Since 2005, the Securities and Exchange Commission (SEC) mandates that all publicly-traded firms inform investors about material risks that may impact future performance. These are referred to as risk factor disclosures, which are discussed within Item 1A of 10-K filings. These risk disclosures have been criticized as lengthy and boilerplate (IRRC 2016; SEC 2016); it is not uncommon for such disclosures to exceed 20 pages in length. Assessing whether risk disclosures are useful to investors is challenging, particularly due to inherent difficulties in measuring and quantifying text: Researchers historically have applied coarse measures such as word or sentence counts and associated these with measures of risk (e.g., stock return volatility).
Our study differs from prior research in two key ways. First, we innovate textual measurement of risk factor disclosures. Second, we use a research design providing a strong theoretical link between the firm’s disclosure of risk (the risk factors) and a market measure of the firm’s risk (the so-called variance risk premium, or VRP).
Recent theory predicts that risk factor disclosures should lower uncertainty about firm risk. Note, this is distinct from risk itself. “Risk” relates to the variance of outcomes a firm faces, whereas “uncertainty about risk” relates to investor uncertainty as to the nature of the risks a firm faces. Theory argues that information about exposure to risk factors (i.e., what risks a firm faces) helps investors assess the uncertainty about the risk of a firm’s future cash flows.
Testing this prediction requires two key proxies: one for a risk factor disclosure and one for uncertainty about risk. To measure risk factor disclosures, we use the year-over-year additions and removals of individual risk factors within 10-K filings. To measure investors’ beliefs about uncertainty regarding risk, we use the VRP, which utilizes data from traded option contracts and represents an estimable value directly related to the uncertainty about risk.
The traditional measurement of risk disclosures relies on word or sentence counts (i.e., the so called “bag-of-words” approach). For example, the nominal amount of textual information in a risk disclosure can be assessed by counting the total number of words, sentences, or subsets of keywords within the same risk disclosure. Since our goal is to extract information about firm’s exposure to individual risk factors, we alternatively measure individual risk factors by exploiting the SEC disclosure requirements within Item 1A. We choose Item 1A because: (i) It is the only comprehensive source of all material risks the firm faces; (ii) it benefits from managers’ private information that can help identify and characterize the firm’s risks in a way that external users are unlikely able to replicate on their own; and (iii) the SEC mandate requires that each disclosed risk factor be labeled with a descriptive heading.
Our analyses exploit this last disclosure requirement through a textual analysis algorithm. Specifically, the SEC requires that each individual risk factor be organized under a relevant subcaption. Using Microsoft’s 2017 10-K, this is the first risk factor heading: We face intense competition across all markets for our products and services, which may lead to lower revenue or operating margins. This caption typically is a lead sentence summarizing the nature of the risk, usually highlighted through bold-face, italics, or numerical listing. We use textual analysis techniques to identify each risk caption heading in firms’ 10-K filings. We then use string similarity algorithms to compare the text of the individual risk factor captions from two adjoining years. This enable us to isolate individual risk factors that have been added from those that have been removed relative to the previous year. We assume that additions and removals of risk factors reflect management beliefs about changes in the exposure to the various risks a firm faces. Therefore, these additions and removals serve as a proxy for the risk factor signals we wish to measure.
We build on prior research, which studies the association between narrative risk disclosures and measures of risk (e.g., stock return volatility, beta) by focusing on uncertainty about risk as opposed to risk per se. Specifically, we use the “variance risk premium” or “VRP.” As the name suggests, VRP can be thought of as a premium (or compensation) that option traders require to bear variance risk. In theory, the VRP arises only if there is uncertainty about risk itself. Accordingly, we use the VRP as our measure of market participants’ beliefs about uncertainty regarding the risks faced by the firms.
Using a broad cross-sectional sample of US-listed firms spanning 2006-2019, we document a lower VRP for firms exhibiting higher additions and removals of individual risk factor disclosures. Restated, this finding suggests investors use changes in the disclosure of individual risk factors in firm’s 10-K reports to update their beliefs about the uncertainty around the risk that firms face. Further, it suggests a way to circumvent the alleged boilerplate limitations of these risk factor disclosures by isolating risk factors that are newly added or removed.
Our research design minimizes potential biases in two key ways. First, we employ recent matching procedures—matching firms exhibiting “high” to firms exhibiting “low” additions and removals of risk factors on several performance and risk dimensions. This ensures that the effect we observe is directly attributable to the risk factor disclosures as opposed to other characteristics. Second, we control for traditional measures of textual risk disclosures used in prior research (such as word counts). Of note, these latter aggregate textual measures fail to exhibit associations with the VRP. This suggests our textual measure captures a distinct dimension of information content. Combined with our primary findings, this suggests that additions and removals of individual risk factors provide market participants with insights into the risk exposures faced by the firm beyond aggregate text disclosure measures (such as word counts).
In additional analyses, we also find that the reduction in VRP is greater for firms that are larger or file their 10-K reports earlier in the fiscal year. We further document that our findings hold for separate analysis of both risk factor additions and removals. We also show that the effect is greatest for risk factors exhibiting the largest changes in materiality, which we measure using risk factors that move the most places within the risk factors section from one year to the next (i.e., exhibiting the largest “jumps” in position).
Our work offers new insights regarding the decision-relevance of risk factor disclosures: Company-sourced risk signals not only map onto future realized variance, but also reduce, ex ante, uncertainty surrounding future risks. Critically, our findings suggest that individual risk factors allow for a more complete characterization of the information underlying risk disclosures, as the effects we document appear incremental (and seemingly unique) compared to more commonly-applied aggregate textual measures (like word counts) and non-text measures (like volatility of income). Our analyses also highlight the potential of using our setting (i.e., Item 1A of 10-Ks) to better understand the information content of risk disclosures. In fact, our study is the first to explore characteristics of individually disclosed risk signals such as the change in their relative positioning within the risk factor section. Overall, we highlight a new dimension of decision-relevance for risk factor disclosures—the textual analysis of individual risk factors—and demonstrate the importance of their granular measurement to researchers, regulators, and practitioners.
Mathew R. Lyle is an Associate Professor of Accounting Information and Management in the Kellogg School of Management at Northwestern University.
Eddie Riedl is the John F. Smith Professor of Management, Professor of Accounting, and Chair of the Accounting Department in the Questrom School of Business at Boston University.
Federico Siano is an Accounting Ph.D. Candidate in the Questrom School of Business at Boston University.
This post is adapted from their paper, “Changes in Risk Factor Disclosures and the Variance Risk Premium,” available on SSRN
The views expressed in this post are those of the author and do not represent the views of the Global Financial Markets Center or Duke Law.