A Long-Run Perspective on US Economic Growth

For those not familiar with the Economic Report of the President, it is published each year by the White House Council of Economic Advisers. In turn, the CEA is led by academics, who are appointed by the president but typically plan to head back to their ivory towers in a few years. Thus, they are clearly a partisan and pro-administration group, but they also have reason to care about their own reputation for expertise and for relatively dispassionate analysis. This tension plays itself out each year in the report.

The parts of the ERP each year that offer a partisan defense of the president aren’t that interesting to me, no matter who the president is, because of how one-sided and perfunctory such defenses tend to be. Of course, if you are looking for talking points to support the economic policies of the Biden administration or if you want to take target practice against those policies, you may be attracted to those parts of the report. But each year, the report also includes facts and nuggets about the US economy and its trends and patterns that have emerged from discussion from thoughtful academics, and some of these can be worth passing on. Here’s one from the first chapter of the 2023 Economic Report of the President, showing the long-term average for US economic growth going back to 1790.

The figure breaks down overall economic growth into three chunks: growth of population (which means more workers and consumers), changes in labor force participation (the share of the adult labor force that either has a job or is looking for a job), and output per worker, which changes according to improvements in human capital (education and skills), physical capital available to workers, and “total factor productivity,” which is econo-speak for productivity improvements. Here are a few reaction to the figure:

1) The slowdown in overall economic growth in in the 2000s is readily apparent. But in a broad historical perspective, it’s also apparent is that a lot of this slowdown is due to a slower rate of population growth (shorter dark blue bars) and also a decline in labor force participation due in part to the aging and retirement of the “baby boom” generation born in the 15 years or so after the end of World War II (the light blue bars in negative territory on the graph). At least in the last decade, output per worker hasn’t been rising at an especially slow rate.

2) For political scientists and those interested in global politics, the sheer size of the US economy matters–the total height of these bars. But for economists, what matters more is a gradually rising standard of living for the average person, which is roughly captured over time by the gain in output-per-worker.

3) The future of US economic growth isn’t likely to come from population growth; instead, it will need to be generated by higher output per worker. The US economy had a mass expansion into high school education from about 1910 to 1940, and a mass expansion of higher education after World War II, but no mass expansions of education since then. US capital investment seems OK, but a lot of one’s thinking around that issue revolves around placing an economic value on information technology and internet access, which isn’t easy to do. Productivity gains are calculated as the residual of what is left over, unexplained, by forces like labor force growth, human capital and physical capital–and by that measure, the US economy isn’t doing especially well since the early 2000s.

4) The 1870s appear on the graph as a time of rapid growth. I confess that I don’t understand this. The report emphasizes that the 1870s are a time of expanding the railroad and the telegraph, along with new inventions. However, the standard dating of US business cycles suggests that the US economy was in the “Long Depression” from October 1873 to March 1879. Maybe it was just a really extraordinary economic boom in the early 1870s in the aftermath of the Civil War?

5) From the perspective of decade averages, the Great Depression of the 1930s looks less “great,” in the sense that overall growth during the decade of the 1930s was similar to that of the 1910s and 1920s. In part, this is probably because we tend to understate the multiple deep recessions of these earlier decades, including three recessions in the 1910s and another three in the 1920s, as well as understating how the US economy recovered from the Great Depression in the later part of the 1930s (albeit with a recession in 1937-38). It may seem odd that labor force participation doesn’t fall noticeably in the 1930s, given the ultra-high unemployment rates of the time. However, the unemployed are counted as “participating” in the labor market–to be outside the labor force participation rate, you need to be not looking for work (say, retired or working in the home by preference).

6) In the 1970s and 1980s, you can see that a noticeable chunk of overall economic growth was the rise labor force participation, mainly due to growing participation of women in the (paid) labor force.

There’s a remarkable economic story behind every bar and line in this graph.

Catastrophes and Costs: Some Trendlines

Each year, the Swiss Re Institute has been publishing an annual report on the natural catastrophes of the previous year, of particular interest to specialists, and some long-run trendlines, which are more interesting to me. This year’s report is “Natural catastrophes
and inflation in 2022: a perfect storm”
(Sigma, 2023, No. 1).

Here’s the number of catastrophes around the world. in the last half-century. Natural catastrophes include “floods, storms, earthquakes, droughts/forest fires/heat waves, cold waves/frost, hail, tsunamis, and other natural catastrophes.” Man-made catastrophes do not include war, but instead are divided into the categories of “major fires and explosions, aviation and space disasters, shipping disasters, rail disasters, mining accidents, collapse of buildings/bridges, and miscellaneous (including terrorism).”

It’s wise to interpret these numbers with care. After all, the global population has more than doubled since 1970, which more-or-less fits the rise up to the early 2000s, although it doesn’t explain the spike it he mid-2000s or the more recent decline (which precedes the pandemic). Also, if a natural event that would be a catastrophe if it happens in a heavily populated area occurs instead in a lightly populated area, does it count as a catastrophe of the same scale? The world’s apparatus for discovering and reporting natural catastrophes was clearly less developed a half-century ago. That said, it appears that man-made disasters have been declining in the last decade or so, while natural catastrophes have been trending up over time.

What about the costs? Let’s look first at lives lost. Notice that the vertical scale is logarithmic, not arithmetic (that is, it rises by factors of ten). The report notes: “Worldwide, 35 157 people are believed to have died or gone missing in disaster events in 2022. Natural catastrophes claimed over 32 600 victims, and man-made disasters over 2500.”

Here, the general pattern seems to be that deaths from man-made disasters are both an order of magnitude lower than natural disasters, and also declining. There doesn’t seem to be any particular trend to deaths from natural disasters, just a pattern of big spikes when an especially awful disaster occurs somewhere.

Finally, here’s an estimate of total financial costs of these disasters, and because Swiss Re is a reinsurance company, the share of those costs covered by insurance. These costs are adjusted for inflation. The sharp rise is because of economic growth and rising property values: for example, a hurricane hitting Florida now will have a higher financial cost than the same hurricane hitting in 1970.

It’s also worth remembering that financial costs will depend on local prices, such that financial costs will be higher in high-income, high-price countries. The report notes that the gap between insured financial losses and total financial losses seems to be rising: it was 53% in 2022, down from 59% over the average of the last ten years.

