Intangible Assets, Investment policy, and Capital Structure

Transcription

1 Intangible Assets, Investment policy, and Capital Structure Steve C. Lim, Texas Christian University Antonio Macias, Texas Christian University Thomas Moeller, Texas Christian University June 12, 2013 Abstract: Using unique data of U.S. public targets that were acquired by U.S. public acquirers, we find that priced intangible assets have a positive and economically meaningful relation with leverage. Firms with higher financial frictions seem to benefit more from this positive relation. Both technology-related intangible assets and Research-and-Development expenditures seem to increase debt capacity, particularly in technology industries. In contrast, only in non-technology industries both marketing-related intangible assets and advertising expenditures seem to increase debt capacity. Whereas better R&D investment ability increases the value of technology-related intangible assets and debt capacity, better advertising investment ability does not affect either the value of marketing-related intangible assets or debt capacity. Overall, we document significant interactions between intangible assets, investment policy and capital structure. Acknowledgements: We thank Houlihan Lokey for providing us with M&A purchase price allocation data.

2 I. INTRODUCTION AND LITERATURE REVIEW Theoretical models show that higher asset liquidity increases the optimal debt leverage (e.g. Shleifer and Vishny, 1992, Morellec, 2001). Campello and Giambona (2013) argue that larger easiness to sell tangible assets (i.e., redeployability) also increases debt capacity. Economic intuition suggests that the debt capacity of a firm goes up as long as the firm has saleable assets independent of the physical substance of those assets (i.e., tangible assets or intangible assets). We articulate this intuition by assessing the impact of intangible assets on capital structure beyond the impact of traditional determinants of financial leverage such as firm size, market-to-book, profitability, tangible assets, etc. A major empirical limitation to study the relation between intangible assets and capital structure is that the value and redeployability of intangible assets, because of its nature, is uncertain. Seminal papers aim to account through goodwill for the value of intangible assets, such as trade-names and knowledge generated from research-and-development (R&D) expenditures (e.g., Gabriel and Preinreich (1936), Harold and Avery (1942)). Using regression equations to estimate the value generated by R&D and advertising expenditures, more recent literature finds that equity market seem to underreact to the value produced by such firm s expenditure behavior (e.g., Sougiannis (1994), Chan, Lakonishok and Sougiannis (2001), Kallapur and Kwan (2004), Li (2011)). It is still an open question, however, to what extent capital structure, firm s expenditure behavior (such as R&D and advertising expenditures), and the value of intangible assets relate to each other (see Harris and Raviv (1991), Parsons and Titman (2009)). We study this question by examining the value of intangible assets priced after an acquisition. 1

3 We use a unique data set of 838 U.S. publicly traded target firms that were acquired by publicly traded U.S. firms during the years 2002 through SFAS 141 (2001) requires that the acquiring firm allocate the purchase price to identifiable assets (tangible and intangible) before it is allocated to goodwill. 1 The acquirer classifies target s assets into six categories. They are 1) tangible assets, 2) developed technology (including patents), 3) in-process research and development, 4) trademarks and trade names (including domain names), 5) customer-related assets (including backlog, customer contracts and customer relationships), and 6) other identifiable intangible assets. Prior to SFAS 141, companies used to dump the majority of their purchase price to acquisition goodwill without providing the details on identifiable intangible assets to the investing public. Acquisitions as well as SFAS 141 provide us a rare opportunity to get hold of the fair market value of asset of target firms in arms length transactions because firms (in this case, the target firms) do not capitalize or recognize self-created intangible assets in the balance sheet due to the conservatism tradition in accounting. In this paper we assess whether the components identified in the Purchase Price Allocation (PPA, henceforth) consideration relates to the target firm s leverage and its R&D and advertising investment behavior. We first test whether higher intangible assets, which the acquirer eventually prices, relate to higher leverage. Second, we assess how the two main types of intangible assets, i.e., developed and in-process technology and marketing-related intangible assets, relate to firm s leverage. Third, we examine whether the association between intangible assets and leverage depends on firms having financial constraints or a technology focus. After having used the firm s expenditure behavior as controls in the first three parts of our analysis, the subsequent analysis explicitly examines whether the relation between intangible assets and 1 FASB standards are now incorporated in the FASB s Accounting Standards Codification (ASC) and SFAS 141 Business Combinations (2001) can be found under FASB ASC 805: Business Combinations. However, to be consistent with prior literature, we will use SFAS 141 instead of ASC 805. FASB revised SFAS 141 in 2007 and it is now called SFAS 141R. 2

