Does a decrease in the supply of credit lead to a reduction in corporate growth? : evidence from the 2007-2009 financial crisis

(1)

Does a decrease in the supply of credit lead to a

reduction in corporate growth?

Evidence from the 2007-2009 financial crisis.

Master Thesis Finance

Name:

Joris Koning

University of Amsterdam

Student Number:

10250905

MSc Business Economics, finance track

Supervisor:

R. van Lamoen

(2)

1

Statement of Originality

This document is written by student Joris Koning who declares to take full responsibility for the contents of this document.

I declare that the text and the work presented in this document is original and that no sources other than those mentioned in the text and its references have been used creating it.

The faculty of Economics and Business is responsible solely for the supervision of completion of the work, not for the contents.

(3)

2

Abstract

Financial markets have been heavily disturbed during the 2007-2009 financial crisis which led to a decrease in the supply of credit. This paper studies the short- and long-run effects of the decrease in credit supply during the crisis on corporate investments and firm growth. To identify the effect of the decrease in credit supply, the decline in growth after the onset of the financial crisis of firms with low cash reserves is compared to the decline in growth of firms with high cash reserves. The idea behind

this method is that a firm with a lot of cash on hands during the crisis is less dependent on external funding than a firm with small cash reserves. The results show that asset growth declines more for firms with low cash reserves during the crisis than for other firms. Credit supply only has an effect on

firm growth in the medium-run and the long-run which implies that there is a delay in the response of a firm’s growth level to a negative shock in the supply of credit.

(5)

4

1. Introduction

The 2007-2009 financial crisis, which was caused by consumer defaults on mortgages, had a big impact on the financial sector. However, the increased lending standards and loan spreads also affected non-financial firms that depend on funding by financial institutions (Campello et al., 2010). In a study on bank lending to U.S. firms during the 2007-2009 financial crisis, Ivashina and Scharfstein (2010) found a substantial decline in new lending which provides evidence of a negative shock to credit supply. The scope of the decrease in credit supply during the recent financial crisis emphasizes how important it is to have a good understanding on how the supply of credit affects the real

economy.

The real economy is driven by the performance of firms which are dependent on funding when they want to invest. Firms can fund their investments with retained earnings, with equity, by issuing debt or with bank loans. In this paper the focus is on the implications of a sharp decrease in the supply of debt. When the supply of debt decreases, one source of funding becomes more expensive or even unavailable for firms. When this is the case, a firm may be unable to do an investment which would have created firm growth and thereby economic growth. In this way, the supply of credit can have a negative effect on the real economy through the effect of the supply of credit on firm growth.

The 2007-2009 financial crisis is characterized by an exceptional decrease in the availability of debt. One of the first empirical researches that uses the 2007-2009 financial crisis to examine the relationship between credit availability and investments is Almeida et al. (2009). They used the variability in maturing long-term debt during the first year of the financial crisis to measure how much credit a firm needs during the contraction in credit supply. They find a negative relationship between the credit that a firm needs during the first year of the crisis and direct investments. Later on, Kahle and Stulz (2013) used the same proxy of the credit need during the financial crisis as Almeida et al. (2009), but they separate the crisis period in different stages. Surprisingly, they only find a significant negative effect in the last year of the financial crisis. They argue that a decrease in capital expenditures must be driven by a common shock to the demand of goods after the fall of Lehman brothers. Furthermore, they argue that the significant relationship found in the last year of the crisis doesn’t provide evidence of a relationship between credit supply and investments because they find that debt issuance between credit constraint and other firms is similar in the last year of the financial crisis.

The literature that examines the relationship between credit supply and investments at the firm level is ambiguous (see Almeida et al., 2009; Campello et al., 2010; Kahle and Stulz, 2013). It is remarkable that the effect of a decrease in credit supply during the crisis on investments depends on

(6)

5 the time interval that is used. So far, the literature only examines the static effect of the credit contraction during the financial crisis on investments using short time periods of data. Furthermore, the existing literature hardly uses data after the financial crisis, with the exception of Kahle and Stulz (2013) who use data until one year after the financial crisis. This paper uses data from 2002 until the end of 2015 to examine what the short- and long-run effects are of a negative shock to credit supply on investments and firm growth. A dynamic panel model is introduced to disentangle the long-run effect of credit supply on firm growth from the short-run effect. Furthermore, a response function is used to examine the dynamics of the effect of credit supply on firm growth in more detail. Thereby, the contribution of this paper is that the dynamic relationship between credit supply and real

economic performance is examined at the firm level. Understanding more about the dynamics of the relationship between credit supply and corporate performance contributes to the literature about the relationship between the financial and the real side of the economy. (e.g. Fazarri et al. (1988), Rajan and Zingales (1998) Kroszner et al. (2007)).

The effect of credit supply on firms’ investments, their growth and therefore on the

economic growth is of interest for policy makers such as politicians and central bankers. They should take the effects of credit market conditions on the real economy into account when a new financial crisis arises and a costly policy measure is considered to avoid a sharp decrease in the supply of credit. Therefore, they should be aware of the difference in short-run and long-run effects of their decisions. Furthermore, the examined relationship in this paper is of interest for companies that need to make funding and investment decisions. More knowledge about this relationship creates awareness of the dependency of firms on the credit market and is useful when a firm sets its funding policy. Knowledge about the short- and long-run effects of a decrease in credit supply supports the policy makers when they want to make stable and future proof funding decisions.

The main research question in this paper is:

What is the effect of credit supply on firm growth in the short-run and the long-run?

Because the effect of credit supply on firm growth works through the effect on corporate investments, the following sub question is answered:

What is the effect of credit supply on investments in the short-run and the long-run?

The rest of this paper is organized as follows. In section 2, the related literature is discussed and hypotheses are formed. Section 3 describes the methodology that is used to answer the research questions. The data is described in section 4. Section 5 presents the empirical results. Section 6 presents additional tests for robustness. Section 7 concludes.

(7)

6

2. Literature review

This section will start with an overview of how financial markets are theoretically related to the real economy. The channels through which the supply of credit potentially affects the

opportunity for firms to grow will be described. Thereafter, an overview will be given on the existing empirical literature that examines the relationship between financial markets and economic growth. Third, the implications of the financial crisis on the supply of credit will be described. Fourth, an overview of the literature will be given that examines the relationship between the financial contraction during the financial crisis an corporate investments. Lastly, a conclusion is drawn from the literature and predictions for the empirical research in this paper are made based on the existing literature.

