Answers are grounded in this paper and cite their sources.
This paper aims to figure out whether the monthly-average 10-year minus 2-year U.S. Treasury term spread can be forecast by its yield-curve and extra macroeconomic information. The research uses monthly data from January 1990 to July 2026 and sets 53 recursive origins during this period. Then the spread is forecast one and three months ahead against a driftless random walk, comparing a second-order vector autoregressive (VAR(2)) model in first differences with a second-order vector autoregressive model with exogenous variables (VARX(2)) that adds inflation, the policy rate, the Chicago Board Options Exchange Volatility Index (VIX) and the economic policy uncertainty (EPU) index. VAR reduces root mean squared error (RMSE) by 7.76% at one month and 3.96% at three months, and the one-month gain survives Holm adjustment (p = 0.0136); the VARX-versus-VAR increment is insignificant at both horizons, so the macroeconomic contribution is weak and not robust. Ex-post groupings concentrate the gains where the spread actually narrowed, which describe forecast-error heterogeneity only. Conditional on other yield lags and the macro controls, lagged 2-year changes carry no robust incremental linear information. The results are pseudo-out-of-sample because the data used are revised.
No takes yet. Share an insight, caveat, or question.
Peiyang Yu (2026) studied this question.
Synapse has enriched 10 related papers. Consider them for comparative context: