Examining the impact of research and development expenditures on Tobin's Q.

Bracker, Kevin ; Ramaya, Krishnan

INTRODUCTION AND LITERATURE REVIEW

As the role and influence of technology impacts firms in the economy and financial markets, so too does the importance of research and development (R&D) spending by corporations. The role of research and development (R&D) on firm productivity and growth are well-documented (Griliches, 1984; 1986) and R&D expenditures across different industries have increased significantly over time (Franzen, Rodgers, & Simin, 2007). As firms and industries continue to evolve, R&D has increasingly become a critical element of firm success and survival (Bremser & Barsky, 2004; Tsai & Wang, 2004).

Table one presents the average R&D intensity (R&D as a percentage of sales) by a large sample of publicly-traded firms drawn from the S&P Compustat Database over the 1976 to 2007 time period. Firms are allocating an increasing portion of their budget outlays to R&D spending. The mean (median) R&D intensity for firms in our sample has grown from 1.75% (0.96%) in 1976 to 7.77% (2.71%) in 2007. Given this increased focus on R&D spending by corporations, it is important to look at the impacts of the spending and how it is perceived by investors.

There is a growing body of research that has studied the influence of R&D on firm behavior as well as the market's reaction to the role of R&D. Chan, Lakonishok, and Sougiannis (2001) found that firms with high R&D to equity market value earned high excess returns. Eberhart, Maxwell, and Siddique, (2004) found that while R&D expenditures were beneficial and firms with high R&D expenditures experienced positive long-term returns, markets were slow to recognize the returns. Chan, Martin, and Kensinger (1990) found increased R&D announcements by high-technology firms resulted in positive abnormal returns on average, whereas announcements by low-technology firms were associated with negative abnormal returns.

Studies by Chan, Martin, and Kensinger (1990) and Szewczyk, Tsetsekos, and Zantout (1996) looked at market response to R&D announcements. Both studies found a positive response to increases in R&D spending. Szewczyk, Tsetsekos, and Zantout (1996) found a positive response to increases in R&D spending, primarily for firms with higher values of Tobin's q (the ratio of the market value of the firm relative to the replacement value). Thus, firms that are perceived to be more productive see a greater response than those that are perceived to be less productive. Hsieh, Mishra, and Gobeli (2003) examined the pharmaceutical industry and found that R&D is a significant factor in improving firm performance across a variety of measures.

Connolly and Hirschey (2005) examined the impact of R&D intensity on Tobin's q and found a positive, linear relationship after controlling for growth, risk, profit margin, and advertising intensity. Huang and Liu (2005) examined R&D intensity in Taiwanese firms and found a curvilinear relationship with respect to R&D spending and profitability. Another interesting approach was that of Gleason and Klock (2006) who attempted to look at R&D as a stock variable instead of a flow variable. They found that the value of R&D expenditures accumulated over the previous five years had a significant, positive impact on Tobin's q. Dutta, Om, and Rajiv (2005) took a different perspective and analyzed a firm's R&D capability instead of intensity. They found that firms with a higher level of capability with respect to R&D tended to have higher levels of Tobin's q.

The consensus of the above research is that R&D intensity is associated with higher levels of firm performance and greater valuation in the financial markets. Our paper contributes to this research in three ways. First, we introduce the curvilinear model to US firms using Tobin's q. Tobin's q is a widely used measure of performance (Lee & Tompkins, 1999). A curvilinear relationship found by Huang and Liu (2005) focused on profitability instead of Tobin's q and was based on Taiwanese firms. Second, we consider a variety of classifications to examine how investors respond to R&D spending for firms with different characteristics. Third, in addition to examining how investors value R&D spending ACROSS firms, we investigate how investors respond to changing R&D intensity WITHIN firms. By extending the literature in modeling the response of investors to R&D spending by corporations, we hope to gain a better understanding of the role of R&D.

DATA AND METHODOLOGY

The data is generated using the S & P Compustat database from 1975-2007. While Compustat has a line item entry (Compustat Data Item xrd) for R&D, this item is left blank for many firms (52% of the firms in the original data set). Our first step in collecting the data was to eliminate all firms that did not report R&D expenditures. We identified a total of 51,223 observations (Table 4). Of these, 40,249 (79%) have non-zero values for R&D Intensity and 10,974 (21%) have values of 0. Next, after examining two market indexes (Russell 3000 and the S&P 600) we eliminated all firms with a market capitalization of less than $25 million as these were well below the normal range for even these small capitalization indices. Since our data set includes several observations that are extreme outliers, resulting in significantly skewed variables, we reduced the impact of outliers by requiring firms to exhibit a return on sales between negative 100% and 100%, R&D intensity of less than 100% and annual sales growth of less than 200%. Tobin's q was calculated as demonstrated by Connolly and Hirschey (2005) who based their method upon Chung and Pruitt (1994). Table Two provides a description of the primary variables used in our analysis.

As mentioned above, Tobin's q represents the market value of the firm divided by the replacement value of the firm's assets. As we calculate the replacement value of the firm's assets based on book value, high values of Tobin's q have a couple of related interpretations. Either the balance sheet fails to capture all of the assets employed by the firm or the firm's management is capable of using these assets more productively than their current and potential competitors. One reason why the balance sheet may understate the market value of the firm's assets is if it does not fully capture intangible assets. For instance, R&D may be perceived as an asset in the financial markets in that it can generate future profits; however, it is expensed in the current period. The value of a firm's brand developed through advertising may also fit this description. Based on this, we might expect firms with higher levels of R&D and advertising to have higher levels of Tobin's q.

