FJFuwei Jiang

Fuwei Jiang

*This is an algorithmically generated author record based on publicly available information.

Current Organizations / Institutes

The organizations/institutes from the scholar's latest article

Previous Organizations / Institutes

The organizations / institutes from the scholar's other articles in the past ten years


Fuwei Jiang has a total of 98 co-authors. Here are the top co-authors by number of shared publications:


Latest Publications

  • Open Access

    Inflation Narratives and Expectation Formation: Evidence from Large Language Model Embeddings

    by Yongmiao Hong, Fuwei Jiang, Yuejie Wang, Bowen Xue

    We develop a novel directional inflation narratives (DIN) index using large language model embeddings applied to Wall Street Journal articles from 1984 to 2024. The index captures whether inflation-related news points toward rising or declining inflation. We find that increases in DIN are followed by higher household inflation expectations and contain information about subsequent realized inflation. Using CPI announcement surprises as an instrument for daily DIN, we provide causal evidence that more inflationary narratives raise household inflation expectations. The relationship between DIN and inflation expectations is stronger when attention to inflation is high and among households with greater news exposure. Moreover, DIN helps explain forecast errors in inflation expectations and sheds light on time-varying information rigidities. Energy and labor feature prominently in inflation narratives, while the importance of other factors varies across episodes. The results are robust to concerns about LLM look-ahead bias, and the embedding-based measure consistently outperforms traditional text analysis methods. These findings show that the direction conveyed by news narratives shapes household inflation expectations and highlight the potential of LLM embeddings for extracting economically meaningful information from large-scale text.

  • Open Access

    Geoeconomic Fear and Sovereign Spreads

    by Hui Ding, Dinghao Huang, Fuwei Jiang, Tian You

    We construct a geoeconomic fear (GeoFear) index by applying a large language model to more than one million Wall Street Journal articles from 1984 to 2025. The index measures the fear conveyed in news coverage of six classes of geoeconomic policy instruments and identifies the implementing, targeting, and affected countries in each event. Higher GeoFear predicts wider emerging-market sovereign spreads: a one-standard-deviation increase forecasts an approximately two-percentage-point widening of the average spread one month ahead, and the relation holds after controlling for global risk factors, domestic fundamentals, and sovereign ratings, as well as out of sample. The predictive content is concentrated among speculative-grade issuers, and is stronger when risk aversion and investor attention to geoeconomic topics are elevated. Higher GeoFear also predicts lower portfolio debt inflows to speculative-grade emerging markets and higher inflows to US debt, consistent with a flight-to-safety reallocation of international portfolios.

  • Open Access

    Global Geopolitical Skewness Risk and Bond Risk Premia

    by Fuwei Jiang, Jie Kang, Ruzheng Tian, Tian You

    Using geopolitical risk indices for 44 countries and regions, we construct a measure of global geopolitical skewness risk (GPR-Skew), defined as the cross-country skewness of idiosyncratic geopolitical risk components. GPR-Skew positively predicts future U.S. Treasury bond risk premia in- and out-of-sample and delivers sizable economic value. Its predictive power is not spanned by the yield curve, macroeconomic predictors, and level-based geopolitical risk measures. Within an affine term structure model, the incremental predictive content of GPR-Skew is associated with the market price of level risk, and the model-implied term premia are counter-cyclical. Elevated GPR-Skew also forecasts weaker subsequent real activity and is related to future interest rate expectation errors, credit spreads, and Treasury auction outcomes.

  • Open Access

    Geopolitical Risk Skewness and Commodity Returns

    by Fuwei Jiang, Ruzheng Tian, Qingdong Xu, Can Yang

    We propose GeoSkew, a new measure of global geopolitical risk asymmetry constructed from the cross-sectional skewness of global geopolitical risk across 44 countries and regions. GeoSkew predicts commodity futures returns both in-and out-of-sample, and its predictive power remains robust after controlling for business-cycle, macroeconomic, and uncertainty indicators, as well as competing measures of geopolitical risk. Portfolio analysis shows that GeoSkew generates economically significant gains for mean-variance investors. We further show that elevated GeoSkew predicts lower GDPat-Risk quantiles and higher future global supply chain pressures. Taken together, our findings highlight the importance of crosscountry asymmetry in geopolitical risk for commodity pricing.

