Why Do Individuals Exhibit Investment Biases Henrik Cronqvist

Why Do Individuals Exhibit Investment Biases? Henrik Cronqvist Claremont Mc. Kenna College Stephan Siegel University of Washington National Taiwan University International Conference on Finance December 6, 2012

Genes and Household Finance Research Agenda So Far More risk Participate in stock market Save Less risk Don’t participate in stock market Consume Today: Investment “biases” Investment “styles” – value vs. growth We use twin study research design to examine, for each of the above choices, the effects of i) genetic and environmental influences, and ii) Gx. E moderators. Parallel to molecular genetics and neuroscientific studies by Camerer, Bossaerts, Kuhnen, Laibson, Zak, and others.

Investment “Biases” Long list of investment behaviors that cannot be explained by standard preferences and belief formation: Underdiversify Prefer local securities – home bias Avoid realizing losses Trade a lot Chase past performance Prefer lottery-type stocks Some of these behaviors have been shown to be: Wide-spread; present also among professional investors Related to fundamental psychological mechanisms Costly But, degrees of behavior vary across investors

Biological Basis for Investment Behaviors

Existing Evidence Experimental evidence (previous slide) Capuchin monkeys exhibit loss aversion Capuchin monkeys prefer gambles with good outcomes framed as bonuses over identical pay-off gambles with bad outcomes framed as losses ¨ Loss aversion is part of “decision-making process that evolved before humans and capuchins separated” (Chen et al. (2006), Lakshminarayanan et al. (2011)) ¨ But: No empirical evidence on the genetics of investment “biases” based on real world financial decisions and data

Research Methodology: Twin Researc Identical Twins Josefin and Elin Nordegren Fraternal Twins The Hodgson Twins

Intuition of Methodology Use identical & fraternal twins to decompose the variation in investment behaviors: ¨ Identical (monozygotic, MZ) twins share 100% of their DNA ¨ Fraternal (dizygotic, DZ) twins share on average 50% of their DNA ¨ Twins who grew up in same family share a common environment ¨ Each twin has his/her individual (non-shared) environment If genes matter, then identical twins should be more similar than fraternal twins in terms of, e. g. , their investment behaviors. 7

Methodology Random effect model with genetic effect a, common effect c and individual-specific effect e: Covariance structure implied by genetic theory: MZ DZ

Methodology, cont’d Estimate parameters σ2 a, σ2 c, and σ2 e via maximum likelihood estimation (MLE) with bootstrapped standard errors Derive the variance components: A-share – genetic component: C-share – common environment (parenting): E-share – individual environment & measurement error:

Data Swedish Twin Registry Matched with annual financial data (including holdings of assets and sales transactions) and socioeconomic data from Statistics Sweden (1999 – 2007) Filters: At least 18 years old ¨ Both twins hold some equities (directly or indirectly) in one year ¨ Average all variables over the years that individual is in sample ¨

Measuring Investment “Biases” Diversification Number of stocks in portfolio Home Bias Proportion of equity portfolio in local (=Swedish) equities Turnover Annual sales volume scaled by value of portfolio at beginning of year Disposition Effect Proportion Gains Realized (PGR) – Proportion Losses Realized (PLR) (Odean (1998)) Performance Chasing Proportion of equities acquired with raw returns in top two deciles Skewness Preference Proportion of “lottery” securities (Kumar (2009))

Evidence from Correlations

Variance Decomposition

Variance Decomposition DETAIL

Robustness Opposite-sex twins Model misspecification Allowing for negative variance components Communication Identical twins communicate more with one another Financial decisions are influenced by communication (e. g. Shiller and Pound (1998), Hong, Kubik, and Stein (2004)) Sort pairs into 10 communication intensity bins and randomly drop identical/fraternal pairs until both types are equally often present per bin. Estimate model across all 10 bins. “A” component is somewhat reduced, but overall, results are robust. Equal environments assumption

Two Additional Results Moderators of genetic investment biases Behavioral consistency: Investment biases and behaviors in other, non-investment, domains

Moderating Genetic Effects Environment can enhance or reduce the effects of genetic predisposition ¨ Example: Education seems to reduce genetic variation in health outcomes (Johnson et al. (2009)) Examine how years of education interact with the genetic effects No significant evidence that years of education reduces genetic predispositions to investment biases

Moderator: Years of Education

Financial Experience Does work experience in the financial industry or, e. g. , in a corporate treasury department reduce genetic predispositions to investment biases? Repeat analysis for individuals working with finance: “A” components reduced (generally by >50%). Experience in the finance industry seems to reduce the genetic predisposition to investment biases.

Behavioral Consistency Behavior across different domains is often consistent If genetic factors matter, source of consistency should be genetic è Correlate Home Bias with Distance to birthplace ¨ Indicator whether spouse is from same home region ¨

Behavioral Consistency

Conclusions A long list of investment biases are “human” in the sense that investors are born with pre-dispositions 25 -50% of variation explained by genetic variation Provide empirical support for evolutionary models of investment biases (e. g. , Brennan and Lo (2009)) Education does not reduce the genetic predisposition to investment biases. But finance industry experience reduces genetic effects. Genetic factors influencing investment biases affect behaviors in other, non-investment, domains.