Do technology spillovers affect the corporate information environment?

Do technology spillovers affect the corporate information environment?

Journal Pre-proof Do technology spillovers affect the corporate information environment? Phuong-Anh Nguyen, Ambrus Kecskés PII: S0929-1199(20)30025-...

664KB Sizes 0 Downloads 45 Views

Journal Pre-proof Do technology spillovers affect the corporate information environment?

Phuong-Anh Nguyen, Ambrus Kecskés PII:

S0929-1199(20)30025-0

DOI:

https://doi.org/10.1016/j.jcorpfin.2020.101581

Reference:

CORFIN 101581

To appear in:

Journal of Corporate Finance

Received date:

26 May 2018

Revised date:

2 December 2019

Accepted date:

23 January 2020

Please cite this article as: P.-A. Nguyen and A. Kecskés, Do technology spillovers affect the corporate information environment?, Journal of Corporate Finance(2020), https://doi.org/10.1016/j.jcorpfin.2020.101581

This is a PDF file of an article that has undergone enhancements after acceptance, such as the addition of a cover page and metadata, and formatting for readability, but it is not yet the definitive version of record. This version will undergo additional copyediting, typesetting and review before it is published in its final form, but we are providing this version to give early visibility of the article. 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.

© 2020 Published by Elsevier.

Journal Pre-proof

Do Technology Spillovers Affect the Corporate Information Environment? PHUONG-ANH NGUYEN and AMBRUS KECSKÉS1

Nguyen is at the School of Administrative Studies, York University; and Kecskés is at the Schulich School of Business, York University.

re

-p

ro

of

Abstract Technology spillovers across firms affect corporate innovation, productivity, and value, according to prior research, so information about technology spillovers should matter to investors. We argue that technology spillovers increase the complexity and uncertainty of value relevant information about the firm, which makes information processing more costly, discourages it, and thereby increases information asymmetry between insiders and outsiders. We find that not only does information asymmetry increase, but so does avoidance by sophisticated market participants, uncertainty, and insider trading. We also find that investors do not misestimate short-term earnings, but they underestimate long-term earnings, consistent with the higher future stock returns that we also find. JEL classification: G12, G14, G23, G24, G41, M40, O31, O32, O33, O34 Keywords: Innovation; Technology spillovers; Research and development; Information asymmetry; Valuation; Earnings; Mispricing

lP

"'They just had no idea what they had,' [Steve] Jobs later said, after launching hugely profitable Apple computers using concepts developed by Xerox." (The Wall Street Journal (2012))

na

1. Introduction

It is well known that firms do not innovate and grow solely through their own efforts but

ur

also by building on the innovations of their technological peer firms (e.g., Arrow (1962), Jaffe

Jo

(1986), Romer (1990), and Grossman and Helpman (1991)). This is because the broader social value of an invention often significantly exceeds its narrow private value to its inventor. Furthermore, it is also well established that the innovation activities of technological peer firms

1

This paper is based on Nguyen's doctoral dissertation at Virginia Tech defended on Sept. 16, 2015 and promptly published online. We greatly appreciate the comments of two anonymous referees, Viet Anh Dang, François Derrien, John Easterwood, Vidhan Goyal, Johan Hombert, Yi Jiang, Dasol Kim, Jeong-Bon Kim, Bart Lambrecht (the Editor), Leonardo Madureira, Sattar Mansi, Ron Masulis, Adrien Matray, Peter Oh, Dino Palazzo, Dimitris Petmezas, Alexander Philipov, Jiaping Qiu, Stephen Teng Sun, Amit Seru, Sarah Qian Wang, Toni Whited, Jin Xu, Zilong Zhang, and seminar participants at the 2018 Canadian Academic Accounting Association Conference, the 2017 European Financial Management Symposium, the 2017 Financial Management Association Conference, the 2017 Financial Management Association Asia/Pacific Conference, the 2017 Financial Management Association European Conference, the 2017 Northern Finance Association Conference, Manchester Business School, Surrey Business School, and Virginia Tech. This research was supported by the Social Sciences and Hu manities Research Council of Canada.

Journal Pre-proof can spill over to a given firm and increase its own innovation, productivity, and value (Bloom, Schankerman, and Van Reenen (2013)). Peer innovation activities have also been shown to affect corporate innovation strategies, technology transfers, tangible asset sales, and mergers and acquisitions.2 Since the innovation activities of a firm's technological peer firms capture potential technology spillovers to the firm's own innovation activities, they should be taken into account in the valuation of the firm itself.

of

However, the valuation of innovative firms is made more difficult by information

ro

asymmetry between insiders and outsiders, which is exacerbated in the case of innovation

-p

activities (Hall (1992a, 2002)). Taking as given that the innovation activities of a firm's

re

technological peer firms influence the firm's own innovation and growth, we study how information asymmetry between the firm's insiders and outsiders is affected by these potential

lP

technology spillovers.3 We refer hereafter to these potential technology spillovers, as captured by

na

the innovation activities of technological peer firms, without the "potential" descriptor, for the sake of brevity and without changing our meaning.

ur

The direction of the effect of technology spillovers on information asymmetry is not

Jo

predicted clearly by theory. On the one hand, technology spillovers can decrease information collection costs for investors by increasing the availability of information relevant to the valuation of a given firm (e.g., as in Foucault and Fresard (2014)). In the presence of technology spillovers, information about any one technological peer firm is relevant to the valuation of all the others. Outsiders thus have additional information, available from the firm's technological peer firms, about the firm itself. An increase in the availability of information can decrease

2

See, respectively, Akcigit and Kerr (2018), Akcigit, Celik, and Greenwood (2016), Maksimovic and Phillips (2001), and Phillips and Zhdanov (2013) and Bena and Li (2014). 3 This insiders-outsiders dichotomy refers to more informed managers and inside investors as compared to less informed outside investors (and related market participants such as sell-side analysts).

Journal Pre-proof information collection costs. Consequently, this increases the motivation of outsiders to produce information and can thus reduce their information disadvantage compared to insiders.4 On the other hand, technology spillovers can also increase information processing costs for outsiders by increasing the complexity and uncertainty of information relevant to the valuation of a given firm.5 In the presence of technology spillovers, a firm should be valued not just using its own information but also information about its technological peer firms, which

of

makes valuation more complex and uncertain. As an illustration (others provided later), consider

ro

how the Android operation system might be valued by outsiders of Samsung, one of its largest

-p

users. These valuation exercises are highly complex and uncertain even today, but they were far more difficult a decade previously, when Android was introduced and its eventual success was

re

far from clear.6 Returning to our reasoning, an increase in the complexity and uncertainty of

lP

information can increase information processing costs. Consequently, this decreases the

na

motivation of outsiders to produce information and can thus increase their information disadvantage compared to insiders.7

ur

In summary, technology spillovers can increase the availability of value relevant

Jo

information (reducing information asymmetry), and they can also increase the complexity and uncertainty of such information (increasing information asymmetry). In theory, then, the direction of the effect of technology spillovers on information asymmetry depends on the

4

Litov, Moreton, and Zenger (2012) and Goldstein and Yang (2015) present similar arguments but without reference to technology spillovers specifically. 5 We refer to uncertainty, as distinct from risk, in the Knightian sense: unknown risk as opposed to known risk. 6 To distinguish between complexity and uncertainty, consider an analyst valuing a firm. In the presence of technology spillovers, the analyst must value the firm as a function not just of its own information but also of the information of another firm. Complexity is higher because the analyst must process information about both firms, and uncertainty is higher because the analyst does not know the weights to put on the information about each firm. 7 There is a large body of work on investor aversion to comp lexity and uncertainty. For complexity, see Cohen and Frazzini (2008), Brunnermeier and Oehmke (2009), You and Zhang (2009), and Carlin, Kogan, and Lowery (2013). For uncertainty, see Dow and Werlang (1992), Chen and Epstein (2002), Cao, Wang, and Zhang (2005), and Bossaerts, Ghirardato, Guarnaschelli, and Zame (2010).

Journal Pre-proof tradeoff between the costs of information production (i.e., collection and processing) and its benefits (e.g., detecting mispricing, generating trading profits, etc.). In practice, however, the higher costs of information production appear to outweigh its benefits, ultimately indicating greater information asymmetry. There are several reasons for which the costs of information production may exceed its benefits. First, for an innovative firm as well as its technological peer firms, detailed information

of

about their innovation activities may not be readily available in accessible formats such as

ro

financial statements. This makes such information difficult to use in valuation (e.g., Lev and

-p

Sougiannis (1996)). Moreover, although information about technological peer firms is value

re

relevant from a theoretical perspective, it is rare to see such information used in practice, even by the most sophisticated financial market participants, suggesting that the costs of such information

lP

outweigh its benefits.8 By contrast, information about product market peer firms (and industry

na

information more generally) is widely used in valuation, by retail investors, institutional investors, research analysts, and investment bankers among many others.9 The use of

ur

information about technological peer firms is rare because it requires a deep and evolving

Jo

understanding of the relationships between many technologies and industries, which is oftentimes not the expertise of most financial market participants. Overall, the complexity and uncertainty of information about technology spillovers make it costly to use in valuing a firm. Furthermore, sophisticated financial market participants, such as institutional investors and sell-side analysts, may not be able to solve the information asymmetry problem. Indeed, market participants that are more sophisticated are more likely to recognize that the net benefits

8

For instance, see Asquith, Mikhail, and Au (2005). Additionally, we have been unable to find a substantive treatment of the valuation implications of technology spillovers, whether in analy st reports, the investment policy manuals of major institutional investors, or standard pedagogical resources. 9 For evidence on analysts, for example, see Kadan, Madureira, Wang, and Zach (2012).

Journal Pre-proof of producing information are lower for firms with greater technology spillovers. To the extent that such market participants consequently avoid these firms, they may exacerbate the information asymmetry problem.10 As previously mentioned, real world practice indicates that, even for sophisticated market participants, valuing firms with greater technology spillovers involves high information processing costs.11 Costs aside, these firms may have higher systematic innovation risk because the innovation activities of technological peer firms are

of

interdependent. Sophisticated market participants are likely to avoid such firms because their

ro

information processing costs outweigh the benefits of identifying mispricing, or their innovation

-p

risk outweighs the arbitrage opportunities they present (Healy and Palepu (2001) and Armstrong,

re

Core, Taylor, and Verrecchia (2011)). 12

Finally, corporate disclosure of innovation activities is also unlikely to completely solve

lP

the information asymmetry problem. First, even full disclosure may not eliminate the inherent

na

complexity and uncertainty of information about technology spillovers for the firm's investors. Second, any disclosure can be exploited by all outsiders, including the firm's product market

ur

competitors. The benefits of disclosure to investors may therefore be outweighed by the

Jo

concomitant proprietary costs of disclosure, especially for innovative firms (Bhattacharya and Ritter (1983), Verrecchia (1983), and Thakor and Lo (2019)). On the whole, technology spillovers are likely to increase information asymmetry between insiders and outsiders.

10

In a world with a continuum of sophistication across econ omic agents (i.e., beyond the simple informeduninformed dichotomy), some agents will have higher benefits than costs from collecting and processing information about technology spillovers. However, this need not be the case for the marginal investor or ev en the average institutional investor or sell-side analyst. 11 For example, in the context of analysts valuing complex firms, Plumlee (2003) and Cohen and Lou (2012) provide evidence that higher information processing costs result in greater valuation errors. 12 There is a large volume of evidence pointing in the direction of sophisticated market participants avoiding firms with higher information production costs (e.g., Lang and Lundholm (1996), Francis, Hanna, and Philbrick (1997), and Healy, Hutton, and Palepu (1999)). However, some evidence does point in the opposite direction (e.g., Barth, Kasznik, and McNichols (2001)).

Journal Pre-proof In our empirical analysis, we study the effect of technology spillovers on the corporate information environment using a sample of roughly 700 different innovative publicly traded firms during the 1981-2001 period. Following Bloom, Schankerman, and Van Reenen (2013) ("BSV" hereafter), we capture potential technology spillovers to a firm by accounting for the extent of its technological similarity with other firms as well as the R&D of other firms. Specifically, we calculate our measure of technology spillovers to a firm as the sum of the

of

weighted R&D stocks of other firms, where the weights are the technological proximities of two

ro

firms. We measure the technological proximity of two firms as the distance between the

-p

technology activities of the two firms in the same technology space or similar technology spaces.

re

We capture technology activities and spaces by patents and patent classes, respectively. Our identification of technology spillovers to a given firm relies on the projected R&D of

lP

other firms based on their R&D tax credits, as in BSV. We identify the effect of technology

na

spillovers on the corporate information environment using exogenous variation in federal and state R&D tax credits. For each firm-year, we project R&D stock on R&D tax credits, we

ur

calculate technology spillovers using the projected R&D stock, and we use this projected

Jo

measure in our main regressions. For added rigor, we always account for product market spillovers in our main regressions. We also control for all variation attributable to the firm's own R&D stock and its own R&D tax credits. In addition, we include both firm fixed effects and industry-year fixed effects in our regressions. Consequently, we identify entirely off the timeseries variation in technology spillovers within firms, after eliminating the variation common to firms within a given industry in a given year. In our first set of results, we examine the effect of technology spillovers on information asymmetry. We use five standard proxies for this purpose: the bid-ask spread, the Amihud

Journal Pre-proof illiquidity measure, the returns ratio, the magnitude of earnings announcement surprises, and the volatility of the market reaction to earnings announcements. All five proxies capture information asymmetry between insiders and outsiders. However, the first three proxies may be better at capturing the tension between uninformed outside investors and informed inside investors. The last two proxies may instead be better at capturing the tension between outside investors and

increase in information asymmetry across all of our proxies.

of

informed managers. Our results show that technology spillovers consistently lead to a significant

ro

Next, we examine the contribution of sophisticated market participants to the effect of

-p

technology spillovers on information asymmetry. As a result of their higher information

re

processing costs and innovation risk, firms with greater technology spillovers are likely to be avoided by institutional investors. Indeed, our results show that technology spillovers decrease

lP

both the number and the stake of institutional investors. Our finding of a lower fraction of

na

institutional ownership, or, equivalently, a higher fraction of retail ownership, is consistent with retail investors making less informed investment decisions than institutional investors. They are

ur

less likely to understand or even recognize the technology relationships between firms and the

Jo

resulting implications for valuation and risk, so they may be more willing to invest in firms with greater technology spillovers than institutional investors.13 Additional results show that among institutional investors, the institutions that are likely to be particularly sophisticated (e.g., active investors) have an even stronger tendency to avoid firms with greater technology spillovers. Moreover, sell-side analysts are similarly likely to avoid firms with greater technology spillovers because covering them requires analysts to exert more effort while generating less

13

The relative risk avoidance of institutional investors may be driven by prudent man regulations, particularly during our sample period (Del Guercio (1996)).

