Stable Sentiment and Persistent Dynamics in U.S. Economic News over 45 Years

Battle Count: 88

The paper's findings are relevant to quantitative trading in that they demonstrate news sentiment has become more persistent over time, meaning sentiment shocks leave longer traces. This implies that sentiment-based trading signals should account for the response time of the sentiment process rather than treating it as a short-memory signal. The increasing bimodality and longer residence times in optimistic/pessimistic regimes could inform regime-detection strategies. However, the paper is primarily descriptive/analytical rather than directly proposing trading strategies, and the sentiment index is economic news-specific rather than market-specific.

DFA is a well-established method with available implementations. The rolling window analysis and null model comparisons are standard. The endogenous-memory model is relatively simple (two-component decomposition) but calibration via Latin hypercube optimization over 6 parameters requires careful implementation. The main complexity lies in data preprocessing (deseasonalization, Hampel filtering) and the multi-scale analysis framework. No code is provided, requiring reimplementation from the paper's description.

The sentiment index is publicly available from the Federal Reserve Bank of San Francisco (FRBSF). The DFA methodology is well-documented and standard. However, the Factiva corpus underlying the index is commercial. The endogenous-memory model calibration procedure is described but no code repository is provided. Null model implementations (IAAFT, block bootstrap) are standard. Sensitivity checks are referenced in supplementary information.