Remembering Managerial Dilemmas

A conversation with a friend over brunch reminded me of a really thought-provoking book I reviewed a few years back, namely Gary Miller’s Managerial Dilemmas: The Political Economy of Hierarchy.

The basic idea in Miller is that, if your organization (company, team, university, whatever) is judged on the basis of how it performs overall, then everyone has an incentive to slack and let everyone else do the work. But since everyone is subject to the same incentives, everyone slacks and the whole thing goes to shit (technically speaking).

Likewise, your company has every incentive to screw you and not, say, invest in educating you; after all, why pay for you to get a master’s degree if you’re just going to take their investment to another job?

Somehow both sides need to agree to disarm: the company needs to credibly (and in a certain sense irrationally) signal that it’s going to support its employees, even though it has no guarantee that they’ll reciprocate; and employees need to credibly (and in a certain sense irrationally) signal that they won’t slack, even though they have no guarantee that the company will reward them.

Turns out it’s a hard problem. I don’t recall Miller talking about this at all, but it seems clear to me that government has a role to play here: since companies can’t be trusted to supply me with a pension that will help me in my old age, let’s make Social Security really good. And since companies can’t be trusted to pay for my master’s degree, let’s have the government subsidize advanced degrees. There are obvious problems with this, but it’s not clear that they’re worse than the economy as she already works.

The Davies-Folk Theorem and Gary Miller's Managerial Dilemmas

I found D2‘s excellent gloss on the folk theorem of iterated game theory the other day:

The Folk Theorem in game theory states that any outcome of a repeated game can be sustained as an equilibrium if the minimax condition for both players is satisfied. In plain language, it can be summarised as stating that if we take strategic considerations into account, there is a game-theoretic rationale for practically anything. This formulation leads on to my contribution, the Davies-Folk Theorem, which states that if we take strategic considerations into account,there is a game-theoretic rationale for practically fucking anything (its a fairly simple [corollary]; proof available from author on request).

(The rest of that post is worth reading. So is Thomas Schelling, for that matter. I’ve only read his Micromotives and Macrobehavior, but I want to read The Strategy of Conflict und so weiter.)

Gary Miller refers over and over again to the folk theorem in Managerial Dilemmas, and I could never figure out why — for reasons that D2 makes most clear. Even at the time, Managerial Dilemmas seemed like a strangely theorized take on corporate structure. But then the folk theorem bits were lacquered on top, as though they justified the point he was making. It was very strange to me.

So yeah. Were you thinking your day would start with iterated game theory? Well it has.

Gary J. Miller, Managerial Dilemmas: The Political Economy of Hierarchy

Attention conservation notice: 1500 words on Gary Miller’s Managerial Dilemmas. MD tries to figure out what works and what doesn’t in organization design, but along the way I wonder if Miller is just recapitulating a moral principle or two that we all ought to be feeling anyway.

Working together is a public-goods problem: if all of us on the team work except for me, then I get the advantages of the team’s production without doing any of the work. But all of us on the team have the same incentives, so we all have an incentive to avoid doing work. But if we all avoid doing work, the team produces less than it could, and we all suffer. In the jargon, “everyone shirks” is a Pareto-inefficient Nash equilibrium: it’s where the process would settle because no one has an incentive to change what he does (the “Nash equilibrium” part), but it’s worse for everyone (the “Pareto-inefficient” part) than if everyone worked hard.

Nothing interesting about this cover. It has the words 'Managerial Dilemmas' and the author's name. One way for managers to address this public goods problem is called “Taylorism“: study very carefully how much time and effort it takes people to accomplish a specific task, then insist that they accomplish that task in that much time. If the time-and-motion guys tell you that you can produce ten automobile hoods per day, you’d better produce ten per day or they dock you. This gets rid of one information problem: under the assumption that the workers want to shirk, take away their power to shirk by rigorously defining what their job is.

This “piece rate” system has obvious flaws. The time-and-motion guys need to measure actual workers doing actual work. The workers know this. So when the time-and-motion guys come around to measure, the workers will work slowly. And if a worker produces more than the piece-rate system says he can, his coworkers will not take kindly to it: they know that any consistently higher rate of production will only make them work faster, or it will lose them a job when the company realizes that it can do more with fewer workers. Some companies are real scoundrels about it and adjust the per-piece rate downward, so that the per-day rate that they pay workers stays constant.

So a piece-rate system, unless accompanied by a promise that you won’t be a scoundrel, appears to be a non-starter. And the company has every incentive to be scoundrels, if it will make them more money to do so. Or at least they have “every incentive” to do so if they look at the short term only. In the longer term, worker enmity may well offset these incentives. Workers may quit and go to other companies that aren’t as exploitative. They may institute a work-to-rule strike. There are lots of ways workers can get back at draconian management.

