Download 52-Week High and Momentum Investing

Leverage, Financial Distress and the
Cross-Section of Stock Returns
Thomas J. George
University of Houston
and
Chuan-Yang Hwang
Nanyang Business School
Nanyang Technological University
December, 2006
Motivation
• Almost every textbook shows that raw return and
equity beta increase with leverage.
• Does leverage affect return once equity beta has
been controlled for? Or equivalently, does leverage
affect risk adjusted return?
• Bhandari (1988) showed that leverage affect return
even after controlling for CAPM equity beta. He
explains this result as CAPM is not a right asset
pricing model.
• More and more evidences have shown that CAPM
is not a good asset pricing model. Fama-French three
factors model has become a popular alternative.
• Will leverage affect FF risk-adjusted return? .
Financial Distress Risk Puzzle
•
•
•
Dechev (1998) shows that firms with higher
financial distress risk earns lower return and
concludes that financial distress risk is not a
systematic risk.
Griffin-Lemmon (2002) show that the distress
puzzle mainly resides in stocks with low book
to market value.
Investors mis-pricing is the cause of the
distress risk puzzle.
Main Findings
• Leverage is negatively related to the future return.
• No distress risk puzzle once we control for leverage.
• Our results are not due to mis-pricing.
• Low leverage firms are riskier than high leverage firms and
they fare poorly in financial distress situation.
• Book-to-market factor alone does not capture how bad things
can get for some firms when times are tough.
• Book to market captures the operating distress risk , while
leverage captures the financial distress risk. Both risks are
priced.
• We construct a leverage factor and show that the leverage risk
premium is larger than the book to market risk premium for
small size and low book to market FF portfolio. This is
consistent with low book to market firms have more intangible
assets hence larger financial distress risk.
Data and Methodology
• Monthly data for all NYSE, AMEX and
NASDAQ firms covered by CRSP from 1965
through 2003.
• Leverage is measured as the ratio of book value
of total debt to book value of total asset.
Data and Methodology (cont’d)




Oscore  1.32  0.407log(total asset )  6.03 total liability  1.43 working capital 
total asset 
 total asset 

