Perfectly Rational Addiction
Economic models typically assume that people have well-defined goals and tend to make decisions that further those goals. For example, one’s goal as a commuter driving home from work may be to choose a route to minimize your driving time, and model worlds with many drivers each attempting to achieve that goal are used to predict how changes in a road system would affect traffic patterns. People in a model are ‘rational’ if they make decisions that are consistent with their goals. It is important to emphasize that ‘rational’ in this context is a technical jargon: It loosely means people weigh the benefits and the costs of a given action when making their decision, but it does not make any judgment about what defines a cost or a benefit per se (Mas-Colell et al., 1995).
Consider a simple example of a model of consumer choice invoking this rationality assumption. A person can buy cigarettes or various other goods and services with a given amount of money. Given income and the prices, all affordable combinations of cigarettes and other goods define a ‘menu’ from which the person must choose. If the price of cigarettes is US$10 per pack and people have US$100 to spend, then 11 packs of cigarettes is not on their menu. If they can consistently rank these options from most to least preferred, which implies that if they rank A higher than B and B higher than C, they must rank A higher than C, then they are rational in the economic sense of the word. Given their rankings, how their choices will vary as the economic environment varies can be predicted. Whether a given change in economic environment makes the person better off in a well-defined sense – that is, makes an option available which is preferred to the current choice – can also be deduced from the model.
The rational addiction model extends this sort of analysis to capture special properties of addictive goods and activities. Canonical models of consumer choice take preferences as given: At some time, for example, a consumer has preferences over cigarettes and other goods and services, and chooses accordingly. The rational addiction model is dynamic; it is a model of decisions and outcomes over time, a complication which is necessary to capture the idea that addictive behaviors today affect behavior and outcomes in the future. preferences over the addictive good and other goods and services at a given time are endogenous in the rational addiction model, as they depend on previous behavior.
The standard rational addiction model makes strong assumptions over this dynamic process, although these assumptions are weaker and more realistic than had been previously invoked in the literature. Before Becker and Murphy (1988) some economists had attempted to model consumption of addictive goods as ‘habits’ that have some but not all features of addictive behaviors. In these models, how much you smoke today depends on how much you have smoked in the past, but you do not take into account that your consumption in the future will change if you choose to smoke more today. People in these models are myopic and naive: They are constantly surprised when they discover that how much they smoked yesterday has changed their desire for cigarettes today.
Becker and Murphy (1988) consider the other extreme case: Instead of completely failing to understand that tomorrow’s outcomes depend on today’s behaviors, Becker and Murphy consider a world in which people understand this relationship perfectly. They model addiction as stock, not unlike a capital stock, that increases or decreases over time according to the flow of consumption. Abstinence leads to depreciation in the addictive stock over time – if you quit smoking today, your level of addiction to cigarettes will decay over time. This decay is offset by consuming the addictive good – smoking more today increases your stock of addiction. Whether you become more or less addicted over time depends on whether you consume enough of the addictive good to offset the decay in your addiction. This model is intended to capture stylized facts about the dynamics of addiction: Addiction does not start or stop instantaneously, rather, an addiction is built up over time through use of the addictive substance and addiction decays over time with abstinence or decreased consumption. The rational addicts choose current consumption being fully aware of how their behavior today affects their stock of addiction, and thus their behavior, tomorrow.
Becker and Murphy prove that people in such a world will display behaviors that are typically associated with addiction. The model predicts that a rational addict builds tolerance and goes through withdrawal. Sufficiently strong addictions generate ‘cold turkey’ quitting behavior as opposed to quitting slowly by gradually decreasing consumption over time. Further, the model allows economists to predict how addicts will respond to a change in the price of the addictive good, and hence provides a lens through which tax policy toward tobacco, alcohol, and other addictive goods can be viewed. Finally, the model generates a falsifiable prediction about behavioral responses to price changes: An anticipated future increase in the price of the addictive good will cause rationally addicted people to immediately reduce consumption of the addictive good. This effect follows from the rational addict understanding that a future price increase will make consuming at current levels in the future more costly, and the pain of withdrawal is diminished if consumption is reduced by small amounts over time rather than a large amount in the future. Likewise, a price decrease in the past will result in more past consumption, leading to a higher addictive stock, and greater consumption levels today. A consequence of this model is that all prices, past, current, and future, influence the person’s current consumption decision.
An extension of the rational addiction framework to multiple addictive behaviors can also explain cyclical binging and abstinence. Palacios-Huerta (in press) shows binging behavior is a prediction of the rational addiction model in which there are multiple, substitutable, addictive goods. For example, consider someone who is addicted to both cannabis and to alcohol but considers them substitutes. If the two behaviors deplete one’s health stock in different ways (one through liver damage and the other through lung damage), binging behavior will result as individuals alternate between the two activities, binging on alcohol while lung health recovers and binging on cannabis while liver health recovers.
These models help us understand addictive behaviors in realistic settings in which there is a complicated relationship between consumption of various drugs and policies, which may fail if they ignore these relationships.
The rational addiction model has a number of important consequences with respect to how the policy is evaluated and implemented. First, people who discount the future heavier are more likely to engage in addictive behaviors. This is particularly relevant in explaining why smoking uptake is so much higher among youth rather than adults. Second, the full effect of a permanent change in prices on individual behavior cannot be judged in the short run. Facing a price increase, the addict will reduce consumption gradually over time. Becker and Murphy predict that in the long run addiction leads to a more price-responsive demand, a hypothesis that is confirmed in the empirical literature discussed in Section Imperfectly Rational Models of Addiction. Finally, announcing future change in tax policies will impact current consumption.
The Becker and Murphy (1988) model is subject to two major criticisms. The first is that the model predicts that addicts will never regret their choices. A large body of evidence falsifies this prediction. The second criticism is that strong assumptions are made about information. In particular, people can accurately predict the future effects of their current consumption. A much-criticized welfare implication follows: The rational consumer always makes optimal consumption choices, and policy interventions designed to deter consumption of addictive goods are generally welfare reducing.