The Determinants of Cash Holdings: Evidence from Eurozone companies after the global financial crisis

(1)

The Determinants of Cash Holdings: Evidence from

Eurozone companies after the global financial crisis

Name: Mitchel Vogt Student number: S3274322 Study program: MSc Finance

Supervisor: E.L. Kramer

Keywords: Cash holdings, transaction motive, precautionary motive, agency motive Date: 12-06-2019

Abstract

This thesis examines the determinants of cash holdings for a sample of 1730 Eurozone listed companies over the period from 2010 to 2017. The predictions of the transaction-, precautionary and the agency motive on the determinants of cash holdings are being tested in this research. I find positive relations with cash holdings for the variables cash flow, dividend, the market-to-book ratio and for the industry sigma. While on the other hand, I find negative relations with the variables size, net working capital and the capital expenditures. And for the variables leverage and R&D expenses I find no statistically significant results. Therefore, it can be concluded that three main motives for companies’ cash holding play an important role in explaining the determinants of cash holdings for companies in the Eurozone. Especially the precautionary motive plays an important role here. While there is reasonable support for the transaction motive as well, and for the agency motive there is no support.

(2)

2

1. Introduction

The last decades, the cash holdings of companies have gathered a lot of attention from the media, analysts, investors and researchers. Companies all over the world increased their cash holdings significantly. Bates et al. (2009) even found out that the average cash ratio of companies in the United States doubled over the 1980-2006 period. Therefore, the question why do companies hold more and more cash has gathered a lot of attention. The most popular explanation offers three main motives for companies’ cash holding: the transaction motive, the precautionary motive and the agency motive. Besides the three main motives the explanation also offers two theories about the optimal level of cash: the trade-off theory and the agency theory.

Several researches have been conducted in order to find out what determines the changes in the cash holdings in practice. For instance, Ferreira and Vilela (2004) investigated the determinants of cash holdings for the European Monetary Union during 1987 and 2000. Ozkan and Ozkan (2004) investigated the determinants for the UK during 1984 and 1999. Bates et al. (2009), Custodio et al. (2005) and Opler et al. (1999) investigated the determinants for US companies during 1980-2006, 1971-2002 and 1971-1994. While Garcia-Teruel and Martínez-Solano (2008) investigated the determinants for Spain during 1997 and 2001. At last, Chen et al. (2012) investigated the determinants for China during 2000 and 2008.

As shown in the previous paragraph, several researches have been conducted in order to find out what determines the changes in the cash holdings in practice. The focus of these researches was mainly on the determinants of cash holdings in Asia, the United Kingdom or the United States. However, there are several reasons why there might be differences in the level of cash holdings between European, Asian and US companies. For instance, Nykvist (2008) states that US companies have better access to the capital markets than European and Asian companies. While La Porta et al. (1998) found out that US has better legal protection for investors than Europe. Also there has been less examined about the cash holdings of companies after the financial crisis of 2008. While Ivashina and Sharfstein (2010) stated that the financial crisis of 2008 had a great impact on the economy and companies. Prices across most asset classes (i.e. equities, bonds, and commodities) fell drastically, the borrowing costs of corporates and banks rose substantially, and financial market volatility rose to levels that have rarely, if ever, been seen.

(4)

4 important determinants of cash holdings in the Eurozone after the global financial crisis of 2008, and whether these determinants have a positive or a negative effect on the cash holding. Therefore, the research question of this study is: To which extent do specific firm characteristics

have an influence on cash holding of Eurozone companies after the global financial crisis of 2008?

2. Literature review

This section contains a discussion of the relevant literature in the field. The first subsection is a review of the theoretical research about cash holdings. Afterwards there will be a more detailed overview of the three main motives for companies’ cash holding. After the literature is discussed, the empirical findings of other authors will be discussed. This section will end with the hypotheses development.

2.1 Cash holdings

According to Opler et al. (1999) are cash holdings commonly defined as cash and cash equivalents. In order to make this a comparable variable for research, the cash and cash equivalents will be divided by the total assets. Cash equivalents are assets that can be easily converted into cash (Ogundipe, Ogundipe and Ajao, 2012). In a world of perfect capital markets, companies would not have the need to hold liquid assets, since they would be easily able to obtain funding for its profitable investment projects (Modigliani and Miller, 1958). However, in practice this is not the case because of the presence of financial fractions. The literature offers three main motives for the companies’ cash holding. These motives are the transaction motive, precautionary motive and the agency motive. The following subsections will discuss the concept of the three motives and their predictions, which will help to develop the hypotheses.

2.2 The transaction motive

Keynes (1934) was the first one who came up with the transaction motive. The transaction motive is about the transaction costs that companies face when they convert a non-cash financial asset into cash. According to literature the transaction motive has influence on the following variables: Size, cash flow, net working capital, leverage and dividend.

Size

(5)

5 Cash flow

According to Kim et al. (1998) the cash flow provides a ready source of liquidity, especially when the company has to meet operating expenditures and maturing liabilities. Therefore, the cash flow is negatively related to cash holding.

Net working capital

Opler et al. (1999) suggest that the net working capital could be used as a measure for liquid assets because it allows covering cash shortage at low cost. Therefore, the net working capital is negatively related to cash holding.

Leverage

According to Baskin (1987) the transactions costs of converting a non-cash financial asset into cash is rising along with the leverage ratio. Therefore, leverage is negatively related to cash holding.

