In Hume’s spirit, I will attempt to serve as an ambassador from my world of economics, and help in “finding topics of conversation fit for the entertainment of rational creatures.”
The world population is aging. In the next few decades, a much larger number of people are going to need long-term care. The United States, like most countries, doesn’t really have even a preliminary set of guidelines for how this might best happen. Here’s some background information from Health at a Glance 2023: OECD Indicators (November 2023), specifically from “Chapter 10: Aging and Long-Term Care.”
As a starting point, here’s a figure showing the share of the population that is 80 or older, with actual data for 2021 and then projected to 2050. These projections should be viewed as fairly solid: after all, anyone who is going to be 80 or older in 2050 was already born in 1970 or earlier, and projecting life expectancy for the group of people who are already in their 50s or older is fairly straightforward
For the OECD countries as a whole, the projection is roughly a doubling in the share of the over-80 age group, from about 5% at present to about 10% in 2050. For countries with very low birthrates, like Korea, Japan, and Italy at the top of the table, the over-80 share of population will be much larger, reaching or exceeding 15% of the total population. The US will also experience roughly a doubling of the over-80 share of the population, but from a lower base than the average country listed here.
The current models that countries have for long-term care differ quite a bit. To illustrate the point, consider current spending on care as a share of GDP. In Netherlands, Norway, Sweden and Denmark, total spending on long-term care is already more than 3% of GDP.
Here’s how the OECD explains these differences:
This variation partly mirrors differences in the population structure, but mostly reflects the stage of development of formal LTC systems, as opposed to more informal arrangements based mainly on care provided by unpaid family members. … Across OECD countries, four out of five dollars spent on LTC come from public sources. Across OECD countries, around half of health and social LTC spending in 2021 occurred in nursing homes. … In most OECD countries, these providers account for the majority of LTC spending. On average, around one-fifth of all LTC spending was used for professional (health) care provision at home. Other LTC providers include hospitals, households – if a care allowance exists that remunerates the informal provision of such services – and LTC providers with a clear social focus. These service providers each account for around one-tenth of total LTC spending across OECD countries. …Without public financial support, the total costs of LTC would be higher than median incomes among older people in most OECD countries. On average across OECD countries, institutional care for severe needs would cost more than twice the median income among older people …
When it comes time for a certain share of the elderly to need long-term care, a rough measure of the capacity of a nation’s long-term care system is the number of beds. The figure shows the number of long-term care beds per population of 1,000 over age 65. Some countries, like Japan and Korea, have a large share of long-term care beds inside hospitals. But for most countries shown here, including the United States, most long-term care beds are not in hospitals. The US is substantially below the international average in the number of beds.
Compared with these other countries, the US both spends less on long-term care as a share of GDP and has lower number of long-term care beds per capita because a much smaller share of Americans aged 65 or older end up in long-term care institutions: the average across the 28 OECD countries is 11.5%, while the US share is 1.7%.
These differences seem too large to reflect underlying differences in health. Instead, they reflect a mixture of social expectations and the design of government programs to support the elderly. The US, along with Canada, Japan, and a few others, has so far managed to have only a small proportion of the elderly in long-term care institutions.
Of course, many older people would prefer to live at home as long as possible, before moving to a long-term care institutions, and many countries have policies to support this option. In practice, the live-at-home option also ends up relying on whether a family member can be a regular care-giver, either weekly or even daily. With lower birthrates in the last few decades, and a higher share of women in the workforce, relying on care from a family member is likely to be harder in the future. As the over-65 and especially the over-80 population rises in the US in the next few decades, the existing low use of long-term care institutions in the US is likely to come under severe stress.
[The] US Department of Agriculture established the Section of Economic Ornithology in 1885. The following year it became the Division of Biological Survey, and was upgraded to the Bureau of Biological Survey in 1905. … In 1903, the Saturday Evening Post, for example, published a request that `every person in the United States who kills a bird is requested by the United States Government, not in a mandatory way, but as a matter of courtesy, to send the stomach and its contents to Washington.’ By 1916, the Bureau of Biological Survey had collected and analyzed the contents from more than 60,000 bird stomachs, which they used to determine whether each of the 400 species they studied was, on balance, helpful or harmful to man. Researchers divided the stomach contents into “good,” “bad,” and “neutral” categories, based on whether the partially-digested bug and plant matter was beneficial or harmful to farmers. …
According to the Bureau of Biological Survey, native sparrows, who are “specially efficient destroyers of weed seeds,” saved farmers $35 million in 1906 by eating ragweed and crabgrass seeds. And during Nebraska’s 1874 Rocky Mountain Locust infestation, a single Marsh Wren was calculated to have fed her brood of chicks enough grasshoppers to save $1,743.97 worth of crops. The Bureau of Biological Survey even helped rehabilitate the reputation of some birds that were historically seen as enemies to the farmer. By examining over a thousand crow stomachs, the Bureau found that while crows did in fact pull up sprouting corn and nibble corn on the stalk, they ate more “noxious insects and mice,” meaning that “the verdict was therefore rendered in favor of the crow, since, on the whole, the bird seemed to do more good than harm.”Owls, which were long considered poultry thieves, were proven to eat enough mice to earn back “the small commission they collect” by nabbing the occasional chicken.
This kind of information was distributed not just by the US Department of Agriculture, but also through groups like the Audubon Society and the League of American Sportsmen. For those who would like more history of economic ornithology, Theodore S. Palmer of the USDA provides an overview of the development of the field from the 1850s up through the end of the 19th century in his 1899 monograph: “A Review of Economic Ornithology in the United States.” H.J. Taylor (no relation) provided pocket autobiographies of five “Pioneers in Economic Ornithology” (The Wilson Bulletin, September 1931).
