Master Thesis
Investment sensitivity to interest rate
A comparison of physical and intangible capital investment sensitivities
MSc Finance – Quantitative Finance Pieter de Vries
June 2018
Statement of Originality
This document is written by Student Pieter de Vries 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
Abstract
Theory on investment has been widely debated, from classic theories describing investment as a function of interest rate to recent literature using Tobin’s q. Despite becoming increasingly important in the literature, intangible investment has not been studied broadly. Current literature agrees that was a shift from tangible towards intangible investment after year 2000. This paper discusses existing literature on investment and intangible capital. Firm level data and panel regressions are used to investigate whether or not intangible investments behave different to changes in interest rate than tangible investments. Overall, the results provide no evidence on a difference in behavior between tangible and intangible investments. However, when dividing the sample into subsamples, this paper does find little evidence on a difference in sensitivity to a change in interest rate between tangible and intangible investments.
Table of contents
1. Introduction ... 5
2. Literature Review ... 7
2.1 Investment ... 7
2.2 Intangibles ... 10
2.3 Intangibles in investment theory ... 12
3. Methodology ... 13
3.1 Regression model ... 15
3.2 Hypothesis ... 16
4. Data and descriptive statistics ... 17
5. Results ... 20
6. Robustness checks ... 24
6.1 Fama and French five industries ... 24
6.2 Time period results ... 25
6.3 Different lags ... 27
6.4 Data adjustments ... 28
7. Conclusion ... 30
Reference list ... 32
1. Introduction
On March 2018 the United States Federal Reserve decided to raise its federal fund rate by a quarter point. Widely used macroeconomic theory suggests that an increase in interest rate leads to a decrease in investment spending. Increased interest rates result in investments becoming more costly. Vice versa, a decrease in interest rate would make it more attractive for companies to invest by decreasing the necessary costs of funding.
It is of vital importance to understand the effects of the interest rate as it is one of the main tools used by government institutions to conduct monetary policy. Hall (1977) states that the sensitivity of investment expenditures to the interest rate is the basis of every monetary policy analysis. He claims that a higher sensitivity means a more efficient monetary policy and vice versa. Additionally, Thorbecke (1997) found evidence that monetary policy has real and quantitatively important effects on the economy. Following the classical macro economical theory, a change in interest rate has an effect on investments. However, to efficiently determine a monetary policy, it is important to understand what the effect of a change in interest rate is on investments, and whether or not that effect is equal for tangible and intangible investments.
Investment spending can be divided into two categories: investment in tangible assets (for example machines, tools, etc.) and investment in intangible assets, such as software and knowledge. Several papers have focussed on intangible asset investments and its increasing intensity of total investments. For instance, Peters and Taylor (2017) discuss that the neoclassical theory has mainly been tested on physical investments, however they continue to find evidence that the same theory is also relevant for intangible capital. They conclude their paper with recommendation for future research, including research how tangible and intangible capital respond differently to financial constraints and growth options. Another paper that discusses the importance of intangible investments is Corrado and Hulten (2010). They find that 34% of the firm’s total capital market consisted of intangible assets. Corrado et al. (2009) have estimated that 800 billion dollars is excluded from US published data. Gutiérrez and Philippon (2016), Alexander and Eberly (2018) and Döttling et al. (2017) discuss that after the year 2000 investment shifted from tangible investments towards intangible investments. Despite this trend, intangible investment has not received much attention for further research and has not been readily researched to date. Only recent literature makes a distinction between tangible and
Literature on investment finds mixed evidence of the effect of interest rate on investment. Classical theories from Hicks (1937) and Jorgenson (1963) predict that investment is a function of interest rate. However, recent literature from Sharpe and Suarez (2013) and Zenner et al. (2014) find that investment is determined by the hurdle rate and not by the investment rate. Moreover, Zenner et al. (2014) add that the hurdle rate is not always correlated with the interest rate, since firms believe that the interest rate is kept at a low level by the Federal Reserve. Firms also believe that the interest rate will rise in the near future. Recent literature on investment from Hall (2015), Alexander and Eberly (2018), Gutiérrez and Philippon (2016) and Peters and Taylor (2017) focus more on total q, the ratio of market value of new investments to the replacement costs of those new investments. Recent literature also uses total cash flow in investment models. Not all literature finds evidence for total cash flow in the investment-q regression. Evidence for total cash flow as a determinant is found by Peters and Taylor (2017), but contradicting literature by Erickson and Whited (2000) finds no evidence for total cash flow as explanatory variable. This paper will build further on the above-mentioned theories by adding intangible investments.
This paper will use firm level date from 1975 to 2016. The sample includes variables and characteristic of firms in the United States. By using panel-regressions, this paper will try to answer the following: Is the sensitivity to a change in interest rate equal for tangible and intangible investments? In order to answer that question, existing literature will be discussed and analysed in an attempt to deepen understanding on this subject matter. Furthermore, the existing literature on intangible capital will be discussed, as that is needed to make a distinction between tangible and intangible investment. This will be further explained in section 2. Section 3 will discuss the chosen methodology in this paper. The used data and the adjustments of the data will be discussed in section 4. Section 5 provides the results of the tests and how to interpreted those results. Section 6 provides robustness checks on the regression by dividing the sample into different industries, different time periods, using different lags and when adjusting the data differently. The conclusion and recommendations for future research can be found in section 7.
2. Literature Review
This section will introduce the existing literature and models on investment. The second part will discuss the existing studies on intangibles and its role in investment. Part three combines the theories from the first two parts and will discuss the expectations of this study based on the existing literature.
2.1 Investment
For generations economists have used the macro-economical IS-LM model from Hicks (1937) to explain the investment changes. The IS-LM model is widely used and discussed in the academic world, for instance in papers of Hicks (1981) and Hall (1977). Hicks based his model on the Keynesian nonflexprice model as he explains in 1981. In his initial theory, Keynes used three elements: the consumption function, the liquidity preference and the marginal efficiency of capital to construct his model. All of these elements were analysed with the use of the parameters. According to the Keynesian theory, later adopted in the Hicks model, investment depends on the interest rate. The relation between income and interest rate that expresses investment results in the IS curve of the IS-LM model of Hicks. In this IS curve, interest rate and investment are negatively related, i.e. an increasing interest rate leads to decreasing investment expenditures. The IS-LM investment function is as follows:
𝐼 = 𝛾!+ 𝛾!𝑌 − 𝛾!𝑟, (1)
where I denotes real investment, Y denotes real GDP and r denotes the interest rate.
This IS-LM model is discussed by Hall (1977). In his paper he discusses how interest rates and the IS-LM model are used and what the effect of stabilization policies is. He states that the sensitivity of investment expenditures to the interest rate is the basis of every monetary policy analysis. He claims that a higher sensitivity means a more efficient monetary policy and vice versa. However, Hall (1977) also discusses several flaws of the interest elasticity named by sceptics. For instance, it takes time for firms to adapt to the new interest rate to design, build and buy new equipment. He states that investment in a certain period of time is the result of previous decisions years before. Only a small part of investment expenditures can be a result of a lower
interest rate. Another flaw discussed by Hall is that stabilization policies affect the short-term interest rate, while long-term interest rates are used for investment decisions.
