Share Issuance and Cash Holdings: Evidence of Market Timing or Precautionary Motives? a R. David McLean b First Draft: June 23, 2007 This Draft: March 26, 2008 Abstract Over the past 35 years, the average U.S. firm saved $0.45 per $1 of share issuance, which was substantially larger than the savings rates of other cash sources. I employ a commonly used model to estimate precautionary cash holdings, and show that the savings of share issuance proceeds can be explained by cash needs. Large share issuers do not generate excess cash, and are dependent on share issuance proceeds to meet their cash targets. Firms with large cash needs are more likely to issue shares, and tend to choose equity over debt when raising capital. Post issuance stock returns support the notion that share issuance funds legitimate cash needs, as low post issuance stock returns are not observed in issuers which save the proceeds, but are limited to issuers which spend the proceeds. Taken in their entirety, the results show that precautionary cash needs are one of the primary motivations for share issuance. a This paper has benefited from comments by Felipe Aguerrevere, Harry Deangelo, Mike Hertzel, Mark Huson, Zhi Li, Jeffrey Pontiff, Richard Sloan, Cliff Stephens, Matsuhiro Watanabe, and Seminar Participants at Arizona State University, Barclay s Global Investors, and the University of Alberta. The author acknowledges support from the Southam / Edmonton Journal Fellowship. Any remaining errors are the author s. b Assistant Professor of Finance, University of Alberta, 4-20K Business Building, School of Business, Edmonton, Alberta, Canada, T6G 2R6. Phone: 780-492-8005. Fax: 780-492-3325. E-mail: rdmclean@ualberta.ca.
1. Introduction This paper studies the motivations for share issuance, by focusing on the propensity to save share issuance proceeds as cash. Over the past thirty-five years, U.S. firms on average saved $0.45 per $1 of share issuance; as a comparison, the same firms saved $0.09 per $1 of debt issuance. The share issuance savings rate also increased, from $0.24 per $1 in 1971, to $0.54 per $1 in 2005. Hence, when firms issue shares today, they tend to save the majority of the proceeds. In this paper I try to explain why firms save such a large portion of their share issuance proceeds; doing so should deepen our understanding of why firms issue shares. I consider two theoretical frameworks. The first is the precautionary motive for cash holdings. It contends that firms with volatile cash flows and firms with valuable investment opportunities should accumulate large cash balances for precautionary reasons. The precautionary motive for cash holdings is first proposed by Keynes (1934), and further studied by Holmstrom and Tirloe (1996 and 1998), Opler, Pinkowitz, Stulz, and Williamson (1999), Mikkelson and Partch (2003), Almeida, Campello, and Weisbach (2004), Bates, Kahle, and Stulz (2007), Acharya, Almeida, and Campello (2007), and Han and Qu (2007). In the precautionary cash holdings literature it is generally assumed that firms save cash from operational cash flows, as external finance can be costly. 1 However, some firms may not generate sufficient internal cash flows to meet their cash needs. These firms could adhere to the pecking order of Myers (1984) and Myers and Majluf (1984) and issue debt, but high precautionary motive firms are characterized by high R&D 1 Almeida, Campello, and Weisbach (2004), Acharya, Almeida, and Campello (2007), and Han and Qu (2007) show empirically that firms with high precautionary motives save a larger portion of their operational cash flows. 1
spending and high cash flow volatility, both of which make them unsuitable for large amounts of debt. Hence, share issuance might be the only means by which some of these firms can meet their cash requirements. The second framework is the market timing framework, which is proposed by Loughran and Ritter (1995) and Baker and Wurgler (2002), and asserts that managers issue shares when share prices exceed fundamental values. Loughran and Ritter (1997) and Deangelo, Deangelo, and Stulz (2007) note that pure market timing theories predict that overvalued firms will issue shares to build liquidity; hence the use of share issuance proceeds to build cash balances is consistent with market timing. Pagano, Panetta, and Zingales (1998), Kim and Weisbach (2007), and Hertzel and Li (2007) show that both IPO and SEO proceeds are more likely to be saved as cash if the issuers have high market-to-book ratios, which could be consistent with market-timing. To estimate precautionary cash needs, I employ a model which estimates a fitted cash value for each firm-year observation. The fitted cash estimate accounts for the firm s precautionary motives, cost of raising funds, and other characteristics. This cashholdings model is also used in Opler et al. (1999), Bates et al. (2007), Ditmar and Mahrt- Smith (2007), Harford, Maxwell, and Manis (2007) and Kesckes (2007). I find that the propensity to save share issuance proceeds can be explained by precautionary cash needs. Firms in the highest fitted cash quintile save $0.56 per $1 of share issuance proceeds, whereas firms in the lowest quintile save $0.09 per dollar of share issuance proceeds. The results further show that firms issue shares to meet their optimal cash levels, not to raise excess cash. The difference between the firm s actual cash holdings and its fitted value from the model is the firm s excess cash. Excess cash 2
