This dissertation presents three essays. The first essay finds that the household risky ratio, the ratio of high risk assets over low risk assets directly owned by households, is a strong negative predictor of the equity premium on the US stock market. The predictability is robust to definition of the asset classes, first versus second half of sample, and the finite-sample bias of Stambaugh (1999). The predictability is stronger than, and not subsumed by popular predictors like price-earnings ratios, yield spread, equity share of issues, or consumption-wealth ratios. The main predictive power is decomposed into three similar parts: 1) the household tilt of risky assets, which is novel and generally orthogonal to known predictors; 2) a valuation ratio component; and 3) an issuance component of high risk versus low risk assets. / Economics