Thumb on the scale: correcting the CEA’s corrections re real wage growth.

September 6th, 2018 at 11:28 am

By Jared Bernstein and Larry Mishel

President Trump’s Council of Economic Advisers has a new piece out claiming to show that real wages are growing faster than has been widely reported. Their conclusion stems from numerous adjustments to Bureau of Labor Statistics wage data that otherwise show flat real earnings for most workers.

However, most of CEA’s adjustments, applied accurately, do not change the inconvenient fact that even amidst strong macroeconomic results and a tight labor market, real wage growth for middle-wage workers has been weak over the past two years. That may change, if falling unemployment triggers faster wage growth, but at this point, measurement tweaks make little difference to the conventional wisdom, at least over the short period covered by the CEA report. As former CEA chair Jason Furman tweeted yesterday, the report “does almost nothing to change our understanding of wage growth”

The CEA argues that by adjusting for the following factors, real wages grew at least 1.4 percent higher over the past year, as opposed to zero:

–adding fringe benefits;
–adjusting for the demographics of the job market;
–adjusting/reducing hours worked;
–using a slower growing price deflator.

We find that only the last point is valid, but that even with that adjustment, real compensation is still flat, largely because benefits, properly measured, have not been rising any faster than wages, contrary to CEA’s claims.

Fringe benefits: Even in the data used by CEA, the growth of benefits is hardly outpacing that of wages, so adding fringes doesn’t change the underlying trends. Their own Figure 2 (see their report) shows how benefits as a share of compensation have been flat at around 30 percent over the recent period of flat real wage growth.

The most commonly cited wage and compensation data come from the Employment Cost Index (ECI), and these data show, for example, that over the past two years, nominal wages (private sector workers) are up 5.4 percent while fringe benefits are up 5.1 percent. Thus, in these data, adding in benefits doesn’t change the wage growth story.

The CEA uses a related data series (ECEC data, which unlike the ECI does not keep the industry and occupation mix constant and provides detailed benefits data) that we analyzed to track two different ways of measuring the benefit share (see table below). The first column uses a “W-2 wages” concept, that includes earnings, paid leave, and supplemental pay in the wage category (the second column includes leave and supplemental pay as benefits, as in CEAs figure above). The W-2 measure corresponds more closely to the measures of wages most analysts and journalists track using household and establishment data. But either way, the share is clearly stable. As we show below, a stable benefit share cannot lead compensation to grow faster than wages.


In comparing compensation to wage trends, it is also germane to note that not all workers get key fringe benefits, yet CEA implicitly assigns them to everyone (and we follow this practice too, for comparability). Recent BLS data show that only 23 percent of workers in the bottom quartile participate in health plans and 25 percent participate in retirement plans. For those in the second quartile (25-50th percentile range), about half participate.

Demographic adjustments: Here again, the short time frame works against CEAs conclusions, as, barring a large shock, like a recession, the composition of the labor force doesn’t change that much year-to-year. Though they claim the entry of young workers taking low-wage jobs is dragging down the average, analysis by Heidi Shierholz and Elise Gould show that “since 2013, as the recovery has strengthened, the opposite has been true—low-wage jobs are actually declining on net while middle and high wage jobs are being added, which has the effect of raising average wages. In other words, the composition effect is currently putting upward pressure on wages.”

What about the other end of the age scale? CEA also claims the retirement of high-earning older persons is putting downward pressure on wages, though here again, for this to affect the trend, such retirements would have to be accelerating. As this interesting analysis by economist Adam Ozimek reveals, the opposite appears to be the case. He finds that “…the senior share of the workforce is growing, not shrinking; 2) seniors are not the highest-paid age group; and 3) to the extent the workforce age composition has affected wages over the last decade or even the last year, it’s not about seniors.”

A simple way to dispel the composition explanation is to look at Atlanta Fed’s wage growth tracker, which controls for demographic changes in the workforce by following the same workers over time. The nominal wage growth in the series accelerated steadily from around 2010 to 2016, it has been largely flat since then, hovering around an average annual growth rate of 3.2 percent (this series runs faster than others in part due to its inclusion of a one-year “experience premium,” the extra pay accruing to workers as the gain experience on the job). This suggests that composition is not a factor in the flattening of real wage growth, which has been a function of faster inflation and slow growth in nominal pay.