2.1 Finance and Economic growth: the Theory

According to the theory of Modigliani and Miller (1958), the financial situation of a firm does not affect real investment decisions at all in a world of complete and perfect capital markets. If the assumption of perfect capital markets holds, internal funds are a perfect substitute for external funds and the financial conditions of a firm do not affect its investment decisions. If the theory of

Modigliani and Miller (1958) would hold, the supply of credit does not have any influence on the growth of non-financial firms. However, if there are frictions in the financial markets such as taxes, transactions costs and asymmetric information , the financial conditions of a firm can potentially influence its investment behavior and thereby its growth level.

This paper examines the effect of credit supply on the growth of non-financial firms. The supply of credit does not generate corporate growth by itself. Corporate growth and economic growth can only be obtained from investment opportunities with a positive net present value. However, there are different channels through which financial markets can have an influence on corporate growth and economic growth. In their 1995 paper, Bernanke and Gertler discuss the credit channel through which a drop in asset prices influences the ability of the corporate sector to raise debt. The credit channel consists of the balance sheet channel and the lending channel. The balance sheet channel states that a decrease in asset prices erodes the capital position of a firm which worsens its collateral position and thereby its ability to borrow1. Figure 1 below illustrates the mechanics of the balance sheet channel. The channel through which the supply of credit potentially influences the performance of the corporate sector is the lending channel. When a shock occurs that negatively affects the asset value of lenders such as banks or other debt holders, they become less willing to lend money. This financial shock to debt holders can result in a situation in which they are curtailing credit to borrowers that face an investment opportunity with a positive net present value.

(8)

7 In this case, a profitable investment opportunity that generates growth is not exploited because the supply of credit is not sufficient or the cost of capital is too high. Figure 2 illustrates the mechanics of the lending channel.

Figure 1 – The Balance Sheet Channel

Figure 2 – The Lending Channel

2.2 Financial market conditions and Economic growth: Empirical evidence

This paper is related to the literature about the relationship between financial markets and the real economy. The literature about the relationship between financial market conditions and economic growth on the macroeconomic level is quite extensive already. Levine (2005) gives an extended overview of theoretical and empirical work on this subject.

One of the first papers that examines the effect of financial factors on investments at the firm-level is the 1988 paper of Fazzari, Hubbard and Petersen. They use observed retention practices of firms to identify firms that face high costs of external finance. They find that the cost of debt has more impact on the investment behavior of firms that retain nearly all of their income than on firms that use a large fraction of their income to payout dividends. The main conclusion from Fazzari et al. (1988) is that liquidity variables matter for investment spending at the firm level. Supportive

evidence on this conclusion is found by Gertler and Hubbard (1988), Whited (1992) and Hoshi, Kashyap, and Scharfstein (1991). Furthermore, the finding of Fazarri et al. (1988) indicates that a higher cost of debt has a negative effect on investments for firms that face a financial constraint. The relationship between financial conditions and economic growth has extensively been researched at the macroeconomic level. An influential paper in this field is from Rajan and Zingales

Asset

Prices

Collateral

value

Cost of

Debt

Investment

decisions

Growth

Firm

Asset value

of

Debtholders

Credit

supply

Availability

/Cost

of Debt

Investment

(9)

8 (1998) who examine the relationship between the state of the capital market in a country and the economic output by industry. They make use of cross-country differences and compare the value added by industries that are highly dependent on bank loans to the value added by industries that are not so dependent on bank loans. Their results show a positive relationship between the level of development of the financial sector in a country and the industrial growth rate of its economy. In addition to previous papers that only find a positive correlation between financial development and economic growth2, Rajan and Zingales (1998) show that the positive relationship they found is likely to be causal, which implies that the state of a financial market in a country has a positive effect on the economic growth in that particular country.

More recently, several papers build on the method of Rajan and Zingales (1998) to examine the relationship between the financial system in a country and industry growth during recessions and banking crises. Braun and Larrain (2005) find that recessions have a more negative impact on growth in industries that are heavily dependent on external financing than on industries that are not so dependent on external finance. A similar paper of Dell'Ariccia et al. (2008) confirms the finding of Braun and Larrain (2005). However, they specifically look at the effect of banking crises on the difference in growth between sectors that rely heavily on external sources of finance and other sectors. They find that when a negative shock occurs to a country, bank distress has an additional negative effect on growth in sectors that are heavily dependent on external sources of finance.

Kroszner et al. (2007) compared the effect of banking crises on growth for external finance dependent industries with other industries. Consistent with the findings of Rajan and Zingales (1998), they find that more external finance dependent sectors grow faster in countries with more

developed financial systems. However, in banking crisis periods, Kroszner et al. (2007) find that the growth in sectors that rely heavily on external financial resources is reduced more heavily in countries with more developed or “deep” financial markets than in other countries. This finding is consistent with the existence of a lending channel. Sectors in countries with well-developed financial markets that are dependent on external finance rely heavily on the lending channel to finance investments during normal periods. A crisis that worsens the ability of banks to give out loans

decreases the effectiveness of the lending channel which in turn decreases the growth of sectors that rely heavily on the lending channel. Firms in countries with a less developed financial system rely less on the credit channel to make their investments. Therefore, a crisis that impairs the working of the lending channel has a less sizable effect on these firms’ ability to invest than in countries with a more developed financial system.

(10)

9 Besides Fazarri et al. (1988), all literature described so far uses a method similar to Rajan and Zingales (1998). They use the variability in financial factors between countries and the variability in credit dependency between industries to extract the effect of the credit channel on economic growth from the effects of a change in the demand for goods on economic growth. A new opportunity to investigate the relationship between credit market conditions and economic consequences was created by the 2007-2009 financial crisis. Several papers make use of the decrease in the supply of credit during the 2007-2009 financial crisis to find new evidence on the influence of the credit market on funding, investments and corporate growth.

2.3 The supply of credit during the 2007-2009 financial crisis.

This paper makes use of the 2007-2009 financial crisis to explore the consequences of a decrease in the supply of credit on firm growth. The 2007-2009 financial crisis was in the first place caused by defaults on consumer mortgages in the U.S. and had a major negative effect on the worldwide economy.3 More specifically, the burst of the housing bubble forced banks to write down several hundred billions of dollars in bad loans caused by mortgage defaults (Brunnermeier, 2008). As a result, overall bank performance during the crisis was the worst since the great depression (Beltratti and Stulz, 2012). Ivashina and Scharfstein (2010) empirically explored the influence of the 2007-2009 financial crisis on the supply of credit to the corporate sector. They find that lending to the corporate sector declined substantially during the crisis. However, a decline in lending to the corporate sector does not provide evidence of a decrease in credit supply. The demand for credit by the corporate sector most likely declined during the financial crisis because firms scale back their expansion plans. The challenge to disentangle the causal relationship from credit supply to external funding by firms (the lending channel) from the relationship from corporate demand for credit to provided loans (the demand channel) is often noticed in the literature and is called the “identification problem”4.

