In Law360 (paywall), I make the case for placing new limits on noncompetes.
In The Atlantic, I analyze Apple v. Pepper, the Supreme Court’s latest antitrust case.
My latest paper on the law and policy implications of labor monopsony strikes a more pessimistic note than earlier work. I argue (with Suresh Naidu) that in the best case, antitrust law can only make a dent in the labor monopsony problem. Other legal and policy approaches are needed, and we discuss some of them. Here is the abstract:
Recent literature has suggested that antitrust regulation is an appropriate response to labor market monopsony. This article qualifies the primacy of antitrust by arguing that a significant degree of labor market power is “frictional,” that is, without artificial barriers to entry or excessive concentration of employment. If monopsony is pervasive under conditions of laissez-faire, antitrust is likely to play only a partial role in remedying it, and other legal and policy instruments to intervene in the labor market will be required.
They don’t budge. See the comparisons from 538 below. Unlike every other president’s poll numbers, Trump’s barely change in response to events. Why don’t people update their beliefs–pro or con–in response to his latest triumph or outrage? Possible answers:
If #3 is right, then the Democrats should not expect that impeachment hearings will generate political support for impeachment, and would be wise to resist the temptation.
In the Times, Daniel Hemel and I argue that the Mueller Report shows the presidency’s vulnerability to agency costs (though we don’t put it quite that way for the general reader). Our argument echoes some of the claims made by Neal Katyal in his 2006 article “Internal Separation of Powers,” but we are (or at least I am) more skeptical than Katyal and other authors that institutions within the executive branch can reproduce the “external” separation of powers, which was intended to keep the executive (as well as the other branches) in check in the first place, but seems to have failed.*
As we note in the piece, the mechanism in question—the ability of aides and other official in the executive branch to defy the president’s orders—is ambiguous in its normative implications. They can defy good orders as well as bad orders; they can constrain for good or ill, as the Hoover example was supposed to show.
The literature should pay more attention to within-executive branch mechanisms. Let me suggest two:
1. Selection. The president selects aides from a pool of candidates. The candidates differ along two dimensions: preferences (which might align more or less with the president’s) and ability. A president must make tradeoffs between the two dimensions. However, if a president wants to be effective, he must take account of talent, and if preferences are normally distributed, this will push him in the direction of the median. Thus, more extreme presidents will be moderated, and this seems to be what has happened with Trump. Most of his better aides disagree with Trump’s more extreme views and projects, while the cronies seem pretty incompetent. (Bannon and Miller have been the exceptions: apparently competent, though I’m not sure, as well as ideologically committed.) Of course, where executive-branch positions require senate approval, the extremist president will also be pushed toward the middle (as we are seeing with Fed nominations).
2. Ex post discipline. The president has very limited tools for rewarding or punishing his subordinates. To reward them, he can promote them, but promotions are limited and usually not very attractive to top officials who want to cash out by leaving government after a couple years. To punish them, he can publicly humiliate them, but he can’t, for example, throw them in jail. The president cannot give bonuses to loyal aides, nor can he (as we are learning) credibly offer them pardons in advance of breaking the law since aides rationally anticipate the president may never issue a pardon, which can be politically costly.
The biggest problem for the president is that officials can continue to pursue their careers in the private sector (or, later, in politics) only if they avoid being tainted by their service in the presidency. Lawyers face the risk of being disbarred if they engage in illegal or unethical conduct.
Both mechanisms suggest that the policies of a president like Trump can be moderated by internal executive-branch constraints (that is, agency costs), but they weaken as the elites polarize. If you imagine a bimodal distribution, for example, the president may find adequate talent away from median, and career opportunities might be available even to people whose actions offend much of the public. Moreover, the mechanisms will subvert a “good” president who seeks in good-faith to act in the public interest under conditions where most people (at least temporarily) disagree with him. (FDR ahead of World War II remains the preeminent example.)
The mechanisms, and particularly the second one, show how important it is for the president to maintain credibility. Subordinates who do not share his preferences need to be sure that he will punish them if they defy his wishes, and reward them if they obey. Executive Unbound argued that “for presidents, credibility is power…. Without credibility, the president is a helpless giant.” This seems to be an accurate enough prediction about Trump, and becoming more accurate every day.
