Discretionary Investments and Cash Flow Volatility:
Northern European Evidence
Author: Msc. V.E.A. Cuelenaere
Student number: 10824081
Date final version: January 3rd 2017 Supervisor UvA: Dr. J.E. Ligterink
Institute: University of Amsterdam (UvA)
- 1 -
Preface
“In the business world, everyone is paid in two coins: cash and experience. Take the experience first, cash will come later”
Harold S. Geneen
This study is the final part of the two and a half year curriculum of the Executive Master of Finance and Control at the University of Amsterdam. It had been two and a half inspiring, enriching and above all fun years. I can truly claim that I can connect the dots, within the world of finance and control, much better than when I did at day one of this journey.
With this study I hope to provide the academic world, as well as the world of controllers and CFOs, additional and relevant insights on the crucial role cash and cash volatility plays in the tumultuous times we face today. The quote “cash is king” is more meaningful than ever before. This simultaneously increases the need for better cash management and risk management for companies in the Netherlands and Northern Europe, and with this study I hope to make a contribution.
Foremost I would like to gratefully thank my supervisor Jeroen Ligterink for his supervision, guidance and help during this interesting, but bumpy road to the finish.
Last but surely not least, I would like to thank my mom, fiancé, family and friends for their patience and the sacrifices they have made to make this enriching journey possible! A sincere and big thanks to you all!
I hope you enjoy reading this thesis!
Vincent Cuelenaere Amsterdam, January 3rd 2017
- 2 -
Executive summary
This study researches the influence of cash flow volatility on investment in capital expenditure, research & development and selling, general and administrative expenses, irrespective of the industry firms are in. This study is conducted based on 1.212 Northern European firms over a period of twelve years, from the start of 2003 until the end of 2014. Additional analysis attempts to provide insights for CFOs and controllers with respect to the influence of the level of cash holdings, the impact of recent economic recession and the influence of financial systems on the relationship between cash flow volatility and investments.
The results of this study provide direct evidence that cash flow volatility is associated with lower levels of investment in capital expenditures in Northern Europe, irrespective of the industry firms are in. No robust and significant evidence is found regarding a relationship between cash flow volatility and research & development investment and between cash flow volatility and selling, general and administrative expenses.
Further analysis shows that the economic recession around 2008 doubled the negative relationship between cash flow volatility and capital expenditure investment. This might suggest that the economic recession increased the number of cash flow shortfalls at firms, resulting in double the amount of foregone capital expenditure investments as a % of total assets. Firms did not seem to anticipate this risk of higher cash flow volatility through hedging or by increasing saving to avoid reducing capital expenditure investment. Evidence also showed that the level of cash holding is a significant influence on the level of capital expenditure investment. This seems plausible as the level of cash holdings has a significant influence on the probability of cash flow shortfalls and therefore on the probability firms foregoing capital expenditure investment. Regarding financial systems, the Anglo-Saxon system seem to have an additional negative effect on the relation between cash flow volatility and capital expenditure investments, compared to financial systems in mainland Northern Europe. Theory explains this by stating that a market-oriented system, like the Anglo-Saxon system, is less efficient in dealing with information asymmetry between lenders and borrowers due to the less open and transparent nature of their agreements and relationship. As we know information asymmetry might increase the costs of financing as a result of a risk premium charged due to higher risks incurred by the lender. In a situation of higher cash flow volatility and hence a higher probability of cash flow shortfalls, this additional external capital constraint, might in theory result in more foregone capital expenditure investments, ceteris paribus. Theory suggest that a more relationship-oriented system like in mainland Europe is more efficient in dealing with this information asymmetry, due to the more open and closer relationship between lenders and borrowers.
Regarding research & development investments, this study shows that before the credit crisis a negative relationship was found, however the years after 2008 did not show any significant relationship. This might indicate that, considering the strategic importance of research & development plans, managers engaged more and better in cash
- 3 - and risk management techniques to protect research & development investment against negative influences like higher cash flow volatility, in anticipation of difficult times ahead. Regarding the financial systems in Northern Europe, the Anglo-Saxon system in the UK showed a higher negative relationship between cash flow volatility and research & development investments compared to its counterparts in mainland Europe. This is consistent with the theory that a market-oriented system like the Anglo-Saxon system is less efficient in dealing with information asymmetry between lenders and borrowers.
Concerning selling, general and administrative expenses, a positive correlation was found in this study, but after controlling for five investment explaining control variables, the relationship does not seems to be significant. This indicates that selling, general and administrative expenses do not appear to be a valid proxy for investments in the context of this study.
Taken as a whole, the results of this study suggest that Northern European firms do not completely smooth out cash flow volatility through time to maintain a desired level of capital expenditure, but instead forgo some capital expenditure investment. Cash flow volatility risks seem to be mitigated better when it comes to research & development expenses, especially during the credit crisis. Firms in the UK still show signs of research & development expenses being negatively influenced by cash flow volatility risk. Further research is recommended with regards to the relationship between selling, general and administrative expenses and cash flow volatility in Northern Europe. I would suggest that including advertising expenses over selling, general and administrative expenses as an investment proxy in the regression models could be a fruitful area of future research.
This thesis does not claim that the results of this study imply that firms should definitively aim to reduce cash flow volatility. Rather, firms should assess the costs and benefits of applying risk management techniques to reduce cash flow volatility on a case by case basis. This thesis solely aims at providing additional insights for CFOs and controllers that can contribute to their decision making process.
