Master’s Thesis
The Effects of Competition on Corporate Cash Holdings
Thomas Terstegen (10098321) June 2017
Amsterdam Business School Master Finance
Asset Management Track
Statement of Originality
This document is written by Thomas Terstegen, who declares to take full responsibility for the contents of this document.
I declare that the text and the work presented in this document is 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.
Abstract
This paper examines how competition affects the decision of firms to hold cash. A positive relation between competition and cash holdings follows from the precautionary motive of holding cash. The theory of predation risk, however, supposes that the motive of holding cash to prevent losing investment opportunities, which are proxied by Tobin’s Q, to rivals, is
negatively related to competition. Contradictory with the former, but consistent with the latter theory, the direct effect of competition on cash holdings appears to be negative. Based on a firm-specific measure of competition, firms are assumed to act in competitive, oligopolistic, or monopolistic industries and assigned to them accordingly, whereafter cash holdings-Q
regressions are run conditional on these industries. These show that cash holdings-Q sensitivity is non-linear over the course of competition as it is highest for firms in oligopolistic industries. This is explained by the high predation risk and the urge for strategic actions by firms in that type of industry.
Keywords: Cash holdings, competition, investment opportunities, precautionary savings,
Table of Contents
I. Introduction ... 5
II. Literature review ... 7
A. Cash Holdings Motivations ... 7
B. Competition ... 8
C. Financial Constraints ... 9
D. Financial Policy Decisions ... 10
III. Methodology ... 13 A. Research Design ... 13 B. Variable Construction ... 15 C. Sample Selection ... 16 IV. Data ... 17 A. Data Sources ... 17 B. Summary Statistics ... 20 V. Results ... 24 A. Regression Analysis ... 24 B. Additional Research ... 28 C. Robustness Checks ... 31 VI. Conclusion ... 35 References ... 38 Appendix A ... 41 Appendix B ... 42
I. Introduction
Firms tend to hold large amounts of cash on their balance sheets for several reasons. The precautionary motive of holding cash states that firms hoard cash to be able to invest when investment opportunities arise (Bates, Kahle and Stulz, 2009; Opler, Pinkowitz, Stulz, and Williamson, 1999). In certain industries these investment opportunities are shared among rivals, which provides room for predatory behavior on behalf of those rivals. Firms can become victims of this when they lack the funds to invest, and consequently lose their investment opportunities to their rivals. To manage this predation risk, firms hoard cash to be able to invest (Haushalter, Klasa, and Maxwell, 2007), and in particular those that are financially constrained (Almeida, Campello and Weisbach, 2004).
Competition is one of the key determinants of the interdependence of investment opportunities with rivals (Haushalter et al., 2007) and has a negative relation to predation risk (Kovenock and Phillips, 1997; Zingales, 1998). This implies that for firms in less competitive industries it is of greater importance to protect themselves against predation risk by holding cash. However, from the precautionary motive of holding cash it follows that firms in more competitive industries are forced most to hold cash as they have more trouble raising external funds. This contradicting effect of competition on corporate cash holdings is the puzzle studied in this paper, and leads to the following research question: How does competition affect the decision to hold cash?
To answer this question, the direct of effect of competition on cash holdings is
measured. In addition, the sensitivity of cash holdings to investment opportunities for different levels of competition is examined. A similar research design to the one Akdogu and MacKay (2008) introduce, is used for this. Annual data between 2002 and 2015 is used from a large sample of U.S. firms. Based on the degree of competition firms face, the sample is split in three panels: competitive, oligopolistic, and monopolistic industries. Furthermore, regressions are run for the entire sample, and conditional on the level of competition. The measure of competition used is the Herfindahl-Hirschman Index (HHI) and Tobin’s Q is the proxy for investment opportunities.
Against the expectations following from the literature on the precautionary motive of holding cash, but consistent with those from the theory of predation risk, a negative direct effect is found of competition on cash holdings. In one of the two robustness checks, a positive effect is found, however, affirming that there are two forces working in opposite directions. Firms facing financial constraints turn out to hold more cash, as expected. The main finding of this paper is the non-linearity in cash holdings-Q sensitivity across the different panels. The cash holdings of firms in oligopolistic industries are most responsive to variations in Tobin’s Q. This finding is robust to changes in sectors included and competition measure used. As an additional research, two explanations of this finding are tested and adopted. The first one implies that predation risk is highest for firms in oligopolistic industries, and the second one implies that firms in this type of industry tend to hoard cash strategically, to be able to invest to deter entry and force exit when investment opportunities arise.
This paper contributes in two ways. First, a new approach is used to examine the effect of competition on cash holdings. This approach was used by Akdogu and MacKay (2008) to investigate the investment-Q relation, and forms the foundation of this cash holdings-research. Through running regressions conditional on the level of competition, relations between cash holdings and their determinants get new dimensions. The second contribution is the finding of the non-monotonic cash holdings-Q sensitivity across the different levels of competition. This can be helpful in understanding corporate financial policy decisions and builds on the paper of Haushalter et al. (2007) in the sense that strategic benefits can be achieved by ascertaining the proportions of predation risk.
This paper is organized as follows. Section II reviews the existing literature on cash holdings and their relation to competition and financial constraints, and on financial policy decisions. From there, three hypotheses are formed. In section III the research design is described, together with the variable construction and the sample selection. Section IV
elaborates on the data sources and presents the summary statistics. In section V the results are reported, explained and checked for robustness, and the hypotheses are tested. Finally, section VI concludes and discusses the paper.
II. Literature review
This section reviews the existing literature on cash holdings and their implications to this paper. The studies that focus on the general motivations to hold cash are discussed first, before
switching to the determinants studied in this paper, competition and financial constraints. The last subsection elaborates on the role of investment opportunities on corporate financial policies on cash holdings and investment. Ultimately, this results in the formation of three hypotheses.
A. Cash Holdings Motivations
The puzzle of holding cash is the trade-off between the costs and benefits of it. Among the costs Opler et al. (1999) identify the lower return compared to when the cash would be
invested, the agency problem of managers wasting excess cash, and the double taxation of the money. They describe the benefits of holding cash as lower transaction costs in raising funds for future investments and having the opportunity to invest with the cash held when other sources of finance are unavailable or too costly. Bates et al. (2009) name this second benefit the
precautionary motive and place it next to the aforementioned transaction motive, and the tax and agency motives as the four main reasons for firms to hold cash. Of those the precautionary motive of holding cash is most extensively studied throughout this paper. It implies that firms hold cash so they can still invest when current net income would not be sufficient. Raising external capital is more costly, and even less preferred when it comes down to issuing new equity. Passing up valuable investment opportunities is not an economically sound decision, making firms’ savings the most favorable source of financing, consistent with the pecking order theory, which was first revealed by Myers and Maljuf (1984). The precautionary motive is also the one that is most strongly related to competition. When competition increases, firms’ margins decrease and their profits fall. This decreases their creditworthiness, making it even more difficult to raise external funds. Opler et al. (1999) argue that management tends to accumulate cash mainly for precautionary reasons, finding evidence in the fact that a negative
cash flow has a stronger effect on cash holdings for firms that already have cash holdings above their target level. Furthermore, Bates et al. (2009) observe that cash holdings have increased substantially since the 1980s, while because of increased hedging possibilities a decrease of cash holdings was expected instead. They find that the reason for this is that firms’ cash flows have become riskier, due to increased idiosyncratic risk. This is consistent with the findings of Opler et al. (1999) that firms with riskier cash flows hold more cash. Bates et al. (2009)
additionally state that the increase in cash holdings is due to firms having changed their balance sheets, holding fewer inventories and receivables then before, and having lower capital
expenditures and higher R&D investments.
