The effect of financial flexibility on firm performance during the recent financial crisis in Holland

1

‘The effect of financial flexibility on firm performance

during the recent financial crisis in Holland’

Name: Harry van Sprundel Student number: 10671609 Supervisor: Jeroen van de Ven Study track: Organization Economics ECTS:15

2 Abstract

In this thesis I have tested how financial flexibility has affected firm performance in Holland during the recent financial crisis. I have gathered data from annual reports of 73 firms that are listed at the NYSE Euronext Amsterdam. I have failed to find convincing evidence that financially flexible firms performed better during the crisis than less financially flexible firms. However, when I compared the effects of financial flexibility for small and big firms I found that small firms might have benefitted during the crisis from being financially flexible before the crisis, although the effects did not have a significant impact.

1. Introduction

In 2007 the average profit of firms from Holland that were listed at the NYSE Euronext Amsterdam was a little over €1 billion. In 2009 this average had decreased to €375 million and even in 2013 the average profit was nowhere near the level of 2007 with an average of €506 million. The recent financial crisis has had a big impact on firms in Holland as these numbers show. What could firms have done to minimize the effects of the crisis and knowing this, how should they prepare for future crises?

In this thesis I will look at how financial flexibility has affected firm performance during the recent crisis. I have tested how being financially flexible prior to the crisis has affected firm profits during the crisis. Also, I will test how flexibility matters for firms that operate in unstable industries. I will partly continue on the work of Arslan-Ayaydin et al. (2014) in the sense that they have looked at how financial flexibility has affected firm performance during two crises in East-Asia, whereas I will test the same effect but look at a different economic environment, namely firms in Holland and I will only concentrate on the most recent financial crisis.

I have found some evidence that financially flexible firms have performed better than less flexible firms. This evidence was shown when I looked at how financial flexibility affected firm performance during the crisis for small firms. Looking at the effect of flexibility on firm performance in general, I did not find convincing evidence that showed that firms might have benefited during the crisis from being financially flexible before the crisis.

3 The rest of this thesis is structured as follows: section 2 will contain a literature review that concerns the topic of financial flexibility. After that I will explain how I have gathered my data and present some descriptive statistics in section 3. Section 4 includes the analysis in which I will test my hypotheses and in the final section 5 I will describe my conclusions and possible limitations of this research.

2. Related literature 2.1 Financial Flexibility

Financial flexibility represents the ability of a firm to respond effectively to unanticipated shocks to its cash flows or its investment opportunities (Bancel & Mittoo, 2011). Also, Gambra & Giantis (2008) state that: ‘financially flexible firms are able to avoid financial distress in the face of negative shocks’. There are different measures of financial flexibility used in the existing literature. One of these measures is how much borrowing power a firm has. A firm’s borrowing power could determine how financially flexible a firm is, because when firms face a decrease in sales and need extra cash they can borrow money easier if their debt levels are lower (Denis and Sibilkov (2010), and Marchica and Mura (2009)). Another determinant of financial flexibility is the firm’s cash holdings (Almeida and Campello, 2007) and (Ayaydin et al.,2014)). The same reasoning as with borrowing power applies to the firm’s cash holdings. When firms have lower revenues, they will be better able to handle the drop in demand if they have higher cash reserves. Higher cash holdings means that firms are better able to make required investments and are better able to cover losses. These advantages of financial flexibility are summarized by DeAngelo and DeAngelo (2007). They state that: ‘the importance of financial flexibility stems from the fact that financially flexible firms are better able to raise cash and rearrange their capital structure at a lower cost’. The opportunity to raise cash and restructure the firm’s financing is important when firms face a sudden change in demand, something which happens during an unexpected shock like a financial crisis. There is however a cost which firms have to incur to become financially flexible. Firms that have high cash holdings will have to pay tax over these holdings and a lower level of debt means that firms have a lower tax shield (Gamba and Triantis, 2008). Because of these costs of tax and loss of tax shields, firms will not always chose

4 to be fully financially flexible. When choosing a firm’s financial flexibility the firm will always have to make a trade-off between the costs and benefits of this flexibility.

2.2 Downside of Financial Flexibility

As mentioned earlier, firms might benefit from being financially flexible in times of crises, because they are better able to restructure their finances and should be better able to avoid financial distress. There is however also a downside of being financially flexible during a crisis. Most research concerning financial flexibility assume that when firms face negative shocks and need more lending they can turn to banks to receive new loans. When firms have low levels of debt it will be easier for these firms to attract new loans since they should be better able to pay their loans back. It has been shown however that during the recent crisis it was much more difficult to get a new loan than during normal times (Ivashina and Scharfstein, 2008), (Kahle and Stulz, 2013). Ivashina and Scharfstein (2008) showed that during the recent crisis loans to large borrowers fell by 48%. This large decline means it might be more difficult to restructure a firm’s capital structure than most research assumes. If the benefit of capital restructuring for financially flexible firms is less existent and the firm still has to incur the costs of being financially flexible than the firm might actually have a disadvantage of being financially flexible during a crisis. Also, during a crisis there will not be so many investment opportunities available as in normal times which could also diminish the benefits of being flexible for a firm during a crisis. So what exactly is the effect of financial flexibility on firms during a crisis? I will discuss several papers that have examined the effects of financial flexibility on firm performance during times of crisis.

2.3 Effect of Financial Flexibility during a crisis

Campello et al. (2010) found in a survey amongst CFO’s of firms in the Unites States, Europe and Asia that financially constrained firms suffered more from the credit crisis than less financially constrained firms. In their study of French private and publicly listed firms, Bancel and Mittoo (2011) surveyed CFO’s to answer the question if financially flexible firms have performed better than less financially flexible firms during the recent credit crisis. A questionnaire was sent to 8000 CFO’s of which 34 replied. To increase their dataset Bancel &

5 Mittoo (2011) also conducted 11 one-to-one interviews with CFO’s. The measure of financial flexibility that the authors used in their study was both high cash holdings and a low level of debt. Bancel & Mittoo (2011) found that financially flexible firms indeed performed better during the recent crisis than less financially flexible firms. A strong point of this study is the extensive explanation of financial flexibility and how this flexibility determines firm performance during a crisis. The downside of the use of surveys that the authors chose for is that the dataset Bancel & Mittoo (2011) use is quite small, although the authors have countered this problem by performing robustness checks. What I will contribute to this study is that I will not conduct surveys amongst CFO’s but instead use annual reports of firms to gather data. Because the data is gathered from annual reports I will have a more objective measurement of financial flexibility.

The use of data from annual reports as a measure of financial flexibility is not entirely new tough. Arslan-Ayaydin et al. (2014) have used accounting and market variables from Thomson Reuters DataStream to look at how financial flexibility influenced firm performance in East-Asia in the period 1994-2009. During this period there were two different crises, in 1997-1998 there was the East-Asian crisis and in 2007-2009 there was the global credit crisis. Arslan-Ayaydin et al. (2014) have used data from 1068 listed firms from Hong Kong, Indonesia, Malaysia, South-Korea, and Thailand. The authors have excluded the Philippines, Singapore, and Taiwan from their study, because these countries suffered only mildly from the crisis. The measurement of financial flexibility that the authors used was the leverage-ratio and the cash holdings of a firm. Arslan-Ayaydin et al. (2014) found that firms that are financially flexible performed better during the crises than less flexible firms. This result however was more significant for the 1997-1998 crisis than for the credit crisis of 2007-2009. The most important determinant for better firm performance during crises was a combination of both low-debt levels as well as high cash holdings. The strongest points of this study are the size of the dataset and the use of objective measurements. Because the authors use objective measurements for financial flexibility it is possible to compare the results of firms in East-Asia with firms in other regions. The authors also showed that the results differ per region, which is why I will contribute to this study by

6 looking at how financial flexibility has determined firm performance in Holland, a country that was hit hard by the financial crisis and which has different macroeconomic policies.

