Book review: Cloud Empires

The libertarian ideal of the internet was that it was a place without borders, without gatekeepers, and without government control. However, the modern internet falls well short of that ideal. In the physical world, it is typically governments that make and enforce the rules. However, online it is increasingly large and undemocratic platform firms that make the rules and enforce them. That is the general idea underlying Vili Lehdonvirta's 2022 book Cloud Empires, which I just finished reading.

Lehdonvirta tracks in detail how we ended up in the current situation, noting that:

The Internet was supposed to free us from powerful institutions. It was supposed to cut out the middlemen, democratize markets, empower individuals, and birth a new social fabric based on self-organizing networks and communities instead of top-down authority. "We will create a civilization of the mind in Cyberspace... more humane and fair than the world your governments have made before."... This is what Silicon Valley's visionaries promised us. Then they delivered something different - something that looks a lot like government again, except that this time we don't get to vote.

Lehdonvirta outlines how the platform firms have essentially replicated the process by which governments established rules, because of the same underlying necessity to maintain control. He uses numerous examples including Amazon, eBay, and cryptocurrencies such as Bitcoin and Ethereum, to illustrate his points. These case studies demonstrate the challenges, and the close corollary between the economic institutions established by the platform firms and those established by governments. Lehdonvirta notes that the key difference between governments and platform firms is in the political institutions. Platform firms lack the accountability that is inherent in political systems, and there is little prospect of overturning the 'government' of a platform. Even the most autocratic state risks revolution in a way that is to a large extent impossible for users to achieve within a platform environment.

While Lehdonvirta does a great job of outlining the issues, where the book falls short is in terms of the solutions. The subtitle of the book promises to tell us, "how we can regain control". Lehdonvirta's solution is a 'bourgeois revolution', of the kind that western countries experienced through the late Middle Ages. The growing urban middle class ('burghers') developed significant resources and gradually pushed back against the local lords, helped by powerful allies in the Church and often the monarchy as well. These coalitions led to more devolution of political power and authority, and eventually to the modern political institutions we observe today.

Lehdonvirta notes that, with some creative licence, it is possible to imagine a similar dynamic playing out on the platforms. However, while he devotes a great deal of effort in explaining the problems and linking them to real-world case studies, he doesn't expend the same effort on his proposed solution. The reader receives a few, almost cursory, observations about how a 'bourgeois revolution' may play out in certain situations. I felt like the book needed a more detailed explanation, linking the solution to embryonic real-world efforts and charting a path forward for them. Although speculative, a 'road map' for advocates of returning some power to the platform users would have added significant value to the book.

Aside from that small gripe, I really enjoyed this book, and it was a good follow-up to reading the more textbook treatment of platforms found in The Business of Platforms (which I reviewed last week).

This week in research #121

Here's what caught my eye in research over the past week (a quiet week, as I have been travelling in Europe):

• Three articles published in the prestigious journal Nature by Miske et al., Aczel et al., and Tyner et al., investigate the replicability of research results in social and behavioural sciences (a very important set of papers that have garnered a lot of attention)
• Mišák (open access) investigates the impact of temperature on soccer team performance, and finds that attacking efficiency is enhanced in warmer conditions, leading to increased goal productivity and improved shot conversion rates, defensive performance appears to weaken in warmer conditions, with a decrease in defensive pressure and passing accuracy, and player aggression follows an inverted U-shaped pattern in relation to temperature

The impact of the 2023 Bud Light boycott on alcohol purchases

When a consumer stops buying a particular product for some reason (for example, if a product becomes unavailable), do they switch their spending to another product within the same category, or do they reallocate their spending across all available goods and services? The consumer choice model (or the constrained optimisation model for the consumer) suggests that the consumer should reallocate across all possible goods and services, rather than transferring the exact proportion of spending to the closest substitute product.

This recent article by Aljoscha Janssen (Singapore Management University), published in the journal Economics Letters (ungated earlier version here) provides an interesting test of that expected response. The context is the 2023 boycott of Bud Light in the US:

The boycott began in early April 2023 after Bud Light partnered with a transgender creator, prompting calls from conservative media to avoid the brand... Viral content amplified the message, and the manufacturer responded with advertising that emphasized traditional Americana themes... Sales declines emerged not only in conservative areas but also in regions without strong ideological leanings...

Janssen uses data from the NielsenIQ Consumer Panel from 2021 to 2023, which tracks spending by between 40,000 and 60,000 US households. Janssen drops households that did not buy alcohol, and then categorises the remaining households into three groups based on Bud Light purchases: (1) 'Bud Light households' (that purchased 18 litres of Bud Light in both 2021 and 2022); (2) 'Bud Light-dominant households' (that purchased at least twice as much Bud Light as other beers, in addition to purchasing at least 18 litres of Bud Light in both 2021 and 2022); and (3) 'Non-Bud Light beer households' (that purchased at least 18 litres of light beer in both 2021 and 2022, of which less than one-third was Bud Light). Janssen reports that:

In the full sample there are 34,470 alcohol-purchasing households; 585 qualify as Bud Light households and 439 of those are Bud Light-dominant, while 5130 are non-Bud Light beer households.

Janssen analyses monthly purchase data using a difference-in-differences approach, essentially comparing the difference in purchases between different treatment and control groups before and after the Bud Light boycott in April 2023. In practice, the comparisons show very similar results for the impact on Bud Light purchases, purchases of other beer, and total alcohol purchases. Specifically, Janssen finds that:

Across all designs, treated households reduce Bud Light by roughly 160 ounces per month (34%–37% of their pre-boycott Bud Light volume)...

Households partially replace Bud Light with other beer: other-beer purchases rise by 70–90 ounces per month. The offset is meaningful but incomplete relative to the Bud Light shortfall...

Net of substitution, total ethanol declines by about 3–4 fl-oz per month among treated households, a 5.5–7.5% drop. Converting with 0.6 fl-oz per U.S. standard drink, this equals roughly 5.0–6.7 drinks per month per treated household...I find no significant changes in wine or spirits, indicating that switching is almost entirely within the beer category.

So, the boycott led households on average to purchase less Bud Light (as you might expect from a boycott). They bought a greater quantity of other beer products, but the increase in other beer purchases was less than half the decrease in Bud Light purchases, meaning that consumers substituted to other non-beer products. Consumers also didn't switch entirely to other alcohol products, as total alcohol purchases declined. Instead, some spending appears to have shifted away from alcohol altogether. In other words, consistent with the consumer choice model, when consumers stopped buying (or reduced their purchases of) Bud Light, they reallocated their spending across all goods and services, not just switching their spending to the closest substitute to Bud Light (other beers).

Does this offer anything meaningful for advocates of reduced alcohol consumption? Probably not in any direct sense. These were fairly unusual circumstances, and consumer boycotts of particular alcohol products are uncommon. It is hard to imagine advocates or policymakers being able to engineer similar boycotts on a regular basis in order to reduce alcohol consumption. However, the findings do suggest a broader possibility. Interventions that reduce purchases of particular alcohol products, especially those associated with high levels of alcohol-related harm, may lead to at least some reduction in overall alcohol purchases, rather than consumers simply switching one-for-one to the nearest substitute. That said, this study is about purchases rather than consumption, and more evidence from other types of interventions would be needed before drawing firm policy conclusions.

Taylor Swift, look what you made fans buy

Taylor Swift released 27 versions of her 2025 album The Life of a Showgirl. That sounds excessive, but it offers a nice lesson in economics and pricing strategy, specifically price discrimination.

Price discrimination occurs when a firm charges different prices to different groups of consumers for the same good or service, and where the price differences do not arise from a difference in costs. One form of price discrimination is 'versioning', where the firm offers different versions of a product that each cost the same to produce, but which appeal to different groups of consumers (with different price elasticities of demand). Consumers that are more price sensitive (and have more elastic demand) would buy the version of the product that is less expensive, while consumers that are less price sensitive (and have less elastic demand) would buy the more expensive version.

