Corporate Commercial Commitments and the Real Effects
of the 2008 Financial Crisis
By
Jiyuan Xia
Supervisor: Dr. Tomislav Ladika
September 2014
Master in International Finance 2013-2014 University of Amsterdam
Acknowledgements
I would like to thank my supervisor, Tomislav Ladika, who gave me thorough instruction and supervision during my whole thesis writing period. His guidance was useful and inspiring to help me finish this paper. I would like to also offer my appreciation to my friends, my girlfriend and my manager at Nike for their help during the way. From this paper, I learned not only how to write and think in a scientific way, but also how to communicate with people and face unanticipated situations. In the end, I would also send my regards to the MIF program, providing such a good thesis channel and a great studying program.
Abstract
We investigated a new factor – purchase commitments that could potentially influence industrial firms’ reactions during financial crisis. Many firms in airline, railroad and shipping industry made purchase commitments in advance to secure deliveries and cost. This behavior might become one reason to cause financial constraints to firms during crisis period. We used regression models with difference in difference matching estimator to test whether purchase commitments could cause companies different behaviors during financial crisis. We tested Firm’s Revenues, Cost of Goods Sold, Operating expenses, Capital Expenditures, Cash Reserves, Long-term Debt Issuance, Dividends Payout, Staff expenses and Repurchase/Sale of Equity during crisis time. We found that companies which made more purchase commitment just before crisis repurchased significant less shares during crisis than companies which made less purchase commitments. However, results of other tests are statistically insignificant which did not support our hypothesis. The paper is the first attempt to investigate in this field.
Contents
1. Introduction ... 1 2. Literature review ... 3 3. Empirical Design ... 7 3.1 Research methodology ... 7 3.2 Data collections ... 8 3.3 Independent Variables ... 9 3.4 Dependent Variables ... 14 3.5 Model construction ... 14 4. Results ... 16 4.1 Summary Statistics ... 16 4.2 Regression results ... 16 4.3 Extensions ... 22 5. Concluding Remarks ... 22 References ... 25 Appendix ... 271.
Introduction
The research topic in this paper is to investigate how big commercial commitments in some specific industries would affect firms’ behaviors in the financial crisis. It was all known that most firms were hit by the finance contraction and could not get new finance from banks in 2009. However, due to big commitments which firms from airline industry or railroad industry did in previous years, firms still had to follow the commitments schedule to make the purchase in 2009. Commercial commitments would clearly be one of the reasons to cause financial constraints problem when unexpected credit shock hit financial system. Firms are not free to cancel or postpone purchase commitment because of contractual obligations in the agreement with suppliers. If firms cannot cancel the orders, where do they get the money to carry on such a big investment? Do they cut other expenses such as R&D, wages or marketing expense? Or do they cut other investments such as technology investments or logistic investments? How do firms respond to the crisis when financing is hard to renew in the capital market? Does purchase commitment cause firms more vulnerable to fall in financial constraint situation?
Railroad, airline and shipping companies are very common to make purchase commitments in order to schedule a locomotive, an airplane or a ship for delivery. Indeed, investment schedule for these companies are planned long-term before and it is not easy to alter investment amount for the nearby period. This might not be consistent with corporate finance textbook where capital expenditure is treated as an annual decision.
However, when they make these commitments, do they really consider what the potential consequence is when they do not have enough capital to finance the contract? Do they really consider the situation when they do not have enough capital to finance the contract? Or they already build up a cash buffer to avoid the contract to become a burden particularly during a financial contraction?
In this paper, we test whether commitments to contractual obligations beforehand might influence firms’ performance and liquidity during financial crisis. We made our target as railroad companies that were listed in US because they committed significant amount of cash to maintain and update their fleet such as locomotives, railcars and ties. Standard OLS regression model with a difference-in-differences matching estimator is used to test hypothesis. Difference-in-difference method is widely used and well explained in other papers such as Abadie et al. (2006) and Almeida et al. (2011).
Then we divided railroad companies with two characters. “Treated” group include companies whose purchase commitments account for more compared to their new capital expenditures just prior to the crisis (more than 29% of capital expenditures in 2008). “Crisis” represents year 2009 when the financial shock spread from banking sector to industrial sector. “Treated*Crisis” is the interaction between the two which represents how “Treated” companies behaved in 2009 compared to others. In the last we include these three dummies and Q (Market to Book ratio), Logasset (Scale) and Cash Holding (Cash Level) into regression model.
We consider a number of dependent variables such as Revenues, Cost of Goods Sold, Operating expenses, Capital Expenditures, Cash Reserves, Long-term Debt Issuance, Dividends Payout, Staff expenses and Repurchase/Sale of Equity into our regression. We found that companies which made more purchase commitment just before crisis repurchased significant less shares during crisis than companies which made less purchase commitments did. However, results of other tests are statistically insignificant which did not support our hypothesis.
The paper will be guided as follows. Section 2 goes through all the researches that studied the effect of firms’ behaviors during crisis time and raises the hypothesis. Section 3 shows how data is selected and how research model is constructed. Section 4 is the summary of statistical results. Section 5 includes remarks, limitations and conclusions.
2. Literature review
One hot topic in corporate finance research is financial constraints. There is a lot of evidence that financial constraints matter for firms' investment. For example, firms that had to refinance their debt right after 2008 financial crisis had difficulty getting new funds, and thus had difficulty to survive and had to give up on attractive investment opportunities. The period during 2008 to 2009 when financial crisis happened was a perfect window to observe behaviors of many financial constraint firms together.
