Reading between the lines: An empirical examination of qualitative attributes of financial analysts’ reports

Abstract

This paper examines whether two qualitative attributes of financial analysts’ reports, detail and tone, are significant in explaining how the market responds to analysts’ reports, after controlling for the information contained in the reports’ quantitative summary measures. Report detail is hypothesized to reflect the level of effort expended by the analyst in preparing the report, and therefore the usefulness of their intrinsic firm value estimates. Report tone is predicted to signal the analyst’s underlying sentiment regarding the firm and may be used to assess the extent to which analysts’ conflicts of interest interfere with the mapping of firm value estimates into stock recommendations. Consistent with these hypotheses, we find that the tone of financial analyst reports contain significant information content incremental to the reports’ earnings forecasts and recommendations, and report complexity (one component of report detail) helps explain cross-sectional variation in the market’s response to the reports’ recommendations.

Introduction

Sell-side financial analysts have been the subject of extensive empirical and experimental research in accounting. The high demand for this research stems from several sources; including, the need for earnings expectations proxies in market research, the wide use of their outputs across the investment community, and the fact that analysts can be used as investor proxies to address interesting questions about the usefulness of accounting data. The significant role of financial analysts in evaluating accounting data and disseminating their analysis to the public ensures that they will continue to attract substantial interest from investors and academic researchers. Our knowledge of the influence of accounting information in capital markets is enhanced as we gain a deeper understanding of how financial analysts process accounting data, how they transmit the results of their analysis to investors, and how the market reacts to their reports (Bradshaw, 2009, Huang and Zhang, 2011). In this study, we contribute to research on the outputs generated by analysts to communicate their private information, and the investor response to these outputs.

When examining outputs from analysts, researchers have historically focused on two quantitative summary measures included in the majority of analysts’ reports: earnings forecasts and stock recommendations.² However, as the following quote expresses, analyst reports consist of more information that presumably supports their ultimate recommendation:

“Stock ratings and price targets are just the skin and bones of analysts’ research. The meat of such reports is in the analysis, detail, and tone” (Tsao, 2002; emphasis added).

If all information contained in analysts’ reports is captured by the stock recommendation and/or earnings forecast, then the other content (which make up the majority of the report) becomes unimportant. However, given the significant effort expended by analysts in producing and disseminating their reports, we conjecture that incremental information can be obtained by examining the contents of the reports, including an analysis of non-quantitative information (i.e., detail and tone). For example, analysts are generally constrained to five broad categories when issuing stock recommendations; strong buy, buy, hold, sell, and strong sell, and the most negative categories are rarely used (Barber et al., 2006). Thus, the detail and tone of analyst reports may provide investors with a valuable source of additional information. Indeed, Ramnath et al. (2008) argue that “examining analyst reports based solely on quantitative information may not capture the complex nature of the analyst’s task” (p. 9). Just as prior research has examined the incremental value-relevance of separate earnings components to assess the quality of the summary measure EPS (e.g., Lipe, 1986, Dechow, 1994), in this study, we examine qualitative attributes of analyst reports to assess whether they contain incremental information about the quality of the reports’ summary measures – EPS forecasts and stock recommendations.³

Our sample consists of 2057 analyst reports published in the year 2006 that initiate coverage of a specific firm. We exploit innovations in the content analysis and computational linguistics fields to develop objective and reliable measures for the reports’ detail and tone – the second and third components of the “meat” of the reports listed in the above quote. Report detail is broadly defined as anything that may signal to investors the amount of effort the analyst expended in preparing the report. Due to its multiple dimensions, three aspects of the report are used to assess detail: complexity, length, and visual aids. Report complexity is measured using the Fog Index, a commonly used measure of writing sophistication. The second and third measures of detail are the length of the report and the number of visual aids, including charts, tables, and graphs, used by the analyst. Report tone is measured using the content analysis software General Inquirer, and is intended to represent the analyst’s underlying opinion of firm value.⁴

We hypothesize that report detail and tone provide capital market participants with insights regarding the value of analysts’ earnings forecasts and stock recommendations. Beyond the significant costs expended by analysts to produce their reports, additional support is provided for this hypothesis from evidence in prior research on analysts’ behavior and incentives. Several studies document results consistent with analysts’ herding behavior (e.g., Trueman, 1994, Hong et al., 2000, Clement and Tse, 2005), which is the tendency for analysts to issue forecasts or recommendations close to the consensus. An analyst who lacks sufficient private information to produce an accurate forecast or recommendation, either through lack of effort or ability, is more likely to mimic forecasts issued by strong analysts. We argue that the level of detail in a report represents an analyst’s attempt to convey to clients that he has invested the necessary effort to produce a credible opinion. Thus, we predict a more pronounced market response to the summary outputs of a report (i.e., forecasts and recommendations) when they are supported by more detail in the report (an interaction effect).

