Tobin’s q as a control variable in
investment models: the difference for
young and old firms
Student Number: 6297471
Supervisor: Christiaan van der Kwaak Field: Economics & Finance
Date: 04-03-2014
Table of Contents Page number
1. Introduction 1-2
2. Literature review 2
2.1 Origination of q 2-3
2.2 Market price of equity 3
2.2.1 Interest rate channel 3-4
2.2.2 Exchange rate channel 5-6
2.2.3 Credit channel 6-7
2.2.4 Tobin’s q as a transmission channel 7
2.2.5 Firm specific factors 7
2.3 Replacement cost of capital 8
2.4 Investment models with Tobin’s q 8-10
3. Data 10
3.1 Young versus old companies 11
3.2 Tobin’s q data 11
3.3 Investment data 11
3.4 Cash flow data 11-12
3.5 Profit and debt data 12
3.6 GDP data 13 4. The model 13 5. Results 14-15 6. Conclusions 16 7. Appendix 17 8. Bibliography 18-19
1. Introduction
In the world of corporate finance Tobin’s q is an important market indicator that helps to explain many corporate phenomena. Tobin’s q is defined as the market value of a firm over the replacement costs of its assets. When q is high, firms can issue equity at a high level relative to the replacement costs of capital. Since firms with a high q can acquire new capital by issuing relatively little equity the level of investments in the economy will rise (Mishkin, 1995). The relationship described by Mishkin can be tested empirically and recent studies try to do this. In the existing literature an important application of Tobin’s q is the use in models to explain investment. In models where cash flow is an important explanatory variable q is used to control for investment opportunities (Gilchrist & Himmelberg, 1995). GH, henceforth GH, state that this application of q is not without problems.
The most common Tobin’s q used for decision-making is the market value of all public companies in the US accumulated divided by their total replacement cost. When this q is high the average cost for companies to acquire capital is relatively low. Investors might see this as an implication for overvaluation. This q value can be found at many financial websites under the section “Key Indicators”.
As stated earlier the use of q to control for investment opportunities is controversial. GH summarize some of the concerns associated with the use of q to explain investment. One of the problems they mention follows from the difference in explanatory power of cash flow for financially constrained and unconstrained firms. Fazzari, Hubbard, and Petersen (1988) found that the explanatory power of cash flows is higher for financial constrained firms. Solving this by sample splitting gives a problem for q. The problem that arises follows from the fact that financially constrained firms are typically younger, smaller and faster growing than unconstrained firms. The market value for a company is an important determinant for q, and the market value for young, small and fast growing firms is more volatile than for older firms. This results in less statistical significance for q. In the model where q controls for investment opportunities the statistical significance is now uncertain.
GH do not refer to an empirical research confirming their assumption that explanatory power moves away from q towards cash flow for other reasons than financial constraints. Their theory that a lack of historical stock data influences q through imperfect accumulation of information is the focus of this study. The main question in this study is if the explanatory power of q varies through samples because of other reasons than financial constraints, especially because of a lack of historical stock data. This study differs from previous papers because of the focus on the availability on stock data. Other factors are held constant. In
previous papers several other factors like dividend payments were taken into account to distinguish between different types of firms
Using the approximation method constructed by Chung and Pruitt (1994) to calculate q values this study will test if there is a difference in statistical significance in explanatory power of q as a control variable between young and old firms. The results will either confirm the concern that GH described or the results will imply that there is less reason for concern. The data sample consists of 197 S&P500 firms. Firms for which the stock data is available for more than two years are considered as old. All other firms are considered as young. Using an Arellano-Bond estimation with a dummy separating young and old firms this paper searches for a difference in explanatory power for q as an explaining variable for investment.
In chapter two the theory behind q is explained and previous research papers for q as a control variable are reviewed. The data collection and usage will be discussed in chapter three. The research method will be explained in chapter four. Empirical results are given in chapter five. Conclusions will be discussed in chapter six.
2. Literature review
In this section the theory of Tobin’s q is reviewed beginning with the origination. After the origination the elements of Tobin’s q, Market Price of Equity and Replacement Costs of Capital, are studied respectively. Next several relevant applications and concerns for this study are reviewed.
