Intangible firm IPOs: the effect of intangibles on IPO outcome
University: University of Amsterdam, Amsterdam Business School Name: Andrew Leek Student number: 10561897 Study: MSc FIN: Corporate Finance Supervisor: Dr. T. Ladika Date: June 9, 2018 Over the years, firms have been increasingly moving from physical investment to intangible investment. Firms with more intangibles do not need as much cash for capital expenditures compared to more traditional firms, as intangible firms do not own as many major physical assets. It is unclear why these intangible firms consider an IPO to raise money. With a hand‐collected sample the relation between intangibles, IPO proceeds and management shareholdings is tested to investigate the results of going public. The results show that high intangible firms float less shares to the public and raise less capital in an IPO, while the results on management shares are inconclusive.Statement of Originality This document is written by Student Andrew Leek who declares to take full responsibility for the contents of this document. I declare that the text and the work presented in this document are original and that no sources other than those mentioned in the text and its references have been used in creating it. The Faculty of Economics and Business is responsible solely for the supervision of completion of the work, not for the contents.
Table of contents
1. Introduction ... 4 2. Literature ... 7 2.1. IPOs ... 7 2.2. Cash holdings ... 8 2.3. Investment and ownership ... 10 2.4. Intangible capital ... 11 2.5. Contribution ... 12 3. Methodology ... 13 3.1. Hypotheses ... 13 3.2. Method ... 16 3.3. Data collection ... 17 3.4. Variables ... 18 4. Results ... 21 4.1. Regressions ... 21 4.2. Robustness ... 25 4.2.1. Alternative explanatory variable ... 26 4.2.2. In‐sample regressions ... 27 4.3. Discussion ... 28 4.3.1. Interpretation results ... 28 4.3.2. Comparison results with literature ... 30 5. Conclusion ... 31 6. References ... 32 7. Appendix ... 34
1.
Introduction
In finance, one of the most prestigious possibilities for a company is an initial public offering (IPO). Bringing a company to the public market also opens up many possibilities to create value. An IPO may also positively signal strong historical earnings and thus a potentially attractive investment (Brau & Fawcett, 2006). Traditionally, companies use an IPO to receive a cash influx or raise additional funds according to Brau & Fawcett (2006), although lately there are less firms that go public (Gao et al. 2013). Coinciding with this decrease in IPOs are the changes in firm characteristics. Firms are increasingly moving towards more intangible investment compared to physical investment. These intangible investment firms have increasingly low asset tangibility, which leads to more internal financing of investments (Almeida & Campello, 2007). Intangible investment is more dependent on human capital than physical investment. This makes intangible assets require less upfront capital investment than tangible assets (Doettling et al., 2016). Investment theory has traditionally focused on physical capital (Peters & Taylor, 2017). Firms with more intangibles do not need as much cash for capital expenditures compared to traditional firms, as intangible firms do not need machinery or major physical assets. It is unclear why these intangible firms consider an IPO, when theory predicts that they do not require as much the capital for investment as traditional firms. Investors may therefore be less confident in an IPO of an intangible firm. These theoretical predictions mentioned raise the question of whether intangible firms raise less IPO proceeds than physical investment firms. When intangible firms consider an IPO, they are often highly publicized events, as seen for example with Facebook, LinkedIn and Snap. There are many more examples of these kind of companies in the tech and pharmaceutical industry. These companies do not require capital to make physical investments, but still go public to attract funds. A characteristic outside of investment that often differs between intangible firms and traditional firms, is that the founders are often active as management or in the board of intangible firms. Often, they hold a large amount of shares and have significant influence on the company’s strategy (Fahlenbrach, 2009). Even after the IPO, company founders keep many shares to maintain control over their company, as seen with Facebook and Google for example. Just like IPO proceeds is expected to be influenced by intangibles, the amount of shares held by management is maybe influenced by intangibles. Therefore, the question israised if management holds more shares in intangible firms than in physical investment firms at IPO. The topics discussed above are researched by existing literature, but they have not been tested together. The shift from physical to intangible investment described in Corrado et al. (2010), has not been linked to IPO and founder‐CEO data in existing literature. There is a lot of existing literature on IPOs and founder‐CEOs, while intangible capital literature is currently growing. To contribute to the literature, the difference between intangible and physical investment firms is tested for economical and statistical significance. This is done by investigating intangibles, IPO proceeds and management shares, which will show that they are related and influence each other significantly. Literature written about IPOs, cash holdings, founder‐CEOs and intangible capital are further built upon to answer the research questions. The first hypothesis will look if intangible firms float less public shares in an IPO than traditional firms. Intangible firms are expected to float less shares to the public. The second hypothesis investigates if intangible firms raise less capital from IPOs than traditional firms. Again, intangible firms are expected to raise less capital in an IPO. The third and last hypothesis tests if management of intangible firms hold more shares after IPO than in traditional firms. Management shares of high intangible firms are expected to be higher than for more traditional firms. The hypotheses are tested using OLS regressions to estimate the coefficients. Intangible Investment is the main explanatory variable that will be regressed on Shares Floated, IPO Proceeds and (Post‐IPO) Management Shares. Each regression will use control variables to control for firm and ownership characteristics. Robustness regressions are used to investigate the consistency of the tests by using slightly different intangible measures. Data on these variables is collected from Compustat, Zephyr and SEC S‐1 IPO filing forms. Two samples are collected, the high intangible (HINT) and the low intangible (LINT) sample. For the main results these samples are combined to make a single sample analysis. Variables not provided directly by the data sources are created by hand from existing and hand‐collected data. The results show support for the first hypothesis. Intangible Investment significantly influences Shares Floated negatively in each regression specification. When Intangible Investment increases by one, the ratio of shares floated decreases by a very statistically significant 0.1179. This implies that a one standard deviation increase in Intangible
Investment leads to 0.003 decrease in Shares Floated, which is 1.4% of the mean value of Shares Floated. The second hypothesis is also supported by the results. IPO Proceeds are significantly lower when Intangible Investment increases. When Intangible Investment increases by one‐standard deviation, this leads to a 0.1351 decrease in ln(IPO Proceeds). For the average firm this indicates a decrease in IPO proceeds of $31.85 million. The third hypothesis is not supported by the main results. Intangible Investment does not seem to statistically influence Management Shares. An alternative main explanatory variable R&D Expenses does statistically affect Management Shares, but this is a negative relation which is against the expectation of the hypothesis. Concluding, HINT firms float less shares and raise less capital in an IPO. There is insufficient statistical evidence to say that management shares differ between HINT and LINT firms. Future research can investigate what HINT firms spend their IPO proceeds on and if the lower proceeds influence this. Also, the relation between intangibles and management shares requires more statistical evidence to prove a causal relation. In the upcoming sections the following topics are discussed: Firstly, current literature on this topic will be reviewed. This literature includes: IPOs, cash holdings, founder‐CEOs and intangible capital. Secondly, the collection and corrections of the data are explained. Thirdly, the methodology will be explained and the corresponding hypotheses clarified. Also, the regression methods will be further detailed. Fourthly, the results will be interpreted and these results will then be compared to other results found in current literature. Lastly, a conclusion will be drawn from these results and findings. Also, possible future research topics that may build upon this thesis are suggested.
