Processing Fluency and Investors’ Reactions to Disclosure Readability

Recent work in the archival accounting literature investigates disclosure readability (Li [2008], You and Zhang [2009]) and its effects on the behavior of small investors (Miller [2010]). In my paper, Processing Fluency and Investors’ Reactions to Disclosure Readability, forthcoming in the Journal of Accounting Research, I use a controlled experiment to provide complementary evidence and to address questions that cannot be answered with archival data. I find that, holding constant length and information content, more readable disclosures lead to stronger reactions from investors, so that changes in investors’ valuation judgments are more positive when news is good and more negative when news is bad. Consistent with prior literature in psychology, I find that participants are more likely to feel as though they can rely on a disclosure that is more readable, and that this effect is mediated by processing fluency. Processing fluency represents how easy it feels to process a given piece of information, and is often subconsciously treated as a heuristic cue that information can be relied upon when making related judgments. In my study, the greater reliance on news (be it good or bad) helps to explain the stronger reactions to more readable disclosures that I observe, even though readability does not lead to significant differences in the actual information that participants are able to gather from the press release. Counter to my predictions, I do not find that disclosure readability directly affects perceptions of management credibility. However, I do find evidence of an indirect effect operating through feelings of processing fluency.

In Stage 2 of my experiment, I explicitly make participants aware of the potential for variation in disclosure readability. I find that participants who initially received the more readable disclosures provide valuation judgments in Stage 2 that are less extreme than the valuation judgments they provide in Stage 1, to the point where the effect of readability on investors’ reactions is no longer significant. These results suggest that the stronger reactions in response to more readable disclosures in Stage 1 of my experiment may have been unintentional. This is consistent with prior research showing that individuals do not treat processing fluency as a cue in their judgments once they realize its source (Schwarz [2004], Alter and Oppenheimer [2009]). This result has potentially important implications for the SEC’s push towards improving disclosure readability. While the SEC describes more readable disclosures as though they are unambiguously positive [SEC 1998], my results indicate that more readable disclosures may actually lead investors to overreact to information, particularly those who are the least sophisticated. In other words, investors may be too quick to accept information that is presented in a way that is easy to process. This suggests that the benefits of more readable disclosures may be less clear-cut than has been argued by the SEC [SEC 1998] and in prior accounting literature (e.g., Li [2008], You and Zhang [2009], Miller [2010]).

My findings add to the growing body of literature investigating the style of disclosures as opposed to their content. Content is the literal meaning of the information conveyed, or the concrete facts contained in a disclosure, whereas style captures the methods used to convey meaning to the audience. In response to Core’s [2001] call for greater computational linguistic processing of qualitative disclosures, stylistic factors besides readability have been investigated in the archival literature, including certainty (Demers and Vega [2010]), numerical intensity (Henry [2008]), boilerplate vs. meaningful language (Nelson and Pritchard [2007]), and tone (Davis, Piger and Sedor [2011]). However, experimental investigation of stylistic characteristics is relatively scarce in the recent accounting literature. One exception is Hales, Kuang and Venkataraman [2011], which investigates the effects of vivid vs. pallid language, above and beyond actual information content. Other recent work has looked at the determinants of the style characteristics of verbal communications (Hobson, Mayew and Venkatachalam [2012]), as well as reactions to verbal communications (Mayew and Venkatachalam [2011], Elliott, Hodge and Sedor [2012]).

A limitation of my study is that it is not clear whether investors’ reactions differ in response to individual linguistic and formatting choices. I made this design choice to produce a strong treatment effect. My treatments also correspond with real-world disclosures, because the use of either high- or low-quality linguistic and formatting features is likely to be highly correlated in a given disclosure. However, there is substantial opportunity for future experimental research investigating how individual stylistic features of a disclosure affect investors’ judgments. Experiments are also likely to be well-suited to the investigation of managers’ disclosure decisions (to investors, but also to auditors, other employees, clients, etc.). Sociolinguistics research supports the prediction that choice of linguistic characteristics may vary in line with specific goals and incentives. Based on these findings, future work might investigate whether managers intentionally or unintentionally use personal pronouns and other linguistic features to influence others’ reactions as firm performance and incentives vary (Chatterjee and Hambrick [2007]).

My study also adds to the literature on processing fluency, and suggests that it may be an important mechanism for both understanding results in prior studies and motivating predictions in future research. While processing fluency is mentioned only rarely in the existing accounting literature (e.g., Koonce and Lipe [2010]), a number of prior findings are consistent with outcomes that would be predicted by research on processing fluency. In their review, Alter and Oppenheimer [2009] point out that all tasks range from effortless to demanding, and that there are a number of sources of processing fluency in a given task. As a result, processing fluency effects are likely to be pervasive across accounting settings. The availability of information, or ease of retrieval, is just one manifestation of processing fluency (Alter and Oppenheimer [2009]). Thus, accounting studies showing that professional and nonprofessional investors do not spontaneously recall or consider unfavorable information about a firm shed light on how processing fluency might affect investor behavior (e.g., Kadous, Krische and Sedor [2006], Krische [2005]). Similarly, conceptual processing fluency relates to the ease with which individuals access relevant (and sometimes irrelevant) knowledge structures to complete a task (Schwarz [2004]). Findings in Hopkins [1996] can perhaps then be thought of as evidence that financial statement classification affects conceptual processing fluency by making category-relevant information more or less accessible.

The effects of processing fluency also represent a potentially fruitful opportunity for future research. For instance, the use of “jargon” is likely to feel more fluent to those with more experience in financial reporting settings, suggesting important experience effects of processing fluency. Furthermore, processing fluency research might also suggest potential remedies for biased decision-making in accounting settings. Alter, Oppenheimer, Epley and Eyre [2007] show that in some circumstances disfluent presentation can lead to more systematic processing, presumably because disfluency can act as a signal that greater effort should be expended on the task. This suggests that processing fluency could be manipulated in a variety of ways to induce relatively more cautious or aggressive behavior among managers, investors, auditors, etc.

Finally, my study uses a source of participants that may prove useful for future research. The Amazon.com Mechanical Turk (AMT) platform provides access to more participants than would typically be available in a laboratory setting, and at a substantially lower cost. Furthermore, AMT has controls in place to help alleviate some of the typical concerns raised with online experiments (as discussed in Bloomfield and Rennekamp [2009]). On average, participants in my study have some experience with investing, have taken some accounting and finance courses, have more than a decade of full-time work experience, and take the task seriously enough to correctly answer questions that check for their attention. Future work should consider utilizing AMT (or other online crowd-sourcing platforms) for accounting research, particularly with respect to the judgments of less-sophisticated investors.

The full paper is available for download here.