Exploring Market on Close Facilities, Part 1: The Americas

PETER HAYNES: Welcome to TD Cowen's podcast series, Bid Out, a market structure perspective from north of 49. My name is Peter Haynes. And today, for episode 71, we start our series on market-on-close mechanisms, which we will cover in three parts by region, first with the Americas, then APAC, and finally, we'll end up in Europe. We expect to publish parts 2 and 3 of this series over the next few months but also have planned some other market structure topics that we will sprinkle in amongst these various market-on-close pods.
Joining me today to totally nerd out on market-on-close mechanisms for the Americas are two experts from Vanguard, Nick Birkett, a portfolio manager located in Scottsdale who has been with Vanguard for the past eight years and, prior to that, worked for 11 years on JP Morgan's portfolio trading desk in London. And we have John Kraynak, who's a portfolio manager for Vanguard, located in Malvern, who's been with the firm for the past 15 years. John and Nick, thanks a lot for joining me today.
NICK BIRKETT: Thanks for having us.
JOHN KRAYNAK: Yes, thanks, Peter. It's a pleasure to be here.
PETER HAYNES: Great. So just a little background on the genesis for this deep dive on closing auction mechanics. It came to me as part of the ongoing narrative about natural liquidity patterns and how much the network effect of the closing auction has driven so much flow, not just indexer flow but natural institutional flow, to the close. And despite some common functionalities, in practice, no two closing auction models are the same. And this is because there's always subtle nuances known to local market participants that maybe aren't as widely known and, hence, the reason why I wanted to study this market-on-close phenomena one by one.
So, Nick, I want to start with the question around liquidity. As market-on-close flows increase because of increased indexed assets and more on-close natural liquidity, in theory, this should mean that trading costs for the rest of the regular session are higher. Do you believe that higher market-on-close activity translates into lower intraday liquidity?
NICK BIRKETT: Yeah, good question. In short, yes. But I guess one thing I would state is if we substitute, for the purpose of this question and for the argument, if we substitute volume for liquidity-- I know liquidity can be a deeper topic. I believe you've covered it on a previous podcast where you talk about the nature of liquidity and what makes accessible liquidity. But for the sake of this argument, if we just say increased volume, or lower intraday volume, that's a fact.
As you say, it's a global phenomenon. Cash flows are increasingly going to the close in Europe. The number's approaching 30% of the day's volume of the close. In the US, if you account for off-exchange prints, we're marching towards 20% of the day's volume at the close. So therefore, relatively speaking, there is less volume intraday relative to the close. And, interestingly, even if you look at the volumes intraday, we're seeing an increasing skew towards the close.
So in the US, you can have half of the day's volume is traded after 3:00 PM, 40% after 3:30 PM. And if we follow that through to the really important part of the question, which is regarding trading costs-- and, in this case, we're looking at trading costs in terms of market impact, so the implicit trading cost-- yes, again, it is shown in studies that it is cheaper-- and, again, cheaper in terms of implied trading costs-- to trade at the close.
And, in fact, again, just to delve a little deeper, throughout the day, it is also cheaper to trade later in the day. So it's all relatively intuitive once you account for the fact that, over the course of the day, spreads tend to tighten. As we're discussing, volumes pick up. And price volatility also reduces. But just to the point of comparing close to intraday, yeah, it's cheaper to trade at the close. And depending on which study you're looking at, we're talking 20% to 50% cheaper.
PETER HAYNES: And as you say, 50% of the daily volume happening after 3 o'clock, when you include the close, all that activity concentrating around the end of the day brings back that common question about the regular trading day and whether or not it is too long. That'll be the subject of another podcast down the road, I'm sure, when we start to dig in a little more on 24-hour trading and what its implications will be for the institutional trading day.
So, John, as you sit on the international desk in Melbourne, which essentially means that you trade non-US markets in the Americas, when you compare trading costs in general in Mexico, Brazil, and Canada, the three markets that we're going to focus on for today's discussions, what do you see as the main differences between trading in each of these three markets?
JOHN KRAYNAK: Yeah, it's interesting question, Peter. So if you look at trading costs and spread over the given day, what you'll notice is you have higher cost of tradings, higher spreads, smaller top of book early on from, basically, about the open till about 30 minutes in. And then, similar to Nick's point, what you start seeing as the day goes on is spreads begin to tighten as the day goes on. And you've seen a shift, maybe not as dramatic in the US as Nick may have alluded to, is that volume is moving towards the close.
But Mexico, in particular, due to its significant volume that it has within the last 20 to 30 minutes of the day due to its closing mechanism, you actually see almost 20% to 30%, depending on a given day, trading over the last 30 minutes of that day.
PETER HAYNES: And when you think about Brazil and you think about how trading works in Brazil, generally speaking, are the mechanics-- are there major differences between how trades go? One thing I find interesting about Brazil is when there's going to be a block trade on an individual security intraday, they actually create these auctions for the single stock if somebody wants to trade a large position. I think that's kind of cool. I think it's an interesting idea. Does it work, though?
JOHN KRAYNAK: Yeah. Actually, so Brazil is unique in that sense where, depending on the size that you're trading-- and there's different criteria, which you can find on B3's portal-- but depending on the size relative to notional or shares outstanding or percent of the company, it could trigger an auction. Those auctions could last 5 minutes, 15 minutes, an hour, in some cases, if it's due to volatility, longer, I would say, for the most part, we try and avoid kicking off an auction when trading. But the idea is sound.
