Bennett’s Law

What foods do poor people buy when they have a bit more money? What you might expect, but not as much of it as you might expect.

Episode Transcript

JEREMY: Hello and welcome to another episode of Eat This Podcast with me, Jeremy Cherfas. (00:06):
undefined
Last time we heard about the 3 billion people worldwide who cannotafford a healthy, nutritious diet.
In this episode, a question that wasn't really relevant last time.

(00:26):
When people do have more money, what kind of food do they actuallybuy?
There's a lot of historical interest in this idea, going back atleast to the middle of the 19th century.
That's when Ernst Engels, a German statistician, published hisdiscovery that as income goes up, the proportion spent

(00:48):
on food goes down, even though the total amount spent on food goesup.
That's now known as Engels Law, and an extension of Engels Law iscalled Bennett's Law, formulated by the American
agricultural economist Merrill Bennett in 1941.

MARC: Bennett's law is is the empirical regularity that says that as people go from very poor to less poor, they substitute away from coarse (01:09):
undefined
grains towards finer grains, so away from sorghum and millet andgrains that are just filling, but not necessarily the tastiest,
towards stuff like wheat, rice and corn.

(01:29):
And as they go from merely poor to perhaps middle income or non-poor,they substitute away from carbohydrates altogether and start bringing
in more protein, usually in the form of animal sourced foods.

JEREMY: That's Mark Bellamere, also an agricultural economist and a frequent guest on this podcast. (01:42):
undefined
And did you catch what he said?
That Bennett's law is not so much a law as an empirical regularity.
What he means is that nobody had actually gone and accuratelymeasured the kinds of foods that people buy when they have more money

(02:05):
to spend, just that they notice that they do seem to switch fromcoarse grains to fine, and then from fine grains to protein.
And of course, the reason I'm talking to Marc Bellemare about this nowis that he and his colleagues have just published a paper in which
they do actually measure what kinds of foods poor people buy whenthey have more to spend.

(02:29):
It is, he says, the first credible test of Bennett's law.
The point here is not that Bennett's law might be wrong.
In fact, it's what you might expect.
But nobody had actually looked in detail.

MARC: On the basis of a handful of assumptions. (02:45):
undefined
Micro-theory will tell you that as income goes up, the demand forfood goes up.
So where we started from was, that is theory.
What do we see empirically?
Is that really true empirically?
And can we get an answer to that question of what is the incomeelasticity of food demand?

(03:09):
Because that's a very thorny issue, as you may know.
And if you ...

JEREMY: Let me just stop you for a minute there, because I understand price elasticity of demand. (03:13):
undefined
What is income elasticity of demand?

MARC: The income elasticity of demand measures what happens to quantity demanded in percentage terms for a 1% increase in income. (03:23):
undefined
What's great about an elasticity measure, elasticity is a unit freeunit of measurement.
It tells you for a 1% change in something, how does the other thingrespond in percentage terms.
So it is -- whether you measure it in dollars, in euros in kwacha --you get a measure that you can compare across contexts.

(03:51):
And this is where we kind of come in with this paper because ...
So we looked at what micro-theory says, we looked at what theempirical evidence so far said, and we asked ourselves, how can we do
better? And it's fairly easy to do better.
First off, you know, I take theory seriously, but I don't take it asgospel, meaning that economic theory is often wrong in specific

(04:15):
contexts. It'll tell you one thing, and then you look at the data,you get the best estimate you can get, and you find something that is
not in line with theory necessarily.
But more worryingly for us, is the fact that many of the estimatesyou see in the literature that are bandied about by policy makers,
some researchers, are just poorly identified, meaning that they arenot necessarily an accurate number.

(04:40):
And that is especially true when it comes to income elasticity ofdemand.
That is something that I've taught.
I think in most undergraduate courses I have ever taught, we don'treally have a good sense of what happens to food demand when income
goes up. The reason being the first thing is demand for what?
Because you can measure the demand for food in several differentways.

(05:02):
The most standard way that people think about would be perhaps interms of calories, right?
What we do in the paper is we look at food expenditures, because thatis something that we can harmonize across contexts, and it is fairly
accurately measurable.
You know, I'm not saying that calories are not accurately measurable,but they're often based on imputations that are very old and that are

(05:22):
dating to kind of a time when the calorie content of various foodswas very different from what we have today.
And so you can measure it in dollars or expenditures, you can measureit in calories.
You can say we will measure it in terms of macronutrients.
How many grams of protein, how many grams of carbohydrates do peopleconsume?

