R&D investment, competition, and stock returns in Europe
Abstract
R&D investments tend to be risky investments, which is especially the case in a competitive environment. Therefore, firms that invest in R&D are expected to compensate investors by means of higher stock returns, even more so if they operate on competitive markets. This paper confirms that R&D investments are positively related to stock returns, especially within competitive markets. However, firms in competitive markets tend to earn lower expected stock returns than firms in concentrated markets, which suggests that investors are not compensated for risks that arise from competition.
Master Thesis 2016-2017 Semester 1, University of Groningen
Name: Ruben Verschuren
Student number: s2064022
Study Program: MSc Economics & MSc Finance (combined thesis) Supervisors: Prof. Dr. J.P. (Paul) Elhorst, Dr. T.M. (Dirk) Stelder
2 1. Introduction
This paper explains the relation between R&D investments and stock returns for the European stock market and how this relation is influenced by product market competition. R&D investments are one of the most important instruments for firms to ensure long-term viability, and thus increase firm value. Competition may affect firm value through the potential benefits of R&D investments. R&D investments potentially reduce firm value through reduced future cash flows in the event that a competitor wins the innovation race. Therefore, it follows that an intensive firm deals with more uncertainty than less R&D-intensive firms, even more so if its market is characterized by heavy competition. It is
interesting to see how this affects investors, and whether they are compensated for the risks that accompany R&D-intensive firms in competitive markets. Moreover, this paper tests whether conventional asset pricing models are able to detect the risks that R&D-intensive firms in competitive markets face.
Not surprisingly, R&D and innovation are key policy components of the European Union’s Europe 2020 Strategy, which is the EU’s agenda for growth and jobs. Having more innovative products and services on the market leads to job creation through increased industrial competitiveness, labor productivity and the efficient use of resources; finding solutions to societal challenges such as climate change and clean energy, security, and active healthy ageing. The EU targets combined public (government and higher education sector) and private (business enterprise sector) investment in R&D to be 3% of GDP by 2020.
Although the public investment in R&D is quite large (35% in 2014), the full impacts of R&D are believed to be realized by business R&D (64% in 2014) [Europe 2020 indicators – R&D, 2016].
Before studying the interaction effect of R&D and competition on stock returns, it seems straightforward to investigate the independent effects of R&D and competition on stock returns, respectively. The first set of hypotheses are:
3 Hypothesis 2: Firms with a high R&D intensity earn higher stock returns than firms with a
low R&D intensity.
Hypothesis 3: Firms in competitive markets earn higher stock returns than firms in concentrated markets.
These hypotheses are based on findings of previous literature on the R&D-return and competition-return relations. Similar hypotheses are tested for the US stock market, but not thoroughly for the European stock market. For instance, Chan, Lakonishok, and Sougiannis (2001) address Hypotheses 1 and 2. They find no significant difference between stock returns of R&D firms and non-R&D firms. In contrast, R&D-intensive firms earn higher stock returns than less R&D-intensive firms. Furthermore, Hou and Robinson (2006) address Hypothesis 3. They find a positive relation between competition and stock returns and suggest that this might be caused by more R&D activities among firms in competitive markets. This explanation makes it interesting to test whether the interaction of R&D and competition affects stock returns. This paper distinguishes the following hypotheses on the interaction effect of R&D and competition on stock returns:
Hypothesis 4: The R&D-return relation is positive, but only in competitive markets.
Hypothesis 5: The competition-return relation is positive, but only among R&D-intensive firms.
These hypotheses follow from the findings of Lifeng (2016), who studies the US stock market, and finds a positive relation between R&D intensity and stock returns, but only for firms in competitive markets. Similarly, the relation between competition and stock returns is found to be positive, but only among R&D-intensive firms.
4 allocations, are studied by cross-sectional regressions in the style of Fama and MacBeth (1973).
The findings of this paper confirm that R&D is positively related to stock returns. More specifically, firms that invest in R&D tend to earn higher stock returns than non-R&D firms. Moreover, higher R&D intensity leads to higher expected stock returns, which is especially visible for firms in competitive markets. However, competition is negatively related to stock returns. Although there are signs of a positive competition-return relation among R&D-intensive firms, effectively, stock returns of firms in competitive markets tend to be lower than stock returns of firms in concentrated markets.
The main contribution of this paper is that it validates prior research on stock returns in relation to R&D and/or competition, with a unique focus on the European stock market. Previous literature on the relation between R&D, competition and stock returns mainly focuses on the US stock market. An interesting question that this paper attempts to answer is whether investors are compensated for the risks that firms face due to R&D and competition. Moreover, this paper provides a comparison between two empirical methods to analyze stock returns.
Chapter 2 reviews previous literature on the concepts of R&D, competition, stock returns, and shows how these relate to each other. Chapter 3 explains the methodology used in this paper, Chapter 4 describes the dataset, Chapter 5 presents and discusses the results, and Chapter 6 presents the conclusions.
2. Literature Review
5 2.1. Economical relevance of R&D and Competition
R&D investment is believed to be one of the main drivers of economic growth. Solow (1957) introduces an economic growth model which is driven by technological change. Since then, the economics literature considered technology development as an important component of economic dynamics. Hsu (2009) explains that technological progress comes from
endogenous effort, which can be R&D investments that lead to inventions, and exogenous incidents, which can be accidental discoveries. Moreover, Romer (1986, 1990) and
Greenwood, Hercowitz, and Krusell (1997, 2000) find that both types of technological progress explain economic growth and fluctuations.
Similarly, for firms, R&D investment is one of the most important instruments to ensure long-term viability (Lifeng, 2016). R&D is directly performed through R&D investments, but is also indirectly incorporated in purchased intermediates, capital goods, often referred to as embodied technological change, and in the human capital, which relates to the knowledge of R&D employees. R&D activities can lead to new products or processes. Product innovations increase firms’ cash flows by charging higher prices for new product features or find new buyers. Process innovations can lead to increased productivity of physical capital and a reduction of production costs. For example, Ortega-Argilés, Piva, and Vivarelli (2014) find that R&D investment has a significant positive impact on firm
productivity. Moreover, the model used by Ortega-Argilés, Piva, and Vivarelli (2014) predicts that investment in physical capital, representing embodied technological change, also
positively affects firm productivity.
Furthermore, Hirshleifer, Hsu, and Li (2013) find that firms with high innovative efficiency, measured as the number of patents scaled by R&D expenditures, generally have higher market valuations and superior future operating performance. Bloom and Van Reenen (2002) note that productivity growth is plausibly achieved by innovation and that investors indeed value innovation. Patents and patent citations, as proxies for innovation, are found to have a strong effect on productivity growth and market valuations.
Economists have long been interested in the relation between competition and
6 difficult for firms to exploit innovations. For instance, Teece (1992) explains that a ‘free rider’ problem arises if firms are unable to exclude other firms from using technology they have developed. Patents only partly overcome this problem, as it is likely that similar
innovations, not foreclosed by the patent, allow competitors to obtain some of the benefits of the R&D investment. Furthermore, innovation races may induce a misallocation of resources, as firms may overinvest in an attempt to be the first to invent a new product. A monopolized industry overcomes these problems. Additionally, firms need to be large to afford the cost of R&D programs and large, diversified firms can absorb the failures from innovating across broad technological fronts. Advantages of monopolies that naturally arise are economies of scale and scope. However, as Teece (1992) argues, the monopolist’s output restriction likely cuts R&D investments.
Conversely, the Darwinian view states that competition stimulates innovation. Competition forces firms to be more efficient, which thus forces firms to invest in R&D. Firms in competitive markets have a greater incentive to invest in R&D than monopolies, because gaining technological leadership in a competitive market is more rewarding, which is called the Arrow replacement effect. Aghion and Howitt (1998) explain that competition may force nonprofit-maximizing firms (due to agency problems between owners and managers) to speed up technological adoptions in order to remain solvent.
