Macro Roundup

James Pethokoukis,"The productivity-pay 'gap': a pernicious economic myth, Faster Please!February 10, 2022, https://fasterplease.substack.com/p/-the-productivity-pay-gap-a-pernicious

The productivity-pay 'gap': a pernicious economic myth

I get this a lot: Why do you care so much about trying to accelerate productivity and economic growth? Are you some sort of neoliberal, “Make the line go up,” “What’s good for Wall Street is good for America” obsessive sicko? If you were a regular New York Times reader, you would know all about the “productivity and pay gap,” the long-term delinkage between worker output per hour and worker compensation. Productivity and economic growth don’t matter, redistribution does. I thought you worked at a think tank. Maybe you need to think harder, dude!

Well, something like that, I guess. Anyway, the popular notion that productivity no longer drives compensation — productivity growth may have been slow since the 1970s, but inflation-adjusted pay has been stagnant — is reflected in various versions of this chart:

Productivity growth never stopped mattering for wage growth

In “Productivity and Pay: Is the link broken?” Harvard University’s Anna Stansbury and Lawrence Summers find higher productivity growth is associated with higher average and median compensation growth. They use the median wage and the average wage of production and non-supervisory employees, as well as the average wage of all workers. In order to smooth out the up and downs of the business cycle, they calculate the three-year moving average of the change in inflation-adjusted compensation and labor productivity. Their findings (bold by me):

We find substantial evidence of linkage between productivity and compensation: over 1973-2016, one percentage point higher productivity growth has been associated with 0.7 to 1 percentage points higher median and average compensation growth and with 0.4 to 0.7 percentage points higher production/nonsupervisory compensation growth. … Our results suggest that productivity growth still matters substantially for middle income Americans.

Moreover, boosting productivity growth is at least as good for workers as reducing inequality. Stansbury and Summers note that if inequality had stayed at 1973 levels, the median worker’s pay would have been around 33 percent higher in 2016. But if productivity growth had been as fast over 1973-2016 as it was over 1949-1973 — about twice as high — median and mean compensation would have been around 41 percent higher. (Recall that the supposed “golden age” for American workers in the immediate postwar decades was also a time of rapid productivity growth.) S&S:

These point estimates suggest that that the potential effect of raising productivity growth on the average American’s pay may be as great as the effect of policies to reverse trends in income inequality. Conversely they suggest that a continued productivity slowdown should be a major concern for those hoping for increases in real compensation for middle income workers.

Productivity growth still matters for take-home pay

Likewise, the slowdown in productivity growth also provides a large part of the explanation of slow wage growth in the late 2010s — that, even though unemployment was even lower than what it was in the fast-wage growth era of the 1990s. The problem: Productivity growth has been rising only a quarter as fast as back then.

As economist Jason Furman wrote for Vox in 2019, it’s no mystery as to why wage growth was so fast in the late 1990s, but less so in the years leading up to the pandemic. Furman points out that productivity growth back then rose at a spectacular 3 percent annually versus just 0.7 percent annually more recently. Furman: “That’s part of the explanation for slow wage growth: Based on the productivity numbers alone, one would predict that average wages would be growing about 2.3 percent more slowly than they did in the late 1990s.”

And let me highlight this from Furman (bold by me): “Greater productivity growth holds the potential of being the most powerful source of sustained wage growth across the income spectrum.” (I like that so much, I might start using it instead of the famous Paul Krugman quote, “Productivity isn't everything, but in the long run it is almost everything.”)

Thinking that productivity doesn’t matter or is a mere secondary or tertiary factor is why I use the word “pernicious” in the headline. Misunderstanding the continued value of productivity growing means misunderstanding the foundation of modern prosperity.

To (almost) conclude: Faster productivity growth is superimportant for faster rising living standards. And the sort of scientific discovery, technological invention, and commercial innovation that boosts take-home pay will also accomplish a lot of other pretty important things: increased opportunity, less poverty, healthier lives, a more stable climate, a multi-planetary civilization. I hope that’s one lesson we’ve all drawn from the rapid mRNA vaccine development over the past two years.

Next steps

So what’s the policy advice here? For faster productivity growth, I would point you to the many editions of this newsletter in which I’ve written about deregulation and great science investment, among other things. Moreover, the continued strong connection between pay and productivity argues for making workers more productive, especially those toward the bottom of the income distribution. I’ll give Strain the final word here, from a 2019 analysis:

Policies to increase the skills of, and training available to, workers — for example, reforms to our K-12 education system and the expansion of apprenticeships and other forms of work- based learning — should be enacted. Earnings subsidies should be expanded to draw more people into the workforce. Policies to encourage business investment should also be considered. Labor market regulations that serve as barriers to workers and reduce the quality of matches between workers and jobs should be removed.

Robert Lawrence of the Peterson Institute created both of the above charts a few years back, and he contends the second chart better reflects the tight productivity-compensation relationship. Why the difference between the two charts? Lawrence made several modifications to the first one: (a) adjusted for an overly narrow definition of workers; (b) added in benefits to wages, (c) used an inflation measure that better accounts for what workers actually purchase; and (d) used a more relevant productivity measure.

Stagnant pay?

Key to the productivity-pay gap claim is the notion that since the early 1970s, real wages have gone pretty much nowhere. Maybe just a tiny 5 percent gain since the Nixon administration. Sure looks like stagnation, or a pretty close facsimile.

But using that 5 percent number to encapsulate some five decades of American economic history is problematic. Since the 1990 business cycle peak, wages are up 20 percent in real terms. And if you use the Personal Consumption Expenditures Price Index rather than the Consumer Price Index as the inflation measure — it’s the one preferred by the Federal Reserve and Congressional Budget Office — you find a 32 percent increase in average worker wages and about the same for the bottom fifth of workers. Fun fact: Broader after-tax income measures show middle-class living standards up by 42 percent since 1979 and 70 percent for the bottom fifth. (For more on all of this, I highly recommend Strain’s 2019 book The American Dream Is Not Dead, from which the below chart is taken.)

