The NYSE closing auction is the busiest time in the US equity market trading day. On average, nearly 300 million shares, or just over 8% of total NYSE-listed volume, are executed in the NYSE closing auction. As the only open outcry trading floor for equities, NYSE combines sophisticated technology and human judgment for the smooth functioning of this auction.
One way to access the Closing Auction’s liquidity is through an order type that replicates a Floor Broker’s manual interaction by including discretionary pricing instructions. This order type is currently known as the “d-Quote”, and as the NYSE migrates NYSE-listed trading to the Pillar technology platform in August 2019, it will become known as the “D Order”. For participation in the closing auction, its function will remain the same; it can be used intraday, or a Floor broker can designate that the discretionary pricing instructions will be active only for the closing auction, which will be known as a “Closing D Order”. Closing D Orders provide additional flexibility for auction interaction.
Below is an overview of the Closing Auction order types and functionality, highlighting the use of Closing D Orders.
1Order types are shown using forthcoming Pillar order type names
On average, Closing D Orders have accounted for approximately 35% of total NYSE Closing Auction volume. We see lower share for these orders on event days such as rebalances and month-ends, due to the huge influx of MOC & LOC orders on these days. This indicates that a very high share of volume on these days is trading on the close due to an event, meaning that there are additional opportunities for traders to discover outsized liquidity on these days (see related post on this topic).
Beginning at 3:50 p.m., NYSE publicly disseminates closing auction order imbalance information, providing information about the level of buyers and sellers in a particular security. At 3:55 p.m., the NYSE includes Closing D Orders in the closing auction order imbalance information. The imbalance data currently updates as frequently as every 5 seconds; when NYSE completes its migration to Pillar the frequency will increase to as much as every 1 second.
Key data points include:
|Imbalance Side||Buy/sell direction of imbalance shares|
|Reference Price||Used to calculate Continuous Book Clearing Price (generally last sale)|
|Paired Quantity||Number of shares matched at the Continuous Book Clearing Price|
|Continuous Book Clearing Price||Price where all better-priced orders on the side of the imbalance could be traded|
We previously discussed the impact of volatility on listed options market quality in 2018. While metrics such as indices and trading ranges can provide some measure of volatility, we are introducing quote volatility (QV), a quote-based metric as a more specific, and more flexible, measure of price volatility in the market. We have found that in times of market volatility, the QV metric correlates with real-time changes to execution venue selection, and may be a worthwhile indicator for considering execution scenarios. Likewise, QV may also be a useful metric for determining when the market has returned to normal conditions following volatility-inducing news events.
The QV metric is calculated using second-to-second “quote returns.” This quote return is calculated by averaging the midpoints of all NBBO updates for a security within each second of the day from 9:35am to 4pm, and then calculating the percentage rate of return of these average quote midpoints from one second to the next. The variance of returns are then calculated in aggregated time periods (e.g., 5-minute buckets) and annualized from seconds to 6.5-hour trading days to 252 trading days in the year. Finally, we take the square root of the annualized variance in the aggregated periods, creating the quote volatility metric.
Markets volatility spiked in December 2018, presenting opportunities to apply the QV metric. The VIX had an average closing value of 24.95 in December. However, that month was significantly more volatile between December 17 -31 than it was December 1-14; the index increased 35% (21.08 to 28.44) from the first half of the month to the second half.
In our analysis, we focused on the most active names and measured the QV metric across all components of the S&P 100. We found that the metric increased 24% (23.06 to 28.62) from the 1st half of December to the 2nd half1. While QV increased sharply with the Fed rate decision on December 19, we found that volatility remained elevated throughout the second half of the month.
As expected the QV metric was highly correlated with market volume, and offered insight into execution venue selection. On December 19, market volume leapt at 2pm with the news event. As with QV, trading volume remained high as the market continued to fall throughout the afternoon. The extent of this volatility increase also impacted trading venue selection. Ten minutes before the announcement, TRF market share was 43%; immediately following the announcement, TRF share dropped to 22% and remained around 25-30% for the rest of the day.
We believe the QV metric may be a highly relevant volatility indicator for execution scenarios. As seen in the December 19 example, this metric can help correlate with changes such as execution venue selection and may be useful to determine when the market has returned to “normal” conditions after a news event.
1 For December quote volatility analysis, Dec. 24 was excluded due to the shortened trading day. It was a highly volatile day, but the difference in timing for a shortened day on the 5-minute intervals would have a significant impact on the analysis.
We also omitted 2 symbols on separate days in which there was a period in the day that they were either halted or had unusual quoting activity for several 5-min intervals; inclusion would significantly skew overall results. The exclusions were NVDA on Dec. 11 and LOW on Dec. 12.
The charts below show the aggregate unexecuted interest available during the Closing Auction near the closing price for specified trading days. By sending orders with looser limit prices, traders may gain access to substantially more volume with only minimal impact to the closing price.
For example, in the S&P 500, 36% more volume could have traded within 10 basis points (bp) of the closing price on an average day in March. In the Russell 2000, 161% more volume could have traded within 50 bp of the closing price.
