1

November 2011

Jon Wongswan

Phatra Securities

jon@phatrasecurities.com

What Influences U.S. International Equity Investment:

Equity or Currency Returns?

Stephanie Curcuru, Charles Thomas,

Frank Warnock, and Jon Wongswan

November 2011

Thammasat Finance Conference

2

November 2011 Outline

I. Overview

II. Data Description

III. Empirical Results

IV. Conclusions

3

November 2011 I. Overview

• This paper examines the relationship between U.S. investors’ international equity portfolio rebalancing (country level) and local-currency equity and currency returns.

Equity Return (USD) = Equity Return (Local Currency) + Currency Return

Portfolio Rebalancing

at time t

Equity Return (Local

Currency) at time …, t-1, t,

t+1, …

Currency Return

at time …, t-1, t, t+1, …

4

November 2011 I. Overview (cont.)

• Earlier work by Curcuru, Thomas, Warnock, and Wongswan (2011 AER)—CTWW1—examines the portfolio rebalancing behavior of U.S. investors’ international equity portfolio in U.S. dollar term.

Portfolio Rebalancing

at time t

U.S. Dollar Equity Return

at time …, t-1,

t, t+1, …

U.S. investors do not exhibit return chasing behavior but tend to sell past winners.

U.S. investors rebalance portfolio into markets that subsequently have abnormal returns.

Contributions: Use portfolio holdings data and find results opposite to existing studies

5

November 2011 I. Overview (cont.)

• Research question for this paper: Are the documented effects in CTWW1 coming from local-currency equity or/and currency returns?

• Findings of this paper: We find that the rebalancing of U.S. investors’ international equity portfolio is not related to past or future currency returns.

6

November 2011

Motivation ...From the Practical to the Theoretical

1989

1991

1993

1995

1997

1999

2001

2003

2005

2007

2009

0%

5%

10%

15%

20%

25%

30%

Cross-border equity claims / World market capitalization

1989

1991

1993

1995

1997

1999

2001

2003

2005

2007

2009

0%

5%

10%

15%

20%

25%

30%

35%

40%

45%

50%

U.S. cross-border equity claims / Total cross-border equity

claims

Cross-border equity investment has become more important.

U.S. investors are a major participant.

7

November 2011 Motivation (cont.)

• Empirical results have policy implications.

• Existing theoretical models imply different relationships between currency returns and international portfolio rebalancing. Adler and Dumas ( 1983 JF): No relationship (Hedge through other asset

classes)

Hau and Ray (2006 RFS): Negative relationship (Reduction in currency exposure)

Burnside, Eichenbuam, and Rebelo (2010 JEEA): Positive relationship (Carry trade)

8

November 2011 II. Data Description

1. Portfolio Holdings

• U.S. investors’ country-level foreign equity holdings in 43 countries from January 1990 through December 2008 (monthly frequency).

• Portfolio holdings can not be directly infer from flows data because of the “Financial Center Bias” in the capital flows data.

To correct for the Financial Center Bias, we use:

• Infrequent comprehensive (security-level) surveys of cross-border holdings as fixed points.

• To interpolate positions between survey dates, we use monthly capital flows data (TIC) and country price returns.

9

November 2011 Estimating Portfolio Holdings Data

ttttt gsgprnhnh )1(1

TTT nhbhgap

)()1(1 tTttttt adjgapgsgprhh

• Form naïve holdings estimates

• Doing so will result in a “gap” at time T of a benchmark

• Solve for an adjustment factor (adj) such that at T estimated holdings (h) = benchmark holdings (bh)

• This methodology is originally implemented in Thomas, Warnock, and Wongswan (2004 IFDP) and improved upon and updated by Bertaut and Tryon (2007 IFDP).

10

November 2011

U.S. Holdings of Foreign Equities: U.K.

0

100

200

300

400

De

c-7

6

De

c-7

8

De

c-8

0

De

c-8

2

De

c-8

4

De

c-8

6

De

c-8

8

De

c-9

0

De

c-9

2

De

c-9

4

De

c-9

6

De

c-9

8

De

c-0

0

De

c-0

2

Bill

ion

s o

f d

olla

rs

Survey Date Benchmark-consistent Naive

Example: U.S. Holdings of U.K. Equities

Security-Level Benchmark Survey

11

November 2011 II. Data Description (cont.)

2. Country Returns

We use MSCI country-level total return index for each country.

• Assume that U.S. investors’ holdings composition within each country is similar to that of MSCI country index.

At the end of 1997, the cross-sectional correlation between firm-weights in the MSCI world (excluding the U.S.) and firm-weights in U.S. investors’ international portfolios is 0.77.

• Adjust for the share outstanding availability for the public (float adjustment).

• Adjust for foreign ownership restrictions.

3. Currency Returns

We use monthly exchange rate against USD data from the Federal Reserve Board.

12

November 2011 III. Empirical Results

We apply well-established techniques from the finance literature to characterize the relationship between portfolio reallocations and equity and currency returns.

1. Portfolio reallocations and past returns

Lag Momentum (LM): Grinblatt, Titman, and Wermers (1995 AER)

2. Portfolio reallocations and future returns

Conditional Weight Measure (CWM): Eckbo and Smith (1998 JF) and Ferson and Khang (2002 JFE)

13

November 2011

1. Portfolio Reallocations and Past Returns

• Goal: Measure the relationship between portfolio reallocations and past returns.

• Momentum behavior: Actively reallocate more (less) investment into the country that has relatively higher (lower) past return.

• Contrarian behavior: Actively reallocate more (less) investment into the country that has relatively lower (higher) past return.

