40126, Bologna, Italy b Bangor Business School, Bangor University, Hen Goleg, Room 1.07, College

Transcription

1 This article was downloaded by: [Alma Mater Studiorum - Università di Bologna] On: 25 October 2013, At: 00:52 Publisher: Routledge Informa Ltd Registered in England and Wales Registered Number: Registered office: Mortimer House, Mortimer Street, London W1T 3JH, UK The European Journal of Finance Publication details, including instructions for authors and subscription information: Mispricing and risk of R&D investment in European firms Andi Duqi a, Aziz Jaafar b & Giuseppe Torluccio a a Department of Management, University of Bologna, Via Capo di Lucca, 34, 40126, Bologna, Italy b Bangor Business School, Bangor University, Hen Goleg, Room 1.07, College Road, LL57 2DG Bangor, UK Published online: 24 Sep To cite this article: Andi Duqi, Aziz Jaafar & Giuseppe Torluccio, The European Journal of Finance (2013): Mispricing and risk of R&D investment in European firms, The European Journal of Finance, DOI: / X To link to this article: PLEASE SCROLL DOWN FOR ARTICLE Taylor & Francis makes every effort to ensure the accuracy of all the information (the Content ) contained in the publications on our platform. However, Taylor & Francis, our agents, and our licensors make no representations or warranties whatsoever as to the accuracy, completeness, or suitability for any purpose of the Content. Any opinions and views expressed in this publication are the opinions and views of the authors, and are not the views of or endorsed by Taylor & Francis. The accuracy of the Content should not be relied upon and should be independently verified with primary sources of information. Taylor and Francis shall not be liable for any losses, actions, claims, proceedings, demands, costs, expenses, damages, and other liabilities whatsoever or howsoever caused arising directly or indirectly in connection with, in relation to or arising out of the use of the Content. This article may be used for research, teaching, and private study purposes. Any substantial or systematic reproduction, redistribution, reselling, loan, sub-licensing, systematic supply, or distribution in any form to anyone is expressly forbidden. Terms & Conditions of access and use can be found at

2 The European Journal of Finance, Mispricing and risk of R&D investment in European firms Andi Duqi a, Aziz Jaafar b and Giuseppe Torluccio a a Department of Management, University of Bologna, Via Capo di Lucca, 34, 40126, Bologna, Italy; b Bangor Business School, Bangor University, Hen Goleg, Room 1.07, College Road, LL57 2DG Bangor, UK (Received 8 June 2012; final version received 22 August 2013) We study whether R&D-intensive firms earn superior stock returns compared to matched size and bookto-market portfolios across several financial markets in Europe. Mispricing can arise if investors are not able to correctly estimate the long-term benefits of R&D investment or whether R&D firms are more risky than others. The results confirm that more innovative firms can earn future excess returns. Stocks listed on continental Europe markets and operating in high-tech sectors are more prone to undervaluation. This can be caused in the first case by information asymmetries that are more severe in bank-based countries. No evidence is found for a different risk pattern of R&D-intensive stocks. Keywords: European financial markets; Fama and French; mispricing; R&D; risk; stock returns JEL Classifications: G11; G15; O30 1. Introduction Prior research studies have documented a positive link between R&D investments and firm market value (e.g. Griliches 1981; Hall 1993; Chauvin and Hirschey 1993; Hall and Oriani 2006; Pindado, de Queiroz, and de la Torre 2010). Furthermore, empirical evidence using US and UK datasets shows a significant relationship between R&D and expected stock returns (Lev and Sougiannis 1996; Chan, Lakonishok, and Sougiannis 2001; Chambers, Jennings, and Thompson II 2002; Dedman et al. 2009). This association has been attributed to a possible mispricing effect due to the failure of investors to correctly estimate the effects of R&D on future firm cash flows or to a different risk pattern for investment in R&D vis-à-vis tangible assets. Several factors contribute to the complexity of assessing the present value of R&D investments. First, R&D-intensive firms generally have few tangible assets; the future benefits from R&D programs are far from certain as the cash flows from these projects are difficult to evaluate (Aboody and Lev 2000; Chan, Lakonishok, and Sougiannis, 2001; Kothari, Laguerre, and Leone 2002). This may lead to an undervaluation of the net value of these expenditures if the longterm benefits of R&D are not considered exhaustively, or to overpricing if investors inflate the market value of R&D-intensive firms by being too optimistic about firm growth related to R&D activity (Al-Horani, Pope, and Stark 2003; Chiao, Hung, and Lee 2008). Second, under US GAAP, R&D costs are completely expensed in the year in which they are incurred, whereas, under IASB standards, capitalization is permitted for development expenditures only when a clear connection can be demonstrated between these costs and a hypothetical future product. 1 However, because this rule is difficult to apply, most of listed firms completely expense R&D in the year it is Corresponding author. andi.duqi@unibo.it 2013 Taylor & Francis

