Determinants of Dividend Initiation by IPO Issuing Firms

Transcription

1 Determinants of Dividend Initiation by IPO Issuing Firms By Bharat A. Jain Department of Finance Towson University Towson, MD (410) and Chander Shekhar Melbourne Business School University of Melbourne Carlton, VIC 3053, Australia and Violet Torbey School of Business Bond University Gold Coast, QLD 4229, Australia JEL Classification: G31, G32 Keywords: Initial Public Offerings, Dividend Policy, Logit Models, Operating Performance, Financial Policy. Corresponding author. Phone +(613) ; Fax +(613)

2 Determinants of Dividend Initiation by IPO Issuing Firms Abstract This article studies the determinants of dividend initiations by newly public firms. We compare the post-ipo patterns in sales growth, capital intensity and profit margins between dividend and non-dividend initiating firms to assess whether significant differences exist in their external financing needs. In addition to size, profitability, and investment opportunities, we analyze the impact of variables unique to the IPO market in influencing the probability of post-ipo dividend initiation. We also examine factors influencing the timing of the dividend initiation decision and compare the characteristics of early versus late dividend initiators. We find that industry technological status, venture capital participation, number of uses of IPO proceeds, initial returns, and pre-ipo capital expenditures are significantly related to the probability of post-ipo dividend initiation. In addition, the duration between IPO and dividend initiation increases with venture capital participation, initial returns, investment bank prestige, and use of proceeds for general corporate purposes and decreases with stage of development of issuing firm at the IPO and size of offering. 2

3 Determinants of Dividend Initiation by IPO Issuing Firms 1. Introduction Initial public offering (IPO) firms typically go public on the promise of growth thereby conditioning investors to expect capital gains rather than dividends during the post-ipo phase. In fact, the offering prospectuses of most IPO issuers usually indicate that the firm is unlikely to pay dividends in the foreseeable future. IPO issuing firms are usually in their early stages of development and belong to rapidly growing, technologically oriented industries. They are expected to invest substantially in areas such as R&D, advertising, and capital expenditures, post- IPO, in an effort to gain market share and achieve technological dominance in their rapidly evolving industries. Issuing firms are often cash flow negative at the time of the IPO and are likely to continue to need substantial external financing during the post-ipo phase to sustain their high growth rates as well as help finance acquisitions. Further, the incidence of U.S. firms going public prior to achieving profitability has been growing over time. For instance, Ritter and Welch (2002) document that the proportion of firms going public with negative earnings in the 12 months prior to the IPO increased from 19% in the 1980s to 37% during the period and rose to an astonishing 79% during the internet bubble. Similarly, Fama and French (2001) report a significant dispersion in the profitability of new listings and a tendency for left skewness. Finally, even if issuing firms are profitable at the time of going public and in a position to pay dividends after the IPO, they may constrained from such an action as a result of loan agreements. As such, IPO firms are not expected to pay dividends during the initial post-issue phase. Fama and French (2001) note that characteristics of dividend paying firms include size, investment opportunities, and profitability. Extant empirical evidence indicates that in the post- IPO phase, issuing firms are high on growth and investment opportunities and low on profitability as demonstrated by strong post-ipo sales and capital expenditure growth, but a pattern of declining operating performance relative to the year of the IPO (Jain and Kini (1994), Mikkelson, Partch and Shah (1997)). As a consequence of their growth opportunities and low profitability, capital constrained IPO firms are likely to utilize all available internal capital to finance their investment opportunities. It is therefore, not surprising that one of the reasons for the declining propensity of U.S. corporations to pay dividends is attributed to the large number of new firms among listed companies (Fama & French (2001)). At some point during their post-ipo phase, however, IPO firms should reach a steady state in their growth patterns and therefore be in a position to consider initiating dividends. Fama 3

4 and French (2001) report that 25% of new lists that survive eventually start-paying dividends. The decision to initiate dividends, however, is likely to be contingent on several factors such as ability to generate and sustain free cash flows and investor reaction/expectations. The recent case of Cisco Systems is illustrative in this regard. The shareholders of the company voted ten to one against a proposal requiring the firm to pay dividends, despite the fact the company had accumulated approximately $21 billion of cash on its balance sheet and faced an unfavorable investment environment in the technology sector. Further, even when IPO issuing firms are in a position to initiate dividends, they may instead elect to repurchase shares thereby preserving financial flexibility with regard to future dividend payments (Jagannathan, Stephens, and Weisbach (2000)). A few recent studies have, however, directly addressed issues related to dividend initiations by IPO firms. They have primarily focused on the valuation effects (Lipson, Maquieira and Megginson (1998), McCaffrey and Hamill (2000) and Kosedag and Michayluk (2000)). For instance, Lipson et al. (1998) suggest that initiation occurs when managers believe that the dividends can be sustained by future earnings, and that managers may use initiation as a means to differentiate their firm from other similar, newly public firms. Kosedag and Michayluk (2000) compare market reaction to dividend initiating IPOs with dividend initiating reverse-lbos and find that the positive reaction is confined to dividend initiation by the IPO group. McCaffrey and Hamill (2000), using UK data, find a positive market reaction to dividend initiations by 270 firms (leading them to claim support for the signaling hypothesis), but also report that 90% of the dividend paying firms initiate dividends within the first year of going public. Bessler, Murtagh and Siregar (2001) study dividend policies of bank IPOs and find that although by the tenth year (after the IPO) two thirds of banks pay dividends, approximately 30% of the banks initiate dividends within the first twelve months of going public. An earlier study by Michaely and Shaw (1994) reports that approximately 22% of IPO firms start paying dividends within the first three years of going public. Further, Lipson et al. (1998) report that the average time to dividend initiation in their sample of IPO firms is about two and a half years. Although collectively these results suggest that the timing of the dividend decision may be relevant, none of the studies explore this issue in detail. In this study, we attempt to gain an understanding of the factors that influence both the decision to initiate dividends as well as the timing of this decision for IPO issuing firms. Focusing on factors that influence the probability of dividend initiation by IPO firms provides an understanding of the impact of variables unique to the IPO market on financial policy choices of entrepreneurial firms. In addition, by focusing on the timing of the dividend decision, we attempt 4