2.1 Origination of Tobin’s q
The underlying theory of Tobin’s q originates from an paper by Brainard and Tobin (1968). In their paper, titled “Pitfalls In Financial Model Building”, they set up a model to simulate an economy. In the explanation of the model the theory on Tobin’s q is mentioned for the first time. Below follows a citation from this paper:
One of the basic theoretical propositions motivating the model is that the market valuation of equities, relative to the replacement cost of the physical assets they represent, is the major determinant of new investment. Investment is stimulated when capital is valued more highly in the market than it costs to produce it, and discouraged when its valuation is less than its replacement cost. Another way to state the same point is to say that investment is encouraged when the market yield on equity rK is low relative to the real returns to physical investment.
Brainard and Tobin use the ratio between the market value and the replacement cost as the prime indicator and proper target of monetary policy in their model. For the same use in the real economy other markets, like real estate and cars, should be considered as well. This is hard to do in practice and modern economies do not have monetary policy that targets Tobin’s q.
Following up on this early insight Bernanke and Gertler (2001) argue that monetary policy should not try to affect asset prices. The possibility of central banks to affect asset prices directly will result in unpredictable effects on market psychology. Bernanke and Gertler simulate how inflation targeting works in the event of an asset price shock. In modern economies the most common monetary policy is based on inflation targeting. They find that trying to respond to volatile asset prices has no beneficial addition to the inflation targeting regime.
After the introduction of the theory it was Tobin who started using the letter “q” to indicate the ratio of the market value of capital to the replacement costs of capital (1969). After the application of q in a simulated economy by Brainard and Tobin, Tobin started to further develop the theory and nowadays Tobin’s q is an important market indicator. Tobin’s q is named after the economist James Tobin. The next paragraph reviews the implications that q has on the modern economy.
2.2 Market Price of Equity
The numerator of Tobin’s q is the market value of equity (Brainard & Tobin, 1968). What is this value and what influences this value? Company stocks are traded on asset markets. To see how prices on asset markets are influenced an analysis on different channels has to be made. Most literature focuses on the effects of monetary policy. The most important channels are discussed below. After discussing the channels that affect the general price level firm specific factors are discussed.
2.2.1 Interest rate channel
The interest rate channel is one of the factors that influence equity prices. One of the first questions that arises is which interest rate should be studied to see the effects on equity markets. Taylor argues that it is most relevant to use a short-term private market rate, like the federal funds rate. In an empirical research on this topic decisions on what interest rate to
include in the analysis should be based on economic theory. An average of several relevant rates might also be used (1995).
Interest rates can be influenced through monetary policy. A decrease of the money supply will result in a higher interest rate (ceteris paribus). This is represented by the IS-LM curve. The theory that underlies the IS-LM curve is developed by Keynes in his paper The General Theory of Employment (1937). Hicks (1937) summarized the model into the
Investment Saving – Liquidity Preference Money Supply model. This model is now known as the IS-LM model. Below is the figure that Hicks used to display his model given.
Figure 1. Hicks, J. R., (1937).
The interest rate is on the vertical axis and real income is on the horizontal axis. As the graph shows a decrease of the money supply will shift the LM curve upward resulting in a higher interest rate. An increase of the money supply has the opposite effect. A change in the propensity to consume or the level of investment shifts the IS curve which will also change the equilibrium level of income and the equilibrium interest rate.
Tobin uses Hicks model to explain his capital account approach. This is relevant due to the macroeconomic effects on domestic asset prices. In his approach income is one of the factors that influence asset markets. He states that “asset stock equilibrium corresponds to any tentative assumption about aggregate real income” (1969). Figure 2 is given below. This is the illustration Tobin used to illustrate his approach. This study focuses on the effect of income on asset markets.
Taking in mind that the interest rate, i, is the most relevant rate for the analysis, the effects of interest changes can be captured in a schematic diagram. In his symposium Mishkin stated that the diagram, according to Keynes traditional view, is as follows.
