Background:
Information dissemination on harmful effects of tobacco has always been a major tool for reducing tobacco consumption in all countries. This paper studies how a change in the Perception of Tobacco Risk (PTR) affects the attitude of smokers and non-smokers towards smoking. First, this paper wants to observe how PTR varies among different types of smokers and non-smokers across different socio-demographic backgrounds. Second, it examines among the cigarette smokers across different socio-demographic groups how changing PTR over the years influences their behavior towards smoking.

Methods:
A panel dataset from the ITC Bangladesh surveys is used for the analysis. Using the panel dataset, a Perception of Tobacco Risk Index (PTRI) is developed for all smokers and non-smokers. OLS and Probit regressions are used on different groups based on income, education, age, residence to assess the effects of knowledge and income on smoking behaviour.

Results:
Results show that among all different types of smokers, quitters have the highest PTRI whereas bidi (cheap local alternative to cigarettes) smokers have the lowest. Among the different socio-demographic groups the higher income, more educated, and those living in urban areas display a higher PTRI than their respective counterparts. However, it is also observed that the change PTRI is bigger among the lower income, less educated and those living in rural areas. Analysis of panel data reveals that the change in PTRI is positively correlated with the probability to quit smoking for most socio-demographic groups. However, increase in PTRI does not significantly affect initiation of smoking and reduces cigarette consumption per day only for the more educated group.

Conclusions:
Results of this paper hold strong policy implications. First, they show that changing PTR holds promising implications for controlling tobacco consumption concerning raising quitting probability. Second, they bring into light, specific socio-demographic groups where policies to change PTR should be targeted.