FABART model reveals asymmetric impact of oil supply news shocks
An ECB working paper introduces the FABART framework to analyze US oil supply news shocks from 1974 to 2019. Authors Eoghan O'Neill and Sofia Velasco find that adverse oil supply shocks trigger stronger real economic contractions than the expansions caused by favorable shocks.
Quantifying asymmetric energy shocks
The Factor Bayesian Additive Regression Tree (FABART) model evaluates 62 macroeconomic, financial, and state-level variables to trace oil supply news shocks identified around OPEC announcements.
The authors find pronounced sign asymmetries: adverse shocks that raise oil prices by 10 percent generate persistent contractions in industrial production, equities, and broader economic activity.
In contrast, favorable shocks of equal magnitude yield significantly weaker expansionary responses.
Size non-linearities emerge primarily between small 1.5 percent oil price movements and moderate 5 percent shifts.
Beyond moderate magnitudes, macroeconomic responses level off rather than scaling proportionally with shock size.
Divergent regional labor impacts
Regional employment data demonstrates substantial cross-state heterogeneity in shock transmission.
Manufacturing-heavy states like Michigan and Ohio suffer persistent job losses following adverse oil supply shocks due to high input costs and durable goods demand shocks.
Conversely, energy-producing regions such as Wyoming and Alaska display resilient or even positive employment dynamics after oil price hikes.
Including forward-looking financial variables in the dynamic factor model reduces omitted-information bias and sharpens the estimated speed of real economic contractions.
Data-driven trees outshine rigid models
Combining machine learning trees with dynamic factor models marks a significant methodological advance.
Allowing transmission to emerge from data avoids the flawed parametric assumptions of older models.
Central bankers gain a far more accurate framework to evaluate energy market risks.