Hanlon’s Razor and the Infuriating H-1B Visa System

tangential thoughts on the WSJ piece about immigrant founders of billion dollar startups

I.

Charlie Munger has this story about how world champion squash player Victor Niederhoffer managed to get A’s at Harvard. Victor liked playing checkers and high-stakes card games, on top of round-the-clock champion-level squash and tennis. Not surprisingly, this left little time for the exacting standards of Harvard classwork. But Victor’s father was a stern police lieutenant who demanded A’s of his son, so Victor set about outsmarting the Harvard system.

Rather than defaulting into a pseudo-major (History of Zumba Dance? Athenian Weights and Measures?) as custom would have predicted, Victor opted instead for graduate-level classwork in economics. You see, everyone knew that it was incredibly difficult to become a graduate student at Harvard, so anyone taking graduate classes must obviously be brilliant and hardworking and organized. But what few recognized was that graduate students ingratiated themselves with professors by doing most of the boring work that would otherwise fall on the professors’ shoulders. And grateful people tend to reciprocate a kindness. Combine the halo effect of being in graduate classes with the reciprocity tendency engendered by doing grunt work and, ladies and gentlemen, we have a lollapalooza effect on our hands! Victor got A, after A, after A, after A and was hardly ever near a class. For a while folks thought they had a new economics protégé on their hands. But eventually people caught on. They still call this playbook “Niederhoffering the curriculum.”

Fun story. But Charlie retells it to make a sinister point: all human systems are gamed.

It’s less funny when you apply this logic to the Medicare System or Workman’s Comp or the H-1B Visa Program. When lives and careers and families are on the line. Ignorant legislatures, with their inability to take into account the second and third order effects of their proposals, set up systems that incentivize gaming and reward misbehavior. Much of the fault lies with those who cheat the system and certain people may have a proclivity for unethical behavior regardless of circumstances. But I tend to agree with Charlie that too little attention is paid to those who create the rules of the game, that “the people who design easily–gameable systems belong in the lowest circle of hell.

II.

Let’s consider the H-1B visa program, that much maligned initiative that Ted Cruz and Donald Trump mention in the same breath as ISIS or “the Wall.” The H-1B visa was designed to bring foreign professionals with college degrees and specialized skills to fill jobs for which qualified Americans cannot be found. This is the U.S. visa, mentioned in a recent WSJ piece, behind (most of the) 44 immigrant co-founders that have helped create over $168 billion in startup value and over 33,000 U.S. jobs (roughly 760 jobs per co-founder; note that the total number jumps to ~200,000 when you consider the jobs created by Uber, whose co-founder Garrett Camp is a Canadian immigrant).

Congress imposed a hard cap of 85,000 visas per year on the H-1B program. Well how big is 85,000 really? Definitely larger than a minyan but probably smaller than the Ohio Stadium seating capacity? Do you know how many job openings there were in the United States? 5.5 million. In January. So 85,000 is big. But it’s not crazy big. It’s not Justin Bieber dropping an acoustic version of Love Yourself big. It’s not what Donald Trump means when he puts his hands two-and-a-half feet apart and says “YYYHUUUGGEEE.” Why then does this small visa program draw ire typically reserved for the fiery intensity that comes with debating religion after the third beer?

YYYHUUUGGEEE-ly amusing Gif…

Well, because people are gaming the system, of course. Sure, 44 immigrants created a bazillion dollars of value but 16,573 of the 2014 H-1B visas went to just seven Indian outsourcing firms. The very firms that Ted Cruz (though it pains me to quote him) probably correctly criticizes for “ bringing in medium and low-skilled (information technology) workers and then firing American workers and, adding insult to injury, forcing the American workers to train their foreign replacements.” It’s not pretty. Here’s the NYTimes graphic:

These firms received a disproportionate number of the total H-1Bs by exploiting a design flaw in the system: to prepare an H1-B visa application, employers must first submit a document called a Labor Condition Application BUT companies can apply for more than one employee based on one Labor Condition Application. So the outsourcing firms use one application to apply for a dozen or more workers. This maneuver makes no sense if there is a specific employee you want to fill a specialized role, as the legislature intended, since you need that unique individual to be approved. But if your workers are interchangeable cogs in a vast outsourcing machine then who cares? Apply for as many as you can on each Labor Condition Application and ship over the subset who are ultimately approved for the visas.

The seven Indian outsourcing firms that received the 16,573 visas submitted ~200,000 Labor Condition Applications. H-1B visas are granted by a computer-run lottery if the number of applications exceeds the 85,000 quota, so the more H-1B applications you submit the greater your odds of being approved. There is no meritocracy involved in approving the visas, aside for meeting some banal requirements, and thus no distinguishing between an amazing applicant and someone who just checks the boxes. It’s very hard for a legitimate applicant to be submitted more than once. Turns out holding down full-time jobs at both Goldman Sachs and Morgan Stanley is tricky.

This gaming of the system is infuriating. Good people, as always, get pulverized in the bureaucratic machine. I’ve had many deserving friends denied visas in the H-1B lottery. Indeed, I was given a one in three chance of being approved. I don’t like those odds.

III.

