(Katona and Mueller, 1955). Clearly, nutritional information acquisition can be
influenced by factors which affect diversified consumer segments and households in
different fashions. These factors include time constraints, the perceived role of dietary
intake in maintaining individual health, literacy in English, a rudimentary understanding
of nutrition, and the perceived benefits of nutritional information. These factors also
vary among distinct demographic segments supporting the use of consumer
characteristics in evaluating nutritional label usage.
The logit model was selected for the regression in this analysis because its asymptotic
characteristic constrains the predicted probabilities to a range of zero to one. The logit
model is also favored for its mathematical simplicity and is often used in a setting
where the dependent variable is binary. As the survey utilized in this analysis provided
individual rather than aggregate observations, the estimation method of choice was the
maximum likelihood estimation (Gujarati, 1992). Among the beneficial characteristics
of MLE are that the parameter estimates are consistent and asymptotically efficient
(Pindyck and Rubinfeld, 1991).
The model assumes that the probability of being a frequent user of nutrition labels, Pi,
is dependent on a vector of independent variables (Xij) associated with consumer i and
variable j, and a vector of unknown parameters β. The likelihood of observing the
dependent variable was tested as a function of variables which included socio-
demographic and consumption characteristics.
Pi = F(Zi) = F(α + β Xi) = 1 / [ 1 + exp (-Zi)]