By Nazed Mannan, Senior Product Manager, BMLL
When trading equity markets, one of the challenges is understanding the breadth and variety of participant types. Within any equity market there is a wide range of participants, from proprietary trading firms to asset managers, with holding periods from seconds to years. Understanding how these different firms trade, and what participants are trading a given stock, is essential for anyone trying to optimise execution and transaction quality.
An increasingly important market participant in the stock market is the retail investor. These non-professional investors have grown considerably in recent years since Covid, especially in the US equity market, fuelled by a combination of online brokerage platforms (for example, Robinhood) and the explosion of zero day options. Retail investors often show characteristics that are quite different to institutional investors, such as high momentum and contrarian habits, exemplified by the Gamestop explosion. Knowing whether a stock has a high level of retail activity is critical for investors, whether seeking alpha or improving execution.
US retail trading today
Unlike other properties of trading, such as the price or size of an order in the market, the exact nature of market participants is not disclosed. Instead, firms have often relied solely on very rough heuristics or delayed data to understand participant makeup. In this article, we show how leveraging messages from Level 3 historical data can considerably improve understanding of retail trading in the US markets.
Firstly, it is important to understand how retail trading works in the US equity market. Retail trading is typically done by payment for order flow (PFOF), in the following way:
- An investor places a trade through a broker
- The broker sends the order to a market maker
- The market maker executes the trade on behalf of the investor. This is typically internalised OTC, rather than on exchange
- The market maker pays the broker a fee for handling the trades
Execution has been historically dominated by the largest wholesale brokers, with most executions occurring off exchange. Looking at the two-week delayed FINRA OTC data, we can see that the makeup of these largest brokers typically includes Citadel Securities, Virtu Americas, and Susquehanna (G1 Execution), which respectively account for 26%, 19%, and 12% of OTC volumes. The chart below shows the total weekly OTC shares traded by broker.
Fig 1: Weekly U.S. OTC Volume by Broker
Whilst the FINRA data behind Fig 1 provides much sought after transparency, there are other considerations. Firstly, FINRA data is two weeks-delayed and aggregated at the weekly level, meaning that important changes can get missed. Secondly, it doesn’t show important details such as when retail trades, or if a stock is becoming more retail.
Instead of using delayed, aggregated FINRA data, a trade-by-trade approximation is often done using the SIP feed. By taking odd-lot trades inside NBBO (since retail normally gets price improvement), you can start to build a picture of retail trading (as seen in BMLL liquidity maps). For example, here we can see the growth in retail using SIP data in Gamestop this year:
Fig 2: U.S. Retail Volume by Ticker (Shares)
Fig 3: U.S. Retail Volume by Ticker (% Shares)
The race with retail: moving on exchange
While the majority of retail trading is OTC, a percentage does end up traded on exchange (between 2% and 10% based on SIP data). Exchanges compete for this flow, with different mechanisms and venues. For example, CBOE EDGX offers queue priority for trades explicitly tagged as retail, while other exchanges (such as NYSE) have retail price improvement programs, allowing retail execution inside NBBO via non-displayed trades. Importantly, we are seeing the race for retail increasing, for example with the recent launch of retail flags on Miami Pearl Equities which explicitly identify trades where participants are retail.
Taking a deeper look
Now we can turn to see what Level 3 historical data reveals. While each exchange is different, we can use a combination of order book flags, retail interest indicators and trade types (unlike the SIP, Level 3 data differentiates displayed and non-displayed trades) to build a full picture of retail on exchange.
We see that EDGX, with retail priority as well as rebates, is heavily skewed as a retail venue, with an average of 20% retail volume from January to early August in 2024.
Fig 4: U.S Retail Volume on Cboe EDGX
Let’s take a look at Gamestop (GME), which attracted significant interest between May to June 2024. On EDGX, we see that between April and May, the proportion of retail activity increased (from 23% to 32%) ahead of an increase in retail volumes (jumping from 1 to almost 25 million shares). Interestingly this suggests that institutional volumes were declining ahead of the retail jump in May, signaling contrarian drivers. Subsequently, the retail proportion in GameStop remained steady before a second jump in retail volumes in June. This is shown in Fig 5.
Fig 5: GameStop Retail Volume and Proportion on Cboe EDGX
Whilst EDGX provides explicit flags in the Level 3 data, for other venues with retail price improvement (RPI) programs (such as CBOE, Nasdaq and NYSE exchange), we can estimate retail volumes through a combination of RPI flags and non-displayed trades.
There is a clear correlation between retail volumes on these venues and EDGX. Importantly though, on-exchange retail volumes are not evenly distributed across venues. In contrast to the 20% retail proportion observed on EDGX, based upon an estimation approach, we observe a significantly smaller proportion of other exchange volumes being retail. This is seen in figures 6 and 7.
Fig 6: EDGX Retail Volume and Estimated RPI Volume by MIC
Fig 7: EDGX Retail Proportion and Estimated RPI Proportion by MIC
What is the true picture of on-exchange retail trading?
Understanding US retail is one of the holy grails for most investors, especially with continued growth in US retail trading. This is critical across all parts of the trading lifecycle, from creating alpha strategies to optimizing execution algorithms.
While either aggregated data from FINRA or anonymized data from the SIP can give some information, it doesn’t tell the whole story. By using retail flags and indicators only available in Level 3 data, you can build a much fuller picture of on-exchange retail trading. This is available trade by trade and order by order, giving a clearer and deeper understanding into how and where retail is trading.