In 1963, Jorgenson proposed a different investment theory. In his paper he proposed a theory where investment is based on the theory on accumulating capital. In his paper he describes that investment depends on the change in interest rate. In his theory investment is calculated as follows: 𝐼! = 𝑧!𝐾!∗+ 𝑦 !𝐾!!!∗ ! !!!!! , 𝑣 = 1, 2, … , 𝜏 , 2
where I is gross investment, K* is the desired amount of capital stock, 𝜏 is the amount of periods
we assume K* to be fixed and y is the form of lagged response. The elasticity of gross investment
to the change in interest rate is calculated as follows:
𝜕𝐼 𝜕𝑟= 𝑧!
𝜕𝐾∗
𝜕𝑟 (3)
The disadvantage of this model is that the assumption of exogenously given output is not consistent with the perfect competition assumption (Hayashi, 1982). Another drawback is that the theory cannot determine the investment rate. These drawbacks led to an adjustment in the classic theory by adding installation costs of new investment goods.
By adding the installation costs, Jorgenson’s theory became seemingly equivalent to the proposed model by Tobin (Hayashi, 1982). In 1969, Tobin proposed an alternative framework, where investment is a function of q. Tobin’s q reflects the ratio market value, consisting of new investments to the replacement costs of those new investments. In addition, Tobin’s q indicates whether a firm is overvalued or undervalued. The marginal q, equalling the ratio of the market value of an additional unit of capital to the replacement cost of that capital, is however unobservable. However, the average q is observable, equalling the ratio market value of existing capital to the replacement cost (Hayashi, 1982). However, Hayashi (1982) found that, under certain conditions, marginal q and average q are equal.
Despite being widely accepted in classical theories, not all literature agrees with the statement that investment is dependent on the interest rate. Sharpe and Suarez (2013) surveyed 500 CFOs and found that cash and interest rates are seldomly used in their investment decision-making. In 2012 most firms would not increase their investments if the interest rate would
decrease and only a few firms would decrease investments if interest rates would increase. They find that minimum returns, hurdle rates, are used instead of interest rates. They conclude that monetary policy, therefore, is redundant. Zenner et al. (2014) suggest that this minimum return, the so-called hurdle rate, is comparatively fixed and does not seem to be affected by lower interest rates. Furthermore, Zenner et al. (2014) indicate a big relative difference between cost of capital (interest rate), cost of debt and the hurdle rate, which is in general underestimated. They also find that that there has been a trend of increasing shareholder distributions in terms of dividend and buybacks. Zenner et al. (2014) state that this increased shareholder distribution was on expense of investments and that this affected economic growth, despite the low interest rate.
Recent literature on investment use combinations of the above mentioned frameworks to investigate investment. Several papers, like Hall (2015), Alexander and Eberly (2018), Gutiérrez and Philippon (2016) and Peters and Taylor (2017) research investment based on cash flow and Tobin’s q. Most of these papers explain why investment does not follow the theoretical framework after the financial crisis. Alexander and Eberly (2018) examine the recovery of investment after the great recession using firm-level data on investment. In their paper they discuss the findings of Hall (2015), finding that investment levels have not recovered as was expected after the financial crisis. They find that this is contradicting the existing theory, since interest rate have been held exceptionally low after the crisis. This also challenges the cashflow and q-theory based approaches. An interesting finding is that after the year 2000 there is a downward trend in physical capital and an upward trend in intangible capital. They find that tangible investment and intangible investment are negatively correlated. Theory suggests that investments seemed to have shifted towards service and cognitive skills after the year 2000. In their empirical research they find that Tobin’s q and cashflow are significantly correlated to investment, controlling for firm and year fixed effects. Tobin’s q and cashflow seem to have a positive effect on investment.
Gutiérrez and Philippon (2016) use industry-level and firm-level data to examine physical investment. Like Alexander and Eberly (2018) and Döttling et al. (2017) they find that after year 2000 the investment trend changed. Gutiérrez and Philippon find that investment is weak relative to Tobin’s q after 2000. In their paper they test what drives the underinvestment relative to Tobin’s q. They test for four categories: financial frictions, measurement error, lack of competition, tighter governance. They find some evidence for regulatory constraints and strong evidence that concentrated industries invest less. Döttling et al. (2017) find that investment is structural lower than expected in the United States since the year 2000. The underinvestment is
partly explained by an increase in intangible investments, starting mostly in the 1990s and the early 2000s.
Peters and Taylor (2017) discuss the investment-q relation and intangible capital. They modify the model by Abel and Eberly (1984), which is a unified model of Jorgenson’s and Tobin’s model. In the extended framework of investment under uncertainty of Abel and Eberly, investment is a non-decreasing function of q. They also discuss the limitations of investment regressions explained by Erickson and Whited (2000). In their paper, Erickson and Whited (2002) discuss that investment regressions are not able to find the level of adjustment costs. Peters and Taylor (2017) also provide a new measure to calculate Tobin’s q. They find evidence that this new measure, which accounts for intangible capital, is a better proxy for investment opportunities compared to classical measures. Although Peters and Taylor (2017) find significant results for cash flow in their regression, no all literature agrees. Erickson and Whited (2000) find no evidence that cash flow should be included in the investment-q regression by testing for financial constraints on investments.
2.2 Intangibles
Intangible investment has become one of the most important drivers of productivity growth (Döttling et al., 2017). In the literature, intangible assets have become more present after the year 2000. As mentioned earlier, Alexander and Eberly (2018) and Gutiérrez and Philippon (2016) found that the year 2000 is a start in shift in investment behaviour towards intangible assets. Intangible assets are discusses in papers by Corrado and Hulten (2010), Corrado et al. (2009) Alexander and Eberly (2018), Peters and Taylor (2017), Almeida and Campello (2007) and Brown et al. (2009). Corrado and Hulten (2010) find that 34% of the total capital market consists of intangible assets. However, Corrado et al. (2009) find that approximately 800 billion dollars is excluded from US published data. Including intangible capital significantly influences the economic growth pattern. They also discuss whether or not intangible assets should be capitalized. From an accounting perspective, intangible assets are often produced within a firm and the market does not generate verifiable data to determine the value of the intangible assets. Moreover, Hulten (1979) states that any use of resource that increases future consumption by decreasing current consumption qualifies as an investment. This suggests that any type of expenditures should be treated equal, meaning intangible investments should be treated equally to tangible investments.
Corrado et al. (2009) further discuss characteristics of intangible assets. They explain the non-rivalness and non-appropriability of it. The non-non-rivalness of some intangible assets, like R&D,
implies that it can be used simultaneous by several firms without limiting the proceedings for one of the firms. However, many intangible assets are non-rival. For instance, brand equity and human capabilities. Those can be very firm specific and very valuable in such a way that other firms do not have access to them. The “non-appropriability” implies that not all research and development (R&D) and employee investments may be captured by the firm that invests. However, they state that intangible investments satisfy the criterion that the expenditures are made to increase future consumption. Therefore, tangible expenditures and intangible expenditures should be treated equally.
Alexander and Eberly (2018) describe intangible assets as well. They first see an increase of the share of intellectual property in the total gross domestic investment. They analyse the difficulties to discuss the intangible investment, since intellectual property clearly is a part of it, but firms could also have invested in software or brands and put these on their balance sheet differently. Some firms would put intellectual property as intangible asset on their balance sheet and other firms as a result of merger and acquisition. Taking this into consideration, they do find a significant increase of intangible capital on firm level data. In intellectual industries, such as telecommunications, they find a higher share of intangible capital compared to the intellectual industries, such as the energy and oil industry.
