Habit formation has been proposed as a possible solution for explaining the equity premium puzzle. This paper extends the class of models that support the habits explanation in order to account for heterogeneity in earnings, wealth, habits and consumption. I find that habit formation increases the equity premium. However, contrary to the earlier results in the literature, the habit hypothesis does not imply a price for risk much higher than the one implied by models with intertemporally separable preferences. The main reasons for this are general equilibrium ones. First, with just two assets available, households can smooth out consumption fluctuations very well. Therefore, the higher utility losses of uncertainty imposed by habits will not command a high price of risk because households manage to avoid this risk. Second, the composition of the set of agents pricing the assets is sensitive to changes in the model. In an economy with habits, pricing agents turn out to be households facing very small consumption fluctuations. In addition I characterize three important properties of the model economy that relate to portfolio choice: willingness to hold risky assets (1) increases with wealth, (2) decreases with labor earnings and (3) decreases with habit stock.