Transparency and its discontents

Why PredictIt should prevail over the CFTC, but no-action letters should stay legal.

“Event contracts” are contracts that pay based on the outcomes of future events. For example, John and Jane could enter into a contract in which John agrees to pay Jane $1 if Donald Trump is the 2024 Republican nominee. “Prediction markets” are markets in which event contracts can be bought and sold.

In an efficient market, the cost of such a contract corresponds to the market’s view of the probability of the event occurring. For example, if it costs $0.40 to buy a contract that would return $1 if Donald Trump is the 2024 Republican nominee, that would imply that the market believes there is a 40% chance that Donald Trump will be the 2024 Republican nominee. If you are bullish about Donald Trump being the nominee, you can buy one or more of these contracts, and if your prediction is correct, you will make a profit. Of course, buying these contracts will cause the market price to increase.

In theory, prediction markets can aggregate the wisdom of crowds and yield predictions that are more accurate than the predictions of individual pundits. The world is better off if we can better predict the future. As such, many people (myself included) count themselves as supporters of prediction markets.

PredictIt is a popular market for political event contracts. On PredictIt right now, for instance, you can pay $0.38 for a contract that returns $1 if Donald Trump is the 2024 Republican nominee for President, $0.26 for a contract that returns $1 if Donald Trump wins the 2024 presidential election, and $0.27 for a contract that returns $1 if Kamala Harris is the 47th President of the United States.

The Commodity Futures Trading Commission (CFTC) believes that political event contracts are subject to its regulatory authority. On October 29, 2014, the CFTC’s Division of Market Oversight issued a “no-action” letter to PredictIt, in which the Division agreed not to recommend initiating an enforcement action if PredictIt adhered to certain conditions.

On August 4, 2022, however, the Division issued a letter withdrawing the prior no-action letter. The Division’s letter states that PredictIt’s operator “has not operated its market in compliance with the terms of” the prior no-action letter, but does not explain the basis for this assertion. The letter states that all open contracts “should be closed out and liquidated no later than 11:59 p.m. eastern on February 15, 2023.”

PredictIt sued the CFTC under the Administrative Procedure Act. After failing to obtain relief from the district court, it appealed to the Fifth Circuit. The Fifth Circuit granted a temporary injunction barring the CFTC from giving effect to the August 4, 2022 letter, which means that PredictIt is currently still operating. The Fifth Circuit heard oral argument and a decision is expected shortly.

In this post, I will make the following points:

• Prediction markets are great, and the CFTC’s regulatory efforts have been poor.

• In attempting to shut down PredictIt, the CFTC is harming the public interest and mistreating PredictIt for no good reason. PredictIt should win its suit.

• PredictIt’s victory should be narrow. In general, agencies should be able to issue and withdraw no-action letters without judicial review. This will improve our ability to predict the future—the exact philosophy animating PredictIt.

All hail prediction markets

“Predictions are hard, especially about the future.” —Yogi Berra

It’s hard for any one person to predict the future. Consider the challenges of one person predicting an election outcome. There may be millions of voters. An individual predictor has never met 99.99+% of them. She has never been to the places they’re from. Many are from backgrounds or cultures with which she has little familiarity. She reads the news, but has no idea how other people interpret the news.

Pundits aren’t helpful; they may be confident of their own predictions, but they have the same blind spots that she does. Polls are pretty good, but are expensive and not everything is polled. Also, polls suffer from low response rates, and respondents merely say who they’re voting for, not their degree of confidence in who is going to win.

Enter prediction markets. Like other markets, prediction markets aggregate the wisdom of crowds. Anyone can purchase an event contract, reflecting their assessment of the probability of a particular outcome. According to the efficient market hypothesis, the price of an event contract will reflect all available information, and therefore be more accurate than any one person’s view. This is not merely theoretical; there’s abundant empirical evidence that prediction markets make better predictions than individual forecasters.

Of course, prediction markets need not focus on politics. They can be used to predict anything. Take a look at Metaculus, which allows forecasters to make predictions on a huge variety of issues, ranging from climate change to AI. Perhaps to avoid regulatory scrutiny, Metaculus does not involve event contracts; instead, one earns points for successful predictions. Inevitably, Metaculus allows you to make a prediction on the future of PredictIt. Metaculus is great, and might be even be greater if it was able to use real event contracts.

Here’s a truly excellent FAQ walking through the benefits and drawbacks of prediction markets. Interesting fact: there is a political movement advocating a form of government called “futarchy,” in which all policy is determined by prediction markets. I would not call myself a futarchist, but I still am optimistic about prediction markets.

