1
Master thesis for MSc Finance
Market valuation of tangible and intangible
expenditures
Author:
Supervisor:
Jan Willem van der Graaf
dr. ing. N. Brunia
Student ID:
Date:
2347032
07-06-2017
Abstract:
2
1
.Introduction
If you were to ask a first year finance student: ‘What is Coca Cola’s biggest intangible asset?’ they will probably respond with brand name. It is clearly not its syrup recipe, as Coca Cola tastes different in various countries. If we would ask the same question with regards to Apple, the answer would likely be related to technology or styling. And from an economic perspective these answers makes sense. Coco Cola’s investments into maintaining and increasing its brand name through advertising are crucial to the success of Coca Cola. And Apple’s investments in research and development to create new innovative products are just as important for Apple. However, in reality balance sheets are based on accounting rules and the majority of intangible assets on balance sheets are made up by goodwill as these investments in other intangible assets are often not capitalized. Most of the time these investments are reported as a yearly expense and no further recognition of their value is made by accountants. It is up to the market to accurately price these investments into the stock price of a firm based on the information available on those investments.
In efficient stock markets we would assume the aforementioned investments are correctly priced in the current stock price of a firm. However, previous research has shown many cases of excess returns achievable with investment strategies based on certain intangible firm characteristic; Chan et al.(2001) for R&D and advertising expenses, Deng et al.(1999) for patent citations, Edmans (2013) for employee satisfaction, Gompers et al. (2003) for corporate governance, and Aboody and Lev (1998) for software development expenses.
3
unexpected profits. While if a company engages in a little R&D, for example Coca Cola to create a new piece of software for internal use, it has little relevance for future profitability and thus the potential undervaluation does not matter as much. Those differences in value relevance between intensity levels highlight the potential importance of looking at the intensity of an expense and not just in a binary way if a firm reports an intangible expense or not. This thesis will add to the literature by extending the research on the market valuation of intangible expenses by looking at the importance of a firm’s intensity of investments in intangible assets. The main hypothesis of this thesis is that the market undervalues firms making relatively big investments in intangibles.
The second hypothesis of this thesis is that intangibles are more difficult to value than intangibles. That this is the case is often posed in the literature, such as Chan et al. (2001) and Edmans (2013), but we have found no formal test for this statement. To test for this we will look at portfolio returns based on the tangible intensity measures of capital expenditures(capex) and change in net working capital and contrast those findings with the findings of the intangible portfolios based on R&D and advertising. This contrast also provides insight on the impact of accounting treatment of expenses on the market valuation of those expenses, as capex is capitalized and R&D and advertising are not.
This thesis examines the relation between investments in tangible (net working capital, and property plant and equipment) and intangible (research and development, and advertising) assets and long-run stock returns. Data for these investments and other variables are sourced from Compustat for available North American stocks in the period of 1975-2017. Two yearly intensity measures of these expenses are created by dividing each of the expenditures of a firm by the market value of that firm and the sales of that year. Based on those intensity measures one-year lagged equal-weighted portfolios are constructed and yearly rebalanced by grouping firms together into quintiles, i.e. the lowest 20% intensity of a measure are grouped in quintile 1 and the highest 20% are grouped in quintile 5. Based on the returns for each portfolio alphas are calculated using Fama French's five-factor model which controls for sensitivities of the returns to market risk, size, book-to-market, profitability, and investment factors.
4
et al. (2001). Alphas for the capex portfolios show a similar pattern of increasing alphas with increasing intensity of capex, with a long short portfolio leading to a statistically significant alpha of 8.08%. That the capex alphas show a similar pattern as the alphas based on advertising and R&D draws into question the assertion that intangibles are more difficult to value than tangibles. Alphas for ΔNWC show no clear pattern and the alpha of the long short portfolio is not statistically significant. This implies that either ΔNWC is accurately priced in the by the market or ΔNWC is not very value relevant. These findings contribute to two main strand of research: The equity market valuation of intensive investments in certain tangible and intangible assets; and the discussion on the difference in difficulty in valuing tangibles versus intangibles. For investors these results indicate an opportunity for excess returns and for firms a possible avenue of better external communication to prevent undervaluation.
The contents of this paper are as organized as follows. Section 2 discusses the theoretical motivation and relevant literature for the hypotheses of this thesis. Section 3 discusses and documents the data and methodology. Section 4 presents and discusses the main results. Section 5 concludes.
2.
Theoretical motivation
For the main hypothesis of this thesis, that the market undervalues firms making relatively big investments in intangibles, to be plausible it is required that the market is capable of mispricing a firm. Section 2.1 discusses why the market might not be perfectly efficient and thus incorrectly value a firm. Section 2.2 zooms in on the potential for mispricing of investments in intangible assets in particular, which forms the basis for our second hypothesis. Section 2.3 discusses why the intensity of investments in intangibles matters.
2.1 Can the market be wrong?
5
approximates the market portfolio. This line of reasoning has contributed to the rise of passively managed index funds which safe money by eliminating active managers and transaction fees.
