Panel Paper:
Sexual Orientation and Earnings Differentials in the United States
*Names in bold indicate Presenter
In this paper, Ordinary Least Squares (OLS) is used as a “baseline” comparison with the findings of the early literature which uses datasets in which sexual orientation is assigned or assumed for individuals (and couples). As for improved data treatment and technique: a Heckman selection technique is used to improve upon OLS as the dependent variable, earnings, is likely censored (there exist individuals in the data for which earnings are zero) which can lead to bias in the OLS results. Finally, quantile regression analysis is utilized to examine where along the income distribution these differences in earnings may occur; the usage of quantile regression considers any gap (earnings differentials) found as not constant or uniform along the income distribution, thereby permitting the identification of larger or smaller gaps at specific quantiles.
The motivation behind this work is to improve upon previous findings in the literature in order to produce higher quality evidence for consideration by policy-makers. These improvements to the literature are important to validate the previous penalty and premium findings by exploiting more appropriate data and estimation techniques. If earnings differentials are found, and we expect a gap to exist, these differentials may be attributed to discrimination and this discrimination could be preventable with policies that promote equality. In addition, it is possible that the ‘averaging’ effects of previous studies has masked differentials for previously ignored sexual minorities of bisexual and queer identifying individuals, it is important to include these oft invisible groups to bring to light they plights they may face. Finally, previous studies may have also “averaged” out earnings differentials at specific quantiles of the income distribution (for example at very low wages or very high wages the differential may be different than at a median wage). This research design will eliminate those shortfalls, thus producing more effective and efficient policy recommendations.