This paper assesses the trends of some main macroeconomic and macro-financial variables across different time horizons related to systemic banking crises. Specifically, by gradually shifting the observation horizon of the same statistical model across time, it observes how these variables are associated with banking crises in the past, present and future. The associations vary considerably when shifting horizons. Domestic house price growth increases the probability of observing a crisis in the future, but its effect disappears when moving closer to a crisis. The inverse holds true for the effect of the global credit gap, while global credit growth consistently and significantly increases the probability of a future banking crisis. Also, banking crises seem to be spatially correlated in the very short run. In all, the results can help policy makers by shedding light on the temporal horizon of the variables they monitor in addition to evaluating their predictive power.
banking crisis binary time series cross section data credit time horizons