Capital Redeployment in the Equity Market *

Transcription

1 Capital Redeployment in the Equity Market * Huaizhi Chen Harvard Business School This draft: November 6, 2017 First draft: August 31, 2017 * I thank Lauren Cohen, Robin Greenwood, Dong Lou, Christopher Malloy, Christopher Polk, Andrei Shleifer, and Erik Stafford for their valuable comments and suggestions. The associated research was conducted during my postdoctoral fellowship supported by the Behavioral Finance and Financial Stability Initiative at Harvard Business School.

2 Capital Redeployment in the Equity Market ABSTRACT Payouts, in the form of dividends and buybacks, reached a height of almost a trillion dollars per annum in the recent years. In an equity market where most assets are held by dedicated financial intermediaries, much of these dollars are directly reinvested into the stock market itself. Consistent with demand-driven price pressure, capital repayments are accompanied by predictable excess returns in stocks connected to these payments through investors. Quintile portfolios formed of stocks without any capital return but sorted on their investors exposures to capital redeployment experience a cross sectional spread of 3.27% (t = 3.16) adjusted returns per quarter from Q to Q Due to the persistence of these capital return programs, abnormal returns accumulate over significant holding periods and revert partially only at a very long horizon. Additionally, the exposure to capital redeployment by connected firms is associated with changes on payout and issuance behavior. While firms connected to capital returns negligibly increase their own buyback and dividend activities, they significantly increase their equity issuance in the medium to long horizon. JEL Classification: G10, G14, G23, G31, and G35. Keywords: Mutual Funds, Payout Policy, Dividend Policy, Stock Buyback, and Spillover Effects.

3 Substantial corporate cash were distributed to investors through dividends and stock buybacks in recent years. Observers of the financial system describe a significant relationship between these capital return programs and the booming stock market. The Wall Street Journal and the Financial Times call stock buybacks and dividend payments a key pillar supporting the bull market and pillar of support for the US stock market s dramatic post-crisis recovery. 1 Such statements are puzzling in the context of classical finance literature. Theories of capital structure posit invariance between policies supporting payouts and non-payouts (Miller and Modigliani 1961) and (Black 1976). A condensed version of these authors argument is that because payouts in the hand of investors are the same as their access through equity ownership, capital return should have no effect on the underlying stock price or on the returns of the entire capital market. In practice, cash payouts from equity firms are often redeployed by asset managers back into the stock market. Dividends are automatically disbursed to the portfolios holding the original stock. In mutual fund shares, dividends from the underlying stocks appear as a part of the Net Asset Value (NAV) when such dividends are recorded. After a pre-arrange distribution date, the portion of NAV represented by the dividend dollars are reallocated as new shares if the mutual shareholders elect to keep these dollars within the fund; otherwise the dividend value per share are returned as cash 2. In the event of stock buybacks, where firms repurchase their own shares, participation by investors is less mechanical. However, much like dividends, these buyback events ultimately transfer cash from the public firms to the portfolios holding these assets. The dollars channeled through payouts are substantial. As a reference, 4.25 trillion dollars from dividends and buybacks were sourced from public firms between 2010 and In the same period, only 744 billion dollars were deposited into mutual funds in net by investors. While previous studies have shown that investor flow tends to affect the cross section of price in equity assets, very little attention has been given to cash payouts by public firms. Given that these payouts are significantly larger in magnitude and much more persistent than investor flows, it is worth considering whether the capital redeployment of cash payouts by asset managers may be a principal source of demand effect in the cross section of equity assets. 1 Wallstreet Journal Money Matters: Stock Buybacks Slow. Should We Be Worried? on August 21, 2017 and Financial Times: Buyback outlook darkens for US stocks on June 21, Although the Investment Company Act requires funds to redistribute cash from both capital gains and dividends to the ultimate investors, most mutual funds have programs for investors to reinvest these proceeds automatically. Empirically, I find that 89% of these distribution dollars are reinvested in the mutual fund on average in my sample period. The outflow from distributions are captured as investor capital outflow. See Section 1.3 and Appendix A5.

4 This paper forwards and tests a demand-based hypothesis to interlink capital repayment to stock market returns 3. According to this hypothesis, capital returns drive demand in the stock market through cash redeployment. Public firms, by initiating capital return, force investors to deploy cash into a potentially limited set of assets. The cash deployment in turn drives up short term demand for certain stocks. Prices increase reflecting the non-fundamental demand (Shleifer 1986). Arbitrageurs cannot trade away this price effect due to their limited risk appetite and imperfect specialization (Shleifer and Vishny 1992), (Shleifer and Vishny 1997); therefore, return predictability is observed. To test this demand hypothesis of capital returns, I trace the cash flow from dividend and stock buybacks to individual mutual fund portfolios. I estimate the redeployment of cash flow from payouts back into the equity markets through the quarter to quarter changes in fund portfolio holdings. I show that cash payouts predictably relate to changes in the mutual fund holdings of certain assets. Appetite for stocks with dividends and buybacks is persistent among mutual funds: a portfolio exposed to significant (insignificant) capital returns will continue to have high (low) payouts by its underlying assets for many quarters. When a high capital return portfolio receives dividends, it keeps this cash predominantly invested in a predictable set of holdings similar to their existing assets. Such a fund, also tends to participate in buybacks, and uses the proceeds to purchase other existing assets. In contrast, portfolios with low capital returns tend to disproportionately liquidate their existing holdings from quarter to quarter. This effect is related to the investor capital flow induced price pressure shown in the literature (Coval and Stafford 2007), (Lou 2012), and (Edmans, Goldstein and Jiang 2012), but is significantly more persistent, and predicts cross sectional spread in asset prices at significant horizon. The empirical effects in this paper indicate that cash from capital returns gets predominantly reinvested (in net) in stocks linked by existing mutual fund holdings. The timing of capital redeployment is as follows. Dividend payments are recorded by a mutual fund as part of their Net Asset Values (NAV) immediately on the dividend date of record. On average, this cash is used to purchase assets within the same quarter. A buyback program 3 There are many articles in the popular press speculating how capital return affects stock prices through demand. For a limited representative set of recent articles, see CBS MoneyWatch s How record stock buybacks inflate the markets on November 4, 2015; Bloomberg s Buybacks Back in Hong Kong Stocks, Giving Rally Extra Boost on June 30, 2017; USA today s 'Big Money' Digs Deep to Prop Up Stock Market on Feb 12, 2016; and New York Time s Buybacks by Companies Like Apple May Signal Danger, Not Growth on June 23, 2016.