So if there's going to be a large block of trades, it helps bring other market participants to the auction and allows you to trade. Brazil not only has intraday auctions but it also has closing auction, which is very similar to this auction. The auctions are pretty much exactly the same, for the most part. But I would say, for the most part, we try and avoid trading those auctions, or kicking one off. But market participants will reach out to us, and they'll make us aware of this in the event that we'd like to go along and trade.
PETER HAYNES: Yeah, it makes sense. Because, obviously, as an indexer, generally speaking, as an indexer, your auction you care about is the one that happens at the end of the day the most, over others. So, Nick, just before we dig into some of the, really, weeds associated with these different facilities, I want to start with a more general question. And it plays into your experience, having been in London for as long as you were.
Given your experience in Europe, where auctions are after continuous trading ends, and your experience trading US closing auction flow for the last eight years, where the auction runs parallel to continuous, I'm curious if you have a preference of one of these models versus the other. And does the liquidity of the underlying security, like large-cap Apple versus small-cap 3000 Russell stock, does that play into factoring which of these types of facilities you prefer?
NICK BIRKETT: Well, I hate to do this to you, Peter, but I'm not going to put my neck on the line and state which model I prefer, but I'm definitely happy to talk through the differences and the pros and cons of each. So as you mentioned, I cut my trading teeth in Europe, as it were. And when I look back to close trading, honestly, it kind of gives me shivers down the spine just thinking about this.
Because when I started in '05, we didn't have algos to trade the close. I recall we would have to line up these spreadsheets that we'd upload into GL Arwen. And Europe being Europe, of course, you had different starting times for the auctions. Switzerland would kick off, then it would be Italy, then 5 minutes later, it'd be the UK and others. And we'd just be manually firing in these spreadsheets.
The reason I mentioned that is because, of course, now, thankfully, we do have algos, but the concept still applies, which is to say, you had, as you stated, continuous ends. Then the auction begins. So that is a very clean model. You don't have the complexity that comes with this overlapping window of time. And from a trading perspective, now that we have algos and aren't having to use spreadsheets anymore, the beauty there is if you look at any close algo ticket in Europe, it's very standard to have various limit offsets that you can set from far touch, loss, whatever you make, one, and you can ladder your LOCs.
And now, coming to the US, I remember when I first opened up an algo ticket to trade the US, and I'm searching around for my strike and for my limit offsets. And I couldn't see it. And so I speak with our algo coverage. And that's not standard. And to this day, it still isn't really standard for a close algo, off the shelf, to include these limit offsets.
And I mention that just because it really is a function of the differing models. So it's very simple in the US to have this concept of limit offsets because of the fact that the continuous ends, and then you have a window of time, usually up to 5 minutes, for which you could strike and send in your LOCs and, in the US, not so much because you have these varying cutoffs.
And, of course, you can still set limits, and, of course, we do still set limits, but it is something that you, in essence, customize or you set in your OMS, or it's something you would agree with, with a PT desk you may be trading with. But you need to determine, OK, at what point will I strike my limit to make sure that I'm still going to get in before the cutoff?
So on the face of it, the European model is cleaner and simpler. The US model has that additional layer of complexity. Obviously, I'm not going to finish there because that makes it sound like there's a clear positive over the other because the US, you do have this 10-minute window that overlaps.
And, as a result, you tend to see smoother price action. So if you were to look at any European stock, it's very common that, from last to the closing print, it may move plus, minus 40, 50 basis points or in that range. And of course, if you now look at the US, it's unbelievably rare to see that kind of dislocation from last to close.
Now, of course, listeners are probably thinking, well, hold on, that's not an apples-to-apples comparison. And that is true because the price action is being absorbed over the previous 10 minutes. But hence, you do have a smoother price action into the close. And there's also an argument that that 10-minute window can be an opportunity to absorb liquidity.
PETER HAYNES: Makes for a very interesting debate because it is very difficult because of the fact that that imbalance gets absorbed into continuous trading as you converge the last sale to the imbalance that is required to clear in the US. You were going to say something else, Nick.
NICK BIRKETT: Yeah, I was just going to say, regarding the question of, for each model, does one benefit, say, a more or less liquid name. Again, no surprise, I'm not going to state definitively either way. I would just say we at Vanguard are big proponents of being out loud and early. And that is to say, both in Europe and in the US, we think, for the most part-- obviously, there's always nuanced situations-- it's good to represent your order earlier in the auction and earlier in the dissemination period. And we think that is beneficial for a stable and liquid closing print.
PETER HAYNES: So, John, it's clear that Nick appreciates the fact that you're allowed to tie in soccer because he's given us two tie-breaking answers.
[LAUGHTER]
We always have winners here in basketball and hockey and football, or at least most of the time in football. So Nick, let's shift gears to the US, where we do want to spend most of our time here. When people talk about trading equities in the US, the discussion inevitably turns to fragmentation, segmentation, off-exchange activity, more than 50% in the dark, with exchange trading during the regular session viewed as, essentially, toxic exhaust.
