Transaction costs, liquidity and expected returns at the Berlin Stock Exchange, 1892-1913

Carsten Burhop, Universität zu Köln Sergey Gelman, ICEF, Higher School of

Economics, Moscow

1st ILFE Workshop,

Moscow, September 18, 2010

2

Motivation

Explore effective transaction cost determinants & effects in a ‘friendly environment’: on an early call auction stock market over a long time span

3

Outline

1. Literature review

2. Historical background

3. Data & Methodology

4. Results

5. Conclusion

4

1. Literature Review I: liquidity & asset pricing

• Amihud (2002, JFM)– Positive risk premium for expected illiquidity– Inverse relation of returns and unexpected illiquidity shocks

• Eleswarapu/Reinganum (1993), Brennan and Subrahmanyam (1996) – Negative/insignificant risk premia

• Bekaert et al. (2007, RFS) – Dynamic interdependence of liquidity and returns on the market

level (whereby liquidity only weakly dependent);– Transaction cost adjustment + liquidity risk premium

• Goyenko et al. (2009, JFE)– Effective transaction cost measures capture liquidity (incl. price

impact)

5

1. Literature Review I: economic history

• Rajan & Zingales (2003): German pre-1913 stock market development higher than US

• Baltzer (2006): price differentials across stock exchanges negligible

• Gelman & Burhop (2008)– weak information efficiency on a rather high level– Efficiency worsens during crises 1901, 1913

• Gehrig & Fohlin (2006) – estimate effective transaction costs for Berlin stock

exchange in 1880, 1890, 1900, 1910. Find gradual decline.

– find inverse relationship to size

6

1. Contribution

• Transaction costs were on average low, but rather variable in time and cross-section

• Transaction costs are inversely influenced by size and previous year returns; are higher in crises

• There is a significant positive liquidity premium, which is more pronounced than market risk and size premia

7

2. Historical background I

• Berlin Stock Exchange (BSE) was the major German stock exchange since 1870-s

• Steadily increasing # of traded companies, around 1000 in 1913

• Trading 6 days per week, one price per day

• Call-auction mechanism with a specialist • Presence of informed insiders possible

8

Berlin Stock market performance

0

50

100

150

200

250

300

350

400

450

500

31

.12

.18

91

05

.05

.18

94

14

.09

.18

96

31

.01

.18

99

15

.06

.19

01

24

.10

.19

03

08

.03

.19

06

23

.07

.19

08

30

.11

.19

10

18

.04

.19

13

Daily Index Eube's market index Ronge's DAX-30

Leipziger Bank defaults

Balkan warBank run in US

9

German aggregate stock trading volume (in bln mark)

20

40

60

80

100

18

92

18

94

18

96

18

98

19

00

19

02

19

04

19

06

19

08

19

10

19

12

10

2. Historical background II

• Major crises with impact on efficiency:– Bankruptcy of Leipziger Bank 1901 – Balkan war fear 1913

• Fixed relative transaction costs:– Transaction tax: 0.01% up to 04/1894; 0.02% to

10/1900 and 0.03% until the end of the sample– Broker fee: official 0.05%; private 0.025%– Provisions for intermediaries: 0.1-0.33%– Total round-trip transaction cost: 0.252-0.82%– Tick size 0.05 Mark (by stock prices of 40 Mark and

above) less than 0.125%

11

3. Data• Daily stock prices for 27 stocks (hand-collected

from Berliner Börsenzeitung) 1892-1913, 6692 observations per company– Industries: banking, machinery, chemicals, mining,

textile, etc.– Requirement: listed during the whole period, <30%

zero returns• Trading volume is available only on annual basis

aggregated for all German exchanges! • Daily stock index values (from Gelman/Burhop 2008)• Annual values for market capitalization

– Heterogeneous: from 0.3 bln RM to 32.8 bln RM• Dividend amounts and dates

12

Descriptive statistics (selection)

Name Mean (ann.) Max. Min.

Std. Dev.

Skew-ness

Kur-tosis

zeros (1) Average MCap.