Interview with Annamaria Lusardi on Financial Literacy

David A. Price interviews Annamaria Lusardi “on financial literacy, seniors versus scammers, and learning from the mistakes of NFL players” (Econ Focus: Federal Reserve Bank of Richmond, First Quarter 2023, pp. 24-28). Lusardi notes:

[W]e have witnessed a highly important change in the United States and around the world, which is that more and more, we have shifted the responsibility to save for retirement from the employer to workers. I am talking about the shift from defined benefit pensions to defined contribution pensions, such as individual retirement accounts and 401(k) plans. In the past, it was the employer who had to manage the pension of the employees; the wealth was managed by a CFO or by other financial experts. Now we ask individuals to make these decisions about their wealth. So even more than when I was an assistant professor, there’s the question of whether people have the skill to manage their money.

Lusardi and Olivia Mitchell have designed a 28-question test to measure financial literacy, which has become a widely used research tool. For a flavor of the kinds of questions, consider what they call the “Big Three.” In their early work, an advantage of having just three questions is that it was practical to add three questions to piggyback on preexisting surveys.


  1. Suppose you had $100 in a savings account and the interest rate was 2% per year. After 5 years, how much do you think you would have in the account if you left the money to grow?
  • More than $102
  • Exactly $102
  • Less than $102
  • Do not know/Refuse to answer

2.Imagine that the interest rate on your savings account was 1% per year and inflation was 2% per year. After 1 year, how much would you be able to buy with the money in this account?

  • More than today
  • Exactly the same
  • Less than today
  • Do not know/Refuse to answer

3. Please tell me whether this statement is true or false. “Buying a single company’s stock usually provides a safer return than a stock mutual fund.”

  • True
  • False
  • Do not know/Refuse to answer

NOTE: Correct answers are (1) “More than $102,” (2) “Less than today,” and (3) False.
SOURCE: Annamaria Lusardi and Olivia S. Mitchell

The ongoing research in this area suggests both that financial literacy is low.

Together with a team at the World Bank, I eventually designed questions similar to the big three that were applied to a sample of more than 140 countries. I would say there are several interesting findings. One is that even though the U.S. is the country with the most advanced financial markets, it actually doesn’t score very high in terms of financial literacy. And this has been true in other surveys, as well. The second thing is that overall financial literacy is not high in other countries, either. Overall, the level of financial literacy globally is really low; only one-third of people around the world are financially literate. …

[W]hat we did recently — and it took us a good many years to do this project — is a meta-analysis of financial education programs. … Because the literature was so extensive, we then decided to concentrate on the most rigorous evaluations. So we looked at only the randomized control trials. … So you expose a group to financial education; you don’t expose the other, similar group; and then you compare what happened to the group you treated. What we found, looking at the evidence in as many as 33 countries, is that financial education works and works well — meaning it does translate into higher knowledge and also better behavior in savings and managing credit and in other areas, including insurance and money transfers. And we also found that it is cost effective. This is due to the fact that many educational programs do not cost very much.

This work has implications reaching in a number of directions. Prominent examples include professional athletes who, even in a relatively short career, might earn as much as the average college-educated person will earn in a lifetime. But these athletes are young adults, their financial literacy is no greater than average, and they are easy targets for financial “advisers” and “planners” who charge high fees for high-risk options. A less prominent but much larger group are the elderly near retirement age, at a stage when they probably have the highest level of assets for their lifetime, but again, their financial literacy is no greater than average, and at they can often find themselves to be targets for high-fee and high-risk “advisers” and “planners.” Indeed, Lusardi’s work has found that the current elderly are often taking more debt into their retirement than previous generations.

Mervyn King: “Our Ambition at the Bank of England is To Be Boring”

Back in 2000, the Deputy Governor of the Bank of England, Mervyn King, gave a speech about monetary policy that has been often-quoted by central bankers around the world (“Balancing the Economic See-Saw,” April 14, 2000). He said:

[O]ur ambition at the Bank of England is to be boring. Not, I hasten to add, at events like this. But in our management of the economy where our belief is that boring is best. Macroeconomic policy has, for most of our lifetime, been rather too exciting for comfort. … Our aim is to maintain economic stability. A reputation for being boring is an advantage – credibility of the policy framework helps to dampen the movement of the see-saw. If love is never having to say sorry, then stability is never having to be exciting.

King’s reference to the “see-saw” is pointing out that monetary policy involves movements between looser and tighter monetary policy. Some such movement is inevitable. But of course, the goal is to have the see-saw of macroeconomic policy involve small adjustments, rather than big swings. King argued that central banks should be willing to take small actions sooner, because otherwise they are likely to need to take bigger actions later. He said:

The longer the correction is left, the sharper the required adjustment will be. The higher one end of the see-saw, the greater the subsequent lurch will be. … In one of the most influential contributions to monetary policy in the post-war period, Milton Friedman wrote that the characteristic of most central banks was that “too late and too much has been the general practice”.

The often-heard recent complaint about the Federal Reserve and inflation was that it waited too long after inflation started in 2021, and then had to act more aggressively to raise interest rates starting in 2022 than would otherwise have been needed. The current concern about monetary policy is that perhaps the Fed has already taken sufficient actions to bring down inflation, but it takes some time for the past hike in interest rates to work through the macro-economy. By not waiting to see what happens from its past actions, the Fed runs a risk of overreacting. The best best description of this phenomenon that I know come from Alan Blinder, who was vice-chair of the Fed in the mid-1990s. He wrote in a 1997 article in the Journal of Economic Perspectives:

[H]uman beings have a hard time doing what homo economicus does so easily: waiting patiently for the lagged effects of past actions to be felt. I have often illustrated this problem with the parable of the thermostat. The following has probably happened to each of you; it has certainly happened to me. You check in to a hotel where you are unfamiliar with the room thermostat. The room is much too hot, so you turn down the thermostat and take a shower. Emerging 15 minutes later, you find the room still too hot. So you turn the thermostat down another notch, remove the wool blanket, and go to sleep. At about 3 a.m., you awake shivering in a room that is freezing cold …”

But setting aside the questions of whether the Fed waited too long to act (I think it did) or whether it is currently in danger of overreacting (I think not yet), perhaps the bigger complaint is that when one thinks back over the last 20 years or so of monetary policy, what Mervyn King called the “see-saw” of policy has shown dramatic shifts. It’s not just the rise and fall of the federal funds interest rate–the policy interest rate targeted by the Fed.