B. Competition
The focus of this paper is on how competition affects cash holdings. In perfect capital markets, firms can borrow money at any time without restrictions. However, in reality, firms may have difficulties raising external capital, because competition gives them a higher probability of becoming in distress. Consistent with this reality, several studies have found that firms facing greater difficulties in obtaining external capital accumulate more cash (Opler et al., 1999). Morellec, Nikolov, and Zucchi (2014) show that product market competition has a direct positive effect on firms’ cash holdings, whereas competition decreases profitability, forcing firms to hoard more cash. And against their expectations that large and unconstrained firms hoard relatively most cash as a strategic device to deter entry or force exit, they find that the effect of competition on financial decisions is most evident for smaller and financially
constrained firms. Hoberg, Phillips, and Prabhala (2014) study the influence of product market threats on cash holdings and payout policy. Their measure of these competitive threats is fluidity, which is a text-based measure capturing relative changes between the products of firms and their rivals. They find that product market threats decrease payouts, dividends and share repurchases, and increase cash holdings, again especially for firms with worse access to external capital. Haushalter et al. (2007) study the effects of predatory behavior among industry rivals. When firms that have the same growth opportunities as their industry
competitors are unable to fully invest in these opportunities, they can become victims of the predatory behavior of their competitors. The interdependence of investment opportunities among rivals is found in the commonality of investment opportunities, the similarity of operations, the covariance of the firms’ growth opportunities with those of rivals, and the absence of competition. Predation risk increases in the degree of interdependence. They find that firms manage this predation risk by hoarding cash and using derivatives, which comes down to a negative effect of competition on cash holdings. Zingales (1998) even argues that predatory behavior can only be present in less competitive environments. The long-term recovery of the predator, which is needed from the short-term costs of his behavior, is only possible when certain entry barriers exist. Kovenock and Phillips (1997) support the idea that these problems appear most in concentrated industries, because in these markets
non-optimizing firms are not weed out straight away, whereas they are in competitive industries.
The literature is contradictory on the effect of competition on cash holdings, as it appears that two forces are working in opposite directions. Through predation risk, the indirect effect of competition seems to be negative. However, the direct effect of competition in most recent literature turns out to be positive, consistent with the precautionary motive. For that reason, the first hypothesis is formed as follows.
HYPOTHESIS 1:The direct effect of competition on cash holdings is positive.
C. Financial Constraints
Whited (2006) investigates the effect of financial constraints on the timing of large investment. Using a hazard model he finds that financial constraints have a negative effect on investment. Almeida et al. (2004) also observe the effect of financial constraints on corporate policy, but rather focus on cash holdings. They take the tendency of firms to save cash out of their cash flow as a measure of financial constraints. Estimating this cash flow sensitivity of cash for manufacturing firms from 1971 to 2000, they find evidence for their theory that constrained
firms have a positive cash flow sensitivity, while for unconstrained firms no significant relation is found. Morellec et al. (2014) also find a stronger effect of competition on cash holdings for firms facing financial constraints, leading to the second hypothesis.
HYPOTHESIS 2:The effect of financial constraints on cash holdings is positive.
D. Financial Policy Decisions
Corporate financial policy decisions on cash holdings and investment are to a large extent dependent on investment opportunities. These have implications on the policy decisions in both a financial and a strategic manner. The latter is of most interest when investment opportunities are shared with rivals, and for incumbent firms to protect their market share. Fresard (2010) examines the impact of cash on product market performance and finds that cash policy has an absolute strategic component. He applies a difference-in-difference estimation on shifts in import tariffs to expose an exogenous increase in competition. The analysis clearly distinguishes the competitive effect of cash from the strategic effect of debt as effects on product market outcomes. Large cash holdings result in an increase in future market share at the expense of competitors operating in the same industry. This effect strengthens when these competitors are financially constrained and interactions between industry rivals are large (Fresard, 2010). Lyandres and Palazzo (2016) focus on the importance of the strategic role of the cash policy of innovative firms. They model a competitive situation where firms do not know the level of competition and where firms use cash as commitment device to invest in innovation. They show that firms base their cash assets ratio on the expected level of
competition. In addition, they observe that cash holdings of competitors are negatively affected by firms’ own cash holdings. This effect is even enlarged when competition is expected to be severe (Lyandres and Palazzo, 2016). Minton and Schrand (1999) study how cash flow volatility affects investment and they find a negative relation. They argue that this implies that firms rather delay or forgo investment than try to raise external capital to finance investments when cash flow is insufficient. Cash flow volatility also comes with higher costs of raising external
financing, increasing the sensitivity of investment to cash flow volatility even more. Akdogu and MacKay (2008) study the relation between investment and competition. In strategic real
options literature, they find that there is a wait-lose trade-off in investment policy. Delaying irreversible investment can be valuable because of uncertainty. However, waiting too long can erode the value of the investment opportunity. Also, investing can be preferable when it serves a strategic purpose. Akdogu and MacKay (2008) argue that the value of waiting is highest for firms in monopolistic industries and that the optimal exercise of investment opportunities is also nearly linked to the level of competition. This link is tested by estimating investment-Q sensitivity for firms operating in industries with different structures and by a duration analysis measuring how competition affects investment timing. They find that firms in monopolistic industries indeed have lower investment-Q sensitivity and delay investments more often than firms in competitive industries. However, the most interesting aspect of the Akdogu and MacKay (2008) paper is that this effect is not linear along the levels of competition. They find that both the investment-Q sensitivity and the pace to invest are highest for firms in
oligopolistic industries, representing the mid-concentration industries. This proves that the value of investing strategically can outweigh the value of waiting. This is emphasized by the fact that these industries face least entries and most exits (Akdogu and MacKay, 2008).
This strategic effect of financial policy decisions is extensively investigated through game-theoretic models. In such models, incumbent firms use irreversible investment and capacity to show their credible commitment to the industry, making it unprofitable for potential entrants to enter the product market (Spence, 1977). This strategic action is an
effective way of changing the setting of the post-entry game to the advantage of the incumbent firm (Dixit, 1980). However, overinvestment can also become a strategic handicap as it reduces the adequacy of reacting to the actions of competitors. Therefore, firms should rather be eager to invest instead of already having their investment on a satisfactory level (Fudenberg and Tirole, 1984). As an additional research these strategic effects of holding cash will be observed for the firms studied in this paper.
Akdogu and MacKay (2008) show that investment-Q sensitivity is strongly dependent on the levels of competition. In this paper the cash-holdings equivalent is tested, which examines whether competition affects the responsiveness of cash holdings to investment opportunities. Two factors must be considered when forming a hypothesis on this sensitivity, predation risk and strategic actions. Both occur only sparingly in highly competitive industries, and are expected to increase as competition decreases. However, at the other extremity, towards highly monopolistic industries, it gets less likely that firms lose their investment opportunities to rivals (Akdogu and MacKay, 2008). The effect of these two factors on cash holdings-Q sensitivity over the entire course of competition is yet unknown, but from above mentioned expectations, the following hypothesis is formulated.
III. Methodology
This section describes the methodology used to carry out the research of this paper. The first subsection presents the research design with which the hypotheses are tested. Thereafter, the construction of the variables and the selection of the samples are discussed.
A. Research Design
The methodology used in most cash holdings literature is quite straightforward: regressions are run of cash holdings on multiple variables acting as determinants. The predictions of the first two hypotheses are as discussed, positive direct effects of competition and financial constraints on cash holdings. The third hypothesis, on the cash holdings-Q sensitivity, gets its implications from the predation risk and the strategic motive. These arguments point towards a decreasing sensitivity as competition increases. Guided by these three hypotheses, the effect of
competition on the decision to hold cash is examined.
To be able to answer the third hypothesis, a similar approach is applied to the one used in the Akdogu and MacKay (2008) paper. First, the sample is split in three based on measure of competition. Then, cash holdings regressions are run conditional on the level of competition, and unconditional for the first two hypotheses. The folded form of the regressions looks as follows.