(mill M)

1 AG für Anilinfabrikation 0.0678 0.1257 -0.2270 0.0082 -3.85 126.25 0.1638 0.0008 2727

2 Allgemeine Elektricitätsgesellschaft 0.0336 0.0526 -0.0611 0.0065 -0.18 11.94 0.0807 0.0820 14997

5 Deutsche Bank 0.0294 0.0333 -0.0544 0.0042 -1.64 24.72 0.1001 -0.0119 32778 9 Deutsche Spiegelglas 0.0687 0.0921 -0.0838 0.0080 -0.30 18.32 0.1877 0.0716 643

10 Erdmannsdorfer Spinnerei 0.0001 0.1143 -0.0774 0.0106 0.62 13.70 0.2497 -0.0425 286

22 Schering 0.0162 0.0652 -0.0657 0.0083 0.17 10.17 0.1630 0.0610 1298 23 Schlesische Zinkhütten 0.0336 0.1079 -0.0853 0.0066 -0.52 33.99 0.2360 -0.0582 7947 26 Siemens Glas-Industrie 0.0287 0.0438 -0.0576 0.0058 -1.12 19.17 0.1966 -0.0273 2290

Value-weighted index 0.0687 0.0296 -0.0562 0.0032 -1.68 30.78 n/a 0.165 161344

13

3. Methodology I

• Measure of full transaction costs (fixed costs + price impact):– LOT (1999): information-based measure

* l * li,t i,t i i,t i

l * hi,t i i,t i

* h * hi,t i,t i i,t i

*i,t i m,t i,t

r r if r

r 0 if < r

r r if r

r r e

14

3. Methodology I

• Estimate with MLE

1 0

2

1, , , ,

1

S.T. 0, 0, 0, 0,

l h ll h it i i mt i i mt i i mti i i i it mt

i i i i

hit i i mt

i i

l hi i i i

r r r rL r r

r r

15

3. Methodology I

• Criticism of LOT measure– Zero returns may be due to noise trading– The measure is driven by the market return

volatility– Does not incorporate other factors than market

• Justification– Is the only available measure of the full

transaction costs and not only spreads– Widely used in recent financial literature, e.g.

Griffin et al. (2010, RFS); Lesmond (2005, JFE)

16

3. Methodology II: Determinants

• Cross-section and Panel estimation

• Dependent variable: annual effective TC (LOT measure) of a company

• Regressors:– Market cap (for size)– Previous year returns– Aggregate trading volume or Time dummies

17

3. Methodology III: impact on asset pricing

• Fama-MacBeth(1973) regression – monthly returns– factor loadings & firm characteristics

• Factor: market risk (our index as proxy)• Characteristics:

– Size– Daily return autocorrelation (momentum)– LOT transaction cost measure (for

illiquidity)

18

4. Results: annual transaction costsTABLE 2: ANNUAL AVERAGE OF TRANSACTION COSTS

Year LOT

1892 1.454

1893 1.584

1894 1.072

1895 0.925

1896 0.805

1897 0.814

1898 0.908

1899 0.878

1900 1.029

1901 1.678

1902 0.977

1903 0.848

1904 0.825

1905 0.696

1906 0.658

1907 0.775

1908 0.846

1909 0.731

1910 1.039

1911 0.713

1912 0.883

1913 1.124

Average 0.966

Own calculations based on daily returns for 27 stocks for the period 1892-1913.

Expressed in percent of share price, equally weighted averages. Two outliers were dropped.

19

4. Results: annual transaction costs

LOT

0

0.5

1

1.5

2

1892 1895 1898 1901 1904 1907 1910 1913

20

4. Transaction costs BSE 1892-1913: rolling window

-0.5

0

0.5

1

1.5

2

29.1

2.18

92

21.1

2.18

94

23.1

2.18

96

02.0

1.18

99

03.0

1.19

01

06.0

1.19

03

05.0

1.19

05

10.0

1.19

07

13.0

1.19

09

12.0

1.19

11

15.0

1.19

13

rollgknlot

21

4. Results: time series of transaction costs

• Transaction costs are low: average LOT-measure of 0.97%, – lower than for the upper decile of NYSE

(1.23%) in 1963-1990 (Lesmond et al. 1999)– better than any of the emerging stock markets

in 1990-s (Lesmond 2005)– But a bit above than DJIA costs of 0.6% 1970-

1980 (Goyenko et al. 2009)

• High variation: from 0.66% (1906) to 1.68% (1901)

22

Appendix 1: Average transaction costs of corporations, included in the investigation

