Absorptive capacity, technology spillovers, and the cross-section of stock returns

Absorptive capacity, technology spillovers, and the cross-section of stock returns

Accepted Manuscript Absorptive Capacity, Technology Spillovers, and the Cross-Section of Stock Returns Jong-Min Oh PII: DOI: Reference: S0378-4266(1...

1MB Sizes 3 Downloads 19 Views

Accepted Manuscript

Absorptive Capacity, Technology Spillovers, and the Cross-Section of Stock Returns Jong-Min Oh PII: DOI: Reference:

S0378-4266(17)30202-9 10.1016/j.jbankfin.2017.08.016 JBF 5199

To appear in:

Journal of Banking and Finance

Received date: Revised date: Accepted date:

4 May 2016 30 June 2017 20 August 2017

Please cite this article as: Jong-Min Oh , Absorptive Capacity, Technology Spillovers, and the CrossSection of Stock Returns, Journal of Banking and Finance (2017), doi: 10.1016/j.jbankfin.2017.08.016

This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.

ACCEPTED MANUSCRIPT

Absorptive Capacity, Technology Spillovers, and the Cross-Section of Stock Returns*

This Version: July 2017 ABSTRACT

CR IP T

Jong-Min Oh §

In the presence of potential technology spillovers, I demonstrate that a firm’s absorptive capacity (AC),

AN US

as proxied by R&D investments, is crucial to benefit from spillovers. I find that higher AC firms, when exposed to large potential spillovers, exhibit stronger future real outcomes (cite-weighted patents and operating performance) and market value. Importantly, however, this value-relevant information does not appear to be immediately incorporated into stock prices, leading to high future abnormal stock returns for firms with high AC and spillover exposure. Furthermore, the undervaluation is most

M

pronounced among low investor attention stocks, suggesting that limited attention contributes to the

ED

undervaluation.

PT

JEL Classification: G11, G12, G14, O31, O32, O33 Keywords: Spillover, Innovation, R&D, Limited attention, Market efficiency

*

AC

CE

I am grateful to Geert Bekaert (the editor), two anonymous referees, Samar Ashour, Seong Byun, Wan-Jiun (Paul) Chiou, Bernhard Ganglmair, Harrison Hong, Po-Hsuan Hsu, Robert Kieschnick, Cheolwoo Lee, Jun Li, Arzu Ozoguz, Michael Rebello, Alessio Saretto, Malcolm Wardlaw, Han Xia, Yexiao Xu, and Harold Zhang and conference and seminar participants at the FMA Annual Meeting 2014, FMA Doctoral Student Consortium Special Job Market Paper Presentation 2014, the Southern Finance Association (SFA) Annual Meeting 2014, FMA Asia Annual Meeting 2015, Oregon State University, University of Central Florida, and the University of Texas at Dallas for constructive comments and suggestions. I give special thanks to Nicholas Bloom for his generosity in sharing his resources.

§

Assistant Professor of Finance, College of Business Administration, University of Central Florida, P.O. Box 161400, Orlando, FL 32816. Email: [email protected] Phone: (407) 823-2360.

ACCEPTED MANUSCRIPT

“We have always been shameless about stealing great ideas.” - Steve Jobs

1. Introduction Innovative activities drive a firm’s technological advancements. They also generate technology spillovers since they are often sources of new ideas and opportunities that add to the existing knowledge pool (e.g., Schumpeter, 1934; Arrow, 1969; Romer, 1986). Thus, benefiting from other firms’ innovative

CR IP T

activities, in addition to investing in one’s own innovations, can be an important factor in firm performance. For example, Xerox developed the first graphical user interface (GUI) for computers in the early 1970s, and Microsoft and Apple used this technology to develop profitable new products (Windows and Macintosh). The literature on innovation also documents that technology spillovers

AN US

(“spillovers” hereafter), knowledge that is transferred from technologically related firms, positively affect a related firm’s operating performance and market value (e.g., Jaffe, 1986; Bloom, Schankerman, and Van Reenen, 2013), as well as subsequent stock returns (e.g., Hsu, 2011; Chen, Chen, Liang, and Wang, 2013; Jiang, Qian, and Yao, 2016). These studies rely on the argument that spillovers create

ED

benefits these firms will enjoy.

M

positive externalities such that the more firms are exposed to potential spillovers, the more future

However, despite the positive externality aspects of spillovers, building on the knowledge of others

PT

does not seem to be done without a cost. Ample anecdotal evidence suggests that exposure to large potential spillovers may not necessarily benefit all firms equally. Related to the previous example, Apple

CE

and Microsoft have enjoyed huge benefits when exposed to spillover (GUI technology), while Commodore International has been less successful in the computer market even though it also was

AC

exposed to the same spillover. Likewise, even though many film-industry firms were exposed to the charged-coupled device (CCD) technology invented by AT&T Bell that enables digital imaging, some firms have been extremely successful in producing profitable digital cameras (i.e., Sony, Cannon, and Nikon), whereas some firms have not (i.e., Eastman Kodak). Therefore, it is natural to ask what drives cross-sectional differences in future firm performance even when all firms are exposed to the same level of potential spillover and whether the stock market is able to distinguish ex ante between firms that are 1

ACCEPTED MANUSCRIPT

most likely to benefit from the potential spillovers. Despite their importance, these questions have not been answered. Answering them is very important since identifying the underlying mechanisms will help investors properly allocate their resources toward firms more likely to generate higher returns on their innovations.

CR IP T

To answer these questions, I examine whether or not cross-sectional differences in a firm’s ability to absorb potential spillovers matter and affect the relation between spillovers and subsequent stock returns. I posit that, when exposed to high potential spillovers, firms with higher R&D investments will likely better absorb and convert the spillovers into value-relevant business improvements. My argument builds

AN US

on the intuitive idea that a firm should be up-to-speed on current technologies to benefit from potential spillovers. For example, a firm that already has undertaken significant research in related areas will better understand, and hence effectively absorb, outside knowledge. Existing studies also argue that R&D investments can enhance a firm’s ability to recognize, assimilate, and exploit new external

M

information, namely, “absorptive capacity” (AC) (e.g., Evenson and Kislev, 1976; Cohen and Levinthal, 1989, 1990; Henderson and Cockburn, 1996).1 In the presence of large spillovers, firms with higher

ED

R&D investments will therefore likely have a higher AC, and this will allow them to generate more successful innovative outcomes (i.e., granted patents and patent citations) in the future and exhibit higher

PT

future operating performance. Therefore, a firm’s R&D investments, coupled with the size of spillover

CE

exposure, contain value-relevant information about future real outcomes. Exploring the complementary relation between spillovers and absorptive capacity (“spillover-AC

AC

synergy” hereafter) leads to my main hypothesis. I hypothesize that spillover-AC synergy, rather than just the size of potential spillovers, positively predicts subsequent stock returns. Although firms with greater spillover-AC synergy exhibit higher future innovation-related performance, as well as operating 1

Alternatively, several studies argue that the level of spillover exposure itself may positively affect the performance of peer firms by allowing these firms to incur low R&D spending for innovations since spillovers create positive externalities (e.g., Nelson, 1959; Arrow, 1962; Griliches, 1979; Mansfield, 1977, 1988; Spence, 1984; Bernstein and Nadiri, 1989). Thus, the relation between absorptive capacity and spillovers remains an open empirical question.

2

ACCEPTED MANUSCRIPT

performance, investors may have difficulty in processing such favorable information. That is, investors may have difficulty recognizing the size of potential spillovers and, more importantly, to what extent a given firm can absorb the potential spillovers. From this perspective, limited investor attention likely leads to underreaction to this favorable information. 2 Thus, the positive effects of spillover-AC synergy 1F1F

CR IP T

may not be fully valued by the market ex ante and may lead to positive subsequent abnormal stock returns.

To test these ideas, I follow previous studies and use R&D intensity as a proxy for a firm’s absorptive capacity (e.g., Cohen and Levinthal, 1989, 1990; Henderson and Cockburn, 1996). Specifically, I

AN US

calculate a firm’s absorptive capacity (AC) as R&D capital divided by firm sales in which R&D capital is the five-year cumulative research and development (R&D) expenditures assuming an annual depreciation rate of 20% (Chan, Lakonishok, and Sougiannis, 2001). Next, I use a measure for potential technology spillover pools recently developed by Bloom, Schankerman, and Van Reenen (2013). The

M

measure captures the extent of a firm’s exposure to other firms’ R&D efforts within a similar technological field. Specifically, the technology spillover pool for each firm constitutes the sum of other

ED

firms’ R&D weighted by technology similarities. The technology similarity between each pair of firms is estimated by calculating the distance between each firm’s technology positions, identified as the

PT

composition of each firm’s patent portfolio.

CE

Using these two measures, I first examine whether or not greater spillover-AC synergy is related to significantly stronger effects on a firm’s future real outcomes, such as the quality of innovation

AC

outcomes and firm profitability, and future market valuation. I show that when firms are exposed to large spillovers, firms with higher AC generate more successful innovative outcomes (i.e., the number of citeweighted patents), allowing them to have higher innovation-related performance. Moreover, firms with greater spillover-AC synergy not only have higher innovation-related performance but also exhibit 2

See, for example, Hirshleifer, Hsu, and Li (2013 and 2014), among others. The authors demonstrate that limited investor attention contributes to underreaction to the hard-to-process favorable information related to firms’ various innovative activities.

3

ACCEPTED MANUSCRIPT

superior future operating performance and higher market value. For example, firms with higher AC and higher spillover exposure exhibit significantly higher return on assets (ROA), profit margins (PM), and market-to-book ratio (MTB) over the next year. These results are consistent with the hypothesis that firms with higher absorptive capacity (AC) will likely better absorb and convert the potential spillovers

CR IP T

into value-relevant business improvements and hence will benefit the most from the potential spillovers. Therefore, the results suggest that high AC and exposure to potential spillovers together contain valuerelevant information regarding strong future firm performance.

Next, I examine whether spillover-AC synergy, rather than just the size of spillover exposure,

AN US

predicts the future cross-section of stock returns. Specifically, I examine whether the positive relation between the size of spillover exposure and the subsequent stock returns found in previous studies is concentrated in firms with high absorptive capacity. I find that spillover-AC synergy, rather than the size of spillover exposure per se, positively predicts the cross-section of future stock returns. Specifically,

M

running Fama-MacBeth (1973) cross-sectional regressions of stock returns on lagged AC, spillovers, and additional sets of control variables, I first confirm the existing studies on spillovers that the size of

ED

spillover exposure is positively related to future stock returns (e.g., Hsu, 2011; Chen, Chen, Liang, and Wang, 2013; Jiang, Qian, and Yao, 2016). Importantly, however, when the interaction between AC and

PT

spillovers is considered, I find that the interaction between AC and spillovers, rather than spillover alone,

CE

exhibits significantly positive relation to the future stock returns. For example, a one-standard-deviation increase in exposure to potential spillovers exhibits an approximate 7.5% increase in annualized future

AC

excess stock returns, on average, for the high AC firms compared to only about 0.4% increase in annualized excess returns for the low AC firms. I show that these results are robust to potential riskbased explanations, such as increased market rivalry effect (e.g., Bloom, Schankerman, and Van Reenen, 2013), financial constraints (e.g., Li, 2011), and the Internet (dot-com) bubble period. Furthermore, my findings are also robust to the inclusion of the other innovation-related factors, such as patent intensity (e.g., Deng, Lev, and Narin, 1999), innovative efficiency (e.g., Hirshleifer, Hsu, and Li, 2013), and

4

ACCEPTED MANUSCRIPT

interaction effects between R&D and R&D ability (Cohen, Diether, and Malloy, 2013) or product market competition (Gu, 2016). These findings are consistent with my hypothesis that high spillover-AC synergy, rather than just high exposure to spillovers, is likely undervalued by the market ex ante and that I have uncovered a new innovation-related feature that can positively affect the future stock returns.

CR IP T

I confirm the future stock return results using portfolio analysis. Specifically, forming portfolios by double-sorting firms into absorptive capacity and spillover exposure terciles, I show that a portfolio of firms with both large spillover exposures and high AC (“AbsorbSpill” portfolio) outperforms a portfolio of firms with the same size spillover exposures but low AC (“NoAbsorbSpill” portfolio). For example,

AN US

the value-weighted AbsorbSpill portfolio earns a 0.54% (t = 4.45) monthly four-factor alpha, while the NoAbsorbSpill portfolio earns only -0.03% (t = 0.37). The spread (zero-cost) portfolio that is long in AbsorbSpill firms and short in NoAbsorbSpill firms earns a monthly alpha of 0.57% (t = 4.22), or a yearly alpha of approximately 7%. The results are also robust to different asset pricing models, such as

M

the Fama and French (2015) five-factor model and the Hou, Xue, and Zhang (2015) q-factor model, and to mispricing factors of Hirshleifer and Jiang (2010) and Stambaugh and Yuan (2017). Overall, the

ED

results suggest that the market’s undervaluation, if any, is concentrated on firms with both high spillovers and high AC, rather than on firms with just high spillover exposure.

PT

Finally, I investigate the source of the market’s inability to fully recognize the positive effects of

CE

spillover-AC synergy ex ante. Specifically, I test whether or not limited investor attention contributes to the undervaluation of firms with high AC and large spillover exposures. If limited attention contributes

AC

to the undervaluation, I would expect that the positive relation between spillover-AC synergy and future stock returns will be most pronounced among firms with low investor attention. Running Fama-MacBeth regressions within subsamples of high and low investor attention groups, I find that firms within the low (high) investor attention group have large (small) and significant (insignificant) slope coefficients on the

5

ACCEPTED MANUSCRIPT

interaction between AC and spillovers.3 Furthermore, the difference in the magnitudes between the low and high attention groups is large and statistically significant. These findings support the notion that the market does not fully recognize and process information about the positive real effects of spillover-AC synergy, leading to positive future abnormal stock returns.

CR IP T

This paper contributes to a large and growing body of work showing that the market does not fully incorporate information regarding a firm’s innovative activities (i.e., R&D investments), as well as the innovation efforts of peer firms (i.e., spillovers) into stock prices (e.g., Lev and Sougiannis, 1996; Deng, Lev, and Narin, 1999; Chan, Lakonishok, and Sougiannis, 2001; Eberhart, Maxwell, and Siddique, 2004;

AN US

Hsu, 2011; Hirshleifer, Hsu, and Li, 2013, 2014; Cohen, Diether, and Malloy, 2013; Chen, Chen, Liang, and Wang, 2013; Jiang, Qian, and Yao, 2016). Recent studies by Chen et al. (2013) and Jiang et al. (2016) show an average positive effect of peer firm’s R&D activities on a given firm’s subsequent stock returns. Unlike the findings of Chen et al. (2013) and Jiang et al. (2016), my findings provide evidence

M

that exposure to large potential spillovers is a necessary, but not sufficient, condition for firms to have economically meaningful subsequent positive abnormal returns. Likewise, I provide evidence that

ED

information about a firm’s high level of past R&D investments alone is not sufficient to generate significant positive alpha, consistent with previous studies (e.g., Chan, Lakonishok, and Sougiannis,

PT

2001; Li, 2011; Cohen, Diether, and Malloy, 2013). However, I provide evidence that in the presence of

CE

large technology spillovers, a firm’s high R&D investments contain information not only about high innovative input levels but also about high absorptive capacity such that firms with high R&D (high AC),

AC

in this case, are likely undervalued in the market. Therefore, what seems to be misvalued by investors is not just the information about the size of technology spillovers or the level of R&D investments, but instead the information about the strong complementary relation between the two (spillover-AC synergy).

3

I use analyst coverage, institutional ownership, and firm size as proxies for investor attention. Firms with low analyst coverage, small size (e.g., Hong, Lim, and Stein, 2000; Hirshleifer and Teoh, 2003), or low institutional ownership (e.g., Hirshleifer, Hsu, and Li, 2014) are classified as the low attention group.

