Cowen Market Structure: Retail Trading – What’s going on, what may change, and what can you do about it?

Key points:

• With retail trading on the rise, we estimate that “inaccessible liquidity” in the market has made up more than 37% of total volume so far in 2021. 
• On the other hand, less than 30% of total trading volumes have been single stock, accessible, non-HFT volume – the true accessible liquidity for institutional investors.    
• Retail trends have also led to fundamental shifts in intraday trading – with more volume moving towards the open (and away from the close), increased volumes in single stocks (versus ETFs), and a degradation in the displayed volumes that make up our NBBO.
• The rate of inaccessible liquidity on a stock by stock basis can vary widely, with no real way to discern intraday levels in real-time.  However, Cowen’s research on execution costs indicates that stocks with the highest rates of inaccessible liquidity see significantly higher slippage versus arrival costs when executing liquidity-seeking orders.
  • This indicates that increasing rates of retail trading, and how those orders are executed away from public markets, can have measurable implications on institutional execution costs.
• Due to the challenges retail trading can present for institutional traders, Cowen is incredibly excited to introduce an update to our algos that allows clients to opt in to “haircut” participation rates to account for inaccessible liquidity on a stock-by-stock basis. 
  • This means that the algos can now naturally adjust to mitigate chasing inaccessible volume in stocks where rates of retail trading may be higher.
• Finally, we share some insights on why we don’t see an appetite for regulatory change to the retail trading framework or inaccessible liquidity in the near future. 

INTRODUCTION

The rise of retail trading has been one of the primary stories in market structure over the past two years. This astronomical pick-up in individual investor participation in the market began with the move to free commissions at several retail brokerage firms in late 2019.

It was further fueled by the pandemic, as people began trading from home throughout lockdowns, quarantines, the cancellation of major sporting events, and rounds of stimulus payments. And we have seen retail trading rise to even higher levels in 2021, as headlines, blogs, YouTube videos, and social media personalities all highlighted the incredible volatility in so-called “meme stocks”. When your great Aunt Frannie, your Uber driver, or your neighbor asked you to explain what was going on with Gamestop, it was probably hard for many of you to avoid the feeling that we had reached peak fervor among retail investors – or at least levels of interest not seen since NYC taxi drivers were pitching tech stocks to brokers on their way to work in the late 1990s.

While the rise of mom and pop (or Roaring Kitty) is not a new story for our clients, we hope to provide some new insights on this new market dynamic. We will show how the change in the mix in investors is causing fundamental shifts in U.S. equity markets – from where volume trades, to when it trades, to what vehicles investors are using to access markets.

We also want to show how the rise in “inaccessible liquidity” in the market impacts institutional investors. Our analysis has found that stocks with higher rates of inaccessible liquidity, which is often driven by retail trading, tend to have higher execution costs.

We also will discuss the regulatory landscape, and why we don’t think there is an appetite in DC to tackle some of the key elements of retail trading – like rising levels of individual trading activity on app-based platforms or the payment-for-order-flow model.  Finally, we are also excited to provide some innovative new solutions to our clients – including an automated adjustment for retail trading on a stock-by-stock basis when you’re executing orders.

For the time being it looks like retail trading is here to stay.  So here is the breakdown of how in impacts markets, how it impacts your orders, and what you can do about it.

CURRENT RETAIL TRADING LEVELS

While there’s no indicator on the tape that denotes retail trading and allows us to track it in real time – we have a few metrics that we can look at that point to record levels in recent months. The first is the sheer amount of volume that we are seeing so far in 2021. If we look at a chart of average daily volumes in U.S. equity markets over the prior 20+ years, you can see just how “off the charts” 2021 volumes have been year-to-date. Through last week, we were averaging 15.2 B shares daily, which is almost 50% higher than last year’s levels, which also set records.

If we zoom in and look at individual trading days, 5 of the top 15 days in terms of shares executed (2008 – 2021) occurred in 2021.  If we look at notional value executed, all the top 15 days fall in either 2020 or 2021.

And yet, as you talk to institutional traders, many seem to say that while they have been busier than perhaps the old “normal” it hasn’t felt like they have been experiencing record-setting volume days or even levels we saw back in February/March 2020. The liquidity one would expect, given these aggregate volume numbers, simply is not there.So where is all this volume occurring and why do we point to the retail investor?

ISOLATING RETAIL TRADING LIQUIDITY

Retail trades are unique in how they are handled in the U.S. Most retail orders are sold under “payment for order flow” agreements (PFOF). Market makers buy the retail order flow and they give these orders tiny fractions of a penny of price improvement over the best bid or the best offer. The retail investor gets a slightly better execution price than they would have received if their order had been sent to the exchange. The order is executed immediately and cheaply, and – in return – the market maker captures the majority of the spread.