It is also reasonable to think that there is a point where both R&D and advertising expenses reach diminishing (or even negative) marginal returns. Huang and Liu (2005) found a curvilinear relationship between R&D intensity and profitability in the current year (as measured by return on sales and return on assets) for a sample of 297 Taiwanese firms. This is an interesting finding that represents a starting point for further analysis. We extend their analysis by looking at a larger, US-based sample and focus on valuation instead of profitability. By focusing on valuation (as measured by Tobin's q), we look beyond the impact of R&D on near-term profitability to its perceived net present value in the financial markets. Assuming that there are diminishing marginal returns to R&D expenditures, we should expect to see a curvilinear relationship between R&D intensity and Tobin's q. We can examine this by including a squared R&D intensity term in the regression similar to Huang and Liu (2005). There are two important caveats that should be considered when looking at the coefficient on the squared R&D intensity variable. First, while we anticipate that the coefficient will be negative, indicating diminishing marginal returns, there is reason to believe that the results may be less intense than in Huang and Liu (2005) due to the focus on value. Ceteris paribus, each dollar spent on R&D this period will lower profitability in this period. However, that same dollar spent could still generate significant value to the firm in terms of net present value and therefore increase Tobin's q. Second, a negative coefficient on the squared R&D intensity allows for, but does not imply, managers overspending on R&D. If managers pursue R&D until marginal benefits equal marginal costs, they will be operating in the area of diminishing marginal returns.

In addition to firm specific factors, it is also possible that various economic and market factors might influence Tobin's q. For example, high interest rates will make the future cash flows that are generated from assets worth less to investors and are likely to lower Tobin's q (Faria & Mollick, 2010). Also, anticipation of a strong economy might allow assets to be more productive and increase the value of Tobin's q. There are many additional factors (changing levels of risk aversion, technology, etc.) that may influence Tobin's q over time. Table Three presents average annual values of Tobin's q for our sample.

Based on the analysis above, we have developed two base models (as many firms do not report advertising expense, we examine the impact on Tobin's q both with and without advertising expense). Specifically, the models are as follows:

Tobin's q = [alpha] + [[beta].sub.1](R&D) + [[beta].sub.2][(R&D).sup.2] + [[beta].sub.3](ROS) + [[beta].sub.4](Eq/Asst) + [[beta].sub.5](Growth) + [epsilon] (Equation 1)

Tobin's q = [alpha] + [[beta].sub.1](R&D) + [[beta].sub.2][(R&D).sup.2] + [[beta].sub.3](ROS) + [[beta].sub.4](Eq/Asst) + [[beta]5](Growth) + [[beta].sub.6](Adv) + [[beta].sub.7](Adv).sup.2] + [epsilon] (Equation 2)

When examining the impact of R&D spending, it is important to recognize that the impacts of R&D are likely to be substantially different for different types of firms. Larger firms may be able to more productively employ their R&D expenditures. High growth firms may benefit more from R&D expenditures. Some industries (such as chemical manufacturing) may see a different role for R&D than others (such as financial services). Profitable firms may be able to afford higher levels of R&D expenditures or, alternatively, non-profitable firms may need to spend more on R&D to turn profitable in the future. By looking at the impact of R&D Intensity across a variety of classifications using a variety of dummy variables, we hope to gain a greater understanding of the role of R&D.

When looking at R&D in the manner discussed above, we are looking at how the financial markets perceive differences in R&D intensity across firms. However, it is also important to consider how markets perceive changes in R&D intensity within the firm over time. It is possible to do this using event-study methodology as in Szewczyk, Tsetsekos, and Zantout (1996). However, this only provides information on firms that make an announcement regarding changes in R&D. Another way to examine this issue is to look at changes from year-to-year on a firm-by-firm basis as opposed to examining levels. To accomplish this, we also estimate the following model:

[Delta]Tobin's q = [alpha] + [[beta].sub.1]([Delta]R&D) + [[beta].sub.3](AROS) + [[beta].sub.4]([Delta]Eq/Asst) + [[beta].sub.5]([Delta]Growth) + [epsilon] (Equation 3)

While equation 3 provides a look at how investors view changes in R&D intensity, after controlling for changes in the other variables impacting Tobin's q, there is another issue to consider when looking at changes. One problem investors face when trying to interpret whether a change in R&D intensity is good news or bad news is in evaluating what the "correct" amount of R&D spending should be. An increase in R&D spending for a firm that is not spending enough on R&D should lead to an increase in Tobin's q. On the other hand, an increase in R&D spending for a firm that is already spending too much on R&D is likely to lower the value of Tobin's q. While there is no easy way to know with precision what the correct level of spending should be for each firm, one possible approach would be to evaluate whether the change brings the firm closer to or further from the industry average. We hypothesize that changes towards the industry average will be associated with positive changes to Tobin's q. This leads to equation 4. [Delta]Tobin's q = [alpha] + [[beta].sub.1]([Delta]R&D toward ind. avg.) + [[beta].sub.3]([Delta]ROS) + [[beta].sub.4]([Delta]Eq/Asst) + [[beta].sub.5]([Delta]Growth) + [epsilon] (Equation 4)

RESULTS AND DISCUSSION

Table Four presents the results from our estimation of Equations 1 and 2. Note that all of the variables included in the analysis are highly significant and in the expected direction. It appears that R&D Intensity has a curvilinear relationship with Tobin's q indicating diminishing marginal returns to each dollar invested. These results are largely consistent with previous research as most analysis finds a positive relationship between these R&D Intensity and firm performance. While Huang and Liu (2005) documented a curvilinear relationship between R&D expenditures and performance, this was done for Taiwanese firms and measured performance as profitability (return on sales). Our results extend their findings to the US and to a value-based measure of performance (Tobin's q). The curvilinear relationship is also consistent with expectations regarding R&D expenditures. R&D spending is designed to generate benefits in the future. If a firm spends too little on R&D, then they are passing up positive Net Present Value opportunities. If a firm spends too much on R&D, then they are undertaking negative Net Present Value opportunities. Too much R&D spending can be just as harmful as not enough R&D spending.

One consideration when looking at both Tobin's q and R&D is that both variables have a strong industry component. To control for this, we have looked at the relationship in two additional ways. First, we use dummy variables for industry to control for industry-level differences across Tobin's q. Second, we run separate regressions based on industry. We define industries based on the four-digit SIC codes. In order to do the industry-level analysis, we reduced the sample to only include the top 10 industries in our sample. The list of industries included in the analysis is presented in Table Five.

