Simple tests for cointegration in panels with structural breaks

Abstract

In the article we propose new panel cointegration tests which allow for structural breaks. We show that the panel tests have good size and power. We apply the test statistics to the analysis of the Feldstein–Horioka puzzle for a sample of 16 OCDE countries. After allowing for breaks, we find strong evidence that investment and saving rates are cointegrated.

Notes

¹ More regressors can be included in the Equation 1.

² Note that the tests allow for heterogeneity because the cointegration vector and the breakpoint position might differ across the units in the panel.

³ The results for the other models are pretty similar and they are available upon request.

⁴ We investigate the power of tests for τ that randomly varies across the cross-section units two periods away from the breaks at τ = 0.25, 0.50 and 0.75. The results, not reported for brevity, are similar to those presented in Table 2.