To find out whether the decline in lending is solely due to a decline in investment plans by the corporate sector or also by a decrease in the credit supply by financial institutions, Ivashina and Scharfstein (2010) compared the lending of banks with more uncertainty about their future liquidity with other banks. They find that banks that face more uncertainty about future liquidity decrease their lending more than other banks. Since it is unlikely that the difference in lending arises from a difference in the demand for loans of the costumer of these two groups of banks, this finding is a strong indication of a decrease in credit supply during the financial crisis. Puri et al. (2011) use a unique dataset from Germany to examine the effects of the US financial crisis on global lending to

3

For a more detailed overview of the causes and consequences of the 2007-2009 financial crisis, See Gorton(2008)

(11)

10 retail customers. By using the same methodology and data from Germany, they find evidence of a decrease in credit supply during the 2007-2009 financial crisis.

The existence of a decrease in credit supply during the 2007-2009 financial crisis is crucial for the validity of the research in this paper. Taking the empirical findings of Ivashina and Scharfstein (2010) and Puri et al. (2011) into account, it is unambiguous that the corporate sector faced a decrease in the supply of credit during the 2007-2009 financial crisis. How the identification problem will be dealt with in this paper is explained in the methodology section.

2.4 The effect of the credit contraction during the crisis on firm growth.

The first paper that uses the financial crisis to explore the effects of credit supply on firm performance is from Almeida et al. (2009). They use ex-ante variability in the maturity of a firm’s long-term debt as an instrument of how much a firm is affected by the financial contraction at the onset of the financial crisis. The authors make the assumption that the ex-ante measured proportion of long-term debt that matures during the crisis can be seen as random so that it is not correlated with investment opportunities. Using an experiment like design, they compare similar firms that have a high fraction of long-term debt maturing during the crisis with firms that have no or only a low fraction of long-term debt maturing during the crisis. They find that firms with large fractions of their long-term debt maturing during the first year of the financial crisis have to reduce their investments by 2.5 percent more than similar firms for which only a small fraction of their long-term debt matures during the first year of the financial crisis. This finding provides evidence of a positive relationship between credit supply and investments. A shortcoming of Almeida et al. (2009) is their relatively small treatment group of 86 firms on which they base their core results.

Another paper that uses the financial crisis to explore the relationship between credit supply and firm performance is the paper of Duchin et al. (2010). To solve the identification problem, Duchin et al. (2010) use a difference-in-differences approach where they use a function of a firm’s internal financial resources one year prior to the start of the financial crisis as a proxy of how much the firm depends on external financing during the crisis. The authors focus their research on the first year of the financial crisis (1 July 2007- 30 June 2008). While controlling for firm and time fixed effects, they find that corporate investments declines by 6.4 percent after the onset of the financial crisis. Moreover, they find that corporate investments decline more for firms with a low ex-ante cash position than for firms with a high ex-ante cash position which is consistent with a negative effect of the decrease in credit supply on corporate investments. When Duchin et al. (2010) extend their sample with the second year of the financial crisis, they find that corporate investments continue to decline. However, Duchin et al. (2010) find that the decline in investments during the second year of the financial crisis is mainly due to a change in investment opportunities which is driven by a demand

(12)

11 shock.

The papers by Almeida et al. (2009) and Duchin et al. (2010) are both closely related to this paper because they use ex-ante firm characteristics to instrument the dependence on credit supply during the financial crisis. However, they only examine the direct effect of the financial contraction on corporate investments for which they only use a very short time period. This paper will use a similar method to these papers but uses firm growth as the dependent variable. Furthermore, this paper focuses more on the long-term effect of a decrease in credit supply on firm growth and aims to disentangle the long-run effect and the short-run effect

Contrary to Almeida et al. (2009) and Duchin et al. (2010), Kahle and Stulz (2013) don’t find evidence of a causal effect of credit supply on investments. They compare firms which they classify as bank dependent and firms that are not bank dependent during the 2007-2009 financial crisis with each other. The authors find that firms that rely on bank loans to fund their investments don’t reduce investments significantly more than firms that don’t rely on bank loans to fund their

investments. However, for the year starting in April 2009, Kahle and Stulz (2013) do find a relatively bigger decrease in investments for firms that are classified as dependent on external credit.

Furthermore, they find that credit dependent firms don’t issue less debt than other firms but they increase their cash position. Therefore, Kahle and Stulz (2013) conclude that the found relationship is most likely driven by a demand shock and not by the shock to the supply of credit.

The findings by Kahle and Stulz (2013) are relevant for this paper since they predict that there is no relation between the supply of credit and corporate investments. Therefore, there is no reason to believe that there is an effect of credit supply on firm growth. However, their finding that corporate investments are lower for firms that are dependent on external funds during the last year of the financial crisis creates doubts about their conclusion. There must be a reason why credit dependent firms have to increase their cash position during the last year of the financial crisis. The most obvious reason for this finding is that they have used all their cash in the first year of the financial crisis. If this is the case, the effect of credit supply on corporate investments might not be observed directly but is observed in the last year of the financial crisis. For this reason, this paper will examine the effect of the financial contraction on firm growth during the financial crisis both on the short-term and on the long-term.

In contrast to the papers that use ex-ante firm characteristics to solve the identification problem, Campello et al. (2010) use a different approach. They asked CFO’s whether their companies faced a credit constraints during the financial crisis. The authors have done a survey among 1050 CFO’s in the U.S., Europe and Asia and used the responses as a measure of how credit constraint a firm was during the crisis. They examined the effects of the estimated credit constraint measure on corporate spending plans. According to their analysis, credit constrained firms cut in tech spending,

(13)

12 employment and capital spending. They also find that credit constrained firms sell assets to fund new investments and that they bypass positive NPV investment opportunities due to their limited access to capital. Another interesting finding of Campello et al. (2010) is that CFO’s do use internal sources to fund investments when access to external capital markets is limited. The findings of Campello et al. (2010) give reason to expect a negative effect of a decrease in credit supply on firm growth. Furthermore, their finding that firms use internal resources to fund their investments when external credit is hard to obtain makes it interesting to examine the long-term effects. Since the internal resources are limited, the use of internal resources to fund investments could lead to a decrease in investments in a later period when the firm runs out of internal resources or when the firm wants to bring its cash position back to its desired level.

