Lin, Ji-Chai, Ajai K. Singh, and Wen Yu, 2009, Stock splits, trading continuity, and the cost of equity capital, Journal of Financial Economics 93, 474–489.
Livnat, Joshua, and Richard R. Mendenhall, 2006, Comparing the post-earnings announcement drift for surprises calculated from analyst and time series forecasts, Journal of
Accounting Research 44, 177–205.
Loughran, Tim, and Jay R. Ritter, 2000, Uniformly least powerful tests of market efficiency,
Journal of Financial Economics 55, 361–389.
McNichols, Maureen, and Ajay Dravid, 1990, Stock dividends, stock splits, and signaling,
Journal of Finance 45, 857–879.
Mitchell, Mark L., and Erik Stafford, 2000, Managerial decisions and long-term stock price performance, Journal of Business 73, 287–329.
Muscarella, Chris J., and Michael R. Vetsuypens, 1996, Stock splits: Signaling or liquidity? The case of ADR ‘solo-splits’, Journal of Financial Economics 42, 3–26.
Nayak, Subhankar, and Nagpurnanand R. Prabhala, 2001, Disentangling the dividend information in splits: A decomposition using conditional event-study methods, Review of
Financial Studies 14, 1083–1116.
Newey, Whitney, and Kenneth D. West, 1987, A simple, positive semi-Definite, heteroskedasticity and autocorrelation consistent covariance matrix, Econometrica 55, 703–708.
Pastor, Lubos, and Robert F. Stambaugh, 2003, Liquidity risk and expected stock returns,
Journal of Political Economy 111, 642–685.
White, Halbert, 1980, Heteroskedasticity-consistent covariance matrix and a direct test for heteroskedasticity, Econometrica 48, 817–838.
Table 1
Post-Split Abnormal Returns
This table reports one-year abnormal returns (in %) following split announcements. We measure abnormal returns based on regression alphas of the calendar-time portfolio formed by firms that had stock split announcements in the past 12 months. We use different factor models to control for popular risk factors: Fama–French (1993) three-factor model, Carhart (1997) four-factor model, and a five-factor model that adds the liquidity factor, based on Pastor and Stambaugh (2003), into the Carhart model. We compute calendar-time portfolio returns using both equal-weighted (EW) and value-weighted (VW) methods. Both ordinary least squares (OLS) and weighted least squares (WLS) approaches are used to run calendar-time regressions where the weight for WLS is the number of splits included in the portfolio in each month. Calendar months with less than 10 stocks in the portfolio are excluded from the regression. Numbers in parentheses are t-statistics based on Newey–West (1987) standard errors. ***, **, and * indicate the significance levels of 1%, 5%, and 10%, respectively.
OLS WLS
EW VW EW VW
Factor model alpha t-stat alpha t-stat alpha t-stat alpha t-stat Panel A: Sample period 1984–2011 (9,366 stock splits)
three-factor 0.45*** (3.67) 0.34** (2.47) 0.43*** (4.04) 0.42*** (3.78) four-factor 0.43*** (3.18) 0.21 (1.39) 0.41*** (3.26) 0.25** (2.17) five-factor 0.40*** (2.66) 0.24 (1.56) 0.40*** (3.18) 0.23** (2.05) Panel B: Sample period 1963–2011 (14,388 stock splits)
three-factor 0.47*** (5.83) 0.37*** (4.05) 0.46*** (5.37) 0.43*** (4.84) four-factor 0.41*** (4.21) 0.19* (1.84) 0.40*** (3.75) 0.22** (2.42) five-factor 0.43*** (3.79) 0.23** (2.01) 0.41*** (3.81) 0.21** (2.29) Panel C: Sample period 1926–2011 (15,145 stock splits)
three-factor 0.37*** (5.42) 0.37*** (4.11) 0.44*** (5.35) 0.42*** (5.02) four-factor 0.31*** (3.76) 0.18* (1.96) 0.38*** (3.67) 0.21** (2.43)
Table 2
Post-Split Abnormal Returns by Horizons
This table reports post-split abnormal returns (in %) over different investment horizons. We measure abnormal returns based on alphas of the regressions using a five-factor model that adds the liquidity factor, based on Pastor and Stambaugh (2003), into the Carhart (1997) four-factor model. We form a calendar-time portfolio by including firms that had stock split announcements in the past 3, 6, or 12 months (Panel A) or firms that made split announcements in the x month before (Panel B) where x =1, 2, …, 12. Monthly portfolio returns are either equal-weighted (EW) or value-equal-weighted (VW). Both ordinary least squares (OLS) and equal-weighted least squares (WLS) methods are used to run regressions where the weight for WLS is the number of splits included in the portfolio in each month. Calendar months with less than 10 stocks in the portfolio are excluded from the regression. Numbers in parentheses are t-statistics based on Newey–West (1987) standard errors. ***, **, and * indicate the significance levels of 1%, 5%, and 10%, respectively.