4 leverage changes as a function of the effectiveness of firm s prior expenditure policies, specifically, a good ability to turn R&D and advertising expenditures into subsequent sales (see Cohen, Diether, and Malloy, 2013). This fourth part of our analysis aims to answer Parsons and Titman (2009) open question for examining whether the interaction between firm s policies and leverage creates value (in our case, intangible assets value eventually priced by an acquirer). The first part of our analysis shows that the PPA consideration provides additional information on the economic determinants of leverage beyond variables previously used in the literature (see Harris and Raviv (1991), Frank and Goyal (2003, 2008), and Parsons and Titman (2009) for surveys on capital structure research). We find that the tangible and intangible components identified by the PPA consideration are positively related to firm leverage. The economic magnitude of the impact is large even after controlling for book values of property, plant and equipment, research and development (R&D) expenditures, and advertising expenditures. For instance, a one-interquantile change in PPA s intangible assets is associated with an increase of 58.7% in leverage. This is equivalent to increase leverage from its sample mean of 17.7% to 28.1%. The results are robust to the inclusion of year and industry fixed effects and to different model specifications, such as OLS, Heckman models (which address potential endogeneity of the selection of the firms in our PPA sample), and seemingly unrelated simultaneous equations (which address potential endogeneity concerns for omitted variables and simultaneity biases). Intangible assets, eventually valued in an acquisition, seem to increase debt capacity. After having established that the association between intangible assets and leverage is economically meaningful, we further document both developed and in-process technology and marketing-related intangible assets impact leverage. The developed technology and in-process technology intangible assets have a positive relation with leverage. Developed technology 3

5 intangible assets, however, have a negative relation with leverage if the firm operates in a nontechnology industry. The contrasting effect between developed technology and leverage is consistent with prior research on equity markets not fully recognizing the value of R&D expenditures (e.g., Cohen, Diether and Malloy (2013), Chan, Lakonishok and Sougiannis (2004)). We also find that both marketing and customer-related intangible assets are also positively related to firm leverage, consistent with Larkin (2013), who documents a link between marketing and financing policies. In the third part of our analysis we document that each type of intangible assets impacts leverage differently as a function of the firm having financial constraints and operating in a technology industry. We first study whether firms that face financing constraints benefit from the increase in debt capacity that intangible assets provide. We use three proxies to identify financially constrained firms, namely, small firms, firms with unrated debt and non-dividend payers (see, Murillo and Giambona, 2013). We find that in-process technology intangible assets are particularly important to financially constrained firms higher leverage. Our findings looking at debt markets complement Li s (2011) result that financially constrained firms seem to drive the positive relation between R&D expenditures and equity market value. In contrast, we find that marketing and customer-related intangible assets are not related to leverage when the firm is financially constrained. Unconstrained firms, however, have a positive association between leverage and both trademarks-and-trade-names and customer-related intangible assets. These contrasting results suggest that developed and in-process technology intangible assets ameliorate credit frictions. Marketing related intangible assets, however, seem to relate to higher leverage for other reasons, such as operating decisions (as Parsons and Titman (2009) suggest). Altogether, our analysis on financial frictions provides further support to Murillo and 4

6 Giambona s (2013) finding that asset redeployability (in our case, target s higher intangible assets value) increases debt capacity. Our analysis on technology industries shows additional contrasting results between developed and in-process technology and marketing-related intangible assets on their relation with leverage. Although our results are robust to industry fixed effects and to potential sample selection bias, our PPA sample comprises a higher proportion of firms in a technology industry compared to the universe of firms in Compustat during the same period (29% vs. 21%). This higher proportion of technology firms can indicate a higher redeployability of the firm s tangible and intangible assets in a technology industry hence also affecting capital structure decision. As expected, firms in a technology industry have higher developed in-process technology intangible assets. Firms in a technology industry, however, have lower marketing intangible assets, mainly related to trademarks and trade names. Multivariate analysis shows that firms in technology firms drive the positive association between leverage and both developed and in-process technology intangible assets. In contrast, firms in non-technology industries have a negative association between leverage and developed technology intangible assets. Marketing related intangible assets also show a contrasting effect on leverage, although the sign of the association reverses. We find a negative association between leverage and both trademarks-and-trade-names and customer-related intangible assets when firms are in technology industries and a positive association in non-technology industries. These contrasting findings as function of the value of trade names partially explain the mixed theoretical implications of having higher debt when theoretical models include competition level (e.g., Brander and Lewis (1986), Maksimovic (1986, 1990), Chevalier and Scharfstein (1996), Dasgupta and Titman (1998), and Parson and Titman (2009)). These findings as function of the value of customer-related intangible assets also support Kale and Shahrur (2007) and 5

7 Maksimovic and Titman (1991) contention that when customer-related investments are more specific (as in the case of technology industries), then the firm has the incentives to have lower leverage. Although we find that firm s expenditure behavior affects the value of intangible assets, the fourth part of our analysis documents that the effectiveness of such firm s expenditures has a divergent effect on the relation between intangible assets and leverage. We first examine whether the effectiveness of R&D and advertisement expenditures in generating future sales relates to leverage. In addition, we assess whether such effectiveness affects the relation of intangible assets and leverage. Parsons and Titman (2009) argue that understanding the relation between firm s policies and both the levels and determinants of leverage can shed light on the value creation of such firm s policies. We follow Cohen, Diether and Malloy s (2013) method to estimate R&D investment ability as the effectiveness of R&D expenditure in generating future sales. We apply their method to estimate also advertising investment ability (Larkin (2013), Barth, Clement, Foster, and Kasznik (1998) link advertising expenditures to potential determinants of leverage). We find that R&D ability is negatively related to leverage. Advertising ability is unrelated to leverage. These results suggest an under-reaction in debt markets to the ability to generate future cash-flows and complement Chan, Lakonishok and Sougiannis (2001) and Cohen, Diether and Malloy s (2013) findings that the equity market underreacts to higher R&D ability. Kallapur and Kwan (2004) and Barth, et al (1998) find that brand value, not necessarily advertising expenditures, is partially recognized by equity prices. We find evidence of a complex interaction between R&D investment policies, intangible assets and capital structure. When we examine the interaction between R&D expenditure and R&D ability we find some evidence of a negative relation between leverage and the interaction 6