Summers argues American labor markets are historically tight and ill-wind for further inflationary pressure, "...The results presented in this paper suggest that the firm-side indicators are highly significant for predicting wage inflation, and that the current level of vacancies and quits observed in the labor market correspond to a degree of labor market tightness previously associated with unemployment rates below 2 percent. We also show that firm-side unemployment predicts extremely rapid growth in nominal wages over the subsequent year. Overall, the evidence presented in this paper indicates that U.S. labor markets are currently extremely tight...Figure 6 displays our rough breakdown of the relative contribution of different factors to the employment shortfall. Overall, we estimate that most of the employment shortfall will remain through 2022,with Covid-19 health concerns, immigration restrictions, excess retirements, shifts in workers’ tastes, and changes in the demographic structure explaining the bulk of the labor shortfall...Shifts in Demographic Structure - 1.3 million workers...Covid-19 Health Concerns - 1.5 million workers...Immigration Restrictions - 1.4 million workers...Excess Retirements - 1.3 million workers..Reduced Incentives to Work - 1 million workers..Covid-19 Vaccine Mandates - 0.4 million workers..We present a cursory analysis of the employment shortfall, and conclude that even under optimistic Covid-19 outcomes, the majority of the employment shortfall will likely persist moving forward. Moreover,if employment were to increase due to an increase in labor force participation, it would be accompanied by increases in incomes, and therefore an increase in demand. We therefore conclude that any benefits from the supply-side over the next year are unlikely to substantially mitigate inflation pressures from the labor market. Given these estimates, and the results presented throughout the rest of this paper, we believe that labor markets will continue to be very tight unless there is a considerable slowdown in labor demand. Overall, these findings suggest to us the need for substantial caution about the possibility of inflationary pressures from the labor market moving forward.

“…'... Figure 3 shows timeseries plots of mean wage growth, calculated from CPS-ORG microdata and conditioned on lagged inflation, versus the 4-quarter moving average of each slack indicator, using historical data from 1990 to 2019. The figure shows strong relationships with clear lags between the level of each slack measure and wage inflation...."

Estimating a firm-side equivalent unemployment rate

“…'...To compare the predictive power of the firm-side equivalent unemployment rate with the actual unemployment rate for nominal wage growth, we estimate a wage Phillips curve model that includes the actual unemployment rate and our predicted unemployment rate as the regressors20. Table 5 displays the results. We run 12 different model specifications, varying the wage series used to calculate nominal wage growth, and changing the lag length of our explanatory variables. The results provide strong evidence that the firm-side equivalent unemployment rate has essentially all the explanatory power in predicting wage inflation compared to the actual unemployment rate: across most of our regression specifications, decreases in the firm-side unemployment rate explain nominal wage growth21. We also note that this wage inflation usually occurs with either 4-quarter or 8-quarter lags. Given the extremely low firm-side predicted unemployment rate today, these results provide strong evidence that the current labor market is very tight. The firm-side predicted unemployment rate has fallen from an average of 3.6 percent in the fourth quarter of 2019 to an average of 1.5 percent in the fourth quarter of 2021, which corresponds to an increase in wage inflation from 4.0% to 4.9% (using CPS-ORG mean wages). Figure 4 below shows estimates for nominal wage growth from a wage Phillips curve model using predicted firm-side unemployment as the slack variable and controlling for lagged inflation. The results indicate that estimated wage inflation in the fourth quarter of 2021 is the highest it’s been in the last 20 years across all four of our wage measures. We then simulate nominal wage growth in 2022 and 2023 under the assumption that the vacancy rate, the quits rate, and the inflation rate remain the same (our 3-year weighted lagged inflation raises to 3% by the end of 2022 and to 5% by the end of 2023). The figure shows that nominal wage growth under these assumptions is projected to increase dramatically over the next two years, surpassing 6% wage inflation by 2023 with three of the four wage measures…”

"... Figure 6 displays our rough breakdown of the relative contribution of different factors to the employment shortfall. Overall, we estimate that most of the employment shortfall will remain through 2022, with Covid-19 health concerns, immigration restrictions, excess retirements, shifts in workers’ tastes, and changes in the demographic structure explaining the bulk of the labor shortfall..."

Immigration Restrictions - 1.4 million workers, “…. In 2020 and 2021, the number of immigrants arriving in the United States dropped substantially, due to increased restrictions related to the Covid-19 pandemic. According to data from the Current Population Survey, prior to 2019 the foreign-born population of working age grew at an average rate of about 660,000 people year. Since the pandemic, there has been effectively zero growth in the working-age foreign born population. Peri and Zaiour (2022)24 use data from the CPS and Department of State and find that the number of working-age foreign-born people in the United States by the end of 2021 is 2 million people below the 2010-2019 trend. We apply a foreign-born working-age employment rate of 72% (OECD, 2022), and estimate a shortfall in employment of 1.4 million workers due to immigration restrictions….”
Excess Retirements - 1.3 million workers“…At the beginning of the Covid-19 pandemic, there was a large spike in the share of the U.S. population (age 16 and above) that was retired. We compare the actual number of retirees to the expected number of retirees implied by the age-specific retirement rates observed in 2019, 25 and find that there were 1.3 million excess retirements between February 2020 and Dec 202126. This estimate is consistent with estimates from several other recent models attempting to quantify the labor shortfall due to excess retirements27. Given that this number has been relatively constant since mid-2020, we believe it is unlikely that the number of excess retirees will decline over the subsequent year….”
Reduced Incentives to Work - 1 million workers, “… According to data from the Federal Reserve Bank of New York’s SCE labor market survey, the average reservation wage increased by 9.3% between Nov 2019 and Nov 202128. This increase in the reservation wage can be explained by many factors that have reduced the marginal cost of continued search, including government transfers like the child tax credit, an increase in home values and personal savings, and a shift in work-life preferences during the pandemic. We present evidence on these factors in turn. The American Rescue Plan expanded the child tax credit to a total of $3,600 for children 5 and younger (up from $2,000 per child), and $3,000 for those ages 6 through 17, and made the full credit available to all low and middle-income families regardless of earnings or income. Relying on elasticity estimates consistent with the academic literature, Corinth et al. (2021) estimates that the expanded child tax credit would cause 1.46 million workers to leave the labor force29. While negotiations to continue the expanded Child Tax Credit in 2022 have stalled, we believe it is reasonable that a share of the Corinth et al. (2021) estimate has already materialized, pushing up workers’ reservation wages. An increase in personal savings and property values is also responsible for keeping workers out of employment. Through September 2021, the typical low-income family still had 70 percent more cash on hand than two years prior30, while in November 2021, about 25% of the unemployed reported that they were not urgently searching for a job due to large financial cushion31. Most estimates for the total excess savings accumulated in the United States since the beginning of the pandemic are around $2.5 trillion32. Home values have also appreciated dramatically, hitting a peak year-over year increase of 19.3 percent in July 2021…”
Covid-19 Vaccine Mandates - 0.4 million workers“…As of November 2021, 29 percent of workers say their employer has enforced a vaccine mandate (KFF Covid-19 Vaccine Monitor, November 8-22). A survey by the Kaiser Family Foundation also found that 4 percent of unvaccinated workers have personally left their job because their employer required them to get a Covid-19 vaccine. Assuming these numbers remain constant in 2022, we estimate that 0.4 million workers will leave their jobs due to Covid-19 vaccine mandates…”
Alex Domash and Lawrence Summers, "How Tight Are U.S. Labor Markets," National Bureau Of Economic Research, https://www.nber.org/papers/w29739

