The 2008 financial crisis and firms’ investment commitments

(1)

1

Master Thesis—Business Economics: Finance

The 2008 Financial Crisis

and

Firms’ Investment Commitments

Student Name: Yijie Gu

Student Number: 10256369

Thesis Supervisor: Dr. Tomislav Ladika

(2)

2

Statement of Originality This document is written by Yijie Gu,

who declares to take full responsibility for the contents of this document.

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

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

(3)

3

1. INTRODUCTION

The 2007-2008 financial crisis has left us with an unprecedented scale of damage a crisis can bring. However, besides evident losses from the bankruptcies or acquisition of investment banks including Lehman Brothers and ABN-AMRO (Erkens et al, 2012), less distinct influences in heavy machine-operating industries including airlines, oil companies, semiconductor companies and transportation industry are overlooked (Goyal &Negi, 2014). In IATA 2010 annual report, it stated “Early 2009 marked the low point for international air travel markets. From the early-2008 peak to the early 2009 through, premium travel fell 25%. Economy travel fell 9%, the decline softened by a shift to cheaper seats.” Chart 1 shows profitability of airline industry from 2000-2015 (IATA annual report of 2015).

Chart 1: Profitability of airline industry from 2000-2015, source by IATA Annual Report, 2015

From the above chart, there was a blunt flunk of airlines’ profitability around the year of 2008. Naturally, the microeconomic theory of demand and supply would suggest less demand thus less supply. Nevertheless, in industries where heavy machineries are the center of productivity, especially when these machines take time to manufacture, the effect of demand and supply is no longer instant and obvious, but lagged. An airline company that put an order for planes

(5)

5

in 2002 which takes a few years for the manufacture to complete, would supposedly receive the planes before 2008 and therefore be less affected by the 2008 financial crisis than another airline company that put an order in late 2007. The possible outcome of this influence for the more severely impacted company is either cancelling the orders at some cost, or assuming leverage to fulfill the orders.

Thus, the research question arises: How does the 2008 financial crisis affect investment commitments of firms that are in the industries of airlines, transportation, oil &gas and semiconductor? To state the question more detailed: do firms differentiate from the shocks in terms of their investment commitment? How do firms’ policies, primarily on capital expenditures, working capital and rental commitment, respond to sudden shocks? The scale of firms in interest are those that make advanced investment commitments to ensure company operation. With the economic downturn in the case of the 2008 crisis, firms in the aforementioned industries made inflexible investment plans just before which were afterwards hard to commit to, due to numerous reasons such as the lack of demand from the market and the shortage in the supply of cash flow. Under such circumstance, the hypothesis is that firms which made above average investment plans are supposedly more affected by the crisis than firms which made under average investment plans. Moreover, heavily influenced firms might be either under obligation to lessen their investment commitments, or to accumulate debt position in order to fulfill these commitments. In the thesis ‘investment commitments’ is defined as material investment that are key to an industry’s operation, for example, an airline’s investment is planes and its commitment is the fulfillment of the order of planes an airline makes to the manufacturer, and an oil company’s investment is oiling rigs and its commitment is the fulfillment of the order of rigs an oil company makes to the

(6)

6

manufacturer. Investment plans in the research question are the dependent variables and consist of capital expenditures, rental commitments and working capital. Capital expenditures increase mechanically alongside investments. A company that commits on more future investment will see a growth in its capital expenditures accordingly, unless there is a change in the commitment, for example, a cancellation of orders. Also, a company would see a change in its rental commitments unless the purchase commitments are altered which might be due to again cancellation of orders. However, reduction of long-term debt or working capital might suggest that a company try to fulfill the commitment by taking up more debt or using its working capital as extra funding for the commitment. More detailed reasoning for using these dependent variables are discussed in section 3.3 Variable explanations.

I find that companies with higher than mean future purchase commitments when hit by the 2008 crisis tend to increase their rental position. There is not enough evidence to say whether they reduce the working capital to fulfill the purchase commitments, nor is there a clear result on whether they reduce their capital expenditures.

The rest of this paper is organized as follows. In section two, the literature review is provided with existing theories related to this research. Hypotheses are thereafter derived from related papers to be tested. In section three, the methodology of this paper is adapted, where a difference-in-differences experiment is setup to test whether the investments of above average committed firms are more influenced by the 2008 shock. The dataset of this paper includes 43 companies from selective time frames between 2000 and 2015 and totals 607 observations. The variable that is unique to this dataset is ‘future commitment within 5 years’ that I hand collected from annual 10-k report. More detailed methodology and data collection

(7)

7

process will be provided in this section. In section four, descriptive and summary statistics for both dependent and independent variables are presented with explanation to key information of the dataset. Afterwards, results from the test are provided with their economic interpretation. In section five, robustness check follows to test whether companies replace planned purchases with capitalized leases. Lastly in section six, I conclude, in the context of related literatures, that the results are in line with existing researches. More importantly, I will discuss the limitation of the research approach and provide directions for future research.

2. LITERATURE REVIEW

The following papers shed light on this thesis. This part is conducted by firstly, stating and comparing the papers’ finding and contributions; secondly, relating them to the thesis and lastly, form hypotheses accordingly.

The effects and consequences of a financial crisis are in every aspect and to numerous extent. One distinctive aftermath of a financial crisis is the difficulty of fulfilling financial contracting. Almeida et al. (2011) discussed if firms with larger percentage of their long-term debt maturing just when the crisis happens experiences more pungent results than firms with similar traits which doesn’t require debt refinancing during the crisis by using a difference-in-differences approach. In place of the independent variable being investment commitment within 5 years in this thesis, they use term debt maturity and find that firms with long-term debt maturing pre-crisis cut their investment ratio by 2.5 percent more per quarter than peer firms with long-term debt maturing after the crisis had taken place. Similar to long-term

(8)

8

debt, the variable of interest in this paper—investment commitment within 5 years is hard to renegotiate on short notice during crisis.

In accordance with Almeida et al. (2011), Campello, Graham and Harvey (2009) investigated if firms are credit constrained during the global financial crisis of 2008 and whether firms’ planned expenditures deviated conditionally on financial constraint, by a survey-based approach from 1,050 CFOs in the U.S., Europe and Asia. They found that financially constrained firms not only had hardness to fund their operations and were more likely to face debt overhung which lead to the termination or bypassing positive NPV projects, but also experienced blunter cuts in capital expenditures, where they canceled or delayed their planned investments. Noticeably, capital investment is reduced by 9% in 2009 in all three regions where they conducted the survey. More importantly, the paper states that ‘the firms that are cutting investment the most during the crisis are those that were overinvesting before it’. Despite its limitations as a survey-based analysis, the paper adopts matching estimator introduced by Abadie and Imbens (2002) and Dehejia and Wahba (2002) to avoid endogeneity. This method includes matching firms based on size, ownership form, credit rating, profitability and industry classification. Similar method is adopted in this thesis with one-by-one inclusion of control variables such as firms’ Q, total assets, return on asset and predetermined industry selections. This is extensively discussed in section three.