There are also arguments to be made against the efficient market hypothesis. One such argument is that certain investors, such as Warren Buffet, and portfolio managers have a consistent track record of beating the market. If performance is completely random how can there be people and institutions consistently profiting from the market? Another argument is the evidence for some consistent patterns such as the January and weekend effect, which indicate recurring higher stock returns in January than in other months (Keim, 1982) and lower stock returns on Mondays respectively (Jaffe and Westerfield, 1985). Even in efficient markets there a period of inefficiency as the market incorporate new data such as the existence of aforementioned patterns. The efficient market hypothesis also assumes that all investors in a market value all available information in the same way. If one investor is looking at firms to find undervalued stocks whereas another investor is valuing firms based on their growth potential it is likely that these different perspective will lead to different assessments of the fair market value of those firms. There is not a universal method of valuing stocks that requires no subjective input. Investors are also subject to a great many biases that influence their investing behavior and valuations. A few examples of biases whose names speak for themselves are: the confirmation bias, loss aversion bias, and overconfidence bias.
While many of the arguments above are not without rebuttal they do illustrate that perfectly efficient markets are an unlikely occurrence. So even though one can debate over the degree of inefficiency in markets, it is almost a certainty that some degree of inefficiency and thus over-and-under valued firms exist.
2.2 The market valuation of intangibles
In this section we zoom in on why specifically firms who invest in intangible assets are likely to be undervalued.
6
value mainly through intangible assets instead of tangibles ones. Hall (2001) estimates the total value of intangible capital to be ranging between half to two-thirds of the total market value of publicly traded firms, as measured by the q ratio (market value to replacement cost of physical assets). Fashion companies such as Louis Vuitton are dependent on the power and perceived quality of their brand. Innovative technology companies such as Google try to attract the best employees to create the best products. In the car industry Toyota employs a lean manufacturing philosophy to create an edge. The previous examples are there to illustrate just a few ways in which intangible assets can play a key role in the value creation process of a firm. And according to research these intangibles are more difficult to value than tangibles. Higgins (2016) finds that in valuations done by analysts (and extrapolative models) forecasts errors of earnings are positively associated with intangibles. Providing proof that analysts have a harder time valuing intangibles than tangibles
Existing research into a number of intangible assets finds opportunities to earn risk adjusted excess returns by creating investments strategies based on these intangible assets. Chan et al. (2001) finds that on average firms that engage in R&D earn the same returns as firms that do not engage in R&D. Implying that the market accurately prices value creation through R&D investments. However, when only looking at firms that engage in R&D they find that firms with a higher R&D intensity earn higher excess returns, up to 6.1% for an equal-weighted portfolio consisting of firm in the top quintile of R&D intensity. These firms with the highest intensity also outperform the firms that do not engage in R&D, with a spread of 5.86% between the highest intensity portfolio and the non-R&D portfolio. They find the same patterns for advertising expenses. Edmans (2011) created value weighted and equal weighted portfolios consisting of companies from the Best Companies to work for list, which is used a proxy for employee satisfaction. He finds that this value weighted portfolio earns annual excess return of 3.5%. Apart from R&D, advertising, and employee satisfaction there also evidence for mispricing of several other intangible firm characteristics: patent citations (Deng, 1999), software development costs (Aboody, 1998), and several corporate social responsibility measures (Gregory 2014). Kempf (2007) find portfolios based on social responsible investments criteria using KLD ratings lead to abnormal excess returns of up to 9% per year.
7
produce a tangible result, such as higher than expected profits, and then the market readjusts its low valuation to a higher one leading to the excess returns. Edmans (2013) tests this hypothesis by looking at positive earnings surprises and the stock price reaction to those earning announcements. He finds that firms with high employee satisfaction have statistically significant higher earnings surprises than all others firms at the 1% level. The effect of the earnings surprises on the stock prices of the firms reporting those earning surprises account for a significant portion, 1.4%, of the 2.6% yearly alpha for the equal-weighted portfolio. His results indicate that as soon as an intangible asset produces a tangible result (the positive earnings surprise) the market adjusts its initial low valuation of the intangibles to a higher level (the positive stock price reaction) which causes the excess returns.
A different explanation for the excess returns of the intangibles is that they are the result of a risk premium firms get for employing these riskier intangible assets as opposed to more traditional ways of value creation. However, this explanation is refuted by papers such as Brancha (2010) who find that while there is a risk premium, it is often not adequately priced and thus there is room to exploit. The same goes for Duqi (2013) who find no evidence of different risk profiles in different levels of R&D but do find excess return for high R&D firms in specific sectors.
2.3 Importance of the intensity of investments in intangibles
8
intensity levels of expenses and not the average of all firms engaging in an intangible expense and the firms which who do not.
3.
Data, descriptive statistics, and methodology
Section 3.1 discusses the data we will use and provides descriptive statistics of the key variables. In section 3.2 the methodology is discussed.
3.1 Data sourcing and construction 3.1.1 Data
The main data source is the North American Compustat Capital IQ database for fundamentals. Data for all North American US dollar listed stocks in the period of 1975-2017 is sourced from Compustat. The North American database is the largest and most complete data set. Previous studies such as those of Chan and Edmans also used North American stocks and thus using North American stocks allows for a better comparison of the results. Data items are taken based on calendar year basis, with price, shares outstanding, and cumulative adjustment factor being end of calendar year figures. Table 1 presents an overview of all the sourced variables from Compustat, their respective Compustat codes, and their symbols used in this thesis.
