Presentation on theme: "Introduction, Definition, and Methodology June 30, 2014 Note: Powerpoint deck includes many “hidden slides,” which were not used in actual presentation."— Presentation transcript:
Introduction, Definition, and Methodology June 30, 2014 Note: Powerpoint deck includes many “hidden slides,” which were not used in actual presentation. David Laibson
Outline Very quick introductions: Emily, Leana, Matthew, David Very quick introductions: you –Name –School –Fields of interest –Who you started rooting for in the world cup Definition of Behavioral Economics Methodology Seven properties Thumbnail history (for more details look at slides)
If you ask questions that are too aggressive, we’ll use the following system to let you know. =
Semantics Behavioral economics –name irritates people –are there any economists who aren’t studying behavior? Other names you’ll hear: –Psychology and economics –Psychological economics Subfields: –Behavioral Finance –Behavioral Game Theory –Behavioral Public Finance –Behavioral IO –etc…
Definition: Behavioral Economics Behavioral economics is just like the rest of economics, but also includes psychological factors. Adds psychology to economics, particularly cognitive psychology, social psychology, and neuroscience. Buy texts in these fields to learn the psychology a.Schacter, Gilbert, and Wegner, Psychology b.Ross and Nisbett, The Person and the Situation c.Glimcher et al eds, Neuroeconomics Consider taking a couple of intro psych courses (tastes good and good for you)
An obnoxious definition The Guardian: The study of “how people actually make decisions rather than how the classic economic models say they make them.” We don’t apply ideological litmus tests (like rationality or dynamic consistency). Nothing is ruled out or ruled-in ex-ante.
Definition Pay special attention to these psychological factors: –Imperfect rationality –Imperfect self-control –Imperfect selfishnss (social preferences) –But this list is only a start (e.g. psychological conceptions of personality) Emphasize the importance of microfoundations –Preferences –Beliefs –Cognition Take experimental evidence seriously –but don’t rely exclusively on it Vote for Obama
Naïve quasi-hyperbolic agent
(ex-)Regulator-in-chief Cass Sunstein Administrator of the White House Office of Information and Regulatory Affairs
But we also vote for David Cameron (the conservative Prime Minister of the UK) The Behavioural Insights Team “Set up in July 2010 with a remit to find innovative ways of encouraging, enabling and supporting people to make better choices for themselves.” It turns out that behavioral economics has supporters on both sides of the political aisle – e.g., the (US) Pension Protection Act was bipartisan. This legislation championed the use of defaults and auto-escalation.
We also have a dress code
Distinct from... Experimental economics Psychology Behavioralism (we are not Behavioralists) Evolutionary psychology Evolutionary economics (BE takes preferences and cognition as primitives) Sociology and economics Radical economics ‘Economics sucks’ economics Lazy economics Sloppy economics Ad hoc economics
Is behavioral economics a field? No: Few “pure” jobs Difficult job market No journal Why ghettoize? Applied theory is not a field, so why should applied psychology be a field? Yes: Some courses You can take behavioral orals Some seminars Many conferences Some “methodological” fields do exist: econometrics, theory, experimental economics Future field status uncertain.
Our expectation/wish All economists will eventually incorporate behavioral stuff where appropriate. Psychology is to “normal economics” as game theory is to “normal economics.” Everyone uses it as a matter of course.
Methodology Experimental science What makes a good model? [Beware of multiple-testing bias (and p-hacking)]
An illustrative “experiment” A. Pick one of the two options below: 10 euros right now. 15 euros in a week. B. Pick one of the two options below: 10 euros in 10 weeks. 15 euros in 11 weeks. Between subject design: each subject answers only 1 question Within subject design: each subject answers both questions
Lab empirics (experiments) If experiments are run well, they will have high internal validity –I understand the specific causal mechanism that is driving my result –I can turn the result on and off by manipulating the experimental treatment –My result is robust and replicable (not “fragile”) But even a well-run experiment may have low external validity –The mechanism that I am studying is important for particular real-world behaviors Experiments complement (do not substitute for) field research
Internal validity experimental artifacts demand effects (are the subjects trying to respond to the perceived expectations of the experimenter?) External validity unrepresentative subjects under-experienced subjects missing decision aids under-incentivized tasks non-naturalistic problems Thousands of other ways that lab decisions differ from field decisions Problems with internal and external validity in lab experiments.
“The Rules” Adapted from George Loewenstein
Experimental Debriefing (especially for pilots) Aggressively use debriefing surveys. “Was the experiment confusing?” “What strategies did you use?” “How did you come up with your answer?” “What was the experiment about?” “What were the other subjects thinking?” What would your payoff have been if you had gone UP instead of DOWN?”