does not increase with share issuance, and on average excess cash among high share issuance firms is close to zero. Market timing does not explain the propensity to save share issuance proceeds. To measure market timing motives, I follow Loughran and Ritter (1995) and Baker and Wurgler (2002) and use the market-to-book ratio as a proxy for mispricing. The marketto-book ratio is correlated to the fitted cash estimates, so I decompose the market-to-book ratio into two components: one component is explained by the fitted cash values, while the other is orthogonal to the cash values. I find that firms with high fitted cash marketto-book values save large portions of their share issuance proceeds; however firms with high orthogonal market-to-book values do not save large portions of their issuance proceeds. Hence the relation between market-to-book and share issuance savings can be explained by precautionary cash needs. If firms use share issuance to meet their cash needs, then firms with high cash needs should issue more shares. I show empirically that this is the case: Firms with high fitted cash values are more likely to make large share issuances, and firms with large cash deficits relative to their cash needs are more likely to make large share issuances. Moreover, both measures of cash needs predict choice of equity over debt. One concern with this analysis is that the estimates of optimal cash holdings could be correlated to mispricing. If issuing firms with high fitted cash values have low postissuance stock returns, then their issuances may be the result of market timing. However I find the opposite. Among large issuers, the high fitted cash firms, which save $0.56 per $1 raised, do not have negative alphas post issuance. Yet the low fitted cash firms, which save only $0.09 per $1 raised, have negative 3-factor and 4-factor alphas post issuance. 3
A hedge portfolio, that is long the savers and short the spenders generates 3-factor and 4- factor alphas of 1.070% (t-statistic = 2.49) and 1.139% (t-statistic = 2.67) per month. The rest of this paper is organized as follows. Section 2 describes the sample and variables. Section 3 studies the relation between share issuance and cash holdings. Section 4 compares the post-issuance returns of issuers which save to those of issuers that spend. Section 5 concludes. 2. Data and Measurement 2.1. Data This study uses annual data from the Compustat Industrial Database. The Compustat data are merged with stock return and share price data from CRSP. The sample ranges from 1971 through 2005. Firms are excluded that are either missing cash and marketable securities (data1), or book value of assets (data6). Firms that do not have a positive book value of equity (data60) are excluded as well. Firms with SIC codes between 4900 and 4999 (financial firms) and 6000 and 6999 (utilities) are also excluded, as these firms may carry excess cash for regulatory reasons [this follows Frank and Goyal (2003) Fama and French (2005) and Bates et al. (2007)]. Each variable is winsorized at the 1 st and 99 th percentiles. The final sample has 118,645 firm year observations. 2.2. Variables This Section describes the variables used in this study. Summary statistics for the variables are displayed in Table 1. 4
2.2.1. Sources and Uses of Cash ΔCash. The yearly change in cash and cash and equivalents (data1), scaled by the lagged book value of total assets (data6). ΔCash t+2. The three-year change in cash and cash and equivalents (Cash t+2 - Cash t-1 ), scaled by the lagged book value of total assets. Share Issuance. The sum of all cash flows resulting from the sale of shares (data 108). This share issuance measure is used by Lemmon and Zender (2002), Frank and Goyal (2003), and Bradshaw, Richardson, and Sloan (2006), and captures all share issuance transactions that result in cash flows to the firm. This measure is scaled by the lagged book value of total assets. Share Repurchase. Cash used in share repurchases (data115) scaled by the lagged book value of total assets. Debt Issuance. Cash flows resulting from the sale of debt (data 111) scaled by the lagged book value of total assets. Debt Repurchase. Cash used in debt repurchases (data114) scaled by the lagged book value of total assets. Cash Flow from Operations (CFO). CFO is income before extraordinary items (data14), plus depreciation and amortization (data18), plus R&D expenditures (data46), all scaled by the lagged book value of total assets. R&D expenditures are added back to reflect the fact that R&D expenditures are like capital expenditures in that they are an investment. The main results in the paper are not sensitive to the inclusion of R&D in CFO. 5
Other Cash Sources (Other). The sum of the sale of property plant and equipment (data107), the sale of investments (data109), and other sources of funds (data218), all scaled by the lagged book value of total assets. 2.2.2. Cash Determinants The variables described in this Section are used to estimate fitted cash values. The cash-holdings model used here is developed by Opler et al. (1999), and is used in Bates et al. (2007), Ditmar and Mahrt-Smith (2007), Harford, Mansi, and Maxwell (2007), and Kesckes (2007). One difference in this paper relative to these other studies is that I omit the market-to-book ratio from the model. I do this because I use the marketto-book ratio as a proxy for mispricing. Including the market-to-book ratio in the model does not change the results in this paper. Research and Development Expenditures (R&D). Cash expenditures on research and development (data46) scaled by the lagged book value of total assets. High R&D firms tend to have more valuable investment opportunities; hence not having sufficient funds for investment is more expensive for these firms. The relation between R&D expenditures and cash holdings is therefore expected to be positive. Following Loughran and Ritter (1997), Opler et al. (1999), and others if R&D is missing, then it is set equal to zero. Cash Flow Volatility. Cash Flow Volatility is the ten-year average of the log of cash flow variances within each 2 digit SIC code. Firms with high cash flow volatility have less reliable internal cash flows; therefore these firms are more likely to suffer an adverse liquidity shock, and should hold more cash. 6