We saw an extreme example of versioning last year, executed by the astute economist Taylor Swift. Paul Crosbie (Macquarie University) wrote about it in this article in The Conversation last October:

The Life of a Showgirl was released in dozens of formats, with physical and digital editions tailored to different levels of commitment.

In total, over the first week, there were 27 physical editions (18 CDs, eight vinyl LPs and one cassette) and seven digital download variants.

A range of covers, coloured vinyl, bonus tracks and signed inserts turned one album into a collectable series rather than a single product. Other artists – such as the Rolling Stones – have used this strategy before, but rarely at this scale or with such an intense response from fans.

Taylor Swift fans who are more price sensitive will have tended to buy the less expensive version of the album. More price-sensitive fans will include those who have lower incomes (where the album price is a higher proportion of their income) and those who are more casual Taylor Swift fans (where there are more substitutes available that they might prefer to spend their income on).

Taylor Swift fans who are less price sensitive will have tended to buy the more expensive premium version of the album. Less price-sensitive fans will include those who have higher incomes (where the album price is a smaller proportion of their income) and those who are more diehard Taylor Swift fans (where there is no close substitute for the latest Taylor Swift album).

Crosbie questions whether it is possible to have too many versions of a product. In this case, 27 versions do seem like a lot. It could be a very effective means of segmenting the market. However, that works best if each buyer only buys one version. There will be some fans who bought more than one version, and perhaps a substantial number who bought several. A non-trivial proportion of the most diehard fans probably own all 27 physical editions of the album.

This matters because the usual rationale for price discrimination through versioning is that consumers sort themselves across the available versions - the casual fans buy the standard version, while the diehard fans buy the premium one. But if some consumers buy multiple versions, the strategy is doing more than just segmenting the market. It is also encouraging multiple purchases from the same buyer. In that case, the different versions are not just substitutes for one another, but for some fans they become collectibles, with each version they collect adding a bit of extra value through completeness, exclusivity, or identity. So, the economics of versioning for Taylor Swift are not only about price discrimination between consumers, but also about extracting more surplus from the most committed fans.

There is a limit to how many versions even the most diehard fan is willing to buy, and that limit arises because of diminishing marginal utility. In economics, utility is the satisfaction or happiness the consumer gets from the goods and services they consume. Marginal utility is the extra utility the consumer gets from consuming one more unit of a good or service. Diminishing marginal utility is the idea that marginal utility declines as the consumer consumes more of a good. In the context of Taylor Swift's album, Crosbie notes that:

The first version of an album brings a lot of satisfaction. The fifth or sixth brings less. Eventually, another version does not add enough enjoyment to justify the price. Fans begin to feel they have had enough.

It is clear that there is a balance to be found between maximising profits by price discrimination using versioning, and the number of versions that are offered when some consumers will want to buy multiple versions. Price discrimination can be an incredibly profitable pricing strategy for firms, including for Taylor Swift. Maintaining fan engagement and encouraging diehard fans to spend more by making the versions collectible are also important. As Crosbie notes in his article:

Instead of leaving that money on the table, the strategy turns passion into profit. The cost of creating extra covers or vinyl colours is small, but the willingness of fans to pay more for them is high. That is exactly where versioning pays off.

In theory, it should be possible to work out the optimal number of versions that maximises long-run profit. The profit-maximising number of versions is not necessarily the number that best segments the market for the purposes of price discrimination, because diehard fans may buy multiple versions. It seems likely that Taylor Swift is well aware of this. Would you be willing to bet she hasn’t gotten close to that optimum? I wouldn’t.

Facebook Marketplace forces a change in TradeMe's business model, but will it succeed?

TradeMe is one of the key examples that I use when teaching about platform markets in my ECONS101 class. But competition from Facebook Marketplace is causing TradeMe to change its business model, and those changes are risky.

The reason why TradeMe is such a good example is that, by attracting buyers and sellers to its platform in the 1990s, TradeMe managed to keep eBay out of the New Zealand market. How did that happen? In a platform market, the firm (in this case, TradeMe) acts as an intermediary that brings together two parties (in this case, buyers and sellers) who would not otherwise interact or easily connect. Buyers using TradeMe create value for sellers, and the more buyers there are, the more value is created. Sellers using TradeMe creates value for buyers, and the more sellers there are, the more value is created. Once TradeMe was set up and had attracted a large share of buyers and sellers, it would be difficult for any other platform to set up in competition with TradeMe. And so, eBay couldn't get a foothold in New Zealand, and TradeMe had an effective monopoly over online auctions for many years.

TradeMe profited by charging a 'success fee' to sellers of goods on the platform. Buyers faced no fees. This reflects the principle that a platform firm (like TradeMe) should set a lower price for access to the platform to whichever side of the market has demand that is more elastic, ceteris paribus. In this case, buyers have more elastic demand for access to TradeMe, as there are many other places that they might go to buy things. Sellers, on the other hand, had more inelastic demand for access to TradeMe because no other place had access to the same quantity of buyers. That is, until Facebook Marketplace appeared.

Facebook Marketplace doesn't charge fees to sellers. And through its links to Facebook users, there are a large number of potential buyers on Facebook Marketplace. And so, there is now a viable (and cheaper) alternative to TradeMe for sellers. And now, TradeMe has finally reacted to this competition, as reported in the New Zealand Herald last month:

Trade Me is removing success fees for casual sellers, in a move that one marketing expert says is probably a response to the growing power of Facebook Marketplace.

Sellers have usually been paying 7.9% of the final sales price of items sold via Trade Me.

But a new fee structure will remove them from next week and site spokeswoman Lisa Stewart said casual sellers would be better off.

It is making other changes at the same time: bank transfers will not be possible and Ping will be offered on every listing alongside cash and Afterpay, with a 2.19% transaction fee for the seller. This provides buyer protection up to $5000 if trades go wrong.

Buyers will also pay a new service fee based on the purchase price, if items are more than $20. This will be 99c for goods sold for $20.01 to $100, $1.99 for sales between $100.01 and $250 and $4.99 for items over $250. Stewart said 44% of trades were under $20.

It was a response to customer feedback and what was happening in the market, Stewart said...

Massey University marketing expert Bodo Lang said it was likely to be in response from growth in the use of Facebook Marketplace, which offers no protection for buyers but charges no fees.

With Facebook Marketplace available for sellers, seller demand for using TradeMe has become more elastic. While the 'success fees' have been eliminated, TradeMe will still profit from sellers through mandating the use of its payment service Ping. And notice that buyers will also now pay a 'service fee' to TradeMe on successful purchases over $20. TradeMe is now leveraging its market power to derive revenue from a complementary good (payment services) rather than the auctions themselves. This change makes TradeMe's business model resemble that of Facebook Marketplace. Meta derives revenue from Facebook Marketplace though selling advertising (on Facebook Marketplace, but also on Facebook), rather than deriving revenue directly from the sales on Facebook Marketplace.

It remains to be seen whether Trademe's new business model will be successful. I question the wisdom of charging a service fee to buyers. They are still the more elastic side of Trademe's platform market, and they have a cheaper option available in the form of Facebook Marketplace. TradeMe is banking on buyers valuing the protection that TradeMe offers them, which Facebook Marketplace does not. However, it isn't clear how much buyers value that protection, and if they don't value it in excess of the new service fee, then they will continue to exit to Facebook Marketplace. And if buyers show an increasing preference for Facebook Marketplace, it won't be long before sellers start to reduce their listings on TradeMe. And if that happens, the market could tip and TradeMe could very quickly find itself in an irreversible decline.

