AI-Driven Financial Foresight

Our latest blog on tracking market themes and sentiment for alpha signals. https://lseg.group/457OASE

Some of the excellent feedback on using sentiment in FX modeling at LSEG Data & Analytics' recent China Hackathon:  💡"By incorporating sentiment indicators, we achieved nearly 70% prediction accuracy on EUR pairs, which is quite impressive given the noisy and regime-shifting nature of FX markets.” ➡️ Mr. Hai Lan Head of FX, Financial Markets Department, China CITIC Bank  💡"LSEG’s [MarketPsych] sentiment factor data is both innovative and practical. It transforms traditional qualitative insights into quantifiable features, allowing our models to shift from purely numerical inputs to emotion-driven dimensions. That’s a game-changer—it really impressed me.” ➡️ Dr. Yifeng Shi Executive Director of Quantitative Analysis, Standard Chartered Bank  💡"For our strategy, we leveraged LSEG’s data, including government bond futures and MarketPsych sentiment factors. By integrating these into a deep learning model—specifically CNN-LSTM—we improved our timing accuracy. One of our key innovations was using sentiment data to determine whether the market was trending or range-bound ... sentiment data offers a new path to uncover alpha.” ➡️ Mr. Zhen Song Head of Bond Quant Trading, SPD Bank  💡 "LSEG’s data has proven to be incredibly valuable in building quantitative trading strategies. Especially when we combine MarketPsych’s sentiment factors with market data, it not only enhances model performance but also deepens our understanding of market behavior."  ➡️ Dr. Qiqi Shuai Senior Quantitative Trader, Global Markets, Bank of China (Hong Kong)

A recent academic study tracks the global transmission of fear in the treasury markets: https://papers.ssrn.com/sol3/papers.cfm?abstract_id=4950235 Fear in the "Fearless" Treasury Market (2025) ➡️ "Fear" predicts U.S. Treasury bond risk premia with strong statistical support after controlling for traditional yield curve, macroeconomic, and sentiment predictors.➡️ The paper shows that the fear index explains 22.42% to 25.62% of in-sample variation in bond risk premia, and achieves out-of-sample R² values ranging from 22.05% to 30.24% across bond maturities.

Finding the next big theme is crucial to investing these days. Here's a depiction of the stock price performance of companies involved in "rare earths" according to their Q1 earnings call transcripts, using our analytics: https://www.lseg.com/content/dam/data-analytics/en_us/documents/fact-sheets/lseg-marketpsych-transcript-analytics.pdf

Will the U.S. enter a Debt Spiral?  We did an interactive presentation two weeks ago with some of the largest money managers in the world (the long-term "smart money"). One of the multiple choice poll questions was how high they see 30-year U.S. Treasury rates (Tbonds) reaching over the next 4-years. The answers are compiled in the graphic. Debt spirals are driven by perceptions - if investors believe there is a risk their loans won't be paid back, they demand higher interest rates to compensate them for the risk. That's what we see in the poll results.

We did a simple study that counted the number of times "tariffs" were mentioned in Earnings Call Transcripts of global companies in March 2025. The share price returns of those companies in April and May were then tabulated. Those whose transcripts mentioned tariffs 30x or more fell 4% through May 31st, while those with no mentions (thousands) rose 12.5% from April 1st to May 31st. Among all stock price factors, the number of references to "tariffs" in earnings transcripts was one of the most predictive in 2025.

With advances in large language models, Earnings Call Analysis is entering a new era. LSEG MarketPsych Transcript Analytics leverages fine-tuned NLP to extract sentiment, emotion, and topical insights from corporate transcripts.  See our blog on this new approach to extracting investable insights from executive (and analyst) language: https://lnkd.in/geNCjvCJ

Markets just delivered a textbook “relief rally.”  A 2019 study in PLOS ONE (“Relief from incidental fear evokes exuberant risk taking”) paired mild electric shocks with an investment task. When the shock threat was removed, participants’ risk appetite—and the expected value of the options they chose—jumped ~11 %. Takeaways for today’s market:• Relief fuels risk-on behavior. As macro stressors ease, investors systematically bid up risk assets. • It’s valuation, not probability, that shifts. The brain literally assigns higher worth to the same payoff once fear subsides. • Mind the half-life. If talks stall or other threats re-emerge, the reverse can happen just as fast. Behavioral finance reminds us that emotion isn’t noise around fundamentals; it re-prices them. Managing portfolios means managing the emotional cycle too.