Journal Pre-proof reliable projections (e.g., earnings estimates).14 Indeed, our results show that technology spillovers decrease analyst coverage, and they also show that among analysts, those that are likely to be particularly sophisticated (e.g., those with more experience) tend even more strongly to avoid firms with greater technology spillovers. We also find that technology spillovers increase the dispersion of analysts' earnings estimates, which is consistent with technology spillovers increasing uncertainty about value relevant earnings information and hence about

of

valuation itself (Diether, Malloy, and Scherbina (2002)). In summary, sophisticated market

ro

participants appear to avoid firms with greater technology spillovers, thereby exacerbating

-p

information asymmetry.

re

Additionally, we examine managerial activities for evidence of whether managers use financial reporting to reduce information asymmetry, and for whether managers' stock trades

lP

reflect their information advantage over investors. We do not find evidence of technology

na

spillovers affecting financial reporting quality (specifically, discretionary accruals), although, as we later explain in more detail, this outcome variable is not simple to interpret. However, we do

ur

find evidence of an increase in insider trading for firms with greater technology spillovers. We

Jo

also find an increase in insider net purchases, which is consistent with managers believing that their firm is undervalued by investors. These additional results directly compare outsiders to insiders. As a result, they are distinct from the first three of our information asymmetry results, which use standard market microstructure measures that compare informed investors and uninformed investors. Overall, our results here suggest that managers do trade on their information advantage, but our evidence does not suggest that they use financial reporting to reduce information asymmetry.

14

See Gilson, Healy, Noe, and Palepu (2001), Duru and Reeb (2002), and Frankel, Kothari, and Weber (2006).

Journal Pre-proof Having established that technology spillovers lead to greater information asymmetry, we explore their implications for earnings expectations. Greater information asymmetry should lead to less precise earnings estimates, which is in fact what we find, but any bias in earnings estimates is not predicted clearly by theory. One possibility is that investors as a group are less certain about future earnings, but they are still correct on average. However, it is also possible that investors make systematic mistakes given the complexity and uncertainty of information

of

about technology spillovers. There are numerous behavioral theories that would support such a

ro

bias.15 Moreover, the literature does find that investors tend to underestimate the earnings of

-p

firms with high intangible assets and also to undervalue them.16 Consequently, it is likely that

re

investors underestimate the earnings of firms with greater technology spillovers. We therefore examine earnings surprises in both the short run and the long run. It is

lP

important to consider the long run because innovation activities can take many years to produce

na

results. We find that over a one year horizon, technology spillovers do not affect earnings, whether realized or expected. Over a five year horizon, however, technology spillovers

ur

significantly increase realized earnings, and these realizations are significantly higher than

Jo

expectations. We also examine whether future stock returns are consistent with the pattern of earnings expectations that we document. We find that investors do incorporate into stock prices the increase in earnings resulting from technology spillovers, but the process takes several years.

15

These include limited investor attention (Hong and Stein (1999) and Hirshleifer and Teoh (2003)), ambiguity aversion (Bossaerts, Ghirardato, Guarnaschelli, and Zame (2010)), and various cognitive limitations (see the survey of Barberis and Thaler (2003)). 16 For evidence on R&D spending, see Lev and Sougiannis (1996), Chan, Lakonishok, and Sougiannis (2001), and Eberhart, Maxwell, and Siddique (2004). For patent outputs, see Cohen, Diether, and Malloy (2013) and Hirshleifer, Hsu, and Li (2013).

Journal Pre-proof Overall, our findings for earnings and stock returns are consistent with our earlier findings for information asymmetry.17 In summary, we find that technology spillovers lead to greater information asymmetry. Furthermore, sophisticated market participants avoid firms with greater technology spillovers, thereby exacerbating information asymmetry. Finally, investors appear to undervalue firms with greater technology spillovers, as evidenced by their underestimation of long-term earnings and

of

by the higher future stock returns of these firms.

ro

Our study is the earliest to provide broad empirical evidence that technology spillovers

-p

significantly affect the corporate information environment. The existing literature documents a

re

number of positive consequences of technology spillovers, for firm value (Jaffe (1986)), productivity and innovation (BSV), and financial policies (Nguyen and Kecskés (2019)). We

lP

find that technology spillovers have more nuanced consequences. They increase information

na

asymmetry, which, all else equal, raises the cost of capital and can have various distortionary effects on corporate behavior.18 At the same time, our earnings and stock returns results indicate

ur

that the impact of technology spillovers is subtly different in the long run versus the short run.

Jo

There is a dynamic aspect to certain effects of technology spillovers. Our findings are complemented by additional recent studies on the effect of technology spillovers on stock returns. One study finds that technological links between firms generate a lead-lag relationship in stock returns (Lee, Sun, Wang, and Zhang (2019)), consistent with technology information being hard to process. Another study finds a lower incidence of stock price crashes (Kim, Sun, and Zhang (2018)), which is not only consistent with our finding that 17

Note that our results do not imply that uninformed outside investors earn arbitrage profits. These investors may also bear additional risk not captured by the standard empirical asset pricing models (e.g., information risk). 18 For instance, investment can be affected (Derrien and Kecskés (2013)) as can governance (Chen, Harford, and Lin (2015)), investment efficiency (Chen, Xie, and Zhang (2017)), and tax avoidance (Chen and Lin (2017)), to name just a few possibilities.

Journal Pre-proof technology spillovers increase information asymmetry, on average, but which further suggests they may also decrease its right tail, reshaping its distribution more broadly. Our study also improves our understanding of the information environment of innovative firms (Hall (1992a) and Himmelberg and Petersen (1994)). We distinguish between the innovation activities of the firm itself, which is the focus of the literature, and those of its technological peer firms. We also show that technology relationships between firms contain

of

substantial value relevant information, above and beyond the industry relationships considered in

ro

the literature (e.g., Bhojraj and Lee (2002)).

-p

Finally, we contribute to the literature on the determinants of the corporate information

re

environment. The previous literature focuses on firms in general and examines factors over which firms have some control (e.g., Frankel and Li (2004) and Brown and Hillegeist (2007)).

lP

We show that technology spillovers are a new and important factor over which firms have more

na

limited influence.

The rest of this paper is organized as follows. Section 2 presents the methodology and

ur

identification, while Section 3 presents the sample and data. Section 4 presents the results for

Jo

information asymmetry, while Section 5 presents the implications for earnings expectations. Section 6 concludes.

Journal Pre-proof 2. Methodology and Identification19 2.1. Measuring Technology Spillovers 2.1.1. General Procedure The motivation behind our technology spillover measures is the insight that a firm benefits more from the R&D of other firms the closer it is to them in terms of technology. To be specific, the extent of technology spillovers from firm j to firm i depends on the technological

of

proximity of firm i and firm j as well as the R&D stock of firm j. Aggregating across all other

ro

firms, technology spillovers to firm i equal the sum of technology spillovers from all other firms

-p

j to firm i.

re

There are three general steps involved in calculating technology spillovers. The first step is the calculation of the technological proximity of two firms. There are two measures of

lP

technological proximity used in the literature: the Jaffe measure (Jaffe (1986)) and the

na

Mahalanobis measure (BSV). With the Jaffe measure, technology spillovers are restricted to the same technology space, whereas with the Mahalanobis measure, technology spillovers are

ur

allowed across different technology spaces. The second step is the calculation of the R&D stocks

all other firms.

Jo

of all other firms. The final step is the calculation of technology spillovers to a given firm from

2.1.2. Jaffe Measure of Technology Spillovers To construct this measure of technology spillovers, the Jaffe measure of the technological proximity of two firms is first constructed. Specifically, each of the patents of a given firm is allocated by the USPTO to one or more technology class. Altogether, there are 426 possible 19

The methodology and identification as well as the data and sample of the present paper are closely related to that of BSV. The present paper also has an empirical framework in common with Nguyen and Kecskés (2019), but the latter paper focuses on the financial policy consequences of technology spillovers rather than the information environment. The present paper is written to be self contained and readable without referen ce to lengthy passages from other papers.

Journal Pre-proof technology

classes.

A

firm's

technology

activity

is

then

characterized

by

a

vector

Ti=(Ti1 ,Ti2 ,…,Ti426 ), where Tiτ is the average share of the patents of firm i in technology class τ over the period 1970-1999.20 The Jaffe proximity of firm i and firm j is then defined as the uncentered correlation of the technology activities of the two firms: 1/ 2 1/ 2 TECH ijJaffe  TiT j TiTi T jT j 

The Jaffe proximity measure takes on values from zero to one. Higher values of the measure

of

mean that the technologies of the two firms are closer to each other.

ro

Second, the R&D stocks of all other firms are calculated. The formula used to calculate a

-p

firm's R&D stock is G t = Rt + (1–δ)G t–1 , where Rt is the firm's R&D expenditures in year t and δ

re

is the depreciation rate. As is widely done in the literature, δ is set to 0.15.

lP

Finally, the Jaffe measure of technology spillovers to firm i in year t equals the sum of technology spillovers from all other firms j to firm i in year t:

na

TECHSPILLitJaffe   j i TECH ijJaffeG jt

ur

2.1.3. Mahalanobis Measure of Technology Spillovers

Jo

The Mahalanobis measure of technology spillovers is somewhat more complicated to construct than the Jaffe measure. The reason for this is that the measure of the technological proximity of two firms uses a measure of the proximity of technology spaces as an input. The literature captures the proximity of technology classes using the observed colocation of the technology classes within firms. The rationale for this approach is that technology classes that

20

In calculating the proximity measure, one can either use all available data or only the data within a rolling window. The former approach benefits from greater precision, while the latter approach benefits from g reater timeliness. Both approaches yield similar proximity measures. Since the data on patents allocated to 426 technology classes is understandably sparse for most firms in any given year, it is common in the literature to use all available data. We follow this approach as well.

Journal Pre-proof tend to colocate within firms are the result of related technologies. Therefore, these colocated technology classes reflect technology spillovers across technology classes. In calculating the proximity of technology classes, the allocation of a technology class is determined by the vector Ωτ =(T1τ ,T2τ ,…,TNτ ), where N is the number of firms and Tiτ is the average share of patents of firm i in technology class τ over the period 1970-1999. The proximity of the two technology classes, τ and ζ, is the uncentered correlation (like for the Jaffe proximity

  1/ 2   1/ 2

ro

   

of

measure) of the allocation vectors Ω τ and Ωζ:

-p

A 426×426 matrix Ω is then constructed such that its (τ,ζ)th element equals Ωτζ. This matrix

re

captures the proximity of technology classes.

lP

The measure of the technological proximity of firm i and firm j depends on the technology activities of the two firms (as captured by the vectors Ti and Tj in the Jaffe measure)

na

and the proximity of technology classes. Technological proximity is defined as:







ur

1/ 2 1/ 2 TECH ijMahal  Ti TiTi  T j T jT j

Jo

This measure of the technological proximity of two firms weights the overlap in technology activities between the two firms by the proximity of their technology classes. (In the special case of Ω=I, the implication is that Ωτζ=0 for all τ≠ζ. This means that technology spillovers can only occur within the same technology class. The Mahalanobis technological proximity measure is therefore identical to the Jaffe technological proximity measure in this case.) This finishes the Mahalanobis measure of the technological proximity of two firms. Next, the R&D stocks of all other firms are calculated just like for the Jaffe measure of technology spillovers. Finally, the Mahalanobis measure of technology spillovers to firm i in year t is the sum of technology spillovers from all other firms j to firm i in year t:

Journal Pre-proof TECHSPILLMahal   j i TECH ijMahalG jt it

2.2. Measuring Product Market Spillovers It is possible that the effect of technology spillovers on a firm is contaminated by the effect of product market spillovers because the technological peer firms of a given firm can adopt similar technologies and also produce competing products. Therefore, the R&D activities of other firms have two separate and opposing spillover effects on the valuation of the firm itself:

of

the positive, productivity increasing effect of technology spillovers, and the negative, market

ro

share decreasing effect of product market spillovers. Since we wish to isolate the effect of

-p

technology spillovers, we control for product market spillovers.

re

Our measures of product market spillovers are motivated by the insight that a firm's

lP

market shares in its various product markets are negatively affected by the R&D activities of other firms with which it competes. Like for technology spillovers, the extent of product market

na

spillovers from firm j to firm i depends on the product market proximity of firm i and firm j as

ur

well as the R&D stock of firm j. Aggregating across all other firms, product market spillovers to firm i equal the sum of product market spillovers from all other firms j to firm i.

calculations

are

Jo

For both the Jaffe and Mahalanobis measures of product market spillovers, the analogous to

the

corresponding technology spillover measures.