It’s also the policies of quantitative easing, due to which the Fed now holds about $8 trillion in Treasury bonds and mortgage-backed securities. I It’s the shift toward the use of “forward guidance,” in which the Fed seeks to shift interest rates and financial conditions in the present by making announcements about the likely course of future Fed policy.

It’s the fact that the Fed has fundamentally shifted its tools of monetary policy policy. A couple of generations of economics students were taught about the three tools of monetary policy: open market operations, reserve requirements, and the discount rate. But the Fed abolished reserve requirements in 2020, and open market operations only worked because banks wanted to avoid not holding enough reserves. Instead of discount rates, the Fed now creates of funds for short-term liquidity, which spring up during the Great Recession or the pandemic recession to reassure markets, and then vanish again. The Fed now seeks to contorl the federal funds interest rate now happens through payment of interest on bank reserves held at the Fed, which didn’t start until 2008, and using overnight reverse repurchase agreements, which didn’t start until 2013.

One can add to this some of the recent debates over whether the goals of Federal Reserve policy should reach beyond the standard see-saw of balancing risks of unemployment and inflation, and also try to take into account possible effects of monetary and banking regulation policy on issues like inequality and climate change.

Of course, the most recent example of the Fed policy see-saw is the meltdown at Silicon Valley Bank and its aftermath. Apparently, neither the Federal Reserve’s monetary policy arm nor its bank regulation arm managed to notice the elementary fact that a monetary policy decision to raise interest rates would affect the value of fixed-interest-rate bonds held by banks (as well as the value of similar assets held by the Fed itself). As a result, the Fed ended up making a rather sudden decision to guarantee all bank deposits , even those above the previous limit of $250,000, at certain “strategic” banks–a guarantee which in practice seems likely to apply to just about any bank in trouble.

It remains true in 2023, as Mervyn King said back in 2000: “Macroeconomic policy has, for most of our lifetime, been rather too exciting for comfort.” I have a reasonably good understanding of the the reasons and rationales for the various Fed policy changes in the last 20 years. But it’s worth remembering King’s other ambition as well: The Federal Reserve, along with other central banks around the world, needs to be more boring.

Some Economics of Pandemic Vaccination

The Oxford Review of Economic Policy has published a 10-paper symposium on the “Economics of Pandemic Vaccination.” Here, I’ll focus on the first overview essay, by Scott Duke Kominers and Alex Tabarrok, titled “Vaccines and the Covid-19 pandemic: lessons from failure and success” (Winter 2022, 38:4, pp.  719- 741). They write: “The expected costs of a future pandemic easily exceed a trillion dollars, thus justifying large expenditures and thought. Our knowledge of pandemics and their costs has been hard-won. What have we learned from global failures and successes in combating Covid-19?”

The makers of vaccines capture only a very small share of the benefits of the vaccines. Thus, their incentive to produce such vaccines is much less than the social benefit of the vaccines.

The vaccine industry, however, can only capture a tiny fraction of the gains from successful vaccines. Calculations from Susan Athey, Juan Camilo Castillo, Esha Chaudhuri, Michael Kremer, Alexandre Simoes Gomes, and Christopher M. Snyder (2022, this issue) and Castillo et al. (2021) suggest that vaccines had a value on the order of $5,800 per course when the price was running at just $6–$40 per course. As a result of the ‘enormous gulf between social and commercial incentives’ (Nordhaus, 2004Ahuja et al. 2021), there was an important role for government investment in vaccine research and development, as well as in production capacity.

The government incentives for vaccine production need to include speed, not just volume. Imagine that the government wants a company to produce 100 million doses of a vaccine. However, the preference of the government is to produce all 100 million as soon as possible, which means the company would need to ramp up production very fast–and then production would drop off. Thus, the company would prefer to spread out production over time, so it can make longer-lasting use of smaller production facilities. Thus, appropriate government policies involve both “push” policies to create a vaccine and “pull” policies to encourage rapid and large-scale production.

Every country underinvested in vaccines–in fact, the US did more to encourage vaccines than most.

It is important to note that underinvestment was not simply a US problem—every country underinvested in vaccines (Ahuja et al., 2021). In fact, Operation Warp Speed was by far the largest vaccine investment programme globally, so whatever problems reduced the effectiveness and scale of the US response may have been far larger elsewhere. Global underinvestment in vaccination may in part have been a result of human psychology—voters tend to reward politicians for dealing with emergencies, but not for avoiding them (Healy and Malhotra, 2009), and in the case of Covid-19, the scales in question may have been especially hard to contemplate. Human psychology may also help to explain why it appears to have been harder to spend trillions on a war against a virus than on wars against other people.

The role of the Defense Production Act in requiring firms to produce inputs for the vaccines was mixed, and deserves additional study.

Operation Warp Speed was about more than just spending. The US Department of Defense and the Defense Production Act also played key roles. The Defense Production Act is a 1950 law that gives the US president significant authority to direct civilian production to uses deemed necessary for national defence. This is sometimes mistakenly thought of as a type of command and control—an order to produce—but in practice, the Defense Production Act was mostly used to supplement the market process due to some of the limitations of investments in capacity. …

The Covid-19 pandemic necessitated vaccination on a larger scale than ever before, and thus many inputs were in net under-supply. Yet instead of allowing the price of vaccines to rise and feed into input prices, governments held prices low, and subsidized some stages of production such as clinical trials. … But governments are unlikely to have the requisite knowledge to see and coordinate the entire supply chain and its multiple substitutes, complements, and opportunity costs (Hayek, 1945). Operation Warp Speed used the Defense Production Act to imperfectly and temporarily substitute for the signalling and incentive role of prices. Instead of prices being transmitted along the supply chain, the Defense Production Act authorized priority ratings to be transmitted along the supply chain—thus, a firm given a priority rating of DO (critical to national defence) could (indeed, it had to) pass that priority rating on to its input suppliers, who in turn would pass the rating on to their suppliers, and so forth.

The Defense Production Act may have been useful in this regard, but it was not without cost. Bown (2022) notes that because the Defense Production Act forbade firms from raising prices, it likely reduced incentives to invest in new capacity, muting the long-run supply response—and, as Covid-19 wore on, the ‘long run’ quickly became the present. The Defense Production Act also allowed less substitution across inputs than might have been possible using market-clearing prices. For future pandemics, it will be important to figure out the ideal balance between using market forces and government management to drive vaccine supply.

