Investor Behavior and the Timing of Secondary Equity Offerings

Transcription

1 Investor Behavior and the Timing of Secondary Equity Offerings Dalia Marciukaityte College of Administration and Business Louisiana Tech University P.O. Box Ruston, LA Phone: Fax: Samuel H. Szewczyk Bennett S. LeBow College of Business Drexel University Philadelphia, PA Phone:

2 2 Investor Behavior and the Timing of Secondary Equity Offerings Abstract We examine stock-price performance around secondary equity offerings. Firms whose shares are sold in these offerings have positive abnormal pre-offering performance, negative market reaction to the announcement of the offering, and positive abnormal post-offering performance. Difference in risk, tax-motivated selling, trading before the end of the year, and the momentum effect cannot explain abnormal post-issue performance. Our findings provide support for the disposition effect associated with investor tendency to sell winners and, in particular, investors' expectations of mean reversion. To reconcile the findings about stock-price performance after secondary and primary issues, we offer a simple model based on selection bias and investor and manager overconfidence about their ability to identify firms with good and poor future opportunities.

3 3 The optimal time for investors to sell firm shares they own is when the firm is overvalued. A negative stock market reaction to the announcements of secondary equity offerings (e.g., Mikkelson and Partch (1985) and Asquith and Mullins (1986)) is consistent with this expectation. However, some studies question whether investors time the selling of shares to their benefit. Odean (1998) examines trading records at a large discount brokerage house and finds support for the disposition effect, the tendency of investors to hold losing investments too long and sell winning investments too soon. He also finds that the winning investments that investors sell continue in subsequent months to outperform the losers they keep. To obtain a better understanding of investor timing of their investment sales, we examine the timing of secondary public equity offerings. In such offerings current shareholders sell shares to public. Our sample consists of 644 offerings made during the 1974 to 1995 period. We find that the shares sold in secondary offerings, on average, are not overpriced. Firms having secondary equity offerings even outperform matching firms in subsequent one- to five-year periods. This overperformance is strongest in the first year after the issue. Average abnormal holding period returns over the first post-offering year range from 8.18% to 13.31%, depending on the benchmark. We confirm these results using Fama and French (1993) three-factor model and four-factor model including momentum factor. Difference in risk, tax-motivated selling, or trading before the end of the year cannot explain this abnormal performance. Loughran and Ritter (1997) suggest that underperformance after seasoned equity offerings might be due to investor expectation that increase in earnings before the issue will continue after the issue. For this explanation to be applicable for secondary equity

4 4 offerings, pre-issue performance of secondary equity issuers should be poor or deteriorating. Instead, we find that secondary equity issuers have a very similar operating performance before and after the issue, as primary equity issuers do in the Loughran and Ritter sample; performance is improving before the issue and deteriorating (although somewhat slower than for primary equity issuers) after the issue. In contrast to primary equity issuers, secondary equity issuers have better operating performance after the issue than their matching firms. Prior research shows that on average analysts are overoptimistic in their predictions of firm earnings. Examining analyst forecast errors around equity offerings show that at the time of secondary issues overoptimism about earnings is reduced. After adjusting for the positive bias in forecast errors using forecast errors for matching firms, we find that at the time of issue analysts are overpesimistic about the earnings of issuing firms. Similar to Odean (1998), we find that investors tend to sell winners. Only 19.18% of our sample firms have negative returns over the year prior to the secondary equity offering, while 38.76% of the Center for Research in Security Prices (CRSP) database firms have negative returns over the same periods. We also find that only pastwinners (firms with positive one-year pre-offering returns) have significant positive postoffering abnormal returns. Past-losers record a one-year post-offering holding period abnormal return of only 1.20% (not statistically significant). Although abnormal performance is concentrated among past-winners, past-performance cannot fully explain overperformance of secondary equity issuers. Firms making secondary equity issues significantly outperform firms matched by prior stock-price performance and size.

5 5 Odean (1998) offers two possible explanations for investor tendency to sell winners: the prospect theory (Kahneman and Tversky (1979)) and the expectation of mean reversion (Andreassen (1988)). Under the prospect theory, in contrast to standard utility theory, value is assigned to gains and losses rather than to level of wealth. The value function is steeper for losses than for gains, making investors, in general, risk averse. However, the value function is concave for gains and convex for losses, making investors behave differently in the domain of gains and the domain of losses. According to the prospect theory, once an investment gains value, current owners of this investment become more risk-averse and require higher compensation for the risk associated with this investment. Even though expected future returns and risk may remain the same, current owners may not consider them to be sufficient to compensate for risk and, therefore, are willing to sell such investments. The opposite happens when investments lose value. The prospect theory suggests that winning investments have lower value only to current shareholders; the rest of the market should have no reason to avoid these stocks. If the market believes in the prospect theory, the market should react less negatively to the announcements of secondary offerings of past winners (in this case selling shareholders have personal reasons to sell the shares) than to the announcements of secondary offerings of past losers. We find the opposite; the market reacts more negatively to the announcement of secondary offerings of past winners. On the other hand, investors may prefer to sell winners and hold losers because they believe stock prices exhibit mean reversion: current losers outperform current winners in the future. Andreassen (1988) finds that in experimental settings people buy and sell stocks as if they expect short-term mean reversion. There is some empirical

