Download presentation
Presentation is loading. Please wait.
Published byScott Newman Modified over 9 years ago
1
Statistical Discrimination David L. Dickinson Appalachian State University April 2006: GATE.
2
“Statistical Discrimination in Labor Markets: An Experimental Analysis.” David L. Dickinson and Ronald L. Oaxaca working paper, 2005. Motivation of paper is to look at second- moment distributional discrimination. Productivity distributions have same mean, but variance or other ‘risk’ characteristic may differ. Distributional variable Support of distribution Probability of less-than-average payoffs Note: these measures of ‘risk’ should not matter to risk-neutral decision-makers
3
Motivation Statistical discrimination is accepted/legal in insurance and credit markets, but it is illegal/controversial in labor markets or with racial profiling. Existence of any type of statistical discrimination will bias attempts to measure prejudiced-based discrimination. This research, in general, examines decision- making under uncertainty.
4
Experimental Design Double-sided oral auction design similar to Smith (1965) DOA has strong convergence to equilibrium, in general Productivity draw is independent of physical worker subject We use labor market context, a trade-off for clarity. Treatments
5
Results Employers pay lower wages in all uncertainty treatments (primarily the males) (Table 2) Note from Table 1 that treatments do not control for single variable changes Table 3 shows more controlled regression analysis. Loss Prob matters most (for males) Females actually pay higher wages the larger is the distribution support (?) Female aversion to competitive negotiations does not seem to explain this (though there is evidence that single-gender contracts occur more frequently than mixed-gender contract.
6
Summary Evidence is found that distributional risk affects experimental wage contracts Effect is largely due to male employer sub-sample. Most significant, in terms of magnitude, variable is less- than-expected profits…consistent with loss-aversion For full sample, evidence of discrimination based on distributional risk is significant Ignoring second-moment statistical discrimination will bias estimates of prejudice-based and other statistically- based discrimination
Similar presentations
© 2024 SlidePlayer.com. Inc.
All rights reserved.