Delay discounting has been widely used in studies of health and health behaviors; probability discounting has also been examined, though to a lesser extent. In these prior studies, either money or hypothetical health is discounted. The research presented uses a novel approach by calculating the delay and probability discount rates for the act of smoking itself; the discount rates were used as indices of perceived risk. These discount rates were used to predict smoking status for youth and adults separately, as well as to predict the intent to quit for all ages. Logistic regression modeling was used to create models that controlled for other risk perception variables as well as demographic variables. For comparison, models that did not include these discount rates were constructed for each group. An approach for calculating a combined probability and delay discount rate was also proposed. This combined discount rate was compared to models that contained separate discount rates.
For both youth and adults, a lower delay discount rate indicated higher odds of smoking. For both youth and adults, the probability discount rate was involved in an interaction, but the overall trend was that a lower probability discount rate was associated with higher odds of smoking. Though these findings contradict much of the discounting literature, these associations were as predicted; because a lower discount rate suggests a perception of less risk, this indicates that youth and adults who smoked perceived smoking as less risky than those who did not smoke. Notably, the delay discount rate did not contribute significantly to models already containing the probability discount rate.
Among smokers, those who intended to quit smoking had higher rates for both probability and delay discounting than those who did not intend to quit. This suggests that smokers who intended to quit had elevated perceptions of risk compared to smokers who did not intend to quit.
Most of the variables needed to calculate discount rates were significant predictors in the models that were constructed without discount rates. This suggests that it is appropriate to use discount rates as indices of risk. In addition, most discount rates contributed significantly to models without discount rates, supporting the idea that discount rates are useful in describing smoking status or intent to quit. Trends of risk perception variables between models predicting smoking status or intent to quit are also discussed.
A combined discount rate (CDR), which incorporated the information in both the delay and probability discount rates, was also proposed and examined. The CDR was a more significant predictor in models not containing a discount rate then either or both of the individual discount rates, suggesting that the CDR may be useful when describing choices with consequences that are both temporally distant and uncertain.
Finally, a method for combining information collected using two different rating scales into a single continuous measure was proposed. Though this was not a primary goal of the study, the approach may be novel, as well as useful to other investigators.