(05:43):
You can measure it in terms of micronutrients.
You can say, well, what is the demand for vitamin A, vitamin D?
And because we have so many different measurements of what fooddemand is, we have a construct validity issue, right?
So me talking about food demand may not be the same thing as an FAOcolleague talking about food demand or a nutritionist talking about

(06:06):
food demand. We needed something that could be compared acrosscontexts.
And so we settled on food expenditures, which is by no means aperfect measure, but it is the measure that we have.
It is something that is measurable in monetary terms.
And then the way we got to writing this paper was, I recall having aconversation with Eeshani Kandpal, my co-author, one of my co-authors

(06:27):
on this. The other is Katherina Thomas.
And I remember telling her, I'm very interested in Bennett's Lawbecause, you know, the name notwithstanding, it wasn't really a law up
until our paper.
Bennett's Law was more like, you know, something that we kind ofobserve in cross-sectional and time series data, but we don't really
have an iron-clad estimate for.

(06:49):
And so I recall an early conversation with Eeshani and saying, well,what could we work on together?
And I said, well, we should do a randomized control trial where wetest Bennett's Law.
And Eeshani, who is infinitely smarter than I am, said, wait asecond.
We don't have to spend any money doing this.
There are publicly available data sets of cash transfer, randomizedcontrolled trials.

(07:13):
So randomized trials where ...
Be selected to receive an infusion of cash regularly should you meetsome conditions.
And that kind of gives us internal validity.
That is what allows us to make a causal statement there.
And what's very interesting about that is that we have five RCTs ofconditional cash transfers in four different countries.

(07:35):
So we have two in Mexico.
We have one in the Philippines, one in Uganda, and the last one is inNicaragua.

JEREMY: Okay. So you've got these ... (07:41):
undefined
Effectively they're government programs usually where some people geta cash handout and other people ...
I don't know whether they're actually randomized.
You say they're randomized controlled trials, but there's a group ofpeople who get the cash and a group of people who don't, and you can
look at their expenditures.
So what is the income elasticity of demand for food purchases?

MARC: Well, it's positive. (08:08):
undefined
So in line with your intuition, as poor people get money and this ismoney that's given to them kind of in a random
fashion, they can they take some of it and spend it on food.
In other words, demand for food goes up.
The interesting news is that it's not as much as people expected.

(08:29):
Our estimates are lower than what the literature had hithertopresented.
And that is an interesting fact.
Or, that's an interesting result, because honestly, I was of theschool of thought that, you know, you're going to see them very much
an economist in that sense, but I was of the school of thought thatincome fixes everything.

(08:50):
Money is the, you know, is the answer to a lot of problems.
And so I thought, well, if we just give people more income, right, ifas incomes go up, as the development process takes place, all of those
nutrition issues are going to get resolved because people are justgoing to start upgrading what they consume, and they're going to be,
you know, they're going to be better nourished.
We're going to see fewer hungry people.

(09:12):
So our income elasticity estimates are not as high as what mostpeople kind of would have speculated up to now.
And that was kind of a very sobering fact because it ...
You know, sometimes you find a result that really changes how you viewhow the world works.
And that is one of them.
That is one of those things where I, you know, it hasn't happenedfrequently in the course of my research that I had a finding where I

(09:37):
thought, this is entirely ...
It's not entirely different, right?
If it were entirely different, I would have found that giving peoplemore money means that they spend less money on food.
That is not what we find, but we find that they don't spend as muchas what we expect.
And thus the whole kind of income as solution to so many problemsdoes not hold.

JEREMY: Just to go back. (09:56):
undefined
You said that elasticity is great because it tells you how much doesspending go up for a 1% increase in income.
And naively, you might think 1% increase in income, 1% increase in inspending on food.
So how much do they increase their spending on food.

MARC: Yeah. So so what we do is we find an income elasticity for overall food expenditures of 0.03. (10:16):
undefined
It's about three cents on the dollar.