- 4 -
Index
1. INTRODUCTION ... -5-
2. LITERATURE REVIEW ... -9-
2.1. CASH FLOW VOLATILITY AND INVESTMENTS ... - 9 -
2.2. CASH FLOW VOLATILITY AND CASH HOLDINGS ... - 13 -
2.3. CASH FLOW VOLATILITY AND FINANCIAL SYSTEMS ... -14-
2.4. HYPOTHESIS ... - 15 -
3. RESEARCH METHOD ... -17-
3.1. RESEARCH METHOD ... - 17 -
3.2. DATA ... - 18 -
3.3. MEASURES AND DEFINITIONS ... - 18 -
3.3.1. Cash flow volatility measure ... - 18 -
3.3.2. Proxies for discretionary investment ... - 19 -
3.3.3. Control variables ... - 20 -
3.4. CORRELATIONS AND REGRESSIONS ... - 21 -
3.4.1. Correlations ... - 21 -
3.4.2. Main model regression ... - 22 -
3.4.3. regression models with an interactive variable ... - 22 -
4. RESULTS ... -24-
4.1. DESCRIPTIVE STATISTICS ... - 24 -
4.2. CORRELATIONS ... - 25 -
4.3. REGRESSION ... - 27 -
4.3.1. Main regression models ... - 27 -
4.4. REGRESSION MODELS WITH AN INTERACTIVE VARIABLE ... - 28 -
4.4.1. Impact of Cash holdings ... - 29 -
4.4.2. Impact of the global Credit crisis ... - 30 -
4.4.3. impact of a financial system ... - 31 -
5. CONCLUSIONS AND RECOMMENDATIONS FOR FURTHER RESEARCH ... -34-
5.1. CONCLUSIONS ... - 34 -
5.2. WORD OF CAUTION AND ADVICE ... - 36 -
5.3. LIMITATIONS AND RECOMMENDATIONS FOR FURTHER RESEARCH... - 37 -
6. REFERENCES ... -38-
- 5 -
1. Introduction
How much strategic attention should companies pay to controlling incoming and outgoing cash flows? Some firms engage in costly risk management techniques like hedging, and some do not. This will probably have a lot to do with the level of cash flow volatility and the risk of cash flow shortfalls which is higher for some companies than for others. Higher cash flow volatility implies that a company has a higher chance of periods with internal cash flow shortfalls. These cash flow shortfalls may throw budgets into disarray, it might delay debt repayments, bring bank covenants to alarming levels and it might even defer discretionary investments. It is this latter effect that has been researched extensively in the current literature. This is understandable, as the concept that cash flow volatility impacts investments is very appealing and relatively simple, and furthermore discretionary investments are in themselves probably the single most important determinant of economic growth for firms. It is therefore this effect that will be elaborated upon in this thesis. It is an effect that is of utmost interest for professionals working in the world of financial control, including myself. Certainly during the global credit and economic crisis hit the world economy, cash flow volatility and its consequences, has become a more important subject for study than in previous decades. For years various academic papers claimed there was a relationship between cash flow volatility and investment, but it was Minton and Schrand (1999) who provided the first direct Northern American evidence that cash flow volatility and investments are negatively related to one another. Minton and Schrand (1999) showed that higher cash flow volatility is associated with lower average levels of investment in capital expenditures, research & development expenses and advertising expenses. They argued that companies do not simply react to these cash flow shortfalls by deferring discretionary investments to match their actual cash flows, but rather forgo investments. This is an interesting finding and is fundamentally important if we believe investment is the single most important determinant of economic growth for firms.
A crucial determinant of this basic finding by Minton and Schrand (1999) is the level of financial constraints firms might have. One could argue that companies smoothen their internal cash flow shortfalls in order to smoothen their investment spending by attracting external capital when needed. Assuming that external capital is freely available and at the same costs the same as internally generated cash flows, one would suggest that there would not be any effect on the investment pattern of a firm. However, Myers and Majluf (1984) show us that external capital is more costly than internally generated capital. Consequently, when following the basic net present value (NPV) decision rule for capital budgeting, it shows us that companies which require more (costly) external capital relative to internal capital will have lower investment levels due to a higher NPV required. The NPV criteria is increased with a hurdle rate which should make up for the additional costs of this external capital. This does imply that financial constraints are a crucial determinant of the relationship between cash flow volatility of internally generated cash flows and investment. Two interesting factors that influence financial constraints are the level of cash holding and the financial system a firm operates in.
- 6 - The level of cash holdings is a large determinant of whether companies are financially constrained or not. Companies who have a large cash holding will always be able to make the first-best investment whenever needed. This, obviously, applies to a much lesser extend to companies with a relatively low cash holding, who often rely on external capital as their internal capital might not be sufficient for the first-best investment. As we know external capital is not freely available and not at the same cost as internally generated cash flows (Myers and Majluf, 1984). Cash rich firms are less dependent on this more expensive external capital to make investments and hence their investment pattern is assumed to be less sensitive for the volatility of internal cash flows. This implies that cash rich firms might experience a weaker negative relationship between the volatility of the internally generated cash flows and investments. Firms with smaller cash holdings have limited buffer to respond to cash flow shortfalls and hence are expected to have a stronger negative relationship between cash flow volatility and investment. This study will research this potential influence of cash holdings by splitting the sample into cash rich companies (relative to their own total assets) and cash poor companies (relative to their own total assets) and comparing the results.
On the second factor, the financial system can be a determinant of the level of financial constraints of firms. The concept that financial systems have a large impact on real economic activities like investment is not new (see the work of Irving Fisher and to a lesser extend the work of John Keynes (see Gertler, 1988 for an overview)). The financial system, and in particular certain characteristics of a system concerning the accessibility of external capital can be a large influence on how firms finance themselves. Aspects like the size of secondary markets (i.e. equity and corporate bond markets) in a system or the (dominant) role banks have in providing external capital to the market have a profound influence on the accessibility of external capital. In that sense, financial systems in Northern Europe can be split into a system with strong Anglo-Saxon influence and a financial system with strong Rhineland and Nordic influence. The Anglo-Saxon system is known for having a ‘free-market’ with a large secondary market as a source of finance. One would assume that in a more market-oriented financial system it is slightly more difficult for companies to get access to external capital due to the simple reason that transactions are more at arm’s length. This indicates that the nature of the relationship between lender and borrower is more transaction focussed and probably less open and transparent. This, in turn, implies that there would be a higher level of information asymmetry between lender and borrower. As we know information asymmetry might increase the costs of financing as a result of a risk premium charged due to higher risks incurred by the lender. This market-oriented system is very much in line with the American financial system and will be represented in this study by the United Kingdom. On the other hand there is the more mainland European system1, where banks are usually the primary source of finance and have a relatively smaller secondary market compared to the Anglo-Saxon system. This relationship-oriented system in mainland Europe characterises itself more by the strong relationship companies have with their finance providers like banks. This strong relationship between lenders and borrowers indicates a more open and transparent way of lending and
- 7 - borrowing which implies that there should be lower information asymmetry. This presumably makes it slightly easier and cheaper for firms to access external capital and therefore easier to mitigate cash flow volatility risks by relying on external capital to bridge periods of cash flow shortfalls. This system will be represented in this study by Germany, The Netherlands, Belgium, Luxemburg, Denmark, Finland, Norway and Sweden. In this study we will try to find statistical proof for this hypothesis.
Whereas Minton and Schrand (1999) conducted their Northern American research in the 1990s, this research will focus on Northern European evidence and will include the interesting effect of the credit crisis on this relationship. I hypothesise that the findings of Minton and Schrand in Northern America will also apply to Northern European companies and that the recent credit crisis potentially strengthened this negative relationship. Presumably, the credit crisis had a substantial negative impact on the availability of external capital. The harder it gets for companies to access external capital in order to compensate for their internal cash shortfalls, the stronger the negative relationship between cash flow volatility and investment should be.
