Food Affordability Over Time

On a number of occasions, I’ve written about the Engel Curve, which relates the share of consumer spending on food to the consumer’s income (or total expenditures on all goods). Whether we compare consumers within a country or compare spending across countries, a common relationship holds: the higher a consumer (or country’s) income, the smaller the share of their income they tend to spend on food.*

This relationship indicates that as consumers and countries get richer, we’d expect food expenditure shares to fall, a phenomenon generally thought to be associated with higher consumer well-being. While this relationship is widely known among economists, there is another fact that is not as widely known. In particular, the entire Engel Curve has been shifting downward over time. That is, for any given level of income, consumers today are spending less on food than they were in the past.

To illustrate this phenomenon, I pulled data from the Consumer Expenditure Survey that has been collected annually since 1984 by the Bureau of Labor Statistics (BLS). The BLS report food expenditures and total expenditures by quintiles of income. These data were used to create the following animation.

The video shows that, despite the year-to-year variation, there is a fairly steady shift in the Engel Curve over time downward and to the right. That is, consumers are getting richer over time (i.e., their total expenditures are rising), and for any given level of total expenditures, the share being spent on food is generally falling. There are several possible drivers of this phenomenon, but one likely culprit is technological progress. For any given level of income or total expenditure, innovation and technological change has brought down the price of food such that consumers are able to eat what they want while being able to spend more of their income on other, non-food items. That is, food today is more affordable (at least by this metric) for households of all incomes (or total expenditure categories).

*Note: just because the food share falls, it doesn’t mean total spending on food falls as income increases. In general, richer consumers spend more on food than poorer consumers. However, spending on non-food items tends to increase at a faster rate than spending on food as income rises, leading to a smaller share of income being spent on food.

Producing More with Less

I’ve given a lot of talks over the past couple years about the importance of increasing agricultural productivity. Often these discussions get couched in Malthusian terms related to the need to produce more food for a growing world population. This 2009 document from the United Nations Food and Agricultural Organization, for example, suggests agricultural production in developed countries needs to double by 2050 to meet the demands of expected population growth.

I’ve been in enough of these conversations at this point to know that a common retort is that we already produce more than enough food to feed today’s population. Isn’t this just an issue of distribution rather than supply? I’ve addressed this issue in previous posts. Here, I want to draw out an implication of productivity growth that is probably obvious to many academic economists, but perhaps not as widely appreciated as it should be.

In particular, when we talk about increasing productivity enabling us to “do more with less,” the focus is often on the “do more” part. That is, increase food production. But, one shouldn’t forget the “with less” part. In short, increasing productivity means producing more sustainably.

To illustrate, consider the figure below (this is what we economists call a production function).


The bottom, lighter blue curve shows the relationship between various inputs (land, water, fertilizer, labor, etc.) and output (or food production). The figure shows that we can produce more food by adding more inputs - more land, more water, etc. However, there are diminishing returns. The first few gallons of water (or rain showers) produce a lot of extra bushels, but the next few gallons have a smaller effect. In fact, if we get too much water (i.e., a flood, as some parts of the Midwest are currently experiencing), production can actually fall. Diminishing marginal productivity was at the heart of the Malthusian concern - if we keep adding more population (or workers) to a fixed amount of land, the extra amount of food that will be produced (and available per worker) will fall, and hunger will ensue.

How do we escape this “trap”? Scientific research, innovation, and entrepreneurship allow us to shift up to a higher curve, as shown by the darker blue curve in the above graph. For a given amount of inputs (labor or land), we might actually have more food per person (now and on into the future as long as we continue to innovate and shift the curve outward).

Let’s, say, however, that one already thinks we produce “too much.” We don’t want any more food. Ok. I’ve drawn the vertical dashed line in the above figure to show a constant amount of food production. But look where this line intersects with the production functions. The figure shows that higher productivity curve allows us to use fewer inputs (less land, less water, less fertilizer, fewer pesticides, etc.) to produce the same amount of food as compared to the original lower production function.