M ! i ! I ! Y
M is in this diagram a tightening monetary policy. A decrease in the money supply results in a
higher interest rate, i. A higher interest rate results in less investment I. Equilibrium is
restored at a lower level of real income. With less income less money will be spend on stocks lowering the market value of companies.
Figure 2. Keynes, J. M., (1937).
2.2.2 Exchange rate channel
Taylor (1995) states that changes in monetary policy actions affects short-term interest rates. This affects both the exchange rate and long-term rates. Changes in the exchange rate affect the exports and imports and this has consequences for real income. This relationship is shown in the diagram below.
M ! i ! E ! NX ! Y
Again there is a tightening monetary policy M which decreases the money supply and thus increases i. A higher domestic interest rate makes it more profitable to invest in the domestic currency E, resulting in a capital inflow. When the domestic currency gets stronger domestic product will be more expensive which is denoted by NX. A lower level of net exports results in a lower aggregate income. With a lower income there is less money to be spend. The demand for assets will be lower and thus the market value of companies decreases. For the effects on asset market see figure 2 above.
2.2.3 Credit Channel
In Mishkins (1995) symposium on the monetary transmission mechanism he states that the credit channel has two components. One of the components is the bank lending channel. Edwards and Mishkin (1995) argue that the bank lending channel in Amerika has become less important due to financial innovation. For that reason this paper focuses on the other
component of the credit channel, the balance-sheet channel.
There are two important ways in which monetary policy can affect firms’ balance sheets. Mishkin (1995) states that a tightening monetary policy will result in lower equity prices as discussed above. According to Mishkin this results in an increase of the adverse selection problem. There is less collateral for bank loans and the adverse selection problem increases. This means that the number of loans issued to unsolvable lenders increases and the number of loans to solvable lenders decreases. Mishkin argues that the moral hazard problem also increases. Since there is less equity firms are more likely to engage in risky projects. Riskier projects will result in more defaults, this means that the number of loans that will not be paid back increases. This channel can be displayed schematically as below.
M ! P ! Adverse Selection & Moral Hazard ! Lending ! I ! Y
Again M simulates a tightening monetary policy. Equity prices, P, are lower due to the decreased money supply. As a results the adverse selection and moral hazard problems increase. Lending decreases and this result in low investment I. A lower level of investment result in a lower level of income Y. This schematic display is similar for the other effect that a tightening policy has which is discussed below.
When the money supply decreases the interest rate increases, as explained above. This affects the cashflow of firms, since a higher payment on debt has to be made. This has similar effects on the adverse selection and moral hazard problem (Mishkin, 1995).
2.2.4 Tobin’s q as a transmission channel
All of the channels above influence asset prices due to a change in the level of income. Changes in asset prices influence q, as can been seen from the definition of q given by Tobin (1969). This definition is the market value of a firm relative to the replacement costs of capital. When asset prices are high, the ratio of the market value of a firm over the replacement cost of capital is high. This means that it is relatively cheap to invest in new plants and equipment instead of acquiring existing capital. By issuing equity they get a high price relative to the new capital they are buying. When q is low the opposite happens. Issuing equity results in a low prices relative to the replacement costs of capital (Mishkin, 1995). This relationship can be shown schematically.
M ! P ! q ! I ! Y
A tightening policy M results in lower equity prices and thus a lower q. A low q will result in less investment I and thus a lower level of income Y.
So q is not only affected indirectly through the interest rate, the exchange rate and the credit channel. A decrease in equity prices also directly decreases q. It is obvious that all of the above is important for understanding the value of q in different business cycles. Firm specific factors that influence Tobin’s q on firm level are discusses in the next session.
2.2.5 Firms specific factors
Maybe more important than the factors discussed above are the firm specific factors that influence the market value and thus q. Kalay (1982) for example found a drop in stock prices the day after a dividend pay-out that is larger than the dividend per share.
Other factors like news about the business environment also influence stock prices. Chen, Ross and Roll studied the effect of economic news on stock prices. They found that stocks are priced according to their exposure (1986). In the data sample used in this study firms are selected randomly from the S&P500 which gives a sample with firms with a large variation in exposure. Therefore specific business environments are not further discussed. Due to the large variation in business environment, possible effects are cancelled out.