The Darwinian view is empirically supported by Nickell (1996) and Blundell, Griffiths, and Van Reenen (1999). Total factor productivity growth is found to be higher in competitive markets, confirming that firms in competitive markets are forced to be more efficient. Moreover, increased competition tends to stimulate innovative activity. However, the benefits of innovation, in terms of market valuation, tend to be higher for firms with larger market shares. Additionally, Aghion, Bloom, Blundell, Griffith, and Howitt (2005) find that, in general, competition positively impacts innovation, but too much competition can have a negative impact on innovation.
7 firms. Additionally, the competitive character of agglomerations may incentivize firms to innovate more, in line with the Darwinian view. Conversely, the existence of knowledge spillovers indicates that firms are unable to internalize the full benefit of R&D investments, which is one of the implications brought forward by the Schumpeterian view. Thus, it follows that both the Darwinian view and the Schumpeterian view are supported to some extent.
2.2. Relation of R&D and Competition to stock returns
Lifeng (2016) investigates the interaction between R&D investment and competition in relation to stock returns, by using a theoretical and an empirical approach. Her findings are; (1), the R&D-return relation is positive, but only for firms in competitive markets, and; (2), the competition-return relation is positive, but only among R&D-intensive firms.
The theoretical approach uses a real options model developed by Berk, Green, and Naik (2004) to determine the risks that accompany R&D investments. Based on the model, Lifeng (2016) finds that the risk premiums of R&D-intensive firms are expected to be higher than the risk premiums of less R&D-intensive firms. Therefore, R&D-intensive firms have higher expected returns. Additionally, competition features in the model through a higher obsolescence rate for R&D firms in competitive markets. The risk for obsolescence of R&D projects is higher in competitive markets. Therefore, R&D-intensive firms active in
competitive markets likely have higher risk premiums, and thus higher expected returns. As for the empirical part of the paper, Lifeng (2016) tests the relation of R&D investments and competition on stock returns in the US stock market (NYSE-, Amex-, and NASDAQ-listed securities). Stock return characteristics are studied using asset pricing models on portfolios of stocks sorted on R&D intensity and competition, which is similar to the approach of Fama and French (1993) who study size and market-to-book factors. Lifeng (2016) finds that the abnormal stock returns of firms in competitive industries are increasing with respect to R&D intensity and are significantly positive for medium and high R&D-intensive firms. Moreover, among R&D R&D-intensive firms, the abnormal stock returns are the highest for firms in competitive markets.
8 relate to the successful completion, uncertainty about the future cash flows, and the risk of competitive threats or obsolescence of the innovation.
Empirical research, which generally focuses on US firms, suggests that the relation between R&D intensity and stock returns is positive. For example, Lev and Sougiannis (1996) find large excess returns for R&D-intensive firms. Chan, Lakonishok, and Sougiannis (2001) confirm the large excess returns for R&D-intensive firms, but they find that stock returns between R&D firms match stock returns of non-R&D firms. Both papers state that the method of accounting R&D expenses influences stock returns. Under conservative accounting
methods R&D expenses are directly reducing current earnings, which distorts conventional financial ratios such as the price-earnings ratio and market-to-book ratio. Conversely, the effects from R&D investments persist longer than one year. Thus, capitalizing R&D is more consistent with the nature of R&D investments. Conservative accounting of R&D expenses leads to mispriced stocks if information available to investors is not interpreted properly, i.e. investors might pursue an investment strategy based on financial ratios that are not corrected for firms’ R&D investments.
Moreover, Chambers, Jennings, and Thompson (2002) find that the excess returns among R&D-intensive firms are mainly explained by extra risk bearing by investors, although they also find evidence of mispricing. Evidence of the risk explanation consists of persistent excess returns, higher volatility on stock returns, and greater variability of analysts’ forecasts and future earnings for R&D-intensive firms. Evidence of mispricing relates to the fact that excess stock returns are larger (smaller) for firms that decrease (increase) R&D expenditures in the current year, which may be caused by higher (lower) current earnings. These findings suggests that even though investors earn higher returns from investing in R&D-intensive firms in the long-run, they also appreciate short-term profitability, which follows from positive abnormal returns to R&D expense cuts. On the contrary, Eberhart, Maxwell, and Siddique (2004) find that firms that unexpectedly increase their R&D expenditures by a significant amount receive persistent positive abnormal stock returns and abnormal operating performance, which suggests that R&D investments are considered beneficial investments and that the market is slow to recognize the extent of this benefit. Conversely, the persistent positive abnormal stock returns might also explain higher risk bearing by investors due to the increased risk from R&D investments.
9 find that firms that have been successful in the past and invest heavily in R&D earn
substantially higher future returns than firms that invest identical amounts in R&D but have poor past track records.
Conversely, Lin (2012) provides a non-risk based model that explains why R&D-intensive firms earn higher average stock returns than less R&D-R&D-intensive firms. The model relates R&D investment with physical capital. Intuitively, the expected stock return depends on the return on R&D capital and physical capital. R&D (physical) investments positively affect the marginal productivity of physical (R&D) capital, but they negatively affect the marginal productivity of R&D (physical) capital. Lin (2012) shows empirically that stock return movements are best explained by the return on physical capital, because physical capital tends to be larger than R&D capital and a high portion of R&D investments focuses on productivity increasing innovations, i.e. enhancing the productivity of physical capital. Since the return on physical capital is positively influenced by R&D investments, stock returns are likely to be increasing in R&D investments. Moreover, stock returns tend to be negatively affected by investments in physical capital.
Hou and Robinson (2006) find that stocks of firms in concentrated industries earn lower returns compared to stocks of firms in competitive industries. They provide two R&D related explanations. One explanation is that firms in concentrated industries are less risky, because they engage in less innovation, and therefore earn lower stock returns. Moreover, competition may force managers to invest in R&D to ensure that the firm remains viable. Therefore, stock returns might increase to compensate investors for the extra risk associated with R&D investments. Additionally, firms in concentrated markets are better insulated from shocks to aggregate demand, which explains lower stock returns due to less distress risk. Sharma (2011) confirms the results of Hou and Robinson (2006), and finds that in addition to market concentration, alternative specifications of competition are also suggest a positive competition-return relation.
Furthermore, Gaspar and Massa (2006) find that competition positively affects
idiosyncratic return volatility. They argue that a firm with market power is better able to pass on idiosyncratic cost shocks to its consumers, where a firm in a competitive market is driven out of the market if its costs are too much out of line with those of its competitors.
10 confirms that firms with more market power have less volatile returns, and additionally finds that these firms have lower expected returns, less volatile profits, and higher Sharpe ratios1. These variables are reduced further by the improved information in stock prices among firms with more market power.
The R&D- and competition-return relations that Lifeng (2016) finds, which are based on a sample of public firms, are robust to an alternative specification of competition that includes private firms. However, Ali, Klasa, and Yeung (2009) find that results of
competition-return literature based on public firm samples are often biased. For example, they find no evidence of higher stock returns in competitive markets based on a sample of public and private firms. Furthermore, based on a sample of public and private firms, Bustamante and Donangelo (2016) find several opposing channels through which product market
competition influences stock returns, which eventually lead to lower expected stock returns in competitive markets. For example, expected stock returns of incumbents decrease through value destruction associated with potential entry of new firms. Additionally, riskier industries are less attractive to new entrants and might therefore remain less competitive, which would explain higher expected stock returns in concentrated industries. Conversely, competition increases expected stock returns through lower profit margins, which reduces buffers to adverse shocks and thus increases risk.
An important thing to note is that, when studying stock returns, the efficient market hypothesis is assumed to hold, i.e. all information available to investors is already
incorporated in the stock price. For example, whether a firm invests in R&D or not and how intensively it does should already be incorporated in the stock price. Abnormal returns either imply that the efficient market hypothesis should be rejected and stocks are mispriced or that investors are paid a premium for risk that asset pricing models fail to explain. If stocks are mispriced, investors fail to incorporate all information available properly.