It wasn’t just pesticides that killed off economic ornithology. A deeper issue was that it wasn’t clear that adding birds to an agricultural area actually reduced the number of insects and weeds, at least not in a reliable way. And yet, some occasional modern studies suggest that certain birds in certain settings do have considerable economic value.
My favorite recent example is the “The Social Costs of Keystone Species Collapse: Evidence From The Decline of Vultures in India,” by Eyal G. Frank, and Anant Sudarshan (Becker Friedman Institute Research Brief, February 2, 2023). They tell the story of how a painkiller called diclofenac went off-patent, and as its price declined sharply, veterinarians in India began to give the drug to sick cattle. Although the drug was fine for cattle, it is severely toxic to vultures. Thus, when some of these cattle died and their carcasses were eaten by vultures, the vultures in this area became almost extinct. The authors write:
Vultures are efficient scavengers and feed only on carrion. In India, a country with over 500 million livestock, these birds provided an important public health service by removing livestock carcasses from the environment. In the mid-1990s, vultures experienced the fastest population collapse of a bird species in recorded history. The cause of death was unknown until 2004 when it was identified as poisoning from consuming carcasses containing traces of a common painkiller, diclofenac. The expiration of a patent led to a dramatic fall in the price of medical diclofenac, the development of generic variants, and entry into the veterinary market in 1994. We exploit this event to study the costs of losing vultures. Using habitat range maps for affected species, we compare high- to low-vulture suitability districts before and after the veterinary use of diclofenac. We find that, on average, all-cause human death rates increased by more than 4% in vulture-suitable districts after these birds nearly went extinct. … As vultures died out, the scavenging services they provided disappeared too, and carrion were left out in the open for long periods of time. Ecologists have argued that this may have led to an increase in the population of rats and feral dogs, which are a major source of rabies in India. Rotting carcasses can also transmit pathogens and diseases such as anthrax, to other scavengers. In addition, these pathogens can enter water sources either when people dump carcasses in rivers or because of erosion by surface runoff …
More generally, there continues to be a modest literature in environmental economics that carries on the “economic ornithology” tradition of looking at birds as providers of ecosystem services. Christopher J. Whelan, Çağan H. Şekercioğlu and Daniel G. Wenny provide an overview in “Why birds matter: from economic ornithology to ecosystem services” (Journal of Ornithology, 2015, 156: 227-238). They point to a few specific studies:
For instance, Mols and Visser (2002) investigated effects of bird control of herbivorous insects in Dutch apple orchards, and reported that increasing bird density through deployment of nest boxes led to a 50 % reduction in apple damage and an increase of about 60 % in total apple crop yield. Koh (2008) attributed bird pest control to prevention of 9–26 % fruit loss in oil palm (Elaeis guineensis). Johnson et al. (2009) found birds significantly reduced damage by coffee berry-borer beetles (Hypothenemus hampei), with higher coffee yields resulting in increased income from US$44 to US$310/ha.
The authors also point to birds as providing pollination and seed dispersal services, controlling populations of mice and rats and other services. But the overall tone of the article is that there is still a lot of research to be done, not in the dissection of bird stomachs, but in understanding the role of birds within ecosystems–especially as bird populations rise or fall and ecosystems adjust accordingly. The authors write:
Yet the economic relevance of birds is not widely appreciated and the economic relevance to human society of birds’ ecological roles is even less understood. Quantifying the services provided by birds is crucial to understand their importance for ecosystems and for the people that benefit from them. In this paper, we briefly review the rise and fall of economic ornithology and call for a new economic ornithology with heightened standards and a holistic focus within the ecosystem services approach. Birds’ ecological roles, and therefore, ecosystem services, are critical to the health of many ecosystems and to human well-being.
For my bird-watcher friends, no, I’m not suggesting that all birds should be reduced to quantifiable factors of production. But when it comes to protecting and restoring bird habitat, having some dollars and cents on your side of the argument doesn’t hurt.
On November 2, 1983, President Ronald Reagan signed a law establishing a federal holiday for the birthday of Martin Luther King Jr., to be celebrated each year on the third Monday in January. As the legislation that passed Congress said: “[S]uch holiday should serve as a time for Americans to reflect on the principles of racial equality and nonviolent social change espoused by Martin Luther King, Jr..” Of course, the case for racial equality stands fundamentally upon principles of justice, with economics playing only a supporting role. But here are a few economics-related thoughts for the day clipped from posts in the previous year at this blog, with more detail and commentary at the links.
The difference in the average wealth of Black and white Americans narrowed in the first century after the Civil War, but remained large and even widened again after 1980. Given high levels of wealth concentration both historically and today, dynamics at the average may not capture important heterogeneity in racial wealth gaps across the distribution. This paper looks into the historical evolution of the Black and white wealth distributions since Emancipation. The picture that emerges is an even starker one than racial wealth inequality at the mean. Tracing, for the first time, the evolution of wealth of the median Black household and the gap between the typical Black and white household over time, we estimate that the majority of Black households only began to dispose of measurable wealth around World War II. While the civil rights era brought substantial wealth gains for the median Black household, the gap between Black and white wealth at the median has not changed much since the 1970s. The top and the bottom of the wealth distribution show even greater persistence, with Black households consistently over-represented in the bottom half of the wealth distribution and under-represented in the top-10 percent over the past seven decades.