3 2 A. Duqi et al. incurred. Investors can erroneously produce high multiples for these firms if they fail to correct accounting variables for the long-term benefits of R&D. Finally, they can also be misled by the past performance of stocks, demonstrating an excessive preference for past winners that could suffer momentum reversals in the future (Lakonishok, Shleifer, and Vishny 1994; Chan, Lakonishok, and Sougiannis 2001). The risk-pattern approach draws on Fama and French s (1992) seminal work, which suggests that firm size and the book-to-market (BM) ratio capture a part of the firm s risk that capital assets pricing model (CAPM) is not able to explain. Lev and Sougiannis (1996) show that the inclusion of an R&D variable increases the overall predictive power of the Fama and French model. R&D projects may contribute significantly to a firm s business risk and systematic risk in a way that is not necessarily completely attributable to size and the BM ratio. Hence, adjusting realized returns only for those two factors is unlikely to completely control for firm risk (Chambers, Jennings, and Thompson II 2002). The present paper aims to examine the impact of R&D on stock returns across European financial markets, and to determine whether there is significant mispricing of R&D. There is a scarcity of research examining the effect of R&D on firm value for firms located in these countries and, with the exception of preliminary work by Duqi, Mirti, and Torluccio (2011), there is no empirical study investigating the relationship between R&D investment and abnormal stock returns. The last two decades have witnessed a substantial growth in the number of firms investing in innovation across Europe. Further, R&D is perceived as a strong value driver by European institutions, policy-makers and corporations. 2 Moreover, the institutional factors in the countries in this study are different from one another in terms of economic systems, legal and accounting environments, corporate governance, and dominant shareholder type. Each of these variables influences the development and efficiency of the financial markets, the funding of investment in innovation, and, indirectly, the pricing of the listed firms assets. An appropriate valuation of the R&D impact on stock returns across Europe could assist investors in their evaluations of future investment strategies. It could also provide managers with useful insights into the long-term benefits of R&D for stock prices and returns. From a data set of firms listed in five European countries, i.e. Finland, France, Germany, Sweden, and the UK, our main empirical results show that R&D expenditures are relevant in predicting future stock returns, thus confirming the results of prior studies using US and UK data sets. This relationship is influenced by sector innovativeness; that is, the R&D effect is stronger in highly innovative industries. We find that R&D-intensive firms earn subsequent excess returns for one and two years after portfolio formation, indicating potential stock mispricing. Further, we test for risk differentials based on whether a firm is highly R&D-intensive. The results indicate that R&D is positively related to stock volatility, especially for UK firms. However, there is no evidence that R&D reflects a different source of systematic risk unaccounted for by other models. Investors do not assign a higher risk premium to R&D-intensive stocks after controlling for size and the BM ratio. The paper contributes to the extant literature in several ways. First, to the best of our knowledge, this is the first study to provide a comprehensive analysis of the impact of R&D on abnormal stock returns across major European financial markets. In so doing, we use different models to test whether there is potential mispricing of R&D, and whether this is due to any risk-bearing unaccounted for by other empirical models. We extend the initial study by Duqi, Mirti, and Torluccio (2011) by carefully distinguishing whether a positive effect of R&D on firm returns is due to mispricing or to a different degree of riskiness of innovative stocks. Our empirical analysis also considers both stock volatility and possible variations between market-based and bank-based

4 The European Journal of Finance 3 countries. Furthermore, we document evidence that R&D-intensive stock returns are affected by the sector in which firms operate and by the firms country of domicile. Mispricing is more pronounced in sectors with a higher propensity to innovate. Moreover, highly R&D-intensive firms listed in continental Europe experience higher abnormal returns than those located in the UK. This is plausibly due to information asymmetries, which are more pronounced in stocks in bank-based vis-à-vis market-based countries. The remainder of the paper is organized as follows. Section 2 discusses previous research on the R&D effect on stock returns, and prior literature about the peculiarities of European markets and how they impact on firm innovation. Section 3 focuses on the methodology used. Section 4 describes the sample and reports several descriptive statistics. In Section 5, we exhibit our empirical estimates of R&D mispricing, excess returns, R&D systematic risk, and the R&D effect on price volatility. Additional robustness analyses are presented in Section 6. Finally, Section 7 offers some final remarks and conclusions. 2. The impact of R&D on stock returns Prior studies on the link between R&D and stock returns focus mainly on anomalies in asset pricing models (e.g. CAPM) and almost exclusively use US datasets. Fama and French (1992, 1993) show that beta and CAPM are poor predictors of the stock market returns of US firms, especially after 1980, and that other variables, such as firm size, BM ratio, earnings-to-price ratio, and firm leverage, appear to be more powerful at capturing the variation in stock returns over the period Building on Fama and French (1992, 1993), a plethora of extant research attempts to link intangible assets, in particular R&D, with future stock returns. Lev and Sougiannis (1996, 1999) provide evidence that the Fama and French (1992) model improves considerably when the variable R&D stock is added. As the BM ratio captures all firm growth options, it might be possible to substitute it with a measure of R&D stock, which is a primary driver of innovation and directly affects firm growth. Lev and Sougiannis leave as an open question whether this is due to a mispricing bias regarding R&D expenditures or a failure to account for some risk that is not captured by other factors. It is commonly accepted that intangible assets are more risky and more difficult to evaluate than physical assets. The variability of future earnings consequent to R&D investment is greater than that attributed to capital expenditure (Kothari, Laguerre, and Leone 2002). Chan, Lakonishok, and Sougiannis (2001) report significant abnormal returns for high-r&d firms compared to non-r&d ones, and also a positive association between R&D intensity and future stock volatility, but they reach the conclusion that this mispricing is similar to that explained by Lakonishok, Shleifer, and Vishny (1994). In other words, firms with a high ratio of R&D to market equity are generally poor past performers that tend to be undervalued by investors. Chambers, Jennings, and Thompson II (2002) show that more innovative firms tend to gain abnormal returns that may persist for up to 10 years. This is consistent with the risk scenario, as asset mispricing is unlikely to endure for such a long period. The mispricing scenario is also supported by Eberhart, Maxwell, and Siddique s (2004) study, which shows consistent evidence that investors systematically underreact to announcements of R&D increases. This leads to significant abnormal returns for firms in the top R&D portfolios. Lev, Sarath, and Sougiannis (2005) and Ciftci, Lev, and Radhakrishnan (2011) support an alternative theoretical approach compared to previous research. R&D expenditure may incur higher business risks, similar to that estimated by Kothari, Laguerre, and Leone (2002) or information risk as