Intangible assets can influence investment in a few ways (Döttling et al, 2017). First, measurement errors can occur, since intangible assets and investments are difficult to measure. The accounting rules are different for intangible capital that is created internally than for intangible capital purchased externally (Peters and Taylor, 2017). This would lead to an over-estimation of q (Döttling et al., 2017). Second, Döttling et al. (2016) describe that intangible capital creation relies on the co-investment of human capital. However, human effort is difficult to quantify and is theoretically not counted as firm investment. Lastly, it might be more difficult to accumulate intangible assets (Döttling et al., 2017). An increase in importance of intangible assets could lead to a different equilibrium of q.
One of the difficulties often discussed in the literature is financing intangible investments. Brown et al. (2009) discuss the financial constraints of R&D. R&D is usually developed within a firm and is very firm specific. Therefore, there is no collateral value in case of bankruptcy. Contrary to tangible capital, debtors cannot collect intangible collateral value, since it has no physical presence and monetary value to others. This leads to a decreasing debt capacity, since debt capacity depends on the ability to collect collateral (Hart and Moore, 1994). Brown et al. (2009) found that young, high-tech firms make use of internal cash flow and raise funds by issuing new shares. Another paper that discusses the financial constraints and tangibility of assets
is Almeida and Campello (2007). Firms that are more tangible have less difficulties financing investment with external funds, due to the fact that they invest in assets that can be partly recaptured by liquidation of assets in case of default. Another important characteristic is that payoffs of tangible assets are easier to observe than the payoffs of intangible assets. This reduced asymmetric information problem leads to less financial constraints for tangible investments. Almeida and Campello (2007) and Brown et al. (2009) are in line with Döttling and Perotti (2017). Döttling and Perotti (2017) also state that intangible investments need to be internally financed.
2.3 Intangibles in investment theory
Based on the neoclassical theory, investment and interest rate are inversely correlated. However, papers of Sharpe and Suarez (2013) and Zenner et al. (2014) state that investment is determined by hurdle rates and not by interest rates. Moreover, Brown et al. (2009), Almeida and Campello (2007) Hart and Moore (1994) and Döttling and Perotti (2017) discuss that intangible investment is not predicted by a change in interest rate. Intangible investments are usually financed by internal funds, raised by issuing new shares (Almeida and Campello, 2007). Thus, the expectation of this research is that there is that investment might be sensitive to the change in interest rate. Moreover, based on the literature on intangibles, this study expects that intangible investment is not determined by the change in interest rate. To conclude, the expectation of this paper is that there is a difference in sensitivity to a change in interest rate between tangible and intangible investments.
3. Methodology
This section outlines how the variables are measured or calculated, as will it explain the
regressions used in this paper. This section will conclude with the hypothesis of this study and expectations of the used variables.
3.1 Regression model
The aim of this paper is the analysis of the effect of a change in interest rate on intangible investments. In order to do that, a way to measure total investment is needed first. This paper will use the investment rate from Peters and Taylor (2017) as a dependent variable. Total investment needs to be determined in order to measure investment ratio. Investment rate 𝜄!"! will be calculated as follows: 𝜄!,!!"! = 𝜄 !,! !!!+ 𝜄 !,! !"# (4) 𝜄!!! denotes the tangible investment rate and 𝜄!"# denotes the intangible investment rate. Investment rate is the sum of tangible investment rate and intangible investment rate. Investment rates are calculated as the investment over the replacement cost of capital Ktot. This paper
measures the replacement cost of capital Ktot as the sum of the replacement costs of tangible
capital Kphy and intangible capital Kint.
𝜄!,!!!! = 𝐼!,! !!!
𝐾!,!!!!"! , 𝜄!,!!"# = 𝐼!,!!"#
𝐾!,!!!!"! , 𝐾!,!!!!"! = 𝐾!,!!!!!! + 𝐾!,!!!!"# (5)
Physical investments are fairly easy to calculate by using firm’s capital expenditures. However, as stated in section 2, intangible capital is difficult to determine since accounting rules depend on internally or externally created intangibles (Peters and Taylor, 2017). Internally created intangible capital is usually expensed as R&D on the income statement, while building human capital is expensed as a general or administrative expense. Externally purchased intangible capital is typically capitalized on the balances sheet as a part of Intangible assets. In this paper intangible capital is measured as the sum of a firm’s internally created and externally purchased intangible capital.
By using R&D expenses and 30% of the selling, general and administrative expenses for intangible investments this paper follows Hulten and Hao (2008) and Eisfeldt and Papanikolaou (2014). Following Peters and Taylor (2017), the selling, general and administrate expenses is corrected for “in progress R&D” and research and development expense. This leads to the following formulas: 𝑆𝐺𝐴!,! = 𝑆𝐺𝐴 𝑒𝑥𝑝𝑒𝑛𝑠𝑒𝑠!,!− 𝐼𝑛 𝑃𝑟𝑜𝑔𝑟𝑒𝑠𝑠 𝑅&𝐷 𝐸𝑥𝑝𝑒𝑛𝑠𝑒𝑠!,! − 𝑅&𝐷 𝑒𝑥𝑝𝑒𝑛𝑠𝑒𝑠!,! (6) 𝐼!,!!"# = 𝑅&𝐷 !,!∗ 1 − 𝛿!&! + 0.3 ∗ 𝑆𝐺𝐴!,!∗ 1 − 𝛿!"# (7) 𝐼!,!!!! = 𝐶𝐴𝑃𝐸𝑋!,!− 𝐷𝑃!,! (8)
In this formula, DP represents depreciation expenses and 𝛿 depicts the depreciation rates of research and development (R&D) and selling, general and administrative expenses (SGA).
Combining these formula leads to the following linear regression:
𝜄!,!!"! = 𝛼 + 𝛽
!∆𝑟!!!+ 𝛾! + 𝜋!+ 𝜑!+ 𝜀!,! (9) ∆𝑟!!! is the lagged change in interest rate, measured as the absolute difference between the interest rate 6 months before and the interest rate 3 months before. 𝛾 represents the firms fixed effects, 𝜋 the state-of-location-by-year fixed effects and 𝜑 represents the industry-by-year fixed effects. The results are clustered by industry. By introducing a lag in interest rate, this paper follows Hall (1977) and Galí (1992).