2.
Literature
For this thesis, multiple areas of finance literature are used to build upon. In section 2.1, IPO literature related to HINT firms is discussed. Section 2.2 discusses relevant cash holdings literature, and section 2.3 discusses investment and ownership of HINT firms. Section 2.4 looks at intangible investment, and lastly, and finally, section 2.5 describes the contribution of this thesis to the literature. There is a slight overlap in topics, as some papers mention several aspects of multiple topics. 2.1. IPOs Gao et al. (2013) investigate why the number of IPOs in the US has decreased over the years. They find that from 1980 to 2000, the average number of IPOs per year was 310, while from 2001 to 2012 this were 99 IPOs. Gao et al. (2013) find that their economies of scope hypothesis holds. This hypothesis describes that the advantages of selling out to a larger company have increased. Selling out is more attractive, because there has been a change in many industries whereby products have to be brought to the market quicker. This is correlated with the increasing technology importance (Corrado & Hulten, 2010). The motivation of 336 CFOs for conducting an IPO is investigated by Brau and Fawcett (2006). In their sample, the authors have a separate measure for high‐tech firms. These firms are comparable to the HINT sample in this thesis and can therefore be used to identify the most important reasons for an IPO. According to the survey of Brau and Fawcett, the most important reasons for going public with a high‐tech firm are to create public shares to use in acquisitions, establish a firm price and enhance firm reputation. When compared with firms that are not defined as high‐tech, analyst attention and reputation are significantly more important for high‐tech firms. Factors that are less important for high‐tech firm IPOs are reducing cost of capital and creating cheaper financing because debt is too expensive. Another angle to look at IPOs is from a share ownership perspective. Alavi et al. (2008) investigate managerial ownership specifically. They find that pre‐IPO management share ownership controls management to lower the costs of the IPO compared to other share owners. This is due to management not using the IPO as an exit strategy, while other shareholders are more likely to exit. Jain and Tabak (2008) look at the factors that influence
the choice between a founder and a non‐founder CEO for IPO firms. The choice between a founder or non‐founder CEO can be viewed within the signaling theory. Having a founder‐ CEO gives a signal to the market about strategy, growth and potentially investment (Jain & Tabak, 2008). Founders that are experienced in product R&D more often retain their CEO position at IPO. Connected with the higher management ownership in HINT firms is the public float of an IPO for HINT firms. Michel et al. (2014) find that in the long run a higher public float leads to lower incentives for insiders to perform. A high public float leads to lower post‐IPO returns for insiders, while new public investors gain monitoring and governance. 2.2. Cash holdings Firms may do an IPO to increase their spending capacity (Brau & Fawcett, 2006). IPO proceeds are mostly in the form of cash. The difference of cash holdings between HINT or LINT firms may influence the amount of IPO proceeds for HINT and LINT firms. Table 1 (Appendix 1) shows that HINT firm hold significantly more cash reserves. The literature distinguishes several motives for holding excess cash are: agency, transaction, precautionary and tax (Opler et al., 1999; Bates et al., 2009). The agency motive of excess cash holdings has been investigated by Jensen (1986). The paper describes the agency problem that managers are agents of the shareholder principals. Managers decide what happens to the free cash flow of a firm, while shareholders provide the cash. Debt has a control function of reducing free cash flow that managers can waste. Opler et al. (1999) also predict that managers prefer higher cash holdings, as it may reduce firm risk and increase discretion. According to Harford et al. (2008), firms that have high insider ownership have higher cash holdings. Weak shareholder rights decrease cash holdings. The amount cash held also increases the amount of acquisitions made. R&D is found to be statistically unrelated to cash holdings. Thakor and Lo (2015) oppose this view by finding that R&D intensive firms carry more cash, due to uncertainty of debt financing. Debtholders may force the firm to repay the debt early, causing a problem for the illiquid R&D investment. Holding cash avoids uncertainty of future financing, as the cash is readily available. There are many more papers that prove that excess cash is related to agency theory. Dittmar et al. (2003) prove that firms hold more cash when access to funds is relatively easy. Faleye (2004) proves that proxy contests are
more common when firms hold excess cash reserves. Finally, Dittmar and Mahrt‐Smith (2007) show that each dollar of cash is worth more in firms with good corporate governance. The precautionary motive of high cash holdings is also well established in the literature. Opler et al. (1999) find that firms with risky cash flows and growth opportunities tend to hold larger amounts of cash over time. Bates et al. (2009) confirm this and also show that cash holdings have doubled over their sample time. The increase in cash is mostly due to riskier cash flows and more uncertainty. Firms have higher cash ratios because of lower inventories, lower receivables, higher cash flow and decreasing capital expenditures. Also, cash can be used to finance risky projects or projects that are not backed by assets (Bates et al., 2009). Firms that are financially unrestrained and have easy access to capital markets hold less cash than other firms. Opler et al. (1999) find little evidence that large cash reserves impact capital expenditures, while R&D investment has increased according to Bates et al. (2009). Firms with firm‐specific assets may have trouble getting a loan, as the collateral is difficult to value. This is especially the case in intangible intensive firms, as they have lower assets tangibility that can act as collateral for debt (Opler et al., 1999; Bates et al., 2009). The precautionary motive is significant, while the agency motive is not significant (Opler et al., 1999; Bates et al., 2009). Both papers conclude that growth firms, small firms and risky firms hold more cash than other firms. Firms with good access to capital markets hold less cash. Another motive for excess cash holdings is the transaction motive. Bates et al. (2009) show that holding liquid assets saves the firm transaction costs to raise cash. An IPO gives a firm access to the public market, but this has significant transaction costs (Opler et al., 1999). Opler et al. (1999) argue that the pecking order theory may dictate the amount of cash held. When a firm accumulates more cash, leverage falls. Equity is last in line, because adverse selection costs make equity a costly financing mechanism. Finally, the tax motive for excess cash holdings describes that multinational firms hold foreign earnings to avoid tax expenses. When a firm repatriates these foreign earnings, the tax consequences are very significant (Bates et al., 2009).
