You are on page 1of 33

Was the US Lucky?

The Distribution of Global Equity Returns


Matthew C. Pollard

Australian National University


London Business School, Visiting Scholar

May 7, 2009
1 Luck
Outcomes of many events are uncertain.
Ex-post winners: successful traders, richest individuals, etc.
How much was due to skill (ex-ante) or luck (ex-post)?

In stock markets, focus on the ex-post winner: the US


Many other countries, far worse outcomes.
If random outcomes, one country will look good by chance.

Were US investors simply lucky?


I argue yes
Global equity markets in the 20th Century, 1930 to 1996

Argentina, Chile, Columbia, Venezuela, Mexico, Peru, Aus-


tria, Belgium, Germany, Denmark, France, Finland, Italy, Nether-
lands, Norway, Portugal, Spain, Switzerland, Sweden, USA,
South Africa, UK, Canada, Ireland, New Zealand, India, Egypt,
Israel, Japan, Lebanon

Real index values, in excess of inflation.

Thanks to William Goetzmann, Yale School of Management


10000
US

1000
100
Index

10
1

1930 1940 1950 1960 1970 1980 1990

Figure 1: The United States


10000
US
Ca

UK

1000
NZ
Au
SA
100
Index

10
1

1930 1940 1950 1960 1970 1980 1990

Figure 2: The British Colonies


10000
Gr
Fr

Sz
Dm
Nt
Nw

1000
Bg

Pg
Fi
100

Sp
Index

10

It
1

1930 1940 1950 1960 1970 1980 1990

Figure 3: Mainland Europe


10000
1000 Au
NZ
Lb

Eg
In
Jp
100
Index

Is
10
1

1930 1940 1950 1960 1970 1980 1990

Figure 4: East Asia and the Middle East


10000
1000
Ch

Mx
Ar
100

Vn
Index

10

Co
1

PU

1930 1940 1950 1960 1970 1980 1990

Figure 5: South America


10000
1000
Ch

Mx
Ar
100

Vn
Index

10

Co
1

PU

1930 1940 1950 1960 1970 1980 1990

Figure 6: South America


Stock Markets in 1900 Stock Markets in 2000

South Africa Australia

Latin America Sweden

Japan Canada

Austria-Hungry Switzerland

Other New Markets

India Others from 1900

Russia UK

UK Japan

USA Euroland

Euroland USA

0 10 20 30 40 50 0 10 20 30 40 50
Capitalization, % Capitalization, %

Figure 7: Market Capitalization, % of World Market in 2000


The US is the ex-post winner.

US real-return: ∼ 5-7% pa. Global real-return : ∼ 0-1% pa.

It wasn’t certain in 1900s that US would win financially.


Western Europe: safe; Russia, Argentina the growth stocks.

But: WW1, WW2, Revolution, Cold War, Financial Crises.


Many countries suffered great, unexpected ordeals.
Russia and China, -100% return on investment, 1900 to 2000.

Ex-ante vs Ex-post? Big difference, lots of noise.


2 US Equity Premium Puzzle
The US ex-post equity premium is too large.
Sensible only with extremely high risk-aversion (γ > 50).
Puzzle arises in the Hansen-Jagannathan bound,

|ERt − Rf | σ(mt )
≤ .
σ(Rt ) E(mt )

Many solutions tweak utility specification on RHS


What about the LHS:
Perhaps ex-ante equity premium in the US may might be much
lower than the ex-post premium.
A small ex-ante premium automatically resolves the puzzle.
3 Self-Selection Bias in US premium
Literature is studying the most successful financial market ever.
We should expect the best to appear exceptional. It is not rep-
resentative.

Statistically, let R1 be the US 20th century average excess-


return.

E[R1 |R1 = max(R1 , R2 , ..., RN )] > E(R1 ) .


| {z } | {z }
expected maximum ex−post ex−ante

Self-selection bias equals the difference in the two terms.


4 The Model
Use the simplest possible model. N countries, each country is
ex-ante identical. Ex-post annual returns are drawn by

Ri ∼ N (µ, σ 2 )

All countries have same mean, variance, positively correlated.

Model parameters set to in-sample global averages


N = 40 countries, T = 50 years, σ = 30%

Assumed µ = 0, zero ex-ante equity premium. Pessimistic as-


sumption, close to 20th century global average.
10000
1000
100
Index

10
1

0 10 20 30 40 50

Year

Figure 8: Model simulation, uncorrelated returns, blue: ex-post winner.


