Financial Transaction Taxes, Market Composition, and Liquidity

• Date: 01 December 2017
• Published date: 01 December 2017
• Author: JEAN‐EDOUARD COLLIARD,PETER HOFFMANN
• DOI: http://doi.org/10.1111/jofi.12510

THE JOURNAL OF FINANCE •VOL. LXXII, NO. 6 •DECEMBER 2017

Financial Transaction Taxes, Market

Composition, and Liquidity

JEAN-EDOUARD COLLIARD and PETER HOFFMANN∗

ABSTRACT

We use the introduction of a financial transaction tax (FTT) in France in 2012 to test

competing theories on its impact. Wefind no support for the idea that an FTT improves

market quality by affecting the composition of trading volume. Instead, our results

are in line with the hypothesis that a lower trading volume reduces liquidity and

in turn market quality. Consistent with theories of asset pricing under transaction

costs, we document a shift in security holdings from short-term to long-term investors.

Finally, we find that moderate aggregate effects on market quality can mask large

adjustments made by individual agents.

THE GLOBAL FINANCIAL CRISIS has renewed interest in financial transaction taxes

(FTTs), a development that has been fueled by the combination of strapped pub-

lic finances and public discontent with the financial industry.1Since suggested

by Keynes (1936), the idea of taxing trading activity has been at the center of

∗Jean-Edouard Colliard is at HEC Paris. Peter Hoffmann is at the European Central Bank.

We are grateful to EUROFIDAI for access to the BEDOFIH database (for more information, see

http://www.eurofidai.org). We would like to thank Bruno Biais (the Editor), three anonymous ref-

erees, Fany Declerck, Hans Degryse, Patrice Fontaine, Thierry Foucault, Luc Goupil, Laurent

Grillet-Aubert, Philippe Guillot, Philipp Hartmann, Frank de Jong, Julien Leprun, Laurence Le-

scourret, Simone Manganelli, Sophie Moinas, Alper Odabasioglu, Richard Payne, R´

emy Praz, Paolo

Sodini, Elvira Sojli, David Sraer, David Thesmar,as well as seminar and conference participants

at the French Treasury, Autorit´

e des March´

es Financiers (AMF), ECB, VU University Amster-

dam, University of Vienna, Toulouse School of Economics, the Banque de France Workshop on

Algorithmic and High Frequency Trading, the 6th Erasmus Liquidity Conference, the Arne Ryde

Workshop, the FBF-IDEI workshop “Investment Banking and Financial Markets Value Chain,”

the 2014 conference “The Future of Securities Markets Regulation” at the University of Geneva,

the 2014 Frontiers of Finance Conference, the 2014 CREDIT Conference, the 10th Annual Cen-

tral Bank Workshop on Market Microstructure, the 2014 Paris December International Finance

Meeting, the 2014 conference “Market Microstructure: Confronting Many Viewpoints,” the 2015

Bocconi-CONSOB conference “The Development of Securities Markets Trends, Risks and Policies,”

the 2016 AFA Meetings, and the 2016 Jan Mossin Memorial Symposium for useful comments and

suggestions. The views expressed in this paper are those of the authors and do not necessarily re-

flect those of the European Central Bank or the Eurosystem. The authors do not have any potential

conflicts of interest, as identified in the Journal of Finance’s Disclosure Policy

1In October 2012, 11 EU countries committed to the introduction of a harmonized tax on

financial transactions, initially planned to be implemented by 2016. However, negotiations stalled

repeatedly and were still not finalized at the time of writing. See “EU financial transaction tax

progress stalls,” Financial Times (online), June 5, 2016. The debate on FTTs has also recently

resurfaced in the United States, in the context of the Democratic Party presidential primaries.

DOI: 10.1111/jofi.12510

2685

2686 The Journal of Finance R

a more fundamental debate on whether there is too much trading in financial

markets.

Proponents of FTTs such as Stiglitz (1989) argue that markets are populated

by too many noise traders whose actions are not based on information and

thus generate “excess volatility.” According to this view, FTTs improve market

quality by reducing the proportion of such “futile” trading. Opponents to FTTs,

however, question the relevance of this composition effect, arguing that it is

dominated by a liquidity effect whereby reduced participation, even by unin-

formed agents, has a negative effect on liquidity. This prevents the correction

of mispricing by arbitrageurs, thereby increasing volatility and reducing price

efficiency.

The theoretical literature, surveyed in Section I, rationalizes these opposing

arguments. In line with this tension, empirical work delivers mixed conclu-

sions.2This is largely a consequence of the fact that existing studies focus on

the aggregate impact of FTTs, mainly due to a lack of granular data. Accord-

ingly, while they can shed light on the sum of the composition and liquidity

effects, they are not informative about the individual roles of the two effects.