5 4 A. Duqi et al. introduced by Easley, Hvidkjaer, and O Hara (2002) and Easley and O Hara (2004). The latter can be mitigated by extensive disclosure of the future benefits deriving from the R&D activity. Otherwise, it may lead to a significantly incorrect valuation by investors or analysts due to the difficulty of measuring these benefits correctly. They document that more innovative firms do not gain excess returns due to a failure to account for systematic risk, because the mispricing is observed in the first years after portfolio formation and tends to disappear in the long term. A few studies have attempted to investigate the relationship between R&D and stock returns in other countries but the results are still inconclusive. Al-Horani, Pope, and Stark (2003) report that the cross-section of UK expected stock returns is positively related to R&D activity, after controlling for size and BM ratio. Dedman et al. (2009) extend the work of Al-Horani, Pope, and Stark (2003), studying Fama and French s (1993) three-factor model in cross-industry portfolios. R&D has a positive effect on stock returns for 13 of their 20 industrial portfolios, suggesting that it can be a useful factor in pricing assets. They also confirm that investors are not misled by the fact that R&D expenditures are expensed in the year in which they are incurred because investors consider R&D as an asset. Xu and Zhang (2004) and Nguyen, Nivoix, and Noma (2010) find no evidence of a R&D effect generating abnormal returns for Japanese firms; they find no undervaluation of R&D investments except in the post-bubble period ( ). Chiao and Hung (2005) and Chiao, Hung, and Lee (2008) study the Taiwan stock market; they highlight an evident mispricing of R&D expenditure, which persists for up to three years, especially for electronics firms. Prior studies on the impact of R&D on the market performance of continental European firms have been hampered by the diversity of national accounting rules, which, unlike those in the USA, do not require comprehensive disclosure of this type of investment. Only in the last 15 years has there been a significant effort to harmonize the rules, adopt international best practices, and truly integrate different markets so as to make them more efficient in terms of asset pricing. Further, since 2005, all listed firms on European markets have adopted the IASB financial reporting standards, aimed at reconciling the financial information supplied to investors. 3 Prior studies using European datasets have reported the positive and significant effect of R&D on market value. However, the positive outcomes of R&D investments can be significantly moderated by firm-specific characteristics such as size, corporate governance, cash flows, or firm growth, and by country-specific characteristics (Toivanen, Stoneman, and Bosworth 2002; Hall and Oriani 2006; Pindado, de Queiroz, and de la Torre 2010; Munari, Oriani, and Sobrero 2010). Corporate governance in continental European countries is significantly different from that in the USA and the UK. There is a huge presence of insider shareholders, who usually control more than 50% of the voting rights (Faccio and Lang 2002; Tylecote and Ramirez 2006). These shareholders might be banks in Germany, the state in France, or family shareholders in Italy. The Nordic countries lie somewhere in between Germany and the UK, having a large proportion of insiders as well as institutional investors. The existence of different types of investors with different investment horizons (Bae and Kim 2003; Munari, Oriani, and Sobrero 2010) or different market, legal, or financial frameworks (Hall and Oriani 2006) can influence the market valuation of R&D (Booth et al. 2006). R&D investment contains a high degree of information asymmetry; insiders usually have more information about its potential outcomes in the long term (Aboody and Lev 2000). These complexities are more likely to arise in countries where insiders can control large stakes in companies, such as in continental Europe and in places where the private benefits of control are higher (Dyck and Zingales 2004). Mispricing should be more pronounced in Europe, where disclosure is more problematic because of the financial reporting rules and practices. Investors in Europe could suffer

