Motivation crowding out, the blood donor study

In 1970 Richard Titmuss published The Gift Relationship, comparing blood donation in Britain, which relied on unpaid volunteers, with the United States, where much of the supply was commercial. His argument went further than efficiency. He suggested that paying for blood could reduce the number of people willing to give it, because payment changes what the act means to the person doing it. Economists mostly rejected this at the time, since it contradicts the basic expectation that raising the return on something produces more of it.

Carl Mellström and Magnus Johannesson tested it directly and published the result in the Journal of the European Economic Association in 2008. They ran a field experiment with three groups. The first was invited to register for a health examination to become a blood donor, with nothing offered. The second was offered a payment of fifty Swedish kronor, about seven dollars. The third was offered the same fifty kronor with the option of sending it to a children's cancer charity instead of keeping it.

Among women, 52 percent signed up when nothing was offered and 30 percent signed up when the payment was introduced. That difference is significant at the five percent level, and it is the result the study is known for. Adding the charity option removed the effect completely, returning sign-ups to the unpaid rate.

That third group is what makes the experiment worth more than the headline. The amount of money never changed between the second and third groups. What changed was whether accepting it meant taking it. So the drop cannot be explained by the payment being too small to motivate anyone, which is the usual objection to results of this kind.

Two things keep this on the debated shelf. Across the whole sample the fall from 43 percent to 33 percent did not reach statistical significance, so the finding rests on a subgroup rather than on the population tested. And among men there was no significant difference between any of the three groups, with roughly a third registering in each. A result that appears in one subgroup and not another is weaker evidence than a result that appears across the board, and it should be read that way regardless of how satisfying the mechanism sounds.

The wider literature points the same direction without settling the size. Bruno Frey and Felix Oberholzer-Gee found in 1997 that willingness to accept a nuclear waste facility in a Swiss community fell when compensation was offered. Gneezy and Rustichini found that volunteers collecting for charity door to door raised less money when given a small commission. The direction repeats. The magnitude, and the conditions under which it appears, are still argued over.

Every incentive you attach to a behaviour is competing with whatever was holding that behaviour in place already. If the answer is money, you are adding to it. If the answer is what the behaviour says about the person doing it, you are replacing it, and the replacement is usually worth less to them than the thing it displaced.

Take the referral bonus. Recommending someone puts your judgment on display in front of colleagues, and getting it right is how people come to trust your read on people. Attach five hundred euros and the same recommendation reads as collecting five hundred euros, to everyone including the person making it. The question they are answering has quietly changed from whether this person is worth putting my name behind to whether this person is worth the money.

Or the commission on client check-ins. An unprompted call proves the relationship exists outside the contract, which is the thing that makes a client trust the person on the other end. Turn it into a compensated activity and the call still happens, while the thing it used to demonstrate stops arriving. Clients work this out faster than most companies expect.

The same shape runs through paying for internal knowledge sharing, putting bonuses on behaviours you have named as company values, and formalising the commitment of someone who was already all in. In each case the behaviour continues at first and the reason for it has been swapped underneath.

None of this argues against paying people. It argues for knowing what you are competing with before you set the price. Where the behaviour is genuinely effortful and nobody is getting anything from it beyond the work itself, money adds. Where the behaviour is carrying a signal about the person, money competes with the signal, and the charity condition in the Swedish study is the best evidence available that something other than the amount was doing the work, since the sum never changed. What exactly was being protected is an interpretation rather than a demonstration, and the study does not test what a larger payment would have done. There is also a floor underneath all of this. The signal only competes when the person can afford to refuse the money, so in places and populations where the amount matters materially, payment stops being a statement about identity and becomes what it appears to be. Crowding out is a phenomenon of people with the option to say no. On what actions communicate about the people taking them, see signalling. On the related case where somebody else carries the cost of a decision, see moral hazard.

Source: Mellström and Johannesson, Crowding Out in Blood Donation: Was Titmuss Right?, Journal of the European Economic Association, volume 6, 2008, pages 845 to 863. The original claim is Titmuss, The Gift Relationship, 1970. Related field evidence: Frey and Oberholzer-Gee, The Cost of Price Incentives, American Economic Review, volume 87, 1997, and Gneezy and Rustichini, Pay Enough or Don't Pay at All, Quarterly Journal of Economics, volume 115, 2000.

The Gift Relationship

Richard Titmuss, 1970

The book the experiment was built to test. Titmuss compares paid and unpaid blood systems and argues that payment changes what giving means, which economists spent thirty years disputing before anyone ran the field experiment.