There is a problem in pandemic of how to decide when a certain possible treatment should be pursued further, or set aside. The problem is that there is a limit on the number of treatments that can be evaluated. If you give up on a possible treatment too soon, that’s obviously bad. But if you keep pursuing at treatment that doesn’t work, there is a tradeoff of not pursuing an alternative that might have worked. How do you know when to stop?

When the pandemic began, there were no known treatments for Covid-19. Thus, flexible trial designs that dropped inefficacious treatments and added potentially efficacious treatments in real-time were critical. These sorts of ‘adaptive’ trials—including the British RECOVERY trial—were instrumental in quickly discovering useful treatments, but they also raise very complex statistical issues which are best analysed in advance. For example: how much evidence should one require to ‘prove’ that a treatment is inefficacious? Setting a high standard could mean testing an inefficacious treatment for too long, with deleterious consequences for patients and for treatment discovery. Clinical trial resources are limited, so the opportunity cost of testing an inefficacious treatment is testing a potentially efficacious one. 

Shouldn’t we be preparing to be able to vaccinate for the next pandemic now?

[A]s the pandemic has dragged on, funding for ongoing vaccination and therapeutic efforts, as well as continued R&D, has been lacking (see, for example, US White House (2022)). Next-wave vaccines—most crucially, nasal vaccines (which may be more effective at reducing transmission) and pan-coronavirus vaccines (which would provide immunity against many variants at once)—are in development (Topol and Iwasaki, 2022) but, at least so far, we have seen nothing like an Operation Warp Speed-level push. With Covid-19 continuing to cause significant morbidity and mortality worldwide, not to mention ongoing economic disruption, estimates such as those of Castillo et al. (2021) suggest that an Operation Warp Speed 2.0 to combat the ongoing threat of Covid-19 would be highly cost-effective.

There are a variety of other questions. When distributing vaccines with in a country, for example, does it make sense to prioritize those who are most likely to die of the illness, or those who are most likely to be infected, or those who are most likely to spread the infection? These groups overlap, but are not the same. Also, should the main focus of vaccine distribution be decentralized–say, through pharmacies–or centralized in government distribution centers? When distributing vaccines internationally, is it better to spread the limited supply of available vaccines across countries, with a low proportion of the population getting the vaccine, or to focus the first wave of vaccines on a subgroup of countries, where a higher proportion of the population within those countries can get the vaccine? And what kinds of contracts make sense for international distribution?

For me, one of the oddest things about the US pandemic experience was that, according to the Global Health Security (GHS) Index produced by the Johns Hopkins Center for Health Security, the US was in 2019 the country in the world most ready for a pandemic. But it turns out that this measure of pandemic preparedness was essentially useless.

Robert Tucker Omberg and Alex Tabarrok (2022, this issue) examine the Global Health Security (GHS) Index, a comprehensive measure of pandemic preparedness that was produced by the highly regarded Johns Hopkins Center for Health Security before the pandemic. The upshot of their paper is that countries that were ranked highly in pandemic preparedness did not perform better during the pandemic, whether looking at infections, confirmed deaths, excess deaths, the country-by-country time series of excess deaths, or other outputs. Indeed, almost no aspect of the GHS Index helps to predict pandemic outcomes, even after controlling for a variety of demographic factors.

Surely (he says in a pathetic and pleading tone), we can do better in being prepared when (not if) the next pandemic comes?

The Failure of Free Community College in Oregon

Economists have a knee-jerk negative reaction to proposals that are phrased in terms of “free,” whether its “free” school lunch, or “free” health care, or “free” housing, or “free” tuition. In each case, the issue isn’t whether the program is a good idea or not. It might be. The issue is that just because something is provided at no cost to a user doesn’t mean it is actually “free,” only that the costs are being paid in some other way. Thus, “free” programs need to be evaluated, like any others, based on who receives benefits and pays costs, not on a pretense that saying “free” solves the problem.

Back in 2015, Oregon passes legislation for the “Oregon Promise,” which said the state would pay the average cost of Oregon community college tuition for Oregon high school graduates. The same legislation required that a state agency called the Higher Education Coordinating Commission produce a report evaluating the Oregon Promise every other year, and the 2022 report is now out.

The hope of the Oregon Promise, of course, was that it would encourage greater college attendance, especially among those from families with lower incomes or from backgrounds traditionally under-represented in colleges. Here’s the summary of the findings:

We found that in the first two years of the program, enrollment rates rose, but declined in the last four years, especially due to the impact of the pandemic on college enrollment. The initial implementation of the Oregon Promise was associated with a clear increase in enrollment at the colleges, and early enthusiasm and attention to the program seemed to realize the program’s goals of opening the door to postsecondary education and training wider. After six years, these early increases have not been sustained, as community college and statewide college-going rates are lower than prior to the Oregon Promise program.

We did not find evidence to suggest the Oregon Promise is associated with reducing equity gaps in college-going rates. Racial/ethnic gaps in college-going rates were similar before and after the implementation of the program, at least until the pandemic. Although college-going rose for Black/African American and for Latino/a/x/Hispanic graduates increased, this increase started before the Oregon Promise and therefore is not attributable to the Oregon Promise program. Gaps in college-going rates by geography and gender widened since the program began. …

We found that Oregon Promise recipients are generally representative of their high school graduating class, though they are somewhat more likely to be women and Latino/a/x/Hispanic. Additionally, students with the Oregon Promise are more likely to be from low-income backgrounds and from urban areas. We also noted that because of the last dollar structure of the program, the vast majority of Oregon Promise dollars go to students from middle- and upper-income backgrounds, even though nearly half of the students are from low-income backgrounds. …

By design, the program maximizes federal financial aid coming into Oregon for those who receive the Oregon Promise grant. … For students, the program slightly reduces the percentage of students facing unaffordable costs. Nonetheless, nearly two out every five students receiving the Oregon Promise still cannot meet the expected cost of attendance at their college even with the grant, and almost two-thirds of students from low-income backgrounds cannot meet the cost of attendance even with the grant. …

We found no association between the Oregon Promise program and the number of terms enrolled or credits earned among recent high school graduates. In addition, we found no lasting increases in completion rates coinciding with implementation of the Oregon Promise program, though the number of cohorts and years available to assess this question are still limited.

