Ties de Kok

The timely flow of financial information is critical for efficient capital market functioning, yet we have little understanding of firms' and auditors' collective abilities to maintain timely financial reporting while under duress. We use COVID as a stress test case to examine whether reporting systems can withstand systemic increases in complex economic events and coordination challenges. Despite COVID-related challenges persisting through 2020 and beyond, we document surprisingly modest average delays in financial reports during COVID, and only in Q1-2020. Reporting timeliness reverts to pre-COVID levels no later than Q2-2020. We find no evidence of meaningful declines in actual reporting quality during COVID, but we do find some evidence consistent with declines in perceived reporting quality. Overall, our findings indicate that current financial reporting processes are remarkably robust, and provide insights about financial reporting more broadly. In particular, given that nearly all firms were able to weather the unprecedented disruptions caused by COVID, our findings imply that most material reporting delays observed outside of COVID are likely due to either a firm's strategic choices or to exceptionally fragile reporting processes.

This study evaluates EBITDA as a financial performance measure and investigates the use of EBITDA in financial reporting. First, we take issue with recent comments that both the SEC and the IASB have levied against non-GAAP earnings numbers, and in particular EBITDA. While EBITDA allegedly provides an accurate reflection of the operations and abstracts from how assets are financed, we argue that (net) operating profit already provides this information without the necessity of making subjective adjustments. Also, our evidence suggests that EBITDA paints a rosy picture of the firm's profitability and cash-generating ability. Next, using textual analysis, we investigate the prevalence of EBITDA in financial disclosures based on a large sample of 15,895 annual reports and 51,758 earnings releases from S&P 1500 firms between 2005 and 2016. We find that 14.8% of sample firms disclose and emphasize EBITDA numbers. EBITDA disclosures modestly increase over time and tend to be rather sticky in nature. In our cross-sectional analyses, we find that, consistent with our hypotheses, EBITDA-reporting firms are smaller, more leveraged, more capital-intensive, less profitable and have longer operating cycles than non-EBITDA reporting firms. They also exhibit higher forecast errors and a higher likelihood of missing the analyst forecast benchmark. Additional tests further underscore the opportunistic nature of EBITDA disclosures as we find that these firm characteristics are more strongly associated with the disclosure of adjusted EBITDA measures, and less strongly associated with the disclosure of EBITA and EBIT.

Generative Large Language Models (GLLMs), such as ChatGPT and GPT-4 by OpenAI, are emerging as powerful tools for textual analysis tasks in accounting research. GLLMs can solve any textual analysis task solvable using non-generative methods, as well as tasks previously only solvable using human coding. While new and powerful, GLLMs also come with limitations and present new challenges that require care and due diligence. This paper highlights the applications of GLLMs for accounting research and compares them to existing methods. It also provides a framework on how to effectively use GLLMs by addressing key considerations such as model selection, prompt engineering, and ensuring construct validity. In a case study, I demonstrate the capabilities of GLLMs by detecting non-answers in earnings conference calls, a traditionally challenging task to automate. The new GPT method achieves an accuracy of 96% and reduces the non-answer error rate by 70% relative to the existing Gow et al. (2021) method. Finally, I discuss the importance of addressing bias, replicability, and data sharing concerns when using GLLMs. Taken together, this paper provides researchers, reviewers, and editors with the knowledge and tools to effectively use and evaluate GLLMs for academic research.

Business complexity is an important informational friction but challenging to measure. We use a GPT large language model fine-tuned on narrative disclosures and inline XBRL tags to measure business complexity from a user perspective. Our measure negatively correlates with the speed of capital market price adjustments to financial reports and positively correlates with filing delays even after controlling for existing complexity measures. We exploit the modularity offered by the measure to construct complexity scores for different transaction categories, such as debt, business combinations, and compensation. Using this modular measure, we find that more complex transactions receive more regulatory scrutiny, correlate with greater cash holdings, and predict worse future performance. Overall, our study reinforces complexity’s role as a friction to decision-making and provides a richer and more flexible measure of complexity for future research.

Using direct measures of information acquisition from a business intelligence provider, we examine the actions managers take to learn about their business. We first provide descriptive evidence that managers of small businesses adopt the business intelligence service due to business complexity and growth expectations, and, after adoption, managers become heavy consumers of financial and related nonfinancial metrics. We then show that, despite offsetting economic motivations, the hedonic motivations of managers appear to significantly affect their information diet—i.e., managers consume information, at least in part, based on their expectation of how the information will make them feel. These findings highlight a significant friction for managers’ information processing and inform the design of pervasive data products such as dashboards.

I study the role of product development updates and funder involvement in improving reward crowdfunding outcomes. Reward crowdfunding is a growing source of funding for entrepreneurs with limited access to traditional financing. The limited regulation in reward crowdfunding markets, however, leaves it up to the entrepreneur and the funders to solve their agency problems. Using data from one of the largest reward crowdfunding platforms, I find that entrepreneurs who provide less frequent voluntary updates experience more involvement by funders in the product development process. The entrepreneur responds to this heightened involvement by adjusting their announcement length but not their update frequency. Finally, I find that funders who are responsive in their product development involvement improve project outcomes. Taken together, my results support that voluntary updating and funder involvement can work together to resolve agency frictions in reward crowdfunding markets.

This paper studies whether increasing supervision impairs a loan officer's motivation to produce and share information. We first exploit an organizational change at a large European bank and find that centralization improves the use of soft information in evaluating the creditworthiness of small-to-medium size loan applicants. Next, we conduct a lab experiment and find that participants are more likely to accept an increase in supervision when it is motivated by changes in the external environment of the organization. Taken together, our study informs financial institutions, regulators, and academics on the potential benefits of centralization and provides insights on how to achieve employee acceptance when introducing a new supervision policy.