𝐶𝑎𝑠ℎ 𝐻𝑜𝑙𝑑𝑖𝑛𝑔𝑠𝑖,𝑡
= 𝛽0+ 𝛽1𝑇𝑜𝑏𝑖𝑛′𝑠 𝑄
𝑖,𝑡+ 𝛽2𝐶𝑜𝑚𝑝𝑒𝑡𝑖𝑡𝑖𝑜𝑛𝑖,𝑡 + 𝛽3𝐹𝑖𝑛𝑎𝑛𝑐𝑖𝑎𝑙 𝐶𝑜𝑛𝑠𝑡𝑟𝑎𝑖𝑛𝑡𝑖,𝑡
+ 𝛿𝑋𝑖,𝑡+ 𝛿𝑈𝑖,𝑡+ 𝜑𝑗+ 𝜈𝑡+ 𝜖𝑖,𝑡
Where 𝛿𝑋𝑖,𝑡 and 𝛿𝑈𝑖,𝑡 represent vectors of standard control variables and risk control
The set of control variables that is commonly believed to affect cash holdings (Bates et al., 2009; Opler et al., 1999) consists of: investment, cash flow, financial leverage, firm size, net working capital, R&D expenditures, acquisition activity, dividends, and net income.
𝛿𝑋𝑖,𝑡 = 𝛽4𝐼𝑛𝑣𝑒𝑠𝑡𝑚𝑒𝑛𝑡𝑖,𝑡+ 𝛽5𝐶𝑎𝑠ℎ 𝐹𝑙𝑜𝑤𝑖,𝑡+ 𝛽6𝐹𝑖𝑛 𝐿𝑒𝑣𝑒𝑟𝑎𝑔𝑒𝑖,𝑡+ 𝛽7𝐹𝑖𝑟𝑚 𝑆𝑖𝑧𝑒𝑖,𝑡
+ 𝛽8𝑁𝑊𝐶𝑖,𝑡 + 𝛽9𝑅&𝐷 𝐸𝑥𝑝𝑖,𝑡+ 𝛽10𝐴𝑐𝑞𝑢𝑖𝑠𝑖𝑡𝑖𝑜𝑛𝑠𝑖,𝑡+ 𝛽11𝐷𝑖𝑣𝑖,𝑡+ 𝛽12𝐿𝑜𝑠𝑠𝑖,𝑡
In addition to the standard controls, two risk proxies, volatility and capital intensity, are added to capture the effect of uncertainty on financing decisions (Andrade and Stafford, 2004; Minton and Schrand, 1999).
𝛿𝑈𝑖,𝑡 = 𝛽13𝑉𝑜𝑙𝑎𝑡𝑖𝑙𝑖𝑡𝑦𝑖,𝑡+ 𝛽14𝐶𝑎𝑝𝑖𝑡𝑎𝑙 𝐼𝑛𝑡𝑒𝑛𝑠𝑖𝑡𝑦𝑖,𝑡
Lastly, the regressions control for shocks on industry level and changes over time that might influence results, by including two-digit SIC code (𝜑𝑗) and year (𝜈𝑡) fixed effects
respectively. All together it results in the unfolded form of the regressions. 𝐶𝑎𝑠ℎ 𝐻𝑜𝑙𝑑𝑖𝑛𝑔𝑠𝑖,𝑡 = 𝛽0+ 𝛽1𝑇𝑜𝑏𝑖𝑛′𝑠 𝑄 𝑖,𝑡+ 𝛽2𝐶𝑜𝑚𝑝𝑒𝑡𝑖𝑡𝑖𝑜𝑛𝑖,𝑡 + 𝛽3𝐹𝑖𝑛𝑎𝑛𝑐𝑖𝑎𝑙 𝐶𝑜𝑛𝑠𝑡𝑟𝑎𝑖𝑛𝑡𝑖,𝑡 + 𝛽4𝐼𝑛𝑣𝑒𝑠𝑡𝑚𝑒𝑛𝑡𝑖,𝑡 + 𝛽5𝐶𝑎𝑠ℎ 𝐹𝑙𝑜𝑤𝑖,𝑡+ 𝛽6𝐹𝑖𝑛𝑎𝑛𝑐𝑖𝑎𝑙 𝐿𝑒𝑣𝑒𝑟𝑎𝑔𝑒𝑖,𝑡 + 𝛽7𝐹𝑖𝑟𝑚 𝑆𝑖𝑧𝑒𝑖,𝑡 + 𝛽8𝑁𝑒𝑡 𝑊𝑜𝑟𝑘𝑖𝑛𝑔 𝐶𝑎𝑝𝑖𝑡𝑎𝑙𝑖,𝑡+ 𝛽9𝑅&𝐷 𝐸𝑥𝑝𝑒𝑛𝑑𝑖𝑡𝑢𝑟𝑒𝑠𝑖,𝑡 + 𝛽10𝐴𝑐𝑞𝑢𝑖𝑠𝑖𝑡𝑖𝑜𝑛𝑠𝑖,𝑡 + 𝛽11𝐷𝑖𝑣𝑖𝑑𝑒𝑛𝑑𝑠𝑖,𝑡+ 𝛽12𝐿𝑜𝑠𝑠𝑖,𝑡+ 𝛽13𝑉𝑜𝑙𝑎𝑡𝑖𝑙𝑖𝑡𝑦𝑖,𝑡 + 𝛽14𝐶𝑎𝑝𝑖𝑡𝑎𝑙 𝐼𝑛𝑡𝑒𝑛𝑠𝑖𝑡𝑦𝑖,𝑡 + 𝜑𝑗+ 𝜈𝑡+ 𝜖𝑖,𝑡
In the additional research section, four more regressions are run to estimate the
investment-Q sensitivity. For this research the same methodology is used as in the Akodgu and MacKay (2008) paper. The regressions are of the following form.
𝐼𝑛𝑣𝑒𝑠𝑡𝑚𝑒𝑛𝑡𝑖,𝑡 = 𝛽0+ 𝛽1𝑇𝑜𝑏𝑖𝑛′𝑠 𝑄𝑖,𝑡+ 𝛽2𝐶𝑎𝑠ℎ 𝐹𝑙𝑜𝑤𝑖,𝑡+ 𝛽3𝐶𝑎𝑠ℎ 𝐻𝑜𝑙𝑑𝑖𝑛𝑔𝑠𝑖,𝑡
+ 𝛽4𝐹𝑖𝑛𝑎𝑛𝑐𝑖𝑎𝑙 𝐿𝑒𝑣𝑒𝑟𝑎𝑔𝑒𝑖,𝑡+ 𝛽5𝐹𝑖𝑟𝑚 𝑆𝑖𝑧𝑒𝑖,𝑡+ 𝛽6𝑉𝑜𝑙𝑎𝑡𝑖𝑙𝑖𝑡𝑦𝑖,𝑡 + 𝛽7𝐶𝑎𝑝𝑖𝑡𝑎𝑙 𝐼𝑛𝑡𝑒𝑛𝑠𝑖𝑡𝑦𝑖,𝑡 + 𝛽8𝐶𝑜𝑚𝑝𝑒𝑡𝑖𝑡𝑖𝑜𝑛𝑖,𝑡+ 𝜑𝑗+ 𝜈𝑡+ 𝜖𝑖,𝑡
B. Variable Construction
The main variables of interest for this research are cash holdings, Tobin’s Q, and financial constraints. Cash holdings are computed as the ratio of cash and marketable securities to total assets (Almeida et al., 2004). Tobin’s Q is calculated as market capitalization plus the book values of debt and preferred stock minus deferred taxes, all divided by total assets (Almeida et al., 2004; MacKay and Phillips, 2005). And a dummy variable is constructed to indicate whether or not a firm is financially constrained. This variable equals 1 if the firm is identified as being in the bottom three deciles of the annual distributions of both payout ratio, common dividends and purchase of common and preferred stock over operating income before depreciation, and firm size, and 0 otherwise (Morellec et al., 2014).
Following Bates et al. (2009) and Opler et al. (1999) a set of standard control variables is included. These variables are constructed as follows. Investment is defined as capital
expenditures in property, plant, and equipment normalized by lagged total assets (Andrade and Stafford, 2004; Minton and Schrand, 1999; Whited, 2006). Cash flow is computed as earnings before interest, taxes, depreciation, and amortization over total assets, and financial leverage as total debt, existing of long-term debt, debt in current liabilities and notes payable, over total assets (Almeida et al., 2004). Firm size is defined as the natural logarithm of net sales. Net working capital is computed by subtracting cash and short-term from working capital and dividing it by total assets. R&D expenditures and acquisitions are both deflated by total assets. Following Morellec et al. (2014) dummies are constructed for dividends and net income. The dividend dummy equals 1 if common dividends are reported, and 0 otherwise. The loss dummy equals 1 if net income is negative, and 0 otherwise.