One can of course raise a number of possibilities here. Maybe it was the pandemic that hurt the program–although little bump in community college attendance after the passage of the program was already over by 2020. Indeed, it looks in some of the more detailed data as if the bump in community college attendance was because of a drop in attendance from other Oregon four-year colleges; in other words, the program caused a few students to shift from four-year to the now “free” two-year institutions.

Maybe a bigger Oregon Promise is needed, going beyond tuition and also governing books, living expenses, and other costs? It’s notable that the program had a “top-up” design: that is, you first qualify for federal assistance for low-income families via programs like Pell Grants, and then the Oregon Promise tops up the rest. But the result of this approach was that most of the funding for the Oregon Promise ended up flowing to middle- and high-income families, who weren’t eligible for low-income aid. Maybe a Promise that gave more generous aid to those with low-incomes, and didn’t also subsidize those from middle- and high income families, would make more sense?

Or it may be that a high level of mentoring and advising that continues into the community college year is just as or more important to nontraditional or low-income students than the financial aid itself. After all, getting admitted to community college but then not having a sense of what classes to take, how to do the work, and where to go when you are having troubles may not help much. Or maybe the Oregon Promise should have requirements that high school students take certain classes or have a certain GPA or test scores to be eligible?

The details of program design matter. There are now a variety of “Promise”-style programs: the Kalamazoo Promise, the Pittsburgh Promise, and similar programs across 24 states. Laura W. Perna, Jeremy Wright-Kim, and Elaine W. Leigh look at some differences in design in “Is a College Promise Program an Effective Use of Resources? Understanding the Implications of Program Design and Resource Investments for Equity and Efficiency” (AERA Open, 6: 4, October-December 2020). They write:

Estimates of net benefits for the Kalamazoo Promise and Pittsburgh Promise are likely not transferable, as these programs differ from others in ways that may influence program outcomes, characteristics of recipients, and costs. The Kalamazoo Promise requires students to attend district schools from kindergarten through high school graduation to be eligible for the maximum financial award, does not reduce the award by other grant aid (i.e., “first dollar”), and allows students to use the award at public 4-year institutions across the state (Bartik et al., 2016). Comparatively, the Tennessee Promise, for example, requires students to apply as high school seniors, provides a financial award that is reduced by other grant aid (i.e., “last dollar”), and limits use of the award to attendance at community and technical colleges (House & Dell, 2020Meehan et al., 2019).

The Oregon Higher Education Coordinating Commission obviously isn’t the final word here, but this is how they sum up the existing research on Promise-style programs:

Whether and how College Promise programs affect access to and success in college has been of national interest. Across the country, these programs cover tuition, but they differ in both scope and design. Regarding scope, some apply to a specific college, others apply only to high school graduates in a specific school district, and still others to multiple public institutions for high school graduates statewide. Regarding eligibility, requirements vary around student residency, high school grade averages, application materials and fees, enrollment levels, and income limits.

In the initial years of these various programs, evaluations found increased college enrollment associated with both local and statewide College Promise programs. However, more recent research has found that College Promise programs do not consistently sustain these increases in enrollment, citing differences in eligibility requirements. Programs that have eligibility requirements consistent with students who are most likely to go to college have not produced long lasting enrollment increases.

Though the relative newness of College Promise programs limits the research on college completion, recent studies suggest that student supports (e.g., advising, mentoring, and other educational supports) are an important intervening factor. Programs with more minimal eligibility requirements in particular did not demonstrate increases in postsecondary credential attainment without additional support resources. Prior research has shown limited impacts of College Promise on equity in college access and success. Programs with eligibility requirements that are consistent with the characteristics of those already likely to attend college maintain existing inequities. Those structured as last-dollar programs show minimal to no improvements in equity.

There’s an old saying that states are the “laboratories of democracy,” the place where you can try things out and see what works. After all, you can learn from your failed experiments, too.

From Multilateralism to Friend-shoring

Offshoring is when production by US companies happens in other places. Onshoring is when at least some of that production that had been done in foreign places shifts back inside US borders. Friend-shoring is when countries decide, as a matter of policy, that they will emphasize and focus on trade with friendly nations. The idea of friend-shoring is being discussed at high levels, as T. Clifton Morgan, Constantinos Syropoulos, and Yoto V. Yotov discuss in “Economic Sanctions: Evolution, Consequences, and Challenges” (Journal of Economic Perspectives, Winter 2023; full disclosure, I work as Managing Editor of JEP).

The main focus of their article is on the rising use of trade sanctions and how well they work (as discussed here). But given the rise in US protectionism that started during the Trump administration and has continued into the Biden presidency, the prospect of friend-shoring also comes up. They write:

Finally, as states seek to reduce their vulnerability to sanctions, what ramifications may arise for the international economic and political system? For example, when members of the World Trade Organization mix trade and security policies, as has been the case in the Russia-Ukraine crisis, the survival of WTO and the rules-based approach to policymaking may be at risk. Comments from prominent officials suggest that the United States and the European Union may already be moving away from global multilateralism toward cooperation with limited circles of friends.

In [US Secretary of the Treasury] Janet Yellen’s (2022) words: “[W]e need to modernize the multilateral approach we have used to build trade integration. Our objective should be to achieve free but secure trade. We cannot allow countries to use their market position in key raw materials, technologies, or products to have the power to disrupt our economy or exercise unwanted geopolitical leverage. So let’s build on and deepen economic integration . . . And let’s do it with the countries we know we can count on. Favoring the friend-shoring of supply chains to a large number of trusted countries . . . will lower the risks to our economy as well as to our trusted trade partners.”

In similar spirit, Christine Lagarde (2022), President of the European Central Bank, remarked: “Russia’s unprovoked aggression has triggered a fundamental reassessment of economic relations and dependencies in our globalised economy . . . Today, rising geopolitical tensions mean our global economy is changing . . . [O]ne can already see the emergence of three distinct shifts in global trade. These are the shifts from dependence to diversification, from efficiency to security, and from globalisation to regionalisation.

Given the disruptions of the pandemic, and Russia’s invasion of Ukraine, and China’s military build-up, there are of course reasons for firms everywhere to shorten their supply chains, and practice some combination of reshoring and friend-shoring. But the current favorite flavor of trade policy in the US and Europe does more than that: It’s more about government subsidies for favored domestic industries, rather than about recalculating gains from trade. It looks as if policymakers may need to learn all over again that while some parts of an economy can subsidize other parts, the gains of the favored firms that get such subsidies don’t mean that the economy as a whole is actually better off. In addition, whatever the geopolitical reasons for reducing international trade, such reductions impose economic costs.