The above charts show all interest, both displayed and non-displayed (such as LOC orders). Additionally, monitoring the displayed depth on the NYSE leading into the Closing Auction can reveal substantial interest priced near the eventual close.
The Closing Auction accounts for more than 8% of NYSE-listed volume thus far in 2019. As the above data shows, the auction’s liquidity has attracted a deep book of opportunistic volume priced near the eventual closing price. This creates unique liquidity opportunities for traders to execute block-size liquidity with minimal impact by entering LOC orders with looser limit prices or providing further discretion to their floor brokers.
With a record year behind us, we wanted to follow up our 2018 Options Market Review with an analysis of the options market in Q1 2019. The market experienced a drop in volatility from the highs of Q4 2018, which also brought a reduction in equity option volumes. Market quality benefited from the low-volatility environment as quoted size and some spreads improved over the quarter and options volumes became significantly less concentrated in the top 50 symbols.
Volatility was driven downward as the VIX closed below 13.00 for the first time in over 5 months on March 15, 2019. Lower volatility depressed options volumes, with Q1 2019 equity option ADV finishing at 17.3mm, down 12.9% from the prior quarter. Despite this reduction in volume, Q1 2019 was the second most-active Q1 of all time, trailing only Q1 2018.
As volatility decreased, posted size increased significantly. Notional size in Penny class symbols increased over 75% and in Non-Penny class symbols increased over 80% from Q3 2018 to the end of Q1 2019. Market-wide quoted bid-ask spreads (measured in basis points) stayed relatively flat as spread widths tightened at the same rate as option values decreased. Quoted spreads in Non-Penny symbols were an exception as they tightened by 28.5%.
Data using methodology described in 2018 Options Market Review and excludes SPY.
Q1 2019 also saw reduced equity option volume concentration. We found volume less concentrated in the top 50 symbols by ADV, and market volume attributed to the top 10 symbols dropped 8%. Volumes actually increased from the prior quarter in symbols ranked 51 through 350, which broadened market activity into previously less-liquid symbols.
Options markets currently show an implied probability of VIX being 15 or lower of about 65% 3 months out and 55% 6 months out. The reduced uncertainty yielded an improvement in displayed options market quality with larger posted size and flat or narrower bid-ask spreads. Although volumes have decreased in the most active options, interest in some of the less active symbols increased to this point.
The NYSE carefully reviewed the recently published report from MITRE entitled, " Scaling of Inefficiencies in the U.S. Equity Markets: Evidence From Three Market Indices and More Than 2,900 Securities." While we appreciate the complexity of the data and calculations required to produce the report, we believe there are serious flaws in the methodology. These issues lead the report to overstate the scale of the alleged issue.
The report investigates "latency arbitrage opportunities and realized opportunity costs" from trading in stocks in the Russell 3000® index during 2016. The paper attempts to identify trades that occur when different data feeds show different prices at the same time. The paper estimates that 23.71% of all trades occur when the National Best Bid and Offer (NBBO) calculated from exchange proprietary data feeds is different from the NBBO published by the Security Information Processors (SIPs). The paper estimates that these trades result in an annual "realized opportunity cost (ROC)" of greater than $2 billion. This result is substantially different than the Bartlett and McCrary paper (2017)1, which estimated that the SIP and direct feeds showed a different price for just 3% of all trades and that potential opportunity costs were an order of magnitude lower.
We believe there are a few key issues leading to an inaccurately high cost estimate.
Most importantly, the authors based their ROC calculations on the quantity of each trade without consideration of the actual quoted quantities available on the data feeds. Consider the following example:
The authors informed us that they calculate the ROC on the 200 shares executed at the SIP NBB of $10.00, or $0.01 per share, resulting in a $2.00 ROC. The paper assumes both quotes are equally accessible at the same instant, but disregards the fact that 200 shares could not have executed at $10.01 on Nasdaq; Nasdaq was only showing 100 shares available on its direct feed at the better price.
Extending this example, if multiple trades executed at $10.00 on different venues, before the SIP quote was updated, each of these would be counted towards the ROC. This implies that potentially any number of shares were available for arbitrage, when there were really only 100 shares that could offset any trade. For there to be an arbitrage opportunity, one must be able to execute the opposite leg of the full quantity of the trade.
Our second key issue is that we are not clear how the authors handled odd lots quoted on direct feeds. If they calculated an NBBO based on odd lots they would further exacerbate the over-estimation described above. The authors could avoid this problem by looking for executions in odd lot size at the prices quoted on the direct feeds. If the odd lot trades occur, it is clear that the shares are no longer available and any subsequent trades at the SIP NBBO could not possibly be afforded an arbitrage opportunity.
Third, we learned from the authors that they sourced the unified SIP and proprietary feed data from Thesys based on data they collected in Carteret, NJ. We believe running such a study from a more neutral geographic location, such as Secaucus, NJ, would lead to significantly fewer perceived arbitrage opportunities. While the authors took a measured approach to try an account for the geographic latencies between the three New Jersey exchange data centers, those latencies are not static and can vary based on a variety of factors, such as spikes in market activity and telecommunications performance.
