1. Introduction
Financial analysts are one of the most important information intermediaries between investors and firms. Consistent with this notion, prior research suggests that increase in analyst coverage leads to increase in stock liquidity and institutional ownership and decrease in the cost of equity (
Beyer et al. 2010).
Barth et al. (
2001) suggest that firms with substantial intangible assets (e.g., R&D), most of which are not recognized in financial statements, have more information asymmetry and inherent uncertainty about firm value than other firms. They further argue that in the absence of private information acquisition by analysts, share prices of intangible-intensive firms would less precisely reflect their fundamental values, which will in turn create opportunities for private information acquisition for analysts. Consistent with these arguments,
Barth et al. (
2001) predict and find a positive association between analyst coverage and R&D. However,
Barth et al. (
2001) do not investigate what particular attribute of R&D leads to this positive association. In this paper we aim to bridge this gap in the extant literature and investigate cross-sectional determinants of analyst coverage for R&D firms. We explore four cross-sectional determinants: reporting biases arising from expensing of R&D compared to capitalization of R&D, uncertainty associated with R&D, investors’ attention and scale effects of R&D.
The first cross-sectional determinant of analyst coverage for R&D firms is reporting bias arising from immediate expensing of R&D compared to capitalization of R&D. Prior research suggests that, due to immediate expensing of R&D, financial statements are less informative for R&D firms because accounting numbers do not reflect firm value and performance (
Amir and Lev 1996;
Lev and Zarowin 1999). An alternative source of information are the analyst forecasts.
Kimbrough (
2007) suggests that financial analysts might fill up the information gap (lack of informativeness) in the financial statements resulting from expensing of R&D. Analysts could benefit from private information acquisition by eliminating the information gap in the financial statements caused by reporting biases and increase analyst coverage for R&D firms. Consistent with this argument,
Barth et al. (
2001) suggest that immediate expensing of R&D increases information asymmetry for R&D firms, which leads to greater analyst coverage for these firms. Alternatively, reporting biases arising from expensing of R&D might also have the opposite effect on analyst coverage for R&D firms.
Lang and Lundholm (
1996) suggest that analyst coverage is greater (smaller) for firms with more (less) informative disclosures, because information acquisition costs are smaller (greater) for firms with more (less) informative disclosures. Reporting bias resulting from expensing of R&D is likely to reduce informativeness of financial statements and increase analysts’ information acquisition costs, which could negatively affect analysts’ willingness to follow R&D firms. Consequently, there are two competing hypotheses about the impact of reporting biases resulting from expensing of R&D on analyst coverage for R&D firms. Therefore, we do not have, a priori, a clear prediction about the impact of reporting biases on analyst coverage for R&D firms.
The second cross-sectional determinant of analyst coverage for R&D firms is uncertainty associated with R&D. Prior research suggests that future benefits of R&D are more uncertain than other investments (
Lev 2001). The success rate of innovation activities is often very low, and the distribution of payoffs is highly skewed (
Lev 2001). Consistent with these arguments,
Kothari et al. (
2002) document the fact that volatility of earnings for R&D is much greater than that for capital or advertising expenditures. Greater uncertainty associated with R&D may increase investors’ demand for analyst forecasts. Therefore, greater uncertainty may increase analyst coverage for R&D firms. However, greater uncertainty associated with R&D could also increase analysts’ forecast errors, and in turn decrease analyst coverage for R&D firms (
Weiss 2010). Consequently, there are two competing hypotheses about the impact of uncertainty on the relationship between R&D and analyst coverage. Therefore, we do not have a clear prediction about the impact of uncertainty on analyst coverage for R&D firms.
The third cross-sectional determinant of analyst coverage for R&D firms is investor’s attention.
Owen (
2002) suggests that investment in high-technology companies boomed as people invested large sums of money, even when there was little chance of the company being profitable. If investors direct greater attention toward R&D-intensive firms, this could also affect analyst coverage for these firms. Investment banks that hire financial analysts profit from trading commissions generated from stock trades. Consequently, financial analysts might follow R&D-intensive firms, which might increase trading commissions for their employer. Therefore, we expect investors’ attention to increase analyst coverage for R&D intensive firms.
The last cross-sectional determinant of analyst coverage for R&D firms are the scale effects of R&D.
