VoxEU Column Financial Regulation and Banking

Dealer information sharing in US Treasury auctions

Information sharing has come under increased scrutiny in the context of interbank lending, foreign exchange markets, and US Treasury auctions. This column explores the benefits and drawbacks of information sharing by dealers in US Treasury auctions. Information sharing is found to benefit first and foremost the issuer, i.e. the Treasury. The model provides insight on auction revenue, risk-sharing, and the decision to bid through a dealer, with information sharing having a sizeable effect on each.

Recent financial market misconduct in the interbank lending (Libor) and foreign exchange markets – involving misuse of information about clients' orders – has prompted legal action, incurred record fines, compromised reputations, and prompted international initiatives to curb information sharing. In US Treasury auctions, where trillions of dollars are sold each year largely to, or through, a small set of primary dealers, the use of information has similarly come under increased scrutiny (Bloomberg 2015).  While it may all sound straightforward, it is in fact not obvious who is harmed when information is shared.

Regulatory initiatives to curb information sharing in Treasury and other markets attempt to protect clients from harm, but we find that information sharing among dealers creates value for investors and improves risk-sharing. Investors are harmed from information sharing not when dealers exchange information with other dealers, as one may suspect, but when dealers share information with other investors. For the Treasury, information sharing can boost auction revenues, an effect that we estimate in the order of about $4.8 billion per year. This increased revenues come at a cost of increased revenue volatility and risk of a failed auction, which is when revenues fall well below expectations. 

How dealers use clients’ order flow information is not a new issue for economists or practitioners. In the economics literature, the use of order flow information has been central to our understanding of sovereign auctions (Hortacsu and Kastl 2012), to market making theory generally (Kyle 1985) and to market practice for decades. In describing Treasury market pre-auction activities in the 1950s, Robert Roosa (1956) noted that “dealers sometimes talk to each other; and they all talk to their banks and customers; the banks talk to each other”. Sharing order-flow information – or, colloquially, ‘market colour’ – with issuers is even mandatory for primary dealers both in the US and abroad.

Despite these long-standing discussions, how information is shared in sovereign auctions has recently been an area of active policy debate. The Treasury Market Practices Group is currently reviewing best practices for information handling by dealers, and outside the US, information sharing practices are changing as well. In 2011, the UK Debt Management Office sanctioned that UK primary dealers (called ‘Gilt-edged Market Makers’) “whilst not permitted to charge a fee for this service, may use the information content of that bid to its own benefit”.1 The revised guidelines stated instead “information about trading interests, bids/offers or transactions may be subject to confidentiality obligations or other legal restrictions on disclosure (including pursuant to competition law)”.

In a recent paper, we explore the benefits and drawbacks of information sharing (Boyarchenko et al. 2016). Because regimes with and without information-sharing restrictions cannot be simultaneously observed, existing data do not reveal what effects this policy change might bring. Therefore, we use a calibrated model that matches the bidding behaviour and settlement prices in US treasury auctions. Then we change the information-sharing rules inside the model to measure their effects on treasury revenue and investors’ profits.

The results prove that information sharing first and foremost benefits the issuer, which, in the case of US sovereign auctions, is the Treasury. Based on our model calibration, moving from full information sharing to no information sharing would lower Treasury auction revenues by $4.8 billion annually. When more information is shared, auction risk is reduced, investors and dealers bid more aggressively, which ultimately boost auction revenues. While intuitive, this economic channel is hardly discussed in the policy debate on establishing ‘Chinese walls’.  

Dealers sharing information with their clients doesn’t sound too objectionable. But surely dealers swapping notes with other dealers cannot be good for small investors, right? No. In fact, the model teaches us that this thinking is backward – when dealers swap information with other dealers, investors benefit. The ability for dealers to share information between his/her clients can actually be worse for those clients.

Why? Because the two types of information sharing have opposite effects on information asymmetry. When dealers share their clients’ information with other clients, there is a shared belief among those clients. However, this in turn causes the beliefs of investors across different dealers to become more polarised, or asymmetric. Information asymmetry encourages some bidders to take large positions against others, worsening overall risk-sharing. This result is harmful for investors overall.

