Drug testing and discrimination

If employers assume that minorities are more likely to take drugs than are other groups, and if you think that drug-taking makes for a worse employee, they could be fuelling a really bad equilibrium. Individuals can't easily signal that they're clean without simultaneously seeming weird (by, for example, voluntarily offering to undergo drug testing), so they pool with drug users and are less likely to get a job offer.

This happens whether or not the employer is right about the relative rates of drug use. If the employer is wrong, that employer will bear some losses by failing to hire quality candidates, and especially when labour markets are tighter. But, worse, initial misperceptions on relative drug use can turn into a bad self-reinforcing equilibrium: if everyone is going to think you're a stoner regardless of whether you do drugs, why not at least enjoy the consumption benefits? We then get the same kind of pernicious equilibrium that Glenn Loury talks about in his work on discrimination.

One way of breaking that equilibrium is to allow employers to require drug testing: it lets clean employees quickly demonstrate that they're clean and consequently breaks the pooling equilibrium. Drug-using potential employees are worse off, because they can't mask type as easily, but those who aren't benefit by the separation.

That's the theory. Abigail Wozniak shows that the empirics hold up. Here's an early ungated version of her paper; here's the NBER Working Paper.

Nearly half of U.S. employers test job applicants and workers for drugs. A common assumption is that the rise of drug testing must have had negative consequences for black employment. However, the rise of employer drug testing may have benefited African-Americans by enabling non-using blacks to prove their status to employers. I use variation in the timing and nature of drug testing regulation to identify the impacts of testing on black hiring. Black employment in the testing sector is suppressed in the absence of testing, a finding which is consistent with ex ante discrimination on the basis of drug use perceptions. Adoption of pro-testing legislation increases black employment in the testing sector by 7-30% and relative wages by 1.4-13.0%, with the largest shifts among low skilled black men. Results further suggest that employers substitute white women for blacks in the absence of testing.

It feels good to oppose mandatory testing, thinking that it somehow is nicer for minority groups. But that opposition mostly works to encourage employers to substitute over to groups that they believe are less likely to take drugs. Another for the continuing series here on feeling good while doing harm.

I note this now because Twitter tells me that Radio New Zealand will be talking drug testing in 15 minutes.

Previously:

• Feeling good, doing harm
• Feeling good, doing harm: aid edition
• Hunt it to save it
• Competing for aid

Academic discrimination audits

Profs in fields where the opportunity cost of time are more valuable use tighter filters in deciding whether to reply to an email; that filter may have discriminatory effect.

Milkman et al run a field experiment sending academics in a pile of fields the following email:

Dear Professor [Surname of Professor Inserted Here],

I am writing you because I am a prospective doctoral student with considerable interest in your research. My plan is to apply to doctoral programs this coming fall, and I am eager to learn as much as I can about research opportunities in the meantime.

I will be on campus today/[next Monday], and although I know it is short notice, I was
wondering if you might have 10 minutes when you would be willing to meet with me to briefly talk about your work and any possible opportunities for me to get involved in your research. Any time that would be convenient for you would be fine with me, as meeting with you is my first priority during this campus visit.

Thank you in advance for your consideration.

Sincerely,
[Student’s Full Name Inserted Here] 

If you're an academic, you're probably getting pretty frequent form-letter-style emails from students wishing graduate supervision. You're probably also getting a billion other emails - you need a way to triage. 

I'm not particularly proud of it, but there's one tranche that just gets the delete button: emails sent that are clearly form letters of the following type:

Dear Professor (name in different font)",

I am writing seeking supervision in [field in different font, or named field far different from anything I research] at [University name in different font]. 

[Describes work experience tangentially related to proposed field of study, or a research project coming from his/her undergraduate study. Or suggests willingness to write a thesis in anything, if I have scholarship money. Generally reveals no understanding of what I'm interested in, no particular interest in Canterbury over other places, and provides no expectation that an answer's really expected - zero investment in this application over others. It's clearly a mass-broadcast distribution.]

There are many many potential students in many countries who seem more interested in getting a student visa out than really in pursuing a thesis. My replying would waste everyone's time, as, until recently, I had no potential funding for graduate students, and none of these students could go anywhere without funding.

I don't get dozens of these emails a day; more likely a dozen or more a month. I imagine that if I were at a top US school, I'd be getting rather more of them. I'd definitely reply to a letter like the one they sent out, because very few of the academic spam emails I get ever suggest a personal meeting: if a potential student would take the time to get to Christchurch, why wouldn't I at least reply?

But then I imagine what it would be like if I were in a city of several million people, with loads of internationals students in town seeking extensions of their undergraduate student visas. Signing up at grad school extends the visa: the prof needs some way of distinguishing emails from people who are actually interested in pursuing study, and those for whom it's just a way of extending the visa.

