Anonymous Trading in Equities

• Date: 01 April 2021
• DOI: http://doi.org/10.1111/jofi.12988
• Author: TOM GRIMSTVEDT MELING
• Published date: 01 April 2021

THE JOURNAL OF FINANCE •VOL. LXXVI, NO. 2 •APRIL 2021

Anonymous Trading in Equities

TOM GRIMSTVEDT MELING∗

ABSTRACT

In this paper, I explore a reform at the Oslo Stock Exchange to assess the causal

effect of posttrade trader anonymity on stock liquidity and trading volume. Using

a regression discontinuity approach, I f‌ind that anonymity leads to a reduction in

bid-ask spreads of 40% and an increase in trading volume of more than 50%. The

increase in trading volume is accounted for largely by increased trading activity by

institutional investors, while retail investors do not adjust their trading behavior

in response to anonymity. The results suggest that posttrade anonymity positively

affects standard measures of market quality.

STOCK EXCHANGES CONTINUALLY f‌ine-tune their markets to promote liquid-

ity. A much-used strategy over the last decade has been to alter the degree

to which traders are anonymous. In this paper, I empirically assess the effect

of posttrade anonymity—the concealment of trader identities after completed

trades—on stock liquidity and trading volume. An anonymity reform at the

Oslo Stock Exchange (OSE) allows for causal inference. I f‌ind that posttrade

anonymity increases stock liquidity and trading volume. The increase in trad-

ing volume is accounted for largely by increased trading activity by institu-

tional investors, while retail investors do not adjust their trading behavior in

response to anonymity.

How transparent should trading in equity markets be? Market regulators

have long advocated for more transparency. For example, in 2009, former

∗Tom Grimstvedt Meling is with the Department of Economics, University of Bergen. This pa-

per has benef‌ited from comments from Signe A. Abrahamsen, Oddmund Berg (discussant), Sjur

D. Flåm, Xavier Giroud, Raffaele Giuliana, Ingar Haaland, TerranceHendershott, Hans K. Hvide,

Hugo Hopenhayn, Jack Light, Katrine V. Løken, Rannveig Huus Meling, Kasper M. Nielsen, Carl

Nilsen, Terrance Odean, Christine Parlour(discussant), K onrad Raff (discussant), Ailsa Röell (dis-

cussant), Øvind Schøyen, Eirik Strømland, Jonas Tungodden, Jialin Yu,and Bernt Arne Ødegaard.

I am also grateful to Bernt Arne Ødegaard for sharing data, and to seminar participants at the

University of Bergen, UC Berkeley Haas, Norwegian School of Economics, EEA, and EWFC for

their valuable comments. I am especially indebted to Stefan Nagel (the editor), two anonymous

referees, and an anonymous associate editor for extensive comments that have greatly helped im-

prove the paper.This work was partially supported by the Research Council of Norway through its

Centres of Excellence Scheme, FAIR project number 262675, and through Finansmarkedsfondet’s

grant number 250215. I have read The Journal of Finance disclosure policy and I have no conf‌licts

of interest to disclose.

Correspondence: Tom Grimstvedt Meling,Department of Economics. University of Bergen, 5007

Bergen, Norway; email: Tom.Meling@uib.no.

DOI: 10.1111/jof‌i.12988

© 2020 the American Finance Association

707

708 The Journal of Finance®

Securities and Exchange Commission (SEC) chairwoman Mary Schapiro

stated that “Transparency is a cornerstone of the U.S. securities market (...) We

should never underestimate or take for granted the wide spectrum of benef‌its

that come from transparency” (SEC (2009)).1Market participants, in contrast,

caution that too much transparency frustrates their ability to eff‌iciently work

large orders because transparent markets expose their demands and may in-

crease trading costs—thus, harming liquidity.2At least partly in response to

trader demands, leading stock exchanges such as the Nasdaq, London Stock

Exchange, and Deutsche Börse have recently reduced transparency by increas-

ing posttrade anonymity (the Internet Appendix provides an overview of policy

changes).3

Theoretical predictions on the effects of posttrade anonymity on stock liquid-

ity and trading volume come from Kyle (1985)-type models in which informed

traders strategically trade against an uninformed market maker to exploit

their private information. In this setting, Huddart, Hughes, and Levine (2001),

Buffa (2013), and Yang and Zhu (2017) vary whether trader identities become

known after trades.4Yang and Zhu (2017) f‌ind that traders are more willing

to acquire fundamental information when trading is anonymous, resulting in

increased trading volume due to entry by informed traders. With respect to

stock liquidity, Huddart, Hughes, and Levine (2001) and Yang and Zhu (2017)

predict that increased trading on private information in anonymous markets

exacerbates the market maker’s adverse selection problem, which leads to a

reduction in stock liquidity. In contrast, in a setting with risk-averse informed

traders, Buffa (2013) f‌inds that anonymity can increase stock liquidity.