The only point of the day, which we've stressed here already, were the two current listings exchanges, NASDAQ and the NYSE, control the flows is on the close. What are the main differences between the NASDAQ closing auction model and the NYSE closing auction model?
JOHN KRAYNAK: OK, are we ready for another tie?
PETER HAYNES: And which do you prefer? [LAUGHS]
JOHN KRAYNAK: [LAUGHS] At the highest level, they are fundamentally very similar. Obviously, as we just discussed a minute ago, they both have the same concept of an overlapping period. In both cases, the dissemination starts at 3:50. And NYSE fairly recently made a change to include D orders, which I'm sure we'll talk more about later, also into their 3:50 dissemination.
And as far as the dissemination of the cadence, there's also very regular updates. It can be as much as every second if there are changes to the dissemination between the two. And as far as the data that is contained within that dissemination, again, naming conventions differ but very similar as far as it's in what's paired off, the imbalance, the indicative clearing price.
Where we see a difference, and where we see a difference that's worth noting, is the cutoffs are different. So NASDAQ has a cutoff for regular MOC LOCs at 3:55. NYSE's at 3:50. But as I just mentioned, the D order, the D-quote order, is really the differentiator between the two because that allows participants to trade the close, theoretically, up to 3:59 and 50 seconds. Now, good luck finding a floor broker who would want to commit to that kind of time window. So usually, you might want to get in slightly before then, but that's the hard cutoff. And, in fact, that's actually the most popular order type on NYSE if you're looking at data over recent months.
So the difference there is if we look at the formation of the book in terms of the size of the volume that's paired off and the imbalance and the predictive power of that imbalance information, NYSE forms later. So NASDAQ after 3:55 becomes, relatively speaking, more predictive, whereas NYSE is later, closer to 3:58, 3:59, which, again, intuitively makes sense when you know that the D-quote is a popular order type, and, therefore, these orders are being routed in slightly later in the window.
Now, one other difference, of course, is the human element. And maybe we can talk in a little more detail about what that entails. I would say, day to day, for day-to-day MOC trading, that's not something that I would highlight as a significant differentiator. But, again, we can talk more about that.
PETER HAYNES: Yeah, we're going to dig in on D orders, or D-quote, as they're known, and the changes that have taken place in the last couple years in terms of including that in the auction information, which I'm assuming has been for a benefit. But we'll get to that in just a minute.
So, John, let's just switch gears here and stay away from the US and dig into the other three markets. And I'm going to start with Canada. Like the NASDAQ, which Nick just explained, prior to this century, the closing price for stocks in Canada was calculated as the 4:00 PM last sale, which introduced a fairness problem where 10 million shares-- in example here-- 10 million shares of a stock could trade at $10 right in the last minute of trading. And then, with one lone wolf at the very, very last second hitting the bid at 4 o'clock, the stock could close at 9.98 on 100 shares. So we had a fairness problem.
NASDAQ, in their case, I think, used to take the print that was 15 seconds after 4 o'clock, the one that was reported 15 seconds after 4:00, and that became the close. So NASDAQ created the auction, which Nick just discussed now is in its current form. Canada went to an auction, which was a blind auction model. And it was created in 2004. And I want to tell a quick story here because I was around for that.
We were heavily involved in the creation of that facility, and we wanted the London model. We thought that that was the right way to do things, and we didn't get it. We got this blind auction. And we were very, very upset about that model. At the end of the day, within one month of the launch of that facility, there was a major index trade involving Manulife merging with John Hancock, a US insurance company. And so there were a number of US brokers that sent orders into the Canadian Manulife close because it was upwaiting, and Hancock was being removed.
And in doing so, they created a lot of offsetting liquidity. Well, lo and behold, there was a bug in the system. And Manulife's closing price in Canada was the equivalent of, like, $5 more than Hancock in the US. And we had these US brokers calling us up, saying, where can we take you to dinner? You just made my year. It cost the Canadian market several basis points of performance, which they never got back. And it was all because there was a bug in the system. So thankfully, that hasn't happened in recent years. But I got thinking about that story back in 2004, when they first launched the MOC.
So here we are. We have a new MOC that was created in 2021. Vanguard had a representative heavily involved in that consultation process. And it was a lengthy one and very, very well done. Tell us about the important features of the new TSX market on close that global investors need to know.
JOHN KRAYNAK: Yeah, absolutely. Thanks, Peter. So I think, first of all, we look at the TSX model, the new one, as being incredibly efficient. And they've added some nice new features that help reduce end-of-day volatility. And it's really increased transparency from the old model, which was very important. But yeah, let's do a deep dive through it. And I'll go through the different features that we think are important.
So let's start from the beginning. It starts at 3:50. And this is, really, what's called the imbalance period. This is in line with major US markets. That make sense. Some of these securities trade in the US and in Canada. That imbalance feed's going to run every 10 seconds. But it doesn't only show just paired volume and imbalance but it also shows a reference price, which, in this case, is the midpoint of the central limit order book.
It has a foreign indicative price, which is an estimated close based on only the MOC and the LOC orders. But then, I think more importantly, at least for us, is the near indicative price. So the near-indicative price during this imbalance period shows you the MOC and LOC orders plus all the visible orders in the TSX continuous market. So this represents those shares that can be traded in the continuous market that could eventually be recycled back into an auction.
