Mutations Financières

Download as pdf or txt
Download as pdf or txt
You are on page 1of 37

G.

Capelle-Blancard 30/03/2017

Mutations financières

La taxation des transactions


financières
Gunther Capelle-Blancard

A controversial topic
“And then there’s the proposal for a Financial Transactions Tax...
Even to be considering this at a time when we are struggling to get
our economies growing is quite simply madness”.
David Cameron, British Prime Minister

“And then there’s the idea of taxing financial transactions, which


have exploded in recent decades.The economic value of all this
trading is dubious at best. In fact, there’s considerable evidence
suggesting that too much trading is going on… it suggests that to
the extent that taxing financial transactions reduces the volume of
wheeling and dealing, that would be a good thing.”
Paul Krugman, economist and Nobel Laureate

1
G. Capelle-Blancard 30/03/2017

A controversial topic
List of controversial issues among Wikipedia editors

In the top 250 for the category “Politics/Economics”


Together with Capital punishment, Holocaust, Gun politics,
Same-sex marriage, etc.

In the top 10 if we restrict the list to economic issues


Together with Capitalism or Communism.

Public opinion

European Parliament Eurobarometer (EB77.2)


Coverage: EU27 (26 593 EU citizens)
Population: Europeans aged 15 or over
Methodology: Face-to-face (CAPI)
Fieldwork: 10-25 March 2012 carried out by TNS
Opinion

66% of Europeans are in favour of the principle of


a financial transaction tax
73% in the euro zone, 78% in Germany, 79% in
France, 66% in Spain… 53% in the UK
4

2
G. Capelle-Blancard 30/03/2017

Public opinion
What is the main reason behind your support for a
tax on financial transactions?
Combat excessive speculation and so help to
prevent future crisis: 42%
Make financial players contribute to the costs
of the crisis: 36%
Reduce public deficits: 12%
Finance innovative politics: 7%
Others/Don’t know: 3%

Public opinion
What is the main reason behind your objection to a
tax on financial transactions in the EU?
Weaken the competitiveness of the European financial
market: 23%
Not be feasible, such a tax on financial transactions can only
be introduced at a global level: 23%
Cause and outflow of capital from the EU: 23%
Only make European financial players contribute to the
costs of the crisis: 21%
Others/Don’t know: 10%

3
G. Capelle-Blancard 30/03/2017

Google trend

Outline

I. Stock prices are not open-ended. Stock trading seems


to be
II. Financial transaction taxes around the world: Past,
Present and Future
III. A global theoretical framework
IV. Empirical evidence: Financial transaction taxes have no
impact
V. Fiscal revenue

4
G. Capelle-Blancard 30/03/2017

Stock prices are not open-ended.


Stock trading seems to be

PART I

The growth of stock trading in the US


a. NYSE trading volume (1888-2013, monthly)
80000
Monthly num ber of shares traded (millions)
20000 40000 0 60000

1880 1900 1920 1940 1960 1980 2000 2020

10 computation
Author’s

5
G. Capelle-Blancard 30/03/2017

952 quotes per minute were disseminated


for all stocks on all exchanges in the U.S.

14,319 134,108

Source : Angel, Harris et Chester Spatt, 2013. Equity Trading in the 21st Century

12
Angel, Harris, Spatt (2013). Equity Trading in the 21st Century: An Update

6
G. Capelle-Blancard 30/03/2017

Stylized facts
Total value of stock traded in the world is about…
$100 trillion in 2013
$5 trillion in 1992

Stock traded/GDP is equal to…


130% in 2013
20% in 1992

This growth is even more pronounced for high-


income countries
US stock trading = US GDP x 3

13

Table 1. Total value of stock traded

Data on traded value is from the World Bank for 1992 and from Thomson Reuters (Monthly Market Share Reports) for
2013. Notes: a) US + Europe + Asia Pacific; b) Shanghai + Shenzen.

2013 1992
($ trillion) (% GDP) ($ trillion) (% GDP)
US 49,589 295% 2,081 33%
Europe 24,326 112% 1,223 14%
France 2,712 99% 122 9%
Germany 3,947 109% 446 22%
UK 4,810 191% 383 35%
Asia Pacific 22,570 103% 1,031 18%
China 7,673 b) 83% 17 4%
Japan 8,480 173% 635 16%
World 96,485 a) 128% 4,783 20%

14 computation
Author’s

7
G. Capelle-Blancard 30/03/2017

Too much finance?


A large (and sound) financial system is essential for
development and growth

But…

Excessive growth of credit? (Arcand et al., 2012)


Misallocation of talents? (Philippon, 2013)

15

How much is too much?


What is the optimal volume of stock trading?

Theoretically, several answers


from zero (rational expectation models without noise)
… to infinite (assets pricing models with dynamic hedging
in the absence of trading costs)!

Empirically, very few studies


Evidence suggest that investors trade too much (Odean,
1999), suggesting that trading activity is too high.

16

8
G. Capelle-Blancard 30/03/2017

Total value of stock traded worldwide


100000
(USD billion)
50000 0

1988 1990 1992 1994 1996 1998 2000 2002 2004 2006 2008 2010 2012

World Bank (WDI)


Thomson Reuters (Monthly Market Share Reports)
17
Author’s computation

Stocks Traded over GDP (in %) for selected


countries (1)
400

Data are from the World Bank (WDI,


historical stock exchanges only) from 1988
to 2009 and
300

from Thomson-Reuters (Monthly Market


Share Reports, all trading venues) since
(in %)

2010
200 100
0

1988 1990 1992 1994 1996 1998 2000 2002 2004 2006 2008 2010 2012

France Germany
Italy Spain
Sweden United Kingdom
18
Author’s computation

9
G. Capelle-Blancard 30/03/2017

Stocks Traded over GDP (in %) for selected


countries (2)
500

Data are from the World Bank (WDI,


historical stock exchanges only) from 1988
400

to 2009 and
from Thomson-Reuters (Monthly Market
Share Reports, all trading venues) since
300

