Drinking, Alcohol Policy, and Attitudes Toward a Campus Riot

Journal of College Student Development, Sep/Oct 2004 by Kaplowitz, Stan A, Campo, Shelly

Other responses suggest a less negative attitude. Views were evenly split on the assertions that "the riots are a result of the university administration and the city of East Lansing restricting student rights." Moreover, over 56% agreed that "the University has no right to restrict drinking at Munn Field," and that "the University is trying to control student drinking too much."

Finally for our measures of Enjoyment, 49.2% got close enough to see the riot and 37.8% said they thought it might be fun to be part of it.

Predicting Riot Attitudes

We tested and estimated some predictive causal models, via Structural Equation Modeling (SEM) using AMOS 5. SEM is well suited for a multiple-equation model in which some of the (latent) variables have multiple indicators (Bollen, 1989). Its advantages include the ability to correct for attenuation due to unreliability (see DeShon, 1997) and to estimate total effects.

Omitting cases. We eliminated those (less than 4%) who did not classify themselves as either White, Black or Asian, because these other groups contained too few respondents for us to assess their effects. We also eliminated those (less than 3% of our sample) who reported living with their families. The SEM analyses were based on 1,824 cases.

Confirmatory Factor Analysis. We first tested our measurement model with a confirmatory factor analysis on all latent variables having two or more observed indicators: satisfaction with professors, alcohol consumption, objecting to restrictions, condoning the riot, perceiving negative consequences of the riot, and enjoying the riot (see Table 1). The model fit very well: The Comparative Fit Index (CFI) was .953 and the RMSEA, which takes parsimony into account, was .062. The median absolute value of the standardized loadings was .74 and the lowest was .56. In contrast, other possible measurement models, such as one in which all observed riot attitude items were indicators of one latent variable, fit substantially less well. We also found that the measurement model would fit considerably better if two of the indicators of condone, "The riots are a result of the university administration restricting students' rights," and "The riots are a result of the city of East Lansing restricting students' rights," were assumed to have correlated errors of measurement.

The Full Causal Model. Because of time ordering, we assumed the causal ordering shown in Figure 1. The socio-demographic variables were clearly determined before the campus experiences. In contrast, attitude toward alcohol restrictions were affected by recent experiences with those who enforce these restrictions. Finally, attitudes toward the riot may not have been formed until it was clear that the riot was likely. In addition, the pattern of correlations among (a) Alcohol Consumption, (b) Objecting to Restrictions, and (c) Condoning, clearly suggested a causal chain with the ordering given above.

Three other features of the model should be noted. First, for completeness, the model estimates all of the path coefficients between a latent variable and any other latent variable that is causally later in the model. second, all three riot attitudes were assumed to have correlated errors of prediction. This is because we assumed that these attitudes are an inter-related system and influence each other (Eagly & Chaiken, 1993). Third, the campus experiences were assumed to affect each other, but as with riot attitudes, we cannot definitively specify the direction of causation. For example, one's residence can affect one's drinking habits, but one can also select a residence based on its compatibility with one's inclinations. Drinking may affect one's GPA, but one's academic involvement may also affect how much one will drink.