#128 2005 · Feeding America · Food banking
Feeding America let food banks bid fake money for truckloads instead of taking turns, and paid them to accept what nobody wanted
the problem
Centrally planned food shipments matched no food bank's real needs
background
Feeding America's network moves donated food to more than 200 member food banks nationwide, and for years did it the way a government agency would: a central office decided who got what, in what order, based on incomplete information about which banks actually needed potatoes versus produce versus protein that week. The predictable result was mismatched shipments — potatoes routed to Idaho, which already grows more than it can use, hot dogs arriving without buns — and food quietly spoiling in warehouses whose managers had no real say in what they'd been sent.
The obvious fix was better central planning: more data, smarter dispatchers, tighter logistics. University of Chicago economist Canice Prendergast proposed something else — stop planning it centrally at all.
what everyone would do
Improve the central planning — more data on each food bank's needs, smarter dispatchers, tighter logistics. It fails because no planner, however well-resourced, can actually know what 200-plus food banks need week to week better than each bank knows itself; the mismatch wasn't a data or effort problem, it was that the information needed to allocate correctly lived with the recipients, not the people doing the allocating.
what they saw
Prendergast saw that a food bank's real need was information only that food bank possessed, and the only reliable way to extract it was to make the bank act on it, not report it. Turning allocation into a bidding market forces each food bank to reveal its actual priorities through what it's willing to spend its budget on, which is a far more honest signal than a survey or a request form a planner would otherwise have to interpret.
the move
In 2005 Feeding America replaced its queue-and-allocate system with an internal market. Every food bank gets a budget of a synthetic currency called 'shares' and bids against other banks for truckloads of food as they become available, revealing what it actually wants instead of accepting what a planner assigned. Crucially, bids can go negative: a food bank can be paid in shares to take a truckload of, say, unwanted pickles, so undesirable donations get absorbed by whichever bank values the shares more than it dislikes the pickles, instead of rotting because no one had standing to refuse them.
why it works
Giving each food bank a budget of synthetic shares and letting it bid on truckloads means a bank only spends its scarce shares on food it genuinely wants, automatically routing supply toward whoever values it most without anyone centrally having to know why. Allowing bids to go negative extends the same logic to the awkward case a queue system can't handle: unwanted donations, instead of rotting in a warehouse because no one had standing to refuse them, get paid-for placement with whichever bank values the shares it receives more than it dislikes taking the food — so even undesirable inventory clears through the same mechanism as desirable inventory, rather than needing a separate fix.
the payoff
The market now allocates roughly 300 million pounds of food a year across the network. Food banks self-sort by real local need rather than a planner's guess, and the negative-price mechanism specifically solved the chronic problem of unwanted donations sitting unclaimed until they spoiled.
where it breaks
The market only works if every participating food bank has a real, meaningful budget to spend and something to lose by bidding carelessly — a budget too small or too easily replenished removes the incentive to bid honestly, and a budget too large lets well-resourced banks simply outbid everyone regardless of actual need, recreating the mismatch the market was built to solve. It also depends on there being enough variety and volume flowing through the market for genuine price signals to emerge; a thin market with few participants or infrequent shipments doesn't generate the competitive bidding that reveals true relative need, and negative-price mechanisms specifically require enough overall goodwill in the shares economy that a bank willing to accept unwanted food can find a share balance worth taking it for.
what came after
Published as Canice Prendergast, 'How Food Banks Use Markets to Feed the Poor,' Journal of Economic Perspectives 31(4), 2017, and covered by NPR's Planet Money ('The Pickle Problem'), the system is now a standard teaching example of fixing a matching problem inside a nonprofit through market design rather than more resources or better central planning.
references
- [1]Wikipedia — Feeding AmericaWikipedia, 2026en.wikipedia.org
- [2]Prendergast — How Food Banks Use Markets to Feed the Poor, Journal of Economic PerspectivesAmerican Economic Association / JEP, 2017aeaweb.org
- [3]How Feeding America's "Choice System" facilitates equitable food distributionMarketplace (APM), 2025marketplace.org