International Review of Equity Market Structure Regulation: Phase II (2020)

As to off-exchange trading volumes, we find that only in the E.U. and the U.S. does off-exchange trading constitute a sizable share of overall equity trading volume. As demonstrated by Figure 8, off-exchange trading volume represents 46 percent of overall trading volume in the E.U. as of September 2020,14 and 35 percent of overall trading volume in the U.S. (approximately $28.4 trillion in annual trading volume in 2019). In Japan, off-exchange trading volumes represent only 15% of overall trading volume and in Hong Kong off-exchange trading volumes represent only 1% of overall trading volumes. Off-exchange trading is prohibited in Mainland China.

Multilateral trading facilities (“MTFs”) in the E.U. and alternative trading systems (“ATSs”) in the U.S execute trades off-exchange. Like exchanges, MTFs and ATSs match buyers and sellers, but they are not registered as exchanges and they are not the primary listing venue for public equities. Regulated market makers also execute off-exchange trades. These regulated market makers are referred to as systematic internalizers (“SIs”) in the E.U. and broker-dealer internalizers in the U.S. They do not match buyers and sellers but instead act as principals and execute customer orders against their own inventory of stocks on a systematic basis. As in the E.U. and U.S., off-exchange trading in Hong Kong and Japan takes place primarily on “automated trading systems” and “proprietary trading systems,” respectively, and through regulated market makers.

In the U.S., broker-dealer internalization accounted for approximately 70 percent of off-exchange trading by share volume in Q4 2019, with ATSs accounting for the remaining share. In the E.U., according to Rosenblatt Securities, SI trading accounted for just 14.3 percent of trading by value as of September 2020, while 18.7 percent of trading was conducted via MTFs and 21 percent of trading was over-the-counter, meaning that trades were non-systematic, ad-hoc, irregular and infrequent. It is important to note that there are doubts as to the reliability of E.U. statistical data as it pertains to SI trading volumes and off exchange trading more generally. SI trading volume data particularly, as acknowledged by ESMA themselves, may have been artificially inflated by the missreporting of some financial market participants due to a lack of clarity in the current regulatory framework requirements. This has resulted in a somewhat distorted picture and difficulty in ascertaining the genuine extent of SI trading volumes in the E.U markets which clearly also suffer from the lack of a consolidated tape. However, the real amount of off-exchange trading volume in the E.U. is likely to fluctuate substantially over the coming years as these issues get fixed, and a consolidated tape is implemented.

One additional note is that off-exchange trading often takes place in the “dark,” meaning that there is no public display of order interest or pre-trade transparency. As noted in PIFS’ International Review of Equity Market Structure Regulation, exchanges and off-exchange trading venues can execute trades in the dark, however all trades in each of the five markets are subject to post-trade transparency. In the United States, ATSs and broker-dealer internalizers typically execute trades without pre-trade transparency. However, in the E.U., 64% of trading on MTFs is “lit,” meaning that the trading venue provides pre-trade transparency.

As demonstrated by Figure 9, trading volume in the U.S., Japan and Hong Kong is predominantly related to large institutional orders, rather than trading by retail investors. According to the TABB Group, institutional trading volume in the U.S. accounted for approximately 85 percent of 2018 trading volume. Japan Exchange Group data shows that institutional investors accounted for approximately 83 percent of 2018 trading by value. And a 2019 Schroders analysis reports that 65 percent of Hong Kong trading volume is attributable to institutional investors. However, in Mainland China, a 2019 FTSE Russell report finds that over 80 percent of trading volume is retail. PIFS staff has thus far been unable to identify estimates of the relative share of trading in the E.U. attributable to institutional or retail investors. 

Figure 10 shows the average on-exchange trade size by dollar volume in each of the five markets. Average trade sizes are significantly higher in the U.S., Japan and Hong Kong––predominantly institutional markets––as compared to Mainland China’s retail-heavy market. We note that the average trade size by dollar volume in the E.U. is consistent with average trade sizes in the U.S., Japan and Hong Kong, which suggests that trading volume in E.U. markets may also be predominantly related to larger institutional orders. However, an evaluation of average trade sizes must also note other factors that are not necessarily reflected in the data. First, broker-dealers break up large institutional orders into many smaller orders (often referred to as “child” orders) to achieve the best execution price for their clients. As a result, average trade sizes are smaller than the original “parent” order from the investor. In addition, in the U.S. approximately 90% of retail orders, which are generally smaller than institutional orders, are executed through broker-dealer internalization and, therefore, are not included in the average on-exchange trade size data.

