A Theory of Excess Hiring as Insurance

And why the tech layoffs are possibly socially inefficient.

Right now is a bad time to be looking for a job in the tech job market. The large firms, like Meta, Amazon, Google, and Apple have cut back hiring, and laid off swathes of employees. 

The general consensus is that they overhired in the aftermath of the pandemic, and now find themselves with unnecessary staff. Laying off unnecessary workers would, on the face of it, be an increase in efficiency. I am concerned, though, that there are countervailing effects which make this bad for social welfare. 

We should start with the nature of innovation. Ideas are not capturable by the creator. They are an enormous positive externality on other people. Any theory of growth which incorporates ideas tells us that we are underinvesting into finding new ideas. I have written elsewhere about this. As an approximation, we can assume that new ideas are an increasing function of the number of people working in a field. More people, more ideas. All else being equal, twice as many people working in a field will generate twice as many ideas in that field. If this is so, the socially optimal level of innovation is too low, because of the large positive externalities. But this is bog standard, and applies to every industry. We would have to have reason to believe that more ideas are generated in tech jobs than elsewhere. I believe this is true, but it’s little more than a hunch. What I find more interesting is that, even if all returns were captured by the inventing company. Laying off unnecessary workers can harm productivity.

Let’s refine our model of how innovation happens a little. People have very different abilities to innovate, but nobody — neither the employers nor employees — can tell who will be better ahead of time. In this world, you could conceivably fire the low performers and not harm productivity at all. Suppose half the people are idea generators, with marginal product 2, and zeros, with marginal product zero. Ahead of time, companies can’t tell apart who is who, so they offer all workers a pay of 1, or the expected value of a new hire. Employees can either invest in learning skills, and work in the tech sector, or work in the safe sector, which pays a wage less than or equal to one. There are two periods. In the first, employees invest in skills, and are hired. Prospective employees in the tech world are randomly divided among all the hiring firms. In the second, the type of each employee is revealed, and firms can choose to lay off unproductive employees, and give productive employees a raise. 

If employees are risk neutral, then the optimal number of people are hired. If they are risk averse, however, then they will avoid entering, just as they would avoid entering any lottery. In fact, under the conditions imposed, any degree of risk aversion results in no employees investing in skills at all! The key is that each firm does not bear the cost of their hiring and firing decisions, because we have assumed that people end up randomly at firms. The firms are providing insurance against your investments into acquiring skill not working out by continuing to employ you, even if you’re mediocre. This is not something that can be obtained by buying an insurance policy, because of moral hazard, as Arrow points out in “Economic Welfare and the Allocation of Resources for Invention”. If you know that you will get paid for not working very hard and none of your ideas working, it’s gonna be awful hard to keep you from not working. The upshot is that every firm would prefer a world in which no one was able to fire unproductive workers!

This is the enlightened case for why tenure exists. Academia requires intensive training for an uncertain future. As the joke goes, a doctoral student is one who foregoes present income so that they may forego future income. Universities are interested mainly in the superstars, but have to hire a bunch of promising mediocrities to get the best. If they simply fired everyone who turned out to be mediocre, very few people would ever enter to begin with. 

Randomness may seem like a strong assumption, but I think it’s a reasonable one. You must invest before you know who you will be employed by. After you invest, the marginal cost of working is 0. It is only before the investment occurs that you stand to lose anything. Thus, after entering people are going to apply to as many firms as they can, and since we already assumed identical and indistinguishable applicants, the firms will accept as many as they have openings. You prefer being employed to not employed, even if you would have, before the game started, preferred certainty to uncertainty. 

What this also suggests is that, as production processes get increasingly complex, the more the inability to insure will matter. Kremer, in his paper “The O-ring theory of economic development”, gives a model where production requires many steps, a failure in any one of which dooms the whole project to failure. If we think about error rate as skill, if the number of steps increase, the returns to skill increase. Income inequality should increase as industries get more complex, and indeed they very obviously do. If people want insurance for investing in skill, and cannot obtain it, the social losses should get larger as society gets richer. Welfare can be welfare enhancing, if it serves as insurance. What it also implies is that, under some circumstances, it may be better for firms not to know the productivity of their own workers. Learning the productivity of your workers is like learning that “defect” exists in a prisoners’ dilemma. You would be better off if no one had ever apprised you of the possibility that you could cheat. 

This is just a sketch of the idea; I hope to expand it to a paper. I think it badly needs formalization — I worry that if one firm lays off workers, it would lead, under this model, to precisely enough workers leaving that the cheating firms hire no workers. In this world, the firm does internalize the cost of defecting in later periods, and it becomes standard game theory. Later thinking should clarify this. I highly welcome any comments on this.