All this evidence suggests little to no role for a demographic or composition explanation of the flat real wage trends that have persisted over the past two years. CEA makes a bigger adjustment by applying the a “demographic composition correction…based on the average for the expansion from 2013 through 2018.” This very likely overestimates any near-term compositional shifts, and it is unclear to us why they use an average versus annual values. Moreover, CEA’s demographic adjustments still do not make the early Trump quarters look better than the period that preceded them.

Similarly, their rationale for their hours adjustment—they reduce hours in some series which raises hourly pay—based on the difference between hours worked and hours paid was unclear to us as the paid leave share of wages or compensation has been stable since at least 2014.

Price adjustments: This is the one place where the CEA has a point, though once again, given the short time frame in their final analysis, 2017q2-2018q2, this too doesn’t amount to much. The CEA argues for the use of chain-weighted deflators, which more accurately measure consumer price changes, as they better account for consumer behavior (specifically, chain-weighted deflators better account for changes in the contents of the consumer market basket driven by changes in relative prices). The CPI, which is used by the BLS in their monthly earnings reports, is not chain-weighted.

Therefore, the CEA adjusts wages using the PCE deflator. However, the problem with this choice, as Bivens and Mishel describe here, is that the PCE is considerably less representative of prices faced by consumers. A better choice would be the chained-CPI, which the BLS has produced since 2000. Though the two chain-weighted deflators track each other fairly closely, over the past year, the PCE deflator was up 2.3 percent compared to 2.7 percent for the chained CPI (and 2.9 percent for the standard CPI). In other word, moving to a chained CPI only raises real wage growth over the past year by 0.2 percent.

As a technical note, we find it curious how economists seemingly attentive to measurement details about inflation typically fail to account for the faster than average growth in health insurance prices (relative to the CPI, whatever version) when incorporating benefits data into a compensation measure. Recognizing this faster inflation in health insurance would yield a slower growth in compensation than the measures both here and in CEAs report.

Putting it all together, the table below shows the impact on real hourly compensation for middle-wage workers, using our benefit adjustment, which excludes wage benefits, such as overtime and shift premiums, from health and retirement benefits, along with the chained CPI (see data note for more details). Between 2017q2 and 2018q2, real hourly compensation adjusted by the CPI was essentially flat, down 0.2 percent. After the relevant adjustments, it’s still flat, essentially unchanged over the last quarter.

Sources: see data note below.

The chart below takes a slightly longer-term perspective, showing real hourly compensation for 2014-2018, second quarters. Compensation was growing, in real terms, in earlier years, driven largely by unusually low inflation and some nominal wage acceleration as the jobless rate fell in those years. But the last two years have been dry spells for the real hourly pay of middle-wage workers, a finding that holds when we include their benefits and shift to a slower-growing deflator.

Sources: See data note below.

In other words, there is no evidence of the CEA-suggested acceleration of compensation since Trump took office. To the contrary, the growth of real hourly pay of middle-wage workers has slowed to a crawl since then.

As we said, faster real wages and benefits may occur if the jobless rate continues to fall, giving less advantaged workers more of the bargaining clout they lack. But for now, hourly wage and benefit growth, inflation-adjusted, is close to flat in real terms and considerably less than the 1.3 percent  productivity growth over the last four quarters. We’d thus urge CEA to spend less time with their thumb on the measurement scale and more time on policies to help working people benefit from the ongoing expansion.

Data note: Table 2 starts with the hourly wage of blue-collar factory workers and non-managers in services (this is the BLS production, non-supervisory wage, which covers 82 percent of the workforce). We then scale that up to include benefits using the ECEC benefits data as described in the text (W-2 measure). Finally, we adjust nominal compensation by both the regular and chain-weighted CPI. Note that since ECEC data are not yet out for 2018q2, we use the 2018q1 benefit share (as, apparently, does CEA).

Some thoughts on that new Fed paper everybody’s talking about.

September 2nd, 2018 at 11:38 am

It’s a lovely morning on the back porch, and the mind turns to that new Fed study everybody’s talking about. It’s the one by Erceg et al about monetary policy at moments like this one, with a flat Phillips Curve (PC), u<u*, along with much uncertainty about u* (importantly, I’d argue that uncertainty is asymmetric; the Fed’s estimate of u* looks too high). BTW, ‘u’ is the unemployment rate; ‘u*’ is the estimate of the “natural rate,” the lowest rate associated with stable prices.

I’ve got a longer, less cryptic piece on this study coming out later this week in WaPo (tomorrow, it’s Dean Baker and I celebrating Labor Day with a piece on unions as a potent weapon against inequality). But I wanted to set the table for that piece with a bit of analysis here. The WaPo piece explains any oblique terminology; apologies in advance for any obscurities in what follows.