2.3. Investment and ownership One of the first papers to link investment and cash holdings is Fazzari et al. (1988). They test if there is an effect of Tobin’s q and cash flow on investment. Tobin’s q is found to be significant for firms for mature firms, but not for growth firms. Cash flow is significant for both growth and mature firms. Cash holdings are also found to be significant, especially for growth firms. Kaplan and Zingales (1997) find that investment‐cash flow sensitivity is negatively correlated to financial constraints, while Fazzari et al. (1988) find a positive correlation. Cash holdings and Tobin’s q are both positively correlated to investment. For a sample of manufacturing firms, Almeida and Campello (2007) find that investment‐cash flow sensitivities increase when asset tangibility increases for financially constrained firms. For HINT firms. the results of intangible R&D investment are not clearly observable in the cash flows (Thakor & Lo, 2015). Doettling et al. (2016) expands these results by finding that HINT firms require less upfront capital investment and thus invest less of their yearly cash flows. There are also several papers in the literature that look at both founder‐CEOs and their investment decisions. Fahlenbrach (2009) finds that eleven percent of large public US firms have a founder‐CEO. He distinguishes that founder‐CEOs view the company as their life’s work. Intrinsic motivation is more important in this case than extrinsic motivation. The founder‐CEOs focus more on the long‐term which coincides with shareholder value maximization. Skills that are specific to the firm also often present, which also influences the decision‐making power in a firm. Gao and Jain (2011) also find that power of founder‐CEOs is usually significant in a firm they founded. On the other hand, founder‐CEOs may be entrenched and lack experience in leading a firm. Founder‐CEO firms invest differently, perform differently on the stock market, invest more in R&D and make better acquisitions (Fahlenbrach, 2009). Adams et al. (2009) confirm this by finding improved operating performance and higher stock market valuations of a firm. Gao and Jain (2011) do not find significant long‐run stock performance increases in general, but they do prove that founder‐ CEOs improve post‐IPO performance in high‐tech firms. Adams et al. (2009) also find that founder‐CEOs are not always entrenched. Gao and Jain (2011) mention that a founder‐CEO potentially decreases agency problems.
2.4. Intangible capital The literature on intangible capital financing is rather new. Most of the papers on this topic are from the last five years, coinciding with the technical revolution described in Corrado and Hulten (2010). Peter and Taylor (2017) test how physical investment theory holds for intangible investment. To estimate intangible capital, Peter and Taylor (2017) use a combination of knowledge and organization capital (Corrado & Hulten, 2010; Falato et al., 2013). Employee training (included in organization capital) increases human capital, as this training increases possible future values received. A percentage of SG&A costs are included to account for this (Peters and Taylor, 2017). Falato et al. (2013) use twenty percent of SG&A, while Sun and Zhang (2015) and Peters and Taylor (2017) use thirty percent. Doettling et al. (2016) expand these results by finding that capital investment by key employees is an important driver of intangible capital. Sun and Zhang (2015) also support this view, as they find that human capital is the main financing mechanism of intangible investment. Employee financing has a larger correlation with intangible investment compared to physical investment. Also, Sun and Zhang (2015) find that there is a significant intangible capital overhang effect. This means that firms with more investment in employees use less debt financing. A proxy of Tobin’s q is the market‐to‐book‐assets ratio. Peter and Taylor (2017) find that this is a worse proxy than standard and total q. Eventually, the paper finds that normal Tobin’s q is just as relevant for intangible capital as it is for physical capital. The authors find that Tobin’s q explains both physical and intangible capital equally. Intangible capital is found to adjust slower to investment opportunities than physical capital. Falato et al. (2013) show that the increase in intangible capital is a driver of corporate cash holdings. They find the shift towards intangible capital has decreased the debt capacity of firms, as only tangible capital can be used as collateral for debt financing. Falato et al. (2013) say that the decrease in debt capacity leads to firms holding more cash to stay financially flexible. Higher cash holdings are mostly held to counter adverse shocks and to anticipate investment opportunities (Falato et al., 2013). Investment and firm growth are found to be positively associated with cash; this is even more significant for firms with more intangible capital. Falato et al. (2013) report an increase of the intangible ratio from 5% in 1970 to around 60% in 2010, coinciding with the increase in cash holdings and a decrease of the net leverage ratio over this period. All in all, Falato et al. (2013) prove that
technological change has changed corporate liquidity management. 2.5. Contribution The literature on the topics discussed above is very extensive. Although some papers mention all these topics, the combination of topics is new in the literature. This thesis can add to the literature by providing evidence that HINT and LINT firms differ significantly, and starting a discussion on what kind of consequences this may have for IPOs, investment and ownership. Beforehand, several contributions can already be identified that expand the literature. Intangible firms may raise a different amount of IPO proceeds compared to traditional firms. The fraction of management shares may also be larger in HINT IPO firms. The statements mentioned above may provide insight in how HINT firms differ from more traditional firms at IPO, which is interesting because of the change in firm characteristics. Also, if HINT firms raise less money in an IPO, this result enables future study regarding the relation between share ownership, IPO proceeds and the increasing influence of intangibles.
3.
