Conference Agenda

The existing evidence for predictability of international bond risk premia raises questions

about whether significant statistical in-sample results can be translated into economic gains.

Moreover, limited information is provided for practical applicability of existing findings. This

study examines a broad set of existing bond risk premia models, extends it to international

markets, and highlights the benefits of using a global forecasting approach for investors. Such an

approach, combining information from multiple international markets, better captures drivers

in international bond risk premia than other approaches, including solely local information. The

out-of-sample findings show how government bond investors can utilize the presented approach

to improve their efficiency frontier, although achievable economic gains are rather limited.

This paper revisits and extends the dividend term structure model of Kragt et al. (2020), which consolidates dividend growth, the risk-free rate, and the risk premium into a unified discount factor. Employing recent data from four major indices (S&P 500, Eurostoxx 50, Nikkei 225, and FTSE 100), I evaluate the model’s performance under three specifications: an unconstrained version, a transversality-constrained version, and an unconstrained version excluding the pandemic period. Although the unconstrained estimates remain economically meaningful for the Eurostoxx 50, Nikkei 225, and FTSE 100, the US market persistently presents a puzzle: unconstrained estimates for the S&P 500 imply explosive long-run dividend projections unless constraints are imposed or pandemic data are removed. While either intervention restores theoretical coherence, it also uncovers a sharp divergence between model-implied and observed valuations, suggesting the omission of critical forces—long-run bubbles and convenience yields—that likely drive equity prices. Moreover, short-run dividend expectations exhibit strong sensitivity to inflationary shocks, whereas medium-term projections hinge on the yield curve and broader monetary policy.

We introduce a method to determine the investor's optimal portfolio size that maximizes the expected out-of-sample utility under parameter uncertainty. This portfolio size trades off between accessing investment opportunities and limiting the number of estimated parameters. Unlike sparse methods such as lasso that exclude assets during the optimization step, our approach fixes the optimal number of assets before computing the portfolio weights, which improves robustness and provides greater flexibility in practical implementations. Empirically, our restricted portfolios outperform their counterparts applied to all available assets. Our methodology renders portfolio theory valuable even when the dataset dimension and sample size are comparable.

This study evaluates the effectiveness of specialised large language models (LLMs) developed for accounting and finance. Empirical analysis demonstrates that these domain-specific models, despite being nearly 50 times smaller, consistently outperform state-of-the-art general-purpose LLMs in return prediction. By pre-training the models on year-specific financial datasets from 2007 to 2023, the study also mitigates look-ahead bias, a common limitation of general-purpose LLMs. The findings highlight the critical importance of addressing look-ahead bias to ensure reliable results. Extensive robustness checks further validate the superior performance of these models.