8 of good R&D ability and developed and in-process technology intangible assets. Looking at advertising policies, we find that higher marketing related intangible assets are related to higher leverage regardless of the advertising ability. The insignificant relation between advertising ability, leverage and marketing related intangible assets is robust to using interactions in the models. Using simultaneous equations models to control for joint determination among firm s expenditures characteristics (i.e., level and ability), intangible assets and leverage, we confirm the positive relation between intangible assets and leverage. We find, nevertheless, diverging results on the impact of intangible assets on leverage as a function of R&D and advertising expenditures and ability. As expected, higher R&D expenditures increases the value of both developed and in-process technology intangible assets. The R&D ability, however, increases only the value of in-process technology intangible assets. Looking at the impact of R&D policies on leverage we find that the firms with larger R&D expenditures and higher R&D ability can increase their debt capacity, yet only in technology industries. Similarly, higher advertising expenditures and ability increase debt capacity, a finding consistent with Larkin (2013) who finds that higher brand perception provides higher debt capacity. Interestingly, although higher advertising expenditures increase the value of trademarks and trade names intangible assets, the advertising ability does not affect such value. This last finding complements Green and Jame s (2013) finding that the interaction between financial and marketing-related intangible assets, such as the company name, increases stock liquidity and valuation ratios. We complement prior research by documenting that such marketing-financing relation exists only in non-technology industries. In sum, we document significant links among firm s policies, leverage and the value of intangible assets. The type of industry determines when R&D and advertising investment policies provides higher incentives to 7

9 have higher leverage as a function of the type of intangible assets (technology or marketing related). Overall our results make several contributions to the existing research. First, this study adds to the capital structure literature on the determinants of capital structure. Murillo and Giambona (2013) show that the redeployability of tangible assets is an important determinant for debt capacity. Rampini and Viswanathan (2013) theoretically show that firm s ability to promise to pay, linked mainly to collateral tangible assets, critically determines capital structure. Kimbrough (2007) provides evidence that the value of intangible assets reported in the PPA consideration provides additional information to equity investors. Our study shows that the value of intangible assets, actually priced by an acquirer, also provides additional information that seems to increase debt capacity. The nature of the intangible assets seems to define whether constrained or unconstrained firms use such intangible assets to have higher leverage ratio. Second, our findings on the relation among firm s expenditure behavior, leverage, and intangible assets complement studies that document equity market under-reaction to R&D and advertising expenditures and ability (e.g., Cohen et al. (2013), Li (2011), Chan et al. (2001), Lev and Sougiannis (1996), Kallapur et al. (2004), Barth et al. (1998)). We provide evidence that, although R&D and advertising expenditures increase the value of intangible assets, expenditure ability has a weaker impact on the value of intangible assets. Our study also adds to the literature on the interaction between financial and marketing decisions. We complement Larkin s (2013) finding that brand evaluation ameliorates financial frictions and provide additional debt capacity. We find that both the priced value of trademarks and trade names intangible assets and advertising expenditures are positively related to leverage although the advertising ability does not seem to affect leverage. Looking at debt and product markets, Fresard (2010) find that the level and use of cash affects the interaction between 8

10 product market and corporate financial decisions. We find that the use of cash, specifically through higher advertising expenditures, also ameliorates financial frictions and increases the value of marketing related intangible assets. Finally, prior research is scarce on the link between financial and customer-related decision. Kale and Shahrur (2007) argue that customer-specific investments and lower leverage align incentives between suppliers and customers. Maksimovic and Titman (1991) theoretically show that suppliers and customers might be reluctant to conduct business with highly levered firms. Our study indicates that customers concern on high leverage matters more in technology industries because firms in such industries have a negative relation between customer-related intangible assets and leverage. In the sections below, we develop our hypotheses followed by data and methodology. Test results are presented next and the last section concludes the paper. II. HYPOTHESES DEVELOPMENT Prior studies ignore intangible asset as a determinant of leverage because the intangible asset cannot be repossessed at the time of bankruptcy due to its lack of tangibility and it can be deployed into production process only by the owner since the agency problems in separating ownership from control are too severe (Rampini and Viswanathan (2013)). The following is standard leverage regression in the literature (Campello and Giambona (2013)): Lev = α + βtan + γcon + δ Firm + λ Year + ε, it, it, it, i t it, i= 1 t= 1 N T where the index i denotes a firm, t denotes a year, α is an intercept, Tan is tangible asset, Con is a matrix of control variables, and Firm (Year) is firm (time) fixed effect. 9

