Currently viewing the category: "Econ-o’food"

Our 7th annual econ-of-food class potluck, new and improved: now with an original student recipe!

If the blog post below is too long, don’t read: jump to a small photo album,
or shorter posts from previous years.

We have these dinners just after the class discusses food choice and least-cost diets, meaning the combination of foods that would meet all nutrient needs at the lowest possible expense. For our weekly exercise, students attempt that calculation given current prices in Boston, and compare the result to what very poor people eat in Ethiopia. You can try the exercise yourself using this spreadsheet preloaded with real data, leading to surprising insight about the link between food choice and nutrient needs.

Actual food choices are not least-cost diets: people choose foods for other reasons, and often do not actually meet all nutrient needs. In class we look at global diets using the wonderful Peter Menzel and Faith D’Aluisio “hungry planet” photos, from which one big surprise is how food choices in poorer places get closer to the least-cost source of essential nutrients, as in the corn, beans, oil and fruit diet eaten by refugees in Chad. Most of us eat more of some nutrients than we need, which is good because — as our spreadsheet exercise reveals — we cannot do the math in our heads.

One insight from our least-cost diet exercise is that, even using a spreadsheet preloaded with items from a well-stocked grocery store and precise measures of food composition, hitting nutrient recommendations exactly requires the use of math programming algorithms that were developed during and after WWII by Stigler and Dantzig. Those formulas are still today being used to improve livestock feed, and to inform nutrition assistance for poor people by the USDA, USAID, WFP and others. Most of the time we have to guess at what’s in our food, and choose items that meet our various goals as best we can. The potluck is a chance to celebrate the diverse objectives we actually pursue when choosing what to eat.

To make the potluck fun as well as tasty and nutritious, we invite students to bring dishes that contribute to an overall healthy diet while also pursuing any one of four common aims: (1) convenience, (2) cultural significance, (3) least environmental harm as well as (4) least financial cost. Each student explained their dish, and my esteemed colleagues Nicole Blackstone, Sean Cash, Parke Wilde and Norbert Wilson declared the winners in each category.

As promised, a recipe: Meghan O’Hearn outdid even Google’s chef Anthony Marco with her Savory Pie a la Stigler:

Stigler’s original estimate of a least-cost diet in 1939 contained only enriched wheat flour, evaporated milk, cabbage, spinach and navy beans.  Using Excel to compute the exact least-cost diet in Boston now yields a roughly similar list, including a starring role for depression-era canned spinach, and our latest research from the CANDASA project finds the same kind of items in least cost diets around the world.

Here is Meghan’s recipe — revealing how a little butter, some eggs, garlic, olive oil and spices are enough to turn Stigler’s original mush into a fine quiche:

Stigler’s Savory Pie: yields one 9-inch pie

CRUST (you can use an alternate crust recipe if you prefer)
1 cup wheat flour (plus more for rolling)
1/2 tsp salt
1/3 cup shortening
4 tbsp water

FILLING (adapted from:
1 12-ounce can evaporated milk
1 can navy beans, rinsed and drained
4 tbsp. (half stick) unsalted butter
2 tbsp. wheat flour
4 eggs
100 g spinach (~half a bag of fresh baby spinach)
50 g shredded cabbage
4 cloves of garlic, minced
1/2 white onion
1 sprig of fresh rosemary, minced
1 tsp thyme
salt, pepper to taste
1/2 tbsp olive oil or any cooking oil


PART I: Pie Crust
1. Preheat oven to 400 F.
2. Mix flour and salt in a large bowl. Cut in shortening with a pastry blender until mixture is completely blended and appears crumbly.
3. Mix in water, 1 tablespoon at a time, by lightly tossing with a fork. Add only enough water to form mixture into a ball. The dough will be sticky and tough if to much water is added, and it will crack and tear when rolled if too little is added.
4. Roll out dough into circle 1 inch larger than the inverted pie plate.
5. Fold circle of dough in half, and gently lift. Place into pie plate and unfold. Either prick the entire surface of dough with a fork, or weight the bottom of the crust with pie weights while baking. Pie weights can be uncooked rice, dried beans, small clean pebbles, or small balls sold as pie weights.
6. Bake for 10 min at 400 F, and then remove weights and continue to bake for another 6-8 minutes until the crust starts to get some color. Remove from oven.