Like Minton and Schrand (1999) and Andrade and Stafford (2004) do in their studies, two risk controls are added next to the standard controls. These are constructed as follows.
Volatility is proxied by the standard deviation of cash flow, over total assets, but only if there are observations of at least four years. And capital intensity, also called the capital-labor ratio, is measured by fixed capital stock, being property, plant, and equipment, over the number of employees. A full overview of the computations of the variables is presented in Appendix A.
C. Sample Selection
In this research three samples are used, sample A, B, and C. In sample A the financial sector (SIC codes 6000-6999) is excluded because there the motivation for cash holdings comes from legal capital requirements rather than the corporate financing motivations studied in this research. Also the utilities sector (SIC codes 4900-4999) is excluded because there the motivation for cash holdings is subject to regulatory supervision (Bates et al., 2009). The competition measure used in sample B is only available for manufacturing industries. Therefore, sample B exists only of firms in manufacturing industries (SIC codes 2000-3999). In sample C, the same competition measure is used as in sample A, which is available to firms from all industries, but to enable unbiased comparisons between the competition measures used, sample C is also limited to manufacturing industries only.
For all three samples observations with missing data on cash holdings, Tobin’s Q,
investment, or firm size are removed, and the outliers are deleted as follows. Observations with sales or total assets equal or smaller than 0, a negative value for investments, or a value of Tobin’s Q of more than 10 are dropped. Also observations with an asset-normalized cash flow outside the [-2,2] interval, or asset-normalized cash holdings or financial leverage outside the [0,1] interval are deleted (Akdogu and MacKay, 2008). Finally, observations with negative earnings before interest, taxes, depreciation, and amortization, larger than their total assets are dropped (Bris, Koskinen, and Nilsson, 2009).
The samples are further trimmed by winsorizing Tobin’s Q in the 1st percentile,
investment in the 99th percentile, and the following variables in the 1st and 99th percentile: Firm size, net working capital, R&D expenditures, acquisition activity, volatility, and capital intensity.
IV. Data
This section elaborates on the sources of the data, in particular on the sources of the two different variables for the HHIs. It finishes with the presentation and discussion of the summary statistics of the three samples, and a comparison between the panels within the main sample.
A. Data Sources
Through the Wharton Research Data Services website the key accounting data of active and inactive U.S. firms of fiscal years 2002 to 2015 is extracted from the Fundamentals Annual Compustat North America database. This database provides a wide variety of yearly accounting data for all U.S. public firms and all variables used in this paper other than the measures of competition are constructed out of Compustat data. The variables that measure competition are merged in from other, more specialized databases.
The measure used for competition in this research is the Herfindahl-Hirschman Index (HHI). This data is extracted from the Hoberg-Phillips Data Library. Their text based industry classifications are year and firm-specific and are, unlike data used in most literature on cash holdings, not limited to firms competing in manufacturing industries. This database is based on web crawling and text parsing algorithms that process the text in the business descriptions of 10-K annual filings, required comprehensive summaries of firms’ financial performance, on the SEC Edgar website from 1996 to 2015 (Hoberg and Phillips, 2016). The scale of the HHIs, which ranges from 0 to 1, is based on fractions. The common variable on which this dataset is merged with the Compustat dataset is the gvkey, a six-digit code that identifies unique firms. An
important aspect of this research is comparing financial decision patterns of firms competing in industries with different degrees of competition. Akdogu and MacKay (2008) create three groups of industries based on the HHI cutoffs presented in the Horizontal Merger Guidelines by the U.S. Department of Justice and Federal Trade Commission (1997). In 2010 this organization revised these guidelines and determined new cutoff rates. In this research the old values are applied for the observations from 2002 to 2009 and the new ones for 2010 to 2015. The cutoff
rates are downscaled by 10,000, from percentage points to fractions, in order to match the scale of the HHI data. The three groups are identified: unconcentrated (before 2010: HHI under 0.1, from 2010: HHI under 0.15), moderately concentrated (before 2010: HHI between 0.1 and 0.18, from 2010: HHI between 0.15 and 0.25), and highly concentrated (before 2010: HHI above 0.18, from 2010: HHI above 0.25). For better readability these groups will be referred to as competitive, oligopolistic, and monopolistic industries respectively (Akdogu and MacKay, 2008).
To be able to perform robustness checks, a second set of HHI values is used. Different from the firm-specific HHIs from the Hoberg-Phillips Data Library, this measure of market concentration is on industry-level instead of firm-level and is calculated by summing the squared market shares of the 50 largest firms within that market. The HHIs are collected from the Census of Manufacturers, which reports the results of the U.S. Census Bureau’s Economic Census. This economic census appears once every five years and provides an exogenous measure of industry concentration (Akdogu and MacKay, 2008). It is claimed the broadest coverage available, since it includes both private and public firms, whereas in Compustat data, only the latter is present. The HHIs are up to six-digit NAICS industry-specific and this industry classification code also acts as common variable on which the dataset is merged with the Compustat dataset. The numbers from the report of 2002 are used for observations with fiscal years 2002 to 2006, the 2007-report for 2007 to 2011 and the 2012-report for 2012 to 2015. In this database percentage points are used, which means the HHIs range from 0 to 10,000, but the HHIs are downscaled later by factor 10,000. The distribution is of this HHI measure is substantially different from the first one, whereas it is steeper and even more skewed to the right. This makes an interesting comparison, which will be analyzed as a robustness check. The cutoffs used for the determination of the three groups of industry concentration are the same but for the scaling correction: unconcentrated or competitive (old: HHI under 1,000, new: HHI under 1,500), moderately concentrated or oligopolistic (old: HHI between 1,000 and 1,800, new: HHI between 1,500 and 2,500), and highly concentrated or monopolistic (old: HHI above 1,800, new: HHI above 2,500). The distinction before or from 2010 is due to the revision of Horizontal Merger Guidelines by the U.S. Department of Justice and the Federal Trade Commission (2010).
Table 1: Summary Statistics Across Samples A, B, and C
Reported are summary statistics for firms (2002-2015) in samples A (N=12,819), B (N=24,627), and C (N=4,892). Sample A is not limited to manufacturing industries, whereas samples B and C exist of manufacturing firms only. Samples A and C use the year and firm-specific Text-based Network Industry Concentration (TNIC) Herfindahl-Hirschman Index (HHI) coming from the Hoberg-Phillips Data Library as the measure of competition. Sample B uses the HHI coming from the Census of Manufacturers of 2002, 2007, and 2012 as the measure of competition. Until 2009 (from 2010) the cutoff rates 0-999 (0-1,499), 1,000-1,800 (1,500-2,500) and 1,800-10,000 (2,500-10,000) are used to indicate competitive, oligopolistic, or monopolistic industries respectively. Cash holdings are computed as cash and marketable securities over total assets, and Tobin’s Q (the proxy for investment opportunities) as market capitalization plus book values of debt and preferred stock minus deferred taxes divided by total assets. Financial constraint equals 1 if the firm is in the bottom three deciles of the annual distributions of both payout ratio, common dividends and purchase of common and preferred stock over operating income before depreciation, and firm size, and 0 otherwise. The standard control variables are: investment, cash flow, financial leverage, firm size, net working capital, R&D expenditures, acquisitions, and dummies for dividends and losses. The additional risk controls are: volatility and capital intensity.
