Courtesy of Daniel Blaseg and Christina Bannier
Issuers of financial securities must disclose vast amounts of information to capital markets. The increasing length and complexity of security prospectuses is an unambiguous sign of this development. To raise the comprehensibility of these documents, financial regulators around the world have started to require that issuers release an additional summary. But do these short-form disclosure documents give an accurate and comprehensive representation of the full prospectus’ information? And irrespective of whether they do or not, can investors glean helpful information from the summaries? These are the questions that we try to answer in a new paper. Along the way, we develop an automatic approach for evaluating the quality of summaries in securities prospectuses.
Empirical methodology and data
We use the German small and medium-sized enterprise (SME) bond market as empirical context. This market offers several benefits for our analysis: First, the relatively high number of actual defaults in this market segment allows us to directly differentiate between low- and high-quality bonds instead of relying on estimated default probabilities. Second, SME bonds have been aggressively marketed among retail investors, who are a major focus group for financial regulators. We consider 159 SME bonds issued between 2010 and 2016 with a total issuance volume of more than EUR 7.4 billion.
Our analysis focuses on the risk factor section of SME bond prospectuses as this has been shown to be one of the most important sections from investors’ point of view. While the risk factor section itself is highly regulated in Europe, the requirements for the summary are less strict. Most importantly, there are no requirements regarding the specificity and materiality of content. This enables issuers to decide how and which factors to summarize. Moreover, there is no civil liability for misstatements based solely on the summary, unless the summary is misleading, inaccurate, or inconsistent when read together with the full prospectus. This comparably light regulation and the low litigation riskclearly pose incentives for executives to engage in impression management in the prospectus summary, i.e. to present the security in a way that is intended to manipulate investors’ perceptions and subsequent decisions.
To address our research questions, we measure the quality of the risk factor summary as a combination of the semantic and linguistic quality of the summarized text compared to the full risk-factor section. The closer the summary comes to the full text with regard to these two measures, the higher its quality. Whereas the semantic quality refers to the divergence between the probability distributions of words in the summary and the full text, we measure the linguistic quality based on four different aspects: readability, use of boilerplate, specificity, and tonality. Readability is derived from the average number of characters and syllables per word and the average number of words per sentence. Boilerplate measures the unnecessary repetition of facts within a text. Specificity refers to the number of specific terms or entity names in a text. Tonality evaluates positive and negative sentiment in relation to each other. We employ elaborate, well-established methods to evaluate each of these individual criteria and make sure that we derive robust assessments of textual quality. We then combine the semantic and linguistic quality proxies into a single measure of summary quality in a way that allows us to achieve a high correlation with human evaluation of quality as “gold standard”. For this, we recruit participants from the Clickworker platform (the German equivalent of Amazon’s Mechanical Turk) to evaluate the quality of the summaries manually.
Once we have established this automatic process for measuring a summary’s quality, we employ multivariate approaches to examine whether the quality is associated with various characteristics of the issuing firm. In addition, we investigate how retail investors react to the provision of different quality levels of summaries to gain insights into the incentives for managers preparing the summaries. Finally, we study the informativeness of the summaries by analyzing the relationship between a summary’s quality and the security’s subsequent performance.
Results
Our findings suggest, perhaps unsurprisingly, that poorly-performing firms use the summaries to engage in impression management. These firms strategically select, frame, and present the information in the summary in a way that is intended to positively bias investors’ perceptions and decisions. More precisely, we find that the lower a firm’s return on assets and the higher its implicit default risk, the lower is the readability of the risk factor summary, the more boilerplate is used, the higher is the tonality and selectivity of content compared to the full risk factor section in the prospectus.
Unfortunately, our next set of results indicates that investors appear unable to detect this reporting bias in the risk factor summaries of poorly-performing firms. Rather, investors react positively to the weak summary quality so that these firms are able to reap higher funding amounts over a shorter funding period. Furthermore, we show that these firms are generally perceived as more positive despite their low summary quality. This surprisingly positive investor reaction may be explained by the fact that summaries of lower quality tend to be more positive and diverge more strongly from the full risk factor section, leading them to be easier to understand. At the same time, the lack of an appropriate, negative market response to the release of weak summary qualities from poorly-performing firms demonstrates the incentives for managers to engage in impression management: Not only is litigation risk from biased reporting small, investors also do not seem to punish managers for doing so.
In a final test, we examine whether the summary quality may nevertheless be informative for investors in the sense that it allows inferences to be drawn about the future market performance of the bonds (if understood correctly). Indeed, our results indicate that firms issuing high-quality summaries are less likely to default on their bonds: A one standard deviation increase in the overall quality score of a summary reduces the likelihood of default by around 10.6%. Even though retail investors appear to turn a blind eye to managers’ attempts at presenting their securities in an overly optimistic light, the summary quality turns out to be highly predictive of the future security performance and, as such, could potentially help to improve investment decisions significantly.
Robustness
Our analyses take great care in controlling for a host of additional factors that might affect the relation between firm performance and summary quality on the one hand and summary quality and bond performance on the other. However, there might still be unobservable characteristics that render the latter relation endogenous so that our results may not be interpreted in a causal sense. In order to reduce these concerns and support the causality of our results, we run an instrumental variable analysis as one further robustness check. In a two-step estimation procedure, we instrument the summary quality via litigation risk, measured by the number of fine proceedings issued by the Federal Financial Supervisory Authority during the six months before the bond emission. These additional analyses indeed confirm the robustness of our results.
Key findings
In this study, we use multiple methods to evaluate summarized information in security prospectuses and examine the informativeness of these documents. In particular, we analyze (1) how firms summarize information by developing an automatic approach for evaluating the quality of a summary, (2) the determinants of the summary quality and how retail investors react to different qualities in summaries, and (3) whether the evaluation of the quality of summaries can help to improve investment decisions.
Our findings suggest that poorly performing firms use risk factor summaries to engage in impression management. Such firms strategically select, frame, and present textual information in disclosure documents in a way that is intended to manipulate investors’ perceptions and decisions. The manipulations include statistically significant differences in readability, specificity, tonality, use of boilerplate, as well as selectivity of content. We show that our automatic evaluation method of summaries correlates highly with human evaluations. Further, we demonstrate that investors are not able to detect reporting biases in summaries; on the contrary, they react very positively to low-quality summaries of poorly performing firms. Nevertheless, we can show that discretionary disclosure practice in prospectus summaries is informative for investors, as the summary quality helps to predict the future security performance. It is therefore important for investors to learn to correctly assess a summary’s quality, with particular caution in the case of issuers with low performance.
Conclusion
Our analysis contributes to the literature on financial disclosure in several ways. To the best of our knowledge, we are the first to empirically analyze the use of impression management strategies in summarized information, and to relate firms’ risk-disclosure practices in summaries to future performance. In doing so, we develop a reproducible approach for measuring the quality of a summary. Furthermore, our study informs policymakers and regulators with regard to the effects of summarization in financial disclosures.
The findings suggest that the current regulatory scheme in the European Union leaves room for manipulation. Our work supports recent calls for higher quality in the summarized presentation of risk factors in prospectuses. Moreover, our results provide investors with helpful knowledge on how to evaluate textual disclosure. Given the tremendous amount of information in securities markets, summaries will become increasingly important, and investors need to learn how to assess the summarized information.