5 to identify characteristics of early (within twelve months of going public) versus late (after twelve months of going public) dividend payers as well as factors that influence the timing of the decision to initiate dividends. Our results indicate that approximately 21% of IPOs that survive ten years after the IPO initiate dividends. In addition, we find that for the most part, the decision to initiate dividends occurs early with approximately 71% of dividend initiations in our sample occurring within the first 12 months of going public. Approximately 81% of dividend firms were profitable at the time of the IPO. Regarding dividend initiation, our main results indicate that the probability of dividend initiation declines with number of uses of IPO proceeds, initial returns, risk of the issue, venture capital participation, membership in emerging industries, and pre-ipo capital expenditure intensity. The likelihood of dividend initiation, however, increases with the stage of development of issuing firm, and size of the offering. Further, the duration between IPO and dividend initiation increases with venture capital participation, initial returns, investment bank prestige, and use of proceeds for general corporate purposes and decreases with stage of development of issuing firm at the time of the IPO and size of the offering. In conjunction with these results, we find few significant differences in firm and offering characteristics between early and late dividend initiators they differ from each other only in the number of planned uses of proceeds declared in the prospectus and in the capital intensity at the IPO stage. We also study the post-ipo patterns in sales growth, capital expenditures, leverage, and profitability for both dividend initiating and non-dividend initiating firms. In terms of post-ipo performance, we find that the dividend group exhibits significantly higher post-ipo operating performance compared to similar non-dividend firms. This result is generally consistent with the notion of dividend policy being used to signal better performance in the future. However, the one, two, and three year, post-ipo sales growth rate of the dividend group is significantly lower than that of similar non-dividend firms but the four year growth rate is not significantly different. Finally, there is no significant difference between the dividend and similar non-dividend firms on post-ipo capital expenditure growth rate or leverage. Taken overall, the results support the notion that the likelihood of dividend initiation depends not only on the more obvious factors such as profitability, investment opportunities, and size, but also on aspects related to the IPO process. Further, the results indicate that the post-ipo investment opportunities for both dividend and nondividend firms are strong and therefore dividend initiation is not related to declining growth prospects. 5

6 The remainder of the paper is organized as follows. Section 2 describes the sample. Section 3 provides details of variable selection and methodology. The results are discussed in Section 4. Finally, Section 5 concludes the paper. 2. Sample Description Our initial sample consists of 5151 IPOs issued during the period and identified from Securities Data Corporation s (SDC) New Issues database. Consistent with most IPO studies, we exclude financial firms and unit offerings, resulting in a sample of 3503 IPOs. Next, we exclude best effort issues and require that issuing firms be listed on CRSP immediately after the IPO. These two restrictions further reduce the sample to 1640 IPO firms. Each firm is tracked on CRSP from the IPO date until the end of year 2000 or the delisting date (whichever is earlier) to determine whether the firm initiated an ordinary regular cash dividend. If the IPO firm initiated an ordinary regular cash dividend, the number of additional dividends paid after the initiation is also noted. For issuing firms that survived until the end of year 2000 and did not initiate an ordinary cash dividend, we go back to the IPO date and search whether the firm made any other type of distribution such as stock splits and stock dividends, liquidating dividends, exchanges and reorganizations, subscription rights, and notation of issuance in the post-ipo period. Our search for post-ipo distributions results in three categories of issuing firms; (1) issuers who initiated regular cash dividends, post-ipo, (2) issuers who initiated distributions other than regular cash dividends, post-ipo and (3) issuers who did not initiate any form of distributions from the IPO date until the end of year 2000 or delisting date, whichever is earlier. From the sub-sample of 111 issuers that initiated ordinary regular cash dividends we exclude 12 issuers where the dividends were recorded as non-recurring or extra/special. This results in a final sample of 1628 IPO issuers containing 99 dividend initiating firms and 1529 non-initiating firms. In order to test for robustness of our results we repeat the analysis with two different specifications for the comparison sample of non-dividend initiating firms. These two alternative specifications are; (1) non-dividend group consisting of IPO firms that did not initiate any form of post-issue distributions and (2) an industry and size matched sample of non-dividend firms. In both instances, the results are qualitatively similar to those reported here and are not reported for purposes of brevity, but are available from the authors. The final sample, therefore, includes 99 of the 1628 IPOs (6.08% of the total) that initiated dividends in the post-ipo period. Overall, 913 of the 1628 IPOs representing 56.08% of the sample were profitable prior to going public. Further, 81% of initiating firms were profitable 6

