The Puzzle of Frequent and Large Issues of Debt and Equity

Abstract More frequent, larger, and more recent debt and equity issues in the prior 3 fiscal years are followed by lower stock returns in the subsequent year. The intercept of a q-factor calendar-time regression for the value-weighted (VW) portfolio of firms with at least 3 large issues is −0.63% per month (t-stat. = −4.31). Purging the factor returns of recent issuers increases the magnitude of the estimated underperformance following frequent equity issues. A VW Fama–MacBeth regression shows that firms with 3 equity issues underperform nonissuers by 0.65% per month (t-stat. = −2.65). Earnings announcement returns are low following frequent issues, especially equity issues.


I. Introduction
In this article, we show that frequent and large issues of debt or equity in the prior 3 fiscal years are followed by low average stock returns in the subsequent year.The value-weighted (VW) averages of raw returns during the next year are 12.2% for firms with no significant external financing in the prior 3 fiscal years, 10.8% for firms that have issued debt or equity only once, and only 3.9% for firms that have issued debt or equity at least 3 times.For firms that have at least 3 large issues, the VW average raw return is even lower, at À1.2%.
We show that, using the Hou, Xue, and Zhang (2015) q-factor model, the larger the issue size and the more frequent the issuance, the greater the underperformance.Our results are similar if we use the Fama-French (2015) 5-factor model as the benchmark.The q-factor model time-series regression intercept during 1975-2018 decreases from 0.12% per month (t-stat.= 2.88) in the subsequent year for the VW portfolio of firms with no external financing in the prior 3 years to À0.00%, À0.32%, and À0.63% per month (t-stat.= À0.08,À2.54, and À4.31), for the VW portfolios of firms with 1 debt or equity issue, 3 or more equity or debt issues, and at least 3 large issues, respectively.We call these patterns the puzzle of frequent and large issues of debt and equity.
We also find that more recent issues are followed by lower average stock returns than issues from several years ago.In other words, the abnormal returns decay over time.The VW portfolio of firms that issued equity in fiscal year t has a q-factor alpha of À0.37% per month (t-stat.= À3.11) in t þ 1, but the VW portfolio of firms that issued equity 1 or more times in the prior 3 years has an insignificant q-factor alpha of only À0.08% per month in the subsequent year, suggesting that the use of the 3-year post-event window in many existing studies is less able to detect abnormal returns than the use of the 1-year post-event window.
An economically important proportion of firms engage in substantial external financing activity over a 3-year period.Over 10% of all firm years are preceded by at least 3 issues of debt or equity in the prior 3 years, with a firm classified as an issuer of equity or debt in a (fiscal) year if the equity or debt issue exceeds 5% of assets and 3% of the market cap at the beginning of the year.Almost 6% of all firm years are preceded by at least 3 large issues, with a large issue defined as exceeding 10% of assets and 3% of the market cap.We measure the issue size using Compustat Statements of Cash Flow information, and thus do not include debt acquired in acquisitions or stock issued in stock-financed acquisitions, but do include increases in bank loans as debt issues and private investments in public equity (PIPEs) as equity issues.
Equity issues, on average, are followed by lower raw returns than debt issues.Equity and debt issuers, however, have different characteristics.Although both invest heavily, debt issuers are much more likely to be profitable than equity issuers.Using Fama-MacBeth (1973) regressions that control for several characteristics, we show that equity issues are followed by lower abnormal returns than debt issues.A Fama-MacBeth regression using weighted least squares (WLS) with value (market cap) weights shows that firms that issued equity in each of the prior 3 years underperform nonissuers by 0.65% per month (t-stat.= À2.65), in contrast to insignificant outperformance of 0.02% per month for firms with 3 debt issues.Our calendar-time q-factor regressions, however, provide mixed evidence on the relation between security type and subsequent abnormal stock returns.
Although we follow the practice in the literature of reporting the results of time-series factor regressions, factor regressions using the Hou et al. (2015) or Fama and French (2015) factors have intercepts, which are the abnormal performance measure, that are biased toward 0 in our context.The reason is that firms with low book-to-market, small size, low profitability, and high investment are disproportionately equity issuers.Thus, to some degree, the low returns on issuing firms are being used to explain the low returns on issuing firms.To remove this bias, following Loughran and Ritter (2000), we construct "purged q-factors" that include only stocks that have not issued debt or equity during the prior 3 years.Using these purged factors, we report q-factor model intercepts that are approximately 15 basis points per month more negative for frequent equity issuers, although the purging makes little difference for frequent debt issuers.For the VW portfolio of firms that issued equity in both of the 2 prior fiscal years, the q-factor regression intercept increases in magnitude from À0.61% per month (t-stat.= À3.42)before purging to À0.78% per month (t-stat.= À4.33) after purging.
Our article is not the first to examine abnormal returns after multiple security issues.Using a sample of U.S. firms issuing in 1980-2005, Billett, Flannery, andGarfinkel (BFG) (2011) find that firms that issue multiple types of securities have lower long-run stock performance than those that issue just 1 type of security.We find that it is not the number of types of securities that are issued that matters, but the number of issues, and the recency and size of each issue.
Although there is widespread agreement among researchers that stock returns following equity issues tend to be low, there is conflicting evidence in the extant literature on whether, after controlling for the characteristics of issuing firms using time-series multifactor regressions, there are negative abnormal returns.Much of this literature focuses on including additional factors in factor regressions, but does not emphasize the importance of issue size, frequency, recency, or factor purging.We provide strong evidence that the ability to detect abnormal returns following issuance depends on whether firms that only occasionally raise a small amount of capital are included in the issuer portfolio, how long issuers stay in the portfolio, and whether the factors are purged of recent issuers.In other words, methodological choices affect the power to detect abnormal returns.
We form portfolios constructed at the end of the fourth month after the end of a firm's fiscal year using statements of cash flow information, with the delay motivated by the time that it takes for companies to make their financial statements public.We thus add these companies to the issuer portfolio 10 months after the issuance, on average, assuming that issuance, on average, occurs in the middle of the fiscal year.Ritter (2021) shows on his website that the low returns on stocks after initial public offerings (IPOs) and seasoned equity offerings (SEOs) do not start until about 6 months after issuance.Consistent with our finding on issue recency, he also shows that the low stock returns do not persist for much more than 2 years.Thus, if firms are added to a portfolio of issuers too quickly, or stay in the portfolio for too long, the quantitative magnitude and statistical significance of the average underperformance of the portfolio are reduced.As is also done in many other articles in the literature, Bessembinder and Zhang (2013) construct portfolios of IPOs and SEOs immediately after issuance and keep stocks in the portfolios for 5 years, with both of these timings moving their abnormal returns toward 0.
We do not directly address why there are negative abnormal returns on firms that are frequent, large, and recent issuers.We do provide strong evidence that more frequent and larger issues, especially equity issues, are associated with lower stock returns around the earnings announcements made in the subsequent year.Riskbased theories cannot easily explain the magnitude of the negative abnormal earnings announcement returns (EARs).Hou et al. (2020) suggest that most of the 452 anomalies that they examine are driven by microcap stocks.Importantly, microcap stocks do not drive our major results.Furthermore, the underperformance of frequent and large issuers has not weakened over time.For example, the q-factor regression intercepts for the VW portfolio of firms with at least 3 large issues are both À0.63% per month (t-stat.= À3.22 and À2.74, respectively) during 1975-1996 and 1997-2018.The existing literature on long-run performance following external financing events has focused almost exclusively on which factors or characteristics to control for.We add 4 new findings to the literature.First, we find that, even after controlling for investment and profitability, frequent and large issuers underperform nonissuers by economically and statistically significant amounts.Our second finding is that the abnormal return is more negative in the first year after the fiscal year of the security issuance than in the second or third year.Our third finding is that purging the factors increases the magnitude of the estimated underperformance following frequent equity issues in calendar-time q-factor model regressions.Finally, we explain how methodological choices affect the ability of different studies in the existing literature to find abnormal returns following external financing events.