Schumpeter (
1942) suggests that larger firms are better positioned than smaller firms to implement and successfully exploit R&D efforts.
Henderson and Cockburn (
1996) argue that larger firms enjoy multiple R&D project spillover advantages.
Cohen and Klepper (
1996) suggest that larger firms have “cost spreading” advantages in R&D investment. Consistent with these arguments,
Ciftci and Cready (
2011) document the fact that R&D is associated with greater earnings for larger firms than for smaller firms, suggesting that larger R&D firms are more productive than smaller firms. Given that larger firms have more productive R&D, analyst coverage could be greater for larger R&D firms, because analyst are more likely to cover more R&D-productive firms. Therefore, we expect the scale effect of R&D to increase analyst coverage for R&D firms.
We use I/B/E/S, CRSP and Compustat data from the U.S. over the period of 1985 to 2018. Consistent with
Barth et al. (
2001), we use R&D-expense-to-operating-expense ratio as our measure of R&D intensity. We first investigate the impact of reporting biases from expensing of R&D compared to capitalization of R&D on analyst coverage for R&D firms. We use two proxies for reporting biases from expensing of R&D. The first proxy comes from
Lev et al. (
2005), and is based on the difference between return on equity (ROE) and annualized R&D growth (RDG_ROE). We develop the second proxy on our own, based on the bias in earnings under expensing of R&D compared to capitalization of R&D (DIFROA). We separate firm-year observations into three categories based on reporting biases resulting from expensing of R&D: aggressive (conservative) firms are those in the bottom (top) 25% of measures of reporting bias (RDG_ROE and DIFROA) and neutral firms are those between 25% and 75% of measures of reporting bias. With the first measure of reporting bias (RDG_ROE), we find that the association between R&D and analyst coverage is lower for conservative firms compared to neutral firms. However, there is no statistically significant difference between aggressive and neutral firms in the association between R&D and analyst coverage.
These results suggest that while aggressive reporting has no impact on the association between analyst coverage and R&D, conservative reporting decreases this association. That is, conservative reporting leads to lower analyst coverage for R&D firms, while aggressive reporting has no effect on analyst coverage. Using the second measure of reporting bias resulting from expensing of R&D (i.e., DIFROA), we find that the association between R&D and analyst coverage for aggressive and conservative firms is significantly lower than that for neutral firms. That is, both conservative and aggressive reporting lead to a decrease in analyst coverage for R&D firms compared to neutral reporting. These results suggest that the costs associated with reporting biases from expensing of R&D are greater for financial analysts than the benefits arising from private information acquisition. Overall, our findings suggest that reporting biases from expensing of R&D negatively affect analyst coverage for R&D firms and that the impact is stronger for conservatively reporting firms than for aggressively reporting firms. Moreover, we find that reporting biases from expensing of R&D do not seem to explain the positive association between R&D and analyst coverage documented by
Barth et al. (
2001).
Second, we investigate the impact of the uncertainty associated with R&D on analyst coverage for R&D firms. We use two proxies for uncertainty: standard deviation of earnings over the past five years (from year t to t − 4) and standard deviation of stock returns over fiscal year t. With both proxies, we find that the association between R&D and analyst coverage decreases with the uncertainty, suggesting that costs associated with the uncertainty for financial analysts outweigh the benefits. Moreover, we find that the uncertainty associated with R&D does not explain the positive association between R&D and analyst coverage.
Third, we investigate the impact of investors’ attention on analyst coverage for R&D firms. We use trading volume as a proxy for investors’ attention, following
Barber and Odean (
2008) and
Hou et al. (
2009). Trading volume is commonly used in financial literature as an indicator of investor interest because higher trading volumes typically reflect greater market activity and interest in a particular stock. Prior research has revealed that trading volume correlates strongly with investors’ attention, making it a reliable proxy. We find that the association between R&D and analyst coverage increases with trading volume, suggesting that investors’ attention increases analyst coverage for R&D firms. Furthermore, we find that investors’ attention partially explain the positive association between R&D and analyst coverage, but not fully.
Fourth, we investigate the impact of scale effects of R&D on analyst coverage for R&D firms. We use log of firm size to capture the scale effects of R&D and find that the association between R&D and analyst coverage increases with firm size, suggesting that scale effects of R&D increase analyst coverage for R&D firms. Moreover, we find that scale effects of R&D fully explain the positive relationship between R&D and analyst coverage documented by
Barth et al. (
2001).