Table 1: Overview of sourced variables
Variable Compustat code Symbol
Research and development expenditure
XRD R&D
Advertising expenditure XAD
Capital expenditure CAPX Capex
Change in net working capital* WCAPC ΔNWC
Sales SALE
Stock price PRCC_C P
Cumulative Adjustment factor ADJEX_C A
Dividends DVT D
Shares outstanding CSHO
9
The change in net working capital measure excludes changes in cash, cash equivalents, and debt. This thesis aims to test if intangibles are more difficult to value than tangibles. The impact of changes in cash, cash equivalents, and debt on the value of a firm are relatively easy to value. By excluding these relatively easy to value changes we avoid the possibility that these easy changes are the primary cause for the results.
The cumulative adjustment factor is a ratio that is required to adjust per-share data (price, dividends) and share data (shares outstanding) for all stock splits and stock dividends that occur subsequent to the end of a given period.
3.1.2 Removal of observations
The raw dataset of all stocks contains 411,967 data points and 32,042 different companies, where one data point is 1 year of the sourced variables for a unique company. The final dataset where returns can be calculated for every year consists of 225,126 data points and 24,594 different companies.
This reduction is caused by removing years for which no return or market value can be calculated. The majority of these removals are due to no stock price being available, which happened in 75,909 cases. As one year of a stock price is needed to calculate returns for two years the amount of data lost due to this is bigger than the 75,909. Due to rounding there are some years where adjustment factors and shares outstanding are zero, these years have been removed. Years with no available shares outstanding or adjustment factor are also removed.
3.1.3 Calculating stock returns, market capitalization and intensity measures
Total yearly returns are calculated based on the following formula:
Total Stock Return=(Pt∗At−P(t−1)∗A(t −1)+Dt)
(Pt∗A(t −1))
(1)
where Pt is the stock price at the end of the year and P(t-1) the stock price at the start. Dt is the
dividend per share. Stock prices are adjusted for stock splits and stock dividends using adjustment factors At.
10
For each company we calculate yearly intensity measures for R&D, advertising, capex, and change in net working capital expenses by dividing those expenses by the market capitalization of the firm. As a robustness test we also create those intensity measures by dividing the expenses by sales.
3.1.4 Portfolio creation
Based on the calculated intensity measures one-year lagged equal-weighted portfolios are constructed and yearly rebalanced by grouping firms together into quintiles, i.e. the lowest 20% of a measure are grouped in quintile 1 and the highest 20% are grouped in quintile 5. Each quintile of firms is one portfolio.
Edmans (2011) found marginal differences in alphas when using value or equal weighted portfolios. Since this difference appears to be negligible we will solely focus on equal-weighted portfolios which is in line with the methodology of Chan et al. (2001) who only use equal-weighted portfolios.
3.1.5 Descriptive statistics
Table 2 presents the descriptive statistics of the key variables and constructed intensity measures.
Table 2: Descriptive statistics
Variables # Obs Mean Median Std.Dev. Market Cap ($ mln) 225,126 2,094 78.036 12,950 Share Price ($) 225,126 17.153 9.375 79.196 Return (%) 225,126 14.131 5.656 65.413 Intensity measures R&D 86,411 0.042 0.784 91.269 Advertising 70,528 0.024 0.697 87.589 Capex 191,692 0.058 0.979 115.882 ΔNWC 50,018 0.026 -0.014 4.139
11
The number of observations for the key variables is the same for each variable as any year for which these numbers cannot be calculated is removed. The number of observations for the intensity measures reports how many total years each intensity measure is calculated. The capex intensity measure is available almost every year while the intensity measures of advertising and R&D are less commonly available. As almost all companies have capital expenditures and not nearly as many have R&D or advertising expenses this difference between observations is to be expected. The ΔNWC data ranges from 1975-2003 hence the lower number of observations.
Descriptive statistics for the intensity measures based on sales are included in table A.1 in appendix A. They resemble the statistics of table 2.
Table 3 provides an overview of how many years on average each intensity measure is calculated for each firm. The mean of 9.161 for R&D, for example, means that the average number of years an R&D intensity measure is calculated is 9.161 years. Implying that if an R&D intensity measure can be calculated for a firm at least once that firm is on average placed in one of the 5 R&D intensity portfolios 9.16 times.
Table 3: Descriptive statistics of the average number of years intensity measures are calculated
# Obs Mean Median Std. Dev. Min Max
R&D 86,411 9.161 6 8.927 1 42
Advertising 70,528 7.061 5 6.884 1 42
Capex 191,692 9.768 7 9.318 1 42
ΔNWC 50,018 6.099 5 4.223 1 20
The number of observations, the mean, median, standard deviation, minimum and maximum amount of years intensity measures are calculated for each intensity measures based on market capitalization for research and development expense, advertising expense ,and change in net working capital.
3.1.6 The possible impact of mean reversion of price earnings ratios
12
these portfolios might be influenced by this mean reversion of P/E ratios. If a firm has a low market value because of a low P/E ratio then the intensity measures of those firms will be higher than those of a firm who have a normal or high P/E ratio with the same amount of expenses. This would cause firms with a low P/E ratio to be over-represented in the highest intensity portfolios. Then if because of mean reversion of P/E ratios the market value of those firms increases to a normal or high level market value of these firms increases which is caused by an increase in price and thus return of that stock, which could influence the returns of the highest intensity portfolio.
