Here is the abstract:

The industrialization of global food systems has led to dietary changes that harm both health and the environment. If global food systems are to meet the needs of a growing population for healthy, environmentally sustainable, and affordable diets, substantial changes will be required. In this Review, we synthesize growing empirical evidence on the complexity of factors that influence consumer dietary and farmer production choices, especially the roles of public and private entities that shape food environments. We outline promising interventions to help facilitate beneficial global dietary transitions, including research and development for product innovation, regulation of food environments, and food assistance and food-as-medicine programs. Understanding and aligning the motives and incentives of various food system actors is essential to achieve improved health, environment, and equity outcomes.

In the article, we steer clear from a lot of could-should-would statements in the extant literature to consider the question of how we can ensure a transition toward sustainable, healthy, and equitable diets (SHE diets) for countries at all levels of income. What I like about the piece is that we take seriously the idea that midstream actors in the food value chain (e.g., processors, distributors, wholesalers, retailers) have an important role to play in getting use to SHE diets, and we have a cool infographic to go along with that idea. The editor, Bianca Lopez, summarizes our work as follows (the emphasis is mine)

Our current food systems are major sources of pollution, greenhouse gases, and land-use change while also struggling to provide adequate nutrition equitably to more than 8 billion people. As development and incomes increase, people are shifting their diets toward more meat and processed foods, with negative effects on both human health and the environment. Yang et al. synthesized research on leverage points for more sustainable and healthy dietary transitions. They highlight the outsized role of midstream actors, including manufacturers, retailers, and restaurants, in influencing consumer choice and farmer actions. A combination of research and development, regulation, education, and public assistance could help make healthy and sustainable foods more tasty, available, and affordable.

Of course, any progress on getting to SHE diets will necessary require a better understanding of the black box that is the midstream of agri-food value chains, both from an econometric perspective, but also from a data availability perspective. Both will be the focus of my 2026-2027 sabbatical.

It is nice to be able to share this manuscript, which started with the observation that criticisms on efficiency grounds of policies like farm subsidies, while correct within the narrow framework of the textbook model, broadly tend to miss the mark because they miss an important externality, viz. that being able to protect its citizens from hunger in times of conflict is important for a state. That is, when economists tend to criticize such policies (and, more broadly, agricultural protection), they usually suffer from tunnel vision in that they tend to ignore what the textbook model assumes away, and end up worshipping the symbol (i.e., the model) for the thing symbolized (i.e., reality).

Bernhard and I had been thinking and talking about this for quite a while (if I recall correctly, since he was a postdoc here, sparked by the Russian attack on Ukraine), and recent events (e.g., Canadian PM Mark Carney’s speech at the World Economic Forum, the war in Iran) have made this even more relevant than we thought it was going to be. This is also timely because geoconomics (i.e., the study of the use of economic policy to pursue strategic aims) is hitting the big time; Mohr and Trebesch had an overview of the topic in last year’s Annual Review of Economics).

Here is the abstract:

Food has long served as an instrument of statecraft. Yet agricultural economics typically analyzes policy through the lens of consumer and producer welfare, which neglects security externalities. We review the literature at the intersection of agricultural economics and political science, examining how food systems both shape and are shaped by geopolitical forces through the two channels of (i) domestic instability with international spillovers, and (ii) the deliberate use of food in statecraft. Our synthesis of key findings in the literature suggests that (i) food prices relate to riots and instability while ambiguously relating to civil war and violence, (ii) the wide geographic spread of agriculture limits but does not eliminate unilateral coercive leverage compared with other strategic sectors, and (iii) domestic food policies are strongly related to national security goals. Moreover, we point out the many extant data sets one can use to do work in this area and identify several research gaps in the literature. As global uncertainty intensifies, integrating geopolitical analysis into agricultural economics is essential for policy relevance.

As the abstract indicates, our aim was to identify a number of open questions in what we view as an area of research that is too important to ignore. As such, we hope this review launches a thousand ships.