7 prior to the IPO. Therefore, firms that are unprofitable at the time of the IPO have an extremely low propensity to initiate dividends during the post-ipo phase. We also evaluate the timing of the dividend initiation for the 99 dividend firms in our sample. The mean (median) time in months from the IPO to dividend initiation is 15.61(6.00) months. Further, approximately 71% of dividend initiating IPOs (70 firms) initiated dividends within 12 months of going public. Table 1 provides the yearly distribution of the 1628 IPO issues in the sample. The results are reported for the overall sample and segmented by dividend versus non-dividend IPO firms. The full sample indicates significant variation in the volume of IPOs issued each year, with the largest number of issues in 1996 followed by 1993 and the lowest number of issues in Overall, 6% of issuers in the sample initiated dividends in the post-ipo period. Not surprisingly, the percentage of issuers in each year that subsequently initiate dividends increases with the time interval between the IPO year and the end of year Table 1 about here --- For instance, 14% of 1990 issuers initiated dividends during the 10 year post-ipo period. Similarly, 5% of 1995 IPO issuers initiated dividends in the five year post-ipo period. Further, 21% of surviving 1990 IPO issuers in our sample initiated dividends in the ten year post-ipo period indicating that one in five survivor IPO firms initiate dividends during the ten year post- IPO window a result that is consistent with Fama and French (2001) who find that approximately 25% of survivor firms among new lists eventually initiate dividends. 3. Variable Selection and Methodology 3.1 Description of Variables Industry Technological Focus Firms going public come from a diverse set of industries that can be broadly classified as emerging versus mature. Issuers in emerging, high technology industries face an innovation oriented environment, with no dominant technologies, shorter product life cycles, evolving product markets, and criticality of knowledge capital, Liu (2000). On the other hand, issuers in mature industries face a stable economic environment, with well-developed product markets and established industry leaders. Unlike mature industries, high technology industries are more susceptible to shocks as a result of changes in technology, government policy or supply and 7

8 demand conditions thereby creating additional demand for external financing as firms in the industry strive to adapt to structural changes occurring in the industry. In order to capitalize on industry growth opportunities and finance new and emerging technologies and product market strategies, issuing firms in high technology industries are likely to need substantial external financing, post-ipo. Further, financing growth opportunities affects the variability and permanence of a firm s cash flow stream which in turn influences its financial policy choices. Issuing firms in emerging industries need to keep their debt and borrowing capacity within manageable limits so as to be in a strong position to capitalize on both internal and external growth opportunities. Therefore, issuing firms in high technology industries are less likely to distribute dividends post-ipo and instead attempt retain available internal funds for their investment needs. By contrast, issuing firms in more mature industries are likely to be presented with fewer investment opportunities and are likely to be in a position to generate free cash flows which may then be paid to shareholders as dividends. Finally, issuing firms in emerging industries have a higher propensity to indulge in substantial stock option grants for incentive pay, recruitment, and retention of key employees. Therefore, even if IPO firms in emerging industries are in a position to distribute cash to shareholders they are likely to prefer a repurchase program over dividends. The above discussion leads us to expect a negative relationship between industry technological status and probability of post-ipo dividend initiation. Venture Capitalist Involvement Venture capitalists (VCs) generally demonstrate industry and product market preferences in the selection of firms they decide to finance. VCs, typically seek young, high growth, risky companies that have the potential to produce breakthrough products and services and achieve strong growth. Typically, they tend to make their investments at an early stage of development when the prospects of success are far from certain. In addition to providing financing, VCs tend to play an active role in the firm(s) by participating in activities such as strategic and operational planning, marketing, personnel selection, supplier management, capital structure choices and other facets of managerial decision-making as their portfolio companies evolve from start-ups to public corporations. VCs, therefore, have an influential role in the strategic evolution of the company and its investment and financing decisions. In addition, their investment horizon is generally not long term and they view the IPO as a mechanism to facilitate their eventual exit, with the investment objective being the realization of capital gains either at the IPO or shortly thereafter, rather than collecting on a stream of dividends out into the future. Recent empirical evidence supports the notion of early exit by the VCs - as noted by Field and Hanka (2001), venture capitalists sell more aggressively than other officers on the expiration of the lock-up 8