 current liability 
 1.72(1 if

current
asset


0.076 



2.73 net income  1.83
 total asset 

total liablity  total asset , 0 otherwise)
funds from operation 

total liability

0.285(1 if a net loss for the last two


years, 0 otherwise)  0.521



net incomet  net incomet 1 

net incomet  net incomet 1 

Data and Methodology (cont’d)
We estimate the leverage effect via the following regressions.
Ri,t  b0 jt  b1 jt Ri,t 1  b2 jt ( BM i,t 1 )  b3 jt (Sizei,t 1 )  b4 jt 52wkhWi,t  j  b5 jt 52wkhLi ,t  j  b6 jt LevHi ,t  j 
b7 jt LevLi,t  j  b8 jt OscHi,t  j  b9 jt OscLi,t  j  ei , j ,t
 According to Fama(1976), these coefficients are the minimum variance portfolio
return
 b0 jt is the month t return a “neutral” portfolio that was formed in month t-j and has hedged
(zeroed out) the effects of book-to-market, size, bid-ask bounce and momentum, as well as
the effects of Leverage and Oscore dummies.
 B6jt can be viewed as the return in excess of b0jt earned by taking a long position j months ago
in a “pure” high leverage portfolio.
 LevLi,t  j =1 if firm i is in the bottom 20% of the leverage in month t-j, 0
otherwise
 LevHi,t  j =1 if firm i is in the top 20% of the leverage in month t-j, 0
otherwise
Data and Methodology (Cont’d)
A strategy to buy high leverage every months and hold on to
it for T months will generate the following portfolio return in
month t.
1 T
S6t   b6 jt
T j 1
 T=12 in this study.
 Unlike traditional approach which uses only high (low)
leverage stocks to estimate high (low) leverage portfolio
returns, we use all stocks.
 Our methodology allows us to control for other effects.
Correlation of variables
Table 1
Correlation Matrix
Using monthly data from January 1965 and December 2003, we construct indicator variables for each of the measures described in the text. The High and Low
Leverage variables are dummies for whether individual stocks are in the top and bottom 20% of leverage as measured by book value of total debt to book value of
assets prior to the portfolio formation month. High and Low FC Index are dummies for stocks ranked in the top and bottom 20% by the financial constraints index
of Whited and Wu (2006). High and Low O-Score are dummies for stocks ranked in the top and bottom 20% by Ohlson’s (1980) O-Score. Details of the
computations of FC Index and O-Score are provided in the Appendix. Numbers reported in the table are time-series averages of cross-sectional correlations.
Low
Leverage
High
Leverage
Low
FC Index
High
FC Index
Low
O-Score
Low Leverage
1.000
High Leverage
-0.250
1.000
Low FC Index
-0.131
0.022
1.000
High FC Index
0.092
-0.031
-0.250
1.000
Low Oscore
0.476
-0.241
0.075
-0.113
1.000
High Oscore
-0.175
0.412
-0.198
0.268
-0.250
High
O-Score
1.000
Table 3
Leverage and O-Score
Raw Returns
Monthly
return
(1,12)
Intercept
Ri,t-1
Book to Market
Size
52 Wk High
Loser
52 Wk High Winner
Low Leverage
High Leverage
1.38
(5.58)
-6.87
(-15.59)
0.30
(3.29)
-0.19
(-4.49)
-0.20
(-1.37)
0.31
(5.88)
0.10
(1.54)
-0.27
(-4.76)
Monthly
return
(1,12)
Jan. excluded
0.98
(4.05)
-6.11
(-14.67)
0.33
(3.61)
-0.07
(-1.81)
-0.57
(-4.26)
0.40
(7.58)
0.11
(1.70)
-0.27
(-4.73)
Low O-Score
High O-Score
Nobs
3228
3228
Monthly
return
(1,12)
1.35
(5.32)
-6.87
(-15.47)
0.20
(2.24)
-0.21
(-5.19)
-0.18
(-1.28)
0.35
(6.49)
Monthly
return
(1,12)
Jan. excluded
0.94
(3.81)
-6.10
(-14.54)
0.23
(2.60)
-0.10
(-2.57)
-0.52
(-4.21)
0.44
(8.44)
0.07
(1.20)
-0.14
(-2.10)
2617
0.07
(1.12)
-0.23
(-3.74)
2617
Monthly
return
(1,12)
1.38
(5.39)
-6.90
(-15.75)
0.27
(3.02)
-0.20
(-5.00)
-0.16
(-1.14)
0.34
(6.55)
0.08
(1.26)
-0.26
(-3.52)
-0.00
(-0.04)
-0.01
(-0.16)
2617
Monthly
return
(1,12)
Jan. excluded
0.97
(3.87)
-6.13
(-14.85)
0.31
(3.39)
-0.09
(-2.34)
-0.50
(-4.07)
0.43
(8.62)
0.11
(1.60)
-0.25