Market makers print most of these shares internally at their firm, so they trade off-exchange. One way we have for isolating retail volume is to look at the share of volume that trades off-exchange, but not in a dark pool. We refer to this as “inaccessible liquidity.” This is because most institutional orders – whether they are executed via algos directly or by high touch desks – primarily go to exchanges and dark pools. So, from the perspective of the algo or the client trying to participate with volume, a lot of this off-exchange / non-ATS (dark pool) volume is “inaccessible” to them.

If we look at total off-exchange volumes, they have been on an absolute tear over the past 2 years.  We now consistently see over 40% of volume executing away from the 16 stock exchanges, and we are quickly approaching a new norm, where the majority of volume trades away from the lit markets.

If you break down this volume further, you see a slightly different story. Despite the rise in off-exchange market share, ATSs/dark pools, which are incredibly important sources of liquidity for institutional investors, have been consistently losing share over the past 5 years. In the first several weeks of 2021, ATSs have only made up 8.7% of total U.S. equity volumes. The non-ATS volumes, on the other hand, now make up as much market share as all off-exchange trading did in 2019. So far in 2021, over 37% of total volume executed was what we would call “inaccessible liquidity.”

If we further drill down into the data and look at the specific firms where this off-exchange/non-ATS volume is trading, a large chunk of it is executing at so-called “wholesalers” – the firms that execute payment-for-order-flow trades.  Analyzing 5 of these larger firms – their internalized volume (so volume that executed internally at their firm, not in an ATS) made up 30.6% of the total 37.7% of inaccessible liquidity. If a good portion of that volume is retail liquidity, we can see in the chart below that these numbers have been on a steady upward trajectory since the move to free trading at several retail brokerages in October 2019. We did see a slight dip in activity in the fall of 2020, but it appears that retail returned with a vengeance as we entered 2021.

One final way that we try to isolate retail liquidity is by looking at the universe of names that institutional investors tend to traffic in (versus the names that perhaps more individual investors tend to trade). Some have suggested that a good deal of institutional flow tends to be concentrated among Russell 3000 names that are priced above $5. What we have seen over the past 2 years is that R3 names over $5 have become a smaller and smaller share of overall U.S. equities volumes. Normally around 50% of all trading, Russell 3000 stocks priced above $5 have made up less than 35% of total volume so far in Q1 2021.

On the other hand, stocks under $5 have represented a rising share of the total market, and an increasing percentage of total Russell 3000 volumes (see below).  The migration of trading to lower-priced stocks has been, at least in part, driven by individual investor activity in these names.

We wanted to see just how unique the sheer volume of trading occurring in low-priced stocks has been in recent months. To do this, we looked at index performance and average execution prices over the past 12+ years. Historically, there has been a consistent trend that the average execution price across all U.S. trading is closely correlated to the performance of major indices. This makes sense – as the markets go up, the average execution price per share also goes up, and vice versa.

Over the past few months, however, we see this correlation break down (circled below).  Despite rising markets, the average execution price across all U.S. trading (total notional value traded / total volume traded) has actually been dropping.  We’re now below $40, which is around where we were in late 2016/early 2017, when the market was nowhere near recent highs. The sheer volume of trading in low priced stocks, some of which is driven by retail investors, has been driving this new trend.

CHANGING DYNAMICS IN THE MARKET AS A RESULT OF RETAIL TRADING

Many of the metrics we look at are pointing to unprecedented levels of retail involvement in the market. We couldn’t help but think that this must be having an impact on the very structure of the market. We pulled a few different metrics and found there have been some fundamental shifts in not only where volume is trading, but when it is trading, what types of securities investors are using, and the very quality of the price formation in the market.

THE VOLUME SMILE IS MORE LIKE A VOLUME “SMIRK”

One of the most telling shifts has been a reversal in the slow but steady shift in trading to the end of the day. The volume smile has long-since resembled more of a volume “smirk”. Over the past several years we have heard people talk about this as a self-fulfilling prophesy – more volume trades at the close, so more people implement strategies to target the close, scheduled models adjust to account for more volume around the close, and then even more volume moves to the close, and so on and so on.

It became very common to see names trading 30% or 40+% of their total daily volume from 3:45 to the close. As some people began to contemplate moving to a 15-minute trading day, it felt like this was a trend that could not end. And yet, if we look at the past year, we have seen that the volume curve is starting to flow back in the opposite direction – towards the open.