9 period after an IPO. Additionally, Allen, Bernardo, and Welch (2000) show that, in general, the presence of any shareholders who can mitigate the agency/information concerns but who do not value dividends, we would expect target companies to substitute re-investment, share repurchases, or other means of payout for dividend payments. For newly public firms, the venture capitalists fulfill this role during the early years, leading us to expect that venture capital involvement is negatively related to the probability of post-ipo dividend initiation. Uses of IPO proceeds Issuing firms are required to specifically identify the various planned uses of the IPO proceeds as well as the amount to be allocated for each use in the IPO offering prospectus. The usual categories of uses of IPO proceeds cited in offering prospectuses include expenditures for items such as general corporate purposes, acquisition financing, marketing and sales, product development/r&d, retiring/refinancing debt, etc. The planned uses of IPO proceeds can provide useful insights regarding aspects of financial and investment policies that in turn influence the firm s ability to pay dividends. The number of planned uses of IPO proceeds can be viewed as a measure of the extent of the issuing firm s financing needs. Therefore, firm s that indicate a larger number of uses of the IPO proceeds are generally likely to re-invest all their earnings, post- IPO and in addition turn to the external capital markets to meet any unfulfilled needs. By contrast, firms with a smaller number of planned uses for the IPO proceeds are less likely to require additional rounds of external financing. This lack of dependence on external financing may flow into the firms dividend policies and therefore, we expect a negative relation between the number of uses of the IPO proceeds and the probability of dividend initiation. In addition to the number of planned uses of the IPO proceeds, issuing firms also specify their plan regarding primary use of the IPO proceeds. Information regarding the primary use of IPO proceeds provides an indication of the firm s strategic intent in areas such as capital structure choices, growth strategy and product market strategy. For instance, issuing firms that indicate that the primary use of the IPO proceeds is for acquisitions are signaling their intent to pursue an aggressive growth strategy through acquisitions rather than internal growth. Aggressive serial acquirers are likely to retain all their earnings and in addition, if required turn to the capital markets for additional financing. Similarly, issuers that plan to use the IPO proceeds to rebalance their capital structure or reduce leverage are less likely to seek external financing and are in a better position to initiate dividends. Therefore, we expect the probability of dividend initiation, post-ipo to be related to the primary use of the IPO proceeds. 9

10 Stage of Development Issuing firms tend to go public at various stages of their development. Some firms tend to go public at a relatively early, developmental stage, while still not profitable, whereas others go public after several years of operations and after attaining profitability or being very close to attaining profitability. Firms going public relatively early face considerable uncertainty regarding future viability and the ability to attain and sustain profitability. These uncertainties arise from factors such as extent of product demand, technological risk, access to distribution channels, ability to secure future financing, and sustainability of growth rates. Firms going public at a later stage in their development are either profitable or are in a position to demonstrate a clear path to profitability and many of the above uncertainties have been resolved or minimized. Such firms are likely to be in a better position to initiate dividends post-ipo. Therefore, we expect a positive relationship between the stage of development at the time of the IPO and probability to initiate dividends post-ipo. Size As pointed out by Fama and French (2001) dividend payers are significantly larger than non-dividend payers. Kahle (2002) also notes that size is a proxy for financing costs and that larger firms have less information asymmetry, stronger cash flows and lower financing costs. Therefore, we expect that the probability of dividend initiation is positively related to size of the firm. Leverage The extent of debt on the balance sheet is likely to have a significant impact on the IPO firm s dividend initiation decision. Smith and Watts (1992) argue that firms with greater growth opportunities will tend to rely less on debt financing. Therefore, we expect a positive relationship between leverage and the probability of dividend initiation by IPO firms. Capital Expenditures Investment opportunities clearly determine the probability of dividend initiation by IPO firms. The investment opportunities available to the IPO firm can be assessed by the extent of managerial commitment to capital expenditures. Prior empirical research in the IPO market has documented that issuing firms substantially increase their capital expenditures post-ipo. Jain and Kini (1994) find that even after adjustment for industry effects, IPO firms have substantially higher growth in post-ipo capital expenditures in comparison to similar firms. Bommel and Vermaelen (2003) report that ceteris paribus, IPO firms that receive positive feedback from the market during the process of going public will spend on average 16.9 % more on capital 10