Asset pricing models adjust stock returns for common risks that are found to explain differences between stock returns. Therefore, excess returns can be assumed to explain the specific return characteristics of the sample, such as an implied risk premium for R&D-intensive firms. The simple CAPM model (Sharpe, 1964; Lintner, 1965) adjusts stock returns for market risk. Stocks are expected to reward investors with a return in excess of the risk free rate and conform the riskiness of the stock in relation to the market portfolio. The CAPM
1 The Sharpe ratio is an information ratio that is measured as the ratio of expected excess returns to standard
11 model lays the foundation for multiple other asset pricing models. Fama and French (1992) find that market beta is not sufficient to explain stock returns and add the factors size and book-to-market to create the three-factor model. In general, small firms are more risky than big firms, which results in higher expected stock returns. The book-to-market ratio contains information regarding a firm’s earnings. Firms with high book-to-market ratios tend to have low earnings on assets, whereas low book-to-market ratios are associated with high earnings. Carhart (1997) adds the momentum factor to the three-factor model. Stock returns tend to have a persistent character, stocks with high returns in one year tend to perform well in the next year and vice versa. Whereas other factors are related to long-term explanations of variation in stock returns, momentum explains the short-term variation in stock returns. Fama and French (2015) introduce the five-factor model that, in addition to size, book-to-market and market beta, also includes factors for investment and profitability. These factors correct stock returns for firm characteristics such as high- or low investments, and robust- or weak profitability. Fama and French (2015) argue that investment and profitability are natural additions to the three-factor model, since both factors directly influence the dividend discount model (Miller and Modigliani, 1961). Market anomalies are often studied using these asset pricing models, see for example Fama and French (2016).
2.3. Europe versus the US
The research on the relation between stock returns and R&D investments and/or competition has mainly focused on US stocks, ignoring how R&D and competition affect the European stocks. Shifting the focus to the European stocks leads to new challenges such as data coverage and market definition, i.e. the relevant market needs to be defined to study the competition effect. For example, R&D and stock data is better available in the US.
Ortega-Argiles, Piva, and Vivarelli (2014) test whether the difference between R&D investments in the EU and the US explains the transatlantic productivity gap. Private R&D expenditures in the EU are equal to 60% of the private R&D expenditures in the US over the period investigated. Not only does the EU invest less in R&D, these investments seem less rewarding, suggesting that firms from the US are more capable of translating R&D
investments into productivity increases than European firms.
Hall and Orianni (2006) show that, in general, R&D capital positively impacts firms’ market values. However, they find that this relation differs between the Anglo-Saxon
12 arise due to different legal regimes, ownership structures, looser stock market discipline, R&D incentive schemes, and data availability. For example, Hall and Orianni (2006) find that R&D-intensive firms without large shareholders tend to have higher market values. They point out that in countries offering weak legal protection to financial investors, such as Germany, France, and Italy, in contrast to Anglo-Saxon countries, large shareholders might exploit information asymmetries created by R&D investments, which results in lower market valuations. Moreover, Aboody and Lev (2000) show that R&D firms are more likely to be subject to insider trading than non-R&D firms. Thus, outside investors may not be able to benefit from higher stock returns associated with R&D firms.
The transatlantic productivity gap and country specific differences may have
implications for the stock return characteristics among R&D-intensive firms in Europe. Hsu (2009) extends the R&D-return research to European markets, such as France, Germany, Italy, and the UK, and finds that technological innovations have positive and predictive power on market returns and risk premiums. Thereby implying that risk premiums for
R&D-intensive firms on European stock markets are similar to those on US stock markets.
3. Methodology
13 Figure 1: Summary of the methodology.
3.1. Method one: Portfolio analysis
This section describes the first empirical method to test the hypotheses, which is portfolio analysis. First, measures of R&D intensity and competition are described, which are used to allocate stocks to portfolios. Next, the asset pricing models are introduced, which adjust the portfolio returns for risks that are commonly associated with stocks.
3.1.1 Measuring R&D intensity
This subsection describes how R&D intensity is measured, which estimates the importance of R&D to a firm and is used to allocate firms’ stocks to portfolios.
R&D can be measured by input variables or output variables. Input variables are R&D expenditure, R&D employment, and R&D capital. A commonly used output variable is patents. R&D input variables are most appropriate in this study, because the potential R&D premium is most likely related to the uncertainty regarding R&D investments. More
specifically, R&D input variables reflect the firm’s decision regarding the level of R&D investments, which leaves investors with uncertainty. A disadvantage of using an input variable to measure R&D activity is that it does not show how successful a firm’s R&D is. Some studies suggest that the successfulness of a firm’s R&D is the main reason that these firms earn abnormal stock returns (e.g. see Cohen, Diether, and Malloy, 2013; Hirshleifer, Hsu, and Li, 2013). Patents can be used to show how successful a firm’s R&D is. However,
Collect European stocks and calculate stock
returns. Method 1: Allocate stocks to portfolios based on rankings of R&D and/or Competition. Analyze (abnormal) portfolio returns Method 2: Analyze stock returns directly using Fama-MacBeth
cross-sectional regressions.
Regress on stocks' allocated portfolio, which is based on R&D
14 the value of different patents is hard to compare and needs to be scaled by R&D investments to be informative.
Theoretically, a firm’s R&D activity is best captured by its R&D capital, because R&D investments have persistent effects and do not just affect current year’s operations. R&D expenditures and R&D employment are one-year inputs for R&D activity and do not incorporate previous years’ R&D investments that still affect current year’s operations. Unfortunately, R&D capital is rarely specified on a firm’s balance sheet. It is often hidden as part of the intangible assets on a firm’s balance sheet. Moreover, accounting standards for some countries allow firms to capitalize some of their R&D expenditures, but the conditions that firms have to meet in order to capitalize their R&D expenditures ensure that most firms incur costs related to R&D immediately. That means that if R&D capital is used as measure to define R&D activity, the value of R&D capital has to be estimated. Several studies show how this can be done, but the amortization rate has to be estimated, which may differ among firms and industries. Therefore, many studies use R&D expenditures to calculate R&D intensity. Note that if R&D expenditures are constant over several years, they are identical to R&D amortization if R&D capital is amortized linearly, which makes an estimation of R&D capital unnecessary. Additionally, R&D expenditure is part of firms’ income statement and therefore widely available. R&D expenditure and R&D employment are highly correlated and can be used interchangeably, but data coverage of R&D expenditure is better. Additionally,
limitations of using R&D expenditures are relying on firms’ disclosure of R&D expenditure, which may differ among countries, inclusion of engineering activities in some countries, and a firm’s headquarters location may differ from their R&D location2.
The literature uses several scaling factors to measure R&D intensity: market capitalization, book equity, sales, total assets, and employment. Lifeng (2016) scales R&D expenditures by market capitalization, based on its ability to forecast returns, unlike other measures of R&D intensity. This paper uses the same variables to measure R&D intensity, be it for other reasons. Thus, R&D intensity is calculated by dividing R&D expenditures (R&D) by market capitalization (MC):
𝑅&𝐷 𝑖𝑛𝑡𝑒𝑛𝑠𝑖𝑡𝑦 =𝑅&𝐷
𝑀𝐶 . (1)
15 Market capitalization, as opposed to other scaling variables used in the literature, measures firm value: the value of current assets and future activities, among them future cash flows from R&D projects. Furthermore, market capitalization is often used to measure firm size. Additionally, sales are commonly used to scale R&D activity, even though R&D expenditures probably do not affect current year’s sales. In other words, current year’s sales is just a
measure of a firm’s current size, which does not take potentially higher future sales due to R&D into account.
3.1.2 Measuring the degree of competition
This subsection describes how the degree of competition of a firm or a firm’s market is measured, which is used to allocate firms’ stocks to portfolios.