2010
(in %)
200 100
0

1988 1990 1992 1994 1996 1998 2000 2002 2004 2006 2008 2010 2012

United States Canada


19
Author’s computation

Stocks Traded over GDP (in %) for selected


countries (3)
800

Data are from the World Bank (WDI,


historical stock exchanges only) from 1988
to 2009 and
600

from Thomson-Reuters (Monthly Market


Share Reports, all trading venues) since
2010
(in %)
400 200
0

1988 1990 1992 1994 1996 1998 2000 2002 2004 2006 2008 2010 2012

Japan Hong Kong


Australia China
Korea, Dem. Rep. South Africa
20
Author’s computation

10
G. Capelle-Blancard 30/03/2017

Financial Transaction Tax Around the


World: Past, Present and Future

PART II

21

Securities transaction tax (STT)


STT exist for a very long time.

Since the introduction of the stamp duty at the


London Stock Exchange in 1694 (the oldest existing
tax in the UK), there has been nearly 50 STT
implemented in very different countries

Very few information on the taxing scheme

An inventory: Capelle-Blancard (2014) “Financial Transaction


Taxes Around the World” (Version 1.0).

22

11
G. Capelle-Blancard 30/03/2017

Around the World: Past, Present


and Future

23
Author’s computation

Stocks Traded over GDP (in %)


and Securities Transaction Tax

24
Author’s computation

12
G. Capelle-Blancard 30/03/2017

Number of Countries with Securities


Transaction Tax: 1989-2015
40
30
20
10
0

1988
1989
1990
1991
1992
1993
1994
1995
1996
1997
1998
1999
2000
2001
2002
2003
2004
2005
2006
2007
2008
2009
2010
2011
2012
2013
2014
2015*
OECD Non-OECD
2015*: Expected
25
Author’s computation

Share of stock traded subject to


Securities Transaction Tax
100
80
60
(in %)
40
20

1988 1990 1992 1994 1996 1998 2000 2002 2004 2006 2008 2010 2012

In % of the world's total value of stock traded


In % of the world's total value of stock traded (except the US)
26
Author’s computation

13
G. Capelle-Blancard 30/03/2017

Empirical evidence: Financial


transaction taxes have no impact

PART III

27

What is the impact?


Past evidence

Critical overview
Endogeneity problems
Reverse causality
Simultaneity and omitted variable biases

Impact on liquidity? Negative,


Impact on volatility? Inconclusive.
6: inconclusive or statistically insignificant results
4: increase for some subsamples
1: decrease

28

14
G. Capelle-Blancard 30/03/2017

Table 1. Previous studies on the impact of an increase1 in STT (or akin transaction costs)

Studies Sample Type of Control group Measure of Results for Measure of Results for
reform liquidity liquidity1 volatility volatility1

Roll (1989) 23 countries STT Countries Std .dev. of returns Not significant
(1987-89) without STT
Umlauf (1993) Sweden STT None or Variance of returns Inconclusive
(1984-86) NYSE-FTSE
Jones & Seguin (1997) U.S. Fixed Nasdaq Std .dev. of returns, Positive impact
(1974-1976) commissions absolute returns
Saporta & Kan (1997) G.B. STT ADRs Variance of returns, Not significant
(1963-86) GARCH
Hu (1998) H.-K., Japan, STT None Turnover Inconclusive Std .dev. of returns Inconclusive
Korea, Taiwan
(1977-93)
Baltagi et al. (2006) China STT None Volume Negative Variance of returns, Positive impact
(1997) impact GARCH
Chou & Wang (2006) Taiwan STT None Volume, bid-ask Negative Realized volatility, Not significant
(1999-2001) impact high-low range
Liu (2007) Japan STT ADRs Volume, Negative
(1989) Autocorrelation impact
Phylaktis & Aristidou Greece STT None GARCH Positive impact
(2007) (1998-00) for highly traded
stocks
Foucault et al. (2011) France Cost of Spot market Autocovariance, Negative Std .dev. of returns, Negative impact
(1998 2002) forward Amihud impact
trading illiquidity
Pomeranets & Weaver U.S. STT None or Nasdaq Holden spread, Negative Std .dev. of returns Not significant
(2012) (1932-81) Amihud impact
illiquidity
Liau et al. (2012) Taiwan (1998-07) STT None GARCH Positive impact
1
Although reviewed studies analyze increases or decreases in transaction costs, the reported results must be interpreted as a reaction to an increase in STT.

29

What is the impact?


New evidence

France and Italy: quasi-natural experiments


The contribution is threefold.
Reliable samples
Several “close” control groups
Suitable variables
Several dimensions of liquidity and volatility
Trustworthy methods
DiD, propensity matching, discontinuity analysis

30

15
G. Capelle-Blancard 30/03/2017

Table. Comparison between the French and the Italian FTT schemes

French Italian
Date of intro. August 2012 March 2013 September 2013
a) a)
Tax base Shares HFT Shares HFT Derivatives
French resident companies Italian resident companies
Market cap. > € 1 billion Market cap. > € 500 million
Tax rate 0.1% Fixed amount:
0.2% 0.01% 0.02%
0.2% (OTC) b) between
Net daily balance of Modified orders > Net daily balance of Modified orders > € 0.01875 and
settled transactions 80% total orders c) settled transactions 60% total orders c) € 200 d)

Territorial scope FTT applies


irrespective of FTT applies irrespective of the buyer’s and seller’s
the buyer’s and The transactions residence
seller’s covered by the
residence tax are those
FTT applies carried out a
irrespective of company
the place where operated in FTT applies irrespective of the place where the
the transaction France transaction is executed and settles
is executed and
settles