Payout. Payout is the sum of cash dividends (data21) and stock repurchases (data115), all scaled by the lagged book value of total assets. One could argue that firms with low payout ratios tend to be financially constrained, so these firms should hold more cash. One could also argue that firms which have generated large amounts of cash have a better ability to make payouts. Assets. The natural log of the book value of total assets (data6). Firms with fewer assets may have greater difficulty in posting collateral to raise cash if needed. Hence smaller firms should hold more cash. Cash Flow from Operations (CFO). As defined in the previous section. Firms with high CFO have a greater ability to generate and accumulate cash; however firms with low CFO have more of an incentive to save cash for precautionary reasons. Net Working Capital (NWC). Net working capital is net working capital (data179) minus cash (data1), all scaled by the book value of total assets (data6). NWC can be a cash substitute as it represents assets that are relatively liquid; hence NWC should be negatively related to cash. CAPEX. Capital expenditures (data128) scaled by the lagged book value of total assets. All else equal, firms that do more CAPEX should have lower cash balances. Leverage. Long-term debt (data9) plus short-term debt (data34), all scaled by scaled by the book value of total assets. Acquisitions. Cash spent on acquisitions (data129) scaled by the lagged book value of total assets. All else equal, firms that do more acquisitions should have lower cash balances. 7
2.2.3. Cash Measures Cash. Cash is the sum of all cash and cash and equivalents (data1) scaled by the book value of total assets. Cash(xb). This is an estimated cash value, based on fitted values from crosssectional regressions. 2 The dependent variable in the regressions is Cash, while the independent variables are those described above in the Section 2.2.2. Each year, for each firm, a fitted cash value is estimated. This fitted value is referred to as Cash(xb). The regression results are reported in the Appendix, and are similar to those reported in the aforementioned studies which use the model. Cash(xb) = t + β1 t R&D + β2 t Cash Flow Volatility + β3 t Payout + β4 t Assets + β5 t CFO + β6 t NWC + β7 t CAPEX + β8 t Leverage + β9 t Acquisitions Cash(e). This is the residual from the regressions used to estimate Cash(xb). Cash(e) is orthogonal to the precautionary and transaction cost proxies used in the regression model, and is an estimate of excess cash holdings. Cash(e) = Cash Cash(xb) 2.2.4 Market-to-book Measures Market-to-Book (MB). Market-to-book is the natural log of the ratio of the market value of equity to the book value of equity (data60). Market value of equity is the product of price and shares outstanding. Market value of equity is measured six-months subsequent to the calendar month in which the firm s fiscal year ends. Lakonishok, 2 Similar results are obtained using a panel regression (see the Appendix). Bates et al. (2007) show that the model s coefficients are stable over time, and report that they also get similar results using either crosssectional or panel regressions 8
Sheleifer, and Vishny (1994), Loughran and Ritter (1997), Pagano, Panetta, and Zingales (1998), Baker and Wurgler (2002), and Kim and Weisbach (2007) use the market-tobook ratio as a proxy for mispricing. Fitted Market-to-Book (MB(xb)). The fitted value from a cross-sectional regression of MB on Cash(xb) (see the Appendix for the regression results). MB(xb) = t +β t Cash(xb) Residual Market-to-Book (MB(e)). The residual from the regression used to estimate MB(xb). This is the component of MB that is orthogonal to the firm s cash needs. MB(e) = MB MB(xb) 3. Do Firms Keep a Larger Portion of Share Issuance Proceeds as Cash as Compared to Other Cash Sources? 3.1 The Propensity to save Cash Flows as Cash: Methodology In this Section I estimate the propensity to save share issuance proceeds as cash, and compare this propensity to those of other cash sources. I use a methodology similar to that used by Kim and Weisbach (2007) and Hertzel and Li (2007). The regression equation is: ΔCash it = + β 1 Share Issuance it + β 2 Share Repurchase it + β 3 Debt Issuance it + β 4 Debt Repurchase it + β 5 CFO it + β 6 Other it + Year Dummies + ε it (1) Equation 1 is also estimated by replacing ΔCash with ΔCash t+2. This is done to test whether the propensities to save are robust over longer horizons. 9
The Share Issuance measure used in this paper captures all share issuances that result in cash flows to the firm, so it is different than those used in Kim and Weisbach and Hertzel and Li, as these studies focus on the use of SEO and IPO proceeds. Another difference is that the sample in this study includes all firms, not just share issuers, while the samples in Kim and Weisbach and Hertzel and Li are limited to IPO and SEO issuers. My broader sample allows me to make more general statements regarding where firms get their cash from, which is not the purpose of these other studies. I also study the use of debt, operational cash flow, and other cash sources separately, whereas Kim and Weisbach and Hertzel and Li group them into one variable. In two of the regressions the model is augmented with four interaction terms, which are the product of a variable denoted Time and each of the four sources of cash measures. Time takes on a value of 1 in the first year of the sample, and increases by 1 in each additional year of the sample. Hence the interaction terms test whether the propensities to save cash flows as cash have increased over time for each of the four cash sources. The regressions are panel regressions, with year fixed effects. The coefficients can be interpreted as the amount of cash saved per dollar of cash flow. Following Petersen (2007) the standard errors are estimated by clustering on the firm. 3 3.1.1 The Propensity to save Cash Flows as Cash: Results Table 2 shows that firms save a larger portion of the proceeds from share issuance as compared to other cash sources. In regression 1, the dependent variable is ΔCash and 3 Petersen (2007) shows that if there are at least 1,000 permnos, then clustering either on permno, or on both time and permno produces identical results (see Figure 7 in Petersen (2007)). Each of the panel regressions in this paper has over 4,000 permnos. Petersen also notes that time fixed effects may be preferable to clustering on time if the number of time clusters is relatively small, and in this paper T = 35. 10
