tibsfox/gsd-skill-creator/behavioral-economics

Behavioral Economics

Standard economics assumes agents are rational: they have stable preferences, process information correctly, and maximize expected utility. Behavioral economics documents the systematic ways real humans depart from this ideal and builds models that incorporate these departures. The field was launched by Kahneman and Tversky's work on heuristics and biases in the 1970s and formalized by prospect theory (1979). It is not a rejection of economics but an enrichment -- the rational model is the benchmark from which behavioral findings are measured.

Agent affinity: varian (pedagogical exposition, connecting behavioral findings to standard theory), sen (welfare implications, capability approach as alternative to revealed preference), ostrom (behavioral foundations of cooperation in commons)

Concept IDs: econ-marginal-thinking, econ-trade-offs, econ-opportunity-cost, econ-market-failures

The Behavioral Economics Toolbox at a Glance

Topic 1 -- Heuristics

What they are. Heuristics are mental shortcuts that reduce complex problems to simpler judgments. They are often accurate and always efficient, but they can produce systematic errors (biases) in specific, predictable circumstances.

Anchoring. People's estimates are influenced by irrelevant starting points. Tversky and Kahneman (1974) asked subjects whether the percentage of African countries in the UN was higher or lower than a randomly generated number, then asked for their best estimate. Subjects anchored heavily on the random number. In economic contexts, anchoring affects negotiations (the first offer sets the anchor), pricing (suggested retail price), and financial forecasts.

Availability. People judge the probability of events by how easily examples come to mind. Plane crashes are vivid and memorable, so people overestimate aviation risk relative to driving risk -- even though driving is far more dangerous per mile. Availability distorts risk assessment, insurance demand, and policy priorities (resources flow to dramatic risks rather than statistical ones).

Representativeness. People judge probabilities by similarity to a prototype rather than by base rates. The "Linda problem" (Tversky and Kahneman, 1983): Linda is a former philosophy student concerned with social justice. Is she more likely to be (a) a bank teller, or (b) a bank teller and active in the feminist movement? Most people choose (b), which is a conjunction fallacy -- the conjunction of two events cannot be more probable than either event alone. In economic contexts, representativeness drives stereotyping in hiring, overreaction to salient narratives in financial markets, and the "hot hand" fallacy.

Worked heuristic example. An investor reads that a tech startup has tripled revenue for two consecutive years. Using representativeness, the investor judges the company as "the next Google" because the growth pattern matches the prototype of a breakout tech company. Base rate neglected: 90% of startups with early rapid growth fail within five years. The investor overweights the salient narrative and underweights the base rate, overpaying for the stock. This is a general pattern in financial markets -- stories move prices more than statistics, especially for individual investors. Professional investors are not immune but markets partially correct through arbitrage.

Base rate neglect. The general form of representativeness errors. A medical test with 99% sensitivity and 1% false positive rate sounds highly reliable. But if the disease prevalence (base rate) is 1 in 1,000, a positive result is more likely to be a false positive than a true positive: P(disease | positive) = 0.99 0.001 / (0.99 0.001 + 0.01 * 0.999) = approximately 9%. Doctors, judges, and jurors routinely make this error when evaluating diagnostic and forensic evidence. In economic contexts, base rate neglect leads to overconfident credit assessments, insurance mispricing, and investment mistakes.

The affect heuristic. People make judgments based on emotional reactions rather than deliberate analysis. If a technology feels exciting (AI, blockchain), people overestimate its benefits and underestimate its risks. If a technology feels scary (nuclear power, genetic modification), the opposite occurs. The affect heuristic explains asymmetric risk regulation: nuclear power is regulated far more stringently than coal power per unit of mortality risk because nuclear feels scarier.

Topic 2 -- Biases

Loss aversion. Losses loom larger than equivalent gains. Kahneman and Tversky estimated the loss aversion coefficient at approximately 2: losing $100 feels about twice as bad as gaining $100 feels good. This explains the endowment effect (people demand more to sell an object than they would pay to buy it), the disposition effect in finance (investors sell winners too early and hold losers too long), and resistance to policy changes (the prospect of losing existing benefits outweighs the prospect of gaining new ones).

Status quo bias. People disproportionately stick with the current state of affairs. This is partly loss aversion (change involves potential losses), partly the effort cost of switching, and partly a heuristic that the status quo has survived this long so it must be acceptable. Status quo bias explains low rates of organ donation in opt-in systems, inertia in retirement savings, and resistance to institutional reform.