6 6 evidence of mean reversion in stock returns. Jegadeesh (1990) and Lehmann (1990) find short-term mean reversion in stock returns. Also, DeBondt and Thaler (1985, 1987) show that long-term past losers outperform long-term past winners in the subsequent three to five years. However, Jegadeesh and Titman (1993) show that over an intermediate horizon of three to twelve months, past winners on average continue to outperform past losers. In our sample of secondary equity issuers, past winners continue to outperform past losers over a horizon of one to five years. Although past winners' overperformance is strongest in the first post-offering year, there is no subsequent underperformance in the later years. Expectation of mean reversion, however, does not have to be based on empirical evidence; investors could have an irrational expectation. If investors (rationally or irrationally) expect mean reversion in stock returns, not only current owners, but also other investors should disfavor past winners. Finding more negative market reaction to the announcements of secondary offerings of past winners is more consistent with the expectation of mean reversion than with the prospect theory. Our findings also highlight the need to understand secondary equity offerings as separate from primary equity offerings. Some previous studies of post-offering performance do not differentiate between secondary and primary equity offerings. Lee (1997) finds that the sample of secondary and combination (primarily secondary) equity issuers do not have abnormal post-offering performance. Brav, Geczy, and Gompers (2000) find only weak evidence of abnormal post-offering performance for the sample of equity issues that include secondary equity offerings. A secondary equity offering is a similar event to a primary equity offering in the sense that firm's shares are offered to the public. However, in a primary equity offering, the seller is the firm itself and proceeds

7 7 from the offering accrue to it, while in a secondary equity offering, sellers are current owners of shares and proceeds accrue to them. In a primary equity offering, managers make the decision to offer shares. Unless managers are the sellers in a secondary offering, they are not the ones making the decision to offer shares. The timing of secondary issues seems to be very different from the timing of primary equity issues. Loughran and Ritter (1995) and Spiess and Affleck-Graves (1995) find that primary equity issuers tend to underperform matching firms in the post-offering period, suggesting that firms are overpriced at the time of primary equity offerings. We find that secondary equity issuers tend to outperform matching firms in the post-offering period, suggesting that firms are underpriced at the time of secondary equity offerings. If these two events are combined in one sample, tests of post-offering performance will not show the true picture. Existing theories cannot explain why primary equity issues(or primary issues of other securities) are followed by underperformance while secondary equity issues are followed by overperformance. We offer a simple model to reconcile these findings. The model is based on selection bias and manager and investor overconfidence in their ability to identify firms with good and poor future prospects. Firms making primary security offerings tend to be firms with expected good future prospects. Under certain assumptions, we show that such firms on average are overvalued firms. Firms making secondary security offerings will tend to be firms with expected poor future prospects; such firms on average are undervalued. Together with the disposition effect, this model also explains why abnormal positive performance is concentrated among past winners.

8 8 The remainder of the paper is organized as follows. In the next section, we describe the sample and methodology. We examine the stock price and operating performance of offering firms before, during, and after the offering in Section II. In Section III we examine whether abnormal returns are due to the higher risk of offering firms, tax-motivated selling, or trading at the end of the year. The disposition effect, possible explanations of it, and momentum effect are examined in Section IV. In Section V we present a model explaining abnormal post-issue performance after primary and secondary offerings. In the last section we conclude. I. Sample and Methodology A. Sample The sample is selected from secondary equity offerings made during 1974 through 1995 reported in the Securities Data Corporation (SDC) database. Sample selection criteria are: (1) the firm is not a regulated utility (SIC codes ), holding firm, or other investment firm (SIC codes ); (2) no primary shares of common stock or other financing instruments are offered jointly with the secondary equity offering; (3) data for the firm are available on the CRSP daily return database at the time of the offering. To avoid statistical problems associated with interdependence of observations, we also require that secondary equity offerings of the same firm be at least five years apart. If two offerings are less than five years apart, we include only the earlier offering. The full sample consists of 644 secondary equity offerings. However, the number of offerings used in the tests varies because of data availability.