This week in research #120

Here's what caught my eye in research over the past week:

• Büyükeren, Makarin, and Xiong (with ungated earlier version here) find that the full-scale launch of Tinder led to a sharp, persistent increase in sexual activity among college students, but with little corresponding impact on the formation of long-term relationships or relationship quality
• Chollete et al. find that recreational marijuana legalisation by US states increases property crime, although the effect disappears when they control for state-specific time trends
• Picault (open access) describes a method to introduce authentic group projects in senior undergraduate economics courses

Book review: The Business of Platforms

As I teach my ECONS101 class, a platform market occurs when a firm acts as an intermediary and brings together two (or more) groups, who otherwise would not connect or easily interact (platform markets are also known as two-sided markets, because the intermediary brings together two sides of the market). We think of platforms as mostly an invention of the digital age, because most of the examples that come to mind (TradeMe or eBay, Facebook, the Android operating system) are digital platforms. But in truth, platforms are everywhere. Credit card companies are platform firms - they bring together merchants and cardholders. Malls are platform firm - they bring together stores and customers. And so on. Once you know what to look for, you recognise just how endemic platforms have become in the modern economy.

I was interested to learn more about the business of platforms, so a few years ago I bought the 2019 book The Business of Platforms by Michael Cusumano, Annabelle Gawer, and David Yoffie. I finally got around to reading it last month. It was not a moment too soon either. I was inspired by the book to greatly expand on the platform market content in my ECONS101 class, specifically in the topic we covered last week. I'm sure they won't thank me for adding to the quantity of ideas that they may be assessed on, but the material from the book added a lot of depth to what was previously a fairly cursory description of two-sided markets.

The importance of understanding platforms is exemplified by their growing importance in the (global) economy. Cusumano et al. write that:

In short, managers and entrepreneurs in the digital age must learn to live in two worlds: the conventional economy and the platform economy.

To that, I would add that consumers and policy makers also need to understand the fundamentals of platform markets. It is there that this book excels. Cusumano et al. provide a clear description of what platform markets are, the 'winner-take-all (or most)' nature of those markets, and the different types of platform markets. They use a wide array of examples to illustrate the concepts, from Android to YouTube, and everything in-between. I'm sure that they could have easily turned the exercise into a textbook treatment. However, the numerous examples they use give more depth and provide more interesting perspective than you would get from a textbook.

After outlining the basics over the first few chapters, the book turns to common mistakes that platform firms make. Many of their examples will be familiar as exemplars of business failure, such as how Microsoft first captured the browser market with Internet Explorer, before subsequently losing their dominance to Firefox and ultimately to Google's Chrome browser. Next, the book looks at how firms can develop a platform, again carefully illustrating the pitfalls of the different options available to firms with real-world examples.

Finally, the last section looks to the future of platforms, but also takes a more normative view, advocating that platform firms should "harness platform power, but don't abuse it". Cusumano et al. have written their book with managers and entrepreneurs in mind, and this last section is an appeal to those future leaders of platform firms. In particular, they focus on antitrust issues, privacy concerns, fairness towards the workforce ("not everyone should be a contractor", and in particular that firms should self-regulate. While this section does paint a picture of how platform firms can easily become bad actors, it seems unlikely to move the needle on platform firms' worst abuses.

Overall, I really enjoyed this book. It is rare these days that a single book adds significant new content to one of the papers I teach, and I really appreciated the clarity that Cusumano et al. bring to this topic, and the way they structured their ideas in a way that was easy to follow. If you are looking to understand platform markets, this book seems essential to me, and I highly recommend it.

The tone and expression of academics on X (or Twitter)

In my previous post, I highlighted the apparent contribution of X (formerly Twitter) to toxicity on the Economics Job Market Rumors (EJMR) website. A natural follow-up question is whether and to what extent academics on X contribute to the toxicity on that platform and, by extension, to other forums such as EJMR. This recent article by Prashant Garg (Imperial College London) and Thiemo Fetzer (University of Warwick), published in the journal Nature Human Behaviour (open access), goes some way towards providing an answer.

Garg and Fetzer constructed a dataset of nearly 100,000 academics, including all of their Twitter [*] activity from 2016 to 2022. They then use large language models (ChatGPT-3.5 and GPT-4) to characterise each tweet in relation to content and tone. They assess each academic's stance on climate change, economic policy, and cultural issues. In terms of tone, they measure egocentrism (how often the academic refers to themselves in the first person), toxicity (based on the probability a tweet is classified as toxic by Google's Perspective API), and the balance between reason and emotion (measured as a ratio of 'affective terms' to 'cognitive terms' based on the Linguistic Inquiry and Word Count tool). The analysis is then largely descriptive, but nonetheless interesting.

Garg and Fetzer first find that:

...leading academics are not typically social media influencers... We found weak correlations between citation counts and Twitter metrics: citations and likes... citations and followers... and citations and content creation...

Garg and Fetzer observe that:

The weak correlation underscores that many prominent public intellectuals online gain visibility through public engagement rather than scholarly achievements, often holding lower academic credentials while commanding significant public attention, thus widening the gap between social media influencers and established academic experts.

I think that Garg and Fetzer overstate the case here. The weak correlations suggest that Twitter includes a cross-section of academics (in terms of academic quality), rather than that the top academics eschew Twitter (which would instead lead to negative correlations between measures of academic quality and Twitter engagement).

I'll put aside their results on political expression, which I round rather uninteresting. In contrast, the results in terms of tone demonstrate some interesting correlations. First, in terms of egocentrism (using self-referential terms such as 'I', 'me', 'my', and 'myself'):

Female academics... exhibit higher egocentrism than male academics...

Egocentrism increases with university ranking: academics at top-100 institutions... exhibit higher egocentrism than those from institutions ranked 101-500... US-based academics... show higher egocentrism than non-US academics

Then, in terms of toxicity:

Academics with high reach but low academic credibility... exhibit lower toxicity than those with the contrasting profile, that is, ones with low reach but high credibility...

Academics at top-100 universities... exhibit higher toxicity than those at institutions ranked 101-500... Moreover, US-based academics... exhibit higher toxicity than non-US academics...

And in terms of emotionality (or reason):

Emotionality is significantly higher among female academics... than male academics... In terms of reach and credibility, high-reach/low-credibility scholars... show significantly higher emotionality than low-reach/high-credibility scholars...

Finally, US-based academics... exhibit higher emotionality than non-US scholars...

Many of those differences will surprise no one, such as US-based academics being more egocentric and toxic in their expression on Twitter. Other differences seem to confirm familiar stereotypes, such as female academics using more emotional language than male academics. No doubt, some of the differences relate to differences in norms across different disciplines in terms of communication styles (both on Twitter and in general academic discourse). Garg and Fetzer don't control those other factors that might affect tone and expression. And before we get carried away about how toxic academics are on Twitter, Garg and Fetzer provide an important comparison with the general population. From Figure 6 in the paper:

Notice that academics (the blue line) exhibit far less toxicity (in the graph in the top middle) than the general population of Twitter users (the red line). Moreover, the trend in toxicity is downwards (for academics over the whole period from 2016 to 2023, and for the general population from 2021 to 2023). So, academics are not the main problem in terms of toxicity in the discourse on Twitter.

Nevertheless, there are important differences across academics, and one difference in particular stands out. Academics with high reach (those that are very active on Twitter) but low academic credibility (they are not highly credible academics, as measured by citations) exhibit less toxic expression on Twitter than other academics, particularly those who have low reach but high academic credibility. In their conclusion, Garg and Fetzer focus on this as a problem because:

...those with the greatest public reach may not represent top scholars, potentially distorting public perceptions

However, I see the opposite problem. In terms of tone, the top scholars with the lowest reach have the most toxic expression. Are those the sorts of academics that we want to promote even further on social media? I would suggest not.

What is a better option? First, more highly credible academics should be encouraged to engage in the social media discourse. However, it is important to recognise that credibility alone is not enough. What is needed are credible academics who also model constructive discourse without the toxicity, raising the standard of debate. However, as noted in yesterday's post, many high-quality (especially female) scholars are targets of hostility on social media. These are not separate issues.

