Friday, November 3, 2017

How Food Banks Use Markets

"Imagine that someone gave you 300 million pounds of food and asked you to distribute it to the poor—through food banks—all across the United States. The nonprofit Feeding America faces this problem every year. The food in question is donated to Feeding America by manufacturers and distributors across the United States. As an example, a Walmart in Georgia could have 25,000 pounds of excess tinned fruit at one of its warehouses and give it to Feeding America to distribute to one of 210 regional food banks. How should this be accomplished?"

Contemplate your answer for a moment. Canice Prendergast discusses how Feeding America used to tackle this problem, and how it switched to a market-oriented solution, in "How Food Banks Use Markets to Feed the Poor," which appears in the Journal of Economic Perspectives (Fall 2017, 31:4, pp. 145-62). 

One piece of context that is useful here is that the 210 regional food banks all have local donors, and they typically get a majority of food from those donors. The question here is how to allocate the additional food donations received at the national level. Here's how Prendergast describes the earlier system: 
"Until 2005, Feeding America had a method of allocating resources that is fairly common among not-for-profits: a “wait your turn” system, where it gave out food based on a food bank’s position in a queue. The queue was determined by the amount of food that a food bank had received compared to a measure of need called the “Goal Factor,” which is (roughly) the number of poor in a food bank’s area compared to the national average. The formula is more nuanced than a simple head count, as it distinguishes between usage rates for those below the poverty line, between 100 and 125 percent of the poverty line, and between 125 and 185 percent. When a food bank’s position in the queue was high enough, it would receive a call or email from Feeding America to say that it had been assigned a “load.” The load had to be collected from the donor, and food banks were (and remain) liable for transportation costs. The food bank had 4–6 hours to say “yes” or “no.” After a food bank was offered food, its position in the queue would be recalculated, as its measure of food received relative to need would change. If it turned down the offer, the load would go to the next food bank in the queue. This mechanism had been used since the late 1980s, and it allocated 200–220 million pounds of food each year from 2000 to 2004. Feeding America did not distinguish much between different kinds of food, so that each food bank on average got a similar product mix from them (though randomly a food bank could get lucky or unlucky in whether it would get food that was popular among participants)."
The rationale for a system like this one is pretty clear, and so where the practical difficulties. For example, it was quite possible for a food bank in Idaho to get offered a large donation of potatoes, when it already had lots in stock.  It could take a few days to work out who might get a certain donation of, say, fresh produce--in which time it could spoil. Some food banks have lots of local donors, while others do not, but the Goal Factor approach--based on number of poor people in the area--doesn't take this into account. And so on. 

Feeding America put together a committee to consider alternatives. "The group consisted of eight food bank directors, three staff from Feeding America, and four University of Chicago faculty." Prendergast gives gives a tone of the early interactions, when the Chicago faculty started talking about market approaches, in this way:
John Arnold, a member of the redesign group who was for many years Director of the Feeding America Western Michigan Food Bank said to me once near the start of the process: “I am a socialist. That’s why I run a food bank. I don’t believe in markets. I’m not saying I won’t listen, but I am against this.’’
 This situation is clearly not one in which a pure cash-based market is going to serve the desired function. But the committee came up with a market-related approach, which it called the Choice System.  An internal currency called "shares" was created, which were given to food banks using the same Goal Factor criterion.

Now what happens is that Feeding America holds an internal sealed-bid auction twice a day, Monday through Friday, at 12 and 4 o-clock. On a typical day, there can be 50 truckloads of food donated, 25,000 pounds apiece. The loads are posted online at least two hours before the auction.

The practical advantages of this approach are manifold, and here are a few of them

  • Food banks can bid on the specific things they need, rather than being offered stuff they don't need. For example, food banks often put a higher value on dry goods that will last well,  like cereal or pasta, or on supplies like disposable plates and tableware. 
  • Food banks have some ability to borrow "shares," or for several smaller food banks to bid jointly. 
  • If a food bank has extra local donations, it can offer them to the Choice System and receive additional shares. 
  • Food banks can bid more on donations that are geographically close to them (remember, the food bank is responsible for transportation costs). 
  • If a food bank doesn't need anything that is being donated right now, it doesn't bid, and carries its shares over to the next auction. 
  • A food bank in an area with a low level of local donations can focus its bidding on loads that have a  high nutritional value and calorie count, but aren't as attractive to other food banks. 
  • In a few cases, no food bank really wants a certain donation, but it's important not to upset potential donors, so food banks can bid negative shares--that is, they can receive shares in exchange for picking up that particular load. 
  • Each night, the "shares" that were spent that day are reallocated among all the food banks, using a formula related to the "Goal Factor." 
  • However, if a food bank with lots of local donations--and thus no need to bid--accumulates a certain number of shares, then it doesn't receive any additional shares above that level, on the basis that it clearly doesn't need them. 

Prendergast describes in more detail in his paper how the system has worked in practice, with specific empirical details. But for many economist-readers, the key theme here will be the interaction of local information, incentives, and bidding, working together as a mechanism for efficient allocation. It may not be possible to have central authority with better altruistic intentions than Feeding America. But when it comes to allocating scarce resources, the decentralized market-like mechanism performs considerably better.