9 9 Panel A of Table 1 shows the distribution of the offerings by calendar year. As shown in Panel B of Table 1, more than 50 percent of the sample firms trade on the New York Stock Exchange. Panel C of Table 1 provides summary statistics for firm size and the size of secondary equity offerings. The mean firm size, as measured by market value of equity before the issue announcement, is $1,130 million. The average size of the offerings, calculated by multiplying the number of shares offered with the offer price (if offer price is not available, we use closing price on the issue day) is $64 million. The mean relative size of the offerings is 12.98% of the firm's total equity value. The distribution of the secondary equity offerings by industry is shown in Table 2. Industries are identified by the Standard Industrial Classification (SIC) codes provided by CRSP. For comparison purposes, we include an overall distribution of CRSP firms by industry. This sample of CRSP firms satisfies the same industry, period, and CRSP data availability requirements as the sample of offering firms. The distribution by industry of the secondary offerings sample is similar to the overall distribution of CRSP firms. B. Estimation of Announcement Period Market Reaction We use a standard market model event test to estimate abnormal returns during the announcement period. The market model parameters are estimated over 150 trading days, beginning 30 days after the offering's announcement day. The announcement day (Day 0) is the earliest of the following dates: the first announcement of the offering found on Dow Jones Newswires or the Wall Street Journal, the SEC registration date for the offering, and issue date of the offering. Cumulative abnormal returns (CAR) are measured over a three-day announcement period including Day -1, Day 0, and Day 1.

10 10 Market model parameters should be estimated during normal performance period. Because of abnormal pre-offering and post-offering performance, we cannot find such period. To avoid this problem, we also estimate the abnormal returns as the difference between the return on the firm's stock and the contemporaneous return on the CRSP value weighted index. This estimation procedure produces results that are similar to those reported for the market model event test. C. Estimation of Abnormal Long-term Stock-price Performance We employ two different methodologies to evaluate abnormal long-term stockprice performance. First, we estimate buy-and-hold abnormal returns relative to the returns of four matching samples. Second, we estimate calendar-time abnormal returns using two factor models. C.1. Matching Samples To evaluate the long-run stock-price performance of firms whose equity is sold in secondary offerings (offering firms), we create benchmarks consisting of matched control firms. Four samples of matching firms are considered: (1) book-to-market and size, (2) prior returns and size (3), industry and size, and (4) size. Offering firms are excluded from the matching samples for the five years before and five years after the secondary offering. The book-to-market and size matched sample is created by forming twenty size portfolios from CRSP database firms each month. Here, size is defined as a market value of equity. The number of firms in each size portfolio is the same. To each offering firm,

11 11 the size portfolio to which the placing firm belongs is assigned. The firm from the assigned group with the book-to-market ratio closest to the offering firm book-to-market ratio is selected to the matching sample. If book-to-market ratio is not available for a particular offering firm, we match by size only. We estimate book-to-market ratio at the end of the fiscal year and market value of equity at the end of the month that ends before the first announcement of secondary equity offering. The prior return and size matched sample is constructed in a similar way as the book-to-market and size matched sample. Only, instead of book-to-market ratio here we use six-month stock returns estimated before the issue announcement. Loughran and Ritter (1995) argue that matching by industry reduces the ability to identify abnormal returns due to industry-wide misvaluations. Although we acknowledge that possibility, we believe that industry matching may provide better control for events unrelated to the equity offering. Companies from the same industry will be affected by other events in more similar way (for example, if the whole industry is in a recession, this is not the effect of equity issue that the offering firm from that industry is performing poorly). The industry and size matched sample is created by finding all the firms with the same first two digits of the Standard Industrial Classification (SIC) code for each offering firm, and then selecting the firm with the closest size to the offering firm size. SIC codes are obtained from CRSP. Matching is performed at the beginning of the first announcement month.

12 12 The sample matched by size is created by selecting firms from the CRSP database with market values of equity closest to those of the offering firms. Matching is performed at the beginning of the first announcement month. To ensure that matching firms are not lost because of unavailable data, we require that matching firms satisfy the same CRSP data requirements as the offering firms. If a matching firm does not satisfy these requirements, we choose the next best match. C.2. Estimation of Buy-and-hold Abnormal Returns To estimate abnormal pre-offering and post-offering performance, we use holding period abnormal returns. We follow the procedure used in Spiess & Affleck-Graves (1995). The holding period return (HPR) for each security in the offering and matching sample is calculated as: b HPR i, a, b = ( 1 + Ri, t ) 1, t= a (1) where HPR i,a,b is holding period return for firm i during the period from a to b; R i,t is the monthly return on common shares of firm i in month t. Holding period abnormal return (HPAR) is calculated as the difference between the holding period returns for the offering firm and matching firm. We are using a matching firm and not a reference portfolio for each offering firm. Barber & Lyon (1997) suggest that in such case a conventional t-statistic can be used to evaluate statistical significance of average HPARs. However, Kothari and Warner (1997) suggest using bootstrapping. To evaluate statistical significance of HPARs we estimate both,