II. Sample Construction and Distribution
Our sample starts with nonfinancial and nonutility firms with information from Compustat and CRSP.We require cash flow information over the 3 fiscal years from t À 2 to t.All returns are from CRSP, and include capital gains, dividends, and other distributions.Because the cash flow information is available only from fiscal year 1971 and CRSP does not include returns on Nasdaq-listed stocks before Dec. 1972, our final sample starts from fiscal year 1974.Since we examine stock returns from month 5 to month 16 after each fiscal year, our sample period ends at fiscal year 2017.We require net equity and net debt issue amounts in year t, t À 1, and t À 2, as well as the book value of assets and the market value of equity at the beginning of each year. 1 We further drop firm-year observations for which the book value of equity at the end of year t or operating income before depreciation in year t has a missing value.Our final sample includes 141,064 firm-year observations from fiscal years 1974-2017.
A firm is defined to have an equity issue or a debt issue in a year if the net equity issue amount or the net debt issue amount in the year is at least 5% of the book value of assets and at least 3% of the market value of equity at the beginning of the year.2A firm is defined to have a large equity issue or a large debt issue in a year if the net equity issue amount or the net debt issue amount in the year is at least 10% of the book value of assets and at least 3% of the market value of equity at the beginning of the year.Because statements of cash flow are used, a firm making a large acquisition financed by issuing stock to the shareholders of the target firm would not necessarily be classified as an equity issue, nor would a firm that increases its book value of equity by retaining earnings.Our definition of debt issues includes both increases in bonds and increases in bank loans, although bank loans are technically not securities.
We use security issuance information in years t, t À 1, and t À 2 to assign a firm into an issuance category and examine its stock return in the subsequent year from the fifth month after the end of t.For example, assume that a firm has an equity issue in year t À 2 and another equity issue in t, but no equity issue in t À 1, t þ 1, and t þ 2. The firm will be defined as issuing equity 2 times for the 3-year window

III. Average Firm Characteristics and Post-Issuance Buy-and-Hold Stock Returns
Table 2 reports the mean firm characteristics (see Appendix A for the definitions).Panel A reports the means categorized by the number of equity issues in the prior 3 years.Firms with more equity issues, on average, are smaller, and have a higher market-to-book ratio and much faster asset growth.More equity issues are also associated with much lower operating income before depreciation divided by the book value of total assets (OIBD÷ASSETS) and much lower return on equity (ROE).Although we only report means, DeAngelo, DeAngelo, and Stulz ((2010), Table 2) document that there is a large amount of heterogeneity among firms conducting SEOs, confirmed in Table IA-1 in the Supplementary Material.Panel B reports the means categorized by the number of debt issues.More debt issues are associated with much larger investment, as measured by the asset growth rate.The number of debt issues is not strongly related to firm size, market-to-book, or profitability.Comparing Panels A and B of Table 2, although equity issuers and debt issuers are quite different in every other characteristic, they both invest heavily.
Panel C of Table 2 reports the average firm characteristics sorted by the total number of issues, from 0 to a maximum of 6.Also reported are the means conditional on at least 3, or at least 4, issues in the prior 3 years.The number of issues has a strong and positive relation with investment.
Panel D of Table 2 reports the average firm characteristics double-sorted by the number of equity issues and the number of debt issues.Conditional on the number of debt issues, the number of equity issues is positively related to the marketto-book ratio and investment, and negatively related to profitability.Conditional on the number of equity issues, the number of debt issues is positively related to   .A firm is defined to have an equity issue (a debt issue) in a year if ΔEQUITY (ΔDEBT) in the year is at least 5% of the book value of beginning-of-year assets and at least 3% of the market value of beginning-of-year equity.A firm is defined to have a large equity issue (a large debt issue) in a year if ΔEQUITY (ΔDEBT) in the year is at least 10% of the book value of beginning-of-year assets and at least 3% of the market value of beginning-of-year equity.No. of equity (debt) issues equals the number of fiscal years with equity (debt) issues in fiscal years t À 2, t À 1, and t.No. of issues equals the total number of equity or debt issues in fiscal years t À 2, t À 1, and t.No. of large equity (or large debt) issues equals the number of fiscal years with large equity (debt) issues in fiscal years t À 2, t À 1, and t.No. of large issues equals the total number of large equity or large debt issues in fiscal years t À 2, t À 1, and t.See Appendix A for the definitions of ΔEQUITY and ΔDEBT.investment.On average, firms with 3 equity issues and 0 debt issues have the highest market-to-book ratio and are the smallest and the least profitable.

Panel
In the Supplementary Material, Table IA-4 reports the mean firm characteristics sorted by the number of large issues.Relative to security issuers in Table 2, large security issues in Table IA-4 are generally slightly smaller and less profitable, grow more rapidly, and have a higher market-to-book ratio.Other than that, the patterns in Table IA-4 are similar to the patterns in Table 2.
Table 3 reports the mean post-issuance stock returns.Because microcaps have a large influence on equal-weighted (EW) averages while large caps have a    Huang and Ritter 177 large influence on VWaverages, we report both EWand VWaverages.We also report both 1-year and 3-year buy-and-hold returns in the table, but will focus on 1-year returns in the following discussions.We measure the returns starting at the end of 4 months after the end of fiscal year t (May 1 for a Dec. 31 fiscal year) in order to allow the release of financial statements for year t before portfolios are formed.Panel A of Table 3 reports the mean returns sorted by the number of equity issues in the previous 3 years.For firms with 0-3 equity issues in the prior 3 years, the EW mean 1-year buy-and-hold returns in the following year are 18.0%, 11.3%, 2.3%, and À7.3%, respectively, a spread of 25.3% between nonissuers and 3-time issuers of equity.The very large spread of 25.3% and the very low return of À7.3% per year for this last category are unlikely to be explained by riskbased theories.The negative correlation between equity issuance frequency and subsequent stock returns is not limited to microcaps.The VW mean buy-and-hold returns in the following year have a spread of 16.1% between the nonissuers and 3-time issuers of equity, suggesting that the pattern is weaker but also exists for nonmicrocaps.The corresponding EW and VW mean market-adjusted buyand-hold returns in the following year have the spreads of 22.9% and 13.4%, respectively, between the nonissuers and 3-time issuers of equity.The 3-year spreads are even wider.3Panel B of Table 3 reports the EW and VW mean buy-and-hold returns sorted by the number of debt issues.The EW results suggest that more frequent debt issues are followed by low stock returns.However, the VW results show only a weak negative relation between debt issuance frequency and subsequent stock returns.
Whether using EW or VW raw returns or market-adjusted returns, the spread in 1-year subsequent returns between the most frequent issuers and nonissuers is more than twice as large when sorted on equity issuance as the spread when sorted by debt issuance.The similarity of the spreads when either raw returns or market-adjusted returns are used suggests that most of the action is due to abnormal returns rather than the ability to time general movements in debt and equity markets.
Panel C of Table 3 reports the EW and VW average buy-and-hold returns sorted by the number of issues, with frequent issuers generally having lower returns.Firms with no debt or equity issue in the previous 3 years have an EW average raw return of 18.8% and a VW average raw return of 12.2% in the following year.In contrast, firms with 6 issues have an EW mean raw return of À13.0% and a VW mean raw return of À12.8% in the subsequent year. 4The spread in the subsequent EW mean 1-year raw returns between firms with 0 and 6 issues is a stunning 31.8%.The spread in the EW mean 3-year buy-and-hold return for firms with 0 and 6 issues is 73.2%!Panel D of Table 3 reports the average returns double-sorted by both the number of equity issues and the number of debt issues.Conditional on the number of debt issues, more equity issues are generally followed by lower stock returns.Conditional on the number of equity issues, more debt issues are generally followed by lower stock returns.
Panels E-H of Table 3 report the EW and VW mean returns following large issues, which are a subset of all issues.The patterns for large issues are often more extreme than those for all issues.Because large issuers are more likely to be small firms that are unprofitable (at least for the equity issuers) with aggressive investment, in the next section, we will control for these characteristics in multifactor time-series regressions.

