Amsterdam Business School
MSc Business Economics, Finance and Real estate Finance (dual tracks) Master Thesis
Real Estate prices and Firm cash holdings
Name Chuxiang Jiang
Student Number 10711163
Supervisor Vladimir Vladimirov
Completion July, 2015
Abstract
What's the relationship between real estate prices and firm cash holdings? Real estate assets could be pledged, and thus affect firms' cash holdings. Taking 1309 listed firms in U.S. through 1993 to 2014, this paper finds that 1 percent increase in real estate price will lead to 0.0044 to 0.0355 percent increase of firms' cash holdings. The results affirm the significance of real estate assets in corporate finance and offer
suggestions for firm's cash management.
Statement of Originality
This document is written by Student Chuxiang Jiang, 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.
Introduction
Firms do not need to hold any cash in a perfect market as they can always borrow money from the outside market without cost. This is not the case in the real world. Take the channel of transferring assets into cash for example, the selling transaction involves with high cost (Opler et al, 1999). As a result, firms have to balance between the benefit and cost of holding cash. So the aim of this paper is to determine the optimal cash holdings when real estate price varies.
Real estate asset accounts a huge part in total assets. Existed summary shows that 59 percent public firms in U.S. own real estate assets in their balance sheets and these assets take up 19 percent of the whole firms' value in 1993(Chaney, Sraer & Thesmar ,2010). Also, it plays dual roles in corporate finance. On the one hand, it's an element of producing activity. On the other hand, its high tangibility and redeployability make it good collateral for firms (Shleifer & Vishny, 1992).
Collateral could improve firms' financing ability due to the imperfection of contract. From lenders' view, they could receive the collateral and liquid it for payment even borrowers default. Otherwise, they will receive the payment as scheduled. Therefore, collateral guarantees lenders a minimum payment and thus improve their confidence.
Federal Housing Finance Agency publishes quarterly housing price index (HPI) which is a transaction based measure of single-family house prices. According to Appendix 2, the housing price is similar to the economy cycle that includes expansion, prosperous, depression and recovery. U.S. housing price has experienced a complete cycle through 1993 to 2014. The initial ascending trend is reversed near 2006. Then housing market is recovering and the price is increasing at present. The most direct of house price cycle is the changing value of real estate assets. Therefore, the value of real estate assets also experiences cycle.
One direct result of the shifty real estate price is that their collateral value varies, too. When real estate price increases, their corresponding value increases simultaneously. For outside lenders, the increasing value of real estate assets implies that their earnings through liquidation will increase. Consequently, they are willing to lend loans. On the contrary, lenders could receive less from the liquidation of real estate assets when their price declines. Therefore, the collateral value of real estate assets is positively linked with the assets' price.
Since the issuance of new debt bring extra cash for firms, debt is positively correlated with firms' cash holdings. Considering the relationship between debt and collateral, this paper suggests there maybe some relationship between cash holdings and real estate price.
To examine the relationship between real estate price and cash holdings, this paper will look into all the U.S. listed firms through 1993 to 2014. Based on corporate data and real estate price data, real estate assets'
market value could be obtained. Then, cash holdings would be regressed by the value of real estate assets. To avoid the noise from real estate cycle and and land supply elasticity, sub samples will be divided, researched and compared.
The problem of endogeneity may threaten this paper since real estate price is possible to relate with firms' cash holdings. Following Himmelberg, Charles, Chris Mayer, and Todd Sinai (2005), instrument variable will be introduced. Land supply elasticity will interact with long-term interest rate to instrument real estate prices.
This paper contributes to the researches of contract and collateral. Previous literatures have explained this question both theoretically and empirically. Borrowers are encouraged to repay loans in case of losing collateral. While lenders are ensured payment equals to the value of collateral. Therefore, contracts could reduce the probability of default as well as guarantee lenders (Barro, 1976; Kaplan & Zingales, 1997). Even though this paper will not research this relation throughly, it empirically prove it during the robustness check. When real estate price increases, firms are able to increase debt issuing because of higher value of collateral. This paper is an extension of the emerging researches focusing on real estate and corporate financing. Almeida & Campello (2007) adopt the differences-in-differences approaches to test the influence of assets' tangibility. They found that the sensitivity of investment cash flow varies with the tangibility of assets under different financial constrained regimes. In addition, Chaney, Sraer & Thesmar empirically finds that as long as collateral increases by $1, firms will increase investment by $0.06. Based on these literatures, this paper is an extension. Firstly, the existed papers have limited sample and time periods. For example, Almeida & Campello (2007) look into the manufacturing firms from 1985 and 2000, and Chaney, Sraer & Thesmar's paper focus through 1993 to 2007. Secondly, previous researches are mainly focuses on investment. Thirdly, the conclusions are general and small factors have not been researched
This paper pay attention to the U.S. market, and all qualified listed firms are included. Also, the most recent data of 2014 is used so that the researching period is through 1993 to 2014. The extended sample and time period enlarge the observations, and thus improve the representativeness of the final results. This paper mainly focuses on the relationship between cash holdings and real estate. Besides, the impacts from real estate cycle and land supply elasticity are examined.
Last but not least, this paper belongs to the works of cash holdings. Firms have to decide the best cash holdings in the fricative market, and this is vital decision for firms. Taking transaction cost, asymmetric information and agency cost into consideration, the trade-off theory indicates an optimal cash holding could be achieved by trading off costs and benefits (Opler et al, 1999). While the theory of pecking-order and free cash flow suggests that the optimal cash holdings do not exist at all (Hardin III et al, 2009; Jensen, 1986). After the empirical model, this paper will examine the relationship between cash holdings and real estate
price. In addition, other accounting variables will be tested. Thus, the findings could help figure the best cash holdings given real estate prices.
The remainder of this paper is organized as follows. The next section is about literature review. Related literature are introduced and analyzed. Section 2 provides the hypotheses and research designing. Section 3 researches the relationship between real estate prices and firms cash holdings. Data, main results and robustness are presented in this part. The final section summary and conclude the paper.
Ⅰ. Literature Review
As mentioned above, this paper identifies the relationship between real estate price and corporate cash holdings. This relationship has not been researched before. But, illustrative literatures do exist and they will be referred.
A. Collateral and corporate finance
The relationship between collateral and corporate finance has been researched for a long time and many researchers believe collateral could enhance firms' financing ability (Barro 1976; Kaplan & Zingales, 1997). Barro (1976) explains this view by two dimensions. On the one hand, borrowers are simulated to repay the loans, otherwise they will bear the loss of collateral. On the other hand, even if default occurs, the ownership of collateral is transferred to lenders immediately. It means lenders could receive the value of collateral in the worst case. Thus, collateral could not only reduce the possibility of default but also ensure lender's proceeds. And the collateral is optimal as long as it's more valuable to borrowers than lenders. (Lacker, 2001). Also, collateral is thought to play the role of signaling because its ability of distinguishing firms. Only firms of high quality are able to practice this due to the high cost of providing collateral. For those smaller firms, they have to bear higher external financing cost then (Chan, & Kanatas, 1985).
Furthermore, other researchers claim that real economy is possible to be impacted by the the asset market through "collateral channel". Kiyotaki & Moore (1995) theoretically demonstrate that asset prices may fluctuate hugely and persistently due to the small shocks of income or technology. In the dynamic economy they constructed, lenders could only foreclosure secured debt. As for financial constrained firms, assets play the dual role of production and collateral. If a productivity shock happened, the assets' value will decreases. Then, the impaired collateral value would reduce firms' loans, investment and profit. Eventually, a small price decline of assets' value at period t could lead to a persistent and severe damage to the firm in the following periods. This is a simple case for "collateral channel". To make it complex, image the situation when the price shock occurs in the national level, uncountable firms would be harmed because of the reduced collateral value. That's the reason why Benanke & Gertler (1989) claim that the "collateral channel" is the source of economic depression.