Sample A N Mean Std. Dev. Minimum Maximum Cash Holdings 12,819 0.193 0.196 0.000 0.999 Tobin’s Q 12,690 1.903 1.519 0.302 9.996 Competition: TNIC HHI 12,819 0.241 0.215 0.021 1.000 Investment 12,691 0.063 0.091 0.000 0.690 Cash Flow 12,819 0.018 0.233 -1.000 1.697 Financial Leverage 12,819 0.214 0.235 0.000 0.995 Firm Size 12,563 5.212 1.939 -1.088 9.574 Net Working Capital 12,428 0.007 0.155 -0.522 0.461 R&D Expenditures 8,180 0.113 0.133 0.000 0.714 Acquisitions 12,499 0.029 0.082 -1.165 0.999 Volatility 10,985 0.103 0.105 0.009 0.781 Capital Intensity 12,256 0.275 0.999 0.001 9.970 Financial Constraint (dummy) 12,819 0.033 0.178 0.000 1.000 Dividends (dummy) 12,819 0.169 0.375 0.000 1.000 Loss (dummy) 12,819 0.459 0.498 0.000 1.000 Sample B Cash Holdings 24,627 0.179 0.194 0.000 0.999 Tobin’s Q 24,381 1.818 1.506 0.269 9.983 Competition: HHI 24,627 0.074 0.058 0.000 0.347 Investment 24,381 0.042 0.046 0.000 0.353 Cash Flow 24,627 0.011 0.242 -1.000 1.697 Financial Leverage 24,627 0.202 0.211 0.000 0.999 Firm Size 24,135 5.042 2.554 -2.096 11.157 Net Working Capital 24,023 0.085 0.168 -0.578 0.539 R&D Expenditures 19,243 0.103 0.125 0.000 0.703 Acquisitions 23,959 0.020 0.067 -1.165 1.028 Volatility 21,989 0.131 0.205 0.010 2.240 Capital Intensity 22,881 0.077 0.120 0.002 1.211 Financial Constraint (dummy) 24,627 0.015 0.122 0.000 1.000 Dividends (dummy) 24,627 0.298 0.457 0.000 1.000 Loss (dummy) 24,627 0.416 0.493 0.000 1.000 Sample C
Cash Holdings 4,892 0.239 0.220 0.000 0.999 Tobin’s Q 4,843 2.111 1.655 0.314 9.838 Competition: TNIC HHI 4,892 0.225 0.214 0.023 1.000 Investment 4,844 0.043 0.053 0.000 0.422 Cash Flow 4,892 -0.056 0.280 -1.000 1.697 Financial Leverage 4,892 0.182 0.213 0.000 0.994 Firm Size 4,795 4.687 2.249 -2.112 9.572 Net Working Capital 4,795 0.049 0.154 -0.449 0.497 R&D Expenditures 4,093 0.156 0.161 0.000 0.796 Acquisitions 4,829 0.020 0.075 -1.165 0.999 Volatility 4,197 0.131 0.131 0.010 0.849 Capital Intensity 4,737 0.077 0.127 0.002 1.165 Financial Constraint (dummy) 4,892 0.024 0.153 0.000 1.000 Dividends (dummy) 4,892 0.137 0.344 0.000 1.000 Loss (dummy) 4,892 0.537 0.499 0.000 1.000
B. Summary Statistics
Table 1 presents the summary statistics for samples A, B, and C. The final samples exist of 12,819, 24,627, and 4,892 unique firm-years respectively. Although samples B and C both exist of manufacturing firms, their sizes are not identical. This is because the firm-specific HHIs are not available for all firms that are in the Compustat dataset, while the alternative HHIs are applied per industry covering all firms in the Compustat dataset.
Differences in the samples are due to two main factors. The first factor is the industries in which the firms operate. Sample A is not limited to firms from manufacturing industries, whereas samples B and C are. The source of the measure of competition used is the second factor. Mean cash holdings in samples A, B, and C are 19.3%, 17.9%, and 23.9% respectively. The means of Tobin’s Q deviate across the samples to a similar extent, as values of 1.903, 1.818, and 2.111 are reported. The most noticeable differences in the variables of interest is found in the HHI. Samples A and C use data on this variable from the Hoberg and Phillips Data Library, whereas sample B uses data from the Census of Manufacturers. Mean HHI in samples A and C is 0.241 and 0.225, substantially higher than the mean value in sample B: 0.074. Furthermore, table 1 shows that investments are over 30% lower for firms in manufacturing industries (sample B: 0.042, sample C: 0.043) compared to firms from a variety of industries (sample A: 0.063). The dummy variable indicating whether a firm is financially constrained equals 1 for 3.3% of the firms in sample A, 1.5% in sample B, and 2.4% in sample C.
To display the differences in distribution between the HHIs from the Hoberg and Phillips Data Library and HHIs from the Census of Manufacturers, two histograms are computed. In the regressions run with sample C the first measure is used, but since this sample also exists of manufacturing firms only, it is possible to collect the values of both measures for firms in this sample. This makes sample C suitable to show the differences between the two measures. As can be seen from figures 1 and 2, both distributions are clearly skewed to the right, but the distribution of figure 2 is most. This explains the fact that in sample B the distribution of firms over the three panels is more unequal, since the same cutoffs are used for categorization. The line running through the figure indicates the normal distribution. In figure 1 this distribution is
clearly flatter than in figure 2, because in figure 1 observations only get fewer in number more to the right, whereas in figure 2 there are no observations at all observed after 3,470.
Figure 1: Distribution of the TNIC HHI Figure 2: Distribution of the HHI
This figure shows the distribution of the TNIC HHIs of the firms in This figure shows the distribution of the industry HHIs of the sample C that come from the Hoberg and Phillips Data Library. The firms in sample C that come from the Census of Manufacturers. y-axis shows the observation density. The scale of the measure is in The y-axis shows the observation density. The scale of the fractions and runs on the x-axis from 0 to 1. Mean value of TNIC HHI measure is in percentage points and runs on the x-axis from 0 to is 0.23 and median value is 0.13. 10,000. Mean value of HHI is 776 and median is 617.
Figure 3 shows the distribution of the firm-years over the three panels for each sample. Due to the differences in HHIs explained above, the panels in samples A, B, and C have very different sizes. In samples A and C, where HHIs from the Hoberg and Phillips Data Library are used, the firm-years are rather evenly distributed over the three panels respectively with 35.5% and 42.6% in competitive industries, 25.1% and 24.4% in oligopolistic industries, and 39.4% and 33.0% in monopolistic industries. In sample B on the other hand, where HHIs from the Census of Manufacturers are used, the firm-years are more unequally distributed over the panels. A whopping 81.8% falls inside the competitive industries, whereas oligopolistic and monopolistic industries only account for 10.7% and 7.5% respectively. However, when these numbers are compared to their equivalents from the paper on investment-Q sensitivity by Akdogu and MacKay (2008), they are accepted more easily, whereas their panel-distribution is 70.9%, 15.1%, and 14.0% respectively. 0 2 4 6 8 Den sity 0 .2 .4 .6 .8 1
HHI from the Hoberg and Phillips Data Library
0 5 .0 e -0 4 .0 0 1 .0 0 1 5 Den sity 0 2000 4000 6000 8000 10000
Figure 3: The Firms’ Level of Competition per Sample
This figure shows the panel distributions, based on the levels of competition, over the three samples. Samples B and C are limited to manufacuring firms, whereas sample A is not. Sample A and C use TNIC HHI, sample B uses HHI. The following cutoff rates are used: HHI<1,000 (<1,500 from 2010) for competitive, 1,000≤HHI≤1,800 (1,500≤HHI≤2,500 from 2010) for oligopolistic, and HHI>1,800 (HHI>2,500 from 2010) for monopolistic industries.
Table 2 presents the summary statistics across the panels in sample A. Mean cash holdings are 20.6%, 19.7%, and 17.9% for firms in competitive, oligopolistic, and monopolistic industries. This reflects a linear relation, where firms in competitive industries hold 4.6% and 15.1% more cash than firms in oligopolistic and monopolistic industries respectively. The values of Tobin’s Q also display a monotonic positive relation with the level of competition.
Investment does too with values of 7.5%, 6%, and 5.6% respectively across the panels, and so does R&D expenditures. Cash flow, financial leverage, and net working capital have a negative, but linear relation with the intensity of competition. Noteworthy are the statistics on firm size and financial constraint. Firm size unexpectedly shows a positive relation to competition. Besides, it appears that firms in competitive industries are least likely to be financially constrained with a mean of 3%, compared to 3.5% and 3.4% for firms in oligopolistic and monopolistic industries respectively. Volatility and capital intensity, both concerning risk, show a positive monotonic relation between competition and uncertainty. Finally, no significant differences are observed in the means of acquisitions, and the dummies indicating a dividend payment or a loss. The summary statistics tables of samples B and C are in Appendix B.
