A state-price security, also called an Arrow-Debreu security, is a contract that pays one unit of money, conventionally $1, if a specified state of the world occurs at a future date and nothing in any other state. Its price today is the state price for that state. It is a theoretical building block of financial economics: any asset whose payoff depends on which state occurs can be viewed as a bundle of such securities.
The idea is named after Kenneth Arrow, who introduced securities contingent on states of the world in a paper first published in 1953, and Gérard Debreu, who developed contingent commodities in his 1959 book Theory of Value. Both later received the Nobel memorial prize in economics, Arrow in 1972 and Debreu in 1983. A market in which a state-price security exists, or can be built, for every possible state is called a complete market. The nearest real instruments are a binary option and an event contract, which pay a fixed amount when a stated outcome occurs.
Pricing with state prices
If there is a state price for each state , the price of any asset is its payoff in each state times the state price, summed:
- is the price of the asset today.
- is the asset's payoff in state .
- is the price today of a security paying 1 in state only.
Buying one of every state-price security guarantees a payment of 1 whatever happens, so the sum of all state prices equals the price of a riskless bond paying 1, or , where is the risk-free rate for the period. Dividing each state price by that sum gives a set of weights that add to one, known as risk-neutral probabilities.
Worked example
Consider a world with two states next year, up and down, and a risk-free rate of 5 percent. A stock priced at 100 today will be worth 120 in the up state and 90 in the down state. The bond condition gives , and the stock condition gives .
Substituting the first into the second gives , so and . Then . A call option on the stock with a strike of 105 pays 15 in the up state and 0 in the down state, so its price is . The same answer comes from the binomial option pricing model, which rests on this logic. The risk-neutral probabilities here are for each state.
State prices in practice
Real markets have far too many possible states to trade a security for each, but state prices can be estimated. In 1978 Douglas Breeden and Robert Litzenberger showed that a butterfly spread of call options with closely spaced strikes approximates a state-price security for prices near the middle strike. Repeating this across strikes recovers the market's implied probability distribution for the underlying asset, adjusted for risk, from option prices.
A strike ladder of event contracts on the same variable works in a similar way: each contract pays if the outcome falls in or above a range, and their prices together describe the distribution of outcomes. This is why prediction market prices are often read as probabilities, with the caveat that they also reflect risk preferences and fees.
Uses in theory
- Arbitrage pricing. If prices admit no arbitrage, a set of positive state prices exists, a result known as the fundamental theorem of asset pricing.
- Derivative valuation. Models such as the Black-Scholes model can be read as assigning a continuous density of state prices.
- Insurance and risk sharing. Insurance and a catastrophe bond pay mainly in bad states, and their prices reflect the high state prices attached to those states.
- Macroeconomics. General equilibrium models use state-contingent claims to describe how households share risk.
Limitations
Actual markets are incomplete. Many risks, such as individual job loss, cannot be traded, and estimated state prices depend on assumptions about how to fill gaps between strikes and beyond the last listed strike. Option prices also carry bid-ask spreads and are thin far from the current price, which makes the tails of an estimated distribution unreliable.
State prices combine probability with the value placed on money in each state. A dollar paid in a recession is worth more to most investors than a dollar paid in a boom, so state prices for bad outcomes are high relative to their true likelihood. Risk-neutral probabilities derived from them are therefore not pure forecasts, and reading them as such tends to overstate the chance of crashes and other adverse events.