(-3.37)
0.02
(0.39)
-0.13
(-1.59)
2617
Table 3 .1
Leverage and O-Score
Risk Adjusted Returns
Monthly
return
(1,12)
Intercept
Ri,t-1
Book to Market
Size
52 Wk High
Loser
52 Wk High Winner
Low Leverage
High Leverage
Low O-Score
High O-Score
0.07
(1.17)
-6.38
(-14.56)
0.25
(3.49)
-0.13
(-4.12)
-0.36
(-2.79)
0.43
(8.42)
0.23
(3.97)
-0.31
(-5.86)
Monthly
return
(1,12)
Jan. excluded
-0.02
(-0.39)
-5.97
(-14.40)
0.30
(4.15)
-0.06
(-2.09)
-0.63
(-5.37)
0.47
(9.22)
0.21
(3.55)
-0.31
(-5.79)
Monthly
return
(1,12)
0.06
(0.86)
-6.38
(-14.43)
0.15
(2.28)
-0.16
(-5.22)
-0.33
(-2.69)
0.46
(9.07)
Monthly
return
(1,12)
Jan. excluded
-0.05
(-0.78)
-5.96
(-14.26)
0.20
(3.07)
-0.09
(-3.21)
-0.58
(-5.25)
0.50
(10.11)
0.20
(3.69)
-0.15
(-2.41)
0.17
(3.05)
-0.23
(-3.65)
Monthly
return
(1,12)
0.08
(1.19)
-6.42
(-14.72)
0.23
(3.30)
-0.15
(-4.77)
-0.31
(-2.54)
0.45
(9.09)
0.19
(3.26)
-0.30
(-4.83)
0.07
(1.32)
-0.01
(-0.09)
Monthly
return
(1,12)
Jan. excluded
-0.02
(-0.39)
-6.00
(-14.58)
0.28
(4.02)
-0.08
(-2.72)
-0.56
(-5.09)
0.49
(10.20)
0.20
(3.29)
-0.27
(-4.33)
0.04
(0.74)
-0.09
(-1.30)
The hypotheses for the negative leverage effect
• Pricing Error: Even the firms with high (low) debt have
suffered from low (experienced high) return in the past
few years, investors may be still too optimistic
(pessimistic) about the earnings.
• Risk: Low (high) debt firms have high return because
they are inherently riskier (less risky), and they fare
much worse (better) in distress situation.
Tests of the Pricing Error Hypothesis for the
Negative Leverage Effect
• Prediction of Pricing Error Hypothesis: There
should be a more positive (negative) earning
surprise, hence a higher (lower) abnormal
earning announcement, for low (high) leverage
firms.
• We follow La Porta et al (1997) methodology
which they use to test the pricing error
hypothesis for the book to market effect
Tests of the Pricing Error Hypothesis for the
Negative Leverage Effect
• We follow La Porta et al (1997) methodology which they use
to test the pricing error hypothesis for the book to market
effect.
• We benchmark each earning announcement return by the firm
with median book-to market in the same decile as the announcer.
• Every June, we sort firms independently into five groups by Oscore and three groups by debt/asset ratio (top 30%, middle 40%
and bottom 30%), and form portfolios based on these groupings.
For each firm, we then compute the average cumulative three
day abnormal return over the four quarterly announcement
returns following portfolio formation and annualize this number
by multiplying by four.
Table 7
Three-Day Cumulative Abnormal Return around Earnings Announcements
for Portfolios Sorted on Debt/Asset and O-Score
Number of
stocks
Debt/Asset
OScore
L
M
L
-0.62
2
H
H-L
P-value
O-Score
-0.27
2.79
3.42
0.025
L
-0.52
-0.07
0.13
0.64
0.181
3
0.19
0.23
0.23
0.03
4
-0.10
-0.12
-0.34
H
-0.54
-0.59
-0.30
L
M
H
417
141
2
2
187
345
47
0.950
3
87
301
165
-0.24
0.744
4
56
199
284
0.24
0.789
H
41
110
289
Evidences of High Debt firms Are Less Risky
• The range of the return from high to low distress
is smaller for the high debt firms.
• STD of the return on asset and on equity are
smaller for high debt firms.
• STD is calculated under the assumption that
return on asset (or equity) follows a seasonal
random walk with drift.
E (Qi ,t )   i  Qi ,t  4
Company attributes (cont’d)
Oscore
L
M
L
2
3
4
H
ALL
Return on Asset
Year 0 (per cent)
11.28
9.22
6.79
7.01
4.83
5.66
-1.31
3.95
-19.25
-1.98
8.88
5.64
L
2
3
4
H
ALL
Number of firms
210
165
80
286
38
304
24
228
23
147
375
1130
H
L
1.72
6.71
6.00
3.73
0.85
3.73
1
8
37
120
206
372
Debt/Asset
M
Return on Asset
Year 1 (per cent)
10.00
8.46
6.77
6.65
5.49
5.49
3.23
4.21
-5.17
2.09
8.17
5.50
H
4.45
6.81
5.60
3.71
2.11
3.98
