The Geography of Venture Capital Financing: An Empirical Analysis

Corporate Scandals and Household
Stock Market Participation
Mariassunta Giannetti, Stockholm School of Economics, CEPR
Tracy Yue Wang, University of Minnesota
The Idea

Effect of fraud on market participation and spillover effect on
firms



Corporate frauds may decrease demand for equity and households’
willingness to (directly or indirectly) participate in the stock market;
Through this channel, fraud may generate large indirect value losses
by increasing the cost of capital for other firms (i.e., firms with no
revealed fraud).
Households are important yet relatively vulnerable
participants in the stock market.



Low household participation rate (Guiso & Sodini 2013)
Incentive to participate is sensitive to past stock market experience
and the level of trust
Corporate frauds likely have a large negative impact on household
participation
2
Empirical Design

Corporate frauds → household participation →
firms’ cost of capital, valuation, liquidity...

Identify variation in households’ exposure to fraud


Likely larger exposure to frauds in firms located in the
same state.
Local bias in households’ investment


Grinblatt & Keloharju 2001, Ivkovic & Weisbenner 2005,
Seashole & Zhu 2010.
Local news coverage, personal connections,…
3
Empirical Design

Causal impact of fraud on household participation



The sudden demise of Arthur Andersen as an exogenous
shock to fraud revelation;
Exogenous variation in households’ life-time experience
of corporate fraud.
Spillover effect on firms with no revealed fraud


Geographic spillover through local demand for equity vs.
Information spillover in an industry
4
Main Results

Corporate frauds in a state significantly reduce
equity holdings and stock market participation of
households in that state.


The impact of fraud is causal.
Significant spillover effect of fraud: other firms
headquartered in states hit by corporate scandals
experience



Decline in # of shareholders (particularly retail shareholders)
Increase in cost of capital
Decrease in valuation and liquidity
5
Relation to Fraud Literature


Most work focus on consequences to shareholder
value of the fraud-committing firms: loss of firm
value, cost of equity, cost of debt…
Limited work on spillover effects


Information spillover among industry peers: Gleason,
Jenkins, and Johnson (2008), Goldman, Peyer, &
Stefanescu (2012).
We document a distinct and potentially more
pernicious spillover effect through the demand for
equity by local households.
6
Relation to Participation Literature

Effect of past stock market experiences (Malmendier and
Nagel 2011)




More recent stock market returns have larger effects;
Younger households’ participation is more affected by past
experiences.
We find that corporate scandals have more lasting effects than general
stock market experiences, affecting older households more.
Effect of trust (Guiso, Sapienza, & Zingales 2008)



Trust that is rooted in cultural origin;
Less financially-literate households are more influenced by trust.
We find that corporate scandals tend to affect households who are
more likely to be in the market in the absence of frauds.
7
Data on Households

Panel Study of Income Dynamics (PSID)



State of residence
Infor. on equity holdings: 1984, 1989, 1994, 1999, 2001,
2003, 2005, 2007, 2009
“Equity Participation” =1 if hold stocks in public firms,
mutual funds, or investment trusts in a given year





Exclude equity holdings in retirement accounts
“Equity Participation_IRA”
Participation rate: 22% (w/o IRA), 30% (w/ IRA)
Equity value, net equity purchase, equity-wealth ratio
Household characteristics: family income, wealth, number
of family members, age, years of schooling, marital status8
Data on Corporate Fraud

Federal Securities Regulation (FSR) database (by
Karpoff, Lee, and Martin)




Enforcement actions by SEC and DOJ: 1980-2009.
Select 711 cases involving 702 US publicly traded issuers
with charged securities fraud under the Securities Act or
the Exchange Act.
Time of fraud revelation
Headquarter state infor. from COMPUSTAT and
hand collection


Covers 47 out of (50 states + DC)
Variation in fraud revelation across states over time
9
Data on Corporate Fraud

“Fraud in State”: cumulative fraud revelation
intensity in the past 4 years in a state



Fraud revelation intensity = # of revealed frauds divided by
total # of public firms in a state;
Mean: 1.02%, S.D.: 2.31%
Alternative measures:



“Fraud in State 2”: larger weights for frauds in larger firms;
“Fraud in State 3”: larger weights for frauds with more
negative market reaction upon detection;
“Fraud in State 4”: larger weights for frauds in firms with
more retail shareholders.
10
Data on State Conditions

State economic conditions could be the omitted
variables



Bereau of Economic Analysis (BEA)


Economic condition affects fraud revelation (Povel et al.
2007, Wang et al. 2010);
May affect households’ stock market participation too.
State GDP growth, employment growth, population, total
personal income (and growth), per capital personal
income (and growth)
State stock market condition

Buy-and-hold VW return of stocks in the state
11
Empirical Model
Participat ionijt    FraudInState jt  X ijt  hi  s j  t   ijt



Participation(ijt): equity participation, log(equity value),
log(net equity purchase), equity-wealth ratio
X(ijt): time-varying controls for the households and the states
Fixed effects: household, state, year



Year : absorb national changes in demand for equity ;
State : absorb average differences in fraud revelation and participation
across states;
Household: absorb any unobservable household characteristics that
affect participation.
12
13
14
Identification Strategy 1

A shock to fraud revelation due to the sudden demise
of Arthur Andersen (AA)



The probability of fraud detection increases for AA clients
relative to other firms during 2002-2004 (Dyck, Morse,
Zingales 2013);
The effect of the shock is different across states,
depending on the fraction of firms in a state that were AA
clients in 2001-2002 (“AA Shock”), which should be
exogenous to our question.
Use “AA Shock” as the instrument for “Fraud in State”
15
Exclusion Restriction
(before 2001)
Validity of the Instrument
16
Identification Strategy 1
17
Identification Strategy 2

Households living in the same state at a point of time
may have different corporate fraud experiences




Differences in household age
Some households have moved across states
Variation across households’ experiences due to age and
moving is exogenous to our question.
Allow us to absorb any state level shocks by including
state-year fixed effects.
18
Identification Strategy 2

Use the methodology in Malmendier & Nagel (2011):
19
Identification Strategy 2
c
20
Which Households are More
Affected?
21
Spillover Effects on Local Firms


Corporate frauds in a state have long-lasting negative
impact on equity holdings of local households
Households are more likely to invest in local stocks.



Local bias implies segmentation in the supply of capital by
geographic areas.
Frauds could lead to a lasting decrease in demand for all
local stocks.
The decrease in demand could generate economic costs
on local firms that are not involved in fraud.
22
Change in Shareholder Base
23
Cost of Capital
24
IV Estimation


Use “AA Shock” to instrument for “Fraud in State”
Also examine valuation and liquidity
25
Conclusion


Corporate fraud has a long-lasting negative
impact on households’ incentive to participate
in the stock market.
The decrease in households’ demand for
equity due to fraud generates significant costs
on local firms that are not even involved in
fraud.
26