OLS WLS
Table 3
Regressions of Post-Split Three-Month Abnormal Returns on Standardized Unexpected Earnings (SUEs) This table reports cross-sectional regressions of post-split three-month buy-and-hold abnormal returns (in %) on standardized unexpected earnings (SUE) and control variables. The dependent variable is the three-month return starting from day +3 after split announcements minus the return over the same horizon of the control portfolio matched on size, book-to-market (B/M), and momentum. We construct 125 size–B/M–momentum control portfolios as follows. We retrieve all firms with available data in the stock universe. In each month from July of year t to June of year t+1, we sort firms into five size portfolios using NYSE cutoffs where the size is the market value at June-end of year t. In each size portfolio, we classify five B/M quintiles where B/M is the book value of equity at the fiscal year t–1 divided by the market value of equity at December end of year t–1. Within each size-B/M group, we further sort firms into five momentum quintiles where momentum is the prior 11-month return cumulative from month –12 to month –2 and is updated monthly. The control portfolio with the same quintile rankings of size, B/M and momentum as of the sample firm is chosen as the benchmark to estimate abnormal returns. SUE is the earnings surprise after stock splits based on I/B/E/S reported analyst forecasts and actual earnings as in Livnat and Mendenhall (2006). SUE is winsorized at 1% level and expressed in percentage. Pre-split SUE is SUE prior to stock splits. Size is the natural logarithm of pre-split market capitalization in millions. B/M is the pre-split book-to-market ratio from Compustat. Run-up is the pre-split 12-month cumulative stock return. Pre price is the natural logarithm of the stock price at day –3 relative to the split announcement. ILLIQ is Amihud’s (2002) illiquidity measure defined as the average ratio of the daily absolute return to its dollar trading volume. Change in ILLIQ is the difference between pre-split ILLIQ estimated over three months before the split announcement date and post-split ILLIQ estimated over three months after the effective date. Numbers in parentheses are t-statistics based on White (1980) heteroskedasticity-adjusted standard errors. ***, **, and * indicate the significance levels of 1%, 5%, and 10%, respectively.
Table 4
Monthly Cross-Section Regression on Split Dummy
This table shows the monthly return regressions on split dummies and control variables. Models 1 and 2 are Fama–
Macbeth cross-section regressions, and Models 3 and 4 are panel regressions. The dependent variable is the monthly return (in %). Split dummy is a dummy variable, which equals 1 if the firm makes a split announcement in the past three months, and zero otherwise. Split dummy is further divided into high, medium, and low dummies based on split factors. If the firm makes a split announcement in the past three months, High split dummy equals 1 for a split factor ≥ 1; medium split dummy equals 1 for a split factor between 0.5 and 1; and low split dummy equals 1 for a split factor < 0.5. All three dummies equal zero elsewhere. SUE is the earnings surprise (in %) in the most recent quarter based on I/B/E/S reported analyst forecasts and actual earnings as in Livnat and Mendenhall (2006). Size is natural log of the market value at the end of June of year t+1. B/M is natural log of book-to-market ratio, defined as the book value of equity at the fiscal year-end of year t divided by the market value of equity at December end of year t. Run-up is prior 11-month cumulative return from month –12 to month –2. Asset growth is annual growth rate of the total assets. Net share issue is annual share issuance, defined as the logarithm of the ratio of split-adjusted shares outstanding at the end of December of year t to split-adjusted shares outstanding at the end of December of year t-–1, where split-adjusted shares outstanding are CRSP shares outstanding divided by the CRSP cumulative factor to adjust shares outstanding. Accruals are the change in noncash current assets less the change in current liabilities excluding the change in debt included in current liabilities and the change in income taxes payable, minus depreciation at the fiscal year-end of year t, all scaled by average total assets at the fiscal year-end of year t and t–1.