Methodology
In the following section the hypotheses are introduced and explained. Section 3.2. provides the accompanying regressions to give a clear view of the dependent and independent variables. In section 3.3. the data collection process is explained, and in section 3.4. the dependent and independent variables are further expanded. The main variables of interest are intangible investment, fraction of shares floated to the public, shares held by management and IPO proceeds. Control variables are added depending on the regression and hypothesis used. 3.1. Hypotheses Before an IPO, firms decide how many shares they want to float as public shares. Other shares are still held by the company, initial investors or management. The difference between these share distributions is again tested for the HINT and LINT firms. The expectation is that HINT firms float a lower fraction of shares to the public. Brau and Fawcett (2006) motivate that HINT firms float shares mostly because of future acquisitions or to set a firm price. On the other hand, floating shares also increases the probability of being acquired, as a public market is created to be acquired. This could be a reason for HINT firms to float less shares compared to LINT firms. Another reason that the amount of public shares floated is lower, is the amount of intangible investment of a firm. If a firm has more R&D or other intangible assets, there may be less demand for public equity, because of the uncertain nature of these assets (Falato et al., 2014). The amount of debt and the availability of it may also influence the shares floated. Less financing may also be needed, because of the intangible capital overhang effect (Sun & Zhang, 2015). Firms hold more cash, decreasing the need for equity (Thakor & Lo, 2015). On the other hand, more shares may be floated because of the demands of debtholders (Thakor & Lo, 2015). These debtholders may want early repayment of their loan, leading to the firm needing more cash. More shares may be floated to finance the debt repayment. This effect is not expected to dominate, because the HINT firms have low outstanding debt (see Table 1). Also, management may hold more shares leading to a lower amount of shares available to the public. HINT firms more often have a founder as CEO compared to LINTfirms (see Table 2) The effect of this status may influence the lower shares floated to the public.
Hypothesis 1. Intangible firms float less public shares with an IPO than traditional firms.
H0: Shares FloatedINT = Shares FloatedTRAD H1: Shares FloatedINT < Shares FloatedTRAD IPO proceeds are also analyzed to gain insight in the effect of high intangibles in a firm. These IPO proceeds consists out of the number of shares floated and the share float price. Following from the previous hypothesis, the IPO proceeds of HINT firms are expected to be lower than for LINT firms. Lower IPO proceeds imply that less money is raised from the IPO. This may be caused by a lower IPO share price in combination with a lower fraction of shares floated. Agency problems may also lower IPO proceeds. HINT firms hold larger amounts of cash than more traditional firms. Going public may increase this even more, leading to even more excess cash for HINT firms. HINT firms more often have deeply invested management or even a founder‐CEO leading the firm (see Table 2). Agency problems may arise with these large amounts of cash, especially because management in HINT firms usually have more control over the firm. On the other hand, because management is more invested in to the firm, badly used cash holdings may lower firm value leading to a value decrease for management. Agency problems or the presence of a founder‐CEO is not expected to lower the IPO proceeds significantly. Investors may also doubt the uncertain nature of R&D done by HINT firms. The cash flows gained from R&D are more uncertain than the cash flows from capital expenditures (Kaplan & Zingales, 1997). Investors may be more hesitant to invest in the riskier investment firms, leading to less shares or a lower price for a HINT IPO. This can also be linked to the precautionary motive of cash holdings. Because of the uncertain R&D, more cash is held to finance potential investment opportunities. Total q shows that the HINT firms have more investment opportunities available than traditional firms (see Table 1). Investors may doubt the quality and rate of return of these investments and invest less in the IPO, thus lowering the IPO proceeds of a firm. Just as with first hypothesis, firm‐specific accounting variables may influence IPO proceeds. The price factor of IPO proceeds can also be influenced by these factors, but may
also be dependent on how the overall market performs. Taking in to consideration the points above, IPO proceeds are expected to be lower for HINT firms.
Hypothesis 2. Intangible firms raise less capital from IPOs than traditional firms.
H0: IPO ProceedsINT = IPO ProceedsTRAD H1: IPO ProceedsINT < IPO ProceedsTRAD Total shares held by management are compared between the HINT and LINT samples. Management is seen as the three most prominent executive managers of a firm. In most cases this are the CEO, CFO and another C‐level executive. An explanation might be that the presence of high intangibles requires more control of management to be successful. Firms with high R&D expenditures often require specific knowledge from management to be successful. Also, HINT firms have a founder in the management more often than LINT firms (Table 2). Founders may hold more shares, because they are more attached to the firm they founded. (Fahlenbrach, 2009; Gao & Jain, 2011). An argument against the hypothesis may be that management uses an IPO to exit the firm. This effect is not expected as lockup periods for insiders exist. Following from the expectations of the first two hypotheses and evidence from Michel et al. (2014), a lower public float leads to higher incentives for executive management to perform. The lower public float also increases post‐IPO returns for management (Michel et al., 2014). Because of the increased risk of intangibles, this lower public float increases the reward for management. Management shares are therefore expected to be higher in HINT firms. Taking in to account these expectations, the following hypothesis is defined: Hypothesis 3. Management of intangible firms hold more shares after IPO than in traditional firms.
H0: Management SharesINT = Management SharesTRAD H1: Management SharesINT > Management SharesTRAD
All in all, the hypotheses are related to each other and the main research question. Combining the expectations from the hypotheses gives a prediction for the research questions. The expectation is that less shares are floated to the public due to the nature of
intangible investment. This leads to less proceeds raised in the IPO. The amount of management shares is then expected to be higher because there are more shares available to management and incentives are better for management. Overall, these effects are expected to be caused by the presence of high intangibles. 3.2. Method The baseline regressions use Ordinary Least Squares (OLS) to estimate the coefficients. Multiple regressions with different dependent and control variables are used to test hypotheses: (1) ε (2) ε (3) ε Specification (1) is the regression model for the first hypothesis, specification (2) for the second hypothesis and specification (3) for the third hypothesis. These regressions test the effect of the main explanatory variable Intangible Investment on the dependent variables. The dependent and main explanatory variable are explained further in section 3.4.. Control variables Xi in these regressions are given in Table 1, and further defined in section 3.4.. The robustness results apply the measure suggested by Sun and Zhang (2015) in two different forms: R&D expenses scaled by total investment and R&D expenses scaled by total assets. (4) & ε (5) & ε The regression models (4) and (5) differ from earlier models by excluding SG&A expenses in the main explanatory variable. Furthermore, the complete sample is separated in to separate HINT and LINT samples. The same OLS regressions used in the main results