Briefly

described, the Jaffe measure of product market proximity is constructed as follows. First, the sales of a given firm are allocated to one or more industry segments based on data from Compustat. Altogether, there are 597 industries covered by the sample. Next, a firm's product market activity is characterized by a vector Si=(Si1 ,Si2,…,Si597 ), where Sik is the average share of the sales of firm i in industry k over the period 1993-2001. (The period is shortened because the industry data are not available over the full sample period.) Finally, the Jaffe distance, the R&D

Journal Pre-proof stocks of all other firms, and the product market spillover measure are all calculated as previously described. 2.3. Brief Discussion of Spillovers The literature contains many good examples that illustrate the concepts underlying technology spillovers (see Nguyen and Kecskés (2019) but also Rosenberg (1979) and Bloom, Schankerman, and Van Reenen (2013)). However, it is worth noting here that the manner in

of

which technologies diffuse throughout the economy shows that information asymmetry is an

ro

inherent feature of technology spillovers. By their very nature, relationships between firms,

-p

especially along technological lines, are more easily understood by insiders than outsiders. These

re

relationships can be explicit (as in the case of Apple and Xerox in consumer and business electronics, for example) or implicit (for instance, in the case of the many software developers

lP

that benefit from technological advances in operating systems and vice versa). Similarly, it can

na

take a long time for technology spillovers to be widely recognized by outsiders of the affected firms. This is readily illustrated by the decades it took for lasers (invented in 1960) and

ur

microprocessors (invented in 1971) to spread from their original military aviation and ultra high

Jo

end computer applications, respectively, to their ubiquitous presence in personal and business products. All in all, information asymmetry tends to go hand in hand with the technology diffusion process. 2.4. Identification Strategy To identify the causal effects of technology spillovers on the corporate information environment, we use variation in federal and state R&D tax credits. A large body of accumulated evidence suggests that changes in R&D tax credits are appropriate for identification in our setting because they do affect corporate policies, they are plausibly exogenous to corporate

Journal Pre-proof policies, and they vary across firms. In greater detail, first, a substantial literature shows that R&D tax credits generate large increases in R&D spending. This is the case both in the U.S. and internationally (Hall (1992b), Berger (1993), Hines (1993), and Bloom, Griffith, and Van Reenen (2002)). It is therefore well established that R&D tax credits affect investment. Second, the literature also demonstrates the exogeneity of these tax policies to corporate policies. For example, BSV provide convincing evidence that changes in economic or political

of

conditions cannot explain changes in R&D tax policies. The same conclusion is arrived at by

ro

other studies that perform similar analyses (Cummins, Hassett, and Hubbard (1994), Chirinko

-p

and Wilson (2017), Moretti and Wilson (2017), and Hombert and Matray (2018)). Indeed, the

re

impact of R&D tax credits on government finances is relatively modest, which limits concerns about potential anticipation effects. Instead, there is a general pattern of R&D tax credits

lP

gradually increasing across states and over time. Still, there is considerable variation in R&D tax

Finally,

na

credits across states and over time, and likewise at the federal level. R&D tax credits vary substantially across firms. The reason for this

ur

heterogeneity at the federal level is that effective federal tax credits are determined by the

Jo

difference between the actual R&D expenditures of the firm and a base amount that varies across firms and time according to the applicable federal tax rules. Furthermore, the amount that a firm can claim depends on the extent to which credits exceed profits as well as other factors that include deduction rules and the corporate tax rate. At the state level, the reason for heterogeneity in tax credits is that state tax credits are determined by the location of the firm's R&D hubs, and these can vary across states for a given firm along with its state R&D tax credits. In Section 2.1, we construct spillover measures to which we refer as "raw" to distinguish them from the "purged" spillover measures to which we now turn. The purged measures are

Journal Pre-proof constructed below to remove the variation in R&D investment that is endogenous to corporate policies and to retain the variation that is exogenous. BSV provide a detailed description, but we summarize it here. Federal and state R&D tax credits are calculated at the firm-year level using the Hall-Jorgenson user cost of capital approach (Hall and Jorgenson (1967)). For firms that operate in more than one state in a given year, tax credits are aggregated to the firm-year level. This is done by summing the weighted state-level tax credits for the firm-year in question, where

of

the weights are the average shares of the firm's inventors located in a given state.

ro

Next, using a firm-year panel, R&D expenditures are regressed on federal tax credits,

-p

state tax credits, and firm and year fixed effects. Predicted R&D expenditures are then calculated

re

using this regression. The remaining calculations are the same as in Section 2.1. The exogenous R&D stock for each firm-year is calculated using predicted R&D expenditures. Finally, the

lP

purged spillover measures are calculated like the raw spillover measures but the exogenous R&D

na

stocks of other firms are used instead of their raw R&D stocks. Further details are provided by BSV in their Appendix B.3 as well as Wilson (2009) and Falato and Sim (2014). It is worth

ur

observing that our identification of technology spillovers to a given firm relies on the projected

Jo

R&D of other firms based on their R&D tax credits and not on the firm's own R&D tax credits.21 2.5. Main Regression Specifications In all of our empirical analyses, we use four regression specifications for all of our outcomes of interest. In the first two specifications, we capture spillovers with the raw and purged Jaffe measures. We do so for both technology and product market spaces. In the last two specifications, we capture spillovers with the raw and purged Mahalanobis measures. We use 21

Our identification strategy follows the literature and focuses on R&D stocks because there is plausibly exogenous variation available from R&D tax credits. By contrast, technology proximity is much more difficult to identify. Nevertheless, the technology proximity between two firms is time-invariant by construction, and our rigorous regression specifications, which notably include firm fixed effects, should mitigate concerns about firms choosing the technology spaces in which they operate.

Journal Pre-proof both the Jaffe and Mahalanobis measures because each has its advantages. The advantage of the Jaffe measure is that it has been extensively used in the literature since its popularization by Jaffe (1986), but it restricts technology spillovers to the same technology space. The advantage of the Mahalanobis measure is that it is a more recent contribution to the literature (BSV), but it allows technology spillovers across technology spaces rather than only within the same space. There are several features common to all of our regression specifications. Specifically,

of

we include technology spillovers, which is our variable of interest, and product market

ro

spillovers, which is our control variable for the product market spillovers of other firms' R&D.

-p

Additionally, we control for the firm's own R&D. In our specifications that use purged spillover

re

measures, we also control for the firm's own federal and state tax credits. The rigorousness of our specifications means that the effect of technology spillovers is identified off the variation in the

lP

R&D of the firm's technological peer firms that is orthogonal to the R&D of the firm's product

na

market competitors. This identification is further refined, in the case of the purged measures, to the projected R&D of peer firms based on their R&D tax credits, and specifically the component

ur

that is orthogonal to the firm's own R&D tax credits. Consequently, our results cannot be

Jo

explained by variation in R&D tax credits that is common to a firm and its technological peer firms, nor can they be explained variation in R&D tax credits that is common to the firm and its product market competitors. We also control for firm age to capture life cycle effects that may be associated with technology and product market spillovers. By doing so, we can rule out such possibilities as firms with greater technology spillovers being more mature and therefore having less information asymmetry. We also include additional control variables that are standard in the

Journal Pre-proof literature for the outcome of interest. These are indicated in the analyses to which they correspond. The independent variables are lagged. All variables are defined in Appendix Table 1. Additionally, we include firm fixed effects and industry-year fixed effects. Consequently, we identify entirely off the time-series variation of technology spillovers within firms across time, and within an industry and a given year across firms. Finally, we cluster standard errors by

3. Sample and Data

ro

3.1. Sample Construction

of

industry- year. We generally multiply the dependent variables by 100 for expositional simplicity.

-p

We construct our sample as follows. We begin with all publicly traded U.S. firms in

re

CRSP and Compustat. We keep U.S. operating firms defined as firms with CRSP share codes of 10 or 11. We drop firms that are financials or utilities. Next, keep only those firms for which we

lP

have data on technology and product market spillovers. This means that our sample is restricted

na

to firms that were issued at least one patent since 1963. Even so, our sample firms account for a substantial proportion of the R&D expenditures in the U.S. For example, in 1995, our sample

ur

firms accounted for 62% of U.S. R&D spending (BSV). Our final sample comprises 12,118 firm-

Jo

year observations corresponding to 694 unique firms between 1981 and 2001. We end our sample in 2001 because of data limitations. First and most importantly, the NBER patent database is already poorly populated by the mid-2000s, and it ends completely in 2006. Patents are included in the database based on grant dates rather than filing dates. The NBER patent database becomes sparse by the mid-2000s because numerous patents filed in the early 2000s were not granted by 2006. Since we use patent data to calculate technological proximity and hence technology spillovers, we end our sample in 2001 to ensure that we use accurate patent data in our calculations. Second, we require data for up to five years into the

Journal Pre-proof future in some of our analyses, so this requirement also limits our ability to extend our sample period. However, we do have a large sample of innovative firms that spans over two decades. 3.2. Data Sources We obtain data on raw and purged technology and product market spillover measures from Nick Bloom's website (see BSV). We obtain patent data from the USPTO patent assignment database and from Noah Stoffman's website (see Kogan, Papanikolaou, Seru, and

of

Stoffman (2017)). Our stock trading data are from CRSP, our accounting data are from

ro

Compustat, our institutional ownership data are from Thomson's 13f filings, and our data on

-p

analysts are from I/B/E/S (the data begin in 1982). We winsorize all continuous variables at the

re

1st and 99th percentiles. The definitions of all variables are provided in Appendix Table 1. 3.3. Descriptive Statistics

lP

[Insert Table 1 about here]

na

In Table 1, we present descriptive statistics for our sample, starting with technology spillovers. Since technology spillovers to the typical firm are large in dollar value and right

ur

skewed, we use them in logarithmic form throughout the paper. However, it is more natural to

Jo

interpret them here in level form. For the raw Jaffe measure, the average firm has technology spillovers worth about $25 billion (median of $20 billion), with a standard deviation of roughly $20 billion. Comparing the magnitudes of our measures, the purged Jaffe measure is similar, and the raw and purged Mahalanobis measures are about five times larger. The Jaffe measures are smaller than the Mahalanobis measures because the former are defined over a more restricted technology space than the latter. Turning to general firm characteristics, our sample firms are innovative by construction, so naturally they spend heavily on R&D and they produce a large number of patents. Our firms

Journal Pre-proof are large, with mean and median total assets of $2.5 billion and $338 million, respectively. They are also mature, with mean and median age of 25 and 20 years, respectively. Since much of the innovation in the economy is carried out by mature public firms (Baumol (2002)), the composition of our sample firms is very much as expected. Overall, while our sample firms are larger, older, and more innovative than the typical publicly traded firm, they are generally comparable in terms of our outcomes of interest during

of

our sample period. While our first three measures of information asymmetry are higher, on

ro

average, than in the years after our sample period, the difference is the result of decimalization,

-p

which took place right at the end of our sample period. Our last two measures of information

re

asymmetry are very much in line with the literature (e.g., Kelly and Ljungqvist (2012)). Additionally, institutional ownership of our sample firms is 109 investors on average

lP

(median of 54) with a stake of 41% on average (median of 43%). Among institutional investors,

na

70% are active (median 73%) rather than passive. Analyst coverage is 11 analysts on average (median of 7), and the typical analyst has over 4 year of experience covering firms (mean and

ur

median). The dispersion of analysts' earnings estimates is also consistent with the literature (e.g.,

Jo

Diether, Malloy, and Scherbina (2002)). As for managerial activities, the magnitude of accruals is 6% relative to total assets (median of 4%). There are 3 insider trades on average per thousand shareholders (median of 1), and there are $7 of insider trades per million dollars of market capitalization (median of $1). Insiders are sellers, on average, and their net sales are 20% of their total trading (median 17%) based on the number of trades and 33% (median of 70%) based on the value of trades. Turning to earnings, the short-term realized earnings of our firms are somewhat lower than expected, but it is their long-term earnings growth rate realizations that fall significantly

Journal Pre-proof short of expectations, by 6 percentage points (mean and median). However, such a large shortfall, even more so for earnings in the long run than in the short run, is well documented, and it is generally attributed to analysts being overly optimistic about the growth prospects of firms that they cover (e.g., Chan, Karceski, and Lakonishok (2003)). At the same time, our firms tend to have high abnormal stock returns, with a mean of 7% (median of 6%). It is also well documented that firms with high intangible assets, typically captured by R&D spending and

of

patent outputs, tend to have high future returns because they are initially undervalued by

ro

investors (e.g., Lev and Sougiannis (1996)).