• We use Lag Momentum (LM) measure of Grinblatt, Titman, and Wermers (1995 AER) to measure this relationship.

14

November 2011

1. Portfolio Reallocations and Past Returns (cont.)

Portfolio reallocation

• Definition: Active changes in portfolio holdings net of valuation effect (Xi,t)

• Calculation:

• Xi,t equals to zero for a buy-and-hold portfolio.

tp

titititi r

rwwX

,

,1,,, 1

1

Implied New Weight due to Valuation EffectActual New Weight

15

November 2011

Example of Active Weight Change (Xi,t) for a Buy-and-Hold Portfolio

A buy-and-hold portfolio weights in period 2 are WA,2 = 48% and WB,2 = 52%,

XA,2 = 0.48 – 0.50*[(1+0.2)/(1+0.25)] = 0.48 – 0.48 = 0

XB,2 = 0.52 – 0.50*[(1+0.3)/(1+0.25)] = 0.52 – 0.52 = 0

Period 2

Country Return (1->2) Holdings Value Weight

A 20% $200*(1+0.2) = $240 240/ (240+260) = 48%

B 30% $200*(1+0.3) = $260 260/ (240+260) = 52%

Period 1

Country Holdings Value Weight

A $200 50%

B $200 50%

Portfolio return = (0.5 * 0.2) + (0.5 * 0.3) = 25%

16

November 2011

1. Portfolio Reallocations and Past Returns (cont.)

• LM is based on the covariance between Xit and relative returns in country i at lag k:

• LM > 0 Momentum trading behavior

• LM < 0 Contrarian trading behavior

• LM can also be computed for buy or sell only, limiting observations to only times in which Xit is positive (BM) or negative (SM).

T

t

N

iktpktitik

T

rrXT

LM1 1

,,, )(1 Relative Past

Return

17

November 2011

Portfolio Reallocations and Past Returns

U.S. Dollar Equity ReturnsLocal-Currency Equity Returns

Currency Returns

Lag 1 Lag 2 Lag 3 Lag 1 Lag 2 Lag 3 Lag 1 Lag 2 Lag 3

LM (Buy and Sell)

0.005 (0.144)

-0.169 (0.152)

-0.179(0.146)

0.146 0.132)

-0.059(0.135)

-0.135(0.145)

-0.122(0.064)

-0.082(0.071)

-0.027(0.069)

BM (Buy Only)

0.159(0.123)

0.064(0.114)

0.036(0.103)

0.249* (0.107)

0.147(0.099)

0.082(0.107)

-0.084(0.052)

-0.067(0.051)

-0.039(0.051)

SM (Sell Only)

-0.154*(0.056)

-0.232*(0.065)

-0.215*(0.065)

-0.103* (0.055)

-0.207*(0.057)

-0.217*(0.060)

-0.039(0.027)

-0.016(0.035)

0.012(0.029)

Newey and West (1987) standard errors are in parentheses.* Statistically significant at the 5 percent level.

• U.S. investors can be characterized as selling past winners.

• No evidence of relationship between portfolio reallocations and currency returns.

18

November 2011

2. Portfolio Reallocations and Future Returns

• Goal: Measure portfolio reallocations and future abnormal returns.

• Good portfolio reallocation: Actively rebalance the portfolio into (out of) a country that subsequently earns positive (negative) abnormal return.

• Abnormal return: Return component that cannot be predicted from using public information.

• We use the Conditional Weight Measure (CWM) of Eckbo and Smith (1998 JF) and Ferson and Khang (2002 JFE) to measure this relationship.

19

November 2011

2. Portfolio Reallocation and Future Returns (cont.)

• The measure is based on the conditional covariances between changes in portfolio weights and future abnormal returns.

where the buy-and-hold benchmark holding is:

• CWM > 0 Good country reallocation ability

TN

itttiti

btitik rErwwECWM

11,1,,, ||

t

ktp

i

r

rkti

bti ww 1

,

,

1

1,,

Abnormal ReturnActive Reallocation

20

November 2011

Portfolio Reallocations and Future Returns

k=1 k=2 k=3

U.S. Dollar Equity Returns 0.369*(0.147)

0.649*(0.211)

0.735*(0.274)

Local Currency Equity Returns

0.037(0.065)

0.037(0.201)

0.037(0.276)

Currency Returns 1.122*(0.382)

1.122*(0.575)

1.122*(0.747)

Newey and West (1987) standard errors are in parentheses.* Statistically significant at the 5 percent level.

• U.S. investors switch into markets that subsequently have abnormal returns.

• No evidence of U.S. investors switching into currencies that subsequently have abnormal returns.

21

November 2011 IV. Conclusions

• Using portfolio holdings data, we do not find significant evidence of the relationship between portfolio reallocations and currency returns (past or future).

• Academic perspective: Results are consistent with existing studies on return predictabilities- Well-documented predictabilities in country-level equity returns (Harvey JF

1991; Ferson and Harvey JF 1993; Ang and Bakeart RFS 2007)

- Limited predictabilities in currency returns (Meese and Rogoff JIE 1983)

• Practitioner perspective: Majority of international equity funds do not make investment decisions based on views on currency movements because- Low ability to predict future currency movements

- Hard to know each firm’s total exposure to currency

- Hard to know each firm’s currency hedging positions and policy

22

November 2011 IV. Conclusions

• The findings do not imply that investors are not concerned about currency movements because- Investors may have hedged the exposure in other asset classes

- Investors may want to take on an exposure as a source of portfolio diversification

• To fully understand the role of currency returns, we may need to examine investors’ overall portfolio across different asset classes.