The competition-return literature often uses Herfindahl-Hirschman Index (HHI) to measure the degree of competition. The HHI represents the industry concentration based on market shares. However, it does have several disadvantages, especially for a sample that only contains listed firms. As unlisted firms are not in the sample, the HHI is estimated by the market shares of the publicly listed firms in the industry. Moreover, the HHI requires a firm’s relevant market, which is difficult to determine for each firm. Some firms compete on a global level, where other firms compete on a country level. More specifically, defining Europe as the relevant market overstates the market shares for firms that compete at the global level and understates the market shares of firms that compete at a country level. Moreover, the HHI does not incorporate competitive threats from other sources than market shares and is not able to detect collusion between firms.
16 related to competitiveness are also taken into account. The PCM of firm i in year t is
calculated by the following formula:
𝑃𝐶𝑀𝑖𝑡 = 𝑂𝑃𝑖𝑡
𝑆𝐴𝐿𝐸𝑆𝑖𝑡, (2)
where SALES stand for net sales and OP stands for operating profit, which is defined as:
𝑂𝑃𝑖𝑡 = 𝑆𝐴𝐿𝐸𝑆𝑖𝑡 − 𝐶𝑂𝐺𝑆𝑖𝑡− 𝑆𝐺𝐴𝐸𝑖𝑡, (3)
where COGS are ‘cost of goods sold’ and SGAE are ‘selling, general and administrative expenses’. The directly reported ‘operating income’ is used as proxy for operating profit if data on COGS and SGAE is missing.
Moreover, it might be the case that not individual competitive threats underlie the competition premium, but rather the competitiveness of a firm’s industry. Therefore, the analyses are also done with industry PCM (IPCM) and HHI as competition measures, which also improves comparability with previous competition-return literature. The IPCM for industry j in year t is calculated as:
𝐼𝑃𝐶𝑀𝑗𝑡= ∑ 𝑃𝐶𝑀𝑖𝑡× 𝑠𝑖𝑗𝑡 𝑁𝑗
𝑖=1 , (4)
where negative values of PCM are set to zero to avoid distortions of the industry average, and 𝑠𝑖𝑗𝑡 stands for the market share of firm i in industry j in year t, which equals the ratio of net sales (SALES) of firm i to the total net sales of industry j:
𝑠𝑖𝑗𝑡 = 𝑆𝐴𝐿𝐸𝑆𝑖𝑡
𝑆𝐴𝐿𝐸𝑆𝑗𝑡. (5)
Additionally, the HHI is calculated according to the following formula:
𝐻𝐻𝐼𝑗𝑡 = ∑𝑁𝑗 𝑠𝑖𝑗𝑡2
𝑖=1 , (6)
17 3.1.3 Portfolio allocation
This subsection describes how stocks are allocated to portfolios. The portfolio
allocation is based on rankings of R&D intensity and/or degree of competition. In order to test the hypotheses, the performance of the portfolios is compared to find whether R&D and competition, independently or in interaction, explain stock returns.
Firstly, for Hypothesis 1, stocks are either allocated to an R&D portfolio or a non-R&D portfolio. Secondly, for Hypothesis 2, in July of each year t stocks are allocated to five portfolios based on the ranked values of R&D intensity in year t – 1. The stocks are sorted to quintiles using either market equity or sales as scaling variable in R&D intensity, see
Equation 1. Similarly, for Hypothesis 3, in July of each year t stocks are allocated to five portfolios based on the ranked values of the degree of competition in year t – 1, which is measured by PCM, IPCM, and HHI. The portfolios are rebalanced in July of each year.
18 firms independently based on 50% breakpoints for both R&D intensity and the degree of competition.
3.1.4 Asset pricing models
This subsection explains how the portfolio returns are adjusted for risks using three conventional asset pricing models. First off, the portfolio returns (𝑅𝑝𝑡) are constructed as follows:
𝑅𝑝𝑡 = ∑𝑁𝑖=1𝑝𝑡𝑅𝑖𝑡× 𝑊𝑖𝑡, (7)
where 𝑁𝑝𝑡 refers to the number of stocks in portfolio p in month t, and 𝑊𝑖𝑡 refers to the portfolio weight of stock i in month t, which depends on whether portfolios are value- or equal-weighted, see Appendix A for more details. The risk adjustment of these portfolio returns leads to abnormal portfolio returns (𝛼𝑝) that represent the risk premium of the respective R&D and/or competition portfolio. The asset pricing models are the Fama and French (1993) three-factor model, the Carhart (1997) four-factor model, and the Fama and French (2015) five-factor model, respectively:
𝐸𝑅𝑝𝑡 = 𝛼𝑝+ 𝛽1𝑝𝑅𝑀𝑡+ 𝛽2𝑝𝑆𝑀𝐵𝑡+ 𝛽3𝑝𝐻𝑀𝐿𝑡+ 𝜀𝑝𝑡, (8)
𝐸𝑅𝑝𝑡 = 𝛼𝑝+ 𝛽1𝑝𝑅𝑀𝑡+ 𝛽2𝑝𝑆𝑀𝐵𝑡+ 𝛽3𝑝𝐻𝑀𝐿𝑡+ 𝛽4𝑝𝑀𝑂𝑀𝑡+ 𝜀𝑝𝑡, (9)
and
𝐸𝑅𝑝𝑡 = 𝛼𝑝+ 𝛽1𝑝𝑅𝑀𝑡+ 𝛽2𝑝𝑆𝑀𝐵𝑡+ 𝛽3𝑝𝐻𝑀𝐿𝑡+ 𝛽4𝑝𝑅𝑀𝑊𝑡+ 𝛽5𝑝𝐶𝑀𝐴𝑡+ 𝜀𝑝𝑡, (10)
19 The asset pricing models presented in Equations 8 to 10 are estimated by time-series OLS regressions for each portfolio. The asset pricing models imply that 𝛼𝑝 equals zero, since the factors should explain all risk associated with the stocks in the portfolio. Thus, in order to find whether a portfolio earns abnormal returns the corresponding null Hypothesis tests whether 𝛼𝑝 = 0 holds. Subsequently, in addition to observing the patterns that the abnormal returns might reveal, the difference between abnormal returns of high- (𝛼𝑝𝐻) and low (𝛼𝑝𝐿) portfolios is tested by the null Hypothesis 𝛼𝑝𝐻 = 𝛼𝑝𝐿. Tests for raw portfolio returns are also
performed, which suffice under the assumption that the portfolios do not differ with respect to common stock risks. More specifically, Hypothesis 1 tests whether 𝛼𝑅&𝐷 = 𝛼𝑛𝑜𝑛−𝑅&𝐷 holds. Secondly, Hypothesis 2 tests whether 𝛼𝑅&𝐷5 > 𝛼𝑅&𝐷1 holds, i.e. R&D-intensive firms earn higher stock returns than less R&D-intensive firms. Similarly, Hypothesis 3 tests whether 𝛼𝐶𝑂𝑀𝑃5 > 𝛼𝐶𝑂𝑀𝑃1 holds, i.e. firms in competitive markets earn higher stock returns than firms
in concentrated markets. Furthermore, Hypothesis 4 tests whether 𝛼𝑅&𝐷𝐻 > 𝛼𝑅&𝐷𝐿 holds in
competitive markets, which is compared to the outcome for concentrated markets. And lastly, Hypothesis 5 tests whether 𝛼𝐶𝑂𝑀𝑃𝐻 > 𝛼𝐶𝑂𝑀𝑃𝐿 holds for R&D-intensive firms, which is
compared to the outcome for less R&D-intensive firms.
Following Cochrane (2001), simple t-tests test the hypotheses, which require that the standard errors are uncorrelated and homoscedastic. Brooks (2008) provides two solutions for heteroscedasticity: using logarithmic variables and heteroscedasticity-consistent standard error estimates. Therefore, the regressions use logarithmic returns for portfolios and factors, which also makes statistical tests more reliable, see Appendix A for more details.