Dividend

According to Bates et al. (2009) and Opler et al. (1999) can dividend paying companies raise funds at low cost. To do this these companies have to lower their dividend payments. Therefore, the expectation is that dividend paying companies hold less cash.

2.3 The precautionary motive

The precautionary motive implies that companies hold cash, so they will not miss future investment opportunities and to hedge the company against risks associated with possible shocks in cash flows. According to literature the precautionary motive has influence on the following variables: Size, cash flow, net working capital, leverage, dividend, market-to-book ratio, R&D expenditures, capital expenditures and the industry sigma.

Size

(6)

6 Cash flow

Almeida et al. (2004) found that the cash holding of financially constrained companies is positively related to the cash flow, while for financially unconstrained companies there is no relationship. Therefore, the overall conclusion of Almeida et al. (2004) is that the cash flow is positively related to cash holding.

Net working capital

Net working capital is calculated as working capital minus cash and marketable securities, and then divided by the book value of assets. Assets, which can substitute for cash, are included in the net working capital. Therefore, a negative relation is expected.

Leverage

In the transaction motive leverage is negatively related to cash holding, while in the precautionary motive it is positively related to cash holding. This because companies with higher leverage tend to hold more cash in order to reduce the possibility of financial distress. Therefore, leverage is in the precautionary motive positively related to cash holding.

Dividend

According to Bates et al. (2009) dividend paying companies are likely to be less risky, and therefore have better access to the capital market. So, their prediction is that dividend paying companies are holding lower levels of cash compared to non-paying dividend companies.

Market-to-book ratio

Ferreira and Vilela (2004) predict that companies with a high market-to-book ratio have greater financial distress costs. Because of this prediction companies with a high market-to-book ratio are expected to hold more cash.

R&D expenses

Just like companies with a high market-to-book ratio, companies with high R&D expenses have greater financial distress costs. Therefore, R&D expenditures are positively related to cash holding.

Capital expenditures

(7)

7 Industry Sigma

The industry sigma measures the standard deviation of the cash flows for companies in the same industry, using the two-digit SIC code. When the industry sigma is high, this means that the cash flows of the company have a high volatility. Because of this high volatility companies tend to hold more cash. Therefore, the industry sigma is positively related to cash holding.

2.4 The agency motive

The agency motive deals with the separation of ownership and the control of the company. According to Jensen and Meckling (1976) agency costs may arise, since management may not always be inclined to act in accordance with the purpose of shareholder wealth maximization. According to literature the agency motive has influence on the following variables: Size and leverage

Size

Bates et al. (2009) stated that companies with a larger size have more dispersed ownership, and therefore are more likely to have agency problems. Therefore, managers of larger companies should have more power over the company and its investment policy and that is why larger companies have higher cash holdings.

Leverage

According to Ferreira and Vilela (2004) companies with a low leverage are monitored less. Therefore, leverage is negatively related to cash holding.

2.5 Optimal level of cash

(8)

8 2.6 Empirical findings

As mentioned in the introduction, several researches have been conducted in order to find out what determines the changes in the cash holdings in practice. This subsection will give an overview of the outcomes of the most important empirical findings. Later on the empirical findings and the explanation of the three main motives for companies’ cash holding will be used for the hypotheses development.

Ferreira and Vilela (2004) investigated the determinants of cash holdings in EMU countries. Their sample consists of a total of 6387 company-year observations over a period from 1987 to 2000. Their results suggest that cash holdings are negatively related to asset’s liquidity, leverage and size, while they found a positive relation with the cash flows and the investment opportunity.

Ozkan and Ozkan (2004) investigated the determinants of corporate cash holdings for UK companies over a period from 1984 to 1999. They found out that the cash flow, liquid assets, leverage, growth opportunities and bank debt are important determinants of cash holdings.

Bates et al. (2009) used a sample of 117,438 observations for 13,599 unique US companies over a period from 1980 till 2006. Their results suggest that the industry sigma, the market-to-book ratio and the R&D expenses are positively related to cash holdings. While net working capital, size, capital expenditures, leverage, dividend and the acquisition activity are negatively related. For the cash flow they found no significant relation with the companies’ cash holding.

Opler et al. (1999) examined the determinants and implications of holdings of cash by publicly traded US companies in the 1971-1994 period. Their results suggests that cash holding is negatively related to company size, leverage, dividend, capital expenditures and net working capital. While cash holding is positively related to the market-to-book ratio, cash flow, R&D expenses and the industry sigma.

(9)

9 2.7 Hypotheses development

In this section the results of some of the most important researches from the past are summarized in Table 1. And in Table 2 the predictions of the transaction-, precautionary- and the agency motive on the company-specific factors are summarized. After that the hypotheses for each variable will be developed.