Fourth, the executions used to calculate potential ROC were sourced from the SIP rather than from the proprietary feeds. Exchange proprietary feeds publish executions in addition to quotes, and we believe that using executions from these proprietary feeds would indicate far fewer potential arbitrage opportunities.
Due to these issues, NYSE concludes that the results presented in the paper are flawed and do not reflect actual latency arbitrage opportunities and costs.
1. How Rigged Are Stock Markets? Evidence from Microsecond Timestamps, Robert P. Bartlett II, Univ. of California, Berkeley and Justin McCrary, Univ. of California Berkeley, Columbia University, NBER
The jump in market volatility in 2018 has provoked several concerns regarding the nature of today's electronic trading environment. The NYSE's unique hybrid market model, which features a Designated Market Maker (DMM) obligated to provide liquidity and facilitate auctions, offers unmatched stability relative to other global markets, especially during times of market turbulence.
NYSE's auctions differ from those at other equities exchanges because they combine human judgment exercised by the DMM who oversees trading activity, with automated mechanisms. This enables NYSE to conduct a manual auction if deemed necessary by market conditions. To conduct the auction, the DMM aggregates buy and sell interest and executes the maximum volume at a price intended to be reflective of market supply and demand.
The manual auction therefore gives the DMM both price and time flexibility, unlike electronic auctions which because of their automated nature are more susceptible to large price dislocations due to temporary supply and demand imbalances — a scenario which occurred recently during an auction on the Singapore Stock Exchange (SGX). The risk of fast, automated executions has led many global exchanges, including NYSE Arca, NYSE American, and Nasdaq, to implement auction "collars" with others coming on line shortly1.
These collars set price thresholds to help to dampen auction price movement. However, the NYSE Opening Auction consistently outperforms electronic auctions, especially during volatile periods. The NYSE Closing Auction, which is more than 7.5% of total NYSE-listed trading activity, similarly outperforms other auctions.
The chart below shows the weekly count of symbols that experienced price dislocations2 above 1% for both NYSE and Nasdaq Closing Auctions, and the difference between the two measures. As the chart shows, while auction price dislocation tends to increase during periods of heightened volatility, the dislocation in the NYSE Closing Auction is far less severe compared to the dislocation that occurs in the Nasdaq Closing Auction. This is attributable to the DMM's ability to help find relevant closing prices even in challenging market conditions3.
2018 was a historic year for equity markets, with major indices reaching record highs before retreating to finish down for the year. Volatility made a comeback, but how did it compare to recent years?
Volatility returned to US markets in 2018, leading to record equity options trading volumes. At the same time, there was increased interest in options market quality, particularly as it relates to the displayed quote.1 Given NYSE’s advocacy for robust displayed markets, and the unique nature of the year’s volatility and volume, 2018 was an ideal year to study the influence of volatility on quoted spread and size. We found that adjustments for option prices can dramatically alter the interpretation of market quality metrics: after converting quoted bid-ask spread values from cents to basis points and converting quoted size from contracts to dollars, the data shows that displayed market quality was fairly consistent during some of 2018’s volatile periods.
Our analysis leveraged NBBO quote data provided from S3 and was limited to the top 15% of symbols by options ADV. We focused on options with underlying symbols that had at least 1,000,000 shares of ADV, removed any new listing and delisting, and removed symbols that changed Penny Pilot groups during the year. Finally, we excluded SPY so as to not skew the results. SPY accounts for nearly 20% of options volume, and this scale likely gives it market structure attributes not common to the rest of the tradable universe. The resulting list of symbols accounted for 71% of total equity options volume.
We calculated our normalized spread as a percent of average option price in basis points:
NBBO Spread (% bps) = Avg NBBO Spread Width ($) / Avg Option Trade Price ($)
Size was converted from contracts to dollars by calculating notional size:
NBBO Notional Size ($) = Avg NBBO Size × Avg Option Trade Price
In the charts below, we plotted monthly average spread and size for Penny Pilot (PP) and Non-Penny Pilot (NPP) symbols. In 2018, NPP names had a spread of roughly 3x the spread of PP names, accompanied by 60% greater size compared to PP names. NPP posted size fluctuated over the year, ending the year higher than it began. We also observed that spread widths for NPP symbols fluctuated more than PP symbols in nominal terms, but less so using normalized measures.
Nominal spreads followed the fluctuations of the VIX closely. We saw nominal spreads widen with the elevated VIX levels in February and March, as well as in October through the end of the year. When we adjust the spread calculation to show percent of option price in basis points, spreads still appeared wider in February, but were more muted the rest of the year. We hypothesize that the speed at which VIX spiked in February, moving from 13.47 to 37.32 in two trading days, had a greater impact on spread widths compared to the more gradual rise of VIX across months late in the year when it rose from 11.61 to 36.07 in 56 trading days. The normalization shows that although nominal spreads increased, they increased proportionally and in tandem with option prices. In our population, average option prices increased by 31.5%, whereas spread widths only increased by 25.5% in Q4.