One reason this piece, which I found to be a thoughtful/useful bit of work, is getting a lot of attention is because its key finding is counterintuitive. Given that unemployment has been well below the Fed’s estimate of u* of 4.5% and inflation’s (PCE core) just now hitting their 2% target, many of us have argued that the optimal monetary policy is to downweight the unemployment gap and focus on the lack of wage or price inflation.

Consider, e.g., the strong version of this view from EPI’s Josh Bivens: “…the definition of labor market slack is wage growth too weak to put upward pressure on the Fed’s price inflation target. If this wage growth is not happening, there is labor market slack. So, simply looking at some quantity-side measure of the labor market (say the unemployment rate) and thinking ‘hmm, that’s low, we must be at full employment” is substituting gut feeling for economic reasoning.’”

In a similar vein, Baker and I have argued that you know you’re at full employment when extra demand generates not jobs and real wages, but inflation.

But the Fed study comes to a different conclusion, arguing that even if u* is uncertain, it’s “better” to target the employment than the inflation gap. The definition of “better” is key, of course, and the authors are explicit that their definition bakes in their result in ways with which reasonable critics may disagree (more on that in a moment).

The paper does a bunch of macrosimulations of unemployment and inflation outcomes using a set of monetary rules that apply stronger or weaker weights to the employment and inflation gaps. The find that “because monetary policy acts with a lag, waiting for inflation to materialize before reacting is undesirable, particularly when economic conditions are such that outsized deviations of inflation from its target are a plausible outcome.”

This is interesting. While camp Bivens sees the combination of the flat PC and overestimated u* as a reason for accommodative monetary policy, their simulations suggest that because of the flat PC, over-weighting the inflation gap will lead to wide and damaging (to demand) swings in monetary policy.

In fact, conditions in the current economy partially drive their result. Suppose the Fed listens to Bivens et al and targets inflation instead of unemployment. Because inflation has long undershot the Fed’s 2% target and the PC is so flat, it would take historically very low unemployment to juice inflation. Conversely, suppose some shock to the system…like, um, a trade war…led inflation to spike; then, the authors argue, it would take really high unemployment to bring inflation back down.

The study’s simulations thus find that if the Fed weighted up its inflation target relative to its unemployment target, the jobless rate could fall so low or climb so high that it could generate “risks to financial stability and more generally to the sustainability of macroeconomic outcomes.”

One way they end up there is by scoring success through a “loss function” that penalizes policy makers for letting the jobless rate fall below u*. But with u* higher than it should be, this approach doles out undeserved penalties for running a hot labor market (when they plug in a u* of 3.7%, upweighting the employment gap looks less favorable; compare Table 3, column F, rows 3 and 6). Their symmetric loss function (being below u* is as bad as being above it) also discounts the extremely valuable benefits of super-tight labor markets to less advantaged workers, a benefit that is especially worth tapping right now given the lack of price pressures. I’d want a loss function to reflect these benefits, one that treats being below u* as preferable to being above it.

As noted, the authors are explicit about this point, and the loss function they use is standard fare. Still, the paper is replete with so many variants, why not add one more? I urge the authors to run the results through a loss function that meets the criteria just noted.

I’ve got two more objections to the findings.

First, at least as I read it, the paper seems to suggest the Fed is unable to look past inflation perturbations caused by supply shocks. As just noted, the simulations appear to combine this inability with the flat PC to generate sharp, yet unnecessary (because it’s a temporary shock, not a shift in demand), accommodation or tightening. But this seems demonstrably wrong, as just recently, Fed statements have included many references to temporary shocks to prices, including energy, cell phone pricing, and Trump’s trade mishegos (the latter of which could eventually whack demand).

Also, what about all those years of hard work by Fed officials to anchor expectations? That too leads people to look through temporary shocks and assume stable, long-term prices. (See the bottom panel of their Figure 1 for evidence of well-anchored inflation expectations.)

Second, in numerous places, including the quote above, the paper argues that it’s better to be a bit more hawkish to avoid financial instability. This seems like step backwards. Former Chair Yellen and others have been very clear on this point: when we use tighter monetary policy to regulate bubbles in financial markets, we penalize the great many to hold back the reckless few. It is macroprudential policy and Dodd-Frank style regulation that should be the first line of defense against excesses in financial markets.

I get that Powell recently (wisely) argued that, given their far-reaching potential damage, the Fed should put financial excesses high on its watch list. But, if the real economy is not overheating, that doesn’t imply that fighting them with higher rates is preferable to regulation, “irrational-exuberance”-style forward guidance, and higher capital buffers.