The below chart compares the overall market VWAP curve from August 2019 (before the flood of “free trading”), February 2020 (just before COVID-related volatility began), and from the past few weeks. You can see that the first hour of the day has picked up market share over this time, while there has been a corresponding drop in volumes in the final hour. Retail investors have historically placed a lot of their orders at the beginning of the day – at night to execute the next day or before they begin work in the morning. The increase in retail volumes seems to be, at least in part, causing a transition of intraday volumes back to the beginning of the day.

The below chart zooms in on the first and last 15 minutes of the day. You can see the drop in volumes at the end of the day in greater detail. This is not to say that institutional investors are changing their behaviors because of this shift. We are still seeing many clients favor the close for its tighter spreads, depth of book and reduced execution impact. We’re also still seeing many clients avoid the open – particularly as names seem to whipsaw more than ever with the rise in retail investing. That makes this trend even more surprising. Despite the fact that the largest investors aren’t necessarily changing their behavior, the sheer volume of retail orders seems to be enough to reverse the concentration of volume around the close – a trend that almost seemed irreversible to many market participants 3 years ago.

ETFS LOSING SHARE

Another seemingly irreversible trend has been the movement to passive investing. While assets may be holding steady in index funds, we looked at ETFs as a percentage of total trading volumes to get a sense of what vehicles new investors are using to access the markets. Whereas ETFs have been holding steady around 15-20% of total volume the past 10 years or so, in 2021 we have seen that number drop to 12.2%. This seems to suggest that this new wave of new retail investors appear to be more focused on individual names versus broad sector or index exposure. Regardless of whether day-trading single stocks is the best point of entry for a novice investor, new retail participants entering the market seem to be placing their “chips” on single stock bets as opposed to ETFs.

INCREASED ODD LOT TRADING

The final trend we want to point out is the increase in odd lot trading, which corresponds with retail activity. A lot of trades from individual investors are smaller orders – many below 100 shares. Some retail firms now even offer fractional share trading.  With the increase in retail investing, we would expect to see a rise in odd lot executions. The SEC’s MIDAS data (below) includes odd lot trading on a symbol-by-symbol basis, so we reviewed the 2-year trends in S&P 500 names.

While we haven’t seen a sustained upward trend in odd lot market share, this is at least in part due to the fact that most retail trading occurs off-exchange, where we don’t have MIDAS data. However, we can still glean important information from this. The key element of odd lots that trade on exchange is that many of them are not displayed on the consolidated tape. Unless they can be aggregated up into a round lot, an order for 99 shares or less – even if it is priced several cents better than the best bid or offer – will remain hidden for most investors. We often see this in high-priced stocks, where there is a great deal of odd lot liquidity priced within the spread. There are proposals in place to add odd lots to the SIP, but in the interim, high rates of odd lot trading driven by retail participation (especially in combination with high rates of off-exchange retail trading) means higher rates of hidden volumes.

ANALYZING THE COMPOUNDING EFFECT OF ODD LOTS AND RISING OFF-EXCHANGE TRADING

We wanted to analyze the compounding effect of odd lots and rising off-exchange trading, so we dug a bit deeper into S&P 500 volumes in 2020. If we look at the volume that traded off exchange, and then added up the non-displayed trading on-exchange (between both hidden orders and odd lots), we get the below chart.

We estimate that on-exchange displayed trading only made up 41.7% of total trading in S&P 500 names in 2020. Displayed trading on exchanges is what makes up the quotes you see on your screen. It’s what makes up the National Best Bid and Offer – or the reference prices that dark pools use to execute your orders, market makers use to price-improve retail orders, and all best execution analyses are based off. As the percentage of displayed volume that is traded on-exchange declines, the quality of our market data degrades. This has a real impact on all investors – retail and institutional alike – who rely on those displayed quotes to price and execute orders.

HOW MUCH OF THE RECENT VOLUME HAS BEEN REAL LIQUIDITY?

The above dynamics have real implications for our clients – whether it’s the quality of the prices they see on their screen or the time they choose to enter their orders. But retail trading has other direct impacts on institutional investors – impacts that are perhaps more tangible as clients work to execute larger orders. Retail executions generate growing volumes that can both move markets (sometimes substantially) and yet are largely entirely inaccessible to institutional traders.

Clients regularly reach out and ask if we have any insights into whether retail investing is driving a particularly sharp move in a stock or the market as a whole. Behind those questions is a sense that it’s not institutions. With retail and wholesaling and other non-institutional forms of trading on the rise, it begs the question – just how much of the activity we’ve seen so far in 2021 has been institutional activity, or even just single stock liquidity that institutional clients could actually access?