Now, during the inbound period, you're not able to cancel MOC and LOC orders. You can only amend LOCs aggressively. And these orders can be entered throughout the day. You don't have to wait till the auction begins. But we think that's very important. So not being able to cancel an order, or fade, is very important, as that could have a large impact on your price formation at the end of the day.
Now, another new important feature that the TSX has created is the freeze period. Now, the freeze period starts randomly between 3:56 and 3:57. The goal for that is to incentivize being out loud in the auction early, which we agree with. But the freeze period is meant to, really, mitigate price volatility and allow for offsetting imbalance orders.
So if you think about that, any order that's placed during the freeze period is essentially a pegged LOC. So if it's an aggressive limit on close, it's going to be repriced back to the reference price at that point in time. So during this period, you can actually enter both sides. If there's a buy-in battle, you can add more shares to buy. And that would be viewed almost as an advertisement. If you'd like to step in and offset any of that, you can enter an order on the other side and offset some of that. In some cases, that allows for more shares to clear on a given day.
Now, at the end of the day, when the price is struck-- so it's the CCP, the Closing Calculated Price-- it's a combination of the MOC book and central limit order book. Now, an important feature that we see here, which we think is great, is price moving extensions. I don't think we see it all that much, but it's a nice feature to have.
So basically, what that means is if that calculated closing price is outside of the price movement extension range, an extension will kick in that can range up-- I believe it can run till 4:10. And then it allows for offsetting orders to come in and help, hopefully, improve price or, at least, offset the imbalance. When it's all said and done, the allocations are based on price, broker, and time.
PETER HAYNES: You did a great job explaining the Canadian market. Wow, I couldn't have thought of it, just remembering some of the details that you just went through there, myself. One of the biggest problems we had in the old facility prior to the one you just discussed was these flip-flops where you had a sell imbalance, but it ended up trading up at the end of the day, as no one could see what was going on in the book formation, which you just described a second ago, can't happen in Canada because you can't reverse an imbalance after that freeze period that you discussed. So that was a great explanation.
Interesting stuff going on in Mexico here, John, which you're involved in, or aware of. I want to talk about what is happening there, where the local primary listings exchange does not currently have a closing auction. Instead, they use an average price over that window of time, which you discussed earlier. And they're working through a proposal currently to consider implementing a market on closed facility. And I think they're looking to north of the United States border, up here in Canada, as a potential model being given some consideration. Can you explain exactly how stocks closed today in Mexico? And what can you tell us about the status of that MOC project?
JOHN KRAYNAK: Yeah, sure. So Mexico is currently one of two markets that we trade-- I think we trade, globally, 49 markets right now-- that does not have a closing auction. The only other market's India. We're actually very active there, as well. And it's really been this way for the past 25 years in Mexico. Prior to that, it was last price. Obviously, there were issues with that, so they moved to a VWAP period to help reduce volatility at the end of the day.
But let's maybe dive into the actual features that they have. So the official price in Mexico is set by a 20-minute VWAP on the BMV. The bull set's their main exchange. It's the one that's been around the longest and the one that commands the most amount of shares trade over a given day, Although BBVA has been seeing some volume flow to them, as well. On a rebound day, it's actually a 30-minute VWAP, so they extend it to try and help spread that out and reduce end-of-day volatility.
But they also have this concept of a board. And that's their closing facility in Mexico where you can post-- let's actually jump back. The market opens at 9:30 for continuous trading. But you can post on this board, if you need to have the closing price, beginning at 9:00 AM. So beginning at 9:00 AM, you can post your order to this board. It's visible to the entire market. And if someone else would like to take the other side of that trade, they can match.
Once you're matched on this board, you're guaranteed the closing price, which is the 20- or 30-minute VWAP at the end of the day, depending on what type of day it is. And once it's matched, you can't cancel. It's rather finite in that sense, where you'd better be sure that you want to trade if you're going to match on the board.
With that being said, we continuously look at markets. And it's not the first time, like I said, that we've actually looked at a market and said, hey, maybe an auction could help this market, help create more efficient price discovery at the end of the day. We've actually done it out of our APAC office in Hong Kong, Japan, currently doing it in India. For us, we try to take a very thoughtful approach in this proposal. We did a significant amount of research and analysis. We did that in-house. We have a great TCA team that helps us with that.
We partner with a sell-side broker to actually have some independent analysis, so it just wasn't ours. And then we actually had many conversations with different local Mexican market participants and tried to incorporate their feedback into our proposal. We also created a working group. This working group consists of asset managers in a sell-side firm to help formulate this proposal, to make sure we had differing opinions, to make sure that it wasn't just one asset manager's thoughts going into this.
And right now, we're at a point where, you're right, we landed on a model similar to the TSX. We did try and incorporate some features of the local market because we think that's important. And the proposal's in front of the AMIB. And the AMIB is basically an association of financial intermediaries in Mexico. I believe there's about 30 members at the AMIB right now.
And it kind of operates like a group that oversees trading functions and best practices in Mexico for the market, a self-regulating organization. And any change to trading in that market actually has to be voted on by the AMIB, which would then flow up to the exchange. So this proposal has been sent to the AMIB. We're waiting to hear feedback. And then we'll move on from there.
