Non-Equivalent Research Question 2

Non-Equivalent Control Group Question #2

The Division of Motor Vehicles has some flexibility in what state to choose as a comparison. Remember the principle that you want the two groups to be comparable on all variables other than the independent variable that defines the group.

What variables would you think are important? State size, density of the population, average age of the population, weather conditions, quality of roadways, number of police officers and the history of enforcement of drunk driving laws, and drinking age could all be important. It is unlikely that any one state will be well matched on all of these variables, so it may be advisable to use two or three states to provide control for the many possible sources of confounding.

Another variable that must be considered is the possibility that the changes in the penalties for drunk driving may be accompanied by other changes that confound the results. For example, the police may be encouraged by localities to be more on the lookout for drunk drivers. Some police officers may be less willing to write drunk driving tickets if they feel that the consequences are too extreme. Others might be more willing because they feel that their efforts will now make a bigger difference in the safety of the highways. How might you control some of these possible sources of confounding?