In order to check whether effects are equal for tangible and intangible investments, this paper introduces an additional variable. This additional variable is the share of intangible capital on total capital per firm, calculated as follows:
𝐼𝑛𝑡𝑎𝑛𝑔𝑖𝑏𝑙𝑒 𝐼𝑛𝑡𝑒𝑛𝑠𝑖𝑡𝑦!,! = 𝐾!,! !"#
𝐾!,!!"! (10) To check for equal effects the intangible intensity variable will be interacted with the lagged change in interest rate. This will lead to linear regression 2:
𝜄!,!!"! = 𝛼 + 𝛽
!∆𝑟!!!+ 𝛽! 𝐼𝑛𝑡𝑎𝑛𝑔𝑖𝑏𝑙𝑒 𝐼𝑛𝑡𝑒𝑛𝑠𝑖𝑡𝑦!,! + 𝛽!𝐼𝑛𝑡𝑎𝑛𝑔𝑖𝑏𝑙𝑒 𝐼𝑛𝑡𝑒𝑛𝑠𝑖𝑡𝑦 ∗ ∆𝑟!!! + 𝛾! + 𝜋!+ 𝜑!+ 𝜀!,! (11)
In order to prevent omitted variable bias, all explanatory variables should be added. As explained in section 2 there is theoretical basis to add total q and cash flow as explanatory variables. This is in line with previous research from Hall (2015), Alexander and Eberly (2018), Gutiérrez and Philippon (2016) and Peters and Taylor (2017). Several studies also found an underinvestment from the start of the year 2000. Therefore, variables for total q, cash flow and a dummy variable for the period starting from the year 2000 are added. This paper used the total cash flow as calculated by Peters and Taylor (2017). In their paper, they introduce an adjusted measure for the cash flow measure of Almeida and Campello (2007) and Erickson and Whited (2012):
𝑐!,!!"! = 𝐼𝐵!,! + 𝐷𝑃!,!+ 𝐼!,!!"# 1 − 𝜅
𝐾!,!!!!!! + 𝐾!,!!!!"# (12)
ctot denotes the total cash flow, IB is income before extraordinary items and 𝜅 represents the
marginal tax rate. This is used because expensing intangible investments is allowed by accounting standards. Following Peters and Taylor (2017), we assume the marginal tax rate to be 30%. Adding these variables will lead to linear regression 3:
𝜄!,!!"! = 𝛼 + 𝛽
!∆𝑟!!!+ 𝛽! 𝐼𝑛𝑡𝑎𝑛𝑔𝑖𝑏𝑙𝑒 𝐼𝑛𝑡𝑒𝑛𝑠𝑖𝑡𝑦!,! + 𝛽!𝐼𝑛𝑡𝑎𝑛𝑔𝑖𝑏𝑙𝑒 𝐼𝑛𝑡𝑒𝑛𝑠𝑖𝑡𝑦 ∗ ∆𝑟!!! + 𝛽! 𝑇𝑜𝑡𝑎𝑙 𝑄!,!+ 𝛽! 𝑐!,!!"!+ 𝛽
! 𝑀𝑖𝑙𝑙𝑒𝑛𝑛𝑖𝑢𝑚 + 𝛾! + 𝜋!+ 𝜑!
+ 𝜀!" (13)
Where the variable millennium equals 1 when the year is 2000 or later.
Since the Federal Reserve has kept the interest rate equal after the financial crisis, investment cannot be explained by the change in interest rate. This can lead to distorted results. Therefore, only using data up until 2008, the regression 3 is repeated for comparison again. By excluding the years after 2008 possible distorted results are prevented.
3.2 Hypothesis
To answer the research question, this study will test for significance of the interaction variable. This leads to the following hypotheses:
𝐻!: 𝛽!"#$"%&'() !"#$"%&#'∗ ∆!!!! = 0 𝐻!: 𝛽!"#$"%&'() !"#$"%&#'∗ ∆!!!! ≠ 0
Guided by the theory in section 2, the expectation is that there will be significant results for the interaction variable. Theory from Brown et al. (2009), Almeida and Campello (2007) and Döttling and Perotti (2017) suggests that intangible investments are paid with internal funds. Therefore, intangible investment is theoretically uncorrelated with interest rate. Furthermore, theory on investments and interest rate is contradicting. Classic theory suggests that interest rate and investment are inversely related, expecting a significant negative coefficient for the change in interest rate. However, recent literature from Zenner et al. (2014) suggests that managers take hurdle rates into account in the decision-making on investment. Based on that theory, the expectation is that no significant results will be found for the change in interest rate.
The results for total q are expected to be positive. Theory suggests that a higher ratio of market value of new investments to the replacement costs of those new investments would increase investment. The theory on cash flow is contradicting. Following Erickson and Whited (2000) there is no evidence that total cash flow should be added into the investment-q regression and therefore the results for total cash flow are expected to be insignificant. However, Peters and Taylor (2017) do find evidence for total cash flow. Based on their paper, the results are expected to be significantly positive.
4. Data and descriptive statistics
The data used in this paper is retrieved from the Wharton Research Data Services. The sample of this paper contains data from 37.275 firms from Compustat from 1975 to 2016. 1975 marks the first year firms were required to report R&D by the Federal Accounting Standards Board. Annual data on US firms is downloaded from Compustat and data on total Q is downloaded from Peters and Taylors total Q section of WRDS. Interest rates are retrieved from the Federal Reserve. Information on the variables used from Compustat can be found in appendix 1. Following standard literature proceedings, regulated utilities (Standard Industrial Classification codes 4900-4999), financial firms (6000-6999) and non-operating and public services (9000-9999) are dropped from the sample. Firms with missing or less than 5 million worth of physical assets are excluded from the sample, as are firms with missing values for total q and intangible capital.
Section 3 defines Ktot as the sum of the replacement costs of tangible and intangible
capital. In this paper we use a firm’s Property, Plant and Equipment (PP&E) as the replacement cost of tangible capital Kphy. The database of Peters and Taylor in Compustat contains the
replacement cost of intangible capital of firms from 1975 to 2016. This measure is used as the replacement cost of intangible capital Kint. Depreciation expenses are reported because of
accounting rules and are extracted from Compustat. For R&D depreciation 𝛿R&D, this paper
follows Li (2012) and use industry-specific depreciation rates. Industry names are mapped into NAICS codes using data from the Bureau of Economical Analysis. As explained in section 3, this paper uses the marginal tax rate of 30%, found by Peters and Taylor (2017). For the depreciation rate of SGA, this paper follows the papers of Lev and Radhakrishnan (2005) and Falato et al. (2013) with 𝛿SGA = 20%. With regards to measuring selling, general and administration expenses,
this paper follows Appendix B.1 from Peters and Taylor (2017). Missing values for SGA, depreciation expenses and “in progress R&D” are set to zero. As is standard in literature, all
ratios are winsorized at a 1% level to remove outliers. Table 1 gives a summary of the winsorized data.
Table 1 (Summary Statistics)
Statistics are based on the sample of Compustat firms from 1975 to 2016. The physical capital stock is measured as property, plant, and equipment (PP&E). Intangible capital stock and Total Q are retrieved from Peters and Taylor. Intangible intensity is calculated as the ratio of intangible capital to total capital. Delta Interest Rate is measured as the absolute difference between the interest rate of 3 months and 6 months before the date. Tangible investment is calculated as capital expenditures minus depreciation expenses and Intangible Investment is calculated by R&D expenses plus 30% of SGA, minus depreciation. Investment rates are calculated by investment over the capital of one year before. Investment rate is measured as the sum of tangible and intangible investment rate. Cash flow is calculated as ((Income before extraordinary items + depreciation expenses + Intangible Investment*(1-marginal tax rate))/total capital of 1 year before.