-p

[Insert Table 2 about here]

re

In Table 2, we present descriptive statistics by industry. Specifically, we group firms by their primary industries, and then we sort industries by technology spillovers. We then compute

lP

descriptive statistics for each industry. Spillovers are in logarithmic form as usual. Clustered at

na

the top of the table (high technology spillovers) are industries that tend to be innovative (e.g., communications), whereas at the bottom of the table (low technology spillovers) are industries

ur

that do not tend to be innovative (e.g., food). Technology spillovers are positively correlated with

Jo

both R&D and product market spillovers, which makes it important to control for both. However, there is also considerable residual variation in technology spillovers as well as significant variation within industries in technology spillovers compared to their variation across industries. Since the analysis in the table is at the industry level, this leaves plenty of room at the firm level for independent variation in technology spillovers and industries. 4. Results for Information Asymmetry Our main argument is that the effect of technology spillovers is, on balance, to increase the complexity and uncertainty of value relevant information, which increases information

Journal Pre-proof processing costs. This, in turn, decreases the motivation of outsiders to reduce their information disadvantage relative to insiders, which leads to greater information asymmetry. In this section, we first examine the effect of technology spillovers on information asymmetry between insiders and outsiders. We then examine the contribution of sophisticated market participants, such as institutional investors and sell-side analysts, to the effect of technology spillovers on information asymmetry. Finally, we examine managerial activities for evidence of whether managers use

of

financial reporting to reduce information asymmetry, and for whether managers' stock trades

ro

reflect their information advantage over investors.

-p

4.1. Information Asymmetry

re

We predict that technology spillovers should increase information asymmetry. To test this prediction, we use five standard proxies for information asymmetry from the literature. These

lP

proxies are: the bid-ask spread; the Amihud illiquidity measure; the returns ratio (the proportion

na

of trading days with zero or missing stock returns); the magnitude of earnings announcement surprises; and the volatility of the market reaction to earnings announcements. In all our

ur

empirical analyses of information asymmetry, our regression specifications follow the empirical

Jo

literature on the subject (e.g., Ferreira and Laux (2007) and Hong and Kacperczyk (2009)). Furthermore, in addition to the features common to all of our regression specifications (Section 2.5), we control for market capitalization, market-to-book, cash flow, stock returns, and stock return volatility. Before proceeding to our results, we should note that the most relevant of our control variables are product market spillovers and the firm's own R&D. Accordingly, we always report results for these two variables. Nevertheless, we minimize the interpretation of these results because these two variables are not the focus of our study.

Journal Pre-proof [Insert Table 3 about here] Table 3 presents the results. Information asymmetry increases significantly as a result of technology spillovers.

Our coefficient estimates are all positive, they are economically

significant, and they are generally also statistically significant. Spillovers are in logarithmic form, and the dependent variables are multiplied by 100. To provide some examples, for a 10% increase in technology spillovers, the bid-ask spread (Panel A) increases by around 0.1

of

percentage point (the coefficient estimate on technology spillovers of roughly 1 (e.g., 0.98 in

ro

Column 1) multiplied by 10%), or by about 3% of its unconditional standard deviation of 3%

-p

(Table 1). The comparison is similar for earnings announcement surprises (Panel D) (increase of

re

around 0.1 p.p., standard deviation of 3%). For earnings announcement volatility (Panel E), its 1.5 p.p. or so increase is about 5% of its unconditional standard deviation of 31%. In the four

lP

instances (out of 20) in which the coefficient estimates are not statistically significant, they still

na

have the correct sign, and their magnitudes are roughly comparable to those of the statistically significant coefficients in the same panel.22

ur

Unlike technology spillovers, product market spillovers do not reliably affect information

Jo

asymmetry. The firm's own R&D is likewise not reliably related to information asymmetry. The lack of significance of our results for R&D is an artifact of our rigorous regression specifications (see Section 2.5), but it is also consistent with technology spillovers playing a bigger role in the firm's information environment than the firm's own R&D.23

22

To check whether our inferences are affected by using predicted regressors, we compute bootstrapped standard errors. We draw 1,000 random samples with replacement and run our regressions on each sample. We then use the standard errors that correspond to the empirical distribution of the estimated coefficients. We find that the results are similar (not tabulated). 23 As an alternative to using product market spillovers constructed using SIC codes and sales weights, we use the Hoberg-Phillips product similarity measure (Hoberg and Phillips (2010) and Hoberg and Phillips (2016)). Our construction of product market spillovers is the same as before except that we us e as weights the pairwise similarity scores between two firms before multiplying by R&D stock and aggregating across firms. While the sample size does shrink due to data availability, our inferences remain unchanged.

Journal Pre-proof 4.2. Sophisticated Market Participants We also

examine whether sophisticated market participants, including institutional

investors and sell-side analysts, avoid firms with greater technology spillovers. As a result of their information processing costs and innovation risk, firms with greater technology spillovers are likely to be avoided by these market participants. We therefore predict that these firms should have a lower number of institutional investors, lower institutional ownership or, equivalently,

of

higher retail ownership, and less analyst coverage.

ro

[Insert Table 4 about here]

-p

We first test our predictions for the number and stake of institutional investors. Table 4

re

presents the results. Both the number and stake of institutional investors decrease significantly as a result of technology spillovers. For a 10% increase in technology spillovers, the number of

lP

institutional investors (Panel A) decreases by about 2%-4% relative to its mean. Similarly, the

na

stake of institutional investors (Panel B) also decreases, by roughly 1 percentage point. This is equivalent to about 2.5% of the 41% unconditional mean of institutional ownership and 5% of its

ur

22 p.p. unconditional standard deviation (Table 1).

Jo

We also test whether institutional investors that are likely to be particularly sophisticated have an even stronger tendency to avoid firms with greater technology spillovers. 24 To proxy for greater sophistication among institutional investors, we use the proportion that is active as opposed to passive. Table 4 shows that the proportion of institutional investors that is active decreases by roughly 1 percentage point for a 10% increase in technology spillovers (Panel C). This 1 p.p. decrease is small but still meaningful compared to the mean proportion of institutional investors that is active (70%) or its standard deviation (17%) (Table 1).

24

We thank an anonymous referee for suggesting this analysis for institutional investors as well as the corresponding analysis for sell-side analysts.

Journal Pre-proof We then test our predictions for the coverage of sell-side analysts. Table 4 shows that there is a decrease in analyst coverage (Panel D), of 2.5% relative to its mean for a 10% increase in technology spillovers. To proxy for greater sophistication among analysts, we use the mean experience of analysts covering the firm, where an analyst's experience is measured as the number of years since the analyst began covering firms. Table 4 shows that, for a 10% increase in technology spillovers, analyst experience decreases by about 2%-4% relative to its mean

of

(Panel E). Alternatively, if we proxy for analyst sophistication using earnings forecast errors (as

ro

in Table 3), our inferences are similar.

-p

Finally, we examine whether technology spillovers increase uncertainty about earnings

re

information. We test this proposition using the dispersion of analysts' earnings estimates (Diether, Malloy, and Scherbina (2002)). Note that since earnings information increases in

lP

complexity owing to the very nature of technology spillovers, this proposition about complexity

na

is very natural but also very difficult to test empirically, differently from the proposition about uncertainty. It is the latter proposition that we test.

ur

Table 4 shows that uncertainty increases significantly as a result of technology spillovers.

Jo

The coefficients are economically significant in all specifications, with dispersion increasing by 1 p.p. for a 10% increase in technology spillovers (Panel F). This is equivalent to about 3% of its unconditional standard deviation of 37% (Table 1). The coefficients are also statistically significant in three of the four specifications, and only marginally insignificant in one. Taken as a whole, our results thus far indicate that technology spillovers lead to greater information asymmetry. The increase in information asymmetry is exacerbated by sophisticated market participants avoiding firms with greater technology spillovers. Finally, information

Journal Pre-proof asymmetry increases in part because of the greater uncertainty about earnings information that results from technology spillovers. 4.3. Managerial Activities Additionally, we examine managerial activities for evidence of whether managers use financial reporting to reduce information asymmetry relative to investors, and for whether managers' stock trades reflect their information advantage versus investors.25 Financial reporting

of

quality as an outcome is difficult to predict, partly because it is a choice variable for managers,

ro

and partly because it is unclear whether managers will, on balance, choose higher or lower

-p

financial reporting quality as captured by accruals. One possibility is that managers use higher

re

accruals (i.e., lower quality) in combination with higher information production costs resulting from technology spillovers to better hide their extraction of private benefits. Another possibility

lP

is that managers use lower accruals (i.e., higher quality) to reduce or even eliminate the higher

na

information production costs resulting from technology spillovers. Furthermore, managers may use their information advantage to increase their stock trading, and to trade against the

ur

misvaluation that they perceive. They may do so in conjunction with higher accruals, which

Jo

would better mask their trading against misvaluation, or they may trade without the cloak of higher accruals. Consequently, our prediction for accruals is ambiguous. [Insert Table 5 about here] We test our first prediction using discretionary accruals to capture financial reporting quality. Table 5 presents the results in Panel A. These results do not provide evidence to suggest that technology spillovers affect financial reporting quality in either direction. However, as we have explained above, indecisive results are not unexpected for this outcome.

25

We thank two anonymous referees for suggesting this analysis.

Journal Pre-proof We test our second prediction using both the number and value of trades to measure both the volume and the direction of insider trading. Table 5 presents the results in Panel B through Panel E. These results show that the volume of insider trading increases as a result of technology spillovers. For a 10% increase in technology spillovers, there is an increase of roughly 0.3 trades per shareholders (Panel B), which is equivalent to about 11% of the unconditional mean number of insider trades (2.8) and 6% of its unconditional standard deviation (5.2) (Table 1). Similarly,

of

for the value of insider trades, there is an increase of roughly $2 per thousand dollars of market

ro

capitalization (Panel C), which is equivalent to about 28% and 9% of its unconditional mean and

-p

standard deviation, respectively (Table 1).

re

The results in Table 5 also show that insiders are net purchasers of stock in their firm as a result of technology spillovers. For a 10% increase in technology spillovers, the net purchase

lP

ratio increases by 3-6 percentage points, depending on whether we consider the number or value

na

of trades (Panel D or Panel E, respectively). This is equivalent to about 4%-8% of the unconditional standard deviation of the number or value of trades (Table 1). We also find that the

ur

purchase and sales ratios, respectively, tend to increase and decrease significantly (not

Jo

tabulated). Overall, our results for insider trading are consistent with managers believing that their firm is undervalued by investors. 5. Implications for Earnings Expectations In the previous section, we established that technology spillovers result in greater information asymmetry. In this section, we first examine the implications of technology spillovers for earnings expectations. We then examine whether future stock returns are consistent with earnings expectations.

Journal Pre-proof Technology spillovers to a firm have been established to increase its productivity and innovation, so its realized earnings should also increase as a result of technology spillovers. The increase in information asymmetry resulting from technology spillovers should lead to investors being less certain about future earnings, but it is unclear whether it should lead to investors accurately predicting future earnings or being systematically biased in their expectations about them. While we do not have clear predictions here, the literature does find that investors tend to

ro

may also apply to firms with greater technology spillovers.

of

underestimate the earnings of firms with high intangible assets and also to undervalue them. This

-p

5.1. Expected and Realized Earnings

re

We begin our analysis by examining whether technology spillovers affect realized earnings. We then test whether the earnings expectations of market participants are affected by

lP

technology spillovers. Finally, we test whether technology spillovers affect the difference

na

between realized and expected earnings, i.e., earnings surprises. Since it can take many years for innovation activities to produce results, we examine

ur

earnings in both the short run and the long run. To this end, we use earnings for the next year and

Jo

long-term earnings growth rates for the next five years. In our regression specifications, we follow the literature on realized and expected earnings (e.g., Core, Guay, and Rusticus (2006), Edmans (2011), and Giroud and Mueller (2011)). In addition to the features common to all of our regression specifications (Section 2.5), we control for market capitalization and market-to-book. [Insert Table 6 about here]

Journal Pre-proof Table 6 presents the results.26 Starting with the short run (one year horizon), Panel A shows that realized earnings as a proportion of total assets are higher, by as much as 0.3% for a 10% increase in technology spillovers, but they are statistically significant in only two of the four regressions. Panel B shows that the market appears to expect comparably higher earnings as a result of technology spillovers, by as much as 0.2% for a 10% increase in technology spillovers, but these estimates are again only significant in two regressions. Indeed, Panel C shows that the

of

market is approximately correct, and earnings surprises are economically and statistically

ro

insignificant in the short run. By contrast, short-term earnings estimates unconditionally tend to

-p

be overly optimistic as a rule (e.g., Hong and Kubik (2003)), i.e., without considering the effect

re

of technology spillovers.

Moving on to the long run (five year horizon), Panel A shows that the realized earnings

lP

growth rate is significantly higher, by roughly 1 percentage point for a 10% increase in

na

technology spillovers. By comparison, its unconditional standard deviation is 23% (Table 1). However, the market expects a lower earnings growth rate, by as much as 0.3 p.p., as shown in

ur

Panel B. The net effect is that, in the long run, realizations are significantly higher than

Jo

expectations, by approximately 2 p.p., as shown in Panel C. This is about 10% of the unconditional standard deviation of the earnings growth rate surprise (21%). By contrast, longterm earnings estimates unconditionally tend to be overly optimistic (e.g., Chan, Karceski, and Lakonishok (2003)), i.e., without considering the effect of technology spillovers. It is noteworthy that technology spillovers lead to positive earnings surprises not in the short run but rather in the long run. The absence of earnings surprises in the short run is consistent with the increase in valuations documented in the literature (e.g., BSV). However, 26

The difference between realized and expected earnings is not exactly equal to the earnings surprise because the sample sizes are different, more so five years out than one year out. However, we find similar results if we only use an overlapping sample.