A. Stock Returns Following Equity Issues
Table 4 reports the monthly excess returns and calendar-time factor regression results for portfolios formed on the basis of the frequency of equity issues for all issues and for large issues, using monthly VW and EW returns from Jan. 1975 to Dec. 2018. 5We report the coefficients from Hou et al.'s (2015) q-factor model and Fama and French's (2015) 5-factor model (see Appendix B for the models).The multifactor models allow us to test whether there are independent issuer effects after controlling for cross-sectional patterns related to size, value, investment, and profitability. 6anel A of Table 4 reports the monthly abnormal returns for portfolios sorted by the number of equity issues.Beginning in the fifth month after the end of its fiscal year, a firm is in a portfolio for 12 months or until its delisting date, if this date is earlier.For example, a retailer with a fiscal year-end in Jan. 2012 would be in the portfolio from June 2012 to May 2013.A coefficient is highlighted in bold to signify that it is statistically different from the corresponding coefficient in the first column (no issuance) of the same subpanel at the 5% level.
Consistent with the results in Table 3, the VW and EW average monthly excess returns on portfolio p, R pt -R ft , decrease as the number of equity issues increases.The VW average excess return is 0.75% per month for the portfolio of firms with no equity issue in years t -2 to t and À0.23% per month for the portfolio of firms with 3 equity issues, a spread of nearly 1%.This spread is significantly different from 0 at the 5% level.Correspondingly, the spread in the EW average monthly excess returns for the 2 portfolios is 1.80%.
For the VW portfolio of firms with no equity issuance in the past 3 years, the q-factor intercept is a significantly positive 0.07% per month, but the 5-factor intercept is indistinguishable from 0. For the EW portfolio of firms with no equity issuance, the q-factor and 5-factor intercepts are 0.45% and 0.28% per month, respectively.
For the VW or EW portfolio of firms that issued equity only once in the past 3 years, a category that represents 66.7% of firms that have issued equity 1 or more times, the q-factor model or 5-factor model intercept is close to 0, consistent with studies that find no abnormal returns for equity issuing firms in multifactor models or Fama-MacBeth regressions that control for important firm characteristics, such as Lyandres, Sun, and Zhang (2008), Bessembinder andZhang (2013), andBessembinder, Cooper, andZhang (2019).However, the q-factor intercepts are À0.40% per month for the VW portfolio of firms with 2 or more equity issues and À0.46% per month for the VW portfolio of firms with 3 equity issues.The corresponding 5-factor intercepts are À0.49% and À0.74% per month,

Calendar-Time Factor Regression Results: Equity Issues
The dependent variable in Table 4 is the portfolio monthly value-weighted (VW) or equal-weighted (EW) percentage excess return from Jan. 1975 to Dec. 2018, with equity issues beginning in fiscal 1971.If there are fewer than 10 stocks in the portfolio in a month, the corresponding observation is dropped.In Panel C, equity issues of (0,0,1), e.g., denotes that the firm conducted no equity issue in t and t À 1 but did an equity issue in t À 2. t-Statistics using a Newey-West correction with 3 lags are in parentheses, with *, **, and *** signifying statistical significance at the 10%, 5%, and 1% significance levels, respectively.A coefficient in bold is statistically different from the corresponding coefficient in the first column (no issuance) of the same panel (or subpanel, if available), at the 5% significance level.See Appendices A and B and Table 1 for variable and factor model descriptions.
Panel A. Frequency of Equity Issues (1975-2018, no. of Average Monthly Excess Return on the Portfolio    Huang and Ritter 181 respectively.The spreads in the intercepts between the VW portfolios of nonissuers and frequent equity issuers (≥2) are 0.47% in the q-factor model and 0.51% in the 5-factor model.
In both the q-factor and 5-factor models, the slope for the size factor is strongly positive for firms with 1 or more equity issues, consistent with our Table 2 results that equity issuers tend to be smaller than other firms.The slope for the value factor in the 5-factor model is negative for equity issuers, suggesting that equity issuers tend to be growth firms rather than value firms.The negative slopes (factor loadings) on the q-factor model's ROE factor, b ROE , or on the 5-factor model's operating profitability factor, r, for equity issuers are consistent with our Table 2 findings of low profitability for equity issuers.The negative slopes on the investment factors, b I/A and c, suggest that equity issuers invest more than other firms.Surprisingly, the slope coefficients on the investment factors do not differ much between firms that issued equity once vs. 2 or more times in the past 3 years.
Motives for large equity issues could include large investment needs (including paying for R&D expenses) and market timing.Panel B of   1,0,1),  (1,1,0),  or (1,1,1   results of the q-factor model regressions for the portfolios sorted by the number of large equity issues.Results for the Fama-French 5-factor model are qualitatively similar, and are reported in Table IA-5 in the Supplementary Material.The VW portfolio of firms with 3 large equity issues has a q-factor intercept of À0.59%.
Although the intercept of À0.32% for the EW portfolio of firms with 3 large equity issues is not statistically different from 0, the spread between the EW portfolios of firms with 3 large equity issues and firms with no large equity issue is a statistically significant À0.75% per month.For firms conducting 2 or more equity issues in the prior 3 years, the abnormal returns are generally more negative for large issues than for all issues.
Our results in Panels A and B of Table 4 show that firms with frequent and large equity issues have negative slope coefficients on the investment and ROE factors.As a result, the abnormal returns following frequent and large equity issues are less anomalous (closer to 0) once we control for the factors.We now address how much the recency of issuance matters.
Panel C of Table 4 reports the average monthly excess returns and q-factor regression results for the portfolios sorted by the frequency and recency of equity issues in years t, t -1, and t -2.The first column in Panel C, with no equity issues in the prior 3 years, is the same as the first column in Panel A. In the second to last column of Panel C, we also pool the firm years with (1,1,0) with (1,1,1) to have a better diversified portfolio.These are the firms with equity issues in both of the last 2 years.As shown in Table IA-2 in the Supplementary Material, the pooled portfolio includes 6,189 firm years, with an average of almost 141 stocks in the portfolio each month.In the second to last column, the VW and EW intercepts of the q-factor model are À0.61%, and À0.28% per month, respectively.More recent (e.g., (1,1,0) relative to (0,1,1)) equity issues are followed by lower returns in year t þ 1, indicating a gradual diminution of abnormal returns: The VW intercept of the q-factor model is À0.58% per month in the column of (1,1,0) and À0.26% in the column of (0,1,1).The last column shows that an equity issue in t is followed by a q-factor abnormal VW return of À0.37% per month in t þ 1 with a t-statistic of À3.11.In comparison, the q-factor VW intercept in Panel A for firms with at least 1 equity issue in the prior 3 years is À0.08% per month with a t-statistic of only À0.83.Thus, the use of the 3-year post-event window in many existing studies has less power to detect abnormal returns than the use of the 1-year post-event window.These findings suggest that if low stock returns following equity issues reflect a low required rate of return, the low rate is only temporary.