However, other counter arguments do exist. After separating the aggregate debt into short-term, long-term and convertible debt, Titman & Wessels (1988) find that firms' debt ratio has no relationship with the collateral value of assets. But the authors explain that the imperfect model may be the reason for the irrelativeness.
It needs to be highlighted that collateral is one kind of asset basically. B. Real estate and corporate finance
Real estate asset is a specific segment of collateralizable assets. From a quantitative point, Chaney, Sraer & Thesmar (2010) shows that 59 percent of listed firms in US hold more or less real estate assets in 1993. And real estate assets take up 19 percent of firms' total market value.
However, real estate assets own some special characteristics. Shleifer & Vishny (1992) notice that less firm-specific assets are more likely to be sold within the same industry. Therefore, higher financing ability relates with the versatility.
Recently, Campello & Giambona(2013) divide tangible assets into their identifiable parts, which contains machinery and equipment, land and building, other assets. Instrument variables are introduced to proxy the liquidity. After doing empirical tests, they find that leverage is huge affected by the redeployability of tangible assets. And this effect is more prominent for financial constraint firms and during economy of tight financing. Besides, the Defense Base Closure and Realignment Act of 1990 released large supply of land and buildings by disposing military bases and supporting facilities. This paper takes this event as a natural experiment of supply shock. It confirms the conclusion capital structure is impacted by tangibility assets' redeployability.
Researchers have noticed the specialty of real estate assets and try to figure its role in corporate finance separately. Chaney, Sraer & Thesmar (2010) take real estate as the only collateral and examine the collateral value shock’s influence on debt structure. Taking a sample of 5584 U.S listed firms and 50858 firm-year observations, they find that investment will increase by $0.06 if the value of firms' real estate assets appreciate by $1. Further work shows that the increase of investment is financed by extra debt. Similar to Campello & Giambona (2013), this paper finds that finance constrained firms is more likely to be influenced by the shock of real estate price.
Cvijanović (2014) extends this work by examining the relationship between real estate price and debt. After merging real estate price with MSA-level land supply elasticity, this paper computes the market value of real estate assets. The paper adopts the instrument variable method and indicates that real estate assets could mitigate financing frictions, including the problem of asset substitution and risk-shifting. The function of real estate assets is due to its collateral value. Quantitatively, firm's leverage will increase by 3 percent in
response to 1 standard deviation increasing price of collateral value. On the basement of previous literatures, this paper studies debt more comprehensively. It also finds that for a given debt level, the liquidation value of collateral could decrease the average cost of debt. And heterogeneous debt differs in their relationship with collateral: Secured debt is more closely related with collateral value.
Similar researches include Lin (2015), he believes that local real estate price variation will affect firms' real estate value. So, he takes the price variation as an exogenous shock to real estate value. And then measures the causal relationship between collateral value and firm's preference of debt. Lin (2015) also focuses on the heterogeneity of debt. He distinguish debt into bank debt and public debt, while Cvijanović (2014) breaks debt into secured debt and unsecured one. And their results are different to some degree. Lin (2015) finds that collateral value encourages bank debt, and this influence works identically no matter how restricted the firm's credit is. On the country, Cvijanović (2014) finds that public and bank debt do not reflect differently to collateral value shock. In addition, the degree of financial constraint will affect the relationship between real estate value and firms' debt.
This paper tries to analyze the reasons for the different results. First of all, these papers may have different definition of real estate asset. Some researchers define it as land and building (Campello & Giambona, 2013). And others include construction in process into real estate assets (Cvijanović ,2014; Chaney, Sraer & Thesmar,2010). The different ranges of object may account for the different results. Secondly, these papers aim at different time period and samples. Campello & Giambona (2013) use samples from 1984 to 1996, which include both active and inactive firms. On the contrary, Cvijanović (2014) exploits strict filtration and many unqualified firms are deleted. The last possibility is about the methods and empirical analysis. Take the instrument variables for example, even though the majority of work adopt this method, they play different functions. Some are used to measure liquidity, and some are designed to control the whole price. Also, the identification strategy may affect the final conclusion.
C. Cash holdings determinants
There is no need for firms to hold cash in a frictionless world. Once their operating or investment cash is too low to run the business, they can always get extra money at the financial market at the zero cost. However, there are transaction costs in the real world, so that cash holding management is vital (Drobetz & Grüninger, 2007).
Holding cash has both advantages and disadvantages. On the one hand, holding cash could improve firm's liquidity and flexibility. Firms are able to respond immediately and seize the profitable investment if they own sufficient cash. Besides, if firms own large amount of cash, they will always be able to pay debt in time. Since financial distress has high cost, the improved liquidity could help firms to save cost due to the reduced possibility of financial distress. On the other hand, holding cash may harm firms to some degree. For
example, compared with firms facing financial constraints, firms with large cash may face serious agency problem. The more cash firms hold, the more control their managers have because cash is valuable resource. As managers have different interest from shareholders, they tend to use the extra cash to benefit themselves rather than maximize the value of shareholders. Thus, extra cash holdings could intensity the agency problem (Jensen, 1986). Therefore, it's difficult to determine the optimal cash holdings.
Due to its huge significance, many scholars have researched the determinants of cash holdings. The previous papers include both theories and empirical results. The most widely acknowledged theories include trade-off, free cash flow and pecking-order theories.
The trade-off theory considers transaction costs, asymmetric information and agency costs. It indicates an optimal cash holding could be achieved by trading off costs and benefits (Opler et al, 1999). Firms have to bear cost when they short liquid assets, so that the optimal level is achieved when the marginal benefit of holding assets equals the marginal cost of it. After analyzing the publicly traded firms in U.S. from 1971 to 1994, they find that large firms with higher credit ratings have a tendency to hold lower ratio of cash to total-non-cash assets because they have better access to capital market. Additionally, excess cash do not have a huge impact on shareholder payout, capital expenditure and etc.
On contrary to the optimal cash holding, pecking-order theory believes debt is negatively related with internal funding. External financing has high costs so it would be used after the internal ones. Therefore, firms would prefer using internal cash holdings, and cash holdings are affected by internal factors (Hardin III et al, 2009). Similar to the pecking-order theory, free cash flow theory asserts there is no optimal cash holding. The reason is that nothing is changed if firms increase cash and debt simultaneously (Jensen, 1986).
In addition to these theoretical researches, empirical works could provide lights on econometrical modeling of this paper. Opler et al (1999) figures the determinants of cash holdings by comparison. 87135 U.S. listed firms are separated into four quantiles based on their cash holdings. Then statistic indictors of firms’ characteristics are compared among different quantiles. To testify the effect, cash ratio is calculated as the dependent variable, and panel regression conducted. The results show cash ratio is related with basic firm characteristics, including firm size, leverage, industry, capital expenditures, cash flow. Bates et al. (2009) complements this research by finding investment opportunity is also a factor.
In addition, previous literatures adopt two indicators to measure cash. Opler et al (1999) defines it as the cash divided by assets less cash holdings, while Kim, C.S. et al (1998) just divides cash by assets.
Ⅱ. Hypotheses and Methodology A. Hypotheses
Illustrated and motivated by previous literature, this paper aims to research these hypotheses.
Hypothesis 1 : Real estate value is positively correlated with firms' cash holdings.