Table Six presents the results of the regression analysis with the inclusion of industry dummy variables. While this does not result in any significant changes in the primary independent variables relative to the results from Table Four, it greatly improves the model fit by allowing for industry differences across Tobin's q.

Table Seven presents the results of fitting the model to each industry. Advertising intensity was not included in this analysis as it greatly reduced the usable observations per industry. This allows for changes across industry for both Tobin's q and R&D intensity. At the same time however, it also reduces the number of observations, and thus reduces the power of the significance tests. In examining Table Seven, we see that the results are consistent with the same general curvilinear relationship between R&D intensity and Tobin's q that we saw in the broader sample. The primary difference is that the R&D intensity squared variable loses statistical significance in three of the ten industries. However, it is important to note that the results are not contradictory to our previous findings, just not as strong. The only true anomaly in Table Seven is the negative and statistically significant coefficient for Return on Sales within the Retail--Eating Places industry.

Tables Eight through Eleven allow us to examine how the impact of R&D intensity (along with advertising intensity) vary across firms based on characteristics such as industry average R&D intensity, firm size, profitability, and growth rates. In order to analyze this issue, slope dummies for R&D intensity and advertising intensity were introduced.

Table Eight presents the results of the comparison between firms in industries with low R&D intensity versus industries with high R&D intensity. The mean R&D intensity for our sample was 6.19%. Therefore, any industry with an average R&D intensity greater than 6.19% was considered a high R&D intensive industry while industries with an average R&D intensity below 6.19% were considered low R&D intensive industries. The variable Highrd is a slope dummy variable that takes the value of 1*R&D Intensity for firms in high R&D intensive industries and 0 otherwise. The value of this variable is positive and significant at the 1% level indicating that firms in high R&D intensive industries receive greater benefits from their R&D expenditures.

Table Nine segments the firms based on firm size with small cap being classified as those firms with a market capitalization of less than $500 million and large cap referring to firms with a market capitalization of more than $3 billion. Four slope dummy variables are introduced to capture the differential impact of both R&D intensity and advertising intensity for small cap and large cap firms. Small firms have a negative and significant coefficient on the slope dummy for both R&D intensity and advertising intensity. Alternatively, large firms have a positive and significant coefficient for both dummy variables. This provides evidence of economies of scale in R&D spending and advertising spending. The results with respect to R&D intensity are consistent with the findings of Connolly and Hirschey (2005).

Table Ten focuses on the distinction between profitable versus non-profitable firms. Given that changes to either R&D intensity or advertising intensity should have a direct impact on profitability, it is reasonable to expect that investors might value this spending differently based on profitability. Our results indicate that investors are focused more on the long-term impacts as profitable firms get less benefit from R&D expenditures and advertising expenditures. This may indicate that investors see profitable firms as more likely to overspend on R&D and advertising while firms that are not yet profitable are going to be more careful with these expenditures (are operating under tighter budget constraints) and get a bigger return on their investment

Table Eleven examines the role of sales growth in valuation of R&D and advertising expenditures. Spending on R&D (and advertising) is designed to build value by impacting future sales. Firms spend on R&D to help them develop new products which will generate sales in future periods and spend on advertising both to help current sales along with building an image that will enhance sales in years to come. Investors may want to see firms with negative-growth rates increase spending on these areas in order to create growth down the road. Alternatively, investors may feel that these firms have not been productive with R&D/advertising expenditures in the past (resulting in negative growth rates now) and want these expenses kept low. Firms with high-growth rates may need to have significant expenditures in these areas to maintain their current levels of growth and/or receive higher valuations on these expenditures based on their past productivity. To analyze these issues, we again developed four slope dummies to look at the impact of R&D and advertising intensities based on growth rates. The evidence seems to support the idea that higher levels of R&D intensity benefit both negative-growth firms and high-growth firms more so than low growth firms. The slope dummy for advertising with negative-growth firms is not significant. When looking at high growth firms (greater than 15% sales growth in the previous year), both slope dummies are positive and highly significant. This is supportive of the idea that high growth firms are seen as more productive with respect to their expenditures on R&D and advertising. The results are similar for alternative cutoff points for defining high-growth firms.

The evidence from Tables Eight through Eleven indicates that it is important to consider characteristics of the firm when evaluating the impact of R&D and advertising expenditures. Large firms, firms in high R&D intensive industries, and high-growth firms appear to generate a better return on their R&D expenditures. Small firms and firms in low R&D intensive industries are not as efficient with respect to R&D spending. Similar patterns are seen for advertising intensity.

Impact of R&D Intensity Changes

While the above analysis provides lots of insight into how investors value R&D and advertising intensity across different firms, it does not address how investors value changes in R&D intensity within a firm from one year to the next. In order to do this, we have to change our focus from looking at the level of R&D intensity to looking at the change in R&D intensity (along with changes in the other key variables) for each firm from one year to the next. Table Twelve present the results of this analysis for all firms in the sample and for just those firms in the top ten industries.

The results from Table Twelve present an interesting comparison to the results presented earlier in this paper (and much of the other research on the relationship between R&D intensity and Tobin's q). There appears to be conclusive evidence that higher values of R&D intensity are associated with higher values of Tobin's q. Based on this, it would be tempting to conclude that firms are not spending enough on R&D and could increase their market valuations by increasing expenditures in this area. However, when we look at changes in R&D intensity we see a different outlook. There appears to be a negative relationship between changes in R&D intensity and changes in Tobin's q, indicating that increases (decreases) in R&D intensity are associated with a drop (increase) in Tobin's q. This is also the only variable that switches signs from our analysis of levels across companies to looking at changes within companies.

What is the explanation for this apparent discrepancy in market response to R&D expenditures? When looking at the relationship between levels of R&D intensity and Tobin's q, we find a positive, curvilinear relationship. However, this does not necessarily imply that increasing R&D spending is viewed positively by the financial markets. Instead, we are seeing evidence that managers are overspending on R&D. While the average productivity of R&D is positive, many firms are operating in the area of negative marginal productivity.