Table 1: Summary of the results of major studies in the past Variable – Research Ferreira and

Vilela (2004 Ozkan and Ozkan (2004) Bates et al. (2009) Opler et al. (1999) Chen et al. (2012)

Size minus minus minus

Cash Flow plus plus plus

Net Working Capital minus minus

Leverage minus minus minus minus plus

Dividend minus minus plus

Market to book ratio plus plus

R&D expenditures plus plus

Capital expenditures minus minus minus

Industry Sigma plus plus plus

*Only significant results are presented in this table

Table 2: Summary of the variables and their expected sign

Variable – Motive Transaction Precautionary Agency

Size minus minus plus

Cash Flow minus plus

Net Working Capital minus minus

Leverage minus plus minus

Dividend minus minus

Market to book ratio plus

R&D expenditures plus

Capital expenditures plus

Industry Sigma plus

Starting with the first variable, which is the company size. When looking at Table 1 we see that the results of Ferreira and Vilela (2004), Bates et al. (2009) and Opler et al. (1999) all suggests that size is negatively related to corporate cash holding. When looking at the predictions of the three motives we see that both the transaction and precautionary expect a negative relation, while the agency motive expects a positive relation between size and the level of cash holding. Taking together the results represented in Table 1 and the predictions of all three motives represented in Table 2, it is reasonable to predict a negative relation between the level of cash holding and size.

(10)

10 For the second variable, which is the cash flow. Ferreira and Vilela (2004), Ozkan and Ozkan (2004) and Opler et al. (1999) all suggest a positive relation between the cash flow and the level of cash holding. The precautionary motive supports the outcomes of those researchers, while the transaction motive supposes a negative relation. Taking all these results and expectations of the theory together it is reasonable to hypothesize the following:

Hypothesis 2. There is a positive association between cash flows and cash holding

Coming to the relation of the net working capital and the level of cash holding. Bates et al. (2009) and Opler et al. (1999) both found a negative relation between the net working capital and the level of cash holding. Also both the transaction motive and the precautionary motive expect a negative relation. Therefore, it is reasonable to predict a negative relation between the level of cash holding and the net working capital.

Hypothesis 3. There is a negative association between net working capital and cash holding

For the next variable, which is leverage, almost all the researchers found a negative relation between leverage and the level of cash holding. Only Chen et al. (2012) which used a sample of Chinese companies over a period from 2000 to 2008 found a positive relationship. When looking at Table 2, which summarizes the predictions of the three motives, we see that the transaction- and agency motive predicts a negative relation, while the precautionary motive predicts a positive relation between leverage and the level of cash holding. Taking together all the predictions of the motives and the results of the major studies, it is reasonable to predict a negative relation.

Hypothesis 4. There is a negative association between leverage and cash holding

Both the transaction and precautionary motive predict a negative relation between dividend paying companies and the level of cash holding. Also when looking at the outcomes of the major studies from the past, most of the researchers found a negative relationship between dividend paying companies and the level of cash holding. Only Chen et al. (2012) found a positive relation. Therefore, it is reasonable to predict a negative relation between dividend paying companies and the level of cash holding.

(11)

11 For the next variable, which is the market-to-book ratio, only the precautionary motive predicts the relation between the variable and the level of cash holding. The precautionary motive predicts a positive relation. When looking at the outcomes of some of the major studies, we see that both Bates et al. (2009) and Opler et al. (1999) found a positive relation between the market to book ratio and the level of cash holding. Therefore, the following hypothesis is created:

Hypothesis 6. There is a positive association between market-to-book ratio and cash holding

Also for the next variable, only the precautionary motive predicts the expected relation. The motive predicts a positive relation between research and development expenditures and the level of cash holding. Which is also the same as with the relationship between the market to book ratio and the level of cash holding, is that only Bates et al. (2009) and Opler et al. (1999) found significant results for the relation between the research and development expenditures and the level of cash holding. They both found a positive relation, which leads to the following hypothesis:

Hypothesis 7. There is a positive association between R&D expenditures and cash holding

Coming to the relation of capital expenditures and the level of cash holding. The precautionary motive predicts a positive relation between them. But all the researchers who investigated this relationship and found significant outcomes found a negative relation between capital expenditures and the level of cash holding. The researches which had these outcomes are Bates et al. (2009), Opler et al. (1999) and Chen et al. (2012). Therefore, the hypothesis is as follows:

Hypothesis 8. There is a negative association between capital expenditures and cash holding

For the last variable, which is the industry sigma, also the only motive which expects a relationship between the industry sigma and the level of cash holding is the precautionary motive. The precautionary motive expects a positive relation between the industry sigma and the level of cash holding. When looking at the outcomes of some of the major studies from the past, we see that Bates et al. (2009), Opler et al. (1999) and Chen et al. (2012) all found a positive relation between the industry sigma and the level of cash holding, which leads to the following hypothesis:

(12)

12

3. Data description

A panel dataset that consist of annual fundamentals of Eurozone companies for the year 2010 to 2017 will be used to investigate the hypotheses on the determinants of cash holdings. The dataset is a panel dataset because the dataset contains a cross-sectional and a time series dimension. The year 2010 is chosen to avoid the effects of the global financial crisis. Data for this research will be collected from the WRDS COMPUSTAT database and the Datastream database. Companies with a Standard Industrial Classification (SIC) code between 6000 and 6999 (financials) will dropped. Since those companies include marketable securities in their businesses, which involves cash, and they are obliged to meet statutory capital requirements. Also companies with a SIC code between 4900 and 4999 (utilities) will be dropped, because their cash holdings are subject to regulatory supervision. After these specific companies are removed from the sample, there will be taken care of the outliers by winsorizing them. Winsorizing means limiting the extreme values in the sample to reduce the effect of spurious outliers. The variables which are being winsorized are cash, size, cash flow, dividend, net working capital, the market-to-book ratio, capital expenditures, R&D expenses, leverage and the industry sigma at 1% at each tail. In the end the sample consist of 13332 observations. Missing explanatory values reduce the panel to 11633 observations for 1730 unique companies spread over 19 countries. The details of the sample are presented in the upcoming two tables. Table 3: Number of observations and unique companies per country

Country Number of observations Number of unique companies

(13)

13 Table 3 represents the number of observations and the number of unique companies per country. As shown in table 3, countries such as France, Germany, Italy and Greece are represented very well in the sample. For Greece this is remarkable, because only 11 million people are living in Greece at the moment. While in France, Germany, Italy and even in Spain there are living at least five times the amount of people. Looking at the lower number of observations, we see that Slovakia, Malta and Estonia are represented with only a few unique companies.

