Close up of a map pin with an apple in the center

Muti

Market Forces Can Help Fill Food Banks

Use them to allocate almost anything more effectively.

You needn’t be a food-insecure person to have some familiarity with the United States’ network of organizations devoted to fighting hunger. You may have volunteered at one of the country’s 60,000 food pantries and meal programs. Or donated to one of its 200 food banks. Or know one of the 10 million–12 million Americans who visit a food pantry in a given year. This network is as vast as it is vital.

This vastness, and the heterogeneity it implies, presents considerable challenges for Feeding America, a not-for-profit organization that distributes food to many of those food banks. Feeding America is the country’s third-largest not-for-profit, after the Red Cross and the United Way, and it distributes about 4 billion pounds of food every year. A lot of that food is earmarked for specific areas or populations by the people who donate it, but hundreds of millions of pounds every year are distributed according to Feeding America’s discretion. Getting the right food, and the right amount of food, to the right people in a reasonably efficient manner is incredibly difficult—so difficult that Feeding America asked a team of academics about 15 years ago to help devise a more effective solution than the one it was using. We did so, using every economist’s favorite phenomenon: market forces.

Together with my colleagues—including Chicago Booth’s Harry L. Davis, Donald D. Eisenstein, and Robert S. Hamada, as well as directors from numerous food banks and staff at Feeding America—I spent 18 months hashing out every detail of a new food-allotment system for Feeding America. Along the way, we created a new currency, discovered the true price of breakfast cereal, and helped provide thousands of additional meals to hungry individuals.

Who needs the food?

Much of the food that goes to food pantries doesn’t actually come from a check that you write to your local food depository. Instead, it comes from either manufacturers or distributors. Feeding America will get a phone call from a big manufacturer that has a truckload’s worth of mac and cheese in its warehouse in Wisconsin that it’s offering to give away. This presents Feeding America with a couple of surprisingly complicated questions: Do food banks want it, and if so, who gets it? 

They’re complicated for a number of logistical and other reasons, but also because of one fundamental problem that Feeding America always faced, which was that it never really knew who needed that truckload most. If it just asked, “Who wants this truck of mac and cheese?” everyone would stick their hands up. But if there’s one thing that markets are good at, it’s allowing desire to be expressed. And markets’ most basic trick is that when somebody raises her hand and says she wants something, it means she doesn’t get something else. 

Prior to our market-based approach, Feeding America’s process for deciding who should get that truckload of mac and cheese that’s sitting in Wisconsin started by looking at its partners—food banks, food pantries, etc.—and those partners’ goal factors, or the size of their potential client bases. For each partner, it took the size of the client base relative to the national average, and the total amount of food it expected to collect nationally, and figured how much food each client should receive: if Feeding America was expecting to distribute 4 billion pounds of food, and a partner served a population that was 5 percent of the national client base, it planned to send that partner 5 percent of 4 billion pounds. From this calculation, it compiled a list of partners ranked according to the volume of food they’d be receiving.

Some food banks are systematically better off than others.

When the call came in that a truckload of food was available, it would dial the first partner on the list. That organization usually would have about six hours to make a decision. If it said yes, it was responsible for going to get the food—it would use its own truck and pay the cost of transporting the food itself. If it said no, it’d be moved down to the bottom of the list for the next call. (Feeding America didn’t move through the list exactly in sequence, since some partners covered big areas and others small areas, some covered wealthier areas than others, and so on.)

The partners who declined were moved down on the list in part because there was some imperative to keep the donors happy. It turns out that if a distributor or manufacturer calls again and again with offers that aren’t accepted, it stops calling. So Feeding America created a penalty for declining donations.

The shortcomings of central planning

There were a couple problems with this system. One big one was that three-quarters of food banks’ food comes through direct donations, such as from local retailers, and therefore Feeding America had no idea what its partners had and how much of it they had. This was particularly a problem because of spoilage: not only did sending, say, yogurt to some place that already had yogurt mean that it didn’t go to some place that needed or wanted it more, but it also created some likelihood that the yogurt would spoil and go to waste. 