To answer this, we took the following steps:

• First, we removed ETF volumes, which accounted for 12% of volume in January 1 – February 5, 2021 (the time period for which Finra dark pool data was available).
• That left us with 88% of volume in single stocks.
• Using dark pool data from Finra, we pulled out what percentage of total volume in single stocks we would consider “inaccessible” – or volume that didn’t trade on exchanges or in dark pools. This accounted for nearly 40% all the single stock volumes, which amounted to 34.7% of the total volume pie.
• Of the remaining 53% of volume, we assumed that 55% of that volume was pure intermediation (automated market making that simply buys from a seller to sell to a buyer) or high frequency trading (HFT) strategies.  At 55% of the remaining 53% of volume, we would place intermediation / HFT in single stocks at 24% of total volume.
• That leaves us with just 29.3% of volume in single stock, accessible, non-HFT volume.  See below for the breakdown for January 1^st through February 5^th.
• In other words, we estimate that less than a third of all volumes so far in 2021 has been single stock liquidity that institutional clients could participate with.

INACCESSIBLE LIQUIDITY IN SINGLE STOCKS AND RISING EXECUTION COSTS

The above chart shows inaccessible volumes in the overall market.  The challenging part of today’s trading is that these numbers have huge implications for execution quality on a given order, yet they vary widely from stock to stock, and there is no way to identify “inaccessible” liquidity on the tape in real time.  If 35% of volume is the market-wide average, how can you know if the stock you’re trading a stock where the rate of inaccessible liquidity may be closer to 50%?

The 2 below tables show the top 25 names in terms of inaccessible liquidity in both the S&P 500 and the Russell 2000.  In general, we have historically told clients that popular “name brands” tend to consistently top these charts.  However, as we’ve seen above, today’s retail investors seem to be changing up their tactics and are focusing on more diverse investments (sub-$5 names, single stocks versus ETFs, Reddit recommendations, and so on).  With retail on the rise, it’s becoming harder than ever to isolate how much inaccessible liquidity is a factor when executing an order.

Why is it important to know these rates of inaccessible liquidity?  Because we have found – in analyzing all of our liquidity-seeking flow over the past year – that there is a correlation between high rates of inaccessible liquidity and increased slippage versus arrival.

The below chart shows a breakdown of S&P 500 names by decile (decile 1 being the names with the most inaccessible liquidity). We analyzed all liquidity-seeking flow executed via Cowen’s algos on a stock-by-stock basis over the past year. What we found was that decile 1 names, which had an average rate of inaccessible liquidity of 30.6%, saw over 10 bps of slippage per 1% of ADV executed.  That was significantly more than any other decile. We also see increased costs in decile 2 and 3 names, before slippage figures start to level out at deciles with around 15% inaccessible liquidity and below.  We somewhat assumed this would be the case – of course “chasing” volume that is not fully accessible to you would cause impact.  However, we were still surprised to see the trend so clearly demonstrated in our execution data across such a wide range of orders and stocks.  What this indicates is that the increasing rates of retail trading, and how those orders are executed away from public markets, can have measurable implications on institutional execution costs.

ADUSTING FOR INACCESSIBLE LIQUIDITY IN COWEN ALGOS

So, the final question is, what can we do about inaccessible liquidity and the costs associated with it? Cowen is incredibly excited to offer a new solution for clients. We now have an option for clients to adjust their trading for inaccessible liquidity on a single-stock basis as they execute orders in our algos. On a weekly basis, Cowen will calculate the percentage of inaccessible liquidity in each U.S. security using the prior 3 months’ FINRA and off-exchange data. Clients will be able to opt in to adjust participation rates in the algos, reducing aggression on a single stock basis to account for that specific stock’s rate of inaccessible liquidity. 

The algos all use a calculation of “eligible volume” to determine minimum, maximum and target participation rates. They decide when to trade or how much to trade based on this eligible volume “clock”.  By subtracting an estimation of inaccessible volume from eligible volume, the algos will naturally adjust their aggression for inaccessible liquidity on a name-by-name basis as they work.  Below is an example of how this would work in our SEEK algo for two different stocks – TSLA and DIS.  Essentially the algo is “haircutting” eligible volume to account for the stock’s inaccessible percentage.

Without the client having to know specifically how much inaccessible liquidity is a factor in TSLA versus DIS, the algo has automatically adjusted to haircut the TSLA order’s participation rate by 45%, whereas the DIS order has only been cut by 30%.