6 The European Journal of Finance 5 from information risk more than those in Anglo-Saxon markets, because such risk increases in the presence of relevant inside information and low disclosure. Information risk significantly affects asset pricing because uninformed investors will require a higher rate of return for holding stocks with a higher degree of private information (Easley and O Hara 2004; Ciftci, Lev, and Radhakrishnan 2011). We would expect the stock price volatility of continental European firms that invest more in R&D to be higher than that of non-r&d firms but lower than that of innovative firms located in the UK, ceteris paribus, because in Anglo-Saxon economies institutional investors put more pressure on managers and are more unwilling to accept the short-term losses that are frequent in high-tech sectors, which are commonly considered as the most R&D intensive (Tylecote and Ramirez 2006). These investors might suffer from a so-called myopic view, which forces them to overreact by selling loser stocks and buying winners. In bank-based countries, inside shareholders are more inclined to accept long-term investments; they do not tend to calibrate their portfolios frequently following periodic releases of information from firm managers. Hence, stock price variability should be lower. In the literature, Sias (1996) and Bushee and Noe (2000) provide empirical evidence that higher institutional ownership is associated with higher stock return variability. There is also evidence that in several European countries, including those considered in the present paper, investors assign a risk premium to size and the BM ratio, but the results are still inconclusive. Fama and French (1998) find that their model can explain stock returns across countries, including European markets. Value stocks with high BM ratios outperform growth stocks in 12 of the 13 international markets they study. Similar results are reported for the French stock market by Lajili-Jarjir (2007) and Chahine (2008). However, Malin and Veeraraghavan (2004) do not produce such results when they compare the UK with French and German stock exchanges. They demonstrate a small size effect in France and Germany, but no value effect for any of the three markets. Previous research has pointed out that the mispricing of R&D could derive from a failure by investors to control for a non-diversifiable source of risk that is intrinsic in the asset (R&D) and cannot be captured by other factors. If this is true, then the moderating effect of R&D in the Fama and French (1993) model should not be any different for continental European firms than that observed in the USA or the UK, because of the similar risk characteristics of highly innovative firms in the USA and Western Europe. 3. Research methodology In order to detect any mispricing of R&D on the stock market and the potential impact of R&D on stock returns, we initially follow Lev and Sougiannis (1996, 1999) approach. We regress (cross-sectionally for each country) future stock returns on R&D expenses, after controlling for firm size, beta, and the BM ratio. The model is as follows: ( ) RD R i,t+6 = α 0 + α 1 β i,t + α 2 Ln(ME) i,t + α 3 Ln(BM) i,t + α 4 Ln + ε i,t+6, (1) ME i,t where R i,t+6 is the stock return of firm i six months after the disclosure of its financial statements for the year t, beta is estimated based on the Fama and French (1992) approach, using between 24 and 60 monthly returns in the five years up to the end of June of the year t, ME i,t is the market capitalization of firm i at the end of year t, BM i,t is the BM ratio for firm i in year t, and RD i,t is the annual R&D expenditure by firm i in year t. We run monthly regressions from 1999 to 2010 following the Fama and MacBeth (1973) model. In a successive set of regressions, we

7 6 A. Duqi et al. augment model (1) by including leverage, E(+)/ME, which is earnings/price ratio when earnings are positive and zero otherwise, and E( )/ME, a dummy variable which is equal to 1 if earnings of firm i are negative and zero otherwise. Next, we estimate the excess returns over and above R&D stocks. Following prior research (Chan, Lakonishok, and Sougiannis 2001; Chambers, Jennings, and Thompson II 2002), we control for size and the BM ratio. In particular, at the end of June of each year t in the period , stocks for each country sample are allocated into five size-based portfolios, using market capitalization (ME). Then, the stocks in each size quintile are sorted into five BM-based portfolios, using the BM ratio at the end of year t 1. All portfolios are rebalanced annually. Every stock is matched to a benchmark portfolio containing firms of a similar size and BM ratio. Buy-and-hold abnormal returns are calculated in June of the first, second, and fifth year as the difference between the stock s return and the return of the portfolio to which it is matched. Afterwards, the stocks are split into four groups based on their RD/ME ratio, and the average one-year, two-year, and five-year excess returns are calculated for each group. Non-R&D firms are treated separately. Following this step, we examine whether there is a potential mispricing of R&D stocks deriving from a systematic source of risk that is not captured by Fama and French s (1993) three-factor model. To do so, at the end of June of each year from 2000 to 2010, each country sample is sorted by market capitalization and the stocks are allocated into one of two size groups, small (S) and big (B), based on whether the firm s market equity is below or above the median. The stocks are also allocated independently into three BM ratio groups, low (L), medium (M), and high (H). Six portfolios are created from the intersection of these size and BM groups (S/L, S/M, S/H, B/L, B/M, B/H), and value-weighted returns are calculated from July of year t to June of year t + 1. The portfolios are rebalanced in June of each year t, so that the BM ratio in year t 1 is known (Fama and French 1993). The small minus big (SMB) portfolio is then defined as the simple monthly average of the differences between the returns on the S/L, S/M, S/H portfolios and the returns on B/L, B/M, B/H portfolios in an attempt to measure the effect of size, since the BM ratio effect should be nullified by taking the difference. In the same way, high minus low is the monthly average of the differences between the returns on S/H and B/H and the returns on S/L and B/L. It represents the BM risk factor. We then sort each country sample by RD/ME and divide each into four groups (1 being the least R&D-intensive and 4 being the most). Non-R&D firms are allocated to a separate portfolio. Portfolios are created in June of each year t, and the returns are calculated over the next 12 months. The time series equation for each portfolio i is as follows: r i,t r f,t = α i + β i (r m,t r f,t ) + γ i SMB t + δ i HML t + ε i,t, (2) where r i,t is the average monthly return for portfolio i, r f,t is the risk-free rate of return for each country 4 in month t, and r m,t is the market value-weighted return for month t. Finally, we include in our analysis the standard deviation of each stock, in order to assess whether more innovative stocks incur more business risk and whether this differs across countries. Following Chan, Lakonishok, and Sougiannis (2001), we estimate the effect of R&D on stock variability and control for size and the BM ratio. The economic rationale behind the idea of an R&D impact is that, ceteris paribus, more R&D-intensive firms should be riskier in terms of returns variability. At the end of June of each year t, we calculate the stock return variance for each firm over the next 12 months. We include in the sample all stocks in each country (whether they invest in R&D or not). The square root of this variance is defined as the total risk for firm i, in year t.