6

ACCEPTED MANUSCRIPT

My findings also add to the literature about how firms’ innovative activities impact another firm’s managerial decisions. For example, Qiu and Wan (2015) provide evidence that potential spillovers have a positive relation with a firm’s cash holding since the firm wants to meet the potential needs of future innovations influenced by the technological opportunities. Bena and Li (2014) and Sevilir and Tian

CR IP T

(2012) document a firm’s incentive to acquire an innovation-intense targets to create synergy on innovations. By providing evidence that firms need to maintain high absorptive capacity (high own R&D) to better benefit from external technological opportunities, the findings in my paper suggest that a firm’s incentive to hold more cash or incentive to acquire the innovation-intense targets may be a rational

AN US

managerial response to benefit from its own, as well as peers’, innovative activities.

Finally, this paper is related to the limited attention literature that shows that the market fails to fully incorporate information into stock prices when that information lacks saliency (e.g., Klibanoff, Lamont, and Wizman, 1998; Huberman and Regev, 2001; Hirshleifer and Teoh, 2003; Lev, 2004; Peng and

M

Xiong, 2006; DellaVigna and Pollet, 2009; Hirshleifer, Lim, and Teoh, 2009, 2011; Hou, Peng, and Xiong, 2009). I provide evidence in line with these studies by finding that the complementary relation

ED

between a firm’s AC and spillovers is most pronounced among low investor attention stocks. The rest of the paper is organized as follows. Section 2 discusses the data and key variables. Section 3

PT

examines the value-relevance of spillover-AC synergy. Section 4 presents the results for the positive

CE

relation between spillover-AC synergy and future cross-section of stock returns. Section 5 provides

AC

evidence for limited investor attention. Section 6 provides the robustness results. Section 7 concludes.

2. Data, Measures, and Descriptive Statistics 2.1. Data

I build spillovers and innovation-related measures primarily by using the National Bureau of Economic Research (NBER) U.S. Patent Data. The NBER U.S. Patent Data contain detailed information about all utility patents granted by the U.S. Patents and Trademark Office (USPTO) from 1976 to 2006.

7

ACCEPTED MANUSCRIPT

The use of patent data restricts my sample to either firms that have received patent grants during my sample period or firms that have been confirmed to have zero patent grants. The NBER patent data provide additional matching data linking the patent assignee classification to COMPUSTAT firm-level identifiers, which I use to merge patent data with the COMPUSTAT and CRSP data (e.g., Hall, Jaffe,

CR IP T

and Trajtenberg, 2001).

I use the data for common stocks trading on the NYSE, AMEX, and NASDAQ exchanges from the CRSP monthly stock file. Firm characteristics are from COMPUSTAT. I add delisting returns to each firm’s monthly returns if it is delisted during my sample period. I filter firm-year observations from the

AN US

COMPUSTAT database following the standard procedures implemented in previous studies by dropping financial firms (SIC codes 6000–6999). I mitigate backfilling bias following the previous literature: Firms must be listed on COMPUSTAT for at least two years before they are included in my sample. I require firms to have nonnegative and nonmissing total book value of assets or equity and also to have

M

non-missing R&D expenditures. Following Fama and French (2006) and Ciftci and Cready (2011), I exclude firms with either extremely small size or sales volume to mitigate the influence of these small

ED

firms. 4 I remove firm-year observations with abnormally large jumps in either sales or number of 4F4F

employees since these are likely to reflect significant restructuring activities, such as mergers and

PT

acquisitions, even if the company keeps the same firm identifier (e.g., Bloom, Schankerman, and Van

CE

Reenen, 2013). Specifically, I follow Bloom, Schankerman, and Van Reenen (2013) and drop a firmlevel observation if either the sales growth or employee growth exceeds 200% or falls by more than 66%

AC

(approximately the top and bottom 0.5 percentiles). I finally merge the resultant data with the NBER patent data. For constructing variables, I use a sample period that ranges from 1976 to 2006. However, for all of the analyses, I use a sample period from July 1982 to June 2007, since some measures are constructed using multiple years of past data.

4 I exclude firms with total book value of assets or sales less than US $5 million or book value of equity below US $2.5 million. However, I find similar results even if I include these firms in my sample.

8

ACCEPTED MANUSCRIPT

2.2. Absorptive Capacity (AC) Measure Since my hypothesis focuses on a firm’s ability to absorb the potential spillovers (absorptive capacity), I need a proxy for a firm’s absorptive capacity (AC) that enables a given firm to benefit from potential spillovers effectively. Previous studies have documented that firms that conduct their own

CR IP T

R&D are better able to recognize, assimilate, and exploit externally available information (e.g., Tilton, 1971; Allen, 1977; Mowery, 1983; Evenson and Kislev, 1976; Cohen and Levinthal, 1989 and 1990; Henderson and Cockburn, 1996). Thus, AC is more likely generated as a by-product of a firm’s R&D investment (Cohen and Levinthal, 1990). Therefore, a firm that has conducted more R&D is more likely

AN US

to stay up-to-speed on current technologies and, by doing so, enable the firm to benefit more from spillover exposures, compared with a firm that is not up to speed (e.g., Cohen and Levinthal, 1989, 1990; Henderson and Cockburn, 1996).

Following Cohen and Levinthal (1989, 1990) and Henderson and Cockburn (1996), I use a firm’s

M

R&D intensity as a proxy for the absorptive capacity (AC). Since a firm’s current AC is a manifestation of the firm’s knowledge accumulation through past and current R&D efforts (e.g., Cohen and Levinthal,

ED

1989), absorptive capacity in year t ACt is defined as R&D capital (RDC) divided by firm sales (RDC-tosales ratio) in year t, where R&D capital (RDC), following Chan, Lakonishok, and Sougiannis (2001), is

PT

calculated as five-year cumulative R&D expenditures, assuming an annual depreciation rate of 20%.5

CE

R&D capital is widely used in finance literature (e.g., Chan, Lakonishok, and Sougiannis, 2001; Li, 2011; Hirshleifer, Hsu, and Li, 2013; Gu, 2016). While firm sales is one of the most widely used variables for

AC

scaling in R&D intensity measures (e.g., Cohen and Levinthal, 1989 and 1990; Lev and Sougiannis, 1996; Chan, Lakonishok, and Sougiannis, 2001; Li, 2011; Cohen, Diether, and Malloy, 2013; Almeida, Hsu, and Li, 2013), I also use market value of equity (ME) to scale for robustness (RDC-to-ME ratio). Even though previous studies have documented that a firm’s R&D investments can serve as a good proxy for a firm’s ability to absorb external knowledge generated by peer firms, R&D investments also

5

Firm i’s R&D capital in year t is

.

9

ACCEPTED MANUSCRIPT

can be correlated with various innovation-related abilities that already have been used in previous studies.6 In Section 6.1, I discuss this issue in more detail and show that the AC measure (RDC-to-sales ratio) is not just a mere reflection of the existing innovation-related abilities, but, rather, captures the

CR IP T

incremental effect to the existing innovation-related ability measures.

2.3. Technology Spillover Exposure Measure

To capture a firm’s exposure to technology spillovers, I estimate a technology spillover exposure measure based on the similarity of technologies among firms (e.g., Jaffe, 1986; Bloom, Schankerman,

AN US

and Van Reenen, 2013).7 Following Bloom et al. (2013), I first construct a vector of a firm’s patent shares within each technology class assigned by the USPTO in order to identify the firm’s position within the technology space. I then measure the technology similarity by calculating the Mahalanobis distance between each pair of firm positions within the technology space. However, to prevent any

M

potential look-ahead bias, I depart from Bloom et al. (2013) and use only information about a firm’s

ED

patent shares up to time t in constructing a spillover pool measure at time t. Specifically, I first define the matrix (K, N) at time t as Xt = [

,



], where Xi,t is a vector

PT

of a firm’s patent shares within each K technology class, Xi,t = (xi1,t, xi2,t … xiK,t), and xik,t is firm i’s (i = 1, 2, …, N) proportion of patents in technology classification k over the period up to time t.

CE

Each row of Xt contains a firm’s patent shares within the K technological classes, and each column stands for firm i’s composition of patent shares across all K patent classes. I then normalize Xt with a

AC

firm’s patent share dot product obtaining the (K, N) matrix ̃ such that each element of ̃ ̃ is the

6

For example, AC (RDC-to-sales ratio) may just be correlated with R&D ability (Cohen, Diether, and Malloy, 2013), innovative efficiency (Hirshleifer, Hsu, and Li, 2013), or innovative capacity (Kumar and Li, 2016). 7 Using technology similarities to consider related firms rather than only considering firms within the same industry has the advantage of capturing the technology spillovers generated by firms outside the same product market industry (e.g., Bloom et al., 2013). For example, Xerox Corp. and Apple Inc. compete in different product market industries, but share patents in conductors and insulators (USPTO Class 174), circuit makers and breakers (200), radiant energy (250), and many other categories reflected in the measure of technology spillovers.

10

ACCEPTED MANUSCRIPT

uncentered correlation between firms.8 For example, element (i, j) of ̃ ̃ is the correlation between firm i and j based on their patent portfolio similarity. ̃ ̃

correlation measure between patent classes. Accordingly,

in which each element is now the captures technology spillovers across

CR IP T

Next, I define a (K, K) matrix at time t as

patent classes within a firm. For example, if patent classes i and j frequently coincide within the same firm, then

will be close to one (with

). Putting all together, the technology closeness

using Mahalanobis distance is

where the (i, j) element in matrix

̃

( )

AN US

̃ (

) measures the correlation weights of the overlap in

patent shares between firm i and j at time t by how close their technological focuses (based on their patent shares) are to each other.

M

Finally, given the technology closeness, technology spillover exposure for firm i in year t is defined



( )

is firm j’s R&D-to-sales ratio in year t. 9

then represents the potential technology

PT

where

ED

as

CE

spillover exposure for firm i in year t.10

AC

2.4. Descriptive Statistics Panel A of Table 1 presents the average characteristics for the different firm groups formed by

independently double-sorting on a firm’s absorptive capacity (AC) and technology spillovers (Spill). 8

For example, ̃ can be expressed as ̃

[

⁄(

)

9

⁄(

)

⁄(

) ].

I thank the referee for suggesting this approach of using R&D-to-sales ratio for RDjt. I scale R&D expenditure of firm j by firm sales to be consistent with the AC measure in which I also use firm sales for scaling. Thus, I use R&D divided by market value of equity (ME) for RDjt in Equation (2) when I test robustness using market value of equity for scaling in the AC measure (RDC-to-ME ratio). I thank the referee for suggesting this approach. 10

11

ACCEPTED MANUSCRIPT

Specifically, I form groups every year at the end of June for year t by independently double-sorting based on the 30th and 70th percentiles of a firm’s AC and Spill levels from the fiscal year ending in calendar year t-1. I then hold these portfolios over the next 12 months (July of year t to June of year t+1). The average number of firms is 91, 76, 45, 55, 100, and 75 for the AClow/Spilllow, ACmid/Spilllow,

CR IP T

AChigh/Spilllow, AClow/Spillhigh, ACmid/Spillhigh, and AChigh/Spillhigh groups, respectively. The average size of the firms in the AChigh/Spillhigh portfolio, which is the main focus of this paper, is measured by the annual average of the monthly median market capitalization of firms in the portfolio at the end of June of year t and is $572.29 million. Moreover, untabulated results show that the annual average of the monthly

AN US

median aggregated market capitalization of firms in the AC high/Spillhigh portfolio is $455,467.80 million, which represents about 7.5% of the total U.S. market capitalization. Therefore, firms in this portfolio constitute an economically meaningful portion of the U.S. stock market considering that previous innovation studies (i.e., studies on the effects of both R&D and R&D spillover), as well as studies on the

M

effects of “small-cap value,” have focused on smaller portfolio sizes in terms of combined market capitalization.11 In addition, AChigh/Spillhigh firms tend to have lower book-to-market ratio compared with

ED

firms in the AClow/Spillhigh group. Within each AC group, there seems to be significant variation in the spillover exposure. Spill for the Spilllow and Spillhigh groups is 16.66 and 441.55, respectively, within the

PT

AClow group, and 21.11 and 804.84, respectively, within the AChigh group.

CE

Panel B of Table 1 reports pairwise correlations among absorptive capacity, spillover, and other key firm characteristics. A firm’s absorptive capacity (AC) and spillovers (Spill) are approximately 23%

AC

correlated, indicating that significant variations in absorptive capacity (AC) exist unrelated to a firm’s 11

For example, Chen et al.’s (2013) high R&D spillover portfolio contains 52 firms, on average (1,512 firmyear observations over the 29-year sample period), with an average firm size of $235 million, which represents a smaller amount of the total market capitalization in the U.S. markets. Similarly, Jiang, Qian, and Yao’s (2015) R&D spillover portfolio has 1,020 micro-cap firms, on average (37,770 firm-year observations over the 37-year sample period) with average size of approximately $90.92 million (average ln(Size) = 4.51), which also represent a smaller portion of total U.S. stock market compared to that used in this paper. In R&D-related literature, Cohen, Diether, and Malloy (2013) focus on the high R&D ability portfolio that contains an average of 10 firms that represent about 0.71% of the total U.S. stock market capitalization. Finally, the small-cap value portfolio, which has been widely studied, contains small and value firms that represent only about 0.5% of the total U.S. stock market capitalization.

12

ACCEPTED MANUSCRIPT

technology spillover exposures (Spill). One possible explanation is that the amount spent on R&D (AC) and the amount a firm should allocate to its R&D spending across different technology areas (exposure to spillovers) are two different dimensions of R&D spending. I exploit this variation in AC across firms to examine the effect of a firm’s increased absorptive capacity on subsequent stock returns. Lastly, one

CR IP T

thing worth noting is that Spill and Patents are both increasing with a firm’s market value, suggesting that it is important to control for firm size.

3. Value-Relevance of Spillover-AC Synergy

AN US

3.1. Effect of Spillover-AC Synergy on Subsequent Innovative Activities

In this subsection I test whether or not spillover-AC synergy has positive effects on a firm’s subsequent innovative activities. If there exists some heterogeneity across firms in absorbing outside knowledge, then firms with higher absorptive capacity (AC) are likely to better absorb and convert the

M

available spillovers into value-relevant innovative outcomes (e.g., Evenson and Kislev, 1976; Cohen and

ED

Levinthal, 1989, 1990; Henderson et al., 1996). I therefore examine the effect of spillover-AC synergy on future cite-weighted patents (CWP) by running the following annual Fama-MacBeth regression: )

(

PT

(

where

)

(

(

) )

( )

CE

(

)

is firm i’s total cite-weighted patents in year t. CWP, as opposed to a number of

AC

(unadjusted) patents, is widely used in the literature to better capture a firm’s innovative activities since this measure also incorporates the quality of each patent granted (e.g., Trajtenberg, 1990; Bloom, Schankerman, and Van Reenen, 2013). Since the previous literature on innovation documents that a firm’s measures for innovative activities (i.e., the number of citations received) show large variation across different technology classes and application years, I follow the literature and adjust for these issues (e.g., Seru, 2010; Bena and Garlappi, 2012; Almeida, Hsu, and Li, 2013; Hirshleifer, Hsu, and Li,

13

ACCEPTED MANUSCRIPT

2013). Specifically, I calculate the adjusted citations by scaling the number of citations received from each patent by the average number of citations received by the patents applied in the same year and assigned to the same technology class by USPTO.12 AC at year t-1;

(

(

) is the natural logarithm of one plus

) is the natural logarithm of one plus Spill at year t-1; and the Xi,t-1 is a

CR IP T

vector of control variables. I use firm size as measured by the natural log of market capitalization at the end of year t-1, ln(MEi,t-1), book-to-market as measured by the log of book value to the market value of equity as ln(B/Mi,t-1), firm leverage at the end of year t-1 as ln(1+levi,t-1), and, finally, firm age at the end of year t-1 as measured by the years since a firm’s first appearance on COMPUSTAT ln(agei,t-1). I also

AN US

include industry fixed effects based on the Fama and French 48 industries classifications. To deal with potential outliers, I winsorize the explanatory variables used in both models at the top and bottom 1% levels.