Alternatively, we could raise the standard of academic discourse on Twitter more generally, without changing who is represented on the platform. That would reduce the toxic nature of the interactions. Stop laughing! It could happen. The tone and expression of academics on X (or Twitter) matters. Academics can set the standards for everyone else. We don't need to descend into the toxic culture wars that play out each day on social media. We are better than that, and if we show ourselves to be such, maybe more people will listen.

[HT: Marginal Revolution, last year]

*****

[*] I refer to the platform mostly as Twitter, because it didn't change names to X until July 2023, after Garg and Fetzer's dataset ends.

More on the toxic environment in Economics Job Market Rumors

The Economics Job Market Rumors (EJMR) website began as a forum for PhD students to discuss the economics job market, but it has long since become notorious for misogyny, racism, and other toxic behaviour (see this post, for example), due in large part to the anonymous nature of the platform. And even though the user community at EJMR has been called out for their behaviour, it doesn't seem to have gotten much better over time. This is documented by this 2025 article by Florian Ederer (Boston University), Paul Goldsmith-Pinkham, and Kyle Jensen (both Yale University), published in the journal AEA Papers and Proceedings (ungated earlier version here).

Ederer et al. analyse content from EJMR over the period from January 2012 to May 2023, documenting a number of changes. First:

...starting in 2018, EJMR saw an explosion in discussions initiated by references to Twitter posts. This shift mirrors Twitter’s growing importance as a real-time source of information and debate in academic and public policy circles.

Twitter (now X) essentially took over from YouTube as being the source of initial references on EJMR from about 2018, which is about the time of the earlier research on toxicity and misogyny on the platform. There were also surprising declines in Marginal Revolution and NBER links as the starting point for EJMR discussions. Given the predominance of Twitter as a source, Ederer et al. then look in more detail at which Twitter accounts were most referenced, reporting that:

These accounts can be broadly categorized into three main groups: economists, right-wing commentators, and journalists. The group of economists (e.g., Claudia_Sahm, jenniferdoleac, and JustinWolfers) includes academic and professional economists from diverse institutions whose tweets often serve as springboards for debates on research findings, policy implications, and professional conduct. The second group includes polarizing and predominantly conservative commentators and agitators (e.g., realChrisBrunet, RichardHanania, and libsoftiktok) and reflects EJMR’s right-wing slant and engagement with contentious political and social issues. The third group is a collection of news sources and journalistic accounts, many of which have a conservative slant (e.g., visegrad24, disclosetv, and nypost).

Finally, Ederer et al. characterise the posts linking to each Twitter account in terms of 'hate speech', 'negativity', 'misogyny', and 'toxicity' (based on measures from their companion paper here), finding that:

Among the 10 most frequently mentioned Twitter accounts, there are four economists, including three female economists. EJMR posts referencing two of these female economists (Claudia_Sahm and jenniferdoleac) have very high average z-scores of 1.974 and 2.598 for the Misogynistic classifier, indicating that EJMR posters discuss them in strongly misogynistic terms compared to all other Twitter accounts mentioned on EJMR... The only other large average z-score for the Misogynistic measure is for EJMR posts referencing elben (z-score Misogynistic = 0.956), an academic economist who has championed LGBTQ-inclusive policies in the economics profession.

In other words, since 2018 EJMR has remained a hostile and misogynistic platform, with its toxicity increasingly fed by same antagonism and culture-war discourse on Twitter/X. EJMR is not just an academic forum, but has become part of that broader hostile ecosystem.

Economists need places where they can share research in progress, ideas, and practical advice, especially early in their careers. In its early days, EJMR served that purpose. However, it has long since become a space that early career economists are better off avoiding.

[HT: Marginal Revolution, in January last year]

Read more:

• The toxic environment for women in econjobrumors.com

This week in research #119

Here's what caught my eye in research over the past week (another very quiet week, it seems):

• Clemens et al. analyse the effect of California's $20 fast food minimum wage, which was implemented in 2024, and find that food away from home prices increased by 3.3 to 3.6 percent in areas subject to the minimum wage relative to control areas (so firms passed on their cost increase to consumers)

Evidence that artificial intelligence is increasing the impact, but narrowing the scope, of research

There is growing evidence of positive impacts of generative artificial intelligence on productivity. This includes productivity in research (see this post, for example), including my own. However, some have questioned whether increasing research productivity comes at a cost of narrowing the scope of research.

So, I was interested to read this article by Qianyue Hao (Tsinghua University) and co-authors, published in the prestigious journal Nature (ungated earlier version here) late last year. They look at the impact of AI tools (not limited to generative AI) on the productivity of researchers and the quality of research. Specifically, they look at authors publishing in six representative fields: biology, medicine, chemistry, physics, materials science, and geology, across three 'eras': (1) the 'machine learning era ' (from 1980 to 2014), the 'deep learning era' (from 2015 to 2022), and the 'generative AI era' (from 2023 onwards). Hao et al. compare authors who publish 'AI augmented papers' with those who do not. An 'AI augmented paper' is one that uses methods such as:

...support vector machines and principal component analysis from the machine learning era, and convolutional neural networks and generative adversarial networks from the deep learning era. Large language models, which have emerged in recent years, also rank among the most frequently used methods...

Using a dataset that includes over 27 million papers with complete records that were published between 1980 and 2025, of which about 310,000 were 'AI augmented', Hao et al. find that:

...annual citations to AI papers are 98.70% higher than those to non-AI papers on average...

So, AI augmented research gathers more citations, which suggests that authors using AI in their research achieve greater impact. This is reinforced by evidence that AI augmented papers are published in higher quality journals (with Q1 journals being the highest ranked). Hao et al. report that:

...the proportion of AI papers in Q1 journals is 18.60% higher than that of non-AI papers in all journals; in Q2 journals, the AI proportion is 1.59% higher; whereas Q3 and Q4 journals hold a relatively lower proportion of papers with AI... These results indicate a heterogeneous distribution of AI-augmented papers across journals, with a higher prevalence in high-impact journals.

And AI appears to make authors more productive, as:

On average, researchers adopting AI annually publish 3.02 times more papers... and garner 4.84 times more citations... than those not adopting AI, with consistency.

All of these results seem to hold across all of the disciplines that Hao et al. consider. However, it is not all good news. Hao et al. use machine learning to create a measure of the 'breadth of scholarly attention'. Using that measure, they find that:

Compared with conventional research, AI research is associated with a 4.63% contracted median collective knowledge extent across science, which is consistent across all six disciplines... Moreover, when dividing these disciplines into more than two hundred sub-fields, the contraction of knowledge extent can be observed in more than 70% of them...

Of course, some of the differences here may be due to selection, as the types of researchers, and the types of research, involving AI use may be meaningfully different from those that don't. However, putting the selection issues aside, Hao et al. note that there is a tension between the individual researcher's incentive to produce a greater quantity of research that has higher impact, which would suggest greater use of AI, and the social incentive to produce a greater breadth of research.

So, the takeaway from this paper is that we need to consider researcher incentives, not just productivity. Specifically, this research suggests that the use of AI in research is leading to a 'prisoners' dilemma' outcome: each individual researcher acting in their own best interests (and using AI in their research) leads to an outcome that is worse for society overall (less breadth of research and more incremental gains).

Hao et al. conclude that:

The substantial academic benefits of AI use may be a driving force behind its accelerated rate of adoption; however, we also find unintended consequences from the increased prevalence of AI-augmented research. In all fields, AI-augmented research focuses on a narrower scope of scientific topics and reduces the scientific engagement of follow-on research, leading to more overlapping research work that slows the expansion of knowledge. Further, with a greater concentration of collective attention to the same AI papers, the adoption of AI seems to induce authors to converge on the same solutions to known problems rather than create new ones.