2.5 Conclusion and hypotheses

The literature on the implications of the financial crisis provides evidence of a negative shock to credit supply during the 2007-2009 financial crisis. The literature that examines the relationship between the state of the financial markets and economic growth on a cross-country base shows a positive relationship between the state of the financial market and economic growth. This paper examines the relationship between credit supply and (economic) growth at the firm level. Since it is likely that more developed financial markets result in a higher supply of credit and a lower cost of capital, the positive relationship that is found between the state of the financial market on economic growth is expected to hold for the relationship between the supply of credit and corporate growth as well. The literature also examined the relationship between the supply of credit and capital

expenditures with the 2007-2009 financial crisis as negative shock to credit supply. Almeida et al. (2009) and Duchin et al. (2010) provide evidence in favor of the credit supply shock theory while the results of Kahle and Stulz (2013) are not in line with the credit supply shock theory.

This paper will contribute to the existing literature since it is the first paper that examines the short-run and the long-run effects of a decrease in credit supply on corporate investments and firm growth using firm level data. Furthermore, this is the first paper that examines the effect of the negative shock to credit supply during the 2007-2009 financial crisis on firm growth. The literature on the macroeconomic level gives reason to expect a decrease in firm growth for firms that are highly influenced by the negative shock to credit supply. Research at the firm level is interesting because at the firm level, firm specific controls can be included in the model. Furthermore, research at the firm level gives the opportunity to control for unobserved firm specific factors. Compared to the

aggregated data that is used in research at the macroeconomic level, firm level data contains more detailed information.

(14)

13 The literature that examines the effect of the negative shock to credit supply on corporate investments at the firm level is ambiguous on the existence of the credit supply shock theory. Almeida et al. (2009), Duchin et al. (2010) and Campello et al. (2010) find that firms that are more credit constraint during the financial crisis are forced to decrease capital expenditures. Kahle and Stulz (2013) don’t find evidence for a decrease in capital expenditures for more credit constraint firms. However, because there are more papers which use different methods that find a positive relationship between credit supply and capital expenditures, and because a positive relationship between credit supply and corporate investments is in line with findings in the macroeconomic literature, the following hypothesis is formed.

First hypothesis: There is a positive relationship between credit supply and corporate investments.

A positive relationship is found in the literature between the development of financial markets and economic growth at the macroeconomic level. Furthermore, at the firm level, there is a reason to expect a positive relationship between credit supply and firm growth. Campello et al. (2010) find that firms need to sell assets and cut employment when they are credit constraint during the crisis. Based on these findings, the second hypothesis is formed.

Second hypothesis: A positive relationship exists between credit supply and firm growth .

Evidence on the dynamics of the effect of credit supply on corporate investments and firm growth is scarce. Kahle and Stulz (2013) found that the decrease in credit supply negatively affected corporate investments only in the last year of the financial crisis. Furthermore, firms can make themselves less dependent on external credit when the credit supply is low. Therefore, it is expected that the medium-short run effect is larger than the long-run effect of credit supply on investments and firm growth. Based on these expectations, the third hypothesis is formed.

Third Hypothesis: The effect of credit supply on investments and on firm growth is greater in the medium run/ short run than in the long run.

(15)

14

3. Methodology

As described above, the main contribution of this paper is to analyze the short- and long-run dynamics of the effect of credit supply on firm growth. To analyze this effect, a

difference-in-differences setup is used that compares different measures of firm growth before and after the negative shock to credit supply during the financial crisis as a function of a firm’s pre-crisis cash holdings. This section will start with an explanation of the methodology that is used to solve the identification problem5. Second, the basic regression specification that is used to analyze the effect of credit supply on corporate investments is given. Third, the basic regression specification that is used to examine the effect of credit supply on firm growth is given. Thereafter, a dynamic panel data model will be given that is used to disentangle the short-run effect from the long-run effect. Fourth, an alternative approach to disentangle the short-run from the long-run effect is given.

3.1 The credit supply shock and the identification problem

What makes it challenging to examine the effect of the credit supply shock on firm growth is the so called identification problem as described in section 2.3 above. The credit supply shock is not the only shock that affect firms during the financial crisis. Besides the negative shock to credit supply, firms are affected by a negative shock to demand and by the balance sheet multiplier effect. To extract the effect of the credit supply shock from the other channels that affect firm behavior after the onset of the financial crisis, a measure of how much a firm is affected by the credit supply shock is needed which is not related to one of the other channels through which the financial crisis

influences firm performance. Similar to Duchin et al. (2010) and Almeida et al. (2009), this paper uses a firm’s financial position prior to the crisis as a determinant of how much the firm is affected by the credit supply shock. Following Duchin et al. (2010), a firm’s cash holdings one year prior to the onset of the financial crisis is used as the variable to measure how much the firm is affected by the

negative shock to credit supply. The idea of using a firm’s cash position one year prior to the onset of the financial crisis is that a firm which has lots of cash on hand does not depend on external credit when making its investment decisions. In contrast to firms with high cash reserves, a firm that barely has cash on hand at the onset of the financial crisis is more dependent on external credit when making its investment decisions. However, a firm without cash on hands can also fund its

investments with internal funds. To make sure that a firm uses external credit as a source of funding, firms with a leverage lower than ten percent are excluded from the sample. Alternative thresholds for the minimal leverage level will be used as robustness checks. Excluding firms with a low leverage from the sample makes the pre-crisis cash holdings of a firm a stronger measure of the dependency of a firm on external credit. However, it does not eliminate the fact that firms can use alternative

(16)

15 sources of funding that makes them less dependent on external credit.

To make the pre-crisis cash holdings of a firm a valid measure of a firm’s dependence on external credit, it is of main importance that the cash holdings are not related to unobserved

variation in investment- and growth opportunities. Therefore, the cash position of a firm at one year before the onset of the crisis is used and not the cash position directly before the onset of the crisis. Duchin et al. (2010) show that it is reasonable to assume that the cash holdings of a firm one year before the crisis are not related to variation in investment opportunities after the onset of the financial crisis. Furthermore, a firm’s cash position at the third quarter of 2006 has to be a good prediction of its cash holdings in the third quarter of 2007 to be a good measure of a firm’s

dependency on external credit during the crisis. The initial cash position is defined as the amount of cash and marketable securities divided by total assets in the third quarter of 2006.

3.2 The effect of credit supply on investments

The first part of this study examines the effect of the negative shock to credit supply during the financial crisis on investments at the firm level. It is expected that a positive relationship will be found between a firm’s cash position just before the financial crisis started and its capital

expenditures thereafter. To model the effect of the negative shock to credit supply during the crisis on corporate investments, a similar regression to Duchin et al. (2010) is used. A static model will be used to examine whether credit supply has an effect on investments. Thereafter, a dynamic model as described in section 3.4 will be used to disentangle the short- and long-run effect of credit supply on investments. The regression specification that is used to examine the static effect of credit supply on investments is given in regression equation (1)6.