Additionally, Newey-West standard errors are used in statistical tests, which corrects standard errors for heteroscedasticity and autocorrelation, and makes the hypothesis tests more
conservative if these effects are indeed present.
3.2. Method two: Cross-sectional regressions
20 Following Fama and MacBeth (1973), cross-sectional regressions3 are performed in each month across all sample firms. The coefficients are averaged over months to represent time-series coefficients for the whole sample. The dependent variable is the simple stock return. R&D and competition feature in the cross-sectional regression through portfolio allocation or directly. The portfolio allocations are described in subsection 3.1.3. A dummy variable specifies to which portfolio a stock is allocated. The cross-sectional regressions are run from July 1990 to June 2016, except for the regressions based on three-by-three portfolio allocations, which are run from July 2000 to June 2016. The direct cross-sectional regressions are run directly on R&D intensity and degree of competition. Moreover, several specifications of R&D and competition are used. R&D expenditures are scaled by market capitalization (R&DMC) or sales (R&DS) to measure R&D intensity. The degree of competition is
measured by the price-cost-margin (PCM), the industry PCM, and the Herfindahl-Hirschman Index (HHI). In addition to simple t-tests on the coefficients, Wald tests are used to compare coefficients of high- and low groups of R&D and competition.
First off, Hypothesis 1 predicts that stock returns of R&D firms equal stock returns of non-R&D firms, which is tested by the following cross-sectional regression:
𝑅𝑖𝑡 = 𝛼𝑡+ 𝛽1𝑡× 𝑅&𝐷𝐷𝑖𝑡+ 𝛾𝑡𝑋𝑖𝑡+ 𝜀𝑖𝑡, (11)
where 𝑅&𝐷𝐷𝑖𝑡 is a dummy indicating whether firm i invests in R&D in month t, and 𝑋𝑖𝑡 contains the control variables. Hypothesis 1 tests whether 𝛽1𝑡 = 0 holds, which would indicate that R&D investments do not affect stock returns. The time subscript t indicates that the cross-sectional regressions are repeated each month and that the estimated coefficients are time-series averages.
Furthermore, Hypothesis 2 predicts that R&D-intensive firms earn higher stock returns than less R&D-intensive firms, which is tested by the following cross-sectional regression:
𝑅𝑖𝑡 = 𝛼𝑡+ 𝛽1𝑡× 𝑅&𝐷𝐻𝑖𝑡+ 𝛽2𝑡 × 𝑅&𝐷𝐿𝑖𝑡+ 𝛾𝑡𝑋𝑖𝑡+ 𝜀𝑖𝑡, (12)
3 Specifically, the Stata add-in “xtfmb” version 2.0.0 made by Daniel Hoechle (2011) is used to perform the
21 where 𝑅&𝐷𝐻𝑖𝑡 and 𝑅&𝐷𝐿𝑖𝑡 are dummies that indicate whether a stock is allocated to the high- or low R&D-intensive portfolio, holding stocks of the top 30% R&D-intensive firms and bottom 30% R&D-intensive firms, respectively. The medium R&D intensity portfolio
functions as the reference group of Equation 12. Hypothesis 2 tests whether 𝛽1𝑡 > 𝛽2𝑡 holds. The regressions are run for portfolio allocations based on R&D relative to market
capitalization and sales, as described by Equation 1.
Hypothesis 3 predicts that firms in competitive markets earn higher stock returns than firms in concentrated markets, which is tested by the following cross-sectional regression:
𝑅𝑖𝑡 = 𝛼𝑡+ 𝛽1𝑡× 𝐶𝑂𝑀𝑃𝐻𝑖𝑡+ 𝛽2𝑡× 𝐶𝑂𝑀𝑃𝐿𝑖𝑡+ 𝛾𝑡𝑋𝑖𝑡+ 𝜀𝑖𝑡, (13)
where 𝐶𝑂𝑀𝑃𝐻𝑖𝑡 and 𝐶𝑂𝑀𝑃𝐿𝑖𝑡 are dummies that indicate whether a stock is allocated to the high- or low competition portfolio, respectively. The medium competition quintile functions as reference group in Equation 13. Hypothesis 3 tests whether 𝛽1𝑡 > 𝛽2𝑡 holds. The
regressions are run for portfolio allocations based on PCM, IPCM, and HHI, see Equations 2, 4, and 6, respectively.
Hypotheses 4 and 5, which predict that stock returns are highest among R&D-intensive firms in competitive markets are jointly tested by the following cross-sectional regression:
𝑅𝑖𝑡 = 𝛼𝑡+ 𝛽1𝑡(𝑅&𝐷𝐻𝑖𝑡× 𝐶𝑂𝑀𝑃𝐻𝑖𝑡) + 𝛽2𝑡(𝑅&𝐷𝐻𝑖𝑡× 𝐶𝑂𝑀𝑃𝐿𝑖𝑡) + 𝛽3𝑡(𝑅&𝐷𝐿𝑖𝑡× 𝐶𝑂𝑀𝑃𝐻𝑖𝑡) + 𝛽4𝑡(𝑅&𝐷𝐿𝑖𝑡× 𝐶𝑂𝑀𝑃𝐿𝑖𝑡) + 𝛽5𝑡× 𝐶𝑂𝑀𝑃𝐻𝑖𝑡+ 𝛽6𝑡× 𝐶𝑂𝑀𝑃𝐿𝑖𝑡+ 𝛽7𝑡× 𝑅&𝐷𝐿𝑖𝑡+ 𝛽8𝑡× 𝑅&𝐷𝐻𝑖𝑡+ 𝛾𝑡𝑋𝑖𝑡+ 𝜀𝑖𝑡
(14)
which is based on three-by-three portfolios and where the dummies indicate each extreme R&D or competition group. Combining the R&D and competition dummies yields three-by-three portfolio allocations and allows to study interaction effects. For example,
22 tests whether 𝛽1𝑡 > 𝛽3𝑡, i.e. the R&D-return is positive for firms in competitive markets. For comparison, the R&D-return relation is also tested for firms in concentrated markets.
Hypothesis 5 tests whether 𝛽1𝑡 > 𝛽2𝑡, i.e. the competition-return is positive among intensive firms. For comparison, the competition-return relation is also tested for less R&D-intensive firms.
In contrast to the three-by-three portfolio specification, the two-by-two portfolio specification uses the whole sample period. The cross-sectional regression in Equation 14 is translated to the two-by-two portfolio specification as follows:
𝑅𝑖𝑡 = 𝛼𝑡+ 𝛽1𝑡(𝑅&𝐷𝐻𝑖𝑡× 𝐶𝑂𝑀𝑃𝐻𝑖𝑡) + 𝛽2𝑡(𝑅&𝐷𝐻𝑖𝑡× 𝐶𝑂𝑀𝑃𝐿𝑖𝑡) +
𝛽3𝑡(𝑅&𝐷𝐿𝑖𝑡× 𝐶𝑂𝑀𝑃𝐻𝑖𝑡) + 𝛾𝑡𝑋𝑖𝑡+ 𝜀𝑖𝑡, (15)
which is tested for Hypotheses 4 and 5 in a similar way to Equation 14.