"...There is a large literature on earnings and income volatility in labor economics, household finance, and macroeconomics. One strand of that literature has studied whether individual earnings volatility has risen or fallen in the U.S. over the last several decades. There are strong disagreements in the empirical literature on this important question, with some studies showing upward trends, some showing downward trends, and some showing no trends. Some studies have suggested that the differences are the result of using flawed survey data instead of more accurate administrative data. This paper summarizes the results of a project attempting to reconcile these findings with four different data sets and six different data series--three survey and three administrative data series, including two which match survey respondent data to their administrative data. Using common specifications, measures of volatility, and other treatments of the data, four of the six data series show a lack of any significant long-term trend in male earnings volatility over the last 20-to-30+ years when differences across the data sets are properly accounted for. A fifth data series (the PSID) shows a positive net trend but small in magnitude. A sixth, administrative, data set, available only since 1998, shows no net trend 1998-2011 and only a small decline thereafter. Many of the remaining differences across data series can be explained by differences in their cross-sectional distribution of earnings, particularly differences in the size of the lower tail. We conclude that the data sets we have analyzed, which include many of the most important available, show little evidence of any significant trend in male earnings volatility since the mid-1980s...."

The evidence:"...The six data series used in this project are shown in Table 1: the PSID, the CPS survey, CPS linked Social Security earnings records, the SIPP survey, SIPP-linked Social Security earnings records, and UI earnings from the LEHD. The PSID has been analyzed many times before, so the primary purpose of including it is only to provide a baseline estimate using the same sample definitions, measures of volatility, and other analysis features as those in the other five data series. For all data sets, only men 25-59 in each year are included..."...All analyses use simple and transparent summary measures of gross earnings volatility, calculating the earnings change from one year to a subsequent year, either one or two years later, depending on the data set..."

“…Figure 1 shows our baseline results, using the samples and earnings variables listed in Table 1, for men working both periods. The PSID shows patterns mostly consistent with prior work, with rising volatility from the 1970s to the mid-1980s, then following a stable trend around significant fluctuations through about 2002, then rising in the period leading up to and including the Great Recession, and then falling post-Recession from 2012 to 2016.The last four years of PSID data are new to this project and show that volatility has declined back to its pre-Recession level in 2006, which was somewhat above its level in the mid-1980s. The series for the other five data sets are often different from the PSID in level but not always in trend. The SIPP administrative data series, which starts in 1980, is higher in level than the PSID but follows a similar slight decline from 1982 through about 1999, but with fluctuations over that period much milder in magnitude than for the PSID (perhaps the result of a larger sample size—see Table 1). It then also rises before the Great Recession and falls afterwards, although again not always of the same magnitude and at exactly the same time points as the PSID. The SIPP survey data have a lower level of volatility than the PSID (and much lower than the SIPP administrative data series) but has an approximately similar pattern in the first half of the period—a rise then fall from 1985 to 1999, but occasionally moving in opposite directions (e.g., 1988-1990).16 But the main difference with the SIPP survey is that it rises much less before and during the Great Recession than the PSID and the SIPP administrative data series…”

Key quote, “…We conclude that average volatility growth rates in the 2000s and partway into the 2010s were quite similar for most of the data sets, with some exceptions. Combined with our finding of very little volatility trend in the three data sets covering the period from the mid-1980s to the 1998- 2002 period, we also conclude that there is little evidence for significant trends upward or downward over the last 20-30 years….”

Ruy Teixeira argues the left should live by a simple political acid test, and implies if they don’t they will die by it, “…. In the 2020 Presidential election, despite a slight improvement over 2016, Democrats still lost white working class (noncollege) voters in 2020 by 26 points (Catalist two party vote). Since 2012, nonwhite working class voters have shifted away from the Democrats by 18 margin points, with a particularly sharp shift in the last election and particularly among Hispanics….Since the 2020 election, the situation has only worsened. Signs of continued slippage among working class voters were unmistakable in the 2021 elections, most notably among Hispanic and Asian working class voters. In the latest Monmouth poll, Biden’s approval rating among the multiracial working class was an abysmal 32 percent vs. 59 percent disapproval, compared to 52 percent approval among the college-educated….This performance among working class voters should be unacceptable for a party of the left. After all, what is the point of a left party that cannot command the loyalty of the working class and therefore plausibly claim to represent its interests?..I suggest a simple test Democrats should be continually making on both their policies and rhetoric: What Would the Working Class Say?…Voters in general, especially working class voters, tend to think the following: Equality of opportunity is a fundamental American principle; equality of outcome is not. Discrimination and racism are bad but they are not the cause of all disparities in American society. No one is completely without bias but calling all white people racists who benefit from white privilege and American society a white supremacist society is not right or fair. There are underlying differences between men and women but discrimination on the basis of gender is wrong. There are basically two genders but people who want to live as a gender different from their biological sex should have that right and not be discriminated against. However, there are issues around child consent to transitioning and participation in women’s sports that are complicated and not settled. Racial achievement gaps are bad and we should seek to close them. However, they are not due just to racism and standards of high achievement should be maintained for people of all races. Language policing has gone too far; by and large, people should be able to express their views without fear of sanction by employer, school, institution or government. Good faith should be assumed, not bad faith. Compare these with the current outlook and obsessions of the college-educated left within the Democratic party, which tends to have a kind of cultural hegemony within institutions associated with the party. It doesn’t match up too well….”

Ruy Teixeira, “What Would the Working Class Say? (WWWCS),”The Liberal Patriot, January 27, 2022, https://theliberalpatriot.substack.com/p/what-would-the-working-class-say

What Would the Working Class Say? (WWWCS)

All across the Western world the working class is deserting the Left. Thomas Piketty and his colleagues among others have copiously documented this trend. The United States is no exception to this trend.

In the 2020 Presidential election, despite a slight improvement over 2016, Democrats still lost white working class (noncollege) voters in 2020 by 26 points (Catalist two party vote). Since 2012, nonwhite working class voters have shifted away from the Democrats by 18 margin points, with a particularly sharp shift in the last election and particularly among Hispanics. This latter development is particularly important since Democrats have hitherto comforted themselves that losses among the working class were just among whites, who they presume to be motivated by retrograde racial and cultural attitudes. That is no longer a tenable view.

Since the 2020 election, the situation has only worsened. Signs of continued slippage among working class voters were unmistakable in the 2021 elections, most notably among Hispanic and Asian working class voters. In the latest Monmouth poll, Biden’s approval rating among the multiracial working class was an abysmal 32 percent vs. 59 percent disapproval, compared to 52 percent approval among the college-educated.