Anybody who really had come all the way to my town and wanted desperately to meet with me would find some way of signalling that in the intro email. Like "I liked your work on [fill in the appropriate blank] and want to extend it [proposes some relevant extension]." Not "I have considerable interest in your research."

Based ONLY on time constraints and visa-seeking behaviour, I would expect the following:

• The greater the number of these solicitations received, by field, the more likely there'll be no reply;
• Fields more likely to have scholarship money or RA work will get more such solicitations;
• Profs in fields getting more of these kinds of solicitations will use tighter filters in deciding what to delete;
• One not-nice but non-crazy filter is to downweight emails from foreign-sounding names where the correspondent reveals zero real knowledge of the prof's work or field. The slightest signal that this indicates actual interest in the prof's work would go a long way. Otherwise, a Bayesian reckons the person sending the email either doesn't know enough to signal an actual knowledge of the prof's work (a bad signal in itself), or really is more interested in something other than the prof's work - like a visa. We should expect then no particular gender effects, no particular difference between Caucasian and Black names, but a difference between those and ones that have a higher probability of being a student on international visa. 

Here's what the paper finds:

Here is the rank order of the gap, on average across all disciplines:

• Chinese female at Private School: -29%
• Indian Male, Private School: -21%
• Indian Female, Private School: -19%
• Hispanic Male, Private School: -18%
• Chinese Female, Public School: -17%
• Black Female, Private School: -16%
• Black Male, Private School: -14%
• Indian Male, Public School: -13%
• Chinese Male, Private School: -11%
• Chinese Male, Public School -10%

All others below 10%. 

If all that were going on were filtering based on "weed out students who have indicated no interest in my research and who are more likely to be international", I'd have expected Chinese males to have fared worse than Blacks or Hispanics; I doubt that my story is all that's going on. I do expect that it's an important part of what's going on, but only part of it. In particular, I can't see how this story explains within-race gender gaps. But I do note that the only substantial discrimination among those at public universities was against foreign-sounding names. 

I wonder what would have happened if the field experimenters had spent a couple minutes, when getting the contact and other details for each prof in the study, also getting the abstract from that prof's most recent paper and saying something about it in the letter. If my story's explaining part of what's here going on, then the gap between foreign-sounding and domestic-sounding names should drop. 

Finally, I always wonder about these kinds of studies and whether race is the only thing signalled. Would the results have been the same if one of the white-sounding names were Cleetus?

Gender quotas

Annick Masselot of Canterbury's Management department wants gender quotas on boards of companies listed on the NZX.

UC associate professor Annick Masselot said the NZX should go further with the introduction of quotas.

"In fact I think they have a duty to do that . . . Diversity is not an option."

She said NZX's diversity rules should at least match rules from the Australian sharemarket that required companies to disclose whether they had a formal diversity policy, and "if not, why not".

Masselot said several studies found that having women on boards led to an improvement in financial metrics.

Countries including Norway had introduced gender quotas in company boards using hard legislative measures. Others, including Australia had increased the number of women on boards with the use of "soft" self-regulation, she said.

The NZX would have the power to introduce gender quotas, but did not consider it appropriate to introduce them and had no intention to do so, a spokeswoman said.

I wouldn't be surprised if there were some studies finding benefits, but I doubt that they'd be sufficient to overturn the general conclusion that quotas aren't necessarily that hot an idea.

Adams and Ferreira, in the Journal of Financial Economics, found that mandating gender quotas for directors reduces firm value for well-governed firms. Forcing Boards to take on more women can help performance where those firms had existing problems that could be solved by greater Board monitoring. But on average, greater diversity yielded worse firm performance in the set of S&P firms (500, MidCaps and SmallCaps) studied.

Given that our previous findings suggest that more gender-diverse boards have stronger governance, these results imply that, on average, tough boards do not improve firm value. But they do not imply that tough boards never add value. There is no reason to expect tough boards to add value in all firms. The value of a tough board should depend on the strength of the other governance mechanisms. If firms have otherwise strong governance, having a tough board could lead to overmonitoring. But if firms have otherwise weak governance, we would expect tough boards to be particularly valuable.

I read this as a strong argument for shareholders' demanding greater female board representation if they think the Board has governance issues. But as for quotas:

Our results highlight the importance of trying to address the endogeneity of gender diversity in performance regressions. Although a positive relation between gender diversity in the boardroom and firm performance is often cited in the popular press, it is not robust to any of our methods of addressing the endogeneity of gender diversity. The true relation between gender diversity and firm performance appears to be more complex. We find that diversity has a positive impact on performance in firms that otherwise have weak governance, as measured by their abilities to resist takeovers. In firms with strong governance, however, enforcing gender quotas in the boardroom could ultimately decrease shareholder value. One possible explanation is that greater gender diversity could lead to overmonitoring in those firms.