Furthermore, additional tests for direct- or interaction effects of R&D and competition are done by running cross-sectional regressions directly on R&D intensity and the degree of competition:
𝑅𝑖𝑡 = 𝛼𝑡+ 𝛽1𝑡𝑅&𝐷𝑖𝑡 + 𝛽2𝑡𝑀𝑃𝑖𝑡 + 𝛽3𝑡(𝑅&𝐷𝑖𝑡× 𝑀𝑃𝑖𝑡) + 𝛾𝑡𝑋𝑖𝑡+ 𝜀𝑖𝑡, (16)
where 𝑅&𝐷𝑖𝑡 stands for R&D intensity of firm i in month t, scaled by sales or market
23 Furthermore, several control variables ensure that the regressions allow for firm
specific differences. These control variables are size, book-to-market ratio, leverage, market beta, and momentum, which commonly feature in asset pricing models. Additionally, country fixed effects control for country specific differences in stock returns. Leverage is calculated as total assets minus book equity divided by market capitalization. Firm size is measured by the logarithm of the market capitalization. The book-to-market ratio is calculated by dividing book equity by market capitalization. Similar to R&D and competition related data,
accounting data and market capitalization in year t – 1 are linked to stock returns from July in year t to June in year t + 1. Moreover, momentum is the cumulative return over the period t – 12 to t – 2 (where t stands for month):
𝑅𝐼𝑖𝑡−2−𝑅𝐼𝑖𝑡−13
𝑅𝐼𝑖𝑡−13 . (17)
Following Fama and MacBeth (1973), a firm’s market beta (𝛽̂𝑀𝑖) is calculated from monthly returns in the previous 60 months, given that at least 24 monthly returns are reported, according to the following formula:
𝛽̂𝑀𝑖= 𝑐𝑜𝑣(𝑅𝑖,𝑅𝑚)
𝜎2𝑅
𝑚 , (18)
where the covariance of a stock return and the market return, the MSCI Europe index, is divided by the variance of the market return. The estimated market beta equals the slope coefficient of a simple OLS regression of stock returns on the market return. Following Fama and French (1992), who argue portfolio betas are more precise, each stock is assigned to a portfolio based on size and beta ranking. Each year in June, stocks are divided into ten size groups. Sequentially, the size groups are further divided into ten groups based on the ranking of betas, which results in 100 portfolios based on size and beta ranking. The portfolio beta is the equal-weighted average of the firm betas and is updated each month.
4. Data
24 data is collected for European firms from 16 countries from 1985 to 2016. ORBIS is used to find all publicly listed firms in these countries. Firms that are not found in DataStream or have no ISIN number in ORBIS are excluded from the sample.
To ensure that accounting information can possibly be incorporated in firms’ stocks, Fama and French (1992) is followed to match accounting information and market
capitalization for all fiscal year-ends in calendar year t – 1 with stock returns from July of year t to June of year t + 1. Therefore, a minimum time gap of six months exists between the fiscal year-end data and the stock returns. The time gap is larger for firms that have a fiscal year-end date before December.
The sample period and countries are mainly based on Fama and French (2012), who study several asset pricing models internationally. The sample period is relatively short compared to similar studies on stock returns that focus on the US, but it ensures a broad coverage of the stock and accounting data, and the use of factor returns from the Kenneth French website4. Specifically, the sample period of the stock returns in the regressions is from July 1990 to June 2016, unless otherwise notified, and stock returns from July 1985 onwards are used to calculate market betas for the cross-sectional regressions. Moreover, the
accounting data and market capitalization are collected for years 1989 to 2015.
The countries in the sample represent the large countries from Western Europe. These countries are relatively more developed, more innovative and are expected to have higher data quality relative to excluded European countries. In general, European countries that are
excluded are small or eastern European countries5. Norway and Switzerland are in the sample, but are not member states of the European Union. However, both countries are geographically part of Western Europe and participate in most of the EU’s open market provisions.
Furthermore, the United Kingdom is a member of the European Union throughout the whole sample period. Thus, it is a fair assumption that the sample countries are well integrated. For a complete list of sample countries, see Table 13 in Appendix B.
The sample contains all publicly listed firms in ORBIS from the countries listed in Table 13 in Appendix B. Data is collected for both R&D firms and non-R&D firms, while focusing on R&D firms for the analysis of R&D intensity and stock returns. Non-R&D firms are used to determine portfolio breakpoints for the degree of competition, calculate the degree of competition at the industry level, and in the independent competition-return analyses. A
4http://mba.tuck.dartmouth.edu/pages/faculty/ken.french/data_library.html
5 Notable exceptions are: Bulgaria, Croatia, Cyprus, Czech Republic, Estonia, Hungary, Latvia, Lithuania,
25 large firm sample is necessary to measure the degree of competition at the industry level, since the industry competition measures rely on market shares. Financial firms and firms from (previously) regulated markets are excluded, following common practice in the literature on competition. Therefore, firms from industries such as rail- and air transportation,
telecommunications, and energy are not taken into account. These industries may be less competitive, due to the fact that they were regulated during the sample period or were not privatized long enough before the sample period to be considered competitive markets.
The US dollar is used as currency for the total return index, market capitalization, and the accounting data. Fama and French (2012) also use US dollar returns and a US risk free rate to create factor returns for the European stock market. Therefore, using the US dollar as currency is not only for comparison of data, it also ensures that the stock returns are consistent with the factor returns. Arguably, this approach might induce more exchange rate risk than choosing national currencies, one of the national currencies (e.g. the euro would be the most appropriate currency for this sample), or focus on countries with one shared currency (i.e. all euro countries). According to Fama and French (2012), however, exchange rate risk can be ignored, under the strong assumptions that purchasing power between currencies does not change and that stocks cannot be used to hedge exchange rate risk.
The factor return data and the risk free rate, which are used in the asset pricing models to analyze portfolio returns, are from the Kenneth French website. The factor returns are based on stocks from the European countries used in Fama and French (2012). The construction of the factor returns is explained in more detail on the website and in the corresponding papers (i.e. Fama and French, 1993; 2012; and 2015). The factor returns are simple returns, which are translated to logarithmic returns following Equation 23 in Appendix B. The risk free rate is the US one month T-bill rate. Instead of the market return from the Kenneth French website this paper uses the MSCI Europe index as market return.
Furthermore, annual accounting data is necessary to measure firms’ R&D intensity and degree of competition. R&D intensity requires R&D expenditures, and market
capitalization or sales. The degree of competition is derived from market shares and implied market power, which means that data such as sales, operating profit, and industry
classification codes must be available for each firm. Firm-year observations are dropped if any of the necessary accounting data is missing or inconsistent. The Thomson Reuters
Business Classification (TRBC) is used to classify a firm’s main industry, which distinguishes 136 different industries. The TRBC allows for an industry classification that is not too
26 can be falsely assumed to operate in different markets. However, if the industry classification is too broad, non-competing firms can be falsely assumed to operate in the same market. For comparison, the two digit SIC (Standard Industry Classification) and the NACE (EU’s industry classification) result in a too broad industry classification, whereas the three digit SIC results in a too narrow industry classification. Another disadvantage of a too narrow industry classification is that diversified firms are allocated to one industry, ignoring activities in other industries.
5. Analysis
This chapter presents the findings on the relation between R&D and competition to stock returns using two empirical methods. Section 5.1 presents the findings of the first method; the portfolio analyses. Section 5.2 presents the findings of the second method; the cross-sectional regressions. Lastly, section 0 describes whether the results are robust to alternative specifications.
5.1. Results method one: Portfolio analyses
This section presents the findings of the portfolio analyses on the R&D-return relation, the competition-return relation, the R&D-return relation in concentrated and competitive markets, and the competition-return relation for different portfolios with respect to R&D intensity.
5.1.1 R&D and stock returns
This subsection describes the relation between R&D and stock returns, ignoring the degree of competition. More formally, this subsection tests Hypotheses 1 and 2 using portfolio analysis.
First off, Hypothesis 1 is tested by comparing the stock returns of R&D firms and non-R&D firms. Table 1 reports the (abnormal) returns for portfolios containing stocks of non-R&D firms and non-R&D firms.