Table 4: Number unique companies per industry

Industry SIC Codes Number of unique companies

Agriculture, Forestry and Fishing 01-09 18

Mining 10-14 50 Construction 15-17 71 Manufacturing 20-39 1028 Transportation 40-48 93 Wholesale Trade 50-51 69 Retail Trade 52-59 92 Services 70-89 292 Nonclassifiable 99 17 Total 1730

(14)

14

4. Research method

This section starts with discussing the methodology which is used in this study. Through careful argumentation it will be explained why and how the respective methods are necessary with testing this particular panel dataset. At last the measurements of the variables will be explained.

4.1 OLS Regression

According to the bulk of literature, the most common way to deal with a panel dataset is to perform an OLS regression. The OLS regression is a type of a linear least squares method for estimating the unknown parameters in a linear regression model. OLS chooses the parameters of a linear function of a set of explanatory variables by the principle of least squares: minimizing the sum of the squares of the differences between the observed dependent variable in the given dataset and those predicted by the linear function.

According to Wooldrigde (2013) the biggest advantage of the OLS regression is that the sample size can easily be increased by pooling observations from different time periods. This is especially helpful when the function to be estimated contains many explanatory variables, but within a period only contains a small amount of cross sectional data. On the other hand, the OLS model has also some major disadvantages. The first disadvantage is that the OLS model is very sensitive to outliers. Because the OLS model is concerned with minimizing the sum of the squares of the differences between the observed dependent variable in the given dataset and those predicted by the linear function, will any extreme data point have a large effect on the outcomes of the OLS regression. A second major disadvantages of the OLS model is that in the presence of heterogeneity, which means that there is variability in the data, the estimators of the regression will become inconsistent and biased (Wooldridge, 2013).

An OLS regression would look as follows:

𝑌𝑖𝑡 = 𝛽0+ 𝛽𝑗 𝑥𝑖𝑡+ 𝜖𝑖𝑡 ( 1 )

Where 𝑌_𝑖𝑡 represents the dependent variable, 𝛽_𝑗𝑥 the independent variables and ϵit the error term. And where 𝑖 =1, 2, 3… N companies and 𝑡= 1, 2, 3…𝑇 time periods.

Therefore, the OLS regression in this particular research would look as follows: 𝐶𝑎𝑠ℎ𝑖𝑡 = ∝ + 𝛽1 𝑆𝑖𝑧𝑒𝑖𝑡+ 𝛽2 𝐶𝑎𝑠ℎ𝐹𝑙𝑜𝑤𝑖𝑡 + 𝛽3 𝑁𝑊𝐶𝑖𝑡+ 𝛽4 𝐿𝑒𝑣𝑒𝑟𝑎𝑔𝑒𝑖𝑡+ 𝛽₅ 𝐷𝑖𝑣𝑖𝑑𝑒𝑛𝑑 𝐷𝑢𝑚𝑚𝑦_𝑖𝑡+ 𝛽₆ 𝑀𝑇𝐵_𝑖𝑡+ 𝛽₇ 𝑅&𝐷 𝐸𝑥𝑝 𝐷𝑢𝑚𝑚𝑦_𝑖𝑡 +

(15)

15 4.2 Between-Effects-Model

The Between-Effects-Model is not used very often within researches about the determinants of cash holdings. However, according to Wooldridge (2010) the Between-Effects-Model is important because the estimator of the Random-Effects-Model is a weighted average of the Fixed-Effects and the Between-Effects coefficient. Therefore, researchers like Fujiki and Kitamure (1995) also decided to include the Between-Effects-Model in their research. The Between-Effects-Model regresses means by cross-section of the dependent variable on means by cross-section of the independent variables.

4.3 Fixed- and Random-Effects-Model

With the OLS model, there is an assumption that in each period the error term is uncorrelated to the explanatory variables. But, according to Wooldridge (2002) for some datasets this assumption is too strong. Wooldridge (2002) points here at the fact that the primary motivation of the panel data models is to deal with the omitted variable problem. According to the bulk of literature there are two important models that take into account these unobserved individual or company specific factors. The two models are the Fixed-Effects-Model and the Random-Effects-Model. If we look at empirical research papers, we see that most of the authors used the Fixed-Effects-Model in their analysis. Authors which used the Fixed-Effects-Model are for instance, Bates et al. (2009), Opler et al. (2009) and Ozkan and Ozkan (2004). To see which of the two models is appropriate for a specific research, Hausman (1978) developed the Hausman Test. In this test, the idea is that the Random-Effects-Model is used unless the tests rejects the null hypothesis. This null hypothesis is that the company specific unobserved factors are uncorrelated with explanatory variables:

𝐸 (𝑎𝑖|𝑥𝑖𝑡) = 0

And the alternative hypothesis is that the company specific unobserved factors are correlated with the explanatory variables:

𝐸 (𝑎_𝑖|𝑥_𝑖𝑡) ≠ 0

(16)

16 The Fixed-Effects-Model analyses the impact of variables that vary over time. The model explores the relation between the predictor and the outcome variables within a company. Where each company has it is own individual characteristics that may influence the predictor.