Overconfidence. People systematically overestimate their knowledge and abilities. Calibration studies show that when people say they are 90% confident, they are right about 70-80% of the time. Overconfidence drives excess trading in financial markets (Barber and Odean, 2001), overoptimistic business plans, and insufficient precautionary behavior. The planning fallacy -- systematic underestimation of how long tasks will take -- is a special case.

Sunk cost fallacy. Rational agents ignore sunk costs -- what is spent is spent and should not affect future decisions. But people persist in failing endeavors because they have already invested time, money, or effort. "We've come this far" is not a valid economic argument for continuing, but it is a powerful psychological one. The Concorde fallacy (named after the supersonic jet that kept flying despite never being economically viable) illustrates the cost: billions were spent because governments could not bring themselves to write off earlier investment.

Worked bias example. A company has invested $10 million in a software project that is now clearly failing -- the technology is obsolete and a competitor has released a superior product. A rational analysis says: the $10 million is sunk. The question is whether future investment will generate returns exceeding future costs. If not, abandon the project. But managers feel the $10 million as a loss and continue investing to "justify" the sunk cost, eventually losing $20 million instead of $10 million. The bias is strongest when the same person who authorized the initial investment makes the continuation decision -- they are "throwing good money after bad" to avoid admitting an error. De-biasing strategies: separate the continuation decision from the initial decision (have a different person evaluate), use pre-commitment rules ("we will evaluate this project at month 6 with a fresh cost-benefit analysis"), and normalize failure ("killing a bad project is a success, not a failure").

Confirmation bias. People seek information that confirms existing beliefs and avoid information that contradicts them. In economic contexts, investors seek news that validates their holdings, managers seek data that supports their strategies, and policymakers seek evidence that justifies their positions. Confirmation bias is especially dangerous in financial markets, where it can sustain bubbles: investors who bought at high prices seek bullish analysis that confirms their purchase was wise, creating demand for optimistic forecasts.

Topic 3 -- Prospect Theory

The framework. Prospect theory (Kahneman and Tversky, 1979) is the most influential alternative to expected utility theory. It has three components:

1. Reference dependence. People evaluate outcomes as gains or losses relative to a reference point, not as final wealth levels. The reference point is typically the status quo but can be an expectation, an aspiration, or a social comparison.

1. Diminishing sensitivity. The value function is concave for gains and convex for losses -- the difference between $0 and $100 feels larger than the difference between $1,000 and $1,100. This produces risk aversion for gains and risk seeking for losses.

1. Probability weighting. People overweight small probabilities and underweight large ones. This explains both lottery purchases (overweighting the tiny chance of a large gain) and insurance purchases (overweighting the small chance of a large loss) -- behaviors that are inconsistent under expected utility theory.

The value function. The value function is S-shaped: concave above the reference point (gains), convex below (losses), and steeper for losses than gains (loss aversion). This single function explains a wide range of anomalies: the endowment effect, risk-seeking in the domain of losses, the equity premium puzzle, and asymmetric reactions to good and bad news.

The isolation effect. People simplify choices by ignoring components that alternatives share. This can produce inconsistencies: the same final outcome is evaluated differently depending on how the problem is framed (whether identical components are included or stripped out).

Topic 4 -- Intertemporal Choice

Hyperbolic discounting. Standard economics assumes exponential discounting: a person who prefers $100 today over $110 tomorrow should also prefer $100 in 30 days over $110 in 31 days. But people often reverse their preference as the earlier date approaches -- they are patient when both options are in the future but impatient when the earlier option is now. This "present bias" is captured by hyperbolic (or quasi-hyperbolic) discount functions.

Self-control problems. Present bias creates a gap between what people plan to do and what they actually do. People plan to save, diet, and exercise -- then do not follow through. Sophisticated agents recognize their future self-control problem and seek commitment devices (automatic payroll deductions, gym contracts with penalties, deadline commitments). Naive agents do not anticipate the problem and are perpetually surprised by their own behavior.

Thaler's mental accounting. People compartmentalize money into mental accounts (rent money, vacation money, emergency fund) and apply different decision rules to each. This violates fungibility -- a dollar is a dollar regardless of which mental account it comes from -- but it serves as a self-control mechanism. Blowing the vacation budget feels different from raiding the emergency fund, even if the economic consequence is identical.