Although there are of course costs that a firm must incur to become financially flexible (Gamba and Triantis, 2008) most literature assumes firms benefit during a crisis from being financially flexible. Campello et al. (2010), Bancel & Mittoo (2011) and Ayaydin et al. (2014) have all shown that firms that are financially flexible perform better during crises than less flexible firms. My contribution to these studies will be that in contrary to the studies of Campello et al. (2010) and Bancel & Mittoo (2011) I will use data from annual reports of Dutch firms in the years 2004-2013. By using data from annual reports I will have a more objective measure of financial flexibility than the measures obtained from surveys that the previously mentioned papers have. Arslan-Ayaydin et al. (2014) have also used annual reports and my contribution to this paper will be that I use data from Holland, a country which has different macroeconomic policies than East-Asian countries and a country that was hit hard by the recent crisis. Arslan-Ayaydin et al. (2014) stated in their paper that the results from their study differed greatly per country and region which makes it interesting to see how these results will be in Holland.

2.4 The Dutch Economy

The Dutch economic environment is different than the countries examined by Arslan-Ayaydin et al. (2014) in a few ways. First of all the size of the economy is quite different. In 2013 GDP per capita in Holland was $41,711 compared to $17,748 in Malaysia, $9,875 in Thailand and $5,214 in Indonesia (Worldbank data) which were 3 of the 5 countries from the dataset of Arslan-Ayaydin et al. (2014). Another important aspect of the Dutch economic environment is the amount of investments made (Hagima, 2013). Holland is placed 6th on the list of the Worldbank data regarding foreign direct investment received compared to South-Korea’s 25th, Thailand’s 32th and Malaysia’s 34th place. The effects of the global financial crisis on Holland were quite large. Holland’s GDP decreased with 4% in 2009 and consumer confidence plunged dramatically between 2007 and 2009 (Bronner and de Hoog, 2012). Because the effect of the crisis was so big and because the economic environment of Holland is quite different from the economic environment of the countries examined by Arslan-Ayaydin et al. (2014), I think it will be

7 interesting to look at the effect of financial flexibility on firm performance during the recent crisis in Holland.

2.5 Uncertain industries

Also, I think it will be interesting to see if firms operating in an uncertain industry benefit more from being financially flexible than firms that operate in more stable industries. Merschmann & Thonemann (2011), found in a study amongst German supply chain flexibility that firms that operated in an uncertain environment benefitted from being flexible. It seems logical that it will be more important for firms that work in industries where profits fluctuate a lot on a year-to – year basis to be flexible than for firms where each year’s profit is more predictable. As a variable of uncertainty I will use the industry’s profit variation. Another reason why firms that operate in an uncertain industry might benefit more from being financially flexible is explained by Eppink (1978) and Gerwin (1994). They state that: ‘firm flexibility is defined as the ability of a firm to deal with uncertainty’ (Gerwin, 1994). Also: ‘a firm that has a more flexible organization structure should be better able to handle unforeseen changes in its environment’ (Eppink, 1978). Firm flexibility and financial flexibility are not the same, but financial flexibility is part of a firm’s flexibility. To my knowledge there is no clear, objective measure of firm flexibility, but there is a measure of financial flexibility which I will be using and I think it could help me to show if firms that operate in uncertain industries can benefit from being financially flexible. 2.6 Lessons from the related literature

What can we learn from these studies? With respect to the effect of financial flexibility it becomes clear a lot of studies have shown the positive effect financial flexibility can have on firm performance during a crisis. However, most studies use questionnaires to measure flexibility and this might be biased. Arslan-Ayaydin et al. (2014) use an objective measure of financial flexibility namely the cash- and debt-ratios of a firm. The advantage of this measure is that these ratios are visible for everyone and are thus not biased by managers who overstate their firm’s flexibility. The disadvantage of this measure is that it may be too narrow. A firm’s assets could also be part of its flexibility since firms can sell these assets in times of financial distress and they can thus act as some sort of cash reserves. Furthermore it could be that a

8 firm’s reputation is also important in times of financial distress. Existing literature assumes that firms who have a lower amount of existing debt are better able to attract new debt than firms who are more leveraged. It could however also be argued that a firm’s reputation is more important when attracting new debt. Especially in difficult times such as a financial crisis banks may be more critical when providing loans and they could favor more reputable firms over less reputable firms, even if the more reputable firms are higher leveraged. Finally, Arslan-Ayaydin et al. (2014) showed that the effects of financial flexibility on firm performance were much less visible in the global financial crisis than in the East-Asian crisis. It could be that because I am studying the same crisis as the crisis in which Arslan-Ayaydin et al. (2014) found financial flexibility to be less important, I could find that financial flexibility does not have such a great impact on firm performance. There are different studies that argue that firms who operate in uncertain environments can benefit from flexibility. However, studies such as Eppink (1978), Gerwin (1994) and Marschmann & Thonemann (2011) do not concentrate solely on financial flexibility but more on firm flexibility as a whole. It seems logical that if firm flexibility is important in uncertain areas that financial flexibility can benefit firms that operate in uncertain industries, but my analysis will still have to prove this.

3. Data 3.1 Data collection

The data is collected from firm’s annual reports. I have started by looking at all firms that are listed at NYSE Euronext Amsterdam. In total, 177 firms were listed at that time. From this I have excluded firms that are not based in Holland. I have also excluded financial institutions as they have to oblige to certain regulations regarding their cash- and debt-ratios which would cause an unnatural comparison. There were also 2 firms (OCI and Ziggo) that had only recently been working under their current structure and therefore did not have any activities prior to the crisis. Because of this lack of data in the years 2004-2008 I have excluded these two firms from the dataset. Finally, there are 2 firms listed at the NYSE Euronext Amsterdam at the moment which were 1 entity before 2011, namely PostNL and TNT Express. Before 2011 these 2 firms operated under 1 name; TNT Express. Because of this I have chosen to only use the data of TNT

9 Express in my analysis. After these exclusions there are 73 firms left in the dataset. From these 73 firms I have looked at their cash-to-asset ratio, their debt-to-asset ratio, their profit, their size and their age in the years 2004-2013. I have used return on assets as a performance measure since it will be very likely that bigger firms have bigger profits. Therefore, if I use the return on assets I will be better able to make a good comparison in firm performance. It was essential for me to have data from some years before the crisis to be better able to check the effect of financial flexibility on firm performance. I have therefore chosen to use the years 2004-2007. I might have tried to use data from years before 2004, but a lot of firms hadn’t recorded data from before 2004 which would make it more difficult to make a good comparison. In total I have 730 observations and 0 missing observations.