ROA is earnings before interest, taxes, depreciation and amortization at the fiscal year-end of year t divided by average total assets at the fiscal year-end of year t and t–1. ILLIQ is Amihud’s (2002) illiquidity measure defined as the one-year average ratio of the daily absolute return to its dollar trading volume. For Models 1 and 2, numbers in parentheses are t-statistics based on Newey–West (1987) standard errors. For Models 3 and 4, numbers in parentheses are t-statistics based on the firm and year double clustered standard errors (Petersen, 2009). ***, **, and
* indicate the significance levels of 1%, 5%, and 10%, respectively.
Fama–Macbeth regression Panel regression
31 Table 5
Three-month Abnormal Returns following Quarterly Earnings Announcements: Two-Way Sorts by Splits and Standardized Unexpected Earnings (SUEs) This table shows the three-month buy-and-hold abnormal returns (BHARs; in %) following quarterly earnings announcements sorted by SUE quintiles and by whether firms make split announcements. All quarterly earnings announcements during our sample period are first sorted into SUE quintiles by fiscal quarter. In each SUE quintile, we classify split firms if firms announce stock splits in the three months before quarterly earnings announcements, and non-split firms otherwise. BHAR is the three-month return starting from day +3 following earnings announcements minus the return over the same horizon of the control portfolio matched on size, book-to-market, and momentum. High – Low shows the difference in BHARs between the top and bottom SUE quintiles. Split – Non-split shows the difference in BHARs between split firms and non-split firms. The bold number in the right bottom corner reports the difference in BHARs between split firms in the top SUE quintile and non-split firms in the bottom SUE quintile. N is the number of quarterly earnings announcements. Numbers in parentheses are t-statistics based on two-side t-tests. ***, **, and * indicate the significance levels of 1%, 5%, and 10%, respectively.
SUE ranking
32 Table 6
Two-Way Sorts of Three-month Abnormal Returns following Quarterly Earnings Announcements: Split Firms vs. Matching Firms.
This table shows the three-month buy-and-hold abnormal returns (BHARs; in %) following quarterly earnings announcements sorted by standardized unexpected earning (SUE) quintiles and by whether firms make split announcements. All quarterly earnings announcements during our sample period are first retrieved. We then pick up split firms that make split announcements within three months before quarterly earnings announcements. For each split firm, we identify its corresponding matching firms matched on size, book-to-market, prior return and SUE quintile that do not have split announcements in the prior three months. We drop all other quarterly earnings announcements from the analysis. BHAR(%) is the three-month return starting from day +3 following earnings announcements minus the return over the same horizon of the control portfolio matched on size, book-to-market, and momentum (prior 11-month return cumulative from month –12 to month –2). High – Low shows the difference in BHARs between the top and bottom SUE quintiles. Split – Matching shows the difference in BHARs between split firms and non-split matching firms. The bold number in the bottom corner reports the difference in BHARs between split firms in the top SUE quintile and matching firms in the bottom SUE quintile. N is the number of quarterly earnings announcements. Numbers in parentheses are t-statistics based on two-side t-tests. ***, **, and * indicate the significance levels of 1%, 5%, and 10%, respectively.
SUE ranking
Table 7
Long–Short Trading Strategies Based on Splits and Earnings Surprises
This table shows the monthly abnormal returns of different trading strategies. The abnormal returns (in %) are measured by the alphas in calendar-time portfolio weighted least squares regressions using the five-factor model.
