Avoiding the Pitfalls: Can Regime-Switching Tests Reliably Detect Bubbles?

Our paper uses simulation methods to examine the size and power of regime-switching tests for bubbles. We find that even with several hundred observations, the tests show sometimes considerable size distortion. This distortion makes the tests conservative; they understate the significance of the evidence of bubbles. Despite this, the tests display considerable power to detect bubbles even when using the conservative asymptotic critical values. We also find that the frequency with which bubbles collapse has an important influence on the tests' power. An application to monthly Canadian and American stock-price data provides mixed evidence of bubbles.