The absence of a negative relation in Northern Europe would suggest that either a) the investments of Northern European firms are not at all sensitive to fluctuations in cash flows or b) that the European companies, operating two decades later, have matured and have a much higher level of control over their investment pattern and have a much higher level of control over their internally generated cash flows. Frankly, I do not see any economic explanation for the first argument. I would even hypothesise the opposite was more likely, as on average American companies are cash richer compared to their European counterparts, respectively 17% compared to 10%-15% (Uyar and Kuzey, 2014). Richer companies should be presumably less prone to fluctuations of internal cash flows, as hypothesised before. As far as the second argument is considered, an economic explanation could be that due to the time difference of two decades between the study of Minton and Schrand (1999) and this study, firms might have matured and indeed have a substantial better investment and cash flow planning. The average firm today is in better control of their cash flows and investments due to technological advancements, like financial planning systems, treasury systems and ERP systems. The CFO of today has substantially better tools and availability of (real-time) data to control cash flows and investments over a shorter period of time. The better a firm is at financial control the higher the probability that investments can be planned and the lower the probability that unforeseen cash flow shortfalls will lead to foregone investments. In that case, one would not expect a significant negative relationship between cash flow volatility and investments.
The objective of this thesis is to provide CFOs and controllers with relevant insights into the relationship between cash flow volatility and discretionary investments. The recent credit crisis increased the relevance of cash flow and cash flow volatility insights tremendously. Besides providing insights on how the cash flow volatility relationship with
- 8 - investments behaved through the current credit crisis, additional insights will be given regarding the level of cash holdings and the financial system in which the firms operate.
With this objective, this thesis proceeds as follows. Chapter 2 will elaborate on the status of the current academic discussion on this subject and will provide the hypothesis of this paper. In chapter 3, the research method will be discussed including the conceptual framework, definitions and regression models used in this study. Chapter 4 will touch upon the descriptive statistics before giving an overview of the results of this empirical research. Finally, chapter 5 will provide the conclusions and practical implications of the results of this thesis, including limitations and recommendations for further research.
- 9 -
2. Literature review
Before the empirical research starts and the findings of Minton and Schrand (1999) will be tested on Northern European data, the current body of literature will be touched upon in this chapter. The first paragraph of this chapter will elaborate on general theories regarding cash flow volatility and investments. The second paragraph will elaborate on the influence of cash holdings on a potential relationship between cash flow volatility and investments as discussed in the introduction. The third paragraph will further research current literature concerning the theoretical and empirical results of the influence of differences in financial system on the relationship between cash flow volatility and investments.
2.1. Cash flow volatility and investments
There is a large literature that estimates the impact of cash flow volatility on the investment behaviour of firms, beginning with the classical paper of Fazzari et al. (1988). They classify firms according to whether they were likely to be financially constrained based on company size, capital structure and dividend pay-outs and whether this characteristic (being financially constrained, or not) determines if investments are more sensitive to the supply of internal generated cash flows. In other words, whether investments are related to cash flow volatility. They find that the highest sensitivities to cash flows are found for firms categorized as financially constrained. Therefore they conclude that the sensitivity of investment to cash flow volatility is related to the wedge between internal and external financing costs. Hence, if external financing is more costly than internally generated funds, investments of financially constrained firms are more dependent on internal cash flows than the investments of unconstrained firms.
Later literature raised new objections to this approach. Kaplan and Zingales (1997, 2000) argue that Fazzari et al. (1988) tend to assign firms incorrectly in their approach. Kaplan and Zingales (1997, 2000) demonstrate that the link between cash flow volatility and financial constraints is sensitive for the classification of constrained firms. In other words, if one makes a different classification for constrained and unconstrained firms, the results can be opposite. In their study, classification based on multiple qualitative and quantitative criteria yields higher cash flow sensitivity estimates for firms that are perceived as least likely to be financially constrained. Never the less, the direction of the relation between cash flow volatility and investment is still negative.
Decades after these contributions and despite the extensive research done in recent years on the influence of financial constraints on the relation between cash flow volatility and investment patterns of companies, there is still no consensus on this topic. Some authors have tried to settle the debate by rejecting or confirming findings on either side, other researchers tried to provide the academic world with explanations for the conflicting results in the literature by focussing on small differences between both sides. It is the simple rationale behind Fazzari et al. (1988) their theory that continue to initiate a large number of contributions. Often using a different definition of financial
- 10 - constraints, but broadly confirming their results that there is a stronger relation between cash flow volatility and investments for financially constrained firms.
Other classical work regarding the relation between cash flow volatility and investments is done by Froot, Scharfstein and Stein (1993). They propose a negative relationship between cash flow volatility and investment in their search to explain why firms engage in costly hedging strategies. They argue that not hedging at all might in some cases lead to an increase in volatility of internally generated cash flows2. This increase in volatility implies that the company is exposed to a higher risk of cash flow shortfalls. They claim that cash flow shortfalls must eventually result in either a) a higher variability in funds generated externally, or b) a variability in current or future investments to be made. Assuming there are no market imperfections and that the cost of capital is independent of the debt-equity ratio as demonstrated in the classic paper of Modigliani and Miller in 1958, the first option seems the most appealing. A higher variability in investments is highly undesirable as long as there are positive NPV investment opportunities. However, we do know that there are market imperfections like taxes, transaction costs, costs of financial distress, subsidies, etc. The cost of capital is thus dependent on the debt-to-equity ratio and external capital might be more costly compared to internal capital based on this Modigliani and Miller theory. In addition, Myers and Majluf (1984) show us that external capital is more expensive compared to internally generated cash flows due to the existence of market imperfections, like information asymmetry. Hereby, this first scenario will lose its attractiveness. Now the second scenario is more likely to happen, or at least a combination of both scenario’s. This implies that a shortfall in cash increases external financing as well as decreases investments.
Fazzari et al. (1988), Froot, Scharfstein and Stein (1993) and Kaplan and Zingales (1997) find a negative contemporaneous relation between annual investment levels and liquidity. These studies however, cannot distinguish whether firms with volatile cash flows time their investment decisions to match internal cash flow realisations or actually decrease their overall investment levels. It where Minton and Schrand with their paper in 1999 providing direct evidence for the first time that cash flow volatility is related to lower investment and not vice versa. Minton and Schrand (1999) their most fundamental finding of their analysis is that higher cash flow volatility is related to lower investments in average annual capital expenditures, research & development and advertising expenses, even after adjusting for the industry and controlling for the level of a firm’s average cash flow and for its average asset growth. They argue that a higher cash flow volatility implies that there is a higher likelihood that companies experience internal cash flow shortfalls. They find that these cash flow shortfalls do not cause a delay in investments, rather companies forgo investments.
2 This depends on the correlation of the risk factor with the investment. In some cases this statement might not hold, for example for firms in the oil & gas industry where returns of investments depend on the oil price, but at the same time where the investments to be made are also determined by that same oil price. In this case hedging does not lead to a lower cash flow volatility.