8 The European Journal of Finance 7 This variable is regressed on a measure of firm size (Ln(ME)), the logarithm of the BM ratio (Ln(BM)), and RD/ME. We use the Fama and MacBeth (1973) methodology to estimate the regression coefficients. The equation is as follows: σ i,t+6 = α 0 + α 1 Ln(ME) i,t + α 2 Ln(BM) i,t + α 3 RD/ME i,t + ε i,t+6. (3) In a further specification of this model, we deflate R&D expenditures by firm sales and total assets to ensure the robustness of the results. 4. Sample description Our initial sample comprises all listed firms in four EU countries, i.e. Finland, France, Germany, and Sweden, with coverage in the Datastream database 5 for the years In addition, a sample of UK firms was added for comparison purposes as the UK stock exchange has the highest number of listed stocks in Europe and also has a high percentage of R&D-intensive firms. We focus on these countries because they account for almost 70% of European listed firms, and the number of firms reporting R&D activity is higher as compared to other European countries. 6 Furthermore, we select the period after 1999 as the proportion of firms in continental Europe with R&D outlays is extremely low prior to that, i.e. only less than 20% of listed firms on average disclose R&D activities for each country in the period We excluded from the sample all financial, insurance, and real estate firms as the financial statements of these firms differ significantly in terms of financial reporting conventions and R&D investments. Following Dedman et al. (2009), we also dropped outliers by winsorizing the top and bottom 1% of all regression variables and firms with a negative price or negative book value of equity. Finally, we used the Bureau Van Dijk database to gather any missing data, especially R&D expenditure. At the end of this process, the final sample comprised 12,911 firm-year observations, of which only 4755 observations reported R&D activity. This consists of 79 Finnish, 278 French, 293 German, 122 Swedish, and 444 UK unique firms. Table 1 provides descriptive statistics for both the R&D and non-r&d firm-year observations for each country of domicile. It can be observed that R&D firms are on average larger than non-r&d ones across all stock markets. This is consistent with prior studies (e.g. Hall and Oriani 2006), which also document that R&D disclosure is related to size. The BM ratios of the R&D firms are lower; innovative firms tend to be growth stocks. Swedish and UK firms have the lowest BM ratios on average (0.390 and 0.469, respectively). French and German firms are the most leveraged, whereas UK firms use more equity. High debt-to-equity (D/E) ratios are common in bank-based countries such as France and Germany where firms typically finance their activities with debt. The fact that non- R&D firms use more debt is in line with economic theories that hold that R&D activity presents a high risk of moral hazard and information asymmetry, and thus it is preferable to finance it using equity (Williamson 1988; Hall 2002). The t-tests for differences in means between R&D and non-r&d groups are highly significant, except for leverage in the case of Finland and France, and the earnings-to-price ratio for Finland. On average, less than 55% of the firms in the sample disclosed that they had carried out R&D during the study period. As mentioned earlier, this might be due to the fact that the disclosure of R&D is not compulsory across Europe. Nevertheless, the overall trend of disclosing R&D activities is increasing in all countries. For instance, the percentage of firms that reported R&D in 1999 was 35% and 29%, and at the end of our sample period, it grew to 67% and 50% for firms located in Finland and Germany, respectively.

9 Table 1. Descriptive statistics. N Ln(ME) BM E/P D/E Country of domicile R&D Non R&D R&D Non R&D p-value R&D Non R&D p-value R&D Non R&D p-value R&D Non R&D p-value FIN FRA GER SWE UK Notes: This table reports mean values of R&D and non-r&d firms by country. N is firm-year observations for firms reporting positive R&D (R&D) and those that do not report any R&D spending during the year (non-r&d); Ln(ME) is the natural log of market capitalization; BM is the book-to-market ratio; E/P is the earnings-to-price ratio; and D/E is the total debt-to-equity ratio. p-values are for t-tests of differences in means between R&D and non-r&d firms. The sample period is Table 2. RD/ME by country of domicile. Country of domicile Mean Median Std. Dev. Min Max FIN FRA GER SWE UK Notes: This table reports descriptive statistics of RD/ME by country of domicile. RD/ME is a ratio of R&D expenditures to market capitalization. The sample period is A. Duqi et al.