11 We now add intangible asset as an additional independent variable to see whether it serves as an additional determinant of capital structure in the following regression: N Lev = α + βtan + γcon + δ Firm + λ Year + ωint + ε, it, it, it, i t it, it, i= 1 t= 1 T where Int is intangible asset. Our null hypothesis is that ω =0. Hence, our alternative hypothesis is that ω >0. Intangible assets is positively related to leverage. We will decompose intangible asset into its components such as marketing related intangible, technology related intangible, and others to see which component of intangible assets loads up positive and significant. Prior research shows that the sensitivity between leverage and the nature of the assets matters more for firms with financial constraints and in industries that require specialized knowledge such as technology industries (e.g., Harris and Raviv (1991), Parsons and Titman (2009), Li (2011), Campello and Giambona (2013). Firm s expenditure and product-market policies and the ability to create value from such policies can also affect the value of the assets the firm use to pledge as collateral when raising debt (e.g., Larkin (2013), Morellec (2011), Rampini and Viswanathan (2013)). We expect that financial frictions, technology industry, firm s expenditure policy and firm s ability affect the sensitivity of the relation between leverage and intangible assets. We discuss the specific expectations in the sections in which we control for these factors. IV. DATA and METHODOLOGY The sample comprises of 838 firms that became U.S. public targets of completed acquisitions by U.S. public acquirers for the period of 2002 through We thank Houlihan Lokey for providing the original data set of 4,652 completed acquisitions of public firms and 10

12 assets with purchase price allocation information. Houlihan and Lokey collect the purchase price allocation from 10K or 10Q filed to Securities and Exchange Commission. In order to be included in the sample, the firm (i.e., the acquirer in the acquisition) had to disclose sufficient information on the target to calculate the purchase consideration (see Figure 1). We first matched the 4,652 acquisitions with COMPUSTAT using company names of the target and this matching reduce the dataset to 1,089 acquisitions, mainly derived from keeping in our sample only the public targets for which we can find information in COMPUSTAT. Next, to calculate the control variables used in our analysis, we require annual data available on the target firm as of the most recent year prior to the acquisition announcement. The final sample results in 838 target firms. To control the impact of extreme observations, we winsorize the extreme upper and lower one-percent of our return and firm-characteristics variables. Table 1 presents the descriptive statistics for the control variables we use in our analysis. Panel A describes our final sample of 838 target firms and Panel B the COMPUSTAT whole universe, 115,189 firm-year observations during the same period. Total assets is the book value of total assets reported in COMPUSTAT. Long-term leverage is the ratio of book value of longterm debt and liabilities. LT Leverage/Assets (i.e., leverage, which is the main dependent variable in our multivariate analysis) is the ratio of Long-term leverage to Total assets. Market Capitalization is the book value of equity, subtracting the book value of common equity and adding the market capitalization of common stock, estimated by multiplying the number of outstanding shares times the stock price at the closing of the prior quarter before the acquisition announcement. Market-to-book is the ratio of the market capitalization of common stock to the book value of common equity. Sales are the net annual sales. Operating Profitability is the ratio of Earnings-before-interest-taxes-and-depreciation (EBITD) to total assets (EBITD/Assets). Cash liquidity is the ratio of Cash and cash-equivalents to assets (Cash/Assets). Marginal tax rate is 11

13 Graham s (2000) marginal tax rate. Cash flow volatility is the standard deviation of the ratio for the prior 3 years (StDev [EBITD/Assets]). Book tangibility is the ratio of Property, plant and equipment to total assets (PPE/Assets). R&D expenditures and Advertising expenditures are the three-year average ratio of research and development to sales (R&D/sales) and advertising expenditure to sales (Advertising Exp./Sales )respectively. Good R&D ability and Good Advertising ability are binary variables that take the value of one if the R&D ability and advertising ability is above the median, and zero otherwise (see Cohen, Diether and Malloy s (2013)). R&D ability and advertising ability are estimated as the 5-year average ratio of sales divided to lagged R&D expenditures and lagged Advertising expenditures respectively. Technology Industry is a binary variable that takes a value of one if the firm operates in a technology industry, zero otherwise. Accounting Profit Margin is the ratio of net income to sales. Asset turnover is the ratio of sales to assets. Equity Multiplier is the ratio of assets to book value of common equity. Sales growth is the natural logarithm of the ratio of sales to lagged sales. When we compare with the whole COMPUSTAT universe we find that the firms in our PPA sample have statistically similar leverage ratios, market-to-book ratios, R&D expenditures and advertising expenditures. The firms in the PPA sample, however, have larger size (total assets, market capitalization and sales), larger leverage, higher operating profitability, lower marginal tax rate, lower cash-flow volatility, lower book tangibility, higher R&D and advertising ability, higher proportion of firms in a technology industry, higher accounting profit margin, lower asset turnover, higher equity multiplier, and lower sales growth. The similarity on the standard deviation of all the variables in both panels suggests that the PPA sample provides adequate variation for our cross-sectional analysis. Table 2 shows the purchase price allocation. Panel A of Table 2 shows dollar amounts of purchase allocated while Panel B of Table 2 shows the percentage of purchase price allocated to 12