Table 2 provides the results. Column (1) includes only AC and Spill variables. The result seems to

M

suggest that a firm’s absorptive capacity and exposure to spillovers are both independently and positively related to subsequent CWP. However, once I include the interaction term between AC and

ED

Spill, as in Column (2), the effect of spillovers on a firm’s subsequent number of cite-weighted patents

PT

appears to be a strong function of its absorptive capacity. That is, firms with higher absorptive capacity (higher AC) are more impacted by spillovers in relation to their future successful innovative outcomes.

CE

The coefficient on the interaction term, ln(1+AC)*ln(1+Spill), is also statistically significant at 5% significance level (t = 2.17). The magnitude of the coefficient on spillovers alone ln(1+Spill), however,

AC

is reduced by a large amount, while the magnitude of the coefficient on the interaction term is high, suggesting that a firm’s future innovative activities can vary significantly depending on its absorptive capacity, even though the firm is exposed to the same level of potential spillovers. That is, the spilloverAC synergy effect is economically significant; a one-standard-deviation increase in exposure to potential

12

In untabulated tests, I also use unadjusted citations in calculating innovation-related productivity. Specifically, I calculate the cite-weighted patents measure as the total number of patent citations received from grant year t to 2006. The results remain the same.

14

ACCEPTED MANUSCRIPT

spillovers exhibits an approximate 16% difference in the firm’s subsequent number of cite-weighted patents generated depending on its level of absorptive capacity (i.e., AC high versus low). Dass, Nanda, and Xiao (2015) find that there seems to exist a truncation bias in the later sample years of NBER patent citation data since application information is released only after the patent is granted.

CR IP T

Thus, following Dass et al. (2015), I exclude the period from 2004 to 2006 from my sample and run the same regression as in Column (2) as a robustness check of my results. Column (3) shows that the results statistically and economically remain the same. As a final robustness check, I also use dummy variables for the AC (AClow being below the 30th percentile and AChigh being above the 70th percentile of AC) and

AN US

run the same regressions as in Columns (2) and (3). Columns (4) and (5) provide the respective results. As shown in Columns (4) and (5), I find results consistent with those in Columns (2) and (3). Overall, the results in Table 2 suggest that even though a firm is exposed to the same level of potential spillovers, the firm’s absorptive capacity significantly influences its subsequent innovative

M

activities.

ED

3.2. Effect of Spillover-AC Synergy on Future Operating Performance and Market Valuation

PT

In this subsection I examine whether or not spillover-AC synergy has positive effects on a firm’s subsequent operating performance (OP) and market valuation. Testing whether spillover-AC synergy has

CE

a positive impact on future operating performance and market valuation will provide further evidence that spillover-AC synergy contains value-relevant information regarding future firm fundamentals.

AC

Thus, I first run the following annual Fama-MacBeth regressions of a firm’s future operating performance measures on the previous year’s absorptive capacity, spillovers, and firm-control variables: (

)

(

)

(

)

(

) ( )

15

ACCEPTED MANUSCRIPT

is the operating performance measures at year t, ln(1+AC i,t-1) is the natural logarithm of one

where

plus AC at year t-1, ln(1+Spill i,t-1) is the natural logarithm of one plus Spill at year t-1, and the Xi,t-1 is a vector of control variables. I use two operating performance measures: return on assets (ROA), measured as a firm’s earnings before interest and taxes (EBIT) plus depreciation compared to lagged total book

CR IP T

value of assets, and profit margin (PM), defined as a firm’s earnings before interest and taxes (EBIT) plus the depreciation divided by sales. Since Hirshleifer, Hsu, and Li (2013) document that a firm’s innovative efficiency (IE) is positively related to subsequent operating performance, I also include ln(1+Innovative Efficiencyt-1) to control for this effect. Following Hirshleifer, Hsu, and Li (2013),

AN US

Innovative Efficiencyt-1 is defined as the total number of patents granted in year t-1 scaled by the stock of R&D in year t-3 (

).

Other firm controls in vector Xi,t-1 include patent intensity, ln(1+Patent/ME)i,t-1, advertising intensity, ln(1+Advertising/ME)i,t-1, capital expenditure, ln(1+CAPX/ME)i,t-1, change in operating performance

M

between year t-2 and year t-1, and one-year lagged operating performance (i.e., ROA‒1 and PM‒1), size as measured by the natural log of market capitalization at the end of year t-1, ln(MEi,t-1), and, finally,

ED

book-to-market as measured by the log of the book value to the market value of equity, ln(B/Mi,t-1).13

PT

To test whether spillover-AC synergy is positively related to future market valuation, I next employ an annual Fama-MacBeth regression of a firm’s future market-to-book ratio on the previous year’s

)

AC

(

CE

absorptive capacity, spillovers, and control variables. I accordingly run the following regression: (

(

) )

(

(

) )

13

( )

In untabulated test, I also include control variables, such as a dummy variable indicating whether or not the past change in the operating performance is negative, and the interaction with the past changes in the profitability, to deal with the nonlinear mean reversion process of profitability (e.g., Fama and French, 2000). The results remain the same.

16

ACCEPTED MANUSCRIPT

where

is the market-to-book ratio at year t calculated as the market value of equity at the end of

December of year t divided by book equity for the fiscal year ending in year t; ln(1+AC i,t-1) is the natural logarithm of one plus AC at year t-1; ln(1+Spill i,t-1) is the natural logarithm of one plus Spill at year t-1; and the Zi,t-1 is a vector of control variables. Following the previous literature (i.e., Hirshleifer, Hsu, and

CR IP T

Li, 2013), I include the following vector of control variables: innovative efficiency, ln(1+Innovative Efficiencyt-1), patent intensity, ln(1+Patent/ME)i,t-1, advertising intensity, ln(1+Advertising/ME)i,t-1, capital expenditure, ln(1+CAPX/ME)i,t-1, and (1/Book Equity)i,t-1. I also include the one-year lagged market-to-book ratio to control for persistence in the market-to-book-ratio (e.g., Fama and French, 2006).

AN US

In both Equations (4) and (5), I also include industry fixed effects based on the Fama and French 48 industries classifications. To deal with potential outliers, I winsorize the explanatory variables used in both models at the top and bottom 1% levels. In both Equations (4) and (5), a key variable of interest is the interaction between absorptive capacity and spillovers ln(1+AC)*ln(1+Spill) since this study focuses

M

on the additional role of a firm’s absorptive capacity (AC) when the firm is exposed to the potential in both equations.

ED

spillovers. I therefore focus on the coefficient

Table 3 provides the results. As hypothesized, the effect of technology spillovers on subsequent

PT

operating performance appears to be a strong function of a firm’s absorptive capacity (AC). Specifically, Columns (1) and (2) of Table 3 show that the coefficient on the interaction between AC and spillovers

CE

exhibits a strong positive relation with subsequent operating performance. For example, the coefficients on the interaction term ln(1+AC)*ln(1+Spill) are positive and statistically significant for both ROA (at

AC

10% level) and PM (at 1% level). The economic significance is also high; firms with high potential spillover exposure can have an approximate 6% (3%) difference in change in PM (ROA) depending on the level of ACs (high-AC versus low-AC). These results suggest that spillover-AC synergy exhibits a strong positive relation to future firm fundamentals and therefore contains value-relevant information. For a robustness check, I also use dummy variables for the AC (AClow being below the 30th percentile, AChigh being above the 70th percentile, and ACmid being between the 30th percentile and the 70th

17

ACCEPTED MANUSCRIPT

percentile of AC sort) and run the same regressions as in Columns (1) and (2). The results are provided in Columns (4) and (5), respectively. I find consistent results. Next, I turn my attention to the effect of spillover-AC synergy on a firm’s subsequent market valuation. Column (3) of Table 3 presents the results. As shown in Column (3), the effect of spillover

CR IP T

exposure alone does not seem to be positively related to future market value of a firm. Rather, the effect of spillovers on future firm value appears to depend strongly on a firm’s absorptive capacity. For example, the coefficient on the ln(1+Spill) has very low magnitude and is also statistically insignificant. On the other hand, the coefficients on the interaction term ln(1+AC)*ln(1+Spill) are positive and

AN US

statistically significant at the 5% significance level. The economic significance is also high; even though firms are all exposed to high spillovers, they can have an approximate 13% difference in their future market valuations depending on their level of ACs. Column (6) presents results using the AC dummy variables for robustness and shows results similar to those in Column (3). Thus, Columns (3) and (6)

M

suggest that firms’ subsequent market values can differ significantly even though they are exposed to the same level of potential spillovers. Furthermore, the result in Column (3) that spillover-AC synergy,

ED

rather than just spillover exposure alone, is positively related to future market value implies that the value-relevant information of spillover-AC synergy is not fully incorporated into stock prices

PT

immediately. I will further explore this point by testing the effects of spillover-AC synergy on firms’

CE

subsequent abnormal stock returns.

AC

4. Absorptive Capacity, Technology Spillovers, and Future Stock Returns In this section I examine whether or not spillover-AC synergy positively predicts future cross-section

of stock returns. I use Fama-MacBeth (1973) cross-sectional regression approach and a calendar-time portfolio regression approach.

4.1. Fama-MacBeth Cross-Sectional Regression Results

18

ACCEPTED MANUSCRIPT

In this section I examine whether spillover-AC synergy, rather than just the size of spillovers has predictability in future cross-section of stock returns. I employ a monthly Fama-MacBeth (1973) crosssectional regression of individual excess stock returns (individual stock returns over the risk-free rate) on lagged absorptive capacity (AC) and spillovers, along with firm characteristics from July of year t to

CR IP T

June of year t+1. The independent variables, except for size, momentum, return reversal, and idiosyncratic volatility, are from the fiscal year ending in the previous calendar year (t-1). I do this to allow enough time for information about the key variables to be incorporated into stock prices. Size, momentum, return reversal, and idiosyncratic volatility are from a previous month.

AN US

The key independent variables are ln(1+AC), ln(1+Spill) and the interaction between these two variables ln(1+AC)*ln(1+Spill). The coefficient of interest is the coefficient on the interaction term. I then use the familiar cross-sectional controls shown to predict future stock returns. The control variables I use are firm size, book-to-market, momentum, return reversal, idiosyncratic volatility, market leverage,

M

ROA, and asset growth. Firm size ln(ME) is the natural log of the market capitalization at the previous month; book-to-market ln(B/M) is the natural log of the book value of equity in the fiscal year ending in

ED

calendar year t-1 divided by the market value of equity at the end of calendar year t-1; momentum (Momentum) is the prior 12-month returns with a one-month gap between the holding period and current

PT

month; return reversal (ret‒1) is one-month lagged stock return; idiosyncratic volatility (IVol) is the

CE

residual returns from the Fama-French three-factor model; market leverage (Leverage) is calculated as book value of debt divided by market value of assets; ROA is measured by ratio of a firm’s earnings

AC

before interest and taxes (EBIT) plus depreciations to lagged total book value of assets; and Asset growth is the change in total assets divided by lagged total assets. For some model specifications, I additionally include industry fixed effects based on the Fama and French 48 industries classifications. Table 4 presents the results for the monthly Fama-MacBeth regressions. Column (1) shows the results without the interaction term in order to compare these results to those of previous studies that focus mainly on just the size of spillover exposure (e.g., Chen, Chen, Liang, and Wang, 2013; Jiang, Qian, and

19

ACCEPTED MANUSCRIPT

Yao, 2016). By examining the coefficients on the ln(1+Spill), the level of the potential spillover exposures alone seems to positively predict future stock returns. However, when I include the interaction term as in Columns (2), the result shows that a firm’s level of absorptive capacity (AC) plays a crucial role in determining spillover’s future stock return predictability. The coefficient on the

CR IP T

ln(1+AC)*ln(1+Spill) term is positive and statistically significant at 1% significance level, suggesting that the marginal value of potential spillovers increases in a firm’s absorptive capacity. Additionally, Column (3) reports the results when running the same model as in Column (2), but including industry dummies to control for any industry-level characteristics that may be driving my results. Adding industry

AN US

fixed effects reduces both economic and statistical significance to some extent, suggesting that it is important to control for the industry effect. However, the coefficient on the interaction term still remains highly significant at the 5% significance level. The economic significance is also high: a one-standarddeviation increase in exposure to potential spillovers exhibits an approximate 0.62% increase in monthly

M

future excess stock returns, on average (7.5% annually), for high AC firms compared with only an about 0.03% increase in monthly excess returns (0.4% annually) for low AC firms.

ED

Finally, I check the robustness of the results. First, I test whether the high future stock returns concentrated on firms with high AC and high spillover exposure shown in Table 4 are simply driven by

PT

the effect of the size differentials between high AC firms and low AC firms within each spillover group.

CE

As evidenced by Table 1, high AC firms tend to be smaller in size within the high spillover group. Thus, in addition to firm size control, I accordingly include control variable, the interaction between size and

AC

the spillovers ln(ME)*ln(1+Spill), to combat this potential concern. Column (4) of Table 4 shows that including the ln(ME)*ln(1+Spill) term does not affect the coefficient on ln(1+AC)*ln(1+Spill). Second, I exclude the Internet (dot-com) bubble period (from calendar year 1999 to 2001) from my

sample and run the same model as in Column (4) to mitigate the concern that many high-tech firms may drive the results in Table 4 since they were experiencing unusually high stock returns during this period. The results are presented in Column (5) of Table 4 and remain the same.

20

ACCEPTED MANUSCRIPT

Third, I also use dummy variables for the AC (AClow being below the 30th percentile, AChigh being above the 70th percentile, and ACmid being between the 30th percentile and the 70th percentile of AC sort) and run the same regressions as in Column (4). The results are provided in Column (6) and are consistent with the results in Column (4).

CR IP T

Fourth, I use a different definition of absorptive capacity, as well as spillover exposure. Specifically, I use a firm’s R&D capital (RDC) scaled by the firm’s market value of equity (ME) instead of firm sales as an alternative proxy for a firm’s absorptive capacity (ACAlt) since R&D-related results can be different when R&D is scaled by market value of equity (e.g, Hou, Xue, and Zhang, 2015).14 I also change the

AN US

spillover measure accordingly. I use R&D-to-ME ratio for the RDjt in Equation (2) instead of R&D-tosales ratio to be consistent with the scaling in the ACAlt and construct the alternative spillover measure SpillAlt. The Appendix provides the results. As shown in Table B1, the results are consistent with the results in Table 4 that only the interaction term between the absorptive capacity and the spillovers,

M

ln(1+ACAlt)*ln(1+SpillAlt), rather than the spillover variable alone is positive and statistically significant. In sum, the results in Table 4 suggest that firms exposed to large potential spillovers outperform in

ED

the future only when these large spillover exposures are coupled with high absorptive capacity (AC). That is, firms without sufficient AC do not seem to enjoy the full benefits of technology spillovers.

CE

PT

Therefore, stock return predictability seems to be concentrated among firms with spillover-AC synergy.

4.2. Portfolio Analysis Results

AC

In this subsection, I confirm the cross-sectional stock return results in Table 4 by implementing portfolio analysis and present a possible trading strategy. I calculate the monthly value- and equalweighted portfolio returns. At the end of June of year t, I sort firms independently into absorptive capacity (AC) terciles (i.e., AClow and AChigh below the 30th percentile and above the 70th percentile of

14

In untabulated tests, I also try to construct the absorptive capacity measure by using only the current R&D (instead of R&D capital) and scale it by market value of equity. The results are also robust to this alternative approach.