10 The European Journal of Finance 9 Table 2 provides descriptive statistics for the firms RD/ME for each country of domicile. The mean values are not remarkably different from one another because of the deflative effect of ME. The other statistics show a strong value asymmetry; that is, the median is close to zero for almost all countries, much smaller than the average values. The RD/ME values are most variable in France and least in the UK. French and German firms have the highest levels of RD/ME in terms of absolute value; the UK firms on average invest less relative to their market value of equity. 5. Empirical analysis 5.1 Cross-sectional regression of stock returns on R&D To analyze whether R&D is a significant factor in explaining future stock returns over and above firm beta, size, and the BM ratio, we apply model (1) asinlev and Sougiannis (1996) and first regress the monthly returns on beta, Ln(ME), and Ln(BM), and then include firm leverage, E(+)/ME, which is the earnings-to-price ratio when earnings are positive and zero otherwise, and E( )/ME, which is a dummy equal to 1 for negative earnings and zero otherwise (Table 3, PanelA). We include Ln(RD/ME) in Panel B to assess whether the inclusion of an R&D variable improves the significance of the basic model introduced by Fama and French (1992). The regressions are run monthly from June 2000 to June 2010, following the methodology of Fama and MacBeth (1973). Previous research for US markets (Lev and Sougiannis 1996, 1999) shows that, if the impact of R&D on stock returns is significant, there could be a potential mispricing effect caused by R&D expenditure. All coefficients are robust to cross-sectional correlation and heteroscedasticity. They indicate that the market model is not significant in predicting future stock returns. In particular, the beta coefficients are not significant and are close to zero for all countries for the period considered. This is in line with Fama and French (1992), who show that the relationship between beta and stock returns disappears in the US market during the period. Firm size has a negative and significant effect on returns at the 1% level for each country, but this effect generally disappears when additional variables are included. The BM ratio is positive and significant for all countries. The inclusion of the R&D variable in the regression improves the predictability of the model only for Germany and the UK (β GER = 0.003, t-test = 3.17, β UK = 0.002, t-test = 2.54). The loadings on the R&D variable for German and UK firms are similar to those evidenced by Lev and Sougiannis (1999) for the US market. Our empirical results show that R&D is not a useful predictor of future stock returns in the other European countries. We confirm previous evidence about the UK, including that regarding the effects of size and the BM ratio. Taken together, when the base model is augmented by three additional variables, the overall predicting power increases across countries. Nevertheless, only the dummy variable for negative earnings impacts negatively on future stock returns. The effect of R&D disappears for German firms. 5.2 R&D and portfolio abnormal returns Our earlier analysis showed a positive relationship between R&D and future stock returns for two of the five countries under study. When other variables are added in the regression, the relationship remains positive only for UK firms. Here, we test whether investors correctly price the firms shares, or whether instead there is evidence of a possible mispricing of R&D investment across Europe. We follow the procedure described in Fama and French (1993) and Chambers, Jennings, and Thompson II (2002) to detect average excess returns, as illustrated in Section 3.

11 10 A. Duqi et al. Table 3. Cross-sectional regressions of monthly future stock returns. FIN FRA GER SWE UK Panel A: Full sample β ( 1.17) ( 0.64) ( 0.14) ( 0.04) ( 0.28) ( 0.41) ( 0.38) ( 0.71) (0.10) (0.19) Ln(ME) ( 2.11) (0.03) ( 2.65) ( 1.85) ( 4.46) ( 2.21) ( 2.86) (1.39) ( 4.36) ( 3.38) Ln(BM) (2.79) (1.72) (2.94) (4.03) (3.12) (2.82) (3.17) (2.25) (5.21) (5.74) E(+)/ME (1.15) ( 0.96) (1.89) (1.42) (0.83) E( )/ME ( 2.90) ( 5.03) ( 5.11) ( 3.30) ( 5.46) D/ME (0.23) (0.33) (1.27) ( 1.24) ( 0.22) Intercept ( 1.88) (0.11) ( 246) ( 134) ( 4.24) ( 1.91) ( 2.76) ( 142) ( 3.72) ( 2.90) N R Panel B: Firms with RD/ME β ( 0.98) ( 0.93) ( 0.45) ( 0.58) ( 0.01) ( 0.03) (0.42) (0.37) (0.03) (0.03) Ln(ME) ( 1.82) ( 0.66) (1.62) (0.35) ( 3.69) ( 2.20) 1.41 ( 0.55) ( 4.06) ( 3.34) Ln(BM) (1.74) (1.12) (0.89) (2.35) (2.01) (2.62) (2.59) (2.04) (3.72) (2.41) Ln(RD/ME) (0.70) (0.55) ( 0.63) ( 1.58) (3.17) (1.64) ( 0.82) ( 0.67) (2.54) (2.24) E(+)/ME ( 0.38) ( 111) (0.17) ( 161) (1.48) E( )/ME ( 0.93) ( 1.33) ( 2.87) ( 120) ( 2.17) D/ME ( 0.85) ( 0.09) (0.60) (0.69) ( 0.96) Intercept ( 1.67) ( 0.18) ( 1.74) ( 0.33) ( 2.80) ( 1.51) ( 1.77) ( 0.15) ( 3.04) ( 2.61) N R Notes: Regressions are estimated using the Fama MacBeth (1973) model. The sample period is , but the estimation of future stock returns is from July 2000 to July β is estimated from the market model using past 60 month returns. Ln(ME) is the natural log of market capitalization; Ln(BM) is the natural log of book-to-market ratio; Ln(RD/ME) is the natural log of R&D to market capitalization ratio; E(+)/ME is earnings to price ratio when earnings are positive and zero otherwise; E( )/ME is a dummy equal to 1 for negative earnings and zero otherwise; and D/ME is debt to equity ratio. T-statistics are in parentheses. significance at the 10% level. significance at the 5% level. significance at the 1% level. Table 4 reports these returns for four equal portfolios, based on RD/ME, where the first portfolio is the least R&D-intensive and the fourth is the most. The results generally reveal that portfolio excess returns increase with respect to RD/ME. The returns are superior to those of matched size and BM portfolios for the first, second, and fifth year

