The argument over equity in salaries between males and females is a debate that lingers on in the modern era. Changes in legislation are designed to ensure that racial, gender and potential ethnic inequalities are addressed by employers who must do self-evaluation to ensure fairness.
Multiple regression is still favored by many companies and although it has its own set of complications as a subjective element can never be overlooked, it is a preferred method when recruiting personnel or measuring and predicting. It allows companies to perform more realistic self-evaluation which cannot necessarily be said of other methods.
For correlation to work, there would need to be a cross-sectional study and accurate comparisons rely on males and females who have similar backgrounds, education levels, and so on. Variables that cannot be measured because they cannot be observed, cause discrepancies and create bias with this method and self-evaluation would need to take place based on the observed external characteristics of a general study. Context can then become a problem.
There are various T-tests that can be performed to measure differences between male and female salaries. Confusion can arise and experienced statiticians need to be involved when using T-tests to evaluate differences. It can be technical and somewhat time-consuming, eliminating it from some smaller businesses as a chosen method of evaluation.
Cross tabulation methods do allow for various outcomes from the analysis of data but they can be misleading as factors not considered in measuring the data, which are important factors, therefore do not contribute to an unbiased review.
To answer the question, it seems that multiple regression is preferred by the largest range of companies in evaluating differences between male and female salaries.