The point? Even if one believes the problem of production is “solved”, don’t still want to find innovative ways to increase productivity to reduce our use of scarce natural resources?

So, how has US agricultural productivity fared? Here is data from the USDA Economic Research Service.


The figure shows that agricultural output has grown by factor of about 2.7 (i.e., we’re producing about 170% more food) since 1948, while use of agricultural inputs, in aggregate, have grown very little and is essentially flat. The gap between the output line at the top and the input line on the bottom is the definition of productivity.

How will this graph look in 2050? Is it possible the trend lines for outputs and inputs can flip? That is, flat output and falling inputs? If total output stays relatively constant, but we can find ways to improve productivity, then total input use will fall. That would be a great sustainability story.

Crop Yields and Taste

That modern agriculture is incredibly productive - much more than the past - is undeniable. These USDA data, for example, suggest we produce about 170% more agricultural output now than in the late 1940s. I have argued that these these increases in agricultural productivity are signals of improved sustainability. Some people believe the the productivity improvements have been accompanied with offsetting externalities or degredations in animal welfare. A different kind of critique is that modern crops - despite being more productive - aren’t as high “quality.” For example, this piece in Politico by Helena Bottemiller Evich, titled “The great nutrient collapse” discuses evidence that vitamin content of crops has fallen as yields have increased, and there is the often-heard complaint that tomatoes don’t taste as good as they once did.

There is some biological basis for these latter concerns. If a crop breeder selects plants for higher yields, they are selecting plants that are spending their energy and nutrients into producing bigger seeds and fruits, which is energy that could have gone (in lower yielding plants) to growing leaves or roots or other compounds that affect taste and vitamin content.

I had these thoughts in the back in my mind when I came across the Midwest Vegetable Trial Report put out by researchers at Purdue and other Midwestern universities. The report compares different vegetable varieties in terms of yield and other output characteristics. I noticed for a couple vegetables - green beans and sweet corn - there were also measures of taste for each variety. Granted, these were not full-on scientific sensory evaluations and they involved small numbers of tasters, but still I thought it would be useful to test the conjecture that higher yielding varieties taste worst.

Some researchers from University of Kentucky put together the green bean report. They compared the performance of 19 different varieties of green beans. The most productive variety (named “Furano”) yielded 785 bushels over six harvests, whereas the lowest yielding variety “Slenderette” only produced 233 bu/acre in six harvests. As the image below reveals, however, there was only a weak correlation between taste and yield. The correlation was negative (-0.26), but not particularly large. About 6.6% of the variation in yield is explained by taste. The best tasting variety “Opportune“ had a taste score of 4.1 (on a 1=poor to 5=excellent scale) and a yield of 557; the worst tasting variety “Bronco” had an average taste score of 2.3 and a yield of 543. So, the best tasting bean had better yield than the worst tasting bean. Overall, the results below provide some weak support for a yield, taste trade-off.


The report also provided production and taste data on supersweet corn (this part was authored by Purdue researchers Elizabeth Maynard and Erin Bluhm). They compared 16 different types of bicolored supersweet corn (they also evaluated two varieties of white and two varieties of yellow, which I’m ignoring here). They had tasters rate “flavor” on a 1 to 5 scale. As the figure below shows, there is actually a positive correlation between flavor and yield, as measured by ton/acre. The correlation is 0.15, but the relationship is weak. The authors also report yield in a slightly different way, ears/acre, and by this measure the correlation is slightly negative (-0.09).


These results don’t necessarily negate the idea that the taste of vegetables has declined over time as higher yielding varieties have been adopted, but they do suggest that in 2017, among the particular varieties tested and among the few tasters asked, there is only a very weak correlation between taste and yield for green beans and supersweet corn.