5 creates an exchange between asset holders and the public firm. I observe that during the quarter of a buyback event, there is a dramatic decrease in mutual fund holdings of the buyback stock and no significant decrease in these funds overall holdings, indicating that the proceeds from this event are immediately reinvested into other assets. This redeployment mechanism by mutual funds is related to changes in stock prices. I show this by aggregating capital returns to the stock level. Capital return induced price pressure, CCCCCCCC, is calculated for each stock by assuming proportional investment in existing assets (Frazzini and Lamont 2008)/ (Lou 2012) from capital repayments. A key innovation in testing price pressure from capital returns is that I examine correlation between CCCCCCCC and returns specifically in stocks from firms that do not conduct capital returns. These growth stocks are attractive laboratories as they share redeployment inflows through investors but have not explicitly changed dividend or repurchase policies. Stocks associated with large amounts of capital returns tend to appreciate in the same quarter. One standard deviation in CCCCCCCC implies 1.79% (t = 2.98) excess return in the underlying stock. Because CCCCCCCC is extremely persistent- past 1 quarter values of CCCCCCCC explain 44% of CCCCCCCC s current variation- stocks associated with large capital returns predictably experience large excess returns in the following quarters. One standard deviation in CCCCCCCC forecasts a 1.16% (t = 2.33) increase in excess returns in the next quarter. This effect is increased to 1.29% (t = 2.71) once I control for contemporaneous investor flows. Furthermore, CCCCCCCC continues to forecast 0.88% (t = 2.62) increase in quarterly excess returns over an entire year. The price predictability associated with capital returns indicates a potential calendar time trading strategy. Quintile portfolios sorted on lag CCCCCCCC have large return spreads in the short to medium horizon. A strategy holding the top quintile and shorting the bottom quintile CCCCCCCC sorted portfolios of non-capital returning stocks yields a return of 3.27% (t = 3.16) per quarter. This strategy can be held for multiple quarters; however consistent with non-fundamental demand, it reverts to statistical insignificance at the very long horizon (See Figure 3). In sum, this paper forwards the demand hypothesis of capital redeployment and presents strong empirical results consistent with the implications of this channel. The last portion of this paper associates this spill-over channel of return predictability with future capital structure decisions. Ex ante, it may be that investors are redeploying cash into firms with better expected future payouts than their peers; however, the ex post changes in capital structure decisions measured over up to 12 years does not support this argument. Firms measured

6 to have the greatest amount of payout induced price effect have economically insignificant changes in their own buyback and dividend payout policies. Instead, I find that these firms significantly increase their issuance activities over other firms on average at the 12 quarter to 48 quarter horizons. This is suggestive of opportunistic behavior by firms to the potential market mispricing. Contrasting the demand hypothesis of capital redeployment, works studying the informational content of capital returns dominate the existing academic literature. Beginning with (Ross 1977) and (Bhattacharya 1979), many argue that capital return policy signals information about the underlying firm. The idea is that in order to overcome information asymmetry with investors, firms with strong expected cash flows commit to payouts because payouts are costly signals of these firms future opportunities, while firms with weak expected cash flows cannot commit to a return policy. In equilibrium, investors price firms by their future cash flow as implied by their current payout policy. Consistent with these theories, empirical research finds that stocks appreciate significantly during announcements of capital return (Vermaelen 1981). However, these stocks also tend to have abnormal returns long into the future- up to 4 years (Ikenberry, Lakonishok and Vermaelen 1995), which indicates either that investors underreact to the information content of payouts or that signaling may not fully explain the price discovery mechanism. These signaling theories also neglect the popularity of buybacks as a form of capital return. The tax shield advantages of buybacks over dividends were eliminated in , and yet buybacks still became more popular than dividends as the preferred method of capital return. In surveys of CEOs, there is a widespread consensus among executives of public firms that stock buybacks are advantageous over dividends because they are a more flexible way of returning capital (Graham and Harvey 2002) and (Brav, et al. 2005)- firms can decrease their buyback activities without suffering significant investor outcry. These stylized facts suggest the existence of unexplored mechanisms originating from capital payouts. Signaling alone cannot rationalize price predictability, and must be joined with a form of investor under-reaction to explain the main empirical facts outlined in the paper. In other words, if firms signal their type through payout policies, investors must also underreact to this signal because prices do not adjust immediately. While I cannot rule out all potential mechanisms involving signaling and investor under-reaction that are consistent with observed price effect on 4 The Jobs and Growth Tax Relief Reconciliation Act of 2003 effectively ended the spread difference between the capital gains and the dividend tax rates.