there are no interaction terms. The Share Issuance coefficient is 0.446 (t-statistic = 88.04), which is significantly larger than each of the other coefficients. The coefficient shows that a $1 increase in share issuance leads to a $0.446 increase in ΔCash. The second largest coefficient is CFO, it is 0.298 (t-statistic = 60.84). As expected, each of the source of cash coefficients are positive and significant, showing that cash balances increase when sources of cash increase. The Debt Issuance coefficient is 0.092 (t-statistic = 14.48), which is substantially smaller than both the CFO and Share Issuance coefficients. This shows that when firms issue debt, they tend to spend the proceeds rather quickly. This supports the notion that debt holders prefer to buy debt from issuers which have a specific and immediate purpose in mind for the proceeds. The Other coefficient is 0.018 (t-statistic = 2.60), showing that asset sales and other non reoccurring events are not major sources of cash. Regression 2 is like regression 1, only the dependent variable is ΔCash t+2. Regression 2 serves as a robustness check, especially to the results regarding share issuance, as it could be that firms issue shares for projects that may unfold over several years. If this is so, then the Share Issuance coefficient in regression 1 could represent earmarked investment, rather than precautionary cash holdings. The Share Issuance coefficient in regression 2 is 0.570 (t-statistic = 42.90), which is larger than the coefficient in regressions 1, and larger than each of the other coefficients in the regression, suggesting that firms do keep a larger portion of their share issuance proceeds as cash when compared to other cash sources. 4 4 Cash holdings are right skewed, and increased over the sample period. This might explain why the source of cash coefficients are larger in regression 2 as compared to regression 1. 11
3.1.2 The Propensity to save Cash Flows as Cash: Time Interaction Results Regressions 3 and 4 include the Time interaction terms. In 1971, the first year of the sample, Time = 1. Time increases by one in each year of the sample, so in 2005, the final year of the sample, Time = 35. In these regressions, the total propensity to keep share issuance proceeds as cash is equal to the Share Issuance coefficient plus the Share Issuance * Time interaction coefficient. We might expect the Time interaction coefficients to be positive, as Bates et al. (2007) show that cash holdings have increased over the sample period. In regression 2 the Share Issuance * Time interaction coefficient is 0.009 (tstatistic = 13.03), showing that the Share Issuance coefficient has on average grown by this amount each year. The Share Issuance coefficient s value is 0.227, so the results show that in 1971 $1 of share issuance resulted in $0.236 cash savings, while in 2005, $1 of share issuance resulted in $0.542 of cash. The propensity to keep CFO as cash was $0.202 per $1 in 1971, similar to that for share issuance, while in 2005 $1 of CFO resulted in $0.372 of cash. Hence firms are keeping more cash from both sources, but are increasingly getting their cash from share issuance, rather than operations, as the propensity to keep share issuance proceeds as cash has increased by 130% (from 0.236 to 0.542), while the propensity to keep operational cash flows as cash has increased by 84% (from 0.202 to 0.372). The Debt Issuance* Time coefficient is also positive and significant. Its coefficient is 0.002 (t-statistic = 5.75). Taken together with the Debt Issuance coefficient, the results show that the propensity to save debt issuance proceeds as cash increased by from 0.040 to 0.105 over the last 35 years. Despite this increase, the 12
propensity to save debt issuance proceeds is still relatively low in 2005 as compared to the propensities to save cash from both operational cash flows and share issuance. Taken in their entirety, the results show that firms keep a larger portion of their share issuance proceeds as cash when compared to other cash sources. This could be because firms are using share issuance to meet legitimate cash needs. Alternatively, it could that firms are issuing shares when they are overvalued, and store the proceeds as cash because they do not have any viable projects to invest in. The subsequent sections will attempt to differentiate between these two hypotheses. 3.2. Can Firm Characteristics Explain the Propensity to save Cash Flows as Cash? The previous Section showed that firms are more likely to keep share issuance proceeds as cash as compared to other cash sources. In this Section I test whether the propensity to save share issuance proceeds as cash can be explained by either legitimate needs for cash, or market timing motives. 3.2.1. Precautionary Cash Needs (Cash(xb)) and the Propensity to Cash Flows as Cash I this Section I test whether the propensity to save share issuance proceeds as cash can be explained by precautionary cash needs. My estimate of cash needs is Cash(xb). Cash(xb) is described in Sections 2.2. and 2.3.; it is an estimate of how much cash a firm should hold, after accounting for its precautionary motives, costs of raising capital, and other factors. High Cash(xb) firms may use external finance to meet their cash needs if they do not have sufficient internal cash flows. As mentioned previously, debt probably is unsuitable for most high Cash(xb) firms. High Cash(xb) firms have high cash flow 13
volatility, so it might be difficult for them to service large amounts of debt, and debt increases cash flow volatility, which is what motivates these firms to hold cash in the first place. High Cash(xb) firms also engage in large amounts of R&D, so when the cash holdings of high Cash(xb) firms are eventually spent, they will most likely be spent on projects and expenses related to R&D. Hall (2002) and Brown, Fazzari, and Petersen (2007) point out that debt financing is atypical for R&D projects, due to information problems between the firm and potential bondholders, uncertainty regarding project payoffs, and lack of collateral. To test whether firms save share issuance proceeds to meet legitimate cash needs, I first sort firms annually into Cash(xb) quintiles. Equation 1 is then estimated within each of the Cash(xb) quintiles, and the coefficients are compared across quintiles. 5 The share and debt repurchase variables are used in the regressions, but the coefficients are not reported for the sake of brevity. The regressions in this section use ΔCash as the dependent variable, although the results are similar if ΔCash t+2 is the dependent variable. If firms issue shares to raise cash to meet their liquidity needs, then the Share Issuance coefficient should be largest in the high Cash(xb) quintile. Panel A of Table 3 shows that the propensity to save share proceeds as cash is monotonically increasing in Cash(xb) (see Figure 1). The Share Issuance coefficient is 0.558 for the high Cash(xb) quintile, and 0.094 for the low Cash(xb) quintile. The difference is statistically significant; its t-statistic is 25.16. The propensity to save debt issuance proceeds as cash is also monotonically increasing in Cash(xb), although the coefficients are smaller than the share issuance 5 In an earlier version of the paper I conducted this analysis by sorting on cash flow volatility, asset size and payout ratio, rather than Cash(xb); the results were similar. 14