This performance among working class voters should be unacceptable for a party of the left. After all, what is the point of a left party that cannot command the loyalty of the working class and therefore plausibly claim to represent its interests? And in raw electoral terms, worsening performance among working class voters makes the Democrats’ quest for political dominance essentially impossible, since the share of working class voters in the country is 70 percent larger than the share of college-educated voters.

To help remedy this situation, I suggest a simple test Democrats should be continually making on both their policies and rhetoric: What Would the Working Class Say? (WWWCS). This test is not so hard to do but it does entail getting outside of the liberal college-educated bubble so many Democrats live within, particularly as experienced on social media, in activist circles and within advocacy, nonprofit, media and academic institutions. Look at actual public opinion data—not as summarized by someone you know or something you read. Look at focus group reports. Talk to actual working class people—there are lots of them! Listen to your intuitions about how working class people would likely react to policies and rhetoric currently associated with the Democrats —not how you think they should react. Think of family members or people you grew up with who are working class. Try to get inside their heads. They are less ideological, more focused on material concerns, more likely to be struggling economically, less interested in cutting edge social issues, more patriotic and generally more culturally conservative. All this makes a difference.

That WWWCS test can tell you a lot about whether Democrats are on track with their approach. If the test indicates that Democrats are advocating or saying something that is likely unpopular, offputting and/or just lacks salience with working class people, that policy or rhetoric is probably on the wrong track. Conversely, if the test indicates that working class people are likely to view what Democrats are advocating/saying as desirable, in tune with their values and actually important to their everyday lives, that is a very good sign.

Here are some examples.

Return to Normality. Covid and the economy remain the top issues in the country. Biden and the Democrats are viewed as doing a good job on neither, despite having been elected primarily to fix these twin crises. Biden’s approval ratings on both these issues are underwater across surveys; the latest Pew survey shows a precipitous decline since last March in confidence that Biden can handle the coronavirus outbreak and make the right decisions about the economy.

In a recent CBS News poll, 58 percent said the Biden administration’s focus is not enough on the economy and 65 percent said the same thing about inflation. Only a third said Biden and the Democrats were focusing on issues they “care a lot about”. Among Biden disapprovers—the majority of the country—63 percent said getting inflation down would improve their opinion of Biden while just 24 percent said the same thing about passing the “Build Back Better” bill.

Despite the mountain of polling evidence suggesting that the Democrats should primarily be focusing on covid, the economy and getting the country back to normal—and communicating that to voters—the party is having difficulty tearing their attention away from a Build Back Better bill. To many Democrats it seems that their dreams of “transformation” are fading into the distance. But that was not what they were elected to do, particularly as far as working class voters are concerned. In the new Monmouth poll just 20 percent of working class respondents indicated they thought a Build Back Better bill should be the top priority of Congress in the coming year.

Evidently, to the typical working class voter, a return to normality in the coming year looms far larger than an endlessly debated “transformative” bill whose contents are almost impossible to keep track of. This makes sense when one keeps in mind the pragmatic, close to the ground perspective of working class voters whose lives have endured far more disruption than the liberal, college-educated Democrats who are the most fervent supporters of pushing the largest possible Build Back Better bill through Congress.

Here indeed in an area where the Democrats should be taking the WWWCS test.

Voting Rights/Crisis of Democracy. The Democrats’ kamikaze pitch for a comprehensive voting rights bill that was sure to fail was a puzzler. The fact of the matter is that voting rights barely register as a public concern. In a recent AP-NORC poll, where respondents could mention up to five problems for the government to work on in the coming year; just 6 percent of the public placed voting rights anywhere in their top five. This issue may be a top priority for the progressive wing of the Democratic party but it’s just not for ordinary voters—and I think we can safely say, for ordinary working class voters.

Even among the black population, a Morning Consult poll found that only 41 percent think the bigger problem with American democracy is that it’s too hard to vote, rather than voting regulations are not strict enough (among Hispanics the analogous figure was even lower at 34 percent). In a Monmouth poll, 84 percent of nonwhites said they supported requiring a photo ID for voting.

All this for voting rights bills that would do little to counter the threat of election subversion from vote decertification for partisan purposes, which was the original Democratic concern. A far more promising approach is reforming the Electoral Count Act of 1887 to seal off a number of channels for election subversion. This approach has some bipartisan support and therefore some chance of actually happening. A pragmatic working class, desirous of both fair elections and moving on to action on their everyday concerns, would likely welcome this. Again: WWWCS.

Crime and Policing. One of the clearest areas where Democrats have lost touch with working class sentiments is on the issue of crime. Working class voters tend to live in areas that have more crime and are therefore unlikely to look kindly on any approach that threatens public safety. And yet Democrats, particularly in the wake of activism around George Floyd’s murder, have persistently eschewed a “tough on crime, tough on the causes of crime” that would effectively marry their concerns with social justice to the demand for public safety.

Instead, voters heard about “defund the police” and a wave of progressive public prosecutors who seemed quite hesitant about keeping criminals off the street, even as a spike in violent crime swept the nation. If Democrats had paid closer attention to the polls, they would have seen that even nonwhites were strongly opposed to reducing police presence and, if anything, favored increasing it. In Minneapolis, where Floyd’s murder took place, the closest thing to defunding the police actually got on the ballot…and was soundly defeated, especially by black working class voters. Consistent with this, a recent Pew poll found that black and Hispanic Democrats—who are far more working class–are significantly more likely than white Democrats to favor more police funding in their area.

As London Breed, Mayor of San Franciso, said about her disagreements with the criminal justice ideology of progressive prosecutor, Chesa Boudin: “hese ideologies have been what has failed our city, what has failed Black people in our city, and what continues to be about what beliefs are rather than how those beliefs are going to translate to an actual real difference in someone’s life and the ability to keep people safe….I think a lot of people, like some members of the board, like Boudin, did not grow up in poverty in San Francisco. They did not grow up in these kinds of conditions. They have a theory as to what they believe based on their ideology.”

So, perhaps especially on this issue: WWWCS.

Race/Gender/Language. It’s not exactly a state secret that Democrats have adopted positions on matters or race and gender that are far away from those of the median voter, especially the median working class voter. Voters in general, especially working class voters, tend to think the following:

· Equality of opportunity is a fundamental American principle; equality of outcome is not.

· Discrimination and racism are bad but they are not the cause of all disparities in American society.

· No one is completely without bias but calling all white people racists who benefit from white privilege and American society a white supremacist society is not right or fair.

· There are underlying differences between men and women but discrimination on the basis of gender is wrong.

· There are basically two genders but people who want to live as a gender different from their biological sex should have that right and not be discriminated against. However, there are issues around child consent to transitioning and participation in women’s sports that are complicated and not settled.