N Mean Median Deviation Standard Skewness Min Max
Tangible Capital (in million U.S. dollars) 108.579 1.990,83 99,18 11.897,79 17,63 5,00 487.365,99 Intanigble Capital (in million U.S.
dollars) 108.579 1.143,83 73,89 6.166,02 14,81 0,00 278.770,60 Intangible Intensity 108.579 0,44 0,46 0,28 -0,01 0,00 1,00 Total Q 108.579 0,94 0,56 1,39 3,14 -0,61 8,60
Measures
Delta Interest Rate 103.466 -0,11 0,00 1,17 -1,21 -10,17 4,56 Tangible Investment 107.086 49,95 0,67 603,59 20,66 -24.514,73 36.169,00 Intangible Investment 108.579 122,99 10,55 612,92 12,41 -67,92 21.820,80 Tangible investment Rate 107.085 0,04 0,01 0,11 3,31 -0,11 0,64 Intangible Investment Rate 108.577 0,08 0,06 0,08 1,75 0,00 0,43 Investment Rate 107.085 0,12 0,09 0,13 2,22 -0,11 1,07 Cash Flow 108.326 0,13 0,13 0,19 0,03 -0,55 0,82
Table 1 shows that the minimum amount of tangible capital is 5 million U.S. dollars. As explained before, it is standard in literature to exclude firms with less than 5 million worth of tangible capital. The average intangible intensity is 44%. This means that almost half of the total capital in our sample is intangible. By construction, the value of intangible capital has to be within 0 and 1, being measured as the ratio of intangible capital to total capital. Furthermore, average intangible investments are 122,99 million dollars, while average tangible investments are 49,95 million dollars. This also leads to a significantly higher investment ratio for intangibles than for tangibles. Although the standard deviations for both intangible and tangible investment ratios are equal, this is relatively significantly higher for the tangible investment rate.
The mean change in interest rate is negative, meaning an average decline in interest rates. This is supported by figure 1, the development of interest rates in our sample. Figure 1 shows that interest rate increased from 1975 to 1981 up to a maximum of 22,36% on July 22nd 1981.
After that the interest rate steeply declined, almost reaching 0% in response to the financial crisis of 2008. It is interesting to see is that the Federal Reserve has kept the interest rate low since the previous financial crisis of 2008. As can be seen in table 1, the biggest negative change in interest rate is greater than 10 percentage points.
The mean total Q measure by Peters and Taylor (2017) is 0,94. This implies that the average ratio of the market value of new investments to the replacement costs of those new investments is 0,94. On average, the firms in this sample are perceived to be undervalued.
0 5 10 15 20 25 In te re st R a te 1975 1980 1985 1990 1995 2000 2005 2010 2015 Date
5. Results
This section will provide an overview of the regression results. Table 2 shows results for the firm-panel regressions discussed in section 3 of this paper. All regressions are controlled for firm fixed effects, state-of-location-by-year fixed effects and industry-by-year fixed effects. Standard errors are adjusted for clustering at industry level.
Regression 1 shows the effect of the change in interest rate on the investment rate. The results show a coefficient of -0.000452. However, the coefficient is not significant at a 10% significance level. Therefore, the results of this model show that the investment rate is not explained by the change in interest rate. This follows recent literature of Sharpe and Suarez (2013) and Zenner et al. (2014) confirming that the interest does not affect investment rate. They state that companies base their hurdle rates on their risk-adjusted cost of capital. However, a lower interest rate does not particularly lead to a lower hurdle rate. Companies believe that the interest rate is held low artificially and they expect it to increase soon. Interesting is that the adjusted R2 is 0.606. This means that, despite the insignificance of the coefficient, the regression
model explains 60.6% of the variance of the dependent variable.
As discussed in section 3, the intangible intensity is added into regression 2. Also, an interaction variable of change in interest rate and intangible intensity is added to the regression. Interesting to see is that, when comparing the models, the adjusted R2 only increases 2 base
points, while regression 2 contains two more variables. The change and interest rate remains insignificant at a 10% significance level. However, the coefficient of intangible intensity coefficient is significant at a 1% level. The coefficient of -0.0719 can be interpreted as follows: one percentage point increase in intangible intensity will decrease the investment rate by 0.000719, ceteris paribus. The coefficient of the interaction variable in regression 2 is not significant at a 10% significance level. Therefore, the results of regression 2 say that the investment rate is not affected by the interaction variable. Considering this results, this regression
does not find evidence that intangible investments have a different sensitivity to interest rate change compared to tangible investments. This contradicts the theory that intangible capital is not purchased by external funds, but with internal funds. Furthermore, the coefficient of the change in interest rate is insignificant at a 10% significance level, meaning that a change in interest rate does not explain the investment rate. The results of regression 2 are in line with the results from regression 1 and, therefore, follow the theories of Sharpe and Suarez (2013) and Zenner et al. (2014). The results contradict the classical theory on investment, stating that decreasing interest rate leads to an increase in investment (Hicks, 1937, Jorgenson, 1963).
More recent literature, as described in section 2, sees cash flow and total q as determinants of investment. Therefore, regression 2 is expanded with control variables of total q and cash flow, calculated as described in section 3. Since papers from Gutiérrez and Philippon (2016), Alexander and Eberly (2018) and Döttling et al. (2017) find a change in investment behaviour after the year 2000 a dummy variable is included. This dummy variable Millennium equals 1 when the observation is later than the year 1999. The results of regression 3 show an insignificant coefficient for the change in interest rate at a significance level of 10%. Like in regression 2, the interaction variable is not significant at a 10% significance level and intangible intensity is significant at a 1% level. The coefficient for intangible intensity decreases in magnitude from -0.0719 in regression 2 to -0.0488 in regression 3. A one base point increase in intangible intensity will now lead to a 0.000488 decrease in investment rate, ceteris paribus. Furthermore, coefficients for total q and cash flow are significant at a 1% significance level. These results are in line with Peters and Taylor (2017). The coefficient of total q equals 0.0216, meaning that a percentage point increase in total q leads to a increase of investment ratio by 0.00216, ceteris paribus. The total cash flow coefficient is 0.125. This means that, holding everything else constant, a percentage point increase in total cash flow leads to an increase of 0.00125 of investment rate. This significant result for total cash flow contradicts the theory from Erickson and Whited (2000) that they find no evidence for significance of total cash flow.
However, the result does follow Peters and Taylor (2017). Furthermore, the dummy variable for observations after the year 1999 has a negative coefficient. However, the coefficient is insignificant, meaning that the coefficient has no significant influence on the dependent variable. Despite the insignificance of the dummy variable, 64.4% of the variance of investment rate is explained by regression 3. Therefore, we can say that regression 3 explains more of the variance of investment rate than regression 1 and 2.
In regression 4, only data until the financial crisis in 2008 is used. After the financial crisis of 2008 the Federal Reserve has kept the interest rate steady at almost 0%. Since the delta of the interest rate for that period is zero, investment cannot be explained by the change in interest rate and that can possibly give distorted results. Therefore, in regression 4, only data until 2008 is used. Almost 20.000 observations are excluded from the sample. It is interesting to notice that the adjusted R2 slightly increases to 65,8%. This means that regression 4 both predict the variance
of investment rate slightly better than regression 3. Despite having a smaller sample, the results are quite equal. The results show an insignificant coefficient for the change in interest rate at a significance level of 10%. Like in regression 2 and 3, the interaction variable is not significant at a 10% significance level and intangible intensity is significant at a 1% level. Besides some small changes in magnitude, coefficients of all variables are alike. A percentage point increase of intangible intensity leads to a decline of investment rate of 0.000606, ceteris paribus. Coefficients for total q and cash flow are significant at a 1% significance level. The coefficient of total q equals 0.0202, meaning that a percentage point increase in total q leads to a increase of investment ratio by 0.000202, ceteris paribus. The total cash flow coefficient is 0.150. This means that a percentage point increase in total cash flow leads to an increase of 0.00150 of investment rate, ceteris paribus. Likewise to the results in regression 3, the coefficient of the dummy variable for observations after the year 1999 is insignificant at a 10% significance level. Interesting to notice is that the adjusted R2 has changed only slightly and that all coefficient results remain the
investment rate. All regressions that included the interaction variable give insignificant results for the interaction variable; meaning that there is no evidence that intangible investment has a different sensitivity to the change in interest rate than tangible investment. Furthermore, it is worth noticing that the change in interest rate is consistently insignificant in all regressions. This can be explained by the theory of Zenner et al. (2014), which explains that investment is chosen based on hurdle rates and is not directly linked to the interest rate. However, classical theories of Hicks (1937) and Jorgenson (1963) explain that investment is a function of interest rate. The results in table 2 contradict those classical theories. The significant results for total q and cash flow are in line with results from Peters and Taylor (2017).