Table 1: Summary of statistics Observations: 730

Variable Mean Std.Dev. Minimum Maximum

Cashratio 0.089 0.1099 0 1 Debtratio 0.216 0.1591 0 0,92 Assets (€mln) 8,460 44,135 2.8 464,766 Profit (€mln) 627.6 3,7284 -2,106 41,504 Age 69.048 68.942 3 360 3.2 Descriptive statistics

A summary of the most important statistics can be found in table 1. As the table shows the mean cash-ratio between 2004 and 2013 was 8.9%, while the mean debt-ratio was 21.6%. The lowest debt- and cash-ratio in the timespan was 0% while the maximum debt- and cash-ratio was 0.92 and 1 respectively. The average firm size was €8.46 billion in assets and the average realized profit between 2003 and 2013 was €627.6 million per year. The smallest firm had a total assets worth of €2.8 million, while the biggest firm had at one point a total assets worth of over €464 billion in total assets. The biggest loss a firm had incurred between 2004 and 2013 was €2.1 billion in a year, while the largest profit a firm had realized was €41.5 billion in a year. The average age of the firms in the dataset I’ve used was 69 years. The youngest firm in my

10 dataset was 3 years old and the oldest firm in my dataset was 360 years old. After the summary of the variables I wanted to see how the values of cash-ratio, debt-ratio and profits have evolved over the time-span 2004-2013 so I summarized the variables again, but this time looked at the values per year instead of the total timespan. As table 2 shows, all important values for debt-ratio hardly change over the years. The mean debt-ratio, the standard deviation and the minimum don’t really change over the years. The mean changes from 0.21 in 2004 to 0.20 on 2013. To test the significance of this difference we can use a simple t-test. The t-test looks as follows: (0.21-0.20)/ ((√(0.15x0.15/73+0.15x0.15/73)=0.01/0.0245= a t-value of 0.408 meaning that the change in the means is not significant. The only value that changes a bit is the maximum value; it decreases somewhat over the years from 92% of the total assets in 2004 to 79% of total assets in 2013. Apparently, firms manage on average to maintain around the same levels of leverage with respect to their assets. The latter makes sense as firms often have a certain level of debt as a goal in order to fully use their tax shield.

Table 2: Change in Debt- and Cash-ratios per year Observations per year: 73

Year Variable Mean Std.Dev. Minimum Maximum

2004 Debt-ratio 0.21 0.15 0 0.92 Cash-ratio 0.11 0.12 0 0.52 2005 Debt-ratio 0.21 0.16 0 0.90 Cash-ratio 0.11 0.13 0 0.73 2006 Debt-ratio 0.21 0.16 0 0.91 Cash-ratio 0.10 0.15 0 1.00 2007 Debt-ratio 0.21 0.16 0 0.90 Cash-ratio 0.09 0.13 0 0.75 2008 Debt-ratio 0.24 0.17 0 0.89 Cash-ratio 0.08 0.10 0 0.61 2009 Debt-ratio 0.23 0.16 0 0.90 Cash-ratio 0.09 0.10 0 0.56 Debt-ratio 0.22 0.15 0 0.87

11 2010 Cash-ratio 0.07 0.07 0 0.30 2011 Debt-ratio 0.22 0.16 0 0.85 Cash-ratio 0.07 0.07 0 0.36 2012 Debt-ratio 0.21 0.17 0 0.81 Cash-ratio 0.07 0.07 0 0.32 2013 Debt-ratio 0.20 0.15 0 0.79 Cash-ratio 0.09 0.11 0 0.76

Table 2 also shows how the cash-ratios of the firms in the dataset have evolved in the years 2004-2013. The mean value of the cash ratios has slowly decreased from 0.11 in 2004 and 2005 to 0.7 in 2010, 2011 and 2012 before slightly increasing to 0.09 in 2013. The t-value of the change in the mean value of the cash ratios between 2004 and 2010 is 2.5 (0.11/√(0.12x0.12/73+0.07x0.07/73), which means this change is significant at a 5% level. The minimum ratio of 0 stays the same over the entire period. The maximum cash-ratio shows a sharp decline after 2006. From 2004 until 2006 it rises from 0.52 to 1 in 2006. After 2006 it decreases to 0.32 in 2012 and in 2013 there was a sharp increase in the maximum value. The value of 1 in 2006 seems strange, but this was a very young firm with a small amount of assets and had solely cash as assets. Unlike the debt-ratio, the cash-ratios do show the impact of the crisis. Apparently firms use their cash holdings to cover losses or use for investments which is shown by the significant decrease in the mean value of the cash-ratios over the years.

3.3 Impact of the crisis

The impact of the financial crisis is best shown in table 3. All values of firm profit show that firms have suffered from the crisis. The mean value of the profit was increasing in 2004-2007 from €503 million in 2004 to over €1 billion in 2007. After this the mean value dropped heavily to 540 in 2008 and even further to 375 in 2009. 2010 and 2011 show a small recovery but in 2012 and 2013 the average profit declined again. The minimum loss also shows the impact of the crisis. Before 2008 the biggest loss a firm in this dataset had was €40 million. This increased a lot after 2007. In 2008 the biggest loss was over €1 billion and in the years after that, with the

12 exception of 2010, the biggest loss was very high with the biggest loss in 2012 of over €2 billion. Even the best performing firm suffered from the consequences of the crisis. In 2007 the best performing firm realized a profit of €41.5 billion while this decreased to €16.5 billion in 2009 and €26.6 in 2010. After 2010 the maximum amount of profit increased a bit again but in 2013 the amount decreased again to a lower level than in 2004.

The crisis has hardly affected how the firms in the dataset have grown in terms of total assets. All values show an increase of total assets every year except for 2009, in that year the average total assets slightly decreased as well as the maximum amount of assets. The standard deviation also increased a lot over the years, indicating that the differences per year got bigger every year.

Table 3: Change in profit and assets in € millions per year Observations per year: 73

Year Variable Mean Std.Dev. Minimum Maximum

2004 Profit 503 2.789 -12.6 23.637 Assets 6.142 28.982 2.8 243.679 2005 Profit 680 3.935 -40 33.394 Assets 6.857 33.787 3 285.370 2006 Profit 766 3.944 -24.7 33.074 Assets 7.233 36.119 3.4 305.859 2007 Profit 1024 4.987 -5 41.504 Assets 8.016 41.220 3.6 350.311 2008 Profit 540 4.033 -1021 34.419 Assets 8.581 45.575 3.7 370.600 2009 Profit 375 2.002 -222.6 16.533 Assets 8.400 43.137 4 367.121 2010 Profit 610 3.162 -29 26.616 Assets 9.448 49.227 4.1 419.328 2011 Profit 683 4.756 -1456 40.451 Assets 9.940 52.678 4.1 448.834

13 2012 Profit 590 4.142 -2106 35.048 Assets 9.825 53.458 4.1 455.383 2013 Profit 506 2.591 -695.9 21.483 Assets 10.169 54.546 4.2 464.766

3.4 Tradeoff in cash- and debt-ratio

As Gamba and Triantis (2008) have explained, firms have to find a balance for the right amount of cash holdings and leverage. When a firm has too high leverage and too short cash holdings, it faces the possibility of financial distress. However, firms pay taxes on their cash holdings and debt causes their tax shield to increase. Also, cash and debt that is not used for investments may cause firm performance to be suboptimal. It is therefore possible that firms with very low cash ratios do just as bad as firms with very high cash ratios and the same reasoning might hold for leverage. I have therefore created two new variables which are the squared ratios of cash and debt. A graph of these variables should show a parabola if the reasoning of a balance in cash- and debt ratios applies. Graphs 1 and 2 show that most observations are concentrated in the lower levels of cash- and debt-ratios. It does not show the expected results that should show that a certain level of cash- and or debt ratio is best for maximizing profits.

What is striking from these figures is that the variance in profit is much larger for the lower levels of financial flexibility. Of course there are much more observations in the lower levels of flexibility which could explain why there are more outliers on the left of the figures. The figures show that the right side of the figure, which represents higher levels of cash- and debt ratios, have much fewer observations than the left side of the figure.