13 13 conventional t-statistic and bootstrapped skewness adjusted p-value. Bootstrapped skewness adjusted p-value is calculated using the procedure described in Lyon, Barber, and Tsai (1999). Significant positive (negative) post-offering abnormal returns indicate overperformance (underperformance) only if the offering firm's risk is not higher than that of the matching firm. To address this bad model problem, we use four different matching samples. A number of matching samples are used in the tests as it is not possible to control for many risk factors simultaneously in one matching sample and obtain close matches by those factors. Matching methodologies often suffer from new listing or survivorship bias (Lyon, Barber, and Tsai (1999)). To avoid this problem, we treat matching and offering samples in the same manner. For example, if an offering firm is delisted from CRSP after three years, this firm and its matching firm are not included in the four- or five-year abnormal return calculation. We apply the same rule when a matching firm is delisted, thereby treating offering firms and matching firms in the same manner. This reduces the sample size, but avoids biasing the results. C.3. Estimation of Calendar-Time Abnormal Returns Statistical significance of average holding period abnormal returns may be overstated because of cross-sectional dependence of observations. This problem can be avoided by using multifactor models (Mitchell and Stafford (2000)). We use Fama and French (1993) three-factor model and four factor model including momentum factor. To

14 14 estimate abnormal returns using Fama and French three-factor model, we follow the procedure described in Mitchell and Stafford (2000): R pt R = α + β ( R R ) + β SMB + β HML + ε (2) ft m mt ft s t h t t where R pt is the average return on the portfolio of offering firms for month t, Rft is the risk-free rate, Rmt Rft is the excess return on the market, t SMB is the difference in returns between a portfolio of small and large stocks, and HML t is the difference in returns between high and low book-to-market stocks. We obtain factor values from French website. The four-factor model in addition to Fama and French factors includes momentum factor: R pt R = α + β ( R R ) + β SMB + β HML + β PR1YR + ε (3) ft m mt ft s t h t p t t where R pt the average return of offering firms for month t, Rft is the risk-free rate, Rmt R ft is the excess return on the market, SMBt is the difference in returns between a portfolio of small and large stocks, HML t is the difference in returns between high and low book-to-market stocks, and PR1 YR is the difference in returns between high and low momentum stocks. To estimate momentum factor, we follow the methodology described in Carhart (1997). Prior research shows that in some cases factor models are unable to sufficiently explain stock returns. Mitchell and Stafford (2000) suggest estimating adjusted intercept relative to the expected intercept. The expected intercept is estimated as an average intercept from 1,000 factor models of random portfolios with the same size and book-tomarket characteristics as the issuing firms. The adjusted t-statistic is estimated as:

15 15 αˆ E( αˆ) t = (4) se where αˆ is the intercept from the factor model for the offering sample, E(αˆ ) is the average intercept from 1,000 regressions for the generated portfolios, and se is the estimated standard error from the factor model. We estimate adjusted t-statistic for Fama and French three-factor model and four-factor model. Although calendar-time methodology avoids problems related to cross-sectional dependence of observations, it has some other disadvantages. Using simulation, Loughran and Ritter (2000) show that multifactor models have low power. Also, it is hard to interpret what exactly these models are measuring. Although Fama and French suggest that their three factors are risk factors, it can be as easily argued that these factors are behavioral factors. II. Stock-price Performance of Secondary Equity Issuers Panel A of Table 3 presents announcement period cumulative abnormal returns and holding period returns for the year prior to the secondary offering. On average, the market reacts negatively to the announcements of secondary equity offerings. Average three-day CARs range from -1.57% (estimated as a difference from CRSP value weighted index returns) to -1.70% (estimated using the market model) and are significant at the 0.01 level. This finding is consistent with earlier studies (e.g., Mikkelson and Partch (1985), Asquith and Mullins (1986)). Stocks sold in secondary offerings perform very well over the year prior to the offering. The average holding period return over the prior year is 49.48%. Although

16 16 matching firms also perform well during the same period, offering firms significantly outperform the matching firms. The average one-year HPAR measured relative to bookto-market and size matched firms is 16.47% and significant at the 0.01 level (t = 4.75). Similar abnormal performance is observed when matching firms are industry and size matched, and size matched (not presented in a table). To evaluate post-offering performance of secondary equity issuers, we examine HPRs and HPARs ranging from one to five years. HPARs are estimated relative to firms matched by book-to-market and size, prior return and size, industry and size, and size. Panel B of Table 3 presents the test results for post-offering performance. Stocks sold in secondary equity offerings earn 24.94% in the first post-offering year and % over five years. Moreover, offering firms outperform matching firms in all the holding periods examined. The evidence of abnormal performance is strongest in the first postoffering year. One-year average HPARs range from 8.18% to 13.31%, depending on the benchmark. They are statistically significant at the 0.01 level. Over the five-year holding period, offering firms outperform the benchmarks by 15.23% to 36.77%, on average. We estimate statistical significance of HPARs using conventional t-statistics and bootstrapped skewness adjusted p-values. The significance of the results is about the same with both statistics. To test if abnormal post-issue performance can be explained by industry-wide missvaluations or momentum effect, we estimate abnormal performance relative to prior returns and size, and industry and size matched portfolios. Using these portfolios slightly (not statistically significantly) reduces the size of the abnormal returns, but that does not eliminate the abnormal returns.