Of course, each investor individually prefers more information to less. But information acquisition is a prisoners’ dilemma problem – an example of the Hirschleifer (1971) effect. In contrast, when dealers share information with each other, the advice they provide to their clients is similar. Clients with more similar beliefs share risk more efficiently, which benefits the investors. In short, dealer information sharing with other dealers has an opposite effect on investor utility from sharing with clients.

We are not dismissing the potential fact that dealers who share information with each other can also collude with each other. While information sharing does not imply collusion, it makes it possible. That being said, the benefits for information sharing outweigh the prevention of information sharing in fear of collusion. If collusion is the only problem, then the obvious remedy is to aggressively enforce anti-collusion laws.

But what if, when information is shared, some collusion cannot be prevented? Even if this is the case, information sharing may be its own remedy. The victim of collusion is the Treasury, which loses revenue when bidders collude on low bids. Since the Treasury is also the main beneficiary of information sharing, when enough information is shared, the revenue benefits of information sharing exceed the revenue costs of collusion. If in addition to sharing information with other dealers and colluding, dealers also share enough information with clients, the net effect on Treasury revenue is positive in our calibrated model.

While information sharing has many upsides, it also has important downsides. One of the most troubling of these is that the combination of information sharing and mixed auctions amplifies negative news, potentially resulting in failed auctions, and more likely reduction in revenues well below expectations.

Treasury auctions are mixed auctions, meaning that investors can either place bids through a dealer or bid directly, without dealer assistance. An investor with good news, who will want to buy many Treasuries, should want to keep that information to themselves. If the information were shared with a dealer who observed their bids, inferred their beliefs, and in turn shared that knowledge with clients, those clients would bid more aggressively upon learning the good news and push the price higher. Avoiding a higher auction price requires not sharing their information or their demand with a dealer and instead, bidding directly. But an investor with negative information typically expects to bid for only a few shares. If the investor expects to end up only with small amounts of the new issue, the consequence of sharing the negative news and pushing the price down is small. But the benefit of getting information (market colour) from their dealer and learning that perhaps the outlook is not as bleak as (s)he thought can be large. As negative-news investors bid through dealers, their negative news is disseminated by the dealer, resulting in weaker demand. If positive news is kept private with investors bidding directly and negative news is shared with dealers, then auction prices will be more responsive to bad news than good news. This asymmetry manifests itself as an increased risk of a failed auction.


Using a model to measure the potential effect of unobserved policy requires that one believes the model to be a reasonable representation of reality. Of course, reality will always be more complex. But the model provides a guiding map by suppressing details. Our model clarifies the various effects of information sharing on auction revenue, on risk-sharing, and on the decision of whether to bid through a dealer or not. Because the size and nature of these effects are ambiguous absent a model, they have not received much attention in the current policy discussion. Our calibration suggests, however, that each effect is sizable. As new policies are crafted and considered, we hope that the model’s insights into who wins and who loses from information sharing can inform this important debate.


Bloomberg (2015) “As US probes $12.7 trillion Treasury market, trader talk is a good place to start”.

Boyarchenko, N, D O Lucca and L Veldkamp (2015) "Intermediaries as information aggregators: An application to US Treasury auctions", FRB of New York Staff Report, 726.

Hortaçsu, A and J Kastl (2012) "Valuing dealers' informational advantage: A study of Canadian treasury auctions", Econometrica, 80(6): 2511-2542.

Hirshleifer, D (1971) “The private and social value of information and the reward of inventive activity”, American Economic Review, 61.

Kyle, A S (1985) “Continuous auctions and insider trading”, Econometrica: Journal of the Econometric Society, 1315–1335.

Roosa, R V (1956) Federal Reserve operations in the money and government securities markets, 332, Federal Reserve Bank of New York.


[1] GEMM Guidebook, 2011 and 2015 versions.

735 Reads