Table 2 (Regression Results)
This table contains firm-panel regressions as discussed in section 3. Regression 1 is the effect of the change in interest rate on the investment rate. Intangible intensity and an interaction variable of delta interest rate and intangible intensity are added in regression 2. Total Q and Total cash flow are added as control variables in regression 3, as is a dummy that equals 1 for the year 2000 and later. In regression 4, only data until the financial crisis in 2008 is used. All regressions are controlled for Firm fixed effects, state-of-location-by-year fixed effects and industry-by-year fixed effects. The full sample contains firm-level data from 1975 to 2016. Standard errors, which were adjusted for clustering at industry level, are reported in parantheses.
Regression (1) (2) (3) (4)
Dependent Variable Investment Rate Investment Rate Investment Rate Investment Rate
Delta Interest Rate -0.000452 -0.000560 -0.000234 -0.000181
(0.000296) (0.000807) (0.000752) (0.000774)
Interaction Variable 0.000244 0.000180 -9.12e-05
(0.00183) (0.00176) (0.00180)
Intangible Intensity -0.0719*** -0.0488*** -0.0606***
(0.0116) (0.0104) (0.0127)
Total Q 0.0216*** 0.0202***
(0.00280) (0.00263)
Total Cash Flow 0.125*** 0.150***
(0.0278) (0.0232) Millennium -1.512 2.112 (97,890) (72,830) Observations 87,626 87,626 87,626 69,374 R-squared 0.606 0.608 0.644 0.658
Firm FE Yes Yes Yes Yes
State-year FE Yes Yes Yes Yes
Industry-year FE Yes Yes Yes Yes
6. Robustness checks
In order to check for robustness, subsamples will be checked. This section will discuss results of tests when the sample is divided into subsamples. First, the full sample as described in section 4 will be divided into the Fama and French five industries. Second, the full sample will be divided into different time periods. Third, a different measure for the change in interest rate will be used. Lastly, some of the adjusted data from section 4 is dropped in order to check for different results. 6.1 Fama and French five industries
When checking for robustness in this section, the full sample as explained in section 4 is used. It is interesting to see the different results when the sample is divided into subsamples based on industry. The five industries used for robustness checks are manufacturing, consumer, health, high-tech and other and the Standard Industrial Classification codes per industry can be found in Appendix 1. The results are shown in table 3.
The results show that the coefficients for intangible intensity, total q and cash flow are significant at a 1% significance level for all industries. Furthermore, the coefficients for the dummy variable and the interaction variable are both insignificant for all industries. These results are in line with the results from table 2, discussed in section 5. It is interesting to see is that the results for the change in interest rate is different for the health industry. At a significance level of 10% the results show that the change in interest rate has an effect on the investment rate, following the classic investment theory. The results show that a base point increase in change in interest rate predicts a 0.00879 decrease of investment rate, ceteris paribus. However, the coefficient for the change in interest rate is insignificant for all other industries in table 3. In conclusion, the results show that the significance of the change in interest rate depends on the industry. Furthermore, when the adjusted R2 of all regressions are compared, the model predicts
Table 3 (Industry Results)
This table shows results when dividing the firms into industry subsamples. The Fama and French five-industry definition is used. Deviation can be found in Appendix A. Unlike in table 2, standard errors (reported in parantheses) are notadjusted by clustered at industry level. Remaining details are the same as in Table 2
Regression (1) (2) (3) (4) (5)
Industry Manufacturing Consumer High-Tech Health Other
Dependent Variable Investment Rate Investment Rate Investment Rate Investment Rate Investment Rate
Delta Interest Rate 0.000502 -0.00113 -0.00337 -0.00879* -0.000122
(0.00122) (0.00118) (0.00258) (0.00532) (0.00224) Interaction Variable -0.000812 0.000899 0.00558 0.0141 -0.00145 (0.00289) (0.00235) (0.00456) (0.00935) (0.00526) Intangible Intensity -0.0635*** -0.0474*** -0.0472*** -0.114*** -0.0412*** (0.0105) (0.00755) (0.00905) (0.0186) (0.0145) Total Q 0.0316*** 0.0173*** 0.0186*** 0.0211*** 0.0231*** (0.00120) (0.000933) (0.000685) (0.00108) (0.00127)
Total Cash Flow 0.162*** 0.248*** 0.117*** 0.0495*** 0.0423***
(0.00694) (0.00611) (0.00514) (0.00915) (0.00828) Millennium -0.656 0.757 3.799 1.139 -0.552 (414,284) (225,372) (362,827) (498,172) (345,742) Observations 23,970 21,467 18,628 7,212 14,501 R-squared 0.622 0.717 0.707 0.675 0.630
Firm FE Yes Yes Yes Yes Yes
State-year FE Yes Yes Yes Yes Yes
Industry-year FE Yes Yes Yes Yes Yes
Standard errors in parentheses; *** p<0.01, ** p<0.05, * p<0.1
6.2 Time period results
In section 5 of this paper results are shown for regressions for the full sample and the sample from 1975-2008. To check for robustness of the model the sample is divided into four time periods. Table 4 shows the results of those regressions. The results show that in the periods from 1975 to 1996 the results are equal to the results found in section 5. Following the results in table 2, the change in interest rate is insignificant. Despite the fact that the change in interest rate also does not predict investment from 1997 to 2006, the coefficient for the interaction variable is significant at a 1% significance level. This means that there is a difference in sensitivity to interest rate for intangible and tangible investments. A base point increase in the change of interest rest will lead to a decrease of the investment rate by 0.0207 multiplied by the intangible intensity,
ceteris paribus. Moreover, the coefficient for the interaction variable is also significant in the period from 2007 to 2016, this time at a 5% significance level. The coefficient for the change in interest rate is significant at a 5% significance level as well. This is interesting, considering the fact that the Federal Reserve has kept the interest equal at almost zero since the financial crisis of 2008. Despite getting significant results for all variables but the dummy variable, regression 4 in table 4 does not have the highest adjusted R2. When comparing the R2 of the regressions we can
see that the model best explains the variance of investment rate in the period from 1975 to 1986.