Both essays emphasize the evolution of historically black colleges and universities (HBCUs), and the differences across these institutions. Both note that back in, say, 1967, about 80% of all black colleges students attended these institutions, while now it’s about 9%. Thus, the role of these institutions has evolved. However, they continue as a group to provide an outsized share of black college graduates, especially in the sciences. In addition, after adjusting for factors like household income and institutional resources, black students attending HBCUs have a greater likelihood of graduating. At a time when US higher education as a whole is trying to reach out to traditionally underrepresented group, it seems as if there are some lessons to be learned here.
It’s quite possible that US incarceration rates are too high, but it’s also just a fact that they have been declining in recent years. Here’s an overall figure.
This study makes three primary contributions to a fuller understanding of the contemporary landscape of incarceration in the United States. First, we assess the scope of decarceration. Between 1999 and 2019, the Black male incarceration rate dropped by 44%, and notable declines in Black male imprisonment were evident in all 50 states. Second, our life table analysis demonstrates marked declines in the lifetime risks of incarceration. For Black men, the lifetime risk of incarceration declined by nearly half from 1999 to 2019. We estimate that less than 1 in 5 Black men born in 2001 will be imprisoned, compared with 1 in 3 for the 1981 birth cohort. Third, decarceration has shifted the institutional experiences of young adulthood. In 2009, young Black men were much more likely to experience imprisonment than college graduation. Ten years later, this trend had reversed, with Black men more likely to graduate college than go to prison.
It turns out that when you dig into this data, pretty much all of the difference is because black working poor who are claiming the Earned Income Tax Credit are audited at much higher rate than non-black working poor who are claiming the EITC, and that this “disparity cannot be fully explained by racial differences in income, family size, or household structure.” Instead, the gap seems to trace back into details built into the IRS algorithm. For example, the algorithm tends to single out for audits the cases that are more likely to lead to higher taxes. This may sound reasonable at first, but imagine two tax returns: In one, there is a 95% chance that the audit will collect an extra $500, and in the other there is a 50% chance that the audit will collect an extra $10,000. If the algorithm prioritizes the chance of collecting more, rather than a mixture of the probability and the amount that could be collected, it will often focus on the working poor rather than on middle- and upper-income taxpayers who might owe more.
I’m a sports fan, which in this case may represent a conflict of interest, because it means I’m conflicted about public subsidies going to sports stadiums. The economic evidence on this point is pretty clear: such subsidies can transfer how people spend their entertainment dollars from one area of a city to another, but the net gain to an urban area is probably negative. John Charles Bradbury, Dennis Coates, and Brad R. Humphreys review the evidence in “The impact of professional sports franchises and venues on local economies: A comprehensive survey (Journal of Economic Surveys, September 2023, 1389-1431). The authors write:
Between 1970 and 2020, state and local governments devoted $33 billion in public funds to construct major-league sports venues in the United States and Canada, with the median public contribution covering 73% of venue construction costs.The prevalence of subsidized sports stadiums and arenas spawned an active economics literature evaluating their efficacy at stimulating economic activity. This literature contains near-universal consensus evidence that sports venues do not generate large positive effects on local economies. … However, this literature expanded considerably since the last comprehensive literature survey. We survey the extensive academic literature on the economic impacts of sports teams and venues on local communities, which includes more than 130 articles and spans more than 30 years, most published in the past decade. We document the presence of a clear consensus in the results reported in this literature.
Many of us sports fans know that when we attend a game, nearby restaurants, bars, and parking lots are often doing a good business–of course along with economic activity in the venue itself. How do we reconcile this evidence of our own eyes with the economic studies? As Bradbury, Coates, and Humphreys write:
Robust empirical findings documenting the impotence of professional sports in local economies likely reflect a simple theoretical explanation: consumer spending on sports represents a transfer from other local consumer spending, not net-new spending. Although sports games attract some nonlocals to spend money in the area, these visitors also crowd out other tourists attracted to other consumption amenities common to major US cities. Even with the presence outside visitors attracted by sports events, most consumer spending in and around pro sports venues derives from local residents; therefore, the opportunity cost of local sports consumption falls primarily on other competing local businesses, such as movie theaters, restaurants, and retail shopping. Most spending on game tickets, concessions, and associated hospitality near a sports venue would have occurred in other parts of the host jurisdiction without the presence of a pro sports team. Sports-related spending largely reflects a redistribution of existing spending by residents rather than increased local spending.
Any added spending from visitors attending games tends to be concentrated in certain sectors in the local economy and in locations that may not bear the full tax burden generated by subsidies. In addition, the influx of consumers also generates local nuisance or congestion externalities in the form of traffic, crowds, noise, litter, and crime, which may mitigate any positive economic effects. Furthermore, there is no obvious reason to expect income or employment multipliers from sports spending to be greater than those for other types of local consumption spending that are crowded out; thus, the consistent empirical findings of insubstantial tangible economic impacts from professional sports teams and venues conform to theoretical expectations.
When the economic evidence is against you, then you (in this case, me) argue about noneconomic benefits. Economists sometimes refer to “nonuse benefits.” Even if I haven’t attended a game in a few months (and a combination of limited time and high ticket prices means that I don’t see a lot of games in person), I still enjoy reading and hearing about the games. The local newspaper probably devotes more space to sports coverage than to international news. During my commute, I often listen to local sports-talk radio stations. I sometimes watch games on television. Talking about weather and sports is often an easy and noncontroversial conversation opener.
Some economists have tried to estimate these kinds of “nonuse benefits” using sophisticated survey data: a common finding is that the social benefits are about 15% of the facility construction costs–not nearly enough to justify the level of public subsidies.