The five-factor model adds the liquidity factor, based on Pastor and Stambaugh (2003), into the Carhart (1997) four-factor model. Standardized unexpected earnings (SUE) are sorted into quintiles by fiscal quarter. We take the bottom two SUE quintiles as low SUE firms, and top two SUE quintiles as high SUE firms. Split/high SUE – Non-split/low SUE is the long–short trading strategy by buying split firms with high SUEs and selling size, book-to-market, and momentum matched non-split firms with low SUEs. Split – Non-split is the trading strategy by buying split firms and selling size, book-to-market, and momentum matched non-split firms. High SUE – Low SUE represents the strategy by buying firms with high SUEs and selling firms with low SUEs. Firms are included in portfolio from the month following split or earnings announcement and held from 3 to 12 months. Monthly portfolio returns are equal-weighted (EW) or market value weighted (VW). Calendar months with less than 10 stocks in the portfolio are excluded from the regression. t-statistics are based on the standard errors using Newey–West (1987) method. ***, **, and * indicate the significance levels of 1%, 5%, and 10%, respectively.
Split/high SUE –
Non-split/low SUE Split – Non-split High SUE – Low SUE
Period alpha t-stat alpha t-stat alpha t-stat
Panel A: Equal-weighted
(+1, +3) 1.63*** 5.54 1.20*** 6.99 0.74*** 9.42
(+1, +6) 1.23*** 5.71 0.78*** 5.81 0.44*** 8.29
(+1, +12) 0.84*** 5.03 0.45*** 4.08 0.24*** 7.03
Panel B: Value-weighted
(+1, +3) 0.96*** 2.60 0.64** 2.14 0.21* 1.89
(+1, +6) 0.86*** 3.45 0.52** 2.46 0.14* 1.71
(+1, +12) 0.72*** 3.42 0.37** 2.11 0.17** 2.42
Table 8
Calendar-Time Portfolio Alpha: Long Only Trading Strategies
This table shows the monthly abnormal returns of different trading strategies. The abnormal returns (in %) are measured by the alphas of weighted least squares regressions by regressing calendar time portfolio excess returns against five factors. The five-factor model adds the liquidity factor, based on Pastor and Stambaugh (2003), into the Carhart (1997) four-factor model. Standardized unexpected earnings (SUE) are sorted into quintiles by fiscal quarter.
We take the bottom two SUE quintiles as low SUE firms, and top two SUE quintiles as high SUE firms. Split/high SUE is the trading strategy by buying split firms with high SUEs. Split includes all split firms in the sample. High SUE represents the strategy by buying firms with high SUEs. Firms are included in portfolio from the month following split or earnings announcement and held from 3 to 12 months. Monthly portfolio returns are equal-weighted (EW) or market value equal-weighted (VW). Calendar months with less than 10 stocks in the portfolio are excluded from the regression. t-statistics are based on the standard errors using Newey–West (1987) method. ***,
**, and * indicate the significance levels of 1%, 5%, and 10%, respectively.
Split/high SUE Split High SUE
Period alpha t-stat alpha t-stat alpha t-stat
Panel A: Equal-weighted
(+1, +3) 1.26*** 5.83 1.18*** 7.14 0.49*** 5.79
(+1, +6) 0.98*** 5.05 0.84*** 5.92 0.36*** 4.48
(+1, +12) 0.63*** 3.56 0.53*** 3.78 0.31*** 3.93
Panel B: Value-weighted
(+1, +3) 0.54* 1.76 0.42** 2.02 0.13* 1.91
(+1, +6) 0.46* 1.89 0.33** 2.06 0.07 1.34
(+1, +12) 0.32* 1.74 0.19* 1.65 0.09* 1.82
Figure 1
Post-Split Monthly Abnormal Return
Fig. 1. This figure shows the post-split monthly abnormal return from the first to the twelfth month relative to the split announcement. The monthly abnormal returns are measured by the alphas in calendar-time portfolio regressions using the five-factor model. The monthly portfolio returns are equal-weighted. Both ordinary least squares and weighted least squares approaches are used to run regressions.
-0.5 0.0 0.5 1.0 1.5 2.0 2.5
1 2 3 4 5 6 7 8 9 10 11 12
OLS WLS
Month Relative to Split Announcement
Abnormal Return (%)
Figure 2
Long-Run Performance of Portfolios based on Splits and SUEs
Fig. 2. This figure shows long-run performance of portfolios formed based on splits and standardized unexpected earnings (SUEs) over the period from January 1984 to December 2011. Split/high SUE – Non-split/low SUE is the long–short portfolio formed by buying split firms with high SUEs and selling size, book-to-market, and momentum matched non-split firms with low SUEs. Split – Non-split is the long–short portfolio formed by buying split firms and selling size, book-to-market, and momentum-matched non-split firms. High SUE – Low SUE represents the long–short portfolio formed by buying firms with high SUEs and selling firms with low SUEs. Splits and SUEs are based on split announcements and earnings surprises in the prior three months. The portfolios are rebalanced every month. We take the monthly value-weighted returns and compound them over time. For long–short portfolios, we compute the long-run return as the difference in compounded returns between long and short positions. We plot the long-run performance measured as the natural logarithm of 1 plus the long-run returns. The S&P 500 index return is spliced into the series when no eligible stocks in the monthly portfolio. Monthly returns are winsorized at the top and bottom 1%.