French Italian
Main exemptions Market making activities and transactions effected in the course of liquidity assistance
Transactions of newly issued shares (primary market)
Transactions between related parties, intra-group and restructuring transactions
Transfers of ownership taking place by inheritance or gift
Repurchases of securities by the issuer
Collateralized transactions, temporary transfers of securities (repos)
Transactions by clearing houses or central depositories
Employee saving scheme Pension funds subject to supervision under Directive
transactions 2003/41/EC and certain EU social security institutions
Exchange or conversion of bonds Transactions on qualifying “ethical” or “socially
into shares responsible” financial products
Transactions with the EU, ECB, Central Banks of EU
member states or institutions established by
International Agreements recognised by Italy
Revenue € 600 million
0 na na na
(2013)

16
G. Capelle-Blancard 30/03/2017

A difference-in-difference approach
A generalized version of the difference-in-difference (DiD) :

= + + + + , (1)
where 0Vit is1a measure
2
of market
3
liquidity or volatility for the
firm i at time t, is a firm dummy variable, is a time dummy
variable, is a dummy variable that is equal to 1 for large
French firms after the introduction of the FTT.
Firm-level clustering of the errors

For Italy:

= + + + + + + ,
(2)
33

Treatment and control groups

FTT French Italy


Data Euronext 100 & Next 150 Borsa Italiana & Deutsche Börse
Groups Treated Control Treated Control
CAC+ AEX, BEL, LX MIB DAX & MDAX
Large caps
(#59) (#40) (#34) (#30)
FRA small ITA small
FRA (#49) ITA mid (#34)
Mid- and small caps
(#29) Foreign mid (#30) MDAX & SDAX
(#66) (#90)

34

17
G. Capelle-Blancard 30/03/2017

French large caps versus foreign large caps


(Euronext 100)

.4
.8

.3
.6

Density
Density

.2
.4

.1
.2

0
0

8 9 10 11 12 6 8 10 12 14
ln(market capitalization) ln(daily trading value)

FR_STT_100 noFR_noSTT_100 FR_STT_100 noFR_noSTT_100

35
Author’s computation

Italian large caps versus German


large caps
.4
.5
.4

.3
.3
Density

Density
.2
.2

.1
.1
0

6 8 10 12
ln(market capitalization) 12 14 16 18 20 22
ln(daily trading value)
MIB (treated) DAX & MDAX (control)
MIB (treated) DAX & MDAX (control)

36
Author’s computation

18
G. Capelle-Blancard 30/03/2017

Correlation of returns between treatment and control groups

This table presents the coefficient of correlation of daily (portfolio) returns between treatment and control groups for
previous studies and for our sample. For our sample, Panel A (59 French large caps) and Panel B (40 non-French large
caps) compose the Euronext 100 index. Panel C (29 French large midcaps), Panel D (49 French small midcaps) and
Panel E (66 non-French midcaps) compose the Next 150 index.

Study Period Treatment group Control group Correlation


of returns
MSCI UK 19.54%
Umlauf (1993) 1980-1987 Sweden
MSCI US 8.83%
Pomeranets &
1971-1981 NYSE Nasdaq 85.93%
Weaver (2012)
MSCI Belgium 78.95%
MSCI France
MSCI Netherlands 91.03%
Euronext 100 Euronext 100
90.61%
Panel A (FR, STT) Panel B (noFR, noSTT)
Next 150
90.15%
Feb. 2012- Next 150 Panel D (FR, no STT)
This study
Jan. 2013 Panel C (FT, STT) Next 150
84.42%
Panel E (noFR, no STT)
Euronext 100
DAX 30 92.62%
Panel A (FR, STT)
Next 150
MDAX 50 83.42%
Panel C (FT, STT)
37

Measuring market liquidity


Sarr and Lybek (2002), “number of measures must be considered
because there is no single theoretically correct and universally
accepted measure to determine a market’s degree of liquidity”.

Different aspects of market liquidity


Breath
Depth
Tightness
Resiliency

38

19
G. Capelle-Blancard 30/03/2017

Measuring market liquidity


Volume: Vi,t = Number of shares traded * closing price
for the stock

Turnover: Ti,t = Number of shares traded / total number


of shares available to ordinary investors

Bid-ask spread: Si,t = 2*(PAi,t–PBi,t) / (PAi,t+PBi,t) where PAi,t


and PBi,t are the asking price and the bid price

Liquidity Ratio: LRi,t = Vi,t / | Ri,t |

Price Reversal: PRi,t is minus the coefficient of a


regression of Ri,t on Vi,t–1*sign(Ri,t–1), controlling for Ri,t–1.
39

Measuring market volatility


Engle and Gallo (2006): “the concept of volatility itself is somewhat
elusive, as many ways exist to measure it and hence to model it”.

Squared Return: SRi,t = (Ri,t)² where Ri,t= log(Pi,t / Pi,t–1)


Absolute Return: ARi,t = |Ri,t| √̟/2
Conditional variance: CVi,t is proxied with a GARCH(1,1)
model.
High-low range: HLRi,t = (log PHi,t – log PLi,t)² / 4 log(2) where
PHi,t and PLi,t are the highest price and the lowest price achieved
for the stock i on the day t.