2.3 Replacement cost of capital
The replacement cost of capital as the denominator in q are the costs for replacing the existing capital. Capital can by acquired in two ways. By investing or by acquiring existing capital from other firms. When capital is acquired from existing firms this is not considered as investment. When the replacement costs of capital are high relative to the market value it is more likely for firms to acquire capital from other firms (Mishkin, 1995). Investment spending will thus be low when it relatively cheap to acquire capital from other firms.
The replacement cost of capital has some specific features. Some capital for instance cannot be reproduced. Tobin and Brainard (1977) state that the following applies for
reproducible assets:
In the case of reproducible assets, the current cost of producing identical or competitive goods is obviously an important factor in the valuation of an existing asset. Thus a rise in residential construction costs can be expected to raise the value of existing homes, and rise in the price of new cars is “good” for the price of previous year’s models.
From this it follows that general supply-demand theory influences the replacement costs of capital through asset markets.
Tobin and Brainard imply that for non-reproducible assets like land and mines valuation depends on specific models. Each non-reproducible asset is valued individually since each asset has its own specifications.
Following up on the section above existing literature with models where q is used as a control variable is discussed below.
2.4 Investment models with Tobin’s q
Fazzari, Hubbard and Petersen developed a model of investment where large variations in q have little effect and investment is constrained by cash flow. One explanation is that
investment is directly dependent of available funds (1988). Another explanation, less
plausible according to GH, is that shocks in net earnings affect future net worth and thus the terms for lending (1995). Costs of funds are lower when profits are high. Lending to a highly profitable firm is not risky and this results in low interest costs. The weighted average cost of capital is low for highly profitable firms.
GH developed a new empirical framework that tried to answer the question if cash flow really explains investment by reducing the costs of capital or if increases in cash flow are more a forecast for investment opportunities. They state that existing studies have problems with using Tobin’s q as a control variable for investment.
It is well-known in the literature that Tobin's Q has low explanatory power and yields implausibly high estimates of the adjustment cost parameters. Thus, it is hard to make the case that Tobin's Q adequately controls for the information content in cash flow. If Tobin's Q is for any reason not a sufficient statistic for investment, then the evidence on excess sensitivity fails to provide convincing proof for the existence of capital market imperfections.
Nevertheless the standard approach in recent literature is to use Tobin’s q as a control variable for investment.
Previous studies, like the study by Fazzari, Hubbard and Peterson, showed that the predictive power of cash flow is higher for financially constrained firms. According to GH this can take away some of the concerns that Tobin’s q has as a control variable. Splitting data samples might result in a higher explanatory power of q since the criteria used to identify financially constrained firms are not related to the shortcomings of q.
Another study where q is included in a model to explain investment is done by Alti (2003). Alti found that investment is more sensitive to cash flow for young firms than for old firms where the a priori criteria to distinguish young from old firms was dividend behavior. The study in this paper focuses on the question if the results will be similar when availability of historical stock data is used as a priori criteria to distinguish young from old firms. This is important to perform sample splitting in a way that does not result in new problems.
Splitting data gives some new concerns. The first concern given by GH follows from the criteria used to identify financially constrained firms. Since the data is split a priori the firms that are marked as financially constrained typically are newer, smaller and faster growing than financially unconstrained firms. It is likely that the stock market has
incorporated less information on these firms than on firms for which stock data is available for a longer time. If this concern is legitimate, the explanatory power of q will decrease for young firms. The explanatory power of cash flow will then increase for young firms.
The second concern involves the learning period that companies have in which they learn about their profitability. Jovanovic’s (1982) model shows that companies that are still
learning about their profitability react stronger to fluctuations in cash flow than companies that know how profitable they are. This follows from the theory that a fluctuation in earnings for a young firm results in a great revision of expected profitability. GH state that in this case the explanatory power again moves away from q towards cash flow.