If I’m reading Managerial Dilemmas right, these sort of utilitarian calculi have never worked because economists have taken the short-term static view: we should do X to prevent Y, because that maximizes our income in the short term. The better approach is to take a long-term dynamic view. One way to model these sorts of problems is as an iterated Prisoner’s Dilemma: you (as my boss) and I (as the worker) are going to have to interact over and over again, so we can’t always act selfishly. If you always act selfishly, it’s in my interest to do the same. So then we both act selfishly, and we both lose.

Here’s the point where every exposition of the iterated prisoner’s dilema has to mention Robert Axelrod, specifically The Evolution of Cooperation. Gary Miller does not disappoint. Here’s the big message that I’ve now had beaten into my brain approximately 100 times: The Tit-For-Tat Strategy Does Really Well. That is, if you start your repeated game by not shirking and only shirk if your boss is a scoundrel, you’ll do all right. When he’s a scoundrel, retaliate. When he treats you well, quickly forgive his earlier trespasses. Everyone and his brother has made something huge out of Tit-For-Tat’s success. Bowles and Gintis even turn it into a moral about the politics of fairness.

The message for managers seems to be that they should learn from iterated game theory and apply what they learn to the structure of their workplace. Chief among these is that they need to find some way to make employees cooperate rather than defect. The only way to do this, in the long term, is to solve a commitment problem — in other words, for the employees to promise that they’ll go against their short-term self-interest and not shirk, and for the managers to promise that they’ll go against their short-term self-interest and treat the employees as valuable people. There are various ways to do this; Miller goes through several in the final third of the book. One is to convey a culture of openness: Hewlett-Packard apparently left doors to its labs unlocked, and encouraged employees to take equipment home. During economic downturns, some companies convince employees of their good faith by refusing to lay anyone off; they may impose pay cuts, but no one loses his job. The general way to describe these commitment strategies is that they are costly and difficult to fake. “Irreversible” might fit in there too: restructuring a factory to accommodate team-focused production is hard to reverse; the employees know you’re in it for the long haul.

All of this sounds to me like fancy game-theoretic garb for a not-very-complicated idea: be good to your employees, and they will be good to you in return. I’m not convinced that these economic arguments will actually make anyone be nice to his employees or his superior. If you were already morally committed to the golden rule, you will still be morally committed to the golden rule. If you weren’t, it’s unlikely that this economic argument will suddenly swing you over. If game theory doesn’t already support the golden rule, so much the worse for game theory. On the one side we have sensible, deeply-felt moral principles; on the other we have an economic model that’s highly sensitive to assumptions about human behavior that we’re only just coming to understand. The moral principles win in a rout.

I felt the same way about Bowles and Gintis’s Recasting Egalitarianism. It might sway some people at the margins — people who believe in helping out the less fortunate but worry about the economic pitfalls of doing so. I wonder how many people are in that margin, though. I suspect that a great many people are morally committed to asset redistribution; Recasting Egalitarianism will not change their minds, though it will give them more reasons to advance at those asset-redistribution cocktail parties you hear so much about. On the other side from the egalitarians are those folks who are rightly suspicious of grand plans to put wealth where wealth wouldn’t go of its own accord; RE will do little to allay their fears of heavy-handedness.

For that matter I feel the same way about John Rawls: if you need to travel to a notional world in which you could be less fortunate than you are, in order to justify helping out the impoverished, will reading a few hundred dry academic pages really sway you? Isn’t this part of Rawls just a restatement of Matthew 25:35-40?

For I was an hungred, and ye gave me meat: I was thirsty, and ye gave me drink: I was a stranger, and ye took me in:
Naked, and ye clothed me: I was sick, and ye visited me: I was in prison, and ye came unto me.
Then shall the righteous answer him, saying, Lord, when saw we thee an hungred, and fed thee? or thirsty, and gave thee drink?
When saw we thee a stranger, and took thee in? or naked, and clothed thee?
Or when saw we thee sick, or in prison, and came unto thee?
And the King shall answer and say unto them, Verily I say unto you, Inasmuch as ye have done it unto one of the least of these my brethren, ye have done it unto me.

The first 2/3 of Managerial Dilemmas is quite useful and thought-provoking. It’s an analysis of exactly why designing hierarchies that work is difficult. The analysis of what’s wrong with piece-rate systems is a good example: few other authors go into much detail about exactly which organizational structures have been tried, which have failed, and what’s worked.

One result, which recapitulates Miller and Hammond’s paper “Why Politics is More Fundamental Than Economics: Incentive-Compatible Mechanisms Are Not Credible, is particularly captivating. It argues that in any hierarchy, it will always be necessary to “bind the king’s hands,” because the king (or the CEO) will always have an incentive to amass more wealth rather than do what’s efficient for his kingdom (or his company). No self-guiding economic mechanism will get you out of the fundamental political problem.

I’d recommend reading the first 2/3, then politely returning Managerial Dilemmas to the library.