So, what is the solution here? Society probably wants research to be higher quality and have a broad scope. But individual researchers' incentives to use AI in their research appears inconsistent with that outcome. The traditional prisoners' dilemma is a repeated game (see here or here, for example), and the players of that game can avoid the worst outcome by cooperating. In this case, the researchers could cooperate by agreeing not to use AI in their research. The problem is that every researcher has an incentive to cheat on that agreement, since if they use AI, then that will be good for their career. This prisoners' dilemma is more difficult to ensure cooperation in than the traditional game, because there are not just two players who need to cooperate, but thousands (or millions). Ensuring cooperation in a prisoners' dilemma game with many players, each of whom is far better off cheating than cooperating, is almost impossible (which is why solving the problem of climate change is so difficult).

My own view is that the answer is not to keep AI out of research. That is not realistic, in the same way that it's not realistic to expect students not to use generative AI. The incentives need to be redesigned, but this will be no easy task. As long as universities, research funders, and publishers reward researchers for quantity, citations, and publication in top-ranked outlets, then we should expect more AI-augmented work, with a narrower scope than society might prefer. If we want AI to expand knowledge rather than simply accelerate competition within narrow foci, then we need institutions that also reward novelty, breadth, and the discovery of new questions. That is the economic challenge we must face up to.

[HT: Marginal Revolution]

The relationship between obesity of politicians and corruption is correlation, not causation

Not every correlation between two variables represents a causal relationship. Even if we can tell a compelling story about why a change in one variable might cause a change in another, that doesn't make the relationship causal. Sometimes a correlation actually results from something other than the story you tell. Sometimes the correlation is just random noise (a spurious correlation). So, we should be cautious when interpreting correlations.

I was reminded of this when reading this 2021 article by Pavlo Blavatskyy (University of Montpellier), published in the journal Economics of Transition and Institutional Change (sorry, I don't see an ungated version online). The article even generated a small debate, with a comment by György Márk Kis, and then a reply by Blavatskyy, appearing in the same issue of the journal.

In the original article, Blavatskyy looks at the relationship between the body mass index (BMI) of politicians in a country and the Corruption Perceptions Index by Transparency International. The data Blavatskyy uses is for 2017, and the sample of countries is limited to 15 post-Soviet countries (Armenia, Azerbaijan, Belarus, Estonia, Georgia, Kazakhstan, Kyrgyzstan, Latvia, Lithuania, Moldova, Russia, Tajikistan, Turkmenistan, Ukraine, and Uzbekistan). The argument for why this correlation matters is explained in Blavatskyy's reply to Kis:

One common form of corruption/lobbying is inviting governmental officials to lavish banquets with excessive consumption of food and drinks... Corrupt politicians frequenting such banquets might risk gaining extra weight. This ‘hedonic theory of corruption’ postulates the existence of a positive relationship between median body mass index of public officials and the level of grand political corruption in society.

So, Blavatskyy is able to tell a good story for why greater corruption would cause higher BMI among politicians. However, that doesn't mean that the relationship is causal. Even though the correlation between perceived corruption and median politician BMI is clear, from Figure 1 in the original paper:

Low numbers in the Corruption Perceptions Index represent higher levels of perceived corruption. So, this figure shows that countries where the politicians have higher median politician BMI have higher levels of perceived corruption.

Kis took issue with a number of things in the paper. First, why those 15 countries? Why not all countries? Kis shows that if you separate the 15 countries in Blavatskyy's sample by their geographic location, you get different relationships within each subsample. However, the broader question is not what happens when you look at subsamples, but does this relationship hold if you add more countries to the sample? Neither Blavatskyy nor Kis answer that question. We should also wonder whether there is something special about 2017 that leads to this correlation. Does it hold in other years?

In his reply, Blavatskyy doesn't really address those two points (narrow sample, and a single year) in a convincing way. Instead, he narrows the sample even further to look at changes in politician BMI and perceived corruption for just one of the countries in his sample, Ukraine. In that analysis, he again shows a correlation between corruption perceptions and politician BMI, in this case over time for Ukraine. However, that simply raises the question of: why Ukraine? Why didn't he look at other countries in his sample in that way? And just because Ukraine shows a correlation over time, that still doesn't demonstrate a causal relationship.

Kis also takes issue with the machine learning algorithm that Blavatskyy uses to estimate the BMI for politicians in his sample. Kis notes that the accuracy of the algorithm is quite dubious (my words, not Kis's), with:

...errors of at least 5.5 in 21.1% of the time.

That's an error in the estimated BMI of 5.5 in over 20 percent of cases. That extent of measurement error would be problematic. To that, I would add that it is unclear whether the training sample that the machine learning algorithm was trained on included people from post-Soviet countries. The relationship between facial features and BMI could well be ethnic-specific, in ways that systematically bias the results. We have no way of knowing. And Blavatskyy didn't address this point in his reply.

Now, the point of this post is to focus on correlation or causation. From what I have seen, this seems a likely candidate for confounding. There are any number of variables that might increase politician BMI and increase corruption, without corruption being a cause of higher politician BMI. As one example, a country with high inequality might simultaneously have high corruption (with petty officials willing to take bribes to supplement their low incomes) and high politician BMI (since politicians would likely be among the wealthy class in society). Blavatskyy doesn't consider confounding variables such as inequality, or differences in age distribution, or differences in average BMI in the population, or regional differences in diet, in his analysis.

Now, to be fair to Blavatskyy, he doesn't adopt a causal interpretation of his results (except in his response to Kis, as I quoted above). Instead, Blavatskyy argues that, if BMI and perceived corruption are correlated, then we might infer how much corruption is being experienced in a country by looking at the median BMI of its politicians. However, even that inference is problematic, and Blavatskyy should know why. He gives the example of Swiss watches in China as a proxy for corruption, but then notes that:

...the rise of social media and Internet anti-corruption platforms in 2011–2012 made it no longer possible to measure grand political corruption through visible luxury Swiss watches. Luxury Swiss watches could still be a popular expenditure of corrupt governmental officials, but these officials are now more careful not to reveal their Swiss watches to the general public.

When politicians realised that their Swiss watches were giving away their corruption, they stopped showing off their Swiss watches. If politicians realised that their expanding waistlines were giving away their corruption, wouldn't they invest more in personal trainers (or liposuction)? As soon as this correlation was used for inference, the correlation would likely start to break down. This again illustrates the limited usefulness of such proxies.

Correlation does not imply causation. And sometimes, correlation today does not imply correlation in the future. We need to be much more cautious when considering analyses like this one.

The impact of Taylor Swift on the Kansas City Chiefs' TV ratings

In 2023, Taylor Swift began a relationship with Kansas City Chiefs tight end Travis Kelce. After that, Kansas City Chiefs broadcasts seemed increasingly eager to cut to shots of Taylor Swift in the corporate boxes, rather than fans in the stands. The NFL was clearly trying to appeal to Swift's fans, but did it work? In a new article published in the Journal of Sports Economics (sorry, I don't see an ungated version online), Kerianne Rubenstein (Syracuse University) and Frank Stephenson (Berry College) show that it did.

Rubenstein and Stephenson collated data on 247 NFL games played in the 2022 and 2023 seasons, noting that the first Chiefs game that Taylor Swift attended was in the third week of the 2023 season. They apply a difference-in-differences analysis, comparing the difference in TV ratings between before and after Week 3 of 2023 for the Chiefs, with before and after Week 3 of 2023 for other teams, while controlling for other variables expected to affect TV ratings. In other words, Rubenstein and Stephenson check whether the Chiefs' TV ratings increased by more than the average before-and-after change that other teams experienced. They find that:

...Chiefs’ games after Taylor Swift started attending see an increase of 2.15 ratings points, which is an approximately 32% increase relative to the mean Nielsen rating... total viewership increased by about 4.8 million after Taylor Swift started attending Chiefs’ games.