Journal Pre-proof presence of earnings surprises in the long run is inconsistent with valuations quickly and completely reflecting earnings information about technology spillovers. Rather, it suggests that while initial valuations may be higher, they are still too low. Additionally, it is worth noting that there is a relatively large effect of technology spillovers on earnings in the long run (and likewise for future stock returns, as we also find). The effect that we document is necessarily estimated using firms that survive in the long run, and we

of

do not (and cannot) capture the effect for firms that disappear from our sample (through mergers

ro

and acquisitions, bankruptcies, etc.). However, our results do indicate that technology spillovers

-p

lead to higher payoffs (earnings to the firm, stock returns to investors) for surviving firms.

re

5.2. Abnormal Stock Returns

We continue our analysis by examining whether future stock returns are consistent with

lP

the pattern of earnings expectations that we document. We predict that investors will incorporate

na

into stock prices the increase in earnings resulting from technology spillovers, but the process may take several years. Following the literature, we examine abnormal stock returns and account

ur

for the extent to which they are explained by both risk factors and firm characteristics

Jo

(Faulkender and Wang (2006), Dittmar and Mahrt-Smith (2007), and Denis and Sibilkov (2010)). Specifically, we run regression of future abnormal stock returns on our variables of interest as well as the standard explanatory variables in the empirical literature on stock returns. These latter variables include market capitalization, market-to-book, cash flow, stock returns, and the volatility of stock returns. Since the effect of technology spillovers may take time to be reflected in stock prices, we examine stock returns over horizons of one to five years. In addition to the features common to all of our regression specifications (Section 2.5), we control for the aforementioned asset pricing variables.

Journal Pre-proof [Insert Table 7 about here] Table 7 presents the results. Abnormal stock returns increase significantly as a result of technology spillovers. At the one year horizon (Panel A), stock returns are higher by approximately 1.5 percentage points for a 10% increase in technology spillovers. By comparison, the unconditional standard deviation of stock returns is 38% (Table 1). The results are statistically significant in three of the four specifications, and just marginally insignificant in one.

of

At the five year horizon (Panel B), the results are higher by a comparable magnitude, again

ro

roughly 1.5 p.p., and they are statistically significant in all four specifications. The results are

-p

similar at the two, three, and four year horizons, and they are all statistically significant as well

the earnings of the corresponding firms.27

re

(not tabulated). The magnitude of abnormal stock returns is also consistent with the magnitude of

lP

Our results for technology spillovers are consistent with the evidence in the literature that

na

investors tend to undervalue firms that are innovative (e.g., see Chan, Lakonishok, and Sougiannis (2001)). Turning to our control variables of interest, our results indicate that product

ur

market spillovers do not reliably affect stock returns. Similarly, while the firm's own R&D is

Jo

positively related to stock returns, consistent with the literature, the relationship weakens as the horizon increases, and it is generally not statistically significant. Taking our results together, we find that technology spillovers lead to higher earnings surprises in the long run as well as persistently higher future stock returns. The effect of technology spillovers on information asymmetry appears to also result in underestimation by investors of the earnings of firms with greater technology spillovers and their undervaluation.

27

For a 10% increase in technology spillovers, the earnings growth rate surprise is about 2 p.p. (Table 6), and abnormal stock returns are about 1.5 p.p., both on a per annum basis estimated using a five year horizon. In a simple discounted cash flow type valuation model, our future stock returns roughly correspond to our earnings growth rate surprises.

Journal Pre-proof 6. Conclusion Motivated by the effect on firm value of technology spillovers documented by the prior literature, we study the effect of these technology spillovers on the corporate information environment. We argue that spillovers of innovation activities across technological peer firms have the effect, on balance, of increasing the complexity and uncertainty of value relevant information about the firm. This increases information production costs, which discourages the

of

production of information, and information asymmetry between insiders and outsiders increases

find

that

technology

spillovers

do

indeed

increase

information asymmetry.

-p

We

ro

as a result.

re

Furthermore, sophisticated market participants, such as institutional investors and sell-side analysts, avoid firms with greater technology spillovers, thereby exacerbating information

lP

asymmetry. Moreover, the dispersion of earnings estimates increases, consistent with technology

na

spillovers increasing uncertainty about value relevant information. Additionally, managers do appear to trade on their information advantage, but our evidence does not suggest that they use

ur

financial reporting to reduce information asymmetry.

Jo

Finally, we examine the implications of technology spillovers for earnings expectations. We find that technology spillovers do not affect realized or expected earnings in the short run, but they lead to both higher realized earnings in the long run and positive earnings surprises. Additionally, future stock returns are also significantly higher as a result of technology spillovers. Overall, investors appear to underestimate the effect of technology spillovers on earnings and firm value.

Journal Pre-proof References Akcigit, Ufuk, and William R. Kerr, 2018, Growth through heterogeneous innovations, Journal of Political Economy 126, 1374-1443. Akcigit, Ufuk, Murat Alp Celik, and Jeremy Greenwood, 2016, Buy, keep or sell: Economic growth and the market for ideas, Econometrica 84, 943-984. Armstrong, Christopher S., John E. Core, Daniel J. Taylor, and Robert E. Verrecchia, 2011,

of

When does information asymmetry affect the cost of capital?, Journal of Accounting

ro

Research 49, 1-40.

-p

Arrow, Kenneth, 1962, The Rate and Direction of Inventive Activity: Economic and Social

re

Factors, Princeton University Press.

Asquith, Paul, Michael B., Mikhail, and Andrea S. Au, 2005, Information content of equity

lP

analyst reports, Journal of Financial Economics 75, 245-282.

na

Barberis, Nicholas, and Richard Thaler, 2003, A survey of behavioral finance, in George M.

Finance.

ur

Constantinides, Milton Harris and René M. Stulz, eds.: Handbook of the Economics of

Jo

Barth, Mary E., Ron Kasznik, and Maureen F. McNichols, 2001, Corporate disclosure policy and analyst behavior, Journal of Accounting Research 39, 1-34. Baumol, William J., 2002, The Free-Market Innovation Machine: Analyzing the Growth Miracle of Capitalism, Princeton University Press. Bena, Jan, and Kai Li, 2014, Corporate innovations and mergers and acquisitions, Journal of Finance 69, 1923-1960. Berger, Philip G., 1993, Explicit and implicit tax effects of the R&D tax credit, Journal of Accounting Research 31, 131-171.

Journal Pre-proof Bhattacharya, Sudipto, and Jay R. Ritter, 1983, Innovation and communication: Signalling with partial disclosure, Review of Economic Studies 50, 331-346. Bhojraj, Sanjeev, and Charles M. C. Lee, 2002, Who is my peer? A valuation-based approach to the selection of comparable firms, Journal of Accounting Research 40, 407-439. Bloom, Nicholas, Mark Schankerman, and John Van Reenen, 2013, Identifying technology spillovers and product market rivalry, Econometrica 81, 1347-1393.

of

Bloom, Nick, Rachel Griffith, and John Van Reenen, 2002, Do R&D tax credits work? Evidence

ro

from a panel of countries, 1979-1997, Journal of Public Economics 85, 1-31.

-p

Bossaerts, Peter, Paolo Ghirardato, Serena Guarnaschelli, and William R. Zame, 2010,

re

Ambiguity in asset markets: Theory and experiment, Review of Financial Studies 23, 1325-1359.

lP

Brown, Stephen, and Stephen A. Hillegeist, 2007, How disclosure quality affects the level of

na

information asymmetry, Review of Accounting Studies 12, 443-477.

paper.

ur

Brunnermeier, Markus, and Martin Oehmke, 2009, Complexity in financial markets, working

Jo

Cao, H. Henry, Tan Wang, and Harold H. Zhang, 2005, Model uncertainty, limited market participation, and asset prices, Review of Financial Studies 18, 1219-1251. Carlin, Bruce Ian, Shimon Kogan, and Richard Lowery, 2013, Trading complex assets, Journal of Finance 68, 1937-1960. Chan, Louis K. C., Jason Karceski, and Josef Lakonishok, 2003, The level and persistence of growth rates, Journal of Finance 58, 643-684. Chan, Louis K. C., Josef Lakonishok, and Theodore Sougiannis, 2001, The stock market valuation of research and development expenditures, Journal of Finance 56, 2431-2456.

Journal Pre-proof Chen,

Tao,

and

Chen Lin,

2017,

Does information asymmetry affect corporate tax

aggressiveness?, Journal of Financial and Quantitative Analysis 52, 2053-2081. Chen, Tao, Jarrad Harford, and Chen Lin, 2015, Do analysts matter for governance? Evidence from natural experiments, Journal of Financial Economics 115, 383-410. Chen, Tao, Lingmin Xie, and Yuanyuan Zhang, 2017, How does analysts' forecast quality relate to corporate investment efficiency?, Journal of Corporate Finance 43, 217-240.

of

Chen, Zengjing, and Larry Epstein, 2002, Ambiguity, risk, and asset returns in continuous time,

ro

Econometrica 70, 1403-1443.

-p

Chirinko, Robert S., and Daniel J. Wilson, 2017, Tax competition among U.S. States: Racing to

re

the bottom or riding on a seesaw?, Journal of Public Economics 155, 147-163.

Finance 63, 1977-2011.

lP

Cohen, Lauren, and Andrea Frazzini, 2008, Economic links and predictable returns, Journal of

383-400.

na

Cohen, Lauren, and Dong Lou, 2012, Complicated firms, Journal of Financial Economics 104,

ur

Cohen, Lauren, Karl Diether, and Christopher Malloy, 2013, Misvaluing innovation, Review of

Jo

Financial Studies 26, 635-666. Core, John H., Wayne R. Guay, and Tjomme O. Rusticus, 2006, Does weak governance cause weak stock returns? An examination of firm operating performance and investors' expectations, Journal of Finance 61, 655-687. Cremers, K. J. Martijn, and Antti Petajisto, 2009, How active is your fund manager? A new measure that predicts performance, Review of Financial Studies 22, 3329-3365.

Journal Pre-proof Cummins, Jason G., Kevin A. Hassett, and R. Glenn Hubbard, 1994, A reconsideration of investment behavior using tax reforms as natural experiments, Brookings Papers on Economic Activity 2, 1-60. Dechow, Patricia M., Richard G. Sloan, and Amy P. Sweeney, 1995, Detecting earnings management, Accounting Review 70, 193-225. Del Guercio, Diane, 1996, The distorting effect of the prudent-man laws on institutional equity

of

investments, Journal of Financial Economics 40, 31-62.

ro

Denis, David J., and Valeriy Sibilkov, 2010, Financial constraints, investment, and the value of

-p

cash holdings, Review of Financial Studies 23, 247-269.

re

Derrien, François, and Ambrus Kecskés, 2013, The real effects of financial shocks: Evidence from exogenous changes in analyst coverage, Journal of Finance 68, 1407-1440.

lP

Diether, Karl B., Christopher J. Malloy, and Anna Scherbina, 2002, Differences of opinion and

na

the cross section of stock returns, Journal of Finance 57, 2113-2141. Dittmar, Amy, and Jan Mahrt-Smith, 2007, Corporate governance and the value of cash

ur

holdings, Journal of Financial Economics 83, 599-634.

Jo

Dow, James, and Sérgio Ribeiro da Costa Werlang, 1992, Uncertainty aversion, risk aversion, and the optimal choice of portfolio, Econometrica 60, 197-204. Duru, Augustine, and David M. Reeb, 2002, International diversification and analysts' forecast accuracy and bias, Accounting Review 77, 415-433. Eberhart, Allan C., William F. Maxwell, and Akhtar R. Siddique, 2004, An examination of longterm abnormal stock returns and operating performance following R&D increases, Journal of Finance 59, 623-650.

Journal Pre-proof Edmans, Alex, 2011, Does the stock market fully value intangibles? Employee satisfaction and equity prices, Journal of Financial Economics 101, 621-640. Falato, Antonio, and Jae W. Sim, 2014, Why do innovative firms hold so much cash? Evidence from changes in state R&D tax credits, working paper. Faulkender, Michael, and Rong Wang, 2006, Corporate financial policy and the value of cash, Journal of Finance 61, 1957-1990.

ro

information flow, Journal of Finance 62, 951-989.

of

Ferreira, Miguel A., and Paul A. Laux, 2007, Corporate governance, idiosyncratic risk, and

-p

Foucault, Thierry, and Laurent Fresard, 2014, Learning from peers' stock prices and corporate

Francis,

Jennifer,

J.

Douglas

Hanna,

re

investment, Journal of Financial Economics 111, 554-577. and

Donna

R.

Philbrick,

1997,

Management

lP

communications with securities analysts, Journal of Accounting and Economics 24, 363-

na

394.

Frankel, Richard, and Xu Li, 2004, Characteristics of a firm's information environment and the

ur

information asymmetry between insiders and outsiders, Journal of Accounting and

Jo

Economics 37, 229-259.

Frankel, Richard, S. P. Kothari, and Joseph Weber, 2006, Determinants of the informativeness of analyst research, Journal of Accounting and Economics 41, 29-54. Gilson, Stuart C., Paul M. Healy, Christopher F. Noe, and Krishna G. Palepu, 2001, Analyst specialization and conglomerate stock breakups, Journal of Accounting Research 39, 565-582. Giroud, Xavier, and Holger M. Mueller, 2011, Corporate governance, product market competition, and equity prices, Journal of Finance 66, 563-600.

Journal Pre-proof Goldstein, Itay, and Liyan Yang, 2015, Information diversity and complementarities in trading and information acquisition, Journal of Finance 70, 1723-1765. Grossman, Gene M., and Elhanan Helpman, 1991, Trade, knowledge spillovers, and growth, European Economic Review 35, 517-526. Hall, Bronwyn H., 1992a, Investment and research and development at the firm level: Does the source of financing matter?, working paper.

of

Hall, Bronwyn H., 1992b, R&D tax policy during the 1980s: Success or failure?, in Tax Policy

ro

and the Economy 7, 1-36.