17 17 Finding that significant negative announcement effects are followed by significant positive post-offering HPARs is inconsistent with market efficiency. Positive postoffering HPARs are also inconsistent with sellers in secondary equity offerings timing their sales to take advantage of overvaluation of the stock by the market. Rather, sellers appear to be selling undervalued shares. On the other hand, the observed pattern of significant positive pre-offering abnormal performance followed by significant postoffering abnormal performance is consistent with the Odean s (1998) findings. We also estimate abnormal one-year post-issue performance using Fama and French (1993) three-factor model and four-factor model including momentum factor (Panel C of Table 3). Again, we find that secondary equity issuers have positive abnormal post-issue performance. Using Fama and French three-factor model, we find that monthly abnormal return is 0.5% (statistically significant at less than one percent level) and one-year abnormal return is 6.2% [( ) 12-1]. Four-factor model shows that momentum factor is highly significant in explaining post-issue performance of secondary equity issuers. However, Fama and French factors together with momentum factor cannot fully explain abnormal post-issue performance. III. Operating Performance and Analysts Earnings Forecasts Loughran and Ritter (1997) show that primary and combination equity offerings tend to occur after the periods of very good operating performance. The authors suggest that underperformance after these offerings might be due to investor giving too much weight to recent earnings in predicting future earnings. For this explanation to be

18 18 applicable for secondary equity offerings, pre-issue performance of secondary equity issuers should be poor or deteriorating. Panel A of Table 5 shows operating income to assets ratio, return on assets (net income divided by assets), capital expenditures and research and development expenses to assets ratio, and market to book ratio (market value of equity divided by book value of equity) for offering firms and the difference in these ratios of offering and matching firms before and after the issue. These variables are calculated in the same way as in Loughran and Ritter study. We find that, despite different stock-price performance, secondary equity issuers have a very similar operating performance before and after the issue as primary equity issuers in Loughran and Ritter sample; performance is improving before the issue and deteriorating (although somewhat slower than for primary equity issuers) after the issue. However, differently than primary equity issuers, secondary equity issuers have better operating performance after the issue than their matching firms. III. Potential Explanations of Abnormal Stock-price Performance: Difference in Risk, Tax Motivated Selling, and Trading at the End of the Year A. Risk Positive post-offering HPARs indicate overperformance only to the extent that the risk of offering firms is not greater than that of matching firms. If offering firms are riskier, their survival rates will be lower. To ascertain that our results are not biased because some firms are delisted from CRSP database within five years of the offering, we examine how many firms in offering and matching samples are delisted and the reasons

19 19 for their delisting. We place the delisted firms into two groups based on the reason for their delisting: 1) mergers and exchanges (delisting codes ), and 2) liquidations and dropped issues (delisting codes 400 and above). In the mergers and exchanges group, the most common subgroups are issues acquired in merger (delisting code 200) and issues acquired by exchange of stock (delisting code 300). When a firm is acquired, the acquisition price is typically set higher than the market value of the firm prior to the merger announcement. If offering firms are more (less) often delisted because of mergers and exchanges than matching firms, abnormal post-offering returns will be understated (overstated). We expect the effect of liquidations and dropped issues to be the opposite. If offering firms are more (less) often delisted because of liquidations and dropped issues than matching firms, post-offering abnormal returns will be overstated (understated). Table 5 shows that matching firms tend to be more often delisted because of liquidations or dropped issues than offering firms (statistical significant only for industry and size matched sample). Offering firms are more often delisted because of merger or exchange than matching firms (significant at least at 5% significance level for all matching samples). These findings suggest that actual positive abnormal performance after secondary equity offerings may be understated in our study. We also examine size and market-to-book ratios of offering and matching firms (not presented in a table). As size is a matching factor for all matching samples, median values for market value of equity are similar for all offering and matching samples. On the other hand, the median book-to-market ratio is lower for offering firms than for matching firms. Lower book-to-market ratio might mean higher risk if it is due to more of the company value coming from high growth opportunities. However, we find that