· Racial achievement gaps are bad and we should seek to close them. However, they are not due just to racism and standards of high achievement should be maintained for people of all races.

· Language policing has gone too far; by and large, people should be able to express their views without fear of sanction by employer, school, institution or government. Good faith should be assumed, not bad faith.

Compare these with the current outlook and obsessions of the college-educated left within the Democratic party, which tends to have a kind of cultural hegemony within institutions associated with the party. It doesn’t match up too well. The language problem is a notably glaring: how can you speak for working class people if you can’t speak with them without annoying the hell out of them? Jeff Maurer puts it well in one of his substack posts: “The language thing is absolute poison. Words and phrases like “latinx”, “whiteness”, “lgbtqia2s+”, or even — God help us — “uterus-haver” make the speaker sound ridiculous and make the listener feel dumb. The speaker is also obliquely calling the listener a bigot and signaling their membership in a group that nobody wants to join. It’s pretentious and high-handed, and — again — everyone notices. There are ways of being welcoming and inclusive without sounding like an Orwellian robot that fell into a puddle.”

Once again, with feeling: WWWCS.

I’ll grant you that the WWWCS test may be hard for some Democrats to pull off. But they should try if their vision of a just society includes a central role for the Democratic party. In fact, if they’re really serious about that, WWWCS should become an everyday reflex. Otherwise, the working class will still have its say and Democrats might not like what that say is.

David Card argues that there is a growing consensus that firms have wage setting power, but the source of the power is unclear, "....in this paper I will try to make the case that the time has come to recognize thatmany - or even most - firms have some wage‐setting power. Such a shift was made with respect to firm’s price‐setting power many decades ago. Economists now routinely accept that the prices of products like gasoline, breakfast cereal, and ketchup are set with some degree of market power, even in on‐line markets. In the past few years we may have reached a tipping point for a similar transition in labor economics, driven by the combination of new (or at least post‐1930) theoretical perspectives, newly available data sources, and accumulating evidence on several different fronts....Four new types of evidence have accumulated in the past 25 years that suggest employer wage‐setting power is non‐negligible: evidence on quit and recruiting responses to wages; evidence on the relationship between wages and firm productivity; evidence on the concentration of employment in small numbers of employers; and evidence of conspiracies and other forms of firm behavior targeted at suppressing firm‐to‐firm mobility and wage growth...."

David Card, "Who Set Your Wage?" National Bureau Of Economic Research, January 2022, https://www.nber.org/papers/w29683

What Happened in the 1990s?

“…Four new types of evidence have accumulated in the past 25 years that suggest employer wage‐setting power is non‐negligible: evidence on quit and recruiting responses to wages; evidence on the relationship between wages and firm productivity; evidence on the concentration of employment in small numbers of employers; and evidence of conspiracies and other forms of firm behavior targeted at suppressing firm‐to‐firm mobility and wage growth….”

Quit, recruiting, and application elasticities

“…Though many economists acknowledge that quit and recruitment rates vary with wages, the connection between these responses and the elasticity of supply that is relevant for a monopsonistic wage setter does not seem to have been fully appreciated until the seminal paper by Burdett and Mortensen (which circulated for many years prior to its publication). Card and Krueger (1995) noted that in any steady state, the elasticity of labor supply is just the sum of the absolute values of the elasticities of recruiting and quitting. Manning (2003) showed that in a simple job ladder model the two are equal: thus, an analyst can estimate one or the other and double it to yield an estimate of the overall supply elasticity. Manning’s insight provides a tractable method of estimating labor supply elasticities that has been implemented in many different settings. Perhaps the most compelling evidence based on this approach comes from the experiment on public sector hiring conducted by Dal Bo, Finan, and Rossi (2013). These authors randomly varied the salaries announced at different job sites to potential job applicants for a position in the office of the Regional Development Program in Mexico. Taking account of the combined impact of higher wages on application rates and on the probability of accepting a job, they calculate that the elasticity of recruiting with respect to wages is around 2.1 (though rather imprecisely estimated). Using Manning’s shortcut, the implied (steady state) elasticity of labor supply is around 4.2.11 In a simple monopsonistic model such an elasticity implies that wages are marked down relative to marginal revenue products by about 20%. Observational studies of the partial correlation between wages and quit or recruiting rates tend to yield elasticities in the same range (see Sokolova and Sorensen, 2021 for a metaanalysis of this literature and Ashenfelter et al. 2022 for an overview of a studies published in a recent issue of the Journal of Human Resources). For example, Bassier, Dube and Naidu (2022) study job ending rates of workers using administrative data from the State of Oregon, and estimate elasticities in the range of ‐1 to ‐2.5, with a preferred point estimate of ‐2.1, implying a steady state labor supply elasticity of 4.2. Azur, Berry, and Marinescu (2019) use data from a large online job posting service to study the application choices of job searchers. Adapting a “micro BLP” estimation strategy (and using instrumental variables for posted wages) they infer that the firm‐specific elasticity of applications with respect to wages is around 2.9. Assuming the recruiting elasticity is the same as the application elasticity this implies a steady state labor supply elasticity of just under..”

The Relationship Between Wages and Firm Productivity

“…In a competitive labor market, more and less productive firms pay the same wages for workers, even if the more productive firms are larger. In imperfectly competitive labor markets, however, more productive firms will generally have to pay more to maintain a larger workforce. Card et al. (2018) developed a simple partial equilibrium model where workers have MNL preferences over different firms and firms set wages without accounting for strategic interaction effects (i.e., a model of monopsonistic competition). They then calibrated the model to (roughly) match the observed degree of pass‐through from value‐added per worker to wages. In the existing literature researchers typically find that wages are about 0.5 to 1.5 percent higher at firms with 10% higher productivity. In the parameterization of preferences adopted by Card et al., this degree of pass‐through is consistent with firm‐specific supply elasticities of about 4. Lamadon et al. (2022) present a more extensive analysis of the pass‐through of firmspecific and market‐wide value added per worker to firm‐specific wages and interpret the effect in a model of monopsonistic competition with nested MNL preferences over firms and markets. Their estimate of the parameter determining the elasticity of supply to each firm is 5, broadly consistent with the calibration by Card et al. and with the evidence from quit and recruiting elasticities. A related method of estimating the degree of wage‐setting power is to look at establishment‐level responses of employment and wages to an exogenous shock (similar to the pioneering study by Sullivan, 1989). Berger, Herkenhoff, and Mongey (2021) uses evidence on firm‐specific reactions to state tax changes to infer the degree of oligopsony power in a setting with strategic interactions among wage setters (based on Atkeson and Burstein, 2008). They estimate that the average markdown of wages relative to marginal revenue products is around 25% (equivalent to the markdown in a simple monopsonistically competitive model with firmspecific elasticities of around 3.5). Kroft et al. (2020) extend the setup in Lamadon et el. (2022) using information on successful bids in government procurement auctions as firm‐specific demand shocks that affect employment and wages at larger construction firms. They estimate labor supply elasticities in the range of 4‐5…”

