One reason regulators all seem to want to create and expand consolidated tapes is because public prices are good at protecting investors. We’ve also seen that tight spreads and low trading costs lower the costs of capital and increase company valuations.
But just because public prices are good for the public doesn’t make them a public good (public goods are usually free).
As we recently discussed, some exchanges are more focused on market quality than others. The data showed that those exchanges need to pay to support competitive quotes. In addition, the traders providing those quotes are also trying to profit from trades and spread capture (not providing a free service to others).
In short, quotes aren’t provided for free – so nor should they be free for those who don’t set them.
This raises an interesting question…
Who sets quotes?
This sounds like an easy question. The trader who sets the NBBO first is the one setting the price.
But working out who is setting prices isn’t so simple, especially in fragmented markets. Consider the example below and a market with geographical fragmentation:
Chart 1: Looking at order arrival and trades and who is “contributing” to public prices
On the offer, we see two separate traders. One is setting the price (Offer 1), and the other is a “pegged” order that joins once the price is set. Most would agree that Offer 1 is setting the price. In fact, it’s hard to argue that the pegged order isn’t using the Offer 1 price to do their trading.
Similarly, after the offer ticks down, we see an algo seller join the market at the “near touch.” Again, it’s hard to argue that the pegged order isn’t using the Offer 1 price to trade.
However, who is contributing most to price discovery when Offer 1 is repriced higher? If the Offer 1 trader is looking at futures or other correlated assets, that might be a better indication of the “right” price for the stock. However, the algo might have a large order behind it and prefer to trade at the original price rather than increase opportunity and wait costs. In that sense, the algo, which knows it has a sell order behind it, is also contributing to price discovery.
Interestingly, when Trade 3 does occur, Quote 1 doesn’t participate. But the algo and the peg orders that did not reprice do contribute to price setting and liquidity. Also, at that specific point in time, the offer that trades matters much more than the bid (that still hasn’t), especially as those quotes are now contributing liquidity.
The problem is harder if we look at the bid. Although we have two separate bids, they arrive at almost indistinguishable times. If the time difference is due to different geographies, it is possible, in actual space-time, that Bid 2 arrived first but wasn’t seen by the SIP until later. Then, who sets the price?
Finally, let’s consider the value of trades.
In this example, there are three trades — the first two at midpoint, then the last one at the offer. The midpoint price is based on the trader at Bid 1 and Offer 1 – who was also trying to capture spread – not support others trading away from them. But from the market’s perspective, the midpoint trades may help with price discovery, especially given no trades happen on the bid, as the “right” price may be closer to the offer.
The work of others, using tick-level data, suggests not all data is equal
A number of well-known academics have also tackled this problem.
The answer, based on their work, suggests that actionable quotes are much more important to price formation than trades. And dark trades don’t add much at all. It would be economically more efficient if the traders and exchanges setting those quotes could be rewarded more, too.
One approach was pioneered by Joel Hasbrouck from NYU. He models the lead-lag relationship between multiple prices while presuming the existence of a theoretical “right price,” as we mentioned in the previous paragraph.
For instance, to estimate the price discovery contributions of two venues trading the same stock, the model assumes that prices observed in each venue can deviate but are always linked together by arbitrage. Price discovery in this setting refers to changes in the “right” price, and information share refers to the contribution of each of the observed data points to prices.
This methodology can be used to calculate an information share for each venue, order type, or message, depending on the research setting. In the results below, the information share ranges between 0% (for no contribution) to 100%, representing the total information share across all venues or order types.
Over the last 30 years, several academic studies have built on this approach to measure the price discovery contribution of trades vs. quotes, spot vs. futures markets, individual venues trading the same security, lit vs. dark trades, and exchange feeds vs. SIP.
Here is what they found:
On Exchange: Quotes Dominate Trades
Some looked at the relative importance of quotes versus trades on exchanges.
A 2019 study by Brogaard, Hendershott and Riordan looked at the Canadian market. They calculated that:
- Only 40% of price discovery comes from transactions vs.
- 60% from quotes, although they were also able to show that orders improving the NBBO added most to lit price information. Amazingly, new orders at the NBBO or worse accounted for less than 4% of price discovery.