Finally, we investigate the combined effect of all cross-sectional determinants (i.e., combined effect of reporting biases, uncertainty, investors’ attention and scale effects of R&D) on analyst coverage for R&D firms. We find that the association between R&D and analyst coverage decreases with reporting biases and uncertainty, consistent with the stand-alone results discussed above. We also find that the association between R&D and analyst coverage decreases with scale effects of R&D, consistent with stand-alone results. However, we find that the impact of investors’ attention on analyst coverage for R&D firms is either marginally significant or insignificant in the combined analysis. These findings suggest that the impact of cross-sectional determinants are not subsumed by the presence of other cross-sectional determinants (except investors’ attention), indicating that the cross-sectional determinants considered in this study are independent from each other.
This paper contributes to several streams of extant research. Our study contributes to the literature that investigates the relationship between analyst coverage and R&D.
Barth et al. (
2001) document a positive association between R&D and analyst coverage. However, they do not investigate what particular attribute of R&D leads to this positive association. In this paper we complement
Barth et al. (
2001) in two ways. First, we identify cross-sectional determinants of the relationship between R&D and analyst coverage such as reporting biases, uncertainty associated with R&D, investors’ attention and scale effect of R&D. Second, we identify which particular cross-sectional determinant explains the positive relationship between R&D and analyst coverage documented by
Barth et al. (
2001).
Our paper further contributes to the prior literature about the negative consequences of immediate expensing of R&D (
Lev and Zarowin 1999;
Lev 2001;
Lev et al. 2005). Several researchers argue that the immediate expensing of R&D negatively affects the informativeness or usefulness of financial statements for equity investors because it biases accounting numbers (
Amir and Lev 1996;
Lev and Zarowin 1999;
Lev et al. 2005). While prior research primarily focuses on the demand for accounting information (i.e., the usefulness or informativeness of accounting information for equity investors), we complement prior research by providing evidence on the supply of information by financial analysts (i.e., analyst coverage). Our evidence suggests that reporting biases resulting from expensing of R&D negatively affect not only the demand for information by equity investors, but also the supply of information by financial analysts. Prior research suggests that financial analysts are likely to fill the information gap (lack of informativeness) in financial statements resulting from expensing of R&D (
Barth et al. 2001;
Kimbrough 2007). However, our evidence suggests that financial analysts themselves are negatively affected by reporting biases. Consequently, they are less likely to fill the information gap in the financial statements arising from expensing of R&D. Therefore, our findings suggest that the negative consequences of expensing of R&D could be more severe than that suggested in the prior literature (
Amir and Lev 1996;
Lev and Zarowin 1999;
Lev et al. 2005). The International Financial Reporting System (IFRS) requires the capitalization of development costs for internally developed intangibles such as R&D (International Accounting Standard 38). Given that the U.S. is on the path to converge to IFRS, our findings about the impact of reporting biases of expensing of R&D on analyst coverage for R&D firms could be useful to standard setters in evaluating the consequences of expensing of R&D as compared to capitalization of R&D.
This paper also contributes to the ongoing discussion about the scale effects of R&D. Prior research documents scale the effects of R&D impact productivity, future earnings, stock returns and number of patents generated from R&D investments (
Cohen and Klepper 1996;
Henderson and Cockburn 1996;
Ciftci and Cready 2011). We extend the prior literature by documenting that the scale effects of R&D also affect analyst coverage for R&D firms.
This paper also enriches the existing literature that investigates the implications of uncertainty associated with R&D. Prior research suggests that uncertainty associated with R&D affects numerous attributes of R&D firms such as stock market and bond market valuations, information asymmetry and insider gains, value of analyst recommendations and stock return volatility (
Aboody and Lev 2000;
Chan et al. 2001;
Shi 2003;
Palmon and Yezegel 2012).
1 We contribute to this stream of research by documenting the fact that the uncertainty associated with R&D also affects analyst coverage for R&D firms. Finally, while prior research focuses primarily on the impact of investor attention on stock market valuation of R&D firms (
Owen 2002), this paper complements this literature by documenting the impact of investor attention on analyst coverage for R&D firms.