Collateral affects corporate finance through many dimensions. On the one hand, collateral could mitigate the imperfection of contract and thus enhance its quality (Lacker, 2001; Barro, 1976; Kaplan & Zingales, 1997). Also, collateral could help reduce the cost of debt and extend the duration of debt (Cvijanović, 2014). Therefore, as a specific asset with collateral value, real estate could affect corporate finance through its shifty price. When real estate price appreciates, its appreciated value enhances firms' collateral and provides firms with more extra financing. The final result is that firms will seize the favorable financing and increase its cash holdings. So this paper suggests that firms' cash holdings have a positive relationship with their real estate value.
Hypothesis 2 : Real estate cycle affects the relationship between collateral and cash holdings .
"Collateral channel" indicates that a trivial shock could be amplified into huge depression. When real assets depreciate, firms could gain less financing through pledging assets. Financing restricted firms are then forced to reduce investment and operation in the following periods. ( Benanke & Gertler ,1989).
This paper chooses the period from 1993 to 2014. According to the house price index from Federal Housing Finance Agency (Appendix 2), this period includes a complete cycle for real estate price. Through 1993 to 2005, the house price tends to increase stably and this is the expansion stage. After the peak around 2006, the house price enters the recession stage with the declining price. The macro economy provides firms with different level of credit constraints. It's assumed that firms will face less financing constraints during real estate boom. And the constraints will grave for depressed time.
Since the researching period implies the variation of financing constrains, this paper will excavate it. The logic is explained as follows: the expansion stage of real estate cycle eases financing because real estate assets could bring more value in the future. Conversely, when real estate assets are experiencing recession, there are more credit limitations for firms. Combined with the theory of "collateral channel", this paper supposes that real estate assets could be pledged for more loans during the expansion stage.
Hypothesis 3：The elasticity of land supply affects the relationship between collateral and cash holdings. Land supply elasticity affects the price of real estate assets. For cities with high elastic lands, they are able to provide extra real estate assets whenever demand outweighs supply. In the long run, the house price for these states will be stable. On the contrary, for cities with inelastic land supply, the increasing demand could not be absorbed by extra construction. The only solution is that price will be driven up by the increasing demand. Therefore, the more inelastic land supply a city has, the more likely its house price will increase.
Normally, the supply elasticity of land is predetermined and people feel it. As the elasticity determines the increasing potential of real estate assets, it may further affect the relationship between collateral and cash holdings. For example, real estate assets locating in low elastic land have higher potential of increase. Thus lenders will expect higher collateral value and offer much cash. So the relationship between collateral and cash holdings may be affected by land supply elasticity.
B. Methodology Data and Resources
To test above hypotheses, this paper researches all U.S. listed firms excluding those operate in finance, insurance, real estate and mining. The time period is from 1993 to 2014, which contains a complete cycle of real estate. After merging with Metropolitan State level house price index, there are 1273 qualified firms and 15295 observations.
To explain the relationship between real estate value and cash holdings, we need four classes of data. The first one is about real estate value. Real estate value is gained by inflating the value of real estate asset in 1993 by real estate price index. Specifically, real estate asset data could be gained and computed through COMPUSTAT. House price index (HPI) is accessible from Federal Housing Finance Agency. The second one is cash holdings. This paper downloads cash, cash and short term investment from COMPUSTAT. The third one is about control variables. To improve the explanatory ability of models, this paper will add control variables of corporate accounting based on the existed papers. For example, leverage ratio, working capital and so on. In addition, interest rate data, which matches the mortgage loan is gained from the World Bank. The summary of data and resources are presented in Table 1.
Basic assumption
It's highlighted that this paper is based on some basic assumptions. To improve its validity, this section will analyze and explain the most important two at the beginning.
The first one is that house price index (HPI) from Federal Housing Finance Agency is used to inflate the real estate assets. This index measures the single-family house price movements and reflects the market of residential properties. Since the real estate assets owned by firms belong to commercial real estate, commercial real estate index would better describe the variation of values.
The most pioneering index for commercial property is NCREIF. This index is collected on private market for investment purpose quarterly. However, it has limitations compared to the residential house price index. On the one hand, this data is available from 1978, which is less than HPI. On the other hand, this index measures house price in the national level so that it cannot be merged with other MSA-level data. In addition, this index is more useful to measure the investment return rather than value appreciation.
Table 1 Data and Resources
Sample U.S. listed firms Period 1993-2014
Variable Resource
Real estate value
real estate assets (building, leases, land and improvement)
COMPUSTAT
House price index Federal Housing Finance Agency
Cash holdings cash
cash and short term investment
COMPUSTAT
Control variable
accounting( asset, leverage etc) COMPUSTAT
Others interest rate Board of Governors of the Federal
Reserve System
Therefore, to ensure the effectiveness of HPI, this paper runs the regression with the national level commercial real estate index. The significance of the interest does not differ a lot with that regressed with HPI. Thus, HPI is a good indicator to measure firms' real estate assets. In sum, even though residential house price index is not the best indicator for commercial buildings, it's a premium one for this paper.
The second one is that firms will only operate real estate assets in the states where they locate headquarters. This is driven by the fact that firms do not report the detail locations for their real estate assets. However, to get the market value of real estate assets, the historical value provided by COMPUSTAT needs to be inflated with local house price index which diverges among different states. To solve this dilemma, this paper assumes firms own real estate assets in the same state of headquarter. And COMPUSTAT offers the state information where firms locate headquarters.
Combined with the past literatures, this paper can prove this assumption is practical. Chaney, Sraer & Thesmar (2010) manually collects the headquarter information from 10K in 1997. Among all the accessible data in SEC, 1815 firms own real estate assets. As 44 percent firms own their headquarter buildings, and headquarters is included as real estate assets. It means that at least 44 percent firms run real estate assets in the same state of their headquarters. By assuming that firms allocate the majority of real estate assets in the same state of headquarters, this paper seems overstate the impact of house price index. To measure the influence, a new dummy indicates whether firms own their headquarter buildings is used to replace real estate value in the regression model. It turns out that the coefficient and significance are close before and after the replacement. As a result, this assumption is acceptable.
Problem of endogeneity
The problem of endogeneity exists because real estate price may alter the real estate value and cash holdings at the same time. Himmelberg, Charles, Chris Mayer, and Todd Sinai (2005) interact land supply and the national interest rate to conduct instrument variable. The chances of real estate boom are less for cities with high elastic land supply. The reasons are as follows. The demand of real estate asset would increase when the interest rate decreases. Then, for cities whose land is lack of elasticity, its housing price is easily to increase. If the land is of high elasticity, the increasing demand of real estate would be transformed into new construction. Therefore, the instrument could measure the whole impact of real estate price and it is independent of firms' real estate
Empirical designing
To describe the relationship between real estate value and cash holdings, this paper will adopt different specifications of model. For firm i, at year t, whose headquarter locates in state s, the basic model is as follows. Cash holding is explained by real estate value.
𝐂𝐚𝐬𝐡 𝐑𝐚𝐭𝐢𝐨𝐢,𝐭𝐬 = 𝛂𝐢+ 𝛃 ∗ 𝐑𝐞𝐚𝐥 𝐄𝐬𝐭𝐚𝐭𝐞 𝐑𝐚𝐭𝐢𝐨𝐢𝐭+ 𝛈 ∗ 𝐇𝐏𝐈𝐭𝐬+𝛄𝐢𝐭∗ 𝐗 + 𝛉𝐭 +𝛆𝐢𝐭 (1) Here, in order to reduce the noise from firms' size, cash and real estate value are scaled by total assets. Cash Ratio is the ratio of cash to total assets. Real Estate Ratio it is for firm i in year t, which is gained from real estate value divides total estate. As for real estate value, it's computed by inflating the real estate value in 1993 with house price index. HPI is the house price index, which measures real estate price growth for state i in year t. X stands for control variables.