* Note that there is a sense in which the separation of powers might be vindicated. The president was constrained by subordinates who feared being prosecuted under the obstruction of justice statute—which requires the involvement of the judiciary, as well as Congress, which needed to pass the statute in the first place. But this is not what people usually mean by the checking power of other branches. The statute remains dormant, and the judiciary uninvolved, unless the “executive branch” acts—by launching the investigation, cooperating with the investigation, and (ultimately) refusing to pardon.
1. As understood even 20 years ago, tech firms need to obtain network- (and/or IP-based) monopolies in order to earn returns that would justify the financial investment.
2. To obtain these monopolies, firms offered their products and services for free (Google, Facebook), or below cost (Amazon, Uber). While the companies did not so intend, their strategy had the short-time side-effect of generating enormous goodwill among consumers. (Remember this? “Amazon, as best I can tell, is a charitable organization being run by elements of the investment community for the benefit of consumers.”)
3. Once they obtained their monopolies, the tech firms rationally engaged in what economists call “rent-extraction”—raising prices or reducing costs by degrading service, which often took the form of violating promises to protect data privacy, bombarding users with ads, failing to maintain earlier quality standards, or all three. Consumers felt betrayed.
4. Meanwhile, the tech firms took advantage of consumers’ unfamiliarity with the new commercialized internet. People didn’t (and mostly still don’t) understand that:
a. When they reveal data about themselves in piecemeal, seemingly harmless ways, they enable tech firms to aggregate the data in ways that may compromise their privacy, autonomy, and even safety.
b. When they use convenient services on the internet, they are often being subtly manipulated in ways that make it difficult for them to stop using those services, even when those services are no longer convenient.
c. When they express their opinions on the internet where normal social cues and expectations are obscured, they often do so in ways that deeply offend, even enrage, people that they don’t want to offend or enrage, and would never try to do so in other settings.
People have been socialized in how to use the telephone, how to write letters, and how to talk to people in professional and social settings. They (still) haven’t been socialized in how to use the internet, especially social media. That will take years.
It’s the combination of monopoly power, technological manipulation of boundedly rational consumers, and consumers’ unfamiliarity with a new technology that makes the internet the hell that it is today. While most people will eventually figure it out, and both social and commercial institutions will socialize us and help protect us from manipulation, only government regulation can counter market power.
Mueller’s group believed that the president committed obstruction of justice. Mueller and his team interpret the law broadly (which is to say, correctly) and they recount numerous actions by the president that interfered with the investigation, and provide plenty of evidence of “corrupt” intent—meaning a desire to protect the presidency from embarrassment and possible criminal liability. Trump tried to cajole or intimidate Manafort, Cohen, and probably Flynn (some information was privileged), among others; and he tried to influence officials like Comey, Sessions, and Mueller himself. Moreover, Mueller (correctly, in my view) does not believe that separation-of-powers and other constitutional considerations give the president license to corruptly interfere with federal investigations.
Whey didn’t Mueller indict Trump? The answer is simple. Under Justice Department regulations, Mueller was bound by the (questionable) OLC decision that a president may not be indicted or criminally prosecuted while in office. While Mueller could have nonetheless announced his belief that Trump violated the law without filing an indictment, he believe that it would be unfair to make such an announcement. Without a speedy trial, the president would be left twisting in the wind, unable to defend himself before a jury of his peers. Thus, “while this report does not conclude that the President committed a crime, it also does not exonerate him.” I would say: the report does not officially conclude that the President committed a crime, but it clearly implies that he did.
Hemel and I reply to Andrew McCarthy’s defense of the Barr memo. Was our original op-ed “surprisingly vapid,” unsurprisingly vapid, or neither?
I have been thinking about labor monopsony lately. Recent academic research suggests that thousands of labor markets are highly concentrated. Labor monopsonists are on notice that any actions to maintain or advance their monopsonies are illegal. Surely such actions are common. Consider, for example, the thousands of mergers among employers, resulting in workplace closures in shared labor markets. Or a monopsonist who uses noncompetes to maintain or advance its control over a labor market.
So where is the litigation? Aside from a handful of cases brought by the government and private parties involving only the most explicit restraints of trade under section 1 of the Sherman Act, almost nothing. Why is antitrust litigation so much more common for product markets than for labor markets? Section 2 seems all but a dead letter for labor markets. Can it be resurrected for a world of ubiquitous labor monopsony?
Searching for answers, I have coauthored an academic article arguing that mergers should be reviewed for their labor market effects. Here is some testimony, further developed as a blog post. Also: a policy paper, an op-ed, a longer op-ed, and a podcast on Akerman WorkedUp with Matt Steinberg in which we discuss noncompetes.