While capturing specific firm R&D capability is beyond the scope of this paper, one way to segment the marginal impact of changes in R&D intensity is to consider the relationship between the firm's R&D intensity and the industry average. The challenge to managers is to find the optimal level of R&D intensity (the point where average productivity is positive and marginal productivity has just fallen to zero). This level will vary dramatically based on the firm and the industry. One way to examine this issue is to consider the industry average as a benchmark. Firms that have R&D intensity greater than the industry average are more likely to be overspending while firms that are below the industry average are more likely to be under spending. Based on this, we introduce a new variable to capture the change in R&D intensity relative to the industry average. This variable is calculated by taking the absolute value of the difference in R&D intensity for the firm and its industry average last year and subtracting the absolute value of the difference in R&D intensity for the firm and its industry average this year. As firms move closer to (further from) their industry average, this variable will be positive (negative).

Table Thirteen presents the results of this analysis. With this adjustment, we see a positive and significant coefficient indicating that changes towards the industry average are associated with positive changes in Tobin's q. This suggests that, on average, firms that are spending less than the industry average are seeing positive marginal benefits to R&D spending while firms that are spending more than the industry average are seeing negative marginal benefits to their R&D expenditures.

Another interesting way to examine how changes in R&D intensity impact a firm's Tobin's q value is to consider the role of leverage. Ross (1977) introduced the idea of signaling with debt. When firms increase the level of debt financing, this is considered to be a positive signal due to the potential bankruptcy costs associated with debt. Myers and Majluf (1984) suggested that firms with good projects will be more inclined to use debt financing so the benefits of these projects belong to existing shareholders as opposed to being spread among new shareholders. Given these concepts, we might expect firms with strong R&D investment opportunities to increase their leverage. Therefore, we introduce a dummy variable (levinc) that takes the value of 1*R&D Intensity when the equity/asset ratio declines and 0 otherwise. The results of this analysis are included in Table Fourteen and are consistent with the idea that increasing leverage provides a positive signal to investors regarding a firm's R&D expenditures.

CONCLUSION AND FUTURE DIRECTIONS

Over the past 30 years, industry has seen significant growth in R&D intensity. A possible explanation for this growth can be attributed to the ever increasing importance of R&D expenditures as a critical component influencing firm performance. Utilizing a variety of approaches and classifications we examined the relationship between R&D intensity and Tobin's q. Important findings from our analysis suggest that first, there is a strong, curvilinear relationship between R&D intensity and Tobin's q. This is consistent with the concept of diminishing marginal returns to R&D expenditures. Second, industry influences play a strong role in explaining both Tobin's q and the impact of R&D intensity. Third, the impact of R&D intensity on Tobin's q changes significantly based on key characteristics of the firm. Specifically, R&D intensity appears to offer greater benefits to larger firms, firms in industries that are research intensive, and high-growth firms. Fourth, there appears to be an inverse relationship between changes in R&D intensity within a firm and changes in Tobin's q. This suggests that, on average, firms are overspending (at least from the perspective of investors) on R&D and operating within the area of negative marginal productivity. Fifth, when firms change their R&D intensity to move closer to the industry average, this appears to be beneficial as it is associated with increases in Tobin's q. Sixth, there appears to be evidence of debt signaling with respect to the quality of R&D expenditures when looking at changes in R&D intensity.

There are two key related limitations to this study. First the study relied on archival data, and second any use of aggregate R&D expenditures does not reveal the underlying purpose of the actual R&D activity and depending on the firm and industry may involve multiple activities that are not captured by aggregate analysis. A logical extension of this study would include a more in-depth analysis of the effects of intra-industry firm differences with respect to R & D intensity. A systematic disaggregation of actual R&D spending into different categories may facilitate a better understanding of their influence on R&D outcomes and performance.

REFERENCES

Bremser, W. G., & Barsky, N. P. (2004). Utilizing the balanced scorecard for R&D performance measurement. R&D Management, 34(3), 229-238.

Chan, S.H., Martin, J. D., & Kensinger, J. W. (1990). Corporate Research and Development Expenditures and Share Value. Journal of Financial Economics, 26(2), 255-276.

Chan, L. K., Lakonishok, J., & Sougiannis, T. (2001). The Stock Market Valuation of Research and Development Expenditures. Journal of Finance, 56(6), 2431-2456.

Chung, K., & Pruitt, S. (1994). A Simple Approximation of Tobin's Q. Financial Management, 23(3), 70-74.

Connolly, R. A., & Hirschey, M. (2005). Firm Size and the Effect of R&D on Tobin's q. R&D Management, 35(2), 217-223.

Dutta, S., Om N., & Rajiv, S. (2005). Conceptualizing and Measuring Capabilities: Methodology and Empirical Application. Strategic Management Journal, 26(3), 277- 285.

Eberhart, A. C., Maxwell, W. F., & Siddique, A. (2004). An Examination of Long-Term Abnormal Stock Returns and Operating Performance Following R&D Increases. Journal of Finance, 59(2), 623-650.

Faria, J. R., & Mollick, A. V. (2010). Tobin's q and U.S. Inflation. Journal of Economics & Business. 62(5), 401-418.

Franzen, L. A., Rodgers, K. J., & Simin T. T. (2007). Measuring Distress Risk: The Effect of R&D Intensity. Journal of Finance, 62(6), 2931-2967.

Gleason, K. I., & Klock, M. (2006). Intangible Capital in the Pharmaceutical and Chemical Industry. The Quarterly Review of Economics and Finance, 46(2), 300-314.

Griliches, Z. (1984). Interindustry Technology Flows and Productivity Growth: A Reexamination. Review of Economics & Statistics, 66(2), 325-329.

Griliches, Z. (1986). Productivity, R&D, and Basic Research at the Firm Level in the 1970's. American Economic Review, 76(1), 141-154

Hsieh P. H., Mishra, C. S., & Gobeli, D. H. (2003). The Return on R&D Versus Capital Expenditures in Pharmaceutical and Chemical Industries. IEEE Transactions on Engineering Management, 50(2), 141-150.

