Financial bubbles are often described as the result of behavioral biases and market frictions. The theoretical model presented in this paper shows how a rational expectation equilibrium (REE), where asset prices can deviate from their fundamental value, can still exist in a world with agents rationally and unbiasedly evaluating an asset's future return. Price deviation is the result of risk-averse fund managers endowed with different information sets who compete to outperform each other. A higher degree of competition leads to larger price deviations and worse risk-adjusted returns. However, competition enhances investors' ability to separate different types of managers, thus creating a trade-off between hampering excessive asset price growth and fostering informed trade identification.

I propose a new benchmark to evaluate hedge fund performance: the returns to shorting CBOE Volatility Index (VIX) futures. The informativeness of this benchmark leads to a new methodology that is able to predict hedge fund performance. Specifically, it separates hedge funds, ex-ante, into one group that delivers higher sharpe ratios and positive skewness (SR of 0.52 and Skew of 4.30) while the other group has lower sharpe ratios and negative skewness (SR of 0.15 and Skew of -0.83), out-of-sample (OOS). I refer to the former group as those hedge funds with edge, in contrast to the latter group as those hedge funds that are without edge. This approach cannot be explained or replicated by previously known methods. Lastly, I show that my empirical findings can be explained by a model that features traders with extrapolative expectations.

This paper hypothesizes that investors in long-short multi-factor strategies focus on variances in their short-term evaluation of portfolio risk, while they neglect correlations, which they perceive as close to zero - in line with the findings of various studies supporting this narrative. Using a set of 14 factor return time series over multiple asset classes, factor styles, and a long historical data period, this paper sheds light on the matter and finds empirical evidence in support of the hypothesis. By decomposing the variance of an equally-weighted multi-factor benchmark into two components, average variance (AV) across and average correlation (AC) between factors, the paper shows that the product of both variables reasonably spans contemporaneous multi-factor variance in-sample, indicating that both components are needed to explain current benchmark risk. However, the subsequent forecasting exercise reveals that only AV predicts future multi-factor risk and return over short horizons (one month up to two years). In sum, the findings support the idea of short-term persistence in risk and the typically assumed variance-in-mean relationship, suggesting that higher AV (in contrast to AC) leads to significantly higher future risk and returns for multi-factor investors, even after controlling for other macroeconomic and market candidate predictors, aswell as alternative risk approximations.