With respect to tone, anecdotal and empirical evidence suggests that analysts’ stock recommendations are often compromised by conflicts of interest. Analysts are influenced by their own compensation structures and will tend to issue more favorable recommendations to facilitate potential underwriting relationships (Groysberg et al., 2011). In addition, analysts have been accused of issuing overly positive recommendations in order to curry favor with management, which presumably offers the analyst greater access to management’s private information (Das et al., 1998, Libby et al., 2008, Mayew, 2008). Michaely and Womack, 1999, Barber et al., 2007 demonstrate that these conflicts of interest adversely affect the quality of analysts’ stock recommendations. Given these conflicts of interest, report tone may better reflect the analyst’s underlying sentiment about the firm, and therefore can be used as an additional piece of information to assess the analysts’ underlying opinion of the firm. We examine the effect of tone on investor response to analyst reports as a main effect and interacted with the reports’ quantitative summary measures. A significant positive coefficient on the tone interaction effect indicates that tone provides investors with incremental information that is supportive of the quantitative summary measures; whereas a significant positive main effect suggests that tone by itself provides value-relevant information.

The above arguments suggest that investors will assess higher levels of credibility to outputs associated with more detailed reports, and that positive report tone will result in a favorable market response. However, an alternative argument could also be made for the opposite result. Specifically, less-informed analysts could use detail in their reports as a means to obfuscate their lack of firm-specific information, and report tone could reflect analysts’ cognitive biases and linguistic preferences. While our ex-ante expectation is that these alternative arguments are not as persuasive as the above predictions, the validity of our hypotheses is ultimately an empirical question.

Our results are consistent with our hypotheses that the detail and tone of financial analyst reports provide the market with incremental information content beyond the reports’ quantitative summary measures. Specifically, one component of report detail; namely, report complexity, causes a stronger market reaction to the news contained in the analyst’s stock recommendation, as measured by three-day size-adjusted returns centered on the reports’ publication dates. This interaction effect suggests that report complexity provides support to the quantitative measures issued by analysts. We also find that more optimistic reports are associated with larger returns independent of any quantitative news provided by the reports. This main effect indicates that investors view report tone as a valuable source of incremental information by itself, regardless of the recommendation or forecast contained in the report.

In supplementary analysis, we examine whether abnormal returns can be earned through long-term trading strategies based on report detail and tone, controlling for the analysts’ stock recommendations and firm characteristics. These investment portfolio tests provide evidence on whether the market fully impounds in prices the valuation implications of report detail and tone at the time they are disclosed. If the market fails to properly value this information, then trading strategies that exploit this knowledge may yield abnormal returns. Our evidence suggests that abnormal returns cannot be earned through long-term trading strategies based on these qualitative attributes, which supports the notion that investors recognize and correctly respond to the value of these attributes when the reports are published.

Our results are robust to a battery of sensitivity tests. We employ two alternative measures of report complexity, as well as an alternative measure of tone designed specifically for use in a financial context, and find qualitatively similar results. We also continue to find similar results when we examine “abnormal” levels of report complexity and tone after controlling for the average levels of these qualitative report attributes across industries and recommendation levels.

Our study is related to Asquith et al. (2005), who manually examine 1126 analyst reports and find that the written analysis made by analysts to support their opinions provides incremental information content to the market. Our study complements and extends Asquith et al. along several dimensions. First, whereas Asquith et al. examine specific analysis contained in analyst reports, we study the other two distinct qualitative attributes that comprise the “meat” of financial analyst reports; detail and tone (see above quote by Tsao). Second, our measures of detail and tone are objective and amenable to replication, and avoid any bias that might be inherent in human classification.⁵ Third, our measures are based on an analysis of the entire report instead of focusing on only a few elements contained therein.

This study has implications for investors and researchers interested in the process through which financial analysts transmit their private information to their clients, as well as how the market reacts to their reports. By providing evidence that investors are concerned about not only the quantitative news contained in the reports of information intermediaries, but also the qualitative content of their reports, this study should be of broad public interest. Additionally, extant research has documented situations where the stock return around an information event and the quantitative news provided by the event are of opposite signs, particularly with regards to firms’ earnings announcements (Kinney et al., 2002, Johnson and Zhao, 2011). By examining the “soft” information contained in one frequently studied information event, namely the publication of financial analyst reports, we provide evidence regarding one potential determinant of this interesting phenomenon. The study should also be of interest to analysts because it shows that levels of report complexity and tone affect how investors interpret and respond to their reports.

The remainder of the paper proceeds as follows. Section 2 develops hypotheses regarding the information content of these qualitative report attributes. The empirical measures of detail and tone, as well as the research design, are introduced in Section 3. Section 4 presents the empirical results related to our main tests, and Section 5 describes additional analyses. The final section presents our conclusions.