(regression models (1), (2) and (3)) are applied to the separate samples to provide more evidence for the main results and support for the control variables. Endogeneity issues that may occur are reverse causality and omitted variable bias. Due to the inherent characteristics of OLS, causality is inferred from theory. Panel data cannot be applied in this thesis, as the data does not change over time (Adams et al., 2009). Other empirical methods are also difficult to use, because of how the data is structured. The data looks at each firm at a single point in time, making OLS or IV regressions the most effective methods. Omitted variable bias is a potential problem, but firm characteristics, dummy variables and macro‐economic conditions are added to avoid obvious biases. Control variables that have been proven in literature to explain the dependent variables are added to the regressions. A potential problem with regression model (3) is reverse causality. The relation could also be reversed, as managers with more control over a firm can choose the amount of intangibles. Following from the third hypothesis, this reverse relation is not expected. 3.3. Data collection To investigate the hypotheses, data is collected from Compustat, Zephyr and SEC S‐1 filing forms. All data is then divided in to two groups, the high intangible (HINT) and the low intangible (LINT) firms sample. From the S‐1 IPO filing forms, data can be found on company share ownership structures. In the US, it is mandatory to file an S‐1 form when planning to go public. This form contains much information about the company, including a detailed financial overview and ownership structure. An analysis is done using a control group of LINT firms. These LINT firms are regarded as more traditional firms that have lower intangible investment and that use more capital expenditures relative to HINT firms. HINT firms contain more intangibles and do more R&D than LINT firms. The sample firms in the HINT and LINT sample are then compared to each other using various independent, dependent and control variables. To collect the sample firms, 2‐ and 3‐digit SIC codes are analyzed by hand to find industries that are known to have high intangibles (Kile & Phillips, 2009). Nearly all firms in the HINT sample are in the tech or pharmaceutical industry. The LINT sample contains several different industries that are regarded as industries with lower intangibles. The sample firms are collected from the Zephyr database by using the following criteria:
1. Sample firm does an IPO 2. Sample firm is listed in the US 3. Time period from 2007 until and including 2016 4. For high intangibles the 3‐digit SIC codes used are: 283, 366, 737, 873. For low intangibles the 2‐ and 3‐digit SIC codes used are: 01, 02, 08, 10, 12, 13, 14, 15, 16, 17, 20, 22, 281, 285, 29, 30, 32, 33, 40. For each sample, the 100 largest relevant public offerings are selected. The SEC EDGAR database is used to collect share information from S‐1 filing reports at IPO, and the COMPUSTAT database is used to collect accounting data of the sample firms. Furthermore, only IPOs listed on major US stock exchanges are considered. This are mostly large deals, which are comparable to each other in size and listing rules. In addition, the EDGAR database only contains companies that have to publish the S‐1 report. Firms that issue equity on OTC‐markets do not have this obligation. To check if data is available for the firms in COMPUSTAT, the GVKEY identifiers are put in to the database. The IPOs found in Zephyr give ISIN keys. These keys are then subsequently converted to GVKEYs. If identifiers are missing, these are added manually by checking the CIK identifier provided by the SEC EDGAR database. Furthermore, spin‐offs are excluded from the sample, as they do not report shares held by management in their S‐1 reports. All demergers, foreign stock exchange listings and cancelled IPOs are also excluded. Demergers do not have to file S‐1 reports and therefore miss sufficient data. Finally, the company business descriptions and financials are checked by hand to identify that the firms are not misclassified. 3.4. Variables Most variables are collected from Compustat and the remaining variables are created using hand‐collected data. Variables regarding firm characteristics are normalized by total assets of the specific firm to make the firms comparable. Descriptive statistics are given in Table 1. The dependent variables are Shares Floated, IPO Proceeds and Management Shares. These variables are constructed with hand‐collected data. Shares Floated is the number of shares offered in an IPO to the public (Class A) divided by the total outstanding shares. The amount of shares offered are published in the S‐1 IPO filing reports. IPO Proceeds is the total
proceeds of the shares offered in the IPO. It is calculated by multiplying the shares offered in the IPO by the offer price. This variable uses the same shares offered number as Shares Floated and uses the offer price provided by the Zephyr database. Management Shares is the post‐IPO shares held by the three most important executives of a firm divided by the total shares outstanding. Usually, this are the CEO, CFO and most of the time the CTO or COO. The main explanatory variable is Intangible Investment. By first estimating intangible capital this variable is created. To measure the amount of intangible capital for publicly traded firms is difficult, as these firms do not report expenses consistently as a single item. An intangible capital variable is therefore estimated using R&D expenses plus 30% of SG&A expenses commonly used in the literature (Peters & Taylor, 2017; Sun & Zhang, 2015). This intangible capital variable is then scaled by intangible capital plus capital expenditures, creating a universal measure comparable between companies and industries. An alternative definition is proposed by Sun and Zhang (2015). They propose to use the accounting line R&D expenses excluding SG&A expenses. According to them, this data is more widely available and because R&D expenses already incorporates human capital investment and innovation activities. This alternative main variable is only used in further robustness tests. The control variables include firm characteristics and ownership information. Variables created by hand are explained in further detail below. The Current Ratio is used to control for firm liquidity. The quick ratio is not used, as the databases miss values and do not provide enough data to calculate it manually. IPO proceeds are calculated by multiplying the number of shares floated in the IPO with the share offer price. The Founder‐CEO dummy is only equal to one if the CEO at the date of IPO is a (co‐)founder. Directors in non‐ management positions are not included in the dataset. The Class B dummy equals one if a dual‐class share structure is present. Class B stock are always considered as stock with superior voting rights. Sometimes companies issue Class A stock with superior voting rights to Class B stock, but this is corrected in the dataset. If there is no dual class stock structure, common stock is reported as Class A stock. Subordinated shares in the LINT firm sample are considered as Class B stock. These subordinated shares have the same voting rights as common stock, but are secondary to the common stock in regard to dividend payout and usually bring some control rights with them. Convertible preferred stock is converted to common stock at IPO, and therefore does not need to be corrected. Tobin’s q is estimated
using the Peters and Taylor (2017) methodology named total q. Total q is measured by the firm market value divided by the replacement cost of physical plus intangible capital. The total q variable is available from WRDS. Total q is winsorized at the 1st and 99th percentile to remove the large negative and positive outlier. Management Shares is also winsorized due to an extreme outlier. Current ratio is not winsorized even though the standard deviation is large. This variable has no large outlier and a lot of variance in the data, therefore winsorizing will change the results too much.
4.