PETER HAYNES: Any idea on timeline?
JOHN KRAYNAK: Not yet. So we've been checking in. We are working very closely with our local Mexican office on this. The AMIB's confirmed that they've received it/ and we're going to probably check in once or twice more in the coming weeks, see where they're at on this.
PETER HAYNES: OK. We'll watch closely. I want to just hone in on, before we get to Brazil, the can't-cancel thing. So in Canada, if I put an order into the MOC facility at 10:00 AM, no one knows it's there because there's no early look. But at 11 o'clock, the client cancels the order, so I cancel. Are you saying that, even prior to any public dissemination, at 3:50, under a proposed Mexican model, that I would not be able to cancel that order I just described a second ago?
JOHN KRAYNAK: So, Peter, is this the current model, or do you mean the proposed model?
PETER HAYNES: No, the proposed model going forward. The current model you explained about the can't cancel on the board. I just was wondering whether-- you mentioned local market conventions. And I know they'll have to exist. We have to appease the local conventions. So I'm curious if that's one of the things that was done to appease, or is that not part of the model?
JOHN KRAYNAK: Yeah, actually, it is. So it's part of the model in the sense that it would be like an order type. So we looked at this, and this was very important to the local market participants. A lot of them have built a lot of their business around trading the close in Mexico. So there'd be an order type where, if you put in an irrevocable MOC and it matches against another irrevocable MOC, you cannot cancel.
PETER HAYNES: It's only once it's matched? I get it.
JOHN KRAYNAK: Yes, only once it's matched. But, look, you still have the option to just go straight MOC. And at that point in time, you can cancel anytime over the day up until the inbound period starts, which we have at 3:50. But let's be honest, we want to be very flexible with this. We're not set on hard time frames or cutoffs. This is more of a conversation, more of a catalyst to start this conversation on, is it possible for us to help implement an auction in Mexico to basically create efficient price formation and benefit all investors in Mexico?
PETER HAYNES: And would you assume that there would be a 10-minute auction similar to the one in NASDAQ and the one in Canada? Is that what people are coalescing around?
JOHN KRAYNAK: I think that's a topic of conversation. Does 10 minutes work for Mexico? Right now, they're used to a 20-minute VWAP period, which gives them more time to trade.
PETER HAYNES: And a 30-minute on the big re-bells, right?
JOHN KRAYNAK: Exactly. So I think that's where we need to keep an open mind. I think everyone involved in this would have to keep an open mind where, even if the proposal says it's going to be a 10-minute auction, maybe it needs to be 15 minutes in Mexico. Maybe the freeze period, or I think we call it, in a sense, the offset period, needs to be longer. Or maybe the whole auction, in general, needs to be a little bit longer. That would be OK. I think the important thing is if we can land on some type of model that works for Mexico.
PETER HAYNES: Yeah, we used to be 20 minutes in Canada. We went to 10 as part of the new model. So absolutely, yo have to stay within the parameters that work locally. So just before we come back to the US, I want to talk about Brazil, which is what we'll call an outlier by American standards. Because it adopted a model similar to the one that Nick was explaining earlier that exists in the European markets, which is to have an auction that takes place after regular trading ends at 4:55 PM. How long has Brazil had an auction, and what are the need-to-know features about this local Brazilian market on close?
JOHN KRAYNAK: Yeah. How long have they had an auction? That's a good question. So it's actually something that I've investigated before as part of the work that we've been doing on our proposal, part of that research. And as far as I could find, they've had an auction, the current model, since September of 2005. Prior to that, my understanding is that it was an open outcry, which is much different than where it is now. But it's moved to a fully electronic model. And it does have more of that European-based style where continuous market trading stops.
PETER HAYNES: So when we talk about what goes on after the regular trading ends, is there any transparency before the regular trading session of what might happen in the MOC? Is there a way to advertise that I'm a buyer in the MOC? Or is it all just, opens up the new trading session, which is the close, I can roll my orders from continuous in there, and then we just fight for price discovery over the next 5 minutes? Is that how it works?
JOHN KRAYNAK: In a sense, yes. There's no early look. So let's put it that way. You can enter your orders throughout the day. Once the auction kicks off, those orders will roll into the auction book. But you're not getting an early look at it like you might in the US. And you're not having any type of imbalance feeds prior to the auction kicking off. Yet something important you said, too-- 5 minutes. I think it's pretty rare in Brazil for the auction to actually go 5 minutes. Usually, they become extended. They can extend till about 4:15.
PETER HAYNES: So they can be an extra 20 minutes, then-- is that what you're saying-- after the regular closing?
JOHN KRAYNAK: Yeah. Actually, it closes regularly after 4:00 PM, I would say, depending on the day. It could be a couple minutes after 4:00. Large days, large flows, index rebalances or index changes, any corporate events that could be driving higher volume in a given name, you could see that actually close closer to 4:15.
PETER HAYNES: You mentioned you like the PME, Price Movement Extension, feature in Canada, which allows stocks to extend if there's a big imbalance that's not clearing within a reasonable price, which, at the end of the day, will be 10% in Canada. Is there anything-- you say extend these MOC, these auctions, in Brazil. But is there a limit? If there just literally is no offsetting liquidity, is there a limit on how much the stock can possibly move during that auction?