Since other accounting variables help to explain firms' cash holdings, they would be adopted as control variables. According to relevant literature, possible control variables include: firm size, leverage, capital expenditure and etc (Opler et al ,1999; Bates et al, 2009). In addition, fixed-effects are included to capture the irrelevant influence. αi measures firm fixed-effect and δt is for time fixed-effect. The combination of two fixed-effects could reduce the problem of endogeneity as well.
The endogeneity problem exists because house price index may affect real estate value and firms' cash holdings simultaneously. The instrument variable and Two Stage Least Squares are adopted to solve this problem. Land supply elasticity and interest rate is multiplied to instrument house price index. As a result, HPI (instrumented) controls the whole impact of real estate prices. Land supply elasticity is available from Saiz (2010) at the MSA-level. Thus, HPI is instrumented in the first stage, and then be used in the second stage.
𝐇𝐏𝐈𝐭𝐬= 𝛌𝐢+ 𝛅 ∗ 𝐋𝐚𝐧𝐝 𝐄𝐥𝐚𝐬𝐭𝐢𝐜𝐢𝐲𝐬∗ 𝐈𝐑𝐢𝐭+ +𝛇𝐢𝐭∗ 𝐗 + Ⅰ𝐭+𝛆𝐢𝐭 (2) 𝐂𝐚𝐬𝐡 𝐑𝐚𝐭𝐢𝐨𝐢,𝐭𝐬 = 𝛂𝐢+ 𝛃 ∗ 𝐑𝐞𝐚𝐥 𝐞𝐬𝐭𝐚𝐭𝐞 𝐑𝐚𝐭𝐢𝐨𝐢𝐭+ 𝛈 ∗ 𝐇𝐏𝐈𝐭𝐬+𝛄𝐢𝐭∗ 𝐗 + 𝛉𝐭 +𝛆𝐢𝐭 (3)
These models are conducted to figure the relationship between cash holdings and real estate assets. Real estate assets are regarded to be preferable collateral which could be used to generate cash. Under this condition, real estate assets are convertible of loans, and it's a weak substitution of cash. So, firms with large value of real estate assets have more cash because they are accessible to loans. The appreciation of real estate assets increase firms' collateral and possible loans, thus firms would hold much amount of cash.
When real estate assets appreciate, real estate value and total assets will increase by the same amount. The ratio of real estate value to total assets will increase mathematically. It's underscored that the appreciating real estate assets will increase real estate ratio in the end, so that this paper may not distinguish them. The coefficientβmeasures the impact of real estate ratio on cash ratio. Based on the assumption that appreciated real estate enables firms to hold more cash, βis assumed to be positive. To measure the influence of time and firm, both firm fixed-effect and time fixed-effect are introduced in the model.
Ⅲ. Real estate prices and firm cash holdings A. Data
This paper chooses sample ranging from 1993 to 2014. The starting point is 1993 because it's the last year that firms report accumulated depreciation of buildings. And this data are necessary for calculating real estate value. Firstly, we gain a sample of 9907 firms by including US listed firms with total assets data in 1993 and . Secondly, special industries are excluded according to the SIC code, including finance, insurance, real estate and mining. 6645firms remain after that. Then, this dataset is merged with U.S. MSA-level residential land prices. Firms with all required data for at least three years are kept, so the final sample has 1309 firms and 15043 observations.
Real estate value
To measure firms' real estate value, this paper collects real estate assets data in 1993. Then variation across MSA and time are accounted to obtain real estate value.
This paper defines real estate as buildings, leases, land and improvement. And COMPUSTAT database provides value for these three subtypes: building, leases, land and improvement (FATB, FATC, FATP). Therefore, firms' real estate asset in 1993 is calculated as the sum of these three indicators.
Real estate asset1993s =FATB1993s + FATC1993s + FATP1993s
Here, s: MSA-level state; FATB: Property, Plant, and Equipment Buildings at Cost; FATC: Property, Plant, and Equipment Construction in Progress at Cost; FATP: Property, Plant, and Equipment Land and
Improvements at Cost.
One critical point is that the data provided by COMPUSTAT is historical cost rather than market value. Many factors including inflation and depreciation affect real estate's value and thus make it differs from the historical one. To convert the historical cost to market value, this paper will follow Chaney, Sraer & Thesmar (2010) and adopts accumulated depreciation as well as MSA-level house price index.
Firstly, building's average age (Average age) could be calculated based on the accumulated depreciation of buildings (DPACB). This calculation has two assumptions. One is that under the U.S. GAAP, only buildings are depreciated. The other one is that buildings have a depreciable life of 40 years (Nelson, Porter and Wilde, 2000). When Buildings at Cost (FATB) is divided by accumulated depreciation of buildings, the result indicates the proportion of depreciated asset to the original value. Next, the average age equals the product of depreciated proportion and 40.
Average Age =DPACB/FATB*40
Here, DPACB: accumulated depreciation of buildings; FATB: Property, Plant, and Equipment Buildings at Cost.
Secondly, the constructed year (Constructed year) could be calculated as the difference between 1993 and average age. The constructed year should be limited to 1975 because MSA-level house price index starts from 1975.
Constructed year=1993-Average age
Thirdly, real estate market value (REV) in 1993 could be computed by inflating the asset price with MSA-level house price index. Federal Housing Finance Agency provides quarterly all-transaction index from 1975. That's the reason this paper limits the above constructed year to 1975. The annual house price index (HPI) is calculated as the mean of the quarter data. The specific computation is like this:
REV93i,s = Real estate asset93i,s ∗ HPI93s /HPICyears Here, i: individual firm; s: state; t:year.
Finally, in order to calculate real estate value in the following years, the market value in 1993 is inflated with the growth of MSA-level house price index of year t. The logic is the same as the third step.
REVti,s = REV93i,s∗ HPIts/HPI93s
To control the difference between states, state name is matched between real estate value and house price index. Due to the fact that COMPUSTAT only provides the location of firms' headquarter, this paper assumes that firms would only hold real estate assets in the city where their headquarters locate. Therefore, the real estate assets are assumed to be located in firm's headquarters.
Table 2 shows summary statistics. Panel A reflects the whole 15043 observations with 1309 firms. In panel B, the initial data in 1993 is presented. There are 6645 observations because the initial data doesn't experience much selection. If a company holds real estate asset in 1993, the dummy variable RE Owner is 1. Otherwise, it's zero.
This table shows that real estate assets take up 27.72% share of total asset for samples, and 41% firms own real estate assets in 1993. Therefore, real estate assets are important assets for firms.
Cash holding data
This paper is expected to find out how firms react to the real estate price variation by adjusting their cash holdings. In order to solve this question, I use the variable of cash ratio (Cash Ratio), in which measures the relationship between cash holdings and total asset. For the specific indicator, existing literature provides the simple division, CaRatio = cash asset� (Kim, C. S. et al ,1998). This paper will adopt this form.
Accounting data
To support the empirical analysis, this paper will use other accounting variables besides real estate values. Relevant literatures provide some typical variables, including firm size, leverage, investment and so on (Opler et al, 1999; Bates et al, 2009). For leverage, we first define total debt as the sum of total debt in current liabilities and total long-term debt. Then, leverage is computed as the ratio of total debt to total asset. Working capital equals to current asset minus current liabilities. It shows whether a company's short term asset is able to cover its short tern term. This paper calculates working capital ratio (Working capital Ratio) as total assets divided by working capital. Companies use capital expenditure to maintain or expand their physical assets and thus it is related with cash. This paper use Capital expenditure Ratio, which is the ratio of capital expenditure to total assets. The detail definitions of variables are summarized in Appendix 1. Hose price index (HPI)
This paper adopts the real estate prices (HPI) to evaluate the house price variation in U.S.