-p

Hall, Bronwyn H., 2002, The financing of research and development, Oxford Review of

re

Economic Policy 18, 35-51.

Hall, Robert E., and Dale W. Jorgenson, 1967, Tax policy and investment behavior, American

lP

Economic Review 57, 391-414.

na

Healy, Paul M., Amy P. Hutton, and Krishna G. Palepu, 1999, Stock performance and intermediation changes surrounding sustained increases in disclosure, Contemporary

ur

Accounting Research 16, 485-520.

Jo

Healy, Paul M., and Krishna G. Palepu, 2001, Information asymmetry, corporate disclosure, and the capital markets: A review of the empirical disclosure literature, Journal of Accounting and Economics 31, 405-440. Himmelberg, Charles P., and Bruce C. Petersen, 1994, R&D and internal finance: A panel study of small firms in high-tech industries, Review of Economics and Statistics 76, 38-51. Hines, James R., 1993, On the sensitivity of R&D to delicate tax changes: The behavior of U.S. multinationals in the 1980s, in Alberto Giovannini, R. Glenn Hubbard, and Joel Slemrod, eds.: Studies in International Taxation, University of Chicago Press.

Journal Pre-proof Hirshleifer, David, and Siew Hong Teoh, 2003, Limited attention, information disclosure, and financial reporting, Journal of Accounting and Economics 36, 337-386. Hirshleifer, David, Po-Hsuan Hsu, and Dongmei Li, 2013, Innovative efficiency and stock returns, Journal of Financial Economics 107, 632-654. Hoberg, Gerard, and Gordon Phillips, 2010, Product market synergies and competition in mergers and acquisitions: A text-based analysis, Review of Financial Studies 23, 3773-

of

3811.

ro

Hoberg, Gerard, and Gordon Phillips, 2016, Text-based network industries and endogenous

-p

product differentiation, Journal of Political Economy 124, 1423-1465.

re

Hombert, Johan, and Adrien Matray, 2018, Can innovation help U.S. manufacturing firms escape import competition from China?, Journal of Finance 73, 2003-2039.

lP

Hong, Harrison, and Jeffrey D. Kubik, 2003, Analyzing the analysts: Career concerns and biased

na

earnings forecasts, Journal of Finance 58, 313-351. Hong, Harrison, and Jeremy C. Stein, 1999, A unified theory of underreaction, momentum

ur

trading, and over reaction in asset markets, Journal of Finance 54, 2143-2184.

Jo

Hong, Harrison, and Marcin Kacperczyk, 2009, The price of sin: The effects of social norms on markets, Journal of Financial Economics 93, 15-36. Jaffe, Adam B., 1986, Technological opportunity and spillovers of R&D: Evidence from firms' patents, profits, and market value, American Economic Review 76, 984-1001. Kadan, Ohad, Leonardo Madureira, Rong Wang, and Tzachi Zach, 2012, Analysts' industry expertise, Journal of Accounting and Economics 54, 95-120. Kelly, Bryan, and Alexander Ljungqvist, 2012, Testing asymmetric-information asset pricing models, Review of Financial Studies 25, 1366-1413.

Journal Pre-proof Kim,

Jeong-Bon,

Stephen Teng Sun, and Zilong Zhang, 2018, Technology spillovers,

information externality, and stock price crash risk, working paper. Kogan, Leonid, Dimitris Papanikolaou, Amit Seru, and Noah Stoffman, 2017, Technological innovation, resource allocation, and growth, Quarterly Journal of Economics 132, 665712. Lang, Mark H., and Russell J. Lundholm, 1996, Corporate disclosure policy and analyst

of

behavior, Accounting Review 71, 467-492.

ro

Lee, Charles M.C., Stephen Teng Sun, Rongfei Wang, and Ran Zhang, 2019, Technological

-p

links and predictable returns, Journal of Financial Economics 132, 76-96.

re

Lev, Baruch, and Theodore Sougiannis, 1996, The capitalization, amortization, and valuerelevance of R&D, Journal of Accounting and Economics 21, 107-138.

lP

Litov, Lubomir P., Patrick Moreton, and Todd R. Zenger, 2012, Corporate strategy, analyst

na

coverage, and the uniqueness paradox, Management Science 58, 1797-1815. Maksimovic, Vojislav, and Gordon Phillips, 2001, The market for corporate assets: Who engages

ur

in mergers and asset sales and are there efficiency gains?, Journal of Finance 56, 2019-

Jo

2065.

Moretti, Enrico, and Daniel J. Wilson, 2017, The effect of state taxes on the geographical location of top earners: Evidence from star scientists, American Economic Review 107, 1858-1903. Nguyen, Phuong-Anh, and Ambrus Kecskés, 2019, Technology spillovers, asset redeployability, and corporate financial policies, working paper. Phillips, Gordon M., and Alexei Zhdanov, 2013, R&D and the incentives from merger and acquisition activity, Review of Financial Studies 26, 34-78.

Journal Pre-proof Plumlee, Marlene A., 2003, The effect of information complexity on analysts' use of that information, Accounting Review 78, 275-296. Romer, Paul M., 1990, Endogenous technological change, Journal of Political Economy 98, S71S102. Rosenberg, Nathan, 1979, Technological interdependence in the American economy, Technology and Culture 20, 25-50.

ro

the biopharmaceutical industry, working paper.

of

Thakor, Richard T., and Andrew W. Lo, 2019, Competition and R&D financing: Evidence from

-p

The Wall Street Journal, 2012, Gordon Crovitz: Who Really Invented the Internet?, July 12.

re

Verrecchia, Robert E., 1983, Discretionary disclosure, Journal of Accounting and Economics 5, 179-194.

lP

Wilson, Daniel J., 2009, Beggar thy neighbor? The in-state, out-of-state and aggregate effects of

Haifeng, and Xiao-jun Zhang, 2009, Financial reporting complexity and investor

ur

underreaction to 10-K information, Review of Accounting Studies 14, 559-586.

Jo

You,

na

R&D tax credits, Review of Economics and Statistics 91, 431-436.

Journal Pre-proof Table 1 Descriptive Statistics This table presents descriptive statistics for technology spillover variables, firm characteristics variables, and all dependent variables. The sample comprises 12,118 firm-year observations corresponding to 694 unique firms between 1981 and 2001. The firms in the sample are publicly traded U.S. operating firms excluding financials and utilities. All variables are defined in Appendix Table 1. All variables are multiplied by 100 except for the technology spillover variables, the stock of patents, firm age, total assets, the number of institutional investors, the number of analysts, analyst experience, and the two insider trading net purchase ratios . The number and value of insider trades are multiplied by 1,000. Standard deviation

25th percentile

Median

75th percentile

Technology spillover variables in levels - Raw Jaffe ($ billions) - Purged Jaffe ($ billions) - Raw Mahalanobis ($ billions) - Purged Mahalanobis ($ billions)

25 21 110 105

19 16 81 75

10 9 51 50

20 17 90 88

34 29 149 140

Technology spillover variables in logarithms - Raw Jaffe ($ millions) - Purged Jaffe ($ millions) - Raw Mahalanobis ($ millions) - Purged Mahalanobis ($ millions)

9.7 9.6 11.3 11.3

1.1 1.0 0.9 0.8

9.2 9.1 10.8 10.8

9.9 9.8 11.4 11.4

10.4 10.3 11.9 11.8

Firm characteristics variables - R&D (%) - Patent stock - Firm age (years) - Total assets ($ millions)

44.9 611 24.6 2,507

68.9 1,935 18.1 6,366

0.0 5 11.7 90

19.9 28 20.1 338

59.5 175 31.5 1,648

2.9 106.7 18.3

2.9 361.3 13.2

1.1 0.3 8.3

1.9 3.2 15.4

3.6 35.6 25.4

1.0

2.9

0.1

0.2

0.7

46.6

30.5

25.5

37.8

58.5

109.3 41.1

139.2 22.1

17.0 23.0

54.0 42.9

148.0 58.7

70.1

17.3

59.7

72.6

82.8

11.3 4.4 18.5

12.0 3.1 36.8

2.0 1.9 3.7

7.0 4.4 7.5

17.0 6.7 16.9

5.8 2.8 7.1

5.9 5.2 21.3

1.8 0.1 0.1

3.9 0.9 0.9

7.6 2.9 4.4

-20.1

68.7

-92.0

-16.7

16.7

-33.2

73.9

-99.9

-69.8

0.0

Jo

Sophisticated market participants variables - Number of institutional investors - Fraction owned by institutional investors - Proportion of institutional investors that is active - Number of analysts - Analyst experience - Earnings estimates dispersion Managerial activities variables - Discretionary accruals - Number of insider trades - Value of insider trades - Net purchase ratio based on number of insider trades - Net purchase ratio based on value of insider trades

ro

-p

re

lP

ur

na

Information asymmetry variables - Bid-ask spread - Amihud illiquidity measure - Returns ratio - Magnitude of earnings announcement surprises - Volatility of earnings announcement reaction

of

Mean

Journal Pre-proof

6.9 8.6 -1.7 9.0 14.9 -6.1

7.6 6.3 5.1 23.3 6.4 21.4

3.2 4.7 -2.4 -3.2 10.9 -16.9

6.5 7.5 -0.5 8.0 13.5 -6.0

10.4 11.2 0.4 17.9 17.3 3.0

Stock performance variables - Abnormal stock returns

7.3

37.7

-14.5

5.8

26.5

Jo

ur

na

lP

re

-p

ro

of

Earnings variables - One year realized earnings - One year expected earnings - One year earnings surprise - Five year realized earnings growth rate - Five year expected earnings growth rate - Five year earnings growth rate surprise

Journal Pre-proof Table 2 Descriptive Statistics by Industry Sorted by Technology Spillovers This table presents descriptive statistics by industry sorted by technology spillovers. The sample comprises 12,118 firm-year observations corresponding to 694 unique firms between 1981 and 2001. The firms in the sample are publicly traded U.S. operating firms excluding financials and utilities. Only industries with at least five unique firms are included (97% of the sample). All variables are defined in Appendix Table 1. R&D is multiplied by 100.

Industry

Obs.

Mean of raw Jaffe technology spillovers

Communications (SIC=48) Transportation equipment (SIC=37) Chemicals and related products (SIC=28) Electronic equipment excl. computers (SIC=36) Construction products (SIC=32) Consumer and business instruments (SIC=38) Business services incl. technology (SIC=73) Machinery and equipment incl. computers (SIC=35) Paper and related products (SIC=26) Rubber and plastic products (SIC=30) Metal mining (SIC=10) Primary metal industries (SIC=33) Wood products excl. furniture (SIC=24) Fabricated metal products (SIC=34) Petroleum refining and related industries (SIC=29) Textile mill products (SIC=22) Oil and gas extraction (SIC=13) Wholesale durable goods (SIC=50) Food and related products (SIC=20) Printing, publishing, and related industries (SIC=27) Furniture and fixtures (SIC=25) Miscellaneous manufacturing industries (SIC=39) Wholesale non-durable goods (SIC=51) Apparel and related products (SIC=23) Leather and related products (SIC=31)

61 727 1,226 1,876 258 1,086 166 1,806 425 261 52 392 84 735 183 185 196 216 517 280 236 318 69 224 122

10.50 10.30 10.24 10.11 10.04 9.98 9.94 9.88 9.85 9.79 9.70 9.59 9.56 9.42 9.40 9.34 9.29 9.16 9.14 8.97 8.94 8.54 8.34 8.27 7.05

Jo

n r u

l a

r P

f o

ro

p e

Standard deviation of raw Jaffe technology spillovers 1.09 0.74 0.57 0.74 0.69 0.69 0.78 0.86 0.94 1.01 0.46 0.86 0.83 0.97 1.52 1.12 1.28 1.03 0.96 1.16 1.07 1.36 1.53 1.29 1.41

Mean of raw Jaffe product market spillovers

Mean of R&D

9.42 8.25 8.54 8.53 6.02 8.15 7.73 7.89 7.13 7.74 4.52 6.47 4.77 6.74 8.81 4.06 7.48 7.66 5.69 6.69 4.50 7.11 3.91 1.64 0.96

56.8 31.0 52.8 70.4 16.4 101.4 74.9 76.4 16.0 25.1 0.8 9.7 0.0 17.4 4.7 9.5 6.4 20.2 4.8 3.7 15.6 12.3 11.8 0.7 16.5

Journal Pre-proof Table 3 The Effect of Technology Spillovers on Information Asymmetry

of

This table presents the results of regressions of information asymmetry proxies on technology spillovers. The sample comprises 12,118 firm-year observations corresponding to 694 unique firms between 1981 and 2001. The firms in the sample are publicly traded U.S. operating firms excluding financials and utilities. For each dependent variable, four regressions are run, one for each measure of spillovers. In each regression, the same measure is used for technology spillovers and product market spillovers. The four spillover measures are the raw and purged Jaffe and Mahalanobis measures. The independent variables are as follows: technology and product market spillovers; R&D; federal and state tax credits, but only in specifications using purged spillover measure s; the natural logarithm of firm age; the natural logarithm of market capitalization; market-to-book; cash flow; stock returns; and stock return volatility. All variables are defined in Appendix Table 1. All dependent variables are multiplied by 100. The independent variables are lagged. Spillovers are measured in natural logarithms. Fixed effects are included for firms and industry-years. Standard errors are clustered by industry-year. ***, **, and * indicate statistical significance at the 1%, 5%, and 10% levels, respectively. Only selected results are tabulated.

