Abstract:
This paper investigates the in-sample predictability of debt ceiling and government
shutdown for real stock returns in the U.S, using rolling window Granger non-causality
estimation. Causal links often evolve over time so the use of the bootstrap rolling window
approach will account for potential time variations in the relationships. We use monthly time
series data on measures of debt ceiling and government shutdown, and real stock returns,
covering the period of 1985:M2 to 2013:M9. Since the debt ceiling and government shutdown
variables under analysis are exogenous, the use of the in-sample predictability to analyse the
relation-ship running from debt ceiling to real stock returns, as well as, from government
shutdown to real stock returns will provide evidence of not only whether in-sample
predictability exists, but also how predictability varies over time i.e. significance in episodes of
high values of index. The full sample bootstrap non-Granger causality test results suggest
existence of no in-sample predictability of debt ceiling or government shutdown for real stock
returns in the U.S. economy. The stability tests show evidence of parameter instability in the
estimated equations. Therefore, we make use of the bootstrap rolling window (24 months)
approach to investigate the changes in the in-sample predictability of the relationship, and detect signifi-cant in-sample predictability of debt ceiling and government shutdown for real
stock returns at different sub-periods, corresponding especially after the phases where there
were sharp increases in the indexes of debt ceiling and government shutdown.