coefficients, again suggesting that debt issuance proceeds are more promptly spent. The Debt Issuance coefficient is 0.373 for the high Cash(xb) firms and 0.010 for the low Cash(xb) firms. The difference is statistically significant; its t-statistic is 19.77. Like Almeida, Campello, and Weisbach (2004), Acharya, Almeida, and Campello (2007), and Han and Qu (2007) I also find that firms with greater cash needs save a greater portion of their operational cash flows as cash. The CFO coefficient is 0.437 for the high Cash(xb) firms and 0.069 for the low Cash(xb) firms, and this difference is also statistically significant. The share issuance coefficients are greater than the CFO coefficients in each of the Cash(xb) quintiles. The results therefore show that the propensity for high Cash(xb) firms to save from share issuance is greater than their propensity to save from operational cash flows. 3.2.2 Market Timing and the Propensity to Save Cash Flows as Cash Market timing could also explain why firms issue shares and then keep the proceeds as cash. Loughran and Ritter (1997) and Deangelo, Deangelo, and Stulz (2007) note that pure market timing theories predict that overvalued firms will issue shares to build cash balances. Pagano, Panetta, and Zingales (1998) and Kim and Weisbach (2007) provide evidence that both IPO and SEO proceeds are more likely to kept as cash if the issuers have high market-to-book ratios, which could be evidence of market timing. I follow Loughran and Ritter (1997) and Baker and Wurgler (2002) and use the market-to-book ratio (MB) as a proxy for overvaluation. MB is correlated to Cash(xb) (correlation is 0.36, not reported in tables), so it could be that high MB firms keep share issuance proceeds as cash for legitimate, rather than market timing motives. Therefore I 15
test whether the propensity to keep share issuance proceeds as cash is related to MB(xb) and MB(e) separately. As discussed in Section 2, MB(xb) is the portion of MB that is explained by Cash(xb), whereas MB(e) is orthogonal to Cash(xb). Both MB measures are lagged on year to ensure that the issuing manager had access to the share price information. 6 Panel B of Table 3 shows that the MB(xb) quintiles produce a similar, but less pronounced, pattern to those created by the Cash(xb) quintiles. The Share Issuance, Debt Issuance, and CFO coefficients are each largest in the highest MB(xb) quintile, and in general the coefficients increase across the MB(xb) quintiles in a near monotonic fashion. The difference between the high and low MB(xb) quintiles in Share Issuance coefficients is 0.299 (t-statistic = 14.97). Panel C displays the coefficients across MB(e) quintiles. The Share Issuance coefficient is 0.469 for low MB(e) firms and 0.463 for high MB(e) firms. Moreover, there no observable pattern in the share issuance coefficients across the MB(e) quintiles, as there is with the Cash(xb) and MB(xb) quintiles. These findings suggest that the tendency for high MB firms to keep share issuance proceeds as cash is explained by legitimate needs for cash holdings. 3.3. Do Firms Use Share Issuance to Generate Excess Cash? The previous Section shows that firms with high precautionary motives keep a large portion of their share issuance proceeds as cash. This could be because these firms are attempting to meet their optimal cash level, as measured by Cash(xb). Alternatively, it could be that these firms are market timing and issuing an excessive amount of shares, 6 If it is the firm s first year in the sample, then the current values of MB(xb) and MB(e) are used. 16
which should inflate their excess cash levels, Cash(e). Market timing therefore suggests that Cash(e) and share issuance should be correlated positively. I test this hypothesis in the following sections. 3.3.1. Share Issuance and Excess Cash: Results In Panel A of Table 4 firms with positive values of net share issuance (cash generated by share issuance minus cash spent on share repurchases) are sorted annually into share issuance quintiles. 7 The amount of share issuance in the first four quintiles is very low, the values range from 0.001 to 0.043. The amount of share issuance in the fifth quintile is 0.409, which is very large relative to the fourth quintile. Pontiff and Woodgate (2006) and Fama and French (2007) also report that share issuance is highly right skewed. If share issuance is the result of market timing, then Cash(e) should increase with share issuance across the quintiles, however there is no observable pattern in Cash(e) across the share issuance quintiles (see Figure 3). Cash(e) for the highest quintile of issuers is only 0.009, so these firms did not generate large amounts of excess cash, especially relative to their share issuance proceeds (0.409). Simulated Cash (e), which is what Cash(e) would have been if the net proceeds from share issuance were zero, is - 0.400, or almost the size of the entire share issuance proceeds. In other words, if these firms had not issued shares, then they would have had large cash shortfalls. The average total cash (Cash(xb) + Cash(e)) held by the highest quintile of issuers is 0.211. This is less than the average issuance proceeds of these firms, so these firms could not have undertaken their chosen operations and investments had they not 7 52% of the firms in the sample have net share issuance values that are positive. 17