House price index (HPI) is provided by Federal Housing Finance Agency and is regarded as an accurate and timely indicator of house price trends at different geographic levels. As a broad measure of the movement of single-family house price in the U.S, HPI reviews repeat mortgage transactions on single-family properties whose mortgages have been purchased or securitized by Fannie Mae or Freddie Mac. Thorough measuring the same properties' refinancing or repeat sales, it shows the average price change. This house price index is chosen in this paper because of its advantages. On the first place, it begins in 1975 and has the longest history over all the indexes. That's also the reason why we limit the buildings' constructed year later than 1975. On the second place, HPI is claimed to be a broad index and contains more useful information. On the last place, HPI contains house price figures for the nine Census Bureau divisions and Metropolitan
Statistical Areas. Therefore, the house price figures could be merged with real estate value on the MSA-level. And the MSA-level data is vital for this paper.
Table 2 Summary statistics
Panel A: Firm data
Variable Mean Min Max p25 p75 Obs
Total asset 3092.8600 0.4890 212949.0000 111.7580 1782.6250 15043
Real estate Ratio 0.2772 0.0000 0.9999 0.1004 0.3954 15043
Cash Ratio 0.0727 -0.0058 0.9619 0.0131 0.0970 13687
Lagged cash Ratio 0.0722 -0.0058 0.9619 0.0131 0.0958 13957
Leverage 0.2875 0.0000 11.2811 0.1197 0.3914 15035
Working capital Ratio 0.1969 -11.8072 0.9886 0.0697 0.3298 15004 Capital expenditure Ratio 0.0635 0.0000 0.8153 0.0274 0.0784 14394 House price index 138.3492 89.2563 293.7163 105.8830 164.0600 15025 Land supply elasticity 1.7852 0.7500 5.4500 1.0728 2.4757 15043
Panel B: Initial data (1993)
RE Owner 0.41 0 1 0 1 6645
This table provides summary statistics for important variables in this paper. Panel A： Total asset is gained from COMPUSTAT. Real estate ratio is the market value of real estate assets divides total assets. The market value of real estate value is computed by inflating real estate value in 1993, and the detail progress is presented above. Cash Ratio is defined as the ratio of cash to total assets. Lagged cash ratio is the one year lagged value for cash ratio. Working capital Ratio is the ratio of working capital to total assets. Capital expenditure ratio is computed as the ratio of capital expenditure to total assets. House price index comes from Federal Housing Finance Agency. Land supply elasticity comes from Saiz (2010). Panel B: As this is the initial data rather than the final sample, the observations here experience less filtration and are bigger. RE owner is a dummy variable which equals to 1 if the firm owns real estate asset in 1993. Otherwise, RE owner is 0.
Land supply elasticity
Endogeneity problem may exist because local economy affects cash holdings and real estate value at the same time. Instrument variable (IV) method is introduced to solve this issue. A new exogenous variable, which could measure the impact from real estate price variations is needed.
Charles, Chris Mayer, and Todd Sinai (2005) construct the interaction term of long term interest rate and local land supply elasticity as an exogenous instrument variable. This paper will follow their methods. It will take the "contract rate on 30-year, fixed rate conventional home mortgage commitments" as the long-term
interest rates. This data is available from Board of Governors of the Federal Reserve System. This paper chooses the 30-year interest rate because the majority of mortgage loans have long repayment periods so it matches with the real estate assets. As for land supply elasticity, Saiz (2010) computes land elasticity for 95 MSAs and this paper will use his results. He estimates the amount of development land in U.S. metropolitan areas by generating satellite-generated data on terrain elevation and presence of water bodies, finally obtains the land supply elasticity data observed states.
The interaction would be a good instrument because it not only affects real estate prices but also has little correlation with cash holdings. The reason is as follows. On the one hand, when interest rate decreases, investors would prefer real estate assets than other investments, for example, bank savings and notes. Therefore, firms would increase their real estate assets. Compared with cities whose land is inelasticity, a city of high elastic land is less possible to experience house price increase. Facing increasing demand, the city with available land can enlarge the supply of real estate property by new construction and stabilize the real estate price in the long term. Conversely, for cities whose lands are lack of elasticity, the increasing demand will drive the real estate price up.
Correlation analysis
Table 3 Cross-correlation Table
Cash Ratio Real estate Ratio Working capital Ratio Capital expenditure Ratio Lagged cash Ratio Leverage HPI Cash Ratio 1.0000 Real estate Ratio 0.0030 1.0000 Working capital Ratio 0.0072 -0.7099 1.0000 Capital expenditure Ratio -0.0087 0.4364 -0.1105 1.0000 Lagged cash Ratio 0.7289 0.0025 0.0128 -0.0046 1.0000 Leverage -0.0055 0.7095 -0.9998 0.1102 -0.0117 1.0000 HPI 0.1506 0.0054 0.0040 0.0003 0.1365 -0.0045 1.0000 17
Table 3 shows the cross-correlation between variables. Cash ratio is positively correlated with real estate ratio, and the correlation is 0.003. This is the simple but direct confirmation for Hypothesis 1: Firms' cash holdings are positively affected by real estate value. Also, cash ratio is heavily determined by its own value in the last year because their correlation is as high as 0.7289.
As for leverage, its correlation with real estate ratio is 0.7095, which means that the two variables impact mutually. Literature suggests that real estate assets are collateral and could encourage loans, the positive correlation is consistent with this supposition (Barro 1976; Kaplan & Zingales, 1997). Furthermore, this paper has affirmed this with empirical model in later section. On the contrary, the negative correlation between leverage and cash ratio is strange. The increasing leverage increases debts and cash at the same time. As cash is less than total assets, when both items increase by the same amount, their ratio (cash/total asset) becomes larger. Leverage should be positively correlated with cash ratio. In addition, when leverage is included as independent variable to explain cash ratio, its coefficient is not significant at all. Thus, leverage would not be a good factor for cash ratio and it will be dropped in the regression analysis.
To decide the proper variables, many dimensions could be considered comprehensively, including the number of correlation, the economical rationality, the possibility of multicollinearity and so on. Finally, this paper chooses working capital ratio, capital expenditure ratio and lagged cash ratio as controls to regress cash ratio.
B. Main Results
As has been mentioned above, there are two models. One uses the Ordinary Least Squares, and the other one is Two Stages Least Squares (IV) to avoid endogeneity. In addition, this paper will test the impact of fixed effects including time level and firm level. The basic model specification is as follows.
OLS model
𝐂𝐚𝐬𝐡 𝐑𝐚𝐭𝐢𝐨𝐢,𝐭𝐬 = 𝛂𝐢+ 𝛃 ∗ 𝐑𝐞𝐚𝐥 𝐄𝐬𝐭𝐚𝐭𝐞 𝐑𝐚𝐭𝐢𝐨𝐢𝐭+ 𝛈 ∗ 𝐇𝐏𝐈𝐭𝐬+𝛄𝐢𝐭∗ 𝐗 + 𝛉𝐭 +𝛆𝐢𝐭 (1) 2SLS model
In this model, land supply elasticity and interest rate is multiplied to instrument house price index. Due to the fact that house price index changes firms' real estate value and firms' cash holdings simultaneously, this instrument variable is expected to alleviate this problem.