B. Stock Returns Following Debt Issues
Table 5 reports the results from calendar-time factor regressions of VW and EW portfolio returns following debt issues.In Panel A, the average monthly excess return does not vary substantially across the columns sorted by the number of debt issues for the VW portfolios, although it substantially decreases with the number of debt issues for the EW portfolios.This pattern suggests that small firms, but not big firms, that issue debt have low subsequent returns.For the VW portfolio of firms with no debt issues, the intercepts of the q-factor model and the 5-factor model are reliably positive 0.12% and 0.07%, respectively.However, for the VW portfolio of

Calendar-Time Factor Regression Results: Debt Issues
The dependent variable in Table 5 is the portfolio monthly value-weighted (VW) or equal-weighted (EW) percentage excess return from Jan. 1975 to Dec. 2018, with debt issues beginning in fiscal year 1971.If there are fewer than 10 stocks in the portfolio in a month, the corresponding observation is dropped.In Panel C, debt issues of (0,0,1), for example, denote that the firm conducted no debt issue in t and t À 1 but did a debt issue in t À 2. t-Statistics using a Newey-West correction with 3 lags are in parentheses, with *, **, and *** signifying statistical significance at the 10%, 5%, and 1% significance levels, respectively.A coefficient in bold is statistically different from the corresponding coefficient in the first column of the same panel (or subpanel, if available), at the 5% significance level.See Appendices A and B and Table 1 for variable and factor model descriptions.
Panel A. Frequency of Debt Issues (1975-2018, no. of months = 528) Average Monthly Excess Return on the Portfolio    firms with at least 2 debt issues, the q-factor and 5-factor intercepts are À0.12% and À0.20%, respectively, both of which are statistically different from the intercepts for the VW portfolio of firms with no debt issues.In comparison, the corresponding q-factor and 5-factor intercepts for frequent equity issuers in Panel A of Table 4 are À0.40% and À0.49%.For the VW portfolios, the slopes on the operating profitability factor or the ROE factor are positive or close to 0 for frequent debt issuers, in contrast to the negative slopes for frequent equity issuers.These findings are consistent with the summary statistics of Table 2, which show that equity issuers are less profitable than debt issuers, and are intuitively plausible: Profitable firms find it much easier to borrow than money-losing firms.Panel B of Table 5 reports the results of q-factor regressions sorted by the number of large debt issues, with 5-factor results reported in Table IA-6 in the Supplementary Material.There is evidence of reliable underperformance following frequent large debt issues.For the VW portfolio of firms with at least 2 large debt issues, the q-factor intercept is À0.29%.The intercept is a much lower À0.83% for the VW portfolio of firms with 3 large debt issues.Panel C of Table 5 reports the q-factor results sorted by the frequency and recency of debt issues.There is weak evidence that more recent debt issues are associated with more negative abnormal stock returns.For example, the VW intercept of the q-factor model is À0.23% per month in the (1,1,0) column and À0.15% in the (0,1,1) column.The second to last column, which pools (1,1,0) and (1,1,1) firm years, includes 9,143 firm years, and reports q-factor VW and EW intercepts of À0.20% and À0.34%, respectively.For the VW portfolios, the slopes of the ROE factor, b ROE , are slightly positive in most cases.The spread in the VW intercepts between columns (0,0,0) and the second to last column is 0.32%.As shown in the last column, a debt issue in t is followed by a q-factor abnormal return of À0.11% per month in t þ 1, with a t-statistic of À1.70.In comparison, the q-factor intercept of À0.06% in Panel A for firms that issued debt at least once in the prior 3 years has a t-statistic of only À1.18.These results again suggest that using a 1-year post-issuance window is better able to detect abnormal returns than using a 3-year post-issuance window.
Our results in Tables 4 (equity) and 5 (debt), showing that the abnormal returns following issuance are lower the more frequent, the larger, and the more recent are the issues, have implications for the power of various specifications to detect abnormal returns.Figure 1 provides a summary of the q-factor model intercepts for the portfolios sorted by issue frequency, size, and recency.In most of the analysis of BFG, e.g., the effect of an issue has been specified to last up to 71 months.Bessembinder and Zhang ((2013), Panel E of Table 4) use portfolios composed of firms that conducted an event within the prior 60 months.Their abnormal returns would presumably be stronger if they used a shorter window.These articles also do not address the importance of the size of each issue in explaining subsequent stock returns.

C. Factor Contamination and Purging
The intercepts in the multifactor models that we, and other authors, have used are biased toward 0 because of what Loughran and Ritter (2000) refer to as factor contamination.As our Table 1 shows, almost 26% of all firm years are preceded by at least 1 equity issue and over 45% of all firm years are preceded by at least 1 debt issue.Our Table 2 shows that both debt and equity issuers, on average, invest heavily, and equity issuers, on average, have low profitability.Thus, the portfolio of firms with heavy investment and the portfolio of firms with low profitability are composed of many equity issuers.These portfolios are used to construct the investment and profitability factors.
To construct the purged q-factor series, we start with replicating Hou et al.'s (2015) q-factor series.From 1975 to 2018, the correlation between our replicated and their original size factor monthly series is 0.996, the correlation between the replicated and original ROE factor monthly series is 0.992, and the correlation between the replicated and original asset growth factor monthly series is 0.981.The purged q-factor model produces more negative intercepts with VW returns following equity issues than the nonpurged q-factor model, suggesting that without purging the factors, the intercepts are biased toward 0.7 This is as expected, since without purging the factors of equity issuing firms, low returns on equity issuers are being used to explain low returns on equity issuers.For firms with at least 2 equity issues, the VW intercept is À0.40% per month (t-stat.= À2.85) for the q-factor model and À0.53% per month (t-stat.= À3.79) with the purged q-factor model.For firms with 3 equity issues, the VW intercept is À0.61% per month (t-stat.= À3.42) for the q-factor model and À0.78% per month (t-stat.= À4.33) with the purged q-factor model.

FIGURE 1
q-Factor Intercepts Sorted by Issue Frequency, Size, andRecency (1975-2018) Figure 1 shows the intercepts (monthly percentage abnormal returns) of the q-factor model for the value-weighted portfolios sorted by the firm's equity or debt issue frequency, size, and recency in the prior 3 years.A firm is defined to have an equity issue (a debt issue) in a year if ΔEQUITY (ΔDEBT) in the year is at least 5% of the book value of beginning-of-year assets and at least 3% of the market value of beginning-of-year equity.A firm is defined to have a large equity (debt) issue using the same definitions except with 10% replacing 5%.In Graphs A and B, the number of equity (debt) issues on the horizontal axis equals the number of fiscal years with equity (debt) issues in fiscal years t À 2, t À 1, and t.In Graphs C and D, (0,0,1) for equity issues, e.g., denotes that the firm had no equity issue in t and t À 1 but had an equity issue in t À 2. Unlike the equity issue regressions, the debt issue regressions in Table 6 show a much smaller difference between the q-factor and purged q-factor VW results.The lack of an effect for debt issues can be attributed to the low absolute values of factor loadings on all but the market factor for firms that do or do not issue debt.Because the slopes are close to 0, whether the factor returns are contaminated with the returns on debt issues has little effect on the intercepts.

D. Stock Returns Following Equity and Debt Issues
So far we have examined equity (Table 4) and debt (Table 5) issues separately.Table 7 examines equity and debt issues together and evaluates the importance of the total number of issues.In Panels A and B of Table 7, we do not distinguish between equity and debt, and report both VW and EW results.In Panels C and D of Table 7, we distinguish between equity and debt, and, in order to save space, report only VW results.
Panel A of Table 7 reports the results sorted by the number of issues.For firms with no issues in the prior 3 years, the VW average excess return is 0.75% per Calendar-Time Factor Regression Value-Weighted Results: Purged q-Factors Table 6 reports the value-weighted (VW) results using purged q-factors.To construct the purged q-factor series, we start with replicating Hou et al.'s (2015), ( 2019), ( 2020), (2021) q-factor series.The purged size, asset growth, and ROE factors used in columns 1-3 are computed after purging stocks of firms with 1 or more equity issues in years t À 2 to t, and those in columns 4-6 are computed after purging stocks of firms with 1 or more debt issues in years t À 2 to t.The market factor is not purged.The dependent variable is the monthly VW percentage portfolio return minus the risk-free rate.If there are fewer than 10 stocks in the portfolio in a month, the corresponding observation is dropped.To facilitate comparison, Panel A reports the nonpurged results.See Appendices A and B and Table 1 for variable and q-factor model descriptions.t-Statistics using a Newey-West correction with 3 lags are in parentheses, with *, **, and *** signifying statistical significance at the 10%, 5%, and 1% significance levels, respectively.month, while the EW average excess return is a much larger 1.36% per month.For firms with at least 3 issues in the prior 3 years, the VW and EW average excess returns are 0.43% and 0.31% per month, respectively.These results are generally consistent with the buy-and-hold results in Panel C of Table 3.