Duchin, Ozbas and Sensoy (2010) studied on the recent 2008 financial crisis on corporate investment with a focus on non-financial firms. They applied panel data of corporate investment and controls for firm fixed effects and time-varying measures of investment opportunities. Their finding is consistent with Campello, Graham and Harvey (2010) such that

(9)

9

the corporate investment decline is greater in firms which depend more on external finance. Likewise, their finding assists Almeida et al. (2011) with the insight that non-financial firms also suffer from the negative shock of 2008 crisis with respect to their investments. The difference of these two papers is the identification strategies that they employed. In order to avoid endogeneity, their measurement of firms’ financial positions dated up to four years prior to the 2008 crisis. They then conclude that the results for corporate investment decline in 2008 are not similar to those of prior years’, which provide this research with evidence that the 2008 crisis is a distinctive event that influences companies’ investment. Unique to Duchin’s research is ‘excessive cash holding’ and it positively relate to post-crisis investment that suggests excessive cash holding before the crisis play a precautionary role. Thus, this thesis incorporates excessive cash holding as a control variable which will be discussed in the latter parts.

Regardless of the 2008 crisis, Rauh (2006) investigates how corporate expenditures based on capital expenditures and firms’ R&D and acquisitions respond to the change in their defined pension plans. He finds that capital expenditures declines are more evident among firms that experience financial constraints with obligatory defined benefit pension plans. Particularly, when estimating the effect of required contributions on investment, he controls for Tobin’s Q and cash flow, arguing that the procedure is defensible with unseen investment opportunities related to the pension funding status functionally for the reason that the correlation between cash flow and investment might be influenced by the omitted variable of investment profitability. Furthermore, his research indicates that smaller firms are more likely to be financially constrained, which suggests that instead of regressing arbitrary investment

(10)

10

commitments, this thesis should look at the proportion of firms’ investment commitments to total assets to gauge the effect of the 2008 crisis on their expenditure policies.

Yet another early research on corporate capital expenditure decisions by McConnell and Muscarella in 1985 inform about investment commitment, which show that not rarely that firms have investment plans that last more than 1 year. This is an important premise of my study for the reason that, without the confirmation that a number of firms make advance investment, this thesis would fail in convincing that economic crisis affect firms differently with regard to their investment commitments. Precisely, they study the relationship between firms’ future capital expenditure plans and stock returns and find that for industrial firms, excessive stock returns and changes in planned capital expenditures are positively correlated.

Lemmon and Roberts (2010) also take their focus of research to whether some firms’ investments are affected more by an external shock than other firms. Their main finding is the investments of below-investment-grade firms (firms with lower than BBB ratings) are more affected by the external shock—contraction in the supply of credit. While this finding is less relevant to my study in the sense that firms’ credit rating don’t necessarily have a causal relationship with capital expenditure, its empirical setting of a difference-in-differences experiment can be implemented in my research.

To summarize, previous literatures are in agreement that corporate investment decisions are likely to be affected by external shocks, with a differentiation of firms’ financial constraints. Therefore, my hypothesis is: Within the time-frame of 2000-2015, firms with above average future investment commitments of five years in the industries of airlines, oil& gas,

(11)

11

semiconductor and transportation are more affected by the 2008 financial crisis revealed in their investment policies measured by their capital expenditures, rental commitment in 5 years, working capital and growth rate of long-term debt.

3. METHODOLOGY

In this section, I start with describing the design of the difference-in-differences approach by stating and explaining the equation and reasoning of including each variable, and then the dataset that I use later on in the tests.

3.1 Difference-in-differences Approach

As is mentioned in the literature of Almeida et al. (2011), they adopt matching estimators strategy to test whether firms that have to refinance their long-term debt when a financial crisis happens change corresponding corporate behaviors. The rationale for this strategy is that a randomized experiment would give stronger and more robust causal relationships than results from observational data. By using matching estimators in their tests, they isolate treatment group which consists of firms with debt maturing during the 2008 crisis, after which they search for control groups which consists of firms that fit the treatment group’s trait the best. They carefully make matches from certain dimensions including firm size, profitability and long-term debt to ensure that the treated and control groups are as identical. This gives an important point that I must look into, that for this study of firms’ investment commitments, firms’ characteristics being as close to each other as possible is very crucial. Without this stand

(12)

12

point, the casual relationship would fail to prevail. Thus, difference-in-differences approach in this thesis, in order to establish causal results instead of only correlations, utilizes this procedure. Inferences about the treatment interest (highly committed firms’ investments during 2008 crisis) are indicated by the in-differences estimator. In the difference-in-differences estimation, I focus on the variables that could affect both treatment group and observed outcomes, instead of mainly concerning about providing a model that aim to completely explain the endogenous variable. To illustrate, the outcome observed in my equation is investment policies, which can be measured differently than in this thesis according to numerous theories. Nevertheless, I include in my estimations covariates that have higher possibility to be involved in simultaneity between the treatment and outcomes. For example, my categorical identification for companies is firm’s industrial classification with SICs and under that category are firms’ Gvkeys. This match in the industrial category offers the comparison later on in the regression more standardized explanation where firms have similar operating nature. However, bond ratings are not presented are a categorical variable for the reason that investment in machineries has to be conducted as long as the firm is still in business. My non-categorical variables are based on Almeida et al. (2011) with the inclusion of firms’ Q—market to book ratio, cash holding, firms’ size and leverage ratio. These covariates are commonly accepted to capture firm heterogeneity.

3.2 Equation

The primary test is whether firms’ capital investment policies are more influenced by having more investment plans outstanding just before the 2008 crisis. Therefore, I employ firms’ future investment commitments in five years as an independent variable, alongside the

(13)

13

dummy variable of the year being 2008 and thereafter; and the difference-in-difference estimator. Beneath is the equation that I work with.

Yit = β0 + β1(Gi ∗ Dt) + β2Gi + β3Dt + Uit

Yit is the outcome variable which depicts firstly, capital expenditures (COMPUSTAT’s capx/at); secondly, working capital (COMPUSTAT’s wcap/at); rental commitment for 5 years in total (COMPUSTAT’s mrct/at) and finally growth of total long-term debt (COMPUSTAT’s ((dd1+dltt)/at)/(lagged (dd1+dltt)) with i understating companies’ identifiers using gvkey and

t understating year ranging from 2000-2015. Gi is a dummy variable denoting 1 when the

company is in treatment group consists of firms with above average future investment commitments in 5 years (control group is then firms with under average future investment commitments in 5 years). In the baseline analysis, future investment is only committed material purchase investment, excluding lease commitments. However, in the robustness check, capitalized lease commitments is included in the independent variable alongside purchase commitments. This is discussed later on in section 5. Robustness check. Dt is another dummy variable denoting 1 when the year is 2008 and thereafter (the treatment). Coefficient of interest β1 , which is an independent variable describing firms’ future investment commitment in 5 years. The procedure of this particular variable collection will be explained in detail later on. Uit consists various control variables such as firms’ Q, size, performance based on firms’ return on asset, long-term debt level, cash holding and cash flow.