The impact of these price normalization steps can be particularly meaningful at an individual symbol level. In nominal terms, high-priced symbols Google (GOOG) and Amazon (AMZN) had some of the worst spreads and posted size. However, in notional terms, GOOG had the tightest quoted spread of all non-penny symbols (422 bps) and moved much closer to the average posted notional size of $85,000; AMZN experienced similar results. Additionally, household names such as McDonald’s (MCD), Caterpillar (CAT), and Stanley Black & Decker (SWK) exhibited tightened quoted spreads after we converted from nominal terms to basis points. We believe that converting spread and size to notional terms may change perceptions of market quality and improve the ability to compare among symbols.
Unlike the US listed options market, many global markets of various asset classes use notional values for quote and trade increments. The above analysis details the merit of using notional values for assessing US listed options market quality, especially in highly volatile times. This perspective shows that the US listed options market handled 2018’s volatility fairly well, particularly when the elevated volatility was sustained over a period of time.
At last month's hotly debated SEC market data roundtable, there was passionate discussion over both the nature of competition and fees charged for market data and market access. Some around the table, including IEX, chimed in that exchanges should only charge transaction fees. Market data and connectivity services, they said, should be provided for free.
We disagree. Exchanges are integrated enterprises that bring broad value through trading and technology services, but not all clients use every exchange offering. Allowing investors to order a la carte empowers each to select the blend of services they need. By contrast, charging high transaction fees but "giving away" data and connectivity is like requiring all diners to order a prix fixe tasting menu. Importantly, the current competitive market structure allows investors both choices. While the blend of exchange revenues has shifted from purely transactional to a mix of execution and technology, the overall costs on NYSE markets remain a bargain. In fact, the All-in Cost to Trade on NYSE Group exchanges is lower than on IEX or on many dark pools.
Keep your eyes on the prize: the All-in Cost to Trade on NYSE Group exchanges is less than $0.0007 per share. For our members, compare that to the $0.0009 transaction fee per share that IEX charges for most of its trading, and to a typical ATS fee of $0.0010 per share.
In fact, NYSE's competitiveness is understated in the comparison above as 1) the NYSE group calculation includes some fees associated with multiple NYSE markets, including NYSE Options, and 2) there are still datacenter and telecom fees associated with trading on IEX and dark pools, even if they are earned by unregulated datacenter and telecom providers instead of the trading venue operators.
Some market participants have understandably asked for a full breakdown of our calculation. We are happy to oblige, and stand ready to discuss any market participant's specific All-in Cost to Trade upon request.
Applying separate fees for market data and connectivity allows members to choose the right bundle of products for their specific business. For example, members of multiple NYSE exchanges do not necessarily subscribe to the same market data products or use the same number of logical order ports on each venue. Some participants add and remove liquidity evenly, while others are skewed more toward one style of trading. Because market participants choose different trading, data and connectivity services, it's entirely possible that any given member's All-in Cost to Trade may be lower or higher than the average in our calculation.
What's your own All-in Cost to Trade?
Reach out to us at firstname.lastname@example.org to find out.
In today's U.S. equities markets, roughly 60% of volume executes on an exchange and 40% executes off-exchange in dark pools and other broker-dealer facilities. All activity, both on and off exchange, relies on the quoted prices from exchanges to inform transaction pricing. This makes the National Best Bid and Offer (NBBO), reflecting the best quoted prices from all exchanges, a key benchmark for all types of trading, including midpoint trading leveraged by institutional investors and price improvement offered to retail investors. We have seen that the quality and quantity of quotes contributing to the NBBO can vary dramatically between exchanges, which we can measure using a new metric called "Enhanced Quoted Spread."
The NBBO that facilitates midpoint trading and retail price improvement arises from robust competition among exchanges to provide the highest bid and lowest offers for the longest portion of the day. One standard calculation for quoting performance is the exchange's average quoted spread. Taking a simple average of quotes published by an exchange, however, can hide an important fact: many exchanges offer two-sided quotes for only a small portion of the day in many stocks. This means that an exchange's "average" quoted spread may exist for only fleeting moments of the day, and market participants looking to execute on such an exchange may frequently find the venue does not offer a competitive quote (or sometimes any quote at all).
The Enhanced Quoted Spread (EQS) measure addresses this by replacing any missing quotes with the value of the Limit Up Limit Down (LULD) band.1 If an exchange has a one-sided quote, or no quote at all, we assign that exchange the LULD band price rather than drop the observation. With this method, exchanges with occasional or periodic quotes incur a penalty for their lack of displayed liquidity rather than misrepresenting a tight but infrequent displayed market as narrow on average.