- 11 - Their main interpretations of their results is that cash flow volatility, on average, leads to lower investment. However, an alternative explanation could be that different levels of investments lead to different volatilities due to the nature of the investments. To some degree, indeed, investment pattern may determine the nature of the cash flows. However, this concern is partly mitigated by providing evidence in their paper that investment-cash flow volatility sensitivities are related to the costs of accessing external capital. One would not expect this relation if investments determine cash flow volatility. In addition, Minton and Schrand (1999) conducted various sensitivity analysis to examine the causality of the relation between cash flow volatility and discretionary investments. Even though no conclusive evidence regarding this causal relationship can be provided and endogeneity will always be a reason of caution, they did provided additional results that are consistent with their interpretations. First, they show that cash flow volatility is not highly correlated with proxies for growth. One would expect a significant and positive correlation if investments determine cash flow volatility, for the simple reason that investments are likely to be positively correlated with growth3. Second, they show that the cash flow volatility is positively and significantly related to investment volatility across all cash flow levels. One would expect this positive relation if cash flow volatility leads to lower investments. If different levels of investment produce different volatilities, one would expect no association between volatility of investment and cash flow volatility. Thirdly, they show that earnings volatility is not related to average investment levels and that the inclusion of earnings volatility does not change the negative relation between cash flow volatility and investment. One would expect that if investments influence cash flows, they would also influence earnings and hence that there should be a significant relation between earnings volatility and average investment levels. In sum, these three additional analysis are consistent with their interpretation that cash flow volatility is related to lower average investments because it measures the incidence of cash flow shortfalls. Besides suggesting the causality of the relation with well-reasoned evidence, Minton and Schrand (1999) provide evidence for a significant magnitude of the negative relation they suggest. For example, they show that capital expenditures by firms with high cash flow volatility (in the highest quartile) are 19% below the mean level of capital expenditures. Companies classified as having a low cash flow volatility (in the lowest quartile) are 11% above the mean. This means that firms with a high cash flow volatility invest 30% less compared to firms with a low cash flow volatility.
Also in more recent literature evidence can be found in favour of the findings of Minton and Schrand and in line with Fazarri et al. (1988). Even though there is limited direct empirical evidence in more recent literature, the negative relation is regularly suggested in other papers. For example, Han and Qiu suggest a similar negative relation between cash flow volatility and investment in their study in 2007. Their model predicts that future cash flow volatility has a negative impact on current investments, while it has a positive impact on current cash holdings. This specifically
- 12 - holds for financially constrained firms. They explain this by suggesting that there is an intertemporal trade-off between current and future investments. They suggest that when marginal returns on investments are convex, an increase in future cash flow volatility makes the expected marginal return on future investments higher for given cash holdings. So for constrained firms, an increase in future cash flow volatility makes them more prudent, to increase cash holdings for more future investment by decreasing current investment.
Later evidence in line with Kaplan and Zingales (1997) are the findings of Cleary (2006). Using international panel data for companies from seven of the world’s largest economies, Cleary demonstrated that investments of firms experiencing a higher cash flow volatility are less sensitive to a given period’s internal cash flow. This implies that the relation between cash flow volatility and investment is not more negative for firms with a high cash flow volatility compared to firms with a low cash flow volatility. Cleary (2006) argues that firms with a higher cash flow volatility simply will build up more financial slack anticipating for these cash flow volatilities. Also Mulier et al. (2013a) follows the line of reasoning of Cleary (2006). Based on a large European sample set, they state that for a given cash flow shock, it is more likely that this signals firms with low cash flow volatility as the probability is higher for them that this shock falls outside their normal cash flow fluctuations. It is more likely that those firms adjust their future cash flows (i.e. reduce their investments) after receiving such a signal. On the contrary, firms with a higher cash flow volatility are less likely to receive this cash flow shock signal and hence are less likely to amend their investment pattern. Even though both studies (Cleary, 2006 and Mulier et al. 2013a) elaborate on the sensitivity of investments to cash flow volatility, Minton and Schrand (1999) researches the direction of the relationship, which is not the same. Yet their line of reasoning suggests that firms with a higher cash flow volatility do not decrease their investment pattern, rather firms with low cash flow volatilities do so. This is contrary to the line of reasoning of Minton and Schrand (1999).
In summary, a large part of the academic discussion around the relation between cash flow volatility and investment is not directly concerning the direction of the relationship, moreover about the sensitivity of the relationship. Irrespective of the sensitivity of the relation between cash flow volatility and investment, at least both schools of thought (with the founding fathers: Fazarri et al. (1988) and Kaplan and Zingales (1997)) are in agreement that there is a certain relation between cash flow volatility and investments, were the majority of scholars indicates this relation to be negative. In addition, both schools of thought are in agreement that the level of financial constraints plays a crucial part in this relation. In this light, two interesting factors could be further elaborated on when explaining the relation between cash flow volatility and investment in the light of financial constraints. Both the level of cash holdings as well as the financial system, might be of influence on the level of financial constraint and hence on the relation between cash flow volatility and investment. The following paragraphs will elaborate on these factors.
- 13 -
2.2. Cash flow volatility and cash holdings
An important factor discussed in length of literature that impacts the relationship between cash flow volatility and investments is the holding of cash and cash equivalents by firms. Before we elaborate on the influence of this factor an important question is: why do firms hold such substantial percentages of cash and cash equivalents in their assets even though there is an opportunity cost associated with cash and cash equivalent? In literature percentages of cash holdings by firms vary quite substantially, but on average Northern European findings vary between 10% and 15%4. According to Keynes (1936), there are two major benefits to cash holdings. First, a firm can save transaction costs by using cash to make payments without having to liquidating assets. This suggests that firms are likely to increase their cash balances when the cost of raising external funds are higher. Second, and possibly more important with regards to this study, a firm can reserve cash to hedge for the risk of future cash shortfalls, this is referred to as the precautionary motive or hedging motive for cash holdings. The essence of the precautionary motive of cash holdings is best described by Myers and Majluf (1984), whom define cash on hand and marketable securities as “financial slack which could be used to overcome the problem of financial constraints and to meet the need for future investment expenditures”. This definition implies that the precautionary motive for cash holdings does not apply for financially unconstrained companies.
Recent evidence regarding the relation between cash flow volatility and investments with respect to cash holdings is provided by Han and Qiu (2007). They try to explain the influence of this factor by distinguishing between financially constrained and unconstrained firms. According to Han and Qiu (2007), a firm is financially unconstrained if it has enough financing capacity (internal and external) to make the first-best investment, regardless of the realization of future cash flow. For these companies there is not a real incentive to precautionary hold additional cash. However, financially constrained firms cannot make additional future investment without reducing current investments because it has exhausted till a certain extend their external financing resources. Therefore, a firm can invest more in the future only by reducing current investments (Han and Qiu, 2007). This is the intertemporal trade-off as explained before. Companies build up a certain cash holding by reducing current investment in order to be able to make more future investments. They will do so by assuming that marginal returns of investments follow a convex, an increase in future cash flow volatility makes the expected marginal return on investments in the future higher for a certain cash holding. This assumes a strong link between cash holdings and cash flow volatility influencing investments for financially constrained firms.