PART II: Pie filling and bake
1. Saute onion and garlic in cooking oil on medium heat. Once translucent, add shredded cabbage. Cook down cabbage until wilted and add in spinach, rosemary, thyme, and salt and pepper to taste. Cook until spinach is also wilted– about 4-5 minutes.
2. In a blender, combine drained/rinsed navy beans, evaporated milk, butter, flour, eggs and salt and pepper to taste. Blend on medium-to-high until smooth.
3. Add sauteed vegetables to the the blender ingredients. Blend on low for 20-30 seconds, or until well combined. You can also do this part outside of the blender.
4. Pour filling into the pre-baked pie crust (see above). I ended up with about 1 cup too much filling so I saved that to bake in the future (without a crust perhaps!!). I also poured off some of the liquid (I think in hindsight may be better to use less evaporated milk?)
5. Bake in the oven at 350 F for about 60 minutes or until the filling has firmed and started to brown slightly on top.
6. Let rest for at least 15 minutes before serving.
7. Enjoy your Stigler pie masterpiece!


Long post alert!

The short answer to the title question is:  Yes, some of the time.

The long story starts with the opening phrase of the first modern textbook on principles of economics. That line, from Alfred Marshall in 1890, defines economics as the study of people’s “actions in the ordinary business of life”.  More than a century later, we’re still drawing some of the same diagrams from that book, applying a few basic principles to explain, predict and evaluate people’s choices.

Economists study ordinary life using well-established methods, but many students find those methods to be profoundly weird.  For example, economists take it for granted that people have at least some stable preferences, from which we can interpret their level of well-being.  Are students willing to go there, based on their prior beliefs before the first day of class?

This year I am testing the use of instant polling, and thought I’d try a warmup exercise before the first class to get at the wide variety of perspectives that students bring — starting with the most basic question of whether our choices are even predictable enough to explain.

To reveal students’ baseline beliefs about decision-making, I created a survey asking them what fraction of last year’s choices we “regret and would do differently” if we “had the same options and circumstances as last year”.

This question reveals the degree to which people think our actions can be predicted at all from year to year, based on data we might collect about past choices and their circumstances.  I wanted to compare food choice to other decisions, and to compare levels of regret about meals at home versus food choices in restaurants.  (More about that later.)

If peoples’ choices are predictable, they might also be ‘rational’. To the extent that we are predictably irrational, we can learn from our mistakes and not make them again.  And we might also be entirely predictable but lacking free will.  Economics traditionally focuses on ‘rational’ choices that we don’t regret and wouldn’t change, adding insights from psychology in the vast field of  behavioral economics, and collaborating with people from other fields to explain all kinds of behavior like we do in the economics of nutrition.

Much of life — including dietary intake — may be entirely random, or entirely deterministic.  Economists are interested in those actions which are actually choices, because those actions might reveal stable preferences.  Only when preferences are stable can we explain and predict choices, and allow outcomes to be ranked in terms of how far people get towards the outcomes they consistently prefer.

The kinds of choices I wanted to include in my survey, in addition to food choice, concern other decisions we make frequently with varying degrees of regret and changing our minds.  My highest priority was to include a question about gifts to others, to be sure that students don’t confuse being ‘rational’ with being selfish.  (Econ 101 models often avoid the topic of generosity, but that’s just because the math is tricky.)  I also wanted to include once-only experiences, like movies and live shows that we don’t know much about until it’s too late to change, and also alcohol since it’s entire purpose is to blur things.  The final list, in alphabetical order, was choices about alcohol, clothing, food at home, food in restaurants, gifts, and travel.

The survey was designed in part to reveal whether students thought differently about their own choices than about the choices made by other people.  Perhaps there would be a kind of self-confidence bias, by which students believe their own decisions to be more predictable than other peoples’ choices. A belief in “rationality for me but not for thee” would myopic but understandable:  since we know more about our own circumstances than those of other peoples, it’s hard to see why others behave as they do.  Anyhow I was too rushed to register my analysis plan with AsPredicted and in any case I want to be clear that this poll was not human subjects research — it was done purely for teaching purposes in this class, and its sample size is too small to be generalized.