Technology spillovers (t-1)

0.08 (0.65)

R&D (t-1)

-0.05 (-0.90)

lP

re

Product market spillovers (t-1)

Raw Mahalanobis 1.06** (2.21)

Purged Mahalanobis 0.94 (1.29)

-0.56* (-1.95)

-0.17 (-0.86)

-0.66 (-1.18)

-0.05 (-0.83)

-0.04 (-0.70)

-0.03 (-0.46)

6,624 6,624 6,624 6,624 0.763 0.764 0.763 0.763 Panel B: Amihud Illiquidity Measure Dependent variable is Amihud illiquidity measure (t)

na

Observations Adjusted R2

Purged Jaffe 1.70*** (3.10)

-p

Raw Jaffe 0.98** (2.50)

ro

Panel A: Bid-Ask Spread Dependent variable is bid-ask spread (t)

Purged Jaffe 54.49 (1.03)

Raw Mahalanobis 186.11*** (3.84)

Purged Mahalanobis 95.30 (1.53)

Product market spillovers (t-1)

-51.69*** (-4.06)

-65.82** (-2.45)

-51.51*** (-2.64)

-106.92** (-2.29)

R&D (t-1)

-23.18** (-2.11)

-18.16 (-1.64)

-20.44* (-1.86)

-17.27 (-1.56)

11,176 0.569

11,176 0.568

11,176 0.568

11,176 0.568

Observations Adjusted R2

Jo

Technology spillovers (t-1)

ur

Raw Jaffe 209.28*** (5.00)

Journal Pre-proof Panel C: Returns Ratio Dependent variable is returns ratio (t) Raw Jaffe 9.04*** (6.62)

Purged Jaffe 6.73*** (3.64)

Raw Mahalanobis 7.58*** (4.34)

Purged Mahalanobis 6.28*** (2.96)

Product market spillovers (t-1)

0.26 (0.88)

-1.49** (-2.20)

1.81*** (3.39)

-2.19 (-1.55)

R&D (t-1)

0.20 (0.78)

0.36 (1.37)

0.26 (0.98)

0.45* (1.67)

Technology spillovers (t-1)

Technology spillovers (t-1)

-0.17 (-1.46)

R&D (t-1)

0.05 (0.49)

Purged Mahalanobis 1.65* (1.79)

0.82 (1.50)

-0.43** (-2.14)

0.36 (0.43)

0.05 (0.46)

0.05 (0.50)

0.06 (0.54)

8,169 8,169 8,169 8,169 0.504 0.505 0.505 0.504 Panel E: Volatility of Earnings Announcement Reaction Dependent variable is earnings announcement reaction volatility (t)

na ur

Technology spillovers (t-1)

Jo

Product market spillovers (t-1)

Observations Adjusted R2

Raw Mahalanobis 2.04*** (3.20)

lP

Observations Adjusted R2

R&D (t-1)

Purged Jaffe 0.91 (1.14)

re

Product market spillovers (t-1)

ro

Raw Jaffe 1.37*** (2.82)

of

11,586 11,586 11,586 11,586 0.746 0.744 0.745 0.744 Panel D: Magnitude of Earnings Announcement Surprises Dependent variable is earnings announcement surprise magnitude (t)

-p

Observations Adjusted R2

Raw Jaffe 12.98*** (3.41)

Purged Jaffe 16.40*** (3.61)

Raw Mahalanobis 15.59*** (3.39)

Purged Mahalanobis 14.41*** (2.61)

1.50 (1.50)

3.04 (1.35)

2.64 (1.65)

7.02* (1.84)

-0.13 (-0.12)

-0.14 (-0.14)

-0.13 (-0.13)

-0.06 (-0.05)

11,413 0.538

11,413 0.538

11,413 0.538

11,413 0.538

Journal Pre-proof Table 4 Technology Spillovers and Sophisticated Market Participants

Jo

ur

na

lP

re

-p

ro

of

This table presents the results of regressions of proxies for sophisticated market participants on technology spillovers. The sample comprises 12,118 firm-year observations corresponding to 694 unique firms between 1981 and 2001. The firms in the sample are publicly traded U.S. operating firms excluding financials and utilities. For each dependent variable, four regressions are run, one for each measure of spillovers. In each regression, the same measure is used for technology spillovers and product market spillovers. The four spillover measures are the raw and purged Jaffe and Mahalanobis measures. The independent variables are as follows: technology and product market spillovers; R&D; federal and state tax credits, but only in specifications using purged spillover measures; the natural logarithm of firm age; the natural logarithm of market capitaliza tion; market-to-book; cash flow; stock returns; and stock return volatility. All variables are defined in Appendix Table 1. In Panel A, Panel E, and Panel F, natural logarithms are taken after adding one to the dependent variables. All dependent variables are multiplied by 100. The independent variables are lagged. Spillovers are measured in natural logarithms. Fixed effects are included for firms and industry-years. Standard errors are clustered by industry-year. ***, **, and * indicate statistical significance at the 1%, 5%, and 10% levels, respectively. Only selected results are tabulated.

Journal Pre-proof Panel A: Number of Institutional Investors Dependent variable is ln(number of institutional investors) (t)

Technology spillovers (t-1) Product market spillovers (t-1)

Raw Jaffe -43.41*** (-6.84)

Purged Jaffe -28.74*** (-3.37)

Raw Mahalanobis -29.04*** (-3.77)

Purged Mahalanobis -23.47** (-2.47)

5.99*** (4.64)

15.56*** (4.22)

1.43 (0.56)

17.23*** (2.74)

1.24 (0.84)

0.38 (0.26)

0.62 (0.42)

-0.01 (-0.01)

R&D (t-1)

Technology spillovers (t-1)

1.56*** (3.33)

R&D (t-1)

1.15*** (2.84)

Purged Mahalanobis -10.84*** (-4.20)

0.89 (0.86)

-0.00 (-0.00)

-0.56 (-0.33)

1.01** (2.49)

1.11*** (2.71)

1.00** (2.44)

11,595 11,595 11,595 11,595 0.816 0.815 0.815 0.815 Panel C: Proportion of Institutional Investors that is Active Dependent variable is proportion that is active (t)

na ur

Technology spillovers (t-1)

Jo

Product market spillovers (t-1)

Observations Adjusted R2

Raw Mahalanobis -11.84*** (-6.37)

lP

Observations Adjusted R2

R&D (t-1)

Purged Jaffe -8.90*** (-4.06)

re

Product market spillovers (t-1)

ro

Raw Jaffe -11.35*** (-7.95)

of

11,595 11,595 11,595 11,595 0.950 0.949 0.949 0.949 Panel B: Stake of Institutional Investors Dependent variable is fraction owned by institutional investors (t)

-p

Observations Adjusted R2

Raw Jaffe -6.25*** (-3.13)

Purged Jaffe -9.13*** (-3.97)

Raw Mahalanobis -9.36*** (-3.94)

Purged Mahalanobis -10.89*** (-4.00)

0.19 (0.37)

2.06* (1.66)

1.42 (1.52)

0.75 (0.34)

1.84*** (4.42)

1.82*** (4.48)

1.81*** (4.35)

1.80*** (4.40)

11,351 0.566

11,351 0.567

11,351 0.566

11,351 0.567

Journal Pre-proof Panel D: Analyst Coverage Dependent variable is ln(number of analysts) (t) Raw Jaffe -24.43*** (-3.65)

Purged Jaffe -27.88** (-2.47)

Raw Mahalanobis -24.75*** (-2.98)

Purged Mahalanobis -29.98*** (-2.80)

Product market spillovers (t-1)

5.43*** (2.96)

17.27*** (3.16)

0.36 (0.12)

7.71 (0.88)

R&D (t-1)

-4.48*** (-3.01)

-4.83*** (-3.19)

-4.51*** (-3.01)

-4.77*** (-3.15)

Technology spillovers (t-1)

Technology spillovers (t-1)

2.42 (1.48)

R&D (t-1)

-1.84 (-1.18)

Purged Mahalanobis -21.06* (-1.65)

3.16 (0.63)

0.54 (0.16)

-11.77 (-1.45)

-2.22 (-1.39)

-1.79 (-1.13)

-2.15 (-1.35)

9,210 9,210 9,210 9,210 0.914 0.914 0.914 0.914 Panel F: Uncertainty Dependent variable is earnings estimates dispersion (t)

na ur

Technology spillovers (t-1)

Jo

Product market spillovers (t-1)

Observations Adjusted R2

Raw Mahalanobis -38.81*** (-3.97)

lP

Observations Adjusted R2

R&D (t-1)

Purged Jaffe -28.21** (-2.49)

re

Product market spillovers (t-1)

ro

Raw Jaffe -31.43*** (-4.71)

of

10,704 10,704 10,704 10,704 0.898 0.898 0.897 0.898 Panel E: Analyst Experience Dependent variable is ln(analyst experience) (t)

-p

Observations Adjusted R2

Raw Jaffe 8.54 (1.64)

Purged Jaffe 16.03** (2.03)

Raw Mahalanobis 12.03* (1.81)

Purged Mahalanobis 19.87** (2.37)

-1.32 (-0.95)

-8.66 (-1.64)

-4.44* (-1.94)

-11.45* (-1.86)

4.48** (2.02)

4.62** (2.07)

4.58** (2.07)

4.70** (2.12)

7,795 0.417

7,795 0.417

7,795 0.417

7,795 0.417

Journal Pre-proof Table 5 The Effect of Technology Spillovers on Managerial Activities

of

This table presents the results of regressions of discretionary accruals and insider trading on technology spillovers. The sample comprises 12,118 firm-year observations corresponding to 694 unique firms between 1981 and 2001. The firms in the sample are publicly traded U.S. operating firms excluding financials and utilities. For each dependent variable, four regressions are run, one for each measure of spillovers. In each regression, the same measure is used for technology spillovers and product market spillovers. The four spillover measures are the raw and purged Jaffe and Mahalanobis measures. The independent variables are as follows: technology and product market spillovers; R&D; federal and state tax credits, but only in specifications using purged spillover measures; the natural logarithm of firm age; the natural logarithm of market capitaliza tion; market-to-book; cash flow; stock returns; and stock return volatility. All variables are defined in Appendix Table 1. All dependent variables are multiplied by 100. The independent variables are lagged. Spillovers are measured in natural logarithms. Fixed effects are included for firms and industry-years. Standard errors are clustered by industry-year. ***, **, and * indicate statistical significance at the 1%, 5%, and 10% levels, respectively. Only selected results are tabulated.

Technology spillovers (t-1)

-0.04 (-0.17)

R&D (t-1)

0.26 (0.99)

lP

re

Product market spillovers (t-1)

ur

Jo

Technology spillovers (t-1)

Product market spillovers (t-1) R&D (t-1) Observations Adjusted R2

Raw Mahalanobis 0.55 (0.48)

Purged Mahalanobis 0.50 (0.36)

0.42 (0.66)

-0.87* (-1.95)

0.38 (0.33)

0.21 (0.79)

0.26 (0.99)

0.22 (0.83)

11,300 11,300 11,300 11,300 0.193 0.194 0.194 0.194 Panel B: Number of Insider Trades Dependent variable is number of insider trades (t)

na

Observations Adjusted R2

Purged Jaffe 0.75 (0.62)

-p

Raw Jaffe -0.52 (-0.57)

ro

Panel A: Discretionary Accruals Dependent variable is discretionary accruals (t)

Raw Jaffe 0.32*** (3.67)

Purged Jaffe 0.22* (1.92)

Raw Mahalanobis 0.44*** (4.03)

Purged Mahalanobis 0.40*** (3.15)

-0.02 (-1.30)

0.02 (0.26)

-0.04 (-0.94)

-0.03 (-0.29)

0.09*** (3.47)

0.09*** (3.51)

0.09*** (3.47)

0.09*** (3.53)

8,537 0.509

8,537 0.508

8,537 0.509

8,537 0.508

Journal Pre-proof Panel C: Value of Insider Trades Dependent variable is value of insider trades (t) Raw Jaffe 1.63*** (3.32)

Purged Jaffe 2.24*** (3.07)

Raw Mahalanobis 2.32*** (3.34)

Purged Mahalanobis 1.98** (2.36)

Product market spillovers (t-1)

-0.22** (-2.08)

-0.15 (-0.46)

-0.25 (-1.39)

0.36 (0.67)

R&D (t-1)

-0.00 (-0.03)

-0.01 (-0.09)

-0.01 (-0.05)

-0.01 (-0.04)

Technology spillovers (t-1)

Technology spillovers (t-1)

0.04 (0.01)

R&D (t-1)

-5.59** (-2.23)

Jo

Product market spillovers (t-1)

Observations Adjusted R2

-7.14 (-0.64)

-6.58 (-1.18)

-0.19 (-0.01)

-5.93** (-2.36)

-5.66** (-2.25)

-5.90** (-2.36)

lP

ur

Technology spillovers (t-1)

R&D (t-1)

Purged Mahalanobis 36.99 (1.60)

8,662 8,662 8,662 0.306 0.307 0.307 Panel E: Net Purchase Ratio Based On Value of Insider Trades Dependent variable is net purchase ratio (t)

na

Observations Adjusted R2

Raw Mahalanobis 29.62* (1.95)

re

Product market spillovers (t-1)