  • Open Access

    More Than Words: Visual Sentiment in Social Media Videos and Stock Returns

    by Fuwei Jiang, Fujing Jin, Tian Ma

    This paper investigates whether sentiment conveyed through videos predicts stock returns. Using a large language model framework, we construct a daily market-level visual sentiment index (SVS) from facial expressions in short videos posted by influencers on Douyin (China's version of TikTok) and compare it with a caption-based textual sentiment measure (STS). We find that SVS predicts future market returns and an initial positive return response is followed by a partial reversal over subsequent trading days, with the predictability concentrated in intraday returns. 𝑆𝑉𝑆 contains predictive information distinct from STS. Textual sentiment attenuates the initial positive return response associated with visual sentiment. Additional tests support a sentiment-based interpretation of SVS: its predictive effects are more pronounced in up markets, retail-oriented portfolios, and active trading environments. Overall, visual sentiment in short videos carries market-relevant information beyond accompanying text and helps explain the timing and dynamics of sentiment-driven price responses.

  • Open Access

    Equity Return Anomalies in the Machine Age

    by Fuwei Jiang, Lingchao Meng, Yiming Wang, Tian You

    We investigate how machine readership, proxied by machine downloads from EDGAR, affects prominent equity return anomalies. Anomalies constructed from stocks with high machine-download intensity earn substantially lower average returns, consistent with higher market efficiency. This relation is stronger when human investors face greater attention constraints and more biased expectations. Evidence from firm-specific information activity and algorithmic trading suggests that both information processing and execution speed may contribute to the relation. Our findings are more consistent with a mispricing-based explanation than with a risk-compensation explanation. Overall, the evidence highlights the growing role of machines in financial markets.

  • Open Access

    What Do Firms Signal to Analysts? Macroeconomic Expectations and Analyst Forecasts

    by Peng Guo, Fuwei Jiang, Mengru Li, Can Yang

    We investigate how firms' macroeconomic expectations influence analyst forecast revisions. Using large language models to analyze annual disclosures of China's A-share listed firms, we construct a novel firm-level measure of macroeconomic expectations. We find that analysts issue more favorable forecasts for firms with more optimistic macroeconomic expectations, especially for those with higher information transparency or greater earnings uncertainty. Our evidence reveals that these expectations positively predict firms' future real activity, profitability, and market reactions. Beyond upwardly revising their forecasts, analysts also translate these signals into investment recommendations, which retail investors incorporate through optimistic sentiment and increased net buying. Our study highlights analysts as key information intermediaries in transmitting firms' macroeconomic expectations to the market.

  • Open Access

    Generative AI for Finance: A New Framework

    by Bailin Chai, Fuwei Jiang, Lingchao Meng, Tian You, Guofu Zhou

    We propose a generative-AI framework for finance that leverages Large Language Model (LLM) architectures to encode the cross-sectional structure of equity markets. Our approach, Risk Premia BERT (RPBERT), reframes the cross-section as characteristic-ordered sequences where firm identities serve as tokens within economic sentences. Motivated by the attention mechanism that underlies modern LLMs, our approach learns context-dependent representations of firms by selectively weighting characteristics and cross-sectional interactions. Rather than imposing a fixed mapping from characteristics to returns, attention endogenously determines how information enters pricing. Using a BERT-based two-stage implementation, we show that this representation-first approach substantially outperforms leading machine-learning and factor-pricing benchmarks in both statistical fit and economic performance. More broadly, our results demonstrate that attention-based generative models provide a general and powerful paradigm for asset pricing. Our proposed new model can be easily applied to analyze other asset classes as well as corporate or accounting issues.

  • Open Access

    Large Language Model and Bond Risk Premia in China

    by Jiazi Chen, Fuwei Jiang, Tian You

    Using a large language model (LLM), we construct a narrative sentiment index to investigate the predictability of risk premia in the Chinese Treasury bond market. Both in-sample and out-of-sample analyses demonstrate that narrative sentiment significantly and negatively forecasts future bond excess returns, yielding significant economic value for mean-variance investors. By incorporating this measure into an affine term structure model (ATSM), we identify a dual-channel effect: narrative sentiment relates positively to expected future short rates but negatively to excess returns, with the impact intensifying at longer maturities. Risk premia estimated via this sentiment-augmented ATSM exhibit more pronounced countercyclical dynamics. Finally, we establish that narrative sentiment carries forward-looking signals regarding both subjective expectations and future fundamentals, providing a coherent economic foundation for its predictive power in the bond market.