STD of Return on Asset
1.91
1.39
6.32
2.52
1.46
1.22
3.18
1.44
1.22
4.20
1.87
1.32
6.67
3.74
2.42
2.40
1.60
1.73
L
M
Return on Asset
Year 2 (per cent)
9.46
7.97
6.30
6.46
5.21
5.20
3.28
4.28
-1.42
2.86
7.68
5.37
H
8.54
6.19
5.49
3.78
2.68
4.28
STD of Return on Equity
2.88
2.64
33.28
3.95
4.36
3.81
5.42
3.28
4.37
12.79
4.25
3.86
20.79
10.08
8.69
3.72
3.40
5.73
Table 12.1
Leverage and Oscore by Subperiod
Risk Adjusted Returns
Intercept
Intercept
Ri,t-1
Book to Market
Size
52 Wk High Loser
52 Wk High Winner
Low Leverage
High Leverage
Low O-Score
High O-Score
Jun 1966-Dec 1979
Jan 1980-Dec 2003
Monthly
return
(1,12)
Monthly
return
(1,12)
0.04
(0.62)
-7.87
(-10.78)
0.14
(1.40)
-0.06
(-1.35)
Monthly
return
(1,12)
Jan. excluded
0.04
(0.56)
-7.73
(-11.43)
0.14
(1.49)
-0.02
(-0.42)
-0.53
(-4.20)
0.30
(3.47)
0.03
(0.44)
-0.04
(-0.39)
0.12
(1.54)
0.05
(0.45)
-0.60
(-5.12)
0.31
(3.85)
0.01
(0.10)
-0.04
(-0.36)
0.09
(1.13)
0.01
(0.05)
Entire Sample
0.10
(0.98)
-4.76
(-10.04)
0.33
(3.81)
-0.16
(-3.86)
Monthly
return
(1,12)
Jan. excluded
-0.07
(-0.79)
-4.11
(-9.59)
0.40
(4.78)
-0.07
(-2.01)
0.09
(1.36)
-5.88
(-14.22)
0.22
(3.35)
-0.13
(-3.94)
Monthly
return
(1,12)
Jan. excluded
-0.01
(-0.15)
-5.47
(-14.55)
0.27
(4.07)
-0.05
(-1.80)
-0.35
(-1.74)
0.58
(8.73)
0.30
(3.97)
-0.50
(-6.81)
0.03
(0.38)
0.00
(0.02)
-0.72
(-4.00)
0.66
(10.02)
0.34
(4.26)
-0.46
(-6.26)
0.02
(0.24)
-0.10
(-1.05)
-0.40
(-3.03)
0.47
-(.93)
0.19
(3.26)
-0.30
(-4.83)
0.07
(1.32)
-0.01
(-0.09)
-0.67
(-5.64)
0.52
(10.18)
0.20
(3.29)
-0.27
(-4.33)
0.04
(0.74)
-0.09
(-1.30)
Monthly
return
(1,12)
Four-Factor Model
• We create a leverage factor (LEV) in addition to FF
three factors (Market, SMB, HML) to form a fourfactor pricing model.
• Since FF (1993) also hypothesized that their HML
captures financial distress risk, can our LEV displace
their HML?
• We use four-factor model to explain the return of FF
100 portfolios. We find that most of the FF 100
portfolios have significant loadings on both LEV and
HML factors.
• This indicates both LEV and HML risks are priced.
Figure 1
LEV Risk Premium Minus HML Risk Premium
0.60%
0.40%
0.20%
0.00%
-0.20%
-0.40%
-0.60%
-0.80%
Largest
-1.00%
7
Book-to-Market
Decile
Lowest
3
4
5
6
7
8
9
Smallest
2
-1.20%
4
Highest
Size
Decile
Operating Distress Risk Vs. Financial Distress
Risk
• Operating distress risk measures the difficulty in
reversing physical investment in bad economy, hence
high book to market firms have high operating distress
risk.
• Financial distress risk measure the loss of asset value
when in financial distress. We have shown low
leverage firms have high financial distress risk.
• HML captures operating distress risk, while LEV
captures financial distress risk. This can explain the
LEV risk premium are high for small size and low
book to market value firms since these firms have more
intangible assets hence larger financial distress risk.
Conclusion
• We document a negative leverage effect : Book
leverage is negatively related to both raw return and
risk adjusted return.
• There is no distress risk puzzle after controlling for the
leverage.
• The negative leverage effect can not be explained by to
pricing error; Instead it can be explained as low
leverage firms have higher financial distress cost.
• Low leverage firms indeed have higher financial
distress cost; we show they fare much worse in
distress.
Conclusion
•
•
We create a leverage factor that capture financial
distress risk which helps explain the FF portfolio
return that book to market factor does not explain.
The leverage factor risk premium is much larger for
small size and small book to market firms. This is
consistent with the fact that low book to market firms
have more intangible assets hence have larger
financial distress risk