An econometric technique to remove unobserved variables that bias the relationship between alcohol and blood pressure

We use econometric techniques to consider whether the statistical association between drinking and blood pressure among men may be due, in part, to the constitutional hypothesis. The constitutional hypothesis holds that the same unobserved genetic or personality factors that affect blood pressure will affect the amount of alcohol consumed. Our sample is restricted to men because most investigations for women have revealed a weak to nonexistent positive, and sometimes a negative, association. Data are drawn from the National Health and Nutrition Examination Survey, 1971-1975. The econometric technique requires fitting three equations using two-stage least squares or multiple regressions. The first equation explains how much people drink. The second and third equations explain fluctuations in systolic and diastolic blood pressures using information on the predicted values of the drinking variable from the first equation. Our results suggest that, after accounting for unobserved constitutional factors as well as other observed covariates such as obesity, salt intake, schooling and so on, the strength of the statistical association between high blood pressure and heavy drinking in men drops only slightly by 8% (diastolic) and 23% (systolic). Thus, a strong statistically significant result remains after removing the unobserved variables bias.