14 each component. Almost half (i.e., 46.6%) of the purchase consideration comes from tangible assets, 22.1% from intangible assets. In terms of the decomposition of the intangible assets acquired, developed technology (DT), in-process R&D (IPRD), trademarks and names (TTN), customer-related assets (CRA),and other intangibles acquired (OTHER) have mean values of 5.2%, 2.8%, 3.2%, 6.8%, and 3.8% of the purchase consideration. When combined into technology (DIPT=DT+IPRD) and marketing (MR=TTN+CRA) related groups, they represent 8% and 10% of the total PPA, respectively. After allocating purchase price to the fair market value of identifiable assets of target firm (including the fair market value of tangible assets and unrecognized identifiable intangible assets), the mean of acquisition goodwill was 31% of the total purchase consideration. This 31% is quite equivalent to the 30.6% mean acquisition goodwill reported in Henning, Lewis, and Shaw (2000) using purchase price allocation data prior to SFAS 141. Consistent with prior literature, in untabulated correlation analysis we find a negative association of leverage with R&D expenditure and a positive association with book tangibility (i.e. PPE/Assets). The signs of the correlations among book tangibility, R&D expenditure, advertising expenditure, PPA s tangible and intangible assets have the expected signs. For instance, book tangibility is positively related to the PPA s tangible assets and negatively related to the PPA s Intangible assets, both as a whole and as individual component measures. The R&D expenditure is positively correlated with the PPA s intangible assets, both as a whole and as individual technology components. The advertising expenditure is positively correlated with the total PPA s intangible assets and to the trademarks and trade names measure. We defer further discussion on the associations to the multivariate analysis in the next section. 13

15 V. MULTIVARIATE RESULTS ON LEVERAGE AND INTANGIBLE ASSETS In this section, we examine the first main question we address in this paper, namely, whether intangible assets relate. We first present Tobit regressions. Then, we present OLS regressions and Heckman models (to address potential endogeneity of the selection of the firms in our PPA sample. Next, we examine the role of financial constraints and of technology industry. Our main objective is to assess whether intangible assets provide additional information beyond the leverage determinants already reported in the literature. A. Tobit Regressions on Leverage Table 3 reports the results from Tobit regressions on the long-term leverage of the firms in the PPA sample. Following Larkin (2013), we use a Tobit model specification, with truncation at zero leverage, because we find that 21% of the firms in our sample (and 25% in the universe of firms in Compustat) have zero leverage, which is consistent with Strebulaev and Yang (2013) who document that 22% of the US non-financial firms in their sample have a leverage ratio lower than 5%. We control for conventional leverage factors (size, market-to-book, profitability, cash-holdings). 2 We use Eicker-Huber-White-Sandwich heteroskedastic-robust standard errors clustered by industry using the 12 Fama-French industry classification (results are robust to clustering by industry using the 48 Fama-French, the SIC2-digits and SIC4-digits industry classifications, and also to double clustering for both year and industry). Following Murillo and Giambona (2013), we also report the interquantile (IQT) changes estimated as the percentage change in leverage relative to its sample mean as each continuous regressor increases from the 25th to the 75th percentile, while other regressors are kept at their sample mean. To assess whether the PPA elements provide any additional information, we present pairs of model 2 The literature we follow in our variable selection process includes Barclay and Smith (1995), Rajan and Zingales (1995), Graham (2000), Baker and Wurgler (2002), Frank and Goyal (2003), Korajczyk and Levy (2003), Hovakimiam, Hovakimian, and Tehranian (2004), Faulkender and Petersen (2006), Flannery and Rangan (2006), Lemmon, Roberts and Zender. (2008), and Murillo and Giambona (2013). 14

16 including or not the Book tangibility and the Book R&D expenditure and Book Advertising expenditure variables. Each pair of model includes the PPA elements at different detail level for the PPA s Intangible Assets percentage. As a benchmark, Model 1 includes only firm characteristics used in prior research as explanatory variables. 3 Consistent with prior literature, all the models that include the Book tangibility, R&D expenditures and Advertising expenditures variables show positive and negative significant coefficients respectively (e.g., Baker and Wurgler (2002), Hovakimiam et al. (2004), Strebulaev and Yang (2013), Larkin (2013)). Models 2 and 7 indicate that the PPA elements are significantly related to leverage. Consistent with Kimbrough (2007), who finds that the equity prices contain additional information related to the PPA considerations beyond what accounting measures contain, debt markets seem to recognize that PPA s Tangible Assets % provide additional debt capacity, even after controlling for Book tangibility. PPA s intangible assets (as a whole and as individual components) are positively related to leverage, even after controlling for R&D expenditures and Advertising expenditures (i.e., Models 2, 4 and 6). The interquantile changes in Intangible Assets % are economically significant. Model 3 shows that a one-iqt change in Intangible Assets % leads leverage to increase by 0.094, which is a 58.7% increase relative to the predicted leverage of This IQR change is similar in magnitude to the IQT change that Book tangibility provides and below to the other control variables, except for Cash liquidity. When we decompose the intangible asset into its main components in Models 4 and 5, both the Developed and In-Process Technology % and the Marketing Related % are also significantly positive. The IQT changes reported in Model 3 The coefficients for the traditional determinants of leverage have, in general, the expected signs. Consistent with Kim and Sorensen (1986) and Merhan (1992) and survey by Parsons and Titman (2009), we do not find any effect of size on leverage. Chen and Zhao (2006) argue that few small firms with very large market-to-book ratios drive the negative relation between market-to-book and leverage documented in prior literature. By construction our sample excludes small firms because of lack of data, which then provides a potential explanation for the lack of effect that market-to-book has on leverage we find in our models. 15

