A conversation withAlexander Hillert, SAFE

„Overly optimistic CEOs can be a warning sign“

At the Leibniz Institute for Financial Research (SAFE) in Frankfurt, an index is currently being developed measuring the sentiment of CEOs and CFOs – based on text analyses. It will soon go live.

„Overly optimistic CEOs can be a warning sign“

A new index measuring the sentiment of managers in German companies is in its final preparations. „We are in the advanced stages of evaluation“, reports Prof. Alexander Hillert, Professor of Finance and Data Science at the Leibniz Institute for Financial Research SAFE in Frankfurt, in an interview with Börsen-Zeitung. „The index will go live soon.“

Unlike other barometers, this new index is not based on surveys of executives. Instead, Hillert and his colleagues are developing an index that measures the sentiment of board members in German companies through text analysis techniques. To achieve this, they analyse the texts of quarterly and annual reports, as well as the transcripts of analyst conferences, including the Q&A sessions with CEOs and CFOs.

Counting positive and negative words

Specifically, the researchers examine how positively or negatively managers communicate. „We use a straightforward, transparent approach“, explains Hillert. There are two well-known lists of terms developed by two US researchers, Tim Loughran and Bill McDonald, which categorise common terms used in corporate reporting into positive and negative categories – terms like „loss,“ „revenue decline,“ or „impairment“ are classified as negative.

„We count the positive and negative words in each text, calculate the difference, and then divide by the length of the text“, says Hillert. The data is aggregated over a month to produce a metric that reflects the overall sentiment.

The index is inspired by a research paper from the US that puts forward the thesis manager sentiment is a leading indicator of market developments. Notably, it suggests that heightened optimism today correlates with lower returns in the future. „In other words, overly optimistic CEOs can be warning signs“, cautions the academic.

Managers tend to overinterpret the status quo

The seemingly surprising connection relates to the tendency of managers to be excessively optimistic or pessimistic. During periods of high optimism, there is a risk that these managers may overinvest and build excess capacity, which can lead to declining profits. „This, in turn, results in lower returns in the stock market“, says Hillert.

Managers run the risk of overinterpreting developments. For instance, during the pandemic, airline executives assumed that demand would not recover in the foreseeable future, leading them to significantly cut capacity. But people wanted to fly again sooner than expected.

Dax, MDax, SDax, and TecDax

„Our project aims to further develop this thesis from the US for Germany“, notes Hillert. The analyses will focus on stocks listed in the Dax, MDax, SDax, and TecDax or those that have been listed in the past 20 years. The evaluations will begin from 2003, as this is when transcripts from analyst conferences became available. Last year, initial analyses were conducted based on historical datasets. „We found indications that the thesis also seems plausible for the German market“, says the economist. Since spring 2024, current data is aggregated monthly to update the index.

The data from the transcripts of analyst conferences, made available by Refinitiv, is an important input for the index. „Goethe University and SAFE offer excellent access to this data“, says Hillert. SAFE can also support the project in further developing the index through large language models, thanks to its high-performance computing resources. „What distinguishes SAFE is its extensive data offerings, strong computing infrastructure, and outstanding colleagues with whom project staff can collaborate", he adds.

Artificial intelligence plays a role, but not a central one. Currently, the index is supported by automated IT systems. „But we keep it simple“, Hillert explains, adding that AI could indeed play a larger role in the future to measure sentiment more finely.