JOHN KRAYNAK: Yeah, they do, actually, Peter. So similar to different-- they actually call them tunnels. So they have volatility tunnels. They have auction tunnels, multiple different tunnels, which is basically a set of bands that, if breached, would create an extension. But really, if you look at why extensions occur in Brazil, it's not necessarily because a volatility tunnel has been breached. During the auction, when the book builds-- so at 3:55, the auction starts, continuous trading stops.
Then you see a book build. And it's very transparent, actually. It's easy to navigate that auction. But you see that book build. And during the book build, there's a theoretical auction price and shares that are calculated. And this price is basically the largest-- is the price at which the largest number of shares can trade on that given day. But any order that comes in-- so a new order comes in. If that changes what that theoretical price or shares would trade at, then it would get extended. And it'll extend for 1 minute.
Also, in Brazil, it's important to say that if you are part of price formation, so if your order is helping drive price formation on that given day, you cannot cancel. So once you're in that auction-- MOC, LOC-- if you're part of price formation, you're not able to cancel. But you can amend your LOCs more aggressively to either help improve price or the amount of shares that are trading. So even if you amend your price, and that would have an impact on your theoretical auction price or shares that are going to trade, that would also create a price extension.
So I think it's pretty common for that to happen. And it's not uncommon for your theoretical auction price to seem far away from the last traded price once the auction starts until the book starts to build. But it is, actually, a fairly orderly auction and has worked out fairly well for us.
PETER HAYNES: It sounds like there's some built-in anti-gaming devices that you just discussed there about not being able to cancel or move the imbalance last second here. OK, Nick, let's go back to the United States and focus on some of the subtleties of these two US models, which you indicated at the start were pretty similar. And I want to start with the NYSE. The NYSE still plays up the role of the human in its market model. What is the role of the floor specialist in closing a stock on the NYSE?
NICK BIRKETT: So the Designated Market Makers, the DMMs, they do oversee the closing auction. Now, in reality, it's largely electronic. And, therefore, it largely goes off without human intervention. But they do have the ability to pause and uncross if there has been a large price move. Now, that's not necessarily saying it needs to be 10% or the threshold to 10 by 90. It could just be a significant price move from last. So at that point, they can pause the uncross.
And the DMM actually has the ability then to help stabilize the price and take the other side of the significant imbalance that still exists. Or they can open up-- the term they use is they open up the pinhole, which is a way for them to solicit interest from other institutions to, again, help stabilize the price at the close. That's the case, which doesn't happen too often. Let's be honest. So it's very rare to see this kind of human intervention because, for the most part, the closing auction passes relatively uneventful, and it's pretty much fully electronic. But there is that ability for the human to step in.
PETER HAYNES: And we talked earlier about D-quote, or D order. I want to dig in a little bit about this order type that is unique to the New York Stock Exchange MOC. Tell us what is a D-quote order, who can enter D-quote orders, and how did the process of entering D-quotes change in the past couple years, once it became completely electronic?
NICK BIRKETT: So basically, the D-quote has replaced what, historically, was the verbal interest. And now we're going a decent way back, certainly, before when I started trading in the US. But previously, that's where the floor brokers would verbally be able to represent their order right up to the last minute when they were approaching the pitchers there in NYSE. So the D-quote has now replaced that mechanism.
You need a floor broker, so that's the first thing to say. So to access the D-quote, you need to have a broker on the floor that is able to route those orders on. But it is, basically, electronic. So it's something that, again, assuming you have the floor broker, it gets routed directly through. The cutoff is, as I think I mentioned earlier on, the cutoff is 3:59 and 50 seconds. And right up to that time, you can modify, cancel, D-quotes.
PETER HAYNES: It is an order type unique to the brokers. Given that these floor participant brokers can also get a look inside the book, pre-imbalance publication, which is a feature that seems to be unique to New York, when you combine it with the fact that you can have these D-quote orders coming in right to the very last 10 seconds, do you feel like this gives market makers, or floor brokers, an asymmetric advantage, such as the imbalances cannot really be trusted?
NICK BIRKETT: I'm not sure I'd go that far. So you're right. The floor brokers have visibility of the book prior to 3:50. So from 2:00 PM onwards, they can actually check how the book is forming and, therefore, the nature of the imbalance up till that point. But the studies we've ran and we've looked at show that there is some predictive power in the imbalance at 3:50 that is immediately absorbed by the market, within milliseconds. We haven't seen evidence that there is price movement prior to 3:50, let's say. That could be a function of having visibility of those order books, or should I say the auction book.
Now, bearing in mind, you cannot do that systematically. It's not the same way the feed is systematically published after 3:50. That's not the case with visibility of the books prior to. But I guess the one, maybe, defining argument as to why I wouldn't go as far as saying it can't be trusted or there's some asymmetric advantage that we should be concerned about is the imbalances shift. It's very common on NYSE to see imbalances flip, which, again, if we then think back to the D-quote and that time frame, that is a function of the fact that this order type, which is the most popular order type, is still regularly being used right up to 3:58, 3:59. And hence, the volume is still building all the way through that time frame.