14 -1 .5 -1 -. 5 0 .5 Pro fi t/ Asse ts 0 .2 .4 .6 .8 debtsquared

scatterplot of debtsquared on return on assets

Graph 1

Scatterplot of cashratio-squared on return on assets

15 4. Analysis

The two hypotheses that I will test in this analysis are: ‘financially flexible firms have performed better than less financially flexible firms during the recent crisis in Holland’ and financial flexibility has had a bigger impact for firms operating in uncertain industries’. Studies such as Arslan-Ayaydin et al. (2014), Bancel & Mittoo (2011) and Campello et al. (2010) all found that financially flexible firms performed better during a financial crisis. The second hypothesis is based on the study of Merschmann & Thonemann (2011) which showed that German manufacturers that operated in uncertain industries benefitted from supply chain flexibility. 4.1 Choice of model

The most important assumption of OLS is that the independent variables are uncorrelated with the error term. In this case however I will probably have problems with endogeneity. It is very likely that I will have an omitted variable bias, because there will be a lot of variables that are correlated both with my dependent as my independent variables which I will be unable to capture in a model. Another cause of endogeinity could be reverse causality. It could very well be that firms that have performed relatively better are also better able to be financially flexible simply because they have more resources to save or lack the need to borrow more money. To be able to make a causal claim between financial flexibility and firm performance I would then need an exogenous variable which is uncorrelated with both the independent as well as the dependent variables. However, this variable has to be uncorrelated with all the variables of the model, observed and unobserved. In the model I use there will most likely be a lot of unobserved variables which will be correlated somehow to the exogenous variable, a phenomenon called unobserved heterogeneity. Because of this problem it will be impossible to find a truly exogenous variable.

Because my dataset contains panel data I am able to solve these previously mentioned problems by using a fixed or a random effects model. These types of models can help counter the previously mentioned problems and should help me to find a causal relationship between financial flexibility and return on assets. A fixed effects model describes the relationship between independent and dependent variables over time within an entity, whereas a random

16 effects model describes the relationship both within and between entities. The entities are the companies in my dataset. In a random effects model the variation across entities is assumed to be random and uncorrelated, unlike the fixed effects model. In a fixed effects model it is assumed that if an unobserved variable does not change over time, any changes in the dependent variable must come from changes in the independent variables. By using a fixed effects model I should be better able to check for a causal relationship between financial flexibility and firm performance. However, if the variables change slowly over time a fixed effects model would not work very well, but I assume in my model that they don’t. Also, in a fixed effects model you can’t control for time invariant changes because a fixed effects model controls for changes within an entity and time invariant changes are assumed to be constant for each entity. The latter can be solved by a random effects model though.

It is most common in the related literature to use a fixed effects model, but there is a test which should help me decide whether to use a fixed or a random effects model. To be sure which model to use I will use a Hausman test. This test checks whether the error terms of the entities are correlated or not. If they are not correlated, I will use a fixed effects model. If they are, I will use a random effects model. The null hypothesis of this test is that the errors are not correlated with the independent variables. The prob>chi-squared value of the Hausman test should be lower than 0.05 and in this case the value was 0.00 which strengthens my belief that I should be using a fixed effect model to test my hypotheses.

The model I will test looks like this: Yit = β1X1t + β2X2t + β3X3t+ β4X4t αi + uit. In this model Y stands for profit in terms of assets (ROA), X1 stands for a firm’s cash ratio, X2 stands for a firm’s debt ratio, X3 stands for the control variable age and X4 stands for the control variable 4 firm size, α stands for the unknown intercept of each entity and uit stands for the error term. The two assumptions I make here are that the individuals may contain something that may impact the predictor or outcome variables. By using a fixed effect model I should be able to control for this and predict the effect of the independent variables on the dependent variable. The second assumption I make here is that the error terms of the entities are not correlated, something which I have tested with the Hausman test.

17 4.2 The determinants of firm performance in 2004-2013

The first thing I will test is what the effect of a firm’s debt- and cash-ratio has been on return on assets in the years 2004-2013, while controlling for firm age and firm size. The latter will be measured as total assets. The results of this can be found in table 4. I have used robust standard errors to control for heteroskedasticity. After I had run the fixed effects regression I performed a test in Stata with the command testparm to check if I needed to add time fixed effects. The null hypothesis was that I needed to add time fixed effects against the alternative hypothesis that I didn’t need to add them. The prob>F value was 0.1924 which is higher than 0.05 so I will not add time fixed effects in my model.

Table 4: FE regression of ratios on ROA Observations: 730

Variable Coefficient Standard Error

Cash-Ratio 0.091 0.099

Debt-Ratio -0.139 0.084

Assets 6.93e-08 9.89e-08

Age -0.007 *** 0.0015 *=sig. at 10% level, **=sig.at 5% level, ***=sig. at 1% level

Table 4 shows how the different predictors have affected profits in terms of assets during 2004-2014. The coefficient of 0.091 of cash-ratio means that for every increase of 1 in cash-ratio the return on assets of a firm increases on average with 0.091. This means that if firms have a 10% increase in cash ratio they had an average extra return on assets of 0.9%. The positive sign of cash ratio is as expected, just as the negative sign of debt ratio. However, both variables did not appear to have a significant effect on ROA during the years 2004-2014 in my dataset. The variable age did have a significant effect on ROA in this regression. For every 1 year in extra age that firms had, they would have an average of 0.7% decrease in ROA. It seems that this effect is quite small and it also doesn’t really make any sense, but apparently relatively younger firms performed a little bit better in terms of return on assets.

18 4.3 The impact of financial flexibility prior to the crisis on firm performance during the crisis In the previous section I have examined the determinants of firm’s return on assets in the years 2004-2013. However, there might very well be causality in this case. It could be that firms that perform better are more financially flexible because they don’t need to use their cash holdings or extra debt to survive. This would mean that firm performance in one year explains financial flexibility in the same year. I will therefore test how firms that were financially flexible prior to the crisis performed during the crisis. This should show how financial flexibility affects firm performance in the future. Different studies such as Maier et al. (2011) and Yerkes et al. (2011) said that the financial crisis started between the years 2007 and 2008 in the Netherlands. Table 3 shows that the biggest impact of the crisis on firm profit has been in 2008. It will therefore be interesting to see how firms that were financially flexible before 2008 have performed during 2008-2013. To test this I have created two new variables; cash- and debt-ratio before 2008. Table 2 shows that on average; the cash-ratio prior to 2008 was 10% and it also shows that on average, the debt-ratio prior to 2008 was 21%. I have created two dummy variables in Excel. The variable cash-high measures 1 if a firm had a higher than average cash-ratio prior to 2008 and the variable debt-low would measure 1 if a firm had a lower than average debt-ratio prior to 2008.

Table 5: Summary of cashprecrisis and debtprecrisis Number of obs.: 730

Variable Mean Std.Dev. Minimum Maximum

Debt-Low 0.493 0.500 0 1

Cash-High 0.384 0.487 0 1

As table 5 shows, there are about the same amount of firms that had a higher than average debt-ratio before the crisis as there were firms with a lower than average debt-ratio which is represented by the mean value of 0.493 which is almost 50%. The mean value of cash-high is 0.384 which indicated that there were more firms that had a lower than average cash-ratio prior to the crisis than there were firms that had a higher than average cash-ratio.

