boost average pay, by temporarily taking lower-paid workers out of the sample. In fact, this is believed to have occurred last September during Hurricane Irma, a storm that disrupted a much larger area than Florence.

The household survey is a phone survey to individuals, so it too could be disrupted by evacuations. Moreover, its employment concept is different from the Establishment Survey as it counts those temporary absent from their jobs due to weather as employed, even if they are not paid. In fact, the Bureau tracks this number, which, as can be seen here, always spikes during big storms.

As noted, a few sectoral job losses suggested a Florence impact, but the BLS reported that survey response rates did not seem much diminished by the storm, so its impacts in today’s report are probably small. Also, such impacts are temporary and storm-induced losses in the labor market tend to be quickly reversed. In addition, repairing the damage from the storms can often show up as a plus for employment, e.g., construction, in later years.


The USMCA is not a free trade deal. That’s because there are no free trade deals.

October 1st, 2018 at 5:43 pm

I’m doing my best to work through the text of the new NAFTA, now called YMCA, I mean CAMUS, no…wait…USMCA!

Trade agreements make for a dense read…here’s a snippet from the Ag chapter, one of the 34 chapters, followed by “annexes” and “side letters”:

“Parties recognize that under Article XI:2(a) of the GATT 1994, a Party may temporarily apply an export prohibition or restriction that is otherwise prohibited under Article XI:1 of the GATT 1994 on foodstuffs to prevent or relieve a critical shortage of foodstuffs, subject to meeting the conditions set out…”

It’s just saying that in a food emergency, a party to the agreement can restrict food exports and remain in compliance, but I print that little example to make a larger point: There is no such thing as “free trade” and there are no “free trade agreements.” FTAs are mythical creatures.

In the real world, agreements exist between trading partners that comprise hundreds of pages of rules by which they will engage in trade. These rules can be straightforward, like the one above, or seemingly obscure (have a look at the side letter on cheese names; as far as I know, I’ve never had Emmentaler cheese, but if this deal becomes law, I can enjoy it free of tariffs!).

Trade rules can favor workers, like the new language in support of independent Mexican unions, or investors, like the dispute settlement procedures that are somewhat weakened in the new agreement. It’s all about who got a seat at the table when the deal was drawn up.

Before getting into some weeds on the new deal, I underscore this point so you don’t confuse trade deals with trade, and especially with the trade balance (I’m talking to you, Trump). The CBO, which certainly doesn’t have a protectionist thumb on the scale, recently pointed out that estimates of the impact of “trade agreements on the U.S. trade balance are very small and highly uncertain.” The flows of goods, services, money, and financial assets will continue apace whether or not this deal is approved.

That doesn’t mean trade deals don’t matter. Instead, it means they have a lot more to do with who wins and loses from trade than whether cargo ships continue to sail the seas and trucks go back and forth across the borders of North America.

Many journalists have done nice work unpacking what’s in the deal, so I’ll just highlight some parts of interest.

Is this really that different a deal than the NAFTA? There are, as noted, differences, but they are not big enough for Trump to credibly claim that NAFTA was a horrible disaster while the USMCA is incredible. As noted, I don’t see the deal, should it become law, changing the flows of goods, services, money, or people in ways that would change economic outcomes. New rules for Mexican auto production, including requirements for a) more production in the trade zone and b) a subset of Mexican auto workers to get paid $16 per hour (2-3 times their current wage), have led some to predict higher car prices.

It’s possible, but nothing is that simple in international trade. Auto exporters trying to sell cars here in the US can forgo the duty-free benefits of the trade deal and pay the existing (WTO) auto tariff amounting to a mere 2.5 percent. And exchange rate movements can eventually swamp such price differences, especially as the $16 is not indexed to inflation.

It also must be underscored that Mexico has a lousy record of implementing and enforcing labor rights, so these changes—ones I view as clear improvements in the deal—will require close monitoring. I don’t trust this administration to follow through on that.

What’s good, what’s bad in the deal? Some of the auto requirements just noted are intended to reduce the trade-induced wage arbitrage opportunities that have long hurt production workers exposed to export competition.

The new rules on Mexican unions are also a positive change. It’s not just that these rules potentially make it easier for Mexican workers to form unions. It’s that the unions could finally gain some true independence, as too often, Mexican unions have been Potemkin unions, fronts for management to impose harsh labor conditions with impunity. Again, this is where enforcement is especially crucial.

Though Canada got to keep one part of the dispute system they wanted—the one they use to fight over dumping and countervailing fees—the ISDS process (investor state dispute settlements) will be phased out for Canada and limited in Mexico. Trade expert Lori Wallach and I have discussed the serious problems with ISDS, as it provides a mechanism by which corporate rights could preempt sovereign rights at the expense of taxpayers. But before we get too excited about this change, we need to learn more about the carve-outs from the new rules. If it’s too easy for favored industries to prosecute investment cases using the old NAFTA and TPP tribunals, then this change won’t be meaningful.

What’s bad in the new deal is the same stuff that was bad in the old one (same with TPP): protectionist measures that further belie the idea of “free trade.” I’m talking here about patent and IP extensions that American lobbies like Big Pharma insist on as the price of their support (Dean Baker has some details here). Surely those complaining about higher car prices based on labor protections should be lodging the same complaints here.

Will Congress Approve the Deal?

Which Congress are you referring to? If the D’s take the majority in the House, look for a whole lot of activism around improving enforcement and blocking ways for countries to get around any labor-friendly rules in the proposed deal. For one concrete example, they’re want to raise that 2.5 percent tariff on auto imports to block companies from skirting the new, higher origin requirements in the USMCA.

But aside from those details—and who knows, I could see Trump and his trade rep, Bob Lighthizer, getting behind the House D’s on preserving those protections—and based on what we know now, I suspect a majority in both chambers will want this deal to go through.

Min wg panel model

September 26th, 2018 at 10:56 pm

I’ve got a piece in the WaPo today that presents findings from the model described in more detail here. What follows is for those who want the technical details of the data and model. As is unlike my wont, I make no effort to explain in layperson’s terms.

The model is a panel of the 50 states plus DC from 1979 to 2017 of wage, labor market, and minimum wage data. The results derive from fixed effects, panel regressions of the change in the log of the state real, 10th percentile wage on state unemployment and the change in the minimum wage variable. For this variable, I used the ratio of the state minimum wage in state i, year t, divided by the federal minimum wage in year t. Standard errors are clustered at the state level (results below).

The article mostly focuses on the first two rows of results below: outcomes for the log change in the real 10th percentile wage. But I also include the same regression for the 90th percentile wage, as a confirmation that the minimum wage variable is picking up variation in low wages. In fact, it is insignificant for the 90th percentile wage.

The WaPo article describes the results from a few simulations of the model over the past few years. To capture the impact of falling unemployment, all else equal, I held the unemployment rate at its 2015 level for each state in 2016-17, while the other variables took their actual values. I take simulated real wage levels and compared them to the path of actual wages.

The same procedure was followed for the change in the minimum wage ratio (state over federal) i.e., I did not allow it to go up in states that raised their minimum wages in 2015 or 2016. I then compared these simulated wages for states where the ratio rose to the actual wage trends for those states (which included any wage effects for the higher state minimum wage).

The full model for the wage results below had 1,938 observations and R-sq of 0.43. Restaurant employment data were only available for 21 states since 1990 (518 observations).

Wage data were derived by the Economic Policy Institute from CPS ORG files; state unemployment is from the BLS; state employment data were from the BLS CES survey; minimum wage data from Ben Zipperer (thanks to Ben for helpful comments, though any mistakes are mine).

Some key results follow (***=significant and <1% level):