Non-Standard Beliefs I

DellaVigna (2009) - three deviations from the standard model

1) Non-standard preferences (U, delta). 2) Non-standard beliefs (p(s)). 3) Non-standard decision making (the max).

Non-standard beliefs - examples

Projection bias, belief in the law of small numbers (gambler's fallacy, overinference), overconfidence.

Projection bias - definition

People project their current tastes onto their future preferences and underestimate how much those preferences will change (Loewenstein, O'Donoghue and Rabin, 2000).

Projection bias - model

Predicted future utility = (1 - alpha) u(c, s) + alpha u(c, current state). alpha = 0: no bias. alpha = 1: thinks future tastes will equal current tastes.

Read and Van Leeuwen (1998) - design

Office workers choose a healthy or unhealthy snack to receive in a week, at a time they will be hungry or satiated. Asked either when hungry (late afternoon) or satiated (after lunch).

Read and Van Leeuwen (1998) - test

Hold expected future hunger fixed and compare by current hunger. Rationally only the future state should matter.

Read and Van Leeuwen (1998) - result

People hungry now chose unhealthy snacks far more often, for both future states. Current state affects future choices, so alpha > 0. Future state still matters too, so the bias is partial (alpha < 1).

Projection bias - underappreciating adaptation

Non-patients predict lower quality of life for conditions like chronic dialysis than patients actually report, because they underestimate adaptation.

Projection bias and the endowment effect

Future owners underestimate how much more they will value an object once they own it (failure to predict adaptation to ownership).

Conlin, O'Donoghue and Vogelsang (2007) - question

Are catalogue buyers over-influenced by the weather on the day they order cold-weather clothing? Rationally, order-day weather should be irrelevant.

Conlin et al. (2007) - data

Over 12 million cold-weather items from a large US outdoor clothing firm, with order and return decisions.

Conlin et al. (2007) - prediction and result

Prediction: colder order day or warmer return day means more returns. Result: lower order-date temperature and order-date snowfall raise return probability.

Conlin et al. (2007) - alternative explanations ruled out

No pattern for non-cold-weather items (so not mood). Robust to household fixed effects (not customer differences). Current weather predicting future need would reduce returns, not raise them.

Bayes' rule

P(A|B) = P(B|A) P(A) / P(B). Posterior = likelihood x prior / probability of the evidence.

Representativeness heuristic

Judging probability by how much something resembles a familiar pattern or category (Tversky and Kahneman, 1974). E.g. shy, tidy Steve judged a librarian even though farmers are more common.

HTHTTH vs HHHTHH

Both equally likely (1/64), but the first 'looks' more random. Shows representativeness.

Belief in the law of small numbers

Mistaken belief that small samples should closely resemble the population. People underestimate random variation in small samples.