JEREMY: That's really not very ... (10:25):
undefined

MARC: It is really not very much, exactly right. (10:28):
undefined
It is way less than you would expect.
I mean, I would not expect a 1 to 1.
I would not expect a one dollar on the dollar because we know thatfood is a necessity, which means or we know, we know from theory --
again, see, I'm taking theory a little bit too seriously -- what weknow from theory is that food ought to be a necessity, meaning that

(10:48):
the income elasticity of food ought to be less than one, for foodoverall.
That is, for luxury goods, for stuff like caviar, we would expect theelasticity to be greater than one, because those are luxury goods or
they're deemed to be luxury goods.
But for food overall, we would expect it to be between 0 and 1, andwe certainly find between 0 and 1, but we find it much closer to zero

(11:11):
than than I expected going into this project.

JEREMY: And it's also ... (11:15):
undefined
I mean, I have ...
My expectation would be that it would be higher, especially becausevery often cash transfers are targeted at women because they're
kind of responsible for feeding the family, and they're oftenaccompanied by education on nutritional questions.

(11:37):
So maybe that targeting is not necessary either.

MARC: That is something we can't quite get into, right? (11:46):
undefined
We can't speak to the targeting of those cash transfers.
What we have is this exogenous variation in who gets cash and whodoesn't.
I would like to clarify, too, that this is an intent to treatestimate, meaning that when we give people money, we can't force them
to take it up and do stuff with it, right?

(12:07):
So this is, we intend to treat those people or the people who ranthose randomized control trials in the first place intended to treat
the people who receive the cash.
What they do with it is entirely up to them.
So it's what we call in the literature an ITT estimate.
That being said, Jeremy, I want to go back a step.
What we find is, yes, overall for food, we find that it's 0.03.

(12:29):
So three cents on the dollar.
But of course staples, right, stuff again going back to like mydefinition of coarse versus fine stuff like sorghum and millet exhibit
the most inelastic demand.
And their estimated income elasticity is not statistically differentfrom zero, meaning that for a change in income, we don't see people
changing in a statistically meaningful fashion their expenditures oncourse staples.

(12:55):
What has the highest -- which is in line with so many reports and somuch of the conventional wisdom in this kind of economics
of food demand and nutrition literature -- is that the most elasticdemand is animal sourced protein.
And so in that sense, we do find stuff that's in line with people'sexpectations of, yes, as people get these kind of as people get these

(13:20):
random infusions of -- random across people, not over time -- theythey will spend, you know ...
The demand of food, sorry, the category of food that responds most tothis influx of cash is animal source protein.

JEREMY: So that's getting to Bennett's law or Bennett's Observational Regularity. (13:33):
undefined

MARC: Correct? Yeah. (13:39):
undefined

JEREMY: Can can you can you actually do the three steps, of course to fine staples and fine staples to meat? (13:40):
undefined

MARC: So you got us there. (13:47):
undefined
So, I am very, very happy to report that yes, we can get to thosethree steps.
What I'm slightly less ...
What I was kind of more surprised by, not disappointed necessarily,is that we find partial support for Bennett's Law.
The average household in our, across our five contexts substitutesfine staples for coarse staples, substitutes protein for coarse

(14:12):
staples, both of which are consistent with Bennett's Law.
But you'll notice that there is a glaring omission, right?
We don't find that they substitute protein for fine staples.
So this is kind of like this middle step where people might, whereyou think, well, they would move away from rice and consume more

(14:33):
chicken perhaps, or more fish.
We don't see that step.
What we do find is that people will substitute chicken and maybe porkand beef for sorghum and millet, and they substitute rice for sorghum
and millet. But we don't find that middle step, which may just ...
I mean, we don't know whether that's kind of a noise problem in thedata, where, you know, it's the data are just a little bit too noisy

(14:56):
to detect that. We find the right sign in that direction, but we justdon't find significance, is what we find.
But it is exciting.
I mean, you know, I said I was disappointed.
That is not entirely true.
I mean, I'm very excited about providing the first credible test ofBennett's Law in the literature.
It's something that's been kind of on my mind very much since Ilearned about it.
I thought, this is fantastic.

