computer repair insurance

Repair / December 31, 2021

You might not guess that economists spend a whole lot of time thinking about auto mechanics, but they actually have a technical term for the service that the mechanics provide: credence goods. Credence goods are typically services where the person who receives the service is incapable of evaluating whether the service was provided quickly and efficiently or even if the needed service was provided at all.

These types of services aren't limited to auto mechanics—other examples include medical care and cab rides in unfamiliar cities. Economists find these services interesting because we can learn from them about human financial interactions in cases where one of the parties might have an incentive to be dishonest.

A study of Austrian computer repair shops shows that plenty of said parties succumb to that incentive. The study also shows a common feature of modern economics that makes matters worse: insurance that covers the cost of the repairs.

The researchers behind the new work, based at the University of Innsbruck, came up with a study design that's ingenious in its simplicity. They bought a series of identical refurbished computers and disabled one of the two RAM chips in them. With the operating system they had installed (Windows 7 Pro), this created a self-diagnostic message on boot that correctly identified a memory problem. They then took these computers into randomly selected repair shops throughout Austria.

The person that brought the computer in had a set script for their interactions with the shop staff: “When starting my computer an error message appears and I am not able to boot the computer. I have no idea what this means and I would like you to repair it, please.” The person then said they'd like an itemized bill. Half the time, however, they added that they needed the bill because the repairs would be insured.

Insurance creates what economists call a moral hazard: neither of the parties involved in the repair have any incentives to minimize its cost. So, if anything, it's likely to make repair shops behave even less ethically.

Of the 61 shops they went into, three of them said that the computer couldn't be repaired or that the repairs were so expensive it made more sense to buy a new laptop. It's not clear whether this was incompetence or an attempt to get the person with the damaged computer to buy a new one in the same shop; in either case, these three were excluded from the analysis. Still, the authors note the one that thought the customer was uninsured did the diagnosis for free; the other two both charged over €50 for the diagnosis.

Another four shops billed for parts they didn't install, but these were evenly distributed between the experimental and control groups.

There were five cases where a shop billed for a repair that wasn't needed (the authors termed this "overprovisioning"). This never happened in the control group, where insurance was never mentioned. This overprovisioning helped explain part of the largest effect the authors saw: while uninsured repairs cost an average of €70, insured ones averaged nearly €130, a difference that was highly statistically significant.

This difference in price is largely the product of billing for hours involved in repair. For shops that thought the customer was uninsured, repairs were billed as taking an average of a half hour less than the insured cases. Charges for this extra labor accounted for 70 percent of the price difference in repairs.

While the authors can't tell for sure why this overbilling happened, they actually went back to shops and told staff members of their findings, and they asked what the employees thought might cause this behavior. The explanation favored by the shop employees themselves is simply that everyone assumed that an insured customer would be unconcerned about the price they paid for repairs.

A couple of very obvious cautions apply here: the study was done in Austria, and it involved a relatively small number of shops. But the effect was so large that it's unlikely to be some sort of statistical fluke. The key questions, really, are how widespread this behavior is geographically and what other credence goods it applies to.