That said, I strongly recommend the paper to those of us calling for heavier relative targeting of inflation as opposed to employment. It offers some high-calorie food for thought.

ISDS and the US

August 28th, 2018 at 1:51 pm

I’ve been touting the fact, i.e., as I understand it, that this new US/Mex NAFTA agreement just struck yesterday largely gets rid of investor dispute rules (investor state dispute settlement, or ISDS) that many progressive have long complained about. (To be clear, whether this deal is going anywhere is a whole other story; I’m skeptical.)

I’m working on a piece about how the new deal looks a lot better for workers on both sides of the border than prior agreements, but re ISDS, the very knowledgeable Lori Wallach tell me it “ends the possibility of any future U.S.-Canada ISDS cases. This is huge given major US-Canada cross investment.” For Mexico, where domestic courts are less reliable, investors who want to bring a case must first exhaust domestic court and administrative remedies, before turning to new procedures that significantly raise the bar to investor compensation (the fact that the Business Roundtable is already complaining about this part of the deal is revealing in this regard). There is apparently a carve out for investments in Mexican energy production that would allow a small group of U.S. investors the same protections as in earlier agreements, but this looks to have been the negotiating price for the larger advances just noted.

My friend Jay Shambaugh, presumably implying that ISDS ain’t so bad, asks the reasonable, though rhetorical (if not snarky: surely Jay, a former Obama-admin economists who’s one of my go-to peeps on international trade, knows the answer). Has a company used ISDS under NAFTA to overturn a US law or regulation?

No, meaning fears about the process overriding US sovereign laws have not been realized. If that’s Jay’s point, it’s a relevant one with which I agree.

But their are still at least two big, existing problems. First, ISDS has been used by corporate bullies of rich countries to extract millions in fines and fees from poorer countries, and not for investor takings (which would be legit) but for protections prohibited by trade deals (examples here and here). What I want to see much more in U.S. trade agreements–and Jay might agree–is less protectionism of the advanced countries’ investor class and its IP and drug patents, and more lifting of standards in poor countries.

The second problem with ISDS is broader:

Through the backdoor of trade agreements, the ISDS process imposes extreme property rights’ concepts rejected repeatedly by Congress and U.S. courts, such as the notion that governments should pay “regulatory takings” compensation to property owners for the right to enforce environmental, health and other safeguards that could undermine the value of their property or investment. We must not solve the problem of weak rule of law among our trading partners by having the broad public bear investment risk or by changing fundamental principles of U.S. law. Instead, investment risk must be borne by the investors themselves; it is their skin, not ours, that should be in the game.

Final point. While the US hasn’t lost a case, a country is only really exposed to ISDS risk when partner countries have substantial investments in the other countries in the deal. That’s why, according to Lori, “54 of the 56 NAFTA ISDS cases to date attacking U.S. or Canadian laws were brought by investors from the Canada or the U.S., not from Mexico.”

Surely, this makes no sense. ISDS isn’t in place–or at least it shouldn’t be–to be invoked in advanced countries with mature legal systems. Let the nationally-sanctioned, highly functional court systems work it out! (Jay: agree or disagree?.)

In fact, recent journalistic research reveals speculation by financial investors in ISDS cases, wherein investors either purchase companies with the express purpose of filing an ISDS claim or directly bankrolling the cases in order to claim a share of the fine. (Investors refer to this practice as “third-party funding of international arbitration against foreign sovereigns”.) Gus van Harten, a law professor who has studied these activities, finds that investors “…can get an award for billions of dollars when that award would never come out in domestic law. It’s just a jackpot for speculators.”

End of the day, the fact that ISDS hasn’t overridden any U.S. laws is comforting and Jay’s right to “ask” about it. But that doesn’t mean it’s non-evil!


Links; some wage thoughts

August 23rd, 2018 at 6:06 pm

A few WaPo entries for your entertainment:

A deeper dive into similarities shared by Trump and Erdogan. And yet, despite their…um…sub-optimal leadership, our economy booms and Turkey’s tanks. I focus on what that has to do with a) dollars, and b) Fed independence.

My old pal Kudlow is sounding off on how the US economy is just “crushing it!” under Trump. Yeah…not so much. Co-written with Cong. Ro Khanna, who has some highly worthy legislative proposals in the mix which we link to in the piece.

