Knowledge that Transforms

This study examines whether we can learn from the behavior of blockchain‐based transfers to predict the financing of terrorist attacks. We exploit blockchain transaction transparency to map millions of transfers for hundreds of large on‐chain service providers. The mapped data set permits us to empirically conduct several analyses. First, we analyze abnormal transfer volume in the vicinity of large‐scale highly visible terrorist attacks. We document evidence consistent with heightened activity in coin wallets belonging to unregulated exchanges and mixer services—central to laundering funds between terrorist groups and operatives on the ground. Next, we use forensic accounting techniques to follow the trails of funds associated with the Sri Lanka Easter bombing. Insights from this event corroborate our findings and aid in our construction of a blockchain‐based predictive model. Finally, using machine‐learning algorithms, we demonstrate that fund trails have predictive power in out‐of‐sample analysis. Our study is informative to researchers, regulators, and market players in providing methods for detecting the flow of terrorist funds on blockchain‐based systems using accounting knowledge and techniques.

Using an international sample of 2,113 initial coin offerings (ICOs), we explore the role of disclosure and information intermediaries in the unregulated crypto‐tokens market. First, we document substantial cross‐sectional variation in the voluntary disclosure practices of ventures seeking to raise capital through ICOs, such as the extent of information released in a prospectus‐type document called a white paper; releasing the technical source code; and communicating through social media platforms. Second, we find that, even with limited disclosure verifiability, ventures with higher levels of disclosure have a greater ability to raise capital. Finally, we find that this association is stronger in the presence of mechanisms that lend credibility to ventures’ voluntary disclosures, such as internal governance practices or external scrutiny from information intermediaries. Overall, our results suggest that voluntary disclosure and information intermediaries facilitate the functioning of ICOs as an alternative capital market.

We use machine learning methods and high‐dimensional detailed financial data to predict the direction of one‐year‐ahead earnings changes. Our models show significant out‐of‐sample predictive power: the area under the receiver operating characteristics curve ranges from 67.52% to 68.66%, significantly higher than the 50% of a random guess. The annual size‐adjusted returns to hedge portfolios formed based on the prediction of our models range from 5.02% to 9.74%. Our models outperform two conventional models that use logistic regressions and small sets of accounting variables, and professional analysts’ forecasts. Analyses suggest that the outperformance relative to the conventional models stems from both nonlinear predictor interactions missed by regressions and the use of more detailed financial data by machine learning.

We examine how tax cuts that benefit some firms are related to the economic performance of their direct competitors. Consistent with tax cuts decreasing the cost of initiating competitive strategies, we find that a decrease in the tax burden for only a specific group of firms in the U.S. economy (i.e., “rivals”) has a negative economic effect on the performance of its direct competitors not directly exposed to the same tax cut (i.e., “competitors”). This negative externality is stronger when the relatively higher taxed competitors (1) are financially constrained, (2) operate in more competitive markets, (3) have similar products to their lower taxed rivals, (4) face rivals that retain more of their cash tax savings due to lower dividends and share repurchases, and (5) face lower taxed, but financially constrained, rivals. We also find that shareholders and lenders price the negative externality manifested in these competitors’ economic performance.

We develop a measure of how information events impact investors' expectations of risk. The measure is broadly applicable and simple to implement. We derive it from an option‐pricing model, where investors anticipate an announcement that simultaneously conveys information on the announcer's expected future cash flows and risk profile. We empirically implement the measure using firms' earnings announcements, showing that it closely aligns with our model's predictions and offers strong forecasting power for firms' risk profiles, costs of capital, and future investments. We further highlight pitfalls of using simple changes in option‐implied volatilities to study information gleaned from earnings announcements. Finally, we apply our measure to study disclosure regulation, the efficacy of text‐based proxies, and market‐wide events, which we use to illustrate our measure's uses, and illuminate its potential limitations.