I won’t rehearse those arguments here. But remember when President Trump first imposed substantial trade sanctions on China? He tweeted in March 2018: ”When a country (USA) is losing many billions of dollars on trade with virtually every country it does business with, trade wars are good, and easy to win. Example, when we are down $100 billion with a certain country and they get cute, don’t trade anymore-we win big. It’s easy!” He imposed tariffs, and here’s the pattern of the US trade deficit since then (red line is trade deficit for goods only; green line is for goods and services combined. At this point, it’s pretty clear (as it has been many times before) that imposing tariffs is not the answer to reducing the US trade deficit, nor to jump-starting US productivity and competitiveness.

Less trade and more government subsidies to favored industries and firms is not a likely path to improved trade balances, or to prosperity in general.

Lessons from James Tobin: The Virtues of Strategic Sloppiness

For those who do not follow the ins and outs of academic macroeconomics, it is perhaps useful to say that there has been an ongoing struggle in recent decades between what is sometimes called freshwater and saltwater economics.

At risk of being struck dead by lightning for gross oversimplification, I’ll just say that freshwater economists tended in the 1970s and 1980s to cluster around places like the University of Chicago, the University of Minnesota, the University of Rochester, and Carnegie Mellon University. Their version of macroeconomics tended to emphasize the themes that economic fluctuations were caused by shocks to supply (like technology) and that discretionary federal macroeconomic policy was likely to have weak or even counterproductive effects. In contrast, the saltwater economists at that time tended to congregate at places like Harvard, Yale, and Berkeley. Their version of macroeconomics tended to emphasize that economic fluctuations were caused by shocks to demand (like bank failures or credit boom-and-bust cycles) and that discretionary federal macroeoconomic policy was not only useful but also necessary in helping to offset such shocks.

Over the decades, the two schools have become somewhat intertwined in the form of what is sometimes called “New Keynesian” economics (for discussions, see here and here). But the spirit of the old dividing lines still remains. The saltwater economists accuse their freshwater siblings of being slaves to models that assume excessively rational people and excessively perfect markets; in response, the freshwater economists accuse their saltwater kinfolk of promiscuously adding theoretical restrictions for immediate convenience, without digging deeply enough into their foundations and implications.

Paul Krugman, as a certified salt-water economist, offers a thoughtful explanation of the merits of this approach as exemplified in the macroeconomics of another certified salt-water economist in “The Godley–Tobin Memorial Lecture: The Second Coming of Tobinomics ” (Review of Keynesian Economics, Spring 2023, vol. 11: issue 1)). Krugman writes:

James Tobin was, obviously, a Keynesian in the sense that he believed that workers can and do suffer from involuntary unemployment, and that government activism, both monetary and fiscal, is necessary to alleviate this evil. But he wasn’t what people used to call a hydraulic Keynesian, someone who imagined that you could analyse the economy by positing mechanical relationships between variables like personal income and consumer spending, leading to fixed, predictable multipliers on policy variables like spending and taxes. …

Instead, Tobin was also a neoclassical economist. That is, he believed that you get important insights into the economy by thinking of it as an arena in which self-interested individuals interact, and in which the results of those interactions can usefully be understood by comparing equilibria — situations in which no individual has an incentive to change behaviour given the behaviour of other individuals.

Neoclassical analysis can be a powerful tool for cutting through the economy’s complexity, for clarifying thought. But using it well, especially when you’re doing macroeconomics, can be tricky. Why? It’s like the old joke about spelling ‘Mississippi’: the problem is knowing when to stop.

What I mean is that it’s all too easy to slip into treating maximising behaviour on the part of individuals and equilibrium in the sense of clearing markets not as strategic simplifications but as true descriptions of how the world works, not to be questioned in the face of contrary evidence. Notably … perfectly clearing markets wouldn’t have involuntary unemployment. So if you’re a neoclassical economist who doesn’t know when to stop, you end up denying that there can be recessions, or that, say, monetary policy can have real effects, even though it takes only a bit of real-world observation to see that these propositions are just false.

So part of the art of producing useful economic models is knowing when and where to place limits on your neoclassicism. And strategic placing of limits is a large part of what Tobinomics is about.

What do I mean by placing limits? Tobin was, first of all, willing to ditch the whole maximisation-and-equilibrium approach when he considered it of no help in understanding economic phenomena — which was the case for his views on labour markets
and inflation, which I’ll get to later in this paper.

Where he did adopt a neoclassical approach, he did so using two strategies that economists need to relearn. First, he was willing to be strategically sloppy — to use the idea of self-interested behaviour as a guide to how people might behave without necessarily deriving everything from explicit microfoundations. Second, he was willing to restrict the domain of his neoclassicism — applying it to asset markets but not necessarily to goods markets or the labour market.

Krugman illustrates his argument with a detailed example from Tobin’s work, but for my purposes, I’ll stop there.

I like the old joke about the problem with spelling “Mississippi,” which seems applicable to me in a number of real-world and academic situations. In a number of situations it can be a useful exercise to start with a pure theory, and then take a steam shovel to dig into its foundations and a telescope to look out at its possible implications. But when you reach the stage of bringing a pure theory to data, especially in the social sciences, it becomes necessary and practical to introduce a degree of strategic sloppiness; for example, the data or the setting you have to work with often will not exactly match the pure theoretical assumptions. The choice of which kinds of real-world strategic sloppiness are most relevant to a given question will often be central to the real-world controversy.

America’s Problems with Building Transit Infrastructure

Building mass transit infrastructure in the United States is considerably more expensive than in other countries. Why? Eric Goldwyn, Alon Levy, Elif Ensari and Marco Chitti dig down into case studies to get some answers in Transit Costs Project: Understanding Transit Infrastructure Costs in American Cities (New York University, Marron Institute of Urban Management, February 2023).

The main focus of the report is to look in some detail at extensions of mass transit in Boston and New York, and also at projects in Sweden, Italy, Istanbul and Istanbul. The disadvantage of a case study method, of course, is that you need to be cautious about drawing conclusions from a small number of examples. The counterbalancing advantage is that you can dig deeply into the details of each individual case. But they do offer some big-picture statistics as well..