Number Name Average

LOT measure

1 AG für Anilinfabrikation 0.943

2 Allgemeine Elektricitätsgesellschaft 0.520

3 Berlin-Anhaltinische Maschinenbau 0.902

4 Bochumer Bergwerk (Lit C) 3.164

5 Bank für Handel und Industrie 0.543

6 Deutsche Bank 0.384

7 Dresdner Bank 0.446

8 Deutsche Jute Spinnerei und Weberei 1.109

9 Deutsche Spiegelglas 1.097

10 Erdmannsdorfer Spinnerei 1.689

11 Gelsenkirchener Bergwerksgesellschaft 0.427

12 Gerresheimer Glashütten 1.284

13 Hallesche Maschinenfabriken 1.112

14 Harpener Bergbau AG 0.425

15 Kattowitzer AG für Bergbau und Eisen 0.667

16 Maschinenfabrik Kappel 1.239

17 Norddeutsche Wollkämmerei 1.135

18 Oberschlesische Portland-Cement AG 1.094

19 Rheinische Stahlwerke 0.781

20 Rositzer Zuckerfabrik 1.053

21 Schaaffhausen'scher Bankverein 0.572 22 Chemische Fabrik vormals Schering 1.001

23 Schlesische Zinkhütten 0.959

24 Schlesische Leinen-Industrie 1.183

25 Schultheiss Brauerei 0.684

26 Siemens Glas-Industrie 0.776

27 Stettiner Chamottewaren 0.905

23

4. Determinants of transaction costs: Cross-sectional results

Average LOT measure

0

2

4

5 7.5 10

1892

0.62 0.03

2 2

ˆ4.92 0.19 ln ,

ˆ0.64, 0,0.20

LOTi i i

i

S MC e

R e

ln(MCap)

24

4. Determinants of transaction costs: panel (1) FE (2) FE (3) FE (4) FE (5) RE (6) GMM

Constant 1.00***

(0.06)

0.96***

(0.05)

0.94***

(0.07)

1.79***

(0.19)

1.85***

(0.20)

Sit-1 0.44***

(0.01)

MCit/ MCit -3.12**

(1.57)

-2.59**

(1.19)

-2.59

(1.91)

-2.58

(1.59)

-4.29***

(0.73)

-0.28

(1.50)

lnPit-1 -0.25

(0.16)

-0.34**

(0.16)

-0.45***

(0.16)

-0.44***

(0.16)

-0.11***

(0.04)

lnTVt -0.20***

(0.05)

-0.20***

(0.05)

-0.22***

(0.01)

t1901 0.26***

(0.08)

t1913 0.25***

(0.07)

Time

effects

Y Y N N N N

R2 0.56 0.60 0.48 0.46 0.27

25

4. Determinants of transaction costs: panel (1) FE (2) FE (3) FE (4) FE (5) RE (6) GMM

Constant 1.00***

(0.06)

0.96***

(0.05)

0.94***

(0.07)

1.79***

(0.19)

1.85***

(0.20)

Sit-1 0.44***

(0.01)

MCit/ MCit -3.12**

(1.57)

-2.59**

(1.19)

-2.59

(1.91)

-2.58

(1.59)

-4.29***

(0.73)

-0.28

(1.50)

lnPit-1 -0.25

(0.16)

-0.34**

(0.16)

-0.45***

(0.16)

-0.44***

(0.16)

-0.11***

(0.04)

lnTVt -0.20***

(0.05)

-0.20***

(0.05)

-0.22***

(0.01)

t1901 0.26***

(0.08)

t1913 0.25***

(0.07)

Time

effects

Y Y N N N N

R2 0.56 0.60 0.48 0.46 0.27

26

4. Determinants of transaction costs: results

• Inverse relation with size – explains about 2/3 of transaction cost variation in

cross-section and 23% in a panel set-up– One std increase in share of m. cap. (0.05) leads to

0.125-0.2 decrease in transaction costs– significance vanishes in FE set-up if we include past

returns

• Inverse relationship with previous year returns explains about 10%– One std decrease in past returns (0.126) leads to

apprx 0.05 increase in LOT

27

4. Determinants of transaction costs: results

• Transaction costs are about 0.25 percentage points higher in crises years

• Transaction costs are inversely related to trade volume– One std increase in log trading volume (0.25)

induces 0.05 decrease in transaction costs

28

4. Effects of transaction costs on asset pricing

(1) (2) (3) (4)

Constant .0018

(.0014)

-.0024

(.0021)

.0024

(.0107)

.0034

(.0108)

Market beta -.0003

(.0019)

.0016

(.0020)

.0013

(.0021)

-.0001

(.0022)

Transaction cost

lagged TC

.3266**

(.1324)

.3068*

(.1773)

.3055*

(.1771)

Size S -.0002

(.0004)

-.0002

(.0004)

Momentum M .0105*

(.0058)

Average R2 0.07 0.12 0.16 0.20

# of cross-sections T 264 252 252 252

29

4. Asset pricing results

• We find support of Amihud (2002):– Lagged transaction costs increase expected

return– Contemporaneous TC – decrease returns

• CAPM doesn’t work• Size effect is absorbed by ex-ante

transaction cost measure• Momentum is positive with tendency to

significance

30

4. Asset pricing results

• Different specifications of liquidity risk do not yield significant results

31

5. Conclusion

• Transaction costs of the Berlin Stock Exchange were on average rather low as early as 1892-1913

• Size and past returns were negatively and crises were positively related to transaction costs

• Illiquidity was the primary concern of investors by asset pricing, levied a positive premium

32

Thank you for your attention