Big data and financial reporting
Big data permeates almost all areas of business. Financial reporting is no exception. I discuss selected academic research that makes use of novel technologies and data.
Being cognizant what users can do – and actually do – is useful when designing the contents of financial reporting. New information technologies facilitate generating and accessing a host of new data and provide sophisticated tools to analyze and combine such data to find new associations and patterns. For users of financial reporting opportunities abound: investors use big data to make superior investment decisions, analysts to improve forecasts, auditors to assess audit risk, enforcement agencies to identify red flags, and researchers to find novel results. Here are two areas that recent account-
ing research has addressed.
Company performance tends to affect readability
It seems intuitive that the readability of a financial report has to do with the content it purports. One might suspect that readability is lower if a company faces poor performance because it tries to obfuscate unfavorable financial information. Alternatively, companies that experience bad results may feel pressure to better explain that performance. Psychological factors, such as attribution – justifying low performance by events outside management’s control –, can play a role. And impression management may be at work.
Simple measures of readability are the length of the report or the portion of boilerplate disclosures, which adds clutter, or asymmetric reporting about risks and chances. Linguistics has developed several measures of readability. A well-known measure is the fog index that counts words per sentence and the percentage of complex words. More advanced tools measure causation, exclusiveness, tone, and emotion through defined dictionaries.
Among others, research finds the following associations:
Financial reports of companies with low earnings are harder to read, and companies with easy-to-read reports have more persistent positive earnings
Declining performance and non-verifiable information, such as forecasts, come with a more positive and optimistic tone. Better corporate governance reduces these associations (Melloni et al., 2016).
Liquidity, institutional ownership, and analyst following are higher for companies with better readable financial reports (Lang and Stice-Lawrence, 2015).
– Companies with hard-to-read, complex financial reports provide more voluntary disclosures to “guide through the fog” (Guay et al., 2016).
Personal traits of executives influence reporting quality
Borrowing from psychology, one notion is that narcissistic executives are more willing to exploit their power and information asymmetry to engage in misreporting. Another one is that overconfident managers overestimate their abilities and skills and assign an inﬂated subjective probability to future outcomes. Studies find innovative ways to measure personal traits.
A study uses the prominence of the photograph of the CEO and his or her cash pay relative to second-highest paid executive and finds more narcissistic CEOs tend to overstate reported EPS (Olsen et al., 2014).
Using the size of the CFO’s signature in the annual report, a study finds that more narcissistic CFOs are associated with more earnings management, less timely loss recognition, weaker internal controls, and
a higher probability of restatements (Ham et al., 2015).
Measuring CEO overconfidence by their willingness to hold deep-in-the-money options close to maturity, a study finds that overconfidence is positively related with earnings management and benchmark-beating (Hsieh et al., 2014).
– Not exactly a trait, but nevertheless exciting, a study uses CFOs’ golfing records as proxy for leisure consumption and finds that financial reports of companies that have CFOs with higher golfing ranks are of lower quality (Biggerstaff et al., 2016).
Big data as the new normal
Not all studies make a lot of sense. Often do they lack theoretical backing, use questionable proxies, and fail to establish causality. They are typically large-sample studies with low explanatory power. Yet big data make it dramatically easier to explore novel associations. Preparers of financial reports should be sensitized of unconventional uses (and misuses) of corporate financial information.