The problem of high costs is nationwide. According to our database (Transit Costs Project .d.) of more than 900 projects in 59 countries, including Hong Kong, the United States is the sixth most expensive country in the world to build rapid-rail transit infrastructure. This is slightly misleading, however, because construction costs scale with the percentage of tunneled track, which is more expensive than building rail at grade. The five countries with greater average costs than the United States are building projects that are more than 65% tunneled. In the United States, on the other hand, only 37% of the total track length is tunneled (Figure 1). … Therefore, it is valuable to understand what it is about the physical, institutional, and social situation of American cities that frustrates subway expansion dreams.

What kinds of factors explain the difference?

The US rail transit stations are often “overbuilt,” meaning that they are built much longer than the actual train platforms, and are also sometime built with nice high vaulted ceilings. In addition, the stations in Boston and New York were not standardized: for example, “three stations [in New York] used two different escalator contractors and have a different number of exits, crossovers, and elevators, all of which raise design costs because each station needs to be customized rather than using a standard design that is modified slightly.” Thus, when you compare the costs of “systems and station finishes” to the costs of “tunnels and station civil works,” the ratio of those costs is about 50:50 in New York, but 25:75 in Paris Milan, and Sweden.

Labor is a much larger share of transit construction costs in the US: “In New York as well as in the rest of the American Northeast, labor is 40-60% of the project’s hard costs, according to cost estimators, current and former agency insiders, and consultants with knowledge of domestic projects. Labor costs in our low-cost cases, Turkey, Italy, and Sweden are in the 19-30% range; Sweden, the highest-wage case among them, is 23%. The difference between labor at 50% of construction costs and labor at 25%, holding the rest constant, is a factor of 3 difference in labor costs …”

Overlapping and redundant bureaucracies also add to US transit construction costs. “[W]e also found overstaffing of white-collar labor in New York and Boston (by 40-60% in Boston), due to general inefficiency as well as interagency conflict, while little of the difference (at most a quarter) comes from differences in pay. … We have identified numerous cost drivers that stem from procurement norms in the United States. These include a pervasive culture of secrecy and adversarialism between agencies and contractors, a lack of internal capacity at agencies to manage contractors, insufficient competition, and a desire to privatize risk that leads private contractors to bid higher. Overall, this raises costs by a factor of 1.85, with the extra money going to red tape, wasted contingencies, paying workers during delays, defensive design, and, owing to contractor risk, and profit.”

There are also “soft costs”: “Soft costs include design, planning, force account, insurance, construction management, and contingencies; breakdowns differ by city. Nonetheless, we harmonized definitions around third-party design, planning, and project management costs. Those add 5-10% on top of the hard contract costs in our comparison cases, most commonly 7-8%. But in English-speaking countries, soft costs add much more; for Second Avenue Subway, it was 21%.”

Thus, if you want to know why it’s so hard and costly for US cities to build transit systems, and why such systems seem to be built and extended in other cities much more easily, a big part of the issue is that the model that US cities have used for building such infrastructure has led to high costs.

A common proposal here is to try to set up construction contracts so that the risk of higher costs and cost overruns would be carried by private firms, not by taxpayers. The authors find little evidence that these kinds of reforms work:

Moreover, many ongoing reforms hailed as steps forward, which we call the globalized system in the Sweden report, at best do nothing and at worst are actively raising costs; these reforms all aim to privatize risk and have been popular throughout the English-speaking world, and while consultants, managers, and large contractors like them, costs grow sharply wherever they are implemented, such as England, Singapore, Hong Kong, and Canada. … The good news is that high-cost countries can adopt the practices of low-cost countries and build subways at costs more in line with those of low-cost Scandinavia, Southern Europe, and Turkey. To do this, it requires rethinking design and construction techniques, labor utilization, procurement, agency processes, and the use of private real estate, consultants, and contingencies. If it implements the best practices we detail in the rest of the overview, the highest-cost city in our database, New York, can reduce its construction costs to match those of Italy and match or even do better than Scandinavia.”

Of course, the US inability to build transit at a reasonable costs is part of a bigger problem: along a variety of dimensions, the US economy has a diminished ability to build, whether we are talking about mass transit, green energy, or housing.

For example, if the United States is serious as a society about a vast expansion of solar and wind energy (a proposition I find myself doubting), then we will need not only very large commitments of land to solar and wind projects, but also a vast physical expansion of the electrical grid–along with facilities for manufacturing and recycling the equipment and a willingness to mine for the copper and other raw materials needed. A report on some of the permitting reforms needed for those steps to happen is here.

Similarly, if we are serious as a society about affordable housing (a proposition it is also possible to doubt), then many cities need a vast expansions of their housing stock, not mostly built at the extreme outer edges of cities where land is cheap, but instead focused on locations within city boundaries where some combination of underused office and commercial space, surface parking lots, and even certain areas in residential neighborhoods (say, those near mass transit or the empty airspace above some local shops) will need to be committed to additional housing units.

None of this needs to mean a free-for-all of building. Zoning can still matter! But current US rules for construction in many areas often look remarkably like the rules you would have in place if you wanted to discourage and slow construction, and raise its costs, rather than the roles you would have in place to facilitate such projects.

Whither Quantitative Tightening?

Starting in the Great Recession of 2007-9, the Federal Reserves started a policy of “quantitative easing,” in which the Fed held substantial quantities of financial assets like US Treasury bonds and federally guaranteed mortgage-based debt. The problem at the time was that the Fed had reduced its policy interest rate to near-zero. The idea was that if the Fed held such assets, then interest rates could be somewhat lower, compared to a situation where these assets were being sold in financial markets.

An obvious question raised at the time was whether the Fed was, in effect, just printing money to cover government borrowing. “Monetizing” the debt it this way is typically thought to be a lousy idea, because it leads to inflationary pressures, and even to doubts about the true value of the debt on financial markets. But Fed economists at the time drew a clear distinction between quantitative easing and monetizing the debt. Here are David Andolfatto and Li Li of the St. Louis Fed on the subject (“Is the Fed Monetizing Government Debt?: February 1, 2013).

What is usually meant by “monetizing the debt,” however, is the use of money creation as a permanent source of financing for government spending. Thus, to ascertain whether the Fed has in fact monetized its purchases of $1.2 trillion in government bonds since 2008, we have to know what the Fed intends to do with its portfolio of assets over time.