Where is firm i’s level of capital expenditures in period t. The variable is a firm’s cash plus marketable securities divided by total assets in the third quarter of 2006. is a dummy variable that is equal to one for observations after the second quarter of 2007.

is defined as income before extraordinary items and depreciation and amortization normalized by lagged total assets. controls for a firm’s investment opportunities and is defined as a firm’s market value over its book value. The model controls for unobserved firm specific factors

6

Because there can be omitted factors that affect the capital expenditures of a firm over time that are not captured by the quarterly dummies. (E.g. downturn in the local economy or a local regulation that stimulates investments). HAC standard errors will be used which are clustered by firm.

(17)

16 that don’t change over time by including firm fixed effects. Controlling for firm fixed effects also takes firm characteristics such as the industry in which it operates into account. Furthermore, the model controls for unobserved factors that are the same for all firms but not over time such as cyclical effects, nationwide regulations, seasonal effects and other unobserved factors by quarter fixed effects. For to be a consistent estimator, the following crucial assumptions must hold.

(2) The assumption in equation (2) states that the error term of regression equation (1) should be independent of all the included independent variables in the regression. Furthermore, large outliers have to be unlikely and there can be no perfect multicollinearity. Since the

variable doesn’t change over time, there will be multicollinearity between and .

Therefore, won’t be included as a variable by itself but will be absorbed by . The interaction term between and does contain time variation.

To answer the first hypothesis, the model will first be applied on a restricted sample period that is similar to the sample period used in Duchin et al. (2010) to find out whether similar results are found. Second, a regression will be performed with only time fixed effects and firm fixed effects as control variables. Third, the regression equation as in equation (3) will be applied on the complete sample period.

3.3 The effect of credit supply on firm growth

The second part of this study focuses on the effect of credit supply on firm growth in the short- and the long-run. A Firm’s growth can be measured in many ways. To make the answer on the research question in this paper less dependent on the choice of the measure of firm growth, three different measures of firm growth are used: asset growth, employment growth and sales growth. The basic regression specifications that are used to examine the effect of credit supply on firm growth are specified in regression equations (3), (4) and (5).

(18)

17 Where the variable is defined as the growth of total assets in quarter t.

is the relative growth in level of employment in quarter t and is

the relative growth in total sales in quarter t. The variable is a firm’s cash plus marketable securities divided by total assets in the third quarter of 2006. is a dummy variable that is equal to one for observations after the second quarter of 2007. The interaction term is included to examine whether corporate growth after the onset of the financial crisis depends on the cash reserves during the financial crisis. Because there could be reversed causality from a firm’s growth to its cash reserves, the cash reserves of a firm one year prior to the crisis is used as a proxy for its cash reserves during the crisis. is defined as the natural log of total assets. controls for the financial health of a firm and is defined as the total book value of debt over total assets. Model specification (3) which models the growth in assets also controls for the sales growth of a firm. All the growth regressions control for firm and time fixed effects. The firm fixed effects control for some important determinants of firm growth which are initial age, industry and the productivity of a firm (Rahaman, 2011). The time fixed effects controls for the time specific factors that influence firm growth such as seasonal factors. For the regression coefficients to be consistent, the same assumptions should hold as for the investment regression in equation (1). Standard errors will also be clustered at the firm level. The regressions in equations (3), (4) and (5) will be performed with and without control variables.

3.4 Fixed or random effects

So far, fixed effects are used in every model specification. However, when the fixed effects are independent of the independent variables included in the model, the use of random effects is more efficient. In a random effects model, the coefficient that captures firm fixed effects is included in the error term of the regression so that: . To use random effects the following assumption must hold:

The assumption that must hold to use random effects implies that the firm specific effects that are absorbed by the error term in a random effects model are independent of the independent

(19)

18 variables included in the regression. To test whether this is the case, a Hausman test is performed for every model specification7. The Hausman test results will be shown in the results section.

3.5 Disentangling the short- and long-run effects of Credit supply on Investments

and Firm growth

A shock to credit supply is likely to have very different short- and long-run effects. At the moment of the shock, firms that depend on external credit are not able to obtain enough funding in order to reach their optimal investment level. However, in the long-run, firms can substitute debt financing by alternative forms of funding. They can increase their retained earnings or use equity financing for example. Therefore, the main contribution of this paper is to disentangle the short-run effect from the long-run effect of credit supply on investments and firm growth. Two different methods will be used to examine the dynamics in this relation. First, the lagged value of the

dependent variable will be included in the regression equations. The basic regression equation that is used to disentangle the short-run and the long-run effect of credit supply on corporate investments and corporate growth is given in regression equation (6).

Where is a measure of investment or firm growth. is a firm’s cash plus marketable

securities divided by total assets in the third quarter of 2006. is a dummy variable that is equal to one for observations after the second quarter of 2007 and includes the control

variables that are relevant for the included dependent variable. In the dynamic model, the short-run effect of the credit supply shock will be given by . The long-term effect of the credit supply shock will be given by the long-run multiplier which is described in equation (7) below.

An econometrical issue in the estimation of regression model (6) arises because the lagged value of the dependent variable is by construction correlated with the firm fixed effects (Verbeek, 2010). Therefore, a so called standard Within regression estimator will lead to inconsistent parameter estimates of the lagged dependent variable. To deal with this issue, the lagged value of the dependent variable in each model will be instrumented by further lags of the dependent variable using the instrumental variables regression approach. The validity of the lags that are used to

7

The Hausman test compares the regression coefficients in a random effects model with the regression coefficient in a fixed effect model. If the coefficients are significantly different when random effects are used, the null hypothesis under which random effects can be used is rejected.

(20)

19 instrument the lagged value of the dependent variable is tested with use of the Sargan-Hansen test for instrument validity which tests for correlation of the instruments with the error term of the regression model. If the instruments are not correlated with the error term in the regression model, the instruments are valid and can be used to instrument for the lagged dependent variable. First differences of the included variables are taken to control for fixed effects.