Ehhh… let’s just ban everything.

The CFTC does not share my admiration of prediction markets.

First, some background. Futures contracts, as the name suggests, are agreements to perform trades in the future. For example, one can agree to buy or sell an asset, like soybeans or oil, for a particular price at a particular future date. People trade futures contracts for lots of reasons, ranging from hedging to diversification to speculation. Event contracts—contracts that pay based on future events, like the winner of a presidential election—are one type of futures contract.

Futures trading is generally legal, but is regulated by the CFTC. However, in Dodd-Frank, Congress authorized the CFTC to ban certain types of event contracts if the CFTC found them to be contrary to the public interest. Here’s what Congress said:

The Commission may determine that such agreements, contracts, or transactions are contrary to the public interest if the agreements, contracts, or transactions involve—

(I)activity that is unlawful under any Federal or State law;

(II)terrorism;

(III)assassination;

(IV)war;

(V)gaming; or

(VI)other similar activity determined by the Commission, by rule or regulation, to be contrary to the public interest.

See 7 U.S.C. § 7a–2(a)(5)(c)(i). This seems pretty reasonable. It doesn’t directly ban anything, but instead gives the CFTC narrow authority to ban certain types of trading that seem particularly bad, such as trading in which people profit from terrorist attacks.

In implementing Dodd-Frank, the CFTC’s regulatory efforts got off to a bad start. The CFTC quickly promulgated a regulation saying that all event contracts falling within categories (I), (II), (III), (IV), and (V) are categorically prohibited. 17 C.F.R. § 40.11(a)(1). The statute says the CFTC should exercise discretion and consider the public interest; the CFTC instead went ahead and preemptively banned everything it is capable of banning.

Things went from bad to worse. In 2012, the CFTC issued the so-called “Nadex Order.” The Nadex Order banned an exchange from listing political event contracts, such as contracts turning on the outcome of presidential or congressional elections.

If there was a Hall of Fame of arbitrary and capricious agency decisionmaking, the Nadex Order would be an early inductee. In a clipped four-page order, the CFTC simultaneously managed to:

• Unreasonably interpret a federal statute.

• Ignore the CFTC’s own rules.

• Adopt a speculative and implausible theory that prediction markets will cause harm, while ignoring all counter-arguments.

First, Dodd-Frank narrowly authorizes the CFTC to ban event contracts based on three specific types of events: acts of terrorism, assassinations, and wars. Elections are really different from acts of terrorism, assassinations, and wars, so under a natural reading of Dodd-Frank, the CFTC doesn’t have the authority to ban event contracts based on elections, or so one would think.

Unfortunately, the CFTC did not see it this way. The CFTC concluded that political event contracts qualify as “gaming,” apparently because “gaming” covers all contracts turning on the outcome of future events. The CFTC arrived at this dubious statutory interpretation by consulting some random state statutes plus an irrelevant federal statute having nothing to do with elections or event contracts.

The implication of the CFTC’s statutory interpretation would seem to be that, under the CFTC’s overly-broad regulation, political event contracts are categorically illegal. However, the CFTC then elected to ignore its own regulation and proceed to a public interest analysis, without giving any explanation for why it was doing this.

Then, the CFTC reasoned that political event contracts are not in the public interest, because: (a) they can’t be used for “hedging,” as if hedging is the only good reason in the world for markets to exist, and (b) they might create “monetary incentives to vote for particular candidates even when such a vote may be contrary to the voter’s political views of such candidates.” Yes, apparently there are an abundance of voters out there who will buy event contracts in favor of the candidate they don’t like, and therefore feel forced to vote for that candidate in view of the 0.00000001% chance that their vote will decide the election, which is bad because the government’s job is to ensure that people cast their votes for the “right” reasons. Meanwhile, the CFTC appeared completely unaware of the many excellent reasons for legalizing prediction markets, notwithstanding a vast literature on this topic.

Oh well. But a couple of years later, some good news arrived. In 2014, PredictIt asked the Division of Market Oversight for a “no-action letter.” A “no-action letter” is a letter in which a division of the agency agrees not to recommend initiating an enforcement action to agency leadership. The letter isn’t binding on the agency leadership, it’s simply a promise by staff members not to make a particular recommendation to their boss.