Daily Price Amplitude: DPAi,t = 2*(PHi,t–PLi,t)/(PHi,t+PLi,t)


40

20
G. Capelle-Blancard 30/03/2017

Graphical analysis

Daily Volume (million euros)


Euronext 100
70
60
50
40
30
20

01 Feb 12 02 Jun 12 02 Oct 12 01 Feb 13

FR_STT noFR_noSTT

41
Author’s computation

Daily Turnover (%) Daily Turnover (%)


Euronext 100 Next 150
.007

.01
.006

.008
.005

.006
.004

.004
.003

.002
.002

01 Feb 12 02 Jun 12 02 Oct 12 01 Feb 13


01 Feb 12 02 Jun 12 02 Oct 12 01 Feb 13
FR_STT FR_noSTT
FR_STT noFR_noSTT noFR_noSTT

Bid-Ask Spread (%) Bid-Ask Spread (%)


Euronext 100 Next 150
.16

.8
.14

.6
.12

.4
.1

.2
.08

01 Feb 12 02 Jun 12 02 Oct 12 01 Feb 13


01 Feb 12 02 Jun 12 02 Oct 12 01 Feb 13
FR_STT FR_noSTT
FR_STT noFR_noSTT noFR_noSTT

Liquidity Ratio Liquidity Ratio


Euronext 100 Next 150
50000 100000150000200000250000300000

10000 20000 30000 40000


0

01 Feb 12 02 Jun 12 02 Oct 12 01 Feb 13


01 Feb 12 02 Jun 12 02 Oct 12 01 Feb 13
FR_STT FR_noSTT
FR_STT noFR_noSTT noFR_noSTT

42
Author’s computation

21
G. Capelle-Blancard 30/03/2017

Absolute Return Absolute Return


Euronext 100 Next 150

4
2.5

3
2

2
1.5

1
1
01 Feb 1 2 02 Jun 12 02 Oct 12 01 Feb 13
01 Feb 12 02 Jun 12 02 Oct 12 01 Feb 13
FR_STT FR_noSTT
FR_STT noFR_noSTT noFR_noSTT

Conditional Variance Conditional Variance


Euronext 100 Next 150
5

14
12
4

10
8
3

6
4
2

01 Feb 1 2 02 Jun 12 02 Oct 12 01 Feb 13


01 Feb 12 02 Jun 12 02 Oct 12 01 Feb 13
FR_STT FR_noSTT
FR_STT noFR_noSTT noFR_noSTT

Daily Price Amplitude (%) Daily Price Amplitude (%)


Euronext 100 Next 150
3.5

5
3

4
2.5

3
2

2
1.5

01 Feb 1 2 02 Jun 12 02 Oct 12 01 Feb 13


01 Feb 12 02 Jun 12 02 Oct 12 01 Feb 13
FR_STT FR_noSTT
FR_STT noFR_noSTT noFR_noSTT

43
Author’s computation

Table 4. The impact of the French STT on stock market liquidity

Model (1) (2) (3)


Sample Euronext 100 Next 150
Nb. of firms FR STT (59) FR STT (29) FR STT (29)
no FR, no STT (40) FR, no STT (49) no FR, no STT (66)
ln(Volumei,t)
STT -0.2158*** -0.2671*** -0.3395***
(s.e.) (0.0456) (0.0834) (0.0900)
Nb. of obs. 25016 19802 24125
adj. R2 0.897 0.827 0.835
Turnoveri,t (%) × 100
STT -0.1275*** -0.1460 -0.1907**
(s.e.) (0.0245) (0.1120) (0.0791)
Nb. of obs. 24975 19802 24126
adj. R2 0.447 0.279 0.256
Bid-ask spreadi,t (%)
STT 0.0232*** 0.0530** -0.0186
(s.e.) (0.0071) (0.0254) (0.0830)
Nb. of obs. 24838 19762 24040
adj. R2 0.272 0.497 0.540
Liquidity ratioi,t × 1,000
STT -1.4628 -3.7223 -5.2771**
(s.e.) (10.5141) (2.4368) (2.4889)
Nb. of obs. 24800 19118 23351
adj. R2 0.099 0.058 0.057

44

22
G. Capelle-Blancard 30/03/2017

Table 5. The impact of the French STT on stock market volatility

Model (1) (2) (3)


Sample Euronext 100 Next 150
Nb. of firms FR STT (59) FR STT (29) FR STT (29)
no FR, no STT (40) FR, no STT (49) no FR, no STT (66)
Squared returni,t (%)
STT -0.1795 0.8566 -1.1502
(s.e.) (0.3904) (1.3615) (0.9570)
Nb. of obs. 25018 19812 24130
adj. R2 0.165 0.034 0.063
Absolute returni,t (%)
STT -0.0341 -0.0450 -0.0994
(s.e.) (0.0767) (0.1182) (0.0954)
Nb. of obs. 25018 19812 24130
adj. R2 0.296 0.168 0.190
Conditional variancei,t
STT 0.0341 0.0141 -0.4725
(s.e.) (0.4568) (1.0874) (0.8709)
Nb. of obs. 25146 19812 24130
adj. R2 0.589 0.281 0.362
High-low rangei,t
STT 0.0000 0.0001 -0.0001
(s.e.) (0.0000) (0.0001) (0.0001)
Nb. of obs. 25016 19802 24126
adj. R2 0.276 0.111 0.161
Price amplitudei,t (%)
STT 0.0091 -0.0130 -0.1304
(s.e.) (0.1006) (0.1329) (0.1445)
Nb. of obs. 25016 19802 24126
adj. R2 0.441 0.329 0.348
* **45
***
, , indicates a coefficient statistically different from zero at the 10%, 5%, 1% level, respectively.