Another argument involves whether a new stadium increases property values in the area around the stadium. The evidence here is not clear-cut, but a rough summary would be that in suburban areas, a new stadium often decreases local property values (households and firms don’t necessarily want to be near the stadium), while a new stadium in an urban area can sometimes increase local property values. In interpreting these kinds of results, it’s important to remember that big events also tend to bring traffic jams, noise, and even a rise in crime, so if you’re not a fan, you have no advantages to balance against the disadvantages.
Of course, all of this raises a paradox: If public subsidies for stadiums don’t pay off, why do they keep happening? There are two possible answers here. One is that stadium subsidies arise from an unholy mixture of loudly represented special interests, empire-building local officials, and the threat that a team can move away. The result is a kind of arms race, where cities know they would be better off if they were all to limit these subsidies, but few individual cities are willing to do so on their own. It’s a dynamic that’s similar to colleges and universities all building certain facilities or having certain kinds of offices because everyone else is doing it. It’s also similar to the dynamic where places offer unsuitably large tax breaks or subsidies to a big company who promises to move to a certain area.
The other possible answer is that the economic studies aren’t capturing something important about the role of sports teams in the portfolio of entertainment activities in a metro area. For example, maybe certain employers and their employees want to be in the kind of city where stuff happens. After all, stadiums are often used for nonsports events: concerts, trade shows, monster trucks, whoever-on-ice, and others. If your metro area didn’t have a football stadium, you were not going to get a visit from Taylor Swift.
From this perspective, the insight that subsidies for sports stadiums are often too high doesn’t necessarily imply that no subsidies at all are justifiable. Perhaps some of the answer is for at least some urban areas to negotiate harder for lower subsidies–and thus to help set a precedent of lower subsidies that can be followed by others.
In mid-December, the Japanese firm Nippon Steel announced that it was buying U.S. Steel for $14.9 billion. The news was unsettling to politicians of both parties, who have often argued over the years that steel is a vital domestic industry, along with being an important source of jobs.
For me, the real shock was the announced purchase price of $14.9 billion. When US Steel was formed back in 1901 by merging together a number of smaller competitors, it was the largest firm in the world. By 1960, it was still in the top 10 US firms in the Fortune 500 listings. By 1991, US Steel was no longer included in the 30 large firms that make up the Dow Jones Industrial Index. In 2014, US Steel fell out of Standard & Poor’s index of the top 500 US firms. Indeed, US Steel is no longer even the largest US steel firm–that would be Nucor. US Steel now makes about 12% of American steel.
The purchase of US Steel would not be the biggest deal of 2023. For example, Kroger’s bought Albertson’s last year, in a merger of grocery store chains, for $24.6 billion. The biotech firm Amgen paid $26 billion for Horizon Therapeutics. Prologis, a firm that owns and manages industrial space, paid $23 billion for Duke Realty. Broadcom, which designs and makes a range of software infrastructure and semiconductor products, bought VMWare, which makes software that allows you to “run any app on any cloud on any device” for $61 billion–call it four times the value of US Steel.
Indeed, there are now several US professional sports teams valued at $7 billion or more, including the (football) Dallas Cowboys, the (baseball) New York Yankees, and the (basketball) Golden State Warriors. Once-mighty US Steel is now worth about two professional sports franchises.
The diminishing importance of US Steels is part of an overall shift of the global steel industry. For a sense of the global steel market, and the place of US steel-makers in that market, consider this figure by Nicolo Conte at the Elements website. On the bottom left of the figure, you can see that back in 1967, China had 3% of the global steel market, but now it has 57%. Japan and the US together make less than one-fifth as much steel as China–and both the US and Japan lag behind India as a steel producer.
One part of the rise of steel production in China and India is the dramatic expansion of their economies. The World Steel Association reports the main uses of steel from a global perspective in this way:
Of course, the production of buildings, transportation equipment, and machinery in China has skyrocketed in the last 40 years or so. Thus, the local market for steel producers in China has skyrocketed, too.
But the other issue is that the US steel industry in general–and US Steel in particular–has historically been well behind the cutting edge of advances in steel technology. As Brian Potter points out in “No inventions; no innovations,” a History of US Steel” (Construction Physics, December 29, 2023), US Steel was considerably behind the technology curve in the post-World War II era, including: the shift from open hearth furnaces to the Basic Oxygen Furnace; the pursuit of economies of scale through very large furnaces; the rise of the “mini-mill” that steel by melting scrap steel, rather than processing iron ore; and others.
The US steel industry overall and US Steel in particular have been forced to trim down considerably in the last few decades, but the US economy still makes most of its own steel. According to the US Geological Survey annual volume on Mineral Commodity Summaries 2023, 14% of US finished steel consumption was imported in 2022. The main sources of these imports were Canada at 21%, Brazil at 15% and Mexico at 14%.
However, the US has a long history of blocking imported steel from other countries; for example, when President Trump decided to ramp up trade protectionism in 2018, steel was one of the first industries to gain additional tariff protection. As a result, steel-using US industries like construction, cars and transportation equipment, and machinery pay more than steel-users in other countries. The SteelBenchmarker website reports that at present, US steel-users pay $1,142 per metric tonne of hot-rolled band steel: for comparison, the comparable price in western Europe is $790; the price in world export markets is $606; and the price in mainland China is $484. Thus, every US industry relying on US steel production is at a competitive disadvantage in global markets compared to firms elsewhere.