21

ACCEPTED MANUSCRIPT

AC) and spillover (Spill) terciles (i.e., Spilllow and Spillhigh below the 30th percentile and above the 70th percentile of Spill). AC and Spill are measured in the fiscal year ending in the calendar year t-1. I consequently form portfolios from the intersection of the firm’s AC and Spill sorts. I then hold these portfolios from July of year t until June of year t+1 and calculate the value-weighted

CR IP T

monthly returns for each portfolio. Following previous studies, I require firms to have stock prices higher than US $5 at the time of portfolio formation. I also calculate the equal-weighted portfolio returns following Hou, Xue, and Zhang (2014 and 2015) and Gu (2016).15 Specifically, to mitigate the impact of microcap firms, I first exclude stocks with market equity below the 20th percentile of NYSE breakpoints

AN US

and then calculate the AC and Spill tercile breakpoints within this sample. Finally, I calculate equalweighted monthly portfolio returns for each portfolio. Given these value- and equal-weighted portfolio returns, I employ the Carhart (1997) four-factor model, the Fama and French (2015) five-factor model, and the Hou, Xue, and Zhang (2015) q-factor model to adjust for style or risk differences among the

M

portfolios. Furthermore, to mitigate the effects of industry or well-known characteristics on portfolio returns, I also calculate industry- and DGTW characteristics-adjusted portfolio returns using industry-

ED

matched and DGTW characteristics-matched returns respectively for the benchmark instead of the riskfree rate.16

PT

I provide portfolio analysis results in Table 5. Panel A of Table 5 presents the monthly value-

CE

weighted portfolio alphas for each portfolio using absorptive capacity (AC) and spillover (Spill) sorts. First, the monthly alpha patterns within AChigh group across all the factor models are consistent with the

AC

previous spillover literature that shows that the level of exposure to spillovers positively predicts subsequent stock returns (e.g., Hsu, 2011; Chen, Chen, Liang, and Wang, 2013; Jiang, Qian, and Yao, 2016). For example, the last column of Table 5 shows that the spread between Spillhigh and Spilllow portfolio within the AChigh group is 0.43% per month (t = 2.76) based on the Carhart four-factor model.

15

I thank the referee for suggesting this approach. The DGTW characteristics are from Daniel, Grinblatt, Titman, and Wermers (1997) and Wermers (2004). The DGTW benchmarks are available via http://www.smith.umd.edu/faculty/rwermers/ftpsite/Dgtw/coverpage.htm. 16

22

ACCEPTED MANUSCRIPT

However, the level of spillover seems to have positive stock return predictability only for firms in the AChigh group. A portfolio of firms exposed to large potential technology spillovers that also have high absorptive capacity (“AbsorbSpill” portfolio hereafter) outperforms a portfolio of firms with a high amount of

CR IP T

spillover exposure but low absorptive capacity (“NoAbsorbSpill” portfolio hereafter). The valueweighted monthly Carhart four-factor alpha for the AbsorbSpill portfolio (0.54% with t = 4.45) is substantially greater in magnitude compared with the monthly alpha of the NoAbsorbSpill portfolio (0.03% with t = 0.37). Moreover, the monthly portfolio return of the NoAbsorbSpill portfolio not only

AN US

demonstrates lower abnormal returns but is also statistically insignificant at the 10% significance level. Finally, Column (7) of Table 5 presents the monthly alphas of a spread portfolio that is long in AbsorbSpill firms and short in NoAbsorbSpill firms. This zero-investment portfolio earns a 0.57% monthly alpha (or approximately 7% annually) with t = 4.22 based on the Carhart four-factor model. The

M

results using other factor models or using industry- or DGTW characteristics-adjusted portfolio returns are consistent with these results. Additionally, the equal-weighted portfolio results (Panel B of Table 5)

ED

are highly consistent with the value-weighted portfolio results presented in Panel A. These results suggest that firms that have low AC, do not appear to be able to effectively absorb and benefit from

PT

potential spillover pools even though the pool is large.

CE

The results presented in Table 5 seem to suggest that the spillover-AC synergy is undervalued relative to various existing factor model benchmarks. That is, the market seems to misvalue the positive

AC

effect of the spillover-AC synergy ex ante. However, it could be possible that the existing asset pricing models may not be suitable in controlling for mispricing factors. Therefore, I further test whether the abnormal returns from the AbsorbSpill portfolio can be explained by the well-known mispricing factors. To test this, I add a mispricing factor, the UMO factor from Hirshleifer and Jiang (2010), to the Carhart (1997) four-factor model and run the time-series regression of portfolio excess returns as in Table 5. Alternatively, I employ the mispricing factor model of Stambaugh and Yuan (2017) that contains

23

ACCEPTED MANUSCRIPT

mispricing factors (MGMT and PERF), in addition to the market and the size factors. I use both valueand equal-weighted portfolios as in Table 5. The portfolio analysis results using the existing mispricing factors are presented in Table 6. As shown in both Panels A (value-weighted portfolios) and B (equal-weighted portfolios), the results are highly

CR IP T

consistent with the four-factor model results shown in Table 5. Specifically, both the magnitudes of alphas and statistical significance remain similar to those for the four-factor model in Table 5, suggesting that adding the existing mispricing factors does not have a significant effect on the abnormal returns of the AbsorbSpill portfolio. Therefore, these findings, combined with findings in Table 5, suggest that the

AN US

positive relation between the spillover-AC synergy and subsequent abnormal stock returns is not explained by the existing risk or mispricing factors.

Overall, the results in Tables 4, 5, and 6 demonstrate that a strong complementary relation exists between a firm’s absorptive capacity (AC) and exposure to spillovers (“spillover-AC synergy”); this

M

spillover-AC synergy, rather than just the size of the spillover pool, positively predicts the subsequent cross-section of stock returns. This implies that spillover-AC synergy is most likely undervalued relative

ED

to various factor model benchmarks (using both existing risk and mispricing factors). That is, the market does not seem to distinguish between spillover-AC synergy and merely high exposure to spillovers. My

PT

findings therefore suggest that the positive future abnormal stock returns found in the previous spillover

CE

literature (e.g., Hsu, 2011; Chen et al., 2013; Jiang, Qian, and Yao, 2016) are mainly driven by firms

AC

enjoying spillover-AC synergy.

5. Limited Investor Attention and Evidence of Misvaluation In this section I test whether or not the positive relation between spillover-AC synergy and

subsequent abnormal stock returns found in Section 4 comes from the market’s undervaluation of spillover-AC synergy effects. A large number of recent studies document that the limited attention of investors results in the market’s underreaction to value-relevant public information and therefore return

24

ACCEPTED MANUSCRIPT

predictability (e.g., Klibanoff, Lamont, and Wizman, 1998; Deng, Lev, and Narin, 1999; Hong, Lim, and Stein, 2000; Huberman and Regev, 2001; Hirshleifer and Teoh, 2003; Lev, 2004; Peng and Xiong, 2006; DellaVigna and Pollet, 2009; Hirshleifer, Lim, and Teoh, 2009 and 2011; Hou, Peng, and Xiong, 2009; Hirshleifer, Hsu, and Li, 2013, 2014). The literature on limited investor attention documents that

CR IP T

investors tend to undervalue favorable information if such information is less salient. Specifically, investors view more complex information that has low-processing fluency more skeptically, so the complexity of information may lead to investor undervaluation.

My hypothesis is that the market may have difficulty in processing information about other firms’

AN US

innovative activities, as well as their similarities and complementarities to a given firm’s own innovations. More importantly, investors also will have difficulty in recognizing to what extent potential spillovers are absorbed by a given firm. The difficulty of processing complex information therefore likely results in the market’s undervaluation of the positive effects of spillover-AC synergy on enhancing

M

a firm’s future innovation-related performance and operating performance.

If an investor’s limited attention and low information processing power are the source of the positive

ED

relation between spillover-AC synergy and subsequent abnormal stock returns, I would expect the positive relation between spillover-AC synergy and subsequent stock returns to be most pronounced

PT

among stocks that have low investor attention. The previous literature on limited attention documents

CE

that firm-specific information spreads more slowly for small-firm stocks and stocks with low analyst coverage (e.g., Brennan, Jagadeesh, and Swaminathan, 1993; Hong and Stein, 1999; Hong, Lim, and

AC

Stein, 2000; Hirshleifer and Teoh, 2003; Hong, Torous, and Valkanov, 2007; Hou, 2007; Cohen and Frazzini, 2008). Also, previous studies have used institutional holdings as the proxy for the degree of investor attention with low institutional holdings as a low investor attention stocks (e.g., Hirshleifer, Hsu, and Li, 2014; Jiang, Qian, and Yao, 2016). Accordingly, I use analyst coverage, institutional ownership,

25

ACCEPTED MANUSCRIPT

and firm size as proxies for limited attention. These proxies are all widely used for testing limited attention.17 I test my limited attention hypothesis by dividing my sample into high and low investor attention stocks and then running the same Fama-MacBeth cross-sectional regression as in Section 4.1 within

CR IP T

these subsamples. Size subsamples are obtained by splitting my sample into small and large size groups based on the median NYSE breakpoints at the end of June in year t.18 Alternatively, I obtain analyst coverage subsamples by dividing my sample into high and low analyst coverage stocks based on the median analyst coverage at the end of year t-1. I use the Institutional Brokers’ Estimate System (I/B/E/S)

AN US

to obtain the analyst coverage for each stock; analyst coverage is calculated by averaging the monthly number of analysts providing current fiscal year earnings forecasts over the previous year. Institutional ownership subsamples are obtained by dividing the whole sample into high and low institutional ownership stocks based on the median institutional ownership at the end of year t-1. Institutional

M

ownership denotes the fraction of firm shares outstanding owned by institutional investors in year t–1. I obtain the institutional holdings data from the Thomson-Reuters 13f institutional holdings dataset.

ED

Table 7 reports that the positive relation between spillover-AC synergy and future stock returns is economically and statistically significant, but only for the firms with low investor attention. The effect of

PT

spillover-AC synergy is much weaker in magnitude and statistically insignificant for firms with high

CE

investor attention. For example, Columns (1) and (2) of Table 6 report that the coefficients of the interaction term ln(1+AC)*ln(1+Spill) are 1.66% (t = 2.03) and 0.71% (t = 1.57) for low and high

AC

analyst coverage, respectively. For the institutional ownership subsamples, the coefficients for the interaction term ln(1+AC)*ln(1+Spill) are 1.16% (t = 2.33) and 0.52% (t = 1.41) for low- and highinstitutional ownership firms, respectively. For the size subsamples, the coefficients for the interaction term ln(1+AC)*ln(1+Spill) are 1.09% (t = 2.19) and 0.67% (t = 1.34) for small and large size firms, 17

See, for example, Hong, Lim, and Stein (2000), Hirshleifer and Teoh (2003), and Hirshleifer, Hsu, and Li (2013, 2014), among others. 18 I obtain NYSE market capitalization breakpoints from Kenneth French’s Web site.

26

ACCEPTED MANUSCRIPT

respectively. Moreover, the χ2-statistics on the difference of coefficients between the low and high attention subsamples show that the differences are statistically significant across all the limited attention proxies. These differences between low and high investor attention subsamples provide evidence that limited investor attention contributes to the undervaluation of firms with high AC and high spillover

CR IP T

exposures.

6. Robustness

In this section I test the robustness of the absorptive capacity measure (RDC-to-sales ratio) used in

AN US

this paper. Additionally, I address alternative explanations that also can drive the positive relation between spillover-AC synergy and the subsequent abnormal stock returns shown in Section 4.

6.1. Robustness tests on RDC-to-sales ratio as a proxy for absorptive capacity (AC)

M

Even though the previous studies and the evidence documented so far in this paper support the notion

ED

that the R&D intensity is a good proxy for the absorptive capacity (AC) in the presence of technology spillovers, R&D intensity of a firm may still be correlated with other innovation-related ability measures

PT

that already have been used in the previous studies. For example, AC may just be reflecting a firm’s R&D ability (Cohen, Diether, and Malloy, 2013), that is, a firm’s ability to convert R&D investments

CE

into tangible outcomes (i.e., sales growth). Likewise, AC can be correlated with a firm’s innovative efficiency (IE) of Hirshleifer, Hsu, and Li (2013) or innovative capacity (IC) of Kumar and Li (2016).

AC

Finally, Gu (2016) documents that product market competition has a complementary effect with a firm’s R&D on future stock returns. So it also could be possible that AC just reflects the level of product market competitiveness in the presence of spillovers. In this subsection, I therefore test whether R&D intensity used in this paper (RDC-to-sales ratio) to proxy for AC is just a reflection of these previously documented innovation-related abilities, rather than a firm’s absorptive capacity. Specifically, I control for the interaction effects of spillovers with (1) R&D

27

ACCEPTED MANUSCRIPT

ability, Ability, calculated as the average of the coefficients from the regressions of sales growth on the past five-year R&D-to-sales ratio (Cohen, Diether, and Malloy, 2013); (2) innovative efficiency, Innovative efficiency, of Hirshleifer, Hsu, and Li (2013), defined same as in Section 3.2; (3) innovative capacity, Innovative capacity, measured by R&D-active firms’ asset growth (AG) in the previous year,

CR IP T

where R&D-active firms have non-missing R&D-to-sales ratio (Kumar and Li, 2016); and (4) the Herfindahl-Hirschman index, HHI, defined as the sum of squared market shares of each firm in the industry. I repeat the previous analyses (as in Tables 2, 3, and 4) and control for these interaction terms. Table 8 presents the results. Column (1) presents the results for the annual Fama-Macbeth cross-

AN US

sectional regression of firms’ future cite-weighted patents (CWP) on absorptive capacity (AC) and the level of technology spillovers, like in Table 2. Columns (2) through (4) present the results for the annual Fama-MacBeth cross-sectional regressions for various subsequent firm performance on absorptive capacity (AC) and spillovers, like in Table 3. Finally, Column (5) reports a monthly Fama-Macbeth

M

cross-sectional regression of firms’ excess stock returns on absorptive capacity (AC) and the level of technology spillovers, like in Table 4.19 As shown in Table 8, the interaction term ln(1+AC)*ln(1+Spill)

ED

remains positive and statistically significant at either the 5% or the 1% level even after controlling for the interaction between the alternative innovation-related abilities and spillovers. The results in Table 8

PT

are consistent with the argument that R&D intensity (RDC-to-sales ratio) captures a firm’s ability to

CE

absorb technology spillovers and has an incremental effect to the existing innovation-related ability

AC

measures.

6.2. Effect of Product Market Rivalry on Spillover-AC Synergy and Stock Return Relation Although the technology spillover exposure measure (Spill) used in this paper incorporates

technological similarities (i.e., holding similar patent portfolios) rather than product-market similarities 19

In untabulated robustness test, I also control for the interaction between spillovers and the alternative innovation-related abilities (i.e., R&D ability, innovative efficiency, innovative capacity, and HHI) examined in this subsection for the limited attention tests (as in Table 7). The results remain the same (for both magnitudes and statistical significance) even after controlling for these alternative explanations.

28

ACCEPTED MANUSCRIPT

(i.e., operating within similar product-market segments), the Spill measure still contains the R&D investments of firms that compete within the same product market. Although technology spillover represents a positive externality, capturing pure positive externality effects within the data is a challenging task for researchers (e.g., Jaffe, 1986; Bloom, Schankerman, and Van Reenen, 2013). Since

CR IP T

the technology spillover measure is a sum of the R&D investments of all firms (including product market competitors’ R&D), the high level of the spillover pool may pick up an increase in productmarket competition that will negatively affect a given firm’s future cash flows in the long-run (e.g., Bloom, Schankerman, and Van Reenen, 2013). From this perspective, firms that face high spillover

AN US

pools will likely be considered risky, leading to higher expected stock returns.

To combat this potential concern, I follow Bloom, Schankerman, and Van Reenen (2013) and disentangle product market rivalry effects from the pure technology spillover effects by constructing an additional spillover measure that captures potential spillovers stemming from product-market

M

competitors. Accordingly, I construct the measure Spillsic, which is calculated analogous to the Spill, but differs in calculating the weight of each firm’s R&D. That is, Spillsic incorporates a firm’s sales

ED

portfolio (i.e., sale shares across different industries), rather than its patent portfolio. The detailed estimation steps are provided in Appendix A.

