An experiment was designed to account for intention-behavior discrepancies by applying the theory of planned behavior to contingent valuation. College students (N = 160) voted in hypothetical and real payment referenda to contribute $8 to a scholarship fund. Overestimates of willingness to pay in the hypothetical referendum could not be attributed to moderately favorable latent dispositions. Instead, this hypothetical bias was explained by activation of more favorable beliefs and attitudes in the context of a hypothetical rather than a real referendum. A corrective entreaty was found to eliminate this bias by bringing beliefs, attitudes, and intentions in line with those in the real payment situation. As a result, the theory of planned behavior produced more accurate prediction of real payment when participants were exposed to the corrective entreaty.