PETER HAYNES: It's interesting, Nick, because that model of flipping, you have a D-quote. You have an order comes in. There's a 50,000 buy imbalance. And then there's a D-quote. Sell 100. Now you're a 50,000-share sell imbalance. Now you can only enter D-quote to offset that. The flip-flopping that goes on at the end of the day, which used to happen in Canada and elsewhere, was frowned upon.
But it seems to be an understood convention in New York, whereas on NASDAQ, you have some cutoffs that prevent that type of flip-flopping happening towards the end of the day. I want to get into what are the important order types on the NASDAQ MOC that we should know about. When are those cutoff times that guarantee a close, both on the demand and supply side of an auction? And how does the exchange ensure that an imbalance can't flip from the buy to sell side in those last few seconds?
NICK BIRKETT: So I'm going to continue to sit on the fence here whilst I'm answering these.
PETER HAYNES: 0-0 tie in the soccer game.
NICK BIRKETT: Yeah, hopefully no draw. They do have that earlier cut-off, 3:55. And similar, it would seem, to the convention in Canada as far as your ability to set an indicative offset after that is constrained, as in you can't set too-aggressive limits. And, obviously, it needs to be offsetting the imbalance.
All of that being said, yes, there is more stability to the imbalance in NASDAQ than there is at NYSE because of the fact that the significant portion of the volume for the close has already formed at 3:55. Because, afterwards, it is just continuous order book price action, as well as some indicative offsets that would potentially cause an imbalance to flip. But it's far less common than it is in NYSE.
Six of one, half a dozen of the other. Let's be honest. The argument would be, well, if it's more stable, is this more supportive of less volatile price action? We haven't seen the studies yet that would allow you to make that assertion. So there is a difference. And fundamentally, it, therefore, results in the order books forming in different time windows. But that's as far as I would go.
PETER HAYNES: It's very possible we'll have a third US exchange, perhaps with its own unique model, down the road if Texas has success bringing some listings to its marketplace, which it plans to launch sometime later this year and into next year. We're all watching that with very close interest. So, John, I want to ask a couple of questions about the Toronto Stock Exchange MOC. Unlike the United States, the model in Canada does not include exchange-traded funds. There has been discussion, especially around when we launched the MOC in Canada, the revised MOC in '21, about adding ETFs. Do you have a position on whether they should or should not be included in a MOC?
JOHN KRAYNAK: That's an interesting question. It's one that I think we need to research more here before formulating an official opinion. But if you look at it, on one hand, we know the closing auctions help create more efficient price formation at the end of the day. It helps create a liquidity event for people to show up and trade. But we also know that ETFs are not single stocks, and they utilize the arbitrage mechanism in the primary market to help facilitate efficient trading over the day.
So, Peter, what I can say here is when I speak to our US capital markets team, they regularly recommend to our clients to not trade the open or the close in the US for ETFs, mainly due to increased volatility or the potential for price dislocation. So generally speaking, I think more research needs to be done. I think we look forward to reviewing that and, hopefully, being part of that as it goes forward. I look forward to seeing where this goes.
PETER HAYNES: Yeah, John, it's an interesting question. When we did our research on this, wondering why there's no problem with an imbalance on a US ETF on ARKA being filled, and in Canada, everyone's worried, well, I can't get the ARB trade off because I don't know if I'm going to get fills because, as you say, it's based on the underlying, we got the impression that, on ARKA, if there was an imbalance, the market makers are willing to take the mean reversion trade, just sell it off ringside or where it closes, assuming they're going to be able to win on the mean reversion trade in aggregate. And yet there doesn't seem to be that same comfort level in Canada if and when there would have been ETF MOC imbalances. So you're right, more study needed.
The other topic that's interesting in Canada is how much of our on-close auction volume is paired. And we've learned all about this since 2021. Do you think there should be some form of a marker that identifies these pairs so as to help with trade cost analysis? Because these trades, these paired volumes, are not accessible volume, but they are included in the volume calculations and can distort on close volume. What do you think about how that should be handled?
JOHN KRAYNAK: If you look at pretrade analysis, to start, you want to be certain what percent of the median closing volume, or average closing volume, whatever metric it is that you look at, that you are around that amount that you intend to be, or you could actually have market impact. And if you look at trading-- say, electronic trading, for instance-- schedule-based algos, they could potentially be impacted by this.
So VWAPs will usually default to participating in the closing auction. And for that, they're going to use forecasts over the day on the volume and then, also, historical volume on the close and probably try and forecast out that day what that closing volume is going to be. And if those forecasts are off of volumes that could potentially be inflated, let's say, due to paired-off volume that could impact it.
Another type of algo that we use would be a close-based algorithm. This type of algorithm, we would designate how much of the potential close that we'd want to be. It would calculate out when it needs to start trading to have that order finished, begin trading into the close, and then go into the close with, let's say I want to target 15% of the median closing volume-- Same impact. So if that closing volume, the forecast is higher than it actually is what I can access, I could end up being more of that volume than I intended to be.
Now, if you look at this from the post-trade standpoint-- it's also just as important. The post-trade analysis, typically done by our TCA team, reviews algo execution, quality, and performance and also helps provide trader coaching to traders and provides feedback. But if the numbers that they're looking at are inconsistent-- 5% order and a 15% order have similar impacts-- that could cause issues. So I think it warrants a look. I think it's important for all market participants to have a clean look at what that actual closing volume is that they can access.