Email me your insights. More to come.
[Socrates:] And amid evils such as these will not he who is ill-governed in his own person–the tyrannical man, I mean–whom you just now decided to be the most miserable of all–will not he be yet more miserable when, instead of leading a private life, he is constrained by fortune to be a public tyrant? He has to be master of others when he is not master of himself: he is like a diseased or paralytic man who is compelled to pass his life, not in retirement, but fighting and combating with other men.
Yes, he [Glaucon] said, the similitude is most exact.
Is not his case utterly miserable? and does not the actual tyrant lead a worse life than he whose life you determined to be the worst?
He who is the real tyrant, whatever men may think, is the real slave, and is obliged to practise the greatest adulation and servility, and to be the flatterer of the vilest of mankind. He has desires which he is utterly unable to satisfy, and has more wants than any one, and is truly poor, if you know how to inspect the whole soul of him: all his life long he is beset with fear and is full of convulsions, and distractions, even as the State which he resembles: and surely the resemblance holds?
Very true, he said.
Moreover, as we were saying before, he grows worse from having power: he becomes and is of necessity more jealous, more faithless, more unjust, more friendless, more impious, than he was at first; he is the purveyor and cherisher of every sort of vice, and the consequence is that he is supremely miserable, and that he makes everybody else as miserable as himself.
No man of any sense will dispute your words.
“There can be no worse foe of mankind in general, and of his own country in particular, than the demagogue of war, the man who in mere folly or to serve his own selfish ends continually rails at and abuses other nations, who seeks to excite his countrymen against foreigners on insufficient pretexts, who incites and inflames a perverse and aggressive national vanity, and who may on occasion wantonly bring on conflict between his nation and some other nation.”
— Theodore Roosevelt (1905)
The graphic comes from the New York Times, based on data from a paper written by Lee Epstein and me, plus new data for Trump. The Trump data confirms the trend that we discovered in our (pre-Trump) paper: the decline in the win-rate for presidents before the Supreme Court since Reagan. The question is what accounts for it. Our two preferred hypotheses are the growth of the private Supreme Court bar, and an increase in the self-confidence of the Court. Will the trend continue after Trump’s next appointment?
One frequent response so far is skepticism that ordinary people can handle the cognitive load some of our proposals would impose on them. Here’s Ryan Avent, for example:
[T]he book seems to dramatically underestimate the cognitive load that is likely to be associated with its proposals, and the likely resistance to those programs on that basis. The authors reckon that apps can be used to make management of these markets as easy as possible. Even so, they are asking people to begin thinking in market-oriented ways about lots of things which don’t currently require such thinking. That, after all, is the point: that aggregating the considered, distributed reasoning of lots of people is likely to produce better outcomes. But contributing to that considered, distributed reasoning is a pain; even if it can all be done on an app, you have to sit, and weigh your actions, and worry that you made an error of judgment. To give just one example: Uber has become far more pleasant to use since surge pricing went away. The system “worked better” in some sense, when riders and drivers had to think harder about how much they actually valued the trip. But that thinking was itself a cost of the service.
This is an important point, and I agree with the Uber example [N.B, updated: actually, I don’t; the airline example below is a better one, as Uber continues to adjust pricing but in a more obscure way]. The same point can be made about the way airlines package and disaggregate different aspects of the service in their pricing decisions. But the problem turns out to be trickier than it first seems. Our COST proposal, for example, requires people to estimate the values of (say) their home, but people have to do that anyway when they sell their home, and also when they buy a home in the first place—and, at least in principle, when they take out mortgages, plan for their retirement, rent out space, etc. And while the COST requires repeated valuations over time, which may enhance the cognitive load for the possessor, this also means that the cognitive load is reduced for all potential buyers, who no longer need to bargain with sellers.
In the case of quadratic voting, the cognitive load is reduced in a more direct way. Because QV effectively allows people to trade political influence over different domains, it allows me to focus on the issues I care about (say, data privacy), or the campaigns that matter to me, or the geographic unit of politics I’m comfortable with. I may be deeply immersed in, and affected by, decisions of the local schoolboard, and I can (implicitly) trade for influence over it with someone who cares much more about the identity of the next president. You might think of it this way: we are all currently generalist producers of democratic outcomes, where QV allows division of labor and specialization in a natural, decentralized way.
