Risk Aversion

Level I formalises risk aversion through a single-period utility score: expected return minus one-half times the risk-aversion coefficient times the variance of returns. The coefficient scales how heavily volatility is penalised. A positive coefficient describes a risk-averse investor, zero describes risk neutrality where only expected return matters, and a negative value describes a risk seeker who is willing to pay for volatility. Higher coefficients mean each unit of variance destroys more utility, so a given portfolio receives a lower score.

Two mechanical points matter in calculations. First, the inputs must be consistent: if expected return is entered in decimal form, variance must be in decimal form squared, and mixing percentages with decimals inflates or deflates the penalty by orders of magnitude. Second, the score is an ordinal ranking device, not a return forecast. A utility of 0.06 does not mean the investor expects 6 percent; it means that portfolio ranks above one scoring 0.05 for that particular investor.

Plotting all portfolios that deliver the same utility score in expected-return-and-standard-deviation space produces an indifference curve. Because variance grows faster than standard deviation, these curves are upward sloping and convex: each additional unit of risk requires progressively more expected return as compensation. The steepness of the curves is the visual signature of the coefficient. A highly risk-averse investor demands a large return increment for a small volatility increment, producing steep curves; a mildly risk-averse investor produces flatter ones.

The optimal portfolio is where the highest attainable indifference curve just touches the capital allocation line. Algebraically, the weight placed in the optimal risky portfolio equals the excess return of that portfolio over the risk-free rate divided by the product of the risk-aversion coefficient and the portfolio's variance. This is why two investors facing identical capital market expectations hold the same risky portfolio but in different proportions against cash: risk aversion determines the mix, not the composition.

Aggregate risk aversion is observable in the premiums embedded in market prices. Equities trade at expected returns above government bills because investors as a group require compensation for cash-flow uncertainty. Longer-maturity government bonds typically offer a term premium over short bills for accepting interest rate and inflation uncertainty. Corporate spreads compensate for default and liquidity risk beyond expected loss. Insurance premiums exceed expected claims because policyholders willingly pay to convert an uncertain large loss into a certain small cost.

Crucially, aggregate risk aversion is not constant. Periods of stress widen credit spreads and equity risk premiums sharply even when fundamental default expectations move little, which is interpreted as a repricing of risk aversion itself. Falling premiums during calm expansions reflect the reverse. This variability is why practitioners treat required returns as time-varying inputs rather than fixed constants.

Expect to compute utility scores for several candidate portfolios given a coefficient, then identify the highest-ranked one, and to reverse the exercise by inferring which coefficient makes an investor indifferent between two portfolios. Expect to determine the allocation between the risk-free asset and the optimal risky portfolio from the coefficient, the excess return and the variance, including cases where the resulting weight exceeds one and implies borrowing at the risk-free rate to lever the risky portfolio.

Also maintain clean vocabulary. Risk aversion is a preference parameter. Risk tolerance is its practical inverse and, in an investment policy statement, blends willingness with the ability to bear risk, the latter driven by wealth, horizon and liabilities. Loss aversion, from behavioural finance, is asymmetric sensitivity to losses relative to gains around a reference point and can coexist with risk-seeking behaviour after losses, so it is not a synonym for the mean-variance coefficient.