Chapter 31 Quiz: Decision Markets and Futarchy

Instructions

Select the best answer for each question. Some questions may have multiple correct aspects, but choose the most complete or accurate response.

Question 1

What is the core principle of futarchy as proposed by Robin Hanson?

A) All government decisions should be made by financial markets B) Citizens vote on welfare metrics, and prediction markets determine which policies maximize those metrics C) Expert panels should replace democratic voting for policy decisions D) Markets should advise governments, but elected officials retain all decision-making power

Answer: B Explanation: Hanson's futarchy proposal specifically separates the "values" question (what should we optimize, decided democratically) from the "beliefs" question (which policy achieves that optimization, decided by markets). The slogan is "vote on values, bet on beliefs."

Question 2

In a conditional prediction market, what happens to positions in the market for a decision that is NOT chosen?

A) Traders lose their entire investment B) Positions are resolved at the current market price C) Positions are voided and traders are refunded at their purchase price D) Positions are automatically converted to the winning decision's market

Answer: C Explanation: Conditional markets that correspond to decisions not taken are "called off" -- positions are returned at cost. This is essential to the mechanism because it ensures traders only bear risk in the market that actually resolves.

Question 3

The price in a conditional prediction market $P_A = \mathbb{E}[Y \mid D = A]$ represents:

A) The causal effect of decision A on outcome Y B) The market's belief about Y in states of the world where A is chosen C) The unconditional expected value of Y D) The cost of implementing decision A

Answer: B Explanation: The conditional expectation conditions on everything correlated with D=A, not just the direct causal effect of A. This means it reflects both the causal effect and any selection effects -- the market's belief about Y in the states of the world where A would be chosen.

Question 4

Which of the following is the primary obstacle to interpreting conditional market prices as causal effects?

A) Market liquidity is always too low B) Traders are irrational C) The decision may be correlated with confounders that also affect the outcome D) Conditional markets cannot handle continuous outcomes

Answer: C Explanation: The core causal inference problem is that the decision D may be correlated with unobserved factors (confounders) that independently affect Y. This is the selection bias problem: E[Y|D=A] mixes the causal effect of A with the baseline conditions under which A is chosen.

Question 5

Under what condition does a decision market's conditional price equal the true causal effect?

A) When the market has at least 100 traders B) When the decision is independent of potential outcomes conditional on market information (ignorability) C) When all traders are risk-neutral D) When the LMSR liquidity parameter is sufficiently large

Answer: B Explanation: The ignorability (unconfoundedness) assumption Y(d) ⊥ D | I_market is the formal condition under which conditional expectations equal causal effects. This is analogous to the identification assumption in the Rubin causal model.

Question 6

In the Hewlett-Packard corporate prediction market experiments, which finding was most notable?

A) Markets required at least 100 traders to be useful B) Markets with only 8-12 traders significantly outperformed official sales forecasts C) Real money was essential for market accuracy D) Markets failed when participants were from different departments

Answer: B Explanation: The HP experiments (Chen and Plott, 2002) showed that even very small markets (8-12 traders) could outperform HP's official forecasts, primarily by aggregating information from different departments that was not effectively shared through normal channels.

Question 7

What is the "buying a decision" problem in decision markets?

A) The cost of participating in the market is too high B) A trader with outside interests can profit by manipulating the market to change the decision, even if they lose money on the market itself C) The decision-maker does not have enough money to buy shares D) Decisions are too expensive to implement

Answer: B Explanation: The "buying a decision" problem is unique to decision markets. A manipulator's cost of distorting market prices may be small compared to the benefit they receive from the resulting policy change, making manipulation profitable on net even though the market trades themselves lose money.

Question 8

Which of the following is Hanson's argument for why manipulation of decision markets is less problematic than it appears?

A) Manipulation can be legally prohibited B) Manipulators subsidize information aggregation by trading against informed traders, and counter-manipulation is equally cheap C) Market makers can detect and reject manipulative orders D) Manipulation requires coordination among multiple traders, which is inherently unstable

Answer: B Explanation: Hanson argues that manipulation is costly (manipulators lose money to informed traders, effectively subsidizing information production), improves liquidity, and is a double-edged sword (counter-manipulation by harmed parties is equally feasible).