The rest of the paper is organized as follows.
Section 2 outlines the relevant literature and hypotheses.
Section 3 briefly discuss the U.S. adoption of IFRS through the lens of its implications for R&D.
Section 4 introduces the research design and provides the sample selection, descriptive statistics and Pearson correlations.
Section 5 reports the empirical results, and
Section 6 concludes the paper.
4. Research Design
We investigate the relationship between R&D intensity and analyst coverage using the following pooled cross-sectional ordinary least squares (OLS) regression:
The definitions of the variables in Equation (1) are provided in
Table 1. We include year and industry indicator variables in all estimations to control for year and industry fixed effects. Industry indicator variables are based on the
Fama and French (
1997) 48 industry definitions. Instead of including industry fixed effects,
Barth et al. (
2001) use industry-adjusted measures of R&D and other intangibles. We prefer to use industry fixed effects rather than industry-adjusted measures because industry-adjusted measures produce both positive and negative values which may not be possible to interpret when we interact industry-adjusted R&D with log of firm size. For example, when we interact industry-adjusted R&D with log of market value (LMV), both variables will have positive and negative values. Consequently, interpretation of the interaction term will not be possible. In addition, we cluster firm-year observations by firm to eliminate autocorrelation, as suggested by
Petersen (
2009). To alleviate the influence of outliers, we winsorize all continuous variables (COVRGE, RND, ADV, DEPR, INTANA, GDWAL, LMV, GROWTH, VOLUME, STDRET, ROA, STDROA) except RET at the top and bottom 1% of respective annual distributions. Absolute change in earnings (absEARN) is winsorized at the top and bottom 1% of signed change in earnings before taking the absolute value.
We deflate RND, ADV, and DEPR with operating expenses following
Barth et al. (
2001). RND in Equation (1) shows the association between R&D and analyst coverage.
Barth et al. (
2001) documents the fact that R&D is positively associated with analyst coverage. Therefore, we expect the coefficient estimate of RND in Equation (1) to be positive. UNCRTNY in Equation (1) shows the association between uncertainty and R&D. We have two measures of uncertainty: STDROA and STDRET. The two measures of uncertainty associated with R&D are used to examine the impact of uncertainty on analyst coverage for R&D firms. Following
Barth et al. (
2001), we include ADV, DEPR, INTANA, GDWLA, LMV, VOLUME and GROWTH in our model. We also add RET, absEARN and ROA into our model, following
Healy et al. (
1999). Consistent with
Amir et al. (
2003), we add LOSS into our model.
In addition to the relationship between analyst coverage and R&D,
Barth et al. (
2001) also investigate the relationship between analyst effort and R&D. We do not investigate the relationship between R&D and analyst effort, for several reasons. The decision to omit measures of analysts’ effort is motivated by the following key considerations. First, analyst coverage represents an ex ante decision made in anticipation of factors like intangible assets. In contrast, measures of analysts’ effort, such as the average number of firms covered (
Barth et al. 2001) or analysts’ forecast accuracy (
Harford et al. 2019), are ex post outcomes influenced by the decision to cover a firm, and both represent actual rather than expected measures of analysts’ efforts. By focusing on analyst coverage, we aim to capture the intrinsic motivation that precedes analysts’ decisions. Second, measures of analysts’ effort (e.g., analysts’ forecast accuracy) are inherently subjective and may be influenced by various factors beyond the incentive to cover a firm, such as analysts’ experience and knowledge. This subjectivity introduces complexity and potential confounding factors to the study. Third,
Shon and Young (
2011) explore factors influencing an analyst’s choice to drop coverage. Their findings, highlighting the impact of economic incentives like analyst compensation and attracting business for the brokerage house on the decision to discontinue coverage, align with our decision to exclude analyst effort. Finally, our study is specifically designed to explore the cross-sectional determinants of the association between R&D and analyst incentives to cover a firm. Excluding measures of analysts’ effort allows us to focus on the primary determinants of analyst coverage (reporting biases, uncertainty, investor attention, and scale effects), aligning with our research objectives.