The third concern follows from the assumption that smaller, younger firms can adapt more quickly to news about future profitability. GH argue that young and small firms have less bureaucracy and hierarchical problems, which allows them to adapt more quickly than old firms. If Tobin’s q is a sufficient statistic the coefficient for q should rise. If q is not a sufficient statistic the explanatory power again moves away from q towards cash flow. This means that investment is more sensitive to cash flow.
The concerns above show that splitting the sample based on a priori information is not a solution for q without problems. The purpose of distinguishing firms based on their access to capital is blurred due to the concerns mentioned above. Even when data samples are split based on a priori information for financial constraints q varies across sub samples due to reasons that have nothing to do with access to external funds. Phrased differently, when for example dividend behavior is used to distinguish between financial constrained and
unconstrained firms in order to control for the variation across q there is still a possibility that q varies in the subsample due to other reasons.
In the next section of this paper the first concern is tested empirically. The results will either confirm the legitimacy of the concern or imply less reason for concern
3. Data
In this section the different variables used in the model are explained. All data is gathered using Datastream. In each subsection the exact Worldscope code is given. The exact data used is also explained. The frequency for all data is yearly. The companies used are selected
randomly. From a list with all the ticker symbols for S&P500 firms, ticker symbols where randomly imported in Datastream. The timespan for the analysis made in this study is 1990-2013. In table 1 at the end of this section summary statistics for all relevant variables are given. In appendix 7.1 data for 10 randomly picked companies is given.
3.1 Young companies versus old companies
As stated in the introduction a concern with the use of q as a control variable comes from the amount of information incorporated in stock prices. Gilchrist and Himmelburg (1995) argued that explanatory power might move away from q towards cash flow for young companies. When stock data is available for less than 2 years the company is considered as new. A dummy variable is used. When the company is young a “1” is assigned. When the company is considered as old a “0” is assigned. This means that when stock data is available since 1996 the company is considered as new for the years 1996 and 1997. After 1997 the dummy changes to 0 and the company is considered as old. The sample used consists of 197 S&P500 companies that are randomly selected.
3.2 Tobin’s q data
In Datastream data on q is not directly available. Using the approximation formula that was developed by Chung and Pruitt (1994) the q values are calculated. The source of the data is Worldscope. The data frequency is yearly. The formula used is given below (1). Tobin’s q is one of the two independent variables of interest in this study.
(1) Tobin’s q = (Equity Market Value + Liabilities Book Value)/(Equity Book Value + Liabilities Book Value)
3.3 Investment data
The variable used for investment is capital expenditures as a percentage of total assets. The resulting value is used as the measurement for investment. This data is available directly so there is no need to construct a formula.
3.4 Cash flow data
To get cash flow the operating, financing and investing are combined. Next the total cash flow is normalized over total assets. Formula (2) shows what the calculation looks like in words.
(2) ((Operating Cash flow) + (Investing Activities) + (Financing Activities)) / (Total Assets)
As discussed in the previous section cash flow is an important variable for explanation of investment. In this research the focus lies on finding a movement of explanatory power away from q towards cash flow. Cash flow is the second independent variable of interest in this study.
3.5 Profit and Debt data
For the profit variable the Operating Profit is normalized over Total Assets. The formula is written down below (3). Profit is used as a control variable in the model. According to economic theory highly profitable firms have lower costs of funds and can thus invest more. The coefficient in the model is expected to be positive, otherwise the results are not in line with the theory.
(3) (Operating Profit) / (Total Assets)
Data on debt is also used as a control variable in the model. A higher debt ratio might lead to less investment. Total Debt is normalized over Total Assets. Formula (4) gives the calculation in words. Debt is used as another control variable. From economic theory it follows that firms with a higher debt ratio have less funds available for investment due to interest payment. The coefficient in the model is expected to be negative.
Table 1 gives summary statistics for the variables discussed above.