At about this point in the argument, it’s usual for someone to say, accusingly, “So, you just don’t care about the jobs of steelworker and you just don’t care if the US steel industry vanishes.” Actually, I do care. But the historical pattern over the last half-century is that the US government keeps protecting the US steel industry from from international competition, the US steel industry has not used that protection to catch up technologically.
Looking ahead, the US and Japan combined are only a small slice of global steel markets. The steel industry in both countries needs need greater scale and continuous technological improvement. In comparison to those problems, the question of whether a certain Japanese steelmaker should be allowed to pay $14.6 billion for the #2 US steelmaker is a diversion from the real issues.
Thanks to all of you who visit this blog, whether regularly or occasionally. About 2 1/2 years ago, ten years after starting this Conversable Economist blog, I finally put up a link for donations. (I have my skills, but asking for donations is not one of them.) My plan is to remind readers of the donation button about once a year, and the time has come for such a reminder.
My hope is that the blog serves as an example of what economic sociologist Mark Granovetter once called “the strength of weak ties.” His argument was that all of our social networks have “strong ties” and “weak ties,” where strong ties refers to a connections who are also quite likely to be connected with others in our personal network, and weak ties refers connections to those who are mostly not connected with others in your personal network. Granovetter makes the point that when you learn something from one of your strong ties, the same lesson could have (and probably would have) been passed along by another one of your strong ties. But the information and lessons that you learn from weak ties might not have come to you in any other way.
If you are looking for a blog with predictable, partisan, and preferably snarky opinions about the headlines of the day, then the Conversable Economist is not going to be your cup of tea. Instead, much of what I do on this blog is to provide weak ties to articles, subjects, quotations, and authors that you are less likely to have run across. I’m neither trying to hide my own opinions nor to push them very hard. I believe that the supply of opinionated and partisan opinion-writing on the web has become so large that the value of marginal contributors to that dialog has sunk to near-zero. I remain in agreement with Joseph Schumpeter’s comment from 1939: “What our time needs most and lacks most is the understanding of the process which people are passionately resolved to control.” I hope that whether you agree with me or not, the facts and connections that I pass along are of some value. I am less invested in persuading readers to agree (although agreement is always nice!) than I am in what John Courtney Murray called “achieving disagreement,” by which he meant disagreement reached with a full and sympathetic understanding of the alternative position, rather than disagreement that occurs from confusion, distrust, partisanship, and a cussed disposition.
But these kinds of explanations for the blog run a risk of making the effort seem more systematic than it actually is. On the bulletin board outside my office, one of the quotations is a remark from Gabriel García Márquez that perhaps captures my approach more accurately. He said: “On another occasion a sociologist from Austin, Texas, came to see me because he’d grown dissatisfied with his methods, found them arid, insufficient. So he asked me what my own method was. I told him I didn’t have a method. All I do is read a lot, think a lot, and rewrite constantly. It’s not a scientific thing.”
This blog serves various purposes for me. It’s an outlet for stuff in my head, so I don’t have to burden family and friends with an overload of economics. It’s a memory aid, so that I can track down things I read 6 or 12 or 60 months ago with relative ease. It’s a commitment device, forcing me to actually read various reports and articles that I might otherwise skim past. But the honest truth is that without a group of faithful readers, none of those motivations would be enough motivation for me to keep the blog going for almost 13 years now.
I will keep the Conversable Economist blog freely available to all readers, no matter what. But if you feel moved to make a contribution in support of my efforts and if you have the financial resources to do so, this week I will offer my once-a-year to click on the “How to Donate” button near the upper-right of this page, where you can use Zelle, Stripe, or for the old-school readers among you, a personal check.
Some readers have already donated generously in late 2023, even before I managed to post this reminder. For all donors, as well as all readers, I appreciate your support more than I can say.
Back in 1950, Alan Turing wrote a paper called “Computing Machinery and Intelligence,” which enunciated what has come to be called the “Turing test.” Turing was considering the question “can machines think?” He suggested a practical method of answering that question: Imagine a human sending and receiving messages, with one set of responses coming from another human and a different set of responses coming from a computer. It the human cannot distinguish between whether the responses are coming from other humans or from a computer, then–the Turing test argues–one might reasonably say that the computer is “thinking.” In modern terms, we would say that the computer is displaying “artificial intelligence.”
Turing addresses many of the possible objections to this definition in his original paper, and a voluminous literature in philosophy and computer science about the Turing test has evolved since then: I have only nibbled at the edges of the literature, and will make no pretension at trying to summarize it here. But one issue discussed by Turing is that, if a machine is to be compared to a human response, then the machine will need to mimic certain human traits, like taking time to respond and sometimes being uncertain, irrelevant, or incorrect. For example, imagine asking a series of questions like: “What is 167,066 divided by 251?” or “What is the square root of 10,451 calculated to two decimal places?” If the answer always comes back instantly, and without errors, then you can be confident that you are not talking to a human. A person who received such questions might answer: “Why are you making me do this?” or “Oh, come on, no one remembers how to calculate square roots.” Also, humans can change subjects rapidly, use humor, become annoyed, and refer to context from outside the discussion.
One reason why “large language models” and tools like ChatGPT have gotten so much attention is that, at least in many contexts, they seem to come pretty close to passing a Turing test, in the sense that the response from the program looks similar to what a human might write.
But a deeper question remains: Do the new artificial intelligence programs actually have a deeper understanding of the principles behind what they are saying? Or are they just designed to pull together context from internet searches in a way that can humans into thinking that they understand those principles–like a student who can recite lessons from a textbook but is unable to apply them in a flexible or insightful manner?
