Tim Alexander Kroencke Understanding and Harvesting Expected Returns of Asset Classes ...

244 Pages·2013·1.42 MB·English

Tim Alexander Kroencke
Understanding and Harvesting Expected Returns
of Asset Classes, Investment Styles, and Risk Factors
Inauguraldissertation
zur Erlangung des akademischen Grades
eines Doktors der Wirtschaftswissenschaften
der Universita¨t Mannheim
Vorgelegt im Fr¨uhjahrssemester 2013
Dekan: Dr. Ju¨rgen M. Schneider
Berichterstatter: Prof. Dr. Erik Theissen
Prof. Dr. Stefan Ruenzi
Tag der mu¨ndlichen Pr¨ufung: 28. Mai 2013
Understanding and Harvesting Expected Returns
of Asset Classes, Investment Styles, and Risk Factors
Thesis Advisor: Professor Dr. Erik Theissen Author: Tim Alexander Kroencke
Introduction
The Tale of Ketchup Economics. P = E [M X ], or the price of an asset today should
t t t+1 t+1
be equal to the future expected discounted value of its payoffs given the marginal utility of the
investor as the discount factor. This simple concept is central in asset pricing. However, there
are two polar approaches of asset pricing. Both can be clarified using Larry Summers’ famous
tale of“ketchup economics”, summarized as follows (Summers (1985)):
There are two groups of economists studying the ketchup market. The first group is
calledgeneraleconomistsandthesecondgroupiscalledfinancialeconomists. General
ketchup economists try to understand the prices of bottles of ketchup by looking
at the market of tomatoes. They try to understand the fundamental sources of
macroeconomic risk to determine the prices of bottles of ketchup. They are addicted
to deep models with utility and production functions.
Financial ketchup economists try to understand the prices of bottles of ketchup by
looking at the ketchup market more deeply. They try to understand prices of bottles
of ketchup relative to the prices of other bottles of ketchup. They love to exploit
powerful arbitrage concepts.
In this thesis, I study investment returns using both perspectives. I deal with the“financial
ketchup economics”perspective in chapter 3 and chapter 4. The central asset pricing equation is
1
tested for a set of interesting test assets, whereby I approximate investor’s marginal utility with
the returns of some benchmark assets. This approach allows me to quantify if the test assets
provide economically sizeable and significant returns which are not feasible with the benchmark
assets (i.e. diversification benefits) and to learn how investors can efficiently harvest them.
Importantly, the bottom line of Summers (1985) is that there is insufficient research in the
interstices between the financial and the economics approaches:
“It is unfortunate [...] that researchers in economics pay so little attention to finance
research, andperhapsmoreunfortunatethatfinancialeconomistsremainsoreluctant
to accept any research relating asset prices and fundamental values.”
Thus, after studying how to harvest returns, I turn to understanding returns. In chapter 1
and chapter 2, I investigate which macroeconomic sources of risk can explain average investment
returns.1 In particular, these chapters are devoted to the construction and measurement of the
fundamental risk factors driving expected returns of asset classes as well as popular investment
styles. To this end, I follow Larry Summers’ suggestion and make use of concepts and methods
from both approaches of“ketchup economics”.
Harvesting Returns. I analyze how investors can gain some extra return from their interna-
tional investment portfolio. International diversification is intended to improve the return per
unit of risk, in the hope that returns across countries are not perfectly correlated (e.g. Grubel
(1968) and Solnik (1974)).
The chapter“International Diversification with Securitized Real Estate and the Veiling Glare
from Currency Risk”showsthattheriskandreturncharacteristicsofglobalstocksandbondsare
different from those of global real estate. Due to its local nature, real estate is subject to rather
local factors and less so to global factors (Eichholtz (1996)). Thus, diversification benefits from
international (listed) real estate are potentially larger than the benefits from common stocks.
1I present the papers of my thesis in reverse chronological order.
2
After accounting for systematic FX risk (Lustig and Verdelhan (2007), Verdelhan (2011)), which
adds inter-asset and inter-country correlation, the paper finds evidence for this hypothesis.
The recent literature has documented that particularly style-based investment strategies in
FX markets provide attractive returns (e.g. Lustig and Verdelhan (2007), Pojarliev and Levich
(2008), Menkhoff, Sarno, Schmeling, and Schrimpf (2012b), Asness, Moskowitz, and Pedersen
(2012)). Thus, the FX component may not only add risk to a portfolio but can also be a source
of additional diversification benefits.