11 expenditures compared to companies that receive negative feedback. Kahle (2002) argues that firms with high capital expenditures have better investment opportunities and less free cash flow and therefore are less likely to be in a position to initiate dividends. Lower capital expenditures, on the other hand, leads to lower external financing needs and therefore results in a higher likelihood of dividend payments. 3.2 Methodology Multivariate logistic regression analysis is used to predict the probability of dividend initiation by an IPO firm in the post-ipo period. The basic form of the logit model is as follows: Prob (Dividend Initiation) = β 0 + β 1 TECH + β 2 VC + β 3 UNDPR +β 4 STAGE + β 5 NUMUSES + β 6 ACQ + β 7 REF +β 8 GEN + β 9 DEBT +β 10 CAPEXA + β 11 LAMT + β 13 RISK + β 14 RANK The dependent variable is the probability that an IPO issuing firm will initiate dividends, post-ipo. Using the SDC definition of high- technology industries, the IPO sample is segmented into two groups based on whether the issuing firm belongs to a high or low tech industry. The variable TECH takes on the value 1 if the IPO firm belongs to a high technology industry and 0 otherwise. The variable VC represents venture capital participation and takes on the value 1 if the issuing firm is VC- backed and 0 otherwise. The variable, NUMUSES measures the number of uses for the IPO proceeds listed by the firm in its IPO prospectus. Three dummy variables are constructed to measure the primary use of the IPO proceeds as follows - ACQ takes on the value 1 if the issuing firm primarily intends to use the proceeds for acquisitions and 0 otherwise, REF takes on the value 1 if the primary use of proceeds is intended for debt reduction/refinancing and 0 otherwise, and finally GEN takes on the value 1 if the primary use of proceeds is for general corporate purposes and 0 otherwise. The stage of development of the issuing firm is proxied by pre-ipo profitability. Firms that are not profitable at the time of the IPO are considered as developmental stage firms while those profitable at the IPO are more mature, later stage firms. The variable STAGE takes on the value 1 if the IPO issuing firm attains profitability prior to the IPO and 0 otherwise. Initial returns are measured by the variable UNDPR expressed as the percentage difference between market price at the end of the first day of trading and the offer price. The initial returns are a measure of the market s reaction to the offering and a stronger reception should lead to higher post-ipo investment activity and therefore lower probability of dividends. Leverage is represented by the variable DEBT measured as the long term debt divided by total assets. CAPEXA is the capital expenditure divided by the total assets in the year of the IPO. Finally, we include offer size, investment bank prestige, and risk of the issue as control variables. These three variables have been widely used in the IPO literature. The variable LAMT proxies for 11

12 size and is measured as the the gross proceeds raised at the IPO. The variable RISK is measured as the standard deviation of the first thirty days returns in the aftermarket. Our proxies for size and risk are generally consistent with the IPO literature (see for example, Ritter (1991), Jain and Kini (1994,1999)). Consistent with the previous literature, we use the Carter, Dark and Singh (1998) nine-point investment banker prestige ranking system to measure investment bank prestige (RANK). In results not reported in the paper, we also evaluate the impact of additional control variables such as managerial ownership retention with qualitatively similar results. The overall model fit is assessed by the likeliihood ratio and the significance of the individual coefficients is assessed by the p-values. 4. Results 4.1 Summary statistics In Table 2, the industry distribution of IPO firms in the sample is reported. The industry groupings are defined on the basis of two digit SIC codes and the industries are listed in order of decreasing number of IPOs that occurred in each industry during the sample period. --Table 2 about here -- The largest number of IPOs from a particular industry occurred in the Advertising and Business Services sector with 403 firms or 25% of the sample drawn from this sector. The next two highest IPO industries are Electronics and Communication with 146 IPOs (9% of the sample) and Medical and Photo Equipment with 139 IPOs (8.5% of the sample). The last column of Table 2 reports the proportion of dividend initiating IPO firms in each industry. The distribution of proportion of dividend initiating IPO firms varies from a low of zero percent for the Motion Pictures /Entertainment Services industry to a high of 36% for the Metal Products industry. Three other industries exhibit a relatively high proportion of dividend initiating firms - Utilities (30%), Transportation and Shipping (16.28%), and Mining/Construction (14%). As expected, the above results indicate that the propensity to pay dividends varies depending on the specific industry environment. Table 3 provides a comparison of IPO firm and offering characteristics for dividend versus non-dividend IPO firms. The results indicate that dividend firms raise significantly higher proceeds at the IPO with a mean(median) value of $81.69($50.40) million compared to a 12

13 mean(median) of $32.02($24.00) million for the non-dividend group and the difference is statistically significant at the 0.01 level. -- Table 3 about here -- By contrast, the initial returns of the dividend group are significantly lower than that of the nondividend group. Since initial returns have been associated with the extent of information asymmetry surrounding the prospects of the issuing firms, these results are indicative of lower uncertainty at the IPO for the dividend group. There is no significant difference between the two groups on managerial ownership retention in the post-ipo firm. Dividend firms have significantly lower planned uses of the IPO proceeds compared to the non-dividend group. Further, the dividend group is characterized by significantly higher investment banking prestige and significantly lower risk of the issue compared to the non-dividend group. The results in Table 3 also indicate significant variation between dividend and nondividend IPOs on firm characteristics. Although the dividend firms have significantly higher levels of total assets, sales, and leverage at the time of going public compared to non-dividend IPO firms, the capital intensity at the time of the IPO (measured as the capital expenditures divided by total assets) is similar across the two groups. Dividend firms are also significantly more profitable at the time of going public compared to non-dividend firms. Further, there are significant differences between dividend and non-dividend IPO firms on aspects such as industry technological status, venture capitalist participation, and choice of exchange listing. For instance, 13% of dividend firms are in high technology industries compared to 44% of non-dividend firms. Additionally, 13% of dividend IPO firms attracted venture capital financing compared to 46% for non-dividend firms. Therefore, the evidence indicates that non-dividend firms are more likely to belong to high technology industries and attract venture capital financing compared to dividend firms. There are also substantial differences between the two groups in terms of exchange listing -- 40% of dividend firms list on NYSE compared to only 5% of non-dividend firms. Similarly, 82% of non-dividend firms list on NASDAQ compared to 53% of dividend firms. Finally, in results not reported here, we also gauge investor interest in the proposed IPO of dividend versus non-dividend firms by examining the relation between final IPO offer price to the anticipated offer price in the preliminary prospectus. The results suggest that 9.47 % of dividend firms are priced above the initial filing range compared to % for the non-dividend group. Almost two-thirds (66.3 %) of issuers in the dividend group were priced within the initial filing range compared to % for the non-dividend group. Similarly, about one-quarter 13