And the Division agreed! The no-action letter says that as long as PredictIt adheres to various conditions (e.g., being a not-for-profit enterprise and limiting markets to 5,000 traders per contract with a $850 investment limit per participant in any contract), staff members won’t recommend an enforcement action. According to the no-action letter, PredictIt is different from Nadex because it is an academic, not-for-profit enterprise. I’m not sure, why a matter of either statutory interpretation or policy, that makes PredictIt better than Nadex, but all’s well that ends well I suppose?

Wrapping up the regulatory summary, there’s another company called Kalshi currently trying to get approval from the CFTC to operate prediction markets (not just a no-action letter, but actual, formal regulatory approval). The CFTC still thinks this might be “gaming” and opened up a comment period on whether these markets are in the public interest, over the dissent of Commissioner Caroline D. Pham, whose dissent is so persuasive that I honestly cannot figure out any arguments against it. Robert Schiller, who won the 2013 Nobel Price in Economics, among others, submitted comments supporting prediction markets. As far as I know this is still pending in the CFTC, so we will have to wait and see.

We will now destroy your business and refuse to explain why, have a nice day!

PredictIt’s salad days did not last. On August 4, 2022, out of nowhere, the CFTC’s Division of Market Oversight issued a letter withdrawing the prior no-action letter. The letter states that all open contracts “should be closed out and liquidated no later than 11:59 p.m. eastern on February 15, 2023.” Goodbye, PredictIt, it was nice knowing you!

Why? The Division’s letter states that PredictIt’s operator “has not operated its market in compliance with the terms of” the prior no-action letter. No further explanation was provided.

At the end of the letter, the Division pleasantly says: “Should you have any questions, please do not hesitate to contact” the Division. Unfortunately, the Division does not actually answer questions, including basic questions such as why are you doing this?! The Division has still not explained why it did this. No one knows.

I had mentioned earlier that the Nadex Order should be enshrined in the Arbitrary and Capricious Hall of Fame. But really the Nadex Order is more of a Harmon Killebrew, certainly an excellent inductee worthy of praise but not really a GOAT candidate.

The Division’s letter, by contrast, is a Babe Ruth. If the Division is going to hurl a thunderbolt and destroy PredictIt’s business, shouldn’t it at least give a reason? Why wouldn’t it give a reason? There are no justifications for not giving a reason.

Worse, the intended effect of this letter is to wipe out PredictIt, which is a useful and great website. I am unaware of any evidence that PredictIt has ever harmed anyone. The Division did not cite any. Why ban it?

Also, the no-action letter merely reflected the Division’s commitment not to recommend an enforcement action to the Commissioners, so the withdrawal of the no-action letter merely means that the Division won’t not make this recommendation. In principle, the withdrawal merely should have returned the world to the status quo ante, in which the Division hadn’t made a commitment one way or another to recommend or not recommend anything and the Commissioners hadn’t decided anything whatsoever. Yet the withdrawal letter is written in such a way that it resembles a command: open contracts “should be closed out and liquidated no later than 11:59 p.m. eastern on February 15, 2023.” The Division doubtless knew that PredictIt users, who would read the Division’s letter and who didn’t have lawyers to interpret it, would perceive it as a direct order to PredictIt to shut down or else.

Being a regulator is often hard. It requires making difficult predictive judgments, balancing costs and benefits, and addressing the interests of multiple stakeholders. However, one regulatory principle that I feel should be uncontroversial is: “Regulators should not write misleading letters that wipe out businesses and harm the public interest, with no countervailing benefits, while also giving no reasons for what they are doing.” Unfortunately, the CFTC does not agree.

Preserving the right to be bullied by our regulators

PredictIt sued the CFTC. After failing to obtain relief in the district court, it sought and obtained a temporary injunction from the Fifth Circuit, where the case is now pending. So PredictIt is currently hanging on by a thread.

As the above discussion should make clear, I am rooting for PredictIt to win its suit, and indeed it is winning so far.

And yet. Things are more complex than they seem.

The CFTC’s primary defense is that APA suits are limited to “final agency action,” and the withdrawal of the no-action letter is not “final agency action.” As regulatory lawyers can recite by heart, final agency action requires that the challenged conduct be the consummation of the agency’s decision-making process, and be conduct from which legal consequences will flow. In this case, the CFTC contends, the withdrawal of the no-action letter merely means that staff members at the CFTC won’t not recommend something to their supervisors; the agency hasn’t made any final decision on anything and the withdrawal has no concrete legal consequences.