Summary of the impact assessments of


the French STT

Authors Period Data Sample Methodology


Becchetti, Source: Bloomberg Sample: 106 French firms
90 trading days
Ferrari & Non parametric individual tests,
surrounding the Type: daily data (Paris Control group: 220 French Difference-in-difference
Trenta,
event Bourse) midcaps and small-caps
(2013)
Source: Thomson
Colliard & 4 months Sample: 90 French firms Difference-in-difference
Reuters
Hoffman surrounding the Control group: 28 Dutch Sub-samples: 20 stocks above and 20
event Type: tick data
(2013) firms + 20 French midcaps stocks below the threshold
(Euronext)
Source: Thomson Sample: 36 French blue
Haferkorn & 20 or 80 trading Difference-in-difference
Reuters chips
Zimmerman days surrounding Sub-samples: clusters by market
the event Type: tick data Control group: 30 German
n (2013) capitalization and price level
(Euronext & Chi-X) blue chips
Meyer, Sample: 94 French firms
80 trading days Source: Thomson Difference-in-difference, Matching
Wagener, &
surrounding the Reuters Type: tick data Control group: 99 UK
Weinhardt
event (Euronext & Chi-X) Sub-samples: size effect
(2013) firms

Source: Thomson
Capelle- Reuters Sample: 88 French firms Difference-in-difference, Propensity score
Blancard & 1 year surrounding
Havrylchyk the event Type: daily data Control group: 106 foreign matching
(2013) (Euronext & Deutsche firms + 49 French midcaps Robustness and sub-samples: size effect
Boerse)

46

23
G. Capelle-Blancard 30/03/2017

Traded Number Turnove Freq. of Spread Depth Low- Price Liquidity Price
value of trades r quotes latency impact ratio reversal
Becchetti,
Not Not
Ferrari & Trenta,
significant significant
(2013)
Colliard & Not
Decrease Decrease Decrease
Hoffman (2013) significant
Haferkorn &
Not
Zimmermann Decrease Increase Decrease
significant
(2013)
Meyer, Wagener,
Not Not
& Weinhardt Decrease Decrease Decrease
significant significant
(2013)
Capelle-
Blancard & Not Not
Decrease Decrease
Havrylchyk significant significant
(2013)

Standard Squared Conditional Daily high- Realized Price


deviation returns variance low range volatility dispersion

Becchetti, Ferrari &


Decrease
Trenta, (2013)
Colliard & Hoffman
Not significant
(2013)
Haferkorn &
Zimmermann Not significant Not significant Not significant Increase
(2013)
Capelle-Blancard &
Not significant Not significant Not significant
Havrylchyk (2013)
47

Figure B. Historical trading volume at the “Bourse de Paris”


Panel A. Annual turnover
2 500 000
€ million

2 000 000

1 500 000

1 000 000

500 000

0
1969
1971
1973
1975
1977
1979
1981
1983
1985
1987
1989
1991
1993
1995
1997
1999
2001
2003
2005
2007
2009
2011

48

24
G. Capelle-Blancard 30/03/2017

Figure B. Historical trading volume at the “Bourse de Paris”


Panel B. Annual variation
150%

100%

50%

0%
1970
1972
1974
1976
1978
1980
1982
1984
1986
1988
1990
1992
1994
1996
1998
2000
2002
2004
2006
2008
2010
2012
-50%49

The impact of the IOB on stock market liquidity and volatility


This table presents difference-in-difference econometric tests. The sample contains the stocks included in the
Euronext 100 (excluding 7 firms) over the 2007-2008 period. ln(Trading valuei,t) = ln(Number of shares traded for the
stock i on day t * Pi,t) where Pi,t is the closing price for the stock i on the day t; number of shares is expressed in
thousands. Estimated bid-ask spread, ESi,t = 2*(eα – 1) / *(1 + eα) with α, β and γ defined by Corwin and Schultz (2012)
and based on the highest and the lowest prices over one-day and two-day intervals. Liquidity Ratioi,t = Vi,t / | Ri,t | with
Ri,t = 100*log(Pi,t / Pi,t–1), for the stock i on the day t, respectively; liquidity ratio is expressed in thousands euros of
trades for a price change of 1%. Absolute Returni,t = |Ri,t| √π/2. Conditional variancei,t is estimated with a GARCH(1,1).
High-low range, HLRi,t = (log PHi,t – log PLi,t)² / 4 log(2) where PHi,t and PLi,t are the highest price and the lowest price
achieved for the stock i on the day t, respectively. Robust standard errors are clustered at the firm level and are
presented in parentheses.

Liquidity ln(Trading valuei,t) Estimated Liquidity Ratioi,t


bid-ask spreadi,t

IOB –0.0798 0.0001 0.26e+04


(s.e.) (0.0544) (0.0003) (2.66e+04)
Nb. of obs. 47,151 46,965 47,151
adj. R2 0.882 0.185 0.080

Volatility Absolute returni,t (%) Conditional variancei,t High-low rangei,t

IOB 0.0969 –0.0721 0.0001


(s.e.) (0.1292) (0.3510) (0.0001)
Nb. of obs. 47,151 45,123 47,151
adj. R2 0.399 0.496 0.336

* ** ***
, , indicates a coefficient statistically different from zero at the 10%, 5%, 1% level, respectively.
50

25
G. Capelle-Blancard 30/03/2017

Table 6. The impact of the Italian OTR, STT and FTT on stock market liquidity
This table presents difference-in-difference econometric tests. The sample period extends from 2011 to 2013 (821
days). The sample period extends from 2011 to 2013 (821 days). All the data are daily. OTR is a dummy variable = 1
from April 2012 if the firm is subject to the OTR and 0 otherwise. STT is a dummy variable = 1 from March 2013 if the
firm is subject to the STT and 0 otherwise. FTT is a dummy variable = 1 from September 2013 if the firm is subject to
the FTT and 0 otherwise. Pi,t is the closing price for the stock i on the day t. NBSTi,t is the number of shares traded for a
stock i on the day t. NOSHi,t is the total number of ordinary shares for the stock i on day t. NOSHFFi,t is the percentage
of shares available to ordinary investors for the stock i on day t. Trading valuei,t = NBSTi,t * Pi,t. Turnover ratioi,t =
100*NBSTi,t / (NOSHi,t * NOSHFFi,t). Bid-ask spreadi,t is the average difference between the asking price and the bid
price offered for the stock i on day t, in percent. Liquidity ratioi,t = Vi,t / | Ri,t |; liquidity ratio is expressed in million
euros of trades for a price change of 1%. Time and firms dummies are included but not reported. Standard errors are
clustered at the firm level and are presented in parentheses.