7 stocks, the results in this paper can reject several explicit versions of this mechanism. The most basic form of signaling requires firms to pay dividends or conduct stock buybacks to signal their own underlying fundamentals. Since capital return implied price pressure affects firms that do not return capital, this version of the signaling hypothesis cannot explain the documented return predictability mechanism. Another version of the signaling/under-reaction hypothesis states that the capital return program by one firm signals future capital return by other firms operating in closely related industries (Massa, Rehman and Vermaelen 2007). There is return predictability on stocks connected to capital returns because they are undervalued in accordance to the available information about their peers, but investors are slow to react to this information. There are multiple reasons why the price effect found in this paper is incompatible. Here are a few first order ones: 1) I show that the degree of future capital return by firms connected to repurchases and dividends is negligible and quantitatively miniscule with their respective increase in prices. 2) Since announcements to change dividend policy and to conduct open market buybacks occur potentially a year prior to the actual program, the signaling hypothesis indicates that the timing of the price predictability on the connected firms should follow the timing of the firm announcements. The documented price effect coincides with the timing of the actual cash redeployment activity. 3) If stock payouts are a signal of the profitability of related firms, then this signal is available to all investors. Instead, I observe that investors with capital return inflows significantly scale up these holdings over investors with low capital return inflows. 4) If the return predictability of stocks associated with capital returns is simply an under-reaction to information contained in the buybacks of similar firms, then the price effect should be strictly positive. The price effect documented tends to revert in the very long run for value-weighted portfolios. A set of related studies investigates the timing of stock issuance and buybacks, the latter of which is a large component of capital return. These works conclude that firm managers initiate stock repurchases (issuance) when they believe their firms are undervalued (overvalue) or when they have incentive misalignment with investors (Loughran and Ritter 1995), (Baker and Wurgler 2000), and (Kahle 2002). I abstract from the timing of buybacks by focusing on firms that do not conduct stock buybacks and shed light on the mechanism of capital redeployment. However, a study in the field of stock market timing that is particularly related to this paper is (Greenwood and Hanson 2012), which finds that firms with negligible buybacks and issuances have factor returns correlated with the net issuance pattern of firms with similar characteristics. The empirical

8 results presented in paper are consistent with their finding as investors tend to have style portfolios related to stock characteristics; however, with a bottom up approach, this paper sheds light on the underlying pricing mechanism in several ways. 1) It documents the association of investor portfolio rebalancing patterns with a style characteristic (capital return), and that this rebalancing pattern is linked to return predictability. 2) It shows that dividends, in addition to buybacks, have predictive power on the returns of related firms. 3) It documents the very long-term reversal of excess returns in these non-capital-returning firms. This paper is also related to a literature on how investors use dividends. The fact that exposures to dividends is a persistent characteristic about investors complements the dividend disconnect phenomenon, which describes the tendency of investors to treat dividend-based returns differently than price-based returns, documented in (Hartzmark and Solomon 2017). The rest of this paper is divided into four sections. The next section analyzes the capital redeployment mechanism at the investor level and shows how dividend and buyback cash are channeled through mutual funds. Section 2 demonstrates price predictability by aggregating the capital return variables into the stock level and calculating the capital return implied price pressure on each stock. I show that this variable is extremely persistent and particularly informative about stock returns at the short 1-quarter to medium 1-year horizon. The positive price effect reverts after significant holding periods. Section 3 reviews the characteristics of the stock receiving redeployed capital and show that the spread between returns is quantitatively large compared to the changes in future payout policies. However, firms under the influence of redeployed capital tend to significantly increase their own issuance activities. Section 4 concludes and discusses the results of the paper. 1. Capital Return and Mutual Funds This section examines the economic magnitudes of capital returns and how this source of cash flow is related to investment by money managers. 1.1 Capital Return in Aggregate

9 In aggregate, capital returns are significant and persistent sources of cash inflow for investors. To show this, I obtain stock-related data from the Center for Research in Securities Prices (CRSP) Stock Security Files. Dividends and buybacks are calculated quarterly. Dividend yield per stock is calculated as the difference between total returns (RRRRRR ii,tt ) and price returns (RRRRRRRR ii,tt ) each quarter, that is DDDDDDDD ii,tt = RRRRRR ii,tt RRRRRRRR ii,tt. Percentage buybacks is calculated as the decrease in shares outstanding. The lower limit for the decrease is set to -10% to limit the exposure of the sample to potential mergers and acquisition. Specifically, this is BBBBBBBBBBBBBB ii,tt = SShaaaaaaaaaaaaaaaaaaaaaaaaaaaaaa ii,tt SShaaaaaaaaaaaaaaaaaaaaaaaaaaaaaa ii,tt [ 10%, 0), where SShaaaaaaaaaaaaaaaaaaaaaaaaaaaaaa ii,tt is the percentage change in split adjusted shares outstanding. The dollar values of dividends and buybacks per stock are calculated by multiplying the stock s buyback and dividend yields with its t-1 market capitalization. Equity mutual fund flows, which have been previously demonstrated in finance literature to affect asset returns through demand, are calculated as TTTTTT ii,tt TTTTTT ii,tt 1 + RRRRRR ii,tt MMMMMM ii,tt, ii where MMMMMM ii,tt is a compensating term for fund mergers. Figure 1 plots the aggregate capital returns from common stocks traded on the AMEX, NASDAQ, and NYSE exchanges and the aggregate net investor capital flow for equity funds. Between 1990 and 2002, annual buyback cash flow ranged from $17 to $159 Billion, while dividend payouts ranged from $92 to $161 Billion. Capital return programs increased dramatically after , paying investors $160 to $497 Billion with buyouts and $172 to $419 Billion with dividends. Investor flows to equity funds, which have been demonstrated to affect stock prices through demand (Coval and Stafford 2007) and (Lou 2012), is plotted as a benchmark. The two sources of capital flow, in aggregate, are quantitatively similar in absolute quarterly magnitude; however, capital returns are significantly less volatile and much more persistent than investor 5 The Jobs and Growth Tax Relief Reconciliation Act of 2003 reduced the overall tax rate for capital gains and dividends.