No. of Equity
In Panel A of Table 7, there is robust evidence that more frequent security issues are followed by more negative abnormal stock returns.The q-factor intercepts are positive and statistically significant for both VW and EW portfolios of firms with no security issues.The intercepts are statistically insignificant for the

Calendar-Time Factor Regression Results: Equity and Debt Issues Combined
The number of issues in Table 7 represents the number of the last 3 fiscal years in which a debt or equity issue occurred, with a maximum of 6 potential issues.The dependent variable is the monthly percentage value-weighted (VW) or equal-weighted (EW) portfolio return minus the risk-free rate.If there are fewer than 10 stocks in the portfolio in a month, the corresponding observation is dropped.t-Statistics using a Newey-West correction with 3 lags are in parentheses, with *, **, and *** signifying statistical significance at the 10%, 5%, and 1% significance levels, respectively.A coefficient in bold is statistically different from the corresponding coefficient in the first column of the same subpanel at the 5% significance level.See Appendices A and B and Table 1 for variable and factor model descriptions.See Table 1    VW portfolio of firms with 1 security issue or the VW portfolio of firms with 2 issues.For firms with at least 3 issues, the VW and EW intercepts are À0.32% and À0.34%, respectively.Firms with at least 4 issues do worse.For these firms, the VW and EW intercepts are À0.44% and À0.66%, respectively.
In Panel A of Table 7, the differences in the VW and EW average excess returns, respectively, between the first column (=0) and the last column (≥4) are 0.59% and 1.53% per month.In comparison, the differences in the VW and EW q-factor intercepts, respectively, between the first column (=0) and the last column (≥4) are 0.56% and 1.26% per month.There is a substantial spread in abnormal   returns between nonissuers and those with at least 4 securities issues after controlling for the q-factors.Table IA-8 in the Supplementary Material reports the market model, 3-factor model, and 5-factor model results.Given the emphasis in the recent literature about using multifactor models to calculate abnormal returns, it is surprising how little difference the choice of model makes in our results.Panel B of Table 7 reports the results for large issues.There is strong evidence that firms with more frequent large security issues have lower subsequent performance.As shown in Panel A of Table 1, 8,843 firm years (6.3% of the sample) are preceded by at least 3 large security issues in the previous 3 years.For these firms, the VW and EW intercepts of the q-factor model are À0.63% and À0.53%, respectively.
BFG find that the number of different types of securities issued is related to post-issuance stock returns.For example, a firm that issues equity via both an IPO and an SEO, and issues debt via both a debt issue and increasing its bank loans, would be deemed to have engaged in 4 external financings.In their Table 3, they report abnormal returns that are insignificantly different from 0 if there has been only 1 external financing event in the prior 36 months, but reliably negative abnormal returns if there have been 2 or more different types of financing.
To see whether our finding on the number of issues is driven by the number of types of securities, in Panel C of Table 7, we distinguish between the number of issues and the number of types of securities.In Panel C1, we estimate time-series regressions for the VW portfolios sorted by the number of types of securities, regardless of the number of issues.The results in Panel C1 are generally consistent with those in BFG.Firms that issue more types of securities are associated with lower abnormal stock returns.In Panel C2, we examine the relation between the number of issues and stock returns, conditional on issuing only 1 type of security.In Panel C3, we examine the relation between the number of issues and stock returns, conditional on issuing both types of securities.There is some evidence of a negative relation between the number of issues and future stock returns, even after controlling for the number of types of securities.
The number of issues is 2 in both columns 5 and 7, but firms in column 5 issue 1 type of security (only debt or only equity), whereas firms in column 7 have 1 debt issue and 1 equity issue.Column 7 has a slightly higher q-factor intercept than column 5, suggesting that firms with 1 debt issue and 1 equity issue do not necessarily underperform those with only 2 debt issues or only 2 equity issues, inconsistent with BFG's conclusion.The number of issues is 3 in both columns 6 and 8, but firms in column 6 issue 1 type of security (only debt or only equity), whereas firms in column 8 issue both debt and equity.Consistent with BFG, column 8 shows a lower intercept than column 6.
After controlling for the number of types of securities, BFG find that security type is not reliably related to long-run performance.However, Baker and Wurgler (2000) and Lewis and Tan (2016) find that equity issues are followed by lower stock returns than debt issues.To shed light on the debate, Panel D of Table 7 examines the relation between security type and subsequent stock returns, after controlling for the total number of issues.
Panel D of Table 7 shows that the number of equity issues is negatively related to future VW average monthly excess stock returns after controlling for the total Huang and Ritter 191 https://doi.org/10.1017/S0022109021000636Published online by Cambridge University Press number of issues.However, when the q-factor model is used, the evidence on the relation between security type and VW abnormal stock returns is mixed.When the total number of issues equals 1 or 2, there is generally a positive relation between the number of equity issues and the q-factor model intercepts for the VW portfolios.For example, the q-factor model intercept for the VW portfolio of firms with only 2 equity issues is þ0.16%, which is higher than the intercept of 0.02% for the VW portfolio of firms with only 2 debt issues, although they are not statistically different.8However, when the total number of issues is 3, there is generally a negative relation between the number of equity issues and abnormal returns.

E. Stock Returns in 2 Subperiods
As is well known, abnormal returns often become less anomalous after the publication of an anomaly (e.g., McLean and Pontiff (2016)).To understand whether our results continue to hold after the publications on negative abnormal returns following equity issues (e.g., Loughran and Ritter (1995), Spiess and Affleck-Graves (1995)), we estimate the factor model regressions separately for the subperiods of 1975-1996 and 1997-2018 (the calendar year of the stock return month).Table 8 reports the VW average excess returns and q-factor results, with

Calendar-Time Factor Regression Value-Weighted Results: Subperiod Analysis
The dependent variable in Table 8 is the monthly percentage value-weighted (VW) portfolio return minus the risk-free rate.If there are fewer than 10 stocks in the portfolio in a month, the corresponding observation is dropped.t-Statistics using a Newey-West correction with 3 lags are in parentheses, with *, **, and *** signifying statistical significance at the 10%, 5%, and 1% significance levels, respectively.See Appendices A and B and Table 1 for variable and factor model descriptions.1975-1996 (no. of   results for other factor models and EW results reported in Table IA-9 in the Supplementary Material.Frequent equity issues are generally associated with low subsequent year abnormal stock returns in both subperiods.For the portfolio of firms with at least 3 large issues (whether debt or equity), the q-factor intercepts are both À0.63% (t-stat.= À3.22 and À2.74, respectively) during 1975-1996 and during 1997-2018.
Our major results generally hold in both subperiods.Fu and Huang ((2016), Table 1), who examine firms conducting SEOs rather than all equity issuers report calendar-time regressions with VW abnormal returns of À16.20% per year during 1980-2002 and À0.36% per year during 2003-2012.In Tables IA-10 and IA-11 in the Supplementary Material, we confirm that the abnormal returns on frequent equity issuers were close to 0 during the 10 years from 2003 to 2012, but are generally similar to those for our overall sample period when either 2000-2002 or 2013-2018 is added to these 10 years.