Sample All firms Large caps Mid- and small caps


(No. firms) (100 IT + 120 GE) (36 IT + 74 GE) (64 IT + 90)
ln(Trading valuei,t)
OTR 0.0008 0.0407 –0.2682***
(s.e.) (0.0791) (0.0754) (0.0858)
STT –0.1096 –0.0428 –0.0662
(s.e.) (0.0806) (0.0555) (0.1009)
FTT 0.0020 0.1911*** 0.1435*
(s.e.) (0.0548) (0.0527) (0.0827)
Nb. of obs. 167,420 82,188 91,320
adj. R2 0.906 0.876 0.798

51

Table 7. The impact of the Italian STT on stock market volatility


This table presents difference-in-difference econometric tests. The sample period extends from 2011 to 2013 (821
days). The sample period extends from 2011 to 2013 (821 days). All the data are daily. OTR is a dummy variable = 1
from April 2012 if the firm is subject to the OTR and 0 otherwise. STT is a dummy variable = 1 from March 2013 if the
firm is subject to the STT and 0 otherwise. FTT is a dummy variable = 1 from September 2013 if the firm is subject to
the FTT and 0 otherwise. Pi,t is the closing price for the stock i on the day t. Squared returni,t, SRi,t = (Ri,t)² with Ri,t=
100*log(Pi,t / Pi,t–1). Conditional variancei,t is estimated with a GARCH(1,1). High-low range, HLRi,t = (log PHi,t – log
PLi,t)² / 4 log(2) where PHi,t and PLi,t are the highest price and the lowest price achieved for the stock i on the day t,
respectively. Time and firms dummies are included but not reported. Standard errors are clustered at the firm level and
are presented in parentheses.

Treated group (No. firms) MIB (34) ITA (30)

Control group (No. firms) DAX+MDAX (74) MDAX+SDAX (90)


Filters Mk cap. > €100 bl.; Trading > €100 bl.
Squared returni,t (%)
OTR 2.3206*** 5.7382*
(s.e.) (0.6621) (3.2707)
STT –1.7423*** –5.4422
(s.e.) (0.5569) (3.9035)
FTT –0.6891* –0.1091
(s.e.) (0.3888) (0.2850)
Nb. of obs. 82,080 91,200
adj. R2 0.112 0.027
Conditional variancei,t
OTR 1.2601*** 1.5374***
(s.e.) (0.4206) (0.4168)
STT –0.8860*** –1.2802***
(s.e.) (0.2649) (0.3865)
FTT –0.4425** –0.3037
(s.e.) (0.1925) (0.1986)
Nb. of obs. 82,163 91,254
52 adj. R2 0.567 0.470

26
G. Capelle-Blancard 30/03/2017

Several robustness tests


Dynamic effects
Propensity score matching (market capitalization and turnover)
Nearest neighbor matching
Kernel matching
Discontinuity analysis
Others control groups
The impact on derivative market (in progress)

A global assessment
Only countries with a stock market capitalization >
US$2 billion.
=> 44 countries

The sample represents (in 2008)


86.5% of the world’s GDP
95.8% of the world’s stock market capitalization
98.9% of the world’s total value of stock traded.

54

27
G. Capelle-Blancard 30/03/2017

Figure 5. Heteregoneity across countries and over time


The sample is restricted to 44 countries with market capitalization > US$2 billion.
Data are from the World Bank (WDI, historical stock exchanges only)
A. Stocks Traded over GDP
800

800
600
600

400
400

200
200

0
0

1988
1989
1990
1991
1992
1993
1994
1995
1996
1997
1998
1999
2000
2001
2002
2003
2004
2005
2006
2007
2008
2009
2010
2011
2012
AUS
AUT
BHR
BRA
CAN
CHL
CHN
DNK

DEU

IRL

LUX

NLD

PHL
PRT

CHE
ARE
USA
ARG

BEL

CZE

FRA
GRC
HKG
IND
ISR
JPN
JOR
KWT
MYS
MEX
NZL
NOR
PAK

QAT
SAU
SGP
ZAF
KOR
ESP
SWE
THA
GBR
FIN

ITA

Country (alphabetical order) Year

Value traded/GDP mean Value traded/GDP mean

55
Author’s computation

Figure 5. Heteregoneity across countries and over time


The sample is restricted to 44 countries with market capitalization > US$2 billion.
Data are from the World Bank (WDI, historical stock exchanges only)
B. Turnover Ratio
500

500
400
400

300
300

200
200

100
100

0
0

1988
1989
1990
1991
1992
1993
1994
1995
1996
1997
1998
1999
2000
2001
2002
2003
2004
2005
2006
2007
2008
2009
2010
2011
2012
AUS
BHR
BRA
CAN
CHN

DEU
CHL

DNK

LUX

NLD

PHL

CHE
ARE
USA
ARG
AUT

ISR
JPN
JOR
BEL

CZE

FRA
GRC
HKG
IND
IRL

KWT
MYS
MEX

NOR
NZL
PAK
PRT
QAT
SAU
SGP
ZAF
KOR
ESP
SWE
THA
GBR
FIN

ITA

Country (alphabetical order) Year

Turnover Ratio mean Turnover Ratio mean

56
Author’s computation

28
G. Capelle-Blancard 30/03/2017

The empirical drivers of stock trading


and the impact of STT
Panel data econometric model with country fixed-
effects models
The fixed-effects model controls for all time-invariant
differences between the individuals, so the estimated
coefficients of the fixed-effects models are presumed to not
be biased because of omitted time-invariant characteristics.

