Risk aversion is the well documented psychological bias that makes us refuse to participate in zero-sum lotteries: to accept a 50 % chance loss of one dollar, we need to be offered a 50 % chance gain of more than 1 dollar, e.g. two dollars instead of one – a compensation for taking risk. Through the same mechanism, investors must be rewarded for accepting to take on financial risk: they are compensated by receiving greater average returns than risk-free assets – the equity risk premium. We have evidence that this compensation for risk varies over time: investors can expect to receive greater returns when the indice levels are low relative to distributed dividends than when they are high. Does it come from time variations in risk aversion or are there other forces at play? Indirect evidence suggests not only the compensation for risk but also the quantity of risk in financial markets may vary over time. Variations in risk aversion would thus not be the sole channel for changes in equity risk premia over time, though it may still play an important role. Disentangling the two sources of variations in the price of risk – risk aversion and quantity of risk – and understanding why and how they may vary over time is at the core of the recent research in asset pricing theory.