0.00% 10.00% 20.00% 30.00% 40.00% 50.00% 60.00% 70.00% 80.00% 90.00% 100.00%
Competitive Industries Oligopolistic Industries Monopolistic Industries Sample A Sample B Sample C
Table 2: Summary Statistics Sample A
Reported are summary statistics for firms in sample A (N=12,819). Sample A, 2002-2015, is not limited to manufacturing industries and uses the Text-based Network Industry Concentration (TNIC) Herfindahl-Hirschman Index (HHI) coming from Hoberg-Phillips Data Library as the measure of competition. This data is year and firm-specific. Until 2009 (from 2010) the cutoff rates 0-999 (0-1,499), 1,000-1,800 (1,500-2,500) and 1,800-10,000 (2,500-10,000) are used to indicate competitive, oligopolistic, or monopolistic industries respectively. Cash holdings are computed as cash and marketable securities over total assets, and Tobin’s Q (the proxy for investment opportunities) as market capitalization plus book values of debt and preferred stock minus deferred taxes divided by total assets. Financial constraint equals 1 if the firm is in the bottom three deciles of the annual distributions of both payout ratio, common dividends and purchase of common and preferred stock over operating income before depreciation, and firm size, and 0 otherwise. The standard control variables are: investment, cash flow, financial leverage, firm size, net working capital, R&D expenditures, acquisitions, and dummies for dividends and losses. The additional risk controls are: volatility and capital intensity.
Panel 1: Competitive Industries N Mean Std. Dev. Minimum Maximum Cash Holdings 4,555 0.206 0.214 0.000 0.999 Tobin’s Q 4,532 2.030 1.572 0.305 9.838 Competition: HHI 4,555 0.083 0.030 0.021 0.150 Investment 4,469 0.075 0.104 0.000 0.677 Cash Flow 4,555 -0.000 0.260 -1.000 1.697 Financial Leverage 4,555 0.205 0.223 0.000 0.995 Firm Size 4,444 5.281 2.053 -1.058 9.574 Net Working Capital 4,478 -0.004 0.141 -0.522 0.457 R&D Expenditures 2,976 0.145 0.151 0.000 0.714 Acquisitions 4,445 0.030 0.085 -0.203 0.999 Volatility 3,883 0.116 0.117 0.010 0.732 Capital Intensity 4,215 0.501 1.463 0.001 9.948 Financial Constraint (dummy) 4,555 0.030 0.170 0.000 1.000 Dividends (dummy) 4,555 0.185 0.388 0.000 1.000 Loss (dummy) 4,555 0.471 0.499 0.000 1.000 Panel 2: Oligopolistic Industries
Cash Holdings 3,211 0.197 0.188 0.000 0.990 Tobin’s Q 3,177 1.872 1.472 0.302 9.743 Competition: HHI 3,211 0.157 0.038 0.100 0.250 Investment 3,201 0.060 0.084 0.000 0.670 Cash Flow 3,211 0.013 0.225 -0.996 0.655 Financial Leverage 3,211 0.212 0.241 0.000 0.990 Firm Size 3,135 5.261 1.912 -1.013 9.544 Net Working Capital 3,113 -0.000 0.154 -0.521 0.461 R&D Expenditures 2,118 0.126 0.139 0.000 0.703 Acquisitions 3,138 0.028 0.083 -1.165 0.827 Volatility 2,754 0.095 0.099 0.009 0.732 Capital Intensity 3,125 0.175 0.638 0.001 9.970 Financial Constraint (dummy) 3,211 0.035 0.183 0.000 1.000 Dividends (dummy) 3,211 0.158 0.365 0.000 1.000 Loss (dummy) 3,211 0.471 0.499 0.000 1.000 Panel 3: Monopolistic Industries
Cash Holdings 5,053 0.179 0.182 0.000 0.997 Tobin’s Q 4,981 1.807 1.490 0.303 9.996 Competition: HHI 5,053 0.437 0.224 0.180 1.000 Investment 5,021 0.056 0.081 0.000 0.690 Cash Flow 5,053 0.038 0.208 -0.980 1.508 Financial Leverage 5,053 0.223 0.240 0.000 0.994 Firm Size 4,984 5.121 1.847 -1.088 9.570 Net Working Capital 4,837 0.022 0.166 -0.518 0.459 R&D Expenditures 3,086 0.074 0.096 0.000 0.698 Acquisitions 4,916 0.030 0.079 -0.395 0.973 Volatility 4,348 0.095 0.097 0.009 0.781 Capital Intensity 4,916 0.145 0.574 0.001 8.934 Financial Constraint (dummy) 5,053 0.034 0.182 0.000 1.000 Dividends (dummy) 5,053 0.162 0.368 0.000 1.000 Loss (dummy) 5,053 0.440 0.496 0.000 1.000
V. Results
In this section the results of the research are presented and discussed, and the hypotheses stated in the literature review are tested. Then, as an additional research, evidence is sought for the arguments made in explaining the results. Finally, robustness checks are performed by comparing the regression results to those of the two other samples.
A. Regression Analysis
Table 3 reports OLS cash holdings regression results for firms in competitive, oligopolistic, and monopolistic industries. In the fourth specification cash holdings are regressed on the
determinants for the entire sample, irrespective to which panels the firms belong. From here, the first two hypotheses can be answered. HYPOTHESIS 1 reads that the direct effect of
competition on cash holdings is positive. The estimated coefficient of competition as regressor of cash holdings equals -0.026, moderately significant. This implies that the indirect effect of competition through predation risk dominates the direct effect of the precautionary motive, on the basis of which HYPOTHESIS 1 is rejected, at the 5% level. HYPOTHESIS 2 reads that cash
holdings increase in the case of financial constraints. The coefficient answering this hypothesis equals 0.024, weakly significant. This proves the positive relation, leading to an acceptance of HYPOTHESIS 2, at the 10% level. The fourth specification also shows a strongly significant positive
sensitivity of cash holdings to Tobin’s Q. The coefficient of Tobin’s Q equals 0.017, which is consistent with the findings of Bates et al. (2009). A standard deviation growth in Tobin’s Q increases cash holdings with 2.9%. The effects of cash flow and investment on cash holdings are insignificant. The strongly significant negative effects of financial leverage, firm size, net
working capital, and acquisitions, and the positive effect of R&D expenditures are all as expected from the literature. The firm-specific risk factors show ambiguous results. Cash holdings are positively affected by volatility, but also show a negative relation to capital intensity, significant at the 5% and 1% level respectively. The coefficients of the dummy
variables, indicating dividend payments and operational losses, both have the expected negative sign, and are significant at the 5% and 10% level respectively.
In the first three specifications of table 3 the regression results are shown conditional on each of the three levels of competition. With these estimates the validity is tested of HYPOTHESIS
3, which states that the sensitivity of cash holdings to Tobin’s Q decreases as competition increases. The coefficients of Tobin’s Q show a break from the expected linear relation between competition and cash holdings-Q. Comparing firms in competitive and monopolistic industries, the former even show a higher sensitivity than the latter with 0.016 compared to 0.015, both significant at the 1% level. These findings show that the assumption that firms in less
competitive environments are more driven by predation risk to hoard cash to be able to utilize potential investment opportunities does not hold over the entire course of competition. The coefficient of Tobin’s Q for firms in oligopolistic industries, however, equals 0.021, also strongly significant. This indicates a 31.3% and a 40.0% higher responsiveness of cash holdings to
variations in Tobin’s Q than for firms in competitive and monopolistic industries, respectively. Based on these findings HYPOTHESIS 3 is rejected. Inspired by the explanations that Akdogu and
MacKay (2008) give for their similar finding in investment-Q sensitivity, two arguments are put forward, the reactive and the proactive, that might explain the non-monotonic relation of cash holdings and investment opportunities across different levels of competition.