Furthermore, it is worth noting that in previous research, uncertainty, one of the determinants of analyst coverage in our paper, has been gauged using analyst forecast errors (
Gu and Wang 2005;
Weiss 2010). For example,
Gu and Wang (
2005) argue that greater uncertainty was related to R&D results in terms of increased analyst forecast errors. Additionally,
Barron et al. (
2002) document the fact that analyst consensus tends to decrease with the level of a firm’s intangible assets. In a similar vein, the accuracy of analyst forecast errors has also been utilized in other studies to measure analysts’ effort (
Harford et al. 2019). Therefore, we posit that uncertainty may capture, in part, the analysts’ effort in our study.
To test for H1, we use two proxies of reporting bias from expensing of R&D compared to capitalization of R&D. Consistent with
Lev et al. (
2005), the first proxy is based on the difference between return on equity (ROE) and annualized R&D growth (RDG). We develop our second measure, DIFROA, based on the difference in earnings in the cases of capitalization and expensing of R&D. When a firm’s R&D is growing over time, earnings under capitalization of R&D will be much greater than those under expensing of R&D. Consequently, DIFROA will be positive, indicating that expensing of R&D understates earning compared to capitalization of R&D (i.e., expensing of R&D leads to conservative reporting). However, when a firm’s R&D is decreasing over time, earnings under capitalization of R&D will be smaller than those under expensing of R&D. Therefore, DIFROA will be negative, indicating that expensing of R&D overstates earnings (expensing of R&D leads to aggressive reporting). We separate firm-year observations into three categories based on reporting biases resulting from expensing of R&D: aggressive (conservative) firms are those in the bottom (top) 25% of measures of reporting bias (RDG_ROE and DIFROA) and neutral firms are those between 25% and 75% of measures of reporting bias. We estimate the following equation to test H1:
The only difference between Equations (1) and (2a) is that we add CONSRV and AGRSV into our model and interact them with RND in Equation (2a). Our variables of interest are RND and its interactions with CONSRV and AGRSV. The coefficient estimate of RND in Equation (2a) shows the association between R&D and analyst coverage for neutral firms. The coefficient estimate of RND×CONSRV (RND×AGRSV) shows the difference in the coefficient estimate of RND between neutral and conservative (aggressive) firms. If the association between R&D and analyst coverage is lower (higher) for conservative firms than that for neutral firms, then the coefficient estimate of RND×CONSRV should be negative (positive). H1 predicts that reporting biases from expensing of R&D compared to capitalization do not affect analyst coverage for R&D firms. Therefore, H1 predicts that there is no significant difference between neutral firms and conservative (or aggressive) firms in the association between R&D and analyst coverage. Thus, H1 anticipates that both interaction terms, RND×CONSRV and RND×AGRSV, should be insignificant. To test H2, we estimate the following equation:
The only difference between Equations (1) and (2b) is that we interact UNCRTNY with RND in Equation (2b). H2 predicts that uncertainty does not affect the association between analyst coverage and R&D. Therefore, the interaction term, RND×UNCRTNY, in Equation (2b), is expected to be insignificant. To test H3, we estimate the following equation:
Following
Barber and Odean (
2008) and
Hou et al. (
2009), we use trading volume (VOLUME) as a proxy for investors’ attention. To test H3, we interact VOLUME with RND in Equation (2c). H3 predicts that the association between R&D and analyst coverage increases with investor attention. Therefore, the interaction term RND×VOLUME is expected to be positive. To test H4, we estimate the following equation:
To measure scale effects of R&D on analyst coverage, we use log of firm size (LMV), consistent with the prior literature (
Cohen and Klepper 1996;
Henderson and Cockburn 1996;
Ciftci and Cready 2011). If R&D has scale effects on analyst coverage, the association between R&D and analyst coverage should be greater for large firms than for small firms. To test the impact of scale effects of R&D on the association between R&D and analyst coverage, we interact RND with LMV in Equation (2d). If the association between analyst coverage and R&D is increasing with scale effects of R&D, as predicted by H4, then the interaction term, RND×LMV, should be positive.