Trends in Farm Land Acreage

I hear a lot of talk about the impacts of federal farm policy on our food system. It is sometimes suggested that farm policy is to blame for “cheap food” and thus obesity (see this nice twitter response by Tamar Haspel) or that many of our purported modern day farm and food ills can be traced back to Earl Butz, who as Secretary of Agriculture in the early 1970’s encouraged producers to plant “fence row to fence row.”

One way to evaluate these sorts of claims is to look at how much (or little) crop acreage in the U.S. has changed over time. Here is data according from the USDA, National Agricultural Statistics Service on the amount of land planted to nine major commodity crops over time (note: vegetable acreage, which comprises only about 1% of all acreage is not included; nor is fruit or nut acreage, which is also a very small share of the total).

The figure below shows the cumulative acreage in the U.S. planted to nine major commodity crops over 93 year time period from 1926 to 2018. Over the entire time period, there was an average of 246 million acres planted to these nine crops each year. Seven out of the 10 highest planting years were prior to 1937 with the remaining three being in 1980, 1981, and 1982.

The coefficient of variation (the standard deviation divided by the mean) is only about 7.5%, implying relatively low variation over time (usually a figure less than 100% would be considered low variation). Since 1990, there have been relatively small year-to-year changes. Over the most recent 28 year time period, about 225.7 million acres are planted each year to these nine commodity crops, with a coefficient of variation of only 1.8%. This lower variation in recent years is interesting because farm policy has been much more market-oriented since 1996, and this is precisely the period over which there has been more stability in planted acreage.

Total land devoted to farming (or crop acreage) today is about 12% lower than the highs of the 1930’s and the early 1980’s. This is amazing in many ways given that the U.S. population is now 130% higher than it was in the 1930’s. Stated differently, twice as many people are now being fed on fewer crop acres.


Moving away from total acreage, it is instructive to look at the mix of acreage (see the two following figures). Here, we can see some significant changes in which crops are planted in the U.S. over time. For example, in 1926, there were only 1.9 million soybean acres but in 2018, for the first year in history, more acreage (89 million acres) was planted to soybeans than any other crop. Prior to that corn had been king every year except 1981-1983, when more acres were devoted to wheat than corn.

Another big change was a reduction in the number of acres planted to oats. Prior to the 1960’s, more than 40 million acres of oats were routinely planted each year. In 2018, only 2.7 million acres were in oats. Why the change? One big reason is that there aren’t as many mules and horses that need to be fed. Cotton also experienced a precipitous reduction in acreage from the late 1920’s to the early 1960s, stabilizing a bit thereafter.


The following figure shows the same data, but with acreage dedicated to each crop expressed as a percentage of total acreage in a given year.

Taken together, these three figures suggests the big change hasn’t been the total farmland planted but rather the change in which crops are planted to the acres. Moreover, this crop mix issue (the rise of soy and the decline of oats) probably had little to do with farm policy.


Given all the concerns expressed these days about mono-cropping, it might be interesting to look at the variation in planted acreage (in terms of the mix of crops planted) today than in the past. To see this, I calculated the coefficient of variation across the number of acreages planted to each of the nine crops in each year. This gives a feel for how much crop variation there in a given year. Here are the results plotted over time.


The coefficient of variation ranges from about 87% to 138%. Comparing this to the coefficient of variation for total acreage planted (which was 7.5%), implies there is more variation in which crops are planted to which acreages in a given year than there is variation in the total planted acreage over time.

The figure above shows that the crop-mix variation (at least among these nine crops) has been increasing since the 1960s, and the variation is higher in the past decade than at any point in the preceding 80 years.

New findings on agricultural productivity

The American Journal of Agricultural Economics has recently published several new and important papers on agricultural productivity.  Whether agriculture is becoming more or less productive is a critical question as it relates to sustainability (are we getting more while using less?), food security (can food production outpace population growth?), and consumer well-being (are food prices expected to rise or fall?). 

These papers focus on "total" or "multifactor" productivity rather than just yield.  Yield is a partial measure of productivity - it is the amount of output per unit of one input: land.  One can increase yield by adding more of other inputs such as water, fertilizer, labor, etc.  What we want is a measure of how much output has increased once we have accounted for uses of all inputs, and this is total or multifactor productivity.