This paper aims to bridge this gap between existing theoretical and empir-

ical work by examining the introduction of a 20 bps tax on the purchase of

French equities on August 1, 2012. We assess the FTT’s causal impact using

a difference-in-differences framework. Importantly, we estimate treatment ef-

fects for different types of market participants, which allows us to assess the

relative roles of the composition and liquidity effects.

Our analysis begins by examining the French FTT’s aggregate impact. We

document that trading volume decreased on average by around 10%, accom-

panied by a moderate decline in market quality. However, the overall impact

masks some significant heterogeneity across stocks. We find that stocks under

Euronext’s “Supplemental Liquidity Provider” (SLP) program, a rebate scheme

aimed at incentivizing high-frequency traders (HFTs) to provide liquidity in the

most active stocks, were only marginally affected by the FTT. In contrast, the

remaining stocks displayed a strong reduction in trading volume (−20%) and a

significant decline in market quality. This discrepancy suggests that an FTT’s

impact depends on the affected stock’s level of liquidity.

We next analyze the FTT’s impact on different trader types using a data

set that groups market participants into three distinct categories according to

their speed. We find that, despite being effectively tax-exempt, HFTs were the

most affected trader type with a decline in trading activity of 35%. This indirect

impact constitutes strong evidence in support of the liquidity effect. We further

find that HFTs were affected due to i) an increase in bid-ask spreads, which

Transaction taxes and administrative charges on trading activity (e.g., the SEC’s Section 31 fee)

are internationally widespread. See Matheson (2011) for an overview.

2Numerous empirical studies estimate the impact of FTTs and transaction costs in general.

A nonexhaustive list includes Roll (1989), Umlauf (1993), Jones and Seguin (1997), Baltagi, Li,

and Li (2006), Hau (2006), Liu and Zhu (2009), and Pomeranets and Weaver (2012). For a more

complete overview, see Matheson (2011).

Financial Transaction Taxes, Market Composition, and Liquidity 2687

hurt the profitability of market orders, and ii) an overall decline in trading

activity, which decreased opportunities for arbitrage or intermediation.

Finally, we use theories of asset pricing under transaction costs to explore

two mechanisms through which FTTs can affect trading volume: first, investors

can react to higher transaction costs by keeping the same portfolio allocation

but adjusting their turnover (intensive margin), and second, a tax can encour-

age some investors to sell the affected stocks to other market participants with

a longer holding horizon (extensive margin). To separate these two effects, we

examine the portfolio holdings and turnover of institutional investors. We find

support for both channels. Consistent with the model of Amihud and Mendel-

son (1986), we find that investors with high portfolio turnover reduced their

holdings of French stocks, whereas investors with low turnover increased them.

Moreover,in line with Constantinides (1986) and Vayanos (1998), investors also

significantly decreased their turnover. We show that this effect was stronger

for investors with high trading activity.

A striking result of our analysis is that the rather muted impact of the

tax on the market quality of the most liquid French stocks masks significant

changes in the portfolio allocations and trading behavior of the affected market

participants. This discrepancy is consistent with the model of Constantinides

(1986). Intuitively, agents adjust to the tax in order to minimize its impact.

As a result, the effect of the tax on aggregate variables is of second-order

importance relative to its effect on individual strategies. Understanding the

full impact of an FTT therefore requires analyzing the reactions of individual

market participants and not only changes in aggregate measures of market

quality. To our knowledge, the French FTT is the first policy experiment for

which data availability permits steps in this direction.

Besides being important for understanding the impact of transaction costs

in securities markets in general, our evidence also informs the policy debate

on transaction taxes. FTTs are frequently motivated by a mix of fiscal (raising

revenue) and Pigovian (correcting externalities) motives. We fail to find evi-

dence in favor of the Pigovian rationale, but the official revenue figures show

that the FTT also fell short of expectations along this dimension.3While the

tax had an overall negative impact on market quality, this effect was rather

muted for the most liquid stocks. Moreover, the tax had an important effect on

the clientele of French stocks that requires looking beyond aggregate measures

of market quality. In particular, it reallocated the ownership of French stocks

toward more long-term institutional investors, an effect that can potentially

improve the corporate governance of the affected companies.4The impact on

market quality for less liquid stocks, however, was considerably more negative.

This paper coincides with a number of other empirical studies of the

French FTT that, despite using a variety of different control groups, reach

3The French FTT yielded 198 million EUR in tax revenues for August to December 2012

(http://www.assemblee-nationale.fr/14/rap-info/i1328.asp), which is considerably less than initial

projections of 1.1 billion EUR per year (see http://www.senat.fr/rap/r12-259/r12-2591.pdf).

4See Stein (1989) for a theoretical argument, and Bushee (2001) and Derrien, Kecskes, and

Thesmar (2013) for recent evidence.