3.6 Response function approach

The second approach to disentangle the long-run from the short-run effect of credit supply on investments and firm growth is the use of a response function as used by Wolfers (2006) and Bos et al. (2013). A time dummy variable for several time periods is introduced to examine the effect of the credit contraction on these specific time periods. This time dummy variable will be interacted with a treatment variable that indicates whether a firm is likely to be credit constrained during the crisis. The treatment variable is an indicator variable which is equal to one for firms that belong to the quartile of firms with the lowest cash and marketable securities in the third quarter of 2006. This identification of a firm’s cash reserves during the financial crisis differs from the continuous measure that is used in equations (1), (3), (4), (5) and (6). The identification variable used in the response function approach divides the sample in a group of firms with low cash reserves during the financial crisis and other firms. An advantage of this approach is that the difference between firms that had low cash reserves during the crisis which made the firm highly dependent on external credit during the financial crisis and other firms is identified. The continuous measure that is used before also identifies the difference between a firm that had high cash reserves and a firm that had really high cash reserves during the financial crisis. A disadvantage of the treatment variable that divides the sample in two groups is that it is not a continuous measure which contains more detailed

information about a firm8_{. Using the interaction variable in the response function approach, a}

distinction can be made between the short-run, the medium long-run and the long-run effect of credit supply on investments and firm growth. The response function is specified in equation (8) below.

Where represents the dependent variable which will be investment, asset growth, employment growth or sales growth. is a treatment variable that is equal to one for the quartile of

8

For transparency, the results of the response function regression with the continuous measure of a firm’s cash reserves during the crisis are presented in appendix B.

(21)

20 firms with the lowest cash plus marketable securities in the third quarter of 2006. is a dummy variable that is equal to one if the interval lies in the time interval k for which the dummy is an indicator. until are the control variables that are included which depends on the dependent variable. The same controls will be used as in the basic regression specifications in equations (1), (3), (4) and (5). The response function and the dynamic panel model are used to test the third

hypothesis.

4. Data & descriptive statistics

This section will start by providing information on the credit market conditions during the 2007-2009 financial crisis. Second, a description of the used sample will be given. Third, the

construction of the dependent variables that are used is explained followed by an explanation of the construction of the control variables. Fifth, the sample selection criteria will be given. Finally,

descriptive statistics on the data will be given and discussed.

Figure 3: Libor and Commercial paper spreads over treasuries

This graph shows the 3 month spread of the LIBOR over treasuries and the 3 month spread of commercial paper over treasuries from January 2000 until October 2015. The data used to generate the graph is obtained from

http://www.federalreserve.gov/datadownload/.

4.1 Credit market conditions

To give an idea of the contraction of the credit supply during the financial crisis, Figure 3 above shows the spreads of LIBOR and commercial papers spreads over treasuries from 2000 until

(22)

21 2015. As can be seen in Figure 3, the financial crisis period had a big impact on credit supply

conditions. The sharp increase in the LIBOR and commercial paper spread over treasuries is caused by a shock in the mortgage backed securities. The sharp increase in spreads during the 2007-2009 financial crisis represents a sharp decrease in the supply of credit after a long period of stable and easy credit. Therefore, the financial crisis provides a unique opportunity to examine the effects of credit supply. To avoid the influence of other shocks to the credit market such as the dot-com burst, the sample period that is used in this paper starts in 2002. From 2002 on, the spreads of the LIBOR and commercial paper are relatively constant compared to the financial crisis period.

4.2 Data collection

The data used in this paper is collected from the Compustat database which consists of balance sheet data from North-American firms. Because this paper examines a long-term effect, all available data from 2002 and onwards will be included in the sample period. This results in a sample period from the beginning of 2002 until the end of 2015. Most firm-level data used in this paper is obtained from the Compustat Quarterly Fundamentals database. Quarterly data is used for three reasons. First, the beginning of the financial crisis is of main importance in this research. With use of quarterly data, the distribution of the data in a pre-crisis and an after start of the crisis group can be made more precisely. Second, the use of quarterly data keeps more variability in the data than the use of yearly data which improves the statistical strength of the findings. Third, the use of quarterly data leaves the opportunity to examine effects over a time-period shorter than a year. Data on employment is obtained from the Compustat Annual Fundamentals database because quarterly data on employment is not available. Data on the U.S. consumer price index is obtained from the OECD9 main economic indicator database.

4.3 Dependent Variables

The dependent variable that is used to test the first hypothesis is a measure for corporate investments. In line with the literature that examines the effects of the financial crisis on investments at the firm level10, capital expenditures is used as measure for corporate investments. Capital

expenditures is reported on a year-to-date base in Compustat. This means that the cumulative value for capital expenditures over the fiscal year is reported in every quarter of the fiscal year. Therefore, the value for capital expenditures(capxy) in the previous quarter is subtracted from the current value for capital expenditures for the second, third and fourth quarter of the fiscal year to obtain the

9

OECD, "Main Economic Indicators - complete database", Main Economic Indicators (Consumer price index),http://dx.doi.org/10.1787/data-00052-en (19-5-2016)

(23)

22 quarterly values for capital expenditures. Following Duchin et al. (2010), the value for capital

expenditures is normalized by total assets(atq).

To test the second hypothesis, multiple measures of firm growth are used as dependent variable. The growth of a firm is measured by the growth in assets, the growth in employment and the growth in sales. However, data on employment is not available on a quarterly frequency in the Compustat quarterly fundamentals database. Therefore yearly data on employment is used. To match the yearly data on employment with the quarterly database, it is assumed that employment grows linear during a fiscal year. Based on this assumption, a prediction can be made for the quarterly value of employment by interpolating the yearly data. Employment growth is defined as the natural log of employment in a current quarter minus the natural log of employment in the previous quarter. To measure the growth in sales, the quarterly sales value(saleq) is normalized by the U.S. consumer price index which is retrieved from the OECD database. Sales growth is calculated as the natural log of the normalized sales level minus the natural log of the normalized sales level in the previous quarter. Assets are normalized in the same way as sales. Asset growth is defined as the natural log of the by the U.S. consumer price index normalized value of total assets(atq) in the current quarter minus the natural log of normalized assets in the previous quarter.

4.4 Control variables

To control for the investment opportunities of a firm, Tobin’s Q ratio is used, which is a measure of a firm’s growth potential. Tobin’s Q represents a firm’s market value over its book value and is defined as total assets(atq) plus market capitalization (Closing price multiplied by total shares outstanding (prccqxcshoq)) minus common equity(ceqq) minus deferred taxes and investment tax credit(txditcq) to total assets. (

atq + prccq×cshoq — ceqq — txditcq)/(atq

). Following Duchin et al. (2010), values for Tobin’s Q that are bigger than 10 are set equal to 10 to eliminate outliers. Following Bliss et al.(2015), Cash flow is defined as income before extraordinary items(iby) and depreciation and amortization(dpq) normalized by lagged total assets. Leverage is defined as total long-term debt(dltt) over total assets. Size is generated as the natural log of total assets(atq). All control variables except for Tobin’s Q are winsorized at the 1st and the 99th percentile to deal with outliers. As robustness checks, variables will be winsorized at the 5th and the 95th percentile to check whether the way in which outliers are handled changes the results significantly.