A Fixed-Effects regression would look as follows:

𝑌_𝑖𝑡 = 𝛽_𝑗 𝑥_𝑖𝑡+ 𝑎_𝑖 + 𝜖_𝑖𝑡 ( 3 )

Where 𝑌_𝑖𝑡 represents the dependent variable, 𝛽_𝑗𝑥 the independent variables, 𝑎_𝑖 the time invariant unobservable company specific effect and ϵit the error term. And where 𝑖 =1, 2, 3… N companies and 𝑡= 1, 2, 3…𝑇 time periods.

4.4 Measures

Dependent Variable

In this thesis, the level of cash holding for each company is the dependent variable. COMPUSTAT defines cash as: “This item represents any immediately negotiable medium of exchange or any instruments normally accepted by banks for deposit and immediate credit to a customer's account.” In order to make this a comparable variable, the following formula will be used for the cash holding:

𝐶𝑎𝑠ℎ 𝐻𝑜𝑙𝑑𝑖𝑛𝑔 =Cash + Cash Equivalent Total Assets Independent Variables

The independent variables used in this research represent the company characteristics. The used independent variables are size, cash flow, net working capital, leverage, dividend, the market-to-book ratio, R&D expenses, capital expenditures and the industry sigma.

Size

The first independent variable is the size of a company. The formula which is used to calculate the size is as follows:

(17)

17 Cash Flow

The formula which is used to calculate the cash flow is as follows:

𝐶𝑎𝑠ℎ 𝐹𝑙𝑜𝑤 =(𝐸𝑎𝑟𝑛𝑖𝑛𝑔𝑠 𝑎𝑓𝑡𝑒𝑟 𝑡𝑎𝑥𝑒𝑠 + 𝐷𝑒𝑝𝑟𝑒𝑐𝑖𝑎𝑡𝑖𝑜𝑛) Total Assets

Net Working Capital

The formula which is used to calculate the net working capital is as follows:

Net Working Capital =(𝑊𝑜𝑟𝑘𝑖𝑛𝑔 𝐶𝑎𝑝𝑖𝑡𝑎𝑙 − (𝐶𝑎𝑠ℎ + 𝐶𝑎𝑠ℎ 𝐸𝑞𝑢𝑖𝑣𝑎𝑙𝑒𝑛𝑡)) Total Assets

Leverage

The formula which is used to calculate the leverage is as follows:

𝐿𝑒𝑣𝑒𝑟𝑎𝑔𝑒 =𝑇𝑜𝑡𝑎𝑙 𝐷𝑒𝑏𝑡 Total Asset Dividend

Dummy variables will be used for this variables. Ferreira and Vilela (2004) and Opler et al (1999) both did the same in their research. Dividend paying companies will have a value of 1, while non-paying dividend companies will have a value of 0.

𝐷𝑖𝑣𝑖𝑑𝑒𝑛𝑑 = 1 𝑖𝑓 𝑝𝑎𝑖𝑑; 0 𝑜𝑡ℎ𝑒𝑟𝑤𝑖𝑠𝑒 Market to book ratio

The formula which is used to calculate the market-to-book ratio is as follows:

(18)

18 R&D expenses

Also dummy variables will be used for this variables. Companies with research and development expenses will have a value of 1, while companies without research and development expenses will have a value of 0.

𝑅𝑒𝑠𝑒𝑎𝑟𝑐ℎ 𝑎𝑛𝑑 𝐷𝑒𝑣𝑒𝑙𝑜𝑝𝑚𝑒𝑛𝑡 𝑒𝑥𝑝𝑒𝑛𝑑𝑖𝑡𝑢𝑟𝑒𝑠 = 1 𝑖𝑓 𝑒𝑥𝑖𝑠𝑡𝑠; 0 𝑜𝑡ℎ𝑒𝑟𝑤𝑖𝑠𝑒 Capital expenditures

The formula which is used to calculate the capital expenditures is as follows:

Capital Expenditures =𝐶𝑎𝑝𝑖𝑡𝑎𝑙 𝐸𝑥𝑝𝑒𝑛𝑑𝑖𝑡𝑢𝑟𝑒𝑠 𝑇𝑜𝑡𝑎𝑙 𝐴𝑠𝑠𝑒𝑡𝑠 Industry Sigma

The industry sigma is measured in the following way: The averages for each year of the standard deviations of cash flows for companies in the same industry, defined by the two-digit SIC code. Industry Sigma

= Avg (Standard Deviation of the Cash Flows of the same industry defined by the two − digit SIC code)

5. Results

This section starts with discussing the descriptive statistics and the correlation matrix. After that, the results of OLS regression, the Between-Effects and the Fixed-Effects Model will be discussed. This section will end with an additional robustness test.