V. Fama-MacBeth Regression Results of Monthly Stock Returns
In this section, we report Fama-MacBeth (1973) regression results using monthly returns.For each of the 528 months from Jan. 1975 to Dec. 2018, we estimate cross-sectional regressions of various model specifications using the return on a stock as the dependent variable.Table 9 reports the time-series averages of the coefficients from the monthly regressions and the Newey-West t-statistics, computed using the time-series standard deviations of the monthly coefficients.To check the robustness of the results, we report ordinary least squares (OLS) results in Panel A and WLS results in Panel B using market cap weights.In Table IA-12 in the Supplementary Material, we also report OLS results after excluding microcaps from the sample.The results excluding microcaps are generally in between the EW OLS and VW WLS results.The dependent variable is the firm's monthly stock return.In each model of Table 9, we also control for the market cap and the market-to-book ratio at the end of year t, asset growth in t, and QTR_ROE tþ .Following Hou et al. (2015), we define QTR_ROE tþ as the most recent quarterly earnings announced prior to the stock return month divided by beginning-of-quarter book value of equity.The other independent variables take on the values from fiscal year t, the firm's most recent fiscal year that ends at least 4 months prior to the stock return month.
Model 1 of Panel A of Table 9 does not include security issue dummy variables.The coefficients on the independent variables are consistent with the literature.In model 2 of Panel A, we include 4 dummy variables for 1, 2, 3, or at least 4 issues, without distinguishing between debt and equity.Consistent with the results in Table 7, more frequent issues are followed by lower stock returns.Firms with 1, 2, 3, and at least 4 issues underperform nonissuers by 0.08%, 0.24%, 0.51%, and 0.78% per month, respectively.
BFG find that the more different types of securities that are issued, the lower are a firm's subsequent abnormal returns.Our model 3 of Panel A of Table 9 shows that, conditional on the number of types of securities, there is a monotone relation between the number of issues and subsequent returns, consistent with the results in  9 and Panel B of Table 7.However, there is mixed evidence on the number of types of securities issued.With 2 issues and just 1 type, the coefficient of À0.21 is higher than the coefficient of À0.31 with 2 issues and 2 types.With 3 issues and 1 type, the coefficient of À0.51 is the same as that with 3 issues and 2 types.Model 4 of Panel A of Table 9 includes the number of issues and the number of types of securities issued in t À 2, t À 1, and t, and the results suggest that more issues are associated with lower returns, but more types are associated with slightly higher returns.Thus, our results do not confirm BFG's findings.Instead, the results in our models 3 and 4 of Panel A of Table 9 suggest that the number of issues is more reliable than the number of types of securities in predicting returns.

Fama-MacBeth Regressions of Stock Returns
Cross-sectional regressions in Table 9 are estimated each month.The dependent variable is the monthly return (in percentage) on a firm's stock.Panel A reports equal-weighted ordinary least squares (OLS) results, and Panel B reports weighted least squares (WLS) results using value (market cap) weights.The WLS results use the market value of equity (the number of shares outstanding Â price per share from CRSP) as the weight.When there are multiple share classes, the market values of all classes of shares are added.The top and bottom 1% values of ln(MARKET_CAP) t , ASSET_GROWTH t , and QTR_ROE tþ are winsorized for each regression sample.QTR_ROE tþ equals the most recent quarterly earnings announced prior to the month of the regression divided by beginning-of-quarter book value of equity.To avoid stale earnings, it is also required that the fiscal quarter that corresponds to the announced earnings ends no more than 6 months prior to the month of the regression.The other control variables have values from fiscal year t ending at least 4 months prior to the month of the regression.This table reports the time-series averages of the monthly coefficients and the corresponding Newey-West t-statistics that correct the for first-, second-, and third-order autocorrelations.tstatistics using a Newey-West correction with 3 lags are in parentheses, with *, **, and *** signifying statistical significance at the 10%, 5%, and 1% significance levels, respectively.See Appendix A and Table 1 for variable definitions.
Panel A. OLS (Equal-Weighted) Results (1975-2018, no. of  Model 5 of Panel A of Table 9 shows that firms with at least 1 debt issue in the prior 3 years underperform nonissuers by 0.16% per month in the next year and firms with at least 1 equity issue in the prior 3 years underperform nonissuers by 0.27% per month. In their Table 3 Fama-MacBeth regressions, BFG find that, after controlling for the number of security types, the security type for the first financing event (IPO, SEO, private equity placement, public debt offering, or bank loan) in a 36-month window is not related to long-run returns.In model 6 of Panel A of Table 9, we distinguish between debt and equity by including 6 dummy variables: Three dummy variables equal 1 if there is, respectively, 1, 2, or 3 debt issues in the previous 3 years, and 3 dummy variables for the frequency of equity issues.More frequent debt issues or more frequent equity issues are followed by lower stock returns in year t þ 1. Firms with 1-3 debt issues underperform those with 0 debt or equity issues by 0.11%, 0.26%, and 0.50% per month, respectively.Firms with 1-3 equity issues underperform those with 0 debt or equity issues by 0.19%, 0.39%, and 0.85% per month, respectively.Inconsistent with BFG, equity issues are followed by lower stock returns than debt issues after controlling for the number of equity issues and the number of debt issues.
The EW OLS results in Panel A of Table 9 could overstate the importance of microcaps (Hou et al. (2020)).To alleviate this concern, we also estimate VW WLS We continue to find that more frequent issues are associated with lower subsequent stock returns.Firms with 1, 2, 3, and at least 4 issues underperform nonissuers by 0.05%, 0.05%, 0.31%, and 0.39% per month, respectively.This pattern appears to be driven by the number of equity issues rather than the number of debt issues.We do not find a robust relation between the number of debt issues and subsequent stock returns, but find that more frequent equity issues are associated with lower subsequent stock returns.Firms with 1-3 equity issues underperform those with 0 debt or equity issues by 0.14%, 0.40%, and 0.65% per month, respectively.There are potentially multiple reasons for the difference between BFG's results and ours.We highlight 2 of the reasons here.BFG focus on the type of the first security issue and the number of types of securities issued (see their Table 3), whereas our model 6 of Table 9 considers the type of each security issue and focuses on the number of debt issues and the number of equity issues in a 36-month window.It is perhaps not surprising that BFG do not find the type of the first issue to be important for explaining the stock return in a month, because our article finds that, other things being held equal, the abnormal return is more negative in the first 2 years after the fiscal year of the security issuance than in the third year.
BFG also distinguish between IPOs, SEOs, and PIPEs and distinguish between public debt and private debt offerings, whereas we do not.9If a firm issued several different types of equity in different years, they would classify the firm as issuing multiple types, whereas we would classify it as a frequent issuer of equity.Thus, the different classification schemes may account for some of the difference in conclusions.
In Table IA-13 in the Supplementary Material, we estimate the regressions using an alternative set of control variables, including market cap, the market-tobook ratio, operating profitability, and asset growth.Our major results are qualitatively similar.In Table IA-14 in the Supplementary Material, we replicate Table 9 using large issues, study the importance of issue recency, and check the results for the subperiods of 1975-1996 and 1997-2018.The Fama-MacBeth results in Table IA-14 are generally consistent with our time-series regression results in Tables 4-8.

VI. Fama-MacBeth Regression Results of Earnings Announcement Returns
The expectational error mispricing story in La Porta, Lakonishok, Shleifer, and Vishny (1997) predicts that a significant portion of mispricings is corrected at subsequent earnings announcements.If investors are overly optimistic about overvalued firms, they will probably be disappointed at the firms' subsequently announced earnings.Alternatively, as Wu, Zhang, and Zhang (2010), Liu andZhang (2014), andZhang (2017) suggest, the investment CAPM also predicts that a significant portion of anomalies should occur around earnings announcements.In this section, we examine stock returns around earnings announcement days.
We estimate regressions to evaluate the importance of issue frequency, size, and recency in explaining EARs after controlling for firm characteristics, such as investment and profitability.The dependent variable is the average 3-day buy-andhold percentage return from 1 day before to 1 day after each earnings announcement made from 123 to 488 calendar days after the end of fiscal year t.We first estimate the cross-sectional regressions for each of the calendar years from 1975 to 2018.