(4) (Total Debt) / (Total Assets)
Table 1
Variable Obs Mean Std. Dev. Min Max
Q 3763 2,60875 3,09218 0,46333 103,82170 INV 3763 0,07203 0,08054 0,00000 1,55760
CF 3763 0,19535 0,20681 -1,66209 1,44855 PROFIT 3763 0,11120 0,12346 -2,72236 1,13492 DEBT 3763 0,24616 0,19633 0,00000 1,63725
All the means, except for q, are between 0 and 1. See the appendix for the methodology to get values between 0 and 1. Furthermore the maximum value for q is very high. This results in a higher standard deviation for q, but this observation cannot be deleted as an outlier since high values for q are not uncommon.
3.6 GDP data
Since investment differs across different business cycles GDP is incorporated as a control variable in the model. Real GDP data in billions of chained 2009 dollars, not seasonally adjusted, is used. GDP is used to control for the business cycle.
4. The Model
In this section an explanation of the econometrical model used is given. The coefficients and dummy implementation are explained as well as the setup of the panel data. Furthermore a short note on how to interpret the relevant coefficients is given.
Arellano and Bond (1991) developed a test that is applicable for estimations with dynamic panel data using the generalized method of moments. This estimation method will be used in this paper.
The ID variable in the panel data is the company name. The time variable is the year. The general model is given below.
(5) Yit = β0 + β1Qit + β2Qit*DUMMYit + β3Cit + β4Cit*DUMMYit + β5Pit + β6Dit + β7Git
+ εit
The dependent variable Y is Investment. β0 is a constant and β1 is the coefficient for q. β2 is
the coefficient for q times the dummy. If the company is young the dummy has value one and the dummy has zero if the company is old. β3 is the coefficient for cash flow. β4 is the
coefficient for cash flow times the dummy. Again the dummy has value one for young companies and zero for old companies. β5, β6, and β7 are the coefficients for profit, debt and
GDP respectively.
According to the concerns of GH the coefficient β2 will be negative and the coefficient
β4 will be positive. The explanatory power then moved away from q towards cash flow.
Phrased differently, the sensitiveness of investment to cash flow increases. By using the dummy variables it becomes easy to investigate if this will happen and thus answer the question if q varies for young and old companies where historical data is used as a priori criteria to distinguish between young and old firms.
The Arrelano-Bond estimation allows gaps in time. This means that when values inside the timespan are missing this does not bias the results. When for a specific firm there is no data for 1999 and 2000 this firm can still be used in the regression.
5. Results
In this section results from the regressions are given. The significance and interpretation of variables is discussed and also the overall fit of the model is reviewed.
Table 2
Arrellano-‐Bond dynamic panel-‐data estimation
Number of obs = 3359 Group variable: ID1
Number of groups = 196 Time variable: YEAR
Obs per group: min = 3 avg = 17,13776 max = 23 Number of instruments = 284
Wald chi2 (8) = 1470,69 Prob > chi2 = 0,0000 One-‐step results
The regression output in table 2 shows that cash flow is indeed an important
explaining variable for investment. The z value is 14,95 which makes the variable significant at the 1% level. The coefficient for q is low relative to cash flow which confirms the results from previous studies that investment is not very sensitive to q. The coefficient for q is significant at the 5% level even when extreme observations are present in the sample. Profit and debt are both significant at the 1% level which makes them significant and relevant to incorporate in the model. They allow for a more isolated analysis of the effect of interest in this study, which is the movement of explanatory power away from q towards cash flow.
GDP is the only variable that is not significant at the 5% level. In the literature review several channels that might affect q are discussed. Al the channels indirectly affected q
INV COEF. Std. Err. z P >|z| 95% Conf. Interval
INV L1. 0,394152 0,0146387 26,93 0,000 0,3654608 0,4228433 Q 0,0015438 0,000645 2,39 0,017 0,0002797 0,0028079 DUMQ -0,0060736 0,0010869 -5,59 0,000 -0,008204 -0,0039433 CF 0,0773238 0,0051718 14,95 0,000 0,0671873 0,0874604 DUMCF 0,0893967 0,0199636 4,48 0,000 0,0502687 0,1285246 PROF 0,0395037 0,0147221 2,68 0,007 0,010649 0,0683584 DEBT -0,0356561 0,0109362 -3,26 0,001 -0,0570906 -0,0142216 GDP 0,0019887 0,002243 0,89 0,375 -0,0024076 0,006385 _CONS 0,0264423 0,0037757 7,00 0,000 0,019042 0,0338427
through the income level. In this model GDP as a proxy for income does not have a significant effect on investment. The use of GDP as a control variable for q is therefore insignificant.