EDIT: Here is a video I had NotebookLM make for policy makers and stakeholders interested in the topic, but who do not have the time or patience to read a 53-page working paper. The prompt was: “Prepare a version of this paper for policy makers and other stakeholders (e.g., people at nonprofits, or in international agencies).”

And here is one Bernhard had NotebookLM make, but using a more general prompt with a view to summarizing the paper:

I forget when Dan and I first discussed this, but this paper was born out of the two of us connecting and bonding over social media during the pandemic. Since about 2014, I had had in mind the idea that if you have, say, 10 years worth of longitudinal data on workers, why was the default way to deal with that to use one individual fixed effect per worker? Why not two fixed effects per worker–one for years 1 to 5, and one for years 6 to 10? Why not more than two? What is the optimal number of fixed effects per unit of observation when you have a “long” panel? At some point, Dan and I discussed this and realized we had both been thinking about the same thing, and so we set to work on this paper. (It’s really Dan’s paper, I was just along for the ride. And while Dan likes to joke that he’s not really an econometrician, he just plays one on TV, I think that’s just humility speaking.)

This is especially important for two reasons. First, if you have only a handful of observations over time per unit of observation (say, you follow workers over two, maybe three years), then yes, you can probably argue that individual fixed effects do a good job of purging the error term of unobserved heterogeneity that is correlated with the covariates because said heterogeneity is arguably time-invariant given that individuals do not change that much over the span of two or three years.

But we are not the 1990s anymore: We now have access to much longer panel data sets, and as one adds additional observations over time to a panel data set, there is a lot less that remains time-invariant, and fixed effects become much less useful for identification. In the limit, as the number of time periods goes to infinity, the fixed effects estimator does no better than a pooled OLS. As we note both in this article and in our article earlier this year in the Journal of Economic Perspectives on Yair Mundlak and the fixed effects estimator, this is something that Mundlak himself recognized in his seminal 1961 article in the Journal of Farm Economics (now the American Journal of Agricultural Economics), in which he brought to economics the first application of the fixed effects estimator.

So far so good. But this brings us to the second problem: Why did the fixed effects estimator become the de facto way to deal with heterogeneity in panel data among the reduced-form applied micro crowd, a crowd that is notoriously picky about and likely to cry foul at identification, especially when there are much better options (e.g., first differences) to account for the fact that, in this sad Heraclitean world of ours, although an individual today might be comparable to herself last year, that same individual today is much less likely to be comparable to herself ten years ago?

For whatever it is worth, this is a particularly egregious problem in political science, where scholars often rely on cross-country fixed effects over periods of time in excess of 25 years. But what does remain constant over time for an entire country? Climate and topography can change. Cultures certainly change as well. Even a country’s borders can change in 25 years! So if you always were skeptical of cross-country longitudinal studies but could never quite put your finger on why, here is your huckleberry.

(The funny thing is that when we submitted this to a leading political science journal, we were told by the editor, with my own emphasis: “[W]e feel that the contribution is not strong enough for a top general interest journal … [w]e also feel that the research design and empirical strategy remain underdeveloped.” I am no political scientist, so I cannot assess how valid the former statement is. But when I wrote to the editor in charge about the latter statement, asking whether we were talking about the same paper since our paper is not your usual application to a research question, and thus does not really have a research design or identification strategy, I never heard back from them…)

In this article, we show how the fixed effects estimator can (and often does) break down with ever-longer panels, and we discuss alternative (and often better) means of dealing with heterogeneity in panel data including first differences, of course, but also interactive fixed effects as well as novel rolling estimators. We then illustrate our point with Monte Carlo simulations and by replicating four sets of results published in leading journals—with only one such set of results turning out to be robust.

Here is the accepted version, and here is the abstract:

Data that span multiple units and time periods allow controlling for time-invariant heterogeneity correlated with the covariates. While researchers can do this in different ways, the fixed effects estimator—also known as the within estimator, and equivalent to the least squares dummy variable approach—has become the default choice. But when time-invariant attributes are not invariant to time—that is, when they are not invariant to the length of the panel—the fixed effects estimator can be considerably biased as researchers incorporate additional time periods. We show that, in finite samples, first-differencing and novel rolling estimators can offer researchers a practical alternative to the fixed effects estimator in this case. These estimators are simple to implement and can significantly reduce bias relative to the fixed effects estimator under certain data-generating processes. Most importantly, researchers should always provide results from multiple estimators. We illustrate this with simulations and four replications.

