Working Paper: CEPR ID: DP1336
Authors: Dan Bendavid; Robin L. Lumsdaine; David H. Papell
Abstract: Recent literature has documented the sensitivity of unit root tests to failure to account for structural change. This paper reconsiders international evidence on the unit root hypothesis while allowing for two structural breaks. We find evidence of two breaks in three-quarters of the data, rejecting the unit root hypothesis in 50% more cases than models that allow for only one structural break. Most of the trend breaks are associated with a change in output levels. As the neo-classical growth model predicts, the magnitude of these level changes is shown here to be related to changes in growth rates during the period following the break.
Keywords: unit root hypothesis; trend breaks; growth
JEL Codes: C22; O1; O5; O47
Edges that are evidenced by causal inference methods are in orange, and the rest are in light blue.
| Cause | Effect |
|---|---|
| Structural breaks in GDP data (N14) | rejection of the unit root hypothesis (C22) |
| Structural breaks (C22) | changes in output levels (E23) |
| changes in output levels (E23) | subsequent growth rates (O41) |
| greater drops in output levels (E32) | higher subsequent growth rates (O49) |
| changes in output levels (E23) | changes in growth rates (O41) |