Results
In the results section the hypotheses will be tested with multiple regressions. The coefficients are then explained, and are given economic significance. Later on, section 4.2 provides several robustness tests that show the consistency of the results. Finally, in section 4.3 other literature will be compared with the results found in this thesis. 4.1. Regressions Table 3. Regressions estimating Shares Floated The regressions include both HINT and LINT samples from 2007 to 2016. Missing explanatory variables decrease the total sample from 200 to 176 firm observations. Variable data is taken from end‐fiscal year after the IPO, with Man. Shares, Class B Dummy and Founder‐CEO Dummy collected at IPO. Standard errors are reported in parentheses with levels of significance indicated by: *** p<0.01, ** p<0.05, * p<0.1. Variable definitions are provided in the Appendix. (1) (2) (3) (4) (5)Variables Shares Floated Shares Floated Shares Floated Shares Floated Shares Floated
Intangible Investment ‐0.118*** ‐0.110*** ‐0.141*** ‐0.138*** ‐0.142*** (0.029) (0.032) (0.032) (0.038) (0.038) Total Q ‐0.004*** ‐0.006*** ‐0.006*** ‐0.006*** (0.001) (0.002) (0.002) (0.002) Current Ratio 0.004*** 0.004*** 0.004** (0.001) (0.001) (0.001) Revenue/Assets 0.070*** 0.070*** 0.071*** (0.013) (0.014) (0.014) Leverage/Assets ‐0.034 ‐0.037 ‐0.030 (0.041) (0.046) (0.048) Cash/Assets ‐0.008 0.011 (0.056) (0.058) Man. Shares 0.063 (0.072) Class B Dummy ‐0.036 (0.025) Founder‐CEO Dummy ‐0.000 (0.022) Dividend Dummy 0.014 (0.025) Constant 0.312*** 0.329*** 0.294*** 0.295*** 0.287*** (0.020) (0.021) (0.025) (0.027) (0.032) Observations 200 183 176 176 176 R‐squared 0.078 0.136 0.279 0.279 0.294 In Table 3 the first hypothesis is tested by looking at the effect of Intangible Investment on Shares Floated. Intangible Investment is expected to negatively affect the amount of Shares
Floated. The univariate regression in column (1) seems to show evidence supporting the hypothesis. When Intangible Investment increases by one, the ratio of shares floated decreases by a very significant 0.1179. This implies that a one standard deviation increase in Intangible Investment leads to a 0.003 decrease in Shares Floated, which is 1.4% of the mean value of Shares Floated. Compared to column (1), Intangible Investment decreases Shares Floated by a slightly larger 2.2% of mean value in column (5). In all specifications Intangible Investment is consistently significant with a stable coefficient. When adding more control variables, the coefficient remains consistent with nearly the same effect size. In column (3), more firm specific accounting variables are added. Total Q, Current Ratio and Revenue/Assets are highly significant at a 1% level, but only Revenue/Assets has economic significance. Leverage and Cash/Assets have no significant effect on Shares Floated. Cash/Assets is against expectations not significant. Column (5) adds several dummies and the Post‐IPO Management Shares (hereafter Management Shares) to the regression. The dummy variables are not statistically significant.
Table 4. Regressions estimating IPO Proceeds The regressions include both HINT and LINT samples from 2007 to 2016. The dependent variable IPO Proceeds is given as the log of IPO Proceeds. Missing explanatory variables decrease the total sample from 200 to 176 firm observations. Variable data is taken from end‐fiscal year after the IPO, with Man. Shares, Class B Dummy and Founder‐CEO Dummy collected at IPO. Standard errors are reported in parentheses with levels of significance indicated by: *** p<0.01, ** p<0.05, * p<0.1. Variable definitions are provided in the Appendix. (1) (2) (3) (4) (5)
Variables IPO Proceeds IPO Proceeds IPO Proceeds IPO Proceeds IPO Proceeds
Intangible Investment ‐0.855*** ‐0.949*** ‐0.814*** ‐0.793*** ‐0.768*** (0.158) (0.177) (0.185) (0.184) (0.181) Total Q 0.010 0.024** 0.022** 0.017* (0.008) (0.011) (0.011) (0.010) Current Ratio ‐0.008 ‐0.008 ‐0.006 (0.008) (0.008) (0.008) Revenue/Assets ‐0.046 ‐0.048 ‐0.077 (0.076) (0.076) (0.074) Leverage/Assets 0.798*** 0.840*** 0.813*** (0.236) (0.237) (0.238) Man. Shares 0.498 0.467 (0.383) (0.389) Market Returns 0.626 0.604 (0.453) (0.440) Class B Dummy 0.576*** (0.136) Founder‐CEO Dummy ‐0.045 (0.121) Dividend Dummy ‐0.017 (0.135) Constant 19.86*** 19.86*** 19.59*** 19.437*** 19.380*** (0.109) (0.116) (0.146) (0.167) (0.176) Observations 200 183 176 176 176 R‐squared 0.129 0.140 0.198 0.213 0.291 Table 4 looks at the second hypothesis by testing the effect of Intangible Investment on IPO Proceeds. The log of IPO Proceeds is used to improve coefficient interpretation. It is expected that Intangible Investment negatively affects IPO Proceeds. A univariate regression confirms this expectation with a highly negative significant coefficient. When Intangible Investment increases by one‐standard deviation, this leads to a 0.1351 decrease in ln(IPO Proceeds). For the median firm this indicates a decrease in IPO proceeds of $25.44 million, which is 12.64% decrease from the median value of $201.32 million. When adding more control variables, the effect of Intangible Investment stays consistently significant, suggesting that HINT firms raise less money in an IPO.