There are many researches investigating how firms behave in the financial crisis to survive. There are basically three types of influences generally. Most of companies had the pressure with liquidity management. First, companies tend to cut or cancel extra expenses (R&D, Employment and Marketing) in several years where they had rights to control during and after crisis time. Second, companies tried to seek external finance from banks or capital market or even private investors to finance their operations. Third, companies tended to sell assets or sell their equity stakes to mitigate short-term liquidity problem.
Campello (2009) interviewed 1,050 CFOs all around the world and indicates that firms which are in financial constraints during that period tend to act deeper to cut the technology investment, wages and capital investment. They also sold a lot of assets to finance their operations. 86% of CFOs from constrained firms claimed that they bypassed attractive projects due to the inability to borrow money during the crisis. And half of all the CFOs say they would cancel or postpone their planned investment in the future years. The paper also found constrained firms also planned to fire employees by 11%, cut technology expenses by 22%, reduce capital expenditures by 9%, marketing expenses 33% and dividend payout(by 14%).
Duchin (2009) studied that firms which have low cash reserve or high portion of short-term loan are tend to cut investment greatly. He also suggested an important precaution motivation to reserve excess cash to prevent financial constraints when
such an unexpected shock hit the whole economy system.
Lemmon (2010) examined how shocks to the supply of finance can influence corporate financing and investment behavior during 1989. He reveals that when there is a shortage of supply of finance from outside shock, firms would reduce net investment and also reduce net debt issuances. But in spite of the sharp change in the policy, firms leverage ratios remain relatively constant.
In this background, this paper mainly focuses on the influence of large firms’ commercial commitments during that crisis period and pin down the consequence for these large commercial commitments during the credit contraction. And we will carefully examine whether contractual obligation is a reason to cause firm financial constraints during crisis time.
Almeida et al. (2011) identified firms whose long-term debt were largely maturing right at the end of 2007 cut their investment/capital ratio by 2.5% (quarter basis) more than firms whose long-term debt were maturing after 2008. The drop was a significant number which represented one third of the firm total investment levels. And they emphasized the importance of setting the right long-term debt maturity for the financial policy. They also consider several reality checks and robust tests for this paper. Duchin (2009) used placebo crises for the same test approach in previous 4 years to make sure they did not find similar results. Almeida et al. (2011) also used several of falsification and placebo tests to confirm results were not confounded with others. Almeida et al. (2011) found that besides cutting investment(12% companies did so), firms also reduced share repurchases(10%), reduced cash holdings(41%), reduced inventories(7%) and reduced cash dividends(1%).
To talk about our special observation, purchase commitment, sometimes written as a contractual obligation, it is a commitment that the company makes to purchase goods, machinery, properties or services from supplier. The reason to make commitments is summarized as locking in a fixed and competitive price, making sure timely delivery, reducing cost and building long-term relationship with suppliers.
Purchase commitments are heavily made by specific industries such as airlines, railway companies and shipping companies. These companies are required to order airplanes, locomotives and cargo ships in advance because these products have a long building cycle and are only provided by a few suppliers.
Purchase commitments are disclosed by listed companies as a note of financial items and Supplementary Data, mostly under Commitment and Contingencies. For example, CSX mentioned the following in 2002 annual report. ‘CSX has a commitment under a long-term maintenance program…this program assures CSX access to efficient, high-quality locomotive maintenance services at settled price levels.’ Through this program, companies can retain timely and professional delivery or services without worrying the cost. However, firms also stated in the annual report that if they violate the contract, there might be a penalty occurred because of this. Firms have to carefully evaluate advantages and disadvantages to sign these types of commitments.
Industrial firms of airline industry, railroad industry and shipping industry are typical firms that make big commercial commitments such as purchasing airplanes, locomotives or ships. Some firms have to schedule the purchasing plan beforehand because it takes couples of years to build a new plane or locomotive. In this paper, we took railroad industry as our sample industry and dig deeper to commercial commitments’ influence during crisis time. Interestingly, some railroad firms would make big purchase commitments out of total capital expenditures in advance, but some firms would make a small portion of commitments of total which provides them a lot of flexibility. The advantage to make a large crunch of commitments is that railroad firms can strategically form a business partnership relationship with their suppliers for a long period and thus reduce the cost of purchase. In the paper we would examine whether there is significant different effects to firms with and without (or little) committed investments.
In this paper, we systematically examine whether purchase commitments bring negative effects to firms during financial crisis period in 2009 and how these
companies who made purchase commitments deal with the crisis. Based on the background information, when railroad firms in the “treated” group experienced unexpected shock and could not refinance the business in 2009, they would cut other types of capital expenditures since purchase commitments are obliged to comply.
To construct “treated” group, we measured purchase commitment percentage by using commitment amount in 2008 divided by capital expenditure in 2008. This will help us find which companies made more purchase commitment than other just prior to the crisis. And we set our crisis time as 2009 because it took some time to hit railroad industry from banking industry. With the background above, we have the following hypothesis to make.
Hypothesis 1: Railroad firms with big purchase commitments would find a hard time to refinance itself from banking and capital market in 2009 than firms with low portion purchase commitments.