(15:16):
This is entirely true, entirely true based on my expectations orbased on my intuition and the patterns that we observe in, kind of, in
our lives.
So if I, if I may just kind of a little personal aside.
My parents, you know, you don't have to go very far up the familytree to find people who were living in rural areas of Quebec in

(15:43):
poverty and people for whom even ...
You know, if I talk to my mother, she will tell you that a meal isnot a meal unless it's got starch.
So a meal is not a meal if it doesn't have potatoes.
A meal is not a meal if it doesn't have bread.
And she will entertain the possibility of pasta and rice as well,which are not exactly endemic to French Canada, but she will entertain

(16:08):
that possibility.
And so these things don't change very quickly, I think, acrossgenerations, because of habit formation and the way we eat.
And that might be something to look into, I think, as a next step of,sure, okay, you see that people themselves will change a little bit at
the margin, but there's a lot of holdovers from history.
There's a lot of holdovers from culture and from these are habits ofconsumption that are formed well in infancy.

(16:34):
Right? I mean, you look at how people eat and they'll say, well, I'vealways eaten this way when I was little, and there's all kinds of
emotions and kind of cultural baggage that is tied into that, andpersonal history that's tied into how people eat that demand might
not switch as easily for food as it does, say, for VCRs or the typeof computer that you're using.

JEREMY: Maybe this is a digression, or maybe ... (16:56):
undefined
Well, if ...
Is there a kind of inverse Bennett's Law that when people are richand educated and have been for a while,
they start eating lower, they start eating whole grains and less meatand quote unquote, healthy.

MARC: That is a fantastic question. (17:21):
undefined
That is where my mind has been for, I would say, about ten years,because when I learned about Bennett's Law, yes, I thought, hey, you
know, I've lived in Madagascar.
I certainly have seen people in a cross section of Malagasy society,right, certainly eat lots and lots of rice at the low end of the

(17:42):
income scale, eat a lot more meat.
So the guys that I was working with, the enumerators that wereworking with me on my, on my survey when I did my dissertation
fieldwork. They they loved meat, right?
Whenever we would go out as a team and I was footing the bill, thoseguys would go all in on meat.
And they were very happy to eat, you know, like more than rice andgreens.

(18:05):
So I've seen that.
And, you know, we see it in our own experience and again, in a timeseries sense of my grandparents and my parents and the way they were
raised and the way they eat.
And so that I asked myself, okay, but this is very interesting.
It tells us what happened in the past from a development sense,living in a high income country.

(18:26):
But quo vadis, right?
What am I looking at for the future?
And so I've got an undergraduate student, a wonderfully talented youngwoman.
And best of all, she is intellectually curious and she's driven, andshe loves to think about food policy.
But what she is doing is, using qualitative evidence to document whatare the consumption patterns of people in this country, in the United

(18:51):
States, across the income domain.
And so my initial speculation -- and she she doesn't have very firmresults yet -- we met yesterday for the, not for the first time, but
for the first time she brought some qualitative results.
But my expectation going in was exactly as you say, right as you getto a certain income level -- and sorry for this very roundabout

(19:11):
answer, it took a while to get here -- but as you get to a certainincome level, you start demanding different food.
You start worrying about health aspects of the food you're consuming.
Sure, Bennett's Law tells you that as you get rich, you're going toconsume more animal sourced foods.
And that's certainly true is true in what EEshani, Katherina, and Ifind in our paper.

(19:35):
But if you look at the data in the United States, for instance, yousee that people consume a lot more fish and seafood, a lot plant, a
lot more plant derived protein.
I mean, I am an example myself.
A colleague is visiting here who is giving our Friday seminar and sheis staying with us.
And this morning I was making her coffee and I said, what do youwant?

(19:57):
We have regular milk, but we also have pea milk and almond milkbecause my wife likes to drink pea milk with her coffee and I drink
almond milk, and we have dairy milk because that's what my daughtereats in her cereal and, you know, little kids should probably get
dairy milk with all the enhancements and enrichments that it has.
So yes, we certainly see that in the data, right?

(20:18):
That as people get very wealthy, they will substitute away from this,away from kind of like straight up animal source protein and go
towards vegetable or plant derived protein.
And even within kind of animal source, they're going to look at fishand seafood more than anything.
So you're entirely right.

JEREMY: Okay. Going going back to the poorer people and their demand for meat and animal protein, if it's kind of, if (20:37):
undefined
it's lower than may be expected, does that mean that fears about theimpact of demand for animal
protein on greenhouse gas emissions-- so more livestock, moregreenhouse gas emissions, more climate change -- are those fears then

(21:05):
exaggerated?