27 between these portfolios is 0.08% with a t-statistic of 0.45. The monthly spread is lower for equal-weighted portfolios, namely 0.02% with a t-statistic of 0.12, which is probably explained by the fact that the non-R&D portfolio is dominated by small firms. By construction, equal-weighted are more affected by the size effect than value-weighted portfolios. A premium for small firms may cause the monthly non-R&D portfolio return (0.36%) to be closer to the monthly R&D portfolio return (0.38%). Thus, the size effect might mitigate the potential R&D premium for the equal-weighted R&D portfolio.
Table 1: Performance of R&D and non-R&D portfolios.
R&D Non-R&D Δ R&D
Panel A: value-weighted returns
Raw Return 0.32*** 0.24* 0.08
FF three-factor α 0.10*** 0.01 0.09**
Carhart four-factor α 0.10*** 0.02 0.08*
FF five-factor α 0.07** -0.03 0.09**
Panel B: equal-weighted returns
Raw Return 0.38*** 0.36*** 0.02
FF three-factor α 0.14*** 0.13*** 0.01
Carhart four-factor α 0.16*** 0.13*** 0.04
FF five-factor α 0.15*** 0.12*** 0.03
Notes: This table provides the results for the portfolio analysis on R&D and non-R&D firms, which tests Hypothesis 1. Logarithmic monthly raw returns and abnormal returns (α) are reported for R&D and non-R&D portfolios, along with the difference between the two (Δ R&D). The raw portfolio returns are calculated according to Equation 7. The abnormal returns are computed by time-series OLS regressions of the asset pricing models described in Equations 8 to 10. The panels differ with respect to value- or equal-weighting of portfolios and portfolio allocation based on R&D intensity scaled by market capitalization (MC) or sales. The sample period is from July 1990 to June 2016 (312 months). An overview of the stock characteristics is presented in Table 2. Returns are presented with significance at the 1%, 5%, and 10% is indicated by ***, **, and *, respectively, accompanied by t-statistics in parentheses.
28 As is clear from Table 2, the average market capitalization of the non-R&D portfolio is much lower than the average market capitalization in the R&D portfolio, which indicates that the return difference could be partly explained by the size effect. In other words, the size premium may overshadow the R&D premium, which is especially plausible among equal-weighted portfolios. Moreover, the fact that R&D portfolio contains bigger firms suggests that big firms are more transparent with respect to the disclosure of R&D expenditures. As a consequence, the equal-weighted portfolios may contain small ‘hidden’ R&D firms. Throughout the sample period, the portfolio of non-R&D firms contains on average 1,346 stocks, where the R&D portfolio averages 623 stocks. More specifically, the ratio of non-R&D firms to non-R&D firms decreases from four to two during the sample period, which indicates that the number of firms that disclose R&D expenditures or the number of R&D firms increases during the sample period.
The findings from Table 1, given the plausible bias towards small firms among equal-weighted portfolios, suggest a rejection of Hypothesis 1. Thus, although the evidence is not strong, the findings indicate that R&D firms receive higher stocker returns than non-R&D firms. For comparison, Chan, Lakonishok, and Sougiannis (2001) find that stock returns of R&D firms match the stock returns of non-R&D firms in the US. However, they also report higher abnormal stock returns for R&D firms.
Furthermore, Hypothesis 2 is tested by analyzing the performance of five portfolios, which are sorted on R&D intensity scaled by market capitalization or sales.
As Table 3 shows, the portfolio returns are positive and slightly increasing as R&D intensity increases, which is confirmed by the positive high-minus-low spread. More specifically, the high-minus-low spread is positive for both R&D intensity specifications. Moreover, the spread is small for value-weighted portfolios, but higher for equal-weighted portfolios, which may be caused by the size effect. Panel A and B present the performance of value- and equal-weighted R&D-to-market capitalization portfolios, respectively. For
29 Table 2: Data description of R&D portfolios.
Non-R&D 1 2 3 4 5 R&D
Panel A: R&D intensity, R&D expenditure to market capitalization ratio
R&D 0 32 64 127 219 370 162 MC 817 8,573 4,304 3,989 3,599 2,382 4,579 Sales 1,156 7,192 3,667 3,166 3,763 6,245 4,805 R&DMC 0 0.4 1.6 3.3 6.7 20.9 6.5 R&DS 0 2.0 4.5 8.1 11.4 18.7 8.9 PCM 10.2 15.4 13.4 12.3 9.9 6.6 11.5 IPCM 12.1 13.5 12.8 13.0 12.9 12.9 13.0 HHI 25.4 26.5 22.7 24.5 23.3 26.1 24.6 B/M 0.88 0.54 0.55 0.59 0.70 0.96 0.67 Leverage 1.63 0.89 0.84 0.96 1.16 2.12 1.19 Stocks 1,346 125 125 125 124 124 623
Panel B: R&D intensity, R&D expenditure to sales ratio
R&D 0 24 51 168 270 310 162 MC 817 6,949 3,515 4,015 3,717 4,704 4,579 Sales 1,156 7,160 4,059 5,532 4,447 2,635 4,805 R&DMC 0 0.9 2.9 5.8 10.3 13.6 6.5 R&DS 0 0.3 1.3 3.0 6.9 35.6 8.9 PCM 10.2 12.2 11.5 12.4 12.4 9.0 11.5 IPCM 12.1 12.0 11.8 12.4 13.1 16.2 13.0 HHI 25.4 24.4 24.0 22.9 24.3 27.7 24.6 B/M 0.88 0.81 0.75 0.62 0.64 0.50 0.67 Leverage 1.63 1.64 1.33 1.21 1.07 0.62 1.19 Stocks 1,346 127 128 128 125 115 623
Notes: This table describes the data for quintile R&D portfolios, R&D firms, and non-R&D firms. Panel A describes the data characteristics of R&D-to-market capitalization (MC) portfolios. Panel B describes the data of R&D-to-sales portfolios. R&D expenditures (R&D), MC, and Sales are presented in millions of US dollars. R&D-to-market capitalization ratio (R&DMC), R&D-to-sales ratio (R&DS), Price-cost-margin (PCM), Industry PCM (IPCM), and Hirschman-Herfindahl index (HHI) are presented in percentages. B/M stands for book-to-market ratio, and is calculated as book equity divided by market capitalization. Leverage is calculated as total assets minus book equity divided by market capitalization. The variables are explained in more detail in Chapter 3. Stocks refers to the monthly average number of stocks in the portfolios. The sample period is from July 1990 to June 2016 (312 months).
Next, the abnormal portfolio returns are characterized by a monotonic increase and a positive high-minus-low spread. This finding holds for value- and equal-weighted portfolios for each R&D intensity measure, but differs in significance. For example, Panel A in Table 3 shows that the monthly abnormal portfolio return spread ranges from an insignificant 0.07% for the three-factor model to a significant (at the 1% level) 0.27% for the five-factor model. As is clear from Panel B, the spread is significant at the 1% level for equal-weighted portfolios sorted by R&D relative to market capitalization. The monthly abnormal return spreads are 0.37%, 0.33%, and 0.40% for the three-, four-, and five-factor model,
30 reports less significant abnormal equal-weighted portfolio return spreads. Only the five-factor abnormal return spread, which equals 0.38%, is significant (at the 1% level).
Table 3: Performance of R&D intensity portfolios.