10 flows. Given a longer measuring span, aggregate capital return accumulates to a significantly larger magnitude compared to investor inflow, which flattens out. For instance, between 2010 and 2015, 4.25 trillion dollars of capital returns accumulated in net from public firms compared to 744 billion dollars of investor inflow to mutual funds. The relative magnitude of firm payouts is almost 6 times as high as investor capital flow. In Figure 2, we also observe that most of the capital returned come from only a small percentage of publicly traded firms. The top panel of figure 2 shows that 80% of the stocks traded participate in quantitatively insignificant amount of capital return; and on average, 5% of stocks conduct more than 50% of the capital return to investors in the financial sector. I will be conducting return predictability tests on specifically assets that do not return capital. In summary, payouts from public firms are significant and large sources of inflow capital into the equity markets. The amount of cash being contributed from these capital return programs are significantly more persistent and less volatile in aggregate than measures of investor demand. The cumulative magnitude of capital return sizably surpasses capital flows from investors and may be a potentially significant source of aggregate demand. 1.2 Exposure to Payouts by Individual Mutual Funds In this section, I show that professional investors (open-ended mutual funds) vary significantly in their exposure to capital returns. This capital return exposure is a persistent characteristic of each portfolio. A fund exposed to large amounts of cash payouts continues to be exposed to new cash payouts, while a fund with little capital return continues to have little capital return exposure. In addition to the previously described data, I use the N-Q quarterly mutual fund holding filings recorded by the CDA/Spectrum database and the Center for Research in Security Prices (CRSP) Survivor Bias Free Mutual Fund database for this section. Domestic open-ended mutual funds are required to disclose their equity holdings each quarter in the N-Q filings and these filings are captured in the CDA/Spectrum database. The holdings data is matched to the Center for Research in Security Prices (CRSP) Survivor Bias Free Mutual Fund database for information on fund style, total net assets, monthly returns, expense ratios, and other fund characteristics. The

11 matching between CDA/Spectrum and CRSP is conducted using MFLinks provided by the Wharton Research Data Services (WRDS). Each portfolio s exposures to dividends and buybacks are calculated as the pro rata implied yield of the portfolio holdings, that is and DDDDDD_RRRRRR jj,tt = WWWWWWWWhtt ii,jj,tt 1 DDDDDDDD ii,tt, ii BBBBBB_RRRRRR jj,tt = WWWWWWWWhtt ii,jj,tt 1 BBBBBBBBBBBBBB ii,tt ii for portfolio j at quarter t. WWWWWWWWhtt ii,jj,tt 1 is the weight of asset i in portfolio j at t-1. The calculation can be interpreted as the dollar dividend return and dollar pro rata buyback for each portfolio as a percentage of the portfolio s Total Net Assets. These two measurements are significantly correlated (ρ = 0.22) for equity fund portfolios implying that a fund exposed to dividends are also exposed to buyback dollars. Table 1 describes the summary statistics on the capital return exposure experienced by equity funds in my sample. An average inflow from dividends is 0.34% of a mutual fund s TNA each quarter, while the pro-rata dollar amount of buyback dollars is 0.42%. These sources of cash are larger in the second half of the sample. The average exposure to these capital return programs is similar in magnitude to the average investor flow, which is about 0.65% of the portfolio s TNA on average. However, capital returns are significantly less volatile and more predictable from quarter to quarter. The autocorrelation coefficients show that exposure to capital return in each mutual fund can be forecasted up to 1 year with significant accuracy. In summary, there is significant heterogeneity among investors in the amount of cash payouts they receive. This heterogeneity is persistent for individual mutual funds. I investigate where all the cash returns are deployed in the next section. 1.3 Changes in Holding Values This section shows that capital returns require investors to redeploy a significant amount of cash into assets. Dividends are invested into holdings in the same quarter as they are recorded.

12 During asset buybacks, mutual funds effectively scale down their positions and redeploy cash from these proceeds into other holdings. Cash from firm payouts are redeployed by investors into assets. In practice, mutual funds have flexibility in managing their dividend and asset selling proceeds. While the Investment Company Act regulates asset managers to return capital gains and dividends to investors at prearranged distribution periods for taxation purposes, each individual fund has its own private distribution management methods. A single fund can keep dividends invested in cash prior to a distribution event; it can invest immediately and re-obtain the needed distribution cash by selling assets before a set distribution date; or it can hedge its cash obligations with option instruments. I show in this section that, empirically, mutual funds on average invest dividend cash into assets as this cash arrives, and leave insignificant amounts of dividend cash as precautionary cash for distributions. This may be because most dividend cash are automatically reinvested by fund investors. Appendix A5 approximates the distribution based flows from investors by calculating the difference between NAV (net asset value) price-returns and net mutual fund returns. The amount of distribution based inflow is roughly 89% of the total distribution (both dividends and capital gains) released by a mutual fund; that is, almost all the distribution dollars are re-invested in the original fund itself, and only about 11% of the distribution dollars are taken out by investors. In the existing literature on mutual fund flows, the cash redistribute to investors are commonly captured as investor outflows. For instance, (Coval and Stafford 2007), (Frazzini and Lamont 2008), and (Lou 2012) make no distinction between capital outflow due to distribution and outflow due to investor redemption of fund shares. I calculate the change in asset holdings per portfolio to describe its reinvestment process. The change in the CDA/Spectrum reported equity holdings of fund j between quarter t-1 and t is calculated as HHHHHHHHHHHHHH aaaaaa jj,tt, HHHHHHHHHHHHHH aaaaaa jj,tt = NN+MM ii=1 NN ii=1 PPPPPPPPPP ii,tt SShaaaaaaaa ii,jj,tt 1, PPPPPPPPPP ii,tt SShaaaaaaaa ii,jj,tt 1 where stocks 1 through N exist in the portfolio at t-1 and stocks N+1 through M are added between t-1 and t. The changes in shares held of stocks with buybacks are calculated as HHHHHHHHHHHHHH bbbbbb jj,tt = ii {bbbbbb} PPPPPPPPPP ii,tt (SShaaaaaaaa ii,jj,tt SShaaaaaaaa ii,jj,tt 1 ) NN, PPPPPPPPPP ii,tt SShaaaaaaaa ii,jj,tt 1 ii=1

