Preferences are endogenous when they are determined by, and may change as a result of, other factors. Preference endogeneity appears as an issue, for example, in explaining behavior driven by altruistic, reciprocal, or envious preferences, conformity to social norms, and the effect of persuasive advertising. Although it may appear obvious to noneconomists that preferences change and that economists should account for them, standard neoclassical economic theory takes preferences as exogenous. There is no doubt that apparent evidence for unstable preferences can be explained by suitable ad hoc assumptions embedded into stable, and exogenous, utility functions. Nevertheless, it is questionable whether ad hoc assumptions can be held as serious competitors to more psychologically realistic views of preferences as subject to change.
But endogenous preferences are unattractive to many economists for understandable reasons: (1) neoclassical economic theory assumes utility maximization of a stable utility function, and the risk of allowing the utility function to change is to be able to explain everything by suitable tinkering, and thus explain nothing; (2) endogenous preferences typically imply a loss of parsimony in the economic model, a loss accepted as essential by some (e.g., Bowles 1998) but not by most; and (3) endogenous preferences also make welfare analysis more difficult since normally welfare is measured in terms of utility, and if the utility function changes as a result of a policy change, the metric on the basis of which welfare is measured also changes. A partial solution to problem three is to show that a policy change is better, or worse, according to the preferences held both before and after the policy change. Problems one and two might be answered by good theory and evidence combined.
There is clear evidence for the instability and endogeneity of preferences in specific setups. For example, when decision makers repeatedly face a new situation, such as a market in the experimental laboratory, a learning process takes place where agents shape their preferences in interaction with the market setup. Gary Becker and Kevin Murphy (1993) reviewed the ten U.S. companies with the largest ratio of advertising expenditures to sales, and noted that many of the products the companies produced—such as chewing gum, beer, or cola—conveyed no or very little information. Among others, Samuel Bowles (1998) and Daniel Zizzo (2003) reviewed a variety of evidence on the endogeneity of altruistic, reciprocal, and envious preferences.
The status of theories of endogenous preferences is, however, less satisfactory. Institutional economics accounts, such as Wilfred Dolfsma (2002), are interesting but generic. Behavioral economics models do little more than arbitrarily postulate certain endogenous relationships. Evolutionary game-theoretical models of altruistic, reciprocal, and envious preferences abound (e.g., Bowles 1998), but they typically involve zero-rational agents (where a modicum of rationality could make a difference) and are liable to criticism as evolutionary just-so stories. Artificial-intelligence approaches (based on neural network or hybrid modeling) may help and may be testable against data, but have so far been mostly neglected.