Firms were having a hard time to go through financial crisis in 2009. And “treated” companies might potentially reduce their COGS (Cost of Goods Sold) and operating expenses, reduce capital expenditures, cut dividend payout pay dividends than firms in “control” group in 2009. And if cutting capital expenditures cannot save enough cash to finance the business in 2009, railroad firms with high portion purchase commitments might tend to cut more wages and fire more workers than firms which do not have to maintain such high purchase levels in 2009.
Hypothesis 2: Railroad firms with big purchase commitments would reduce COGS and increase operating efficiency, reduce dividends payout, sack off more employees and cut more salaries than firms with low portion purchase commitments during 2009 in the financial crisis.
3. Empirical Design
3.1 Research methodology
We will use standard OLS regression to test the hypothesis. We will differentiate firms into two groups-“treated” and “control” group by the weight of commitments amounts (see Section 4.3). We basically select companies that committed most of their capital expenditure as purchase commitments as our “treated” group (more than 30% of new capital expenditure). We then take comparable firms in these two groups with controlling characters size, market to book ratio and cash reserve level into the regression model. Because it has some lags between the start of crisis, spreading between banks and the hit to railroad industry, we would say railroad firms were hit by credit crisis in 2009. Therefore we also make our crisis testing point as 2009.
One would say it is hard to tell whether it is a general reaction to cut expenses during the crisis time for all firms, therefore we introduce several dummy variables to divide all the data into subgroups. The regression model takes observable firm characters and unobservable idiosyncratic factors into consideration. The method of Matching Estimators is largely used in academic world. Abadie et al. (2006) developed new methods for analyzing the large sample size and it is easy to carry out the approach with the software Stata, Matlab and R.
After constructing the model, we will pay special attention on the coefficients and p-value of the three dummy variables. Dummy “Treated” will represent the difference between companies which made big purchase commitments and companies which did not make significant purchase commitments in the whole period from 2000 to 2012. “Crisis” will examine what is the response for all firms at 2009. And “Treated*Crisis” meant what was the difference between companies which made big purchase commitments and companies which did not make significant purchase commitments in the whole period in 2009.
3.2 Data collections
To make sure we work with complete data, we chose all railroad companies for the period 2000 to 2012 from SEC filings1.from US railroad market to be our research
sample. There are nice Railway firms in the sample: BURLINGTON NORTHERN SANTA FE, CSX CORP, KANSAS CITY SOUTHERN, NORFOLK SOUTHERN CORPORATION, UNION PACIFIC CORPORATION, PROVIDENCE AND WORCESTER RR, RAILAMERICA INC, GENESEE & WYOMING INC, and CANADIAN NATIONAL RAILWAY. These films are all Class I Railway Companies which listed in US stock market. They all made proper and comprehensive disclosures annually in SEC website and subject to tight regulations. They had a reliable financial disclosure mechanism and usually promptly disclosed. Importantly, a detailed committed purchase plan would always be disclosed in the annual report because of requested section format in 10-K called “Schedule of Contractual Obligations and Commercial Commitments”. We examined 10-K from SEC filings especially with the item “Commitments and Contingencies”. The purchase commitment schedule is only available in the annual report so we collect the information manually. 5 companies out of 9 in Railway industry made a disclosure of their schemes with clear dollar amounts. In conclusion, we collect 9 companies each with 12 years’ purchase commitment schedule.
The starting point for our dependent variables comes from COMPUSTAT’s North America Annual. We first screen out railway companies (SICs 40) to narrow our target and collect GVKEY for all the companies. Using GVKEY as an identifier, we collect financial items which we are interested such as Capital Expenditures, Cost of Goods Sold, Total Dividends, Long-term Debt Issuance, Employee Numbers and Staff Expenses from the database. The period is from 2000-2012, which includes both internet bubble in 2001 and financial crisis from 2008 to 2009 in order to provide more chances to consider different types of crisis in terms of external shocks to firms. We dropped variables which are missing such as Acquisitions (AQC), R&D (XRD), marketing expenses (XAD). Unlike Almeida et al. (2004) and Frank and Goyal (2003),
the sample is more comparable within Railway Companies compared to cross-sectional companies, we did not drop any companies.
In conclusion, we collect 10 railroad companies’ Commercial Commitments from 117 annual reports and all the other variables (list in Appendix) from COMPUSTAT.
3.3 Independent Variables
After collecting all the purchase commitments from 10-K, we made a summary and converted companies into “treated” and “control” group to account for the company size. For each company and each year, we are using the following four metrics to differentiate companies’ purchase commitment scale by comparing commitments amount with Capital Expenditure or PPENT (Net Amount of Plant, Property and Equipment):
(Committed Purchases in next year)/(Capital Expenditures in next year) (Committed Purchases in next year)/(PPENT in next year)
(Committed Purchases in 3 years + Committed Purchases in 4 years + Committed Purchases in 5 years)/(Capital Expenditures in next year)
(Committed Purchases in 3 years + Committed Purchases in 4 years + Committed Purchases in 5 years)/(PPENT in next year)
The summary of data is as follows.