Question 9

The "thin market problem" in conditional decision markets arises because:

A) Only a few people understand how prediction markets work B) Conditional markets split available liquidity across multiple decision alternatives C) Market makers are unwilling to provide liquidity D) Regulatory restrictions limit market participation

Answer: B Explanation: With k decision alternatives, the total liquidity and trading interest is split k ways. Each conditional market has approximately 1/k of the liquidity that a single unconditional market would have, leading to noisier prices and greater vulnerability to manipulation.

Question 10

In Meta-DAO's futarchy implementation, what serves as the welfare metric?

A) Revenue generated by the DAO B) Number of active contributors C) The META governance token price D) A composite index of DAO health metrics

Answer: C Explanation: Meta-DAO uses the META token price as its welfare metric. This is controversial because it reflects short-term market sentiment and may not capture the long-term health of the DAO, but it has the practical advantage of being continuously observable and relatively hard to manipulate.

Question 11

Which of the following best describes why the welfare metric choice is the "Achilles' heel" of futarchy?

A) Welfare metrics are too difficult to measure B) Defining what constitutes societal welfare is a fundamental values question that futarchy claims to bypass C) Markets cannot predict welfare metrics accurately D) Welfare metrics change too rapidly to be useful

Answer: B Explanation: Futarchy separates "values" (what to optimize) from "beliefs" (how to optimize it), but defining the welfare metric is itself a profound values question. The choice of metric, its components, and their weights embed ethical judgments that cannot be resolved by markets.

Question 12

In an LMSR-based conditional market with liquidity parameter $b = 1000$, approximately how much would it cost to move a binary outcome price from 0.50 to 0.60?

A) About $10 B) About $100 C) About $1,000 D) About $10,000

Answer: B Explanation: In an LMSR, the cost of moving the price is approximately b × |p' - p| for small movements. So 1000 × 0.10 = $100. (The exact cost depends on the full cost function calculation, but this linear approximation gives the right order of magnitude.)

Question 13

Which approach to conditional market implementation preserves liquidity across decision conditions?

A) Separate conditional markets B) Combinatorial market with conditional tokens C) Independent order books D) Sequential auctions

Answer: B Explanation: Combinatorial markets with conditional tokens (like the Gnosis CTF) operate on the joint distribution of decisions and outcomes. Liquidity is shared across all conditions because the market maker prices the joint space, rather than operating separate pools for each condition.

Question 14

The "randomized implementation" defense against manipulation works by:

A) Randomly selecting traders to participate B) Randomly ordering the decision alternatives C) With some probability, ignoring the market and implementing a random decision D) Randomly determining the resolution time

Answer: C Explanation: By implementing a random decision with some probability p (regardless of market prices), the mechanism reduces the expected payoff from manipulation. The manipulator pays the cost of distorting prices but only gets the desired decision with probability (1-p), making manipulation less profitable.

Question 15

Google's internal prediction markets demonstrated which of the following?

A) Markets were perfectly calibrated and always outperformed experts B) Markets were well-calibrated overall but showed an optimism bias C) Markets were poorly calibrated but still useful for decision-making D) Markets were accurate only when real money was used

Answer: B Explanation: Cowgill, Wolfers, and Zitzewitz (2009) found that Google's markets were well-calibrated overall but showed a slight optimism bias -- prices were systematically too high for positive outcomes, possibly because enthusiastic participants self-selected into the market.

Question 16

In the potential outcomes framework, the selection bias in a decision market is defined as:

A) $\mathbb{E}[Y(A)] - \mathbb{E}[Y(B)]$ B) $\mathbb{E}[Y(B) \mid D=A] - \mathbb{E}[Y(B) \mid D=B]$ C) $\mathbb{E}[Y \mid D=A] - \mathbb{E}[Y(A)]$ D) $P(D=A) - P(D=B)$

Answer: B Explanation: Selection bias is the difference in baseline potential outcomes between the group that receives treatment A and the group that receives treatment B. It equals E[Y(B)|D=A] - E[Y(B)|D=B], which measures how the "untreated" potential outcome differs across groups, reflecting the non-random selection into treatment.

Question 17

Which of the following is NOT a proposed solution to the thin market problem?

A) Market subsidization through funded market makers B) Reducing the number of decision alternatives considered C) Increasing the number of conditional markets D) Using combinatorial markets that share liquidity

Answer: C Explanation: Increasing the number of conditional markets would worsen the thin market problem by splitting liquidity even further. The solutions involve concentrating liquidity (fewer markets, combinatorial sharing, subsidization) or attracting more traders (incentives, focused designs).

Question 18

What is the key difference between a "decision market" and an "advisory prediction market"?

