Estimating Policy Effectiveness in a Dynamic Context when Individuals Anticipate Policy Change: The Case of Indoor Smoking Bans and Smoking Behavior

Tuesday, June 14, 2016: 3:20 PM
F45 (Huntsman Hall)

Author(s): Brett Matsumoto

Discussant: Titus J Galama

In this paper, I examine how the introduction of indoor smoking bans affects individual smoking behavior using panel data from the National Longitudinal Survey of Youth 1997.  This paper addresses two main questions.  First, are indoor smoking bans an effective policy tool for reducing smoking, and do individuals anticipate the introduction of the smoking bans? I find that indoor smoking bans are effective at reducing the probability that a nonsmoker starts to smoke but potentially increase the probability that a smoker continues to smoke.  Also, there appears to be some evidence that individuals are able to anticipate and adjust their behavior prior to the introduction of an indoor smoking ban.  Individuals adjust their behavior in response to the implementation of city and county level smoking bans in their state of residence but outside of their own county of residence.  I interpret this response as individuals adjusting their beliefs as to the likelihood of a future state level ban.

The identification strategy commonly used to identify the effect of an indoor smoking ban is to use the variation in the timing of the introduction of indoor smoking bans across states as part of a difference-in-difference estimator.  There are two problems with this identification strategy.  One problem, which is specific to this context, is that the state level bans are commonly preceded by indoor smoking bans at the city and county level.  By only considering the policy at the state level, the causal effect will likely be understated because only a subset of individuals in a given state will be affected by the change in policy at the state level.  The second problem is that the policy change is treated as exogenous.  In reality, the policy is the outcome of a democratic process which the individual is able to observe.  For example, individuals may change their subjective expectation of the likelihood of a state level indoor smoking ban after observing cities and counties within the state impose bans.  If individuals begin to change their behavior prior to the enactment of the ban, then identification methods that rely on the exogeneity of the policy will understate the effectiveness of the policy.  The link between subjective expectations, the information used by the individual when forming beliefs, and the potential bias in standard policy treatment effect estimates is an important and ongoing research topic.  This paper provides evidence that the anticipatory effect in the context of indoor smoking bans is consistent with a change in the individual's subjective beliefs about the likelihood of future policy.