14 (24.21 %) of issuers in the dividend group were priced below the filing range compared to 17.36% for the non-dividend group. Thus, ex-ante there does not appear to be significant differences in investor reception to IPOs from dividend versus non-dividend firms. We also compare the characteristics of dividend versus non-dividend IPOs for a subsample of 913 IPOs that were profitable in the year of the IPO. This sub-sample represents issuing firms that are in a position to initiate dividends in the post-ipo period. However, only 80 firms representing 9% of firms that are profitable at the time of the IPO initiate dividends. We find similar significant differences between dividend and non-dividend firms among issuers profitable at the time of the IPO as was documented for the overall sample. For instance, the pre- IPO profitable dividend group raises significantly higher proceeds at the IPO, is marketed by more prestigious investment bankers and is more leveraged compared to the non-dividend group. IPO firms can initiate dividends soon after the IPO (within the first year of going public) or wait for several years after the IPO before initiating dividends. Approximately 71% of the dividend firms in our sample elected to initiate dividends within 12 months of going public. Firms that initiate dividends early usually state their intent to pay dividends in the offering prospectus. On the other hand, late dividend initiators indicate an inability to initiate dividends within the foreseeable future in the offering prospectuses. We compare the characteristics of early versus late dividend initiators to determine if certain offering or firm characteristics can explain the timing of the dividend decision. The results are reported in Table 4. We find that apart from the number of uses in the IPO proceeds, early and late dividend initiators are indistinguishable. The mean (median) number of uses of proceeds for the early dividend group is significantly lower than that of the late dividend group. --Table 4 about here- 14

15 4.2 Uses of Proceeds Table 5 describes the distribution of the planned primary use of IPO proceeds as indicated by the issuing firm in the offering prospectus. The results are reported for the overall sample and also segmented by dividend versus non-dividend firm. The IPO issuing firms are segmented on the basis of the three most commonly cited primary uses of the IPO proceeds - acquisition financing, general corporate purposes, and retiring/refinancing debt (both bank loans and other debt). -- Table 5 about here All the remaining primary uses of proceeds are included in the Others category. The results in Table 5 indicate that approximately 54 % of issuing firms cite general corporate purposes as the primary use of IPO proceeds while 35 % cite retiring/refinancing debt, 3 % acquisition financing and the remaining 8 % cite reasons included in the Other category. The comparison of the primary uses of IPO proceeds for dividend versus non-dividend firms reveal significant differences between the two groups - 55% of non-dividend firms cited general corporate purposes as the primary use of IPO proceeds compared to 35% for dividend firms. On the other hand, the most commonly cited use of the IPO proceeds for the dividend group was retiring/refinancing bank loans/debt with 42% issuers citing this reason compared to 34% for the non-dividend group. The results indicate that there is little difference in the percentage of issuers that cite acquisition financing as the primary use of the IPO proceeds for the dividend and nondividend group Post-IPO Sales growth, Capital Expenditure growth and Profitability In this section, we explore the differences between dividend and non-dividend IPO firms on their post-ipo sales growth, capital intensity, leverage, and operating performance. The comparison is made over several time windows measured related to the year of the IPO. The performance of the dividend group is compared with a sample of all non-dividend IPOs as well as with an industry and size matched sample of non-dividend IPOs. The matched sample comparison allows us to compare the post-ipo performance of dividend firms versus similar nondividend firms. The matched sample is constructed by identifying a matching non dividend firm for each dividend firm in the sample on the basis of industry and size of offering. Therefore, for 15