PT

I run the same Fama-MacBeth regressions as in Column (4) of Table 4, but include the Spillsic

CE

measure to disentangle the product-market rivalry effects from the technology spillover effects. Column (1) of Table 7 shows that including this product-market rivalry effects does not affect the results found in

AC

the previous sections. This suggests that the positive spillover-AC synergy and stock return relation is unlikely to come from a risk associated with an increase in R&D competition.

6.3. Other Innovation-Related Factors

29

ACCEPTED MANUSCRIPT

To test whether other innovation-related factors drive the results in Table 4, I also run the same Fama-MacBeth regressions as in Column (4) of Table 4 (including Spillsic from previous subsection), but include a host of innovation-related factors shown to be positively related to future stock returns. Existing studies on corporate innovation show that the number of patents granted to a firm (e.g.,

CR IP T

Deng, Lev, and Narin, 1999) and a firm’s innovative efficiency (e.g., Hirshleifer, Hsu, and Li, 2013) have a positive impact on the firm’s future stock returns. It is possible that spillover-AC synergy may just reflect a different level of patent intensity or efficiency in producing patents. Thus, in Table 9, Column (2), I follow Hirshleifer et al. (2013) and use the natural logarithm of one plus the number of

AN US

patents (adjusted for different patent classes and application year) divided by the market value of equity (Patent/ME) and the natural logarithm of one plus innovative efficiency (Innovative efficiency) to control for these effects.20 The coefficient on the interaction term (between AC and spillover) remains highly

patent holdings or innovative efficiency.

M

consistent, suggesting that spillover-AC synergy explains additional effects not identified by a firm’s

Gu (2016) documents that product market competition and R&D investment have a significant and

ED

positive interaction effect on future stock returns. Since high AC firms are R&D-intense firms and high spillover exposure might be related to fierce competition among competitors, the spillover-AC synergy

PT

effect may be explained by the interaction effect of product market competition and R&D. I thus follow

CE

Gu (2016) and use the interaction between the Herfindahl-Hirschman Index (HHI), defined as the sum of squared market shares of each firm in the industry, and absorptive capacity (AC) to control for this effect.

AC

Additionally, Li (2011) finds that R&D-intense firms earn high future returns only when these firms are financially constrained. From this perspective, the risk associated with a firm’s financial constraint may drive the positive relation between spillover-AC synergy and subsequent abnormal stock returns. Hence, I follow Li (2011) and use the interaction between the financial constraint measure (KZ-index)

20 Construction of innovative efficiency is described in Section 3.2 and is described in more detail in Hirshleifer, Hsu, and Li (2013).

30

ACCEPTED MANUSCRIPT

and absorptive capacity (AC) to control for this effect.21 The results controlling for the interaction effect of R&D (AC) with product market competition or financial constraints are presented in Column (3) of Table 7. As shown in Table 7, the effect of the spillover-AC synergy on subsequent stock returns is robust to the inclusion of different R&D interaction effects as shown in Gu (2016) and Li (2011).

CR IP T

Finally, Cohen, Diether, and Malloy (2013) document that a firm’s ability to convert R&D investments into tangible outcomes (i.e., sales growth) matters in explaining the relation between R&D and future stock returns. Specifically, high R&D is related to high future stock returns only when a firm has high R&D ability. Thus, following Cohen et al. (2013), I construct the R&D ability measure (Ability)

AN US

and interact with the absorptive capacity to control for this effect.22 The results are presented in Column (4) to (6) of Table 7 depending on the definitions of financial constraints. As shown in the table, the effect of the spillover-AC synergy on future stock returns is robust to the inclusion of the interaction between R&D (AC) and R&D ability, as well as different definitions of financial constraints.

M

To summarize, I still find a positive and significant spillover-AC synergy and future stock return relation, even after controlling for other innovation-related factors from previous studies, suggesting that

ED

the positive relation between spillover-AC synergy and subsequent abnormal stock returns found in this

PT

paper are not mainly driven by existing innovation-related factors.

CE

6.4. Sensitivity to Down-Market Periods The higher alpha for the AbsorbSpill portfolio also may be explained by downside risk if firms in the

AC

AbsorbSpill portfolio exhibit high sensitivity to the market return, particularly when the market return is very low. Accordingly, I run the same portfolio analysis as in Section 4.2, but I restrict my sample period to times during which the market return is low. These down-market periods are defined as the months in 21

The KZ-index (KZ) is from Kaplan and Zingales (1997). I also try different measures of financial constraints, such as the SA-index (SA) from Hadlock and Pierce (2010) and the WW-index (WW) from Whited and Wu (2006). I report these results in Column (5) and (6) of Table 7. 22 Construction of R&D ability is described in Section 6.1 and is described in more detail in Cohen, Diether, and Malloy (2013).

31

ACCEPTED MANUSCRIPT

which either the market return (CRSP value-weighted market index) is below -3% (e.g., Giglio and Shue, 2014) or the market return is below the Treasury-bill rate (e.g., Chan, Lokonishok, and Sougiannis, 19F19F

2001).23 The untabulated results show that the value-weighted AbsorbSpill portfolio’s beta coefficient on the excess return on the market,

, is 0.30, which is even lower than the beta coefficient in Table 5

CR IP T

(0.50), where I measure downside months as having a market return below -3%. The beta coefficient exhibits 0.33 when the downside months are defined as those in which the market return is less than the Treasury-bill rate. Importantly, the AbsorbSpill portfolio’s beta coefficient on the market excess return is always lower than the NoAbsorbSpill portfolio’s beta. For example, when the downside months are

AN US

defined as those in which the market return is below -3% (below the Treasury-bill rate), then the beta of the AbsorbSpill portfolio is 0.30 (0.33), whereas the beta of the NoAbsorbSpill portfolio is 0.41 (0.43). These results suggest that differential exposure to downside risk does not seem to play a key role in

M

explaining the higher alpha for the AbsorbSpill portfolio.

7. Conclusion

ED

In this paper I provide evidence that a strong complementary relation exists between a firm’s high absorptive capacity (AC) and high exposure to technology spillovers (spillover-AC synergy). I first show

PT

that spillover-AC synergy contains value-relevant information regarding a firm’s future performance.

CE

When coupled with high exposure to potential spillovers, a firm’s high absorptive capacity (proxied by R&D intensity) has a strong positive association with an increase in its ability to generate more value-

AC

relevant future innovative outcomes. I also show that firms with spillover-AC synergy not only perform better in their future innovations but also have higher future operating performance and market value. However, information about the spillover-AC synergy effect does not seem to be fully valued in the

stock market ex ante. That is, spillover-AC synergy seems to positively predict the subsequent crosssection of stock returns. Specifically, I show that spillover-AC synergy, rather than merely high exposure 23

I also use other cutoffs (i.e., -2% or -5%) in defining the down-market months. The results remain similar.

32

ACCEPTED MANUSCRIPT

to spillovers, positively predicts subsequent stock returns by using absorptive capacity measure (R&D capital to sales ratio) and technology spillover pool measures estimated as the technology similarityweighed sum of other firms’ R&D. I further show that the effect of spillover-AC synergy is robust to risk

related factors shown to have positive impact on future stock returns.

CR IP T

characteristics correlated with a high product market rivalry effect or to a host of other innovation-

Finally, I show that the market is likely to undervalue the effects of spillover-AC synergy due to limited investor attention. The difficulty in processing and incorporating information about the details of technology spillover pools combined with complementarities to a given firm’s own innovations can, in

AN US

some part, contribute to the undervaluation of strong, positive spillover-AC synergy effects on a focal firm’s increased future innovation-related productivity and firm profitability.

Technology spillovers are an important source of new ideas in a firm’s ongoing sequence of innovations. However, my findings suggest that exposure to large potential technology spillovers is a

M

necessary, but not sufficient, condition for a firm to earn high subsequent stock returns. Firms must maintain sufficient absorptive capacity so that they can better absorb and convert available spillovers

ED

into value-relevant firm performance. Thus, if a firm’s level of absorptive capacity is not properly taken into account, then the estimated relation between technology spillovers and a firm’s future stock returns

PT

can vary, depending on the extent of absorptive capacity.

CE

Although I provide evidence more aligned with the market’s inability to fully value spillover-AC synergy, I do not completely rule out the possibility that the results found in this paper come from

AC

unknown factors correlated with a given firm’s risk characteristics. For example, Lin (2012) theoretically argues that R&D investments are correlated with a firm’s risk not captured by the existing asset pricing models. It would be interesting to further examine whether the strong positive effect of spillover-AC synergy on future stock returns also can be explained, at least in some part, by risk.

33

ACCEPTED MANUSCRIPT

References Allen, T., 1977. Managing the flow of technology. Cambridge, MA: MIT Press. Almeida, H., P. Hsu, and D. Li, 2013. Less is more: Financial constraints and innovative efficiency. Working Paper, University of California, San Diego.

CR IP T

Arrow, K., 1962. Economic welfare and the allocation of resources for invention. The rate and direction of inventive activity: Economic and social factors. Princeton: Princeton University Press, 609–626. Arrow, K., 1969. Classificatory notes on the production and transmission of technological knowledge. American Economic Review 59: 29–35.

AN US

Bena, J., and K. Li, 2014. Corporate innovations and Mergers and Acquisitions. Journal of Finance 69: 1923–1960. Bena, J., and L. Garlappi, 2012. Corporate innovation and returns. Working Paper, University of British Columbia.

M

Bernstein, J., and I. Nadiri, 1989. Research and development and intra-industry spillovers: an empirical application of dynamic duality. Review of Economic Studies 56: 249–267.

ED

Bloom, N., M. Schankerman, and J. Van Reenen, 2013. Identifying technology spillovers and product market rivalry. Econometrica 81: 1347–1393.

PT

Brennan, M.J., N., Jegadeesh, and B., Swaminathan, 1993. Investment analysis and the adjustment of stock prices to common information. Review of Financial Studies 6: 799–824. Carhart, M., 1997. On persistence in mutual fund performance. Journal of Finance 52: 57–82.

CE

Chan, L., J. Lakonishok, and T. Sougiannis, 2001. The stock market valuation of research and development expenditures. Journal of Finance 56: 2431–2456.

AC

Chen, S., Y. Chen, W. Liang, and Y. Wang, 2013. R&D spillover effects and firm performance following R&D increases. Journal of Financial and Quantitative Analysis 48: 1607–1634. Ciftci, M., and W. Cready, 2011. Scale effects of R&D as reflected in earnings and returns. Journal of Accounting and Economics 52: 62–80. Cohen, L., K. Diether, and C. Malloy, 2013. Misvaluing innovation. Review of Financial Studies 26: 635–666.

34

ACCEPTED MANUSCRIPT

Cohen, L., and A., Frazzini, 2008. Economic links and predictable returns. Journal of Finance 63: 1977– 2011. Cohen, W., and D. Levinthal, 1989. Innovation and learning: the two faces of R&D. Economic Journal 99: 569–596.

CR IP T

Cohen, W., and D. Levinthal, 1990. Absorptive capacity: A new perspective on learning and innovation. Administrative Science Quarterly 35: 128–152. Daniel, K., M. Grinblatt, S. Titman, and R. Wermers, 1997. Measuring mutual fund performance with characteristic-based benchmarks. Journal of Finance 52: 1035–1058.

AN US

Dass, N., V. Nanda, and S. Xiao, 2015. Truncation bias in patent data: Does it explain why stockliquidity seemingly reduces innovation? Working Paper, University of Texas at Dallas.

DellaVigna, S., and J. Pollet, 2009. Investor inattention and Friday earnings announcements. Journal of Finance 64: 709–749. Deng, Z., B. Lev, and F. Narin, 1999. Science and technology as predictors of stock performance. Financial Analysts Journal 55: 20–32.

M

Eberhart, A., W. Maxwell, and A. Siddique, 2004. An examination of long-term abnormal stock returns and operating performance following R&D increases. Journal of Finance 59: 623–650.

ED

Evenson, R., and Y. Kislev, 1976. A stochastic model of applied research. Journal of Political Economy 84: 265–282.

PT

Fama, E., and J. MacBeth, 1973. Risk, return, and equilibrium: Empirical tests. Journal of Political Economy 81: 43–66.

CE

Fama, E., and K., French, 2000. Forecasting profitability and earnings. Journal of Business 73: 161–175.

AC

Fama, E., and K. French, 2006. Profitability, investment and average returns. Journal of Financial Economics 82: 491–518. Fama, E., and K. French, 2015. A five-factor asset pricing model. Journal of Financial Economics 116: 1–22. Giglio, S., and K. Shue, 2014. No news is news: Do markets underreact to nothing? Review of Financial Studies 27: 3389–3440. Griliches, Z., 1979. Issues in assessing the contribution of research and development of productivity growth. Bell Journal of Economics 10: 92–116.

35

ACCEPTED MANUSCRIPT

Gu, L., 2016. Product market competition, R&D investment, and stock returns. Journal of Financial Economics 119: 441–455. Hadlock, C., and J. Pierce, 2010. New evidence on measuring financial constraints: Moving beyond the KZ index. Review of Financial Studies 23:1909–1940.

CR IP T

Hall, B.H., A. Jaffe, and M. Trajtenberg, 2001. The NBER patent citation data file: Lessons, insights and methodological tools. NBER Working Paper 8498. Henderson, R., and I. Cockburn, 1996. Scale, scope, and spillovers: The determinants of research productivity in drug discovery. RAND Journal of Economics 27: 32–59.

AN US

Hirshleifer, D., P. Hsu, and D. Li, 2013. Innovative efficiency and stock returns. Journal of Financial Economics 107: 632–654. Hirshleifer, D., P. Hsu, and D. Li, 2014. Don’t hide your light under a bushel: Innovative originality and stock returns. Working Paper, University of California, Irvine.

M

Hirshleifer, D., and D. Jiang, 2010. A financing-based misvaluation factor and the cross-section of expected returns. Review of Financial Studies 23: 3401–3436.

ED

Hirshleifer, D., and S.H. Teoh, 2003. Limited attention, information disclosure, and financial reporting. Journal of Accounting and Economics 36: 337–386.

PT

Hirshleifer, D., S. Lim, and S.H. Teoh, 2009. Driven to distraction: Extraneous events and underreaction to earnings news. Journal of Finance 64: 2289–2325.

CE

Hirshleifer, D., S. Lim, and S.H. Teoh, 2011. Limited investor attention and stock market misreactions to accounting information. Review of Asset Pricing Studies 1: 35–73. Hong, H., and J. C. Stein, 1999. A unified theory of underreaction, momentum trading and overreaction in asset markets. Journal of Finance 54: 2143–2184.

AC

Hong, H., T., Lim, and J.C., Stein, 2000. Bad news travels slowly: Size, analyst coverage, and the profitability of momentum strategies. Journal of Finance 55: 265–295. Hong, H., W., Torous, and R., Valkanov, 2007. Do industries lead stock markets? Journal of Financial Economics 83: 367–396. Hou, K., 2007. Industry information diffusion and the lead-lag effect in stock returns. Review of Financial Studies 20: 1113–1138.

36

ACCEPTED MANUSCRIPT

Hou, K., L. Peng, and W. Xiong, 2009. A tale of two anomalies: The implications of investor attention for price and earnings momentum. Working Paper, Ohio State University. Hou, K., C. Xue, and L. Zhang, 2014. A comparison of new factor models. Working Paper. Ohio State University.

CR IP T

Hou, K., C. Xue, and L. Zhang, 2015. Digesting anomalies: an investment approach. Review of Financial Studies 28: 650–705.

Hsu, P., 2011. Technological innovation spillovers and stock returns. Working Paper, University of Hong Kong.