17 4 indicate that Developed and In-Process Technology % has almost double the impact than that of R&D expenditures (8.4% vs.4.6%). Model 5 indicates an even larger impact (17% IQT change) if we do not control for R&D expenditures.the economic impact of Marketing Related % is even larger with a IQT change of 19.9% or 22.04% (depending on the controls used in the model). When we further decompose the PPA components in Models 6 and 7, we find that the economic impact of all the detailed PPA components is significant except for Developed- Technology %. The In-process Technology % seems to drive the positive relation between R&Drelated intangible assets and leverage. The stronger significance and larger IQR change of the Customer-related Assets % compared to the Trademarks and Trade Names % suggest that Customer-related Assets % is the main driver for the Marketing Related % large economic impact in Models 4 and 5. Panels B and C replicate the multivariate analysis now using an OLS model specification and including year and industry fixed effects. Results are similar although when we add year and industry fixed effects the Trademarks and Trade Names % becomes insignificant. Overall, Table 3 shows that the PPA consideration provides additional information on the capital structure of the firms in our sample. The pricing of intangibles assets by a third party (the acquirer in the subsequent acquisition) seems to shed additional light on whether such intangible assets had also been important economic elements of the target firm leverage before the acquisition took place. To alleviate concerns that these findings occur only in our smaller sample of PPA firms, we run Heckman regressions to control for sample selection bias in the next section. B. Heckman regressions on Leverage One of the main advantages of our research design, namely, using the information released when an acquirer discloses the PPA consideration after acquiring a target firm, can also 16

18 be the source of a potential sample selection bias, and hence, might limit the generalization of our findings. For instance, although Table 1 reports similarities between the sample of target firms and the rest of the COMPUSTAT universe on our main variables of interest, namely, leverage ratios, R&D expenditures and advertising expenditures, we also find some differences in other firm characteristics. Moreover, the acquirers of the targets in our PPA sample can select firms that have different economic incentives to define their capital structure as a function of their own characteristics that are priced also by the acquirer (and reported in the PPA consideration). We run Heckman regressions to account for the endogeneity of the selection of the firms in our PPA sample which ultimately can affect the estimation of the association between the PPA consideration and leverage. Even though the target firms may not choose when and if to be acquired, Heckman regressions should help assess the economic impact of the PPA consideration and address the lack of randomness in the sample (Heckman, 1979; Li and Prabhala, 2007; and Guo and Fraser, 2010). We do not have any priors, nor to our knowledge prior research suggests suitable determinants, on what causes a firm to become a target; although bidders might consider attractive firms with higher Return-on-Equity ratios. We include as control variables in the first stage the three ratios that a DuPont decomposition of the Return-on-Equity suggests, namely, Accounting profit margin, Asset turnover, and Equity multiplier. For identification purposes, we change the log of market capitalization for total assets, include exclude other controls from Table 3 in the first stage (our results are robust to including all the other control variables from Table 3 and to excluding Sales growth as a control). Table 4 presents the Heckman regressions in which the dependent variable in the second stage models is leverage. The dependent variable in the first stage model is the indicator variable equal to one if the firm is included in the PPA sample. The first stage uses the universe of firms 17

19 in COMPUSTAT and the second stage uses the smaller PPA sample. We report the coefficients after solving the models with maximum likelihood. Our results are robust to using a two-step approach. We report the correlation among the errors in both stages estimated with the maximum likelihood approach and the lambda (i.e., inverse mills ratio) estimated in the two-step approach. Our discussion focuses on the coefficients of our main variable of interest, namely, the PPA components. The first stage estimation in column 1 indicates that Book tangibility decreases, and R&D expenditures increases the probability of inclusion in the PPA sample, and arguably of becoming a target in general since our PPA sample is built precisely on becoming a target. This finding suggests that third parties, such as acquirers, appreciate investing in intangibles, such as in R&D. Alternatively, one mechanism though which firms can redeploy intangible assets is an acquisition. The second stage models of Table 4 confirms the findings in Table 3. Even after controlling for potential sample selection bias, both PPA s tangible and intangible assets, either as a whole or based on its individual components, show a significantly positive association with leverage. Arguably, the acquirers can price intangible assets in a similar way that debtors consider such intangible assets when lending money to the firms. Such pricing is consistent with Morellec (2001) notion that higher asset liquidity increases debt capacity. In any case, neither the Wald test of error correlation or lambda is significant, which suggests that the results we find in the PPA sample can be extrapolated to the COMPUSTAT universe of firms. Accounting for potential simultaneity and omitted variables biases the Heckman regressions suggest that the PPA consideration, specifically, of the intangible assets components, have a significant economic impact on leverage. 18