issued shares. This is consistent with the findings in a contemporaneous paper by Deangelo, Deangelo, and Stulz (2007), who find that most SEO issuers could not have undertaken their operations and capital expenditures without the SEO proceeds.. Panel A also shows that Cash(xb) increases monotonically across the share issuance quintiles, suggesting that cash needs are a motivation for share issuance. Cash(xb) increases from 0.097 in the lowest quintile, to 0.202 in the highest quintile, and the difference between quintiles is statistically significant. 3.3.2. Share Issuance and Excess Cash: High Issuers Only; Sorted into Cash Needs Quintiles Table 3 showed that high Cash(xb) firms saved a high portion of their share issuance proceeds as cash. This could have been done to meet their cash needs, or it could have resulted in excess cash. Cash(xb) and Cash(e) are by design uncorrelated across the sample, however the two variables could be correlated across large issuers; Panel B tests whether or not this is the case. Panel A shows that large share issuances are limited to firms in the highest share issuance quintile. Panel B therefore limits the sample to firms that are in the highest net share issuance quintile, and studies these issuers across Cash(xb) quintiles. Panel B shows that in each of the Cash(xb) quintiles Cash(e) is small relative to the size of the share issuance, suggesting that excess cash was not a motivation for the share issuances. Net Share Issuance ranges from 0.514 to 0.302, while Cash(e) ranges from 0.027 to 0.058. The Simulated Cash(e) for all of the quintiles is very large, it ranges from -0.244 for the low Cash(xb) firms and -0.514 for the high Cash(xb) firms, so there would have been large cash short falls had these firms not issued. 18
Panel B also shows that share issuance is monotonically increasing in Cash(xb), suggesting that cash needs are a motivation for share issuance. The average amount of net share issuance is 0.302 in the lowest Cash(xb) quintile, and 0.541 in the highest Cash(xb) quintile, and the difference is statistically significant (t-statistic = 4.50). Both debt issuance and CFO are negatively related to Cash(xb) across large share issuers. As mentioned before, debt is probably unsuitable for high Cash(xb) firms, which are characterized by their high cash flow volatility and high R&D spending. The debt issuance proceeds of the lowest Cash(xb) quintile is 0.155, while that of the highest Cash(xb) is 0.003; the t-statistic for the difference is -10.74. Hence, firms that need to hold large amounts of cash and issue large amount of shares are dependent on share issuance for their cash inflow. Moreover, for each quintile the total cash of these firms (Cash(xb) + Cash(e)) is less than the amount of share issuance, so these firms could not have undergone their chosen operational and investment activities had they not issued shares. 3.3.3. Share Issuance and Excess Cash: Large Issuers Only; Sorted into Market-to-Book Quintiles Panels C and D limit the sample to firms that are in the highest net share issuance quintile, and study these issuers across MB(xb) and MB(e) quintiles. Looking across the quintiles in both panels, we see that share issuance is monotonically increasing in both MB(xb) and MB(e). This is consistent with Loughran and Ritter (1997) and Baker and Wurgler (2002), both of whom show that high market-to-book firms issue more shares. Do high market-to-book issuers generate excess cash? Table 3 showed that high MB(xb) firms kept a relatively large portion of their issuance proceeds as cash. Panel C 19
shows that it was not too large an amount, as Cash(e) for the high MB(xb) issuers is 0.022, small relative to the level of issuance (0.553). Simulated Cash(e) is -0.510, far larger than the total cash of the high MB(xb) firms, showing that these firms need share issuance not only to meet their precautionary cash needs, but also to undertake their chosen operational and investment activities. Panel D shows that Cash(e) is not excessive for high MB(e) firms, as it is only 0.012, which is small compared to the amount of shares that these firms issued. Moreover Simulated Cash(e) is -0.458, which shows that almost the entire share issuance was needed in order for these firms to end the year with correct amount of Cash(xb). Taken together, the results in Panels C and D show that among large issuers high market-to-book firms issue more shares. However, among these large issuers, the share issuance proceeds were needed to both fund operations and investment, and to end the year with cash at the required level. There is no evidence here that high MB firms, or firms in general for that matter, use share issuance to generate cash beyond their needs, as measured by Cash(xb). 3.4. Cash Needs and Share Issuance: Regression Tests If firms use share issuance to meet their cash needs, then differences in cash needs ought to explain differences in share issuance. I define deficit as the dollar difference between time t cash(xb), the optimal amount of cash that the firm needs to hold at time t, and time t-1 cash, the actual cash held by the firm at time t-1. deficit = cash ( xb) t casht 1 = cash( xb) t cash( xb) t 1 cash( e) t 1 20
The lowercase notation indicates that the variables are not scaled by assets. If we scale both sides by time t-1 assets, then we get: cash( xb) Deficit = t cash( xb) Assets t 1 cash e t 1 ( ) Deficit is therefore the dollar amount by which the firm needs to increase its cash over the year, scaled by the assets at the beginning of the year. If firms use share issuance to meet their target cash levels, then Deficit should be positively correlated to share issuance. Deficit and Cash(xb) are mechanically correlated, as both are increasing in cash(xb). 8 If high Deficit firms also have high Cash(xb), then they are probably not well suited for debt, and will rely on share issuance for external finance. If Cash(xb) firms have on average have larger Deficits, then these firms will also need to issue more shares. Some firms need external finance to maintain their cash balances and fund nearterm operations and investment. In other words, a firm may begin the year with a Deficit of zero, but still need external finance to ensure that it ends the year with a Deficit value of zero. Given that high Cash(xb) firms are poorly suited for debt, share issuance might be necessary to keep them financially solvent; this is another reason that Cash(xb) ought to predict share issuance. 3.4.1. Probit Regressions of Share Issuance on Cash Needs and Controls The regressions in Table 5 are probit regressions. The dependent variable is equal to 1 if the firm s net proceeds from share issuances are either greater than or equal to 5% 8 Note that cash( xb) Cash( xb) t = Assets t t 21
