Question

Studenmund Using Econometrics Chapter 4, page 113 Problem 6 -c The anwser is given at the end of book but I would much appreciate if you coud state whether it is true or not and if it is true why? The answer is given as follows :"The coefficient of ADVICE implies that an individual will drink 11.36 more drinks, holding constant the other independent variables in the equation, if a physician advises them to cut back on drinking alcohol. This coefficient certainly has an unexpected sign! Our guess is that DRINKS and ADVICE are simultaneously determined, since a physician is more likely to advise an individual to cut back on their drinking if that individual is drinking quite a bit. As a result, this equation almost surely violates Classical Assumption III. "

Answer 1

The coefficient of advice should have been negative as if the physician advises them to cut back on drinking alcohol, the individual would cut back on drinks as expected. But as this is not the case, it is a clear case of violation of the classical assumption III which states that there might be simultaneity between the independent variable (advice) and dependent variable (drinks) which leads to correlation of the independent variable with the error term of the linear regression equation. There is definitely simultaniety present in the experiment as those are advised who are drinking more. The experiment might not be collecting data points after a month if the no. Of drinks have reduced which will be explained by a matched pair test. The error term should be random in OLS equation and the independent variable should not be able to predict it. In this case, the OLS model incorrectly attributes some of the variance of the error term in the model to the "advice" independent variable which makes the coefficient absurd.