𝐇𝐏𝐈𝐭𝐬= 𝛌𝐢+ 𝛅 ∗ 𝐋𝐚𝐧𝐝 𝐄𝐥𝐚𝐬𝐭𝐢𝐜𝐢𝐲𝐤∗ 𝐈𝐑𝐢𝐭+ +𝛇𝐢𝐭∗ 𝐗 + Ⅰ𝐭+𝛆𝐢𝐭 (2)
𝐂𝐚𝐬𝐡 𝐑𝐚𝐭𝐢𝐨𝐢,𝐭𝐬 = 𝛂𝐢+ 𝛃 ∗ 𝐑𝐞𝐚𝐥 𝐞𝐬𝐭𝐚𝐭𝐞 𝐑𝐚𝐭𝐢𝐨𝐢𝐭+ 𝛈 ∗ 𝐇𝐏𝐈𝐭𝐬+𝛄𝐢𝐭∗ 𝐗 + 𝛉𝐭 +𝛆𝐢𝐭 (3)
Table 4 Real estate value and cash holdings
Real estate cycle Land supply elasticity
(1) (2) 2SLS (3) (4) 2000-2005 (5) 2006-2010 (6) high (7) low
Real Estate Ratio 0.0044*
(1.87） 0.0058*** (2.44) 0.0355*** (7.25) 0.0508*** (3.8) -0.0112 (0.54) 0.0374*** (6.19) 0.0352*** (4.66) HPI 0.0001*** (8.28) 0.0003*** (6.63) 0.0002 (0.53) 0.0001 (1.07) 0.0002*** (2.83) 0.0003 (0.49) 0.0005 (0.87)
Control Yes Yes Yes Yes Yes Yes Yes
Firm fixed effect No No Yes Yes Yes Yes Yes
Year fixed effect No No Yes Yes Yes Yes Yes
Observations 13687 13687 13687 3686 2306 6870 7017
Adjusted R2 0.5679 0.5631 0.5764 0.3419 0.4200 0.4958 0.5216
This table shows the relationship between real estate value and cash holdings. The dependent variable is cash ratio, which is computed as the ratio of cash divides total assets. Real estate ratio is defined as the real estate value divided total asset. HPI is the single-family house price index at the MSA-level. Control variables include the lagged cash ratio, capital expenditure ratio, working capital ratio. Appendix 1 shows the detail definition for all variables. Except for column 1 and column 2, all regressions consider year fixed effect and time fixed effect. Column 2 is the only regression adopting 2SLS method. In the 2SLS method, HPI is instrumented by the interaction of land supply elasticity and interest rate. In column 4 and 5, observations are separated based on time periods. Column 4 shows the result for real estate boom that is between 2001 and 2005. And column 5 shows the result for real estate recession during 2006 to 2010. Column 6 and 7 separate firms based on whether it locates in a city of high land elasticity or not. All standard errors are clustered at the firm level. T-statistics are included in the parentheses and significances are marked by stars. (*** Significant at 1 percent level; ** Significant at 5 percent level; * Significant at 10 percent level)
In the above models, the coefficient for real estate value is β, which measures the relationship between real estate value and cash ratio. ηmeasures the relationship between house price index and cash ratio. X stands for control variables andγare their coefficient. θt is time fixed effect. αi is the year fixed effect.
According to equation (1) (2) (3), five different specifications are regressed and their results are presented in Table 4. The dependent variable is cash ratio, which is computed as cash divides total asset. Among all the specifications, column (2) is the only one that adopts the 2SLS method. House price index is instrumented by the interaction term of interest rate and land supply elasticity. Other specifications use the OLS methods
The relationship between real estate value and cash holdings
Column (1) is the simplest form for equation (1). It's regressed with the OLS method, and included no fixed effect. Firms' cash ratio is explained solely by real estate ratio, house price index and control variables. β gauges the different exposure of real estate value on cash holdings. Column (1) shows β equals to 0.0044. This coefficient is positive as well as significant at the 10 percent level. The positive sign means that firms will increase cash holdings when real estate price increases. Or firms with appreciating real estate assets decide to hold more cash than firms with no real estate assets. Economically, when real estate ratio increases by 1 percent, firms are able to increase their cash ratios by 0.0044 percent averagely. The adjusted R square indicates that 56.79% change of cash ratio is captured by these independent variables.
Column (3) adds year fixed effect and firm fixed effect on the basement of column (1). Year fixed effect could measure the influence of time and firm fixed effect considers firm's individual traits. In addition, when the two fixed effects are added together, they can mitigate the problem of endogeneity. Both the absolute value of βand adjusted R square are larger compared with regression without fixed effects. The significance of βis improved since it's now significant at 1 percent level. On average, when real estate ratio rises by 1 percent, firms will increase their cash ratio by 0.0355 percent.
Compared with the model of OLS method, introducing instrument variable makes the value of coefficient larger. In column (2), when cash ratio is explained by real estate ratio and the instrumented house price index, the coefficient for real estate value is significant at 1 percent level. The 0.0058 value means that as long as real estate ratio increase by 1 percent, firms' cash ratio will increase 0.0058 percent accordingly. 2SLS method improves the significance of explanatory variable because of less endogeneity. As for the explanatory ability of the whole model, the adjusted R squareis close to that in column (1), so that the introduction of instrument variable does not improve the explanatory ability of whole model.
In sum, despite the various specifications, real estate ratio is always positively correlated with cash holding ratio. In other words, firms can increase cash holdings when their real estate assets appreciate. This
result is consistent with Hypothesis 1. However, the absolute values of the impaction vary, they range from 0.0044 to 0.0355. The different specification may account for this difference.
The influence of real estate cycle
In the above empirical analysis, the researching period starts from 1993 and ends in 2014. This 22 year period covers a complete cycle for real estate prices (see Appendix 2). House price tends to increase since 1993 and comes to the peak in 2006. Similar to the pattern of economy cycle, real estate price drops after the boom. The volatile real estate price thus offers a natural laboratory for testing the different effect of real estate boom and recession.
According to the house price index from Federal Housing Finance Agency, this paper defines the boom period as 2000-2005, and the recession period as 2006-2010. Therefore, each subsample lasts for 5 years and stands for different stage of real estate cycle.
Column (4) and (5) in Table 4 presents the results for different stages of real estate cycle. The dependent variable is cash ratio. The independent variables include real estate ratio, house price index, working capital ratio, capital expenditure ratio and the lagged cash ratio. Besides, both firm fixed effect and time fixed effect are added.
Column (4) includes data from 2001 to 2005, when the real estate is experiencing stable increase. There are 3686 observations included and the explanatory variables account 34.19% change of cash ratio. The coefficient of real estate ratio is statistically significant at 1 percent level with a t-value of 3.8. Therefore, firms experiencing 1 percent increase of real estate ratio will increase their cash ratio by 0.0508 percent on average.
As for the recession periods, column (5) shows the regression. The result is inconsistent with others. This is the only regression with the negative sign for real estate ratio. In addition, the coefficient of real estate is not significant even at 10 percent level.
These results suggest that the impact of real estate value on cash holdings diverges at the different stages of real estate cycle.
During real estate boom, the appreciating real estate value increase its ratio to total assets, and finally increase firms' cash holdings. However, when real estate properties undergo recession, it could not generate extra cash. This confirms the indication of "collateral channel": firms' financing ability will be damaged when real assets depreciated.