16 each dividend IPO firm, a matching non-dividend firm in the same four digit SIC code and closest to the dividend firm in terms of IPO offer size is identified. In instances where a suitable four digit match is not available, we identify a similar firm in the same three-digit SIC code. The results of the comparison of post-ipo sales growth, capital expenditures growth and change in operating performance for the dividend group versus non-dividend group and industry and size adjusted non-dividend group is reported in Table 6 for several post-ipo time windows. The year 0 represents the fiscal year of the IPO while years 1, 2, 3, and 4 represent post-ipo fiscal years. -- Table 6 about here In panel A, a comparison of the sales growth rate of the dividend and non-dividend group is provided. The median percentage change in sales growth is reported for years 1, 2, 3,and 4 relative to year 0 for the two groups. The median one, two, three, and four year post-ipo sales growth rates for the dividend group are 11%, 24%, 43% and 71% compared to 35%, 68%, 103%, and 129% respectively, for the non-dividend group and the differences are statistically significant for all four time windows. Therefore, while dividend and non-dividend IPO firms both demonstrate strong post-ipo sales growth, the non-dividend group demonstrates significantly higher post-ipo growth rates. The comparison of the post-ipo sales growth rate for the dividend and matched sample of non-dividend firms is essentially similar with one exception. While the dividend group has significantly lower one, two and three year growth rates compared to similar non-dividend firms, the four year growth rate for the two groups is not significantly different. The question of whether dividend firms tend to have lower post-ipo capital expenditures compared to non-dividend firms is addressed by comparing the post-ipo capital expenditure growth rate. In panel B, the median percentage change in capital expenditure intensity (measured as capital expenditure over total assets) is reported for dividend and non-dividend groups. The results indicate that while both dividend and non-dividend firms demonstrate substantially high capital expenditure growth rates, for the most part, the capital expenditure growth of the dividend group is significantly lower than the non-dividend group. However, the comparison with the matched sample indicates that there is no significant difference in capital expenditure growth rates between dividend and similar non-dividend firms. In panel C, the results indicate that except for the first year after the IPO, both dividend and non-dividend firms increase their leverage in the post-ipo years. On the other hand, there is no evidence to indicate that the two groups significantly differ in terms of the extent of leverage. 16

17 We also compare the post-ipo profitability of dividend and non-dividend firms to assess whether significant differences exist. Panel D of Table 6 provides a comparison of the operating return on assets. Consistent with the existing literature on post-ipo performance, both dividend and nondividend IPO firms demonstrate a decline in post-ipo operating performance relative to the IPO year. However, the decline in post-ipo operating return on assets is significantly lower for the dividend group compared to the non-dividend group for all time windows. For instance, the median change in operating return on assets for the years 1, 2, 3, and 4, relative to year 0 are %, -2.03%, -5.05%, and % for the dividend group and -6.50%, %, %, and % for the non-dividend group and the difference is statistically significant over all time windows. The comparison with the industry matched sample is essentially similar. In panel E, the median change in operating return on sales for the dividend and nondividend groups is provided. The results indicate that at least in the years +1, +2, +3 relative to year 0, the dividend group shows no decline in operating performance. The dividend group does, however, demonstrate a significant decline in operating performance over a four year window relative to the IPO year. The non-dividend group, on the other hand, demonstrates a significant decline in operating performance in each of the four time windows. Further, the dividend group demonstrates significantly superior performance relative to the matched sample of non dividend IPO firms in each of the four post-ipo time windows. Overall, the results indicate that dividend IPO firms grow at a slower rate and exhibit superior operating performance relative to the non-dividend group in the post-ipo years. However, both dividend and non-dividend firms have strong post-ipo investment opportunities and the growth rates in capital expenditures for the two groups are not significantly different for all post-issue time windows. 4.4 Estimation of Logit Models To identify the variables that significantly influence the probability of post-ipo dividend initiation, we estimate several logit models. These results are reported in Table 7. All models include a common set of three control variables; LAMT, RISK and RANK. The estimated log likelihood coefficients, p-values and significance levels for each variable are reported along with the likelihood ratio for the overall model. -- Table 7 about here In model 1, the impact of industry technological orientation and venture capital participation on the likelihood of post-ipo dividend initiation is explored. In addition to the three 17

18 control variables, the industry technological focus variable (TECH) and the participation of venture capital in an IPO firm (VC) are included as explanatory variables. The coefficient of TECH is negative and significant ( , p-value=0.024), indicating that the probability of dividend initiation declines if the issuing firm belongs to a high technology industry. The variable VC is also negative and significant ( , p-value=0.0002) indicating that the probability of dividend initiation post-ipo declines with VC participation. The coefficient of LAMT is positive and significant indicating that larger offerings (and by implication, larger firms) have a higher probability of initiating dividends. On the other hand, the coefficient of RISK is negative and significant indicating lower probability of dividend initiation by riskier issues. The control variable investment bank prestige is positive but not statistically significant. In model 2, we examine the impact of industry technological conditions as well as the issuer s indicated uses of the IPO proceeds on the probability of post-ipo dividend initiation. In addition to the independent variables in model 1, we include the number of uses of the IPO proceeds (NUMUSES) and the three variables that measure the most commonly cited primary use of the IPO proceeds (ACQ, GEN, REF). Consistent with model 1, both industry technology status and VC participation remain negative and significant. The coefficient of the variable NUMUSES is negative and significant ( , p-value=0.0263) indicating that the probability of dividend initiation increases with the decrease in the number of uses of the IPO proceeds. The disaggregated variables measuring the various primary uses of the IPO proceeds, however, are statistically insignificant, suggesting that such classification by the issuer in the prospectus has no impact on the future likelihood of dividend initiation. In model 3, in addition to TECH, VC, and NUMUSES, variables such as developmental stage of issuing firm (STAGE), leverage (DEBT), pre-ipo capital expenditure intensity (CAPEXA) and initial returns at the IPO (UNDPR) are included as explanatory variables. Consistent with the earlier models, TECH, VC, and NUMUSES are significantly negatively related to the probability of dividend initiation. Although the likelihood of dividend initiation increases with the firm leverage, the results are not significant. The coefficient of UNDPR and CAPEXA is negative and significant. Further, the coefficient of STAGE is positive and significant indicating probability of dividend initiation increases with stage of development of issuing firm. Model 4 includes all the independent variables that were found to be significant in the first three models. The direction of the relationship and significance of the variables are consistent with the earlier results, supporting the notion that these variables influence whether a newly public company chooses to initiate dividends or not. Finally, in model 5, we include all 18