10000
Index

100
1

0 10 20 30 40 50

Year

Figure 9: Model simulation, uncorrelated returns, blue: ex-post winner.


50000
20000
5000
1000
Index

500
200
100
50
20

0 10 20 30 40 50

Year

Figure 10: Model simulation, correlated returns ρ = 0.4, blue: ex-post win-
5000
1000 2000
500
200
Index

100
50
20
10
5

0 10 20 30 40 50

Year

Figure 11: Model simulation, correlated returns ρ = 0.4, blue: ex-post win-
10000
1000
100
Index

10
1

1930 1940 1950 1960 1970 1980 1990

Figure 12: Actual 20th Century Equity performance, blue: USA


5 Tests of the Model
The model has four simple tests:

1. Does it predict the US ex-post premium and Sharpe ratio?

2. Does it predict the distribution of Global ex-post premia?

3. Does it predict the distribution of Global Sharpe ratios?

4. Does it predict the distribution of Global capitalizations?

Strong tests for a simple model.

The Results...
Figure 13: Distribution of price path conditioned on ex-post maximum
6 Results – US Premia
The model predicts the winner’s ex-post premium of
σ N −α
!
ER1 = √ QN
T N − 2α + 1

Expected premium is 6.1%, and Sharpe ratio is 32%


Historical US is 6.2%, and Sharpe ratio is 37% (Shiller, 2008).

Conclusion:
No statistical difference between the US results and the pre-
dicted result. High ex-post premium consistent with zero ex-
ante premium.
Results – Global Premia
The model predicts the cross-sectional distribution of global eq-
uity premia as
σ2
R ∼ N (0, )
T
Empirical distribution of world premia is statistically indistin-
guishable from the model prediction.
A Wilcoxon test was applied, p-value = 43%

Conclusion:
All of the (ex-post) variation in countries performance can be
explained by the simple model.
Figure 14: Distribution of Countries’ ex-post equity premia (broken line).
Average close to zero. Distribution predicted by model drawn in black.
Statistically indistinguishable.
Results – Market Capitalizations
The model predicts a log-normal distributed with
2 2
T σ 2 T σX2
eσ T − 1
 

µX = −log(N ) − + , σX = log  + 1
2 2 N

US: The expected maximum market capitalization with 40 mar-


kets is 29%. The current US market capitalization is 30%.
Global: Distribution of predicted capitalizations matches em-
pirical distribution. Wilcoxon test: p = 18.4%

Conclusion:
Variation in market cap also explained by the simple model.
Figure 15: Distribution of capitalization paths conditioned on ex-post
50

4
40

3
Density

Density
30

2
20

1
10
0

0
0.00 0.10 0.20 0.30 0.0 0.2 0.4 0.6 0.8

Percentage Market Cap Maximum Percentage Market Cap

Figure 16: Distribution of predicted market capitalizations (left) and actual


capitalizations. Distribution of maximum predicted capitalization and the
current US value.
7 The Ex-Ante Equity Premium
Historical US returns over-estimate equity premium due to suc-
cess/selection bias. Should expect lower realized returns than
history (“regress to the mean”, Galton).

The actual ex-ante premium? US and global returns are


consistent with very low premium of 0. Pessimistic?

Best prediction: use consumption asset pricing result

σ(mt )
E(Rt − Rf ) = −ρ(rt , mt )σ(Rt ) ≈ 1%
E(mt )
8 Future Equity Market Performance
Even if ex-ante premium is 0%, we still expect some countries
to have much higher ex-post realizations.

Unlikely that the US will remain 21st century winner. In the


simple model, has 1/N chance, or ∼2%.

The more volatile the country, the larger are ex-post outcomes.
Therefore, riskiest markets have the highest chance being 21st
century winner.
Candidates: Brazil, Russia, India and China?
9 Conclusion
Have tested the hypothesis: US ex-post premium due to luck.
US 5-7% historical premium consistent with self-selection bias.
Use the simplest possible model, ex-ante identical countries.

Simple model explains quite a lot:


US ex-post premium and Sharpe ratio
Distribution of ex-post premia and SR for 40 countries
Distribution of market capitalizations across 40 countries.

Provides the simplest possible solution to the equity premium.


Thank You! Now time for Questions.

Figure 17: A Brownian Motion. If only financial markets behaved this nicely.
Figure 18: US Real S&P500 Returns, 1930 to 2009
Figure 19: Publications on US Equity Premium, 50 yr Rolling Window

You might also like