The influence of land supply elasticity
Land supply elasticity is an important determinant of real estate value because it decides the amount of supply. For states with high elastic land, the increasing demand of real estate could be absorbed by the increasing supply. Consequently, real estate value will not appreciate hugely. On the contrary, the total supply of real estate is restricted for states with inelastic states. Facing increasing demand, the only solution is the increasing price because of limited supply.
Column (6) and (7) is designed to test the difference between firms locating in high elastic land and those not. Observations are divided based on the land supply elasticity of location. If their land supply elasticity is less than the median (1.59), they are circled into the low elasticity group. Otherwise, they are believed to locate in high elastic land. OLS method is adopted for both regressions and fixed effects are considered.
Column (6) reflects the regression for firms locates in high elastic land. There are 6870 observations ranging from 1993 to 2004. Statistically, the coefficient of real estate ratio is highly significant because it's significant at 1 percent level. Economically, when real estate ratio increases by 1 percent, firms will increase their cash ratio by 0.0374 percent.
Column (7) is the regression result for firms locates in low elastic land. The total observation number is 7017. The coefficient for real estate ratio is 0.0352, which means firms will increase cash ratio by 0.0352 percent given real estate ratio increases 1 percent. Besides, it's significant at 1 percent level, too.
The regression results for subsamples with different land elasticity are consistent with the regression for whole samples. The coefficients of real estate ratio are highly significant at 1 percent level. Even though the absolute coefficient of real estate ratio is close, it should be noticed that compared with the low elastic group, the group with high elastic supply land has a higher coefficient. In other words, the higher elastic land a firm locates in, the bigger influence of real estate value to cash holdings.
C. Robustness Tests
This section will conduct some robustness tests and provide further explanations. Heterogeneity of cash
In the above analysis, this paper defines cash as the ratio of cash to total assets. This explanatory variable will be replaced to test the robustness. Cash is usually defined as the immediate negotiable medium of exchange or any instruments accepted by banks for deposits and immediate credit to a firm's account (Musmanno, Thomas E., Joseph A. Marrone, and Laura Carey, 1988).Based on this definition, short term investment is a kind of cash naturally. Thus, this section defines cash as the sum of cash and short term investment. The cash & investment ratio is then defied as cash and short term investment divide total assets. Taking the same strategy above, this new cash ratio is regressed following equation (1) with OLS method.
Column (1) shows the regression results of new dependent variables. In this regression, cash and short term investment ratio is explained by control variables and MSA-level house price index. Similar to the result with cash ratio, real estate ratio has a positive influence on cash and short term investment. Once real estate ratio appreciates by 1 percent, the ratio of cash and short term investment will decrease by 0.0351 percent. This relation is similar to that of cash ratio. In addition, it's still significant at 1 percent level.
Real estate index
This paper researches the relationship between firm's real estate value and cash holdings. As for the real estate value, it includes firms' office, retail and industrial buildings, thus its variation would be better captured by a commercial real estate index. However, this paper picks the house price index from Federal Housing Finance Agency because it measures the MSA-level price and contains the broadest information.
NCREIF Property Index measures the quarterly return of commercial real estate properties in the private market only for investment purposes. It begins in 1978 and could cover the time period. This paper constructs instrument variable as the interaction term of land elasticity and interest rate. Since land elasticity needs to be matched at the MSA-level, NCREIF cannot be modeled with instrument variables. As a result, 2SLS is no longer feasible and the problem of endogeneity may exist
This paper chooses the official return index from NCREIF. The annual index is computed as the sum of quarterly data for each year. The methodology used to calculate real estate value is the same, except that HPI is replaced by NCREIF. In other words, real estate value in 1993 could be computed based on depreciation, construction year and NCREIF. Then, real estate value in the following year could be inflated with NCREIF based on the value in 1993.
Column (2) is the regression result with the national level house price index NCREIF. Real estate ratio is statistically significant at 1 percent level. Measured by the commercial index, firms will increase cash ratio by 0.0323 when real estate ratio ascends 1 percent. Contrary to HPI that is measured at the MSA-level and for residential assets, NCREIF is for the commercial properties all over the country. Though the two indexes differ, their regression results are close. Thus, HPI could be used as a proxy for NCREIF and a good measurement for commercial real estate, too.
Besides, the significance of NCREIF is much higher than HPI. HPI is not significant at 10 percent level while NCREIF is highly significant at 1 percent level. Since the result for real estate ratio is similar, the conclusion that HPI is good proxy for NCREIT holds.
After regressing with different explanatory variable and house price index, the relationship between real estate ratio and cash holding remain stable. Therefore, it's safe to conclude that the regression is robust.
When real estate value appreciates, its ratio to total assets will increase consequently. Finally, the increasing real estate ratio will enhance firm's cash holdings.
Table 5 Robustness Tests
(1) (2)
Dependent variable Cash & short term investment ratio
Cash ratio Real Estate Ratio 0.0351***
(6.41) 0.0323*** (5.9) HPI (MSA-level) 0.0002 (0.07) NCREIF (National level) 0.0128*** (5.56)
Control Yes Yes
Firm fixed effect Yes Yes
Year fixed effect Yes Yes
Observations 13655 13655
Adjusted R2 0.6837 0.6837
This table shows the result for robustness tests. Column (1) and column (2) use OLS method and add time and firm fixed effects. Real estate value is inflated with the MSA-level house price index through its value in 1993. Control variables include working capital ratio, capital expenditure ratio and leverage. Appendix 1 shows the detail definition for all variables.
Column (1) tests the influence of changing dependent variables. Column (2) tests the influence of different house price index. For Column (1), the dependent variable cash & short term investment ratio is computed as cash and short term investment (COMPUSTAT) divides total asset. Column (2) is designed to measure the influence of real estate price index. Instead of HPI, this column takes the national level NCREIF index. All standard errors are clustered at the firm level. T-statistics are included in the parentheses and significances are marked by stars.
*** Significant at 1 percent level; ** Significant at 5 percent level; * Significant at 10 percent level The validity of instrument variable
The existence of endogeneity may threaten the regression as real estate price may influence real estate ratio and cash ratio at the same time. Instrument and 2SLS method are thus introduced. This paper multiples land supply elasticity and interest rate to conduct the instrument variable.
To test the validity of instrument variable, the first stage equation (2) is regressed and the result is presented in column (1) of Table 6. The adjusted R square equals to 0.4852 and the coefficient for the instrument is 3.7160. In addition, this coefficient is of high significance at 1 percent level. Thus, the instrument is highly correlated with house price index and it's a qualified instrument.
Debt and collateral
Table 6 Further tests
(1) (2)
Dependent Variable HPI Leverage
Land supply elasticity*interest rate 3.7160*** (25.19)
Real Estate Ratio 0.0867***
(7.12) HPI
(MSA-level)
0.0003*** (3.54)
Control Yes Yes
Firm fixed effect Yes Yes
Year fixed effect Yes Yes
Observations 13711 14975
Adjusted R2 0.4852 0.4077
This table presents results for additional tests. Column (1) is the regression result for the first stage of 2SLS. House price index (HPI) is explained by real estate ratio, working capital ratio, capital expenditure ratio and the lagged one year cash ratio. Appendix 1 shows the detail definition for all variables. In addition, both firm fixed effect and time fixed effect are contained. Column (2) shows the relationship between leverage and real estate ratio. The dependent variable leverage is defined as the ratio of total debt (equals to the sum of total debt in current liabilities and total long-term debt) to total asset. It's worth noticing that control variables in this regression are different. The lagged one year cash ratio is not used as explanatory variable here. T-statistics are included in the parentheses and significances are marked by stars.