The reactive argument comes from the wait-lose tradeoff of investing, where the value of delaying investment falls if predation risk emerges. Firms in competitive industries, however, might lack the power and control to threaten rivals’ investment opportunities (Haushalter et al., 2007). And predation risk seems less applicable for firms in monopolistic industries as the low competition just gives them time to exercise investment opportunities (Akdogu and MacKay, 2008). The industry structure in between, the oligopoly, might be most suitable for predatory actions, whereas firms are too big to be ignored, and not big enough to ignore (Akdogu and MacKay, 2008). Here the importance of the first mover advantage exceeds the value of waiting when investment opportunities are shared, emphasizing the importance of having access to
funds, which explains the high sensitivity of cash holdings to Tobin’s Q for firms in oligopolistic industries.
Table 3: Cash Holdings Determinants for Different Levels of Competition Sample A: HHIs from the Hoberg-Phillips Database – Not Limited to Manufacturing Industries
OLS regression results for firms in competitive, oligopolistic, and monopolistic industries, not limited to the manufacturing sector (2002-2015). Competition is measured using the Text-based Network Industry Concentration (TNIC) Herfindahl-Hirschman Index (HHI) coming from Hoberg-Phillips Data Library. This data is year and firm-specific. Until 2009 (from 2010) the cutoff rates 0-999 (0-1,499), 1,000-1,800 (1,500-2,500) and 1,800-10,000 (2,500-10,000) are used to indicate competitive, oligopolistic, or monopolistic industries respectively. The dependent variable is cash holdings (cash and marketable securities over total assets). Tobin’s Q (market capitalization plus book values of debt and preferred stock minus deferred taxes divided by total assets) is the proxy for investment opportunities. Financial constraint equals 1 if the firm is in the bottom three deciles of the annual distributions of both payout ratio, common dividends and purchase of common and preferred stock over operating income before depreciation, and firm size, and 0 otherwise. The standard control variables included are: investment, cash flow, financial leverage, firm size, net working capital, R&D expenditures, acquisitions, and the dummies for dividends and losses. The additional risk controls are: volatility and capital intensity. All regressions control for industry (two-digit SIC code) and year fixed effects. Robust standard errors are reported in parentheses. The constants are included in the regressions but are not reported. *, **, and *** indicate significance at the 10%, 5%, and 1% confidence levels, respectively.
Competitive Industries Oligopolistic Industries Monopolistic Industries Entire sample
Panel: 1 2 3
Tobin’s Q 0.016*** 0.021*** 0.015*** 0.017*** (0.003) (0.004) (0.003) (0.001) Competition: TNIC HHI -0.211 0.122 0.011 -0.026**
(0.135) (0.162) (0.014) (0.009) Investment 0.059 -0.041 -0.042 -0.004 (0.076) (0.088) (0.039) (0.039) Cash Flow -0.070* 0.054 -0.058 -0.039 (0.037) (0.049) (0.036) (0.032) Financial Leverage -0.105*** -0.166*** -0.247*** -0.181*** (0.015) (0.021) (0.015) (0.010) Firm Size -0.026*** -0.023*** -0.007* -0.018*** (0.004) (0.004) (0.004) (0.002) Net Working Capital -0.174*** -0.207*** -0.197*** -0.204***
(0.035) (0.038) (0.030) (0.018) R&D Expenditures -0.001 0.188** 0.160** 0.122*** (0.046) (0.068) (0.069) (0.030) Acquisitions -0.223*** -0.234*** -0.258*** -0.229*** (0.051) (0.032) (0.036) (0.025) Volatility 0.183*** -0.006 0.028 0.086** (0.040) (0.060) (0.044) (0.031) Capital Intensity -0.053*** -0.030* -0.024*** -0.027*** (0.017) (0.014) (0.006) (0.006) Financial Constraint 0.012 -0.001 0.067*** 0.024* (0.026) (0.022) (0.019) (0.013) Dividends 0.001 -0.010 -0.034*** -0.016** (0.008) (0.013) (0.009) (0.006) Loss -0.038*** -0.020 -0.032*** -0.032*** (0.009) (0.014) (0.008) (0.005) Year Fixed Effects
Industry Fixed Effects
Yes Yes Yes Yes Yes Yes Yes Yes Observations 2,386 1,655 2,454 6,495 R-squared 0.35 0.31 0.32 0.32
The proactive argument implies that incumbents invest to deter entry and force exit. These strategies are not assigned to firms in competitive industries, since these industries typically have low entry-barriers, and also less to firms in monopolistic industries, since the direct need for these strategies is less just because of the lower level of competition. This makes that this kind of strategic action is also most appropriate to firms in oligopolistic industries. In the next subsection, both arguments are tested by additional research.
Furthermore, the first three specifications of table 3 show that within panels,
competition has no significant effect on cash holdings. For investment and cash flow also no significant coefficients are reported, apart from a weakly significant negative effect of cash flow on cash holdings for firms in competitive industries. The coefficients of financial leverage are all significant at the 1% level across the three panels, and also show a clear linear relation to competition. The effect of leverage on cash holdings is solely negative, but weakens in the level of competition. A standard deviation growth of leverage decreases cash holdings with 5.9% for firms in monopolistic industries, whereas this is only 2.3% for firms in competitive industries. The effect of firm size is negative in all panels and takes a linear course, most negative and strongly significant for firms in competitive industries, less negative and strongly significant for firms in oligopolistic industries, and least negative and weakly significant for firms in
monopolistic industries. The effect of net working capital is negative and strongly significant across the three panels, whereas the effect of R&D expenditures is not significant for firms in competitive industries, but positive and moderately significant for firms in oligopolistic and monopolistic industries. Acquisitions show a strongly significant negative effect on cash holdings, which slightly decreases as competition increases. What stands out about the risk controls is that they are mostly appropriate to firms in competitive industries. As expected, volatility has a strongly significant positive effect on cash holdings for firms in competitive industries, but it does not appear to be significant for firms in more concentrated industries. The coefficient of capital intensity is also strongly significant in the first panel, negative though, and shows significance again in the third panel, although being halved in size compared to the first panel. The effect on cash holdings of being financially constrained is only significant for firms in monopolistic industries, at the 1% level. More specifically, if a firm is financially
constrained, cash holdings increases with 6.7% in that industry. The effect of a dividend
payment is negative and strongly significant, but again only for firms in monopolistic industries. The coefficient of the dummy variable indicating an operational loss is negative, and strongly significant in the first and third panel, while remaining insignificant in the second panel.
The regressions in table 3 include more control variables than the regressions in most existing literature do, increasing the chance of multicollinearity. To check for the presence of this phenomenon the variance inflation factors (VIF) are observed. The rule of thumb is that values of VIF above 5 are alarming, and for the regressions presented in table 3 the VIFs do not come near this number.
B. Additional Research
The reactive argument explaining the high cash holdings-Q sensitivity of firms in oligopolistic industries implies that cash is held to enable firms to rather quickly utilize investment
opportunities when they arise. To prove the validity of this argument, an additional research is carried out. Regressions are run to examine the investment-Q sensitivity of the firms in the three panels.
Table 4 reports the OLS investment regression results for firms in competitive, oligopolistic, and monopolistic industries. Although a different competition measure is used than in the investment regressions of Akdogu and MacKay (2008), the results are quite similar. The coefficients of interest are those of Tobin’s Q, which are all strongly significant. They show that the investment-Q relation is also non-monotonic across different levels of competition. Moreover, for firms in competitive and monopolistic industries the same investment-Q sensitivities of 0.007 are found, whereas for firms in oligopolistic industries the coefficient equals 0.010. This implies that investment of firms in oligopolistic industries is 42.9% more sensitive to changes in investment opportunities than it is of firms in the other industries. These findings are consistent with the Akdogu and MacKay (2008) paper when it comes to the
argument. It shows that for firms in oligopolistic industries the degree of competition has a positive effect on investment, significant at the 1% level, which is evidence that firms react by increasing own investment when the threat of losing investment opportunities increases. For firms in competitive industries, it appears that competition has an opposite effect on
investment, significant at the 5% level. This proves that the reactive argument does not hold for firms in competitive industries and holds even less as the intensity of competition increases, explaining the higher cash holdings-Q sensitivity of firms in oligopolies.
