Voluntary pre-trade anonymity and market liquidity

ABSTRACT

This paper analyses the effects on liquidity of voluntary pre-trade anonymity in the trading process. We confirm previous studies showing that market liquidity improves immediately after anonymous trading. Using the daily percentage of effective volume traded anonymously, we show that the anonymity–liquidity relationship presents a non-linear U-shape. We focus on the voluntary concealment of broker identification introduced by the Spanish Stock Exchange in October 2015. We conclude that, in our sample, anonymity increases stock liquidity but at a decreasing rate; when a considerable part of the effective volume is traded anonymously, additional percentages of anonymous trading deteriorates stock liquidity.

KEYWORDS:

• Voluntary trader anonymity
• market liquidity
• pre-trade transparency
• information disclosure
• market quality

JEL CLASSIFICATION:

• G10

Acknowledgments

We are grateful for comments and useful discussions from B. Alonso, J. Hernani, J. Penalva and J. Yzaguirre. The opinions expressed in this paper are those of the authors exclusively.

Disclosure statement

No potential conflict of interest was reported by the authors.

Notes

1. See Meling (2018) for a summary of the recent stock market´s modifications regarding broker ID.

2. Hence, for instance, on 23 April 2001 for Euronext Paris, and 30 June 2003 for the Tokyo Stock Exchange the broker identifiers from all the limit order were removed; on the contrary, the Spanish Stock Exchange allowed the traders to become anonymous or to continue showing its identifiers. BME however is not the only exchange in which disclosing trader anonymity is voluntary. The London Stock Exchange in 2001, and NASDAQ and the Toronto Stock Exchange in 2002 also offered a similar option to traders.

3. Effects on alternative measures of market quality as depth, volume, or probability of information-based trading have been also studied in the literature. See, for instance, Foucault et al. (2007), Frino et al. (2008), and Grammig et al. (2001).

4. Næs and Skjeltorp (2006) or Martínez et al. (2005) among others, using information of the consolidated LOB build some style measure that shows the properties of liquidity measured beyond the quoted spreads.

5. See, for instance, Madhavan et al. (2005), McInish and Wood (1992), and Stoll (2000).

6. We do not consider intraday estimations because of the lack of intraday data in the dataset.

7. We thank an anonymous referee for pointing out us this line of reasoning.

8. It is noteworthy that hidden orders are not permitted in BME, so effective spread is equal to quoted spread.

9. The values of the Anon variable for the first quartil goes from 0 to 0.72, and to 0.84 and to 0.92 for the second and third quartils, respectively.

10. These results are not reported, but they are available to readers upon request.

11. We use the xtabond Stata command instead of xtabond2. xtabond2 is designed for dynamic ‘small-T, large-N’ panels.

12. We also use a second approach using GMM estimations for the full sample. As Roodman (2009) highlights, an important obstacle emerges when using long panels: the proliferation of instruments. This is because the number of instruments to be generated is directly related to the length of the panel (number of periods). The number of instruments increases, as each regressor is instrumentalised by all their differences and levels (with GMM). This proliferation of instruments initially was seen as favourable, since it increased the efficiency of the estimator (Arellano and Bond 1991), however, it could cause overidentification. Therefore, as the panel grows in periods, the probability of overidentification increases. Even though the Sargan test (including 30 lags), which verifies the validity of the instruments used in the analysis, shows that the estimations are valid joint AR(2) test of Arellano-Bond, and therefore overidentification does not exit. Results of these robustness tests are not reported in the text in order to conserve space, but they are available to readers upon request.

Additional information

Funding

This work was supported by the Comunidad de Madrid [S2015/HUM-3353 (EARLYFIN-CM)]; Ministerio de Ciencia y Tecnología [ECO2015-69205-P]; Ministerio de Economía y Competitividad [ECO2014-51914-P]; Fondos FEDER [UNC315-EE-3636]; Basque Government [IT1336-19].