1984 1987 1990 1993 1996 1999 2002 2005 2008 2011
Split/high SUE – Non-split/low SUE Split – Non-split
High SUE – Low SUE Small
Value Market
Figure 3
Three-Month BHARs of the Long–Short Portfolio Formed by Buying Splits and High Standardized Unexpected Earnings (SUEs) and Selling Non-Splits and Low SUEs
Fig. 3. This figure shows three-month BHAR of the long–short portfolio over the period from 1984 to 2011. At the beginning of each calendar quarter, we form a portfolio by buying firms with split announcements and high SUEs (top two SUE quintiles) in the past quarter and selling non-split firms with low SUE (bottom two SUE quintiles) in the past quarter. The long–short portfolio is rebalanced every quarter.
-30
-20 -10 0 10 20 30 40
1984 1987 1990 1993 1996 1999 2002 2005 2008 2011
BHAR (%)
Appendix Table AI
Post-Split Abnormal Returns by Horizons
This table reports post-split abnormal returns (in %) over different investment horizons. We measure abnormal returns based on alphas of the regressions using a q-factor model consisting of the market factor, size factor, investment factor, and return-on-equity factor as in Hou, Xue and Zhang (2012). We form a calendar-time portfolio by including firms that had stock split announcements in the past 3, 6, or 12 months (Panel A) or firms that made split announcements in the x month before (Panel B) where x =1, 2, …, 12. Monthly portfolio returns are either equal-weighted (EW) or value-weighted (VW). Both ordinary least squares and weighted least squares methods are used to run regressions where the weight for weighted least squares is the number of splits included in the portfolio in each month. Calendar months with less than 10 stocks in the portfolio are excluded from the regression. Numbers in parentheses are t-statistics based on Newey–West (1987) standard errors. ***, **, and * indicate the significance levels of 1%, 5%, and 10%, respectively.
OLS WLS
Appendix Table AII
Long–Short and Long Only Trading Strategies Based on Splits and Earnings Surprises
This table shows the monthly abnormal returns of different trading strategies. The abnormal returns (in %) are measured by the alphas in calendar-time portfolio weighted least squares regressions using the q-factor model as in Hou, Xue, and Zhang (2012). Standardized unexpected earnings (SUEs) are sorted into quintiles by fiscal quarter.
We take the bottom two SUE quintiles as low SUE firms, and top two SUE quintiles as high SUE firms. In Panel A, Split/high SUE – Non-split/low SUE is the long–short trading strategy by buying split firms with high SUEs and selling size, book-to-market, and momentum-matched non-split firms with low SUEs. Split – Non-split is the trading strategy by buying split firms and selling size, book-to-market, and momentum matched non-split firms. High SUE – Low SUE represents the strategy by buying firms with high SUEs and selling firms with low SUEs. In Panel B, Split/high SUE is the trading strategy by buying split firms with high SUEs. Split includes all split firms in the sample. High SUE represents the strategy by buying firms with high SUEs. Firms are included in portfolio from the month following split or earnings announcement and held from 3 to 12 months. Monthly portfolio returns are equal-weighted or market value-equal-weighted. Calendar months with less than 10 stocks in the portfolio are excluded from the regression. Numbers in parentheses are t-statistics based on the standard errors using Newey–West (1987) method.
***, **, and * indicate the significance levels of 1%, 5%, and 10%, respectively.
Panel A: Long-short Trading Strategies based on Splits and Earnings Surprises
Split/high SUE – Split – Non-split High SUE – Low SUE
Panel B: Long Only Trading Strategies
Split/high SUE Split High SUE