20 20 low book-to-market ratios are not associated with higher performance (when controlling for size). To insure that overperformance after secondary offerings is not due to the higher risk of offering firms, we also examine the distribution of post-offering returns for issuing and matching firms. If offering firms are riskier, the worst performing offering firms should have worst performance than the worst performing matching firms. We examine this hypothesis in Table 5. Offering and each matching sample firms are divided into deciles based on their five-year stock price performance. Worst 10% performers in the offering sample perform better than worst 10% performers in any matching sample; the risk based explanation of high offering sample returns predicts the opposite. We find essentially the same results when we examine the worst performing 5% (not presented in a table). The distribution of returns by deciles also shows that very good post-offering performance is not due to the performance of just a few extremely well performing firms; offering firms perform better than matching firms in most deciles. B. Tax Motivated Selling We examine whether there are non-behavioral explanations of investor tendency to sell shares of firms that subsequently outperform matching firms. In some cases it may be optimal for investors to sell undervalued shares even when they know that shares are undervalued if other benefits outweigh this loss. In this section we examine tax motivated selling, and in the next section we examine trading at the end of the year. The tax law was changed in 1986 making selling winning investments more costly after that year. Because of this change, it could be rational for investors to sell

21 21 winning investments in We find that the proportion of past winners in secondary equity offerings (95.83%) is significantly higher in 1986 than in other years (80.20%) (Panel A of Table 7). However, the proportion of past winners remains very high even with the exclusion of We also find that positive abnormal post-offering performance cannot be explained by 1986 sales. As shown in Panel B of Table 7, postoffering abnormal performance is lower for 1986 firms than for the other firms. C. Trading at the End of the Year There may be special reasons for selling at the end of the year. Investors may want to sell past losers at the end of the year to reduce their taxes and to eliminate losing stocks from their investment list. On the other hand, if investors tend to rebalance their portfolios at the end of the year, they may sell past winners to restore the original proportion of these stocks in their portfolios. We examine whether the tendency to sell winners and post-offering performance are different for the offerings made in December. Panel A of Table 7 shows no significant difference between the percentage of winners for the offerings made in December (83.78%) and the percentage for the other offerings (80.70%). We exclude 1986 offerings from this analysis because of the different effect of tax considerations on the investor tendency to sell winners in this year. To take advantage of lower capital gains taxes, investors are expected to sell more past winners in 1986 than usual. In Panel B of Table 7, we show that positive abnormal post-offering performance is higher for the offerings made in December, but it is not significantly different from the performance reported for other offerings.

22 22 IV. Disposition Effect and Momentum in Stock Returns A. Disposition Effect In Panel A of Table 7 we examine the proportion of past winners in our sample of secondary equity offerings. Past winners are defined as firms that had positive returns over the year prior to the secondary offering. Pre-offering returns are available for 524 sample firms. Past winners make up 80.92% of our sample. This percentage could be so high not only because of investor tendency to sell winners, but also because firms in general enjoyed positive annual returns during this period. To rule out this possibility, we examine annual returns for all CRSP firms during the sample period. To make the CRSP sample more comparable to our sample, we use the following procedure. For each year we calculate the percentage of winning stocks in CRSP database in the prior year. To calculate the average percent of past winners, we weight each year by the number of offerings in our sample in that year. Past winners make up 61.24% of the CRSP sample (Panel A of Table 7). The percentage of past winners among secondary equity issuers is significantly higher than the percentage of past CRSP winners. This finding supports the disposition effect. In Table 8, we examine whether the tendency to sell winners is justified by the subsequent stock price performance. Past winners seem to perform especially well in the one to five year post-offering period. Their one-year average HPARs are 12.82%, while past losers one-year average HPARs are only 1.20%. The difference in the abnormal returns is statistically significant for the first four years. We estimate average HPAR for each firm as an average of HPARs estimated relative to book-to-market and size, prior

23 23 returns and size, industry and size, and size matched firms. The tendency to sell winners is not justified by the post-offering stock price performance. In Table 9, we test the relation between pre-offering stock price performance and post-offering performance, while controlling for other variables. The dependent variable in the first regression analysis is post-offering three-year average HPAR. We get very similar results with HPAR estimated relative to book-to-market and size matched portfolio (portfolio that is usually used to control for risk) (not presented in a table). We control for two firm characteristics: logarithm of market value of equity and book-tomarket ratio. We also control for offering circumstances and characteristics. Relative issue size is estimated relative to the market value of equity. We use a December issue dummy equal to one if the offering is made in December and equal to zero otherwise. In some cases, the shares that investors are offering in the secondary equity issue have been acquired as a result of some prior transaction (for example, convertible bonds were exchanged for stock). Keeping such shares may be inconsistent with investors' goals. To control for these cases, we use a prior transaction dummy equal to 1 if shares have been acquired as a result of some prior transaction and equal to 0 otherwise. Sellers can be expected to be especially knowledgeable about the value of offered shares when they are the managers of the offering firm. A manager sale dummy is equal to one if at least one of the sellers is a manager or significant shareholder (owner of at least 30% of shares or founding shareholder) of the firm. We obtain information on the last two variables from Dow Jones Newswires and the Wall Street Journal (WSJ), which we search for the offering month and five months prior to the offering. Even when controlling for these variables, past-winners tend to have significantly better post-offering performance.