13 where {bbbbbb} is the set of stocks with buybacks. The HHHHHHHHHHHHHH aaaaaa jj,tt and HHHHHHHHHHHHHH bbbbbb jj,tt variables can be naturally interpreted as the percentage difference between the value of total and buyback assets held at time t and the value of total assets held at time t-1 if these assets were held to t. Mutual funds increase their total holdings immediately when they receive dividends and sell to stock buyback their existing shares for cash which are reinvested into stocks. When a fund receives large amounts of dividends from its underlying, it increases its total holdings in the same quarter. Panel a of table 2 shows that only 34.7% of mutual funds in the lowest quintile of dividend exposure increase their asset holdings, contrasting 39.7% of funds in the highest quintile. Funds with high exposures to dividend inflows also are less likely to decrease their asset holdings. About 52.7% of funds in the lowest quintile reports a decrease in asset holdings compared to 46.7% in the highest quintile. The average partial effect of dividend exposure to growth in holdings is plotted in figure 4a. Coefficients from the regression of total holdings growth for mutual funds on their dividend exposure during the current and past 3 quarters. The regression coefficients are shown on the y axis, and the x axis represents the lag s of the regressor DDDDDD_RRRRRR jj,tt ss. Holdings growth correlate significantly to dividend exposure in the current quarter, and not to dividend exposure in the past. The coefficient is at an approximate one to one ratio. That is, dividends from a fund s underlying stocks are directly reinvested as they are entered into a portfolio. Buyback programs, in effect, exchange cash from public firms for shares with portfolio managers. Portfolios generally decrease their holdings of the company shares when buyback programs are conduct. A stock that is currently conducting a large buyback program is more likely to be sold by its existing shareholders than a stock with small or no buyback programs. Table 2 panel b shows that for a stocks currently in the highest quintile of buyback size, about 31.5% of their current holders sold assets in net, while only 25.6% bought in net. This is in contrast to funds holding stocks that do not conduct buyback programs, where only 25.9% were net sellers and 28.6% were buyers. Figure 4b plots coefficients from the regression of the change in holdings of stocks with buybacks on buyback exposure. There is an immediate and large reduction in the holdings of firms conducting buybacks by mutual funds. A pro-rata 1 dollar exposure to stock buybacks indicates 2.8 dollars of reduction in position size. There is no significant correlation between buyback exposure and contemporaneous total holdings. Since there is selling in the portion of a portfolio formed of stocks with buybacks, portfolio holdings of assets without

14 buybacks must be increasing else total holdings would decrease. This indicates that all of fund proceeds from selling stocks to buyback programs are reinvested into other assets. These results on dividends and capital return indicate that a mutual fund has to invest cash positions equivalent to roughly a dollar of dividend for each dollar dividend paid and 2.8 dollars for each dollar of pro rata buyback exposure. I investigate where these dollars are invested in the next section. 1.4 Proportional Investment into Assets If there are no trading costs or market frictions, we expect portfolio managers to optimize entirely based on their expectation of risk and return on assets. The degree of capital return based inflow should have very little information on the type of assets a fund purchases. A stock revealed to be undervalued would likely be acquired by a fund regardless of the fund s exposure to payout programs. However, given that funds have individual style mandates and there are liquidation costs to rebalancing, a more practical benchmark may be that capital return inflow is correlated with some measure of a portfolio s existing holdings. I show that mutual funds tend to invest predictably with their payout based inflows. In particular, mutual funds with high capital returns tend to keep invested in existing assets compared to funds with low capital returns. Additionally, a stock s gross purchase by high (low) capital return exposure mutual funds is significantly related to how much the stock is already held by funds with similar capital return exposures. Table 3 panel a compares the changes in the top 5 largest stock positions from mutual funds with low capital inflow with those held by mutual funds with high capital inflows. Although funds in both groups tend to scale down their existing positions on average, there is a large differential in scaling between the two types of funds. On average, mutual funds with the lowest capital returns tend to scale down their largest positions by over 15%, whereas funds with the highest tend to scale down by only 7%. High capital returning portfolios tend to keep their existing holdings. The results from the section 1.3 also indicate that the total holdings of portfolios are growing in proportion to the amount of capital return exposed to each portfolio, a natural question is what stocks do investor channel this capital return? Table 3 panel b shows that gross buying by funds with high capital return can be predicted using the ex-ante percent of assets held by other

15 mutual funds with high capital return exposure. Here I group mutual funds into 5 bins based on their capital return exposure. I calculate the gross buying of each stock in each bin as the total positive change in holdings by the mutual funds in each bin, similar to the buying measure in (Coval and Stafford 2007). Specifically, BBBBBBBBBBBB ii,tt,bbbbbb = jj MMMMMM ΔHHHHHHHHHHHHHH ii,jj,tt, 0 jj GGGGGGGGpp bbbbbb. jj HHHHHHHHHHHHHH ii,jj,tt 1 I find that this buying of assets by each bin, BBBBBBBBBBBB ii,tt,bbbbbb, is significantly related to the exante percentage of asset i held in the same bin, that is: PPPPPPPPPPPPPPPP ii,tt,bbbbbb = jj HHHHHHHHHHHHHH ii,jj,tt 1 jj GGGGGGGGpp bbbbbb. jj HHHHHHHHHHHHHH ii,jj,tt 1 Panel b of table 3 regresses BBBBBBBBBBBB ii,tt,bbbbbb on the percentage holdings for each bin. I find that the best predictor for asset purchased is the ex-ante holding of each stock in each bin. That is, funds receiving large (small) capital returns primarily buy assets already held by funds receiving large (small) capital returns. To summarize, I find that dividends represent significant cash inflows to and that buyback programs contribute significant amount of cash requiring deployment from mutual funds. This cash is deployed into assets already co-held by similar funds based on their cash return characteristic. 2. Stock Price Pressure Given that cash from capital returns stay predominantly invested (in net) in stocks linked by existing mutual fund holdings, it is natural to ask, is there a price effect on these stocks? For market participants with excess inflow from capital returns to keep purchase new assets, they will only be able to scale up these holdings through purchases if these assets are supplied by price sensitive market participants. This occurs when stock prices increase to the increased demand. In this section, I show that stock prices predictably correlate with this capital redeployment based inflow mechanism. 2.1 Capital Return Induced Price Pressure