To keep the survey short and avoid any kind of priming, confirmation bias or social desirability bias, I randomly assigned students to one of two nearly identical surveys.  One asked “for each category below, thinking about your own actions, please guess what fraction of recent choices you regret and would do differently. “, and the other asked the same question but “thinking about other people like you“.  I made it a series of 7 multiple-choice questions, asking if they expected themselves or others to “make all the same choices again“, “change less than one percent (<1%)”, “change less than five percent (<5%)”, “change less than half (<50%)”, or “change most (>50%)” of their own choices.  I had to gather responses to make charts before class, by which time there had been about 85% response rate with 21-22 responses in each arm.  You can see the whole questionnaire and all responses here.

To summarize results, I’ll focus on the extremes:  those who said choices are almost entirely predictable, meaning that they’d change less than 1% or make all the same choices again, versus those that would mostly change (>50%).  Among the students who were asked about their own choices, about a third felt that their food choices are almost entirely predictable, and about a tenth said they would change most of their food choices.  Percentages were the same for food at home and in restaurants, which was contrary to my expectations:  personally, I almost never regret what I eat at home, but do so fairly often when eating at restaurants.  Maybe students eat in restaurants a lot more routinely than I do, so they know what they’re getting.

Respondents expected others’ choices to be less consistently predictable than their own, although responses for movies and shows were similar, and the big exception was alcohol:  more respondents thought that they would learn and change their own drinking choices than that others would learn and change.  That’s a particularly good example of self-confidence, although maybe just a January effect.

We’ll discuss these results in class tomorrow, and learn more throughout the semester.  For now, some takeaways:

(a) Many students think their own choices are likely to change and therefore hard to predict, and an even larger number believe that other peoples’ choices will change, even if their options and circumstances stay the same. Both are probably correct, although I’ll do my best to teach techniques that explain and predict a larger fraction of peoples’ choices.

(b) Food choices are in the same ballpark as other things.  Food is not unique in being subject to random (or seemingly random) whims and fads that change without explanation.

(c) I was very surprised by similarity in responses about food at home and in restaurants, for the reasons mentioned above, and also surprised by respondents’ confidence in their choices about money given to others.  Personally, I second-guess my own charitable donations all the time:  I’m very confident about the effectiveness of some donations, while others are just a guess which I change from year to year.  This is a good example of the final takeaway:

(d) There is a lot of heterogeneity here, with very different answers from different respondents.  In food choice, for example, a third would repeat almost all their choices while a tenth would change most of them.  That’s why business schools teach marketing, so companies can spot their loyal customers and also identify people who might switch.  I hope my economics class works kind of like that, serving all kinds of students in different ways.

(e) For teaching, I’m still experimenting with PollEverywhere but so far it works as advertised and is super fun.  I’ve long done instant polls by asking students for hidden hand signals in front of their chest, but then choices are limited to a thumbs-up, stop or number of fingers, and only I can see their answers.  Also difficult for me to count answers quickly.  Asking students to respond on their touchscreens is much better and has led me to relax my previously draconian no-devices policy (but only on phones – still no laptops, so notetaking is only on paper).  And randomization of polls to compare responses, as in this pre-class survey, seems very promising.  If you’re interested, send me email or respond on twitter:


It’s the first weekend of January — which means the start of job-search season for many 2019 graduates.  I recently looked at U.S. employment trends in agriculture and the food system, and here’s what I found:

1. There are many farms, but few farm jobs
The USDA counts about 2 million farms in the U.S., almost all of which are still owner-operated family enterprises.  Farming is the backbone of any food system.  It requires a huge amount of management skill, but not a lot of time:  the last big survey of labor use on U.S. farms found that they took an average of 3,250 hours per year to run, with family members still doing most of the work as shown for example in Table 4 of this report.

The few jobs in farming that do exist are often seasonal, and they are highly spatially concentrated with limited growth potential.  The chart below summarizes the overall employment data from successive waves of farm labor surveys in the 1990s and 2000s. The only category with continuous measurement are “directly hired” workers, which declined in the 2000s and has not increased since the 2008 recession.

2. Farms provide food, but post-harvest handling provides employment
New employment opportunities in the food sector arise primarily off the farm.  Food manufacturing alone provides about twice as many jobs as farming, and grew significantly from its post-recession bottom of 1.4 million jobs in January 2010 to an all-time high of almost 1.7 million jobs in July 2018.

Data on non-farm jobs are collected by the Bureau of Labor Statistics, Current Employment Surveys.  The chart below compares total employment on farms with the number of jobs classified as food manufacturing, from January 2008 through October 2018 to show post-recession growth in food manufacturing.