19 To test how being financially flexible before the crisis has affected firm performance during the crisis I will probably not be able to use the previously used fixed effect model. If firms had a higher than average debt-ratio before the crisis it could very well be that they also had a lower than average cash-ratio before the crisis. If this were true, the predictors would be correlated and then I will not be able to use a fixed-effects model anymore. In that case I would have to use a random-effects model. To be sure which model to use I will use a Hausman test again. The results of the Hausman test show that Prob>Chi-squared is 0.3487, which is much larger than the 0.05 that is the maximum value which it could be for me to be able to use a fixed effects model. The Hausman test has thus confirmed my belief that I need to use a random effects model to test the effect of financial flexibility before the crisis on firm performance during the crisis. The random-effects model is slightly different from the fixed effects model and looks like this: Yit = β1X1t + β2X2t + β3X3t + β4X4t + α + uit + εit. In this model Yit is the return on assets, X1 is the dummy variable cash-high, X2 is the dummy variable debt-low, X3 is the control variable age, X4 is the control variable assets, α stands for the intercept, uit stands for the between-entity error and εit stands for the within-entity error. The results of the random effects regression can be found in table 6. I have also created two graphs two show how firms that had a higher than average cash-ratio before 2008 have performed over the years compared to firms that had a lower than average cash-ratio before 2008. I’ve done the same for the debt-ratio. The results of this can be found in graph 3 and 4.

Table 6: Random-Effects Regression of Financial Flexibility before 2008 on return on assets after 2008

Observation:365

Variable Coefficient Robust Standard Error

Cash-High 0.0290 0.0189

Debt-Low -0.0096 0.0182

Age 0.0001 0.0001

Assets 1.65e-07 *** 4.96e-08 *=sig. at 10% level, **=sig.at 5% level, ***=sig. at 1% level

20 -1 .5 -1 -. 5 0 .5 2004 2006 2008 2010 2012 2014 Year

ProfitAssets, Cashhigh == 0 ProfitAssets, Cashhigh == 1

Fitted values Fitted values

How cash-rich firms before 2008 performed over the years

Graph 3

ProfitAssets, Leveragelow == 0 ProfitAssets, Leveragelow == 1

Fitted values Fitted values

How low-debt firms have performed over the years

21 What graph 3 shows is that firms that had a higher than average cash-ratio before 2008 performed better over the years than firms that had a lower than average debt-ratio before 2008 as can be seen by the fitted orange line which is slightly above the fitted green line. Graph 4 shows the opposite, in this graph it seems that over the years the green line is just above the orange line. The latter indicates that firms that had a higher than average debt-ratio before 2008 performed slightly better over the years than firms who had a lower than average debt-ratio. To be sure of this I will perform several regressions in a random effects model and the first results can be found in table 6.

Table 6 shows how being financially flexible before the crisis has affected ROA during the crisis. The coefficients of a random-effects model are a bit more difficult to interpret than the coefficients of a fixed effect model. The coefficient describes the impact of x over y when x changes both over time and between entities. In this case this would mean how x changes on average per year and between companies. The variable cash-high is a dummy variable that measures 1 if a firm had a higher than average cash-ratio before the crisis. The variable leverage-low is a dummy variable that measures 1 if a firm had a lower than average debt-ratio before the crisis. The results from the random-effects regression show that having a higher than average cash-ratio before the crisis had a positive impact on return on assets during the crisis. If a firm had a higher than average cash-ratio before the crisis it would on average have a higher ROA of 2.9% during the crisis in comparison with firms that had a lower than average cash-ratio. This effect was not significant however. The debt-ratio variable did not have the expected sign. The results from table 6 show that firms with a lower than average debt-ratio before the crisis actually had a lower ROA during the crisis of 0.96% than firms who had a higher than average debt-ratio. This effect was also not significant tough. The only variable that appeared to have a significant effect on ROA during the crisis is assets, but this effect was very small as is shown by the extremely low coefficient.

To make a good comparison of the impact of being financially flexible before the crisis on firm performance during the crisis I will also test how firms that were financially flexible before the crisis performed in those years before the crisis. Table 6 shows that firms with a higher than

22 average cash-ratio before the crisis performed better than firms with a lower than average cash-ratio during the crisis, but maybe they were performing even better before the crisis. Table 7 shows the results of the random-effects regression where I have looked at how financial flexibility has affected firm performance before the crisis.

Table 7: Random-Effects Regression of Financial Flexibility before 2008 on return on assets before 2008

Observations: 292

Variable Coefficient Robust Standard Error

Cash-High 0.0186 0.0167

Debt-Low -0.007 0.0165

Age 0.000 0.0000

Assets 2.42e-07 ** 8.51e-08 *=sig. at 10% level, **=sig.at 5% level, ***=sig. at 1% level

If we compare the results from tables 6 and 7 we can see that having a higher than average cash-ratio before 2008 already had a positive influence on ROA before 2008. Firms who had a higher than average cash ratio before 2008 had a higher ROA of on average1.86% before 2008 compared to firms with a lower than average cash ratio. Also, firms that had a lower than average debt-ratio before 2008 had on average 0.7% lower ROA than firms with a higher than average debt-ratio. The coefficient of cash-ratio has slightly increased after 2008 from 0.0186 to 0.029. The coefficient of debt-ratio has decreased from -0.007 to -0.0096. This means that firms would be better off during the crisis by having a higher than average cash-ratio before 2008 as well as a higher than average debt-ratio before 2008. It has to be noted tough that all estimates did not show a significant effect. To be able to see if the differences in the coefficients are significant I have ran another regression with an interaction term. I have therefore created another dummy variable called financial crisis, which measures 1 for the years since the crisis. With this dummy variable financial crisis I have created 2 interaction terms with the cash- and debt-ratios that represented firms who had a higher than average cash-ratio and a lower than average debt-ratio before the crisis. With these interaction terms it

23 should immediately become clear how firms that were financially flexible before the crisis have benefitted from this during the crisis. The results of the random effects regression can be found in table 8.

Table 8 shows how being financially flexible before the crisis has affected the change in ROA during the crisis. The variable cash-ratio is an interaction term which measures 1 if a firm had a higher than average cash-ratio before 2008 and looks at the years after 2008. The variable debt-ratio is also an interaction term and this term measures 1 if a firm had a lower than average debt-ratio and it also looks at the years after 2008. I have used robust standard errors to control for heteroskedasticity.

Table 8: Random-Effects Regression of FF on change in ROA with 2008 as start of the crisis

Observations:730

Variable Coefficient Robust Standard Error

Cash-ratio -0.0027 0.0169

Debt-ratio -0.0308 ** 0.0125

Age 5.81e-06 0.0001

Assets 1.41e-07 *** 3.74e-08 *=sig. at 10% level, **=sig.at 5% level, ***=sig. at 1% level

The results from table 8 show that for both measures of financial flexibility, firms are worse off. Firms that had a higher than average cash-ratio before the crisis did not benefit from this during the crisis. In fact, if a firm had a higher than average cash-ratio before 2008 it would have a lower return on assets of 0.27% during the crisis. This means that the difference in return on assets between cash-rich and cash-poor firms decreased by 0.27% during the crisis (table 7 had shown that cash-rich firms performed better than cash-poor firms before 2008) . Firms that had a lower than average debt-ratio before the crisis had a lower return on assets of 3.08% during the crisis. The latter means that the difference in return on assets between low-debt and high-debt firms increased by 3.08% (low-high-debt firms were performing worse before 2008 according to table 7). While the coefficient of cash-ratio did not have a significant effect on return on assets,

24 the coefficient of debt-ratio did have a significant effect. The latter means that the decrease in return on assets for low-debt firms was a significant change.