For retail investors, transaction costs include brokerage commissions and the bid-ask spread. Brokerage commissions entail the direct fees charged by brokers for executing an investor’s order. These fees can vary widely both within markets and across markets, particularly with U.S. brokerage commissions approaching zero for many retail orders. The cost associated with the bid-ask spread is slightly less straight-forward and best described with an example. Suppose that Security X is trading on an exchange with a best bid price of $9.99 and a best offer (or ask) price of $10.01. The midpoint of the best bid and offer is $10.00, which is the implied “fair value” of Security X. Next, suppose that Retail Investor A places a market order to buy 100 shares of Security X. This market order will execute on the exchange at $10.01, the current best offer, because given the small size of retail orders there is typically sufficient available liquidity at the best offer price. The cost associated with the bid-ask spread is the difference between the mid-point of $10.00 and the execution price of $10.01. However, as noted in Part I, in the U.S.,more than 90% of retail orders are executed off-exchange through broker-dealer internalization. These retail orders typically receive “price improvement,” meaning that they are executed at a price better than $10.01 in this example (i.e. closer to the “fair value” of $10.00).

For institutional investors, transaction costs associated with price impact are higher as there is typically insufficient liquidity at the best bid or offer to execute an order. The price impact for institutional orders is best described with an example. Suppose that Security X is still trading with a best bid of $9.99 and a best offer of $10.01. Further suppose that only 2,500 shares are offered at $10.01, while another 5,000 shares are offered at $10.02 and 2,500 shares offered at $10.03. For illustrative purposes, suppose that Institutional Investor B places a market order to buy 10,000 shares of Security X. Given the shares available, the order will be executed at $10.01 for 2,500 shares, $10.02 for 5,000 shares and $10.03 for 2,500 shares, resulting in an average execution price of $10.02 for the full order. The $10.02 execution price is 1 cent or 10 basis points (“bps”) higher than the $10.01 best offer that prevailed at the time of the order. This 10-bp difference is the additional price impact (often referred to as slippage) that occurs as a result of the relatively large buy order (i.e. the 10,000 share order is large relative to the size of offers in the market). Price impact costs are included in the Virtu and AbelNoser datasets. 

Virtu estimates brokerage commissions and price impact cost by collecting trading data from over 180 asset managers on over $13 trillion in trades per year. Virtu trading costs and order size are a weighted average based on the value of shares traded. We note that for the Virtu data we review six markets rather than five, as Virtu separates trading cost data for the U.K. from the E.U.

Figure 11 sets forth overall institutional trading costs in each of the six markets in 2019 measured in basis points. One basis point is equivalent to 1/100th of 1% or 0.01%. The U.S. has the lowest total trading costs at 35.7 bps, followed by E.U. (35.9 bps), the U.K. (46.4 bps), Hong Kong (47.5 bps), Mainland China (52.9 bps), and Japan (52.9 bps). Figure 11 also shows that price impact contributes between 84 and 91 percent of total trading costs.

However, the overall trading cost by value (measured in USD) for institutional investors depends on the size of the institutional order on which the basis point cost is applied. The Virtu data set considers parent orders rather than their constitutent “child orders.” As described in Part I, broker-dealers divide parent orders into many smaller orders to minimize transaction costs and achieve best execution. Table 1 below sets forth the average order size in each market in 2019 (based on the Virtu data set) and the average dollar trade cost for each market. The average dollar trade cost is derived by simply multiplying the average order size by the average basis point cost. For example, the average order size in the U.S. is $327,463. A 35.7 bp (or 0.357%) transaction cost would represent a cost of $1,170 to an institutional investor for executing an average-sized order. And, in Mainland China, the average order size is $329,478, so a 52.9 bp (or 0.529%) transaction cost would represent a cost of $1,744 to an institutional investor.

a. Commission Costs

There was a substantial decrease in commission costs across all six major equity markets from 2015 to 2019. Figure 12 illustrates that average commission costs dropped between 27 and 42 percent in each market during this time.

b. Price Impact Costs

Figure 13 demonstrates that from 2015 to 2019 price impact costs increased substantially in Mainland China (growing 19 percent) and declined substantially in Japan, Hong Kong, the E.U., and the U.S (dropping between 12 and 14 percent in each market). Price impact costs declined a more modest 6 percent in the UK. The increase in price impact cost in Mainland China is likely due to the substantial reduction in trading volume in Mainland China from 2015 to 2019 demonstrated by Figure 2. In general, price impact costs are higher when there is less liquidity. As noted in Part I, the decline in trading volume in Mainland China from 2015 to 2019 represented a reversion to historical trends, as trading volumes in Mainland China experienced a signicant increase leading up to 2015.

c. Overall Trading Costs

Figure 14 illustrates that from 2015 to 2019, overall institutional investor trading costs (commission costs + price impact costs) have fallen across major public equity markets other than Mainland China where they have slightly risen by three percent. The increase in trading costs in Mainland China is primarily driven by an increase in price impact costs (+16.8 basis points), which exceeded the decrease in commission costs (-5.6 basis points). As noted above, the increase in price impact costs in Mainland China is related to a decrease in trading volumes in Mainland China from 2015 to 2019 that itself represents a reversion to historical trends. Critically, this means that the recent spike in transaction costs is likely not due to a reduction in the efficiency of equity market structure.

d. Order Size

Figure 15 and Table 2 illustrate that from 2015 to 2019, the average institutional order size in the Virtu data set increased moderately in the U.S. (10.5%) and the U.K (1.9%), and increased significantly in the E.U. (ex-U.K.) (23.4%). Over the same time period, average order sizes decreased moderately in Japan (-9.8%) and Hong Kong (-17.0%), while dropping significantly in Mainland China (-70.1%).