Purged Jaffe 35.87** (2.00)

8,660 0.143

ro

Raw Jaffe 11.04 (0.97)

of

8,660 8,660 8,660 0.143 0.143 0.143 Panel D: Net Purchase Ratio Based On Number of Insider Trades Dependent variable is net purchase ratio (t)

-p

Observations Adjusted R2

8,662 0.307

Raw Jaffe 27.07** (2.25)

Purged Jaffe 50.98** (2.56)

Raw Mahalanobis 45.24*** (2.65)

Purged Mahalanobis 56.65** (2.17)

0.62 (0.15)

-11.02 (-0.86)

-4.43 (-0.82)

-7.95 (-0.42)

-5.92** (-2.28)

-6.04** (-2.35)

-5.94** (-2.29)

-5.92** (-2.30)

8,662 0.295

8,662 0.296

8,662 0.296

8,662 0.296

Journal Pre-proof Table 6 The Effect of Technology Spillovers on Earnings This table presents the results of regressions of realized, expected, and unexpected earnings on technology spillovers. The sample comprises 12,118 firm-year observations corresponding to 694 unique firms between 1981 and 2001. The firms in the sample are publicly traded U.S. operat ing firms excluding financials and utilities. For each dependent variable, four regressions are run, one for each measure of spillovers. In each regression, the same measure is used for technology spillovers and product market spillovers. The four spillover measures are the raw and purged Jaffe and Mahalanobis measures. The independent variables are as follows: technology and product market spillovers; R&D; federal and state tax credits, but only in specifica tions using purged spillover measures; the natural logarithm of firm age; the natural logarithm of market capitalization; and the natural logarithm of market-to-book. All variables are defined in Appendix Table 1. At the one year horizon, natural logarithms are taken after adding one to the dependent variables. All dependent variables are multiplied by 100. The independent variables are lagged. Spillovers are measured in natural logarithms. Fixed effects are included for firms and industry-years. Standard errors are clustered by industry-year. ***, **, and * indicate statistical significance at the 1%, 5%, and 10% levels, respectively. Only selected results are tab ulated. Panel A: Realized Earnings One year horizon

Raw Jaffe 0.77 (0.59)

Purged Jaffe 0.49 (0.31)

Product market spillovers (t-1)

-0.39 (-1.56)

-0.23 (-0.19)

R&D (t-1)

0.62 (1.59)

0.60 (1.57)

Observations Adjusted R2

8,397 0.533

Technology spillovers (t-1)

J

u o

8,397 0.535

r P

Five year horizon

Dependent variable is realized earnings growth rate (t)

Raw Mahalanobis 3.90** (2.38)

Purged Mahalanobis 3.63* (1.83)

Raw Jaffe 14.20*** (3.59)

Purged Jaffe 7.49 (1.34)

Raw Mahalanobis 13.71*** (2.78)

Purged Mahalanobis 13.33** (2.04)

-0.67 (-1.28)

-0.99 (-0.54)

1.11 (0.97)

0.26 (0.09)

4.28** (2.10)

2.76 (0.53)

0.54 (1.40)

0.56 (1.46)

-0.66 (-0.52)

-0.18 (-0.14)

-0.68 (-0.53)

-0.34 (-0.27)

8,397 0.534

8,397 0.535

8,329 0.248

8,329 0.246

8,329 0.248

8,329 0.246

l a

rn

o r p

e

Dependent variable is ln(realized earnings) (t)

f o

Journal Pre-proof Panel B: Expected Earnings One year horizon

Five year horizon

Dependent variable is ln(expected earnings) (t)

Dependent variable is expected earnings growth rate (t)

Raw Jaffe -0.22 (-0.27)

Purged Jaffe 0.28 (0.24)

Raw Mahalanobis 2.64*** (2.65)

Purged Mahalanobis 2.62** (2.08)

Raw Jaffe -3.30*** (-2.74)

Purged Jaffe -1.32 (-0.98)

Raw Mahalanobis -0.68 (-0.42)

Purged Mahalanobis 1.51 (0.82)

Product market spillovers (t-1)

-0.22 (-1.18)

0.67 (1.14)

-0.73** (-2.01)

0.16 (0.18)

0.88*** (3.58)

2.55*** (3.46)

0.49 (1.21)

2.33* (1.94)

R&D (t-1)

0.65** (2.57)

0.59** (2.38)

0.58** (2.31)

0.56** (2.25)

-0.67** (-2.22)

-0.83*** (-2.77)

-0.77** (-2.58)

-0.88*** (-2.96)

8,490 0.710

8,490 0.711

6,974 0.635

6,974 0.634

6,974 0.634

6,974 0.634

Technology spillovers (t-1)

Observations Adjusted R2

8,490 8,490 0.710 0.711 Panel C: Earnings Surprise One year horizon

o r p

e

r P

Dependent variable is ln(earnings surprise) (t)

Technology spillovers (t-1)

Raw Jaffe 0.75 (0.67)

Product market spillovers (t-1)

-0.18 (-0.70)

R&D (t-1)

-0.01 (-0.02)

Observations Adjusted R2

8,397 0.417

Purged Jaffe -0.41 (-0.27)

l a

Five year horizon

Dependent variable is earnings growth rate surprise (t)

Raw Mahalanobis 0.70 (0.47)

Purged Mahalanobis 0.97 (0.49)

Raw Jaffe 22.87*** (5.00)

Purged Jaffe 12.19 (1.66)

Raw Mahalanobis 23.10*** (4.90)

Purged Mahalanobis 21.53*** (3.42)

0.13 (0.26)

-1.94 (-1.05)

0.86 (0.71)

-7.15** (-2.12)

4.42* (1.84)

-7.59 (-1.06)

0.05 (0.13)

-0.01 (-0.04)

0.04 (0.10)

1.44 (0.90)

2.52 (1.61)

1.54 (0.98)

2.40 (1.54)

8,397 0.418

8,397 0.417

8,397 0.418

5,453 0.238

5,453 0.233

5,453 0.237

5,453 0.233

n r u

Jo

f o

-1.24 (-1.07)

Journal Pre-proof Table 7 The Effect of Technology Spillovers on Abnormal Stock Returns

ro

of

This table presents the results of regressions of abnormal stock returns on technology spillovers. The sample comprises 12,118 firm-year observations corresponding to 694 unique firms between 1981 and 2001. The firms in the sample are publicly traded U.S. operating firms excluding financials and utilities. For each dependent variable, four regressions are run, one for each measure of spillovers. In each regression, the same measure is used for technology spillovers and product market spillovers. The four spillover measures are the raw and purged Jaffe and Mahalanobis measures. The dependent variables are abnormal stock returns es timated using the four-factor model and annualized. Abnormal stock returns are measured over horizons of one and five years in Panel A and Panel B, respectively. The independent variables are as follows: technology and product market spillovers; R&D; federal and state tax credits, but only in specifications using purged spillover measures; the natural logarithm of firm age; the natural logarithm of market capitalization; market-to-book; cash flow; stock returns; and stock return volatility. All variables are defined in Appendix Table 1. The dependent variables are multiplied by 100. The independent variables are lagged. Spillovers are measured in natural logarithms. Fixed effects are included for firms and industry-years. Standard errors are clustered by industry-year. ***, **, and * indicate statistical significance at the 1%, 5%, and 10% levels, respectively. Only selected results are tabulated. Panel A: One Year Horizon Dependent variable is abnormal stock returns (t) Raw Mahalanobis 12.56 (1.64)

Purged Mahalanobis 18.14** (1.99)

-0.29 (-0.19)

-9.46** (-2.06)

3.79 (1.31)

-9.29 (-1.26)

2.41 (1.42)

2.72 (1.60)

2.41 (1.41)

2.86* (1.67)

re

Technology spillovers (t-1)

Purged Jaffe 16.97** (2.07)

-p

Raw Jaffe 13.95** (2.15)

lP

Product market spillovers (t-1) R&D (t-1)

11,582 11,582 11,582 11,582 0.180 0.181 0.181 0.181 Panel B: Five Year Horizon Dependent variable is abnormal stock returns (t)

ur

na

Observations Adjusted R2

Raw Jaffe 18.26*** (8.93)

Purged Jaffe 14.47*** (4.72)

Raw Mahalanobis 16.42*** (6.13)

Purged Mahalanobis 17.61*** (5.32)

Product market spillovers (t-1)

0.01 (0.01)

-5.54*** (-3.43)

3.67*** (3.26)

-3.37 (-1.42)

R&D (t-1)

0.73 (1.26)

1.13* (1.90)

0.80 (1.37)

1.16* (1.96)

Observations Adjusted R2

11,033 0.534

11,033 0.531

11,033 0.533

11,033 0.530

Jo

Technology spillovers (t-1)

56

Journal Pre-proof Appendi x Table 1 Variable Definitions This table presents variable definitions. Variables are computed for every firm-year. Industry is defined using twodigit SIC codes. * indicates that the variable is defined using Compustat data items. † indicates that the variable is computed as in Bloom, Schankerman, and Van Reenen (2013). Name

Definition

Spillover variables - Raw Jaffe - Raw Mahalanobis

of

The Jaffe or Mahalanobis distances in the technology or product market spaces are computed for each pair of firms. Then the stock of R&D is computed for every firm-year. Finally, the spillover variables for a firm are computed as the natural logarithm of the sum of the R&D stock of each of the other firms weighted by the distance between the firm in question and each of the other firms. † Computed like the corresponding raw variables except that the R&D stock of other firms is first purged before weighting and summing. Specifically, R&D tax credits are computed for each firm-year, and the R&D stock is regressed on the R&D tax credits. The resulting predicted values are used as the purged R&D stock corresponding to each firm-year. †

-p

ro

- Purged Jaffe - Purged Mahalanobis

Bid-ask spread computed using daily stock trading data Amihud illiquidity measure computed using daily stock trading data Ratio of the number of trading days with zero or missing returns to the total number of trading days Mean of the absolute value of quarterly earnings forecast errors. Earnings forecast errors are measured as analysts ' earnings estimates minus earnings reported by the firm all divided by the stock price. Annualized mean of the standard deviation of quarterly earnings announcement returns. Earnings announcement returns are measured as raw returns minus market returns during the three days centered on the earnings announcement date. For each quarterly earnings announcement during the year, the volatility of daily returns is calculated. This resulting volatility is then averaged across the quarterly earnings announcement during the year, and the resulting mean volatility is annualized.

- Magnitude of earnings announcement surprises

Jo

ur

na

- Volatility of earnings announcement reaction

lP

re

Information asymmetry variables - Bid-ask spread - Amihud illiquidity measure - Returns ratio

Sophisticated market participants variables - Number of institutional investors - Fraction owned by institutional investors - Proportion of institutional investors that is active

- Number of analysts - Analyst experience

- Earnings estimates dispersion

Number of institutional investors in the firm Fraction of the firm owned by institutional investors Proportion of institutional investors that is active. Active investors are those with active share of 25% or more. Active share is computed as half of the sum across all firms of the distance between the weight of a firm in the investor's portfolio minus its weight in the index. See Cremers and Petajisto (2009). Number of analysts covering the firm Mean experience of analysts covering the firm. An analyst's experience is measured as the number of years elapsed since the analyst began covering firms. Coefficient of variation of analysts ' earnings estimates

57

Journal Pre-proof Managerial activities variables - Discretionary accruals

Absolute value of discretionary accruals divided by total assets. See Dechow, Sloan, and Sweeney (1995) for estimating discretionary accruals. Number of insider trades scaled by number of shareholders measured in thousands of people. Value of insider trades scaled by market capitalization measured in millions of dollars Insider purchases minus insider sales all divided by the sum of insider purchases and insider sales . Trading is computed based on the number of trades. Insider purchases minus insider sales all divided by the sum of insider purchases and insider sales. Trading is computed based on the value of trades.

- Number of insider trades - Value of insider trades - Net purchase ratio based on number of insider trades - Net purchase ratio based on value of insider trades

of

Earnings variables - One year realized earnings - One year expected earnings

ro

Actual earnings times shares outstanding all scaled by total assets Analysts' earnings estimates times shares outstanding all scaled by total assets Difference between realized and expected earnings at the one year horizon Five year growth rate of (IB/CSHO)/AJEX * Analysts' long-term earnings growth rate estimates Difference between realized and expected earnings growth rate at the five year horizon

- One year earnings surprise

re

Stock performance variables - Abnormal stock returns

lP

Abnormal stock returns estimated using the four-factor model implemented using daily returns and then annualized Stock of the firm's R&D accumulated up to a given firm-year adjusted for depreciation and scaled by the firm's stock of physical capital † Natural logarithm of the firm's federal and state tax credits in a given firm-year † Number of years as a publicly traded firm Stock of the firm's patents accumulated up to a given firm-year AT * PRCC_FCSHO * (PRCC_FCSHO)/(TXDITC+CEQ) * OIBDP/AT * Annualized mean daily stock returns Annualized standard deviation of daily stock returns

Jo

ur

na

Control variables - R&D

- Federal tax credits - State tax credits - Firm age - Patent stock - Total assets - Market capitalization - Market-to-book - Cash flow - Stock returns - Stock return volatility

-p

- Five year realized earnings growth rate - Five year expected earnings growth rate - Five year earnings growth rate surprise

Highlights Technology spillovers increase complexity and uncertainty of information Hence information asymmetry between insiders and outsiders should increase We find more information asymmetry, avoidance by sophisticated market participants Also find increase in uncertainty, insider trading Also, investors underestimate long-term earnings, future stock returns are higher

58