Working Paper: NBER ID: w1105
Authors: jeffrey a frankel; james h stock
Abstract: Volatility tests are an alternative to regression tests for evaluating the joint null hypothesis of market efficiency and risk neutrality. Acomparison of the power of the two kinds of tests depends on what the alternative hypothesis is taken to be. By considering tests based on conditional volatility bounds, we show that if the alternative is that one could"beat the market" using a linear combination of known variables, then the regression tests are at least as powerful as the conditional volatility tests.If the application is to spot and forward markets, then the most powerful conditional volatility test turns out to be equivalent to the analogous regression test in terms of asymptotic power. In other applications,the volatility test will be less powerful than regression tests against our chosen alternative. However, these results are not inconsistent with the observation that volatility tests may be more powerful against other alternative hypoth-eses, such as that risk-averse investors are rationally maximizing the present discounted utility of future consumption,with a time-varying discount rate.
Keywords: No keywords provided
JEL Codes: No JEL codes provided
Edges that are evidenced by causal inference methods are in orange, and the rest are in light blue.
| Cause | Effect |
|---|---|
| Regression tests (C52) | Power against the alternative hypothesis (C12) |
| Volatility tests (C58) | Power levels depending on the alternative hypothesis (C12) |
| Volatility tests (C58) | Power against other alternatives (D74) |
| Regression tests (C52) | Equivalent power to volatility tests in spot and forward markets (C58) |