A) Decision markets use real money while advisory markets use play money B) In a decision market, the market price directly determines the decision; in an advisory market, the price merely informs human decision-makers C) Decision markets are for government use while advisory markets are for corporate use D) Decision markets resolve faster than advisory markets

Answer: B Explanation: The critical distinction is whether the market's output is binding. A decision market has a pre-committed rule mapping prices to decisions. An advisory market provides information to a human decision-maker who retains discretion. This distinction matters because it changes the incentive structure for traders.

Question 19

The Gnosis Conditional Token Framework (CTF) allows traders to:

A) Only trade on government policy decisions B) Create tokens representing any combination of conditions and outcomes C) Trade anonymously without any market maker D) Override DAO governance votes

Answer: B Explanation: The Gnosis CTF is a general-purpose framework for creating tokens that represent claims on arbitrary combinations of conditions and outcomes (e.g., "Proposal X passes AND metric Y exceeds threshold Z"). This flexibility makes it suitable for implementing various forms of conditional and decision markets.

Question 20

Scoring-rule-based decision markets differ from traditional market-based approaches in that:

A) They use quadratic scoring instead of logarithmic scoring B) Each trader independently reports a probability distribution, and a scoring rule incentivizes truthful reporting, without a trading mechanism C) They score traders on speed of trading rather than accuracy D) They require all traders to have identical prior beliefs

Answer: B Explanation: In a scoring-rule-based approach, there is no market mechanism. Each trader independently reports their belief distribution, which is scored using a proper scoring rule (like the log scoring rule) based on the actual outcome. The aggregate prediction is formed by combining reports. This avoids some market mechanism issues but loses the price discovery benefits of trading.

Question 21

Which of the following best describes the "Goodhart's Law" concern with futarchy?

A) Markets always predict the wrong outcome B) Once a welfare metric is used as a governance target, agents will optimize for the metric rather than genuine welfare, degrading the metric's quality C) Prediction markets violate Goodhart's original financial regulations D) The law prevents futarchy from being legally implemented

Answer: B Explanation: Goodhart's Law states: "When a measure becomes a target, it ceases to be a good measure." In futarchy, the welfare metric becomes the optimization target. Agents (including policy-makers, citizens, and market participants) will find ways to improve the measured metric without necessarily improving actual welfare.

Question 22

In Meta-DAO's implementation, why is the token price measured at a random time after the decision?

A) To save computational costs B) To prevent manipulation at a known resolution time C) Because blockchain timestamps are unreliable D) To give traders more time to adjust positions

Answer: B Explanation: If the resolution time were known in advance, a manipulator could time a large trade to temporarily move the token price at the exact resolution moment. Random resolution timing makes this manipulation strategy much more expensive because the manipulator would need to sustain the price manipulation continuously.

Question 23

What is the multiple equilibria problem in decision markets?

A) Markets can only reach one equilibrium, limiting flexibility B) There can be self-fulfilling prophecy equilibria where the market "locks in" to a decision regardless of true causal effects C) Multiple market makers compete and destabilize prices D) Traders cannot agree on which equilibrium concept to use

Answer: B Explanation: In decision markets, if traders believe decision A will be chosen and trade accordingly, their trading can make A appear superior, causing A to actually be chosen -- a self-fulfilling prophecy. This can happen even if B has a higher true causal effect, creating multiple equilibria.

Question 24

Which aspect of corporate prediction markets was found to be surprisingly effective?

A) Using real money with large stakes B) Play money performing comparably to real money in many settings C) Requiring mandatory participation from all employees D) Limiting markets to senior executives only

Answer: B Explanation: Research on corporate prediction markets (including at Google and other companies) found that play money markets performed surprisingly well -- often comparably to real money markets. This was unexpected and suggests that competitive motivation and reputation effects can substitute for financial incentives.

Question 25

What is the strongest theoretical argument for why decision markets can recover causal effects despite the selection problem?

A) Markets are always efficient B) When the decision is determined solely by market prices, traders internalize the decision rule and report beliefs about the causal effect rather than the conditional expectation C) Traders can directly observe the causal effect through insider information D) The law of large numbers eliminates selection bias

Answer: B Explanation: The key theoretical result is that when traders know the decision rule (D is chosen based on market prices), they can reason backward: "My trade affects the price, which affects the decision, which affects what I'm conditioning on." In equilibrium, price-taking traders effectively report the causal estimate because the conditional expectation given "this decision is chosen because the market says it's best" converges to the causal effect under certain conditions.