Also, the CFTC says that if the court subjects no-action letters to judicial review, it’s simply not going to issue them anymore, which will make everyone worse off. To quote the CFTC: “If this Court were to suddenly subject no-action letters to judicial review, those letters would predictably become more difficult to attain, forcing market participants to undertake more burdensome and prescriptive statutory procedures that today may be considered unnecessary or unjustified as a use of public or private resources. The same chilling effect may well spread to other parts of the Government, decreasing choice and subjecting more conduct than is necessary to regulatory scrutiny.”

In this case, I’m not persuaded this argument should win. First, the withdrawal of the no-action letter is written in a manner to create the impression that it is a direct order from the CFTC, and I’m not sure it should evade regulatory scrutiny merely because agency lawyers explain after the fact that it isn’t one. Second, PredictIt says that if it keeps operating post-withdrawal, the agency will assert that it is defying an agency command and therefore acting “willfully,” resulting in an increase in punishment, so the withdrawal letter does have concrete legal effect. The CFTC does not disavow that it would rely on the letter in exactly this way.

But let us postulate a hypothetical world in which withdrawals of no-action letters:

1. Explain the Division’s reasoning.

2. Express clearly that they merely reflect the withdrawal of a recommendation to Commissioners not to initiate an enforcement action, and are not an order to do anything.

3. Are not admissible as evidence of willfulness. (In other words, if the Division issues and then withdraws a no-action letter, this would be treated the exact same way as if the Division had never issued a no-action letter in the first place.)

Should withdrawals of no-action letters be judicially reviewable? No.

The whole notion of a no-action letter is weird. There’s no doubt the agency is allowed to issue no-action letters—there’s a specific regulation authorizing them. See 17 C.F.R. § 140.99(a)(2). But it’s quite unusual that a mid-level staff member would issue a public letter declaring that she will not make a particular non-binding recommendation to her boss. Usually, organizations shield internal deliberations and publicly disclose only their final decisions. I could not imagine a law firm associate publicly declaring in writing that if opposing counsel agrees to respond to 80% of a particular set of discovery requests, the associate won’t recommend to the partner to file a motion to compel. I could not imagine a mid-level prosecutor announcing that if a corporation under investigation comes clean and apologizes, she won’t recommend to the D.A. to initiate a prosecution.

But “weird” does not mean “bad.” Judicially unreviewable no-action letters seem like a great idea to me, and entirely in line with PredictIt’s philosophical principles.

No-action letters essentially say: if a company agrees to take a particular set of actions, the probability of an enforcement action will decrease, even though it will not decrease to zero. The commissioners are not bound by the no-action letter, so there is no guarantee that the company is safe, but if the company complies with the no-action letter, its odds are better. This type of probabilistic information about the future—allowing for better, if imperfect predictions—is exactly what PredictIt is in the business of providing.

But why shouldn’t no-action letters be an ironclad guarantee, and why shouldn’t they be judicially reviewable? Well, they could be, but the CFTC says that if those restrictions are imposed, then the CFTC will simply stop issuing no-action letters, leaving everyone worse off. This reasoning makes perfect sense: increased process inevitably deters action.

Empirically, I would say that the type of person who favors prediction markets also tends to be the type of person who decries bureaucratic red tape. If one were to take polls on propositions like: “NEPA makes it too hard to build new projects” or “it is bad that, as a result of over-regulation, it takes longer to build an escalator today than it took to build the Empire State Building in 1930,” I suspect that support among the pro-prediction market community would be somewhere near 100%. Well, if you believe those things, then of course you should believe that injecting formality and layers of review into no-action letters will slow down and deter no-action letters.

Of course, it is possible that people will overreact to withdrawals of no-action letters and incorrectly interpret them as direct orders to stop operating. Should the risk that people incorrectly interpret information be a basis for suppressing that information? No. Regulators sometimes argue that prediction markets should be suppressed because of the dangers that they will be misinterpreted or manipulated. Defenders of prediction markets say that ambiguous information is better than no information, and we should not suppress the flow of information to guard against the risk that it may be misunderstood. I agree with the defenders of prediction markets—and for the same reason, I agree with the concept of judicially unreviewable no-action letters.

Withdrawal of no-action letters, without judicial review, can be framed as a kind of unaccountable regulatory bullying. But everything has costs, including the abolition of unaccountable regulatory bullying. We are all better off if agencies can unaccountably withdraw no-action letters, as long as they are up-front with what they are doing, which the CFTC was not in the case of PredictIt.

Thanks for reading! The topic of the next post is TBA.