If mean reversion of P/E ratios would have a big impact on what firms end up in 5th quintile the actual years spend in that quintile should be lower than the expected years spend in there. Firms with a low P/E value and thus relative low market value are skewed towards the higher intensity portfolios. In other words, everyone gets a turn to be in the top because of a low P/E ratio.
To check if the alphas are impacted by this mean reversion we take a look at the average number of years firms spend in the 5th quintile of intensity and compare those to the expected number of years a firm should be in in the highest intensity portfolio based on chance.
Table 6: Average numbers of years spend in the 5th quintile for each intensity measure
Years in all quintiles Expected years in Q5 Actual years in Q5 R&D 9.161 1.832 3.991 Advertising 7.061 1.412 4.296 Capex 9.768 1.954 4.209 ΔNWC 6.099 1.219 2.093
Years in all quintiles denote the average number of years a firm reports an expense and is thus placed in one of the quintiles. Expected years in Q5 denotes the average number of years every company reporting an expense should spend in the 5th quintile if years spend there are purely random, calculated by taking 20% of the years in all quintiles number. Actual years in Q5 reports the actual average number of years each firm reporting an expense is placed in the 5th quintile if they spend at least 1 year there.
13
that if a firm is in the highest intensity portfolio at least one year that firm is placed in the highest intensity portfolio for an average of 4 years while chance would dictate 1.8 years. As this actual number of years is much higher than the expect number of years it is likely mean reversion plays only a small role in determining what firms are placed in the highest intensity portfolios.
More in depth descriptive statistics on the number of years each firm is in the highest intensity portfolio are reported in appendix D.
3.2 Methodology
Our main hypothesis is that the market undervalues firms making relatively big investments in intangibles. To test this we look at the risk adjusted returns of long short portfolios going long in the highest intensity firms and short in the lowest intensity firms. Section 3.2.1 describes how we calculate the risk adjusted excess returns for each of the quintile portfolios for each intensity measure. Section 3.2.2 describes how we calculate the risk adjusted returns for the long short portfolios going long in the highest intensity portfolio and short in the lowest intensity portfolio. These risk adjusted returns will also be used to test our second hypothesis, that intensity of intangible investments is harder to value than intensity of tangible investments. Section 3.2.3 and 3.2.4 describe additional robustness tests.
3.2.1 Risk adjusted returns for the quintile portfolios
We will calculate risk adjusted excess return (alphas) for the quintile portfolios based on the latest five-factor model of Fama and French (2015). We use the Fama and French five-facor model as this model outperforms the three-factor model in explaining stock returns. Using the five-factor model to calculate risk adjusted abnormal returns is a slight deviation from the methodology of Chan and Edmans, who use the three-factor Fama French model, and the four-factor Carhart model respectively. The choice of model is just a choice of a benchmark to compare the returns of the portfolios in a study. As we apply the same model to all our returns the choice of model has marginal impact on the results.
Specifically this means I run time series regressions in the form of:
Rpt−Rft=αp+βmkt (Rmt−Rft)+βsmbSMBt+βhml HMLt+βrmw RMWt+β cmaCMAt+Ɛpt(2)
for each quintile portfolio p, where:
Rpt - Rft is the yearly return on a portfolio in excess of the risk free rate, the Treasury bill rate, in
14
Rmt - Rft is the yearly excess return on the value-weighted market index.
SMBt is the yearly return on the Fama-French (2015) factor-mimicking portfolio for size (average
return of nine small stock portfolios minus the average return on the nine big stock portfolios).
HMLt is the yearly return on the Fama-French (2015) factor mimicking portfolio for book to market
ratio (the average return on the two value portfolios minus the average return on the two growth portfolios).
RMWt is the return on the Fama-French (2015) factor mimicking portfolio for operating profitability
(the average return on the two robust operating profitability portfolios minus the average return on the two weak operating profitability portfolios).
CMAt is the return on the Fama-French (2015) factor mimicking portfolio for investment (the
average return on the two conservative investment portfolios minus the average return on the two aggressive investment portfolios).
αpis an intercept that measures the abnormal risk-adjusted returns.
The risk free rate, the return on the market and the SMBt, HMLt, RMWt, and RMWt factors are taken
from Ken French’s website.
Standard Errors are calculated using Newey-West (1987) to allow Ɛptto be serially correlated and heteroskedastic. If intensive investments in intangibles are indeed undervalued we expect to find that higher intensity portfolios produce higher alphas than the lower intensity portfolios.
3.2.2 Risk adjusted returns for the long short portfolios for each intensity measure
To test if higher intensity measures lead to higher excess returns we create a long short portfolio where we go long in the highest intensity portfolio and short in the lowest intensity portfolio for each intensity measure.
Specifically this means we run the following time series regressions:
Rpht−Rplt=αp+ βmkt (Rmt−Rft)+βsmb SMBt+βhml HMLt+βrmw RMW111t+ β cma CMAt+Ɛpt(3)
¿
Rpht - Rplt is the yearly return on the highest intensity portfolio minus the yearly return on the
lowest intensity portfolio. Since we create a long short portfolio which can be seen as a zero cost investment we do not subtract the risk free rate from the portfolio returns. The other specifications are the same as those in section 3.2.1.