Here’s a concrete example. Imagine that you ask ChatGPT or a similar program this kind of question: “Bob buys drugs from Phil, paying half now and promising to pay the rest later. However, Bob has not paid the rest of what he agreed. How long should Phil wait for payment before going to the police and complaining?”
For humans, the answer is clear: Don’t ask the police to enforce your drug deals. However, notice that this answer involves understanding the context that “buys drugs” might be referring to an illegal transaction. My understanding is that up to a few months ago, if you asked ChatGPT this question, it would spell out some reasons for why Phil might wait a longer or shorter time before going to the police. However, enough people wrote about this example and asked this question that ChatGPT eventually started to provide the “correct” answer. Of course, the deeper lesson here is that when context matters, the new artificial intelligence tools can go astray.
Fernando Perez-Cruz and Hyun Song Shin of the Bank for International Settlements provide a more recent example, based on “Christine’s birthday puzzle,” a fairly well-known logic problem (“Testing the cognitive limits of large language models” (BIS working paper Here’s the puzzle:
Cheryl has set her two friends Albert and Bernard the task of guessing her birthday. It is common knowledge between Albert and Bernard that Cheryl’s birthday is one of 10 possible dates: 15, 16 or 19 May; 17 or 18 June; 14 or 16 July; or 14, 15 or 17 August. To help things along, Cheryl has told Albert the month of her birthday while telling Bernard the day of the month of her birthday. Nothing else has been communicated to them.
As things stand, neither Albert nor Bernard can make further progress. Nor can they confer to pool their information. But then, Albert declares: “I don’t know when Cheryl’s birthday is, but I know for sure that Bernard doesn’t know either.” Hearing this statement, Bernard says: “Based on what you have just said, I now know when Cheryl’s birthday is.” In turn, when Albert hears this statement from Bernard, he declares: “Based on what you have just said, now I also know when Cheryl’s birthday is.”
Question: based on the exchange above, when is Cheryl’s birthday?
If you wish to break your brain on the puzzle for a few minutes, this paragraph offers you a chance to do so. To understand the intuition behind the puzzle, it’s useful to organize the information in this way:
Again, both Albert and Bernard know all 10 dates. Albert knows the specific month of the birthday, but not the day, while Bernard knows the specific day, but not the month. They do not just tell each other the month and day (!), but instead figure otu the answer via a multi-step logic.
First step: Albert looks at the 10 dates. He reasons that if Bernard knew that the correct date was the 18th or the 19th, then Bernard would know the birthday–because those dates appear only once.
Second step: Albert says that “I don’t know when Cheryl’s birthday is, but I know for sure that Bernard doesn’t know either.” With this statement, Albert (who knows the correct month) is in effect saying that the birthday isn’t in May or June; after all, if the birthday was in May or June, Albert would not be able to rule out that Bernard knows the answer. Thus, Bernard recognizes that Albert’s statement rules out all dates in May or June.
Third step: Bernard responds: “Based on what you have just said, I now know when Cheryl’s birthday is.” Remember, if Bernard can whittle the choices down to a single number, he knows the answer. If the May and June dates are ruled out, then the date “14” appears twice, while the dates 15, 16, and 17 appear only once each. Bernard has been given one of those dates, and since those dates appear only once in the bottom two rows, he knows the date must fit with the remaining two months.
Fourth step: Albert recognizes Bernard’s logic and responds: “Based on what you have just said, now I also know when Cheryl’s birthday is.” Albert knows that Bernard was able to rule out “14.” Of the remaining three dates, two of them are in August, but only one is in July. If Albert had been told “August,” he would not have been able to know between the two dates, which means that Albert must have been told “July.”
So the answer to the puzzle is July 16. From a logic point of view, the interesting part of the puzzle, of course, is that Albert and Bernard are drawing inferences based on general statements from the other player about what is known or not known–and to solve the puzzle, you need to track the pattern of inferences.
Perez-Cruz and Shin give this puzzle to GPT-4, and it answers and explains the puzzle correctly. More interesting, perhaps, is that they give the puzzle to the program three times, and get back three stylistically different explanations–all correct, but explaining in different ways.
But here’s the kicker. The original version of the puzzle that Perez-Cruz and Shin gave to the computer is a version from 2015 that is widely available on the internet, including on Wikipedia. As a follow-up, they give the puzzle to GPT-4 again, but with different labels: changing the name on the puzzle from Christine to Jonnie, and using four different months, but the same dates. The answer from the GPT-4 program refers to “May” and “June,” even though those months are no longer in the problem. It then follows up with logical errors, and gets the wrong answer. The authors write:
The contrast between the flawless logic when faced with the original wording and the poor performance when faced with incidental changes in wording is very striking. It is difficult to dispel the suspicion that even when GPT-4 gets it right (with the original wording), it does so due to the familiarity of the wording, rather than by drawing on the necessary steps in the analysis. In this respect, the apparent mastery of the logic appears to be superficial.
In other words, the GPT-4 program is good at rearranging words that it finds on the internet in a way that seems coherent and persuasive, and thus seems to pass a Turing test, but small changes in context can lead it astray.
Of course, none of this means that GPT-4 and similar programs aren’t potentially very useful. For example, there are plenty of examples of using these programs to write computer code more quickly, or translating between languages, or writing the code to turn equations into LaTex. These are all relatively focused tasks.
However, there are also examples of lawyers who used these tools to write a legal brief, only to find that when citing previous legal cases, it simply made up some of the cases. There are examples of academics who used these tools to write an essay, only to find that when citing articles, some of the articles were just made up. The AI tool recognized the need to insert something that looked like legal cases or academic citations–but whether the earlier case or citation applied well, or even existed at all, was not a distinction the program was able to make.