4.1. Sample Selection
We include all the firm-year observations in CRSP, Compustat and I/B/E/S that have data for estimation of Equation (1) with non-missing value of RDG_ROE. Calculation of RDG_ROE requires R&D data over the past five years (i.e., from years
t to
t − 4). Therefore, we require the firm-year observations to have s positive value of R&D over the past five years. We also delete firm-year observations with a negative average book value of equity (averaged over years
t to
t − 1), because average book value of equity is used in the calculation of RDG_ROE. In addition, we require the firm-year observations to have sales revenue of at least USD 1 million. Following
Barth et al. (
2001), we exclude financial industries (4-digit SIC = 6000–6999) and utilities (4-digit SIC = 4900–4999). Our sample period is from 1985 to 2018. We generate financial statement data from Compustat Annual Files. Analyst coverage is generated from I/B/E/S. Stock returns are generated from CRSP Monthly Files. There are 29,203 firm-year observations in our sample when we use RDG_ROE as the measure of reporting bias and 27,242 firm-year observations when we use DIFROA as the measure of reporting bias. Calculation of DIFROA requires the data availability for R&D over the past six years (from years
t to
t − 5). Therefore, the sample size with DIFROA is slightly smaller than that for RDG_ROE.
4.2. Descriptive Statistics and Pearson Correlations
Table 2 presents the descriptive statistics. The mean and median values of analyst coverage (COVRGE) are 10.37 and 7.00, respectively. Our first measure of reporting bias, RDG_ROE (
Lev et al. 2005), is based on the difference between ROE and annualized R&D growth.
Lev et al. (
2005) separate firm-year observations into aggressive, neutral and conservative firms, using RDG_ROE. To prove that their measure captures reporting biases from expensing of R&D,
Lev et al. (
2005) document the fact that for firms that report aggressively, ROE under capitalization of R&D is smaller than that under expensing of R&D. However, for firms that report conservatively, they document the opposite relation (i.e., ROE under capitalization of R&D is greater than that under expensing of R&D). Based on these results, they conclude that RDG_ROE captures the reporting biases arising from expensing of R&D compared to capitalization of R&D. We develop our second measure of reporting bias, DIFROA, which is based on the difference in earnings in the cases of capitalization and expensing of R&D. DIFROA is designed to capture the reporting bias resulting from expensing of R&D compared to that when R&D is capitalized. When a firm’s R&D is growing (decreasing) over time, earnings under capitalization of R&D will be greater (smaller) than earnings under expensing of R&D. Therefore, DIFROA will be positive (negative). Hence, when R&D is growing (decreasing) over time, a firm is likely to report conservatively (aggressively).
The mean and median values of RDG_ROE are positive: 0.1732 and 0.0468, respectively, in
Table 2. Similarly, the mean and median values for DIFROA are positive: 0.0154 and 0.0071, respectively. These results suggest that, on average, expensing of R&D leads to conservative reporting compared to capitalization of R&D (because RDG_ROE and DIFROA are positive, on average). That is, on average, earnings under capitalization of R&D are greater than those under expensing of R&D, suggesting that expensing of R&D, on average, understates earnings compared to capitalization of R&D. Therefore, the descriptive statistics indicate that expensing of R&D, on average, leads to more conservative reporting compared to capitalization of R&D.
Table 3 presents the Pearson correlations. COVRGE is significantly correlated with all variables (the largest correlations of COVERGE are with LMV and VOLUME) in Equation (1), indicating the need to control for these firm characteristics in our analysis. There is a 0.35 correlation between RDG_ROE and DIFROA, suggesting that our measures of reporting bias resulting from expensing of R&D capture the same phenomena, to a large extent. RND has high positive correlations with RDG_ROE and DIFROA, suggesting that high-R&D firms have a high reporting bias arising from expensing of R&D (i.e., correlations between RND and RDG_ROE and DIFROA are both 0.39). There is a 0.40 correlation between STDROA and STDRET, indicating that our measures of uncertainty are highly correlated. There are also high correlations between RND and our measures of uncertainty, STDROA and STDRET, suggesting that uncertainty increases with R&D intensity (
Lev 2001;
Kothari et al. 2002). Moreover, there is high correlations between some of the cross-sectional determinants. For example, the correlation between LMV and VOLUME is 0.46. This finding raises the possibility that some of the cross-sectional determinants may not be independent. Hence, we need to perform a combined analysis of cross-sectional determinants to test whether one cross-sectional determinant subsumes the other.