15
returns would imply firms making intensive investments in those intangibles are undervalued by the market. We expect the alphas for the ΔNWC and capex portfolios to be closer to zero than the alphas for R&D and advertising, indicating the market values intensive investments in tangibles closer to their true value.
3.2.3 Test for significance of the differences in alphas between long short portfolios
The alphas based on the methodology in section 3.2.2 gives us a rough idea of what intensive investment is mispriced the most. The alpha closest to zero indicates the lowest achievable excess returns and thus the most accurate pricing for that expenditure. The alpha furthest away from zero indicates the highest achievable risk adjusted excess return and thus the most mispriced expenditure.
As a further robustness test to test for the significance of the differences between the alphas based on different expenditures we create a long short portfolio where we go long in the long short portfolio based on one expense and short in a long short portfolio based on a different expense. This means we run the following regressions:
Rpht−Rplt=αp+ βmkt (Rmt−Rft)+βsmb SMBt+βhml HMLt+βrmw RMW111t+ β cma CMAt+Ɛpt(4)
¿
Where Rpht denotes the yearly return on the long short portfolio based on an expense measure and Rplt is the yearly return of the long short portfolio based on a different expense measure. The other
specifications are the same as those in section 3.2.2.
3.2.4 Gibbons Ross Shanken test for joint significance
As robustness test for the joint significance of the alphas of the portfolios based on an expenditure measure we follow the procedure of Gibbons et al. (1989) to calculate the Gibbons Ross Shanken (GRS) statistic. The GRS test tests if the estimated intercepts from a multiple regression model are jointly zero.
16
4.
Results and discussion
In section 4.1 and 4.2 we discuss the results and their implications for our hypotheses. Section 4.3 discusses the results where intensity measures are based on sales. Section 4.4 discusses the results for when restricting the time period to 2002-2016.
4.1 Are intensive investments in intangible assets initially undervalued by the market?
Table 4 presents the core results of this thesis. The expansive results of each regression for the different portfolios with intensity measures based on market value are reported in appendix B.
Table 4: Risk adjusted yearly alphas for portfolios based on the different expenditure intensity measures relative to market value
Portfolio R&D Advertising Capex ΔNWC 1Low -1.975 -3.321* -4.439** 2.132 2 0.562 1.035 1.434 2.984 3 1.759 3.267 2.277 6.579 4 3.467* 4.742* 4.468** 5.063 5High 4.836* 7.913*** 3.639 4.250 High-low 6.811** 11.235*** 8.078*** 2.118
Estimated five-factor Fama French regressions for yearly returns (in percent) on portfolios based on research and development, advertising, capital expenditure and change in net working capital. Quintiles are based on increasing intensity, with 1 having the lowest intensity and 5 the highest. Excess returns are relative to the risk-free rate. P-values smaller than 0.01, 0.05 and 0.10 are indicated by ***, ** and *, respectively.
17
lowest intensity portfolio. These results confirm our main hypothesis that intensive investments in intangible asset are initially undervalued by the market.
The results of the Gibbons Ross Shanken test for the joint significance of alphas of the 5 different intensity portfolios based on an intensity measure are reported in table 5.
Table 5: Results of the Gibbons Ross Shanken test
GRS statistic P-Value
R&D 1.476 0.225
Advertising 2.864 0.030
Capex 13.133 0.000
ΔNWC 0.317 0.897
Results for the Gibnons Ross Shanken test which tests if the alphas for an intensity measure based on market value are jointly zero.
The results indicate that the chance that alphas of the portfolios based on advertising intensity measures are jointly 0 is below 5%. The p-value for R&D is relatively high compared to the result for advertising. However, as the R&D long short portfolio returns a high alpha which is statistically significant at the 5% level we feel that a p-value of 0.225 is not high enough to make us question the validity of the results of the intensity portfolios.
4.2 Are intangibles more difficult to value than tangibles?
18
Portfolio 3 generates the highest alpha of 6.58% and there are decreasing alphas for the highest and lowest intensity portfolio. Furthermore is the alpha of the long short portfolio not statistically significant and the lowest of all alphas. One likely explanation is that changes in net working capital are accurately priced in by the market as we hypothesized. A different explanation that could explain the difference between the alphas of the portfolios based on capex and change in net working capital is that net working capital is not very value relevant while capex is. Changes in net working capital are more likely to be the result of a company conducting its business instead of a firm actively increasing or decreasing its working capital to impact revenue creation which is not, or less so, the case for capex.
The results of the GRS test presented in table 5 with regards to the ΔNWC portfolios are in line with expectations. The result indicates that there is roughly a 90% chance that the alphas should jointly be 0, providing further evidence that ΔNWC is accurately valued by the market or that ΔNWC is not very value relevant. The results for capex are not in line with expectations; we expected no achievable excess risk adjusted returns with an investment strategy based on capex intensity. The result of the GRS test confirms that the alphas of the capex intensity portfolios are jointly different from 0 and thus provides further validity to the alpha generated by the long short portfolio.
We find that the differences between the alphas of the different long short portfolios are not significant when using the methodology described in section 3.2.3. For every possible combination, i.e. long in the long short R&D portfolio and short in the long short ΔNWC portfolio, we find P values above 0.26. This indicates that our hypothesis that investments in tangible assets are more accurately valued by the market intensive than intensive investments in intangible assets is likely to be wrong. And it is probably more accurate to say that intensive investments in value relevant assets are undervalued.