The marginal effect of q for a young firm follows from the equation given below. The marginal effect for an old firm is also given.
(6) Δqyoung = β1 + β2Qit*DUMMYit = β1 + β2 = 0,0015438 – 0,0060736 = -0,0045298
(7) Δqold= β1 + β2Qit*DUMMYit = β1 = 0,0015438
From this marginal effects it becomes clear that q acts different for young firms. Even though the coefficient is small, the effect for q is now in the opposite direction. How is this possible? From the theory q should be higher for young firms since in general they are more credit constrained.
A possible explanation is the difference in marginal effects for cash flow between young and old firms. In the equations below the marginal effects for cash flow are given.
(8) ΔCFyoung = β3 + β4Cit*DUMMYit = β3 + β4 = 0,0773238 + 0,0893967 = 0,1667205
(9) ΔCFold = β3 + β4Cit*DUMMYit = β3 = 0,0773238
From this equations it becomes clear that cash flow has more explanatory power for young firms than for old firms. This makes the concern that GH legitimate since the explanatory power of q is lower for young firms than for old firms. Young firms are less sensitive to q and more sensitive to cash flow.
Alti (2003) comes to a similar result. He developed a model where investment is regressed on cash flow and q. Investment is sensitive to cash flow even when a control variable for profitability, Tobin’s q, is included in the model. In his model the sensitiveness was higher for young firms than for old firms. His a priori criteria to distinguish young from old firms was dividend behavior. He did not study the effect of the availability of historical stock data. The results of the regression in this study imply that the availability of stock data also influences the information content of q. Since availability of stock data is independent from dividend behavior, splitting solely on dividend behavior results in variation across subsamples.
6. Conclusion
This study tried to research one of the concerns about Tobin’s q as stated by GH. The existing literature still uses Tobin’s q as a control variable in investment models even though it is clear that there are problems with q that cannot be solved by sample splitting based on a priori criteria for financial constraints. One concern with Tobin’s q is that a lack of stock data results in less accumulation of information in stock prices and thus blurred values of q. When this is true sample splitting based on a priori information on financial constraints is not sufficient for a significant meaning of q.
This study focused on the concern that a lack of historical stock data results in a blurred value of q. Introducing a dummy for young and old firms gives an opportunity to see what happens with the explanatory power of q and cash flow. The dummy was assigned based on the availability of historical stock data. When data is available for less than two years the company was considered as young. When historical data is available for more than two years it is assumed that the market incorporated all possible information and the company is considered as old.
An Arellano-Bond estimation was performed on a sample of 197 randomly picked S&P500 firms. By analyzing the marginal effects it becomes clear that explanatory power moves away from q towards cash flow. This makes the concern GH have legitimate.
Further research can focus on constructing a model that controls for the variation of q across subsamples. To do this corrections needs to be made for all possible factors that cause variations across subsamples. The existing literature gives a number of these factors and this paper added one more that should be taken into consideration.