If you would like to use the alternative estimators to fixed effects we discuss in the paper, you can find a Stata template in this .zip file.

I am glad that this is finally accepted for publication: Bernhard and I first discussed it in 2022 when he was here for his postdoc, and he began presenting it when he was on the job market that same year, in early 2023. Since then, the paper has only gotten better as a result of comments from colleagues. I really like that it marries Bernhard’s interests in international trade and agricultural value chains with my own interest in food prices and agricultural value chains.

It also helps that the results tell an interesting story: As countries become more involved in global agricultural value chains, food prices go down, but food price volatility increases. Given what we know about how consumers and producers respond to food price volatility, then, it is no surprise that low- and middle-income countries may be reticent to liberalizing their agricultural sector even as high-income countries insist that they do so.

Here is the accepted version, and here is the abstract:

We study the relationship between the extent of participation in global agricultural value chains (GAVCs) and food prices at the country level. Using longitudinal data on a sample of 138 countries for the period 2000–2015 and a shift-share instrumental variable design, we study how the extent of a country’s participation in GAVCs in a given year relates to food price levels and volatility in that same country and in the same year. We document a mean–variance trade-off in food prices, finding that participation in GAVCs is associated with a decrease in consumer food price levels but an increase in food price volatility. Looking at a country’s upstream (i.e., closer to producers) or downstream (i.e., closer to consumers) positioning in GAVCs, we find that food price volatility is associated more strongly with downstream participation than with upstream participation.

The story here is not just that the job market is bad this year; the story is that, with the exception of an uptick in 2022, the job market has been going from bad to worse over the last six years. I knew this anecdotally from looking at how my PhD students on the job market have been doing over that time period, but it is sobering to see my intuition supported by actual job market data.

I have heard a variety of reasons for why this is happening. “The Trump administration doesn’t like experts!” Yeah? Doesn’t explain why the trend predates January 2025 by quite a few years. “AI is replacing us!” Be that as it may, ChatGPT, Claude, and Gemini weren’t exactly household names in 2023 or even early 2024. “There’s an enrollment cliff coming!” Sure, è pericoloso sporgersi, but since when do teaching needs directly dictate hiring needs, under the “if you build it, they will come” logic? And so on, and so forth.

For all I know, it’s a question of… economics. See, for years there was this profession where you were all but guaranteed a really good job—one that would reward you handsomely with money, the freedom to work on whatever you felt like or both, in some rare and much-coveted jobs at top departments—if only you were willing to put in the hard work of getting a PhD. And for a long time, this was true year in, year out.

I know that economic theory is viewed as old-fashioned by and has fallen out of favor with younger people, but anyone who has done a PhD in economics (or even a PhD in something econ-adjacent like business or public policy) has heard of allocative efficiency, the idea whereby resources tend to flow to where their productivity (and thus their wage or rent) is the highest. Given that, is it any surprise that bright young people flocked to econ (and econ-adjacent) PhD programs? Is it any surprise that eventually the gap between demand and supply would be filled, and that there would then be an excess supply of workers in that lucrative profession, especially given the five-, six-, and even seven-year time lag between the start of a PhD and the job market’s verdict on a fresh PhD?

But there’s also another, no-less-important fact that seems to be getting lost in discussions of how bad the job market is. I alluded somewhat snarkily above to the fact that economic theory is viewed as old-fashioned by and has fallen out of favor with younger people. At best, it is seen as yet another entry barrier to be dispensed with via coursework and sundry qualifying exams. I cannot speak for macro or for more structural fields (e.g., industrial organization), but as someone who is working under the broad umbrella of reduced-form applied microeconomics, I have seen the field change from relying heavily on theory to derive testable predictions that were tested almost exclusively with observational data in the early 2000s to something much more atheoretical relying on causal inference methods, with or without experimental data.