When adding more accounting variables in column (3), Total Q becomes positively significant at a 5% level. Against expectations Leverage is positively significant with a relatively large coefficient. A one‐standard deviation increase of Leverage leads to a $42.98 million increase in IPO Proceeds. Column (5) includes dummy variables to the regression. This does not change the consistency and significance of the main explanatory variable, but the effect of Total Q is not significant anymore. IPO proceeds are significantly higher when a dual‐share structure exists. When the Class B dummy increases by one‐standard deviation, IPO proceeds increase by $16.41 million. The significant Leverage supports that LINT firms have more IPO proceeds than HINT firms. Opposing this result is the Class B dummy, as this coefficient suggests that HINT firms raise more money in an IPO. Table 5. Regressions estimating Management Shares The regressions include both HINT and LINT samples from 2007 to 2016. Missing explanatory variables decrease the total sample from 200 to 176 firm observations. Variable data is taken from end‐fiscal year after the IPO, with Man. Shares, Class B Dummy and Founder‐CEO Dummy collected at IPO. Standard errors are reported in parentheses with levels of significance indicated by: *** p<0.01, ** p<0.05, * p<0.1. Variable definitions are provided in the Appendix. (1) (2) (3) (4) (5)
Variables Man. Shares Man. Shares Man. Shares Man. Shares Man. Shares
Intangible Investment 0.013 ‐0.018 0.012 ‐0.002 ‐0.026 (0.032) (0.037) (0.031) (0.033) (0.036) Total Q 0.004* 0.003* 0.004** (0.002) (0.002) (0.002) Current Ratio 0.000 0.001 (0.002) (0.002) Revenue/Assets 0.002 0.010 (0.015) (0.015) Leverage/Assets ‐0.070 ‐0.026 (0.047) (0.047) Class B Dummy 0.032 0.008 (0.026) (0.027) Founder‐CEO Dummy 0.089*** 0.095*** 0.100*** (0.021) (0.021) (0.023) Dividend Dummy 0.034 0.032 0.035 (0.024) (0.025) (0.026) Constant 0.103*** 0.118*** 0.048* 0.046* 0.046 (0.022) (0.029) (0.026) (0.025) (0.032) Observations 200 176 200 183 176 R‐squared 0.001 0.048 0.098 0.129 0.153
Table 5 investigates the third hypothesis by testing the effect of Intangible Investment on Management Shares. Intangible Investment is expected to be positive. From the results of column (1), a negative insignificant coefficient is found. In the other columns, Intangible Investment is also not positive or significant when tested on Management Shares, even when estimated with an unreported regression using HINT Dummy in place of Intangible Investment. When management contains a founder‐CEO, more shares are held by management. In column (5) when the Founder‐CEO Dummy equals one, the fraction of Management Shares increases by 0.1. This effect is logical as a founder usually receives shares when a company is founded. An increase of Total Q by one, increases Management Shares by 0.4% at the 5% level. The results suggest that Management Shares is not statistically influenced by Intangible Investment. Management of HINT firms do not seem to hold significantly more shares than management of LINT firms when using Intangible Investment as the main explanatory variable. 4.2. Robustness In the following section several additional regressions are run to confirm the results found in the previous section. The main explanatory variable Intangible Investment is replaced by R&D Expenses and R&D Investment to check for consistent results. The total sample is also split in to two samples: the HINT and LINT sample. Regressions specified in section 4.1. are used again to verify the results. The findings in the following section are not cross‐sample, so the results found cannot be applied to the full sample.
4.2.1. Alternative explanatory variable Table 6. Regressions with an alternative main explanatory variable The regressions include both HINT and LINT samples from 2007 to 2016. Missing explanatory variables decrease the total sample from 200 to 176 firm observations. Column (1) and (2) use R&D Investment, and column (3) and (4) use R&D Expenses. Variable data is taken from end‐fiscal year after the IPO, with Man. Shares, Class B Dummy and Founder‐CEO Dummy collected at IPO. Standard errors are reported in parentheses with levels of significance indicated by: *** p<0.01, ** p<0.05, * p<0.1. Variable definitions are provided in the Appendix. Table 6 replaces the main explanatory Intangible Investment that was used in earlier regressions with R&D Investment and R&D Expenses. R&D Investment leaves out the 30 percent SG&A factor in the nominator compared to Intangible Investment. R&D Expenses is the main measure used in Sun and Zhang (2015), which is defined as R&D expenses divided (1) (2) (3) (4)
Variables Shares Floated IPO Proceeds IPO Proceeds Man. Shares
R&D Investment ‐0.073** ‐0.635*** (0.033) (0.176) R&D Expenses/Assets ‐1.754** ‐0.227* (0.697) (0.132) Total Q ‐0.006*** 0.016 0.009 0.004** (0.002) (0.010) (0.010) (0.002) Current Ratio 0.003** ‐0.006 ‐0.009 0.001 (0.001) (0.008) (0.008) (0.001) Revenue/Assets 0.056*** ‐0.173** ‐0.155** 0.006 (0.014) (0.075) (0.076) (0.015) Leverage/Assets ‐0.043 0.681*** 0.749*** ‐0.042 (0.046) (0.248) (0.252) (0.048) Man. Shares 0.072 0.513 0.427 (0.073) (0.394) (0.405) Class B Dummy ‐0.032 0.594*** 0.602*** 0.009 (0.026) (0.138) (0.141) (0.027) Founder‐CEO Dummy ‐0.001 ‐0.025 ‐0.055 0.102*** (0.023) (0.124) (0.126) (0.023) Dividend Dummy 0.017 ‐0.045 0.013 0.030 (0.026) (0.139) (0.140) (0.026) Market Returns 0.621 0.653 (0.446) (0.455) Constant 0.249*** 19.247*** 19.144*** 0.053* (0.029) (0.168) (0.166) (0.029) Observations 176 176 176 176 R‐squared 0.241 0.271 0.242 0.165
by total assets. In other literature, 30% of SG&A expenses are included in intangible investment. SG&A expenses contains investments in human capital and organization capital, which is also part of intangible investment. A reason for leaving out SG&A expenses is that this measure contains a lot of noise, because the intangible part of it hard to estimate. In column (1) the first hypothesis is tested with R&D Investment as the main explanatory variable. Statistical significance is found in this regression for R&D Investment, but with a smaller less significant coefficient. Compared to Table 3, two differences are the excluded SG&A factor and Cash/Assets is left out of the regression due to high correlation with R&D Investment. In column (2) and (3) the second hypothesis is tested with an alternative main explanatory variable. Column (2) uses R&D Investment, and here the result is a similar coefficient compared to the results found in Table 4. The coefficient is slightly smaller, but this does not have economical consequences. R&D Expenses is used in column (3) and again finds a negative significant coefficient. Lastly, column (4) is an alternative regression for the third hypothesis. R&D Expenses is found to significantly influence Management Shares negatively at the 10% level. When R&D Expenses increases by one standard deviation, this leads to a decrease of 0.03 in Management Shares, which is a decrease in mean value of 27%. When considering above results, the alternative main explanatory variables find similar results for IPO Proceeds and Shares Floated, but different results for Management Shares. Management Shares is statistically affected by R&D Expenses, while it was not by Intangible Investment in the main results. A potential problem with the alternative variables is that the values for the LINT sample are often very low. The median of R&D Expenses for the LINT sample is equal to zero. This fact may skew the results towards the HINT sample, while Intangible Investment avoids this problem. 4.2.2. In‐sample regressions The separate HINT and LINT regressions provide insight in the direction of the control variables. The effect of some control variables may differ for the HINT and LINT samples, leading to the question which sample drives an effect. The regression tables of this section are provided in Appendix 1. Table 3 is recreated with separate LINT and HINT samples in Table 7. The statistical significance of Total Q and Current Ratio are driven by the HINT sample, while