7. Appendix
7.1 Data for 10 companies
YEAR Investment Cash flow Q Profit Debt GDP DUMQ DUMCF
1 3M 2000 .0808 .178594 3.935.743 .214271 .197302 .04090661 0 0
2 Apple Inc 2000 .0207 .253565 1.143.035 .091136 .044098 .04090661 0 0
3 Adobe Systems Incorporated 2000 .0371 .490348 13.406.418 .81 0 .04090661 0 0
4 United Parcel Service, Inc. 2000 .0932 -.071277 3.628.022 .206029 .166374 .04090661 3.628.022 .071277
-5 Best Buy Co, Inc. 2000 .1437 .255455 2.611.776 .180033 .010233 .04090661 0 0
6 BlackRock, Inc. 2000 .0732 .23685 5.319.576 .264771 0 .04090661 5.319.576 .23685
7 Consolidated Edison 2000 .0672 .121873 1.169.853 .0829 .373603 .04090661 0 0
8 Health Care REIT, Inc. 2000 .0365 -.162056 .920045 .059099 .380111 .04090661 0 0
9 Occidental Petroleum Corporation 2000 .0674 .313022 1.251.746 .150664 .281807 .04090661 0 0
10 Coach, Inc. 2000 .1007 .167864 3.852.071 .201164 .013557 .04090661 3.852.071 .167864
7.2 Worldscope codes
Worldscope code Definition
WC01250 Operating Income WC02999 Total Assets WC03255 Total Debt WC03351 Total Liabilities WC03501 Common Equity
WC04860 Net Cash Flow (Operatin) WC04870 Net Cash Flow (Investing) WC04890 Net Cash Flow (Financing) WC08001 Market Capitalization
WC08416 Capital Expenditure % Total Assets
7.3 Formulas
DPL# at the beginning and the 6 at the end of a Datastream entry specify the number of decimals in the output. The data used in this study thus has six decimals. Below are the formulas in worldscope codes for q, cash flow, profit and debt given respectively.
(1) DPL#((X(WC08001) + X(WC03351)) / (X(WC03501) + X(WC03351)),6) (2) DPL#((X(WC04860) + X(WC04870) + X(WC04890)) / X(WC02999),6) (3) DPL#((X(WC01250)) / (X(WC02999)),6)
8. Bibliography
Alti, A. (2003). How sensitive is investment to cash flow when financing is frictionless? The
Journal of Finance, 58(2), 707-722
Arellano, M., & Bond, S. (1991). Some tests of specification for panel data: Monte Carlo evidence and an application to employment equations, The Review of Economic
Studies, 58(2), 277-297
Bernanke, B. S., & Gertler, M., (2001). Should Central Banks Respond to Movements in Asset Prices?, The American Economic Review, 91(2), 253-257
Brainard, W. C., & Tobin, J., (1968). Pitfalls in Financial Model Building, The American
Economic Review, 58(2), 99-122
Chen, N. F., Roll, R., & Ross, S. A. (1986). Economic forces and the stock market. Journal of
business, 383-403
Chung, K. H., & Pruitt, S. W. (1994). A simple approximation of Tobin’s q,
Financial Management, 23(3), 70-74
Edwards, F. R, & Mishkin, F. S., (1995). The Decline of Traditional Bankin: Implications for Financial Stability and Regulatory Policy, Economics Policy Review, 1(2), 27-45 Fazzari, S. R., Hubbard, G., & Petersen, B. C. (1988). Financing constraints and
corporate investment, Brookings Papers on Economic Activity, 1, 141-195
Gilchrist, S., & Himmelberg, C. P. (1995). Evidence on the role of cash flow for investment,
Journal of Monetary Economics, 36, 541-572
Hicks, J. R., (1937). Mr.Keynes and the “Classics”; A suggested Interpretation,
Econometrica, 5(2), 147-159
Jovanovic, B. (1982). Selection and the Evolution of Industry, Econometrica: Journal of the
Econometric Society, 649-670
Kalay, A. (1982). The ex-‐dividend day behavior of stock prices: a re-‐examination of the clientele effect. The Journal of Finance, 37(4), 1059-1070
Keynes, J. M., (1937). The General Theory of Employment, The Quarterly Journal of
Economics, 51(2), 209-223
Mishkin, F. S., (1995). Symposium on the Monetary Transmission Mechanism, The Journal
of Economic Perspectives, 9(4), 3-10
Modigliani, F., (1971). Consumer Spending and Monetary Policy: The Linkages, Boston:
Federal Reserve Bank of Boston, 9-84
Taylor, M. P., (1995). The Economics of Exchange Rates, Journal of Economic Literature,
Tobin, J., (1969). A General Equilibrium Approach to Monetary Theory, Journal of Money,
Credit and Banking, 1(1), 15-29
Tobin, J., & Brainard, W. C. (1977). Asset markets and the cost of capital, Economic