This paper investigates the effect of mandatory disclosure requirements for private firms on their decision to go public. Using detailed project‐level data for biopharmaceutical firms, we explore the effects of a legal reform that exogenously required firms to publicly disclose information regarding clinical trials. Exploiting cross‐sectional heterogeneity in firms' exposure to the regulation based on their internal development portfolios, we find that affected firms are significantly more likely to transition to public equity markets following the reform. Moreover, firms that go public because of the increased disclosure requirements subsequently reduce the size of their project portfolios while shifting to safer investments acquired externally. We provide additional evidence for the main hypothesis using a second setting: a 2006 German reform which enhanced the enforcement of mandatory disclosure requirements for private firms. The results suggest that private firms' general information environment and disclosure requirements influence the propensity of going public.

Could real‐time big data help unravel material firm events? How would it compare with firm disclosure and traditional media in terms of timeliness and completeness? Could big data provide incremental value‐relevant information for investors? With these questions in mind, we use a novel data set of cell phone “pings” (i.e., geolocation signals from mobile devices) to track production disruptions (outages)––material events for U.S. oil refineries. We first validate the construct by examining the effects of outages on local gas prices and firms’ accounting performance. Our main analyses show that (1) refining firms do not voluntarily disclose refinery outages identified by cell phone pings; (2) traditional media cover only a small portion of ping‐based outages; (3) the stock market finds ping‐based outages to be value relevant but incorporates the information with delay. Further analysis suggests that given the incomplete media coverage and lack of firm disclosure, investors appear to learn the financial impact of such outages through subsequent earnings announcements. Our evidence has implications for regulators such as the U.S. Energy Information Administration and the Securities and Exchange Commission as they continue to evaluate both the compliance and usefulness of disclosures for material firm events such as production disruptions.

In September 2009, Thomson Reuters (TR) discontinued its practice of relying on analysts to determine the treatment of unexpected charges and gains in favor of their immediate exclusion from GAAP earnings. Adopting a difference‐in‐differences approach, we show that this plausibly exogenous change in TR's methodology resulted in street earnings that are more predictive of future performance; and timelier, more accurate, and less dispersed analyst forecasts of future earnings, consistent with TR enhancing the properties of street earnings and analyst forecasts. Finally, using path analysis we show that a significant portion of TR's effect on price discovery is through its effect on analysts; and that the change in TR's treatment of unexpected items increased (decreased) the relative influence of TR (analysts) on the pricing of street earnings. We conclude that forecast data providers like TR are more than a conduit of information from analysts to investors.

We investigate whether non‐GAAP reporting affects the audit process and thereby the quality of the related financial statements. First, we provide evidence that auditors in numerous countries, including the United States and the United Kingdom, rely to varying degrees on non‐GAAP profit before tax as a benchmark for determining quantitative materiality. Then, using Premium Listed companies on the London Stock Exchange, we document that U.K. auditor reliance on non‐GAAP materiality benchmarks often results in a higher quantitative materiality amount and can lower audit quality. Although U.K. auditors appear skeptical of managers’ more aggressive non‐GAAP adjustments, auditors adopt more of management's low‐quality adjustments when auditor independence is weaker. In sum, our results suggest that non‐GAAP reporting can indirectly affect investors by reducing the rigor of the financial statement audit.

Effective financial reporting requires efficient information processing. This paper studies factors that determine efficient information processing. I exploit a unique small business lending setting where I am able to observe the entire codified demographic and accounting information set that loan officers use to make decisions. I decompose the loan officers’ decisions into a part driven by codified hard information and a part driven by uncodified soft information. I show that a machine learning model substantially outperforms loan officers in processing hard information. Loan officers can only process a sparse set of useful hard information identified by the machine learning model and focus their attention on salient signals such as large jumps in cash flows. However, the loan officers use salient hard information as “red flags” to highlight where to acquire more soft information. This result suggests that salient information is an attention allocation device: It guides humans to allocate their limited cognitive resources to acquire soft information, a task in which humans have an advantage over machines.