16 The results in Section 1 demonstrate that portfolios exposed to capital returns predominantly invest in assets held by similar portfolios. Stocks held on average by investors with high (low) levels of capital return, should experience high (low) levels of investor demand. The degree to which price correlates with this demand depends on the stocks price elasticity (Shleifer 1986). To empirically proxy this price effect, I aggregate capital return based inflow to the stock level by assuming proportional investment in assets. While mutual fund portfolios do not literally reinvest proportionally into their existing assets, as there is significant turnover and investment into new positions, this is a simple and commonly used assumption in prior measures of flow exposure by stocks; for example, (Frazzini and Lamont 2008), (Lou 2012), and (Coval and Stafford 2007) use the assumption of proportional flow exposure by assets but none of these studies observe significant increases in existing positions when given capital inflows. An alternative measure of capital return induced price pressure- as the percentage of assets held in the top quintile mutual funds exposed to capital returns- gives qualitatively the same results in this section. Like prior measurements of investor flow induced price pressure in the existing literature, Capital-return Induced Price Pressure is calculated as CCCCCCCC ii,tt = jj SShaaaaaaaaaaaaaaaa ii,jj,tt 1 SShaaaaaaaaaaaaaaaa ii,jj,tt 1 CCCCCC_RRRRRR jj,tt where SShaaaaaaaaaaaaaaaa ii,jj,tt 1 is the number of shares in stock i held by mutual fund j at t-1 and CCaapp_RRRRRR jj,tt is the degree of capital return experienced by portfolio j from t-1 to t, CCCCCC_RRRRRR jj,tt = SS DDDDDD WWWWWWWWhtt ii,jj,tt 1 DDDDDDDDDDDDDDDDDD ii,tt + SS BBBBBB WWWWWWWWhtt ii,jj,tt 1 BBBBBBBBBBBBBB ii,tt. ii DDDDDD_RRRRRR jj,tt ii BBBBBB_RRRRRR jj,tt SS DDDDDD and SS BBBBBB are scaling coefficients chosen to be 1 and 2.8 respectively. Alternative calculation of CCCCCC_RRRRRR jj,tt using different positive scaling coefficients of dividend and buyback exposure do not change the results qualitatively. This is because both dividends and buybacks individually forecast returns (Appendix A3). Table 4 contains summary statistics on CCCCCCCC, and FFFFFFFF - the flow induced price pressure generated by assuming proportional investment of investor flow in existing assets. The crosssectional spread between high CCCCCCCC and low CCCCCCCC stocks is magnitudes smaller than the spread

17 in FFFFFFFF ; however, there are several reasons to suspect that CCCCCCCC can be significantly correlated with stock-level return. 1) Much like capital return on the portfolio level, CCCCCCCC for each stock is extremely persistent, and is predictable by its lagged variable at a 1 year horizon whereas FFFFFFFF at the same horizon, is quantitatively unforecastable. 2) Cash from capital returns stay predominantly invested (in net) in stocks linked by existing mutual fund holdings. 3) While investor flows only affect mutual funds, capital return by firms affect all participants of the financial market. The redeployment by investors and the predictable reinvestment of inflows observed in the paper can very well extend to all existing institutional investors. I conduct several return predictability tests using this capital redeployment inflow variable. For these tests, I restrict the sample of public common stocks traded on the AMEX, NASDAQ, and NYSE exchanges in two ways. 1) Only stocks with no dividend payments in the past year and no stock buybacks in the past 5 years are used. 2) Stocks with market capitalization less than the tenth percentile of NYSE firms and the bottom decile of stocks ranked on mutual fund ownership are excluded to eliminate micro capitalization and liquidity issues. The final firms in my sample do not explicitly commit to capital returns, either through dividends or stock buybacks, and are large enough to abstract from simple microstructure related concerns. There are 61,757 stockquarter observations left to serve as a clean laboratory for testing the effect of capital inflow induced price pressure. In the Appendix Table A4, I relax the first restriction on stocks - that the sample filters out firms with significant capital return - to demonstrate that the identified pricing phenomenon is generalizable to the entire cross section of stock returns. 2.2 Fama Macbeth Regressions Stock prices are significantly correlated with capital return inflows. CCCCCCCC is associated with significant contemporaneous stock level returns, and also forecasts excess returns at the 1 quarter and 1 year horizons. This is because CCCCCCCC is extremely persistent, the lag value of CCCCCCCC forecasts capital return induced price pressure for many quarters into the future. In this section, I conduct Fama Macbeth regression analysis of returns on CCCCCCCC and various common characteristics (Fama and MacBeth 1973). A single standard deviation of CCCCCCCC is associated with 1.79% (t = 2.98) increased excess returns in the same quarter. It also forecasts 1.16% (t = 2.33) increased excess return in the following quarter, and an average quarterly return