3. The biggest source of new jobs is food away from home
By far the fastest growth in new jobs is in food and beverage service provision, for food away from home. Using longer-term Current Employment Survey data allows us to compare the three main kinds of off-farm jobs in the food system:  food manufacturing, food and beverage stores, and food services and drinking places.  The chart below compares these three employment categories to on-farm employment, over the entire time period for which this kind of jobs data is available.

In summary…
Farm jobs are great.  I’ve had one, but it was seasonal (summer of 1984) and when I worked through the winter it was part-time.  Many more employment opportunities start after harvest, like my own high-school job making ice cream.  Since 1990, food manufacturing has consistently offered about twice as many jobs as farming, grocery retailing in food & beverage stores now provides about 4x the number of jobs in farming.  Employment in food retail rose from about 2.7 million to around 3.0 million jobs before the recession, and is now up to an all-time high of 3.1 million jobs.  But almost 3 million jobs have been added over the past decade in food services, including restaurants and workplace dining or other food consumed away from home, which rose from a recessionary low of 9.3 million jobs in January 2010 to an all-time high of over 12 million in October 2018.

Farms, food manufacturing, grocery stores and restaurant services are all interconnected.  Each plays a different role in the food system.  No farms, no food — but most of the labor is engaged after harvest, in processing and retail, especially in your local restaurants and other away-from-home food services.



Research in nutrition and the health sciences is often kept semi-secret until publication.  Why?  Does it matter?
     Scientists in many fields circulate their work in progress as widely as possible, hoping for feedback and citation even before submission to a journal.  Institutions run their own working paper series (like the Tufts economics department), individuals use their own websites (like my personal site), and many use general repositories (like arXiv and SSRN).  Acceptance at a top journal certifies the quality of the final version and facilitates dissemination, but draft work in the physical and social sciences is typically circulated as widely as possible before publication.
     In nutrition and health research, the default rule is secrecy.  Results are typically kept confidential until publication, even for work that will be eventually be published on an open-access basis.  Study designs for human subjects research is disclosed through registries like, and ongoing work may be described at conferences from which brief abstracts are published in outlets like the FASEB supplements, but detailed methods and results are not generally shared until publication day.
     The difference between fields is nicely illustrated by a twitter thread reproduced in this post, in which I experimented with sharing a photo album from an agricultural economics conference so as to see connections between different presentations.  That led to quick reply from a leading nutrition researcher, Purnima Menon, who noted that posting photos of slides could jeopardize publication in top health journals.
     A specific example of how nutrition research is kept hidden until publication comes from a recent conference that I organized at Tufts, called GlobalFood+.  This event was designed around 7-minute speed talks designed for sharing on the internet — but one of the best talks had to be kept off our website.
     Confidentiality of work in progress can be important to prevent theft of ideas, to ensure that scientists receive credit for what they do.  Limiting prior publication might also be important for subscription-based journals, to ensure that institutional libraries want that journal in their collection.  But neither rationale applies when working papers can be cited, and when publication fees are paid for open-access articles.
     In general, the primary reason to discourage prior sharing is to pursue media coverage.  Journal publishers make this clear, as in the statements against ‘pre-publicity’ at the top general outlets, Nature and Science.  Previous posting is especially discouraged in the top medical journals, as explained by JAMA and NEJM.  They encourage publication of teaser abstracts, like a movie trailer, but detailed results are subject to a media embargo until publication day.  Universities and research labs are keen to cooperate, in the hope that science journalists will treat the paper’s arrival as a newsworthy event.
     Embargoes may be needed in some case, but for most studies the scientific community is turning against pre-publication secrecy.  Treating publication as a news event is itself a problem, contributing to ‘study-a-day’ media coverage that exaggerates the importance of new studies relative to previous knowledge.  And limiting prior scrutiny to a handful of referees and editors raises the risk of error.  Policies favoring prior circulation of working papers were adopted long ago in PNAS, and have recently been adopted in the health sciences at BMJ and The Lancet.  In nutrition, the AJCN and Journal of Nutrition still put some limits on prior circulation, discouraging their use.
     Why would different kinds of journals have different policies?  One factor could be audience demand for different kinds of news.  Many people want to know about the latest findings in nutrition and medicine, so media outlets often assign reporters to meet that need.  Universities are happy to supply a curated flow of individual studies in the specific fields that reporters most want to write about, like dietary advice and lifestyle choices.  Managing the flow of news is also important for high-stakes pharmaceutical trials and other controversial studies.  But in many cases, secrecy before publication is sought mainly to protect the economic interests of the publisher.  With JAMA, NEJM and some others, subscriptions are still important so they must restrict prior publication to enforce a paywall.  And some journals like Science and Nature run scientific articles alongside weekly news about science, with significant revenue from ads for lab equipment and materials. These journals need publicity to attract traffic for their journal as a news source, to sell both subscriptions and advertising.
      Scientific work is changing fast and it’s hard to keep up.  For example, my previous foodecon post was about how to limit the plague of academic spam.  Much is being written about the business of scientific publication, including great nonprofit work by Scholarly Kitchen and the EmbargoWatch blog by Ivan Oransky (who also blogs at RetractionWatch), as well as superb reporting on academic life by and  Personally, I hope that the tradition of pre-publication secrecy is soon replaced by wider circulation of working papers, even in nutrition and the health sciences.  Sharing work in progress would raise overall quality, and help break reporters’ study-a-day habit and encourage them to cover the overall flow of knowledge.  Wouldn’t that be nice?