Table 4 (Period Results)
This table shows results when dividing the sample into differnt subsamples.The sample is divided into 4 periods as shown in each regression. Remaining details are the same as in Table 2
Regression (1) (2) (3) (4)
Time Period 1975-1986 1987-1996 1997-2006 2007-2016
Dependent Variable Investment rate Investment rate Investment rate Investment rate
Delta Interest Rate -0.000335 -0.00490 0.00484 -0.0181**
(0.000666) (0.00345) (0.00570) (0.00722) Interaction Variable 0.000508 0.0106 -0.0207*** 0.0237** (0.00142) (0.00649) (0.00713) (0.00991) Intangible Intensity -0.182*** -0.144*** -0.0562*** -0.0685*** (0.0142) (0.0131) (0.0105) (0.0122) Total Q 0.0173*** 0.0183*** 0.0173*** 0.0181*** (0.00164) (0.00109) (0.000697) (0.000882)
Total Cash Flow 0.359*** 0.187*** 0.0768*** 0.0432***
(0.00898) (0.00717) (0.00511) (0.00537) Millennium 66.03 (3.067e+06) Observations 17,674 18,090 26,508 23,200 R-squared 0.745 0.764 0.707 0.683
Firm FE Yes Yes Yes Yes
State-year FE Yes Yes Yes Yes
Industry-year FE Yes Yes Yes Yes
6.3 Different lags
The regressions in table 2 used an absolute change in interest rate for interest rates of 3 months and 6 months before. However, Hall (1977) discussed that it can take more than a year to design, purchase and install equipment after a change in interest rate. He found evidence that between 10% and 30% of the long-run response of capital occurred within the first year. This paper follows Gutiérrez and Philippon (2016) by introducing different lags. Table 5 presents results for different lags. The delta remains 3 months for all regressions.
Table 5 shows results for lags of 3 months, 6 months, 1 year and 2 years. When increasing the lag from 3 months to 6 months, the coefficient for the change in interest rate becomes significant at a 5% significance level. Moreover, the coefficient for the change in interest rate becomes positive, meaning that an increase in interest rate leads to an increase of investment rate. Despite being significant, this cannot be explained by theory. Other results, despite some small changes in coefficients magnitude remain the same. Also the adjusted R2 increases with only one
percentage point.
The same can be said about a lag of 1 year. The coefficient for the change in interest rate becomes significant at a 5% level and the coefficient remains positive. All other coefficients remain the same, besides some minor changes in magnitude. When using a lag of 2 years, the results show a significant negative effect at a significance level of 1%. Furthermore, the interaction variable is insignificant in regressions for all different lags. This table does not find evidence for a different sensitivity to the change in interest for tangible and intangible investments. Results for all other variables remain the same for all regressions. When we compare the R2 of all regressions in table 5, we see that that they are all more or less equal. This means that
all models explain the same amount of variance in investment rate. However, the results of a lag of 3 months and 2 years are supported by the theory discussed in section 2.
Table 5 (Lag Results)
This table shows results when using different lags for the change in interest rate. The delta remains 3 months as in table 2. Different lags are used, from 6 months to 2 years. Remaining details are the same as in Table 2
Regression (1) (2) (3) (4)
Lag 3 months 6 months 1 year 2 year
Dependent Variable Investment Rate Investment Rate Investment Rate Investment Rate
Delta Interest Rate -0.000234 0.00102** 0.000815** -0.00150***
(0.000740) (0.000420) (0.000380) (0.000567) Interaction Variable 0.000180 0.00141 -0.000175 -0.000876 (0.00155) (0.00105) (0.000960) (0.000888) Intangible Intensity -0.0488*** -0.0480*** -0.0477*** -0.0484*** (0.00463) (0.00467) (0.00470) (0.00470) Total Q 0.0216*** 0.0218*** 0.0217*** 0.0216*** (0.000414) (0.000420) (0.000419) (0.000419)
Total Cash Flow 0.125*** 0.123*** 0.124*** 0.124***
(0.00291) (0.00293) (0.00295) (0.00295) Millennium -1.512 -0.949 0.396 -0.874 (161,159) (116,634) (100,935) (127,324) Observations 87,626 86,081 85,435 85,257 R-squared 0.644 0.645 0.646 0.646
Firm FE Yes Yes Yes Yes
State-year FE Yes Yes Yes Yes
Industry-year FE Yes Yes Yes Yes
Standard errors in parentheses;*** p<0.01, ** p<0.05, * p<0.1
6.4 Data adjustments
As described in section 4, the data is adjusted by following Peters and Taylor (2017). In progress R&D, Selling, General and Administrative expenses and depreciation expenses are set to zero in case of missing values. The reason behind that is that sometimes values for those variables are already allocated into other variables and therefore not reported. This study assumes those missing values to be allocated into other variables. However, the fact that they are not reported does not necessarily mean that they are allocated into other variables. Therefore, another regression is done over the whole sample, but missing values for in progress R&D, SGA and Depreciation expenses are excluded. The results are shown in table 6.
When comparing these results to the results of table 2, we find that the number of observations drop to 36,000. The results change compared to the results in table 2. The
coefficient for the change in interest rate becomes significant, at a 5% significance level. The coefficient for the interaction variable becomes significant as well, at a 10% significance level. The results can be interpreted as follows: a one base point increase in change in interest rate will lead to a 0.0156 decrease in investment rate and a increase in investment rate of 0.0151 times the intangible intensity, ceteris paribus. Furthermore, the coefficients for intangible intensity and total q remain equal to the results in table 2, despite some small changes in magnitude. The coefficient for total cash flow becomes less significant when using the adjusted data. The results of table 6 can be explained by the theory described in section 2. The R2 increases to 69.1%, meaning that
the model with the adjusted data explains the variance in investment rate better than the regression model in table 2.
Table 6 (Adjusted Data Results)
This table shows results when excluding all observations with missing values for In Progress R&D, depreciation expensen and SGA.
Remaining details are the same as in Table 2
Dependent Variable Investment Rate
Delta Interest Rate -0.0156**
(0.00717) Interaction Variable 0.0151* (0.00851) Intangible Intensity -0.0582*** (0.00963) Total Q 0.0185*** (0.00257)
Total Cash Flow 0.0607*
(0.0327) Millennium -2.272 (165,427) Observations 36,000 R-squared 0.691 Firm FE Yes State-year FE Yes Industry-year FE Yes
Robust standard errors in parentheses; *** p<0.01, ** p<0.05, * p<0.1
7. Conclusion
This paper examines whether or not there is a difference in sensitivity to a change in interest rate for tangible and intangible investments. Classic theory suggests that investment is a function of interest rate, but they make no distinction between tangible and intangible investments. Recent literature, however, finds contradicting evidence that interest rate does not explain investment. Some literature does make a distinction between tangible and intangible investments. They find that intangible investments are financed by internal funds and therefore are uncorrelated with the interest rate. Recent investment regressions that make a distinction between tangible and intangible investments focus more on total q and cash flow. This paper finds little evidence that investment is explained by the change in interest rate. More evidence is found for the effect of intangible intensity. However, little to no evidence is found that there is a difference in sensitivity between tangible and intangible investments.
The results in this paper are in line with modern theories on investment on tobin’s q and cash flow. Little evidence is found that investment is explained by the change in interest rate, as described in modern theories. We also find strong evidence that total q and cash flow are determinants of investment. The latter is supported by most recent literature (Alexander and Eberly (2018), Gutiérrez and Philippon (2016), Peters and Taylor (2017).