6. Conclusions
Barth et al. (
2001) document a positive association between R&D and analyst coverage. However, they do not explore what particular attribute of R&D explains this positive association. In this paper, we attempt to fill this gap by investigating the cross-sectional determinants of the relationship between R&D and analyst coverage. We consider four cross-sectional determinants: reporting biases under expensing of R&D compared to capitalization of R&D, uncertainty associated with R&D, investors’ attention, and scale effects of R&D. We find that reporting biases resulting from expensing of R&D decrease the association between R&D and analyst coverage and that the effect is stronger when a firm reports conservatively than when it reports aggressively. In addition, we find that the uncertainty associated with R&D decreases the association between analyst coverage and R&D. We also find that investors’ attention and the scale effects of R&D increase the relationship between R&D and analyst coverage. When we perform combined analysis, we find that the impact of investors’ attention is largely mitigated, while the other cross-sectional determinants are not affected. Finally, we find that the scale effect of R&D seems to explain the positive association between R&D and analyst coverage documented by
Barth et al. (
2001).
Our findings have implications for both academicians and standard-setters. Our paper complements
Barth et al. (
2001) in two ways. First, we extend
Barth et al. (
2001) by investigating cross-sectional determinants of the relationship between R&D and analyst coverage and identify four cross-sectional determinants. Second, we are able to identify what particular attribute of R&D explains the positive relationship between R&D and analyst coverage documented by
Barth et al. (
2001). Furthermore, our findings add to the literature on the negative consequences of expensing of R&D. The prior research primarily focuses on the demand for accounting information (i.e., the informativeness or usefulness of accounting information for equity investors). We complement the prior research by documenting the negative consequences of expensing of R&D on the supply of information by financial analysts. While some researchers suggest that financial analysts might eliminate the information gap in the financial statements resulting from expensing of R&D (
Kimbrough 2007), our evidence suggests that analysts themselves are negatively affected by reporting biases. Therefore, they are less likely to eliminate the information gap in financial statements due to reporting biases resulting from expensing of R&D. Therefore, our findings suggest that the negative consequences of reporting biases could be more severe than those suggested in the prior literature. The U.S. is in the process of converging with IFRS, which allows the capitalization of R&D. Thus, our findings could provide useful insights to standard setters about the impact of expensing as compared to capitalizing of R&D.
In addition, our paper enhances the prior literature that investigates the implications of the uncertainty associated with R&D. Prior research suggests that the uncertainty associated with R&D affects many aspects of R&D firms such as stock and bond market valuations, insider gains and information asymmetry, value of analyst stock recommendations, stock return volatility, etc. We add to this literature by documenting the fact that the uncertainty associated with R&D also affects analyst coverage for R&D firms. Moreover, our paper enhances the existing literature about the scale effects of R&D. Prior research documents the fact that the scale effects of R&D affect productivity and future earnings associated with R&D investments. We document the fact that the scale effects of R&D also affect analyst coverage for R&D firms. Finally, our paper contributes to the prior literature about the impact of investors’ attention on high-technology R&D firms. While prior research focuses on the impact of investors’ attention on stock market valuation, we document the fact that investors’ attention also affects analyst coverage for R&D firms.
Based on our findings, we suggest the following practical recommendations for R&D-intensive firms. First, R&D-intensive firms should enhance the transparency of their R&D expenditures and projects, to mitigate the negative effects of reporting biases and uncertainty on analyst coverage and to reduce adverse impact of expensing R&D. Second, R&D-intensive firms should actively communicate the potential and progress of their R&D activities, to attract and retain investor attention and analyst coverage. Third, firms with significant R&D investments should highlight the scale effects and potential future earnings, to positively influence analyst coverage.
While our study sheds light on the cross-sectional determinants of the relationship between R&D and analyst coverage, several areas remain open for further exploration. Future research could build on our findings in the following ways. First, future studies could examine how the relationship between R&D and analyst coverage evolves over time, particularly in response to changes in accounting standards or economic conditions. This would provide insights into the dynamic nature of this relationship. Second, with the U.S. moving towards convergence with the IFRS, which allows the capitalization of R&D, it would be valuable to compare the impact of different accounting treatments on analyst coverage across different jurisdictions. Such comparative studies could inform policymakers and standard setters about the broader implications of accounting choices. Finally, future research could investigate whether the determinants we identified vary across different sectors, such as biotechnology, pharmaceuticals, and technology. Sector-specific analyses could reveal unique factors influencing analyst coverage in these industries.