Thechapter“International Diversification Benefits with Foreign Exchange Investment Styles”
studies the currency component of international stock and bond portfolios. While style-based
investmentsandtheirroleforportfolioallocationinequitymarketshavebeenwidelystudied(e.g.
Eun, Huang, and Lai (2008), or Eun, Lai, de Roon, and Zhang (2010)), there is considerably less
knowledgeabouttheportfolioimplicationsofstyleinvestinginFXmarkets. Thepaperquantifies
economically large and significant diversification benefits from simple FX investment styles like
carry, momentum and value. Almost all previous research focuses on currency “hedging” in a
portfolio context (e.g. Glen and Jorion (1993), Campbell, de Medeiros, and Viceira (2010)). In
contrast, this chapter documents that style-based currency“speculation”can be quite attractive
in a global portfolio.
Understanding Returns. The remaining two chapters are located in the interstices between
the two fields of“ketchup economics”and study the relationship between the real economy and
investment returns. Before Summers (1985) as well as in the subsequent years, this kind of
literature proceeded rather slowly with classic contributions by Chen, Roll, and Ross (1986),
Breeden, Gibbons, and Litzenberger (1989), and Campbell (1996). The main issue is that the
correlation between standard macroeconomic variables and financial returns is low, and thus,
the returns of risky asset classes like equity (Mehra and Prescott (1985)), or specific investment
styles like value and momentum (Fama and French (1993), Jegadeesh and Titman (1993)), seem
to be too large to be justifiable.
3
More recently, starting with the new millennium, more promising attempts have been made
by introducing novel macroeconomic variables as fundamental risk factors which are more suc-
cessful (to name a few contributions, Lettau and Ludvigson (2001), Vassalou (2003), Parker and
Julliard (2005), Petkova (2006), Jagannathan and Wang (2007), Savov (2011), Koijen, Lustig,
and Nieuwerburgh (2012)). Indeed, today there is a “zoo” of fundamental factors which suc-
cessfully explain several kinds of investment returns (Cochrane (2008)). As a consequence, it
is quite difficult to assess how all these variables relate to each other. More specifically, many
of these fundamental factors produce mixed results in different empirical settings, or differently
motivated factors lead to very similar results. Thus, with these two chapters of my thesis at
hand, I aim to analyze in a systematic manner properties, construction and measurement of
several newly introduced fundamental risk factors to gain novel insights.
The chapter “GDP Mimicking Portfolios and the Cross-Section of Stock Returns” studies
popularGDP-basedmacroeconomicfactorsinaunifiedframework. AggregateGDPasameasure
of fundamental risk is a popular risk factor in asset pricing. From an empirical perspective,
aggregate GDP does not correlate well with stock returns and is not useful for explaining average
returns. However, there is a stylized fact in macroeconomics, which has been so far ignored in
finance: some components of GDP lead the aggregate, while some components of GDP lag the
aggregate (e.g. Greenwood and Hercowitz (1991), Gomme, Kydland, and Rupert (2001), Davis
and Heathcote (2005), Fisher (2007), Leamer (2007)). The paper documents that leading GDP
components can explain the size premium and the value premium quite well. The opposite is
documented for the momentum premium. The lagging GDP components explain the return of
momentum portfolios very well. A three-factor model with the market excess return, one leading
and one lagging GDP component compares very favorably with the Carhart four-factor model
in jointly explaining a large cross-section of size, book-to-market, momentum, and industry
portfolio returns.
Finally, the chapter “Asset Pricing without Garbage”, which also serves as my Job Market
Paper, looks more deeply at consumption as a fundamental risk factor. The key contribution
4
is that the paper exploits insights from state space analysis to provide an explanation for the
bad performance of traditional consumption measures (Mehra and Prescott (1985), Breeden,
Gibbons, and Litzenberger (1989)) and the good performance of recently proposed alternative
consumption measures in the classical consumption-based asset pricing model (Parker and Jul-
liard(2005), JagannathanandWang(2007), Savov(2011)). Morespecifically, thepaperprovides
an explanation of why garbage as a measure of consumption implies a several times lower coef-
ficient of relative risk aversion in the consumption-based asset pricing model than consumption
based on the official National Income and Product Accounts (NIPA). Unlike garbage, NIPA
consumption is filtered to mitigate measurement error. The paper applies a structural model of
the filtering process, which allows to revoke the filter inherent in NIPA consumption. “Unfiltered
NIPA consumption”performs as well as garbage in explaining the equity premium and risk-free
rate puzzle. Furthermore, the paper documents that two other popular NIPA-based measures,
three-year and fourth-quarter NIPA consumption, are related to unfiltered NIPA consumption.
