Time Series Signals and Prediction

We propose a generalized multivariate unobserved components model to decompose macroeconomic data into trend and cyclical components. We then forecast the resulting cyclical component series using Bayesian methods. We document that a fully Bayesian estimation, that accounts for state and parameter uncertainty, consistently dominates out-of-sample forecasts produced by alternative multivariate and univariate models. In addition, allowing for stochastic volatility components in variables improves forecasts. To address data limitations, we exploit cross-sectional information, use the commonalities across variables, and account for both parameter and state uncertainty. Finally, we find that an optimally pooled univariate model outperforms individual univariate specifications, and performs generally closer to the benchmark model. 

We introduce a novel approach to generating more precise and adaptive trading signals in pairs trading. The method consists of deriving the full conditional distribution of the hedge ratio (ß1) and using its quantiles as confirmation thresholds for trading signals originated in the standard cointegration strategy. We applied the method to selected pairs of assets across the various markets. The empirical analysis demonstrated that the proposed algorithm significantly enhances trading performance and risk management for several pairs compared to the traditional cointegration method. By adopting a distribution-based framework through Bayesian hierarchical modeling, our method not only enables more timely trading signals but also improves pair selection by effectively detecting a considerable proportion of false positives in cointegration tests as demonstrated via simulations. 

Missing data often significantly hamper standard time series analysis, yet in practice they are frequently encountered. In this paper, we introduce temporal Wasserstein imputation, a novel method for imputing missing data in time series. Unlike existing techniques, our approach is fully nonparametric, circumventing the need for model specification prior to imputation, making it suitable for potential nonlinear dynamics. Its principled algorithmic implementation can seamlessly handle univariate or multivariate time series with any missing pattern. In addition, the plausible range and side information of the missing entries (such as box constraints) can easily be incorporated. As a key advantage, our method mitigates the distributional bias typical of many existing approaches, ensuring more reliable downstream statistical analysis using the imputed series. Leveraging the benign landscape of the optimization formulation, we establish the convergence of an alternating minimization algorithm to critical points. Furthermore, we provide conditions under which the marginal distributions of the underlying time series can be identified. 

This study examines the volatility dynamics between crude oil and gold returns using the GARCH-MIDAS framework, which combines mixed-frequency data to capture both short- and long-term volatility. Though these commodities are key hedges against inflation and uncertainty, their interlinked volatility over time is not fully understood. We apply GARCH-MIDAS and compare it to the traditional GARCH-X model to assess how macroeconomic and market factors influence volatility over quarterly, semi-annual, and annual horizons. Using data from 1980 to 2024, we find that GARCH-MIDAS outperforms GARCH-X, reducing mean absolute errors by 40–60%. Gold plays a critical exogenous role in forecasting crude oil volatility, with its influence varying across forecast periods. Our analysis includes recent disruptions like COVID-19, offering fresh insights into changing volatility patterns. This research advances the literature on cross-commodity volatility and introduces a robust framework for mixed-frequency volatility forecasting. 

The influence and impact of social media campaigns on democratic elections is a critical area of research in modern big-data analytics. While the efficacy of using social media data for forecasting election results has been debated in political and social science literature, this paper introduces a novel modeling approach that combines public opinion polls with Twitter data, incorporating key network structure features of Twitter to enhance prediction accuracy. We developed two models: the Twitter Hashtag based Opinion Survey (THOS) model, which uses hashtag frequency, and the Twitter Hashtag and Network-based Opinion Survey (THANOS) model, which includes network centrality measures. Applying these models to Ireland's 36th amendment referendum and the 2018 US Senate elections yielded promising results. The THOS model effectively predicted outcomes in races with clear frontrunners, while the THANOS model excelled in closely contested races by leveraging network dynamics. These findings underscore the potential of integrating social media data with traditional polls to improve electoral forecasts, demonstrating the robust capabilities of the THOS and THANOS models in providing accurate predictions based on the interplay of public opinion and social media activity.