4.4 Robustness check

As a robustness check I will use a different year as starting point of the financial crisis. Mayer et al. (2011) and Yerkas et al. (2011) among others have stated the financial crisis started in 2007-2008. I used the year 2008 as a benchmark because the results from table 3 showed that in this year the average profit of the firms in my dataset declined the greatest after years of growth. However, it might be interesting to see what the effects of being financially flexible before 2007 has been on ROA before and after 2007 as a robustness check.

Table 9: Random-effects regressions of FF before 2007 on ROA before and after 2007 Variable Before 2007 Obs:219 After 2007 Obs:438

Cash-High Coef. 0.0044 0.0223 Std.Err. 0.0179 0.0194 Leverage-Low Coef. 0.0000 -0.0133 Std.Err. 0.0179 0.0189 Age Coef. -0.0001 0.0001 Std.Err. 0.0001 0.0001

Assets Coef. 2.17e-07 ** 1.67e-07 ***

Std.Err. 7.19e08 4.85e-08

*=sig. at 10% level, **=sig.at 5% level, ***=sig. at 1% level

Table 9 shows two regressions; 1 in which I have regressed being financially flexible before 2007 on ROA before 2007 and the other being financially flexible before 2007 on ROA after 2007. I have once again used robust standard errors to control for heteroskedasticity. The table shows that again having a higher than average cash-ratio before the crisis (in this case the start of the crisis was set at 2007) had a positive effect on a firm’s ROA both before as well as after 2007. Before 2007, firms that had a higher than average cash-ratio had on average a higher

25 ROA of 0.44% than firms with a lower than average cash-ratio. This effect became larger after 2007, the coefficient rose to 0.0223. The latter means that a higher than average cash-ratio resulted on average in a higher ROA of 2.23% compared to firms which had a lower than average cash-ratio before 2007. It has to be noted tough that both coefficients do not have a significant impact. Before 2007, having a lower than average debt-ratio in those years did not affect a firm’s ROA in the same years, but it did have a negative effect on a firm’s ROA after 2007. If a firm had a lower than average debt-ratio before 2007 it would have on average a lower ROA after 2007 of 1.3% in comparison with firms that had a higher than average debt-ratio before 2007. This is roughly the same trend as I observed when I took 2008 as the starting year of the crisis. To be better able to compare the coefficients I have again used a model with an interaction term. The results of this regression with the use of the interaction term can be found in table 9.

Table 10 shows the same trend as table 8 does. As we can see from the table firms are actually worse off during the crisis from being financially flexible before the crisis. The difference in return on assets before and after 2007 for cash-rich firms decreased by 1.37% compared to cash-poor firms. This difference was even greater for low-debt firms, for them the difference in return on assets decreased by 3.76% compared to high-debt firms.

Table 10: Random-Effects Regression of FF on change in ROA with 2007 as the start of the crisis

Observations: 730

Variable Coefficient Robust Standard Error

Cash-ratio -0.0137 0.0181

Debt-ratio -0.0376 *** 0.0139

Age 0.0000 0.0000

Assets 1.46e-07 *** 4.03e-08 *=sig. at 10% level, **=sig.at 5% level, ***=sig. at 1% level

It’s interesting to see that table 9 shows that the coefficients of cash-ratio increased before and after 2007, while table 10 shows that having a higher than average cash-ratio had a negative

26 effect on the change in return on assets during a crisis. Because the coefficients of for cash-ratio in both table 9 and table 10 are not significant I have to be reluctant with jumping to conclusions about this.

4.5 Firm size

After having run 6 different regressions it has become clear that financial flexibility has not had a positive effect on how firms perform during a crisis. Table 8 has shown that when firms had a higher than average cash-ratio before the crisis they would have a decrease in the difference in return on assets during the crisis of 0.27% compared to firms that had a lower than average cash ratio, although this coefficient did not have a significant impact. The coefficient of debt-ratio did have a significant. The regressions from table 8 and 10 showed that firms who had a lower than average debt-ratio before the crisis saw a decrease in their return on assets during the crisis compared to high-debt firms.

Total assets seemed to be a significant variable in almost all regressions. I will therefore make a distinction between smaller and bigger firms, and see how financial flexibility has affected ROA for them. I will test again how the pre-crisis ratios have affected return on assets, but this time I will make a distinction between firms with more than 1€ billion in total assets at any time during 2004-2013 and firms with lower than 1€ billion in total assets. For these tests I will use the random effects model again in which I use the interaction terms to be able to see how the difference in return on assets has changed for financially flexible and financially inflexible firmsI will test again how the pre-crisis ratios have affected return on assets, but this time I will make a distinction between firms with more than 1€ billion in total assets at any time during 2004-2013 and firms with lower than 1€ billion in total assets. For these tests I will use the random effects model again. The results can be found in table 11.

Table 11: Random effects regressions of FF before and after 2008 on ROA for small and big firms

Varia ble

Small Firms before 2008(obs:180)

Small Firms after 2008(Obs:202)

Big Firms before 2008(Obs:112)

Big Firms after 2008(Obs:163)

27 Cash-High Coef. 0.0089 0.0152 0.0207 0.0160 Std.Err. 0.0270 0.0317 0.0186 0.0269 Debt-Low Coef. 0.0211 0.0447 0.0066 -0.0158 Std.Err. 0.0212 0.0311 0.0247 0.0266 Age Coef. -0.0001 0.0001 -0.0001 0.0001 Std.Err. 0.0001 0.0001 0.0001 0.0001 Asset s

Coef. 0.0001 *** 0.0001 1.54e-07 * 1.15e-07

Std.Err. 0.0001 0.0001 1.14e-07 5.91e-08

*=sig. at 10% level, **=sig. at 5% level, ***=sig. at 1% level

Table 11 contains the results of 4 random effects regressions I have ran. I have made a distinction between small firms (firms with a total assets of less than €1 billion) and big firms (firms with a total assets of more than €1 billion) and I have also made a distinction between the effect of being financially flexible on return on assets before and after 2008. The results from table 9 show that small firms could have benefitted during the crisis if they were financially flexible before the crisis. Small firms that had a higher than average cash-ratio before the crisis had on average a higher return on assets of 0.89% before 2008 in comparison with firms that had a lower than average cash-ratio. After 2008 cash-rich firms had a 1.5% higher return on assets than cash-poor firms. The coefficient lower than average debt-ratio has also increased: from 0.02 to 0.0447. This means that firms with a lower than average debt-ratio had on average a 2% higher ROA than firms with a higher than average debt-ratio before 2008, and this increased to 4.47% after 2008. However, both variables do not have a significant effect. The reverse phenomenon is visible for bigger firms. Comparing the coefficients of cash- and debt-ratio of big firms before and after 2008 we actually see a decline in both coefficients. This means that big firms were worse off during the crisis if they were financially flexible before the crisis. Table 9 furthers shows that in 2 regressions the variable assets had a significant effect, but this effect was only very small as can be seen from the very small coefficients.

28 I have again used a model with an interaction effect to look at the significance of the changes in ROA between cash-rich and cash-poor firms and between low-debt and high-debt firms. The results of this can be found in table 12.