(14)

14

Firms’ capital expenditure is a relatively straightforward form of proxy of firms’ investment policy, as Richardson (2006) conducts his study on over-investment of free cash flow with dependent variable being investment expenditure including capital expenditures. Thus , in this thesis, I adopt capital expenditure as one of the measurements of firms’ investment policies. The second measurement of investment policies I use is working capital, which serves as a source of an input and a readily reversible store of liquidity to fund investment. Fazzari and Petersen (1993) states that working capital is current assets including account receivables, inventories and cash minus current liabilities including accounts payable and short-term debt, which counts for firms’ net position of liquidity in asset. They find a negative relationship in the regression of working capital and fixed-investment. Therefore, I expect working capital to be negatively influenced by the 2008 crisis because of the reduced financing options for the committed investments. The third measurement of investment policies for my equation is rental commitments for 5 years in total. This variable represents the minimum rental expenses due in future 5 years under all existing non-cancelable leases. The incorporation of this variable offers an insight on the fact that firms might turn to leases instead of purchases of machineries when facing the crisis. The fourth dependent variable is the change in long-term debt. Besides investment policies adjusting to the crisis and firms’ investment commitments, firms could also assume extra leverage to fulfill those commitments.

In this thesis, firms’ operating expenses is not considered a valid measure of firm’s investment plans. As shown in Chart 2 with airlines’ operating expenses components, almost two thirds of the operating costs are ‘labor costs’ and ‘fuel costs’. Fuel costs are exogenous and the main way to reduce fuel costs is to reduce the profit generating services; whereas labor costs are the first to be reduced when hit by an economic downturn. However, layoffs are not

(15)

15

necessarily related to planned future purchase commitments. They relate more to other endogenous factors such as airlines’ profitability. The potential spillover effect of meeting the crisis on reducing operating expenses is out of the scope of this thesis.

Chart 2: Airlines’ operating costs components

The simultaneity and endogeneity between treatment and outcomes can be discontinued by including various control variables in the regression. In this thesis, the incorporated controls are firms’ market-to-book ratio (firms’ Q), performance, cash flow, size, cash holding and long-term debt level. The rest of this paragraph I explain some variables from above with the support of existing literature. Firstly, Q is considered related to investment opportunities in firm value where a higher market-to-book ratio suggests stronger increase in investment opportunities in firm value (Gompers, 1995). The exclusion of this variable might cause endogeneity problem in the way that higher market value of a firm with considerable investment opportunity may enable the firm to increase the component of its investment

(16)

16

policy regardless of the 2008 crisis or having a huge amount of investment commitments. Secondly, cash flow is also an important determinant of firms’ investment policy. Fazzari et al. (1988) claims that a firm’s internal cash flow may influence its investment policy through ‘financing hierarchy’ in which internal cash flow is more cost efficient than external funding such as new debt and equity and therefore cash flow and firms’ investment decisions are interdependent. The expectation for the relationship between cash flow and investment decisions would be positive. Thirdly, firms’ performance based on ROA could affect their investment decision through the profitability of those firms. Logically, they would assume a positive relationship, although many papers have disapproved traditional performance measurements such as return on asset, return on investment and return on equity. Fourthly, as discussed in literature review section, cash holding is also concerned when gauging firms’ investment decisions. Additionally, Denis and Sibilkov (2007) provides robust evidence that greater cash holdings are related with higher level of investment for both financial constrained and unconstrained firms in their study where higher cash holdings allow constrained firms to take positive NPV projects. In line with the above paper, Arslan et al. show that constrained firms exhibit greater investment–cash flow sensitivities than unconstrained firms and that cash stands as an effective tool for companies, mainly during the crisis period. Thus, I assume that cash holdings are firms’ investment decisions might be positively related. Finally, long-term leverage is expected to have a negative relationship with firms’ investment policies. Evidence from Almeida et al. in 2011 suggests that firms with large amount of debt maturing during the crisis decrease their investment to repay those debts. With the limitation of substitution to bank debt and alternative sources of capital (e.g., equity, cash balances, and trade credit), there is an almost one-for-one decline in net investment with the decline in net debt issuances (Lemmon and Roberts, 2006).

(17)

17

3.4 Data Description

To lead the empirical tests, I use data from COMPUSTAT’s North America Fundamentals Annual and firms’ annual 10-k reports retrieved from SEC EDGAR website (https://www.sec.gov/edgar/searchedgar/companysearch.html). The sampling relies mainly on firms within the industries of oil& gas, transportation (shipping, railroad and trucking), semiconductors and airlines and limited to firms with 10-k reports available on SEC EDGAR. I require firms to expose valid information on their total assets, capital expenditures, total long-term debt, net income, total sales, rental commitments in 5 years, working capital, lease commitments in 5 years, market capitalization, common equity, deferred taxes and investment tax credit, depreciation and amortization, cash and short-term investments on COMPUSTAT. I hand pick the data needed from 10-k reports, which in detail, are total purchase commitments up to the next 5 years of rigs(oil& gas companies), ships and other carrier types (transportation companies), wafers (semiconductor companies) and aircrafts (airlines). The future commitments within the next 5 years of these firms are firstly scaled by each firms’ total asset. These commitments are then categorized into ‘high commitment’ and low commitment according to the mean of the summary statistics of the variable ‘future commitment in 5 years’. Firms with above mean future commitments are in the treatment group and below mean future commitments are in the control group. Ideally, the firms within the sample starts providing information on investment commitments in 2000 till 2015. However, for some firms, the information is limited for reason such as bankruptcy after 2008. Due to the low rate of disclosure of investment commitments of firms, the final sample consists of 43 individual firms and 607 observations.

(18)

18

In the baseline analysis, outcome variables are explained in the former part of equation. Note they are all scaled by firms’ total assets to avoid arbitrary larger investment in larger sized companies, and then are winsorized for later regression. As stated before, the construction of control variables partially follow that by Almeida, Campello, Laranjeira and Weisbenner (2011). Firms’ Q is defined as total assets plus market capitalization minus common equity minus deferred taxes and investment tax credit divided by total assets (COMPUSTAT’s

(at+prcc*csho-ceq-txditc)/at). Firms’ return on asset is defined as net income/loss divided by sales times the

ratio of sales to total assets (COMPUSTAT’s ni/sale*sale/at). Firms’ cash flow is defined as net income/loss plus depreciation and amortization divided by total assets (COMPUSTAT’s

(ni+dp)/at). Firms’ size is defined as the log of total assets (COMPUSTAT’s at). Cash holding is

defined as cash and short-term investment divided by total assets (COMPUSTAT’s che/at). Long-term leverage is defined as total long-term debt divided by total assets (COMPUSTAT’s

(dd1+dltt)/at).

4. RESULTS

In this section, I start with the summary statistics on my samples of treatment and control groups of firms along with control variables. I then present the baseline empirical test results.

