Identify Retail Investors
Retail investors and their trading behaviour attract many research interests. One strand of literature uses proprietary datasets to identify retail investors. The other uses algorithms. A recent JF paper Boehmer et al. (2021) proposes a simple one based only on the trade price, which also signs the trade direction effectively. Even more interestingly, I just read a follow-up work forthcoming on JF by Barber et al. (forthcoming) (SSRN). The authors placed 85,000 retail trades themselves to validate the Boehmer et al. (2021) algorithm.
Boehmer et al. (2021) algorithm
A quick summary of the background info and the algorithm:
- In the U.S., retail investors’ marketable orders are typically not executed on exchanges but filled from a broker’s own inventory.
- These orders are publicly reported to a Financial Industry Regulatory Authority (FINRA) TRF, which is included in the TAQ consolidated tape with an exchange code
D. - These orders often receive minor price improvements (like 0.01, 0.1, or 0.2 cent) relative to the national best bid or offer (NBBO).
- SEC Rule 606 requires brokers to provide summary statistics on their order-routing practices for orders not directed to exchanges, which are mostly retail and often receive price improvement.
- Institutional orders, in contrast, are typically executed on exchanges or dark pools and priced in round pennies, except for midpoint trades.
Therefore, the algorithm to identify retail transactions is simply:
- Retail seller transactions are reported on a TRF at prices slightly above a round penny.
- Retail buyer transactions are at prices slightly below a round penny.
- Transactions at round pennies or near half-pennies are not classified as retail.
For example, for a trade with exchange code D on TAQ,
- if trade price is 10.21 (round penny), it is non-retail.
- if trade price is 10.211 (slightly above a round penny), it is a retail sell.
- if trade price is 10.209 (slightly below a round penny), it is a retail buy.
- if trade price is 10.214, 10.215, or 10.216 (near half-penny), it is non-retail.
Barber et al. (2023) experiment
I find this very interesting. The authors actually conduct an experiment by actually placing 85,000 retail trades themselves to validate the Boehmer et al. (2021) algorithm. They find that the algorithm identifies 35% of their trades as retail, incorrectly signs 28% of identified trades.
There is no need to summarize the paper here. SSRN link is here.