Table 1. Characteristics of ratio of purchase commitments to Capex and PPENT
Panel A: Average
YEAR IN 1 YEAR/Capex IN 1 YEAR/PPENT IN 3-5 YEAR/Capex IN 3-5 YEAR/PPENT
2001 7.04% 0.51% 16.86% 1.31% 2002 8.71% 0.58% 41.80% 2.98% 2003 15.87% 0.93% 34.96% 2.02% 2004 17.76% 1.13% 39.94% 2.49% 2005 23.21% 1.63% 37.72% 2.84% 2006 31.44% 2.42% 32.83% 2.57% 2007 32.33% 2.88% 31.95% 2.82% 2008 26.85% 2.01% 40.14% 2.94% 2009 26.04% 1.56% 47.72% 2.85% 2010 14.93% 1.27% 19.20% 1.63% 2011 21.62% 1.79% 18.11% 1.56% Panel B: Median
YEAR IN 1 YEAR/Capex IN 1 YEAR/PPENT IN 3-5 YEAR/Capex IN 3-5 YEAR/PPENT
2001 0.00% 0.00% 10.20% 0.65% 2002 6.14% 0.37% 44.85% 3.46% 2003 14.38% 0.74% 40.49% 2.45% 2004 13.25% 0.71% 32.85% 2.19% 2005 22.30% 1.52% 28.98% 2.27% 2006 22.10% 1.98% 22.90% 1.77% 2007 23.73% 1.51% 20.30% 1.47% 2008 24.76% 1.80% 24.09% 1.52% 2009 11.35% 0.73% 19.06% 1.26% 2010 19.55% 1.81% 12.29% 1.09% 2011 19.76% 1.50% 8.32% 0.59%
Panel C: Standard Deviation
YEAR IN 1 YEAR/Capex IN 1 YEAR/PPENT IN 3-5 YEAR/Capex IN 3-5 YEAR/PPENT
2001 12.19% 0.93% 20.99% 1.74% 2002 11.65% 0.78% 35.23% 2.48% 2003 22.01% 1.31% 34.49% 2.00% 2004 23.37% 1.57% 28.21% 1.79% 2005 22.50% 1.60% 37.97% 2.77% 2006 42.16% 3.17% 34.98% 2.66% 2007 51.99% 5.28% 37.72% 3.78% 2008 36.30% 3.05% 46.00% 3.73% 2009 49.58% 2.69% 75.62% 4.11% 2010 15.88% 1.30% 16.08% 1.46% 2011 23.99% 2.12% 16.05% 1.49%
Notes: This table contains three parts: the average, median and standard deviation of the ratio of purchase commitment to Capex and PPENT in fiscal year 2000-2011. The data of purchase commitment is manually collected from SEC 10-k files from 9 railway companies. Capex and PPENT data are found in COMPUSTAT. Capex is the capital expenditure of each company in every fiscal year. PPENT is the net amount of property, plant and equipment of each company in every fiscal year. The ratios are the average amount of purchase commitment divided by Capex or PPENT of each company in every fiscal year. The purchase commitment used in 1-year calculation is the amount committed in each fiscal year and divided by the capital expenditure or net amount of property, plants and equipment in next fiscal year. The purchase commitment used in 3-5 year calculation is the total amount of committed amount in 2,3,4 years after the commitment is made and then divided by the capital expenditure or net amount of property, plants and equipment in next fiscal year.
When comparing to the whole base of PPENT, purchase commitments seems only accounting for 1-2%. But from capital expenditure standpoint, purchase commitment is 20% to 30% of new capital expenditure before crisis, which is significant to affect firms’ financial situation during crisis period. This guarantees that purchase commitment is a significant issue for firms to deal with during crisis and firms behaviors could potentially influenced by the level of purchase commitments.
We also rank companies in 2008 in terms of relative commitment amount. From the above results, we choose CANADIAN NATIONAL RAILWAY CO, CSX CORP and UNION PACIFIC CORP as our treated group because their purchase commitments were significantly higher than others just before the crisis. The other six companies are in our control group. The sample selection procedure make sure that our regression model can test hypothesis and rule out the possibility that a firm start to make commitments after crisis.
Table 2. Ranking of companies’ commitments to capital expenditures or total PPE ratios in fiscal year 2008
2008
Ranking 1 Year/Capex 1 Year/PPENT 3-5 Year/Capex 3-5 Year/Capex 1 CSX CORP CSX CORP CSX CORP CSX CORP
36.56% 2.28% 75.74% 4.72%
2 CN RAILWAY CN RAILWAY UNP UNP
32.60% 2.02% 33.85% 2.16%
3 UNP UNP CN RAILWAY CN RAILWAY
29.91% 1.90% 14.34% 0.89%
4 G&W G&W G&W G&W
19.61% 1.70% 2.20% 0.19%
5 NSC NSC NSC NSC
7.70% 0.44% 0.00% 0.00%
- BNSF BNSF BNSF BNSF
- KCS CORP KCS CORP KCS CORP KCS CORP
- RAILAMERICA RAILAMERICA RAILAMERICA RAILAMERICA
- P&W P&W P&W P&W
Notes: this table presents the detail ratio of sample companies in fiscal year 2008. The data of purchase commitment is manually collected from SEC 10-k files from 9 railway companies. Capex and PPENT data are found in COMPUSTAT. The ratios are ranked from high to low. Companies without purchase commitments are presented using NA and without rankings. The purchase commitment used in 1-year calculation is the amount committed in each fiscal year and divided by the capital expenditure or net amount of property, plants and equipment in next fiscal year. The purchase commitment used in 3-5 year calculation is the total amount of committed amount in 2,3,4 years after the commitment is made and then divided by the capital expenditure or net amount of property, plants and equipment in next fiscal year.
3.4
Dependent Variables
The dependent variables are constructed as follows.