More generally, our results show that female directors have a substantial and value-relevant impact on board structure. But this evidence does not provide support for quota-based policy initiatives. No evidence suggests that such policies would improve firm performance on average. Proposals for regulations enforcing quotas for women on boards must then be motivated by reasons other than improvements in governance and firm performance.

Nominalism

An Eleanor by any other name might be as smart, but would have worse odds of winding up at Oxford. A snippet from Greg Clark's work on names and surnames:*

It isn't that the name itself is particularly powerful, or destructive; rather, the sorts of people who wind up naming their daughters Eleanor are very different, on average, from the sorts of people who wind up naming their daughters Kayleigh. From the BBC piece:

However, there is no evidence that it's the names causing such a marked discrepancy, rather than other factors they represent, Clark says. Different names are popular among different social classes, and these groups have different opportunities and goals.

"That's something that's emerged in modern England that didn't exist around 1800," he says. When he re-ran his study, but this time looking at students attending Oxford and Cambridge in the early 19th Century, he found the correlation between names and university attendance far less marked. First names simply weren't the social signifiers they are now.

What's happened since then is a move towards unusual, even unique, names. Before 1800, Clark says, four first names referred to half of all English men. In 2012, according to the Office for National Statistics, the top four names (Harry, Oliver, Jack, Charlie) accounted for just 7% of English baby boys (and the picture was much the same in Wales).

Similarly in the US, in 1950, 5% of US parents chose a name for their child that wasn't in the top 1,000 names. In 2012, that figure was up to 27%.

As late as the 18th Century, it wasn't uncommon for parents to call multiple children the same name - two Johns for different grandfathers, for example. Now parents increasingly look for unique names or spellings of names. As Jean Twenge points out in her book the Narcissism Epidemic, Jasmine now rubs shoulders in naming lists with Jazmine, Jazmyne, Jazzmin, Jazzmine, Jasmina, Jazmyn, Jasmin, and Jasmyn.

Our baby-naming algorithm, which wound up choosing Eleanor for our girl, started with a domain of names in the top 1000 from 1880 to 1930 in the US, then eliminated any names that were in the top-100 any time in the last three decades. So you rule out the made up stuff while also ruling out having too many classmates with the same name. But we're the kind of people who would choose that kind of algorithm.

I especially liked the BBC piece's pointing to Figlio's work on names. It's pretty common for researchers to use audit studies to test for racial discrimination: they send out otherwise-equivalent vitae, but put a 'black-sounding' name on some and a neutral one on others. They tend to find a penalty for having a black-sounding name. The problem with these kinds of studies is that they simultaneously test for black-sounding and for class-linked names: instead of testing black versus white, they're effectively testing lower-class versus upper-class names. Here's Figlio:

This paper investigates the question of whether teachers treat children differentially on the basis of factors other than observed ability, and whether this differential treatment in turn translates into differences in student outcomes. I suggest that teachers may use a child's name as a signal of unobserved parental contributions to that child's education, and expect less from children with names that "sound" like they were given by uneducated parents. These names, empirically, are given most frequently by Blacks, but they are also given by White and Hispanic parents as well. I utilize a detailed dataset from a large Florida school district to directly test the hypothesis that teachers and school administrators expect less on average of children with names associated with low socio-economic status, and these diminished expectations in turn lead to reduced student cognitive performance. Comparing pairs of siblings, I find that teachers tend to treat children differently depending on their names, and that these same patterns apparently translate into large differences in test scores.

I wish that audit studies instead used a cross-cutting method that would test Da'Quan against Dwayne and Cleetus instead of just against Peter. Put it into a 2x2 matrix with race on one axis and class on the other.

First names may just be signals, but "just signalling" hardly means unimportant. If you're economising on time while going through the resumes or university applications, and if a Cleetus is less likely than a Peter to know which fork to use at the important dinner with a client or donor, well, a Cleetus would have to be better than a Peter to get a second look. Because parents who have some minimal social capital expect this to be at least somewhat likely to be true, they make different naming choices than those who don't.

Colby Cosh is right:

The parents who need Crampton’s naming algorithm most will give up long before they hit “Borda count”.
— Colby Cosh (@colbycosh) April 14, 2014

The separating equilibrium separates.

I'd tell you to avoid giving your kids dumb names, but if you're reading this, you almost certainly already know it.

* I really need to get his book. The interview teasers have been very tempting.

Previously:

• Class Etiquette
• Time to re-run the algorithm