This year’s class potluck was especially saboroso, with a delicious Sopa Paraguaya from Gabi Fretes — and a wild Puerto Rican Coquito from Nayla Bezares here being praised by judge Norbert Wilson:

Also meaningful, in a different way: Blackbird Donuts (thank you Ilana Cliffer!).

For context, you might check out posts from previous years, or additional photos from this one here.

As always, respect and thanks to our august jury of distinguished food economists, not just Norbert but also Sean Cash and Parke Wilde.  Time for an econo’food recipe book project, anyone?


A big part of economics is data analysis, which starts with data visualization:  “seeing like an economist” means looking for patterns across many observations, recognizing that the data we see result from peoples’ choices.  In class we practice this through weekly exercises and a course project that start with analytical diagrams (such as supply and demand curves) to show the logic by which we explain each observation, and then download data from authoritative sources to make our own charts and tables that summarize what’s been observed.

This blog post pulls together a few suggestions and links about data visualization for convenient reference.  The dataverse of available information is expanding rapidly, with increasingly sophisticated expectations about data visualization.  That complexity can be daunting, making it hard to get started. My vote for best quick advice about data is to keep it simple, as explained in great posts about how to clear off the table and remove to improve.  Those start with bad examples and show how to clean things up and avoid numbo-jumbo. Successful data visualizations help you tell a story, by making comparisons that highlight both similarities and differences.  Charts and tables offer a kind of language designed to help us communicate clearly.

My favorite guide to data visualization for policy audiences is from the UK:, and it’s useful to compare it to an excellent guide to scientific charts and tables here:  Change over time is usually best shown with line graphs like Figure 1 of that page, while differences among categories is usually best shown with bar charts that are sorted by magnitude, and a cloud of individual observations is best shown by a scatter plot.  It’s useful and fun just to browse through the different charts presented here:, and also click through

Your final reports and presentations weave together a sequence of charts and tables.  To keep things straight, all figures (whether an analytical diagram or a chart of data) should be numbered consecutively as Figure 1, 2, 3…, and all tables should be numbered separately as Table 1, 2, 3…  Each should have a clear title and note describing the nature and source of all data shown in the chart or table, so that a future reader could replicate or update your visualization in the future.  Different fields use different conventions about table or figure titles and footnotes, and have preferred visual styles for how things are presented.  In general, economics and other social sciences use brief titles above the chart and detailed notes below it, while many health science readers expect a single long figure caption that combines both kinds of information.  Examples from my own recent papers include one in health economics style (title and footnote), and one in health-science style (a long caption)

For oral presentation, your charts and tables should appear in ways that help you tell the story.  There are many good guides to using PowerPoint effectively, of which one of my favorites is from a prominent biologist named Susan McConnell:

And finally, if you’re interested in guides to writing in general, my favorite is Steven Pinker’s Sense of Style — especially for his brilliant description of how all communication requires effort to overcome the curse of knowledge, in part by chunking information into digestible units which you can then bundle up into increasingly powerful stories.  I look forward to seeing how you put your pieces together!