(19)

19

Summary statistics are presented in Table 1 with mean, standard deviation, observations, variance, minimum, maximum, and percentiles of 10th_{, 25}th_{, 50}th_{, 75}th_{, 90}th_{of each variable}

without differentiating treatment and control group. All variables are winsorized at 1% to avoid large outliers influences. Independent variable future commitments in 5 years has an mean value of 14% (column 1) of total assets with a standard deviation (column 2) of 0.21, ranging from 0 to 1.33 (column 5&6). 10th_{percentile of value 0 indicates that 10% of the}

sample companies in the lower commitment bound didn’t incur any investment commitment and 90th_{percentile indicates that 10% of the sample companies in the higher commitment}

bound incur 30% investment commitment out of total assets. There are yet even higher commitment of more than 130% of total assets made by companies. Dependent variable capital expenditures has a mean proportion of 11% from total asset with a standard deviation of 0.086 ranging from almost 0 (0.0052) to 0.45. Its 90th_{percentile indicates that 10% of the}

higher capital expenditure bound incur that of 22% from total asset. Noticeably, rental commitment in 5 years has the largest mean proportion of 24% from total asset out of the dependent variables alongside the largest standard deviation of 0.48 ranging from almost non-existing to more than 300% of total asset. This suggests that some sample companies make high rental commitments which would presumably lower their purchase commitments. Regarding control variables, cash holding and cash flow are on average at about 15% of total assets and some companies incur negative cash flow and return on asset.

Here please refer to Table 1

Summary statistics of treatment group and control group separately and test of mean comparison is presented in Table 2. It shows the differences between high investment

(20)

20

commitments firms and low investment commitments firms which I split according to the mean of future investment commitments in 5 years. Testing for the mean difference provides an insight on the effectiveness of the control variables. Having a glance at the summary statistics of Table 2, there are around 420 firms in the control group and around 180 firms in the treatment group. Each control variable may have different observations due to the availability of data. There exist larger differences in the mean value of firms’ Q and return on assets than long-term debt/assets, cash holding/assets, cash flow/assets and firms’ sizes. P-value of mean comparison in long-term debt/assets is 0.303, which suggests that this difference between the treated and control exhibits no significant value than 0 at a 5% level. Similarly, p-value of mean comparison in cash holding/assets and cash flow/asset also suggests no evident support for non-zero value at 5% level. Thus, the above controls variables are potentially effective for the fact that they have the characteristics of Abadie and Imbens’ (2002) matching estimator to ensure that treatment and control groups are not otherwise influenced by the differences in the control variables. However, firms’ Q and ROA might not be as effective as the other four control variables since their test results indicate non-zero differences of the mean value. The paper published in 1997 by Ghalayini et al. explains why conventional measure of profitability such as ROA are to some extent invalid. Therefore, these two control variables are possibly biased, which need to be taken into consideration when conducting baseline analysis.

(21)

21

I start with difference-in-differences experiment of firms’ investment commitments interacting with the 2008 financial crisis; then I add in control variables to test my hypothesis of whether within the time-frame of 2000-2015, firms with above average future investment commitments of five years in the industries of airlines, oil& gas, semiconductor and transportation are more affected by the 2008 financial crisis revealed in their investment policies. In my analysis, firms are highly committed in investment when they have above mean value future investment commitments in 5 years within the sample; otherwise firms are not highly committed in investment. This division is done through a dummy variable being 1 if highly committed and 0 otherwise.

Here please refer to table 3

In this analysis, I conduct the difference-in-differences regression on four different outcome variables of measures of investment plans, namely capital expenditure, rental commitments in 5 years, working capital and growth rate of long-term debt. Table 3 shows the relation between investment plans and highly committed companies being shocked by the 2008 crisis during 2000-2015. Column (1) starts with capital expenditure as dependent variable regressed on being a highly committed company hit by the crisis. The coefficient of difference-in-differences estimator is significant at 10% confidence level. Thus, further testing with addition of control variables is necessary for establishing causal relationship between being highly committed companies during the crisis and having to change investment plans. As expected, being in the treatment group of above mean purchase commitment indicates a great economic effect on capital expenditures with being significant at 1% confidence level. Future purchase commitments and capital expenditures are positively correlated. In other words, if a company is in the treatment group with above average future purchase commitments, then

(22)

22

its capital expenditure is on average 2.5% more than that of a company with below average future purchase commitments. One reason would be that companies in the high commitment group might have already incur capital expenditure such as a deposit payment when they put an order into manufacturing. Also, there is strong significance at 1% confidence level with the constant. This interprets as average firms have positive capital expenditures of 11% of their total assets, which confirms the fact that capital expenditure is a valid measure of firms’ investment plan. Column (4) presents the regression results with rental commitments in 5 years as dependent variable regressed on being a highly committed company in the crisis. The coefficient of difference-in-differences estimator is insignificant at 10% confidence level, which indicates limited economic effect of being a highly committed company in the crisis on the rental commitments. Nonetheless, the 2008 crisis leads to an average increase of 19% on rental commitments. This would happen due to a possible reason that during the crisis, companies on average would convert from purchase commitments to rental commitments because of the shortage of credit or funding, which is in line with former literatures arguing financial constrains would cause companies to deviate from their original investment plans. As well, the constant staying significant at 1% confidence level with an increase of coefficient to almost 30%. This shows that regardless of the 2008 crisis and having a great deal of purchase commitment, an average firm incur 30% rental commitment of its total asset. Column (6) presents the regression results with working capital as dependent variable regressed on being a highly committed company in the crisis. The coefficient of difference-in-differences estimator is significant at 10% confidence level, with a negative relationship between being a highly committed company in the crisis and its working capital. Heavily committed companies during the 2008 crisis reduce about 4% of their working capital compared to below average committed companies. There is also a difference when comparing

(23)

23

the effect of the independent variable on capital expenditure and working capital, with a 3% decrease of capital expenditures and a 4% decrease of working capital. One of the reasons for a more severe decrease in working capital than capital expenditures would be that working capital serves are a funding of liquidity for company operation and most prone to be utilized when facing an external change, concluded by Fazzari and Petersen in 1993. Thus when experiencing a financial shock, firms are more likely to first offset the effect of the shock by the means of working capital; while capital expenditures might include fixed expenses that are hard to manipulate on a short notice. Column (8) presents the regression results with growth rate of long-term debt as dependent variable regressed on being a highly committed company in the crisis. The coefficient of difference-in-differences estimator is insignificant at 10% confidence level, which indicates limited economic effect of being a highly committed company in the crisis on the rental commitments. Other variables are also insignificant in this regression, meaning that none of the independent variables could economically explain the dependent variable. The main reason could be that the standard error of the dataset is large, which in turn gives the insignificance. This can be found from the coefficient of the constant that shows on average firms don’t change their long-term debt and that the growth rate is not a valid outcome variable to conduct the tests on. Therefore, I exclude the growth rate of long-term debt as a dependent variable and the hypothesis that firms with above average future purchase commitment increase their debt position when hit by the 2008 financial crisis cannot be proven evident.