Table 3. Calculation of dependent variables
Variables in model Data from COMPUSTAT
Revenue Log(Revenue Total[REVT])
Capital Expenditures Log(Capital Expenditures [CAPX])
Cash and Short-Term Investments Log(Cash and Short-Term Investments [CHE]) Cost of Goods Sold Log(Cost of Goods Sold [COGS])
Operating Expenses Total Log(Operating Expenses Total [XOPR]) Long-Term Debt Issuance Log(Long-Term Debt Issuance [DLTIS]) Dividends Total Dividends Total [DVT]/Asset Total [AT]
Employees Log(Employees [EMP])
Staff Expense Total Log(Staff Expense Total [XLR])
Purchase of Common and Preferred Stock Log(Purchase of Common and Preferred Stock [PRSTKC])
Sale of Common and Preferred Stock Log(Sale of Common and Preferred Stock [SSTK])
Sale of Property Log(Sale of Property [SPPE])
Notes: this table presents the conversion of variables used in regression models. Variables names are presented in column 1. Calculation methods are presented in column 2. The data used in calculation is found in COMPUSTAT.
3.5
Model construction
We are using a standard OLS regression to process our data. To include our Treated and Crisis variables in the regression, “Treated” “Crisis” and “Treated*Crisis” are created as dummy variables. “Treated” companies include CSX, UNP and CN Railway. “Crisis” represents year 2009 when the financial shock spread from banking sector to
industrial sector. “Treated*Crisis” is the interaction between the two dummy variables which represent how “Treated” companies behave in 2009 compared to others. This structure basically followed Almeida et al. (2011) with independent variable definition for Q (Market to Book ratio), Logasset (Size) and Cash Holdings (Cash Level). Q is defined as the ratio of asset total plus market capitalization minus common equity (AT+MKVALT-CEQ). Logasset is defined as log (AT). Cash Holding is defined as the ratio of cash divided by asset total (CHE/AT). By using these three independent variables, we took out the influence of firm size, cash reserve level and capital market expectation.
The final regression model is:
Dependent variable (Dividends et al.) = β0+β1×Treated+β2×Crisis+β3× (Treated*Crisis)
+β4×Q+β5×Logasset+β6×Cash holding (1)
Here β1 shows whether there is difference of firm’s behaviours between “treated”
and “control” firms during the whole period, i.e. from 2000 to 2012. β2 shows
whether there is difference between crisis and non-crisis time in 2009. β3 shows
4. Results
4.1
Summary Statistics
Table 4. Pre-crisis characteristics of treated, non-treated, and control firms at the end of 2009
Panel A Means for
Variables Treated Control Difference Crisis Non-crisis Difference
Q 1.4087 1.2255 0.1832 1.29 1.291 -0.001
Asset 4.4362 3.3877 1.0485* 3.8498 3.7554 0.0944***
Cash Holding 0.0222 0.028 -0.0058 0.0424 0.0247 0.0177*
Panel B Standard Deviation for
Variables Treated Control Difference Crisis Non-crisis Difference
Q 0.3144 0.3855 -0.0711 0.2614 0.379 -0.1176
Asset 0.1271 0.8974 -0.7703 0.895 0.8831 0.0119
Cash Holding 0.0133 0.3109 -0.2976 0.0336 0.026 0.0076
*, ** and ***indicate statistical significance at the 1%, 5% and 10% (two-tail) test levels, respectively.
Notes: this table compares the characteristics of different companies in fiscal year 2009 and presents the average and standard deviation of the data. For the independent variables, “Treated” and “Crisis” are dummy variables. “Treated” companies include CSX, UNP and CN Railway. “Crisis” represents year 2009 when the financial shock spread from banking sector to industrial sector. “Treated*Crisis” is the interaction between the two dummy variables which represent how “Treated” companies behave in 2009 compared to others. The differences are measured as “Treated minus Control”, “Crisis minus Non-crisis” and “Treated*Control minus Others”.
From the above table, we found that the asset and cash holding level are not the same across group. Therefore simple T test cannot verify our results because we should take out the influence from firms’ scale and cash holding level.
4.2 Regression results
We use standard OLS regression by testing many dependent variables. This table reports results of our baseline regression. And detailed explanations are in the below.
Table 5. Baseline regression results Dependent Variable Independent variables Capital expenditures Cash Balance Revenue Cost of Goods Sold Operating Expense Total Long Term Debt Issuance Dividends Total Employees Staff Expenses Total Purchase of Common and Preferred Sale of Common and Preferred Treated -0.1649 0.1929 .2599*** .2511*** 0.2465 .3711* .6944* .2424*** .5633** .4991*** -0.1282 (0.0309) (0.115) (0.1426) (0.1437) (0.139) (0.1198) (-.2070) (0.1375) (0.1966) (0.2843) (0.2024) Crisis -0.0648 0.3075 -0.0188 -0.0288 -0.0159 0.0042 0.1606 -0.0372 0.2602 -0.6228 0.1169 (0.0571) (0.2095) (0.2644) (0.2665) (0.2578) (0.189) (0.3222) (0.255) (0.3645) (0.5272) (0.3752) Treated*Crisis 0.0485 -0.0509 -0.05565 -0.0598 -0.0805 -0.1889 -0.1386 -0.0169 -0.3547 -1.2993*** -0.0075 (0.097) (0.3617) (0.4494) (0.4529) (0.438) (0.2915) (0.4885) (0.4333) (0.6194) (0.8959) (0.6377) Q 0.0486 .3589* .5074* .4943* 0.4838 0.0457 1.0317* .4398* .6588* .8115** -0.1993 (0.0365) (0.1319) (0.1686) (0.1699) (0.1643) (0.1412) (0.2799) (0.1625) (0.2323) (0.3361) (0.2392) Asset 1.0131* .7833* .5809* .5700* 0.5453 0.1444 .3719* .4848* .4502* .4445* .9270* (0.018) (0.0659) (0.0837) (0.0844) (0.0816) (0.093) (0.1181) (0.0807) (0.1154) (0.1669) (0.1189) Cash Holding 0.241 NA 0.5578 0.2266 0.1761 1.7359 3.0239 -0.4517 0.5229 -2.3029 6.1239 (0.467) NA (2.1633) (2.1802) (2.1086) (2.5498) (3.2041) (2.0859) (2.9818) (4.3128) (3.0696) R2 0.9797 0.7408 0.5723 0.5581 0.5559 0.2152 0.4961 0.5136 0.4336 0.301 0.4362
*, ** and ***indicate statistical significance at the 1%, 5% and 10% (two-tail) test levels, respectively.