R&D Intensity Quintiles
1 (L) 2 3 4 5 (H) H-L
Panel A: value-weighted returns, R&D intensity scaled by MC
Raw Return 0.30** 0.32*** 0.33** 0.26* 0.41** 0.12
FF three-factor α 0.07 0.11** 0.10 0.03 0.14* 0.07
Carhart four-factor α 0.09 0.10* 0.09 0.12 0.19** 0.09
FF five-factor α -0.02 0.12 0.07 0.12 0.25** 0.27***
Panel B: equal-weighted returns, R&D intensity scaled by MC
Raw Return 0.19 0.31** 0.40*** 0.39*** 0.57*** 0.38** FF three-factor α -0.05 0.07* 0.17*** 0.15** 0.31*** 0.37*** Carhart four-factor α -0.01 0.09** 0.19*** 0.17*** 0.33*** 0.33*** FF five-factor α -0.04 0.06 0.16*** 0.19*** 0.36*** 0.40*** Panel C: value-weighted returns, R&D intensity scaled by sales
Raw Return 0.31** 0.36*** 0.31** 0.25 0.35*** 0.03
FF three-factor α 0.07 0.13** 0.05 0.02 0.18*** 0.11** Carhart four-factor α 0.09 0.10** 0.05 0.13* 0.16*** 0.07 FF five-factor α -0.01 0.01 -0.07 0.10 0.23*** 0.24*** Panel D: equal-weighted returns, R&D intensity scaled by sales
Raw Return 0.30** 0.37*** 0.36*** 0.44*** 0.37** 0.06 FF three-factor α 0.03 0.11** 0.11* 0.21*** 0.17 0.14 Carhart four-factor α 0.06 0.14*** 0.13** 0.25*** 0.20** 0.14 FF five-factor α 0.00 0.03 0.06 0.27*** 0.37*** 0.38*** Notes: This table provides the results for the portfolio analysis on the R&D-return relation, which tests Hypothesis 2. Logarithmic monthly raw returns and abnormal returns (α) are reported for quintile portfolios sorted on R&D intensity. The raw portfolio returns are calculated according to Equation 7. The abnormal returns are computed by time-series OLS regressions of the asset pricing models described in Equations 8 to 10. The panels differ with respect to value- or equal-weighting of portfolios and portfolio allocation based on R&D intensity scaled by market capitalization (MC) or sales. The sample period is from July 1990 to June 2016 (312 months). An overview of the stock characteristics is presented in Table 2 and Figure 2. Returns are presented with significance at the 1%, 5%, and 10% is indicated by ***, **, and *, respectively, accompanied by t-statistics in parentheses.
Furthermore, Table 2 provides the data description for the R&D-to-market
capitalization portfolios in Panel A and R&D-to-sales portfolios in Panel B. An interesting fact is that higher R&D intensity portfolios also have higher R&D expenditures. For example, see Panel A, despite having the least market capitalization, most R&D expenditures are done by firms allocated to the fifth R&D portfolio. The high R&D-to-market capitalization
31 Furthermore, the average PCM of the market capitalization based R&D portfolios is
decreasing, which might suggest that R&D-intensive firms face more competitive threats. Moreover, the average R&D-to-sales is affected by outliers (if not bounded at one), which are most likely small firms that wait for future sales following from current R&D projects.
Furthermore, Figure 2 in Appendix B shows the number of stocks underlying the quintile R&D portfolios. The number of underlying stocks increases each month and is always above 22 stocks, ensuring that the portfolios represent the respective R&D intensity group well.
Evidently, higher levels of R&D intensity are associated with higher (abnormal) stock returns, regardless of the measurement of R&D intensity or portfolio returns, which suggests that Hypothesis 2 holds, indicating a positive R&D-return relation. However, only the five-factor model significantly supports the positive R&D-return relation according to each specification. Similarly, Chan, Lakonishok, and Sougiannis (2001) find that R&D intensity explains stock returns in the US, most notably if portfolios are sorted on R&D relative to market capitalization. They find that high R&D relative to sales portfolio tend to pick up ‘glamour’ stocks that do not necessarily have poor past returns, whereas high R&D relative to market capitalization portfolios tend to pick up value stocks that have poor past returns.
5.1.2 Competition and stock returns
This subsection tests whether Hypothesis 3 holds, which implies a positive competition-return relation. Table 4 presents the (abnormal) competition-returns for five portfolios sorted on the degree of competition derived from PCM, IPCM, and HHI, respectively. The first portfolio contains stocks of firms from concentrated markets, i.e. markets with low competition. The fifth portfolio contains firms from competitive markets.
-32 0.02%, respectively. Similarly, as Panel E and F report, the monthly value- and
equal-weighted HHI portfolio returns are -0.09% and -0.06%, respectively. The size effect most likely explains the higher return spreads among equal-weighted portfolios.
Table 4: Performance of competition portfolios.
Competition Quintiles
1 (L) 2 3 4 5 (H) H - L
Panel A: value-weighted returns, PCM
Raw return 0.32*** 0.31** 0.32** 0.22 0.04 -0.29
FF three-factor α 0.12*** 0.08** 0.06* -0.05 -0.22*** -0.34*** Carhart four-factor α 0.12*** 0.08** 0.08** 0.05 -0.11 -0.22**
FF five-factor α 0.08** 0.00 0.05 0.05 0.01 -0.07
Panel B: equal-weighted returns, PCM
Raw return 0.38*** 0.42*** 0.42*** 0.35*** 0.22 -0.15 FF three-factor α 0.15*** 0.18*** 0.19*** 0.11*** 0.01 -0.14** Carhart four-factor α 0.15*** 0.19*** 0.19*** 0.10*** 0.02 -0.12* FF five-factor α 0.12*** 0.15*** 0.17*** 0.08** 0.12 0.00
Panel C: value-weighted returns, IPCM
Raw return 0.34*** 0.33*** 0.27* 0.24 0.19 -0.14
FF three-factor α 0.14*** 0.09* 0.05 0.00 -0.06 -0.20*** Carhart four-factor α 0.13*** 0.07 0.05 0.06 0.01 -0.12** FF five-factor α 0.13*** -0.02 -0.05 0.08 -0.01 -0.13**
Panel D: equal-weighted returns, IPCM
Raw return 0.35*** 0.38*** 0.35*** 0.41*** 0.33** -0.02 FF three-factor α 0.14*** 0.14*** 0.10*** 0.18*** 0.08** -0.06 Carhart four-factor α 0.12*** 0.14*** 0.11*** 0.18*** 0.10** -0.02 FF five-factor α 0.16*** 0.13*** 0.06 0.17*** 0.12*** -0.03
Panel E: value-weighted returns, HHI
Raw return 0.28** 0.36*** 0.29** 0.31*** 0.19 -0.09
FF three-factor α 0.06 0.12** 0.07 0.08* -0.08* -0.14** Carhart four-factor α 0.06 0.12** 0.09* 0.08* -0.02 -0.08*
FF five-factor α 0.06 0.07 0.05 0.01 -0.12** -0.18**
Panel F: equal-weighted returns, HHI
33 Furthermore, the high-minus-low abnormal return spreads are negative or close to zero for every portfolio specification. The value-weighted portfolios in Panel A (PCM), C (IPCM), and E (HHI) show that abnormal returns are clearly declining as competition increases. Also, the abnormal returns are often negative for high competition portfolios, leading to negative high-minus-low spreads. In contrast, the equal-weighted portfolios in Panel B (PCM), D (IPCM), and F (HHI) show declining abnormal returns as competition increases, which are significantly positive to around zero for high competition portfolios. The equal-weighted portfolios are worse explained by the asset pricing models, probably because of the size effect. For example, the monthly three-factor abnormal return spread for the value-weighted PCM portfolio is 0.34% and significant at the 1% level. For comparison, this spread is -0.20% and -0.14% and also significant for the value-weighted IPCM and HHI portfolios, respectively. In contrast, among the equal-weighted portfolios this spread is only significant for the PCM portfolio, namely 0.14% with a 5% significance level, which is 0.06% and -0.09% with a 10% significance level for the IPCM and HHI portfolios, respectively.
Moreover, Table 5 describes the data characteristics of the competition portfolios. Notably, the high competition portfolio contains stocks of relatively small firms, indicated by the low average market capitalization. The R&D-to-market capitalization ratios are similar among the competition portfolios, although average R&D expenditures are higher for less competitive markets. Furthermore, firms in competitive markets tend to have high leverage ratios, which can be explained by lower profits and the corresponding need for external funds. Additionally, Figure 3 in Appendix B shows that the number of stocks in the portfolios is above 80 throughout the whole sample period.