Huang, C. J., & Chun J. L. (2006). Exploration for the Relationship Between Innovation, IT and Performance. Journal of Intellectual Capital, 6(2), 237-252.

Lee, D., & Tompkins, J. (1999). A Modified Version of the Lewellen and Badrinath Measure of Tobin's Q. Financial Management, 28(1), 20-31.

Myers, S. C., & Majluf, N. S. (1984). Corporate Financing and Investment Decisions When Firms Have Information That Investors Do Not Have. Journal of Financial Economics, 13(2), 187-221.

Ross, S. A. (1977). The Determination of Financial Structure: The Incentive Signalling Approach. Bell Journal of Economics, 8 (1), 23-40.

Tsai, K., and Wang, J. (2004). The R&D Performance in Taiwan's Electronics Industry: a Longitudinal Examination. R&D Management, 34(2), 179-189.

Szewczyk, S. H., Tsetsekos, G. P., & Zantout, Z. (1996). The Valuation of Corporate R&D Expenditures: Evidence from Investment Opportunities and Free Cash Flow. Financial Management, 25(1), 105-110.

Kevin Bracker, Pittsburg State University

Krishnan Ramaya, Pacific University & Washington State University, Vancouver Table 1: R&D Intensity by Year Year Mean R&D Intensity Median R&D Intensity 1976 1.75% 0.96% 1977 1.98% 1.05% 1978 2.03% 1.07% 1979 2.03% 1.08% 1980 2.22% 1.21% 1981 2.67% 1.44% 1982 3.58% 1.83% 1983 4.01% 2.03% 1984 4.21% 2.08% 1985 4.58% 2.21% 1986 4.33% 2.20% 1987 4.52% 2.02% 1988 4.66% 1.90% 1989 4.51% 1.94% 1990 4.80% 2.03% 1991 4.96% 2.07% 1992 5.49% 2.18% 1993 5.68% 2.17% 1994 5.80% 2.15% 1995 6.09% 2.06% 1996 6.88% 2.48% 1997 7.45% 2.57% 1998 7.57% 2.60% 1999 7.22% 2.62% 2000 7.90% 2.76% 2001 8.95% 3.16% 2002 8.57% 3.03% 2003 8.44% 3.01% 2004 7.80% 3.10% 2005 8.02% 2.98% 2006 7.89% 2.87% 2007 7.77% 2.71% Table 2: Definitions of Variables Tobin's q (Market Value of Equity + Book Value Assets--Common Equity)/Book Value Assets R&D Intensity R&D Expenditures/Sales Growth (Current Year's Sales--Previous Year's Sales)/Previous Year's Sales Equity/Assets Common Equity/Book Value Assets Return on Sales Net Income/Sales Advertising Intensity Advertising Expenditures/Sales Change in R&D Intensity |R&D [Intensity.sub.FIRM]--R&D Relative to Industry [Intensity.sub.INDUSTRY]|last year-- Average |R&D [Intensity.sub.FIRM]--R&D [Intensity.sub.INDUSTRY]|this year Table 3: Average Tobin's q by Year Year Mean Tobin's q Median Tobin's q 1976 1.31 1.10 1977 1.33 1.05 1978 1.35 1.04 1979 1.47 1.10 1980 1.69 1.18 1981 1.47 1.14 1982 1.71 1.23 1983 1.75 1.38 1984 1.56 1.27 1985 1.70 1.38 1986 1.68 1.37 1987 1.52 1.23 1988 1.63 1.31 1989 1.74 1.33 1990 1.60 1.19 1991 2.07 1.41 1992 2.13 1.54 1993 2.17 1.60 1994 1.97 1.50 1995 2.34 1.61 1996 2.29 1.67 1997 2.33 1.73 1998 2.28 1.53 1999 3.18 1.61 2000 2.28 1.44 2001 2.32 1.59 2002 1.75 1.35 2003 2.33 1.76 2004 2.33 1.78 2005 2.27 1.74 2006 2.27 1.79 2007 2.21 1.67 Table 4: Impact of R&D on Tobin's q All Firms 51,223 Observations F-Value = 1198.08 Adj. [R.sup.2] = 0.1046 Standardized Variables Coefficient T-Value Coefficient Intercept 0.83158 28.03 *** R&D Intensity 8.78221 38.95 *** 0.35466 R&D Intensity Squared -7.06296 -19.23 *** -0.16310 Return on Sales 1.47984 20.76 *** 0.09920 Equity/Assets 0.94559 16.63 *** 0.07718 Growth 1.43281 39 49 *** 0.16743 Advertising Intensity Advertising Intensity Squared Firms with Advertising Intensity Greater than Zero 19,450 Observations F-Value = 281.86 Adj. [R.sup.2] = 0.0918 Standardized Variables Coefficient T-Value Coefficient Intercept 0.76895 14.01 *** R&D Intensity 6.87388 17.63 *** 0.24537 R&D Intensity Squared -5.73691 -7.77 *** -0.10292 Return on Sales 1.06445 7.91 *** 0.06162 Equity/Assets 0.98081 10.10 *** 0.07594 Growth 1.83216 28.66 *** 0.19887 Advertising Intensity 3.64582 4.69 *** 0.05890 Advertising Intensity -0.63688 -0.23 -0.00292 Squared *** Significant at the p<0.01 Level Table 5: Industry Classification SIC Code Observations R&D Intensity Industry Name 2834 1777 0.152195 Pharmaceutical Preparations 3559 616 0.125422 Special Industry Machinery, NEC 3661 811 0.126673 Telephone & Telegraph Apparatus 3663 1065 0.101957 Radio & TV Broadcasting & Communications Equipment 3674 1700 0.168644 Semiconductors & Related Devices 3714 644 0.028804 Motor Vehicle Parts and Accessories 3845 832 0.118152 Electromedical & Electrotherapeutic Apparatus 5812 795 9.64E-05 Retail--Eating Places 7372 2969 0.184256 Services--Prepackaged Software 7373 898 0.114244 Services--Computer Integrated Systems Design SIC Code 6798 (Real