24 24 Positive post-offering abnormal performance is also stronger for small firms. However, overperformance after secondary equity issues is not limited to small firms. Only 7% of our sample offering firms have lower market values of equity than the median market value of equity for CRSP firms in the offering year. A subsample of secondary equity offerings restricted to firms larger than the median CRSP firm still significantly overperform matching firms after the offering (not presented in a table). B. Explaining the Disposition Effect Odean (1998) offers two possible explanations for the disposition effect: the prospect theory (Kahneman and Tversky, 1979) and the expectation of mean reversion (Andreassen, 1988). The prospect theory suggests that winning investments have lower value only to current shareholders, the rest of the market should have no reasons to avoid these stocks. The prospect theory predicts that the market should react less negatively to the announcements of secondary offerings of past winners (in this case selling shareholders have personal reasons to sell the shares) than to the announcements of secondary offerings of past losers. Investors may prefer to sell winners and hold losers because they believe stock prices exhibit mean reversion: current losers outperform current winners in the future. Andreassen (1988) finds that in experimental settings people buy and sell stocks as if they expect short-term mean reversion. There is some empirical evidence of mean reversion in stock returns (Jegadeesh (1990), Lehmann (1990), and DeBondt and Thaler (1985, 1987)). However, Jegadeesh and Titman (1993) show that over an intermediate horizon of three to twelve months past winners, on average, continue to outperform past

25 25 losers. We find that past winners continue to outperform past losers in secondary offerings over a post-offering horizon of one to five years. Although past winners outperformance is strongest in the first post-offering year, there is no subsequent underperformance in the later years. Expectation of mean reversion, however, does not have to be based on empirical evidence; investors could have irrational expectation of mean reversion. Unlike the prospect theory, the mean reversion expectation suggests that not only current owners, but also other investors should disfavor past winners. A finding of greater negative market reaction to the announcements of secondary offerings of past winners is more consistent with the expectation of mean reversion than with the prospect theory. In Table 8, we also report announcement period CARs for announcements of sales of past winners and past losers. We find that market reaction is significantly more negative for the announcements of sales of past winners (-1.64%) than for past losers (- 0.94%). In Table 9, we examine the relation between announcement period CARs and preoffering performance while controlling for other variables. Controlling variables include book-to-market ratio, equity value of firm, relative offering size, December offerings, shares received as a result of prior transactions, sales by managers and significant shareholders and offer discount. Even when controlling for other variables, we find a significant negative relation between announcement period CARs and past stock price performance. This result is consistent with the expectation of mean reversion. C. Momentum in Stock Returns

26 26 The disposition effect by itself cannot explain abnormal post-issue performance. It only suggests that investors have a preference to sell past-winners. However, some studies show that past stock-price performance can predict future stock-price performance. The disposition effect together with momentum effect potentially could explain the abnormal performance after secondary equity offerings. Earlier studies show that over an intermediate horizon of three to twelve months, past winners continue to outperform past losers (Jegadeesh and Titman (1993)), but long-term past losers outperform past winners in the subsequent three to five years (DeBondt and Thaler (1985, 1987). For our sample of secondary equity offerings positive post-issue performance continues beyond three to twelve months that the momentum effect predicts (Panel B of Table 3). Also, if abnormal performance is due to the momentum effect, matching by past stock returns should eliminate this abnormal performance. We estimate holding period abnormal returns relative to the returns of sample matched by size and six month pre-issue returns. This matching sample also has very similar one-year pre-issue returns as offering sample. Inconsistent with the momentum explanation of abnormal returns, such matching does not eliminate abnormal performance (Panel B of Table 3). V. Explaining Performance after Primary and Secondary Issues There are many factors affecting manager and shareholder decisions to offer common stock. However, here we focus on those factors that help to explain the difference in the stock-price performance after primary and secondary issues.