18 of 0.88% (t = 2.68) over the following year. The predictability is increased to 1.29% (t = 2.73) and 1.01% (t = 3.05) once contemporaneous flow induced price pressure FFFFFFPP is added as a control. 2.3 Calendar Time Portfolios and Reversal The Fama-Macbeth regressions indicate a particular calendar time strategy. I sort stocks into calendar time portfolios using. CCCCCCCC Quintile portfolios are formed each quarter and are held for multiple quarters in overlapping portfolios following (Jegadeesh and Titman 1993). As shown in Table 7, the top quintile portfolio rebalanced quarterly and held for 1 quarter experiences a 4- factor adjusted excess return of 1.61% (t = 2.85), while the lowest quintile portfolio experiences excess return of -1.92% (t = -2.59). A strategy shorting the lowest quintile portfolio and holding the highest quintile experiences a return of 3.52% (t = 3.91) each quarter. A strategy that longs the top portfolio and shorts the middle (3 rd quintile) portfolio experiences a return of 1.63% (t = 1.94). CCCCCCCC continues to forecast excess returns in overlapping portfolios for multiple horizons. At the one-year horizon, the top quintile portfolio has a risk-adjusted alpha of 1.78% (t = 3.63) each quarter, while the bottom quintile portfolio obtains -1.55% (t = -2.43). The long-short strategy at this horizon generates an excess return alpha of 3.33% (t = 4.13) per quarter. Return predictability persists significantly over multiple periods. This contrasts the immediate demand pressure phenomenon found in the existing literature. The investor flow induced price effect begins reverting immediately after its measurement date (Frazzini and Lamont 2008). The persistence of this price effect is likely due to the length and scale of capital return programs, which usually last multiple quarters if not multiple years. A stock currently receiving capital redeployment induced demand will likely continue to receive this price pressure over a significant horizon. This persistent demand would push stock prices further from some fundamental value for multiple quarters or even years. Reversals of prices toward fundamental value may occur only after a significant run-up or when this source of demand fades. The calendar portfolios experience a reversal of their abnormal return patterns at the very long holding period horizon. Figure 3 records the average quarterly risk-adjusted returns of strategies that long the top quintile portfolio and either short the bottom or the mid quintile portfolio over various holding horizons. For each horizon, no new overlapping calendar portfolio is initiated unless it can last the whole holding horizon to Q (for example, a new portfolio

19 is not initiated if the holding horizon is 5 and the formation date is Q1 2015). We observe a partial reversal of the cumulative excess returns over the 12-year horizon holding period for the first strategy, and a complete reversal for the second strategy. In summary, the abnormal returns associated with capital returns in the moderate horizon and its long-term reversal is consistent with a demand channel of capital redeployment. 3. Connected Stocks and Corporate Structure This section describes the characteristics of stocks influenced by capital redeployment. First, I summarize the contemporaneous relationship between capital redeployment demand and stock characteristics such as value and size. Second, I summarize the future relationship between the stocks linked to capital redeployment and their future cash payout and issuance patterns. Panel A of Table 7 reports the average market equity and average book equity of the stocks in the calendar time portfolios. Consistent with their factor loadings, the portfolio at the top of the quintile tends to contain larger stocks on average. In 1990, the average market cap of the formation portfolio at the top quintile is $2.079 billion, while that of the lowest quintile is $3.424 billion. The distribution of stock sizes became more skewed to the top quintile over time. In 2003, the average market cap of the formation portfolio at the top quintile increased to $ billion, while that of the lowest quintile decreased to $1.470 billion. This trend is followed to Q Panel B describes the capital return policies of the firms in each portfolio over time. If mutual fund investors are rationally responding to capital payouts in certain firms by purchasing similar stocks with the expectation of future payout, then we should see significant capital return in the high CCCCCCCC portfolios. The highest quintile portfolio according to CCCCCCCC does initiate more capital return and more dividends over the 6 years and 12 years after the portfolio formation period. However, Panel B of Table 7 indicates that the programs initiated by these firms are extremely marginal, and economically insignificant. Despite experiencing cumulative returns of over 11% in the first 12 months of the holding period, these firms on average only bought back 0.047% more of their stock and increased total dividend yield by 0.048% over the lowest quintile portfolio over 6 years. This is decreased to 0.006% in buybacks and 0.030% in dividend payments over a 12-year horizon. The magnitudes indicate that while CCCCCCCC captures some potential increases of capital return programs, the marginal increase cannot be the source of the significant price effect.

20 In table 8, I perform regression analysis to see the average correlation between CCCCCCCC and changes in buyback, issuance, and dividend activities. Once I control for characteristic such as size, past issuance, and past returns, there is no significant correlation between capital return spillover and a firm s own capital return activity. However, in contrast, this price spillover mechanism is significantly correlated with future issuance activities both in statistical significance and economic magnitudes. One standard deviation of CCCCCCCC implies an increased of 0.55% (0.57%) shares outstanding per quarter over 12 (48) quarters. In summary, stocks with this spillover channel of induced price gain only marginally and economically increase their own payout activities. In contrast, they significantly increase their equity issuances, potentially taking advantage of market mispricing. 4. Conclusion This paper examines the redeployment of capital in the equity markets by following the capital return induced trading by professional investors. Investors receiving large amounts of capital repayment tends to predominantly reinvest in existing holdings when compared to investors that receive insignificant capital repayment. This indicates a separation between assets that are connected to capital returns and ones that are not. By taking advantage of a cross section of stocks from firms that do not return capital but vary in their exposure to capital returns, this paper shows that targets of capital return induced purchases tend to appreciate in the short to moderate horizon, and partially revert at the very long horizon. This price effect is consistent with mechanical uninformative demand channel. Although this paper is mainly a test of the demand channel of capital redeployment through the financial markets, our results are suggestive about the mechanism of price discovery for stocks in firms that do repurchase their own assets. Given that demand originating from redeployment of capital causes significant shifts in the prices of related firms, it is likely that this demand is also a part of the pricing mechanism for the original firms. The abnormal announcement day returns of capital repurchase programs have been generally used as support for a signaling hypothesis of firm fundamentals. However, if demand is generated through the actual open market repurchases, then an immediate price reaction may be a response by investors to the expectation of future demand rather than to the expectation of future cash flow. The findings in this paper questions whether a