Estate Investment Trusts) has the second-most Observations in our sample (2822), but is omitted because there is only one observation (one firm for one year) with non-zero R&D expense. Table 6: Tobin's q with Industry (SIC) Dummy Variable All Firms 12,107 Observations F-Value = 1102.91 *** Adj. R2 = 0.5772 Standardized Variables Coefficient T-Value Coefficient R&D Intensity 5.91408 12.07 *** 0.28397 R&D Intensity Squared -4.51154 -6.27 *** -0.09954 Return on Sales 1.40629 10.83 *** 0.07710 Equity/Assets 1.09478 8.63 *** 0.18674 Growth 1.86921 27.50 *** 0.19610 Advertising Intensity Advertising Intensity Squared D2834 1.38582 14.25 *** 0.13965 D3559 0.28194 2.17 ** 0.01673 D3661 0.66972 5.43 *** 0.04559 D3663 0.69676 6.29 *** 0.05436 D3674 0.81995 7.39 *** 0.08082 D3845 0.52530 4.68 *** 0.03187 D5812 1.46583 11.79 *** 0.10107 D7370 0.96312 8.50 *** 0.06492 D7372 1.21800 11.87 *** 0.15865 D7373 1.04011 9.06 *** 0.07451 Firms with Advertising Intensity Greater than Zero 4892 Observations F-Value = 508.62 *** Adj. [R.sup.2] = 0.6382 Standardized Variables Coefficient T-Value Coefficient R&D Intensity 3.95144 5.05 *** 0.18554 R&D Intensity Squared -2.50230 -1.88 * -0.04453 Return on Sales 1.43494 7 71 *** 0.08012 Equity/Assets 1.37274 7 80 *** 0.25109 Growth 2.02390 20.42 *** 0.21700 Advertising Intensity 3.00100 2.28 ** 0.04464 Advertising Intensity Squared 0.20612 0.05 0.00083 D2834 1.09518 7.66 *** 0.13326 D3559 0.40721 1.65 * 0.01593 D3661 0.44584 2.52 ** 0.03158 D3663 0.37453 2.11 ** 0.02416 D3674 0.33062 2.05 ** 0.03159 D3845 0.48118 2.21 ** 0.02084 D5812 0.85718 4 79 *** 0.06141 D7370 0.79996 5 90 *** 0.07664 D7372 1.13197 7.60 *** 0.17752 D7373 0.88423 5.20 *** 0.06284 *** Significant at the p<0.01 level ** Significant at the p<0.05 level * Significant at the p<0.10 level Table 7: Impact of R&D Intensity by Industry Industry Classification by SIC Code Variable 2834 3559 3661 3663 Intercept 1.4661 1.3016 0.6662 0.5200 (8.11) *** (7.50) *** (2.49) *** (2.58) *** 5.2428 4.4217 7.8398 4.0425 R&D Intensity (6.08) *** (3 95) *** (4.22) *** (3.20) *** 0.41088 0.32828 0.31738 0.20480 R&D Intensity -3.9500 -2.9033 -10.0561 -1.3159 Squared (-3.63) *** (-1.28) *** (-2.34) *** (-0.61) *** -0.24282 -0.10214 -0.17810 -0.03825 1.1035 0.2387 2.1283 0.5113 Return on Sales (4 19) *** (0.75) *** (4 51) *** (1.40) *** 0.11379 0.03874 0.18281 0.04886 1.3402 0.1407 0.8753 1.6617 Equity/Assets (4.88) *** (0.52) *** (2.24) *** (5 41) *** 0.11258 0.02192 0.07818 0.16209 1.2075 0.4254 1.9786 1.5898 Growth (7.57) *** (3.45) *** (10.62) *** (9.37) *** 0.17514 0.14480 0.34763 0.27686 Observations 1777 616 811 1065 F-Value 37.36 *** 9.14 *** 44.50 *** 36.81 *** Adj [R.sup.2] 0.0929 0.0621 0.2117 0.1440 Industry Classification by SIC Code Variable 3674 3714 3845 Intercept 0.2585 0.3626 0.3126 (0.74) *** (3.62) *** (0.70) *** 8.0062 14.5935 18.8356 R&D Intensity (4 94) *** (7.04) *** (7.10) *** 0.28898 0.57745 0.62152 R&D Intensity -6.1849 -37.7506 -13.2530 Squared (-2.53) *** (-5.58) *** (-3.67) *** -0.14517 -0.46275 -0.28969 2.9057 0.0052 1.7390 Return on Sales (6.28) *** (0.01) *** (3.05) *** 0.18832 0.00049 0.12546 1.5182 1.5575 0.8297 Equity/Assets (3.07) *** (7.68) *** (1.42) *** 0.07725 0.29845 0.04635 1.8370 0.70449 1.9841 Growth (7.52) *** (4.24) *** (6.78) *** 0.18182 0.15283 0.21727 Observations 1700 644 832 F-Value 35.68 *** 33.66 *** 35.13 *** Adj [R.sup.2] 0.0926 0.2025 0.1704 Industry Classification by SIC Code Variable 5812 7372 7373 Intercept 0.5740 1.64475 1.0655 (3.76) *** (6.88) *** (4.58) *** 829.1786 4.6466 4.6848 R&D Intensity (3 19) *** (3 49) *** (2.85) *** 0.32507 0.15521 0.21632 R&D Intensity -166455 -3.1231 -9.8424 Squared (-2.61) *** (-1.36) *** (-2.64) *** -0.26510 -0.06006 -0.20007 -1.8328 1.6419 0.2317 Return on Sales (-3.33) *** (6.09) *** (0.56) *** -0.11300 0.12373 0.02032 2.3546 0.4010 1.3194 Equity/Assets (8.97) *** (1.30) *** (3.48) *** 0.31143 0.02330 0.12063 0.7113 2.6541 2.3886 Growth (4.17) *** (17.41) *** (10.19) *** 0.14129 0.30959 0.32423 Observations 795 2969 898 F-Value 25.45 *** 82.49 *** 30.38 *** Adj [R.sup.2] 0.1334 0.1207 0.1407 t-values are within parenthesis *** Significant at the p<0.01 level ** Significant at the p<0.05 level Standardized coefficients are presented in bold italics below t-values Table 8: High versus Low R&D Intensive Industries All Firms 51,223 Observations F-Value = 999.26 *** Adj. [R.sup.2] = 0.1047 Standardized