Driving home in my beat-up Honda last week, I listened to an interesting story  on NPR on why good people do bad things. One particularly interesting paper  being discussed was by Francesca Gino and Lamar Pierce on how those folks who test your car emissions may be favoring folks who are more like them – malfeasance but with a potentially social dimension.
The paper goes something like this. They use data from emissions testing from urban areas in “a large northern state.” They have a range of variables – inspector ID numbers, car characteristics, facility information, and time, date and emissions testing results for all cars tested between 2001 and 2004.
Now, it turns out that it’s fairly easy for the folks doing these tests to rig things so that your car passes. And this is not an uncommon event – one study (using covert audits) found that in 10 percent of cases, technicians were testing cars other than the one that was supposed to be tested (apparently it is also possible to rig the equipment as well). Hypothesizing that testing employees are more likely to help someone of their own income bracket, Gino and Pierce use the data from their chosen state to see if this pattern shows up in the data.
They split cars into luxury and standard by make. Anything older than 10 years was automatically classified as standard, “under the premise that a 12-year-old BMW would not be a clear indication of wealth.” Then they take inspectors who have a lot of inspections (more than 3500) so they get enough variation in car type. And then they look for systematically high pass rates and test (with a Wald test) for differences in the pass rate across types of cars by inspector.
Amongst their 249 inspectors, they find a group of people – “Robin Hood discriminators” -- who systematically favor (i.e. pass at a rate that is high) standard cars (i.e. cars that the inspectors, given their wages are likely to drive). There is also a smaller group of people who tend to favor luxury cars (“luxury helpers”).
Of course, this could just all be noise in the data. So what they do is randomly reassign cars to each inspector to create a counterfactual distribution. These simulation results don’t result in a reproduction of much of the luxury helpers, but do reproduce well the Robin Hood discriminators – suggesting that most of the luxury helpers they find are likely randomly generated by the data. Bottom line, with the most conservative confidence levels, they conclude that there are 18 Robin Hoods and 9 luxury helpers out of their 249 inspectors – not a trivial amount. Finally, as a robustness check, they also track a sub-sample of inspectors who change the facilities they work in, and these inspectors’ discrimination appears to be consistent across jobs.
So what’s going on? Two big explanations present themselves. First, these folks have empathy for those who drive standard cars, and illegally tilt things in their favor. Second, given that inspections in this state are done by private contractors (who may be mechanics) there might be some financial gain to be had (and they aren’t powered for facility fixed effects). So it’s into the lab they go -- this is the rare two-for-one paper.
In the lab, they run an experiment with students where they run a scenario that basically involves a fellow student asking to borrow a parking pass for the day, an act which is specifically against the rules. So again, we have malfeasance (duly recognized as such by the participants), and they run a 2 car type (standard/luxury) by 2 gender by 2 race (white and Asian – the main groups amongst their study population) design. Participants were significantly more likely to give their pass over to the fictional student when the car was standard and when the fictional student was female. They then use other questions to tease out the potential motivation for this behavior, and the results seem to indicate that feelings of empathy and envy are at play (interestingly, same sex and same race identification don’t seem to matter).
From a development view point, these results suggest that when we look at the malfunctioning of service providers and bureaucracies, we should also consider the relative role of non-pecuniary related malfeasance and, in particular, the roles of empathy and envy. We also might want to think about how social connections might play into generating/strengthening this empathy (or envy). Of course, there are other questions at play here – how much of the malfeasance we observe in service providers and bureaucracies is due to this, and how much due to sheer financial motives? How does this play out in environments where the wealth differential is more extreme? Are there any papers out there (in development that is – there’s a fair literature in the US) to add to this discussion?