So, it appears that Taylor Swift did increase TV ratings for the Kansas City Chiefs. Good news for the Chiefs (and for other NFL teams, who share in the broadcast revenue). Interestingly, and to be expected given Swift's young fan base, the effect was even larger on TV viewership among those aged 18-34 years, with a 40.1 percent increase in TV rating.

An important question, though, is whether Swift attracted new fans, and whether they stuck around. In terms of the former, Rubenstein and Stephenson find some evidence that games played at the same time as Chiefs games suffered a decrease in TV ratings (although that analysis is based on a sample of only ten games, which limits how much we can take from it). However, they also find an increase in TV ratings when the Chiefs game was the only game in its timeslot. So, while there was some substitution between NFL games, new fans were also attracted to watch. And, they did stick around - Rubenstein and Stephenson find limited evidence that the effect declined over time, with Chiefs games later in 2023 having a similar TV rating as those earlier in the season (it is worth noting that the Chiefs had a particularly good 2023 season though, finishing the regular season 11-6, winning their division, and ultimately winning Super Bowl LVIII).

Celebrities are a common feature of sports games. Rubenstein and Stephenson note the example of the Atlanta Hawks, who make courtside seats available to celebrities with large social media followings in the hopes of increasing game attendance and TV ratings. Not every celebrity has the profile of Taylor Swift. However, the results in this study suggest that the Hawks' strategy might be a sensible strategy for increasing the profile of games. The NFL should take notice. Certainly, this would make much more sense than, as some conspiracy theorists would have you believe, biasing the officiating in favour of particular teams (like the Chiefs). So, leaving conspiracies aside, what we learn from this paper is that celebrity appearances at games can increase demand. That seems to be exactly what happened here, with Taylor Swift’s presence helping to increase the audience for Kansas City Chiefs games.

This week in research #118

Here's what caught my eye in research over the past week (a quiet week, following last week's bumper edition):

• Rubenstein and Stephenson assess the effect of Taylor Swift’s relationship with Travis Kelce on the Kansas City Chiefs’ television audience, and find that viewership increases by about one-third beginning with Swift’s first time attending a Chiefs’ game
• Bussoli and Fattobene (open access) find that Financial Graph Literacy is lower among older adults, those with less education, and lower-income groups, and is significantly associated with a greater likelihood of engaging in proactive financial behaviours such as saving, investing, budgeting, and using digital financial tools

How the 'travelling Pope' affected international trade

Pope John Paul II was known as 'the travelling Pope' because of the large number of international trips ('pastoral visits') he undertook (more than 100 during his reign from 1979 to 2004). He also had a huge following, as you might expect as the leader of the Catholic Church, but the advent of television meant that the public could follow his travels in a much closer way than ever before. And, through his pastoral visits and his following, he exposed Catholics the world over to new places they would otherwise not have seen or, in some cases, even heard of. What effects did that exposure have?

That is essentially the question addressed in this recent article by Alexander Popov (European Central Bank), published in the Economic Journal (ungated earlier version here). Popov focuses on the impact of the Pope's visits on exports from the visited country, and especially exports to Catholic countries. He employs an event study design - looking at how exports changed between the time before and the time after the Pope's first visit to a country, while controlling for GDP growth, population, the US dollar real exchange rate, and the extent of trade liberalisation and democracy. The key results are summarised in Figure 2(a) from the paper:

The figure shows how exports evolve before and after the Pope's visit. Beforehand, there isn't much evidence of a trend (notice that the red line hovers around zero). However, after the Pope's visit, exports increase (the red line is clearly above zero and trending upwards), and the effect is substantial. Popov notes that:

...the point estimate on Year 3 after the pope’s visit to a country is 0.1152, which implies that exports to the rest of the world are higher by 12.2%, relative to the year of the visit.

And the effects are even larger for exports to countries with larger Catholic populations. Specifically:

...exports to a trading partner with 54.3% (75th percentile), relative to a trading partner with 1.1% (25th percentile) Catholics in the population were higher by between 16.5% and 36.9% during years 1 to 5 after a visit by the pope.

Clearly, Catholics were paying attention to where the Pope was visiting. Popov then asks the obvious question: what explains this effect? He examines three hypotheses:

The first one is that during a foreign visit, the pope explicitly encourages Catholics around the world to engage with the host country on economic terms. I analyse 633 speeches given during the pope’s 130 first visits and I find rare occasions when he mentions words like ‘trade’, ‘economic’ or ‘globalisation’.

So, the Pope wasn't explicitly telling Catholics to buy more goods from the countries he was visiting. Then:

The second hypothesis is that, by simply visiting a country, the pope raises its profile, or ‘puts it on the map’ for the global Catholic family, especially if Catholics around the world are for cultural or economic reasons less connected with the visited country. I find that the effect on exports of a pastoral visit to a country is stronger if this country is relatively poor and if it has relatively fewer Catholics and relatively weaker bilateral trade links with the partner country. The third hypothesis is that Catholics around the world are simply buying souvenirs to commemorate the pope’s visit. I analyse data on bilateral trade at the product level, for ten different sectors, and I find that after a pastoral visit, the increase in exports I detect takes place in half of them.

So, the third hypothesis (souvenirs) doesn't have much support. Popov concludes that the second hypothesis shows the likely driver of the increase in exports. This evidence is consistent with the Pope raising the profile of the countries he visited, and those countries benefiting from their higher profile among Catholics in the form of higher exports, especially to Catholic countries.

What makes this paper interesting in an economic sense is that it suggests trade flows don't just depend on prices, trade policy, and distance. They also depend on visibility, familiarity, and the ways that cultural influence can affect economic outcomes. Pope John Paul II's visits appear to have increased visibility and familiarity, which may in turn have boosted trade. The 'travelling Pope' may have also been the 'trade-promoting Pope'.

Seven decades of change in the demographics and research styles of top economics research

Back in 2013, Daniel Hamermesh (University of Texas at Austin) published this article in the Journal of Economic Literature (ungated earlier version here), which summarised changes in the demographics and research styles of top economics research, based on articles published between 1963 and 2011 in three top journals: the American Economic Review (AER), the Journal of Political Economy (JPE), and the Quarterly Journal of Economics (QJE). A new update last year (open access) from Hamermesh extends the analysis to include articles up to 2024.

In terms of demographics, the trends show a continuation and in terms of gender, Hamermesh notes that:

The progression that occurred from the 1960s and 1970s, when only a minute fraction of authors were women, to the early twenty-first century has, if anything, accelerated.

This will be welcome news, given the persistent gender gap in economics (see this post and the links at the end of it). It likely reflects the changing demographics of young economists, with a growing proportion of the young 'stars' in economics being women (and noting that it is young stars who often get published in the top journals that Hamermesh is considering).

In terms of the age structure of authors, Hamermesh reports that:

The changes from 2011 to 2024 continued those that started in the 1980s, but the rate of change has not accelerated. Indeed, most noticeable from 2011 to 2024 was a continuing sharp and statistically significant drop in the representation of the youngest group (and a nearly equal sharp rise among those 36–50)...

...the average age of authorship has increased steadily since 1973. 

Can I change my comment above about the young stars in economics? The increasing median age of authors in top journals seems to be a general trend across academia. Hamermesh then turns to research 'style', documenting a continued dramatic rise in the proportion of articles in those journals that are co-authored:

There were no four-authored papers as recently as 1983; today they account for 17 percent of articles. There were no papers with more than four authors in 2003; today nearly 12 percent of articles have five or more authors (with five articles written by six authors each and one by seven authors). Obversely, sole-authored papers are now quite scarce; and even two-authored papers today only account for slightly more than one-fourth of all articles (compared to a majority as recently as 2003).

