Revolutionising financial decision-making with emerging technologies and interdisciplinary approaches

In today’s rapidly evolving business landscape, emerging technologies and interdisciplinary approaches are transforming how we understand financial decision-making and corporate behaviour.

Our* recent research highlights the power of unstructured data analytics (data that doesn’t fit into traditional databases or spreadsheets, e.g. images, videos, audio, social media posts) in uncovering novel insights about management behaviour and its impact on financial outcomes.

Using image processing to understand how team confidence influences fundraising success

For our first study, we employed image processing and computer vision techniques, combined with psychological experiments, to investigate the impact of management team confidence on the success of fundraising in Initial Coin Offerings (ICOs) - a type of crowdfunding where companies usually raise capital using cryptocurrencies. By analysing photographs of management teams through both AI-powered computer vision and human perception experiments, we demonstrated that the confidence conveyed by the team significantly affects a firm's ability to raise capital.

We designed an image experiment to explore how confidence is communicated through team photos. Research in psychology and neuroscience has shown that humans form initial impressions of others incredibly quickly - in mere milliseconds - and these snap judgements tend to stick, even with longer exposure to the image.

Building on this insight, we developed a systematic approach to evaluate the team confidence projected in photographs. We showed the participants a photograph of an ICO team and asked them to rate the team's overall confidence level to be on a 5-point scale. Importantly, we showed only the team photos without any other technical or financial information about the ICO projects, ensuring that participants' judgements were based purely on visual impressions. To make the experiment robust, each team photo was evaluated by 10 different participants, with each participant rating between 10 to 50 different teams.

What makes these photos particularly interesting is that they contain rich behavioural information beyond just faces - everything from expressions and poses to clothing choices represents intentional decisions by team members about how they want to present themselves to potential investors.

This approach allowed us to quantify the confidence projected by the team based on their photos into concrete data that we could analyse alongside fundraising outcomes. The experiment's design, grounded in established psychological research, helped us draw meaningful conclusions about how visual presentations of confidence influence fundraising success in the ICO market.

In order to understand what signals confidence to potential investors in team photographs, we further examined various visual elements, from clothing choices to pose styles, by using advanced image processing AI tools. We looked at factors such as the percentage of team members wearing suits or smiling, how close people were to the camera, whether team members posed with crossed arms (a popular stance in business photos), and even stylistic choices like using black-and-white photography or uniform backgrounds.

We found that teams appeared more confident when more members wore suits and smiled in their photos. Interestingly, using black-and-white photography also boosted perceived confidence, likely because it creates a unified, professional look. But here's a surprise: the "crossed arms" pose, despite being a staple of business photography, didn't actually increase perceived confidence.

The visual elements serve as alternative information channels to assist investors in making investment decisions in a financial market with significant information gaps and barriers. In the ICO market, where regulation is limited and disclosure standards are underdeveloped, investors face fewer protections and must rely on all available information to make decisions. Understanding blockchain technology and cryptocurrency projects from white papers and project statements requires specialised knowledge that many investors may not possess.

In this environment of high uncertainty, investors naturally seek additional signals to assess project quality and team capability. This is where visual elements become particularly valuable – they serve as an accessible, immediate channel through which teams can communicate professionalism, confidence, and competence to potential investors.

Using advanced mining techniques to analyse how political ideologies shaped Chief Executive Officers’ (CEOs) decision-making during Covid-19

In our second study, we investigated how CEOs made tough decisions during the COVID-19 pandemic, particularly when choosing between protecting shareholder interests and preserving jobs.

We employed sophisticated text mining techniques (the process of using computational tools to extract meaningful patterns and insights from large volumes of textual data) to analyse more than 7,500 mandatory company filings (known as 8K forms) from the USA’s largest public companies during 2020. These documents contained rich textural information about various corporate actions, including operational changes, supply chain adjustments, financial restructuring and employee policies.

It is this use of advanced text mining techniques that makes our study particularly powerful. Instead of just looking at employment numbers, we analysed detailed descriptions of how companies managed their workforce costs. We developed a sophisticated multi-stage approach to extract and analyse this information from thousands of corporate filings.

First, we built a comprehensive keyword detection system that combined human expertise with machine learning. We started by identifying core terms related to labour cost reductions through careful manual analysis, then expanded this vocabulary using word embedding models - an AI technique that can identify semantically-similar words based on their usage context. This approach allowed us to capture and identify not just obvious terms but also subtle variations and related concepts that might otherwise be missed.

The rich textual data revealed various approaches to cost reduction – e.g. temporary furloughs, permanent layoffs, wage reductions, reduced working hours. Through careful manual validation and coding of these extracted texts, we captured corporate behaviours that wouldn't be visible in simple employment statistics.

During the COVID-19 period, traditional factors like industry conditions, the company’s financial health, and market position influenced how companies responded to the crisis. However, CEOs facing similar financial pressures often made surprisingly different decisions about employment and dividends. This is where CEO ideology became particularly interesting.

While previous research examined how CEOs' personal ideologies influence their decisions during stable economic times, the COVID-19 pandemic presented a unique and unprecedented scenario. Unlike the 2008 financial crisis, which came with warning signs, the pandemic struck suddenly and forced CEOs to make rapid, high-stakes decisions about balancing shareholder and employee interests. They had to choose between protecting shareholders through continued dividends, preserving employment levels, or trying to balance both.

Our study shows conservative-leaning CEOs (e.g. CEOs who donated to the Republican Party only) were more likely to aggressively reduce labour costs while still meeting dividend expectations. Conversely, CEOs with fewer conservative ideologies were less likely to meet dividend expectations and less likely to reduce labour costs.

In contrast to previous research describing differences in behaviour between conservative and liberal CEOs, our findings suggest that conservative CEOs as a category are different from all other categories of CEOs (i.e. liberals-leaning CEOs who donated to the Democratic Party only; Non-partisan CEOs who donated to both the Democratic and Republican parties; and Zero-donations, CEOs who did not make any donations), not just liberal ones, in their reaction to an extreme event.

Our big data approach enabled us to uncover patterns in corporate decision-making during crisis periods, thereby disclosing how CEOs' political ideologies influenced their responses to the pandemic's challenges.

Unlocking new insights with unstructured data analytics

Both studies demonstrate the transformative potential of unstructured data analytics in finance research. By utilising tools from computer vision, natural language processing, and machine learning, we can now analyse previously untapped sources of information - from facial expressions to corporate communications - to better understand the human factors that influence financial decisions.

This interdisciplinary approach, which combines behavioural science, data analytics, and finance, opens new avenues for research and illustrates how technological innovation can enhance our understanding of management behaviour and its financial implications.

Read the journal articles:

• *Winifred Huang, Silvio Vismara, Xingjie Wei, Confidence and capital raising, Journal of Corporate Finance, Volume 77, 2022, 101900, ISSN 0929-1199, https://doi.org/10.1016/j.jcorpfin.2021.101900

• *Ali Bayat, Marc Goergen, Panagiotis Koutroumpis, Xingjie Wei, The impact of CEO political ideology on labor cost reductions and payout decisions during the COVID-19 pandemic, Journal of Corporate Finance, 2024, 102692, ISSN 0929-1199, https://doi.org/10.1016/j.jcorpfin.2024.102692

A Leeds University Business School internal research funding scheme supported research undertaken during the first project.