The Number of Competitors for Labor Services

“…In thinking about price‐setting or wage‐setting power many economists turn instinctively to the question of how many potential sellers or buyers are present in a market, or to the degree of market concentration measured by the Herfindahl‐Hirschman Index (HHI). As noted by Berry, Gaynor and Scott Morton (2019), Syverson (2019), and Eeckhout (2021), simple measures of the number of competitors or their concentration do not necessarily provide a clear index of market power. Nevertheless, a common perception (among judges for example) is that the number of potential employers for any given worker is large, and that the market power of employers is therefore negligible. One of the most surprising findings in the recent literature is that for many workers in many local markets the number of potential employers is relatively small, particularly when the “market” is defined by actively searching firms.13 Azar et al., (2020), for example, use data on the near‐universe of U.S. vacancy listings to calculate HHI’s for labor markets at the narrowlydefined occupation‐by‐commuting zone (CZ) level. They estimate that an average labor market has an HHI of around 4300 - equivalent to 2.3 equal sized recruiting firms. This is low enough to possibly raise concerns about the effect of mergers and acquisitions on labor outcomes (see Naidu and Posner, 2022). A growing number of papers study the relationship between average wages for a specific subgroup of workers in a given local market and the HHI of potential employers in that market. These studies differ in how they define the set of potential employers (based on industry or occupation), how they count employment (based on the stock of employment, the number of job openings, or some transition‐probability‐adjusted stock of employment), and whether they use a purely observational approach, or implement a research design that isolates some exogenous component of the local HHI. Despite these differences, most recent studies seem to show a negative effect of higher concentration on wages, with elasticities between the HHI and wages on the order of ‐0.05 to ‐0.15. For example, Azar et al. (2022) use data from a large national employment website to study the relationship between posted wages for jobs in a given occupation and commuting zone and the HHI of employers listing vacancies in that occupation and location. They find smaller elasticities of posted wages with respect to the HHI in simple OLS models, but larger elasticities when they instrument the HHI with the leave‐out mean number of competitors searching for workers in that same occupation in other markets. Rinz (2022) estimates HHI’s from counts of establishment‐level employment by CZ and industry, then relates these to administrative earnings from tax data. In OLS models he finds that wages are slightly higher in more concentrated markets, but in models that use the leave‐out mean of the HHI for the same industry in other locations as an instrumental variable, he obtains negative elasticities in the range of Azar et al. (2022). Recent studies by Arnold (2020) and Prager and Schmidt (2021) use event study designs to look at the effects of merger and acquisition activity on local HHI’s and wages. In my opinion, these designs provide the best available evidence that employer consolidations that raise the HHI have significant negative effects on wages, at least for workers who are highly attached to the affected industry…”

Conspiracies and Other Arrangements to Suppress Competition

“…Adam Smith wrote that employers “are always and everywhere in a sort of tacit, but constant and uniform combination, not to raise the wages of labour above their actual rate”.15 He also noted, however, that “(w)e seldom, indeed, hear of this combination, because it is the usual, and one may say, the natural state of things, which nobody ever hears of.” While discoveries of employer collusion are still relatively rare, in the past two decades there have been a number of lawsuits and public disclosures that provide the details of some agreements to suppress competition. These provide a useful perspective on the mechanisms generating market power for employers. The best‐known lawsuit concerned "no poaching" and "no solicitation" agreements affecting software and animation engineers in Silicon Valley (see Ashenfelter et al. 2022 for more details).16 The agreement originated in the mid‐1980s when Lucasfilm sold its computer animation division to Steve Jobs, who then renamed the company ‘Pixar’. To avoid bidding wars over employees, Lucasfilm and Pixar agreed (1) not to "cold call" each other's employees; (2) to notify the other company should they receive an application for employment; (3) and that all offers to employees at the other company would be "final," with no further bidding. Ultimately this agreement was extended to other high‐tech firms (e.g., Google, Microsoft, and Oracle) and lasted over 20 years, until 2008. The size of the settlement to affected engineers ($585 million in two suits), and other wage adjustments made after the agreement was made public (e.g., a 10% across‐the‐board increase offered by Google to all its employees in November 2010) suggest that the suppression of between‐firm competition was successful - a validation of the idea that at least some labor markets are vulnerable to wage fixing. Another interesting lawsuit concerned a “no hire” agreement between the medical schools at Duke University and University of North Carolina (Seaman v. Duke). This case, which resulted in a settlement of around $10,000 for each member of the medical faculties at the two
schools, reveals how localized competition appears to matter, even for workers who arguably face a national market. While one might be tempted to think that “no hire” and “no poaching” agreements affect only highly skilled workers, Ashenfelter and Krueger (2022) found that no poaching clauses were widespread in U.S. franchise agreements.17 These agreements typically prohibit a franchisee from hiring another franchisee's employees for some pre‐specified period of time after an employee's departure. For example, a standard franchise agreement for McDonald’s as of 2016 had a clause stating: “Franchisee shall not employ or seek to employ any person who is at the time employed by McDonald’s, any of its subsidiaries…or otherwise induce, directly or indirectly, such person to leave such employment” (quoted in Ashenfelter and Krueger, 2022). The prohibition extended to employees for 6 months after leaving another McDonald’s job. Another strand of recent research has focused on the prevalence of non‐compete agreements, which prohibit employees from moving to jobs at “competitor” firms for a specified period (e.g., Starr, 2019; Balasubramanian et al., 2022). Again, a surprising fact is the prevalence of these agreements even for relatively low‐wage workers. Recently, however, a number of states have enacted legislation that prohibits non‐compete agreements for “low‐wage” workers (e.g., earning less than $100,000 per year in Washington State - see Goldstein and Oberlander, 2021). The popularity of no‐poaching and non‐compete agreements seems to confirm the basic insights of a BM‐style job ladder model. Since the quit rate in such models depends in part on the rate at which workers obtain offers at other employers, limits on poaching or firm‐to‐firm mobility will reduce quits and increase monopsonistic power…”

defense spending, adjusted for ppP. In ppP terms the US is being outspent by China and Russia combined and that American number includes funding for ongoing operations in the middle east which aren’t relevant vis-à-vis peers, “… I construct a military-PPP exchange rate for 59 countries, using publicly available data. Intuitively this breaks down each country’s military budget into three broad categories, personnel, operations, and equipment, and then constructs a relative unit cost ratio across counties for each component. The military-PPP exchange rate is then the average of relative equipment prices; relative wages for personnel, adjusted for skill levels; and the GDP-PPP exchange rate to measures operations costs - where the budget share of each component is the appropriate weight. Thus, the military-PPP exchange rate will take account of factors such as low salaries for personnel and appropriately weight these costs in the overall defence budget. The approach yields values that can be interpreted as a first-order approximation to the true relative military unit cost ratio, under standard assumptions of cost minimisation and constant returns to scale...In market exchange rate terms, these three countries have combined military spending in 2019 that amounts to $378 billion, compared to the US’s $734 billion. In military-PPP terms, however, their combined spending is $840 billion (Figure 1). Hence, in real terms, the US military budget in 2019 is not larger than the next eleven countries. Rather it is smaller than the next three and similar to the spending of China and Russia combined….”