Variables Coefficient T-Value Coefficient Intercept 0.84093 28.05 *** R&D Intensity 8.00566 18 99 *** 0.32330 R&D Intensity Squared -7.08199 -19.28 *** -0.16353 Return on Sales 1.46874 20.55 *** 0.09845 Equity/Assets 0.94273 16.58 *** 0.07695 Growth 1.43062 39.42 *** 0.16718 Advertising Intensity Advertising Intensity Squared Highrd 0.80365 2.18 ** 0.03283 Firms with Advertising Intensity reater than Zero 19,450 Observations F-Value = 246.92 *** Adj. [R.sup.2] = 0.0919 Standardized Variables Coefficient T-Value Coefficient Intercept 0.75645 13.62 *** R&D Intensity 7.92158 9.83 *** 0.28277 R&D Intensity Squared -5.67358 -7.67 *** -0.10178 Return on Sales 1.07785 7.99 *** 0.06240 Equity/Assets 0.98704 10.16 *** 0.07642 Growth 1.83454 28.69 *** 0.19913 Advertising Intensity 3.62623 4.67 *** 0.05858 Advertising Intensity Squared -0.64117 -0.23 -0.00294 Highrd -1.08430 -1.49 -0.03958 *** Significant at the p<0.01 level ** Significant at the p<0.05 level Highrd is a slope dummy variable representing 1 * R&D intensity for firms with R&D intensity higher than the mean of our sample (0.061911) and 0 otherwise Table 9: Impact of R&D on Tobin's q by Market Capitalization All Firms 51,223 Observations F-Value = 974.15 *** Adj. [R.sup.2] = 0.1174 Standardized Variables Coefficient T-Value Coefficient Intercept 0.75083 25.34 *** R&D Intensity 10.09124 36.25 *** 0.40753 R&D Intensity Squared -5.79200 -15.75 *** -0.13375 Return on Sales 1.00134 13.73 *** 0.06712 Equity/Assets 1.16555 20.43 *** 0.09514 Growth 1.42417 39.50 *** 0.16642 Advertising Intensity Advertising Intensity Squared SmCap -3.75025 -16.82*** -0.13605 LgCap 4.16249 11.65*** 0.05694 AdSmCap AdLgCap Firms with Advertising Intensity Greater than Zero 19,450 Observations F-Value = 214.46 *** Adj. [R.sup.2] = 0.1077 Standardized Variables Coefficient T-Value Coefficient Intercept 0.66003 12.05 *** R&D Intensity 7.93012 16.07 *** 0.28307 R&D Intensity Squared -5.43509 -7.42 *** -0.0975 Return on Sales 0.22365 1.58 0.01295 Equity/Assets 1.34713 13.71 *** 0.1043 Growth 1.86171 29.32 *** 0.20208 Advertising Intensity 2.77838 2.94 *** 0.04489 Advertising Intensity Squared 7.3351 2.62 *** 0.03364 SmCap -3.2895 -7.99 *** -0.10274 LgCap 1.79028 3.00 *** 0.02646 AdSmCap -4.03043 -4.64 *** -0.05616 AdLgCap 7.95159 6.74 *** 0.0605 *** Significant at the p<0.01 level SmCap is a slope dummy variable set to 1 * rdint for small capitalization firms and 0 otherwise LgCap is a slope dummy variable set to 1* rdint for large capitalization firms and 0 otherwise AdSmCap is a slope dummy variable set to 1* adint for small capitalization firms and 0 otherwise AdLgCap is a slope dummy variable set to 1* adint for large capitalization firms and 0 otherwise Small capitalization refers to firms with a market capitalization of $500 million or less Table 10: Profitable versus Non-Profitable Firms All Firms 51,223 Observations F-Value = 1001.80 *** Adj. [R.sup.2] = 0.1049 Standardized Variables Coefficient T-Value Coefficients Intercept 0.82644 27.84 *** R&D Intensity 9.51788 33.59 *** 0.38437 R&D Intensity Squared -7.72324 -19 40 *** -0.17834 Return on Sales 1.66266 20.02 *** 0.11145 Equity/Assets 0.94294 16.59 *** 0.07697 Growth 1.43808 39.62 *** 0.16805 Advertising Intensity Advertising Intensity Squared Profit -1.04558 -4 29 *** -0.02538 Adprofit Firms with Advertising Intensity Greater than Zero 19,450 Observations F-Value = 220.77 *** Adj. [R.sup.2] = 0.0923 Standardized Variables Coefficient T-Value Coefficients Intercept 0.75808 13 79 *** R&D Intensity 7.51740 14 42 *** 0.26834 R&D Intensity Squared -6.27533 -7.80 *** -0.11258 Return on Sales 1.40978 8.23 *** 0.08161 Equity/Assets 0.97891 10.08 *** 0.07579 Growth 1.83116 28.64 *** 0.19876 Advertising Intensity 6.00343 5.05 *** 0.09699 Advertising Intensity Squared -3.43217 -1.17 -0.01574 Profit -0.97285 -2.10 ** -0.02234 Adprofit -2.40770 -2.63 *** -0.03439 *** Significant at the p<0.01 level ** Significant at the p<0.05 level Profit is a slope dummy variable set to 1* rdint for firms with positive net income Adprofit is a slope dummy variable set to 1* adint for firms with positive net income Table 11: Impact of R&D on Tobin's q by Growth Rate All Firms 51,223 Observations F-Value = 866.53 *** Adj. [R.sup.2] = 0.1058 Standardized Variables Coefficient T-Value Coefficient Intercept 0.85612 28.70 *** R&D Intensity 7.77488 28.10 *** 0.31398 R&D Intensity Squared -7.53896 -20.29 *** -0.17409 