In summary, I find evidence to support the effects of being a highly committed company hit by the 2008 financial crisis on its capital expenditures and working capital, but not rental

(24)

24

commitments or growth rate of long-term debt. Continue on is the multivariate analyses of the sample.

4.3 Multivariate regressions

In this paper, multivariate tests control for a number of firms’ characteristics whose potential correlations with the outcomes variables measuring firms’ investment policies are discussed in the methodology. These control variables are firms’ Q, size, cash holding, cash flow, return on assets and long-term debt. Moreover, firm fixed effects defines at Gvkey level are included in the regressions since the effects on firms’ investment plans are likely to be firm specific.

4.3.1 Difference-in-differences analysis of capital expenditures as

dependent variable

Here please refer to Table 3, Column (2) & (3)

Table 3, Column (2) and (3) conduct regressions with firms’ capital expenditures as a measure of firm’s investment policies. Additional control variables on top of the univariate analysis in section 4.2 seek to avoid endogeneity and provide with reasonable explanations. The 2008 crisis dummy equals 1 if the fiscal year is 2008 and thereafter and 0 otherwise. The high commitments dummy equals 1 if the company is in the above average future purchase investment treatment group and 0 otherwise. Column (2) starts with an addition of firms’ size, long-term debt, Q and return on assets as control variables. The effect of the difference-in-differences estimator drops by 0.1% compared to the univariate analysis but still significant at 10% confidence level. Having higher future investment commitments still exhibits a significant

(25)

25

effect of increasing capital expenditure at 1% confidence level. These figures suggest that capital expenditures decrease by 2.6% more for firms that have above average future investment commitments when the 2008 crisis hit, relative to controlling firms with less than average future investment commitments. Firm size indicates a strongly significant 0.6% decrease in capital expenditure when the total assets of a firm grows by 1 unit when not controlling for cash flow and cash holding. One reason for this could be the larger the firm is, the more future planed investment commitment it has and the more urgent it might have to cancel these commitments because of the lack of funding during the crisis and therefore decreasing total capital expenditures. This finding supports the hypothesis with regard to the effect on capital expenditure. The regression also confirms that ROA explains little about capital expenditures presented in Table 2. The coefficient of ROA shows an insignificant 0.05% increase of capital expenditures when ROA grows by 1 unit. Otherwise, the regression results remain more or less the same as column (1). The reason that return on assets might not affect capital expenditure would be that a firm with a large amount of assets could generate same level return as firms with less assets without necessarily changing its capital expenses. For example, airlines might not differ in return when the demand for flight services is lower (in non-busy seasons) no matter how many aircrafts the company has since a number of aircrafts would suffice to carry out the productivity; however, those big companies wouldn’t necessarily cut their capital expenditures (aircrafts, hub costs) merely because of their return on assets is not sufficiently high. Controlling for long-term debt and Q doesn’t change the regression results dramatically. However, the regression results in column (3) is strikingly different from previous result, where the coefficients of difference-in-differences estimator are borderline insignificant at 10% confidence level with the inclusion of cash holding and cash flow as control variables. But the relationship between being a high commitment company

(26)

26

during the crisis and its capital expenditures follows the same negative trend. These results give strong significance to cash holding and cash flow at 1% confidence level. With 1 unit increase in firm’s cash holding, capital expenditures drop 32%, while with 1 unit increase in firm’s cash flow, capital expenditures rise almost 5%. On the one hand, a possible explanation is that there is little evidence in these pairs of regressions that capital expenditures relates with high commitment and the 2008 crisis; moreover, that capital expenditures are merely negatively driven by cash holding since retaining cash would result in less capital expenditures and are positively driven by cash flow since more liquidity could be provided to fund capital expenditures. On the other hand, one could argue that the coefficient of the difference-in-differences estimator is borderline insignificant at 10% confidence level. This might happen because of the sample data limitation. A possible solution would then be having more observations in the sample which is hard to conduct in this research for the unavailability of future purchase commitment in 10-k reports that can be found online.

To summarize, Table 3, column (2) states 10% significant results when controlling for firms’ size, long-term debt, ROA and firms’ Q, which provides evidence that being a highly committed company hit by the 2008 crisis has a negative effect of almost 3% on its capital expenditures. However, in addition to Table 3, column (1) the univariate tests, column (3) also give a concern that after controlling for firms’ cash status, it is possible that capital expenditure is simply affected by their cash status but not their investment commitment during the crisis.

4.3.2 Difference-in-differences analysis of rental commitments in 5 years as

dependent variable

(27)

27

Here please refer to Table 3, Column (5)

Table 3, column (4) and (5) conduct difference-in-differences analysis with rental commitments in 5 years as dependent variable. The overall setting is similar to that of previous columns. For every regression the effect of being in the years of 2008 and thereafter is positively related to firms’ rental commitments. This can be interpreted as firms in both control and treatment group increase their rental commitment because of the shortage of funding that 2008 financial crisis brings. Also, firms’ size is of great negative significance at 1% confidence level to indicate that firms with more assets tend to have less rental commitment and this might be due to the fact that firms with more assets are more likely to use the current asset instead of renting more. I start with adding firms’ size as control variable and there is no evidence that a company with high investment commitment and hit by the crisis increase their rental commitments; yet the coefficient of the difference-in-differences estimator increases from 0.03% in the univariate test to 0.8% here. Thus, I expect that by adding more control variables would improve the significance of this estimator. Includes ROA as control variables and gives similar results as column (4) with an exception of the coefficient of interest increasing from 0.8% to 1.8% even though it is still not significant. With the inclusion of more control variables, there is significant result for being a higher committed firm or being hit by the 2008 crisis, with a stronger coefficient for the interested effect on rental commitments by adding long-term debt and firms’ Q as control variables where long-term debt shows positively significant effects in both regressions on rental commitments. 1 unit increase in long-term debt would lead to almost 30% increase in rental commitment. Column (5) shows significant effect of being in the treatment group where firms have higher than mean future investment

(28)

28

commitments and are hit by the 2008 crisis at 10% confidence level with the addition of cash holding and cash flow to the current control variables. Long-term debt, cash flow, cash holding are all significant at 1% confidence level. This could be interpreted as firms with higher than mean future investment commitments increase their future rental commitments by approximately 13% after hit by the 2008 financial crisis. In this case, firms might choose to commit rental obligations instead of having to make purchase commitments of machineries. Noticeably, with one unit increase in cash holding, rental commitments grow by almost 100% as shown in the last two columns; while one unit increase in cash flow would decrease rental commitment by 30%. Right opposite to Table 3, column (1) to (3) which have an dependent variable of capital expenditures, this column is with rental commitments excluding capitalized leases. One reason could be that cash holding provides a precautionary saving for firms (Han& Qiu, 2007) and rental expenses are reversible, which companies are prone to invest in when there is increase of cash holding in the company and they would prefer to maintain the cash holding level later on by cutting down the rental commitment, unlike the non-reversible capital expenses. An increase in cash flow would, on the other hand, encourage firms to invest in fixed asset instead of rentals since they might financially be in better health than those which didn’t generate a positive change in cash flow and had to take rental position. Thus, cash holding has positive effects and cash flow has negative effects in Table 3, column (5).

