Short Selling and Earnings Management:
A Controlled Experiment
Vivian Fang, University of Minnesota
Allen Huang, Hong Kong University of Science and Technology
Jonathan Karpoff, University of Washington
SEC, Division of Economic and Risk Analysis
& U of Maryland, Center for Financial Policy
May 1st, 2015
1
Motivation
Does short selling (SS), or its prospect, constrain firms’
opportunistic reporting behavior?
 Short sellers can identify earnings manipulation and fraud
• Dechow, Sloan, and Sweeney (1996), Christophe, Ferri, and Angel (2004), Efendi, Kinney, and
Swanson (2005), Desai, Krishnamurthy, and Venkataraman (2006), and Karpoff and Lou (2010)
 Anecdotal evidence
 David Einhorn (Founder and president of Greenlight Capital)
• “Fooling some of the people all of the time”:
• Six-year fight with Allied Capital between 2002 and 2008
• In June 2007, the SEC found that Allied broke several securities laws
relating to the accounting and valuation of illiquid securities.
2
Motivation (Cont’d)
 There exist various incentives for earnings management (EM)
 Compensation, insider trading, job security, operational flexibility, or control
 SS imposes one constraint on EM
• By reducing benefits: SS increases price efficiency
(e.g., Boehmer, Jones, and Zhang, 2013; Boehmer
and Wu, 2013)
• By increasing costs: SS tracks a firm’s discretionary
accruals and detects EM (e.g., Cao et al., 2006;
Karpoff and Lou, 2010; Hirshleifer, Teoh, and Yu,
2011)
3
MC2
MC1
MB1
MB2
L2
L1
EM
Empirical Approach
 We exploit Regulation SHO’s 202 T pilot program
 Every third stock in the Russell 3000 index ranked by trading volume
within each exchange was drawn and designated as a pilot stock.
 From May 2, 2005 - August 6, 2007 (as scheduled), pilot stocks were
exempted from short-sale price tests.
 Tick test (for exchange-listed stocks): a short-sale can only occur at a price
above the most recently traded price (plus tick) or at the most recently traded
price if that price is an uptick from the last different price (zero-plus tick).
 Bid test (for Nasdaq NM stocks): a short sale can only occur at a price at
least one penny above the bid price if the bid price is a downtick from the
previous bid.
4
Empirical Approach (Cont’d)
 The pilot program facilitates a diff-in-diff (DiD) analysis
 Relevance: an economically meaningful decrease in pilot stocks’
SS costs (increase in SS prospect) relative to non-pilot stocks’
 SEC (OEA, 2007): increase in short-sale trades and short sales-to-shares ratio for
pilot stocks / Diether et al. (2009): more order-splitting for NYSE pilot stocks
 Strong public reaction: Repeal of price tests in July 2007 led to backlash from
practitioners and politicians & Reversal of the policy in Feb 2010 drew sharp
criticism from HFs and short sellers
 Exogeneity: the decrease in pilot stocks’ SS costs is exogenous
 Controlled experiment with random assignment of treatment and control group
 No evidence that the firms themselves lobbied for the pilot program
 The goal was to evaluate the impact of short-sale price tests on market quality
5
Data
• Data Sources
 2004 Russell 3000 index
 Pilot stocks are from SEC’s first pilot order
 Excluded are stocks that were not previously subject to price tests and stocks that went
public or had spin-offs after April 30, 2004
 Sorted by stock’s average daily dollar volume from June 2003 through May 2004
 Every third stock (starting with the 2nd one) within each listing market are then picked
 Compustat Industrial Annual Files for EM and controls
 Excluded are financial and utilities firms
• Sample
 388 pilot and 709 non-pilot firms (balanced sample) over 2001-2003, 2005-2010
 The results are similar using an unbalanced sample.
6
Balanced Sample
Variable
ASSET
MB
ASSETGR
CAPEX
R&D
ROA
CFO
LEV
CASH
N
388
388
388
388
388
388
388
388
388
DIVIDENDS 388
7
Treatment Group
Control Group
(Pilot Firms)
(Non-pilot Firms)
Mean
Median
N
Mean
None is
Significant.
Median
tWilcoxon
statistic z-statistic
3,748.61
2.75
13.42
5.55
4.19
14.37
11.36
29.36
0.21
817.69
1.95
7.88
3.76
0.00
14.51
11.33
26.46
0.12
709
709
709
709
709
709
709
709
709
3,746.25
2.60
13.22
5.50
4.04
14.15
10.56
29.80
0.22
817.42
1.98
7.66
3.65
0.32
14.29
10.46
27.56
0.11
0.00
0.68
0.10
0.14
0.27
0.25
1.10
-0.25
-0.44
0.51
0.03
-0.30
0.91
-0.94
0.22
1.10
-0.24
-0.32
0.83
0.00
709
0.73
0.00
1.14
1.07
Pilot-related Variables
• PILOT: dummy variable to indicate pilot firms
• PRE: dummy variable to indicate whether FYE falls between January 1,
2001 and December 31, 2003
• DURING: dummy variable to indicate whether FYE falls between
January 1, 2005 and December 31, 2007
• POST: dummy variable to indicate whether FYE falls between January 1,
2008 and December 31, 2010
8
EM Variables
• Discretionary Accruals (Kothari, Leone, and Wasley, 2005)
 Discretionary accruals = Total accruals (Earnings – Cash Flows) – Fitted
normal accruals (Performance matched benchmark within industry-year)
• Likelihood of beating earnings targets (e.g., Graham, Harvey,
and Rajgopal, 2005; Bhojraj et al., 2009)
 BEAT_ALY = 1 if the reported EPS falls between the analyst consensus
forecast and that plus 1ct in a quarter and 0 otherwise;
 BEAT_EPS = 1 if the reported EPS falls between prior year same
quarter’s EPS and that plus 1ct in a quarter and 0 otherwise.