ST_GDPj,t = β’ Xj,t + δ STTj,t + αj + uj,t

TRj,t = β’ Xj,t + δ STTj,t + αj + uj,t

57

Which factors?
Financial Depth
e.g. Domestic credit to private sector
Efficiency and banks’ market structure
e.g. Lending-deposit spread
Financial openness
e.g. Loans from nonresident banks
Stability
e.g. banking crisis, stock market volatility
Economics
Legal origin
58

29
G. Capelle-Blancard 30/03/2017

Preliminary diagnostics
Heteroskedasticity
Cross-sectional dependence (contemporaneous correlation)
First-order serial correlation
4 different model regressions with alternative standard error
estimates :
1. A simple fixed-effect regression model;
2. A fixed-effect regression model with heteroscedasticity consistent
(or “Huber-White”) standard errors;
3. A fixed-effect regression model with bootstrap inference;
4. A fixed-effect regression model with the cross-sectional dependence
consistent Driscoll-Kraay estimator.
The preferred estimator is the last one, which provides a
nonparametric covariance matrix estimator which produces
heteroscedasticity consistent standard errors that are robust to
very general forms of spatial and temporal dependence. Indeed,
given the high level of stock market contagion and
interdependences, cross-sectional dependence is very likely.
59

Table A. General diagnostic tests for panel data models


The dataset contains annual data from 1988 to 2011 for a panel of 44 countries with market capitalization larger than
US$2 billion. Stock traded/GDP is the total value of stock traded over GDP (%) and the turnover ratio is defined as the
ratio of annual traded value to the year-end market capitalization. Data are from the World Bank (WDI and GFDD).

Stock traded/ Turnover


GDP ratio
Modified Wald test for groupwise heteroskedasticity
H0: σi² = σ² for all i
Chi2 (38) 59,842.51 48,922.84
Prob. > chi2 0.0000 0.0000
Breusch-Pagan LM test of independence
Chi2(528) 2,558.229 2,475.986
Prob. > chi2 0.0000 0.0000
Pasaran cross-sectional dependence test
Z-test 21.035 11.913
Prob. > Z-test 0.0000 0.0000
Average absolute value of the off-diagonal elements 0.391 0.326
Wooldridge test for autocorrelation in panel data
H0: no first-order autocorrelation
F(1, 43) 94.097 40.897
Prob. > F 0.0000 0.0000
Hausman test (according to Wooldridge, the auxiliary regression is estimated with
Driscoll and Kraay standard errors)
F(11, 23) 66.96 73.58
60 Prob. > F 0.0000 0.0000

30
G. Capelle-Blancard 30/03/2017

The empirical determinants of stock trading


This table provides the coefficient estimates from the regression models in (Eq. 1) and (Eq. 2) estimated by fixed effects
(within) regression. The standard errors (in parentheses) are obtained from the covariance matrix estimators in the
column headings. The dataset contains annual data from 1988 to 2011 for a panel of 44 countries with market
capitalization larger than US$2 billion. The dependent variable in the regression is the total value of stocks traded over
GDP (%) in panel A and the turnover ratio (defined as the ratio of annual traded value to the year-end market
capitalization) in Panel B. Data are from the World Bank (WDI and GFDD).

Panel A (1) (2) (3) (4)


Stocks traded over GDP (%) Fixed-effects Robust Bootstrap Driscoll-Kraay
Credit/GDP ns ns ns ns
Nb listed firms / Pop. 1.1874*** ns ns 1.1874*
Foreign claim/GDP ns ns ns ns
Bank Spread ns ns ns ns
Bank crisis (dummy) –18.0693* –18.0693* –18.0693* –18.0693**
Credit/deposits –0.2877** ns ns ns
Stock Market Volatility 0.9881*** 0.9881* 0.9881* 0.9881**
STT (dummy) ns ns ns ns
GDP (ln) ns ns ns ns
GDP (ln) * legal origin UK ns ns ns ns
GDP/cap. 0.0070*** ns ns 0.0070**
N 477 477 477 477
adj. R2 0.232 0.293 0.232
*
p < 0.10, ** p < 0.05, *** p < 0.01

61
s.e. are not reported here to save space. The trend and the constant are also omitted.

The empirical determinants of stock trading (continued)


This table provides the coefficient estimates from the regression models in (Eq. 1) and (Eq. 2) estimated by fixed effects
(within) regression. The standard errors (in parentheses) are obtained from the covariance matrix estimators in the
column headings. The dataset contains annual data from 1988 to 2011 for a panel of 44 countries with market
capitalization larger than US$2 billion. The dependent variable in the regression is the total value of stocks traded over
GDP (%) in panel A and the turnover ratio (defined as the ratio of annual traded value to the year-end market
capitalization) in Panel B. Data are from the World Bank (WDI and GFDD).

Panel B (1) (2) (3) (4)


Turnover ratio (%) Fixed-effects Robust Bootstrap Driscoll-Kraay
Credit/GDP ns ns ns ns
Nb listed firms / Pop. ns ns ns ns
Foreign claim/GDP ns ns ns ns
Bank Spread ns ns ns ns
Bank crisis (dummy) –16.0990** –16.0990** –16.0990** –16.0990***
Credit/deposits ns ns ns ns
Stock Market Volatility 1.1997*** 1.1997** 1.1997*** 1.1997***
STT (dummy) ns ns ns ns
GDP (ln) ns ns ns ns
GDP (ln) * legal origin UK ns ns ns ns
GDP/cap. 0.0027*** ns ns 0.0027**
N 471 471 471 471
adj. R2 0.033 0.111 0.033
*
p < 0.10, ** p < 0.05, *** p < 0.01

62
s.e. are not reported here to save space. The trend and the constant are also omitted.

31
G. Capelle-Blancard 30/03/2017

The empirical determinants of stock trading with first differences


This table provides the coefficient estimates from the regression models in (Eq. 1) and (Eq. 2) estimated by fixed effects
(within) regression with first-differences. The standard errors (in parentheses) are obtained from the covariance matrix
estimators in the column headings. The dataset contains annual data from 1988 to 2011 for a panel of 44 countries with
market capitalization larger than US$2 billion. The dependent variable in the regression is the total value of stocks
traded over GDP (%) in panel A and the turnover ratio (defined as the ratio of annual traded value to the year-end
market capitalization) in Panel B. Data are from the World Bank (WDI and GFDD).