The results found in this paper differ strongly by using different (sub)samples. Only evidence is found for a significance of the change in interest rate in the health industry. In none of the Fama and French five industries significant results are found for a difference in change in interest rate sensitivity between tangible and intangible investments. However, when the sample is divided over different period, evidence is found that, from 1997 onwards, intangible investment has a different sensitivity to a change in interest rate than tangible investment. However, it must be said that since 2008, the Federal Reserve has kept the interest rate equal, leading to a change in interest rate of zero. Therefore, one needs to be cautious in drawing conclusions. Furthermore,
when different lags in interest rate change are used, somewhat surprising results are found. Unexplained by theory, this study finds evidence that the change in interest rate and investment rate are positively correlated for a lag of 6 months and 1 year. This paper finds no evidence for a different sensitivity of tangible intangible investments when different lags are used. Finally, when different assumptions are made, this paper finds evidence that there is a different sensitivity for intangible investment than for tangible investment.
Overall, this study provides little evidence that intangible investment and tangible investment have a different sensitivity to the change in interest rate. However, not all results provide evidence for a different sensitivity. It is interesting to understand how investment really responds to the interest rate. When using different approaches, different results have been found. Some evidence is found that intangible investment behaves different to a change in interest rate compared to tangible investment. With the increasing amount of intangible investment it is important to understand how intangible investment responds to the market. Institutions can adjust their policy in order to create a more efficient policy. Therefore, future research to intangible investment is needed, by using different assumptions, samples and time period. Recent periods seem to behave different than periods before. Thus, it is important to learn more about the behaviour of intangible investment in recent time frame. Since theory suggests that intangible investments are financed by internal funds, it is interesting to test whether or not intangible investments has a different sensitivity to cash flow than tangible investments. This study could be enhanced further by the inclusion of hurdle rates.
Reference list
Alexander, L., & Eberly, J. (2018). Investment Hollowing Out. IMF Economic Review , 5-30.
Almeida, H., & Campello, M. (2007). Financial constraints, asset tangibility, and corporate investment.
Review of Financial Studies , 20 (5), 1429-1460.
Brown, J. R., Fazzari, S. M., & Petersen, B. C. (2009). Financing External Innovation and Growth: Cash Flow, Equity, and the 1990s R&D Boom. The Journal of Finance , 64 (1), 151-185.
Corrado, C., & Hulten, C. (2010). How do you measure a technological revolution . American Economic
Review .
Corrado, C., Haskel, J., Jona-Lasinio, C., & Iommi, M. (2012). Intangible Capital and Growth in Advanced Economies: Measurement Methods and Comparative Results. WORKING PAPER .
Corrado, C., Hulten, C., & Sichel, D. (2009). INTANGIBLE CAPITAL AND U.S. ECONOMIC GROWTH. Review of Economic and Wealth .
Döttling, R., & Perotti, E. (2017). Secular Trends and Technological Progress. Working Paper , 1-12. Döttling, R., Gutiérrez, G., & Philippon, T. (2017). Is There an Investment Gap in Advanced Economies? If So, Why? Working Paper , 1-44.
Döttling, R., Ladika, T., & Perotti, E. (2016). The (Self-)Funding of Intangibles Robin. Tinbergen Institute
Discussion Paper, No. 16-093/IV , 1-27.
Eisfeldt, A. L., & Papanikolaou, D. (2014). The Value and Ownership of Intangible Capital. The American
Economic Review , 104 (5), 189-194.
Erickson, T., & Whited, T. M. (2000). Measurement Error and the Relationship between Investment and q. Journal of Polical Economy .
Erickson, T., & Whited, T. M. (2012). Treating measurement error in Tobin's q. Review of Financial Studies ,
25 (4), 1286-1329.
Falato, A., Kadyrzhanova, D., & Sim, J. W. (2012). Rising Intangible Capital, Shrinking Debt Capacity, and the US Corporate Savings Glut. SSRN Electronic Journal , 1-55.
Gutierrez, G., & Philippon, T. (2016). Investment-less growth: an empirical investigation. NBER Working
Hall, R. E. (1977). Investment, Interest Rates, and the Efects of Stabilization Policies. Brooking Papers on
Economic Activity .
Hall, R. E. (2015). Quantifying the Lasting Harm to the US Economy from the Financial Crisis. NBER
Macroeconomics Annual , 29 (1), 71-128.
Hart, O., & Moore, J. (1994). A Theory of Debt Based on the Inalienability of Human Capital. The
Quarterly Journal of Economics , 109 (4), 841-879.
Hayashi, F. (1982). Tobin ' s Marginal q and Average q : A Neoclassical Interpretation. Econometrica , 50 (1), 213-224.
Hicks, J. (1981). "IS-LM": An Explanation. The Journal of Post Keynesian Economics , 3 (2), 139-154.
Hicks, J. (1973). Mr . Keynes and the " Classics "; A Suggested Interpretation. Econometrica , 5 (2), 147-159. Hulten, C. R., & Hao, X. (2008). What is a company really worth? Intangible Capital and the "Market to Book value" Puzzle. Working Paper 14548 , 1-36.
Jones, C., & Philippon, T. (2016). The Secular Stagnation of Investment? NBER Working Paper , 1-25. Jorgenson, D. W. (1963). Capital Theory and Investment Behavior. The American Economic Review , 53 (2), 247-259.
Lev, B., & Radhakrishnan, S. (2015). The Valuation of Organization Capital. In C. Corrado, J. Haltiwanger, & D. Sichel, Measuring Capital in the New Economy (pp. 73-110). Chicago: University of Chicago Press.
Li, W. C. (2012). Depreciation of Business R&D Capital. U.S. Bureau of Economic Analysis.
Peters, R. H., & Taylor, L. A. (2017). Intangible Capital and the Investment- q Relation. Journal of Financial
Economics , 123, 251-272.
Sharpe, S. A., & Suarez, G. A. (2013). he Insensitivity of Investment to Interest Rates: Evidence from a Survey of CFOs. SSRN Electronic Journal , 1-40.
Zenner, M., Junek, E., & Chivukula, R. (2014). Bridging the Gap between Interest Rates and Investments.
Appendix
Appendix 1. WRDS variables list
Appendix 1: WRDS Variables
Database WRDS code Variable
Compustat capx Capital Expenditures
cogs Cost of Goods Sold
dp Depreciation Expenses
fyear Fiscal Year
ib Income Before Extraordinary Items
naics North American Industry Classification Code
ppegt Property, Plant and Equipment
rdip In Progress R&D Expenses
sic Standard Industry Classification Code
state State/Province
xrd Research and Development Expenses
xsga Selling, General and Administrative Expenses
Peters and Taylor Total Q k_int Replacement Cost of Firms Intangible Capital
q_tot Total Q
Appendix 2. Fama and French 5 industry codes
1. Consumer: Consumer Durables, Non Durables, Wholesale, Retail, and Some Services (Laundries, Repair Shops)
0100-0999 2000-2399 2700-2749 2770-2799 3100-3199 3940-3989 2500-2519 2590-2599 3630-3659 3710-3711 3714-3714 3716-3716 3750-3751 3792-3792 3900-3939 3990-3999 5000-5999 7200-7299 7600-7699
2. Manufacturing: Manufacturing, Energy, and Utilities 2520-2589 2600-2699 2750-2769 2800-2829 2840-2899 3000-3099 3200-3569 3580-3629 3700-3709 3712-3713 3715-3715 3717-3749 3752-3791 3793-3799 3830-3839 3860-3899 1200-1399 2900-2999 4900-4949