Hasbrouck, in 2021, looked at the U.S. market and estimated that quotes contribution to price discovery ranges between 60% and 70%, with the remainder coming from trades.
Finally, Benos and Sagade (2016) look at the U.K. market and find that quotes contribute to around 57% of price discovery.
Chart 2: Results of studies comparing information in exchange quotes and trades
Primary market matters more than market share
Other calculations show that a higher market share does not result in more price discovery.
For instance, Hasbrouck (2021), using granular tick data from two liquid U.S. stocks, finds that listing venues executed about one-fifth of all U.S. trading volume, but their information share is over 50%. Despite other exchanges doing much more trading, they contributed less than 50% to price discovery. We show that in the chart below, where the height of the bar shows information share, while the width shows market share.
Chart 3: Results of studies comparing listing vs trading exchanges
Consistent with these results, in the Nasdaq Nordic market, we see more than 50% of EBBO improvements originate on Nasdaq. This number is 2x to 3x larger than the next largest venue.
That suggests that listings markets are the primary engines of price discovery. That’s not surprising – listing markets are often designed to focus more on their issuers. That requires designing markets with liquidity and tight spreads – which have been shown to lower costs of capital.
Lit markets contribute more than dark markets
Speaking of market share and price discovery, we know that off-exchange venues have significant market share in the U.S. and Europe. We also know that most publish no quotes, as many also use tiers and customer IDs to segment customers, so their trades have better spread capture.
Even speed bump markets aim to trade more with less informed flow. That, by definition, means the more informed flow is concentrated on other venues. And “informed traders” add more to price discovery.
In addition, dark venues often peg their quotes and trades to prices in the lit venues anyway.
The toolkit above has also been used to examine the relative contributions of lit and dark trades. They find that lit markets contribute 86% or more of all price information.
Chakrabarty, Cox, & Upson (2022) show that lit venues provide over 86% of the information share in the U.S. equity market despite being less than 64% of the volume share (Chart 4).
Other studies have returned even lower results for off-exchange contributions to prices.
For example, Hasbrouck (2021) estimates that, in the U.S., dark markets together have a market share of 32% but contribute only around 0.2% to price discovery (Chart 4).
Similarly, Hagströmer & Menkveld (2023) analyze the U.K. equity market and find that dark pools, systematic internalizers, and periodic auctions together contribute at most 0.34% to price discovery despite representing roughly 30% of the trading activity.
Chart 4: Lit markets add more to price discovery than dark markets
Consistent with this, Comerton-Forde and Putniņš (2015) also find that, in the Australian market, dark venues’ information share increases at a slower ratethan its volume market share.
Why does this matter?
There is a lot of discussion about the value of a consolidated tape to investors. Many countries around the world (from the U.S. and Canada to the U.K. and Europe) are looking to create or enhance their own consolidated tapes.
But there isn’t much discussion about how to reward the prices that make the consolidated tape so valuable - nor which data actually adds the most to the public domain.
“It is relatively expensive for a market to provide a price discovery mechanism…In contrast, once a price has been determined and publicized, it is relatively cheap to provide order matching or crossing functions that simply pair off buyers and sellers at that price.” Hasbrouck (1995, pp 1185)
Those who set up the U.S. SIP, which creates the U.S. consolidated tape, agreed that quotes and trades have value – and users should pay something to those providing important market data. Dark pools, retail traders and even the media pay money to the SIP. However, when determining how to reward data providers, the SIP treats all quotes and trades equally:
- All shares quoted are counted equally each second, earning 50% of the economics.
- All shares traded, dark and lit, count equally for the other 50% of SIP economics.
Based on the results above, that’s not fair either. In fact, many say the SIPssupport fragmentation, rewarding venue competition more than quote competition.
Importantly, the U.S. shows how incentives affect (and sometimes distort) behavior. Some venues have been designed to capture quote credits without providing liquidity or better prices.
Setting the wrong incentives – or even worse, no incentives – risks penalizing the very markets and participants that are setting the prices everyone wants in their consolidated tapes. That not only makes markets less efficient, but it will also widen spreads and increase the costs of capital for issuers, harming investors and the economy in the process.