12 The European Journal of Finance 11 Table 4. Average excess returns for R&D-sorted portfolios across countries and portfolios. p-value Non-R&D Q1 (low) Q2 Q3 Q4 (high) Q4 Q1 FIN 1 year excess returns (%) year excess returns (%) year excess returns (%) BM Ln(ME) E/P (%) FRA 1 year excess returns (%) year excess returns (%) year excess returns (%) BM Ln(ME) E/P (%) GER 1 year excess returns (%) year excess returns (%) year excess returns (%) BM Ln(ME) E/P (%) SWE 1 year excess returns (%) year excess returns (%) year excess returns (%) BM Ln(ME) E/P (%) UK 1 year excess returns (%) year excess returns (%) year excess returns (%) BM Ln(ME) E/P (%) Notes: Firms are ranked and sorted in four portfolios in July of each year t based on the RD/ME ratio of year t 1. Buy-and-hold abnormal returns are calculated for the first, second, and fifth year after portfolio formation and averaged. Size and BM control portfolios are created in each July of year t splitting each country sample in five quintiles based on market capitalization, and then five quintiles of BM ratio. BM is the average book-to-market ratio of equity for each portfolio; Ln(ME) is the log of market capitalization; E/P is the earnings-to-price ratio. Excess returns and E/P are in percentages. p-values are for t-tests of differences in means between Q4 and Q1. after portfolio formation, and this is confirmed for every country. The most R&D-intensive groups have the highest excess returns for every country, but are greatest in Finland (18.33%, 6.11%, and 5.32%, respectively, in years 1, 2, and 5) followed by France (18.45%, 9.14%, and 6.06%) and Germany (10.99%, 7.79%, and 4.43%). In the other two countries, we observe overvaluation for the lowest R&D groups, especially in the Swedish case, but the trend reverses in the second and fifth year. The mispricing of the top R&D group among the UK firms is lower than for the other

13 12 A. Duqi et al. countries (0.52% for the first year, 1.05% for the second, and 1.26% for the fifth). Non-R&D firms do not present a clear pattern across different markets. With the exception of Sweden and the UK, they are undervalued compared to top R&D firms. In the last column of Table 4, the differences in means between high and low groups are tested over the period of one, two, and five years. Except for France, the p-values show that generally only excess returns for the first and second year are significant. The t-tests show that mispricing is unlikely to last in the long term, corroborating previous empirical evidence from US data (Ciftci, Lev, and Radhakrishnan 2011). Generally, our results confirm a pronounced undervaluation of R&D stocks, especially for the top quartiles in France, Finland, and Germany, but a potential overvaluation in Sweden in the first year after portfolio formation. We do not observe any significant mispricing in the UK. This may confirm the view that in continental Europe more innovative firms can suffer from mispricing due to their high information asymmetry in comparison to value stocks. Investors in these countries find it difficult to correctly adjust their expectations about future cash flows related to R&D investment. In the UK, mispricing is much lower because of the higher transparency of financial statements and lower information risk. We observe, though, that mispricing is also low for Swedish firms. Sweden s financial and legal environment and corporate governance is similar to those in other continental European countries such as Germany. Nevertheless, Cooke (1989) and Grey and Skogsvik (2004) have demonstrated that listed firms in Sweden opt for voluntary disclosure so as to give positive signals to the markets about their future performance. This is more pronounced in sectors such as pharmaceuticals, where the R&D intensity is higher. This disclosure helps investors to assess conservative measurement biases in the accounting numbers and enables better predictions of future cash flows, reducing stock mispricing. In rows three to five of Table 4, we report average values for some of the key variables across the RD/ME quartiles. Generally, the top R&D firms have lower market capitalization. This is confirmed for all the countries in our sample. The BM ratios are lower for the most innovative stocks. The economic literature has shown that growth stocks with low BM ratios usually represent firms that innovate the most. This is reflected in high stock prices compared to the book value of equity (Lakonishok, Shleifer, and Vishny 1994; Chan, Lakonishok, and Sougiannis 2001). The same rationale seems to guide the earnings-to-price (E/P) ratio in our sample. It is inversely proportional to RD/ME across all countries. Taken together, the results shown in Tables 3 and 4 suggest that R&D expenditures have a significant impact on future stock returns, but not for all countries and only in the short-term. Investors tend to overvalue firms with a lower RD/ME ratio but seem to have lower expectations regarding the future cash flows of highly innovative firms. In this case, undervaluation is common across all countries, although it is weaker for the UK and Sweden. Mispricing is not persistent; top R&D firms do not earn superior excess returns compared to low R&D firms after five years. These outcomes indicate that mispricing is more pronounced among the highly innovative firms in continental Europe. The results are consistent with previous research showing that information asymmetries are more severe in these countries but they are overcome in the long term (Ciftci, Lev, and Radhakrishnan 2011). This issue is relevant especially when investors have to value R&D projects, whose quality is more difficult for outsiders to evaluate. 5.3 R&D expenditure and systematic risk The analysis above shows that stock mispricing is positively influenced by R&D intensity across different countries. It is plausible that this incorrect valuation by investors reflects additional risk