We know there are times when 15% means you just need to be 15%, so we’ve made this an opt in feature only. Clients can apply the Inaccessible Liquidity Adjustment to certain strategies or aggression levels by default. Or they can apply it on an order-by-order basis, by entering something in the “custom” field of their ticket.  There’s a ton of flexibility in terms of how we can apply this to orders.

We also acknowledge that clients have, to a certain degree, naturally adjusted for inaccessible liquidity across the total market over the past several years by bringing down their target participation rates on a lot of strategies. If you apply a haircut to participation rates that have already been lowered across the board, you run the risk of slowing down the algos too much.  Therefore, one of the ways we would suggest implementing this would be to apply it to a custom strategy that has higher participation rates. The key here is that these rates will be significantly lowered for the names where inaccessible liquidity is a large problem, and they will not slow down orders where inaccessible liquidity isn’t really a factor. That is the beauty of the stock-specific Inaccessible Liquidity Adjustment. 

Below is a sample of a custom BEST algo with relatively high participation rates. It gets slightly more aggressive when a stock moves in your favor, and slightly less aggressive when it moves against you. But the key, as the below TSLA example shows, is that the participation rates in the algo will automatically adjust for inaccessible liquidity. The realized participation rates account for the fact that this is a stock where you don’t want to be chasing 45% of what hits the tape. It makes this adjustment without the trader having to know exactly what that inaccessible percentage is for TSLA at the outset of the order and manually amending aggression, which hopefully eases work flow a bit as clients execute a wide array of names.

REGULATORY LANDSCAPE

We’ve spent a lot of time on retail trading and inaccessible liquidity over many years (and over the past several pages of this note, should you make it this far). We’ve measured the impact it can have on institutional investors. We’ve tried to present innovative solutions to help clients with the unique challenges retail trading can present. We guess the final question that remains is where do we go from here? Where does this retail fever end? Our short answer is that we don’t necessarily see an impetus for change anytime soon. 

From a regulatory standpoint, many people have asked whether there is any risk to the payment for order flow model or free trading. At this point, we (and our Washington Research Group policy counterparts) don’t see a regulatory appetite for curbing free trading. The drawbacks to the model (impact on displayed quotes which lead to wider spreads and less-informed execution prices, increased slippage in your retirement accounts due to inaccessible liquidity, etc.) are so much more difficult to explain than the very, very simple concepts of FREE trading and price improvement.

Beyond that, we think our colleague Jaret Seiberg put it best when he said, “Americans long ago accepted the premise that they will give up data in exchange for free stuff. This happens every time someone uses a free service on the Internet.” People are ok with their orders being sold in exchange for free executions. We don’t see an appetite for a congressional or regulatory intervention that would take that free trading away. This is particularly true at a time when the pool of retail investors has never been more diverse – a fact that was touched upon several times in a recent Senate Banking Committee hearing on GameStop, Robinhood, and the state of retail investing.

WILL REGULATORY ENFORCEMENT INCREASE?

So, if there’s no policy change, what about enforcement? Do we see any enforcement actions coming out of the recent volatility in so-called “meme stocks”?  While enforcement actions are not out of the question, what is important to keep in mind is the timeline on these actions. Typical regulatory investigations and settlements tend to take a long time to come to fruition. There were recent headlines about an approaching regulatory settlement with Robinhood, however the focus of the settlement were outages that occurred last March. That’s at least a year lag, and frankly that’s relatively fast in enforcement time. If market participants are hoping for short-term actions to address retail trading apps or activities, that’s just not the timeline our regulatory agencies work within.

SOME POSSIBLE REGULATORY CHANGES

There are a few changes we see potentially coming down the pipe. There is increasing demand for shortening settlement cycles, as settlement processes and associated capital requirements were largely blamed for the decision to halt trading in single stocks on some retail platforms. However, despite widespread support for this change, it will likely take a few years to implement. The move from T3 to T2 was a multi-year process. 

Another idea we have heard kicked around would be more clearly marking off-exchange trades on the tape – for example, denoting whether a trade occurred on an ATS/dark pool or fell into that non-ATS “inaccessible” bucket. Many people have said that having a better indication of these rates in real time would help clients to adjust their trading strategies for current market conditions (like a swell of retail activity). While this also has a decent amount of support, and we think it’s a worthwhile idea, the unfortunate fact is that regulatory actions like this also take several years. On the bright side, it does feel as though the SEC is actively examining this issue – even in a period of transition – and is seeking ideas from the investment community on how to address rising rates of off-exchange and inaccessible volume.

CONCLUSION

As we have for many years, we will continue to keep a close eye on inaccessible liquidity and other trends in liquidity, client costs, and other market structure metrics.  However, if you have any questions on these issues or this data, please don’t hesitate to reach out.