Won’t we all be worse citizens then? The current system of democracy puts a massive cognitive load on all of us—we are expected to be informed about literally everything—all issues, national, state and local, dozens of candidates, etc., so that we can vote wisely and responsibly in countless elections and referenda. Of course, nearly all of us duck this load by remaining massively ill-informed, just as consumers do when confronted with complicated products and services. I suppose when democracy was first proposed, someone must have said—“you must be crazy: how are people going to bear the cognitive load?” Or the ancient Greek equivalent.
But I think the better way to think about this is to start with the general problem: all the goods and services in the economy must be allocated somehow. If we gave the task to a central planner, the cognitive load would be far too great for any single person. Central planners of the past tried to solve this problem by creating vast bureaucracies, enabling division of labor and specialization. But as von Mises, Hayek, and others pointed out long ago, the cognitive load (or what economists came to call information costs, but is better understood in psychological terms, I think) is best distributed among all citizens via the mechanism of the market. The more broadly shared the burden, the more easily it is borne by individual citizens.
What they didn’t establish is that the market institutions of their time were a superior bearer of the aggregate cognitive load than possible alternative market institutions, as their focus was the critique of central planning. But having accepted their critique, the next question is: what market design does the best job of distributing the cognitive load among the most people, and among the people best able to bear it?
Reviews from Italy, France, Latin America, and Ethereum (and here, and, interestingly, an RM-related help wanted ad). More from Robin Hanson and Daniel Roberts. Matthew Kahn on radical markets and urban economics. Ryan Avent posts a review on Medium. I hope to provide some responses to criticisms & questions in due course.
The first argument is the argument about policy; the second argument is about whether the first argument is permissible.
Thus, for Ezra Klein and Sam Harris, the first argument is the argument about the relationship between race and intelligence, and the second argument is about whether Sam Harris is allowed to interview Charles Murray about the relationship between race and intelligence. For Michelle Goldberg, the first argument is about whether women and men can work together in the workplace, and the second argument is about whether it is appropriate to debate Jordan Peterson about whether women and men can work together in the workplace. For Robin Hanson and Jordan Weissmann, the first argument is about whether sexual opportunities should be redistributed, and the second argument is about whether Hanson should have been allowed to make the first argument. For Jennifer Finney Boylan, the first argument is about the treatment of trans people and the second argument about whether the treatment of trans people may be debated. In many cases, the second argument is couched in terms of the permissible ways of making the first argument (Hanson/Weissman), including who should be allowed to make the first argument (Klein/Harris), rather than the first argument is literally permissible, but ultimately it amounts to the same thing.
It’s enough to make your head spin, but the pattern is pretty clear. Person #1 makes an argument. Person #2 might criticize the argument on the merits, in which case we have a rare single-argument encounter. But with increasing frequency, Person #2 says (in essence) that Person #1’s willingness to make the first argument shows that Person #1 is a “creep,” or a fascist, or is acting in bad faith, which always is meant to imply that Person #1 should be fired from his job, or deprived of a public forum, or not taken seriously—should be publicly shamed. Moreover, because of the vagaries of language and the nature of the current hothouse political and cultural climate, it is easy for Person #2 to mistakenly (or deliberately) believe that Person #1 made the second argument even if Person #1 did not, which leads Person #2 to make the second argument about Person #1, namely that Person #1 is trying to censor and shame Person #2 and for that reason Person #1 should be fired, or deprived of a public forum, or not taken seriously (see Harris/Klein). Sooner or later, someone will make the second argument.
Because our government doesn’t censor people, second-argument makers have sought speech restrictions from the private sector, and have succeeded in many ways:
1. Universities, which have increasingly enacted speech codes.
2. Employers, which have cracked down on dissenters.
3. Social media companies, from Facebook to Twitter, which have imposed numerous restrictions on content.
4. Even content sellers, like Spotify, have gotten into act.
Failing all that, shame campaigns on social media may sometimes lead to 1, 2, 3, or 4.
While not all of this is new (employers have always regulated speech in workplaces), what is new is the contribution of technology, which has raised the stakes for everyone, and given tech giants quasi-monopolies over the public sphere, and ideological shifts, which have produced polarization (at least relative to the last 20-30 years). You can think of the second-argument phenomenon as the result of ideological polarization amplified by technological advances in communication.