21 firm s announcement returns are due to information about the firm s future fundamentals or are due to expectations that the firm will drive future demand through the actual buyback. The mechanism described in the paper is that buyback and dividend programs implicitly generate demand for related firms through the redeployment of capital back into the equity market. Existing finance literature indicates that the executives of public firms initiate stock repurchases for a variety of purposes- from following the belief that their shares are undervalued to acting on payouts incentives. However, there is very little reason that these executives consider the stock prices and investment behavior of related firms when directing their own cash distributions. The spillover channel documented in this paper associates the changes in prices and capital structure of related firms to a manager s capital return decisions for his own firm. If this price-effect channel improves the competitiveness of these related firms, then capital return in the form of cash payouts may be viewed as having caused unintended consequences for the manager who initiated cash payouts. The empirical results outlined here can be a useful account for corporate finance practitioners when considering future capital repayment decisions. This paper s limited scope also suggests another interesting directions for future research. One potential avenue is to understand the nature of stock issuance in the financial markets. A large academic literature argues that initial public offerings and seasoned equity offerings are strategic responses to demand in the equity markets. Issuers, through their information network, may be particularly adept at gauging the persistent capital return from firm payouts. Isolating the connection between investor net flow, which is temporary, and capital return-based inflow, which is extremely persistent, may aid in understanding the strategic response by firms to gear their capital structure toward equity mispricing.

22 References Baker, Malcolm, and Jeffrey Wurgler "The Equity Share in New Issues and Aggregate Stock." The Journal of Finance Barberis, Nicholas, and Andrei Shleifer "Style Investing." Journal of Financial Economics Bhattacharya, Sudipto "Imperfect Information, Dividend Policy, and 'The Bird in the Hand' Fallacy." The Bell Journal of Economics Black, Fischer "The Dividend Puzzle." The Journal of Portfolio Management 5-8. Braun, Matías, and Borja Larrain "Do IPOs Affect the Prices of Other Stocks? Evidence from Emerging Markets." The Review of Financial Studies Brav, Alon, John R. Graham, Campbell Harvey, and Roni Michaely "Payout Policy in the 21st Century." Journal of Financial Economics Brown, Stephen J., and William N. Goetzmann "Mutual fund styles." Journal of Financial Economics Coval, Joshua, and Erik Stafford "Asset fire sales (and purchases) in equity markets." Journal of Financial Economics Edmans, Alex, Itay Goldstein, and Wei Jiang "The Real Effects of Financial Markets: Impact of Prices on Takeovers." The Journal of Finance Fama, Eugene F., and James D. MacBeth "Risk, Return, and Equilibrium: Empirical tests." Journal of Political Economy Frazzini, Andrea, and Owen A. Lamont "Dumb Money: Mutual Fund Flows and the Cross-Section of Stock Returns." Journal of Financial Economics Graham, John, and Campbell Harvey "How do CFOs make capital budgeting and capital structure decisions?" Journal of applied corporate finance Greenwood, Robin, and Samuel G. Hanson "Share Issuance and Factor Timing." The Journal of Finance Hartzmark, Samuel M., and David H. Solomon The Dividend Disconnect. Working Paper. Ikenberry, David, Josef Lakonishok, and Theo Vermaelen "Market Underreaction to Open Market Share Repurchases." Journal of Financial Economics

23 Jegadeesh, Narasimhan, and Sheridan Titman "Returns to Buying Winners and Selling Losers:." The Journal of Finance Lou, Dong "A Flow-Based Explanation for Return Predictability." Review of Financial Studies Loughran, Tim, and Jay R. Ritter "The New Issues Puzzle." The Journal of Finance Massa, Massimo, Zahid Rehman, and Theo Vermaelen "Mimicking Repurchases." Journal of Financial Economics Miller, Merton H, and Franco Modigliani "Dividend Policy, Growth, and the Valuation of Shares." The Journal of Business Ross, Stephen A "The Determination of Financial Structure: The Incentive-Signalling Approach." The Bell Journal of Economics Shleifer, Andrei "Do Demand Curves for Stocks Slope Down?" The Journal of Finance Shleifer, Andrei, and Robert W. Vishny "Liquidation Values and Debt Capacity: A Market Equilibrium Approach." The Journal of Finance Shleifer, Andrei, and Robert W. Vishny "The Limits of Arbitrage." The Journal of Finance Vermaelen, Theo "Common Stock Repurchases and Market Signalling: An Empirical Study." Journal of Financial Economics

24 Figure 1. Aggregate Quarterly Capital Return and Equity Fund Flow The figure plots the quarterly aggregate buyback and dividend payments in the CRSP universe of common stocks traded on the NYSE, NASDAQ, and AMEX exchanges; and net fund flow from the CRSP universe of Equity Funds. Buyback is calculated as the product of adjusted decrease in shares and quarter start prices. Firms whose shares outstanding decreases by more than 10% per quarter are ignored to avoid counting mergers. Dividend payment is dividend yield (the difference between total and price returns) times quarter start market capitalization. Equity flow is calculated from CRSP as the difference between change in TNA and quarter start TNA adjusted by fund returns.

25 Figure 2. Composition of Capital Return and Aggregate Market Capitalization from the Top Capital Returning Stocks. The top left panel depicts aggregate quarterly capital return (Dividend and Composite Buybacks) decomposed to levels by the top capital returning stocks. The top right panel shows the compositions of aggregate Market Capital as attributable to the top capital returning stocks. The bottom left panel depicts the fraction of aggregate quarterly capital return attributable to levels of top capital returning stocks. The bottom right panel depicts the fraction of aggregate market cap attributable to levels of top capital returning stocks.