Capital market regulation often relies on two methods: imposing restrictions on the actions of parties with private information and requiring greater transparency about these actions. The U.S. Securities and Exchange Commission’s (SEC) recent proposal targeting insider trading abuses follows this paradigm, similar to other proposals related to share repurchases and short selling. The insider trading proposal looks to add restrictions on insiders’ ability to use and trade on Rule 10b5-1 plans, plus enhanced disclosures. In a recent working paper, we use a setting analogous to capital markets to provide evidence that greater disclosure of insider trading is likely to have a limited effect in driving “more efficient capital allocation and more informationally efficient prices” (see pg. 56 of the SEC’s insider trading proposal). Rather, based on theory and empirical work on “cursedness,” we expect that the more heavy-handed component of the proposal that restricts trading by insiders will be more effective in promoting price efficiency.
A useful starting point for our predictions is the canonical “lemons problem,” which arises when the seller of an asset (like a defective used car, or “lemon”) knows more about the asset than the buyer does. At a given price, there is an inverse relation between asset quality and the size of the ownership stake offered for sale. The lemons problem results in market failure when buyers recognize this relation and refuse to transact – the average quality of what is offered for sale is too low for the price. Yet, recognizing and assessing this relation can be difficult; cursed buyers (or traders) do not have accurate perceptions of the true link between asset quality and ownership retention. In particular, cursed traders fail to fully appreciate the private information behind others’ actions, which manifests as underreaction to a value-relevant signal. An implication is that disclosure of ownership sales is not enough to bring perceptions in line with reality.
In stylized form, the figure below provides some intuition. It shows the actual relation between the percentage of ownership sold and asset quality (line “A”), with price held constant. It also shows market participants’ perceptions of this relation, which theory commonly assumes are unbiased but could instead reflect overreaction or underreaction. Cursedness predicts that market participants’ perceptions of quality are too insensitive to ownership retention, meaning they underreact to an owner’s decision to sell (line “U” is flatter than line “A”). If traders are cursed, correcting perceptions via disclosure (i.e., shifting line “U” so it overlaps with line “A”) is likely less effective than regulatory interventions that restrict insiders’ ability to act on private information (i.e., shifting line “A” so it overlaps with line “U”).
We use horse race betting as a setting to examine the effectiveness of different regulatory approaches to resolving cursedness and promoting price efficiency. Horse race betting is comparable to capital markets in several ways, including the behavior of individuals and the market and the information available to market participants. Moreover, relative to a capital market setting, features of the horse race setting allow for more refined tests of cursedness that provide new insights on the regulation- and asset-specific factors that contribute to cursedness.
Analysis
We use a 10-year sample of over 30,000 optional claiming races at North American racetracks to test our prediction. In these races, a horse’s owner chooses whether to offer it for sale at a fixed claiming price simultaneously with the decision to enter it into the race, a fact that is always disclosed and easily visible to bettors. The sale price is set by the track and is uniform for all horses in the race, making it impossible for bettors to discern quality of individual horses based on sale price alone. Even so, we confirm that horses offered for sale by their owners are on average lower quality – information that bettors can easily incorporate into their bets.
Because North American horse racing relies on pari-mutuel betting (winning bets divide the money wagered on losing bets, less transaction costs), the setting ensures a tight link between quality expectations and an objective measure of quality. Quality expectations in this setting are win probabilities implied by the race-time odds, and the objective measure of quality is the race outcome. If bettors’ expectations of quality are unbiased, then the implied win probability (based on the betting pool) should, on average, correctly predict whether a horse offered for sale actually wins the race. In contrast, we find that horses offered for sale are less likely to win the race than implied by the betting pool, which suggests that bettors misprice the information contained in an owner’s decision to offer a horse for sale. Portfolio analyses help quantify the economic magnitude and consistency of this effect. Returns to betting on horses for sale are lower than returns to betting on horses not for sale in every year of our sample and are 4 percent lower per bet, on average.
Though the for-sale signal is disclosed in all races, additional analyses suggest that bettors do not fully differentiate between the disclosure’s underlying information content, consistent with cursedness. For example, we find that new assets (i.e., horses with little experience) are mispriced to a greater extent. In contrast, we do not find evidence of mispricing in races that substantially limit owner discretion to offer the horse for sale. For regulators, the message is that market participants are partially cursed, which results in underreaction even in the presence of salient disclosures. But regulations that limit insiders from using nonpublic information are likely to promote price efficiency.
We use several additional tests to speak to cursedness as a mechanism. Inconsistent with limited attention or some other behavioral bias as an alternative explanation, we do not find significant variation in underreaction based on “trading on prices” or the extent of participation by relatively more inattentive or unsophisticated bettors. Additional tests also help mitigate the concern that the results are due to features of the setting that are less likely to apply to capital markets. For example, we do not find that the results depend on market liquidity or bettors’ risk preferences.
Conclusion
Our study contributes to the regulatory debate and literature on how market participants price asset sales by insiders. To the extent that features of capital markets continue to encourage more trading (e.g., as brokerages emphasize payment for order flow), evidence of individual behavior and biases in a market that already has these features should be informative to regulators. Because information asymmetries between insiders and outsiders can be acute in lottery-like stocks in which retail investors often trade, ignoring or underweighting ownership sale signals becomes more problematic for these stocks.
Our findings suggest that expanded disclosures about insider trading are unlikely to fully address regulator concerns about price efficiency. Rather, our evidence supports the efficacy of more heavy-handed interventions (e.g., additional restrictions on Rule 10b5-1 plans) that reduce the scope for managerial discretion and opportunism in instances of significant information asymmetry. At a minimum, regulators could do more to help investors understand disclosures, given that ownership sale signals are easy to acquire yet are still mispriced. In the absence of heavy-handed interventions, disclosure regulation may be more effective if it were to help investors interpret the informationconveyed by an owner’s decision to sell an asset.
This post comes to us from Darren Bernard at the University of Washington, Madi Kapparov at London Business School, Sara Toynbee at the University of Texas at Austin, and John Wertz at Indiana University. It is based on their recent paper, “When Disclosure Isn’t Enough: Evidence on Cursedness in Betting Markets,” available here.