Methodology

The survey sample of 14,174 was selected in systematic fashion by the American Culinary Federation and Readex Research from the American Culinary Federation’s active, domestic membership with email addresses on file, representing 14,238 members (or 63% of the American Culinary Federation's entire 22,767 membership) at the time of sample selection.

Data was collected via electronic survey from May 29 to June 6, 2008. The survey was closed for tabulation with 3,452 usable responses—a 27% response rate based on the net effective mailout of 12,923. From these, a random sample of 1,000 responses was selected for processing. As with any research, the results should be interpreted with the potential of non-response bias in mind. To ensure representation of the audience of interest, results have been filtered to include only the 887 respondents who indicated they are employed or self-employed full time (35 hours or more per week).

The margin of error for percentages based on 887 usable responses is ±2.9% at the 95% confidence level. That is, 95% of the time we can be confident that percentages in the actual population would not vary by more than this in either direction. The margin of error for percentages based on smaller sample sizes will be larger.

The Compensation Calculator

The analysis of the American Culinary Federation Salary Survey data used multiple regression analysis to model the determinants of salary by identifying those variables which, when taken together with appropriate weights, provide the best prediction of any individual's actual salary.

The final salary prediction model is somewhat restricted in its applicability—it represents only full-time professionals (employed year round) who are under 65 years old with salaries in the range of $25,000 to $130,000. Only those whose salary is at least 70% of their compensation were considered.

Statistically speaking, this model is moderately powerful: it explains 41% of the variation in salary (adjusted R-square = .411), and is significant by the F-test at p < .000.

While a model explaining about 41% of the dependent variable’s variation may be described as “moderately powerful,” it still leaves over half of the variation unexplained. It is virtually certain that other variables not captured through this survey also have an effect on salary levels: individual job performance, for example. To the extent that this model does not include variables actually important in determining salary, its conclusions must be interpreted cautiously.