Notes: This table reports results of our baseline regression. Coefficient and standard error (in blanket) are reported in the table. For the independent variables, “Treated” “Crisis” and “Treated*Crisis” are dummy variables. “Treated” companies include CSX, UNP and CN Railway. “Crisis” represents year 2009 when the financial shock spread from banking sector to industrial sector. “Treated*Crisis” is the interaction between the two dummy variables which represent how “Treated” companies behave in 2009 compared to others. The method is standard OLS regression. The interaction variable” Treated* Crisis” is only significant when testing for repurchase of common stock which means “Treated” companies reduced repurchase significantly compared to “control” companies.
A. Capital expenditure
Companies in “treated” group are inclined to spend less than companies in “control” group on capital expenditure when taking out scale factor. And the financial crisis causes a negative effect on capital expenditure. However, it seems that companies in “treated” group spent more than control group during financial crisis period. This is not consistent with hypothesis. One potential explanation is that companies in “treated” group still had to fulfil their purchase commitments, whereas those in “control” group were more flexible to cut their capital expenditures and were more independent with investment policies. The following analysis could further explain the reason.
Table 6. Regression results of Capex taken out commitments part
Dependent Variable
Independent Variables
R2
Treated Crisis Treated*Crisis Q Asset
Cash Holding Capital expenditures# .1411 -.0744 .0854 -.5877* 1.1312* -2.3934 0.8916 (.1043) (1.8913) (.2886) (.1441) (.0612) (1.5434)
Notes: This table reports results of our baseline regression. Coefficients and standard errors (in blanket) are reported in the table. For the independent variables, “Treated” “Crisis” and “Treated*Crisis” are dummy variables. “Treated” companies include CSX, UNP and CN Railway. “Crisis” represents year 2009 when the financial shock spread from banking sector to industrial sector. “Treated*Crisis” is the interaction between the two dummy variables which represent how “Treated” companies behave in 2009 compared to others. The method is standard OLS regression and *, ** and *** represent 1%, 5% and 10% confidence level.
The result is not significant and does not match our hypothesis. It might because companies in railway industry have to make consecutive capital investment to maintain and update their fleets and rails. And it is quite capital intensive therefore “committed” company keeps the investment level high.
B. Cash balance
Companies in “treated” group are potentially inclined to hold more cash than companies in “control” group. One interesting observation is that the crisis did not deteriorate companies’ cash position as we can see the coefficient is positive. One potential explanation is that companies were more cautious to invest new projects during the financial crisis period and they hold the cash as the precaution motivation. However, it seems that cash positions in “treated” companies also were hit more by financial crisis. One potential explanation is that “treated” companies had to deploy cash to purchase commitments according to the results above and purchase commitments were indeed a burden to “treated” companies’ liquidity. However, the coefficient is not statistically significant and the test result cannot support our hypothesis.
C. Sales/ COGS/Operating Expenses
a) Sales
Sales number is an important indicator of company’s performance during financial crisis. From the regression results, companies in “treated” group had a relatively large sales base than those in “control” group. One plausible explanation that why these companies made large purchase commitments is that they had such a large fleet to maintain and renew and they should plan the investment schedule in advance. As we can anticipate, all firms’ revenue was hit by financial crisis at 2009. And companies in “treated” group were hit more than others. However, the coefficient is not statistically significant and the test result cannot support our hypothesis.
b) Operating Expenses & COGS
Operating expenses is another important indicator of company’s performance during financial crisis. We saw the same tendency as sales from the test. Financial crisis hit firms’ operating activities and sales and cost were both declining during the financial
crisis. The interesting finding is that companies in “treated” group had a large operating expenses base, but they managed to control operating expenses more effectively than others. That is also the indication that treated companies had to cut cost and save money for something else, potentially obliged capital expenditure--purchase commitments. Cost of Goods Sold (COGS) had the same pattern as Operating expenses from the test results. However, the coefficient is not statistically significant and the test result cannot support our hypothesis.
D.
Long-term debt issuance
Companies in “treated” group were showing they had a strong ability to refinance from banks and debt market. But during 2009 they seemed to struggle to get new finance. One potential explanation is that bond holders or banks (whose credit positions were already tight) did not want to provide money to these old unattractive committed contracts. However, the coefficient is not statistically significant and the test result cannot support our hypothesis.
