5 years ago

Anthony P. Carnevale Stephen J. Rose Andrew. R. Hanson

Anthony P. Carnevale Stephen J. Rose Andrew. R. Hanson


Appendix B: REGRESSION ANALYSES OF EARNINGS (SIPP AND NLSY) The previous tables demonstrate the difference in earnings between certificate holders and workers with a high school diploma but no postsecondary education. However, in isolated cases, this approach is not accurate because of unusual factors. For this reason, researchers have refined a more robust method for determining earnings differences by education level: multivariate regression analysis. To demonstrate that the results presented above are accurate and not influenced by any unusual factors, these are the results using regression analysis. These results are nearly identical to the other data presented in the text. The standard approach is to use the log of earnings and adjust for demographic differences, experience, and indicators of educational attainment: a series of zero or one “dummy” variables. The coefficients presented in regressions represent differences from the omitted variable. For example, in regressions with all workers, the variable “female” shows how much less women make than men after adjusting for educational attainment and age. In a similar fashion, the race/ethnicity variables represent the difference from white workers. Finally, the comparison group for the education variables is those with a high school diploma and no postsecondary education. Regression analysis also differs from comparisons based on tabular results because there is a test of “statistical significance” of how accurate the estimated effect is. In general, researchers say that a result is statistically significant if the probability value that the coefficient is different from zero at the 95 percent level of accuracy. Consequently, in all of the tables presented below, this probability factor is included and these results are very robust because in most cases this probability is greater than 99.9 percent—the “

42 Table A1: Regression analyses, SIPP 2004/2008 All workers Male workers Female workers Variable Coefficient Probability Coefficient Probability Coefficient Probability Female -0.489

Anthony P. Carnevale Stephen J. Rose Andrew ... - Inside Higher Ed
In J. E. Moody, S. J. Hanson, & R. P. Lippmann (eds.)