Panel A (1) (2) (3) (4)


Stocks traded over GDP (%) Fixed-effects Robust Bootstrap Driscoll-Kraay
Stocks traded/GDP 0.0028*** 0.0028** ns 0.0028**
Credit/GDP ns ns ns ns
Nb listed firms / Pop. ns ns ns ns
Foreign claim/GDP ns ns ns ns
Bank Spread ns ns ns ns
Bank crisis (dummy) ns ns ns ns
Credit/deposits 0.0053** ns ns ns
Stock Market Volatility ns ns ns ns
ΔSTT (dummy) –0.5898* ns ns ns
GDP (ln) –0.7032*** –0.7032** –0.7032** –0.7032*
ΔGDP (ln) ns 0.6890** 0.6890** 0.6890*
GDP/cap. ns ns ns ns
N 471 471 471 471
adj. R2 0.008 0.088 0.008
*
p < 0.10, ** p < 0.05, *** p < 0.01

63
s.e. are not reported here to save space. The trend and the constant are also omitted.

The empirical determinants of stock trading with first differences (cont’d)


This table provides the coefficient estimates from the regression models in (Eq. 1) and (Eq. 2) estimated by fixed effects
(within) regression with first-differences. The standard errors (in parentheses) are obtained from the covariance matrix
estimators in the column headings. The dataset contains annual data from 1988 to 2011 for a panel of 44 countries with
market capitalization larger than US$2 billion. The dependent variable in the regression is the total value of stocks
traded over GDP (%) in panel A and the turnover ratio (defined as the ratio of annual traded value to the year-end
market capitalization) in Panel B. Data are from the World Bank (WDI and GFDD).

Panel B (1) (2) (3) (4)


Turnover ratio (%) Fixed-effects Robust Bootstrap Driscoll-Kraay
Turnover Ratio 0.3672*** 0.3672** 0.3672** 0.3672***
Credit/GDP ns ns ns –0.1570*
Nb listed firms / Pop. ns ns ns ns
Foreign claim/GDP ns ns ns ns
Bank Spread ns ns ns ns
Bank crisis (dummy) ns ns ns ns
Credit/deposits ns ns ns ns
Stock Market Volatility –0.8441*** –0.8441*** –0.8441*** –0.8441***
ΔSTT (dummy) ns ns ns ns
GDP (ln) ns ns ns ns
ΔGDP (ln) ns ns ns ns
GDP/cap. ns ns ns ns
N 464 464 464 464
adj. R2 0.088 0.163 0.088
*
p < 0.10, ** p < 0.05, *** p < 0.01
64
s.e. are not reported here to save space. The trend and the constant are also omitted.

32
G. Capelle-Blancard 30/03/2017

Revenues from FTT

PART IV

65

Revenues from FTT (Europe)


% Total fiscal revenue

66
Author’s computation

33
G. Capelle-Blancard 30/03/2017

% Total fiscal revenue


Revenues from FTT (Asia+SA)

67
Author’s computation

Figure A. Revenues from the French FTT


1 000 0,25%
FTT revenue (€ million), lhs
900 As % of total tax revenue, rhs

800 0,20%

700

600 0,15%

500

400 0,10%

300

200 0,05%

100

0 0,00%
1970
1972
1974
1976
1978
1980
1982
1984
1986
1988
1990
1992
1994
1996
1998
2000
2002
2004
2006
2008
2010
2012

Source: OECD. 1970-2007: “Impôt de Bourse”; 2012-2013 (expected): “Taxe sur les transactions financières”
68
Author’s computation

34
G. Capelle-Blancard 30/03/2017

French FTT – Fiscal revenue

69

Effective (implicit) FTT rate

FTT revenue
Value of share FTT revenue
2013 (% Total fiscal Implicit FTT rate
trading (€ million) (€ million)
revenue)
France 2 243 180 697 0,07% 0,03%
Royaume-Uni 3 622 338 3 656 0,50% 0,10%
Irlande 360 902 251 0,66% 0,07%
Suisse 1 106 411 1 481 1,06% 0,13%
Taiwan 473 307 2 286 4,04% 0,48%
Corée 777 792 2 113 0,89% 0,27%
Hong Kong 1 176 062 2 492 6,39% 0,21%

70

35
G. Capelle-Blancard 30/03/2017

Conclusion

PART V

71

To sum up (1)

These papers analyse the impact of STT on market volatility. This


question is at the heart of economic policy debate about the use of
taxes to curb speculative activity and render financial markets more
stable. The opponents argue that additional taxation of financial
transactions will hurt market liquidity, thus, making markets even
more volatile.
I studied the impact of the STT/FTT introduced in France (2012)
and Italy (2013). Unlike previous studies, I isolate the effect of the
tax due to the unique design of these taxes, which are levied only
on large firms
Diff-in-diff, Propensity Score Matching, Regression Discontinuity

72

36
G. Capelle-Blancard 30/03/2017

To sump up (2)

Overall, results show that the introduction of the STT/FTT has


reduced trading value and volume of stocks

At the same time, there is no effect on theoretically based


measures of liquidity, such as price impact or price reversal.
As to volatility measures, the results are insignificant.

At the end, our results are very much in line with Bloomfield,
O’Hara and Saar (2009) who conclude that “[STT] reduces
activity by noise and informed traders roughly equally (...), and
perhaps as a result it does not alter bid-ask spreads or other price
impact measures of liquidity, and has only a weak effect (if at all)
on the informational efficiency of prices”.
73

Conclusion

Neither a panacea nor a “madness”

There is no significant effect on market volatility and


even liquidity

The design of the FTT is crucial

Overall, FTT seems more a revenue-focused tax than a


corrective tax

74

37

You might also like