To summarize, with the addition of control variables, the treatment group of higher commitments hit by the 2008 crisis exhibits an economically significant effect on rental commitment in 5 years. Treated firm on average would increase their rental position by almost 13%.

(29)

29

4.3.3 Difference-in-differences analysis of working capital as dependent

variable

Here please refer to Table 3, Column (7)

Table 3, column (6) & (7) conduct difference-in-differences analysis with working capital as dependent variable. The overall setting is similar to that of Table 3, column (5). In the univariate tests conducted in Table 3, the difference-in-differences estimator in column (6) shows 10% significance. However, this significance fades later on with addition of control variables, which are almost all significant at 1% confidence level in every regression. This finding strikingly reputes the univariate tests. One reason could be that the highly committed hit by the crisis doesn’t explain the change in their working capital, but the control variables do. This could happen because working capital is defined as firm’s current asset/current liability. It shows whether a firm can sustain its daily operation. When a highly committed firm is hit by the crisis, neither its current assets (accounts receivable, inventory, marketable securities and prepaid expenses) nor its current liabilities (short-term debt and accounts payable) would have to change accordingly. For example, the effect of long-term debt on working capital is significantly negative throughout the regressions while that of cash holdings is significantly positive Long-term debt might not have a direct effect on current liability but some companies lump the current portion of long-term debt in its working capital as well. Thus, working capital decreases with increasing long-term debt’s current portion in current liability. However, cash holding is practically internal cash and cash equivalents, which are important component of current asset. Therefore, the increase in cash holding would also increase working capital.

(30)

30

To summarize, Table 3, column (7) exhibits insignificant result of the difference-in-differences estimator on working capital. This finding provides little evidence that being a highly committed company hit by the 2008 crisis would utilize its working capital to fulfill those future investment commitments. Nevertheless, control variables such as long-term debt and cash holding could be the driving force behind working capital.

5. ROBUSTNESS TEST-- Capitalized Leases

In this section, I conduct the robustness test with the sum of future capitalized leases for 5 years and future purchase commitments in 5 years (capital expenditures, rental commitments in 5 years and working capital) as the dependent variables to see whether the effect of being a highly committed company hit by the 2008 crisis remains. Capitalized leases is constructed using the total amount of COMPUSTAT’s CLD2, CLD3, CLD4 and CLD5. Independent variables are constructed the same as in univariate analysis with the 2008 crisis dummy equals 1 if the fiscal year is 2008 and 0 otherwise and the high commitments dummy equals 1 if the company is in the above average future purchase investment treatment group and 0 otherwise. Again the coefficient of interest is the difference-in-differences estimator. The reason for this robustness test is that capitalized leases are similar to investment commitment in the way that they are mostly non-reversible and that capitalized leases are more intensively used in some industries such as the airline industry than others. Thus, I would want to see how much this effect of capitalized leases can be transferred to the industries in this paper, namely, oil& gas industry, transportation industry and semiconductor industry.

(31)

31

Table 4 presents with relationship between investment policies including capitalized leases and highly committed companies being shocked by the 2008 crisis during 2000-2015. Column (1) has the sum of capital expenditures and capitalized leases as outcome variable. There is a significant difference-in-differences estimator at 5% confidence level. Its coefficient of -0.054 could be interpreted as when companies that have more than average future purchase commitment are shocked by the 2008 crisis, they cut down capital expenditures including capitalized leases by 5.4%. This finding is in accordance with the univariate analysis with stronger significance. One reason could be that often the industries in this research also use capitalized leases intensively instead of solely relying on capital expenditures. Column (2) has the sum of rental commitments in 5 years and capitalized leases as outcome variable. However, there is no economically significant relationship between being highly committed companies experiencing the 2008 crisis. This agrees with the univariate analysis. Yet, the multivariate analysis would suggest with the addition of control variable there might be significant effect of being in the treatment group on rental commitments in 5 years and capitalized leases. Column (3) has the sum of working capital and capitalized leases as outcome variable. The result suggests that being a highly committed firm shocked by the 2008 crisis has a negatively significant effect of 5.3% on the working capital and capitalized leases. This supports the univariate analysis in the way that companies in the treatment group reduce their working capital to fulfill purchase commitments and moreover those companies might also reduce capitalized leases to less the burden they have during that period.

To summarize, the robustness test shows that the univariate analysis has a robust result against the inclusion of capitalized leases as a part of the outcome variables with improved

(32)

32

significance of effects of the coefficient of interest on both capital expenditure and working capital. However, further work is needed to verify the robustness of the multivariate tests.

6. CONCLUSION

The widely spread influence of the 2008 financial crisis not only involves financial institutions such as investment banks, but also heavy machineries operating industries such as airlines, in a different way. Having committed to an purchase order of machineries and right afterwards shocked by the 2008 financial crisis would be indeed catastrophic for certain firms. In this thesis, I seek the answer to the research question ‘How does the 2008 financial crisis affect investment commitments of firms that are in the industries of airlines, transportation, oil &gas and semiconductor?’ My data sample consist 43 firms in the above industries and 607 observations. I gain a deeper understanding of how firms’ investment policies react to the crisis by measuring investment commitments in four different means: capital expenditures, rental commitments, working capital and long-term debt growth. Prior literatures show that financial crisis would affect firms’ funding ability and in turn, to some extent force firms to change their investment plans. I analyze the effect of firms having higher than mean purchase commitments being hit by the 2008 crisis on various firms’ investment policies mentioned before using difference-in-differences approach. I found robust evidence to support the univariate tests that having higher than mean purchase commitments when hit by the 2008 crisis would reduce both capital expenditures and working capital, even with the inclusion of capitalized leases in the outcome variables. To avoid potential endogeneity, I run multivariate analyses controlling for long-term debt, firm’s Q, size, cash holding, cash flow and return on assets. There exists an ambiguous effect of having higher than mean purchase commitments when hit by the 2008 crisis on capital expenditure when including all above control variables.

(33)

33

The reason for this ambiguity could either be sample limitation or the irrelevance of the two variables. However, the effect of having higher than mean purchase commitments when hit by the 2008 crisis on rental commitment improves significantly as I control for more variables, indicating there is evidence for firms taking more rentals in such circumstance. Long-term debt growth rate doesn’t provide with sufficient evidence that it would be affected by the dependent variable in the univariate analysis while working capital would be reduced. However, unfortunately when facing control variables, working capital tends to have little correlation with having higher than mean purchase commitments when hit by the 2008 crisis. Finally, this paper also contributes to the current literatures that certain firm characteristics including firm size, cash holding and cash flow would affect firms’ investment policies. For further study on similar topic, one is recommended to enrich the dataset to focus on various financial crisis; and to take into consideration the potential spillover effect the crisis of non-capital expenses such as operating expenses mentioned before.