Table 12: RE regression of FF on change in ROA before and after 2008 for small and big firms

Variable Small Firms Big Firms

Cash-ratio Coefficient 0.0057 -0.024 Robust Std.Err 0.0251 0.0197 Debt-ratio Coefficient -0.0302 ** -0.0117 Robust Std.Err 0.0139 0.0235 Age Coefficient 0.0000 -0.000 Robust Std.Err 0.0000 0.000

Assets Coefficient 0.0000 4.11e-08

Robust Std.ERr 0.0000 5.01e-08

*=sig. at 10% level, **=sig. at 5% level, ***=sig. at 1% level

Table 12 shows that small firms had benefitted during the crisis from having a higher than average cash-ratio before the crisis as is shown by the positive coefficient of 0.0057. This coefficient means that cash-rich firms had an increase of 0.57% in the difference in return on assets during the crisis compared to cash-poor firms. The table however also shows that small firms with a lower than average debt-ratio were worse off during the crisis as the negative coefficient of 0.0302 shows. This coefficient meant a significant decrease of 3.02% in return on assets. Big firms were worse off during the crisis if they had a higher than average cash-ratio before the crisis and if they had a lower than average debt-ratio before the crisis.

4.6 Summary of findings regarding the first hypothesis

In this section I have tested my first hypothesis, namely that financially flexible firms have performed better during the crisis than less financially flexible firms. The regressions in this section have shown that being financially flexible does not have a positive impact on firm

29 performance during the crisis. Table 8 showed how the return on assets has changed during the crisis from before the crisis for firms that were financially flexible before 2008. This table had shown that financially flexible firms actually performed worse after 2008 if they were flexible before 2008. As a robustness check I picked another starting year for the crisis, 2007. Table 10 shows how firms that were financially flexible before the crisis had performed after 2007 and it became clear that financially flexible still performed worse.

4.7 The impact of financial flexibility prior to the crisis on firm performance during the crisis per industry

Firms that operate in an industry where profits fluctuate a lot per year might benefit more from financial flexibility than firms that operate in more stable industries. As discussed earlier, companies constantly have to find a balance for the right amount of cash holdings and leverage. Having less cash and more leverage has tax benefits, but increases the costs of financial distress. When firms operate in an industry where profits fluctuate a lot every year than the costs of financial distress might be more difficult to predict and firms working in these industries might benefit from having financial flexibility. To test this hypothesis I have created 9 dummy variables, 1 for each industry that the firms in this dataset are categorized in. The 9 different industries are: consumer services, industrial goods & services, basic materials, technology, oil & gas, consumer goods, support services, real estate and construction. The different industries have been categorized by the NYSE Euronext Amsterdam’s website. In total there are 10 firms that operate in consumer services, 11 firms in industrial goods & services, 4 firms that operate in basic materials, 14 firms that operate in technology, 3 firms that operate in oil & gas, 10 firms that operate in consumer goods, 8 firms that operate in support services, 7 firms that operate in real estate and 6 firms that operate in the construction industry.

Table 13: Profit statistics per industry (€millions)

Industry #Firms Mean Std.Dev. Minimum Maximum Std.Dev/Mean

30 IndustrialGoods&Services 11 228 782 -1456 5383 3.430 BasicMaterials 4 413 1551 -2106 9361 3.755 Technology 14 66 259 -872 1494 3.924 Oil&Gas 3 10332 15300 -572 41504 1.481 Consumergoods 10 695 1310 -202 5554 1.885 SupportServices 8 35 125 -696 385 3.571 RealEstate 7 76 193 -410 802 2.539 Construction 6 47 103 -183 352 2.191

A summary of the profit, cash- and debt-ratio statistics per industry are listed in table 10 and table 11.

Table 14: Cash-ratio and debt-ratio statistics per industry

Industry Variable #Firms Mean Std.Dev Minimum Maximum

Consumer Services Cashratio 10 0.093 0.103 0.001 0.750 Debtratio 10 0.185 0.125 0 0.505 Industrial Goods&Services Cashratio 11 0.065 0.089 0.000 0.605 Debtratio 11 0.225 0.122 0.01 0.528 BasicMaterials Cashratio 4 0.078 0.113 0.032 0.458 Debtratio 4 0.190 0.113 0.032 0.458 Technology Cashratio 14 0.149 0.140 0.000 0.760 Debtratio 14 0.188 0.220 0 0.92 Oil&Gas Cashratio 3 0.048 0.022 0.022 0.118 Debtratio 3 0.273 0.069 0.167 0.414 Consumer Goods Cashratio 10 0.061 0.060 0.000 0.322 Debtratio 10 0.236 0.177 0.001 0.634 Support Services Cashratio 8 0.075 0.076 0 0.311 Debtratio 8 0.159 0.103 0 0.390 RealEstate Cashratio 7 0.059 0.159 0 1 Debtratio 7 0.361 0.128 0 0.562

31

Construction Cashratio 6 0.118 0.079 0.025 0.470

Debtratio 6 0.180 0.103 0.013 0.341

Table 10 shows the profit statistics per industry. From this it is clear which industries have been the most profitable in the last 10 years. It should come as no surprise that the oil & gas industry is by far the most profitable industry with an average profit per year of a little over €10 billion. The least profitable industries have been the support & services industry and the real estate industry. In the table I have also included the standard deviation, since I want to test if financial flexibility is more important in industries where profits fluctuate a lot. The oil & gas has by far the highest standard deviation but this is not the variable of my interest. It makes sense that the industry that has the largest average profit also has the largest standard deviation, at least in absolute terms. I have therefore also included a column where I have weighted the standard deviation by the average profit. From this column it becomes clear that the oil & gas industry is actually the most stable industry; the standard deviation here is only 1.48 times the average profit. The most unstable industries are: basic materials, technology and support services. I will keep this in mind when testing my second hypothesis. Table 11 shows the cash- and debt-ratios statistics per industry. With respect to the cash-ratio, table 11 shows that the 2 industries that are the most stable, oil & gas and consumer goods (a value of respectively 1.481 and 1.885 for standard deviation/mean from table 10), are also the two industries that have the lowest average cash ratio. The table further shows that the most unstable industry, Technology (a value of 3.924 for standard deviation/mean), has the highest average cash ratio. Table 10 also shows the debt-ratio statistics per industry. The three industries that had the highest average debt-ratio are: real estate, consumer goods and oil & gas. This again shows that when profits fluctuate less, firms operating in these industries can afford themselves to be not very financially flexible.

Firms that operate in stable industries can thus afford themselves to be financially inflexible whereas firms that operate in unstable industries tend to be more flexible. But how has this flexibility or lack of flexibility worked out for the firms in this dataset? To test this I have

32 regressed how financial flexibility prior to the crisis has affected profits during the crisis per industry. This should show whether firms have benefited or not from adapting industry commons, such as being flexible in unstable industries or not being flexible in stable industries. Table 15: RE-regression with interaction terms to check change in ROA before and during crisis

Industry Cash-ratio Debt-ratio

Consumer Services Coefficient 0.003 0.003

Robust. Std.Err. 0.081 0.083

Industrial Goods & Services

Coefficient 0.039 -0.059 ***

Robust. Std.Err. 0.025 0.0148

Basic Materials Coefficient -0.167 *** -0.143 ***

Robust. Std.Err. 0.063 0.0475

Technology Coefficient 0.0285 -0.0628 ***

Robust. Std.Err. 0.0411 0.0374

Oil & Gas Coefficient - -0.039

Robust. Std.Err. - 0.054

Consumer Goods

Coefficient 0.0061 -0.0177

Robust. Std.Err. 0.0244 0.016

Support Services Coefficient 0.0029 -0.113

Robust. Std.Err. 0.0333 0.0261

Real Estate Coefficient -0.106 *** 0.0819 **

Robust. Std.Err. 0.0279 0.0400

Construction Coefficient -0.0519 *** 0.044

Robust. Std.Err. 0.0201 ** 0.0227

*=sig. at 10% level, **=sig. at 5% level, ***=sig. at 1% level

Table 15 shows how the returns on assets have changed from before 2008 to after 2008 per industry. I have marked the changes that were positive green and the negative changes red to

