Solutions to the Asset Allocation Problem by Informed Respondents: The Significance of the Size‐of‐Bet and the 1/N Heuristic

• DOI: http://doi.org/10.1111/j.1540-6296.2009.01166.x
• Published date: 01 September 2009
• Date: 01 September 2009

C

Risk Management and Insurance Review, 2009, Vol.12, No. 2, 251-271

SOLUTIONS TO THE ASSET ALLOCATION PROBLEM

BY INFORMED RESPONDENTS:THE SIGNIFICANCE

OF THE SIZE-OF-BET AND THE 1/

N

HEURISTIC

Gordon L. Clark

Emiko Caerlewy-Smith

John C. Marshall

ABSTRACT

Asset allocation is a classic topic in the theory of finance and a crucial issue

for investment policy. Noted for its significance in driving pension fund per-

formance, it is also an issue that individual investors consider when designing

their investment portfolios. In theory, Markowitz and those following in his

wake have an optimal solution. In practice, however, we show that when asked

to allocate their own money to a set of asset classes (from relatively low risk

to high risk) in an experimental situation, most of our informed respondents

would vary their investment strategies according to the size-of-bet (the money

value of assets to be invested). Wealso show that most participants in the study

adopted one of three solutions to the posed problem only one of which could be

thought consistent with Benartzi and Thaler’s 1/nheuristic. Since respondent

Gordon L. Clark is with the Centre for the Environment, Oxford University, South Parks Road,

Oxford OX1304, United Kingdom; phone: +44-1865-285197; fax: +44-1865-285073; e-mail: gor-

don.clark@ouce.ox.ac.uk. Emiko Caerlewy-Smith is with the Centre for the Environment, Oxford

University, South Parks Road, Oxford OX1 3QY, United Kingdom; phone: +44-(0)20-7213-8950;

fax: +44-(0)20-7212-4418; e-mail: emiko.caerlewy-smith@ouce.ox.ac.uk. John C. Marshall (de-

ceased) was with the Department of Clinical Neurology,John Radcliffe Hospital, Oxford Univer-

sity, Oxford OX3 9DU, United Kingdom. Support for this article was provided by the National

Association of Pension Funds (NAPF). The authors thank Christine Farnish, David Gould, and

Geoff Lindey for their interest in the project. Helpful comments on the projectand related papers

have been made by Keith Ambachtsheer,John Evans, Simon Ford, Michael Orszag, Roger Urwin,

members of the NAPF Benefits and Investments Councils, participants at the NAPF Investment

Conference, the FT-NAPF Trustee Conference, the annual pensions conference at the University

of New South Wales (Sydney), and the ICPM Rotman School conference at the University of

Toronto.Our thanks go to Amy Dickman who provided assistance in the initial coding and anal-

ysis of the data and to Janelle Knox, Esther Lim, Kendra Strauss, and two anonymous referees

who commented on previous versions of the article. The first-named author benefited from con-

versations on the topic with Ashby Monk and Richard Zeckhauser and the exchange of research

with Anup Basu, Michael Drew,and Gur Huberman. The results and interpretations reported are

the sole responsibility of the authors; none of the above should be held to account for any errors,

omissions, or opinions expressed herein. This article was subject to double-blind peer review.

251

252 RISK MANAGEMENT AND INSURANCE REVIEW

solutions do not seem to be explained by formal education, professional

qualifications, or training, it is suggested that solutions to the asset allocation

problem are a product of strategies that mix intuitive responses to the initial

tranche of money with theoretical cum practical shared conventions. Solutions

to the asset allocation puzzle suggest that the size-of-bet could be a significant

consideration for many informed investors. In conclusion, suggestions are made

about taking forward closer scrutiny of these experimental results.

INTRODUCTION

Asset allocation is a classic topic in the theory of finance (Markowitz, 1952). It is also a

crucial strategic issue when pension funds and individuals plan their long-term invest-

ments. It is widely believed that asset allocation is the single most important decision

an investor can make when planning his or her long-term retirement income. Moreover,

the available evidence in the United Kingdom suggests that “long-run asset allocations,

however modelled, account for the bulk of the time-series variation in returns” (Blake

et al., 1999, p. 459).1

If left to individuals, however, it appears that asset allocation is subject to a variety

of behavioral and cognitive effects. Samuelson and Zeckhauser (1988) noted that new

TIAA-CREF participants tended to allocate 401(k) pension assets on a 50/50 (bonds

and equities) basis with little change thereafter despite the long-term accumulation of

financial assets and changing age and status.2While the simplest of lifestyle and life-cycle

models of asset allocation suggest that participants should revisit their asset allocations

on a regular basis as they approach retirement, many participants do not. Information

awareness, prototypical decision trees, and devices designed to enhance individual

decision making appear to have only modest dampening effects on these shortcomings.

However, Benartzi and Thaler (2001) suggested that well-designed default settings in

pension plans can be effective in mitigating these problems.

There is a vibrant market for information and advice on asset allocation strategies

(see Canto, 2006). Similarly, pension fund consultants have developed techniques and

methods of analysis to cue decision making on this issue; risk budgeting, asset-liability

models, and performance metrics provide formal protocols that integrate asset allo-

cation with long-term decision making (Bauer et al., 2006). Even so, the evidence on

financial decision-making competence and consistency is not altogether encouraging.

For example, we have suggested that many U.K. pension fund trustees lack the funda-

mental knowledge and skills to be expert decision makers (Clark et al., 2006). Judgment

in these situations is highly dependent upon the level of formal education, professional

qualifications, and role-specific training (Clark et al., 2007).

In this article, we report results from our project on U.K. pension trustee decision mak-

ing using the experimental methods that are now familiar when testing (as opposed to

1Blake et al. (1999) believe that this reflects the management and organization of the U.K. invest-

ment industry (an issue of internal and external governance) whereas others, notably Campbell

and Viceira (2006), have a more theoretical argument based on modern portfolio theory appli-

cable to both defined benefit and defined contribution plans.

2TIAA-CREF is a private pension fund operating on behalf of many U.S. universities and foun-

dation employers. It offers participants defined contribution pension savings products with

IRS tax-preferred status (i.e., 401(k) eligible). It is one of the world’s largest private pension

providers.