For many exchanges, the EQS calculation is similar to the average quoted spread calculation, especially in active stocks. For example, in active NYSE names, NYSE, NYSE Arca, and Nasdaq have nearly equal quoted spread and EQS calculations. However, exchanges with low market share and/or dark-oriented trading models fare worse under the EQS approach in both active and less-active stocks. For example, three venues exhibit EQS metrics of several times their standard quoted spread results, indicating they frequently have no displayed quote in the market.
As many investors have focused more attention on off-exchange trading, the exchange contribution to price formation has become frequently overlooked or even derided, even though off-exchange trades rely on exchange quotes to set prices. As the Enhanced Quoted Spread shows, contributions to price formation vary widely among exchanges. Maker/taker venues consistently outperform other venue types in both average quoted spread and in the EQS measure, suggesting that under today's market construct pricing incentives contribute positively to both the quality and reliability of displayed quotations.
1 The NYSE wishes to thank David Weisberger, who gave us the idea for this calculation.
The price of a closing auction on the primary listing exchange determines the official closing price for most liquid corporate stocks and exchange-traded products (ETPs). However, many less-liquid securities do not receive sufficient closing-price interest to generate an auction. In such cases, the official closing price is based on the consolidated last sale price before the end of trading. This presents a distinct problem for less-liquid ETPs: if the underlying index/underlying fund holdings and corresponding market maker quotes have changed, but there have not been any consolidated last sales in the closing minutes of the market, the consolidated last sale value will not reflect current market pricing. This will then cause a greater disparity between the market price and its underlying net asset value (NAV). Around two-thirds of listed ETPs fit into this category.
To address this issue, in June, NYSE Arca introduced a process for setting an official closing price for its listed ETPs that better reflects the true value for ETPs that do not end trading with a closing auction. The new Arca Official Closing Price (AOCP) methodology for NYSE Arca-listed ETPs applies when an ETP does not have a closing auction, or the closing auction is an odd lot. Previously, the AOCP for ETPs without an eligible closing auction was the consolidated last sale, regardless of when the last sale occurred - be that days, weeks or even months prior.
The new AOCP methodology uses both consolidated last sale and National Best Bid and Offer (NBBO) inputs. NYSE Arca tracks the time-weighted average midpoint price (TWAP) of the NBBO over the last five minutes of the trading day. The TWAP and the consolidated last sale are blended based on the last sale time to determine the AOCP, according to the following schedule:
AOCPs derived solely from quoting activity account for 44% of total closing prices, while AOCPs reflecting a mix of trade and quote activity account for 5% of the total.
To assess the new logic's performance, we compared the difference between the AOCP and NAV Price before the logic took effect (May 1st 2018 - June 1st 2018) and after the logic was implemented (June 4th 2018 - June 30th 2018). We measured the average difference between the AOCP and NAV Price (at close) by product for each time period for products without auctions and then compared them across the before and after periods. As expected, the average difference between the AOCP and NAV tightened dramatically with the new logic.
As expected, this performance improvement relative to NAV is most pronounced in the least-active products. For ETPs trading under 10,000 shares per day, the median difference between the AOCP and NAV decreased from 60 basis points previously to 10.6 basis points using the new logic. Because the AOCP sets the reference price for the 'Limit Up Limit Down' (LULD) mechanism on the next trading day, an AOCP more closely aligned with the NAV will reduce the potential for an LULD trading pause to be triggered. LULD is invoked during times of volatility and sets a percentage level above and below which a security can move within a five-minute period. A security is automatically halted if the price would move outside the set percentage levels. An incorrect reference price is likely to lead to more LULD trading pauses because it does not reflect the value of the security.
As these results demonstrate, the new AOCP logic produces closing prices closer to NAV for the nearly two-thirds of ETPs that close without an auction each day. In addition to helping fix hard-to-explain premiums/discounts due to stale data, a more accurate official closing price also helps provide market protection with better reference prices to LULD. Furthermore, since ETPs report market price returns versus returns on NAV, stale closing prices can inaccurately skew performance data. This is of particular concern on month-end, quarter-end, and year-end dates.
Our previous post highlighted how end investors could potentially bear increased costs as a result of the SEC's proposed Transaction Fee Pilot. As we expected, the post triggered a significant amount of public debate, as well as discussion between the Exchange and members of the buy and sell-side. This is an important topic worthy of discussion.
This follow-up post provides additional detail of our original calculations. We have also prepared a sensitivity analysis highlighting that substantial costs would remain for investors even if our reasonable assumptions prove, in practice, to be either too aggressive or conservative. Finally, in the interest of inviting parties to reach their own conclusions, we have created an interactive model that enables readers to input their own assumptions related to venue and liquidity type distributions. By providing their own data, readers can see the resulting estimated impact.
As noted in our introduction, we are providing a spreadsheet that enables users to input their own assumptions so they can arrive at an estimated annual impact from their firm's own data. The model includes a robust set of venue and liquidity action variables, enabling users to customize volume mixes for variables such as add/take, standard/inverted/dark venues, etc. We also include a Yes/No variable for cost-plus or pass-through pricing models. Many of the questions generated by our initial post related to volume and activity assumptions, and we expect that this model will enable readers to review their own activity distribution and see the resulting impact estimate.