Revenue/Assets and Cash/Assets are driven by the LINT sample. Leverage is significant in both samples. LINT firms float significantly less shares when cash holdings are high. Again, Table 8 uses the same specification as the IPO proceeds table in the results section. Leverage is significant for LINT firms at a 5% level and for HINT firms at a 1% level. For both samples increased leverage leads to higher IPO proceeds. The significance of the Class B Dummy is driven by the HINT sample. Dividends seem to decrease IPO proceeds for HINT firms, but this is not found in Table 4. Compared to Table 5, not much changes in the robustness results in Table 9. Only the Founder‐CEO Dummy is significant in both samples. The effect size of the Founder‐CEO Dummy is larger in the LINT sample. This may be due to firms in the HINT sample having a founder in top management, but not with a large amount of shares. The results could also be driven by a skewed sample of founders in LINT firms. All in all, findings found above appear to be in line with evidence found in the main results. Some significant variables are driven by a specific sample as expected beforehand. 4.3. Discussion The following section will look at how the results influence the hypotheses and how existing literature compares to these results. Because this thesis looks at a relatively new topic in the literature, a direct comparison between results is difficult. Therefore, variables and conclusions will be used to make a comparison. 4.3.1. Interpretation results The results in Table 3 provide evidence that HINT firms float less shares than LINT firms. Even after adding multiple control variables, the effect of Intangible Investment stays significant and consistent. Total Q is statistically significant, but only holds economically for a few firms with very large total q values. Current Ratio also economically insignificant effect. The significance of Revenue/Assets means that firms float more shares to the public when revenue is large relative to assets. The negative coefficient of the Class B Dummy suggests that investors dislike dual‐class share structures. Cash held and the amount of leverage do not seem to influence the fraction of shares floated. Considering the arguments above, the null hypothesis of Hypothesis 1 is rejected. Intangible firms float less public shares with an IPO than traditional firms.
Intangible Investment is statistically and economically very significant in Table 4, suggesting that HINT firms raise less IPO proceeds in an IPO than LINT firms. The large Leverage/Assets coefficient is attributable to both HINT and LINT firms, following from the robustness results. Positive Leverage/Assets indicates that the market appreciates it by increasing IPO proceeds significantly, due to its disciplining effect on management. The significant Class B Dummy is in slightly in conflict with evidence found in Table 3, where the presence of a dual‐class structure suggestively lowered the number of shares floated. The descriptive statistics in Table 1 show that the dual‐class share structure is evenly distributed over HINT and LINT firms. Agency problems are not the cause of the lower IPO proceeds following from the results. Market Returns suggests that IPO proceeds increase when the market performs well. Following the evidence above, the null hypothesis of Hypothesis 2 is rejected. Intangible firms raise less capital from IPOs than traditional firms. Evidence supporting the third hypothesis is more contradictive in Table 5. Intangible Investment is not statistically significant in any regression. The Founder Dummy does suggest that management holds more shares, but this effect is for both the HINT and LINT sample. Total Q suggests the same, but again this is for the combined sample. When considering the robustness regressions with alternative main explanatory variables of Table 6, the variable R&D Expenses actually decreases Management Shares at the 10% significance level. Considering the findings above, the intangible measures do not provide enough evidence to reject the null hypothesis of Hypothesis 3. There is not enough evidence to say that management of intangible firms hold more shares after IPO than in traditional firms. The non‐rejection of Hypothesis 3 can be caused by several reasons. The difference in management shareholdings may not be large enough to find a significant effect of intangibles. Management Shares are similar to each other for HINT and LINT firms, although there are more founder‐CEOs in HINT firms. Because is there slight evidence when using R&D Expenses as the main explanatory variable, an instrumental variable regression may decrease the error of estimation. Uncertainty of intangibles and financing difficulties may be an important reason why management does not hold significantly more shares. Finally, reverse causality may have caused the insignificant result. Although unexpected, more Management Shares may cause the choice of Intangible Investment.
4.3.2. Comparison results with literature In the literature it is well established that intangible capital is a driver of increasing firm cash holdings (Falato et al., 2013). Bates et al. (2009) conclude from their results that firm characteristics have changed over time and coinciding with this is the change from capital expenditures to R&D. More cash holdings are found to be required because of the greater R&D intensity. Thakor and Lo (2015) confirm that cash holdings are higher than in the past, and that net debt is therefore lower. The different motives for holding excess cash do not hold in this dataset. Compared with the results found in this thesis, cash holdings for HINT firms are indeed significantly larger than in LINT firms. Coinciding with this is also a lower amount of debt for HINT firms. While these variables increase Intangible Investment, there does not seem to be a significant relation between cash holdings and Shares Floated or IPO Proceeds. Alavi et al. (2008) find that managerial ownership is related to shares offered at IPO. They use a dependent variable very similar to the Shares Floated variable in this thesis. In their paper, Shares Floated decreases significantly when ownership by insiders is larger. They also square this insider variable, and find that this is significant and positive. The negative effect is found to dominate under 50% management share ownership, which thus holds for this thesis as the mean of management shares is 11.07%. Alavi et al. (2008) investigate a different market with a different definition of management ownership, which means it cannot be directly compared to the results of this thesis. Their finding of lower Shares Floated because of higher Management Shares does not hold in the main results, but there is evidence provided when R&D Expenses is used. R&D intensive firms more often have a founder leading a firm (Fahlenbrach, 2009). This is also confirmed by the sample in this thesis. Michel et al. (2014) find that the incentives for firm insiders to perform are lower when the public float of shares is larger. When applied to this thesis, Class B shares prevent the dilution of insider shares. The Class B Dummy is found to be significant and positive for IPO Proceeds, meaning that public investors value incentives for insiders to perform. All in all, there have not been a lot of papers published on the exact topic of this thesis. The characteristics of R&D intensive firms found in other literature also hold for the HINT dataset. The influence of intangible investment on IPO proceeds is not commonly tested in the literature. The results of Alavi et al. (2008) regarding Management Shares are partly supported when R&D Expenses is used in place of Intangible Investment.