Fama-MacBeth Regressions of Earnings Announcement Returns
The dependent variable in Table 10 is the average 3-day buy-and-hold return (in percentage) from 1 day before to 1 day after the quarterly earnings announcement date (Compustat item RDQ) for all earnings announcements made from 123 to 488 calendar days after the end of fiscal year t.We estimate cross-sectional regressions for each of the calendar years from 1975 to 2018, using observations with the fiscal year end date that falls into the prior calendar year.Panel A reports equal-weighted ordinary least squares (OLS) results, and Panel B reports weighted least squares (WLS) results.The WLS results use the market value of equity at the end of fiscal year t from Compustat as the weight, adjusting for inflation within the year for firms with different fiscal year ends.The top and bottom 1% values of ln(MARKET_CAP) t , ASSET_GROWTH t , and QTR_ROE t are winsorized for each regression sample.QTR_ROE t equals earnings in the fourth quarter of fiscal year t divided by beginningof-quarter book value of equity.This table reports the average of the annual coefficients and the corresponding Newey-West t-statistics that correct for first-order autocorrelation.t-statistics using a Newey-West correction with 3 lags are in parentheses, with *, **, and *** signifying statistical significance at the 10%, 5%, and 1% significance levels, respectively.See Appendix A and Table 1  Huang and Ritter 197 The model specifications in Table 10 are similar to those in Table 9.In model 1 of Panel A of Table 10, the results on the control variables are generally consistent with those in Lewis and Tan (2016).The results in model 2 of Panel A of Table 10 show that firms with more issues are associated with lower EARs.During the 3 days around the earnings announcement, firms with at least 4 issues in the 3 prior years, on average, underperform nonissuers by À1.00%.Since there are typically 4 earnings announcements per year, this 3-day abnormal return translates into a cumulative 12-day abnormal return of À4.00% in year t þ 1.In model 2 of Panel A of Table 9, the corresponding coefficient is À0.78% per month or À9.36% per year.Thus, on average, over 40% of the abnormal return in year t þ 1 occurs on the 5% of days around the 4 earnings announcements.The EAR results are consistent with the expectational error mispricing story, but the magnitude of the EAR patterns cannot be easily explained by risk-based theories.
The results in model 3 of Panel A of Table 10 show that, conditional on the number of types of securities, the number of issues is generally associated with lower EARs.However, conditional on the number of issues, there is mixed evidence on the relation between the number of security types and EARs.The results in model 4 of Panel A of Table 10 suggest that the number of issues is negatively associated with EARs, but the number of types of securities is positively related to EARs.In model 5 of Panel A of Table 10, the 3-day EARs in the subsequent year of firms with at least 1 debt issue in t À 2 to t are not statistically different from those without any issue, and the 3-day EARs in the subsequent year of firms with at least 1 equity issue are 0.54% lower than those without any issue.Equity issues are followed by lower EARs than debt issues, consistent with the findings of Lewis and Tan (2016).The results in model 6 of Panel A show that more frequent debt issues or more frequent equity issues are followed by lower EARs.
Our major results on the relations between securities issuance and subsequent EARs are not driven by a large number of microcaps or a few firms with the largest market caps in the sample.The WLS results in Panel B of Table 10 and the OLS results after excluding microcaps in Table IA-15 in the Supplementary Material have similar patterns to the OLS results in Panel A of Table 10, although the statistical significance is typically lower and the absolute values of the coefficients are generally smaller.
In the Supplementary Material, Tables IA-16 shows that our major results are robust to alternative controls.Table IA-17 shows that larger, more frequent, and more recent issues are generally followed by more negative EARs, and our major results are mostly robust in the subperiods of 1975-1996 and 1997-2018.VII.Related Literature and the Importance of Issue Size, Frequency, and Recency

A. The Timing of Portfolio Construction
We find that the abnormal return is lower in the first 2 years after the fiscal year of the security issuance than in the third year.This finding suggests that keeping issuers in the portfolio for fewer than 3 years after issuance could boost the power to detect abnormal returns, and helps reconcile different findings in the literature on post-issuance long-run performance.
A long list of articles use event data on securities issuance to examine stock returns during the 3 or 5 years after issuance. 10For example, Panel E of Table 4 of Bessembinder and Zhang (2013) reports the results of a regression using monthly VW portfolio returns for firms that conducted 1 or more SEOs during the prior 60 months.For a time-series regression from Apr. 1980 to Dec. 2010 (369 months), they report an insignificant alpha of þ0.09% per month using the Fama-French-Carhart 4-factor model.
In contrast to those that use event data, articles that use the change in splitadjusted shares outstanding from CRSP or information from Compustat to identify equity or debt issuance typically examine stock returns during a 1-year window after the fiscal year of issuance.Table 4 of Hou et al. (2015) reports a q-factor model alpha of À0.26 per month with a t-statistic of À1.75, using a sample period of Jan. 1972-Dec. 2012, for a VW long-short portfolio of net stock issuers.The long side is the top decile, and the short side is the bottom decile, of companies ranked by the percentage increase in shares outstanding (roughly speaking, large issuers minus repurchasers).The portfolios are formed at the end of every June.
Table 6 of Fama and French (2016) reports significantly negative 5-factor model intercepts for 3 out of 5 size (market cap)-sorted VW portfolios for companies in the top quintile (top 20%) of net equity issuers.They form portfolios once a year at the end of June.The average intercept of their 5 high net issue portfolios is À0.18% per month, and the average intercept of their 5 repurchasers portfolios is þ0.04% per month, a difference of À0.22% per month.Their sample period is July 1963-Dec.2014.
Why do Fama and French (2016) find a lower abnormal return following equity issuance than Bessembinder and Zhang (2013)?Our finding on issue recency suggests that it is partly, because the latter article keeps the SEO firms in the portfolio much longer than the other 3 articles do, although different definitions of issuing firms and different sample periods may also explain some of the difference in abnormal returns.
On his website, Ritter (2021) shows that, for both IPOs (from 1980 to 2019) and SEOs (from 1970 to 2011), there is no underperformance after controlling for market cap and book-to-market in the first 6 months after issuance, but there is a lot in the next 18 months.The conventional wisdom is that firms avoid negative earnings surprises shortly after an equity issue, both through the guidance of analysts and by earnings management.Ritter also shows that there is not much underperformance in years 3-5 after the IPO or the SEO, consistent with our finding on issue recency.Thus, adding a stock immediately after issuance will move the estimated portfolio abnormal return toward 0, but waiting too long or keeping it in the portfolio too long (more than 2 years) will also move the abnormal return toward 0. In Bessembinder and Zhang (2013) and many other articles, issuers enter the portfolio too quickly, and stay too long, with both of these timings moving abnormal returns toward 0.
Both Hou et al. (2015) and Fama and French (2016) form portfolios at the end of every June and define equity issuers using the change in split-adjusted shares outstanding over the fiscal year ending in the prior calendar year.In their articles, an issuer is first added to the issuer portfolio from 6 to 29 months after issuance.In the fastest case, they add a firm that has a fiscal year ending on Dec. 31 of t À 1 and issues equity on this same date to the portfolio on July 1 of t.In the slowest case, they add to the portfolio on July 1 of t a firm that has a fiscal year ending in January of t À 1 and issues on Feb. 1 of t À 2. Thus, they miss some of the negative abnormal returns in the post-issue months 7-12 that we capture.

