2013-09: The Decision to Lever

We provide an exact performance attribution for levered strategies. The attri- bution includes some familiar elements: return magnification due to leverage, bor- rowing costs, trading costs, and the variance drag; while familiar, these elements are sometimes downplayed in backtests of levered strategies. In addition, we find empirically that the covariance between return of the source portfolio and leverage plays an important role in determining the cumulative return of levered strategies. This covariance is highly unstable over horizons of three to five years. In our exam- ples, risk parity levered to the volatility of a traditional 60/40 portfolio, and bonds levered to the volatility of stocks, the covariances were negative over our 84-year horizon (1929–2012); as a consequence, fixed-leverage versions outperformed the volatility-targeting versions of those strategies. We find no evidence that traditional fully-invested strategies such as 60/40 underperformed risk parity strategies with comparable volatility.