AN US

Huberman, G., and T. Regev, 2001. Contagious speculation and a cure for cancer: A nonevent that made stock prices soar. Journal of Finance 56: 387–396. Jaffe, A., 1986. Technological opportunity and spillovers of R&D: Evidence from firms' patents, profits, and market value. American Economic Review 76: 984–1001. Jiang, Y., Y. Qian, and T. Yao, 2016. R&D spillover and predictable returns. Review of Finance 20: 1769–1797.

M

Kaplan, S. N., and L. Zingales, 1997. Do financial constraints explain why investment is correlated with cash flow? Quarterly Journal of Economics 112: 169–216.

ED

Klibanoff, P., O. Lamont, and T.A. Wizman, 1998. Investor reaction to salient news in closed-end country funds. Journal of Finance 53: 673–699.

PT

Kumar, P., and D. Li, 2016. Capital investment, innovative capacity, and stock returns. Journal of Finance 71: 2059–2094.

CE

Lev, B., 2004. Sharpening the intangibles edge. Harvard Business Review: 109–118.

AC

Lev, B., and T. Sougiannis, 1996. The capitalization, amortization, and value-relevance of R&D. Journal of Accounting and Economics 21: 107–138. Li, D., 2011. Financial constraints, R&D investment, and stock returns. Review of Financial Studies 24: 2974–3007. Lin, X., 2012. Endogenous technological progress and the cross-section of stock returns. Journal of Financial Economics 103: 411–427. Mansfield, E., 1977. The production and application of new industrial technology. New York: W. W. Norton.

37

ACCEPTED MANUSCRIPT

Mansfield, E., 1988. The speed and cost of industrial innovation in Japan and the United States: external vs. internal technology. Management Science 34: 1157–1168. Mowery, D., 1983. The relationship between intrafirm and contractual forms of industrial research in American manufacturing, 1900-1940. Explorations in Economic History 20: 351 -374.

CR IP T

Nelson, R., 1959. The simple economics of basic scientific research. Journal of Political Economy 67: 297–306. Peng, L., and W. Xiong, 2006. Investor attention, overconfidence and category learning. Journal of Financial Economics 80: 563–602.

AN US

Qiu, J., and C. Wan, 2015. Technology spillovers and corporate cash holdings. Journal of Financial Economics 115: 558–573. Romer, P., 1986. Increasing returns and long-run growth. Journal of Political Economy 94:1002–1037. Schumpeter, J., 1934. The theory of economic development. Cambridge, MA: Harvard University Press.

M

Seru, A., 2010. Firm boundaries matter: Evidence from conglomerates and R&D activity. Journal of Financial Economics 111: 381–405.

ED

Sevilir, M., and X. Tian, 2012. Acquiring Innovation. Working Paper, Indiana University. Spence, M., 1984. Cost reduction, competition, and industry performance. Econometrica 52: 101–122.

PT

Stambaugh, R., and Y. Yuan, 2017. Mispricing factors. Review of Financial Studies, Forthcoming.

CE

Tilton, J., 1971. International diffusion of technology: the case of semiconductors. Washington. DC: Brookings Institution.

AC

Trajtenberg, M., 1990. A penny for your quotes: Patent citation and the value of innovations. Rand Journal of Economics 21: 172–187. Wermers, R., 2004. Is money really 'smart'? new evidence on the relation between mutual fund flows, manager behavior, and performance persistence. Working Paper, University of Maryland. Whited, T., and G. Wu, 2006. Financial constraints risk. Review of Financial Studies 19: 531–59.

38

ACCEPTED MANUSCRIPT

Table 1 Summary statistics This table reports the average characteristics of each portfolio formed based on absorptive capacity (AC) and exposure to technology spillovers (Panel A), as well as sample correlations between key variables

CR IP T

(Panel B). At the end of June in year t, I sort firms into three groups based on the top and bottom 30% of the absorptive capacity (AC) measure of the fiscal year ending in the calendar year t-1. Then I sort firms independently into three groups based on the top and bottom 30% of the technology spillovers (Spill) measure of the fiscal year ending in the calendar year t-1. Finally, I keep these groups for 12 months

AN US

until the end of June in year t+1. I follow this process for every year from 1982 to 2007. Average return is the value-weighted average returns. AC is a firm’s absorptive capacity calculated as R&D capital (RDC) scaled by firm sales at the end of each fiscal year, where RDC is the five-year cumulative research and development (R&D) expenditures assuming an annual depreciation rate of 20% (Chan,

M

Lakonishok, and Sougiannis, 2001). Spill is the potential technology spillover exposure for each firm and is measured as weighted sum of other firms’ R&D efforts (R&D-to-sales ratio). The weights are

ED

calculated as a given firm’s technology closeness with other firms using the Mahalanobis distance. A detailed calculation can be found in Section 2.3. Following previous studies (e.g., Seru, 2010; Bena and

PT

Garlappi, 2012; Almeida, Hsu, and Li, 2013; Hirshleifer, Hsu, and Li, 2013), I calculate Patent as the

CE

number of patents within each technology class divided by the cross-sectional average number of patents applied in the same year and assigned to the same technology class by the USPTO. Cite-weighted patent

AC

is calculated by scaling the number of citations received from the grant year until 2006 for each patent by the average number of citations received by patents applied in the same year and assigned to the same technology class by the USPTO. The book-to-market ratio, B/M, is the book equity in fiscal year ending in calendar year t-1, divided by the market equity at the end of calendar year t. Book equity is calculated as the stockholder’s book equity plus deferred taxes and investment tax credit, minus the book value of preferred stocks. The market value of equity ME is computed as shares outstanding multiplied by the

39

ACCEPTED MANUSCRIPT

stock price. Momentum is the previous 12 month’s stock returns excluding the last month’s return. IVol is idiosyncratic volatility, which is the residual returns from the Fama-French three-factor model. Leverage is market leverage calculated as the book value of debt divided by market value of assets. ROA is measured by the ratio of a firm’s earnings before interest and taxes (EBIT) plus depreciations to

Panel A: Average portfolio characteristics

AChigh

91 0.010 0.019 16.658 1.018 3.458 0.747 266.22 0.165 0.315 0.232 0.476 0.134

76 0.008 0.111 19.746 1.691 7.909 0.662 221.11 0.188 0.481 0.156 0.538 0.173

45 0.007 0.351 21.107 1.231 8.366 0.542 235.49 0.222 0.781 0.072 0.605 0.174

AClow

Spillhigh ACmid

AChigh

55 0.007 0.026 441.552 7.504 26.069 0.660 1,463.64 0.162 0.164 0.249 0.430 0.102

100 0.009 0.119 507.713 14.936 60.124 0.551 1,210.85 0.172 0.246 0.185 0.480 0.132

75 0.011 0.500 804.836 13.901 64.765 0.451 572.29 0.214 0.542 0.105 0.422 0.183

M

AN US

Spilllow ACmid

ED

Average number of firms Average return AC Spill Patent Cite-weighted patent B/M ME ($MM) Momentum IVol Leverage ROA Asset growth

AClow

CR IP T

lagged total book value of assets. Asset growth is the change in total assets divided by lagged total assets.

PT

Panel B: Correlation of key variables

AC Spill

CE

AC

Cite-weighted patent

Spill

Patent

ME

1.000 0.228

1.000

-0.004

-0.015

1.000

Patent

-0.004

0.013

0.841

1.000

ME

-0.010

0.107

0.425

0.512

1.000

B/M

-0.086

-0.119

-0.064

-0.066

-0.122

AC

Cite-weighted patent

B/M

40

1.000

ACCEPTED MANUSCRIPT

Table 2 Annual Fama-MacBeth regressions of future cite-weighted patents on AC and spillovers I perform an annual Fama-Macbeth cross-sectional regression of firms’ future cite-weighted patents on absorptive capacity (AC) and the level of technology spillovers. The independent variables are lagged

CR IP T

one year (t-1) in relation to the dependent variable. The sample consists of firm-year observations from 1982 to 2007. Following previous studies, Cite-weighted patent is calculated by scaling the number of citations received from the grant year until 2006 for each patent by the average number of citations received by patents applied in the same year and assigned to the same technology class by the USPTO

AN US

(e.g., Seru, 2010; Bena and Garlappi, 2012; Almeida, Hsu, and Li, 2013; Hirshleifer, Hsu, and Li, 2013). ln(1+AC) is the natural logarithm of one plus AC. The dummy variable ACmid is equal to one if firms are above the 30th percentile and below the 70th percentile of AC sort in year t-1 and zero otherwise. The dummy variable AChigh is equal to one if firms are above the 70th percentile of AC sort in year t-1 and

M

zero otherwise. ln(1+Spill) is the natural logarithm of one plus Spill. Both AC and Spill are defined in Table 1. Columns (3) and (5) provide the results of running the same model as in Columns (2) and (4),

ED

respectively, but excluding the sample period from 2004 to 2006 to further mitigate the citation truncation bias. Market capitalization (ME), book-to-market ratio (B/M), number of patents (Patent), and

PT

leverage (Leverage) are defined as in Table 1. ln(age) is the natural logarithm of firm age, where age is

CE

measured by years since a given firm’s first observation on COMPUSTAT. Finally, I include industry fixed effects based on the Fama-French 48 industries classifications. t-statistics (based on Newey-West

AC

standard errors) are reported in parentheses. The control variables are winsorized at the top and bottom 1%. *, **, and *** refer to significance at the 10%, 5%, and 1% level, respectively.

41

ACCEPTED MANUSCRIPT

ln(1+AC)

0.441*** (2.89)

ln(1+Cite-weighted number of patents) (2) (3) (4) -0.037 (-0.25)

AChigh ACmid ln(1+Spill)

0.545*** (6.43)

ln(1+AC)* ln(1+Spill)

0.007 (0.61) 0.250** (2.17)

-0.010 (-0.06)

0.009 (0.71) 0.291** (2.15)

No No 17,200 0.128

Yes No 17,200 0.719

Yes Yes 17,200 0.720

Yes No 14,141 0.754

AN US M ED PT

Constant

AC

CE

Industry fixed effects Excluding post-2003 Observations Adjusted R2

Yes Yes 14,141 0.755

0.052 (0.54)

ln(ME)

ln(age)

0.088*** (3.27) 0.070*** (3.68) 0.759*** (74.98) 0.211*** (26.67) 0.009 (0.51) 0.006 (0.98) -0.153*** (-9.87) -0.546* (-1.90)

0.774*** (70.47) 0.211*** (26.79) 0.008 (0.48) 0.001 (0.17) -0.151*** (-10.72) -0.493 (-1.68)

ln(1+Patent)

ln(Leverage)

0.084 (1.35) 0.014 (0.34) -0.027** (-2.09)

0.678*** (13.89) 0.182*** (11.82) 0.011 (0.70) 0.000 (0.08) -0.122*** (-6.68) -0.453* (-1.83)

ACmid*ln(1+Spill)

ln(B/M)

0.050 (0.89) -0.019 (-0.47) -0.026** (-2.32)

0.078*** (3.34) 0.065*** (3.93) 0.665*** (13.95) 0.183*** (11.98) 0.011 (0.72) 0.004 (0.83) -0.123*** (-6.31) -0.489* (-2.02)

AChigh*ln(1+Spill)

42

(5)

CR IP T

(1)

ACCEPTED MANUSCRIPT

Table 3 Annual Fama-MacBeth regressions of subsequent firm performance on AC and spillovers This table reports annual Fama-MacBeth cross-sectional regressions for various subsequent firm performance on absorptive capacity (AC) and spillovers. The independent variables are lagged one year (t-

CR IP T

1) in relation to the dependent variable. I use three measures for the firm performance: (1) return on assets (ROA) is measured by ratio of a firm’s earnings before interest and taxes (EBIT) plus depreciations to lagged total book value of assets, (2) profit margin (PM) is a firm’s a firm’s earnings before interest and taxes (EBIT) plus depreciations divided by sales, and (3) market-to-book equity (MTB) ratio at year t is the

AN US

market value of equity at the end of December of year t divided by book equity for the fiscal year ending in year t. ln(1+AC) is the natural logarithm of one plus AC. The dummy variable ACmid is equal to one if firms are above the 30th percentile and below the 70th percentile of AC sort in year t-1 and zero otherwise. The dummy variable AChigh is equal to one if firms are above the 70th percentile of AC sort in year t-1 and

M

zero otherwise. ln(1+Spill) is the natural logarithm of one plus Spill. Both AC and Spill are defined in Table 1. Following Hirshleifer, Hsu, and Li (2013), Innovative efficiency at year t-1 is defined as the total

ED

number of patents granted in year t-1 scaled by R&D stock in year t-3 ( ). Market capitalization (ME), book-to-market

PT

ratio (B/M), and number of patents (Patent) are defined as in Table 1. Advertising is advertising expense,

CE

and CAPX is capital expenditure. I also include the change in operating performance between year t-1 and year t-2, as well as one-year lagged performances (ROA-1, PM-1, ln(MTB)-1). Independent variables are

AC

winsorized at the top and bottom 1%. t-statistics (based on Newey-West standard errors) are reported in parentheses. *, **, and *** refer to significance at the 10%, 5%, and 1% level, respectively.

43

ACCEPTED MANUSCRIPT

PM (2)

ln(MTB) (3)

-0.026 (-0.89)

-0.087** (-2.37)

-0.078 (-0.82)

AChigh ACmid ln(1+Spill) ln(1+AC)* ln(1+Spill)

-0.001 (-0.67) 0.016* (1.80)

-0.003*** (-2.83) 0.037*** (4.54)

0.001 (0.25) 0.082** (2.28)

AChigh*ln(1+Spill)

ln(1+Patent/ME) ln(1+Advertising/ME) ln(1+CAPX/ME)

ln(B/M)

PT

ROA ROA−1

CE

PM PM−1

0.001 (1.54) 0.022 (0.93) 0.081** (2.20) -0.024 (-1.62) 0.001** (2.24) -0.018*** (-4.44)

0.012*** (3.63) 0.108 (1.03) 0.035 (0.34) -0.213*** (-4.34)

ED

ln(ME)

-0.000 (-0.14) 0.082** (2.61) 0.133*** (3.14) -0.049** (-2.66) -0.000 (-0.48) -0.002 (-0.39) -0.094*** (-6.60) 0.856*** (88.77)

M

ln(1+Innovative efficiency)

AC

0.005 (1.29) -0.001 (-0.56) 0.000 (0.38)

0.015** (2.55) 0.010*** (3.18) 0.001 (1.53)

-0.011 (-0.78) -0.009 (-0.85) 0.003 (1.60)

0.003*** (3.29) 0.001 (0.98) 0.000 (0.04) 0.094** (2.10) 0.155*** (4.73) -0.055*** (-4.35) -0.001*** (-4.53) -0.008*** (-3.46) -0.089*** (-9.32) 0.813*** (48.03)

0.004** (2.14) -0.001 (-0.79) 0.002*** (3.39) -0.013 (-0.62) 0.116*** (3.68) -0.038** (-2.43) 0.001 (1.56) -0.026*** (-5.13)

0.016*** (4.19) 0.005 (1.17) 0.012*** (5.46) 0.125** (2.73) 0.030 (0.33) -0.216*** (-4.86)

-0.082*** (-3.26) 0.829*** (46.60)

0.055*** (5.22)

0.051*** (3.09)

-0.243* (-1.78) 0.759*** (31.85) 0.151** (2.60)

Yes 13,139 0.794

Yes 13,139 0.903

Yes 13,139 0.722

ln(MTB)−1

Industry fixed effects Observations Adjusted R2

ln(MTB) (6)

-0.071*** (-2.86) 0.915*** (89.89)

1/(Book Equity)

Constant

PM (5)

AN US

ACmid*ln(1+Spill)

ROA (4)

CR IP T

ln(1+AC)

ROA (1)

44

0.072*** (10.85)

0.054*** (6.34)

-0.232* (-2.04) 0.760*** (33.38) 0.169*** (4.88)

Yes 13,139 0.792

Yes 13,139 0.900

Yes 13,139 0.721

ACCEPTED MANUSCRIPT

Table 4 Fama-MacBeth cross-sectional regressions of stock returns on AC and spillovers This table reports a monthly Fama-Macbeth cross-sectional regression of firms’ excess stock returns on absorptive capacity (AC) and exposure to spillovers. The dependent variable is the individual firm’s

CR IP T

monthly excess stock return, which is calculated as the firm’s individual return minus the risk-free rate. The independent variables, except for size, momentum, reversal (ret-1), and idiosyncratic volatility (IVol), are from the fiscal year ending in the previous calendar year (t-1). Size, momentum, reversal (ret-1), and idiosyncratic volatility (IVol) are from the previous month. The sample consists of monthly returns from

AN US

July 1982 to June 2007. ln(1+AC) is the natural logarithm of one plus AC. The dummy variable ACmid is equal to one if firms are above the 30th percentile and below the 70th percentile of AC sort in year t-1 and zero otherwise. The dummy variable AChigh is equal to one if firms are above the 70th percentile of AC sort in year t-1 and zero otherwise. ln(1+Spill) is the natural logarithm of one plus Spill. Both AC and Spill are

M

defined in Table 1. Size (ME), book-to-market ratio (B/M), Momentum, IVol, Leverage, ROA, and Asset growth are defined as in Table 1. ret-1 is reversal, which is the one-month lagged return. Industry fixed

ED

effects are included based on the Fama-French 48 industries classification codes. The control variables are winsorized at the top and bottom 1%. t-statistics using Newey-West corrected standard errors are reported

AC

CE

PT

in parentheses. *, **, and *** refer to significance at the 10%, 5%, and 1% level, respectively.

