You are here:
ACCOUNTING FOR THE ENDOGENEITY OF HEALTH AND ENVIRONMENTAL TOBACCO SMOKE EXPOSURE IN CHILDREN: AN APPLICATION TO CONTINUOUS LUNG FUNCTION
Citation:
Williamson**,JM, J. ACCOUNTING FOR THE ENDOGENEITY OF HEALTH AND ENVIRONMENTAL TOBACCO SMOKE EXPOSURE IN CHILDREN: AN APPLICATION TO CONTINUOUS LUNG FUNCTION. Presented at 2004 Toxicological and Risk Assessment Conference, West Chester, OH, April 26 - 30, 2004.
Impact/Purpose:
To inform the public
Description:
The goal of this study is to estimate an unbiased exposure effect of environmental tobacco smoke (ETS) exposure on children's continuous lung function. A majority of the evidence from health studies suggests that ETS exposure in early life contributes significantly to childhood morbidity, but an important question with respect to exposure and lung function has yet to be clearly answered in observational studies. That question is one of causality: Does ETS exposure contribute directly to poor health, or are there other factors correlated with ETS exposure and poor health that bias the estimates? A primary assumption in ETS research, with a few exceptions, has been that exposure leads to poor health, even after controlling for other observable factors. When estimating exposure effects, a single stage regression has been considered adequate. That assumption may be invalid if other factors that influence a child's health and at the same time influence his or her likelihood and level of ETS exposure are missing from the health production function. The issue of concern is of selection effects, an underlying cause of non-randomness in observational data, influencing exposure to ETS. For example, parents with children who have existing respiratory illnesses may take steps to avoid exposing children to ETS. Failing to account for the selection effects creates a biased estimate of the effect of ETS on child health. My research attempts to overcome the biases created by selection effects involved in the production of child health. The innovation of this study is that it employs widely used econometric techniques to account for selection effects in the human health production function. Using the considerable work in econometric methods that has produced unbiased treatment effect estimators for non-experimental data, along with the Third National Health and Nutrition Examination Survey, I present an instrumental variables framework. I first estimate a level of ETS exposure using the biomarker serum cotinine. In the second stage of the estimation process, the exposure variable (predicted level of serum cotinine) is used as the instrumented variable in the child health production equation. The outcome of interest is continuous lung function. The results from the two-stage estimation procedure show a downward bias of single stage regression results. My research strengthens the argument that ETS exposure has a detrimental effect on children's continuous lung function. My findings indicate a 1% increase in ETS exposure, as measured by serum cotinine level, reduced lung function by an average of 5% or a reduced forced expiratory volume of between 115-140 mL.