Experimental odds and ends... Run a pilot (debrief pilot) Randomize order of treatments. Consider measuring expectations and other non- observables. Consider collecting demographic info. Consider measuring cognitive process (aka process tracing).
Field empirics High external, low internal validity (unless you run a field experiment or have some other convincing source of exogenous variation). In the field, it is sometimes hard to pin down the causes of phenomena (e.g., problems of reverse causality and omitted variable biases plague empirical studies). Test multiple predictions to rule out competing hypotheses. Make sure you know exactly how your model is identified.
Don’t glibly overlook rational explanations. But, don’t automatically accept rational actor “just so stories” (in practice rational actor model can be just as ad hoc as behavioral models) When faced with competing explanations take parsimony as one leading guide. Behavioral explanations needn’t be the only explanation. Field empirics continued
Field experiments and lab experiments are complementary Neither is the gold standard They feed off (and stimulate) each other in useful ways Avoid making the mistake of thinking that just because you’ve run a well-designed lab experiment you know how the phenomenon will generalize Avoid making the mistake of thinking that just because you’ve run a well-designed field experiment you know how the phenomenon will generalize
Seven Properties Gabaix and Laibson (2008) These properties typically need to be traded off against each other. No social science model achieves all of these goals. 1.Parsimony 2.Tractability 3.Conceptual insightfulness 4.Generalizability (portability) 5.Falsifiability 6.Empirical accuracy 7.Predictive precision: the model makes sharp predictions.
Figure 1: The value of parsimony. The data (squares) is generated by sin(x/10) + ε, where ε is distributed uniformly between -½ and ½. The sold line fits the first 50 data points to a fifth-order polynomial – a non-parsimonious model. The polynomial has good fit in sample. Sample for estimation of a 5th order polynomial
Figure 1: The value of parsimony. The data (squares) is generated by sin(x/10) + ε, where ε is distributed uniformly between -½ and ½. The sold line fits the first 50 data points to a fifth-order polynomial – a non-parsimonious model. The polynomial has good fit in sample and poor fit out of sample (dashed line). Sample for estimation of a 5th order polynomial
Model = “X+Y > 1” = X Y 1 1 Data = Panel A: Model is falsifiable, empirically consistent, and does not have predictive precision. Model = “(X,Y) = (1,5)” = Data = X Y 1 1 5 Panel B: Model is falsifiable, empirically inconsistent, and has predictive precision. Figure 2: Falsifiability, Empirical Consistency, and Predictive Precision
If physicists wrote theorems like economists: Theorem (existence and uniqueness): Given any initial conditions for a set of mass-points in a vacuum, there exists a unique continuation path that obeys the laws of gravity. This is falsifiable (is it interesting or useful?).
Useful classical physics: Theory: At the surface of the earth gravity causes a constant acceleration of g = 9.8 m/s². Predictive precision: An object projected from the surface of the earth will follow a parabolic path, attaining a height of h = v 2 /(2g) before falling back to the surface (where v is the vertical velocity of the object at t = 0 ).
Predictive Precision in Economics Black-Scholes Option Pricing Formula Auction Theory Solow model with the Kaldor facts Quantity theory of money These theories are not exactly right, but they do make precise quantitative predictions that are almost right.
The Role of Assumptions Models use assumptions – including axioms – to make predictions. Scientific models do not have inviolate axioms. Scientific axioms – even seemingly sacrosanct axioms – are usually modified with time. –Earth is flat –Planets and stars rotate around earth Ptolemaeus vs. Copernicus –Space is three dimensional and Euclidean Newton vs. Einstein
Accuracy vs. Tractability Assumptions need not be 100% accurate to be useful. Some false assumptions are maintained because they enhance the properties of Parsimony, Tractability, Conceptual insightfulness, Generalizability, and Predictive Precision. –The earth is not flat, but flat maps are more tractable than globes. –The earth is not even round – it’s an ellipsoid with grooves and bumps – but the round earth model is a parsimonious approximation.
Economic Assumptions Classical economic assumptions are also useful approximations. –Perfect rationality –Dynamic consistency –Revealed preference These assumptions should be continuously judged on their ability to enhance the seven modeling properties enumerated a few slides back.
A few final thoughts on theory Is it cute math, or are you talking about something potentially real? Is it real but minor? Does it generate non-obvious implications (are they true)? –Beware the ‘multiple-testing’ problem (“half” of the sign implications will be true by chance) –The goal is not to fool your advisor and a few referees Does it explain things that you already knew? Only OK. Does it predict new things that you can confirm? Better. Could it become a workhorse for other economists (is your model a tool economists can use)? Does it truly explain an anomaly or is the success a coincidence?
