Drug use and life satisfaction

One of the late Gary Becker's many contributions to economics is a theory of rational addiction (RA; which he developed along with Kevin Murphy (see here, or here for an ungated version). That theory suggests that drug users (and other addicts) are rational utility maximisers, and that drug use today contributes to higher utility from future drug use. For drug users to be rational though, they need to be able to accurate predict their future utility from drug consumption. In contrast with the rational addiction model, the utility misprediction (UM) theory in psychology suggests that people make errors in predicting their utility from consumption. In the context of drug use, they may overestimate the short-term benefits and/or underestimate the long-term costs of their drug use.

In the context of these competing theories, this 2018 article by Julie Moschion (University of Melbourne) and Nattavudh Powdthavee (Warwick Business School), published in the Journal of Economic Behavior and Organization (ungated earlier version here), attempts to determine which theory is more consistent with empirical data. They first note that:

The key distinction between the RA and the UM models is the prediction on the effect of current substance use on life satisfaction. If current use is associated with lower life satisfaction, then the coefficient on Uit [current use] should be negative – and even more negative than the coefficient on Uit+1 [future use]. This would be consistent with the psychology model of UM, which predicts that consuming addictive substances that have harmful properties should lead to a significant deterioration of life satisfaction. However, if the coefficient on Uit is positive – or significantly less negative than the coefficient Uit+1 - then we have evidence that current consumption may have improved users’ life satisfaction compared to the previous period, which would be consistent with the prediction made by the RA model.

Moschion and Powdthavee use data from 1174 respondents who answered all six waves of the Journeys Home survey in Australia over the period from September 2011 to May 2014. The survey included in each wave measures of respondents':

...usage and the frequency of usage in the past 6 months of: tobacco, alcohol, cannabis, and any other type of illegal/street drugs (which might include amphetamines, such as speed and ice, heroin, cocaine, ecstasy, and so on). We focus on the following substance use variables: smoking tobacco daily (versus not smoking or smoking less than daily), drinking 21 or more standard drinks a week (versus not drinking or drinking less than 21 drinks a week)... using cannabis daily (versus not using or less than daily), using illegal/street drugs weekly (versus not using or less than weekly).

The dataset also includes a measure of life satisfaction (measured on a 0-10 scale). Essentially, Moschion and Powdthavee explore the relationship between respondents' current life satisfaction and their past, current, and future drug use. They find that:

• Average life satisfaction drops significantly prior to the consumption of illegal/street drugs, but not prior to the consumption of other types of substances. The evidence on illegal/street drugs is consistent with the idea that less satisfied individuals are more likely to consume addictive substances because of the expectation that by doing so, their experienced utility could improve.
• An excessive consumption of alcohol, the use of cannabis daily and illegal/street drugs weekly is observed together with a significant drop in life satisfaction. This finding is more consistent with the UM framework in which experienced utility is predicted to decrease in periods following individuals’ consumption of addictive substances.
• We find evidence that the consumption of illegal/street drugs 6 months to a year prior to the interview is associated with lower current life satisfaction, independently of current consumption. This suggests that of the substances studied, only illegal/street drugs may have an effect on life satisfaction for more than 6 months.
• Our results thus suggest a vicious circle of illegal/street drug use by which lower life satisfaction increases the propensity to consume illegal/street drugs that in turn further lowers life satisfaction in the future.

They also find that:

...the estimated effects of past, present, and future consumption of illegal/street drugs on current life satisfaction to be noticeably more negative and statistically more pronounced for women and those from a lower educational background.

These results seem more consistent with the UM model than the RA model. On that point, Moschion and Powdthavee note that:

Our estimates could therefore suggest that respondents’ substance use patterns don’t follow a pattern consistent with the RA model although in reality they do if mental health issues explain non-rational patterns of substance use. To test the implications of controlling for mental health issues, we add 5 dummy variables to capture whether the respondent was diagnosed with any of the following mental health conditions between interviews: bipolar affective disorder, schizophrenia, depression, post-traumatic stress disorder and anxiety disorder. Our results are largely unchanged.

However, in spite of those results Moschion and Powdthavee are still unwilling to declare the rational addiction model dead:

Indeed, this drop in average life satisfaction of substance users over the six months following the use of substances is less consistent with the RA model, but does not rule it out completely, provided that the use of substances increases individual’s life satisfaction in the very short term and that individuals have an extremely high discount rate.

So, I guess until we have a dataset that can not only include past, current, and future drug use, along with a measure of utility (like life satisfaction, and probably measured at high frequency in order to pick up short-term effects) and a measure of discount rate, we're not going to be able to definitively satisfy ourselves that the rational addiction model cannot adequately capture drug users' behaviour. In the meantime, the utility misprediction model has the edge.