Peter Robertson, "Debating defence budgets: Why military purchasing power parity matters," Center For Economic and Policy Research, October 9, 2021, https://voxeu.org/article/why-military-purchasing-power-parity-matters

Debating defence budgets: Why military purchasing power parity matters

In his 2016 State of the Union address, US President Obama claimed that “he United States of America is the most powerful nation on Earth. Period. It’s not even close. We spend more on our military than the next eight nations combined.”

Since then, US military spending has only increased, and today, measured at market exchange rates, it is greater than the next 11 countries combined (Stockholm International Peace Research Institute 2021, Peterson Foundation 2021). This apparent difference makes a mockery of US defence budget requests and of claims that any country poses a significant threat.1

In this column, I show that the perception of the US military ‘spending supremacy’ is misplaced. Specifically, the conventional use of market exchange rates to compare across countries dramatically overstates US real military spending relative to other countries and, likewise, understates the real spending in countries such as China, India, and Russia.

What is militaryppP?

The need to adjust for local price differences when making international comparisons of real GDP or standards of living is well recognised. Because relative prices differ across countries, a dollar spent on non-traded goods and services can go a lot further in some countries than in others. This is known as the Penn effect or the Balassa-Samuelson effect, and reflects the fact that the market exchange rate only arbitrages the price of traded goods, so international relative prices of non-traded goods can differ widely from the exchange rate.

The same applies to military spending. Soldiers in the Chinese People’s Liberation Army (PLA), for example, receive far lower salaries than the equivalent US personnel (The Economist 2021). Thus, the PLA’s salary budget, if converted to US dollars at the market exchange rate, would pay for far fewer US army personnel than what it actually pays for in China.2

In response to this, international comparisons of military spending are sometimes made using GDP-purchasing power parity (PPP) exchange rates. These, however, are designed to compare average prices, not military costs; they could overstate military spending since procurement doesn’t necessarily depend on local labour costs (Gilboy and Heginbotham 2012, Frankel 2014).

The alternative views and exchange rate concepts have long been a source of frustration for defence planners, particularly with the enormous variations they imply for China, which has been as much as 200% (Crane 2005).

In fact, neither GDP-PPP exchange rates nor market exchange rates provide the ‘correct’ ratio for converting spending across two countries into real values. The appropriate exchange rate is one that tells us the relative price or unit cost of defence services across countries - that is, a ‘military-PPP’ exchange rate.

The problem in constructing a military-PPP exchange rate is that output prices and quantities of defence services are unobservable and, perhaps unsurprisingly, few countries have shown much inclination to share the necessary input cost data. There is also a problem of substitution bias, which arises because countries will choose different combinations of inputs - such as personnel and equipment - because they have different relative prices.

The challenge, therefore, is to find the best estimate of countries’ real defence budgets given the available data. In Robertson (2021), I construct a military-PPP exchange rate for 59 countries, using publicly available data. Intuitively this breaks down each country’s military budget into three broad categories, personnel, operations, and equipment, and then constructs a relative unit cost ratio across counties for each component. The military-PPP exchange rate is then the average of relative equipment prices; relative wages for personnel, adjusted for skill levels; and the GDP-PPP exchange rate to measures operations costs - where the budget share of each component is the appropriate weight.

Thus, the military-PPP exchange rate will take account of factors such as low salaries for personnel and appropriately weight these costs in the overall defence budget. The approach yields values that can be interpreted as a first-order approximation to the true relative military unit cost ratio, under standard assumptions of cost minimisation and constant returns to scale.3

US military spending is about the same as China's and Russia's combined

Looking at military spending this way reveals that real military spending in many countries is indeed far larger than the values implied by market exchange rates. To see this, rather than looking at the US relative to the next eight countries, we can focus on just three other countries - China, India, and Russia.

In market exchange rate terms, these three countries have combined military spending in 2019 that amounts to $378 billion, compared to the US’s $734 billion. In military-PPP terms, however, their combined spending is $840 billion (Figure 1). Hence, in real terms, the US military budget in 2019 is not larger than the next eleven countries. Rather it is smaller than the next three and similar to the spending of China and Russia combined.

Real defence spending versus military power

Whether the US is spending too much or not enough is a hotly debated issue andppP concepts are sometimes invoked (e.g Korb 2021 and Greenwalt 2021). As tragically demonstrated by the collapse of the Afghan National Army, however, spending isn’t everything. Comparing real military spending across countries amounts to comparing military inputs, not military output. Military output, or ‘power’, also depends on a country’s defence strategy, alliances, force multipliers and other non-budget factors.

This, however, is true of all international comparisons of military spending, irrespective of what exchange rate concept is used. Using military-PPP does not remove the need for an expert eye to how effective a country’s spending is. It does, nevertheless, provide a more solid economic basis for thinking about these issues.

Caution is also required since the military-PPP values discussed here are based on very aggregate data and involve approximations. They can be improved upon, particularly with more disaggregated data. The key strength of the military-PPP approach is it sets out an exchange rate concept that has a clear economic meaning.

With these caveats, the results show that market exchange rates dramatically understate the real military spending of many countries including China, India, and Russia. Defence analysts will have their views on what this means in terms of relative capabilities. But evaluations of security risks can only be improved if they avoid exchange rate misperceptions.

References

Beinart, P (2021), “Let’s cut our ridiculous defense budget”, New York Times, 7 April.

The Economist (2020), “Buck for the bang: Nominal spending figures understate China’s military might”, 1 May.

Frankel, J (2014), “China is not yet number one”, VoxEU.org, 9 May.

FederalPay.org (2021), “United States Army - O-6 colonel”, accessed 25 September.

Gilboy, G J, and E Heginbotham (2012), Chinese and Indian strategic behaviour: Growing power and alarm, New Delhi: Cambridge University Press.

Greenwalt, B (2021), “China already outspends US military? Discuss”, Breaking Defense, 26 May.

Korb, L (2021), “China is not outspending US on defense; ’22 budget is enough”, Breaking Defense, 7 June.