*** Significant at 1 percent level; ** Significant at 5 percent level; * Significant at 10 percent level
The above paper has proved that real estate appreciation could increase firm's cash holdings. This section will further analyze the channel through which this relation works. Real estate assets could generate cash through transaction and renting. As one part of fixed assets, firms tend to hold real estate assets rather than sell and profit from the appreciating real estate frequently. Also, for samples in this passage, which
exclude finance, insurance, real estate and mining, renting real estate properties is not their daily operation. On the other hand, the increasing real estate value could enhance firms' collateral and thus financing ability. Therefore, this paper hypothesizes that the firms are able to increase their cash through debt when real estate assets appreciates.
This paper calculates leverage of total debt divides total asset. Total debt is the sum of current debt and long term debt. Then, leverage is regressed with real estate ratio and some control variables. The correlation table shows that leverage is correlated with working capital and capital expenditure, to simplify the process, this regression scales them by total assets and run the regression. To avoid the noise of different time and firms, year fixed effect and time fixed effect are included.
Column (1) shows the relationship between leverage and real estate value. Consistent with previous hypothesis, firms issue larger amount of debt when real estate ratio increases. 1 percent increase in real estate ratio will lead to 0.0867 percent increase on leverage averagely.
Ⅳ. Conclusion
Real estate is a specific asset that could be pledged for cash. Also, the prices of real estate assets vary all the time. Facing such volatile pricing, firms' cash holding will change correspondently. In order to figure out how firms should react to different real estate prices, this paper researches the relationship between real estate prices and firm cash holdings.
This paper starts with computing the market value of real estate assets through inflating its historical value by real estate price index. Then, real estate value is matched with the MSA-level land supply elasticity. OLS method and 2SLS method are adopted to run the regression where firms' cash holdings are explained by real estate value.
This paper makes contribution corporate finance. The empirical results show that real estate prices are positively related with firms' cash holdings. If the ratio of real estate value to total assets increases by 1 percent, firms that hold real estate assets are able to increase their cash ratio by 0.0044 to 0.0355 percent. And real estate prices work on cash holdings through the channel of issuing extra debt. After comparing different stages of real estate price cycle, this positive relationship works during real estate booms while do not exist when real estate assets depreciate. The results of this paper agree and complement the existed literature of real estate assets. Previous literatures have confirmed that real estate assets are related with higher investment and debt issue. This paper shows that by issuing new debt, firms with real estate assets could hold more cash.
The paper shows that real estate assets have no significant relation with cash holdings during real estate recession. Therefore, real estate assets do not play the role of collateral and bring no cash. This is consistent
with the "collateral channel" which claims depression will be exacerbated once assets value depreciates because of the decreased financing ability. This paper does not take a deep study for this theory, but it's worth attention
Even though this paper clarifies the relationship between real estate prices and cash holdings, it still has limitations and could be further improved. Firstly, the model used in this paper may not be the most proper one. It would be better if extra models are tested and considered. Difference-in-difference model may be one option. Secondly, this paper only examines real estate prices' influence on cash holdings. Similar objects are worth investing, too. For example, how would real estate prices affect the operating cash flow or investing cash flow. Thirdly, this paper fails to give an explanation for its divergence from "collateral channel".
In sum, this paper indicates that real estate prices have a positive influence on firm's cash holdings. In addition, firms could not collateral real estate assets for cash when real estate depreciates. Practically, it indicates that firms should purchase real estate assets and pledge them for cash when real estate assets appreciate. Literarily, this paper improves the understanding of real estate assets and corporate finance.
Reference
Almeida, H., & Campello, M. (2007). Financial constraints, asset tangibility, and corporate investment. Review of Financial Studies, 20(5), 1429-1460.
Barro, R. J. (1976). The loan market, collateral, and rates of interest. Journal of money, Credit and banking, 439-456.
Bates, T. W., Kahle, K. M., & Stulz, R. M. (2009). Why do US firms hold so much more cash than they used to?. The Journal of Finance, 64(5), 1985-2021.
Benanke, B. S., & Gertler, M. (1989). Agency Costs, Net Worth, and Business Fluctuation. AER, 1989, 14-31. Chan, Y. S., & Kanatas, G. (1985). Asymmetric valuations and the role of collateral in loan
agreements. Journal of Money, Credit and Banking, 84-95.
Chaney, T., Sraer, D., & Thesmar, D. (2010). The collateral channel: How real estate shocks affect corporate investment (No. w16060). National Bureau of Economic Research.
Campello, M., & Giambona, E. (2013). Real assets and capital structure.Journal of Financial and Quantitative Analysis, 48(05), 1333-1370.
Cvijanović, D. (2014). Real Estate Prices and Firm Capital Structure*.Review of Financial Studies, hhu035. Drobetz, W., & Grüninger, M. C. (2007). Corporate cash holdings:
Evidence from Switzerland. Financial Markets and Portfolio Management, 21(3), 293-324.
Hardin III, W. G., Highfield, M. J., Hill, M. D., & Kelly, G. W. (2009). The determinants of REIT cash holdings. The Journal of Real Estate Finance and Economics, 39(1), 39-57.
Himmelberg, Charles, Chris Mayer, and Todd Sinai. "Assessing high house prices: Bubbles." Fundamentals and Misperceptions (2005).
Jensen, M. C. (1986). Agency cost of free cash flow, corporate finance, and takeovers. Corporate Finance, and Takeovers. American Economic Review,76(2).
Kim, C. S., Mauer, D. C., & Sherman, A. E. (1998). The determinants of corporate liquidity: Theory and evidence. Journal of financial and quantitative analysis, 33(03), 335-359.
Kiyotaki, N., & Moore, J. (1995). Credit cycles (No. w5083). National Bureau of Economic Research.
Lacker, J. M. (2001). Collateralized debt as the optimal contract. Review of Economic Dynamics, 4(4), 842-859. Lin, L. (2015). Collateral and the choice between bank debt and public debt.Management Science.
Musmanno, Thomas E., Joseph A. Marrone, and Laura Carey. "Securities brokerage-cash management system with short term investment proceeds allotted among multiple accounts." U.S. Patent No. 4,774,663. 27 Sep. 1988.
Nelson, T. R., Potter, T., & Wilde, H. H. (2000). Real estate assets on corporate balance sheets. Journal of Corporate Real Estate, 2(1), 29-40.
Opler, T., Pinkowitz, L., Stulz, R., & Williamson, R. (1999). The determinants and implications of corporate cash holdings. Journal of financial economics, 52(1), 3-46.
Scott, J. H. (1977). Bankruptcy, secured debt, and optimal capital structure.Journal of finance, 1-19.
Shleifer, A., & Vishny, R. W. (1992). Liquidation values and debt capacity: A market equilibrium approach. The Journal of Finance, 47(4), 1343-1366.
Saiz, A. (2010). The geographic determinants of housing supply. The Quarterly Journal of Economics, 125(3), 1253-1296.
Titman, S., & Wessels, R. (1988). The determinants of capital structure choice. The Journal of finance, 43(1), 1-19
Appendix 1
Table 7 Definition of variables
Variable Definition
Real estate Ratio Real estate value/Total asset
Cash Ratio Cash/Total asset
Leverage (Current debt + Long term debt)/ Total asset
Working capital Ratio Working capital/total asset
Capital expenditure Ratio Capital expenditure/total asset
Appendix 2
Figure 1: Housing Price Index: All-Transactions Indexes
Sources: Federal Housing Finance Agency
0 50 100 150 200 250 300 350 400 1993 1994 1995 1996 1997 1998 1999 2000 2001 2002 2003 2004 2005 2006 2007 2008 2009 2010 2011 2012 2013 2014 29