One of my coauthors likes to joke that difference-in-differences is all younger people seem to know about nowadays. That is a bit unfair, but it does contain a kernel of truth: A lot of the work done in applied micro nowadays is almost entirely empirical, often with little to no discussion of the economics of the application—and that’s if and when people look at an application that is actually economic in nature rather than looking at questions that seem more relevant to other disciplines.

As I have been telling graduate students lately: The empirical methods you use—diff-in-diffs, for sure, but also RCTs, IV, RDD, synthetic control, shift-share, etc.—are tools that are accessible to people in other disciplines just as much as they are available to you. In fact, there are social scientists who do excellent work writing about econometrics from outside of economics (a lot of them were trained as political scientists and teach in political science departments; people like Matt Blackwell, Adam Glynn, Cyrus Samii, Maya Sen, and so on), and they often do a much better job of explaining their contributions to applied researchers than your friendly neighborhood Real Rigorous Econometrician. So at the end of the day, if you are going to be doing work that is hard to distinguish from the work done by applied researchers in other disciplines, you are only setting yourself up as a more expensive version of what those other disciplines can offer.¹

So what is to be done, as Lenin famously asked? For starters, one might want to bring economic theory back into what they are doing. The one thing that separates economists from other social scientists is that economic theory gives us a way to analyze the world—one that often leads to conclusions that are counterintuitive or surprising. I’m not arguing for full-blown structural work, just for bringing back economics in (reduced-form) applied micro research.²

1. “But we know about identification and do a better job of identifying causal relationships than other disciplines!” Bless your heart. During those Seven Blighted Years during which I taught at a policy school, first-semester MPP students were taught about causal inference. ︎
2. And again, I am not a macroeconomist nor am I a structural econometrician, and so I do not speak for what those guys do. ︎

In my talk, I shared my experience finding my research niche as a grad student, pre-tenure, and post-tenure.

It was nice to prepare slides for this, as it gave me a chance to reflect on how I have approached the last few decades in my professional life. You can find my slides here in .pdf format.

I have never been under any illusion that my research would influence policy—a belief which has only become stronger these past eight months—and so it is good to know that what I write might at least have an influence on my chosen profession.

Unfortunately, from what I hear and from some of the comments I saw on LinkedIn, where many of us have now retreated,¹ it seems some people took a descriptive discussion of the trade-off between quality and quantity in publishing as a green screen on which to project their insecurities.

As far as I can tell, there was nothing offensive about my describing the quantity approach to being a successful agricultural and resource economist in my earlier post. As I wrote then, quantity has a quality all of its own, and with a greater quantity of articles comes a significantly greater likelihood that a high-quality article will emerge. The only thing I could maybe see as controversial if I squint hard enough was when I wrote that if a department wants to go up the rankings, it should invest in quality rather than quantity. But then again, I don’t know why that’s controversial because, well… just look around for proof.

(In what follows, I will be using originality to differentiate between a high- vs. low-quality article, ceteris paribus. It will soon become obvious why I do that.)

But there is one rather pernicious way in which quantity is a decreasingly useful strategy for success.

With the advent of and constant improvement in generative AI, the returns to quantity will decrease significantly, and the returns to quality will increase significantly.

What I mean by this is that thanks to generative AI, the quantity of unoriginal-but-competently-done-and-written articles will explode. To take an example I know well, there will be many more ho-hum articles looking at the “effects” of participation in contract farming on income with shaky identification strategies.

(This assumes that generative AI cannot come up with original and interesting research ideas on its own, something I believe will remain true for a good long while.)

As a result, we are likely to witness the academic equivalent of Steve Bannon’s “flood the zone with shit” strategy. Journals will see a firehose of unoriginal-yet-competently-done-and-written articles, and their editors (or their own algorithms) will not be able to tell the difference between unoriginal-yet-competently-done-and-written articles written by generative AI and submitted by unscrupulous authors on the one hand and unoriginal-yet-competently-done-and-written articles written by authors who submit their own work.