14 The European Journal of Finance 13 inherent in R&D expenditure that is not exhaustively captured by either the BM ratio or the size effect. To address this issue, we use Fama and French s (1993) multifactor model described in Section 3, sorting each country sample by RD/ME. We explained how the portfolios were constructed in Section 3. We then apply model 2, which we rewrite here for convenience: r i,t r f,t = α i + β i (r m,t r f,t ) + γ i SMB t + δ i HML t + ε i,t. (4) For each country, we have 128 monthly observations, spanning the period 1999 to July This model should capture all variations in firm excess returns. If the intercept is significantly different from zero, there is a high possibility that R&D might actually induce an incorrect valuation of a firm s stock returns due to its systematic risk. The results presented in Table 5 indicate that the three-factor model explains a high proportion of the time series variation in returns for all five portfolios. The adjusted R 2 are higher than 70%, with the exception of the top quartile in the Finnish sample. The loading for the market factor is generally high across countries and R&D portfolios, and significant at the 1% level. Abnormal returns are not particularly consistent across the results; that is, we observe negative and significant intercepts in the lowest quartiles for Finland (α Q1 = 0.007, α Q2 = 0.006, α Q3 = 0.006) and Sweden (α Q1 = 0.011, α Q2 = 0.006, α Q3 = 0.010). This trend could be due to the positive correlation between the two Scandinavian countries (the correlation between the stock returns of Sweden and Finland during the period under observation is 0.46). In the other countries, we notice a significant underpricing for France s Q2 portfolio (α Q2 = 0.008) and overpricing for the lowest UK portfolio (α Q1 = 0.008). The non-r&d portfolios suffer from overvaluation, but only significantly so in the case of Germany, Sweden, and the UK. The overall picture seems to indicate that there are no significant risk-based excess returns for the top R&D firms in France, Germany, and the UK, and significant albeit low overpricing in Finland and Sweden. Size is found to have a positive loading when significant for all countries except France. Moreover, there is a positive trend in SMB loadings from the lowest to the highest R&D portfolios, except for France and for Q1 and Q2 in Finland and Germany. High RD/ME firms are smaller, and this is reflected in higher SMB coefficients for the more R&D-intensive firms because they are supposed to be more risky. BM shows no clear trend across countries or across R&D-sorted portfolios. Its coefficient is positive and significant for certain portfolios in Finland (Q2 and Q3), Germany (Q3 and Q4), and Sweden (Q1, Q2, and Q3). It is generally not significant in the French and UK samples. The results for the UK data confirm previous research by Al-Horani, Pope, and Stark (2003) and Dedman et al. (2009). They also find evidence that, when firms are grouped by a measure of R&D intensity, the BM ratio impact is not straightforward. This may be due to the fact that the R&D effect is often subsumed by the BM risk factor (Lev and Sougiannis 1999). 5.4 Overall variability for R&D-intensive firms Prior research has shown that R&D activity is generally riskier because of the high level of information asymmetry and moral hazard (Bhattacharya and Chiesa 1995). The output from R&D investment is highly uncertain and profits are far from assured. These issues should be reflected in higher stock return variability for firms investing in R&D. Chan, Lakonishok, and Sougiannis (2001) and Xu and Zhang (2004) have shown, for the USA and Japan, respectively, that returns variability is positively affected by R&D expenditure along with a set of controlling variables. The results of our own regressions are exhibited in Table 6.

15 Table 5. Risk adjusted returns on R&D-sorted portfolios. FIN FRA GER R&D R&D R&D Non-R&D Q1 (low) Q2 Q3 Q4 (high) Non-R&D Q1 (low) Q2 Q3 Q4 (high) Non-R&D Q1 (low) Q2 Q3 Q4 (high) R m R f (19.02) (35.97) (23.17) (17.28) (12.56) (30.69) (42.47) (31.42) (27.83) (20.82) (51.83) (30.59) (35.37) (38.95) (27.56) SMB (5.58) (21.19) (9.01) (8.90) (7.59) ( 1.57) (0.09) (0.97) (0.16) (105) (37.37) (10.26) (9.09) (15.52) (11.40) HML (4.80) (0.04) (6.73) (3.33) (0.44) ( 0.38) ( 0.08) ( 0.98) (0.17) (0.07) ( 2.81) ( 5.36) ( 0.93) (4.59) (6.51) Intercept ( 124) ( 4.15) ( 231) ( 1.75) ( 1.00) ( 163) ( 155) (2.25) (0.97) ( 0.69) ( 7.64) ( 144) ( 145) ( 0.42) ( 0.38) N Adj. R SWE UK R&D Non-R&D Q1 (low) Q2 Q3 Q4 (high) Non-R&D Q1 (low) Q2 Q3 Q4 (high) R m R f (50.39) (29.62) (40.07) (35.11) (23.77) (32.71) (25.31) (24.44) (23.02) (14.62) SMB (23.92) (3.94) (7.44) (4.35) (15.12) (27.44) (12.04) (15.16) (18.33) (12.82) HML ( 6.84) (14.30) (12.65) (5.21) (0.58) ( 1.60) ( 2.08) ( 0.21) ( 0.54) ( 117) Intercept ( 3.58) ( 2.66) ( 1.87) ( 2.56) ( 138) ( 5.23) ( 3.66) ( 0.79) ( 0.60) ( 0.80) N Adj. R Notes: Portfolios are created in each July of year t and rebalanced every year based on lagged accounting information. Non-R&D firms are grouped separately. R m is the return on the value-weighted market index; R f is the risk-free rate of return; R m R f is the monthly excess return of the specified market index; SMB and HML are the monthly returns on the two portfolios which proxy for size and BM risk. significance at the 10% level. significance at the 5% level. significance at the 1% level. R&D 14 A. Duqi et al.