Empirical scope in science The scope of productive scientific inquiry is always in debate. It is easy to “get ahead” of measurement technologies. –Phrenology (circa 1900) was a premature and counterproductive effort to study the relationship between the brain and human behavior. Nevertheless, empirical scope has a clear trend. Over time, most scientific fields have productively incorporated progressively smaller units of analysis. –Physics: atoms, subatomic particles, strings –Biology: cells, DNA, molecules –Psychology: neurons, neurotransmitters.
Empirical Scope in Economics We anticipate that economics will also successfully follow a similar path of generalization. Economics will incorporate finer and finer levels of analysis – including measurements of activity in the brain. The positive study of human behavior will be advanced by the study of the human brain – the question is when not whether this will happen. We anticipate that over the next thirty years economics will successfully incorporate methods and measurements from biology, neuroscience and genomics.
Domains of evidence Market behavior Laboratory behavior Beliefs Other self-reports Third-party reports Biological and physiological measurement (e.g., neuroimaging, hormones, and genes) Each domain of evidence complements the others, even if our ultimate goal is only to be able to predict market behavior.
Normative Economics Normative economics depends on some axioms that are not subject to empirical verification. These non-empirical axioms are only subject to evaluation with philosophical arguments. Do positive intertemporal preferences and normative intertemporal preferences coincide? How should society trade-off my welfare with someone else’s welfare?
Desirable axioms for welfare economics. Some classical economic axioms are appealing axioms of normative economics, whether or not they turn out not to be desirable assumptions for positive economics. –Dynamic consistency. –Rational maximization. Other classical economic axioms are not good axioms of normative economics. If revealed preference is going to be an axiom of normative economics, we’ll need to simultaneously incorporate a theory of errors into whatever is being revealed. So what is being revealed is a combination of “normative preferences” and other stuff (including mistakes).
An example of normative economics from a behavioral perspective. Normative framework has four components: 1.Positive model of behavior (e.g. β-β-δ discounting with coefficient of relative risk aversion ρ). 2.Some set of normative axioms mapping positive parameters to normative parameters. –Normative utility function is Σ δ t u(c t ) 3.Structural estimation of parameters in (1), which infers both positive parameters and normative parameters. 4.Mechanism design taking account of both positive model of behavior and normative welfare function. ^
A Behavioral Approach to Revealed Preference: Choice (and other measurements) reveal both the normative preferences and the positive model of behavior. Normative axioms explain how you derive one from the other. These normative axioms are arbitrary, but they are just as arbitrary as the classical normative axioms (i.e. classical revealed preference).
The alarm clock If he can’t jog tomorrow, David sets his alarm for 8:00 AM. Otherwise, David sets his alarm for 7:00 AM so he can go for a jog. The alarm wakes him up, but David always lies in bed until 8:00 AM, thereby missing his jog. David’s self-reported preference ranking (at night). –Wake up at 7:00 AM and go for a jog. –Wake up at 8:00 AM (too late to jog). –Wake up at 7:00 AM and nevertheless fail to jog.
The alarm clock Consider these two modeling options: 1.Analyze David’s behavior as revealed preference. –Ignore David’s self-reports. –Assume that he prefers to set the alarm for 8:00 AM when it’s rainy, or snowy, or very hot and humid, or he has a twisted ankle, or his sneakers are lost, or his jogging shorts are in the laundry. –Assume that he prefers to set the alarm for 7:00 and lie in bed until 8:00 in all other states of nature. –Assume that he prefers to never jog. –Model wouldn’t make good out-of-sample predictions: what would he do the day after he stepped on a rusty nail?
The alarm clock 2.Develop a model in which David prefers to jog tomorrow but does not prefer to jog today. –Make assumptions that map positive model to normative preferences. Option two is relatively parsimonious, it is congruent with David’s own explanation of what is going on, and it makes accurate and precise out-of-sample predictions.
Outline Quick introductions Definition of Behavioral Economics Methodology Seven properties Thumbnail history
Thumbnail history... Bounded rationality of Simon succeeded more as rhetoric than as something for economists to do Satisficing wasn’t a precise theory that could be an alternative to mainstream economics Anomalies of the 1950’s and 1960’s did not stop the rational expectations revolution of the 1970’s “the rational model is a good approximation” 1970’s: heyday of “as-if” economics
1970’s 1974: Heuristics and Biases (K&T) –representativeness (similarity heuristic) –availability –anchoring 1979: Prospect Theory –probability weighting function –risk-seeking in the loss domain –risk-avoidance in the gain domain –loss aversion –framing
Representativeness Decision makers use similarity or “representativeness” as a proxy for probabilistic thinking “Steve is very shy and withdrawn, invariably helpful, but with little interest in people, or in the world of reality. A meek and tidy soul, he has a need for order and structure and a passion for detail.” What is the probability that Steve is a farmer, salesman, airline pilot, librarian, or physician? How similar is Steve to a farmer, salesman, airline pilot, librarian, or physician? Subject rankings of probability and similarity turn out to be the same. OK, if similarity predicts true probability. Why might similarity poorly predict true probability?