Unsurprisingly, the increase in the number of co-authored articles means that the age diversity of author collaborations has increased over time as well. In terms of the types of research, he reports that:

The big changes are the continuing rise in empirical work based on original non-laboratory data and the rapid and even accelerating increase in experimental work. Today these two methods, which both involve collecting original data, account for over half of all published papers, compared to less than 4 percent four decades ago...

These trends are not all unrelated, of course. Experimental research, and the increasing use of large datasets, typically both require larger research teams. They also often require more detailed methods, which may involve both larger teams, and more experienced researchers. Larger teams might be more likely to include female team members. And larger teams often need someone to lead and coordinate all of the team members, and those leaders tend to be more experienced (and older) academics. So, it would not surprise me, if more detailed analysis was conducted, to see that the trends are interconnected.

Now, the interesting thing will be what happens going forward, given the increasing use of generative AI in research (see here, for example). Since generative AI can now do a lot of the work that research assistants and early career researchers previously did, will the trend towards larger research teams be reversed? How will that interact with the gender gap in research (given that the age of female economists skews younger at the moment). And how will it affect the age distribution of researchers (given that men, and younger people, are somewhat more likely to use generative AI). I'll be looking forward to Hamermesh's next update. Hopefully, we don't have to wait another 12 years.

[HT: Marginal Revolution, last year]

Changing their minds could be a good thing for economists

People don't like to change their minds. This may partly be an expression of loss aversion - we really want to avoid losses, including the loss of an idea that we previously thought was true. This leads to status quo bias - we prefer not to change things, and keep them the same, because changing things entails a loss. But what if changing our minds could make us better off? Would we be so reluctant to do so?

This 2025 paper by Matt Knepper (University of Georgia) and Brian Wheaton (UCLA) suggests that economists, at least, should not be afraid to change their minds, because doing so increases the number of citations to their research. Knepper and Wheaton investigate authors who undergo an 'ideological reversal' - previously publishing research that could be considered right-wing, before switching and publishing a paper that draws a left-wing-consistent conclusion, or the reverse (switching from left-wing to right-wing). Their main data source is every economics paper ever published in the top 100 economics journals indexed in Web of Science - some 200,000 articles. They also have a narrower dataset of papers referenced in meta-analyses on policy topics, including:

...the minimum wage, the economics of unions, the taxable income elasticity, the fiscal multiplier, intergenerational transfers, trade and productivity, trade and domestic employment, crowd-out, the gender wage gap, unemployment insurance, disability benefits, universal preschool, childcare and employment, immigration and wages, and more.

Knepper and Wheaton use this narrower dataset to train a machine learning model to categorise the rest of the papers in the dataset, as to how left-wing (or right-wing) the conclusions are. For instance, a paper that concludes that the minimum wage reduces employment is more right-wing, whereas one that concludes that there is no disemployment effect of the minimum wage is more left-wing. Knepper and Wheaton define an author as left-wing if they published more left-wing papers than right-wing ones over the previous five years, and the reverse for right-wing authors. They then use the larger dataset to investigate what happens to each economist who undergoes an 'ideological reversal'. They first outline some descriptive facts based on their dataset, including:

• Fact #1: The typical author mostly publishes results on one side of the political spectrum.

• Fact #2: Ideological reversals are not rare; they occur at least once for 40% of authors.

• Fact #3: Ideological reversals become much more common later in an author’s career, with authors essentially never undergoing a reversal in the first decade of their career.

• Fact #4: Most ideological reversals do not represent a permanent defection to the other side of the political spectrum, but rather the beginning of repeatedly publishing results on both sides of the spectrum.

• Fact #5: Ideological reversals occur much more frequently amongst authors who are (initially) classified as right-wing.

That does seem like a surprisingly high proportion of economists who undergo at least one ideological reversal. However, perhaps we should take comfort in that - if the results point in a particular direction, our conclusions should say that, even if that conclusion is inconsistent with our previous conclusions on the same topic.

Do these ideological reversals matter though? Knepper and Wheaton employ a difference-in-differences analysis, comparing the difference in citations (and other metrics) between authors who did, and did not, undergo an ideological reversal, between the time before, and after, the reversal occurred. In other words, they look at whether citation counts rise more for economists who have an ideological reversal than for otherwise similar economists who do not. The results are striking, with:

...a sharp clear increase in citation count following an ideological reversal with essentially no evidence of pre-trends... The citation boost accumulates to approximately 9 over a one-decade period and 30 over a two-decade period.

The results remain consistent when Knepper and Wheaton limit the analysis to papers published before the ideological reversal, and when they limit the analysis to papers in the meta-analysis only (showing that the machine learning approach doesn't drive the results). Knepper and Wheaton also find evidence consistent with no change in the quality of papers before and after the ideological reversal, and that:

Both left-to-right and right-to-left reversals are rewarded by increased citations of roughly the same magnitude. The boost in citations received subsequent to a left-to-right reversal is mostly driven by citations from right-wing authors, and the boost in citations received subsequent to a right-to-left reversal is mostly driven by citations from left-wing authors. Encouragingly, however, the new right-wing (left-wing) audience garnered by a left-to-right (right-to-left) reversal... also engages with and cites the author's previous left-wing (right-wing) papers. This dynamic suggests that ideological reversals help prevent the formation of echo chambers in economics academia and expose authors to opposite ideological findings.

This last result is particularly important, and I believe it allows us to conclude that economists need not fear ideological reversals. In doing so, they can attract a new audience from the other side of the ideological spectrum, bringing the two sides closer together. Hopefully through that, we end up with higher-quality research overall.

[HT: Marginal Revolution, last year]

Artificial intelligence and the 'age of leisure'

My ECONS101 class covered constrained optimisation last week, and one of the models we looked at was the labour-leisure trade-off for workers. Now artificial intelligence, and in particular generative AI, is likely to have large impacts on the labour-leisure trade-off. As the Financial Times reported last year (paywalled):

The idea that technological progress can enable people to work fewer hours is not outlandish...

But in order to believe a similar trend is going to take hold again, you have to assume three things. First: that AI will deliver a substantial boost to economic productivity...

Second, you have to assume the economic gains will be widely distributed...

Third, you have to believe workers will “cash in” those proceeds in the form of extra leisure, rather than higher income. But will they? In many developed countries, there has been a slowdown in the reduction in working hours in recent decades...

Far from trading income for leisure, it is the people with the highest salaries who tend to work the longest hours.

Will workers trade off higher productivity for more leisure time? Are we about to enter an 'age of leisure'? The constrained optimisation model for the worker (see also this post) can help us clarify the possibilities. In this model, we'll assume that AI increases productivity, and that the increase in productivity is represented by higher wages for workers. [*] The model will then tell us whether workers might respond by consuming more, or less, leisure.

Our model of the worker's decision is outlined in the diagram below. The worker's decision is constrained by the amount of discretionary time available to them. Let's call this their time endowment, E. If they spent every hour of discretionary time on leisure, they would have E hours of leisure, but zero income. That is one end point of the worker's budget constraint, on the x-axis. The x-axis measures leisure time from left to right, but that means that it also measures work time (from right to left, because each one hour less leisure means one hour more of work). The difference between E and the number of leisure hours is the number of work hours. Next, if the worker spent every hour working, they would have zero leisure, but would have an income equal to W0*E (the wage, W0, multiplied by the whole time endowment, E). That is the other end point of the worker's budget constraint, on the y-axis. The worker's budget constraint joins up those two points, and has a slope that is equal to the wage (more correctly, it is equal to -W0, and it is negative because the budget constraint is downward sloping). The slope of the budget constraint represents the opportunity cost of leisure. Every hour the worker spends on leisure, they give up the wage of W0. Now, we represent the worker's preferences over leisure and consumption by indifference curves. The worker is trying to maximise their utility, which means that they are trying to get to the highest possible indifference curve that they can, while remaining within their budget constraint. The highest indifference curve they can reach on our diagram is I0. The worker's optimum is the bundle of leisure and consumption where their highest indifference curve meets the budget constraint. This is the bundle A, which contains leisure of L0 (and work hours equal to [E-L0]), and consumption of C0.