4.5 Sample selection criteria

Firms that operate in the financial industry and utilities are excluded from the sample by dropping firms with sic codes between 6000-6999 and 4900-4999. Financial firms are excluded because they are directly linked to the supply of credit. Utilities are excluded because they have an

(24)

23 exceptional high leverage ratio which can lead to unrepresentative test results. Following Duchin et al. (2010), firms with a market capitalization lower than 50 million dollars in 2002 are dropped out from the sample. This selection criteria eliminates the smallest firms with volatile accounting data. Firms with a market capitalization lower than 50 million in 2002 represent less than 0.2% of the firms in terms of market capitalization. To eliminate firms that are involved in M&A activity from the sample, firms with an asset growth of 100% in one quarter are excluded11_{. Furthermore, only firms}

are kept in the sample that have data available from 2 years prior to the start of the financial crisis until 2 years after the start of the financial crisis. This implies that firms are required to have data available from the 2nd quarter of 2005 until the 2nd quarter of 2009. To make sure that firms make use of external credit, firms are required to have a leverage level of at least 10% one year prior the crisis. Different thresholds for the minimal leverage level will be used as robustness checks. The final sample consists of 909 firms and 48,461 observations.

Figure 4: Firm growth over time

Figure 4 illustrates the firm growth levels for the period from 2002 until 2015 for three different measures of firm growth. Employment growth is defined as the natural log of employment in a current quarter minus the natural log of employment in the previous quarter. Sales growth is calculated as the natural log of the normalized sales level minus the natural log of the normalized sales level in the previous quarter. Asset growth is defined as the natural log of the normalized value of total assets in the current quarter minus the natural log of normalized assets in the previous quarter.

(25)

24

4.6 Descriptive Statistics

Figure 4 shows how the different measures of firm growth evolve over time. As can be seen in Figure 4, the financial crisis clearly has an impact on all measures of firm growth. The impact on sales growth is the biggest, followed by the impact on asset growth, the impact on employment growth is the smallest. The employment growth is the most constant over time while sales growth is the most volatile.

Descriptive statistics of the complete sample are shown in Table 1 below. All values are expressed on a quarterly frequency. The average investment level is 1.4% of total assets per quarter. The average asset growth is similar to the average sales growth which are both 0.7% on average. Employment growth is lower with a growth rate of 0.4% on average. The firms in the used sample have 4,543 million in assets on average, the median value for total assets is 1,255 million. The average employment level is 18,810. Firms have an average cash flow equal to 1.9% of total assets per quarter. The average cash position one year prior to the crisis is equal to 8.4%. The standard deviation of the cash position that firms have one year prior to the crisis is 11.3%.

Table 1: Descriptive statistics

Table 4 reports descriptive statistics for US Compustat firms from the sample period 2002Q1 to 2015Q4. Investment is capital expenditures normalized by total assets. Asset growth is the difference in the natural log of total assets compared to the previous quarter. Employment growth is the difference in the natural log of employment compared to the previous quarter. Sales growth is the difference in the natural log of total sales compared to the previous quarter. Assets is total assets in millions. Employment is the number of employees in thousands. Sales is the total level of sales in millions. Cashposition is the cash and marketable securities over total assets in the 2nd quarter of 2006. Q is a firm’s market value divided by its book value. Cash flow is defined as income before extraordinary items and depreciation and amortization normalized by lagged total assets. Size is the natural log of total assets and Leverage is defined as total long-term debt over total assets.

Standard 25th 75th

Mean Median Deviation Percentile Percentile # Obs.

Investment 0.014 0.009 0.016 0.004 0.017 46,417 Asset growth 0.007 0.002 0.086 -0.021 0.030 45,782 Employment growth 0.004 0.003 0.059 -0.010 0.018 45,671 Sales growth 0.007 0.010 0.261 -0.053 0.078 46,696 Assets 4543 1255 9339 495 3755 47,022 Employment 18.81 5.51 38.90 1.70 16.45 45,809 Sales 1056 300 2175 108 918 47,740 Cash Position 0.084 0.041 0.113 0.016 0.106 46,959 Q 3.565 1.610 3.625 1.172 4.054 48,461 Cashflow 0.019 0.022 0.031 0.012 0.033 45,199 Size 7.192 7.135 1.614 6.204 8.231 47,022 Leverage 0.294 0.256 0.209 0.157 0.385 46,715

(26)

25 To look at the differences in the data between firms with a low cash position and firms with a high cash position at the onset of the crisis, the sample is divided into two groups. The quartile of firms with the lowest cash position at the onset of the financial crisis and the other firms. Table 2 compares investment and growth between the quartile of firms with the lowest cash position at the onset of the crisis with the other firms in the sample before and after the onset of the crisis. It can be observed in Table 2 that investment decreases by 0.1% more for firms that have low cash reserves during the crisis. Asset growth decreases by 1.1% more for firms with low cash reserves than for other firms. In line with this finding, average employment growth decreases by 0.6% more for firms with low cash reserves during the crisis than for other firms and average sales growth decreases by 1% more for firms with low cash reserves during the crisis than for other firms. So for all measures of firm growth it is found that the average growth level decreased more for firms that don’t have much cash on hand during the crisis than for other firms. The level of leverage increased by 2% for firms with a low cash reserve during the crisis while the leverage of other firms remained constant. The increase in leverage for firms with a low cash position is an indication that firms with low cash reserves during the crisis are indeed dependent on the issuance of debt.

Table 2: Mean estimates before and after the onset of the financial crisis

This table reports difference- in means estimates of quarterly investment, asset growth, employment growth, sales growth, size and leverage. Before refers to the period before the third quarter of 2007 and after refers to the period thereafter. Low cash refers to the quartile of firms with the lowest cash position in the third quarter of 2006, others refers to all other firms in the sample. All values are in means per quarter. Difference is defined as the value in the period after the onset of the financial crisis minus the value in the period before the onset of the financial crisis.

5. Results

In this section the results of the analyses described in section 3 are described. First, the Hausman test results that determine whether fixed effects are random effects should be used are given. Second, the results concerning the investment regressions are given and interpreted. Third, the results concerning the firm growth regressions and their implications are given. Fourth, the

Low Cash Others Low Cash Others Low Cash Others

Investment 0.018 0.013 0.016 0.013 -0.002 -0.001

Asset growth 0.021 0.013 0.000 0.003 -0.021 -0.010

Employment growth 0.011 0.007 -0.001 0.001 -0.012 -0.006

Sales growth 0.024 0.016 -0.002 0.000 -0.026 -0.016

Leverage 0.31 0.29 0.33 0.29 0.02 0.00

(27)

26 results concerning the response function approach as described in section 3.5 are given and

interpreted.