(34)

(1) (2) (3) (4) (5) (6) Percentiles mean standard

deviation observations Variance min max 10th 25th 50th 75th 90th

Future Commitments in 5 Years/assets 0.1397 0.2093 605 0.0438 0.0000 1.3272 0.0000 0.0225 0.0782 0.1633 0.2991 Capital Expenditures/assets 0.1105 0.0856 602 0.0073 0.0052 0.4506 0.0178 0.0442 0.0952 0.1519 0.2152 Rental Commitments in 5 years/assets 0.2424 0.4787 553 0.2291 0.0004 3.1499 0.0049 0.0137 0.0772 0.2947 0.5200 Working Capital/assets 0.0914 0.1477 602 0.0218 -0.2353 0.5853 -0.0700 -0.0027 0.0661 0.1685 0.2788 Long-term Debt Growth 0.0356 0.4284 519 0.1835 -1.0000 2.0302 -0.3045 -0.1587 -0.0268 0.1444 0.4147 Long-term Debt/assets 0.4152 0.8841 598 0.7816 0.0000 6.6549 0.0083 0.0993 0.2196 0.3598 0.5402 Q 1.4772 0.7824 511 0.6122 0.5926 4.5759 0.8305 0.9843 1.1919 1.6601 2.5777 Cash Holdings/assets 0.1412 0.1213 605 0.0147 0.0006 0.5566 0.0088 0.0394 0.1151 0.2086 0.3069 Cash Flow/assets 0.1576 0.2704 599 0.0731 -0.2232 2.0803 0.0163 0.0647 0.1227 0.1762 0.2323 ROA 5.0694 9.0990 603 82.7924 -27.0273 40.9770 -2.9191 1.3768 4.4595 8.6839 14.7774 Size 8.1610 1.6784 605 2.8170 5.0193 11.0699 5.8788 6.7244 8.0997 9.8019 10.3832

Years 2000-2015, Observations= 607, Firms= 43

Table 1. Summary Statistics

This table looks at summary statistics for the main sample of 607 firm-year observations. In order to be a valid sample item, a firm-year observation must be

disclosed in both COMPUSTAT North American Fundamental Annual and SEC EDGAR 10-k reports (see Appendix Table I for sample construction). All variables are winsorized at the 1st and 99th percentile of their distribution. Most variables are extracted from fiscal year-end. Independent variable is total purchase

commitments up to the next 5 years of rigs(oil& gas companies), ships and other carrier types (transportation companies), wafers (semiconductor companies) and aircrafts (airlines). Dependent variables: capital expenditures are COMPUSTAT’s balance sheet item capx ; working capital is COMPUSTAT’s balance sheet item wcap ; rental commitments for 5 years in total is COMPUSTAT’s balance sheet supplement item mrct; total long-term debt growth uses COMPUSTAT’s balance sheet items dd1+dltt divided by at, then divided by lagged dd1+dltt. The construction of control variables partially follow that by Almeida, Campello, Laranjeira and Weisbenner (2011): firms’ Q is defined as total assets plus market capitalization minus common equity minus deferred taxes and investment tax credit divided by total assets (COMPUSTAT’s (at+prcc*csho-ceq-txditc)/at); firms’ return on asset is defined as net income/loss divided by sales times the ratio of sales to total assets (COMPUSTAT’s ni/sale*sale/at); firms’ cash flow is defined as net income/loss plus depreciation and amortization divided by total assets (COMPUSTAT’s (ni+dp)/at); firms’ size is defined as the log of total assets (COMPUSTAT’s at); cash holding is defined as cash and short-term investment divided by total assets (COMPUSTAT’s che/at); long-term leverage is defined as total long-term debt divided by total assets (COMPUSTAT’s (dd1+dltt)/at).

(35)

mean observations standard

deviation 25th% 75th%

p -value of mean comparison Long-term Debt/assets Treated 0.4785 182 1.0687 0.1042 0.3770

Control 0.3876 416 0.7898 0.0975 0.3549

Q Treated 1.3144 153 0.6080 0.9842 1.4836

Control 1.5469 358 0.8372 0.9967 1.8192

Cash Holdings/assets Treated 0.1464 184 0.1098 0.0601 0.2202

Control 0.1389 421 0.1261 0.0304 0.2056

Cash Flow/assets Treated 0.1374 182 0.2735 0.0463 0.1553

Control 0.1664 417 0.2688 0.0759 0.1801 ROA Treated 3.3601 183 8.6257 0.5976 6.5663 Control 5.8142 420 9.2088 1.8130 9.8326 Size Treated 8.1023 184 1.5303 7.2841 9.3451 Control 8.1866 421 1.7403 6.6805 9.9610 0.002 0.551 0.303

Table 2: Mean Test for Treatment and Control Groups in Firm Characteristics

The sample of 43 firms are seperated into treatment and control group according to their future investment commitment in 5 years with above mean future investment commitments firms in the treatment group and below mean future investment commitments firms in the control group. The construction of control variables is as follows: firms’ Q is defined as total assets plus market capitalization minus common equity minus deferred taxes and investment tax credit divided by total assets

(COMPUSTAT’s (at+prcc*csho-ceq-txditc)/at ); firms’ return on asset is defined as net income/loss divided by sales times the ratio of sales to total assets (COMPUSTAT’s ni/sale*sale/at ); firms’ cash flow is defined as net income/loss plus depreciation and amortization divided by total assets (COMPUSTAT’s (ni+dp)/at ); firms’ size is defined as the log of total assets

(COMPUSTAT’s at ); cash holding is defined as cash and short-term investment divided by total assets (COMPUSTAT’s che/at ); long-term leverage is defined as total long-term debt divided by total assets (COMPUSTAT’s (dd1+dltt)/at ). The test for a difference in the mean value of a firm characteristics across two groups is conducted by the mean comparison test and on the right most side of the table are the p -values of this test.