Turning to other econ news, from the minutes of the latest Fed meeting, here’s their thoughts on wages:

Many participants commented on the fact that measures of aggregate nominal wage growth had so far picked up only modestly. Among the factors cited as containing the pickup in wage growth were low trend productivity growth, lags in the response of nominal wage growth to resource pressures, and improvements in the terms of employment that were not recorded in the wage data. Alternatively, the recent pace of nominal wage growth might indicate continued slack in the labor market. However, some participants expected a pickup in aggregate nominal wage growth to occur before long, with a number of participants reporting that wage pressures in their Districts were rising or that firms now exhibited greater willingness to grant wage increases.

Let’s take these in turn: Slow productivity growth, check, as I discuss here. But I also point out that if workers had more bargaining power, they’d be claiming a larger share of profits, which are uniquely high in such an allegedly tight labor market (and the wage share of national income is, in turn, uniquely low).

Lags…yeah, well, everybody always blames the lags. What’s more meaningful in this context is potential non-linearities, about which I hope to write more soon. Going from 7 to 6 on the unemployment rate does a lot less for wages than going from 4 to 3.

“Continued slack in the labor market…” I’m very glad to hear them say that, because I think there are a number of signs that there’s still some labor supply left to be absorbed in this expansion. See here re employment rates of prime-age workers.

Finally, they expect wages to pick up. I do too, both nominal and, as high energy prices subside, probably real wages as well. But workers have a lot of ground to make up and, absent clear signs of overheating, it’s important for policy makers not to get spooked by some real wage gains finally reaching many who’ve heretofore been left behind.

On that front, not that unit labor costs–labor costs per unit of production–are coasting at around 2%, the Fed’s target inflation rate, so little sign here, or, more importantly, in core inflation price gauges, of wage-push price pressures.

Source: My wage mashup series (5 wage series smushed together) minus 3-yr avg of productivity growth.


Lynx; Trump/Erdogan: compare and contrast

August 14th, 2018 at 12:43 pm

Recent links to WaPo pieces:

Productivity and wages: They’re connected, of course, but the extent of the connection requires nuanced analysis of wages at different percentiles and movements in labor’s share of national income.

There’s an interesting dichotomy here in how economists and people think about productivity and wages. For many economists, it’s the determinant of wage growth. For many people, it’s irrelevant, in that powerful forces divert productivity growth from paychecks to profits. The truth, especially once you get away from averages, lies in-between. Productivity matters a great deal, but it is not by itself sufficient to drive broadly shared prosperity.

Employment rates also matter a lot: They take the elevator down in recessions and the stairs up in recoveries. They also may carry some info about the arrival of next recession. Plus, their recent movements reveal the disproportionate benefits of full employment to the least advantaged.

Are politicians no longer listening to economists? You wish. In fact, they’re listening to the wrong ones telling them what they want to hear.

Now, a quick note on current events.

As regards the tanking of the Turkish lira, the business press is largely concerned with the contagion question: to what extent will Turkey’s problems spillover into European and American economies? The consensus is “not much,” based on Turkey’s size and financial markets’ limited exposure to Turkish debt, much of which is dollar-denominated, meaning it becomes more expensive to service when the Turkish currency depreciates.

That’s probably right, and Turkey has uniquely weak fundamentals among emerging market economies: “current account deficit of 6.3% of GDP, Corporate foreign exchange debt is 35% of GDP, inflation rate of 16%.” But the situation bears close watching, of course, and the strengthening dollar has important implications for the trade war, i.e., it pushes in the opposite direction of the tariffs (tariffs make imports more expensive; the stronger dollar makes them less expensive).

But another interesting aspect of the Turkish meltdown is how much Trump and Erdogan have in common. In one sense, that’s not surprising, as the strongman, faux populist playbook is pretty straightforward, and history is replete with examples.

In this case, Trump and Erdogan both pursue: reckless fiscal policy, muscling the central bank to keep rates down (though Trump doesn’t use anything like the muscle that Ergodan does), appointing family members to high places (sons-in-law, to be specific), vilifying other countries/media as the source of any woes (in a Trumpian flourish, Erdogan recently blamed “economic terrorists on social media” for spreading misinformation).

And yet, the economic outcomes, particularly via the currency and capital flows couldn’t be more different. In fact, the relative currency moves show foreign exchange traders are pulling out of the riskier emerging markets and buying dollars and U.S. debt.

It’s a reminder of the dominant size and durability of US economy, the role of the dollar as the reserve currency, and our still independent central bank. More broadly, there’s just a lot that insulates the U.S. economy from terrible leadership.

At least there is for now…