5.1 Descriptive statistics

(19)

19 Table 5: Descriptive statistics

Cash Size

Cash

Flow Dividend NWC MTB Capex

R&D Exp. Leverage Industry Sigma Mean 0.12 5.53 0.08 0.32 0.04 1.84 0.04 0.45 0.56 1.03 Median 0.08 5.32 0.10 0.00 0.04 1.13 0.03 0.00 0.47 0.96 Maximum 0.80 11.20 0.41 1.00 0.62 22.52 0.23 1.00 1.00 6.57 Minimum -0.15 0.04 -0.91 0.00 -0.98 0.34 0.00 0.00 0.00 0.00 Std. Dev. 0.15 2.31 0.16 0.47 0.22 2.80 0.04 0.50 0.41 0.53 Skewness 2.12 0.30 -2.89 0.78 -0.99 5.62 2.19 0.22 -0.01 0.88 Kurtosis 8.71 2.75 16.81 1.61 7.08 38.13 9.24 1.05 1.24 5.92 Observations 11633 11633 11633 11633 11633 11633 11633 11633 11633 11633 5.2 Correlation analyses

Table 6 provides the correlation coefficients of all variables. The market-to-book ratio appear to have the highest positive correlation with the independent variable, which is cash. The cash flow appear to have the highest negative correlation with the cash variable. In the correlation matrix the level of correlation can be checked to see if there is multicollinearity. Correlation can take a range of values between -1 and 1, and correlation between -0.6 and 0.6 is acceptable (Lawrence and Lin, 1989). For this dataset, the level of correlation does not exceed -0.28 and 0.34. This leads to the conclusion that there is no multicollinearity in the sample.

Table 6: Correlation matrix

Cash Size

Cash

Flow Dividend NWC MTB Capex

R&D Exp. Leverage Industry Sigma Cash 1 -0.15 -0.27 -0.04 -0.12 0.22 -0.07 0.16 -0.02 0.13 Size -0.15 1 0.34 0.30 -0.08 -0.16 0.10 0.22 -0.19 -0.14 Cash Flow -0.27 0.34 1 0.29 0.18 -0.28 0.22 0.00 -0.05 -0.11 Dividend -0.04 0.30 0.29 1 0.13 -0.07 0.09 0.12 -0.04 -0.09 NWC -0.12 -0.08 0.18 0.13 1 -0.13 0.02 0.08 0.01 0.00 MTB 0.22 -0.16 -0.28 -0.07 -0.13 1 0.00 0.06 0.05 0.10 Capex -0.07 0.10 0.22 0.09 0.02 0.00 1 0.04 0.01 -0.05 R&D Exp 0.16 0.22 0.00 0.12 0.08 0.06 0.04 1 -0.04 0.09 Leverage -0.02 -0.19 -0.05 -0.04 0.01 0.05 0.01 -0.04 1 0.02 Industry Sigma 0.13 -0.14 -0.11 -0.09 0.00 0.10 -0.05 0.09 0.02 1 5.3 Regression results

(20)

20 Table 7: Regression estimating the determinants of cash holdings

The sample includes all Compustat and Datastream company-year observation from 2010 to 2017. Financial companies (SIC code 6000-6999) and utilities (SIC code 4900-4999) are omitted from the sample, yielding a panel of 13332 observations. Missing explanatory values reduce the panel to 11633 observations. Absolute value of t statistics is reported in parentheses. * stands for significant at 10%, ** stands for significant at 5% and *** stands for significant at 1%.

1 2 3 4 5 6 7

Model OLS OLS 2010-2013 OLS 2014-2017 BE FE Entity FE Time FE Both

Dependent Variable Cash / Assets Cash / Assets Cash / Assets Cash / Assets Cash / Assets Cash / Assets Cash / Assets

(21)

21 Model 1 is a regular Ordinary Least Square regression, where the cash to assets ratio is used as the dependent variable. Dividend, the market-to-book ratio, R&D expenses and the industry sigma all have positive signs and are statistically significant at the 1% level. This means that the cash ratio increases with these variables mentioned above. As seen in table 7, size, cash flow, net working capital, leverage and the capital expenditures all have a negative sign and are statistically significant at the 1% level. Looking at the results from the OLS regression in table 7 and the hypotheses development in chapter 2, we see that not all the results are in line with the hypotheses. The signs of the variables size, net working capital, leverage, the market-to-book ratio, R&D expenses, capital expenditures and the industry sigma are all in line with their expected signs. However, the signs from the cash flow and dividend are not in line with their expected signs. If we look at the theories about the transaction-, precautionary- and agency motive and the empirical findings we also see mixed expected signs for these variables. For example for the cash flow, the majority of the empirical findings found a positive relation, which is in line with precautionary motive. However, the transaction motive expected a negative sign. So for those variables which are not in line with the expected sign there always might be an explanation. Kim et al. (1998) mentioned in the literature about the transaction motive that the cash flow provides a ready source of liquidity, especially when the company has to meet operating expenditures and maturing liabilities. And therefore they assume that the cash flow is negatively related to cash holding. And according to Ivashina and Sharfstein (2010) the global financial crisis of 2008 had a great impact on the economy and companies. And prices across most asset classes fell drastically, the borrowing costs of corporates and banks rose substantially, and financial market volatility rose to levels that have rarely, if ever, been seen. Therefore, it is likely that companies used the cash flow as a ready source of liquidity instead of borrowing against high costs in the years right after the global financial crisis of 2008. This means that before the global financial crisis of 2008 the level of cash holding increased along with the cash flow, while after the global financial crisis the level of cash holding decreases when the cash flow increases. The R-squared is 15% for model 1.