B. Issue Size and Frequency
Importantly, we show that large and frequent issues in the prior 3 years are followed by lower stock returns in the subsequent year than small and infrequent issues.Thus, another reason for why some articles find no abnormal returns following debt or equity issuance is that their samples include many firms that occasionally raise a small amount of external capital.
Articles that use event data for issuance information typically ignore issue size, although some articles that use CRSP or Compustat for issuance information do distinguish between small and large issues.However, almost all articles on long-run performance after issuance ignore issue frequency.For example, both Hou et al. (2015) and Fama and French (2016) examine equity issues on the basis of the change in split-adjusted shares outstanding in a 1-year window rather than a 3-year window.Our findings on issue size and frequency suggest that focusing on a sample of firms that raise a large amount of external capital with either 1 large issue or multiple regular-size issues during 2 or 3 years could boost the power to detect abnormal returns following debt or equity issuance.

VIII. Conclusions
The literature on post-issuance stock returns almost always studies 1 type of security issuance without fully controlling for surrounding issuances of the same type of security and other types of securities.This practice makes inferences difficult, especially since frequent issuances are prevalent.For example, without adequately controlling for surrounding debt issues when studying long-run stock returns following SEOs, it is not clear how much of the results on SEOs are driven by debt issues.BFG find that an increase in the number of types of securities issued is related to lower abnormal long-run stock returns.In comparison, we find that more frequent, larger, and more recent issues of debt and equity in the 3 years from t À 2 to t are followed by lower abnormal stock returns in the 12 months starting from month 5 of year t þ 1, with the number of types of securities being relatively unimportant.
Frequent and large issues are followed by lower abnormal returns than infrequent and small issues.The intercept of the Hou-Xue-Zhang (2015) q-factor calendar-time regression for the VW portfolio of firms with at least 3 large issues from t À 2 to t is À0.63% per month (t-stat.= À4.31) in the subsequent year.
More recent issues (e.g., year t issues relative to year t À 1 or t À 2 issues) are also followed by lower average abnormal returns in t þ 1 than less recent issues.The VW portfolio of firms that did an equity issue in fiscal year t has a q-factor alpha of À0.37% per month (t-stat.= À3.11),but the VW portfolio of firms that did at least 1 equity issue in the prior 3 years has an insignificant 5-factor alpha in the subsequent year.A VW portfolio of firms that issued equity in both t À 1 and t has a q-factor alpha of À0.61% per month (t-stat.= À3.42) in t þ 1.
Purging the q-factors by excluding firms with equity or debt issues in the prior 3 years in the construction of the q-factors increases the power to detect abnormal returns.For example, for the VW portfolio of firms that issued equity in both t and t À 1, the purged q-factor regression intercept decreases to À0.78% per month (t-stat.= À4.33)from À0.61% per month.
Our findings suggest that studies that find no abnormal returns for issuing firms in multifactor models (e.g., Lyandres et al. (2008), Bessembinder and Zhang (2013)) have low power to find abnormal returns for several reasons: i) abnormal returns following issuance do not start immediately after issuance, but then decay over time, and thus studies that add issuers to a portfolio too soon or too late or keep issuers in a portfolio for too long will move abnormal returns toward 0; and Huang and Ritter 201 LARGE_EQUITY_ISSUE: A firm is defined to have a large equity issue in a year if ΔEQUITY in the year is at least 10% of the book value of assets and at least 3% of the market value of equity at the beginning of the year.DEBT_ISSUE: A firm is defined to have a debt issue in a year if ΔDEBT in the year is at least 5% of the book value of assets and at least 3% of the market value of equity at the beginning of the year.
LARGE_DEBT_ISSUE: A firm is defined to have a large debt issue in a year if ΔDEBT in the year is at least 10% of the book value of assets and at least 3% of the market value of equity at the beginning of the year.ASSET_GROWTH: The growth rate of end-of-year total assets (AT).
QTR_ROE: Quarterly ROE, defined as income before extraordinary items (Compustat quarterly item IBQ) divided by the beginning-of-quarter book value of equity, when the denominator is positive.As in Hou et al. (2015), (2020) and Hou, Mo, Xue, and Zhang (2019), (2021), the book value of equity is defined as shareholders' equity (SEQQ) þ deferred taxes and investment tax credit (TXDITCQ) À the carrying value of preferred stock (PSTKQ).If TXDITCQ is missing, it is set to 0. If SEQQ is missing, shareholders' equity equals common equity (CEQQ) þ the carrying value of preferred stock (PSTKQ) if available, or equals total assets (ATQ) À total liabilities (LTQ).
RETURN tþ1 : The total return on a stock in fiscal year t þ 1, measured from month 5 to month 16 after the end of fiscal year t so as to allow the release of fiscal year t's numbers before returns are measured and a firm is classified.If the stock gets delisted before 1 year, the return until delisting is used.
RETURN tþ1, tþ3 : The total return on a stock from fiscal year t þ 1 to fiscal year t þ 3, measured from month 5 to month 40.If the stock gets delisted before 3 years, the return until delisting is used.
MARKET_ADJUSTED_RETURN tþ1 : Total return on the firm's stock in fiscal year t þ 1, measured from month 5 to month 16 after the end of fiscal year t, minus the VW return on the market in the same year.
MARKET_ADJUSTED_RETURN tþ1, tþ3 : Total return on the firm's stock from fiscal year t þ 1 to fiscal year t þ 3, measured from month 5 to month 40, minus the VW return on the market in the same 3 years.
ln(MAKRET_CAP): The natural logarithm of the market value of equity (Common Shares Outstanding (CSHO) Â Price Close Fiscal Year (PRCC_F)) from Compustat, measured in Dec. 2018 dollars using the Consumer Price Index from the Bureau of Labor Statistics.MARKET_TO_BOOK: The market value of equity (Common Shares Outstanding (CSHO) Â Price Close Fiscal Year (PRCC_F)) divided by the book value of equity.The book value of equity is defined as the book value of assets (AT) À Total liabilities (LT) À Liquidating Value of Preferred Stock (PSTKL) þ Deferred Taxes and Investment Tax Credit (TXDITC).When PSTKL is missing, the redemption value (PSTKRV) is used.When PSTKRV is also missing, the carrying value (PSTK) is used.

TABLE 1 Sample Distribution
Table 1 reports the sample distribution A. Sample Distribution by No. of Issues and No. of Large Issues, Independently https://doi.org/10.1017/S0022109021000636Published online by Cambridge University Press

TABLE 2
Average Firm Characteristics Table 2 reports the averages of several firm characteristics.See Appendix A and Table 1 for variable definitions.The top and bottom 1% values of the firm characteristics are winsorized.

TABLE 3
Average Post-Issuance Percentage Buy-and-Hold Returns

TABLE 4 (
continued) Calendar-Time Factor Regression Results: Equity Issues Panel C. Frequency and Recency of Equity Issues(1975-2018, no. of months = 528)

TABLE 5 (
continued) Calendar-Time Factor Regression Results: Debt Issues Panel C. Frequency and Recency of Debt Issues(1975-2018, no. of months = 528) Table 6 reports nonpurged VW results, and Panel B reports purged VW results, with EW results reported in Table IA-7 in the Supplementary Material.

TABLE 7
for the definition of issuer or large issuer.

TABLE 7 (
continued) Calendar-Time Factor Regression Results: Equity and Debt Issues Combined Panel C. 2-Way Sort by No. of Types of Securities and No. of Issues(1975-2018, no. of months = 528) Panel D. 2-Way Sort by No. of Issues and Type of Security(1975-2018, no. of months = 528)

TABLE 9
://doi.org/10.1017/S0022109021000636Published online by Cambridge University Press regressions.The WLS results in Panel B of Table 9 are qualitatively similar to the OLS results, although the absolute values of the coefficients are generally smaller. https Table 10 reports the time-series averages of the annual coefficients and the corresponding Newey-West t-statistics.Overall, the results in this table cannot be easily explained by risk-based theories.

TABLE 10
for variable definitions.

TABLE 10 ( continued )
Fama-MacBeth Regressions of Earnings Announcement Returns