In line with Han and Qiu (2007) Belghitar and Khan (2011) find similar results for UK firms. They show that firms with greater cash flow volatility hold more cash on their balance sheet. Their line of reasoning is similar that this is
4 Ferreira and Vilela (2004) stating average cash holdings of 14,8% for European companies. Ozkan and Ozkan (2004) find an average of 10% cash
- 14 - resulting from the precautionary motive firms have to hold cash. The higher the uncertainty due to a high cash flow volatility the higher the need to be precautious, the higher their cash holdings.
Contrary to the logic of Han and Qui (2007) and to the hypothesis of this thesis is the logic of the outcome of research by Kim et al. (1998). Kim et al. (1998) researched the optimal amount of cash holding when external financing is costly (i.e. for financially constrained firms). They define the optimal amount of cash holdings as a trade-off between low return earned on liquid assets and the benefits of minimizing the need for costly external financing. Their model, based on a large US sample set, predicts that the optimal cash holding is increasing in the cost of external financing, the volatility of future cash flows, and the return on future investment opportunities. This implies that a) cash flow volatility is positively related to cash holdings, and b) that cash holdings are positively related to investments. This is in sharp difference with Minton and Schrand (1999).
In summary, a lot of the findings and discussions regarding the role of cash holdings in the relation between cash flow volatility and investments concerns the level of being financially constrained. This thesis will not split firms into financially constrained and financially unconstrained firms, but will follow the same logic and test whether higher (lower) cash holdings lead to a lower (higher) negative relationship between cash flow volatility and investments.
2.3. Cash flow volatility and financial systems
The second factor researched in this thesis is the role of financial systems in the relation between cash flow volatility and investments. The concept that financial systems in countries have a large impact on real economic activities like investments is not new. This is easily traceable to the time of the great depression in the beginning 1930’s (see the work of Irving Fisher and to a lesser extend the work of John Keynes (see Gertler, 1988 for an overview)). Various more recent research has been conducted on the impact financial systems have on financial constraints. As discussed before, these financial constraints have influences on the level of investments firms are able to make. Important research on this topic is based on the work of Bond et al. (2003). Bond et al. (2003) researched the relation between cash flow volatility and investments for four European countries, being United Kingdom, Germany, Belgium and France. They show that investments of UK firms appear to be more sensitive to cash flow volatility than investments in the three continental European countries, in particular Germany. Their explanation is that market-oriented systems like the UK, where secondary markets like equity markets, corporate bond markets and commercial paper markets are a far larger source of funding. These secondary markets as suppliers of finance have a much more arm’s length relation with firms. These market-oriented systems tend to be less effective in dealing with asymmetric information problems. These asymmetric information problems are the commonly suggested cause of higher levels of financial constraints on investments and make external capital more expensive due to a risk premium. A financial system like in other parts of Northern Europe, is characterized by a more relation-oriented market where firms and suppliers of finance have much more transparent and closer arrangements that allows the suppliers of finance to
- 15 - exercise greater scrutiny over borrowers. This relation-oriented markets tend to be more effective in dealing with asymmetric information between firms and suppliers of finance and hence
Mizen and Vermeulen (2005) proceeded on this research of Bond et al. (2003) and studied the root cause of this difference in cash flow volatility and investment relation between UK and German firms. Mizen and Vermeulen (2005) confirm the results of Bond et al. (2003) that UK firms are more sensitive to cash flow variability when it comes to making investment decisions. They confirm the explanation provided by Bond et al. (2003) that market-oriented financial systems, with larger secondary markets, are more likely to show greater volatility of cash flow when it comes to investments due to the nature of the relation between lender and borrower. Relationship-oriented financial systems show a smaller sensitivity of cash flow volatility to investments.
In summary, current literature suggests for this thesis that the more market-oriented financial systems like in the UK will show a higher negative relation between cash flow volatility and investments compared to the relation-oriented financial systems like in mainland Europe. This due to the types of primary financial sources per system and the level of efficiency they obtain in dealing with asymmetric information between the lender and the borrower.
2.4. Hypothesis
Based on this literature regarding the relationship between cash flow volatility and investments, this thesis will attempt to enrich this body of literature with Northern European evidence. In addition this thesis will add to this by comparing the difference of impact of several factors influencing this relationship, like cash holdings, the credit crisis and financial systems. The central hypothesis of this thesis is:
H1: There is an negative relation between volatility of operational cash flows and discretionary investments for Northern European firms
In order to answer this main hypothesis (H1), first bivariate Pearson correlations will be ran followed by panel data regressions5. These regression models will provide insights in the strength of the (possible) relation between cash flow volatility and investments.
After these main regression models are ran5, the results will be tested on robustness and additional insights will be given by conducting three additional regression models5. These additional regression models will include an interactive variable per model which will a) test whether the results of the main regression model still hold and b) provide insights in the influence of factors potentially influencing on the relationship between cash flow volatility
- 16 - and investments. The first regression model with an interactive variable concerns the influence of cash holdings. Based on the literature in paragraph 2.2, I hypothesize that companies with relative lower (higher) cash levels will experience a higher (lower) impact of cash flow volatility on their discretionary investment spending. This brings us to the second hypothesis:
H2: The negative relation between cash flow volatility and discretionary investments is higher (lower) for firms with lower (higher) cash levels
Closely related to cash holdings is the impact the credit crisis had on the investment pattern of Northern European firms. Although not discussed in literature due to the simple reason that the majority of the research in this field is conducted before the 2008 credit crisis and hence, simply could not research it. Conducting this research in 2016, I cannot neglect potential effects this historical recession. I hypothesis that investments are more (less) negatively related to cash flow volatility during the credit crisis compared to the period before (before). This leads to the third hypothesis:
H3: There negative relation between cash flow volatility and discretionary investments for the period after 2008 is higher compared to an equal period before
Finally, based on the literature in paragraph 2.3, this study will test whether there is a difference in relationship between cash flow volatility and investments among two financial systems in Northern Europe. Based on the assumption by Bond et al. (2003), the more market-oriented Anglo-Saxon financial system will probably be less efficient in dealing with information asymmetry compared to the more relationship-oriented financial systems in mainland Europe, I hypothesize that cash flows of companies in the Anglo-Saxon system will be more negatively related to investments compared to countries in mainland Europe. This brings the fourth and final hypothesis:
H4: There negative relation between cash flow volatility and discretionary investments for firms operating in Anglo-Saxon models is higher compared to companies in mainland Europe.
- 17 -
3. Research method
In this chapter the research method will be described. In the first paragraph, the research method will be explained and the conceptual framework depicted. The second paragraph will briefly touch upon the sample data in this study. The third paragraph concerns the measures and definitions including choices made regarding the control variables. The fourth paragraph will describe the correlation definition and the regression models.
3.1. Research method
In order to test the hypothesis, the research method of this thesis will mainly be based on the research method as used by Minton and Schrand (1999) in their article: “The impact of cash flow volatility on discretionary investment and the costs of debt and equity financing”.