0.005 0.458 0.231

(36)

(1) (2) (3) (4) (5) (6) (7) (8) Growth rate of long-term debt

2008 Crisis dummy -0.002 0.006 0.0014 0.193*** 0.105*** 0.013 0.03*** 0.009

(0.008) (0.009) (0.008) (0.045) (0.032) (0.015) (0.009) (0.048)

High commitments dummy 0.025*** 0.038*** 0.041*** 0.124 0.122** -0.019 -0.013 -0.064

(0.011) (0.12) (0.01) (0.077) (0.06) (0.016) (0.011) (0.057) Diff-in-diff estimator -0.027* -0.03* -0.023 0.0003 0.122* -0.038* 0.001 0.072 (0.015) (0.016) (0.015) (0.083) (0.07) (0.023) (0.016) (0.076) Long-term Debt/assets -0.007 -0.005 0.31*** -0.034** (0.009) (0.006) (0.11) (0.015) Q 0.005 0.017*** -0.02 0.04*** (0.008) (0.006) (0.03) (0.006) Cash Holdings/assets -0.32*** 0.97*** 0.61*** (0.033) (0.19) (0.044) Cash Flow/assets 0.049*** -0.3*** -0.05** (0.014) (0.12) (0.023) ROA 0.0002 0.0002 0.002 0.004*** (0.001) (0.0004) (0.004) (0.001)

Capital expenditure Rental commitments in 5 years

This table represents the relation between investment policies and highly commited companies being shocked by the 2008 crisis during 2000-2015. Dependent variables are four different measures of firms' investment plans. Column (1) is baseline analysis with capital expenditure as investment measure. Column (2) is multivariate regression on capital expenditures with firms' size, ROA, long-term debt and Q as control variables. Column (3) is multivariate regression on capital expenditures with firms' size, ROA, long-term debt, Q, cash holding and cash flow as control variables. Column (4) is baseline analysis with rental commitments in 5 years as investment measure. Column (5) is multivariate regression on rental commitments in 5 years with firms' size, ROA, long-term debt, Q, cash holding and cash flow as control variables. Column (6) is baseline analysis with working capital as investment measure. Column (7) is

multivariate regression on working capital with firms' size, ROA, long-term debt, Q, cash holding and cash flow as control variables. Column (8) is baseline analysis with growth rate of long-term debt as investment measure. Diff-in-diff estimator is the coefficient of interest in the regression. All variable definition can be found in Section 3. Methodology. Standard errors are in the parentheses and they are heteroskedasticity-robust and clustered by firm.

Table 3: Baseline and Multivariate Regressions with Difference-in-Differences Approach

(37)

Size -0.003 -0.002 -0.044*** -0.022***

(0.002) (0.002) (0.011) (0.002)

Constant 0.108*** 0.118*** 0.133*** 0.294*** 0.43*** 0.096*** 0.13*** 0.042

(0.006) (0.02) (0.02) (0.040) (0.113) (0.011) (0.02) (0.038)

Firm fixed effects Yes Yes Yes Yes Yes Yes Yes Yes

Year fixed effects No No No No No No No No

Psuedo R2 0.01 0.03 0.23 0.05 0.4 0.02 0.61 0.00

Observations 602 500 500 553 461 602 500 519

(38)

(1) (2) (3)

Capital expenditures+capitalized leases Rental commitments in 5 years+capitalized leases Working capital+capitalized leases

-0.004 -0.2*** -0.003 -0.012 (0.05) (0.017) 0.044*** 0.08 -0.016*** (0.017) (0.09) (0.02) -0.054** 0.007 -0.053* (0.02) (0.07) (0.027) 0.14*** 0.32*** 0.14*** (0.008) (0.044) (0.013)

Table 4: Robustness Check with Capitalized Lease and Purchase Commitments as Dependent Variable

This table presents the relation between investment policies including capitalized leases and highly commited companies being shocked by the 2008 crisis during 2000-2015. It uses the total of capitalized lease and different purchase commitments as dependent variable. Independent variables are constructed the same as in univariate analysis with the 2008 crisis dummy equals 1 if the fiscal year is 2008 and 0 otherwise and the high commitments dummy equals 1 if the company is in the above average future purchase investment treatment group and 0 otherwise. Column (1) is with capital expenditures and capitalized leases as investment measure. Column (2) is with rental commitments in 5 years and capitalized leases as investment measure. Column (3) is with working capital and capitalized leases as investment measure. Diff-in-diff estimator is the coefficient of interest in the regression. All variable definition can be found in Section 3. Methodology except for the construction of capitalized leases in section 5. Standard errors are in the parentheses and they are heteroskedasticity-robust and clustered by firm.

* p<0.1; ** p<0.05;***p<0.01 2008 crisis dummy High commitments dummy Diff-in-diff estimator Constant

(39)

39

References

Abadie, A., and Imbens, G., 2002. Simple and Bias-Corrected Matching Estimators for Average Treatment Effects. Technical Working Paper NBER T0283

Almeida, H., Campello, M., Laranjeira B. and Weisbenner. S., 2011. Corporate Debt Maturity and the Real Effects of the 2007 Credit Crisis. Critical Finance Review 1: 3–58

Arslan, O., Florackis, C., Ozkan, A., 2006. The Role of Cash Holdings in Reducing Investment-Cash Flow Sensitivity: Evidence from a Financial Crisis in an Emerging Market. Emerging

Markets Review 7: 320–338

Benmelech, E., and Bergman, N., 2009. Collateral Pricing. Journal of Financial Economics 91.3: 339-360

Campello, M., Graham, J. and Harvey, C., 2010. The Real Effects of Financial Constraints: Evidence from a Financial Crisis. Journal of Financial Economics 97: 470–487

Dehejia, R., and Wahba, S., 2002. Propensity Score-Matching Methods for Nonexperimental Causal Studies. Review of Economics and Statistics 84: 151-161

Duchin, R., O. Ozbas, and B. Sensoy., 2010. Costly External Finance, Corporate Investment, and the Subprime Mortgage Credit Crisis. Journal of Financial Economics 97: 418–435

Erkens, D., Hung, M., Matos, P., 2012. Corporate Governance in the 2007–2008 Financial Crisis: Evidence from Financial Institutions Worldwide. Journal of Corporate Finance 18: 389– 411

Fazzari, S., Hubbard, R., Petersen, B., Blinder, A., and Poterba, J., 1988. Financing Constraints and Corporate Investment. Brookings Papers on Economic Activity 1: 141-206

Fazzari, S. and Petersen, B., 1993. Working Capital and Fixed Investment: New Evidence on Financing Constraints. The RAND Journal of Economics 24. 3: 328-342

(40)

40

Ghalayini, A., Noble, J. and Crowe T., 1997. An Integrated Dynamic Performance Measurement System for Improving Manufacturing Competitiveness. Production Economics 48: 207-225

Gompers, P., 1995. Optimal Investment, Monitoring, and the Staging of Venture Capital. Journal of Finance, 50.5: 1461-1489

Han, S. and Qiu, J., 2007. Corporate precautionary cash holdings. Journal of Corporate

Finance, 13.1: 43-57

Lemmon, M. and Roberts, M., 2010. The Response of Corporate Financing and Investment to Changes in the Supply of Credit. Journal of Financial and Quantitative Analysis 45.3: 555– 587

McConnell, J. and Muscarella C., 1985. Corporate Capital Expenditure Decisions and the Market Value of the Firm. Journal of Financial Economics 14: 399-422

Rauh, J., 2006. Investment and Financing Constraints: Evidence from the Funding of Corporate Pension Plans. Journal of Financial 62.1: 33-71

Richardson, S., 2006. Over-investment of Free Cash Flow. Review of accounting studies, 11.2-3: 159-189

The 2008 financial crisis and firms’ investment commitments

Master Thesis—Business Economics: Finance