(22)

22 significant in the time period 2014-2017, Dividend is in the time period 2014-2017 only statistically significant at the 5% level and capital expenditures is only statistically significant in the time period 2010-2013. The R-squared is 10% for model 2 and 19% for model 3.

The next model in table 7 is model 4, which shows the results of the between effects approach. In the between effects model, the between estimator uses time averages for all the dependent and independent variables. The capex is in this model not significant and also the sign is not in line with model 1, where the capex has a negative sign. For the rest all independent variables are significant and the signs are in line with the outcomes of model 1. The R-squared of this model is 25%.

As mentioned in section 4.3, we can reject the null hypothesis of the Hausman test, which means that the Fixed-Effects-Model is the appropriate model to use in this research. Therefore, the last three models all include different variations of the Fixed-Effects-Model. The three variations of the Fixed-Effects-Model that were applied are the entity fixed effects, the time fixed effects and as last a Fixed-Effects-Model which includes both the entity fixed effects and the time fixed effects.

(23)

23 The second Fixed-Effects-Model, which is model 6, is about the time fixed effects. In this test the period effect specification is set to fixed instead of none. Looking at the regression results in table 7, we see that the results of this model are not completely in line with the results of the most suitable model so far, which is model 5. The variables size, net working capital, dividend, market-to-book ratio and capital expenditures all have the same sign and significance level as model 5. While the cash flow and leverage both have a negative sign now and are statistically significant at the 1% level. R&D expenses now has a positive sign and is statistically significant at the 1% level. And the industry sigma is no longer statistically significant at the 1% level, but now only at the 5% level. A possible explanation for the differences between both the models is the use of a different approach. The entity fixed effects model uses entity-specific intercepts that capture heterogeneities across entities. And the time fixed effects model eliminates omitted variable bias, which are caused by excluding unobserved variables that evolve over time but are constant across entities. The R-squared of this model is 15%.

The last model, which is model 7, includes both the entity fixed effects and the time fixed effects. This combined model allows to eliminate omitted variable bias that evolve over time but are constant across entities and it controls for factors that are different across entities but stay constant over time. When comparing this model with the entity fixed effects model, which is model 5, we see that the results are almost completely in line. Only the variable industry sigma is slightly different. In model 5 this variable is statically significant at the 5% level, and now only at the 10% level. The R-squared of this model is also 80%, so this model is together with model 5, the most suitable model for the estimations on the determinants of cash holdings.

(24)

24 Table 8: Predicted signs versus the determined relation

Variables Predicted Sign Determined relation

Size Minus Minus

Cash Flow Plus Plus

NWC Minus Minus

Leverage Minus not significant

Dividend Minus Plus

MTB Plus Plus

R&D Exp. Plus not significant

Capex Minus Minus

Industry Sigma Plus Plus

5.4 Additional robustness analysis

The additional robustness analysis will be done by comparing separate entity fixed effects regressions from countries and industries with each other and the entity fixed effects regression of the entire sample. Table 9 shows the results of the top three countries, ranked by the number of unique company observations. And table 10 shows the results of the top three industries, ranked by number of unique company observations.

Table 9: Top 3 countries with the most observations

5 5 5 5

Model FE Entity France FE Entity Germany FE Entity Italy FE Entity Dependent Variable Cash / Assets Cash / Assets Cash / Assets Cash / Assets

(25)

25 The first thing that is noticeable when looking at table 9, is that only a few variables are statistically significant. This is the case because a small sample size decreases the statistical power. Secondly, when looking at the statically significant variables we see that two variables are not in line with the results of the entire sample. The variable size in model 1, which includes only companies from France, has a positive sign and is statistically significant at the 1% level. While in the entity fixed effects regression for the entire sample the variable has a negative sign and is statistically significant at the 1% level. And the variable market-to-book ratio in model 2, which includes only companies from Germany, is now slightly positive and statistically significant at the 5% level. While in the entire sample it is slightly negative and statistically significant at the 1% level. Therefore, it can be concluded that size is positively related to cash holdings for French companies, while for most of the companies out of other countries there is a negative relation. And for German companies there is a positive relation between the market-to-book ratio and the level of cash holding, while for most of the companies out of other countries there is a negative relation.

Table 10: Top 3 industries with the most observations

5 5 5 5

Model FE Entity Manufact. FE Entity Services FE Entity Transport. FE Entity Dependent Variable Cash / Assets Cash / Assets Cash / Assets Cash / Assets

(26)

26 The first thing that is noticeable when looking at table 10, is that the industries manufacturing and services both have a lot more statistically significant results than the transportation industry. When looking at the number of unique observations per industry we see that the transportation industry only counts 93 unique observations, while the other industries count 1028 and 292 unique company observations. The second thing that is noticeable, is that almost all the statistically significant results have the same sign in all the models. There are only minor differences in terms of economical value and in the significance levels. For example, the variable R&D expenses is not statistically significant in model 4, while in model 1 it is statistically significant at the 10% level. And the net working capital is in model 3 only statistically significant at the 10% level, while in the other models it is statistically significant at the 1% level.

6. Conclusion

This thesis examines the determinants of cash holdings for a sample of 1730 Eurozone listed companies over the period from 2010 to 2017. With the theories behind the transaction-, precautionary and the agency motive, we can find the determinants of cash holdings. This has led to the following research question: To which extent do specific firm characteristics have an

influence on cash holding of Eurozone companies after the global financial crisis of 2008?

