Formative Experiences and Portfolio Choice: Evidence from the Finnish Great Depression

• DOI: http://doi.org/10.1111/jofi.12469
• Published date: 01 February 2017
• Author: SAMULI KNÜPFER,ELIAS RANTAPUSKA,MATTI SARVIMÄKI
• Date: 01 February 2017

THE JOURNAL OF FINANCE •VOL. LXXII, NO. 1 •FEBRUARY 2017

Formative Experiences and Portfolio Choice:

Evidence from the Finnish Great Depression

SAMULI KN ¨

UPFER, ELIAS RANTAPUSKA, and MATTI SARVIM ¨

AKI∗

ABSTRACT

We trace the impact of formative experiences on portfolio choice. Plausibly exogenous

variation in workers’ exposure to a depression allows us to identify the effects and a

new estimation approach makes addressing wealth and income effects possible. We

find that adversely affected workers are less likely to invest in risky assets. This result

is robust to a number of control variables and it holds for individuals whose income,

employment, and wealth were unaffected. The effects travel through social networks:

individuals whose neighbors and family members experienced adverse circumstances

also avoid risky investments.

THE LARGE DEGREE OF HETEROGENEITY in how individuals construct financial

portfolios poses a challenge for theory and empirical work. Studies of twins

deepen the puzzle by showing that genetically inherited traits and family back-

ground cannot account for the variation in portfolio choice. Likewise, observ-

able characteristics over and above genetic makeup and family environment do

∗Samuli Kn¨

upfer is with BI Norwegian Business School and CEPR. Elias Rantapuska is with

Aalto University School of Business. Matti Sarvim¨

aki is with Aalto University School of Busi-

ness and VATT Institute for Economic Research. We thank Statistics Finland and the Finnish

Tax Administration for providing us with the data. We are also grateful to Ashwini Agrawal,

Shlomo Benartzi, Janis Berzins, David Cesarini, Joao Cocco, Henrik Cronqvist, Francisco Gomes,

Harrison Hong, Kristiina Huttunen, Tullio Jappelli, Lena Jaroszek, Ron Kaniel, Markku Kaustia,

Hyunseob Kim, Juhani Linnainmaa, Ulrike Malmendier, Stefan Nagel, Stijn Van Nieuwerburgh,

Alessandro Previtero, Miikka Rokkanen, Clemens Sialm, Ken Singleton (Editor), Paolo Sodini,

Robin Stitzing, Johan Walden, an Associate Editor, and two referees for insights that benefited

this paper. Conference and seminar participants at the 4 Nations Cup 2013, American Economic

Association Meetings 2014, China International Conference in Finance 2013, Conference of the

European Association of Labour Economists 2013, Duisenberg Workshop in Behavioral Finance

2014, European Conference on Household Finance 2013, European Finance Association Meetings

2013, Helsinki Finance Summit 2013, Young Scholars Nordic Finance Workshop 2012, Aalto Uni-

versity,European Retail Investment Conference 2015, HECER, London Business School, Research

Institute of Industrial Economics, Stockholm School of Economics, and University of California at

Berkeley provided valuable comments and suggestions, and Antti Lehtinen provided excellent

research assistance. We gratefully acknowledge financial support from the Emil Aaltonen Foun-

dation, the Foundation for Economic Education, the Helsinki School of Economics Foundation,

the NASDAQ OMX Nordic Foundation, and the OP-Pohjola Group Research Foundation. Earlier

versions circulated under the title “Labor Market Experiences and Portfolio Choice: Evidence from

the Finnish Great Depression.” The authors have read the Journal of Finance’s disclosure policy

and have no conflicts of interests to disclose.

DOI: 10.1111/jofi.12469

133

134 The Journal of Finance R

little to explain portfolio heterogeneity.1Thesepatterns suggest the answers to

the heterogeneity puzzle may lie in the events and circumstances individuals

experience during their lifetimes.

Experiences may influence portfolio choice through the formation of beliefs

and preferences. Although psychologists and sociologists have long acknowl-

edged that experiences can have a long-lasting impact on people (see Elder

(1998) for a review), economists have only recently started investigating the

impact of such formative experiences on financial decision making (Malmendier

and Nagel (2011,2016)). In this paper, we extend this line of research by in-

vestigating how formative experiences contribute to the large degree of hetero-

geneity in household portfolios.