An example that illustrates how insensitive this system was to individual partners’ specific food needs was provided by one of the people who was on the committee that developed our market system. He was a food bank director in Idaho, and he would routinely get phone calls from Feeding America saying, “Great news, we have potatoes for you.” But of course, he already had warehouses full of potatoes.

Another problem was that there were some areas of the country that were very close to a lot of distributors and food production, and others that were not. If you’re a food bank in Wisconsin, you’re getting a lot of food anyway. If you’re in East Texas, you’re not near a lot of production and distribution. So some food banks are systematically better off than others.

A third problem was the six-hour decision window the partners had when they received an offer. Often what would happen was that after six hours, the partner would say it couldn’t get the food, and then Feeding America would call somebody else, and that partner would have another six hours. This, in the language of economics, is a very illiquid market. It moves really slowly.

And finally, there was a logistical problem. If you pick two random points in the US, on average they’ll end up being about 1,000 miles apart. In other words, donations were often located quite a long way from whichever food bank or pantry they were being offered to.

A market for donations

Our solution to this problematic operating procedure was to create a marketplace for donations and a new type of currency, which we called shares. Twice a day, Feeding America’s partners submit sealed bids for the donations available in the market—on a given day, there’s usually about 60 truckloads of food—via a website, and they receive an email shortly after the auction ends to notify them whether they’ve won. Partners’ budgets of shares vary in proportion to how many clients they need to serve. So if you’re a food bank in Mississippi, which is the poorest area that Feeding America has, you’d get more shares per capita. But if you’re in Los Angeles, you will also get lots of shares, because there’s a huge population to serve. We also created a money-supply rule that, at midnight each day, recirculates whatever shares have been spent that day—again, according to the magnitude of each area’s need. 

There are three broad reasons why a system based on markets and a specialized currency has been useful in addressing Feeding America’s problems.

First, as noted earlier, everyone doesn’t value all types of food equally. The market gives each partner a way to express what it needs and doesn’t need.

Second, money is what’s called “a store of value.” Suppose that I’ve already got a ton of donations this month, and I don’t really want or need any more food. I can take my shares and sit on them and wait until next month, next year—whenever my need is greater—to spend them. 

Finally, the bidding process that partners use to purchase food allows Feeding America to price donations in a way it previously couldn’t. For example, the most valuable type of food is cereal: cereal is great because it’s durable, it doesn’t need refrigeration, and almost everybody likes it. If you had asked the people at Feeding America, prior to the creation of our market, whether food banks preferred to get cereal over apples, they would’ve said sure, of course they do. But they had no way of knowing how much more they preferred cereal. The market has revealed the price dispersion between different types of food, and it’s far greater than anyone would have guessed.

There are other advantages as well. One is that the market is more liquid than the old system—most of the time, the food is picked up from the donor within 12 hours of it going on the market, whereas in the past, it would sometimes be sitting for three days. This is a strong incentive for donors, many of whom are motivated not entirely by altruism, but also by a need to get this excess food out of their warehouses. 

And what about all that excess food I mentioned food banks and pantries had sitting around? Under our system, that partner with an extra truckload of yogurt can take it and put it on our market and become a seller. It’s another way to increase food usage. 

Because each organization is getting the food it actually needs, it values that food more than the food it was getting before—25 percent more, by my calculation.

Of course, there were challenges to implementing this new system. One of the biggest was making sure it would be fair for everybody. And the threat to fairness was mostly from one dimension: simply put, some food banks are really big, and some are really small. The small food banks have, naturally, smaller staffs and fewer resources overall, which makes them less able to monitor something like our donation marketplace. When you combine this issue with the fact that they also have, by design, fewer shares to spend, it could potentially mean that budgeting and making wise spending choices are more difficult.

As a result, we came up with a number of solutions for these small partners. One is to offer them credit: if they want to get something that’s very expensive, but they don’t have enough money in their account, basically we have a banking system and a mechanism for loaning shares that they can then pay back. We also set up what we call joint bidding—since most donations are by the truckload and might entail tens of thousands of pounds of food, more than a small food bank really needs, multiple small food banks can pool their resources, bid jointly, and divide the donation among themselves.