E. Dividends
“Treated” companies paid more than “control” companies in the period that we investigated. It could be a reflection “treated” companies had a relatively strong financial performance (which also leads to big purchase commitments). The interesting finding here is that one would assume companies should cut their dividend payout when they were experiencing financial crisis, but actually only 1 out of 9 companies cut the dividend payout in 2009. It seems that although companies were desperate to keep cash in 2009, they would like to show to the outsider investors they were performing well and did not want their share prices to drop down. Although “treated” group during 2009 seems only to increase dividend a smaller part than “control” companies. However, the coefficient is not statistically significant and the test result cannot support our hypothesis.
F. Employee numbers/Staff expenses
There were already several indications that firms fired employees during financial crisis. In CSX 2009 Annual Report, it stated that they would recruit previous employees back in the future. That is also to say that we fired many employees during crisis time to cut costs. In general, “treated” companies had more employees than “control” companies. Both types of companies seemed to fire workers and employees during crisis period. But it seems that “treated” companies cut more severe than “control” companies. As many of annual reports stated, they took locomotives and railcars out of services during 2009 which also meant employees such as drivers and maintenance workers were sacked off as well. “Treated” companies even fired more workers during crisis time than “control” companies. During crisis, companies seemed to spend more on staff expense which is unexpected. One plausible explanation is that companies had to pay compensation to fire employees. Moreover, “treated” companies seems pay high salaries to employees from the results. But they pay less during financial crisis than others. However, the coefficient is not statistically significant and the test result cannot support our hypothesis.
G. Repurchase / Sale of common shares
As we assumed, “treated” companies repurchased less equity in 2009 compared to “control” companies. At the same time, they found it hard to raise new capital through capital market during crisis time. This is the only significant coefficient (p=0.14) result that we find in the test. Every “Treated” firm cut its repurchase scheme to zero during 2009. It was potentially because repurchasing stock is the easiest and most flexible money expenditure compared to other items. “Treated” firms can save money at least somewhere to fulfil their contractual obligations.
4.3 Extensions
We constructed the same test for 2001 Internet Bubble as well. And the result is not sufficient to prove any hypothesis. One plausible explanation is Internet Bubble is not the same type of crisis and the influence to the railroad and economy is minimum to be tested. We also found that firms who made bigger purchase commitments prior to financial crisis continue to act in the same way. The behaviour means that companies still enjoy the benefit from purchase commitments and clearly the benefit outweigh the possibility that purchase commitments could potentially lead to financial constraints in the crisis time.
5. Concluding Remarks
In this paper, we test whether commitments to contractual obligations beforehand might influence firms’ performance and liquidity during financial crisis. We made our target as railroad companies that were listed in US because they committed significant amount of cash to maintain and update their fleet such as locomotives, railcars and ties. We divided railroad companies into “treated” and “control” group by their commitment level to total PPE (Property, Plant and Machinery). After that we use three dummy variables to differentiate “treated” factor, “crisis” factor and the interaction factor. In the last we include these three dummies and Q (Market to Book ratio), Logasset (Scale) and Cash Holding (Cash Level) into regression model.
We tested Firm’s Revenues, Cost of Goods Sold, Operating expenses, Capital Expenditures, Cash Reserves, Long-term Debt Issuance, Dividends Payout, Staff expenses and Repurchase/Sale of Equity during crisis time. We found that companies which made more purchase commitment just before crisis repurchased significant less shares during crisis than companies which made less purchase commitments. However, results of other tests are statistically insignificant which did not support our
hypothesis. The paper is a first attempt to investigate in this field.
This research is limited to the following points. Most of coefficients for our interaction variables are not statistically insignificant. But the sign seems consist in line with our hypothesis. It might because our sample is small or our measurement system has issues. Or it just meant that our hypothesis does not hold at least not for railroad industry. It is a field that we can improve in the future.
We only examine listed railways company that already registered in SEC for the whole period between 2000 and 2012. Other Class 1 Railroad Companies such as Grand Trunk Corporation, Canadian Pacific Railway and Soo Line Railroad are not included in our sample1. It may create selection bias somehow. In the future, it is
recommendable to include all the railway samples in the paper. Besides, one could also investigate other industries that also make big purchase commitment such as airline industries.
It would be also better to use several other models to test whether firms making big purchase commitments are in financial constraints. Models are already used by researchers to determine financial constraints such as Abadie and Imbens (2002) and Dehejia and Wahba (2002). Because railway companies has a relatively large asset and have a relatively strong financial liquidity, these tests may not be significant.
Other dependents could also make companies suffer more during the crisis time such as the exact maturity of long-term debt. Almeida et al. (2011) found that companies whose long-term debt was matured at 2007 cut more investments than companies whose not. We can develop more robust tests to verify the results. In general, there could more factors other than purchase commitments that influence the financial situation of “treated” companies.
Another potential research aspect is to investigate railroad companies case by case. The most important factor that firms committed to contractual obligations beforehand is that firms can buy assets and receive consultancy service at a lower price. Clearly, benefit of keep purchases commitments outweighs the possibility that
firms maybe have liquidity issue because of this. One could deeply investigate what is average price per locomotive and maintenance fee, but this does need detailed and specific information which are not easy to find.
Our results contribute to the study of financial crisis and railroad firms’ behavior. No paper has looked at purchase commitment particularly before. It would be an interesting topic to look at how these contractual obligations are signed and what is the really benefit from them for railroads, airlines and shipping companies in more details.