20 Taken together, the above results indicate significant interactions between firm s expenditure behavior, the value of intangible assets and leverage. To shed more light on what affects such interactions, in the next section we assess for differences in the impact of intangible assets on leverage as a function of the firm having financial constraints and being in a technology industry C. Role of Financial Constraints The prior sections document that the intangible assets that acquirers price (as reported in their PPA consideration) also affect the capital structure of the target firm. If the target considers its intangible assets as potential collateral when raising debt, then the economic impact of such intangible asset collateral can be larger for firms facing financing frictions. Following Campello and Giambona (2013) we use three proxies to classify firms as financially constrained. First, we classify as financially constrained those firms in the bottom three deciles of the COMPUSTAT universe of firms ranked by total assets, ranked annually (Gilchrist and Himmelberg, 1995, Fama and French, 2002, and Frank and Goyal, 2003). Second, we classify as financially constrained the firms whose public debt (COMPUSTAT s S&P Domestic Long Term Issuer Credit Rating, S&P Subordinated Debt Rating, and S&P Domestic Short Term Issuer Credit Rating) is not rated (See Faulkender and Petersen, 2006). Finally, we classify as financially constrained the firms that do not pay dividends (Fazzari, Hubbard and Petersen, 1988, Fama and French, 2002). Panel A in Table 5 reports univariate analysis after splitting the sample based on our main proxy for financial constraints, namely, small firms. Small firms have a smaller proportion of tangible assets and larger proportion of intangible assets and goodwill. The Developed technology, Customer related and Other PPA components drive the higher proportion of intangible assets in smaller firms. 19

21 Panel B in Table 5 reports Tobit models on long-term leverage divided by assets after including interaction terms between the main variables of interest and our main financial constraint proxy, namely, the Small firm indicator. The comparison between the insignificant coefficients for the interaction terms in Models 1 and 2 and the significant coefficients in many of the interactions in Models 3, 4, 5 and 6 highlight the importance of distinguishing different types of intangible assets. We find that in-process technology intangible assets increase the positive association with leverage in financially constrained firms. Our findings looking at debt markets complement Li s (2011) result that financially constrained firms seem to drive the positive relation between R&D expenditures and equity market value. In contrast, we find a positive relation between Trademarks and trade names and leverage when the firm is not financially constrained but a negative relation when the firm is financially constrained. This contrasting behavior can partially explain non-financially constrained, firms benefit from having higher debt when the firms compete more aggressively, as Brander and Lewis (1986) and Maksimovic (1986, 1990) theoretically argue. The negative relation between market related intangible assets and leverage for financially constrained firms seems consistent with Dasgupta and Titman s (1998) model, in which smaller firms who aim to increase their market share, prefer lower leverage ratios. The results are similar, albeit weaker, in Panel C after including interactions with our other two proxies for financial constraints. Overall, the results in Table 5 are consistent with the hypothesis that intangible assets related to in-process technology can help financially constrained firms to have higher leverage ratios. In contrast, unconstrained firms seem to have larger leverage by exploiting their marketing related intangible assets. Next section examines another factor, i.e., technology industry, which might affect the relation between intangible assets and leverage 20

22 D. Role of Technology Industry In this section we aim to assess how the relation between intangible assets and leverage change depending on the industry. 4 Panel A in Table 6 describes the sample distribution by industry for our PPA sample and for the universe of firms in Compustat. We find a higher proportion of acquisitions within certain industries, specifically in Business equipment, consistent with acquisitions occurring in waves clustered by industry (i.e., Mitchell and Mulherin, (1996) Maksimovic and Phillips (2001), Rhodes-Kropf, Robinson, and Viswanathan (2005), Ahern and Harford (2013)). The type of PPA components that seem to matter for each industry makes economic sense. Panel B in Table 6 reports top four industries with the highest intangible assets are Healthcare, Consumer non-durables, telecommunications, and business equipment. Healthcare has the largest Developed and In-process technology intangible assets, followed by business equipment. Consumer non-durables, Manufacturing, and Consumer durables have the largest Trademarks and trade intangible assets. Telecommunications, Business equipment, and Manufacturing have the largest Consumer related intangible assets. To get a more detailed classification, we use 2-digit SIC codes to identify technology industries. Technology Industry takes a value of one if according to the firm s 2-digit SIC code indicates the firm is in Medical equipment, (12), Pharmaceutical Products (13), Machinery (21), Electrical equipment (22), Defense (26), Computers (35), Electronic equipment (26), and Measuring and Laboratory equipment (37). We get similar results if we use the 48 Fama French industry codes to classify Technology industry. The classification by technology industry reported in Panel C also makes economic sense. Technology industries have larger Developed and In-process technology intangible assets. 4 The Heckman regressions in Table 4 show that our results are robust to sample selection. 21