19 the independent variables that are used in Models 1-4 to test for their significance. As evident from the results in the last column of Table 7, our basic conclusions hold as all significant variables retain their directional influence as well as their statistical significance. Considered overall, the logistic regressions presented here indicate that the probability of dividend initiation declines if the IPO issuer belongs to a high technology industry or receives venture capital financing. Further, the probability of dividend initiation decreases with the increase in the number of uses of the IPO proceeds, pre-ipo investment in capital expenditures, initial returns at the IPO and the risk of the issue. Finally, the probability of dividend initiation increases with the stage of development of the issuing firm and size of the offering. Factors such as pre-ipo leverage and investment bank prestige are unrelated to the probability of dividend initiation of newly public companies. In addition, in results not reported here but available with the authors we re-estimated the logit models described in Table 7 by confining the dividend sample to early dividend paying firms only. The results are qualitatively similar to those reported above Timing of Dividend Initiation Decision Hazard analysis methodology is employed to identify factors influencing the timing of the dividend initiation decision by IPO issuing firms. Hazard analysis is a statistical technique that has been employed to study the occurrence and timing of events (Allison 1984; LeClere 2000). An event represents a change of state and the duration of time between states is referred to as failure or event time (LeClere 2000). The event of interest in our application is that of dividend initiation. Hazard analysis is not only capable of predicting whether an event will occur or not but also when the event occurs. The primary benefits of hazard analysis over regression analysis or qualitative response models lies in their ability to explicitly account for time and handle censored observations and time varying covariates (LeClere 2000; Shumway 2001). Censoring refers to the situation where the event of interest has not as yet occurred at the end of the observation period or the firm has left the sample before the end of the observation period for reasons other than the event of interest. Therefore, the time spent by the firm or individual in the origin state is incomplete and the duration until event is known for only a portion of the sample (LeClere 2000). Specifically, in our application, censoring occurs for firms that do not initiate dividends until the end of the tracking period or leave the sample as a result of being acquired or delisted. Survival analysis models use estimation techniques that incorporate information from both censored and uncensored observations to provide consistent parameter estimates (Allison 2000). Furthermore, Shumway (2001) finds that hazard models are both theoretically and empirically preferable to static models and also perform as well as or better than alternatives. We select the widely used 19

20 Cox Proportional Hazard model to identify variables that influence the timing of the dividend initiation decision by IPO issuing firms. A detailed discussion of CPH models is available in Cox (1972). We define the event in our analysis as the initiation of dividends by IPO issuing firms. In the post-ipo period, issuing firms will either initiate dividends, leave the sample as a result of acquisition or failure or continue to operate until the end of the tracking period without initiating dividends. Censored firms are defined as IPO firms that do not initiate dividends until the end of the tracking period. The event time is measured as the number of months from the IPO until dividend initiation or the end of the tracking period whichever comes first. The dependent variable is the log of the hazard. The explanatory variables include all the variables evaluated in the logistic regression analysis in determining the probability of dividend initiation. The results of the hazard analysis are reported in Table 8. The overall model Chi-square, individual variable coefficients and their associated p-values are reported. Since the dependent variable is the log of the hazard rate, a positive coefficient on an explanatory model indicates that an increase in the value of the independent variable is associated with an increase in the hazard rate and consequently lower duration of the event of interest. Therefore, in the context of our study, a positive coefficient indicates than an increase in the value of the independent variable results in a decrease in the time to dividend initiation. Similarly, a negative coefficient indicates that an increase in the value of the variable leads to a longer time to dividend initiation. Finally, the risk ratios are reported. For indicator variables the risk ratio is interpreted as the estimated hazard of dividend initiation for those with value 1 relative to the hazard of dividend initiation for those with value zero after controlling for other covariates. For continuous variables, the risk ratio indicates the percentage change in the hazard of dividend initiation for a unit increase in the covariate of interest controlling for other factors (Allison 2000) --Table 8 about here-- The results in table 8 indicate that several variables are significantly related to the time to dividend initiation decision. The coefficient of the variables VC, UNDPR, GEN and RANK are negative and significant while the coefficient of the variables STAGE and LAMT are positive and significant. Therefore, the hazard analysis indicates that time to dividend initiation increases with initial returns, VC participation, investment bank prestige, and use of proceeds for general corporate purposes. Similarly, the time to dividend initiation decreases with the stage of development of the IPO firm and amount of proceeds raised at the IPO. The hazard or risk ratios help increase our understanding of the interpretation of the coefficients in hazard analysis (LeClere, 2000). Risk ratios equal to one indicate the variable has no effect on the event while a 20