PETER HAYNES: It can really be distorted, as we've seen in some of the analysis in Canada. So, Nick, there's a lot of noise about broker-sponsored closing auction facilities in the US, which, in practice, as I understand it, are just a way for brokers to pair their own MOC orders internally and then print them, post the close, on the TRF, and therefore avoid exchange fees, which can be high.
Exchanges will argue that this practice harms price discovery. Meanwhile, brokers will argue that the trades hit the tape either way, and whatever residual imbalances they have will be placed in the MOC, so why does it matter? Do you have a position on this issue of broker-sponsored closing auctions, good or bad?
NICK BIRKETT: I'd say we evaluate the merit of these alternative facilities as they come and were generally in favor of innovative approaches. I'm not sure I'd state, in each case, whether it's necessarily good or bad. I personally might have an agreement with the notion that the net trade is what ultimately impacts price versus a cross. But I would probably point out that we're generally concerned about the increasing costs of closing auction. And therefore, competition that can help drive down that cost is certainly not a bad thing.
PETER HAYNES: Yeah, it gives, definitely, some competitive pressures on the incumbent exchanges. I've heard some rumblings from institutional customers about brokers that are managing schedule-based trades, or even sending block orders into their own close matching engine, without notifying their customers. And this is frustrating from the transparency perspective. Do you always have an understanding if a trade that you're executing with a US broker is being sent into, routed to, or being part of an internal MOC? Or do you really care?
NICK BIRKETT: Yeah, we certainly do care. And I would probably start by saying this is where tagging is absolutely critical. So John referenced our TCA team. They do a phenomenal job. And part of their job is verifying venue tags and running analysis on venue tags. So that allows us to have that transparency to know where those executions have come from. And we do engage with brokers in advance to determine whether we want to be part of their internal matching or a guaranteed close facility.
And honestly, there's no right or wrong answer to say you should or you shouldn't. It can be very much order-specific, case by case. I would give, for example, one area where we do not allow brokers to use guaranteed closing facilities, and that's for our target close wheel.
So that is a decision we make for the benefit of our ranking and understanding of how well brokers perform. We do need to see when those orders commence trading. And therefore, we can run analysis to better understand the potential impact that has been had. And that's, therefore, just an example of where we would definitively communicate to our brokers that are in the target close wheel that they cannot use a guaranteed closing facility.
PETER HAYNES: Got it. So we're just going to finish up here, final question for each of you. I'm going to go with you, John, first. What advice do you have for traders that are searching for on-close liquidity in the three main markets we discussed that you trade?
JOHN KRAYNAK: I would say, especially for Brazil and Canada, be out loud in the auction early. So if you need to trade to close and you have some, or all, of it that you're going straight MOC with, makes sense to be out early and advertise your size. This helps bring any offsetting participants to the auction and allows for more orderly close.
I'd say, on top of that, understand what your risk is. So we have a PM trader model here. So I'm a portfolio manager on different portfolios, but then I'm also trading. So I know what the impact of those given trades are on my portfolio. I understand that's not the same model everywhere. But if you understand your risk and how much you have to complete, you can use limits. So you should use limits when applicable to help reduce the impact, manage your risk. Chances are, if the stock moves up 5% on the close, you probably don't need to buy it at that point, and it may revert the next day, anyway.
For Mexico? Mexico is a little unique. So Mexico has the closing board, which we spoke about. It has the VWAP period. Understand if you want to advertise in that board and take that close for a portion of your order. If not, you can utilize the VWAP at the end of the day, whether it's 20 minutes or 30 minutes. You could VWAP without a limit or with a limit if you want to be price-sensitive. But understand your risk, and make your decisions based off of that.
PETER HAYNES: Well, longer term, we're going to have an auction, hopefully, in Mexico. We won't have to worry about the post to the board or deal with VWAP situation. So Nick, over to you. What advice do you have for traders in the US markets who are navigating these end-of-day auctions?
NICK BIRKETT: Since John and I both work at Vanguard, it's probably no surprise that I'm going to echo a lot of his sentiments. And I will just repeat the idea of being out loud and early. The studies show that that's conducive to improved liquidity and price stability. So we are proponents of our orders, again, where appropriate-- MOC and relevant LOC limits.
As John said, this is also all about managing risk. And therefore, you need to intelligently set your limits and use MOC where appropriate. But without being too much of a parrot, just the point about being out loud and early, we think, is a very good way to approach the closing auction, assuming you're looking for outcomes like ours, which are stable liquid closing events.
PETER HAYNES: As I said, and think about what we just talked about for the last 50 minutes here, you have two portfolio managers working for a large asset manager who are basically telling everyone else how to trade the close. But what it shows is you're managing your own orders. Everyone else is managing their orders. But we're in this together to make sure, as an industry, we use best practices. And where we can share those best practices back and forth, without giving away state secrets, then I think we've all won.
So we're in this together, and I really appreciate the fact that you were willing to provide some insights into the markets you're trading for others out there that will benefit from that information and make the markets better, overall. So thank you on behalf of TD Securities for coming on to the podcast. And we'll definitely continue this conversation going forward.
NICK BIRKETT: Thank you.
JOHN KRAYNAK: Thanks, Peter.
[MUSIC PLAYING]