References
Abadie, A. and G. Imbens. 2006. “Large Sample Properties of Matching Estimators for Average Treatment Effects.” Econometrica 74: 235–267.
Abadie, A., D. Drukker, J. Herr, and G. Imbens. 2004. “Implementing Matching Estimators for Average Treatment Effects in Stata.” Stata Journal 4: 290–311.
Acharya, V., T. Philippon, M. Richardson, and N. Roubini. 2009. “The Financial Crisis of 2007-2009: Causes and Remedies.” In Restoring Financial Stability: How To Repair a Failed System, V. Acharya andM. Richardson, eds., Wiley, New Jersey.
Almeida, H., Campello, M., Laranjeira, B., & Weisbenner, S. 2009. Corporate debt maturity and the real effects of the 2007 credit crisis (No. w14990). National Bureau of Economic
Research.
Almeida, H. and T. Philippon. 2007. “The Risk-Adjusted Cost of Financial Distress.” Journal of Finance 62: 2557–2586.
Almeida, H., Campello, M., & Weisbach, M. S. 2004. The cash flow sensitivity of cash. The Journal of Finance, 59(4), 1777-1804
Baños-Caballero, S., García-Teruel, P. J., & Martínez-Solano, P. (2014). Working capital management, corporate performance, and financial constraints. Journal of Business Research, 67(3), 332-338.
Campello, M., E. Giambona, J. Graham, and C. Harvey. 2011. “Liquidity Management and Corporate Investment During a Financial Crisis.” Review of Financial Studies 24: 1944– 1979.
Campello, M., J. Graham, and C. Harvey. 2010. “The Real Effects of Financial Constraints: Evidence from a Financial Crisis.” Journal of Financial Economics 97: 470–487.
Dittmar, Amy, Jan Mahrt-Smith, and Henri Servaes, 2003, International corporate governance and corporate cash holdings, Journal of Financial and Quantitative Analysis 38, 11–134. 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.
Erickson, Timothy, and Toni Whited, 2000, Measurement error and the relationship between investment and Q, Journal of Political Economy 108, 1027–1057.
Fazzari, S. and B. Petersen. 1993. “Working Capital and Fixed Investment: New Evidence on Financing Constrains.” RAND Journal of Economics 24: 328–342.
Guariglia, A. (2008). Internal financial constraints, external financial constraints, and investment choice: evidence from a panel of UK firms. Journal of Banking & Finance, 32(9), 1795-1809.
Hahn, J., & Lee, H. (2009). Financial Constraints, Debt Capacity, and the Cross‐section of Stock Returns. The Journal of Finance, 64(2), 891-921.
Lemmon, Michael, and Michael R. Roberts. "The response of corporate financing and investment to changes in the supply of credit." (2010): 555-587.
Lins, K. V., Servaes, H., & Tufano, P. (2010). What drives corporate liquidity? An international survey of cash holdings and lines of credit. Journal of financial economics, 98(1), 160-176.
Rauh, J. D. (2006). Investment and financing constraints: Evidence from the funding of corporate pension plans. The Journal of Finance, 61(1), 33-71.
Appendix
The data types of accounting measures from WRDS COMPUSTAT
Variables Code Definition
Assets Total AT Total assets/liabilities of a company
Capital Expenditures CAPX Funds used for additions to property, plant, and equipment, excluding amounts arising from acquisitions
Cash Dividends CDVC Total amount of cash dividends for common stock Common Equity CEQ Common shareholders' interest in the company Cash and Short-Term
Investments
CHE Cash and all securities readily transferable to cash as listed in the Current Asset section
Cost of Goods Sold COGS All costs directly allocated by the company to production, such as material, labor and overhead Common Shares
Outstanding
CSHO Net number of all common shares outstanding at year-end, excluding treasury shares and scrip Long-Term Debt Due in
One Year
DD1 Total long-term debt falling due within the first year from the balance sheet date, including all long-term bank, finance lease and other forms of debt
Long-Term Debt Issuance
DLTIS This item includes increase in long-term debt issued for or assumed in an acquisition
Long-Term Debt Total DLTT Debt obligations due more than one year from the company's balance sheet date
Depreciation and Amortization
DP Depreciation is actual cost or other basic value of tangible capital assets over their estimated useful life Dividends Total DVT Total amount of dividends, other than stock
dividends, declared on all equity capital of the company, based on the current year's net income
Employees EMP Actual number of people employed by the company and its consolidated subsidiaries
Market Value Total - Fiscal
MKVALT Sum of all issue-level market values, including trading and non-trading issues
Net Income (Loss) NI Fiscal period income or loss reported by a company after subtracting expenses and losses from all revenues and gains
Property, Plant and Equipment Total (Net)
PPENT Cost, less accumulated depreciation, of tangible fixed property used in the production of revenue
Price Close Annual - Fiscal
PRCC_F
Purchase of Common and Preferred Stock
PRSTKC Any use of funds which decreases common and/or preferred stock
Revenue Total REVT Sales/Turnover (Net) plus Operating Revenues Sale of Property SPPE Funds received or cash inflows from the sale of
property, plant, and equipment Sale of Common and
Preferred Stock
SSTK Funds received from issuance of common and preferred stock
Staff Expense Total XLR Salaries, wages, pension costs, profit sharing and incentive compensation, payroll taxes and other employee benefits
Operating Expenses Total
XOPR
Pension and Retirement Expense
XPR Pension and retirement expense included as an expense in the Income Statement