The Finnish Great Depression of the early 1990s, combined with rich

register-based data, allows us to solve three challenges that plague attempts

to identify the impact of formative experiences on portfolio choice. First, the

events and circumstances people experience should relate to the formation of

beliefs and preferences. We analyze how experiences of labor market distress

influence long-run portfolio choice. Experiencing a job loss, facing difficulties

in job search, and seeing other workers laid off can lead individuals to hold

more pessimistic beliefs, reduce their appetite for risk, and shatter their trust

in financial markets.2

Second, individuals may experience different events and circumstances be-

cause they are different in ways that are unobservable to the econometrician.3

Variation across local labor markets solves this identification problem. The

depth of the Finnish Great Depression and its root causes—the collapse of

the Soviet Union in 1991 and a twin currency-banking crisis4—meant many

local labor markets experienced unexpected and severe disruptions. We show

that the exposure to adverse labor market conditions is plausibly exogenous

to worker characteristics and hence allows us to identify the impact of labor

market experiences on portfolio choice.

1Bertaut and Starr-McCluer (2002), Calvet, Campbell, and Sodini (2007), Curcuru et al. (2010),

Guiso and Sodini (2013), Haliassos and Bertaut (1995),andHeatonandLucas(2000) document

portfolio heterogeneity. Barnea, Cronqvist, and Siegel (2010), Cesarini et al. (2010), and Calvet

and Sodini (2013) use register-based twin data to analyze the determinants of portfolio choice.

2During the 2007 to 2009 Great Recession, approximately one in six workers in the U.S. labor

force experienced a job loss (Farber (2011)). Labor market distress was not solely confined to job

losers: more than one-third of workers expressed anxiety about layoffs, wage cuts, shorter hours,

and difficulties in finding a good job (Davis and von Wachter (2011)).

3For example, more risk-averse individuals may experience fewer adverse events because they

choose a safer environment, but at the same time their risk aversion makes them less likely

to invest in risky assets. This behavior would generate a spurious positive correlation between

adverse experiences and risky investment.

4From 1991 to 1993, Finland’s real Gross Domestic Product (GDP) fell by 10% and its unem-

ployment rate rose quickly from 3% to 16%. The collapse of the Soviet Union in 1991 induced large

output contraction in industries involved in barter trade between Finland and the USSR (Gorod-

nichenko, Mendoza, and Tesar(2012)). The export shock was amplified by a twin currency-banking

crisis that is typically attributed to financial deregulation and credit expansion in the 1980s and

to attempts to defend the currency peg (Honkapohja et al. (2009), Jonung, Kiander, and Vartia

(2009), Gulan, Haavio, and Kilponen (2014)).

Formative Experiences and Portfolio Choice 135

Third, experiences may not only influence portfolio choice through their ef-

fects on preferences and beliefs, but also affect other determinants of financial

decisions, such as income and wealth.5Controlling for contemporaneous in-

come and wealth does not necessarily solve this challenge because income and

wealth are potential outcomes of labor market conditions (Angrist and Pischke

(2009)). We develop an alternative approach that leverages the great amount

of detail in the data to isolate experiences that likely do not correlate with

wealth, income, or other determinants of portfolio choice. These settings ana-

lyze the influence of secondhand experiences gained through the network of an

individual’s neighbors and family members.

Our measure of local labor market conditions stratifies the data according

to each worker’s region and occupation, and calculates how the rate of un-

employment in each region-occupation cell changed compared to years prior

to the depression. We then relate labor market conditions to investment in

risky assets measured more than a decade after the depression. Importantly,

the employment histories that feed into the calculation of labor market con-

ditions and the asset holdings that determine our measure of investment

in risky assets do not suffer from measurement error caused by recall bi-

ases in surveyed employment histories or by imprecise reporting of asset

holdings.6

Our way of measuring labor market experiences captures local labor market

conditions that are unrelated to worker characteristics. Conditional on fixed

effects for regions and occupations, a number of observable worker charac-

teristics measured prior to the depression do not correlate with labor market

experiences during the depression. Most importantly, labor market conditions

during the depression do not relate to investment in risky assets prior to the

depression. This falsification exercise suggests omitted variables are unlikely

to explain our results.

We find that experiences of adverse labor market conditions are associated

with less long-term investment in risky assets. The estimates suggest the stock

market participation rate, more than a decade after the depression, was 2.8

to 3.6 percentage points lower for workers who experienced a one-standard-

deviation deterioration in labor market conditions. The t-values, clustered at

the level of local labor markets, range from –5.9 to –6.0. This effect is large given

that the unconditional stock market participation rate in our sample is 22%.

The reductions in risky investment also extend to other asset classes. Adverse

labor market conditions are associated with less investment in fixed income

5These factors play a key role in determining risky investment in theories of household portfolio

choice (see Campbell (2006) and Guiso and Sodini (2013) for reviews).

6The main survey used to assess job losses in the United States, the Displaced Workers Survey

(DWS), suffers from recall bias due to its long recall period (five years in the early years of

the survey, three in the later years). For example, the number of displaced workers in 1987 is

dramatically different when estimated based on answers to the January 1988 wave (2.3 million)

and the January 1992 wave (1.3 million). See Appendix A in Congressional Budget Office (1993)

and Evans and Leighton (1995).