23 Non- Technology industries have larger Trademarks and trade intangible assets. Consumer related intangible assets do not change as a function of technology industry. Although we expect to find that leverage has a higher sensitivity with both the value of R&D expenditures and technology related intangible assets (see Parsons and Titman (2009), we do not have any priors on the effect of industry on marketing related intangible assets. Prior research on the value of brands does not make any distinction based on the industry (e.g., Kallapur and kwan (2004), barth et al. (1998), Fresard (2010), Larkin (2013)). Kale and Shahrur (2007) finds suggestive evidence that industries with higher R&D intensity (such as those classified in our Technology industry indicator), potentially requiring relationship-specific investments, will have lower leverage. Hence, we expect to find a negative relation between leverage and Customer related intangible assets. Tobit models in Panel A of Table 7 further indicates additional contrasting results between developed and in-process technology and marketing-related intangible assets on their relation with leverage after we consider the role of Technology industry. Models 2 and 3 show that, although firms in technology industry have lower leverage, Developed and In-process technology intangible assets seem to provide higher debt capacity for technology firms. In contrast, firms in non-technology industries have a negative association between leverage and Developed technology intangible assets. This reversal in sign for Developed technology explains the insignificant coefficient for Developed technology in the prior tables. The interaction with technology industry also indicates a contrasting impact on the relation between Marketing related intangible assets and leverage. We find a negative association between leverage and both Trademarks-and-trade-names and Customer-related intangible assets when firms are in technology industries and a positive association in nontechnology industries. These results are consistent with Kale and Shahrur (2007) and 22

24 Maksimovic and Titman (1991) contention that when customer-related investments are more specific (as in the case of technology industries), then the firm has the incentives to have lower debt ratios. Treatment effects regressions in Panel B in Table 7 address potential self-selection bias on the Technology industry treatment effect. 5 Although instrumental variables in two- or three-stage-least-squares models can control for endogeneity in the form of simultaneity bias (Baum, 2006), these techniques do not address self-selection biases (Guo and Fraser, 2010), tend to suffer from omitted variables (Li and Prabhala, 2007), and have a bias towards finding insignificant results (Ettner, 2007). The first stage indicates, as expected, that firms in a technology industry are positively related to R&D expenditures and negatively related to Book tangibility. Again we find that neither the Wald test of error correlation or lambda is significant, which suggests that the results we find in the PPA sample can be extrapolated to the COMPUSTAT universe of firms. The second-stage Models 2, 3 and 4 show that the instrumented Technology Industry is insignificant. After controlling for potential self-selection driven by the Technology Industry we find that the Intangible assets as a whole and as all the four individual components are positively related to leverage. So far, our analysis indicates a robust positive association between intangible assets and leverage. Next section examines whether the ability a firm has to turn its R&D and advertising expenditures into future sale affects the relation between intangible assets and leverage. V. MULTIVARIATE RESULTS ON LEVERAGE, INTANGIBLE ASSETS AND FIRM S INVESTMENT ABILITY 5 Other examples of treatment effect regressions in corporate finance are in Campa and Kedia (2002), Song, (2004), Fang (2005), Bris, Welch, and Zhu (2006), and Jiang, Li and Wang (2012). 23

25 In the last part of our analysis we examine to what extent the ability to invest in R&D and advertising, and not only the simple use of cash in R&D and advertising expenditures, benefits the firm in either increasing the value of its intangible assets or in increasing the sensitivity of the relation between intangible assets and leverage. Specifically, we want to address Parsons and Titman s (2009, pg. 82) call to study the relation between the quality of corporate behavior, leverage and value creation (in our case, examining whether the ability to invest cash in R&D and advertising affects the value of intangible assets). Prior literature has documented mixed results when examining whether equity markets recognize or underreacts to value creation from R&D expenditures and ability (e.g., Lev and Sougiannis (1996), Chan, Lakonishok and Sougiannis (2001), Kallapur and Kwan (2004), Barth et al (1998), Green and Jame (2013), Larkin (2013), Cohen, Diether and Malloy (2013). Hence, we want to assess also whether debt markets show a similar under-reaction to R&D and advertising expenditures, ability and to the value of intangible assets. To estimate R&D ability we follow Cohen et al. s (2013) non-regression-based measure of ability, estimated as the 5-year average ratio of sales divided to lagged R&D expenditures respectively. 6 We apply their method to estimate also the effectiveness of advertising expenditures, which prior literature links to potential determinants of leverage (e.g., Larkin (2013), Barth, Clement, Foster, and Kasznik (1998)). We first examine whether the effectiveness of R&D and advertisement expenditures in generating future sales relates to leverage. Next, we 6 When we replicate their much more restrictive regression-based measure Good R&D ability estimation we find only 34 firms in the whole PPA sample. This small number of firms labeled as Good advertising ability is consistent with the figures reported in Cohen et al. s (2013) Table A3 (online Appendix), in which Panel A reports only 29 High Ability and High R&D cases, out of their whole sample that uses the whole universe of firms in Compustat from 1980 to 2009 using 5 years of prior data. Cohen et al. use a time-series approach which can use firm-year observations, which allows them to use this more restrictive measure. Our results are robust, albeit weaker, to using this this more restrictive measure in our cross-sectional analysis, arguably due to our smaller sample size compared to Cohen et al s (2013) portfolio approach. 24