VoxEU Column Labour Markets

Screening technologies and disadvantaged job applicants

There is concern that worker screening may aggravate discrimination against disadvantaged groups, but its effects may be counterintuitive. This column presents recent evidence on the effect of drug testing on hiring practices by race in the US. Removing this channel of information asymmetry allows employers to distinguish workers by this relevant characteristic rather than relying on their priors; the result is that hiring of non-drug-using African-Americans increases, at the expense of white women.

In recent decades, a wave of new technologies has made it possible for employers to screen job applicants – as well as current employees – for a range of past personal behaviours.

  • Rapid and inexpensive physical specimen tests allow employers to quickly screen workers and applicants for recent drug use.
  • A variety of online providers offer quick access to criminal conviction and arrest histories, sex offender status, and credit scores (Finlay 2009).
  • The advent of computerised testing has allowed employers to develop pre-employment skills tests that can be easily administered and quickly scored.

Large US companies were among the first to adopt these technologies, but advancements in the broader testing industry have lowered costs to bring these procedures within reach of smaller companies. The practices are becoming increasingly widespread in Europe (Verstraete 2005).

It is often assumed that such practices further limit the employment prospects of disadvantaged individuals – those from populations with lower average educational attainment and higher rates of interaction with the justice system. However, Autor and Scarborough (2008) point out that the validity of this assumption depends critically on how employers view applicants from disadvantaged groups in the absence of testing. They show that the adoption of skills testing for prospective employees at a retail chain had no impact on the rates of African-American hiring, even though African-Americans had lower average scores on the test than whites. They show that this is consistent with employers already taking average skill differentials into account prior to test availability. Instead of reducing rates of African-American hiring, the skills test allowed the retailer to hire more productive workers of both races while maintaining the pre-testing hiring ratio.

In the US, drug testing pre-dates the other types of screening; it relies on inexpensive lab tests that were developed in the 1960s and 1970s rather than internet-accessed databases which were not widely available prior to the 1990s. Drug testing also connects to the ‘War on Drugs’ policies of the 1980s, which left a large legislative footprint. As a result, drug testing has long been common in the US, and legislation has generated variation in testing prevalence across states and industries. As a result, we can observe the long-run impacts of this policy on the distribution of employment of disadvantaged groups more generally.

By the mid-1990s, nearly half of U.S. employers were testing job applicants and workers for drug use. This rate has since been stable. Because of persistent media and pop culture stereotypes linking African-Americans with drug use, a common assumption is that the rise of drug testing must have had negative consequences for African-American employment.

In a recent paper, I find that, contrary to this expectation, the rise of employer drug testing may have benefited African-Americans (Wozniak, forthcoming). I use variation in the timing and nature of drug testing regulation to identify the impact of testing on African-Amercan hiring. Specifically, some US states enacted legislation promoting wider testing by employers, while others passed legislation restricting testing. Still others passed no restrictions at all. The result is that some states saw testing increase faster and some slower than the overall upward national trend. Similarly, some industries are required by federal law to test applicants and employees, and the concentration of employment varies across cities in a stable way linked to historic local economic development patterns. The result here is that some cities saw larger increases in the likelihood of testing than did others, even within the same state.

I find that after a pro-testing law is passed in a state, African-American employment increases in sectors that have high testing rates (mining, manufacturing, transportation, utilities, and government). These increases are substantial: African-American employment in these industries increases by 7-30%. Because these industries tend to pay wage premia and to have larger firms offering better benefits, African-American wages and benefits coverage also increase. Real wages increase by 1.4-13% relative to whites. The largest shifts in employment and wages occur for low skilled African-American men. I also find suggestive evidence that employers substitute white women for African-Americans in the absence of testing. Gains in hiring African-Americans in these sectors may have come at the expense of women, particularly in states with large African-American populations.

The likely explanation for these findings is that prior to drug testing, employers overestimated African-Americans' drug use relative to whites. Employers do care about the results of the tests: I also find that employment in high-testing industries increases among self-reported non-users after a pro-testing law is introduced. Drug testing therefore appears to detect a characteristic that employers value but which they had a difficult time analyzing – even at the population level – prior to the availability of testing. This is consistent with some evidence that testing for drug use without using physical specimens is typically inaccurate. By enabling non-drug-using African-Americans to prove their status to employers, drug testing improved hiring of African-Americans in the testing sector.

Concluding remarks

As information on job applicants and workers becomes ever more plentiful, firms and policymakers need to consider the consequences of screening. Lessons from the analysis of screening technologies suggest that these impacts are not always easy to predict. Moreover, the impact of a piece of information will depend both on what it communicates and on what employers knew or believed prior to its revelation. As a result, different screening technologies will differ in their impacts, as illustrated by the contrasting impacts of skills testing versus drug testing. The experience of the US with drug testing policy suggests that screening technologies interact in important ways with the legal environment and the wider culture. It is therefore important for all involved to ensure that employers are learning about characteristics relevant for productivity. Finally, although the impacts for disadvantaged populations as a whole may not always be negative, new screening technologies will undoubtedly have negative effects on some disadvantaged individuals. We must therefore ask what can be done for those individuals who are increasingly screened out.


Autor, David H and Scarborough, David (2008), “Does Job Testing Harm Minority Workers? Evidence from Retail Establishments”, Quarterly Journal of Economics.

Finlay, Keith (2009), “Effect of Employer Access to Criminal History Data on the Labor Market Outcomes of Ex-Offenders and Non-Offenders”, in David H. Autor, (ed.), Studies of Labor Market Intermediation, Chicago: University of Chicago Press.

Linden, Leigh and Rockoff, Jonah (2008), “Estimates of the Impact of Crime Risk on Property Values from Megan’s Law”, American Economic Review.

Stoll, Michael A and Shawn Bushway (2008), “Effect of Criminal Background Checks on Hiring Ex-Offenders”, Criminology and Public Policy.

Verstraete, Alain (2005), “Introduction, Fourth Symposium on Workplace Drug Testing, Dublin, June 16-17, 2005”, Forensic Science International 174(2-3): 89.

Wozniak, Abigail (forthcoming) “Discrimination and the Effects of Drug Testing on Black Employment”, Review of Economics and Statistics, forthcoming.

630 Reads