Both can be viewed as ad hoc unfilter rules.
5
Contents
Asset Pricing without Garbage
1 16
1.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 16
1.2 Measurement Error, Filtering,
and Time Aggregation in Consumption Data . . . . . . . . . . . . . . . . . . . . 21
1.3 Stylized Model . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 22
1.4 Accounting for Time Aggregation . . . . . . . . . . . . . . . . . . . . . . . . . . 24
1.4.1 Simulation Experiment . . . . . . . . . . . . . . . . . . . . . . . . . . . . 25
1.5 Asset Pricing without Garbage . . . . . . . . . . . . . . . . . . . . . . . . . . . 31
1.5.1 Comparison of Alternative Consumption Measures . . . . . . . . . . . . . 31
1.5.2 Equity Premium . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 35
1.5.3 Cross-Section of Stock Returns . . . . . . . . . . . . . . . . . . . . . . . 42
1.5.4 The Best of two Measures: A Combined Approach . . . . . . . . . . . . 45
1.6 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 50
GDP Mimicking Portfolios and
2
the Cross-Section of Stock Returns
52
2.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 52
2.2 Stylized Facts:
Lead and Lag in GDP Components . . . . . . . . . . . . . . . . . . . . . . . . . 55
2.3 GDP-Mimicking Portfolios . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 59
2.4 Model, Estimation, and Data . . . . . . . . . . . . . . . . . . . . . . . . . . . . 65
2.5 Asset Pricing Tests . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 68
2.5.1 Dissecting GDP and the Cross-Section of Stock Returns . . . . . . . . . 68
2.5.2 A GDP-Based Three-Factor Model . . . . . . . . . . . . . . . . . . . . . 73
2.6 Robustness . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 85
6
2.7 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 87
International Diversification Benefits
3
with Foreign Exchange Investment Styles
88
3.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 88
3.2 Related Literature . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 92
3.3 FX Investment Styles . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 93
3.4 Data and Portfolio Construction . . . . . . . . . . . . . . . . . . . . . . . . . . . 97
3.5 Empirical Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 101
3.5.1 Risk and Return Characteristics . . . . . . . . . . . . . . . . . . . . . . . 101
3.5.2 Mean-Variance Efficiency Tests . . . . . . . . . . . . . . . . . . . . . . . 105
3.5.3 Out-of-Sample Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . 111
3.5.4 Summary of Main Results . . . . . . . . . . . . . . . . . . . . . . . . . . 119
3.6 Further Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 121
3.6.1 Value and Momentum Everywhere . . . . . . . . . . . . . . . . . . . . . 121
3.6.2 Accounting for Skewed Returns . . . . . . . . . . . . . . . . . . . . . . . 124
3.6.3 Market States . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 131
3.7 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 135
Securitized Real Estate and
4
the Veiling Glare from Currency Risk
137
4.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 137
4.2 Literature . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 140
4.3 Methodology . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 143
4.3.1 Spanning Tests . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 143
4.3.2 Currency Risk and Hedging . . . . . . . . . . . . . . . . . . . . . . . . . 144
4.4 Data and Descriptive Statistics . . . . . . . . . . . . . . . . . . . . . . . . . . . 147
4.5 Empirical Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 150
7
4.6 Further Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 155
4.7 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 162
A Asset Pricing without Garbage 180
A.1 Simulation: Details on the Asset Pricing Economy . . . . . . . . . . . . . . . . . 180
A.2 Empirical Data . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 182
B GDP Mimicking Portfolios and
the Cross-Section of Stock Returns 184
B.1 Estimation without Intercept . . . . . . . . . . . . . . . . . . . . . . . . . . . . 184
B.2 Growth Rates of Components . . . . . . . . . . . . . . . . . . . . . . . . . . . . 184
B.3 Additional Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 184
C Foreign Exchange Investment Styles 204
C.1 Data Archive . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 204
C.2 Dissecting Currency Hedging and Speculation . . . . . . . . . . . . . . . . . . . 217
C.3 Stochastic Dominance Tests . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 218
C.4 Additional Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 221
8