The four hypothesis will be tested using a solely quantitative research methods. The quantitative research will be conducted as follow. At first, the sample size and timeframe will be defined. After the measures of cash flow volatility, the proxies for discretionary investments and the control variables will be selected based on Minton and Schrand (1999) and other literature. Thirdly the correlation method will be conducted with the Pearson correlation test. Finally, the regression models for all hypothesis (H1 until H4) will be defined and conducted using panel data regressions with YEAR as fixed effect and with restricted maximum likelihood as estimation method. For this, the mixed models functionality in SPSS will be used.
A graphical overview of the conceptual framework for this thesis is provided in figure 1.
- 18 -
3.2. Data
The data will be retrieved from the COMPUSTAT database for companies matching the revenue criteria of > €100 million and < €5.000 million, for twelve sample years, being 2003 till 2014. This revenue criteria is in line with Minton and Schrand (1999). To make sure the results are not driven by country specific elements and to test the impact of financial systems on investment behaviour, nine countries representing two financial systems are included in this study. The United Kingdom represents the Anglo-Saxon system in this study. The Netherlands, Belgium, Luxemburg, Germany, Denmark, Norway, Sweden and Finland represent a more mainland European system
This thesis will only include non-financial firms and hence a filter will be applied to exclude all companies in the financial service industry. The reason being, is that the cash flows and cash flow volatility for companies in the financial service industry is not expected to be related in the same way to discretionary investments as non-financial companies. Further, all proxies for investment and the cash flow volatility will be adjusted for their industry, to eliminate industry difference in the results, this is in line with Minton and Schrand (1999). For that, all (non-financial) firms selected will be categorized and grouped based on a two-digit Standard Industry Code (SIC). The SIC industries with less than 10 samples in the group will be excluded. This to have a representative group of companies per industry as the mean per industry will be used to make industry adjustments on the data. If the industry sample is too little, the mean does not make a representative correction. See appendix I for an overview of all SIC industries included in the sample.
In order to exclude the outliers of these sample sets, the data will be winsorized on both the outer negative and outer positive cases. At a minimum of 2,5% and a maximum of 97,5% of the total number of cases, the data might be winsorized. This method of winsorizing is in line with Minton and Schrand (1999). Without doing so, there is a high probability that the outliers will disturb any potential correlation or relation significantly.
3.3. Measures and definitions
This paragraph will touch upon the different measures and definitions used. 3.3.1. CASH FLOW VOLATILITY MEASURE
The cash flows will be retrieved on a quarterly basis from the COMPUSTAT database as ‘Net Cash Flows of Operating Activities’.
Cash flow volatility is defined as the coefficient of variation of the quarterly cash flows per company over a six year period preceding each of the twelve sample years from 2003 till 2014. For example, the coefficient of variation for the sample year 2014 is calculated using 24 quarters of data from Q1 2008 till Q4 2013. If a company has over six missing observations during the 24 quarters, the company will be expelled from the sample. The coefficient of
- 19 - variation is the standard deviation of net operating cash flow scaled by mean over the same period. This measure of variation has been used by Albrecht and Richardson (1990) and Minton and Schrand (1999).
3.3.2. PROXIES FOR DISCRETIONARY INVESTMENT
The following two items, will proxy for discretionary investments and are based on Minton and Schrand (1999),: I. Capital expenditures (CAPEX)
II. Research & Development expenses (R&D expenses)
The following item will serve as third proxy for investments, but is not based on Minton and Schrand (1999): III. Selling, General & Advertising expenses (SG&A expenses)
SG&A expenses is selected as a proxy for investments even though not all expenses in this group of expenses would be a great proxy for discretionary investments. Minton and Schrand (1999) used advertising expenses as investment proxy, but this data is unavailable for the Northern European market in the COMPUSTAT database or in other available databases. After conducting a linear correlation between advertising expenses and SG&A expenses for 2.864 US companies, a significant positive correlation of 58% is found for these US companies between SG&A expenses and advertising expenses. This means SG&A expenses are highly correlated with advertising expenses and would be a valid replacer of advertising expenses to serve as investment proxy in this study.
See table 1 for the descriptive output of the Pearson correlation of SG&A expenses (dependent variable) and advertising expenses (independent variable) based on a data set of 2.864 US companies of similar size (> 100 million USD & < 5.000 million USD revenue) with data related to a similar period (Q1 2003 until Q4 2014).
N Minimum Maximum Mean Std. Deviation correlation Pearson
Selling, General & Administrative
expenses 16.177 -7.598.000 3.351.470.000 261.113.552 338.821.993 0,578** Advertising expenses 16.177 0 1.200.000.000 26.161.386 60.892.395 0,578**
NOTES:
** and * indicate statistical significance at the 1% and 5% (two-tail) test level. N is the number of cases included in the correlation which is 2.864
times 12 minus the missing cases. Selling, general and administrative expenses and advertising expenses are denoted as absolute numbers
and in US dollars.
Table 1: Descriptive output regression 'SG&A expenses' and 'Advertising expenses'
All three investment proxies are scaled by the firm’s total assets at the beginning of every year. These average CAPEX, R&D expenses and SG&A expenses are calculated over the same rolling six years as the cash flow volatility is
- 20 - calculated6. These three proxies will also be adjusted relative to the industry median for all sample firms with the same two-digit SIC code for the same sample year. Industry-adjusting the proxy variables for investment controls for variation across industries in capital intensity and growth per industry during the sample period.
3.3.3. CONTROL VARIABLES
Based on current literature, there are various factors that might influence the relationship between cash flow volatility and investments. This is in line with the findings of Minton and Schrand (1999) who found a relative low R2 in the relationship (varying between 3,71% and 11,68%). Indicating that there is still a lot of room for various other factors explaining this relationship. Based on Minton and Schrand (1999) the following control variable is selected.
I. The growth rate at which firms grow. This control variable is included as growth is an important indicator of the level of investment. Companies with a high growth rate can be expected to have higher levels of investments to sustain this growth rate. Hence, it is important to control for the influence of growth on the relation between cash flow volatility and investments.
In addition, four more control variables are selected based on other relevant literature. These control variables are selected due to the potential explanatory power they have on investment patterns of firms other than cash flow volatility.
II. Credit constraints (based on Fazzari et al., 1988). Presumably this factor is the most fundamental factor of influence on the level of investments. Literature is till today still divided on the impact and direction credit constraints have on investments. Even though various proxies can be selected as control variable for credit constraints, this study included the dividend pay-out ratio as control variable in order to control for various levels of credit constraints in the sample set.
III. The level of cash holdings (based on Han and Qiu, 2007). As discussed in length in chapter 2, the level of cash holdings is closely related to the level of investments and to the level of financial constraints of a firm. Based on the level of financial constraint, a firm decides to make an investment now or increase its cash holding to ensure future investments. Therefore it is important to control for this factor when explaining investment behaviour based on the level of cash flow volatility.