4.3 Portfolios based on intensity measures using sales
The results of the portfolios based on sales are reported in appendix C.
19
alpha in all three cases, even showing negative alphas for the portfolios based on R&D and capex. Portfolio 5 of R&D, for instance, generates a yearly alpha of -5.31% while portfolio 4 generates an alpha of 5.52% alpha which implies a spread of 10.83%.
This pattern of portfolios 1 to 4 generating increasing alphas with a sharp drop in the alpha of portfolio 5 is also seen in the results of Chan et al (2001) when they use sales instead of market value to calculate their intensity measure. This difference between the alphas of portfolio 4 and 5 when using sales to calculate intensity of investments is not elaborated on by Chan et al. We think that this difference is caused by firms which have very little to no sales and some R&D expenses, which would put them in the highest intensity portfolio. It is easy to imagine how all companies which have close to no sales but which do engage in R&D, such as biotechnology companies for example, will be placed in portfolio 5. Chan et al. find that this decrease in alphas between portfolio 4 and 5 nearly disappears when they wait longer with portfolio formation, i.e. use 3 year lagged intensity measures of sales instead of 1 year lagged intensity measures. We imagine companies which have no sales for several years but invest a lot in R&D projects which might have big potential payoffs later on heavily skew results based on one year lagged intensity measures. As we use one year lagged measures we consider the results based on intensity measures based on market value leading.
4.4 Results for 2002-2017
To test whether the alphas of table 4 are also achievable when limiting ourselves to a more recent time period we have chosen look at alphas from long short portfolios over the period 2002-2017. We chose this period as Chan et al. published their results in 2001, giving the market time to incorporate their findings. When restricting our sample to this period we find yearly alphas of 1.27%, 4.35%, and 6.77% for long short portfolios based on R&D, capex, and advertising respectively. None of these alphas are significant at even the 10% level; however the sample size of portfolio returns is only 15 years when restricted to this time period. This small sample size increases the variability of the returns and thus lower p-values are to be expected.
20
would be that the market has incorporated much of the research done on R&D, advertising, and capex investments and adjusted the initially to low valuation of these investments to a more appropriate level. As the p-values are rather high we feel the differences between the decreases in alphas are not clear enough for us to speculate on.
6.
Conclusion
In this conclusion we start by addressing our main hypotheses in section 6.1 and 6.2 Section 6.3 discusses practical implications and recommendations. In section 6.4 we discuss the most important limitations of this thesis.
6.1 The market valuation of relatively intensive investments in intangible assets
This thesis finds that long short portfolios, going long in the highest intensity firms and short in the lowest intensity firms, based on investments in R&D and advertising generate statistically significant yearly alphas of 6.81% and 11.24% respectively. These results indicate that risk adjusted excess returns are achievable with an investment strategy focusing on firms making relatively large investments in intangible assets. We believe that these alphas are the result of the market readjusting its initial undervaluation of investments in intangible assets as soon as they produce tangible results.
The results of this study are in line with the findings of Chan et al. (2001). The alphas generated by portfolios based on advertising and R&D expenses are roughly the same even though the time period of the data used by Chan and this thesis differ substantially, 1995 for Chan and 1975-2017 for this thesis.
When restricting our sample size to the period 2002-2017 we find much lower but still positive alphas generated by the long short portfolios. To us this indicates that the market is getting more efficient in incorporating investments of firms in intangible assets in their stock price at their fair value. However, due to the low sample size of that time period the accuracy of those alphas is limited. making it hard to say how much more efficient the market has become.
6.2 Are investments in intangibles more difficult to value than investments in tangibles?
21
in intangible ones we would not find this alpha which is higher than the alpha generated by the long short portfolio based on R&D. The differences between the alphas of the long short portfolios based on the different expenditures are also not statistically significant indicating that relatively intensive investments in in tangibles are not valued more efficiently by the market than investments in intangible assets.
Contrary to the alpha generated by the long short portfolio based on capex intensity we find that a long short portfolio based on the intensity of changes in net working capital generates a low not statistically significant alpha. Indicating changes in net working capital are accurately priced in by the market or that changes in net working capital are not very value relevant. This is confirmed by the results of a Gibbons Ross Shanken test which indicates that the alphas generated by the change in net working capital portfolios are jointly very likely to be 0%.
The differences between the alphas of the portfolios based on capex and change in net working capital show that some investments in tangible assets are valued more efficiently by the market than investments in intangible ones but surely not all. It is more appropriate to say that market undervalues relatively big investments in value relevant assets.
6.3 Practical implications and recommendations
We believe our results added to the literature of the market valuation of certain expenditures. We especially build upon the initial research of Chan et al. (2001) regarding the importance of intensity of investments. Furthermore we found we this importance of intensity is not just limited to investments in intangible assets, as our results indicate a similar importance for capex intensity. Further research into the specific size of these effects in a more recent time period would be interesting as our initial inquiry is not accurate due to the limited sample size due to the use of yearly data. There is also room for research to see if other investments, such as employee training expenditures or public relations expenditures, produce similar results.
22 high intensity levels may continue for a while longer.