Return on Sales 1.46171 20.32 *** 0.09798 Equity/Assets 0.93970 16.54 *** 0.07670 Growth 1.30614 32.02 *** 0.15263 Advertising Intensity Advertising Intensity Squared Neggrowth 0.73626 2.79 *** 0.01788 Highgrowth 1.96249 7.92 *** 0.06265 Adneggrowth Adhighgrowth Firms with Advertising Intensity Greater than Zero 19,450 Observations F-Value = 183.59 *** Adj. [R.sup.2] = 0.0936 Standardized Variables Coefficient T-Value Coefficient Intercept 0.80457 14.54 *** R&D Intensity 5.78236 12.15 *** 0.20641 R&D Intensity Squared -6.33086 -8.34 *** -0.11357 Return on Sales 1.09250 7.97 *** 0.06325 Equity/Assets 0.96178 9.91 *** 0.07447 Growth 1.64062 21.34 *** 0.17808 Advertising Intensity 2.46913 2.93 *** 0.03989 Advertising Intensity Squared -1.96031 -0.70 -0.00899 Neggrowth 1.16095 2.37 ** 0.02518 Highgrowth 1.92246 4.21 *** 0.05464 Adneggrowth 1.30643 1.38 0.01229 Adhighgrowth 3.40502 3.95 *** 0.03734 *** Significant at the p<0.01 level ** Significant at the p<0.05 level Neggrowth is a slope dummy variable set to 1 * R&D intensity for firms with negative growth in revenues Highgrowth is a slope dummy variable set to 1 * R&D intensity for firms with greater than 15% growth in revenues Adneggrowth is a slope dummy variable set to 1 * R&D intensity for firms with negative growth in revenues Adhighgrowth is a slope dummy variable set to 1 * R&D intensity for firms with greater than 15% growth in revenues Table 12: Impact of Changes in R&D on Changes in Tobin's q All Firms 43,703 Observations F-Value = 92.44 *** Adj. [R.sup.2] = 0.0083 Standardized Variables Coefficient T-Value Coefficient Intercept -0.05469 -5.51 *** [DELTA] R&D Intensity -1.00439 -6.37 *** -0.03348 [DELTA] Return on Sales 0.19410 4.37 *** 0.02315 [DELTA] Equity/Assets 0.57350 5.17 *** 0.02484 [DELTA] Growth 0.38086 13.64 *** 0.06603 Firms in Ten Largest Industries 12,068 Observations F-Value = 64.45 *** Adj. [R.sup.2] = 0.0206 Standardized Variables Coefficient T-Value Coefficient Intercept -0.11537 -5.83 *** [DELTA] R&D Intensity -1.87449 -7 01 *** -0.06934 [DELTA] Return on Sales 0.17303 2.71 *** 0.02673 [DELTA] Equity/Assets 0.42982 2.20 ** 0.01992 [DELTA] Growth 0.52726 10.69 *** 0.09883 *** Significant at the p<0.01 level Significant at the p<0.05 level Refer to Table Five for a list of the ten largest industries Tables 13: Impact of Changes in R&D on Tobin's q Revisited All Firms 43,703 Observations F-Value = 88.67 *** Adj. [R.sup.2] = 0.0080 Standardized Variables Coefficient T-Value Coefficient Intercept -0.05475 -5 51 *** [DELTA] R&D Intensity Relative 0.82750 5.06 *** 0.02614 to Industry Average [DELTA] Return on Sales 0.22396 5 10 *** 0.02672 [DELTA] Equity/Assets 0.57738 5.20 *** 0.02500 [DELTA] Growth Rates 0.39151 14.06 *** 0.06787 Firms in Ten Largest Industries 17,736 Observations F-Value = 59.26 *** Adj. [R.sup.2] = 0.0189 Standardized Variables Coefficient T-Value Coefficient Intercept -0.11621 -5.86 *** [DELTA] R&D Intensity Relative 1.48890 5.36 *** 0.05175 to Industry Average [DELTA] Return on Sales 0.22280 3.52 *** 0.03441 [DELTA] Equity/Assets 0.45404 2.32 ** 0.02104 [DELTA] Growth Rates 0.55981 11.44 *** 0.10493 *** Significant at the p<0.01 level * Significant at the p<0.05 level Refer to Table Five for a list of the ten largest industries Tables 14: Leverage and the Impact of Changes in R&D on Tobin's q All Firms 43,703 Observations F-Value = 74.94 *** Adj. [R.sup.2] = 0.0084 Standardized Variables Coefficient T-Value Coefficient Intercept -0.05619 -5.65 *** [DELTA] R&D Intensity 1.50274 6 74 *** 0.04748 Relative to Industry Average [DELTA] Return on Sales 0.22921 5.22 *** 0.02734 [DELTA] Equity/Assets 0.56033 5.05 *** 0.02427 [DELTA] Growth Rates 0.39432 14.17 *** 0.06836 Levinc 1.20322 4.45 *** 0.03029 Firms in Ten Largest Industries 12,068 Observations F-Value = 49.72 *** Adj. [R.sup.2] = 0.0198 Standardized Variables Coefficient T-Value Coefficient Intercept -0.11951 -6.02 *** [DELTA] R&D Intensity 2.26464 6.28 *** 0.04748 Relative to Industry Average [DELTA] Return on Sales 0.23065 3.65 *** 0.02743 [DELTA] Equity/Assets 0.44457 2.27 ** 0.02427 [DELTA] Growth Rates 0.57168 11.66 *** 0.06836 Levinc 1.51030 3.37 *** 0.03029 *** Significant at the p<0.01 level ** Significant at the p<0.05 level Levinc is a slope dummy variable that takes the value of 1 * R&D intensity when the equity/asset ratio declines and 0 otherwise Refer to Table Five for a list of the ten largest industries