Peter G Peterson Foundation (2021), “The US spends more on defense than the next 11 countries combined”, 19 July.

Robertson, P E, and A Sin (2017), “Measuring hard power: China’s growth and military capacity”, Defence and Peace Economics 28(1): 91-111.

Robertson, P E (2015), “China’s military and growing political power”, VoxEU.org, 30 March.

Robertson, P E (2021), “The real military balance: International comparisons of defense spending”, Review of Income and Wealth (forthcoming).

The South China Morning Post (2021), “China to boost pay for PLA forces after years of sweeping reforms”, 21 January.

Stockholm International Peace Research Institute (2021), SIPRI military expenditure database, accessed June 2021.

Zakaria, F (2021), “The Pentagon is using China as an excuse for huge new budgets”, Washington Post, 18 March.

Endnotes

1 Fareed Zakaria, writing in March 2021, argued, “America’s ‘edge’ over China is more like a chasm … China spends around $250 billion on its military, a third as much as the United States” (Zakaria 2021). Similarly Peter Beinart criticised US President Biden’s 2022 Department of Defense budget, arguing that “China spends less than one-third as much on defense as the United States” (Beinart 2021).

2 As an example, a colonel in the US army earns a salary of $7,100 to $12,638 per month (FederalPay.org 2021). Converting this to renminbi at the market exchange rate gives enough renminbi to pay for two to four colonels in the People’s Liberation Army, given a monthly salary of RMB 20,000 as reported by The South China Morning Post (2021).

3 Robertson (2021) discusses the implication of various standard index number approaches including bilateral and multilateral indices. The method builds on previous work on China in Robertson (2015) and Robertson and Sin (2017).

4 Military spending is also an annual flow, whereas capabilities depend on past spending assets and experience.

Note Wikipedia’s ranking of defense spending by countryhttps://en.wikipedia.org/wiki/List_of_countries_by_military_expenditures

Every year between 2015 and 2018 an average of 40,641 individuals between ages 18 and 64 died of opioid overdoses. A large majority of them, 37,596, did not have a college degree. The rest, 3,045 of them, were college graduates. The benchmark economy matches these statistics exactly. The downward spiral from opioid use is portrayed in Figure 8. It shows how the college- and non-college-educated' expected utilities steadily decline as they move through the stages of opioid addiction. The descent appears fairly gradual until one hits the addiction stage and, of course, the loss in utility associated with death is large. While utility is always higher for college graduates, the relative utility values of death to nonuse are roughly the same for both types of individuals. Backin 2000, the number of opioid-related deaths was only 8,179 (7,549 deaths among non-college and 629 among college graduates). Between 2000 and 2018, more than 400,000 individuals between the ages of 18 and 64 died from opioid overdoses. What can account for the dramatic rise in opioid use and overdose deaths during the last two decades? The model is now used as a quantitative laboratory to answer this question. Three candidates are entertained; namely, the fall in opioid prices, more powerful prescriptions, and longer length prescriptions…The calibrated model is then used to highlight the forces underlying the recent opioid epidemic. Through the eyes of the model, there were two key forces. The first force isthe decline in prices for both prescription and black market opioids. Thishad a big effect. The second force isthe increase in the dosages per prescription meted out by doctors. Thisalso had a significant impact. The fact that doctors kept pain suffers on prescription opioids for a longer period of time had little effect. Last, an analysis is conducted onmedical interventions that reduce either the probability of becoming addicted or the odds of an addict dying from an overdose. While such interventions are valued by consumers, theyincrease the number of opioid users.Reducing the odds of addiction can result in even more deaths due to the rise in users…The developed framework is taken to the US data for both the college- and non-college educated populations. The quantitative analysis has three key steps. The first step is the estimation of Markov chains characterizing the movements in and out of misuse and addiction. Death is an absorbing state. The model is calibrated in the second step to match the estimated transitions from the Markov chains for both the college- and non-college educated. The framework fits the US data well. A check is performed on the calibration, by examining whether the model's prediction on the relationship between prescription opioid access and opioid deaths is consistent with the cross-state evidence about this. The calibrated framework is then used to decompose the rise in opioid usage. The analysis suggests that drops in the price of both Rx and illicit opioids combined with increases in Rx dosages were the primary drivers of the opioid epidemic.

The evidence, “…In 2019 the age-adjusted death rate from an opioid overdose was 21.6 per 100,000 people. This compares with 12.9 for kidney disease, 14.2 from suicides, 14.7 for influenzas, 21.6 from diabetes, and 161.5 from heart disease (the leading cause of death in the United States). Opioid overdose deaths place in the top 10 leading causes of death in the United States. As can been from Figure 1, prior to 2015 most of these opioid deaths arose from prescription (Rx) overdoses, but after that they came from synthetic opioids; in particular, fentanyl. (The sources for all the data displayed in the figures are presented in Appendix B.) The overdose death rate was much higher for those without a college degree compared with those who had one. The rise in the death rate from synthetic opioids is particularly marked for the non-college educated population….”

Jeremy Greenwood, Nezih Guner and Karen Kopecky, "The Downward Spiral," National Bureau Of Economic Research, https://www.nber.org/papers/w29764

“…Figure 4 shows that the out-of-pocket expense for prescription opioids has fallen by a factor of 3 since 2001. This has been attributed to two factors. First, the advent of generic prescription opioids. Second, the expansion of social programs such as Medicare and Medicaid that subsidized the purchase of opioids, as can be seen from Figure 5. For the college educated Medicaid is less important than private payers while for the non-college educated the reverse is true. The share of opioids prescriptions funded by the government grew from 17 percent in 2001 to 60 percent in 2010. The vast majority of opioids went to people who needed relief from pain caused by either disability or illness. Over the same period the street price of opioids dropped by a factor of 3. This has been chalked up to both the illegal imports of inexpensive powerful synthetic opioids, for example fentanyl, from China and elsewhere. Additionally, the diversion of opioids from legal sources onto the black market via fraudulent prescription, family and friends giving away and/or selling their prescriptions, and theft. The rise of illegal imports is ascribed to the tightening of prescriptions and the introduction of a tamper-proof form of OxyContin. The upshot is that opioids are much less expensive now than they were in 2001. Likewise, the introduction of low-cost heroin at the beginning of the 20th century was due to the banning of smoking opioids and the increased restrictions on the use of cocaine…”

“.. Starting around the year 2000 there was a dramatic increase in number of opioid prescriptions per person for both the college- and non-college-educated populations, as shown in Figure 3. The non-college educated were much more likely to have an opioid prescription than the college educated. The former often work in occupations involving physical labor. Additionally, the amount of Rx opioids consumed, conditional on a prescription, also rose. Again, this was particularly true for those without a college degree. The price of prescription opioids has fallen dramatically since 2000…”