45

ACCEPTED MANUSCRIPT

Dependent variable = ri,t - rf

ln(1+AC)

(1)

(2)

(3)

(4)

(5)

0.726 (1.24)

-1.740 (-1.45)

-1.508 (-1.54)

-1.534 (-1.33)

-1.183 (-1.34)

ACmid ln(1+Spill)

0.075* (1.83)

ln(1+AC)* ln(1+Spill)

-0.034 (-0.71) 1.181*** (3.27)

-0.029 (-0.63) 0.821** (2.49)

ACmid * ln(1+Spill)

ln(B/M) Momentum

CE

ROA

PT

IVol Leverage

Asset Growth Constant

AC

-0.012 (-0.27) 0.860*** (3.58) 0.449** (2.50) -4.002** (-2.51) -0.008 (-0.50) 0.469 (1.39) 1.067*** (5.39) -0.856*** (-3.79) 0.149 (0.33)

-0.000 (-0.01) 0.894*** (3.76) 0.372** (2.17) -4.780*** (-3.19) -0.004 (-0.25) 0.542 (1.61) 0.903*** (5.68) -0.688*** (-4.98) 0.745 (1.26)

No No 226,028 0.084

Yes No 226,024 0.145

Yes No 226,024 0.148

ED

ret−1

-0.014 (-0.26) 0.852*** (3.83) 0.472*** (2.89) -3.928*** (-3.75) -0.009 (-0.69) 0.547 (1.37) 1.037*** (4.65) -0.835*** (-3.23) -0.077 (-0.15)

-0.043** (-2.22) 0.123** (2.03) 0.938*** (4.20) 0.367** (2.38) -4.769*** (-3.72) -0.004 (-0.29) 0.551 (1.43) 0.906*** (5.12) -0.688*** (-4.83) 0.014 (0.02)

M

ln(ME)* ln(1+Spill) ln(ME)

Industry fixed effects Excluding dot-com period Observations Adjusted R2

No No 226,028 0.079

0.202 (1.58) 0.848** (2.08)

AN US

AChigh * ln(1+Spill)

46

-0.641 (-1.50) 0.099 (0.61) 0.081 (0.66)

CR IP T

AChigh

(6)

0.163 (1.39) 0.835** (2.24)

-0.038* (-1.96) 0.132** (1.99) 1.121*** (4.59) 0.330* (1.91) -4.801*** (-2.99) -0.008 (-0.50) 0.903*** (2.69) 1.006*** (6.27) -0.734*** (-5.08) 0.065 (0.10)

0.238** (2.00) 0.008 (0.14) -0.022 (-1.52) 0.054 (0.82) 0.851*** (3.48) 0.375** (2.18) -4.571*** (-3.07) -0.006 (-0.38) 0.546 (1.62) 0.809*** (4.58) -0.681*** (-4.74) 0.285 (0.41)

Yes Yes 195,747 0.140

Yes No 226,024 0.148

ACCEPTED MANUSCRIPT

Table 5 Abnormal returns of portfolios double-sorting on AC and spillovers This table reports the monthly abnormal returns for six portfolios double-sorted on firm’s absorptive capacity (AC) and technology spillovers. At the end of June in year t, I form portfolios by first sorting

CR IP T

firms by their AC of the fiscal year ending in the previous calendar year (t-1) and assigning firms below the 30th percentile as AClow and those above the 70th percentile as AChigh. I then independently sort firms by the technology spillover exposure (Spill) of the fiscal year ending in the calendar year t-1 and assign firms into Spilllow and Spillhigh based on the 30th and 70th percentiles of spillover exposures,

AN US

respectively. Finally, I keep these groups for 12 months until the end of June in year t+1. I follow this process for every year from 1982 to 2007. Both AC and Spill are defined in Table 1. In Panel A, I report value-weighted monthly portfolio returns for the six AC- and Spill-sorted portfolios. In Panel B, I follow the previous studies and exclude stocks with market equity below the 20th NYSE percentile to sort firms

M

based on AC and Spill and then report the six equal-weighted portfolio returns (e.g., Hou, Xue, and Zhang, 2014 and 2015; Gu, 2016). I then run a time-series regression of portfolio excess returns on

ED

factors in the Carhart (1997) four-factor model, the Fama and French (2015) five-factor model, and the Hou, Xue, and Zhang (2015) q-factor model. I also run a time-series regressions of industry-adjusted and

PT

DGTW characteristics-adjusted portfolio returns by employing the Carhart (1997) four-factor model.

CE

The industry-adjusted portfolio returns are calculated using industry-matched (based on Fama and French 48 industry classification) returns for the benchmark instead of the risk-free rate. The DGTW

AC

characteristics-adjusted portfolio returns are calculated using characteristics-matched (based on Daniel, Grinblatt, Titman, and Wermers, 1997) returns for the benchmark instead of the risk-free rate. t-statistics using Newey-West corrected standard errors are reported in parentheses. *, **, and *** refer to significance at the 10%, 5%, and 1% level, respectively.

47

ACCEPTED MANUSCRIPT

Spillhigh ACmid (5)

AChigh (6)

-0.03 (-0.37) 0.50*** (27.18) -0.09*** (-3.47) 0.21*** (5.75) 0.01 (0.50)

0.12 (1.55) 0.64*** (29.57) -0.10*** (-4.18) 0.09* (1.91) -0.01 (-0.25)

0.54*** (4.45) 0.50*** (12.25) -0.16*** (-2.80) -0.31*** (-4.76) 0.06 (1.56)

0.57*** (4.22) 0.00 (0.04) -0.07 (-1.61) -0.52*** (-10.10) 0.05 (1.64)

0.43*** (2.76) -0.08* (-1.93) -0.45*** (-9.23) 0.01 (0.17) 0.14*** (4.15)

-0.04 (-1.12)

0.00 (0.06)

0.30*** (2.89)

0.34*** (3.32)

0.54*** (2.91)

-0.17*** (-3.01)

-0.05 (-0.98)

0.27*** (3.54)

0.44*** (4.39)

0.23** (1.97)

0.09 (0.59)

-0.15* (-1.90)

-0.02 (-0.23)

0.48*** (3.89)

0.63*** (4.57)

0.39** (2.51)

0.07 (0.42)

-0.01 (-0.08)

0.39*** (2.61)

0.50*** (3.19)

0.32* (1.95)

Industry-adjusted portfolio returns

FF-5 factor portfolio returns -0.03 -0.05 (-0.17) (-0.45) q-factor portfolio returns 0.02 (0.15)

-0.11 (-1.30)

ED

0.03 (0.15)

AN US

0.03 -0.03 -0.24 (0.42) (-0.41) (-1.54) DGTW characteristics-adjusted portfolio returns 0.10 0.07 0.04 (0.70) (0.91) (0.34)

Spread portfolio returns (6) - (4) (6) - (3) (7) (8)

CR IP T

AClow (4)

M

Panel A: Value-weighted portfolio returns Spilllow AClow ACmid AChigh (1) (2) (3) Carhart-4 factor portfolio returns 0.11 0.14 0.11 (0.66) (1.29) (0.82) 1.12*** 0.70*** 0.58*** (22.62) (24.54) (14.36) 0.10 0.28*** 0.29*** (0.88) (8.23) (7.10) 0.22*** -0.08 -0.32*** (2.73) (-1.63) (-4.99) -0.10* -0.14*** -0.09** (-1.70) (-4.13) (-2.19)

0.88*** (4.11)

0.74*** (3.17)

0.61*** (2.71)

Industry-adjusted portfolio returns 0.06 0.01 (0.67) (0.13)

CE

PT

Panel B: Equal-weighted portfolio returns (excluding micro stocks) Spilllow Spillhigh AClow ACmid AChigh AClow ACmid (1) (2) (3) (4) (5) Carhart-4 factor portfolio returns 0.09 0.23** 0.27 0.14 0.16 (0.75) (1.97) (1.17) (1.07) (1.46)

Spread portfolio returns (6) - (4) (6) - (3) (7) (8)

0.18 (1.43)

0.11 (1.33)

0.59*** (3.32)

0.41** (2.00)

0.53** (2.24)

DGTW characteristics-adjusted portfolio returns 0.06 0.21** 0.27 (0.67) (2.02) (1.32)

0.06 (0.48)

0.11 (1.25)

0.70*** (4.63)

0.65*** (3.32)

0.43** (2.06)

FF-5 factor portfolio returns -0.18* 0.06 (-1.65) (0.48)

0.16 (0.62)

-0.15 (-1.11)

-0.06 (-0.63)

0.88*** (4.36)

1.04*** (4.84)

0.72*** (3.15)

q-factor portfolio returns -0.03 0.25* (-0.16) (1.70)

0.37 (1.40)

0.01 (0.05)

0.02 (0.14)

0.94*** (3.90)

0.93*** (3.47)

0.57** (2.35)

AC

0.05 (0.24)

AChigh (6)

48

ACCEPTED MANUSCRIPT

Table 6 Abnormal returns of portfolios controlling for mispricing factors This table reports the monthly abnormal returns for six portfolios double-sorted on firm’s absorptive capacity (AC) and technology spillovers. The abnormal returns are obtained after controlling for the well-

CR IP T

known mispricing factors. The six portfolios are formed in a same way as in Table 5. In Panel A, I report value-weighted monthly portfolio returns for the six AC- and Spill-sorted portfolios. In Panel B, I follow the previous studies and exclude stocks with market equity below the 20th NYSE percentile to sort firms based on AC and Spill and then report the six equal-weighted portfolio returns (e.g., Hou, Xue, and Zhang,

AN US

2014 and 2015; Gu, 2016). In each panel, I first run a time-series regression of portfolio excess returns on factors in the Carhart (1997) four-factor model but adding one more factor, the mispricing factor UMO from Hirshleifer and Jiang (2010). Then, I run a time-series regression of portfolio excess returns on mispricing factors MGMT and PERF from Stambaugh and Yuan (2017) in addition to the market (MKT)

M

and the size (SMB) factors. t-statistics using Newey-West corrected standard errors are reported in

AC

CE

PT

ED

parentheses. *, **, and *** refer to significance at the 10%, 5%, and 1% level, respectively.

49

ACCEPTED MANUSCRIPT

Panel A: Value-weighted portfolio returns AClow (1)

Spilllow ACmid (2)

AChigh (3)

AClow (4)

Spillhigh ACmid (5)

AChigh (6)

Spread portfolio returns (6) - (4) (6) - (3) (7) (8)

0.04 (0.31) 0.52*** (12.59) 0.27*** (7.03) -0.29*** (-4.64) -0.08** (-2.18) 0.01 (0.16)

-0.09 (-1.09) 0.52*** (25.80) -0.08*** (-3.24) 0.16*** (4.09) -0.02 (-0.75) 0.09*** (2.68)

0.02 (0.26) 0.68*** (26.70) -0.10*** (-3.89) 0.01 (0.13) -0.05* (-1.75) 0.15*** (3.37)

0.47*** (3.52) 0.53*** (11.77) -0.15*** (-2.76) -0.37*** (-5.61) 0.03 (0.70) 0.11* (1.77)

0.56*** (3.90) 0.01 (0.19) -0.07 (-1.58) -0.53*** (-8.40) 0.04 (1.21) 0.02 (0.33)

0.42*** (2.74) 0.01 (0.23) -0.42*** (-9.17) -0.08 (-1.17) 0.11*** (2.70) 0.10 (1.43)

-0.16* (-1.88) 0.52*** (25.00) -0.05 (-1.62) 0.27*** (7.26) 0.02 (0.61)

0.01 (0.08) 0.67*** (27.02) -0.07** (-2.47) 0.14*** (3.41) 0.04 (1.48)

0.40** (2.58) 0.57*** (11.40) -0.10 (-1.43) -0.11** (-2.05) 0.16*** (4.52)

0.55*** (3.11) 0.04 (0.77) -0.05 (-0.44) -0.38*** (-5.54) 0.14*** (3.64)

0.36** (2.40) 0.06 (1.10) -0.37*** (-3.55) 0.24*** (2.95) 0.10*** (2.65)

0.07 (0.53) 0.72*** (21.55) 0.28*** (8.43) -0.14*** (-2.66) -0.17*** (-4.53) 0.11* (1.84)

AN US

0.11 (0.56) 1.12*** (22.20) 0.10 (0.88) 0.21** (2.26) -0.10* (-1.75) 0.01 (0.12)

0.04 (0.25) 0.50*** (10.30) 0.26*** (5.42) -0.35*** (-5.83) 0.06* (1.82)

ED

0.01 (0.12) 0.70*** (19.64) 0.32*** (7.54) -0.11** (-2.06) -0.02 (-0.65)

M

Stambaugh mispricing factor portfolio return 0.25 (1.19) 1.02*** (20.92) 0.11 (0.79) -0.02 (-0.22) -0.09 (-1.39)

CR IP T

UMO-augmented portfolio return

Spilllow ACmid (2)

CE

AClow (1)

PT

Panel B: Equal-weighted portfolio returns (excluding micro stocks) AChigh (3)

AClow (4)

Spillhigh ACmid (5)

AChigh (6)

Spread portfolio returns (6) - (4) (6) - (3) (7) (8)

0.38 (1.56)

0.04 (0.29)

0.06 (0.55)

1.04*** (4.58)

1.00*** (4.15)

0.66*** (2.78)

-0.04 (-0.26)

0.04 (0.29)

0.79*** (2.81)

0.83*** (2.86)

0.50** (2.11)

-0.02 (-0.16)

0.16 (1.23)

AC

UMO-augmented portfolio return

0.13 (0.93)

Stambaugh mispricing factor portfolio return -0.05 (-0.31)

0.29 (0.91)

50

ACCEPTED MANUSCRIPT

Table 7 Testing limited attention This table reports monthly Fama-Macbeth cross-sectional regressions of firms’ excess stock returns on absorptive capacity (AC) and exposure to spillovers within subsamples divided into high and low investor

CR IP T

attention group. I use analyst coverage, institutional ownership, and firm size to proxy for the investor attention. Analyst coverage subsamples are obtained by dividing the whole sample into high and low analyst coverage stocks based on the median analyst coverage at the end of year t-1. Analyst Coverage is calculated by averaging the monthly number of analysts providing current fiscal year earnings forecasts

AN US

over the previous year. Institutional ownership subsamples are obtained by dividing the whole sample into high and low institutional ownership stocks based on the median institutional ownership at the end of year t-1. Inst. Ownership denotes the fraction of firm shares outstanding owned by institutional investors in year t-1. Size subsamples are obtained by splitting the whole sample into small and large size stocks based on

M

the median NYSE breakpoints at the end of June in year t. The dependent variable is the individual firm’s monthly excess stock return, which is calculated as the firm’s stock return minus the risk-free rate. The

ED

independent variables, except for size, momentum, reversal (ret-1), and idiosyncratic volatility (IVol), are from the fiscal year ending in the previous calendar year (t-1). Size, momentum, reversal (ret-1), and

PT

idiosyncratic volatility (IVol) are from the previous month. The sample consists of monthly returns from

CE

July 1982 to June 2007. All the independent variables are defined as in Tables 1 and 4. Industry fixed effects are included based on the Fama-French 48 industries classification codes. The control variables are

AC

winsorized at the top and bottom 1%. t-statistics using Newey-West corrected standard errors are reported in parentheses. *, **, and *** refer to significance at the 10%, 5%, and 1% level, respectively.