The purpose of this paper is to empirically assess the effect of posttrade

anonymity on stock liquidity and trading volume. To do so, I exploit the fact

that semiannually over the period 2008 to 2010, the 25 most-traded stocks

at the OSE were selected for posttrade anonymous trading, while all others

1Market regulators both in Europe and the United States have recently initiated comprehen-

sive market reforms to increase various aspects of market transparency. For example, the Euro-

pean Securities and Markets Authority (ESMA) recently capped the volumes that can be traded

in the least transparent venues (ESMA (2018)). Similarly, in 2016, the U.S. Financial Industry

Regulatory Authority (FINRA) announced plans to expand its Transparency Initiative by mandat-

ing the public disclosure of block-sized transactions in ATSs, a class of low-transparency trading

venues (FINRA (2016)).

2For example,Øvind Schanke, the head of equity trading at NBIM, the world’s largest sovereign

wealth fund, recently expressed concern about transparent trading to the Wall Street Journal:

“If we sent our orders into the market, we would have to wait days or weeks for our brokers to

execute the trade. Even then, there are risks of information leakage.” See Hope, Bradley, “Upstairs’

Trading Draws More Big Investors,” Wall Street Journal,December 8, 2013.

3The Internet Appendix may be found in the online version of this article.

4A separate theoretical literature studies the impact of pretrade transparency—the extent to

which traders can observe the prices, quantities, or trader identities corresponding to unexecuted

limit orders—on trader behavior and stock liquidity.One strand of the literature f‌inds that trans-

parency increases liquidity by mitigating information asymmetry (e.g., Pagano and Röell (1996)),

while another f‌inds that transparency can reduce liquidity by discouraging informed traders from

placing limit orders (e.g., Rindi (2008), Boulatov and George (2013)). Biais, Glosten, and Spatt

(2005) and Foucault, Pagano, and Röell (2013) survey this literature.

Anonymous Trading in Equities 709

were not. Comparing just-included and just-excluded stocks in a regression

discontinuity design allows for causal estimation. I f‌ind that anonymity signif-

icantly increases stock liquidity and trading volume. For example, the bid-ask

spread, a standard measure of illiquidity and trading costs, is 40% lower for

just-included anonymous stocks than just-excluded stocks, and trading volume

is more than 50% higher. These f‌indings suggest positive effects of posttrade

anonymity on standard measures of market quality.

Improvements in trading volume and stock liquidity may not be due to post-

trade anonymity, but rather to index inclusion effects. Anonymity at the OSE

was determined by membership in the OBX index, a composition of the most-

traded stocks at the OSE. Systematic differences between index and nonindex

stocks, caused (for example) by index benchmarking strategies, can confound

the estimated anonymity effects. To examine whether OBX index stocks are

systematically different from nonindex stocks, I compare index and nonin-

dex stocks in periods before anonymity was introduced. I f‌ind no differences

between marginal index and nonindex stocks in periods without anonymity.

Moreover, index funds typically track the broader Oslo benchmark index, in

which all sampled stocks are included, and not the OBX by itself. For example,

only two index funds track the OBX in the sample period, and their combined

net assets amount to about 5% of the net assets tracking the benchmark index.

Thus, it seems unlikely that index effects are driving the results.

To investigate the mechanisms behind the main empirical results, I f‌irst

note that existing theories posit that anonymity reforms affect market quality

by changing the behavior of informed traders. In the Yang and Zhu (2017)

model, for instance, anonymity increases overall trading volume by inducing

entry by informed traders. To test this mechanism, I use detailed transaction-

level data on the trading of all investors at the OSE. As empirical proxies for

“informed” and “uninformed” investors, I follow Linnainmaa and Saar (2012)

and use institutional and retail investors, respectively. Consistent with the

Yang and Zhu (2017) mechanism, I f‌ind that the observed increase in overall

trading volume under anonymity can be accounted for largely by an increase

in the trading activity of institutional investors, while retail investors do not

adjust their trading behavior in response to anonymity.

Another behavioral response posited by theory relates to how informed

traders execute their orders and underpins informed traders’ attraction to

posttrade anonymous markets. In the Yang and Zhu (2017), Huddart, Hughes,

and Levine (2001), and Buffa (2013) models, anonymity enables informed

traders to split their orders into series of smaller trades to mitigate market

impact and better exploit their private information. This induces positive auto-

correlation in the direction of informed trades, as buys follow buys and sells fol-

low sells. Under transparency, in contrast, such autocorrelated trading strate-

gies may be inferred and exploited by other market participants, and informed

traders must take costly (and thus entry-deterring) actions to conceal their

underlying demands by occasionally buying when the real goal is to sell, and

vice versa. Such “bluff‌ing” reduces the autocorrelation in informed trading.

Brokerage identif‌iers in the transaction-level data allow me to test this