IV. The financial system (based on Mizen and Vermeulen, 2005). Mainly based on the work of Bond et al. (2003) and Mizen and Vermeulen (2005) is the influence financial systems might have on investments. The way financial systems deal with asymmetric information between lender and borrower is of crucial importance here. Due to the multi-country sample set it is important to control for the various financial systems in the sample set of this study.
- 21 - V. The firm size measured in sales (based on Mizen and Vermeulen, 2005). Also the size of a firm is of influence on the level of investment. In other studies firm size has been used as indicator of accessibility to external finance. For example, small firms are generally younger, with higher levels of firm-specific risk and less collateral. This makes them less likely to attract external capital, which is of influence on their investment pattern. Hence, also this factor should be controlled for when explaining what influences the level of investment.
An important limitation to this set of control variables is the probability of a certain level of autocorrelation among the control variables. These five control are till a certain extend interrelated to each other and of influence to each other when explaining the level of investment. This should be considered when interpreting results.
For all these control variables a dummy variable is included in the model. The dummies are defined as follows: I. Growth rate: ‘0’ is included for the 50% firms which grow slowest. ‘1’ is included for the 50% firms growing
fastest. The firm growth rate is measured by calculating each firms years growth rate of their total assets. II. Credit constraints: ‘0’ is included for the 50% firms with the lowest dividend pay-out ratio. ‘1’ is included for the 50% firms with the highest dividend pay-out ratio. The dividend pay-out ratio is calculated by scaling the total annual dividend pay-out by the firm’s EBIT. This proxy for credit constraints is selected based on Cleary (2006).
III. Cash holdings: ‘0’ is included for the 50% firms with the lowest average cash holdings scaled by their total assets. ‘1’ is included for the 50% firms with the highest cash holdings scaled by their total assets.
IV. Financial system: ‘0’ is included for all firms operating in mainland Europe, which characterized itself as a financial system where mostly banks are the primary source of finance in their financial system. This is a so called relationship-oriented system. ‘1’ is included for firms in the Anglo-Saxon model, where mainly large secondary markets (i.e. corporate bond market) are the main source of finance in the financial system. V. Firms size: ‘0’ is included for the 50% smallest firms measured in total annual sales. ‘1’ is included for the
50% largest firms.
Including these control variables in the regression model will increase the robustness of the model and enforces the strength of the model when explaining the relationship between cash flow volatility and discretionary investments.
3.4. Correlations and Regressions
3.4.1. CORRELATIONS
Before the hypothesis will be tested based on the regressions with YEAR as fixed effect, all independent, dependent and control variables will be correlated with each other and see if there are significant and interesting correlations between all variables. A bivariate Pearson correlation test will be used for this.
- 22 - 3.4.2. MAIN MODEL REGRESSION
After the correlations of all variables are known the actual hypothesis will be tested. For the hypothesis testing a regression with YEAR as fixed effect and with restricted maximum likelihood as estimation method will be used. This model is selected for two reasons. First, it theoretically makes sense, as every company in the sample set has observations per year and these observations are not of random influence to each other. A single CAPEX observation for a specific firm in 2003 is not fully random related to a CAPEX observation for that same firm in 2004. Therefore the years are of influence on the regressions and should be considered fixed. For that reason, the regression is run with YEAR as fixed effect. This is done by using the ‘mixed model’ functionality in SPSS. A second reason is this mixed model with YEAR as fixed effect is statistically considered the strongest explanatory model. This fixed effect model is chosen over a mixed model with random effect and over a mixed model with year as repeated variable due to the lower value of the Akaike’s Information Criterion (AIC), corrected for the degrees of freedom which every addition parameters uses up in the AIC score7. This indicates that a panel data regression with YEAR as fixed effects can statistically considered the strongest model to test the relationship between cash flow volatility and investments. The results of the AIC test are included in Appendix II.
First, the main hypothesis which examines the relation between investment and cash flow volatility will be tested with the following regression equation:
INVESTMENTit = a0 + a1 CVFCit + ∑ ak CONTROLit + eit (1) Where:
o
INVESTMENT (dependent variable) is one of the three proxies for discretionary investment: capitalexpenditures, R&D expenses or SG&A expenses. All three the proxies are scaled by the firm’s total assets at the beginning of the year;
o
CVCF (independent variable) is the cash flow volatility variable. The coefficient of variation of the cash flows;o
CONTROL (control variables) are the five control variables that control for effects on a firm’s investmentbehaviour other than cash flow volatility.
3.4.3. REGRESSION MODELS WITH AN INTERACTIVE VARIABLE
After the main hypothesis is tested, the impact of various factors on the sensitivity of investment to cash flow volatility will be tested with additional regression models which have an interactive variable included. These regression models will answer hypothesis H2, H3 and H4.
The following regression model will test H2 and indicates the adjustment of cash flow volatility on investment for firms with a higher cash holding compared to firms with a lower cash holding:
7See: http://www.theanalysisfactor.com/repeated-and-random-2/
- 23 - INVESTMENTit = b0 + b1 HI_CH + b2 CVCFit + b3 CVCF x HI_CHit +∑bk CONTROLit + eit (2)
Where, in addition:
o
HI_CH (indicator variable) is an indicator equal to ‘1’ if a firm is in the highest 50% of the sample firms whenranked on cash holdings (calculated as % of cash and cash equivalents over the beginning of period total assets).
o
CVCF x HI_CH (interactive variable) which measures the adjustment of cash flow volatility on investment forfirms with a higher cash holding compared to firms with a lower cash holding;
The following model will test H3 and indicates the adjustment of cash flow volatility on investment during the 2008 credit crisis (2009-2014) compared to the same six years before the credit crisis (2003-2008):
INVESTMENTit = c0 + c1 CC + c2 CVCFit + c3 CVCF x CCit +∑ck CONTROLit + eit (3) Where, in addition:
o
CC (indicator variable) is a dummy variable equal to ‘1’ if it concerns data during the credit crisis;o
CVCF x CC (interactive variable) which measures the adjustment of cash flow volatility on investment for theperiod before the credit crisis compared to an equal period during the credit crisis;
The final model will test H4 and indicates the adjustment of cash flow volatility on investment for the firms in the Anglo-Saxon financial system compared to firms in the mainland European financial system:
INVESTMENTit = d0 + d1 FS + d2 CVCFit + d3 CVCF x FSit +∑dk CONTROLit + eit (4) Where, in addition:
o
FS (indicator variable) is a dummy variable equal to ‘1’ if a firms is incorporated in the United Kingdom(Anglo-Saxon financial system);
o
CVCF x FS (interactive variable) which measures the adjustment of cash flow volatility on investment for thefirms in the UK (Anglo-Saxon model) compared to firms in mainland Northern Europe (Rhineland and Nordic model);
Similar to the first model (1), also the models (2) & (4) will be estimated annually using a regression with YEAR as fixed effect and with restricted maximum likelihood as estimation method. The third model (3), concerning the credit crisis will not have YEAR as fixed effect, as the factor time is already included in the indicator variable (CC)
k=4,5,6,7,8,9 k=4,5,6,7,8
- 24 -
4. Results
In this chapter all hypothesis (H1 until H4) will be tested per dependent variable as follows. In the first paragraph the descriptive statistics will be described. In the second paragraph the results of bivariate correlations will be shown per variable. In the third paragraph, the results of the regressions with YEAR as fixed effect are shown. In the fourth paragraph, the results are shown regarding the additional regression models. The outcome of the additional regression models will determine the robustness of the main regression equation and verify whether this equation still holds when other interaction effects are incorporated in the model.