We consider the substantial debate around our original post a welcome outcome. We achieved our goal of encouraging discussion of the possible impacts of the SEC's proposed Transaction Fee Pilot. We hope that future commentators will attempt to include substantive and quantifiable data in support of their stance, as we have tried to do here. We welcome feedback and continue to believe that the proposal will result in increased costs to investors due to wider spreads. We agree with the general view of many who have commented that no one can precisely predict the future and that several assumptions are required to model the possible results of the pilot. In our view, we believe that costs for end investors to take liquidity will rise.
After much anticipation, the SEC has proposed a “Transaction Fee Pilot,” which would impose additional price controls on exchange access fees and rebates. As proposed, all equity exchanges (but not alternative trading systems (“ATS”) or other over-the-counter (“OTC”) trading venues) would be required to reduce access fees and/or reduce or eliminate rebates on 3,000 stocks for a period of up to two years. While some commentators equate a lower access fee with a better trade price, we have seen little analysis of the Proposal’s actual cost or benefit to investors. To fill this void, we are presenting two approaches that attempt to roughly quantify the Proposal’s potential impact on investors.
The analysis involves numerous assumptions, and we welcome any and all feedback. First, we assume that a reduction in access fees will result in a reduction in rebates. Second, we assume that with a lower rebate, spreads will widen.
The widening of spreads is generally accepted as a cost to investors because of the related increased transactions costs, particularly for agency liquidity-seeking order flow. Importantly, a wider spread will result in higher trading costs for this type of flow regardless of whether the order trades on an exchange or an off-exchange venue that derives prices from exchanges.
As demonstrated in the chart below, we find that as access fees decline, the cost to investors will increase by at least $1bn, increasing to nearly $4bn should such changes be applied to the entire market. While all investors would absorb the costs of wider spreads, the benefits from the proposed reduction in access fees would accrue primarily to sell-side brokers and proprietary traders
We first estimated costs using a top-down approach, which applies the proposed Fee Pilot changes to current average market-wide statistics. We assumed that rebates on trades in securities in each proposed Trade Groups would fall by the same amount as access fees would fall. For Group 3 (the “no-rebate” group) we assumed that market forces would reduce the access fee to $0.0002. We expect Group 3 to settle at a rate below Group 2’s $0.0005 cap as there is no rebate allowed on the other side of the trade; we also note that flat-fee venues which charge both sides of a trade today are generally priced between $0.0000 and $0.0003. As shown in the following table, this yields a blended access fee reduction of $0.00082 per share.
In order to find the expected new average spread, we identified the following calculation to apply the impact of the rebate reduction to consolidated spreads:
New Consolidated Spread = Current Consolidated Spread + Rebate Reduction * 2
The Current Consolidated Spread is the median market-wide bid-ask spread, and the Rebate Reduction is the $0.00082 blended average fee change. The Rebate Reduction is multiplied by 2 as we anticipate market makers will adjust both their bids and offers to account for the new pricing structure. This calculation results in a 1.1% increase in average spreads, to 28.1 basis points (bps).
As noted by the SEC in its proposal, brokers that are subject to exchange fees and rebates generally do not pass those costs/credits to their customer. We therefore assess principal and agency flow differently as principal flow is impacted by both explicit exchange fees and spread costs, while the ultimate customer behind an agency order incurs spread costs but usually does not pay explicit exchange fees. We also assume that the principal flow benefit from the fee reduction applies to maker/taker activity, but the higher spread cost applies to all principal and agency flow in the market.
Our cost to investors is found by calculating the cost to cross the new, wider spread; our cost to principal traders is found by calculating the cost to cross the new, wider spread netted against the benefit from lower access fees. Spread costs here are considered to be ½ the quoted spread for liquidity-taking flow, per standard transaction cost analysis measurement of performance against arrival price.
Our results show, on net, an estimated cost of $1.08bn to the industry, of which $721MM would be incurred by agency flow.
We believe that this result is somewhat conservative, primarily due to the assumptions of 1) no change in quote size despite the wider spread, 2) no shift in venue market share, and 3) applying the NYSE and NYSE Arca principal/agency ratio despite the fact that the market-wide agency taking share is much higher. This second assumption likely limits our estimated cost substantially, as a quick glance at major retail brokerage firms’ 606 reports indicates that nearly all held market orders are executed OTC. These conservative assumptions are offset by the exclusion of taker/maker (i.e., rebate to take and fee to add) venues’ impact on principal flow, the assumption that all agency flow does not pay explicit exchange fees, and by not assigning any benefit to liquidity-providing agency flow from a wider spread. We also assume a representative amount of volume in each of the pilot groups, which could be incorrect in either direction.
The below chart shows the distribution of the spread cost increase and the access fee decrease for the proposal’s three groups compared to the current market average. This again assumes an even distribution of liquidity characteristics across stocks. The access fee paid by brokers is small relative to spread costs in today’s world, and could fall as much as 93% for Group 3 stocks.