27 27 Equity issues (or other securities issues) are not randomly distributed among firms and in time. We show that manager and shareholder timing of primary and secondary equity issues may explain why firms are on average overvalued at the time of primary equity offerings and undervalued at the time of secondary equity offerings. For simplicity, let s assume that on average managers and investors have correct expectations about the future performance of firms, and that overvaluation and undervaluation of future performance are randomly distributed. We examine the market with six firms; three of these firms have poor future opportunities (P) and the other three have good future opportunities (G) (Panel A of Table 10). As overvaluation and undervaluation are randomly distributed, we can expect the same number of overvalued and undervalued firms among P and G firms. In Panel A, we have one undervalued P firm (P u ), one correctly valued P firm (P), and one overvalued P firm (P o ). The same is for G firms; we have one undervalued G firm (G u ), one correctly valued G firm (G), and one overvalued G firm (G o ). Although misvaluation is present, both P and G firms are on average valued correctly. However, when we look at the perceived instead of actual future opportunities, the situation is very different (Panel B of Table 10). Because of undervaluation, G u firm is perceived as a firm with poor future opportunities and, because of overvaluation, P o firm is perceived as a firm with good future opportunities. Consequently, there are more overvalued firms among firms with expected good future opportunities, and there are more undervalued firms among firms with expected poor future opportunities. It can be expected that primary equity offerings will be concentrated among firms that are expected to have good future opportunities. Managers of such firms believe that they have many good projects available and need money to

28 28 invest in these projects. It also can be expected that shareholders will try to avoid selling shares of such firms and will be more willing to sell shares of firms that they think have poor future opportunities. However, as was mentioned earlier, firms with expected good (poor) future opportunities on average will be overvalued (undervalued) firms. As with time stock prices revert to true values, primary equity offerings (or primary offerings of other securities) will be followed by underperformance, while secondary offerings will be followed by outperformance. If managers and investors would behave perfectly rationally they should understand this process and eliminate any predictable abnormal performance. However, there is plenty of evidence showing that people, including professionals, are overoptimistic about the precision of their knowledge (for example, Fishhoff, Slovic, and Lichtenstein (1977), Gervais and Odean (2001)). If managers and investors are overoptimistic about their ability to identify good and poor future performers, abnormal performance after security issues may persist. This simple model is consistent with the empirical findings of underperformance after external financing both public and private, both equity and debt (for example, Ritter (1991) [IPOs], Spiess and Affleck-Graves (1995) [seasoned equity offerings], Spiess and Affleck-Graves (1999) [public debt issues], Hertzel et al. (2002) [private equity placements], and Billett, et al. (2001) [bank loans]). This model is also consistent with the overperformance of shares that investors choose to sell relative to the performance of shares they choose to hold (Odean, 1998) and overperformance after secondary equity offerings found in our study. Together with the disposition effect, this model can also explain why abnormal performance is concentrated among past winners. As investors are reluctant to sell past

29 29 losers, investors are not likely to sell such securities when they have a choice which securities to sell and if to sell them at all. According to our model if investors are not choosing which securities to sell and when to sell them, there is no reason to expect abnormal performance after the sale. VI. Conclusions We analyze a sample of 644 pure secondary equity offerings made during 1974 through On average, secondary equity issuers significantly outperform matching firms in the one- to five-year post-offering period. Abnormal performance is strongest in the first year after the issue. One-year holding period abnormal returns range from 8.18% to 13.31% depending on the benchmark. Abnormal post-offering stock-price performance cannot be explained by risk, tax motivated selling, or trading at the end of the year. We also find that offering firms significantly outperform matching firms in the year prior to the secondary offering. These two findings are consistent with the disposition effect which suggests that investors have the tendency to sell winners. Our cross-sectional analyses show that past winners have significantly better postoffering performance than past losers. The disposition effect is significant even when controlling for firm size, book-to-market, relative issue size, December offerings, shares received as a result of prior transactions, and sales by managers. Despite positive post-offering performance, market reaction to the announcements of secondary equity offerings is negative. Furthermore, the market reacts more negatively to the announcements of secondary offerings of past winners. It seems that

30 30 not only the current owners, but also other investors do not have confidence in these stocks. Odean (1998) offers two explanations of the disposition effect: the prospect theory (Kahneman and Tversky, 1979) and the expectation of mean reversion (Andreassen, 1988). More negative market reaction to the offerings of past winners is not consistent with the prospect theory. According to the prospect theory, past winners have lower value only to the current owners of these shares. If the market is aware of shareholders having personal reasons to sell the shares, it should react less negatively to the announcements of such sales. Our findings are consistent with the expectation of mean reversion. We also show that the disposition effect together with the momentum effect cannot explain abnormal performance after secondary equity offerings. Secondary equity issuers significantly outperform firms matched by prior performance and size. We show that different abnormal performance after secondary and primary security issues can be explained by manager and investor overconfidence in their ability to identify firms with good and poor future prospects. In a simple model we show that under certain assumptions firms with expected good future prospects (they are more likely to make primary issues) on average are overvalued, while firms with expected poor future prospects (their shares are more likely to be sold in secondary offerings) on average are undervalued. Together with the disposition effect our model can also explain why abnormal performance after secondary equity issues is concentrated among past winners.