In general, changes in the average order size in each market over time should be considered alongside changes in average trading costs in each market. Holding all else equal, as the size of an order increases, so does the cost to execute that order, and vice-versa.45 However, in the U.S., E.U. (excluding UK), and United Kingdom, Virtu data shows that overall trading costs are generally falling despite increases in order sizes. Therefore, other market factors (such as higher liquidity, improved competition or execution technology and strategies by broker-dealers) likely account for lower trading costs in those markets. As to Japan, trading costs have fallen by 14% from 2015 to 2019 whereas average order size has only decreased by 9.8% over the same period of time, which also suggests that market factors other than changes in average order size are contributing to lower transaction costs. In Hong Kong both trading costs and order size have fallen by 17% from 2015 to 2019. It is therefore possible that the fall in trading costs in Hong Kong is partially explained by the similar decrease in average order size.

Mainland China, however, has experienced a contrasting relationship between order size and trading costs over the past five years. Even though the average order size has more than halved since 2015, trading costs have increased 3%, suggesting that factors other than order size are causing trading cost increases in Mainland China. Indeed, the increase in transaction costs is likely driven by the significant fall (52%) in trading volumes in Mainland China from 2015 to 2019, as demonstrated in Figure 2 in Part 1. Less market-wide liquidity is typically associated with higher trading costs.

AbelNoser calculates trading costs by collecting data from over 200 asset managers and analyzing approximately $7.5 trillion in trading volume annually. AbelNoser trading costs and order sizes represent the median by value of shares traded. 47 Also, AbelNoser aggregates transaction costs for the U.K. and the E.U., so we review five markets: E.U. (incl. U.K.), Hong Kong, Japan, Mainland China and the U.S. We observe similarities between the Virtu and AbelNoser data sets at the end of our analysis.

Figure 16 sets forth overall institutional trading costs in each of the five markets in 2019 measured in basis points. The U.S. has the lowest total costs for institutional investors at 20.5 bps, followed by the E.U. and Mainland China (each 20.6 bps), Japan (22.7 bps), and Hong Kong (29.2 bps). Figure 16 also shows that price impact contributes between 62 and 79 percent of total trading costs.

Table 3 sets forth the average parent order size in each market in 2019 and the average dollar trade cost implied by multiplying the average order size by the median trade cost in basis points.

a. Commission Costs

Figure 17 shows that commission costs fell across all five major equity markets between 2015 and 2019. Commission costs fell the most in Mainland China (-47%), followed by the U.S. (-29%), Hong Kong (-23%), the E.U. (-23%), and Japan (-11%). For the period as a whole, commissions in the U.S. remained the lowest, and those in Hong Kong remained the highest.

b. Price Impact Costs

Figure 18 shows that, between 2015 and 2019, price impact costs fell in the E.U. (-16%), Japan (-15%), and Mainland China (-7%), but they rose in Hong Kong (+39%) and the United States (+9%). The rise in Hong Kong is entirely attributable to a sharp increase in 2019.

c. Overall Trading Costs

Figure 19 shows that, between 2015 and 2019, overall trading costs fell in all markets to varying degrees, except Hong Kong. Overall trading costs fell the most in Mainland China (-23%) and the E.U. (-18%), followed by Japan (-14%) and the United States (-2%). Overall trading costs rose 10% in Hong Kong. The U.S. had the lowest overall trading costs during the entire period.

d. Order Size

Figure 20 and Table 4 on the next page illustrate that from 2015 to 2019, the average institutional order size in the AbelNoser data set decreased across all markets with significant declines in Mainland China (-56.7%), Japan (-42.2%), the E.U. (including the UK) (-27.9%), and the U.S. (-20.0%). Hong Kong dropped more modestly with only a 14.0% decline over the same time period. We note that overall trading costs dropped alongside declines in order size in four of the five markets from 2015 to 2019, as expected based on the relationship between order size and trading costs, except for Hong Kong, where trading costs rose and the decline in order size was slight.

Our review illustrates three key patterns across both data sets. First, price impact costs constitute the majority of overall trading costs for institutional investors across all five markets. Second, brokerage commission costs have significantly declined across all five markets and overall trading costs have also generally declined in recent years. Third, the United States has the lowest overall trading costs of the five major markets.

One key difference is also apparent—trading costs (measured in bps or dollars) are substantially lower in the AbelNoser data, as compared to the Virtu data. However, this difference can be explained by the fact that the median order size in the AbelNoser data is significantly smaller than the average order size in the Virtu data. As noted earlier, smaller order sizes correlate with lower trading costs, and vice-versa.