We also estimated changes from eliminating rebates across the market as a whole. We used a “bottom-up” approach that looked at the difference in quoted spreads for each stock trading on Cboe EDGX Exchange, Inc. (“EDGX”, which is a maker-taker venue) and Cboe EDGA Exchange, Inc. (“EDGA,” which is a flat-fee venue). EDGA and EDGX are very similar in that neither is a listing market, and both operate on the same technology in the same location. Accordingly, any differences in spreads between the two markets could be due to the different pricing models available on each exchange.
In aggregate, the EDGA average spread is roughly twice that of EDGX, but there is substantial variation by symbol. To account for this variation, we applied the difference in spread to the current consolidated spread for each symbol, capped that difference to 25%, and then further limited the maximum spread difference to the ratio of the primary exchange spread to the EDGX spread (these limitations were to account for the variance between venues and the fact that we are modeling a world with narrower differences in exchange pricing). The chart below shows the differences in average quoted spread between these two venues, the primary market and the consolidated quote.
We believe that eliminating rebates would widen spreads, as demonstrated by EDGA’s wider spreads relative to EDGX. Accordingly, applying this wider spread to current trading activity of all NMS securities on all equity exchanges would result in an impact of roughly $3.8bn per year, once again born largely by agency liquidity-taking flow. We also checked this result by setting all stocks to group 3 in the first model; our result in that case was a similar $3.7bn impact.
To recap, we have used two different models to assess the impact of reduced fees and rebates on liquidity-seeking flow. We find a $1bn cost from the proposed Transaction Fee Pilot, rising to $3.8bn should such limitations be applied across the market. As stated, any such analysis requires numerous assumptions, and we encourage input from market participants on how we could further refine this assessment of investor cost.
- Kevin Tyrrell and Steven Poser
On April 9, 2018, the New York Stock Exchange broke with 225 years of tradition and began trading stocks listed on other exchanges. By April 25th NYSE was trading more than 8,000 total names across the NMS universe - over 5,400 of which were listed on Tapes B (regional exchanges) and C (Nasdaq). By the beginning of May, NYSE had achieved an average market share of 0.75% in Tapes B & C trading.1
The NYSE market model for Tapes B & C is similar to the existing model for NYSE-listed securities. While Tapes B & C names do not benefit from a Designated Market Maker (DMM), the NYSE Floor Brokers and Supplemental Liquidity Providers (SLPs) continue to play key roles and together account for nearly 30% of liquidity provision.
Across all participants, activity is slightly more concentrated in active names compared to more-established venues. The top 100 most-active Tapes B&C names on NYSE account for about 61% of total volume compared to 56% at other maker/taker venues.
Source: NYSE TAQ and NYSE internal data. May 1 – May 4 2018
NYSE currently offers two benefits for passively trading Tape B and C securities: 1) a relatively thin limit order book while participants adapt their strategies for the venue, and 2) the parity execution model. Orders entered via a Floor Broker share a portion of incoming order flow, resulting in quicker fills than standard price-time priority venues. This means that orders can get filled quicker and incur less immediate reversion relative to other venues.
The chart below shows short-term price reversion statistics for resting orders filled at the full spread (i.e., buying on the bid or selling on the offer). NYSE outperforms all other maker/taker venues on both Tapes B and C, showing results similar to inverted and flat-fee venues.
Source: NYSE TAQ. April 25 – May 4 2018
The NYSE market model has also facilitated strong quoting performance in the new symbols right off the bat. The chart below shows NYSE’s time at the inside for the most-active Tape B & C names. The full symbol rollout was complete on April 25; starting the next day NYSE was at the inside in these names more often, on average, than individual taker/maker or flat fee venues.
Source: NYSE TAQ. Data reflect the 200 most-active stocks on Tapes B & C. April 25 – May 4 2018
We expect market quality statistics such as time and size at the inside to continue to improve as more participants adopt the venue into their strategies, which should drive increased market share. We are going to continue to track our performance in active names, particularly those with long queues, where the differentiated NYSE execution model can add value.
Adding Tapes B & C to NYSE is the latest step in the on-going NYSE Pillar migration, to be followed by the launch of NYSE National and the transition of Tape A trading on NYSE.
1Source: NYSE TAQ. May 1 – May 4 2018
The Treasury Department’s 2017 Capital Markets report recommended that “issuers of less-liquid stocks, in consultation with their underwriter and listing exchange, be permitted to partially or fully suspend UTP for their securities and select the exchanges and venues upon which their securities will trade.”
The argument for eliminating UTP trading is that the limited liquidity in these stocks could aggregate on one venue, reducing the friction associated with accessing a broad number of venues for a shallow pool of volume. However, many of these stocks are already highly concentrated on a small number of venues, so removing UTP trading may have limited impact on available liquidity.
We compared securities in the Wider Tick Pilot Control Group with issues in the S&P 100. As shown in the table below, exchange fragmentation is lower in less liquid securities, with the primary exchanges providing more than 43% of the total displayed quote size. However off exchange fragmentation, as represented by TRF market share, is higher in these same less liquid securities.