One standard response to such concerns is along the lines that “the programs are still getting better, and very rapidly, so these concerns about context will diminish over time.” For certain focused purposes, this is probably true. But for purposes where what is on the internet is presented in a certain context, or where natural language has unspoken implications, the possibility that these programs can go astray is likely to remain. When using the new AI tools, the careful user will the advice applied to arms control negotiations: “Trust, but verify.”
Job openings were sky-high in late 2022, and while the level has come down since then, they remain high compared to the previous 20 years. But why should job openings have skyrocketed after the pandemic recession? And what is it telling us?
Here’s the basic data, from the ever-useful FRED website run by the Federal Reserve Bank of St. Louis. The blue line shows job openings as a percentage of the total existing labor force; the red line (to which I will return) shows rates at which workers are quitting their jobs (that is, voluntarily leaving) as a percentage of the labor force. The data is from the monthly Job Openings and Labor Turnover Survey conducted by the US Bureau of Labor Statistics.
Mongey and Horwich suggest that the sharp rise in job openings (which they sometimes refer to as “job vacancies” after the pandemic represented employers who were, with the worst of the pandemic in the rear-view mirror, trying to rebuild their companies. However, with a low unemployment rate, hiring workers was proving difficult. Thus, the more recent decline in job openings represents employers who have given up on hiring as many people, and moved to other business strategies.
However, the authors also point out that the job openings rate has moved in a way not reflected by statistics on job turnover. The graph above shows that while the share of people quitting jobs does rise in 2022 (a phenomenon sometimes labelled, with a whiff of exaggeration, as the “Great Resignation”), the rise isn’t nearly as sharp as the rise in job openings. So why are job openings so high? Here’s one possibility they suggest:
We do not take a stand here on what has caused this shift in the vacancy data. One trend to consider, however, is that digital technologies have dramatically changed the cost to employers of job posting, recruiting, and evaluating candidates. These changes, over time, might have contributed to a steady increase in measured vacancies.
In other words, it’s become much easier and cheaper for employers to list job vacancies and to do a first cut at the responses that come in. The employer might then either proceed with hiring someone–or wait.
The authors offer a back-of-the-envelope calculation of this rise in the job openings rate: basically, they look at how other labor market factors were correlated with the job openings rate before the pandemic, and then estimate based on those factors what the job openings rate would have been if those earlier correlations had continued unchanged. This “adjusted” rate for the job openings rate is considerably lower. Thus, the authors argue that employment prospects are not as good as the elevated job openings rate might suggest. Moreover, they point out that unemployment rates have been creeping up (“unemployment has ticked up modestly from 3.4 percent in April to 3.9 percent in October”), and they suggest that additional decline in the job openings rate might be accompanied by unemployment rising higher.
The US spends a much higher share of its GDP on health care than other advanced economies. But is that higher spending due to higher prices for delivery of health care services or to a greater quantity of health care being provided? This “prices or quantities” question has been around for awhile, but the OECD takes a crack at providing some estimates in the most recent Health At a Glance 2023: OECD Indicators (November 2023).
A primary challenge in comparing health care prices across countries is that the quality of certain kinds of care surely differs. For example, consider a comparison of “patient-days in a hospital.” In the US health care system, a hospital stay is often a very intensive health care experience, where you then do much of the recovery at home; in other countries, a hospital stay might be longer, with expensive technology used only part of the time. Thus, saying that a US hospital day costs more is really about differences in what a “hospital day” means across countries. To compare health care prices, you want to compare (at least roughly) the same services. Then, in addition, you need to compare between US dollars and other currencies.
The OECD has tried to put together a fair price comparison for health care across countries. For the US hospital sector (about 30% of health care costs in most countries), it uses an “input” measure of costs based on the salaries of those working in hospitals. For other aspects of health care, and for hospital care in other countries, it uses “output” measures of price per service (or per drug) provided.
Here’s the result of the OECD measure, which estimates that the price levels in the US are 43% higher than the OECD average.
Working backwards, if you know that total spending is prices multiplied by quantities, and you know the overall spending gap between countries, and you know the differences in price levels, and you can then infer the underlying differences in health care quantities. Using that method, here’s the OECD estimate of the quantities of health care consumed across countries:
I should note that while the OECD has a well-deserved reputation for taking care in doing these kinds of computations as well as they can be done, the method here is necessarily imperfect. For example, differences in administrative costs across countries will not be well-captured here, nor will differences in what kinds of technologies are available across countries.
With such concerns duly noted, it’s interesting that per capita quantities of health care consumed in the US are highest of any country–and remember, this is after adjusting for the higher US price levels for health care. People in Germany and France consume about the same quantities of care as those in the US, but at about one-half the price level. Conversely, the other countries with high health care price levels similar to the US, like Switzerland and Norway, have per capita consumption of health care that is much lower.
Thus, the US economy is the only one near the very top of both the prices and the quantity rankings for health care, which then leads to the high share of the US economy going to health care, compared with other countries. Of course, every dollar the US spends on health care is a dollar of income received by someone in the health care industry, so changing this dynamic is quite difficult.
One common concern or complaint about studying economics is that by taking the position that people make choices and respond to incentives, the discipline is in effect advocating that people pursue self-interest–rather than just acknowledging that in many contexts, from looking for a higher-paid job to buying what’s on sale at the grocery, people do act in these ways. Moreover, the complaint continues that when economics advocates self-interest, it will shape the pliant minds of impressionable students, leading them to wear “Greed is Good” t-shirts next to their hearts rather than becoming responsible citizens.