of total assets, and zero otherwise. This measure is regressed on both Cash(xb) and Deficit and several control variables, which include MB, PPE, Debt, CFO, (each defined in Section 2) and Size. Size is measured as the firm s revenues as a percentage of the sample s total revenues in the same year. This measurement follows Leary and Roberts (2004). The control variables are measured at the beginning of the year. The regressions also include dummy variables for both year and industry. Industry is defined via the Fama and French 12 industry definition. 9 The coefficients are marginal probabilities and the standard errors are clustered on the firm. In regression 1, the Cash(xb) coefficient is 0.163 (t-statistic = 18.80). Table 1 shows that Cash(xb) has a standard deviation of 0.156. Hence a one standard deviation increase in Cash(xb) increases the probability of a large share issuance by 2.5%. This shows that firms that need to hold more cash issue more shares. These results are consistent with those in Table 4, which showed that large share issuers have large cash needs. In regression 2 Cash(xb) is replaced with Deficit. The Deficit coefficient is 0.265 (t-statistic = 36.98). Deficit has a standard deviation of 0.156 (not in tables), so a one standard deviation increase in Deficit increases the probability of a large share issuance by 4.1%. This shows that if firms have an expected year-end cash shortfall, then they are more likely to issue shares. Hence share issuance is a means by which firms meet their cash targets. Regression 3 includes both Cash(xb) and Deficit. The Deficit coefficient is similar to that in regression 2, however, the Cash(xb) coefficient (0.081) and its t-statistic 9 Thanks to Ken French for making the industry definitions available. The industry definitions can be found here: http://mba.tuck.dartmouth.edu/pages/faculty/ken.french/data_library.html 22
(8.32) are both about half of the values reported in regression 1. Cash(xb) and Deficit are correlated (correlation is 0.21), and the results here show that a good part of the reason that high Cash(xb) firms issue shares is avoid expected cash deficits. Regression 4 is like regression 3, only a Cash(xb)*Deficit interaction term is added. As compared to regression 3, the Cash(xb) coefficient does not change very much, however the Deficit coefficient is now 0.157, about two-thirds of its value in regression 4, and the Deficit coefficient s t-statistic also declined from regression 3 to regression 4, from 36.26 to 12.10. These findings suggest that firms with cash deficits are more likely to use to share issuance to erase their deficit if they also have high Cash(xb). This finding is makes sense. Low Cash(xb) firms can also have expected cash deficits, but these firms are probably better equipped to meet their cash needs by saving either from operations or from debt. 3.4.2. Probit Regression of Equity vs. Debt Choice on Cash Needs and Controls Table 6 also reports the results of probit regressions, only here the dependent variable represents share issuance vs. debt choice. The dependent variable is equal to 1 if the firm s net share issuance proceeds are greater than 5% of the book value of its total assets, and zero if the firm s net debt issuance proceeds are both greater than 5% of the book value of its total assets, and greater than the net proceeds from share issuance. The sample is therefore limited to firms that meet the definition for either share or debt issuance. Like in Table 5, I include dummy variables for both industry and year, and standard errors are estimated by clustering on the firm. 23
In regression 1 the Cash(xb) coefficient is 1.572 (t-statistic = 41.94), showing that a one standard deviation increase in Cash(xb) leads to a 25% increase in the probability that a capital raising firm chooses equity over debt. This supports the argument that due to their high cash volatilities and R&D expenditures high Cash(xb) firms will avoid debt and rely on the equity market if they need external financing. In regression 2 Cash(xb) is excluded, while Deficit is included in the regression. The Deficit coefficient is 0.684 (t-statistic = 29.95), showing that a one standard deviation increase in Deficit leads to a 11% in the probability that a capital raising firm chooses equity over debt. This shows that if a firm uses external financing to meet cash its needs, then it will most likely choose equity. Regression 3 includes both Cash(xb) and Deficit. The coefficients are similar to those in regressions 1 and 2, suggesting that each variable has a distinct effect on debt vs. equity choice. The Cash(xb)*Deficit interaction coefficient is insignificant, and it does not affect the coefficients when compared to regression 3. Hence, Cash(xb) and Deficit seem to have unique effects on capital structure choice, however both cash needs measures strongly predict choice of shares over debt. Taken in their entirety, the results here show that firms with high cash needs tend to choose equity over debt, which suggests that cash needs are an important determinant of capital structure changes. 4. Share Issuance Savings and Post-Issuance Stock Returns The Cash(xb) measure used in this study could be correlated to share price overvaluation. If this is so, then the high propensities of both high Cash(xb) and high MB(xb) firms to issue shares and keep the proceeds as cash might be the result of market 24
