Memo to Elon and Vivek: What Are The Incentives That Drive Inefficient Government?
It’s tempting, especially for a libertarian-verging-on-anarchist like me, to just assume government functionaries are inherently bad people who like doing bad things to good people.
It’s not really true — in general, people in government are doing what they think is good. So why does it seem like it’s true? Why does it so often seem that what government does is at best useless, often counterproductive, and sometimes actively harmful?
Elon Musk posted something on X a few days ago that struck me: incentives drive outcomes.
In psychology, this is called “operant conditioning.” Put simply, it’s that a behavior that gets rewarded, gets repeated.
It’s useful, though, to analyze an existing situation going the other direction: ask what the incentives are that drive a particular behavior.
This is actually a callback to my previous memo to Musk and Ramaswamy. The short version is that government suffers from having vague, unstated, or even contradictory objectives, and no measures of success that can be verified.
As commenters pointed out, often objectives and measures thoughtlessly defined can be counter-productive. One example was customer service reps in a call center. If the number of calls handled is the measure of a CSR's production, they will optimize to shunt difficult calls aside to even hang up on them. This results in good calls-per-hour, but bad customer satisfaction. Or in my own field of software engineering, a common measure used for the "size" of a piece of code is the number of source lines composing the code — "source lines of code" or SLOC. This is a surprisingly good measure of the size of a product — as long as the programmers aren't being rewarded based on something like lines of code an hour. But I had a consulting engagement to try to figure out what had gone wrong with a project being outsourced to a firm offshore, and lines of code per day were a major way in which productivity was being measured.
By the time I was investigating, the project included more than a million lines of code, and a concomitant number of hours billed. On review, I saw one example of code that was more than 600 lines as delivered. I wasn't supposed to rewrite anything but this so offended me that I rewrote it on my own time, and reduced 600+ lines to 10.
In Musk's example, the project was to "save the homeless." Yet, year after year, there were more homeless to be "saved," which meant a bigger budget, more staff, and of course higher compensation for the CEO. The one thing that wasn't happening was a reduction in the number of homeless people in homes and out of the program.
But then, that wasn't where the rewards lie.
This paradox of incentives shows up in government over and over. FEMA brags about the number of temporary housing units it can supply — whether or not disaster victims are even using them or can use them. Amish volunteers build many "tiny homes" in Western North Carolina — that can't be occupied because they didn't get the proper permits. Defense contracts are optimized not to delivering the best widgets, but to seeing that the widgets are produced in as many Congressional districts, with as many jobs as possible. Small contracting firms with the "right" demographics — number or women, number of favored minorities — get contracts that are subcontracted to big firms that can actually do the work.
And the crazy thing is that no one involved is cackling and twirling their mustache like Snidely Whiplash. All of them are doing what they're being told to do by the incentives as they actually stand.
If you can find the incentives that reward a behavior you want to change, you can change the incentives to direct the behavior to a more desirable direction. If you try to impose a change in behavior without understanding the incentives, then you constantly fight what can seem like pushback, even malicious interference.
I suspect this is actually the biggest challenge to DOGE: finding a way to reorient the incentives so that government is actually progressing toward a desired goal beyond the number of employees needed and power point decks presented.
No comments:
Post a Comment