4.1. Descriptive statistics
The descriptive statistics for the sample of firms are shown in table 2. The total sample consists of 1.212 companies representing nine Northern European countries. The sample has a total time lapse of twelve years (from 2003 until 2014). This brings the total number of cases to 14.544 of which not all variables have all data. The number of data points per variable is shown in N. In order to exclude extreme values that distort the correlations and regressions, some variables are winsorized at a minimum of 2,5% and a maximum of 97,5% as percentage of the total of cases per variable. In table 3, the variables are shown, both ‘industry adjusted’ (scaled by the mean of the SIC industry average) and ‘NOT industry adjusted’. The industry adjusted values are used for the correlations and regressions.
N Minimum Maximum Mean Std. Deviation
Capital expenditure 12.723 0,001 0,153 0,045 0,036 Industry adjusted capital expenditure 12.723 0,024 3,303 1,211 0,740 Research & development expenses 5.570 0,000 0,162 0,013 0,029 Industry adjusted research & development expenses 5.570 0,014 17,637 3,032 3,766 Selling, general & admin. Expenses 10.880 0,000 0,731 0,172 0,178 Industry adjusted selling, general & admin. expenses 10.880 0,032 5,348 1,560 1,194 Cash flow volatility 12.979 -8,414 9,888 0,857 2,715 Industry adjusted cash flow volatility 12.979 0,045 41,548 3,526 7,958 Cash holdings 14.316 0,000 1,000 0,500 0,500 Financial system 14.544 0,000 1,000 0,333 0,471 Credit constraints 14.508 0,000 1,000 0,500 0,500 Firm size 14.544 0,000 1,000 0,464 0,499 Growth 14.532 0,000 1,000 0,500 0,500 NOTES:
N is the number of cases included in the regression model. The maximum number of cases is 14.544. Any N lower than 14.544 indicates missing
data for this variable. Capital expenditure, R&D expenses and SG&A expenses are denoted as a % of total assets. For the industry adjusted capital
expenditures, R&D expenses and SG&A expenses, the expense ratio per firm is scaled by the mean ratio for all firms in the same SIC industry. This to adjust for any industry differences in the sample. The cash flow volatility is defined as the coefficient of variation of net operating cash flow.
The industry adjusted cash flow volatility is calculated by scaling the cash flow volatility per firm by the mean of the SIC industry cash flow
volatility.
- 25 -
4.2. Correlations
To examine the association among the variables, including the relationship between cash flow volatility and investment, the Pearson correlation coefficients have been calculated. The Pearson correlation coefficient is the correct measure because all variables contain quantifiable data on at least an interval scale. Nonetheless, this Pearson correlation coefficient, does not indicate anything about the strength of the relationship.
Table 3 summarizes the correlation matrix. Consistent with Minton and Schrand (1999), the matrix shows a (significant) negative correlation between CAPEX and cash flow volatility. This implies that when a firm has higher industry adjusted cash flow volatilities, industry adjusted capital expenditures as a % of total assets are lower. This indicates that the difference in time (two decades later) and the geographical difference (European data compared to US data) does not lead to a significant different direction of the correlation when comparing these results with the results of Minton and Schrand (1999).
Further, a negative correlation is found for R&D expenses, but cannot be considered significant in this bivariate correlation test. This is not in line with Minton and Schrand and not in line with the hypothesis. Two potential explanations could be suggested for this outcome. At first, a potential explanation for this non-significant correlation outcome could be related to the data quality of this sample set. As shown in the descriptive statistics (table 3), R&D expenses have only 5.570/14.544 cases, whereas the other two variables have over 10.000/14.544 cases. This has mostly to do with the fact that not all companies actually have R&D activities. Secondly, this non-significance might have to do with the assumption that European firms, almost two decades later, are managed in such a way that R&D expenses are forecasted and planned much better that irrespective of the cash flow volatility the R&D investments can be done and the R&D activities are not influenced or to a much lesser extend by cash flow influences. The assumption that R&D expenses are of much more strategic importance compared to capital expenditures and hence more management attention is given, might explain why specifically R&D expenses do not show a significant relation and CAPEX does.
Inconsistent with hypothesis H1 and inconsistent with the advertising expenses correlation in Minton and Schrand (1999), is the (significant) positive correlation coefficient of SG&A expenses. This implies that when a firm has higher cash flow volatilities, industry adjusted SG&A expenses as a % of total assets are higher. An economic explanation could be that a higher cash flow volatility means less recurring business, more sales effort and less standard back office work compared to times that cash flows are more stable. Hence, for the same amount of sales, relatively more work has to be done for selling, general and administrative activities. This might result in managers having to increase the resources committed to these additional activities needed, resulting in higher SG&A expenses relative to total assets.
- 26 - Table 3: Correlations matrix
Dependent variables Independent variable Control variables
Industry adjusted capital expenditure Industry adjusted research & development expenses Industry adjusted selling, general & admin. expenses Industry adjusted cash flow
volatility holdings Cash Financial system Constraints Credit Firm size Growth Industry adjusted capital expenditure
Industry adjusted research & development expenses 0,003
Industry adjusted selling, general & admin. expenses ,020* 0,200**
Industry adjusted cash flow volatility -0,051** -0,015 0,034**
Cash holdings -0,057** 0,104** 0,097** 0,028** Financial system -0,072** 0,033* 0,063** -0,080** -0,069** Credit Constraints 0,008 0,049** 0,080** -0,084** 0,062** 0,185** Firm size 0,019* -0,021 0,012 -0,015 -0,083** 0,206** 0,147** Growth 0,005 0,013 -0,078** 0,009 0,115** -0,066** -0,094** -,228** NOTES:
** and * indicate statistical significance at the 1% and 5% (two-tail) test level. Capital expenditure, R&D expenses and SG&A expenses are denoted as a % of total assets and are industry adjusted by scaling the capital expenditure, R&D expenses and SG&A expenses ratio per firm by the mean ratio for all firms in the same SIC industry. This to adjust for any industry differences
in the sample. The cash flow volatility is defined as the coefficient of variation of net operating cash flow and is industry adjusted by scaling the cash flow volatility per firm by the mean of the