5.1 Fixed or random effects

As described in section 3.4, a Hausman test is performed for every regression specification to find out whether random effects or fixed effects are more appropriate. The test results of the Hausman tests are showed in Table 3 below. The Hausman test on the investment regression that is described in equation (1) provides a Chi-squared value of 79.32 and a p-value of 0.03. This means that the null hypothesis under which there is no difference in regression estimators between random and fixed effects is rejected at the 5% level. Therefore, the use of random effects leads to inconsistent estimators and fixed effects must be used. The Hausman test statistics on the

regressions on asset growth and employment growth which are described in equation (3) and (4) respectively both reject the null hypothesis at the 1% level. Contrary to the other model

specifications, the Hausman test statistic on the sales growth regression model that is described in equation (5) doesn’t reject the null hypothesis. With a Chi-squared value of 14.84, a random effects model specification is more efficient to model sales growth.

Table 3: Hausman test statistics

Table 3 reports Hausman test statistics on modelling the dependent variable in regression specifications (1), (3), (4) and (5) respectively using random versus fixed effects. The null hypothesis of the test states that there is no systematic difference in the regression estimators between a random effects and a fixed effects model specification. * means that the covariance matrices used in the test are based on the estimated disturbance variance from the efficient estimator. This option is used when the regular Hausman test leads to a negative Chi-squared statistic.

5.2 Investment regressions

Table 4 shows the output of the investment regressions as described in section 3.2 above. Column 1 of Table 4 shows the results of the regression model that uses a sample period similar to the sample period used in Duchin et al. (2010) which is used to check whether similar results are found. The used sample period in column one is from one year prior to the onset of the financial crisis until one year thereafter which is from the third quarter of 2006 until the third quarter of 2008. The regression coefficient on the interaction term which compares firms with a low cash position at

Dependent Variable Chi-squared P-value Preferred Model

Investment 79.32 0.03 Fixed Effects

Asset growth* 1526.16 0.00 Fixed Effects

Employment growth* 129.58 0.00 Fixed Effects

(28)

27 the onset of the crisis with firms that have a high cash position at the onset of the crisis is equal to 0.00493 and significant at the 5% level. This is in line with the findings of Duchin et al. (2010) who found a regression coefficient of 0.0049 with a significance level of 1%. The coefficient on the crisis variable indicates that investments as fraction of total assets declined by 0.14 percentage points which is significant at the 1% level. The coefficient of 0.00493 on the interaction term implies that a firm needs 28.19%12_{of its assets in cash at the onset of the crisis to mitigate the decline in}

investments after the onset of the financial crisis. Additionally, the standard deviation of a firm’s cash position at the onset of the crisis which is given in Table 1 above is equal to 11.3%. This implies that an increase of a firm’s cash position at the onset of the financial crisis by one standard deviation mitigates the decline in investments after the onset of the crisis by 0.0557 percentage points, or 40% of the total decline for a zero cash firm.

Column 2 of Table 4 presents the results over the complete sample period from 2002 until 2015. The regression in column 2 only controls for firm and time fixed effects. The coefficient on the dummy variable that identifies observations after the onset of the financial crisis is equal to 0.00114 and is not significant. Furthermore, the coefficient on the interaction term between cash position and the crisis dummy is equal to 0.172 percentage points. As expected, the regression coefficient on the interaction term is positive which is in line with the first hypothesis that credit supply has an positive effect on corporate investments. However, the result is not statistically significant when the sample period is extended. Column 3 of Table 4 further controls for a firm’s investment opportunities measured by Tobin’s Q and a firm’s cash flow. Including these controls into the model increases the explanatory power of the model from 2.9% to 3.6%. The regression coefficients on the crisis dummy variable and on the interaction term are still insignificant. The coefficients imply that investments as fraction of total assets declines by 0.00114 percentage points and a cash position of 0.66% of total assets mitigates this decline. However, since both coefficient are not statistically significant, not too much weight should be put on these numbers. In line with economic theory, the cash flow of a firm and its investment opportunities are positively related to its investment level.

When the first hypothesis is tested using the restricted sample period, a significant relation is found between a firm’s cash reserves during the crisis and its investment level. This provides

evidence for a negative effect of a decrease in credit supply on investments. However, when the complete sample period is included in the model, no evidence is found of a positive relationship between credit supply and investments. It is remarkable that the significance on the interaction term disappears when the sample period is extended. A possible explanation for this phenomenon is that

12

The fraction of assets that a firm needs in cash to mitigate the decline in investment after the onset of the crisis is calculated by the decrease in investment after the onset of the financial crisis divided by the increase in investment if a firm holds 100% of its assets in cash: 0.00139/0.00493=28.19%

(29)

28 the effect of credit supply on investments isn’t strong enough to be identified by the experimental design when the sample is extended.

Column 4 of Table 4 shows the results of the dynamic model specification as explained in section 3.4. The regression coefficient on the crisis dummy variable is equal to -0.00169. This finding implies that a firm without cash at the onset of the crisis decreases its investment level by 0.169 percentage points after the onset of the financial crisis. This decrease in investments after the third quarter of 2007 is significant at the 1% level. The coefficient on the interaction term that measures the effect of a firm’s cash position on its investment level after the onset of the crisis is equal to 0.525 percentage points and is significant at the 1% level. These findings imply that a one standard deviation increase in the cash position of a firm at the onset of the crisis mitigates the decline in investments by 0.006 percentage points per quarter in the short-run. The regression coefficient on the first lag of investment is equal to 0.321 which is significant at the 1% level. Since investment is positively related to investment in the previous quarter, the long-term effect of credit supply on corporate investments is bigger than the short-term effect. The long term effect is determined by the long-run multiplier in equation (7) and is equal to 0.00773. This implies that in the long-run, an increase of one-standard deviation in the initial cash position of a firm mitigates the decline in investments by 0.009 percentage points. Therefore, the third hypothesis which states that the short-run effect of a decrease in credit supply is bigger than the long-short-run effect is rejected with respect to the effect on corporate investments. Using the second and the third lag of investment as instruments for the lagged value of investment, a p-value of 0.216 is found of the Sargan-Hansen test which tests for the validity of the used instruments. This implies that the used instruments are valid. The F-statistic for instrument relevance is equal to 303. Since the general rule of thumb states that instruments are relevant when the F-statistic is bigger than 10, it can be concluded that the second and third lag of investment are relevant and exogenous instruments for the first lag of investment.

Does a decrease in the supply of credit lead to a reduction in corporate growth? : evidence from the 2007-2009 financial crisis