The results of this intervention, described in detail in the preceding section, show that a high-intensity social norms campaign improved the accuracy of the target audience's perceived norms and increased its reported healthy, preventative attitudes and behaviors regarding driving after drinking. Compared to changes in the 21-to-34-year-old adults in the control counties, statistically significant differences resulted in the targeted counties (see table 4) demonstrating:

• a 24.8-percent relative increase in recall of campaign messages about the majority norms regarding not engaging in impaired driving (p<=.001);
  
  
• a 7.5-percent relative decrease in the percentage who believed that the average Montanan their age drove after drinking during the previous month (p<=.05);
  
  
• an 11.0-percent relative increase in the percentage who accurately perceived the majority of their peers use a non-drinking designated driver (marginal significance at p = .056);
  
  
• a 13.7-percent relative decrease in the percentage who reported personally driving after drinking (p<=.05);
  
  
• a 15.0-percent relative increase in the percentage who reported using non-drinking designated drivers (p<=.05);
  
  
• a 16.5-percent relative increase in the percentage who would support passing a law to decrease the BAC legal limit for driving to .08 (p<=.01).

Limitations of this data must be acknowledged, however. These results rely solely on the data reported by respondents to the four telephone surveys. Corroboration of these results through changes in the BAC of arrested drivers or numbers of alcohol-related fatalities was not possible, for several reasons.

Montana maintains no statewide database on the BAC of arrested drivers. The city police departments, county sheriffs offices and the Montana Highway Patrol all collect this information independently, without standardized methods of reporting. Collecting and comparing these data would be a years-long project in itself.

In terms of tracking changes in the fatal alcohol-related crashes, because Montana has very high vehicle miles traveled (VMT), a small population, and a correspondingly low total number of crashes and fatal crashes, documenting a statistically significant change is extremely difficult even over a period of several years. Year to year comparisons are not always reliable; with the small number of total crashes, very few incidents can result in large fluctuations in the data. To complicate matters further, State crash databases fail to record driver county of residence, making it impossible to determine whether drivers involved in crashes had been exposed to the media intervention.

This table shows how likely it is that the changes in perception, reported behavior, and support for policies are a result of the implementation of the intervention (social norm campaign) as indicated by the calculation of the p-value (p<=). The table compares survey responses in the Western Counties where the implementation of the intervention took place to the Eastern Counties in which there was no implementation of the intervention (social norm campaign).

The difference of change is calculated by subtracting the change observed in the control counties from the change in the intervention counties. This measure adjusts for change across seasons and years not related to the intervention. Thus, if the change in control counties was a decrease, while the change in the intervention counties was an increase, the overall impact of the intervention is even greater than the simple increase observed in the intervention counties alone.

The p-value, a statistical calculation, determines the likelihood that the connection observed in the data is due simply to chance. The change is more significant if the p-value is smaller. This table shows that the p-values in each response category are equal to or less than 0.05 (a less- than-five-percent likelihood that the results are not by chance). The changes are significant.

For the social norms campaign to be effective in the intervention counties, the data would show an increase in recalling the media messages, decrease in perception of public drinking after driving, decrease in reported drinking after driving, and a greater support for policy. In this instance, all of the changes in intervention counties showed significant change in the desired direction. However, much of the responses showed the opposite effect in the control sites. Possible explanations for these changes are contained throughout the document.