Table 4: Investment Determinants for Different Levels of Competition
Sample A: HHIs from the Hoberg-Phillips Database – Not Limited to Manufacturing Industries
OLS regression results for firms in competitive, oligopolistic, and monopolistic industries, not limited to the manufacturing sector (2002-2015). Competition is measured using the Text-based Network Industry Concentration (TNIC) Herfindahl-Hirschman Index (HHI) coming from Hoberg-Phillips Data Library. This data is year and firm-specific. Until 2009 (from 2010) the cutoff rates 0-999 (0-1,499), 1,000-1,800 (1,500-2,500) and 1,800-10,000 (2,500-10,000) are used to indicate competitive, oligopolistic, or monopolistic industries respectively. The dependent variable is investment (capital expenditures in property, plant, and equipment normalized by lagged total assets). Tobin’s Q (market capitalization plus book values of debt and preferred stock minus deferred taxes divided by total assets) is the proxy for investment opportunities. The standard control variables included are: cash flow, cash holdings, financial leverage, and firm size (defined as the log of total assets). The additional controls measuring risk are: volatility and capital intensity. All regressions control for industry (two-digit SIC code) and year fixed effects. Robust standard errors are reported in parentheses. The constants are included in the regressions but are not reported. *, **, and *** indicate significance at the 10%, 5%, and 1% confidence levels, respectively.
Competitive Industries Oligopolistic Industries Monopolistic Industries Entire Sample
Panel: 1 2 3 Tobin’s Q 0.007*** 0.010*** 0.007*** 0.007*** (0.001) (0.002) (0.001) (0.001) Cash Flow 0.052*** 0.047*** 0.026*** 0.047*** (0.008) (0.008) (0.007) (0.004) Cash Holdings 0.006 -0.006 0.003 0.002 (0.007) (0.009) (0.005) (0.004) Financial Leverage -0.001 0.007 0.001 0.002 (0.006) (0.006) (0.005) (0.003) Firm Size -0.001 -0.004*** 0.001 -0.001* (0.001) (0.001) (0.001) (0.001) Volatility 0.019 -0.031* 0.007 0.002 (0.021) (0.017) (0.016) (0.011) Capital Intensity 0.019*** 0.033*** 0.023*** 0.021*** (0.004) (0.004) (0.005) (0.004) TNIC HHI -0.185** 0.116*** 0.003 0.006 (0.069) (0.034) (0.008) (0.006) Year Fixed Effects
Industry Fixed Effects
Yes Yes Yes Yes Yes Yes Yes Yes Observations 3,541 2,659 4,193 10,393 R-squared 0.46 0.34 0.24 0.35
The proactive argument explaining the high cash holdings-Q sensitivity of firms in oligopolistic industries implies that firms hold cash to be able to invest strategically to deter
entry or force exit when investment opportunities arise, in this paper referred to as strategic hoarding. To prove the validity of this argument, another additional research is carried out. An industry analysis is performed showing the entry and exit rates across the three panels.
In table 5 firms are labeled entrant, incumbent, or exiter. An entrant is a firm that is absent in the first period of the sample (2002-2008), but appears at least once in the second period (2009-2015). The opposite holds for an exiter, and an incumbent is a firm that is present in at least one year in both periods (Akdogu and MacKay, 2008). The proactive argument that strategic hoarding is carried out most by firms in oligopolistic industries is proved when entry and exit rates are observed. The exit rate in panel 2 amounts to 20.1%, whereas for panels 1 and 3 this is only equal to 13.6% and 16.3% respectively. This means that the exit rate of oligopolistic industries is 6.5 and 3.8 percentage points higher than that of competitive and monopolistic industries respectively. The entry rates of panels 2 and 3 are roughly the same with 13.9%, where panel 1 notes a rate of 18.6%. The proactive argument suggests that this 4.7 percentage point difference between oligopolistic and competitive industries is to the credit of strategic hoarding. Taking into account both entry and exit rates, sufficient evidence is provided to adopt the proactive argument.
Table 5: Industry Analysis Sample A – Entry and Exit Rates within Industries
Reported are sample means of cash holdings for firms in sample A (N=12,819). In this sample, 2002-2015, firms are assumed to act in competitive, oligopolistic, or monopolistic industries, based on their firm-specific HHI coming from the Hoberg and Phillips Data Library, and assigned to these industries accordingly. In this table firms are categorized by firm type. Entrants are firms that are present in the sample in 2009-2015, but not in 2002-2008. Exiters are firms that are present in the sample in 2002-2008, but not in 2009-2015. Incumbents are firms that are present at least once in every time period.
Panel 1: Competitive Industries Entrants Incumbents Exiters Total Cash Holdings 0.239 0.187 0.257
Observations 847 3,087 621 4,555 Percent of Subsample Firm-Years 18.59% 67.77% 13.63% 100% Panel 2: Oligopolistic Industries
Cash Holdings 0.200 0.181 0.247
Observations 447 2,120 644 3,211 Percent of Subsample Firm-Years 13.92% 66.02% 20.06% 100% Panel 3: Monopolistic Industries
Cash Holdings 0.166 0.176 0.203
Observations 700 3,532 821 5,053 Percent of Subsample Firm-Years 13.85% 69.90% 16.25% 100% Entire Sample
Cash Holdings 0.205 0.181 0.233
Observations 1,994 8,739 2,086 12,819 Percent of Subsample Firm-Years 15.56% 68.17% 16.27% 100%
C. Robustness Checks
To test the robustness of the regression results with sample A, the same regressions are run for samples B and C, and presented in tables 6 and 7. As described in the data section, the samples differ in sectors included and measure of competition. First the acceptance or rejection of the three hypotheses are reconsidered and then other remarkable differences are discussed.
In samples A and C, where the HHIs from the Hoberg and Phillips Data Library are used, competition has a negative direct effect on cash holdings in the fourth specification, at the 5% and 1% level significant respectively. In sample B, however, where the HHIs from the Census of Manufacturers are used, this coefficient is positive and strongly significant, as expected. This is not enough to adjust the decision to reject HYPOTHESIS 1,however. The effect of financial constraints on cash holdings is not robust across the samples, as it is significantly positive in samples A and B, both at the 5% level, but insignificant in sample C. The decision to accept HYPOTHESIS 2is weakened, but maintained. Noticeable is that in sample A the significance comes from the firms in monopolistic industries, whereas in sample B it comes from the firms in
oligopolistic industries. The non-monotonic relation of cash holdings and Tobin’s Q over the three panels shows to be robust across the samples, with sensitivities of 0.016, 0.021, and 0.015 in sample A, 0.019, 0.024, and 0.017 in sample B, and 0.013, 0.018, and 0.017 in sample C, for firms in competitive, oligopolistic, and monopolistic industries respectively. All these coefficients are significant at the 1% level, irrespective of the sample. This means that firms in oligopolistic industries are 26% to 39% more responsive to deviations in investment
opportunities than firms in competitive industries, and 6% to 41% more than firms in monopolistic industries, which leads to a maintenance of the decision made to reject HYPOTHESIS 3.
Across the samples it shows that several control variables also differ remarkably. The effect of investment on cash holdings is only significant in sample B, negative and at the 1% level, but shows no real significance in the other samples. Cash flow only shows significant coefficients in sample B. Striking is that its overall effect, and the effect for firms in competitive industries is negative, whereas it is positive and statistically and economically significant for firms in monopolistic industries. This implies that firms in competitive industries hoard more
cash in case of disappointing operational results, supporting the theory of precautionary savings. At the same time it proves that cash holdings of monopolistic firms are rather determined by the direction of financial results. The effect of financial leverage on cash holdings is always negative. It is remarkable that in samples A and C this effect decreases as competition increases, whereas in sample B it increases. The signs and significance of the other control variables are mostly similar across the three samples. When the R-squares are
observed, it appears that the regressors in sample B have most explanatory power. The adjusted R-squares would tell the same, since the number of regressors is the same and the sample sizes are of such size that they do not yield any difference.