timing, rather than precautionary motives. In order to distinguish between the two hypotheses, the post-issuance share-price performances of issuing firms are analyzed. If Cash(xb) is correlated to overvaluation, then post-issuance returns ought to be lowest among the high Cash(xb) and high MB(xb) issuers. 4.1. Post Issuance Performance Measurement Portfolio abnormal returns are measured using calendar time portfolios. The calendar time portfolio methodology is used in Mitchell and Stafford (1998), Brav, Geczy, and Gompers (2000), Fama and French (2007), Hertzel and Li (2007), and others, and is advocated by Fama (1998). To form the portfolios, yearly quintiles are formed based on rankings of Cash(xb), MB(xb), and MB(e). Once created, the quintiles are limited to firms which, in the last three years, had at least one share issuance event that placed them in the highest net share issuance quintile among positive issuers in that year. If a firm was a large issuer in more than one of the last three years, then its cash and market-to-book rankings from the most recent year are used. The issuance measure in this study includes all share issuances that result in cash flows to the firm. Use of such a broad issuance measure in market timing tests is supported by the results in Daniel and Titman (2006), Pontiff and Woodgate (2006), Bradshaw, Richardson, and Sloan (2006), Fama and French (2007), and Mclean, Pontiff, and Watanabe (2007), who show that broad measures of share issuance, which include all share issuance transactions, are robust predictors of cross-sectional returns in both U.S. and in international stock markets. Moreover, Pontiff and Woodgate show that such an 25
issuance measure predicts low cross-sectional returns even if SEOs, share buybacks, and stock-financed mergers are excluded from the issuance measure. Abnormal returns are measured four different ways: Excess returns are the portfolio s monthly return minus the risk-free rate. CAPM-alpha is the intercept from a regression of the excess returns on the market factor (CRSP value-weighted index minus the risk-free rate). 3-factor alpha is the intercept from a regression of excess returns on the market, size, and book-to-market factors, while the 4-factor alpha is generated from regressions that include the momentum factor. 4.2. Post Issuance Performance Results Panel A of Table 7 displays the post-issuance performance of issuers across Cash(xb) quintiles. The results suggest that high Cash(xb) issuers are not overvalued, but that low Cash(xb) issuers might be (see Figure 4). In Panel A, the 3-factor and 4-factor alphas are -0.690% (t-statistic = -4.67) and -0.447% (t-statistic = -3.12) in the lowest Cash(xb) quintile. The alphas of the highest Cash(xb) quintile are positive, but insignificant. A hedge portfolio, that is long the high Cash(xb) issuers and short the low Cash(xb) issuers generates 3-factor and 4-factor alphas of 1.070% (t-statistic = 2.49) and 1.139% (t-statistic = 2.67) per month. Table 3 shows that the propensity to save issuance proceeds as cash was monotonically increasing in Cash(xb), so the propensity to save cash and overvaluation appear to be negatively correlated across share issuers. The results here reinforce the notion that firms which issue shares and save the proceeds do so to meet liquidity needs, and are not engaging in market timing. 26
Table 3 showed that the propensity to save share issuance proceeds as cash is monotonically increasing in MB(xb), but Panel B shows that the 3-factor and 4-factor alphas are negative and significant only for the lowest MB(xb) quintile. Moreover, the differences in 3-factor and 4-factor alphas between the high and low MB(xb) quintiles are 0.442 (t-statistic = 2.46) and 0.472 (t-statistic = 2.55). This suggests that the share issuances and subsequent savings by high MB(xb) firms were most likely done to meet legitimate cash needs. Panel C displays the returns of the MB(e) portfolios, and shows that the high MB(e) firms perform worst post issuance. The difference between the high and low portfolios is significantly different for three of the four performance measures. The differences in 3-factor and 4-factor alphas between the high and low MB(e) quintiles are 0.209 (t-statistic = -1.41) and -0.228 (t-statistic = -1.78). The low returns of the high MB(e) firms could be the result of market timing, however Table 3 shows that the propensity to keep issuance process as cash is not higher for high MB(e) firms as compared to low MB(e) firms, so the underperformance of these firms is not related to cash savings. 4.3. Post Issuance Performance Discussion Taken in their entirety, the results in Table 7 are incompatible with pure market timing. Pure market timing, as defined by Loughran and Ritter (1997) and Denagelo, Deangelo, and Stulz (2007), predicts that firms issue overvalued shares to increase their cash holdings. The results here show that firms with the worst post-issuance performance are the firms that spend the proceeds, as the low Cash(xb) and low MB(xb) firms have the 27
worst post-issuance performance and the lowest propensities to save. The results suggest that firms that issue shares and keep the proceeds as cash are most likely not market timing, but raising funds to meet legitimate cash needs. What can explain the poor performance of firms that issue shares and either invest or spend the proceeds? Answering this question empirically is beyond the scope of this paper, but there are a couple of good candidate explanations. If which spend the issuance proceeds spend it on capital expenditures and inventories, then poor post-issuance share price performance of firms which issue and spend could be consistent with the predictions in Carlson, Fischer, and Giammarino (2006) and Lyandres, Sun, and Zhang (2007). These papers contend that when firms issue shares and invest the proceeds they are exercising growth options. Once the exercise has taken place the firm is less risky, which causes the low post-issuance returns. If this type of risk is not captured in the 3- factor and 4-factor models, then it might explain the low post-issuance returns of the firms that issue and spend (low Cash(xb) and low MB(xb)) in Table 7. Firms that issue and save do not exercise growth options, so there is no reason for their returns to be lower post-issuance in a growth options framework. It could also be that the market overvalues some firms, and that the managers of these firms believe the false signal that the market is sending them. Morck, Shliefer, and Vishny (1990) discuss this possibility. Hence overvalued firms issue shares and invest the proceeds, believing that their projects are valuable, and then suffer low post-issuance returns reflecting their pre-issuance overvaluation. This is also consistent with Cooper, Guylen and Schill (2007) who show that growth in non-financial assets and subsequent 28