51

ACCEPTED MANUSCRIPT

ln(1+AC)* ln(1+Spill) ln(ME) ln(B/M) Momentum ret−1 IVol Leverage

ED

ROA

PT

Asset Growth Constant

CE

χ2-stat on Difference

Industry fixed effects Observations Adjusted R2

AC

-0.965 (-0.95) -0.039 (-0.86) 0.712 (1.57) 0.021 (0.40) 0.746** (2.38) 0.443*** (2.61) -5.632*** (-6.11) 0.008 (0.75) 0.421 (1.01) 0.849*** (3.88) -0.354* (-1.95) 0.583 (0.99)

-2.269** (-2.08) -0.057 (-0.87) 1.156** (2.33) 0.061 (1.35) 1.581*** (5.41) 0.378** (2.16) -3.508** (-2.06) -0.019 (-1.08) -0.008 (-0.02) 0.861*** (3.43) -0.631*** (-2.61) -0.500 (-0.60)

1.933

Yes 58,616 0.257

-0.057 (-0.06) -0.057 (-1.01) 0.524 (1.41) 0.003 (0.05) 0.089 (0.32) 0.310 (1.62) -4.736*** (-3.16) -0.004 (-0.26) 1.014*** (2.62) 0.722*** (3.89) -0.594*** (-3.46) 0.747 (1.12)

AN US

ln(1+Spill)

-2.115 (-1.36) -0.106 (-0.98) 1.663** (2.03) 0.136** (2.17) 0.961*** (3.09) 0.252 (1.40) -0.685 (-0.34) -0.048** (-2.24) 0.253 (0.58) 1.027*** (3.26) -1.453*** (-4.28) -0.349 (-0.38)

M

ln(1+AC)

Inst. Ownership Low High (3) (4)

3.664

Yes 167,408 0.186

Yes 91,920 0.210

52

Firm Size Small Big (5) (6) -2.279** (-2.58) -0.074 (-1.15) 1.092** (2.19) 0.148*** (2.65) 1.045*** (4.08) 0.343** (2.14) -4.364*** (-3.08) -0.011 (-0.73) 0.576 (1.54) 1.227*** (6.00) -0.843*** (-4.33) 0.044 (0.06)

0.033 (0.03) -0.026 (-0.39) 0.668 (1.34) -0.079 (-1.14) 0.594* (1.81) 0.094 (0.50) -3.711** (-2.25) -0.005 (-0.31) 0.371 (1.00) 0.372* (1.77) -0.405** (-2.05) 1.024* (1.81)

CR IP T

Analyst Coverage Low High (1) (2)

5.247 Yes 134,104 0.197

Yes 144,733 0.154

Yes 81,291 0.311

ACCEPTED MANUSCRIPT

Table 8 Robustness tests on the validity of RDC-to-sales ratio as a proxy for AC This table reports the results of rerunning same analyses in previous tables but additionally controlling for the interaction effects of spillovers with the other innovation-related abilities documented in the previous

CR IP T

studies. Following Cohen, Diether, and Malloy (2013), Ability is calculated as the average of the coefficients from the regressions of sales growth on the past five-year R&D-to-sales ratio. Following Hirshleifer, Hsu, and Li (2013), Innovative efficiency at year t-1 is defined as the total number of patents granted in year t-1 scaled by R&D stock (five-year cumulative R&D expenditures assuming an annual

AN US

depreciation rate of 20%) in year t-3. Following Kumar and Li (2016), innovative capacity (IC) is measured by R&D-active firms’ asset growth (AG) in the fiscal year ending in calendar year t-1, where R&D-active firms have non-missing R&D-to-sales ratio. HHI is the Herfindahl-Hirschman Index, which is

M

defined as the sum of squared market shares of each firm in the industry. I repeat the previous analyses controlling for the above interaction terms. Column (1) is a result on annual Fama-Macbeth cross-sectional

ED

regression of firms’ future cite-weighted patents (CWP) on absorptive capacity (AC) and the level of technology spillovers as in Table 2. Columns (2) through (4) are results on annual Fama-MacBeth cross-

PT

sectional regressions for various subsequent firm performance on absorptive capacity (AC) and spillovers as in Table 3. Finally, Column (5) reports a monthly Fama-Macbeth cross-sectional regression of firms’

CE

excess stock returns on absorptive capacity (AC) and the level of technology spillovers as in Table 4. All the models include standard controls used in Tables 2, 3, and 4. All the dependent and independent

AC

variables are defined in Tables 2 through 4. The control variables are winsorized at the top and bottom 1%. t-statistics using Newey-West corrected standard errors are reported in parentheses. *, **, and *** refer to significance at the 10%, 5%, and 1% level, respectively.

53

ACCEPTED MANUSCRIPT

HHI* ln(1+Spill) Ability* ln(1+Spill) ln(1+Innovative efficiency)* ln(1+Spill)

Firm controls Industry fixed effects

(1)

(2)

(3)

(4)

(5)

0.383

-0.031

-0.044

-0.061

-1.779

(1.26)

(-1.22)

(-1.14)

(-0.77)

(-1.41)

-0.051

-0.003*

-0.002

0.009

0.091

(-1.22)

(-1.90)

(-0.93)

(1.18)

(0.61)

0.299**

0.019***

0.026**

0.081**

1.111**

(2.28)

(2.80)

(2.67)

(2.07)

(2.38)

-0.091

0.003

0.000

-0.021

0.197

(-1.48)

(0.79)

(0.00)

(-1.30)

(1.09)

0.001

-0.000

0.000

-0.000

0.002

(0.96)

(-1.04)

(0.67)

(-0.41)

(1.01)

0.055***

-0.000

-0.001

-0.002

-0.019

(4.60)

(-0.39)

(-1.18)

(-1.26)

(-0.87)

-0.041

0.003

-0.001

-0.021

-0.119

(-0.64)

(0.71)

(-0.17)

(-1.35)

(-0.67)

Yes

Yes

Yes

Yes

Yes

Yes

Yes

Yes

Yes

Yes

10,598

8,439

8,439

8,439

136,998

0.756

0.835

0.933

0.782

0.200

AC

CE

Adjusted R

PT

Observations 2

ri - rf

ED

Innovative capacity* ln(1+Spill)

ln(MTB)

CR IP T

ln(1+AC)* ln(1+Spill)

PM

AN US

ln(1+Spill)

ROA

M

ln(1+AC)

ln(1+CWP)

54

ACCEPTED MANUSCRIPT

Table 9 Robustness This table reports a monthly Fama-Macbeth cross-sectional regression of firms’ excess stock returns on absorptive capacity (AC), the level of technology spillovers, R&D competition, product market competition,

CR IP T

financial constraints, and R&D ability, along with standard control variables as in Table 4. The dependent variable is the individual firm’s monthly excess stock return, which is calculated as the firm’s individual return minus the risk-free rate. The independent variables, except for size, momentum, reversal (ret-1), and idiosyncratic volatility (IVol), are from the fiscal year ending in the previous calendar year (t-1). Size,

AN US

momentum, reversal (ret-1), and idiosyncratic volatility (IVol) are from the previous month. The sample consists of monthly returns from July 1982 to June 2007. ln(1+Spillsic) is the natural logarithm of one plus Spillsic, where Spillsic is the weighted sum of R&D efforts (R&D-to-sales ratio) of firms that compete in the same product market. The weights are calculated as the similarity of a given firm’s position in the

M

product market space with other firms’ positions by using the Mahalanobis distance. HHI is the Herfindahl-Hirschman Index, which is defined as the sum of squared market shares of each firm in the

ED

industry. Financial constraint is proxied by the KZ index (KZ) of Kaplan and Zingales (1997), the WW index (WW) of Whited and Wu (2006), and the SA index (SA) of Hadlock and Pierce (2010). Following

PT

Hirshleifer, Hsu, and Li (2013), Innovative efficiency at year t-1 is defined as the total number of patents

CE

granted in year t-1 scaled by R&D stock (five-year cumulative R&D expenditures assuming an annual depreciation rate of 20%) in year t-3. Following Cohen, Diether, and Malloy (2013), I estimate Ability as

AC

the average of the coefficients from the regressions of sales growth on the past five-year R&D-to-sales ratio. All other independent variables are defined as in Tables 1 and 4. The control variables are winsorized at the top and bottom 1%. t-statistics using Newey-West corrected standard errors are reported in parentheses. *, **, and *** refer to significance at the 10%, 5%, and 1% level, respectively.

55

ACCEPTED MANUSCRIPT

Dependent variable = ri,t - rf

ln(1+Spill) ln(1+AC)* ln(1+Spill) ln(ME)* ln(1+Spill) ln(1+Spillsic)

(2)

(3)

(4)

(5)

(6)

-1.599 (-1.43) 0.204 (1.65) 0.858** (2.14) -0.044** (-2.31) 0.004 (0.11)

-1.739 (-1.56) 0.151 (1.19) 0.893** (2.24) -0.038* (-1.95) 0.005 (0.11) 1.743 (1.30) 0.007 (0.26)

-0.590 (-0.56) 0.163 (1.31) 0.642* (1.78) -0.039* (-1.86) 0.002 (0.05) 1.332 (1.19) 0.009 (0.42) 0.124 (0.49) -3.917* (-1.73) -0.001 (-0.14) 0.077 (1.24)

-0.630 (-0.49) 0.209 (1.34) 0.884** (1.99) -0.047* (-1.87) 0.017 (0.49) 0.491 (0.35) 0.002 (0.07) 0.545* (1.72) -5.473* (-1.83) 0.005 (0.51) 0.076 (0.74)

-3.021 (-1.48) 0.204 (1.23) 0.938** (2.20) -0.046* (-1.74) 0.010 (0.29) 0.524 (0.37) -0.001 (-0.02) 0.700** (2.22) -6.975** (-2.25)

-1.597 (-1.16) 0.251 (1.53) 0.904** (2.07) -0.054** (-2.11) 0.018 (0.51) 0.642 (0.46) -0.003 (-0.11) 0.560* (1.78) -6.004* (-1.93)

AN US

ln(1+Patent/ME) ln(1+Innovative efficiency) HHI ln(1+AC)* HHI

M

KZ

ED

ln(1+AC)* KZ SA

Ability

CE

ln(1+AC)* WW

0.029 (0.22) -0.762 (-1.27)

PT

ln(1+AC)* SA WW

ln(1+AC)* Ability

AC

CR IP T

ln(1+AC)

(1)

Additional standard controls Industry fixed effects Observations Adjusted R2

Yes Yes 223,397 0.150

Yes Yes 212,148 0.155

56

Yes Yes 200,538 0.165

0.001 (0.49) 0.073 (1.52)

0.001 (0.39) 0.080* (1.65)

-1.637 (-1.01) -4.722 (-1.08) 0.001 (0.48) 0.081* (1.68)

Yes Yes 126,909 0.214

Yes Yes 126,909 0.214

Yes Yes 126,909 0.215

ACCEPTED MANUSCRIPT

Appendices

Appendix A: Estimation of Spillsic

CR IP T

As in the construction of Spill, I depart from Bloom, Schankerman, and Van Reenen (2013) and use only past information about a firm’s sales share. Specifically, I use COMPUSTAT Segment data to construct a vector of a firm’s sales shares in each business segment Si,t = (si1,t, si2,t … siK,t), where sik,t is firm i’s proportion of sales within the three-digit SIC industry k over the period to time t. 24 Similar to constructing the Spill measure, I define the (P, N) matrix at time t as ̃

AN US

20F 20F

similar to ̃ in Section II.B such that each element of ̃ between firms.

25 21F 21F

is the uncentered correlation

I then obtain the product market closeness among firms by using the

Mahalanobis distance measure. It is similar to TECHt in Equation (1): ̃

M

̃

( )

analogous to

ED

̃ ̃ captures the R&D spillovers across different product market industries

where

in TECHt. The (i, j) element in the matrix

measures the correlation

PT

weights of the overlap in sales shares between firm i and j by how close their different sales

CE

shares are to one another. I finally measure firm i’s spillover pool containing the R&Ds of rivals within similar product markets as ( )

AC



where

is firm j’s R&D-to-sales ratio in year t.

24

I also construct a firm’s sales shares in each business segment based on industry classification using fourdigit SIC codes. Using either a three-digit industry or a four-digit industry classification does not change the results. 25 ⁄( ⁄( For example ̃ [ ⁄( ) ) ) ].

57

ACCEPTED MANUSCRIPT

Appendix B: Additional Robustness Tests Table B1. Fama-MacBeth cross-sectional regressions of stock returns using alternative AC and spillovers

(1) -2.782** (-2.51) 0.142 (0.86) 1.793*** (2.74) -0.042 (-1.56)

ln(1+SpillAlt) ln(1+ ACAlt)* ln(1+ SpillAlt) ln(ME)* ln(1+ SpillAlt)

M

ln(1+Spillsic)

ln(1+ ACAlt)* HHI

CE

KZ

PT

HHI

ED

ln(1+Patent/ME) ln(1+Innovative efficiency)

ln(1+ ACAlt)* KZ

-1.833 (-1.19) -0.016 (-0.08) 1.408* (1.83) -0.018 (-0.55) 0.047 (0.83) 0.578 (0.41) 0.012 (0.49) 0.393 (1.35) -2.389 (-1.01) 0.017* (1.77) -0.039 (-0.38) 0.003 (0.85) 0.058 (1.29)

Yes Yes 200,538 0.163

Yes Yes 125,032 0.211

ln(1+ ACAlt)* Ability

Additional standard controls Industry fixed effects Observations Adjusted R2

Yes Yes 200,538 0.149

58

(3)

-2.456** (-2.07) 0.155 (0.93) 1.592** (2.43) -0.041 (-1.53) 0.003 (0.05) 0.728 (0.70) 0.024 (1.16) 0.037 (0.15) -1.060 (-0.55) 0.008 (1.09) -0.003 (-0.04)

AC

Ability

Dependent variable = ri,t - rf (2)

AN US

ln(1+ACAlt)

CR IP T

This table reports monthly Fama-Macbeth cross-sectional regressions of firms’ excess stock returns on the absorptive capacity and the exposure to spillovers as in Table 7, but reports results using an alternative AC measure and spillover measure. ln(1+ACAlt) is the natural logarithm of one plus RDC-to-ME ratio, where RDC is the five-year cumulative research and development (R&D) expenditures assuming an annual depreciation rate of 20% (Chan, Lakonishok, and Sougiannis, 2001), and ME is the market value of equity. ln(1+SpillAlt) is the natural logarithm of one plus SpillAlt, where SpillAlt is measured as weighted sum of other firms’ R&D-to-ME ratio. The weights are calculated as a given firm’s technology closeness with other firms by using the Mahalanobis distance. All the other independent variables are defined as in Table 7. The industry fixed effects are included based on the Fama-French 48 industries classification codes. The control variables are winsorized at the top and bottom 1%. t-statistics using Newey-West corrected standard errors are reported in parentheses. *, **, and *** refer to significance at the 10%, 5%, and 1% level.