The power of fake money

The results of the shift in allocation mechanisms have been dramatic, according to a number of measures. One of the most staggering results from the market’s implementation was on the overall size of the food supply. The more efficient, more liquid market we created was accompanied by a large uptick in how much food Feeding America received. From 2004 to 2006, food allocated by these donations increased by about 130 million pounds, enough to feed about 80,000 extra people every day.

But another striking outcome, especially for an economist, was in price discovery. Under the old rules, a pound of one type of food was given the same value as a pound of any other type of food—there was no rate of exchange between them. But we have found that if you allow people to exercise choice and you assign relative values to different items, those values will vary greatly. For example, on our market, you can get 1 pound of cereal for about 35 pounds of fresh fruit and vegetables. Food banks are very different in the choices that they make. Some organizations want the inexpensive things—the produce, the drinks, the snacks. Some organizations only want the expensive goods, such as cereal. Some partners are getting 15 pounds of food for every share that they spend; others are getting close to 0 pounds per share. 

The ability of Feeding America’s partners to express their preferences this way has improved things for everyone, but perhaps especially for the smaller organizations, for whom we were concerned about fairness. Virtually every small food bank responded to the implementation of this market by buying large quantities of relatively inexpensive food. Larger food banks purchased more expensive goods—but they also saved more of their shares. The fact that those large organizations were no longer having their arms twisted to go pick up donations meant that there was now more food to go around for everyone else. So the resource-poor organizations ended up getting more of both the low-cost and the high-cost items as a result. Regular use of the credit system—12 percent of bids are made on credit—further facilitated this.

We can quantify the benefit of allowing everyone to express their preferences and get the food they actually want, instead of the food they’re assigned to collect, as was the case under the old system. Because each organization is getting the food it actually needs, it values that food more than the food it was getting before—25 percent more, by my calculation, on the basis of the extreme positions they take during bidding and the distance they’re willing to travel for the food they want. 

What does that mean in terms of dollars? We can calculate this by looking at how food banks bid on lots of food that are different distances away. If you have already employed the driver, the extra cost of driving another trucking mile is about 80¢. So if a food bank in Chicago bids one amount for a truckload of potatoes in Iowa and a smaller amount for a truckload of potatoes in Colorado, the difference in shares should be equal to the greater expense of driving to Colorado instead of Iowa. So we can see the rate of exchange between dollars and shares, and the data suggest that the surplus value the food banks are getting from the market system is equal to something like $70 million per year.

The market also addressed the logistical problem I mentioned earlier. Feeding America’s partners still have to travel a reasonably long way, on average, to collect donations, but the distance has been cut from around 1,000 miles to roughly 650 miles each way. This alone represents an additional $14 million savings.

Across many dimensions, then, the market has performed well for Feeding America and its partners. But things are changing in the food-bank world, and my collaborators and I are revisiting the particulars of the system to see whether and how they should adapt to those changes. 

One of the essential premises of the current system is that food manufacturers and distributors will make mistakes, and end up with too much food, and give that food to Feeding America when they can’t sell it themselves. But there’s a secondary market for food that didn’t exist previously, in which discount stores are buying a lot of the food that used to get donated. As a result, the quality of the food that’s donated has gone down over time. And that, in turn, has meant that some of the wealthier food banks have basically dropped out of our market over the last few years.

So despite the market’s successes, the system needs some reengineering. As was the case prior to our intervention, the challenges the system faces are real and considerable. But as our experience with Feeding America has made clear, market forces, properly guided, can do amazing things.

Canice Prendergast is the W. Allen Wallis Distinguished Service Professor of Economics at Chicago Booth. This essay is based on a talk given as part of Booth’s Rustandy Center for Social Sector Innovation’s Perspectives in Social Entrepreneurship Series.

More from Chicago Booth Review

Related Topics

More from Chicago Booth

Related Topics

Your Privacy
We want to demonstrate our commitment to your privacy. Please review Chicago Booth's privacy notice, which provides information explaining how and why we collect particular information when you visit our website.