On Testing the Random‐Walk Hypothesis: A Model‐Comparison Approach