The first paper by Matthew Andersen, Julian Alston, Phi Pardey, and Aaron Smith is worrisome.  They write:

In this paper we have used a range of data and methods to test for a slowdown in U.S. farm productivity growth, and the evidence is compelling. The results all confirm the existence of a surge and a slowdown in productivity but with some variation in timing, size, and statistical significance of the shifts. ... Over the most recent 10 to 20 years of our data, the annual average rate of MFP [multifactor productivity] growth was half the rate that had been sustained for much of the twentieth century. More subtly, and of equal importance, the past century (and more) of statistics assembled here suggest the relatively rapid rates of productivity growth experienced during the 1960s, 1970s, and 1980s could be construed as aberrations (along with the relative rapid rates of growth experienced during a period spanning WWII), with the post-1990 rates of productivity growth now below the longer-run trend rate of growth.

The second paper by Alejandro Plastina and Sergio Lence provides a deeper understanding behind the causes of productivity growth.  They present a straightforward way to decompose multifactor productivity into six different factors: technical change, technical efficiency, allocative efficiency, returns to scale, output price markup, and the input price effect.    They write:

Technical change is the major driver of TFP growth over the long run, and there is evidence that technical progress in the 1990s and 2000s was much slower than in the 1970s. This is a relevant result for policy makers, and begs the question of what is actually causing the slowdown in technical change. This is the first study to show technical regress in the agricultural sector during the farm crisis of the 1980s.

Another novel result is that annual changes in TFP bear no significant correlation with annual rates of technical change but instead are highly correlated with the markup effect, followed by the returns to scale component and allocative efficiency change. These findings suggest that evaluating the effects of research, extension, and other variables on each of the components of our measure of TFP change (rather than solely on an aggregate TFP index) can shed light on the actual channels through which those variables affect agricultural productivity growth in the United States and therefore contribute to policy design.

Finally, there is Julian Alston's fellow's address from last year's AAEA meetings.  In addition to providing an excellent literature review, he makes several important points.  He argues that agricultural research is significantly under-funded relative to the benefits it provides in increased productivity:

Evidence of remarkably high sustained rates of social payoffs to both private and public investments in agricultural R&D testify to a significant failure of government to fully address the underinvestment problems caused by the market failure. Moreover, if anything, in high-income countries like the United States, agricultural R&D policies seem to be trending in the wrong direction, making matters worse.


a reasonable first step would be to double U.S. public investment in agricultural R&D—an increase of, say, $4 billion over recent annual expenditures.4 A conservatively low benefit-cost ratio of 10:1 implies that having failed to spend that additional $4 billion per year on public agricultural R&D imposes a net social cost of $36 billion per year—an order of magnitude greater than the annual $1–5 billion social cost of $20 billion in farm subsidies.

Alston also points out that the main beneficiaries of productivity growth are consumers, and the farmers may or may not benefit.  He writes:

It seems inescapable that the agricultural innovations that made food much more abundant and cheaper for consumers did so to some extent at the expense of farmers as a whole—more than offsetting the effects of growth in demand for output from the sector. This finding is reinforced when we pay attention to the details of the timing. Specifically, the periods of the most rapid decline (or slowest growth) in [net farm income] seem to coincide with the periods of most rapid increase in farm productivity—the 1940s to 1980s, especially 1950–1980, as identified by Andersen et al. (2018)—consistent with the hypothesis that agricultural innovations have reduced net incomes for U.S. farmers as a group.

This suggests something of a paradox.  Farmer groups have often been some of the biggest supporters of agricultural research and are proponents of productivity growth, while consumers have been skeptical if not hostile toward many productivity-enhancing technologies on the farm.  Yet, it is likely food consumers that have received the lion's share of the benefits from increases in agricultural productivity through greater food security and lower food prices.