For firms making relative big investments in value relevant assets we believe it would be beneficial if they communicated the potential future benefits of those investments more clearly. Our results indicate that those relative intensive investments are initially undervalued by the market. Better communication and reporting, especially for long term investments that are expensed instead of capitalized, might prevent the market from underpricing the potential benefits of those investments. This recommendation of course also goes for managers who are considering making a big investment in a long term asset and have to report to their superiors or the shareholders. We acknowledge that estimating and giving a clear picture of uncertain benefits of investing in, for example, employee satisfaction might not be easy to do. But surely is better to give a slightly incorrect picture and try to convey the value of that investment than to expense the investment and know it is going to be undervalued.
6.4 Limitations
We do not include transaction costs, which would decrease any potential excess returns if this strategy were to be implemented in real life as frequent rebalancing of the portfolios is required.
We have not used log returns as arithmetic returns aggregate better across portfolios. This might have implications for the normality of the distribution and time-additivity of the returns.
Returns are based on equal-weighted portfolios, meaning that the return of a small firm has as much an impact as returns of larger firms. Meaning that if one where to apply a real life strategy based on these results they would need to accurately balance their portfolios to ensure it is equal-weighted.
23
References
Aboody, D., Lev, B., 1998. The value relevance of intangibles: the case of software capitalization. Journal of Accounting Research 36, 161–191.
Baruch, L., Sougiannis, T., 1996. The capitalization, amortization, and value-relevance of R&D, Journal of Accounting Economics 21, 107–138.
Brancha, B., Chichiraub, C., 2010. Mispricing vs risk premia in R&D-intensive firms. International Review of Financial Analysis 19, 358–367.
Campbell, J., Shiller, R., 1998. Valuation rations and the long-run stock market outlook. The Journal of Portfolio Management 24, 11-26.
Campbell, J., Shiller, R., 2001. Valuation Ratios and the long-run stock market outlook: an update. Cowles Foundation Discussion Papers 1295, Cowles Foundation for Research in Economics, Yale University.
Carhart, M., 1997. On persistance in mutual fund performance. Journal of Finance 52, 57-82.
Chan, L., Lakonishok, J., Sougiannis, T., 2001. The stock market valuation of research and development expenditures. The Journal of Finance 56, 2431–2456.
Core, J., Guay, R., Rusticus, T., 2006. Does weak governance cause weak stock returns? An examination of firm operating performance and investors’ expectations. Journal of Finance 61, 655– 68.
Deng, Z., Lev, B., Narin, F., 1999. Science and technology as predictors of stock perfor- mance. Financial Analysts Journal 55, 20–32.
24
Edmans, A., 2011. Does the stock market fully value intangibles? Employee satisfaction and equity prices. Journal of Financial Economics 101, 621–640.
Fama, E., French, K., 1993. Common risk factors in the returns on stocks and bonds. Journal of Financial Economics 33, 3-56.
Fama, E., French, K., 2010. Luck versus Skill in the Cross-Section of Mutual Fund Returns. The Journal of Finance 65, 1915-1947.
Fama, E., French, K., 2015. A five-factor asset pricing model. Journal of Financial Economics 116, 1-22.
Gibbons, R., Ross, S., Shanken, J., 1989. A test of the efficiency of a given portfolio. Econometrica 57, 1121-1152.
Giroud, X., Mueller, H., 2011. Corporate governance, product market competition, and equity prices. Journal of Finance 66, 563–600.
Gregory, A., Tharyan, R., Whittaker, J., 2014. Corporate social responsibility and firm value: disaggregating the effects on cash flow, risk and growth. Journal of Business Ethics 124, 633–657.
Hall, R., 2001. The Stock Market and Capital Accumulation. The American Economic Review 91, 1185-1202.
Higgins, H., 2016. Can securities analysts forecast intangible firms’ earnings? International Journal of Forecasting 29, 155–174.
Hong, H., Kacperczyk, M., 2009. The price of sin: the effect of social norms on markets. Journal of Financial Economics 93, 15–36.
25
Jaffe, J., Westerfield, R., 1985. The Week-end effect in common stock returns: the international evidence. The Journal of Finance 40, 433-454.
Keim, D., 1982. Size-related anomalies and stock return seasonality. Journal of Financial Economics 12, 13-31.
Kempf, A., Osthoff, P., 2007. The effect of socially responsible investing on portfolio performance. European Financial Management 13, 908–922.
Miller, M., Modigliani, F., 1961. Dividend policy, growth and the valuation of shares. The Journal of Business 34, 411-433.
26
Appendix A:
Descriptive statistics for intensity measures based on salesThe descriptive statistics for the intensity measures of the expenses over sales are presented in table A.1. Advertising expense, r&d expense, change in net working capital, and capital expenditures over sales ratios. # Obs, Mean, Median, Std. Dev. reporting the number of observations, mean, median and standard deviation respectively.