So who is to blame for this perversion of judicial intention? Well, it’s easy to point a finger at Infosys and Wipro and other outsourcing firms. Or claim a vast government conspiracy to steal jobs from law-abiding American citizens. Or find any number of people or companies that are profiting from this broken system. But I think that misses the point. These villainous folks are abusing the system but they aren’t breaking the poorly designed rules. And it is a red herring focus on obvious villains that causes us to lose our shit over Martin Shkreli hiking the price of Daraprim by 5,556% rather than critiquing the totally broken legislative system that allowed Shkreli to exist in the first place. This distinction is important. It’s easier to change a man-made system than to change a man’s nature.

Presumably the well-intentioned folks who drafted the H-1B visa legislation did not envision it being abused by crafty outsourcing firms. I think this failure can be mostly diagnosed by Hanlon’s Razor: never attribute to malice that which is adequately explained by stupidity.

But that does not mean legislatures are not to blame. Failure to consider second and third order consequences is calamitous.

It’s this failure that crushes people in the jaws of a leaky healthcare system or an iniquitous tax system or a broken immigration system. And we scream “change!” at our politicians until we are blue in the face. But what few politicians admit or probably realize, is that they too are trapped in the blundering machinations of complex human systems. They too are subject to the psychological forces of the systems they create. Most wouldn’t know where to start even if they wanted to affect positive change. Reduce capital gains tax and you lower economic growth. Broaden healthcare coverage and inadvertently increase premiums. Maybe. Who knows? A butterfly flapping its wings and all that good stuff.

In the end we are left with a revised Hanlon’s Razor situation that was described well by Eliezer Yudkowsky:

Never attribute to malice what you can attribute to an enormous complicated System full of conflicting incentives getting stuck in a weird equilibrium. When that weird equilibrium is crushing people in its gears, don’t attribute that harm to a conspiracy of evil powerful people who planned it all and profit from it. There is no master plan behind the US medical system, it’s just an enormous complicated thing that got stuck. Even if there’s a billionaire or politician benefiting from the current setup, they didn’t cunningly plan for the US medical system to be dysfunctional, and they couldn’t make anything be different by choosing otherwise. Conspiracies of evil people plan how to profit from the System’s current stuck state. They don’t decide where it gets stuck.

Which is why it’s so important to consider human psychology when designing systems — why Charlie Munger says “the people who design easily–gameable systems belong in the lowest circle of hell.” We must think beyond intention (first level thinking) and focus on incentives, rewards, and punishments (second level thinking). Once the system is out there it gains a life of its own. What starts off as a simple set of rules becomes an emergent director of behavior that is incredibly hard to fix after the fact.

It’s like this scene from Cube (thanks again Eliezer for the comparison) where Worth says: “This may be hard for you to understand, but there is no conspiracy. Nobody is in charge. It, it’s a headless blunder operating under the illusion of a master plan.

IV.

It all seems rather macabre but maybe it does not have to be. We can design better systems and improve existing ones. The H-1B system could probably be improved by accepting applicants in order from higher to lower salaries, segmented by profession, instead of based on a lottery (since the outsourcing firms pay their H-1B holders around the required salary floor of $60,000, far lower than an American citizen would be paid in an equivalent role) or by limiting the number of employees that can be attached to an individual Labor Certification Application. But there need to be incentives to improve systems.

As Scott Alexander puts it, “people can alter the incentive landscape in order to build better institutions. But they can only do so when they are incentivized to do so, which is not always.”

What worries me is that congressional representatives do not have incentives to create the “best” version of the H-1B visa system or healthcare system or welfare system. Rather, as Scott puts it, “every representative’s incentive is to appeal to his or her constituency while throwing the rest of the country under the bus.” And this strategy seems to work: while only 11% of Americans like Congress, an approval rating lower than Nickelback and traffic jams, around 62% of people who know who their own Congressional representative is approve of them. In other words, people hate Congress but like the congressperson who goes to bat for their interests.

Perhaps we can force new incentives on our representatives by better informing the electorate; remove smoke-and-mirrors rhetoric and replace it with clear, accessible data on job creation, system abuse, etc. Fight the battle that John Oliver fights every Sunday night. Elucidating work done by the New York Times or the Wall Street Journal or the National Foundation for American Policy might move the needle. Though I have my doubts… confirmation bias seems inexorable.

V.

All human systems will be gamed and designing easily game-able systems is a moral betrayal. Munger says, “the people who design easily–gameable systems belong in the lowest circle of hell.Dante calls the lowest circle of hell Tradimento, or treachery. Hanlon’s Razor (the excuse of stupidity) does not morally exonerate.

For me, it all comes back to considering second-order and higher consequences when creating systems or making decisions in the first place — to think not only of the system’s intention, but whether participants are incentivized to use the system as intended. It’s not easy. Unlike first-level consequences, which are simplistic and superficial and just about anyone can spot, second-level consequences are definitionally deeper, more complex, and more convoluted. It’s hard to consider all the ways various stakeholders will interact with a system. To quote Howard Marks, “anyone who thinks it’s easy must be a first-level thinker.

One good starting point is to focus on incentives and avoid underestimating their effect. We should keep in mind Munger’s observation: “I think I’ve been in the top 5% of my age cohort all my life in understanding the power of incentives, and all my life I’ve underestimated it.” In fact, I propose a new eponymous Munger’s law that states: incentives are more important than you think, even when you take into account Munger’s Law.

But, while we wait for our politicians to recognize this point, perhaps you should find a handful of outsourcing firms to submit a bunch of H-1B Labor Condition Applications on your behalf? (half kidding).