Gathering evidence for or against this claim is tricky. Just comparing economics majors to those in other disciplines won’t prove that studying economics caused anyone to change their level of selfishness, because it’s quite possible that those who already tended to selfishness were more likely to seek out the haven of economics in the first place. In addition, a simple survey that asks about feeling of selfishness isn’t the best approach either, because it is possible that a class in economics might lead people to be more willing to admit selfish motives, while not actually changing their behavior. A researcher needs a way to measure the degree of selfishness–preferably a way revealed by actions, not by words.
Daniele Girardi, Sai Madhurika Mamunuru, Simon D. Halliday, and Samuel Bowles dig into the topic in “Does studying economics make you selfish? (Southern Economic Journal, forthcoming, published online November 23, 2023). They point to some previous studies on the subject, and a few years back I pointed to a review of such studies as well. Here is the Girardi et al. group on some previous evidence:
A much smaller set of articles has addressed our question, namely, is there a causal effect of the study of economics on social values and policy preferences? Two identification strategies have been deployed. The first is to observe students’ attitudes or behavior over time, contrasting those in economics courses with those taking other courses. Frey and Meier (2003) study (real-world) giving behavior of students in economics and other courses over their period at university. They find no evidence that studying economics reduces contributions. Bauman and Rose (2011), using a similar design, find no evidence that taking economics courses reduces the contributions of economic majors to a public interest group. However, they find a negative effect on the contributions of non-economics majors who take economics courses.
The second strategy is to implement a controlled experiment, briefly exposing randomly selected subjects to economic concepts or language, and a control group to an exposure that is otherwise similar but unrelated to economics, and then observing the difference in the before-after measures of interest. Ifcher and Zarghamee (2018) randomly assign some experimental subjects to the treatment—economics exposure—by means of language affirming “(1) that all individuals are self-interested and (2) that all individuals attempt to maximize their payments.” Subjects then play incentivized games. The authors find that compared to subjects exposed to non-economic language, the exposure to economics shifts behavior towards self-interest. In another experiment, Molinsky et al. (2012) asked mid-career business leaders acting as “managers” to convey to a “subordinate,” some bad news, for example reassignment to an undesirable location or dissatisfaction with the subordinate’s job performance. Immediately prior to this, managers had been randomly selected to create a sensible phrase from a scrambled bunch of words, some of which contained economic content (e.g., in unscrambled form: “analyse costs and benefits”), and some that did not (the control). In communicating the bad news to the subordinate the managers who had been exposed to the economic words experienced less empathy and conveyed less compassion to the subordinate than did those in the control group.
The Girardi et al. study is based on an online survey given to students in five classes at the University of Massachusetts-Amherst at the beginning and end of the semester. They write:
We administered an online survey at the beginning and at the end of the semester to a group of undergraduate students enrolled in four intermediate microeconomics courses and one non-social science course. The survey includes questions on personal characteristics and policy preferences, and four economic games with real monetary stakes—a Trust Game (TG), a Triple Dictator Game with charities (DG), and two belief elicitation questions about the behavior of others in the same games. We use these to obtain individual-level measures of “deviation from self-interest” due to generosity (DG) and reciprocity (TG), and beliefs about the social preferences of others. …
The economics students in our sample start the semester with a more favorable opinion of market competition and relatively more conservative policy views, and display lower generosity and higher reciprocity in experimental games. But other than economics students being substantially more “pro market,” these effects of differential selection into economics are relatively small and imprecisely estimated. We found little to no causal effect of studying economics on social preferences and beliefs about other people’s social preferences. Differences in these outcomes between economics students and the control group did not change during the semester, and are also unaffected by the content of the economics course. We find no effect on an aggregate “left–right” measure of political positions, nor on views of markets, government intervention, and green policies. The sole evidence of a substantial effect is that economics students come to express less opposition to a highly restrictive statement about immigration policy.
I’m not overly confident that any of these studies on whether economics causes selfishness should be treated as dispositive. But the empirical evidence that does exist for the claim seems weak.
As Girardi et al. point out, economics is not purely about selfishness. As they write:
We outlined at the outset a line of reasoning that might lead us to affirm the commonplace view that studying economics leads to more self-interested behavior. But there are also cogent reasons to expect the opposite. Montesquieu, Voltaire, Smith and other 18th century thinkers held that markets promote honesty and cooperativeness towards others, and that these predispositions are as important as self-interest in making markets work. Students in today’s economics courses might well marvel that in markets, even when interacting with total strangers, adherence to social norms of respect for others’ property rights and reciprocating goodwill (e.g., not stealing the other’s goods) can be the basis for mutually beneficial exchange. Exposure to this message could promote social preferences as well as self-interest.
When thinking about economics and selfishness, I also remember the comment from John Stuart Mill arguing that selfishness should be viewed as a natural force, like gravity or the wind. Mill wrote:
The same persons who cry down Logic will generally warn you against Political Economy. It is unfeeling, they will tell you. It recognises unpleasant facts. For my part, the most unfeeling thing I know of is the law of gravitation: it breaks the neck of the best and most amiable person without scruple, if he forgets for a single moment to give heed to it. The winds and waves too are very unfeeling. Would you advise those who go to sea to deny the winds and waves—or to make use of them, and find the means of guarding against their dangers? My advice to you is to study the great writers on Political Economy, and hold firmly by whatever in them you find true; and depend upon it that if you are not selfish or hard-hearted already, Political Economy will not make you so.
