TRB: “Irrational” Behavior and the Ventral Striatum

One of the most enduring observations of behavioral economics is the reference-dependent nature of utility.  Rarely do we behave as if motivated by the absolute values that might result from possible choices; instead, we appear to assign utility to the amount and direction of change away from some reference point.  Known as prospect theory, this idea was at the center of Daniel Kahneman and Amos Tvesky’s Nobel Prize-winning work in behavioral economics, and it predicts actual human behavior more accurately than the expected utility theory of standard economics.

One implication of prospect theory is an explanation for the endowment effect.  According to standard economic theory, our valuation of a good should not depend on our relationship to it.  Countless behavioral studies, however, have shown this not to be the case.  Once someone gains possession of a good his valuation of it rises, and he will demand a significant premium over the original price when selling it only minutes later.  According to prospect theory, buyers evaluate the transaction in terms of a gain, while sellers perceive it as a potential loss; because people are loss averse, individual behavior will diverge depending on these frames.

A paper published in this week’s Journal of Neuroscience (currently offered as an open-access article) by Benedetto De Martino, Dharshan Kumaran, Beatrice Holt, and Raymond Dolan investigated the neural activity associated with reference depended valuation and the endowment effect.  At the beginning of the experiment, subjects chose one of two decks of lottery tickets.  On the front of the tickets was a printed pound value; the back of the tickets had a scratch-off area that determined the actual reward paid by the ticket, which varied randomly between £0 and the face value.  In other words, the expected value of each ticket was exactly half the face amount.  Once inside the fMRI scanner, subjects negotiated to sell tickets from their deck and to buy tickets from the other one.  To do this, they reported for each ticket either the maximum amount they would pay to buy it or the minimum amount they would accept to sell.  Interleaved with these trials were control evaluations in which subjects judged the value of tickets from an third deck.  On certain other trials, the computer randomly assigned values to tickets from the two decks, and subjects were required to accept the computer’s valuation but otherwise to respond in exactly the same manner as when they were dictating the prices.

As predicted by prospect theory, participants demanded significantly higher prices than they offered to pay for tickets of identical expected value.  In other words, they exhibited a strong endowment effect.  This effect was remarkable in that individual subjects’ buying and selling prices deviated significantly even from their own subjective evaluation of ticket values in the third deck.  The fMRI data revealed activity in a network of regions commonly thought to encode the value of stimuli and to guide economic behavior: orbitofrontal cortex (OFC) and dorsal and ventral striatum.  Because these results were produced by comparing neural response when subjects performed evaluations and when they executed the computer’s valuations, they should represent only the processes that underlie valuation and not the other aspects of the task. Activity in OFC and in the striatum correlated with the participants’ stated prices, which is consistent with the idea that these regions encode representations of economic value.

To isolate the brain regions responsible for reference-dependent valuation, the authors determined which areas responded in concert with the magnitude of deviation between subject’s stated valuations of tickets from the neutral deck and their buying or selling prices for those from the other two.  Of the three regions involved in valuation, only the ventral striatum exhibited a significant relationship with the participants’ behavior.  Differences in ventral striatum response not only predicted the different prices offered by individual subjects accross the course of the experiment but also individual differences in susceptibility to the endowment effect across the group of participants.  Thus, De Martino, et al. concluded that the ventral striatum performs the computations that underlie reference-dependent estimations of utility.

The striatum is a major target for projections from the midbrain dopamine neurons that respond to appetitive stimuli and unexpected rewards.  The dopamine system evolved long before animals could produce a surplus or make confident judgments about the future.  In this moment-to-moment existence, the gains and losses that might result from possible avenues of behavior could be far more important than an animal’s overall welfare, a barely coherent notion in that environment.  Despite the massive gains of evolution, our behavior is still, at its core, motivated by the activity of this system.

The value of neuroscience is that it will vividly invalidate the blank-slate model of human decision-making assumed by and necessary for traditional economic theory.  These neural data tell us little more about prospect theory itself than behavioral economic analysis, but they allow to frame the theory in a way that reveals the true causes of our behavior.  Reference-dependent valuation is not an example of a time when we fail to act as we should; it’s an example of our brain solving a problem as it should but failing to solve the right problem.  We can act like perfectly rational economic agents–witness the subjects’ ability to fairly value the third deck of tickets in this study–but in most circumstances we do not.  Only by learning more about ourselves will we learn to predict those circumstances and understand how to perfect our behavior.

De Martino, B., Kumaran, D., Holt, B., & Dolan, R. (2009). The Neurobiology of Reference-Dependent Value Computation Journal of Neuroscience, 29 (12), 3833-3842 DOI: 10.1523/JNEUROSCI.4832-08.2009