9
First Finding Univariate Plot
First Finding: Pilot firms’ discretionary accruals (and likelihood of marginally
beating earnings targets) decrease during the pilot program, and revert to preexperiment levels when the program ends.
0.002
0.000
-0.002
Pre-pilot
During-pilot
Post-Pilot
-0.004
-0.006
-0.008
-0.010
-0.012
-0.014
-0.016
Discretionary Accruals of Pilot Firms
Discretionary Accruals of Non-pilot Firms
10
First Finding Multivariate DiD
Dependent Variables
PILOT×DURING
Discretionary accruals
BEAT_ALY
BEAT_EPS
-0.010**
-0.079**
-0.074*
(0.004)
(0.040)
(0.044)
PILOT×POST
0.003
0.017
-0.003
(0.004)
(0.043)
(0.033)
PILOT
0.000
0.022
0.045*
(0.003)
(0.029)
(0.024)
DURING
-0.001
-0.219***
0.089**
(0.002)
(0.050)
(0.039)
POST
-0.001
-0.437***
-0.062*
(0.005)
(0.056)
(0.032)
Controls
Yes
Yes
Yes
# of obs.
9,873
28,341
59,589
Adjusted R2
0.40%
2.21%
0.64%
Discretionary accruals is 1 percentage point lower for the treatment group than for the control
group during the 3-year pilot period compared to the 3-year period pre-pilot.
The likelihood of marginally beating analyst consensus (prior year same quarter’s EPS) is 1.8
(0.8) percentage points lower for the treatment group, 11.1% (14.2%) of the unconditional %.
11
First Finding Robustness Checks
 Results are robust to using alternative measures of EM
 Fraud-score of Dechow et al. (2011)
 Results are not explained by
 Growth / Investment
 We consider several ways to control for growth/investment
 Pilot firms’ investment levels do not follow the same pattern as discretionary accruals
 Equity Issuance
 We show similar results for firms not seeking to issue equity as for the overall sample
 Market attention
 We control for various proxies of market attention/investor awareness
 Reverting pattern is not consistent with a market attention story (Chen et al. 2004)
12
Second Finding
Second Finding: For financial misconduct that occurred before the announcement
of the pilot program, pilot firms are more likely to be caught after the pilot program
starts than non-pilot firms. As we sequentially include frauds initiated after the
pilot program begins, the unconditional likelihood of pilot firms getting caught
converges monotonically toward that of non-pilot firms.
Pr(Caught(t + n), Fraud(t)) = Pr(Fraud(t)) × Pr(Caught(t + n)|Fraud(t))
 Pr(Caught(>May 2005), Fraud(<July 2004))
=
 Pr(Caught(>May 2005), Fraud(>July 2004))
Dependent Variables
PILOT
Controls
Industry FE
# of obs.
Pseudo R2
13
(1)
Pre-2004
fraud caught
0.202***
(0.061)
Yes
Yes
2,694
5.90%
(2)
Pre-2005
fraud caught
0.193***
(0.062)
Yes
Yes
2,697
5.92%
(3)
Pre-2006
fraud caught
0.185***
(0.070)
Yes
Yes
2,700
5.18%
(4)
Pre-2007
fraud caught
0.142*
(0.076)
Yes
Yes
2,704
4.44%
Third Finding Current return-future earnings test
Third Finding: Pilot firms’ price efficiency increase during the pilot program.
Annual buy-andDependent Variable
Rt
Aggregated earnings
hold return
X3t
0.270***
for the next 3 years;
(0.027)
Coefficient of current
X3t×PILOT×DURING
0.158***
return on future
(0.060)
earnings (Lundholm
X3t×PILOT
-0.014
and Myers, 2002)
X3t×DURING
Xt-1
Xt
R3t
14
INTERCEPT
# of obs.
Adjusted R2
(0.036)
0.037
(0.030)
-0.676***
(0.054)
0.576***
(0.044)
-0.125***
(0.007)
0.347***
(0.010)
13,844
10.90%
Third Finding PEAD test
Third Finding: Pilot firms’ price efficiency increase during the pilot program.
Boehmer and Wu (2013)
Post-earnings announcement drift, CAR (+2, +11)
Treatment
Control
t-Statistics
Group
Group
Control – Treatment
Earnings surprise
D1 (Most negative)
-0.38%
-1.26%***
2.47**
D2
-0.41%**
-0.41%***
0.01
D3
-0.38%**
-0.57%***
0.88
D4
0.00%
-0.07%
0.35
D5
-0.22%
-0.24%**
0.12
D6
-0.13%
-0.16%
0.13
D7
0.10%
-0.10%
0.95
D8
-0.02%
-0.06%
0.16
D9
0.23%
0.36%**
0.45
D10 (Most positive)
0.97%***
0.83%***
0.43
• When investors fail to fully capitalize on earnings information, returns will drift in the same
15 direction as the earnings surprise (Ball and Brown, 1968; Bernard and Thomas, 1989, 1990).
• Short selling facilitates the incorporation of negative information into stock prices.
Conclusion and Contribution
• Takeaway: An exogenous reduction in short selling costs leads to less EM,
higher conditional probability of detection, and enhanced price efficiency.
• Contribution:
 Literature on the real effects of trading in financial markets
 Bond, Edmans, and Goldstein (2012) for a survey
 Literature on the determinants of earnings management
 Dechow, Ge, and Schrand (2010) for a review
 Literature on short selling and price discovery
 Add to the policy debate on the costs and benefits of short selling
 Short selling, or its prospect, improves financial reporting quality even among
firms that are not charged with financial reporting violations.
16