33 make it easier to observe. I have also controlled for age and firm size, but left these variables out of the table to make it easier to interpret. I have used standard robust errors to control for heteroskedasticity. The table shows that firms in only 1 industry have benefitted during the crisis from being financially flexible before the crisis. That industry is the consumer services industry. The coefficient of 0.003 for cash-ratio means that firms who had a higher than average cash-ratio before the crisis had an increase of a 0.3% difference in their return on assets compared to firms with a lower than average cash-ratio. This means that the difference between cash-rich and cash-poor firms became bigger in terms of return on assets by 0.3%. Firms operating in the consumer industry also benefited during the crisis from having a lower than average debt-ratio before the crisis. Firms that operated in the industrial goods & services industry, technology, consumer goods and support services could have benefitted during the crisis from having a higher than average cash-ratio before the crisis, but they were worse off during the crisis if they had a lower than average debt-ratio. Firms who worked in the basic materials industry were worse off during the crisis if they any measure of financial flexibility before the crisis. Finally, first that operated in real estate or in construction were worse off during the crisis if they had a higher than average cash-ratio before the crisis, but they were better off during the crisis if they had a lower than average debt-ratio before crisis.

The hypothesis that I wanted to test in this section was if firms who operate in unstable industries could benefit during the crisis from being financially flexible prior to the crisis. The most unstable industries in terms of variation in profits are: industrial goods & services, support services, basic materials and technology (table 13). Table 14 shows that firms operating in unstable industries are indeed more financially flexible than firms operating in stable industries. However, when looking at table 15 it becomes clear that firms which operate in the most unstable industries were all were off during the crisis if they had a lower than average debt-ratio before the crisis. Table 15 does however show that with the exception of the basic materials industry, firms from the most unstable industries could have benefitted during the crisis from having a higher than average cash-ratio before the crisis.

34 It is interesting to see that firms which operated in the most stable industries (oil & gas, consumer goods and construction) could partly benefit during the crisis from being financially flexible before the crisis. Firms from the consumer goods could have benefitted during the crisis if they had a higher than average cash-ratio before the crisis, whereas firms operating in the construction business could have benefitted during the crisis from having a lower than average debt-ratio before the crisis.

The results in this section show that I can’t support my hypothesis that firms in unstable industries would benefit during the crisis from being financially flexible before the crisis. However it has to be noted that for most industries I only had a small sample size in my dataset. For example the oil & gas industry only contained 3 firms and the basic materials industry only contained 4 firms. Also, I used the cash and debt ratios as variables for financial flexibility but this may be too narrow. There are of course also other indicators of flexibility that I have not included in my model and that may affect firm performance during a crisis. I could for example also have used assets as an indicator if flexibility and then especially intangible assets as they can be sold in times of financial distress and this have the same advantage as cash reserves.

5. Conclusion & Discussion 5.1 Summary of findings

In this thesis I have aimed to show how financial flexibility affects firm performance during a crisis. I have concentrated on firms from Holland and looked at the years 2004-2013. The analysis has shown that financial flexibility has only mildly affected firm performance during the recent crisis. First, I have looked at the effect of being flexible before the crisis on profits in terms of assets during the crisis. Tables 6 and 7 show how being financially flexible before 2008 has affected return on assets after 2008 in comparison with firms who were financially inflexible. Firms that had a higher than average cash ratio before 2008 had a higher ROA of 2.4% before compared to firms with a lower than average cash ratio. After 2008, cash-rich firms still had a higher ROA than cash-poor firms but this extra ROA had decreased to 1.8%, meaning their advantage had decreased during the crisis. Firms with a lower than average debt ratio before 2008 were performing worse before 2008 than firms that had a higher than average

35 debt ratio before 2008 and after 2008 this difference became even bigger. These changes have been tested with an interaction term in table 8 and this table also showed that firm who were financially flexible before 2008 were worse off during 2008. As a robustness check I picked another starting point of the crisis, 2007. The results of this are shown in table 9 and this table shows that firms who had a higher than average cash-ratio before 2007 performed better before 2007 than firms with a lower than average cash ratio and after 2007 they performed even better (an increase of 0.44% extra ROA to 2.23% ROA). However, this change in return of assets did not have a significant effect and when I used an interaction term in table 10, it showed that firms who had a higher than average cash-ratio before the crisis actually had a decrease in the extra return on assets they had during the crisis. In terms of leverage firms were worse off being financially flexible before 2007. Before 2007 low debt and high debt firms were performing the same in terms of ROA but after 2007 the low-debt firms performed worse. Table 10 showed the same effect for low-debt firms, namely that low-debt firms were worse off during the crisis than high-debt firms.

I have also compared the effects of being financially flexible before the crisis on ROA during the crisis for small and big firms and found that small firms benefitted during the crisis having a higher than average cash-ratio before the crisis, but were worse off during the crisis if they had a lower than average debt-ratio. Big firms were worse off during the crisis if they were financially flexible before the crisis. However, all coefficients for big firms did not have a significant effect and for small firms only the coefficient of the debt-ratio had a significant effect.

Finally, I have tested whether flexibility mattered more in unstable industries. I was unable to show the importance of being financially flexible in unstable industries. I did find that firms that operated in unstable industries are in fact financially flexible as table 10 shows, but this did not positively affect their firm performance during the crisis. I did find that from the 4 most unstable industries only firms operating in the basic materials industry didn’t benefit during the crisis from having a higher than average cash-ratio before the crisis. On the other hand, firms

36 from all 4 unstable industries were worse off during the crisis if they had a lower than average debt-ratio before the crisis.

5.2 Limitations

The global character of the financial crisis which is subject of my research might have caused the lack of significant impact that financial flexibility had on firm performance. Arslan-Ayaydin et al. (2014) found that the impact of financial flexibility on firm performance was much less significant in the recent financial crisis than in the other crisis they looked at, the East-Asian crisis of 1997-1998. Perhaps because the recent crisis had hit firms worldwide it was more difficult for firms to find profitable investment opportunities and attract new debt. Maybe when a crisis is more regional, firms can invest or attract new debt from countries where the economy is still flourishing but when a crisis hits firms worldwide this will become more difficult. This could explain why Arslan-Ayaydin et al. (2014) found that the effect of flexibility was less significant in the recent crisis and why I found little evidence of this effect in Holland. If it’s true that the global character of the crisis has caused that financial flexibility has little effect on firm performance than the same results should be found in other countries. This is something future research might prove.

Financial flexibility comes at the cost of losing tax benefits and not spending in investments. However, it decreases the possibility of financial distress. In my analysis I have only looked at firms that are still active and thus have not gone bankrupt during the crisis. It might be that firms who went bankrupt during the crisis and have thus experienced financial distress might have benefited from being financially flexible. This may cause a selection bias because omitting these firms may bias the analysis towards the benefits of lacking financial flexibility. Future research might include firms that went bankrupt during the crisis.

In my research I have used the cash- and debt-ratios of firms as a measure of financial flexibility. I have used this measure of financial flexibility, because these objective measures are used most in the existing literature (Almeida and Campello 2007, Marchica and Mura 2009, Denis and Sibilkov 2010, Ayaydin et al. 2014). However, I could have also used other measures of financial flexibility that could affect firm performance during a crisis. For example, a firm’s