We also measured market concentration using a standard economic metric called the Herfindahl-Hirschman Index (HHI). This measure is 67% higher for the tick pilot group than S&P 100 issues when measuring share of liquidity provision by exchange at the national best bid or offer (NBBO). An HHI level above 2,500 is considered highly concentrated; less-liquid and active names are charted below against the wireless and supermarket industries to highlight the relative concentration in the quoting of these securities.
As shown above, primary listing venues already display a substantial portion of the quoted size in less-liquid names. At the same time, off-exchange trading is significantly higher in these names relative to active names, meaning the primary exchange is providing more price information to the market but not receiving a proportionate increase in executions. Aggregating displayed liquidity on a single venue, while allowing off-exchange trading to continue in its current form, could exacerbate this situation as queues lengthen and market makers have more incentive to trade off-exchange rather than compete to tighten spreads in the displayed market. We therefore believe that any such program to concentrate trading on fewer exchanges should be accompanied by rules requiring meaningful price and/or size improvement in the OTC market.
- Steven Poser
1We chose the control group, because securities in the test groups have seen a substantial increase in fragmentation as market makers seek to improve their queue position when providing liquidity.
2Sources: NYSE TAQ Data
FCC - Annual Report and Analysis of Competitive Market Conditions With Respect to Mobile Wireless, Including Commercial Mobile Services
Bloomberg: Little Reason to Fear Amazonopoly With Whole Foods Deal
We’ve discussed the importance and liquidity of closing auctions, but during periods of high volatility the opening auction is also a key source of liquidity. What’s more, the NYSE opening process reduces investor transaction costs by tens of billions of dollars per year.
Opening auctions on NYSE, like IPO auctions and closing auctions, are overseen by the Designated Market Maker (DMM). DMMs can open a stock in an automated manner or, depending on the situation, run a manual auction to aggregate interest and open at a more iterative price. This is especially important for stock-specific events such as IPOs or openings after earnings, or when market-wide volatility increases.
To assess the opening auction’s performance, we analyzed price discovery on the most volatile days in Q1 2018 (Feb. 5&6, March 1&2, March 26-28) relative to price discovery on other “standard” days. We measured open price discovery by comparing the open auction price with the market VWAP over the five minutes following the open auction. As expected during volatile periods, slippage vs. the opening price increased, but NYSE’s opening price performance changed less than electronic venues such as Nasdaq. NYSE-listed securities’ slippage increased by seven basis points while Nasdaq price changes following the open increased by 15 bps.
We also did a more granular comparison, looking at similarly-priced NYSE-listed and Nasdaq-listed opens for the January 2, 2018 through March 28, 2018 period. In each price and volume bucket, NYSE-listed issues achieved superior price discovery at the open.
- Kevin Tyrrell
With the quarterly expiration approaching on Friday, we reviewed trading from the December expiration to see shifts in price discovery behavior. One widely-known market structure feature is the increase in volume on inverted venues at the end of the trading day, when inverted venues can approach 20% of S&P 500 volume at certain points.
However, during key trading periods such as quarterly expirations, the shares available at primary exchanges1 far exceeds that of inverted venues. We looked at the December expiration’s average quoted size at the NBBO from primary exchanges and inverted venues, and find that primary exchanges’ quoted size at the end of the day increased 38% vs. a standard day2 while inverted venues increased just 18%.
This additional pre-trade displayed liquidity results in more intra-day trading activity flocking to the primary exchanges. Primary exchanges see an increase in volume every day heading in to the closing auction, and the trend is more dramatic on event days such as an expiration. The below chart shows shares trading in the S&P 500 by time period, exclusive of closing auction volume.
This is the “liquidity begets liquidity” argument on display, with greater displayed size enticing more liquidity-seeking flow. This increasing order volume allows limit order queues to process more quickly, a topic we’ll revisit in a future post.
- Kevin Tyrrell
1“Primary exchanges” refers to trading on the listed market (e.g., NYSE trading NYSE-listed and Nasdaq trading Nasdaq-listed).
2“Standard day” represents an average of December 1st - December 22nd trading activity, excluding December 15th
While most people know that the NYSE Closing Auction is the single largest liquidity event in the U.S. market, accounting for more than 6% of NYSE-listed volume on a regular basis, fewer are aware of the additional auction liquidity available at nearby price points.
To highlight this effect, we show the cumulative additional liquidity available on various days in December for NYSE-listed Russell 2000 and S&P 500 stocks, with colors indicating the distance of the liquidity from the final closing price. The median spread for NYSE-listed Russell 2000 stocks is roughly 29 bps, meaning all additional liquidity represented in green and blue on the Russell chart are on average within 2 spreads of the final closing price. The chart shows that traders who seek out this liquidity, by leveraging a floor broker’s experience and capabilities or by submitting orders at various price points, can find material additional volume.
- Kevin Tyrrell