So what happens then? Journal editors will have to come up with a means of choosing what gets published that goes beyond whether an article is competently done and written well. And what they will use as a rationing device is likely to be whether a research question is original and interesting. In other words, they will look for whether something is of good quality since, in a world where everything is written well and competently done, originality will be the only mark of authenticity—and thus of quality.

My point is this: If I were a graduate student these days, not only would I choose not to do development economics, but I would also make triply sure that I invest in quality rather than quantity. Because many academics become obsolete as soon as they defend their dissertations, this might require bucking the advice of many advisors, but the payoff will in all likelihood be worth it.

1. Twitter long ago lost its usefulness to academics, and Bluesky never had any, because it turns out that the thing that is most hated about Twitter—its algorithm—is the thing that made everyone love Twitter in the first place. It’s almost as though there is money to be made selling outrage. More seriously, going from Twitter to Bluesky is like going from a right-wing Charybdis to a left-wing Scylla, and the last thing I want (or need) in my life are more polarized and polarizing viewpoints. But if you can get past hollow congratulations on your work anniversary, LinkedIn is a good way to share professional stuff. ︎

Here is the abstract, with what strikes me as the most exciting things about this paper in boldface font:

An inverse unconditional relationship between farm or plot size (e.g., hectares) and productivity (e.g., kilograms per hectare) is often observed in low- and middle-income countries that appears to be at odds with economic theory. The traditional approach to studying the inverse relationship regresses yield (i.e., output divided by size) on size as well as control variables, testing the null hypothesis that the coefficient on size is zero. We first show that in many circumstances, the relevant null hypothesis is misspecified because the estimand cannot be zero. Moreover, because size appears on both sides of the equation—indirectly on the left-hand side as denominator, and directly on the right-hand side as a measure of size—inherent issues arise with the identification of the relationship between size and productivity. Specifically, any unobserved production factor, even if independent from size, will introduce bias in the estimated coefficient. We next highlight persistent methodological flaws and contradictions in the literature on the inverse size–productivity relationship, discussing how better controls and more precise measurements are unlikely to ensure unbiased estimates. We further identify the stringent requirements that need to be satisfied to correctly estimate the relationship. Finally, we conduct a meta-analysis of the literature on the inverse relationship, discussing the evolution of empirical specifications and documenting evidence of publication bias in favor of negative and significant estimates of the relationship between size and productivity.

Ars longa, vita brevis. This paper is the fruit of several years of thinking. I remember working on early analytical derivations during the summer of 2018, trying to overcome jetlag while in Tokyo to teach a short course at Waseda University. This paper is also a contribution to a literature that is both old and new. Next year will mark the hundredth anniversary of A.V. Chayanov documenting the existence of an unconditional inverse relationship between farm size and productivity in Russia. But as we document in our meta-analysis, the number of studies on the inverse relationship has practically exploded since 2010. And over the last 100 years, the inverse relationship has captivated the attention of many researchers, including that of a Nobel laureate.

Here is the abstract:

“Economic policy and research rely on the accurate measurement of welfare. In nearly all instances, measuring welfare requires collecting data via long household surveys. If survey response patterns change over the course of a survey to introduce measurement error, this measurement error can be either classical (i.e., changing distributions, leading to noise) or non-classical (i.e., changing expectations, leading to bias). We embed an experiment in a survey by randomly assigning a questionnaire with either the assets module near the beginning of the survey or the assets module at the end of the survey, delaying enumeration of assets by about 60 minutes. We find no evidence in the full sample that survey ordering introduces differential response patterns, either in the number of reported assets or the reported value of those assets. In exploratory analysis of heterogeneity, we find evidence of non-classical measurement error due to survey ordering within sub-samples of respondents who (i) are from larger households or (ii) have low levels of education. Our experimental design can be generalized to serve as an ex-post test of data quality with respect to questionnaire length.”

Here are the slides of that short talk, in which I tried to go beyond what I had already said in Doing Economics.