(27)

27 The only hypothesis, which is not proven to be true is the one with dividend. The transaction- and the precautionary motive both predicted a negative relation between dividend and the level of cash holding. While in practice, we found a significant positive relationship. According to Bates et al. (2009) companies that are dividend paying hold less cash, because they can raise funds at low cost, and have better access to the capital market. Therefore, the conclusion is that dividend paying Eurozone companies hold more cash. One of the possible explanations for this could be the global financial crisis of 2008, where things as the great recession, the European debt crisis and the crisis in the banking system all were parts of. Another possible explanation could be the cultural differences, such as the better access to the capital markets for US companies and the better legal protection for investors in the US.

Limitations and future research

(28)

28

References

Almeida, H., Campello, M., Weisbach, M., 2004. The Cash Flow Sensitivity of Cash. Journal of Finance 59, 1777-1804.

Baskin, I., 1987. Corporate Liquidity in Games of Monopoly Power. Review of Economics and Statistics 69, 323-329.

Bates, T., Kahle, K., Stulz, R., 2009. Why do U.S. firms hold so much more cash than they used to?. The Journal of Finance 64, 1985‐2021.

Chen, Q., Chen, X., Schipper, K., Xu, Y., Xue, J., 2012. The sensitivity of corporate cash holdings to corporate governance. Review of Financial Studies 25, 3610-3644.

Fama, E., MacBeth, J., 1973. Risk, return, and equilibrium: empirical tests. Journal of Political Economy 81, 607-636.

Ferreira, M., Vilela, A., 2004. Why do firms hold cash? Evidence from EMU countries. European Financial Management 10, 295-319.

Fujiki, H., Kitamura, Y., 1995. Feldstein-Horioka Paradox Revisited. Bank of Japan. Monetary and Economic Studies 13, 1-16.

García-Teruel, P., Martínez-Solano, P., 2008. On the Determinants of SME Cash Holdings: Evidence from Spain. Journal of Business Finance & Accounting 35, 127-149.

Hausman, J., 1978. Specification tests in econometrics. Econometrica 46, 1251–1271.

Ivashina, V., Scharfstein, D., 2010. Bank lending during the financial crisis of 2008. Journal of Financial economics 97, 319 – 338.

Jensen, M., Meckling, W., 1976. Theory of the firm: Managerial behaviour, agency costs and ownership structure. Journal of Financial Economics 3, 305‐360.

(29)

29 Kim, C., Mauer, D., Sherman, A., 1998. The determinates of Corporate Liquidity: Theory and Evidence. Journal of Financial and Quantitative Analysis 33, 335-359.

Lawrence, I., Lin, K., 1989. A Concordance Correlation Coefficient to Evaluate Reproducibility. Biometrics 45, 255-268.

Miller, M., Orr, D., 1966. A model of the demand for money by firms. Quarterly Journal of Economics 80, 413–435.

Modigliani F., Miller M., 1958. The cost of capital, corporation finance and the theory of investment. The American Economic Review 48, 261-297.

Nykvist, J., 2008. Entrepreneurship and liquidity constraints: evidence from Sweden. The Scandinavian Journal of Economics 110, 1113-1155.

Ogundipe, L., Ogundipe, S., Ajao, S., 2012. Cash holdings and firm characteristics: Evidence from Nigerian emerging market. Journal of Business, Economics & Finance 1, 45–58.

Opler, T., Pinkowithz, L., Stulz, R., Williamson, R., 1999. The determinants and implications of corporate cash holdings. Journal of Financial Economics 52, 3‐46.

Ozkan, A., Ozkan, N., 2004. Corporate cash holdings: An empirical investigation of UK companies. Journal of Banking and Finance 28, 2103-2134.

Porta, R., Lopez‐de‐Silanes, F., Shleifer, A., Vishny, R., 1998. Law and Finance. Journal of Political Economy 106, 1113-1155.

Rajan, R., Zingales, L., 1995. What do we know about capital structure? Some evidence from international data. The Journal of Finance 50, 1421-1460.

Shin, M., Kim, S., Shin, J., Lee, J., 2018. Earnings Quality Effect on Corporate Excess Cash Holdings and Their Marginal Value. Emerging Markets Finance & Trade 54, 901 - 920.

Wooldridge, J., 2002. Econometric Analysis of Cross Section and Panel Data. MIT Press, Cambridge.

(30)

30 Wooldridge, J., 2013. Introductory Econometrics: A Modern Approach. MIT Press,

(31)

31

Appendix

Appendix 1: Hausman Test

Test Summary Chi-Sq. Statistic Chi-Sq. d.f. Prob.

Cross-section random 542.893566 9 0.000

Variable Fixed Random Var(Diff.) Prob.

Size -0.01290 -0.01331 0.00000 0.82610 Cash Flow 0.09626 0.02297 0.00002 0.00000 NWC -0.11942 -0.10908 0.00001 0.00000 Leverage 0.00046 -0.00475 0.00000 0.00000 Dividend 0.00790 0.00755 0.00000 0.69390 MTB 0.00196 0.00327 0.00000 0.00000 R&D Exp. -0.00469 0.01870 0.00000 0.00000 Capex -0.15668 -0.16729 0.00006 0.17660 Industry Sigma 0.00333 0.00516 0.00000 0.00000

The Determinants of Cash Holdings: Evidence from Eurozone companies after the global financial crisis