More Radical Markets. Glen on TV. Glen on radio (okay, “podcast”). Our screed against economics in the Chronicle. Nice reviews in the Economist (“arresting if eccentric”) and the Irish Times (“It made my head hurt, and then spin”). The Irish Times review was written by a government minister. Can you imagine an American politician admitting he has read a book, let alone reviewing one? Well, there’s this (“the key for me is to keep expectations low,” which is either self-refuting or self-reinforcing, or possibly both).
The enthusiasm for cryptocurrencies, and blockchain more generally, derives from the sense, to enthusiasts, that an old dream is within reach, thanks to advances in computer technology: governance without (human) discretion. Consider our system of property and contract rights. Many people think these bodies of law allow us to organize our affairs in a fair and efficient way, and that the law can be reduced to a set of relatively simple rules that can, in principle, be applied mechanically. The frustration and tragedy is that we must rely on human beings—judges and other bureaucrats—to apply the law, and these human beings unavoidably make errors and, worse, may be biased or corrupt. The tension is even worse for more controversial areas of the law, such as tax law, where the suspicion that IRS officials enforce the law arbitrarily and with bias, illustrated by the Tea party scandal, is widely held.
But if the laws could be enforced by computer, using open-source software that can be examined by anyone, the problem is solved. Hence the excitement about smart contracts, for example. But more than that: what is wanted is a system of law that is entirely self-contained—made, or at least approved, by the public rather than by politicians, and enforced automatically. This is, of course, the blockchain as applied to the (theoretically) self-enclosed system of currency.
But there is a problem—the governance problem. One source of perplexity is who designs the initial software program. What ensures that he or she designs a program in the public interest? Rousseau was so stymied by this problem in the context of political theory that he believed that the lawgiver must be seen as divine. Satoshi Nakamato, who helped get Bitcoin started by shrouding his identity and achieving semi-mythical status, knows his Rousseau!
But even if this initial problem is solved, the second and more interesting problem is how the law (code) can be updated as events change. In the case of bitcoin, miners who collectively hold more than 50% of computer power can change the protocol. But this creates obvious incentives toward consolidation, which appears to have taken place. Bitcoin is a plutocracy. Other cryptocurrencies keep power in the hands of the founders and designated agents, creating, in other words, an oligarchy.
If we try to imagine democratic alternatives, we run into significant problems. It is tempting to think that we could give every (say) bitcoin user one vote. But bitcoin users are anonymous; the network can’t trace transactions to a single user (since a single user can use different keys for each transaction). Perhaps in a different version of the protocol, a vote could be attached to a bitcoin, so whoever happens to possess a bitcoin (or fraction) at any time could vote that bitcoin (or fraction). Still, the wealthy would possess most of bitcoins and hence most of the political power, and strong incentives would exist to accumulate bitcoins in order to maximize power.
A truly democratic cryptocurrency would be one in which all people (presumably, in the world, since national boundaries mean nothing in cyberspace) would possess equal voting power, including people with few coins or none at all.
Global voting seems scarcely imaginable, and also seems incompatible with anonymity. But put aside anonymity and imagine that every person in the world, or in the relevant jurisdiction if voting based on nationality or some other identity were possible, could vote on proposals to update the protocol. Formidable problems would continue to exist. Standard voting procedures—one-person-one-vote—have significant problems, and would be vulnerable to collusion. More sophisticated voting procedures—like quadratic voting—could made headway with this problem, but would still be vulnerable to the classic weakness of democracy: most people would be both incapable of, and uninterested in, making informed decisions on proposals to update a computer program, as the questions would involve technical issues of both finance and programming. Some kind of representative system might be necessary, but the representatives would necessarily have discretion, and be vulnerable to political influence, and we are back where we started.
For these and related reasons, I believe that the dream remains unattainable, at least in the near future, and with respect to currency supply and other public goods. A more realistic possibility is that relatively discrete blockchain systems would be governed based on users’ contributions in maintaining those systems, much as bitcoin is, but these systems would need to be narrowly tailored to some specific commercial problem and couldn’t be used for public goods, like currencies, that affect everyone. Even in the discrete case, however, discretion probably cannot be completely eliminated, as it is in games like chess. People seem to underestimate the extent to which even straightforward-seeming transactions, an interest-rate swap, for example, involve complex contracts with terms that are deliberately ambiguous so that they can be applied in a discretionary manner by judges if unforeseen circumstances arise. But the scope of discretion can be narrowed, and within the remaining field of decision-making, better or worse governance systems can be used. This is the area in which progress can be made.