MARC: I hesitate to say that this is but one estimate. (21:09):
undefined
My sense if I were to go purely with our estimates, I would say yes.
Maybe those fears are a little bit exaggerated.
That doesn't mean that they're wrong, right?
It just means that they are less than what we were expecting.
That being said, the fact remains that income elasticity for a fooddemand for animal sourced foods remains

(21:35):
positive. Which means that as countries that are on the lower tomiddle income scale develop, they are going to again, the sign is
right, right? The sign is what people thought it was.
And so as as the demand for food, as the demand for animal sourceprotein goes up in low and middle income countries, something is going
to have to give if we want -- with the goal of climate -- if we wantto kind of hit those climate targets.

(22:00):
And thus I think it is incumbent on high income countries to kind ofexamine their own consumption patterns, because where I have a
problem as an economist and where I have a problem as a, you know, asa critical thinking citizen, is when I hear policymakers kind of look
at those demand patterns and say, oh, well, people in low incomecountries are going to have to adjust their consumption and move away

(22:25):
from ... And there are a lot of people like that who will kind oflecture low and middle income countries and say, oh, you got to, don't
do what we did, right?
Do as I say, not as I did, which I find very hypocritical becausethere comes a stage of development where we are not
working in offices, typing on computers all day.

(22:46):
Right? People do physical work in the manufacturing sector or stuffthat's not necessarily a services sector, where you need you need to
grow strong and tall.
And the way you do that is by consuming enough animal source protein.
I'm not a nutritionist, but from what I know, you can get enoughprotein from plant derived foods.

(23:07):
But from what I know, you don't get the nine necessary amino acids inplant derived proteins the way you get them as easily available in
animal source foods. And so to deny that to billions of people whenyou're sitting pretty here in your ivory tower, I find this a little
bit, you know ... It's not, you know, it's not just wrongheaded, it'salso hypocritical.

JEREMY: Yeah. I mean, the Eat Lancet, the original Eat Lancet diet was a classic example of, oh, no, poor people mustn't eat meat. (23:31):
undefined
Nobody must eat meat.
Anyway, leaving that aside, finally, in the paper, you actually saythat increased income had a limited impact

(23:52):
on food consumption and in turn, on nutritional well-being.
But you also say that cash transfers are effective, and that'sespecially if you compare them to nutritional supplements and in-kind
transfers. So, I'm a policy maker.
What's your best advice?

MARC: I think ... (24:17):
undefined
So my best advice would be do ...
cash transfers are indeed effective right?
They have, they hit several targets.
We're looking at one slice of what those cash transfers do.
And the good thing about cash transfers is that they're verystraightforward.
They are incredibly simple to administer.
And so if you think about well what's cost effective, what's easy todo is ...

(24:41):
Giving people cash is easy to do and then it gives them ...
And this is where I may differ from many other policymakers or frommany policymakers and other economists or other social scientists ...
But I think when you give people cash, they will use it for what theythink is best for them.
And sure, you're going to have some collateral damage, like someone'sgoing to use the cash to drink it, someone's going to use the cash

(25:03):
for, you know, to buy cigarettes with it, to buy illicit drugs.
That is not something you can fix.
And that is something that, at any rate, as the development processoccurs, you're going to see more and more of that.
So I find the whole cash transfer argument to be very appealing.
It's very administratively simple and it delivers results even if theresults are more humble, even if, or even if the estimates of the

(25:27):
effects are more sobering than what you thought going in, as I didwith this paper.

JEREMY: Marc Bellemare of the University of Minnesota, reaffirming the idea that giving poor people cash is a good idea, even if they don't (25:33):
undefined
spend quite as much of it on food as one might like.
I'll put a link to the paper by Marc and his colleagues in the shownotes at Eat This Podcast.com, along with some other links that you

(25:56):
might find interesting.
I hope you enjoyed this episode, and if you did, it would be great ifyou would tell a friend or even the world at large.
Discovering new things to listen to remains the weakest link inindependent podcasting.
And by spreading the word, you can help.

(26:17):
That's all for now.
Don't forget, you can follow me and ask questions on Instagram andMastodon.
The links are on the website at Eat This podcast.com.
So from me, Jeremy Cherfas and Eat This Podcast, goodbye and thanksfor listening.