If the recent rapid accumulation of Treasury debt on the Fed’s balance sheet constitutes a permanent acquisition, then the corresponding supply of new money would be expected to remain in the economy (as either cash in circulation or bank reserves) permanently as well. As the interest earned on securities held by the Fed is remitted to the Treasury, the government essentially can borrow and spend this money for free. If, on the other hand, the recent increase in Fed Treasury debt holdings is only temporary (an unusually large acquisition in response to an unusually large recession), then the public must expect that the monetary base at some point will return to a more normal level (through sales of securities or by letting the securities mature without replacing them). Under this latter scenario, the Fed is not monetizing government debt—it is simply managing the supply of the monetary base in accordance with the goals set by its dual mandate. Some means other than money creation will be needed to finance the Treasury debt returned to the public through open market sales.

Here’s a figure from the Fed showing the evolution of its asset holdings over time. You can see the build-up in assets after the Great Recession–those first waves of quantitative easing. You can then see a modest decline in Fed asset-holding, as Aldolfatto and Li described above. The idea during this time was that the Fed would simply hold the debt it had purchased until that debt expired, and thus gradually and slowly let its asset holdings phase down. A phrase commonly used was that the process would be about as exciting as watching paint dry.

But then the pandemic recession hits, the US government runs extraordinarily large deficits to fund its pandemic relief programs, and given the uncertainties in global financial markets, the Fed holdings of assets take another dramatic leap. For reference, the numbers are in millions of dollars, so the “8M” on the right-hand axis refers to eight million million–that is, $8 trillion.

So even if the Fed was not monetizing federal debt back in 2013 or so, is it doing so now? What’s the current plan for phasing down the now much-larger Fed holdings of government debt? Huberto M. Ennis and Tre’ McMillan of the Federal Reserve Bank of Richmond lay out part of the plan in “Fed Balance Sheet Normalization and the Minimum Level of Ample Reserves” (Economic Brief No. 23-07, February 2023).

Ennis and McMillan write:

From the beginning of the pandemic through the spring of 2022, the Fed’s balance sheet increased significantly due to the Fed’s efforts to aid market functioning and support the flow of credit to households and businesses. Reserves in the banking system increased to record highs, well beyond levels desired by the Fed in the long run. With financial and economic conditions improving, the Fed started the process of balance sheet normalization in March 2022, whereby it intends to significantly reduce the amount of Treasuries and mortgage-backed securities (MBS) that it holds in its System Open Market Account (SOMA) portfolio.

So what is going to happen, and how far and how fast? Here, the desired reduction in Fed assets over time–basically, the Fed holding less in Treasury bond and in federally guaranteed mortgage-backed securities–intersects with the imperatives of conducting everyday monetary policy. The main tool that the Fed uses for conducting monetary policy is to change the interest rate it pays on reserves that banks hold at the Fed, and in this way to affect its key policy interest rate, called the federal funds rate. A Fed website explains: ,

The FOMC [Federal Open Market Committee] has the ability to influence the federal funds rate–and thus the cost of short-term interbank credit–by changing the rate of interest the Fed pays on reserve balances that banks hold at the Fed. A bank is unlikely to lend to another bank (or to any of its customers) at an interest rate lower than the rate that the bank can earn on reserve balances held at the Fed. And because overall reserve balances are currently abundant, if a bank wants to borrow reserve balances, it likely will be able to do so without having to pay a rate much above the rate of interest paid by the Fed. Typically, changes in the FOMC’s target for the federal funds rate are accompanied by commensurate changes in the rate of interest paid by the Fed on banks’ reserve balances, thus providing incentives for the federal funds rate to adjust to a level consistent with the FOMC’s target.

Thus, as a first step, when Fed wants to raise interest rates, it raises the interest rate it pays on bank reserves. For this policy to work, there need to be “ample” bank reserves. As Ennis and McMillan write:

The idea is for the Fed to maintain a balance sheet large enough to accommodate growth in currency in circulation plus an ample quantity of bank reserves. “Ample” means that reserves are plentiful enough to not carry any significant convenience yield. In other words, banks should value the marginal unit of reserves for the interest on reserves that they earn, but not because that marginal unit facilitates the daily operations of the bank holding it in any meaningful way. 

What does that “enough but not too much” language mean in practical terms? They suggest that the goal should be to get bank reserves back to their level in about 2019, before the run-up of Fed assets during the pandemic. (This seems a commonly held view at the Fed: for example, scroll down and see similar comments from Christopher Waller.) Here’s an illustrative figure with their calculations out to 2029.

For our purposes, here are the important lines. The top line shows the total holdings of Fed assets, as shown above. The light blue lines shows bank reserves actually held; the red line shows their calculation of “minimum ample reserves.” You can see their projection that bank reserves will decline in the next year or two until they reach the “minimum ample” level. The bottom orange TGA line shows the “Treasury General Account” at the Fed: you can see how it bounced up at the start of the pandemic, when the Treasury wanted to have cash on hand to make payments as legislation and emergencies dictated, but then drops again.

All of this raises two questions. The first question is practical: How does the Fed plan to make this happen? The answer is that as the Fed receives interest payments on the Treasury debt and the mortgage-backed securities that it is holding, it will not reinvest that money in new debt. Thus, the value of the debt it is holding will diminish over. Ennis and McMillan estimate that this process will reduce Fed holdings of debt by about $80 billion per month. Thus, the “ample” level of bank reserves would be reached by about 2026.

The second question is harder: Is this goal for reducing Fed assets the right one? Here, it’s worth pointing out that the policy goalposts have been shifted. Back in 2013, you’ll remember, Aldolfatto and Li argued that the first wave of quantitative easing was not “monetizing the debt,” because it wasn’t a permanent step, and debt held by the Fed was being phased down. Now here were are in 2023, and even if the Fed manages to stick to its plan of phasing down its Treasury debt holding in the next three years, total Fed assets would still be about $7 trillion by 2029. What looked like a slow phase-down of Fed assets back in 2013 will be starting to look like a permanent and generally rising pattern of Fed assets by 2029–which was the working definition of monetizing the debt.

I don’t see any easy choices for the Fed here. The optics of having a central bank decide to do a large-scale sell-off of the debt of its own country would not be a good look in financial markets. But the approach of reducing the debt that the Fed is holding by not reinvesting interest payments on the debt that the Fed holds is slow. As I see it, the Fed seems to be hoping that after having experienced two once-in-century events in the last 15 years–the Great Recession of 2007-9 and then the pandemic recession–it just won’t feel a need to do quantitative easing (or “monetizing the debt?”) again in the next decade or so.