Now, let's say that the situation shown above is the situation before the advent of AI. After AI is introduced, productivity increases, and so wages increase (from W0 to W1). This causes the budget constraint to pivot outwards and become steeper (since the slope of the budget constraint is equal to the wage, the slope has increased from -W0 to -W1). The worker can now reach a higher indifference curve, and it is the position of that higher indifference curve that determines the worker's response in terms of whether they consume more leisure or not. If they move to the higher indifference curve I1, then the worker's new optimum is the bundle of leisure and consumption B, which contains leisure of L1 (and work hours equal to [E-L1]), and consumption of C1. For this worker (whose response is shown in red on the diagram), leisure hours decrease as a result of the higher wage. On the other hand, if they move to the higher indifference curve I2, then the worker's new optimum is the bundle of leisure and consumption C, which contains leisure of L2 (and work hours equal to [E-L2]), and consumption of C2. For this worker (whose response is shown in blue on the diagram), leisure hours increase as a result of the higher wage. [**]

Either of these possibilities could happen. In fact, both could happen, with some workers increasing leisure time and others decreasing leisure time. By itself, this model doesn't answer the question of what will happen, but shows that both increased leisure and decreased leisure are possible outcomes.

The key difference here comes down to the size of the income effect of the increase in wages. When wages increase, the opportunity cost of leisure increases. That makes leisure relatively more expensive, and workers should respond by consuming less leisure. That is what we call the substitution effect - workers substitute away from leisure as it becomes more expensive. However, increased wages also lead to an income effect. Leisure is a normal good, which means that as the worker's income increases, they would like to consume more leisure. Notice that the substitution effect and the income effect are working in opposite directions here. For workers who overall decrease their leisure, the substitution effect (which says they should consume less leisure) must be bigger than the income effect (which says they should consume more leisure). For workers who overall increase their leisure, the reverse is true - the substitution effect must be smaller than the income effect.

AI may lead us into an age of leisure. But only if productivity gains lead to higher wages, and the income effect of higher wages more than offsets the substitution effect.

*****

[*] The assumption that productivity gains will lead to higher wages is a strong assumption. Indeed, the FT article questions whether this assumption is valid. If productivity gains don't lead to higher wages, then this model doesn't help us evaluate whether we're about to move into an 'age of leisure', and the impacts might be more macroeconomic than microeconomic. That is, we may end up with leisure, but arising through weaker labour demand, reduced hours, or unemployment rather than through workers voluntarily choosing more leisure as wages increase.

[**] Notice that the indifference curves I1 and I2 are crossing, and indifference curves cannot cross. However, those two indifference curves are for different workers, so there is no problem. I could easily have drawn two different diagrams, one for each worker, but I've kept them both on the same diagram for efficiency.

This week in research #117

Here's what caught my eye in research over the past week:

• Zhang et al. find that Uber’s entry into a US city significantly reduces crime rates, with larger effects in areas facing greater liquidity constraints (less bank credit supply, fewer local job opportunities, higher personal bankruptcy risk, and greater household financial stress)
• Sandorf and Navrud (open access) establish convergent validity between a contingent valuation survey and a discrete choice experiment (meaning that both measures are highly correlated), with the example they use being willingness-to-pay to reduce the spread of invasive crabs in Norway
• Desierto and Koyama (with ungated earlier version here) explain the economics of medieval castles in Europe
• Ordali and Rapallini (with ungated earlier version here) conduct a meta-analysis of the relationship between age and risk aversion, and confirm that there is a positive relationship in studies using survey data and lotteries
• Singh and Mukherjee conduct a replication of an earlier study that established 'action bias' among goalkeepers facing a penalty kick, and find that jumping left or right rather than staying in the centre of the goal is not a sub-optimal action for goalkeepers in FIFA World Cup matches, and so the high frequency of jumping is not indicative of action bias (it is good to see a replication study published in a good journal)
• Lindkvist et al. (open access) investigate attitudes toward research misconduct and questionable research practices among researchers and ethics reviewers across academic fields, and find that researchers and ethics reviewers in medicine, as well as more senior and female researchers and reviewers, took a more negative view of questionable research practices
• Lei et al. use China’s Compulsory Schooling Law as a quasi-natural experiment to investigate the effect of education on HIV/AIDS, finding that mass education significantly enhances knowledge about HIV/AIDS, and that each additional year of exposure to the law reduces HIV/AIDS and mortality rates by 6.51 percent and 2.15 percent respectively
• Daoud, Conlin, and Jerzak (open access) study the differential effects of World Bank and Chinese development projects in Africa between 2002 and 2013, using data across 9899 neighbourhoods in 36 African countries, and find that both donors raise wealth, with larger and more consistent gains for Chinese development projects
• Stoelinga and Tähtinen (open access) find that conflict exposure, on average, increases support for democracy in African countries, but the effects vary by ethnicity and regime type, but interestingly, violence increases trust in ruling institutions in autocratic regimes
• Ruiz et al. (with ungated earlier version here) find that, following the exodus of Cuban doctors from Brazil in 2018, the reduction in doctors was associated with persistent reductions in the care of chronic diseases, while service utilization for conditions requiring immediate care, such as maternal-related services and infections, quickly recovered
• Geddes and Holz (open access) investigate the effect of rent control on domestic violence in San Francisco, and find that there was a nearly 10 percent decrease in assaults on women for the average ZIP code (some good news for advocates of rent control, but it hardly offsets the bad outcomes)
• Clemens and Strain (with ungated earlier version here) add further to the literature on the disemployment effects of minimum wages, this time looking at the difference between large and small minimum wage changes, finding that relatively large minimum wage increases reduced usual hours worked per week among individuals with low levels of experience and education by just under one hour per week during the decade prior to the onset of the Covid-19 pandemic, while the effects of smaller minimum wage increases are economically and statistically indistinguishable from zero

Anticipating higher future petrol prices, consumers actually push up petrol prices now

In his 1984 book The Evolution of Cooperation, Robert Axelrod suggested that people cooperate in repeated games because of 'the shadow of the future'. They alter their behaviour by cooperating now, because they anticipate that will lead to greater gains for them in the future. I really like this analogy of the shadow of the future affecting our decisions now, and not just in the context of game theory and repeated games. In fact, we've seen it play out in a different context this past week, as reported by the New Zealand Herald:

Kiwis are rushing to fill up their cars across the country amid fears of price increases at the pump because of escalating conflict in the Middle East.

Video sent to the Herald of Waitomo Tinakori petrol station in Wellington today showed a queue of cars waiting for fuel, with vehicles spilling out on to the road.

Waitomo Group CEO Simon Parham said there has been a similar increase in demand at stations across the country, with sales increasing by 10-15% this week.

“People are filling up and filling their cars ahead of the price increase that will flow through the market over the coming weeks because of the Iran conflict,” he said.

To see what is going on here, let's consider the retail market for petrol, as shown in the diagram below. Before the current conflict in the Middle East, the equilibrium price of petrol was P0, and Q0 petrol was traded per week. Then the conflict begins. Consumers anticipate that the price of petrol will increase in the future, so they decide to fill up their vehicles now. That increases the demand for petrol from D0 to D1. The equilibrium price of petrol increases to P1, and there is Q1 petrol traded in the week. 

Notice that by trying to avoid the high petrol price in the future, the consumers cause the price to rise today, which is exactly the outcome they were trying to avoid! In effect, when consumers rush to fill up early, they bring some of the future price pressure forward into the present. Expectations about future prices can cause self-fulfilling prophecies like this, which is a point I will make in my ECONS101 class in several weeks, when we talk about financial markets (where self-fulfilling prophecies are a clear and present danger at all times). The shadow of the future matters - consumers' actions based on trying to avoid future price rises make those price rises happen now instead.