Table A.1: Descriptive statistics for intensity measures using sales
# Obs Mean Median Std.Dev.
Advertising 70,511 0.080 0.018 2.705
R&D 86,391 0.129 0.052 111.390
ΔNWC 45,581 0.218 0.014 29.163
27
Appendix B
: Risk-Adjusted Returns for portfolios sorted by expense intensity relative to market value.Estimated five-factor Fama-French regressions for yearly returns (in percent) are presented in table B.1 up to table B.4.Where MKT is the excess return on the market portfolio, SMB, HML, RMW and
CMA are the returns on the Fama-French (1995) factor-mimicking portfolios for size,
book-to-market, profitability and investment, respectively. Quintiles are sorted based on intensity of an expense relative to market value, with 1 having the lowest intensity and 5 the highest. α indicates the abnormal excess returns of the portfolio and aR2 the adjusted R2.
Excess returns are relative to the risk-free rate. P-values smaller than 0.01, 0.05 and 0.10 are indicated by ***, ** and *, respectively.
Table B.1:Risk-Adjusted Returns for portfolios sorted by advertising intensity.
Portfolio α βMKT βSMB βHML βRMW βCMA aR2 1Low -3.321* 0.845*** 0.976*** -0.127 -0.327* -0.396* 0.847 2 1.035 0.856*** 0.932*** 0.107 -0.249 -0.286 0.767 3 3.267 0.817*** 0.933*** 0.179 0.227 -0.138 0.795 4 4.742* 0.818*** 0.853*** 0.507*** -0.177 -0.352 0.731 5High 7.913*** 0.775* ** 0.894*** 0.513** -0.239 -0.386 0.612 High-Low 11.235*** -0.070 -0.082 0.639** 0.088 0.011 0.356
Table B.2: Risk-Adjusted Returns for portfolios sorted by capex intensity.
28
Table B.3: Risk-Adjusted Returns for portfolios sorted by change in net working capital intensity. Portfolio α βMKT βSMB βHML βRMW βCMA aR2 1Low 2.132 0.797* 1.114** -0.177 -1.276 0.104 0.459 2 2.984 0.721** 1.026*** -0.175 -0.837** 0.017 0.559 3 6.579 0.687 0.664* -0.063 -0.594 -0.282 0.513 4 5.063 0.655** 1.090*** -0.578 -0.319 0.690 0.555 5High 4.250 1.065*** 1.188*** 0.657 -0.759* -0.555 0.654 High-Low 2.118 0.268 0.074 0.835* 0.047 -0.659 0.177
Table B.4: Risk-Adjusted Returns for portfolios sorted by R&D intensity.
29
Appendix C
: Risk-Adjusted Returns for portfolios sorted by expense intensity relative to sales.Estimated five-factor Fama-French regressions for yearly returns (in percent) are presented in table C.1 up to table C.4.Where MKT is the excess return on the market portfolio, SMB, HML, RMW and
CMA are the returns on the Fama-French (1995) factor-mimicking portfolios for size,
book-to-market, profitability and investment, respectively. Quintiles are sorted based on intensity of an expense relative to sales, with 1 having the lowest intensity and 5 the highest. α indicates the abnormal excess returns of the portfolio and aR2 the adjusted R2.
Excess returns are relative to the risk-free rate. P-values smaller than 0.01, 0.05 and 0.10 are indicated by ***, ** and *, respectively.
Table C.1: Risk-Adjusted Returns for portfolios sorted by advertising intensity.
Portfolio α βMKT βSMB βHML βRMW βCMA aR2 1Low 2.859 0.859*** 0.905*** 0.072 -0.201 -0.201 0.806 2 3.122 0.919*** 1.052*** 0.196 -0.150 -0.184 0.820 3 2.603 0.864*** 0.875*** 0.212 -0.170 -0.217 0.789 4 3.936 0.759*** 0.819*** 0.352* -0.385* -0.306 0.671 5High 2.511*** 0.735*** 0.921*** 0.338* -0.372* -0.610** 0.681 High-Low -0.348 -0.124 0.015 0.266* -0.171 -0.409* 0.120
Table C.2: Risk-Adjusted Returns for portfolios sorted by change in net working capital.
30
Table C.3: Risk-Adjusted Returns for portfolios sorted by capex intensity.
Portfolio α βMKT βSMB βHML βRMW βCMA aR2 1Low 2.842 0.855*** 0.969*** 0.393** -0.133 -0.267 0.791 2 4.507** 0.835*** 0.936*** 0.257* -0.108 -0.181 0.803 3 3.586* 0.850*** 0.929*** 0.145 -0.022 -0.165 0.823 4 1.733 0.871*** 0.934*** 0.091 -0.044 -0.165 0.813 5High -3.647 0.809* ** 0.972*** -0.022 -0.210 -0.296 0.723 High-Low -6.489** -0.046 0.002 -0.415** -0.078 -0.019 0.282
Table C.4: Risk-Adjusted Returns for portfolios sorted by R&D intensity.
31
Appendix D:
Descriptive statistics of number of years firms are placed in the 5th portfolioTable D.1 presents the number of observations, the mean, median, standard deviation, minimum, and maximum amount of years spend in the 5th quintile for companies placed at least one year in the 5th quintile for each intensity measures based on market capitalization. The 5thquintile is made up by the firms with the highest intensity measure for each expense.
Table D.1: Descriptive statistics of the number of years firms are placed in the 5th quintile
# Obs Mean Median Std. Dev. Min Max
R&D 17,269 3.991 3 3.835 1 35
Advertising 14,085 4.296 3 4.547 1 41
Capex 38,315 4.209 2 4.680 1 41