“Linda is 31 years old, single, outspoken, and very bright. She majored in philosophy. As a student, she was deeply concerned with issues of discrimination and social justice, and also participated in anti-nuclear demonstrations.” Please rank the following statements by their probability, using 1 for the most probable and 8 for the least. –Linda is a teacher in elementary school. –Linda works in a bookstore and takes Yoga classes. –Linda is active in the feminist movement. –Linda is a psychiatric social worker. –Linda is a member of the League of Women Voters. –Linda is a bank teller. –Linda is an insurance salesperson. –Linda is a bank teller and is active in the feminist movement.
– (5.2) Linda is a teacher in elementary school. – (3.3) Linda works in a bookstore and takes Yoga classes. – (2.1) Linda is active in the feminist movement. – (3.1) Linda is a psychiatric social worker. – (5.4) Linda is a member of the League of Women Voters. – (6.2) Linda is a bank teller. – (6.4) Linda is an insurance salesperson. – (4.1) Linda is a bank teller and is active in the feminist movement. Depending on the subject population, 80%-90% rate last item more likely than third to last item K&T call this the conjunction effect (since the conjunctive event receives a HIGHER probability)
Done with naive subjects (undergrads from UBC and Stanford with no background in probability or statistics) Done with intermediate subjects (graduate students in psychology, education and medicine from Stanford, who had taken several courses in probability and statistics) Done with sophisticated subjects (graduate students in the decision science program of the Stanford Business School who had taken several advanced courses in probability and statistics) Results are nearly identical for these three groups Also similarity ranks perfectly coincide with probability ranks
Potential confound: Maybe "Linda is a bank teller," is interpreted as "Linda is a bank teller and is NOT active in the feminist movement." Response: run a between-subject design (in contrast to the within- subject design described above) Specifically, show some subjects (group A) the list without the critical conjunctive event (eighth item). Show other subjects (group B) the list without the critical non- conjunctive events (third and sixth items). Group B ranks "8" higher than Group A ranks "6"
Another experiment (conjunction effect: 68%). Please rank the following events by their probability of occurrence in 1981. (1.5) Reagan will cut federal support to local government. (3.3) Reagan will provide federal support for unwed mothers. (2.7) Reagan will increase the defense budget by less than 5%. (2.9) Reagan will provide federal support for unwed mothers and cut federal support to local governments.
Another experiment (conjunction effect: 72%) Suppose Bjorn Borg reaches the Wimbeldon finals in 1981. Please rank order the following outcomes from most to least likely. (1.7) Borg will win the match. (2.7) Borg will lose the first set. (3.5) Borg will win the first set but lose the match. (2.2) Borg will lose the first set but win the match.
Application: Base Rate Neglect Problem 1: Jack's been drawn from a population which is 30% engineers and 70% lawyers. Jack wears a pocket protector. What is the probability (p 1 ) Jack is an engineer? Problem 2: Jack's been drawn from a population which is 30% lawyers and 70% engineers. Jack wears a pocket protector. What is the probability (p 2 ) Jack is an engineer?
Using Bayes rule can show that, [p 1 /p 2 ][(1- p 2 )/(1- p 1 )] = (3/7)². But, in the lab: [p 1 /p 2 ][(1- p 2 )/(1- p 1 )] = 1. What happens when we give the subjects no information other than base rates? What happens when we change the description to something uninformative like, “Jack went to college.”
Availability (Mechanics were asked...) If a car doesn't start, what is the probability that... battery charge too low starting system defective =.20 fuel system defective ignition system defective =.14 other engine problems mischief or vandalism all other problems =.08
(Mechanics were asked...) If a car doesn't start, what is the probability that... battery charge too low (starting system defective; omitted) fuel system defective (ignition system defective; omitted) other engine problems mischief or vandalism all other problems =.14 (not.42)
Anchoring Kahneman and Tversky's first anchoring experiment: subjects were asked to estimate the percentage of African countries in the UN (% African) first spin Wheel of Fortune to generate random number, R then guess whether “% African” > R then guess “% African” –when spin = 10, mean guess for “% African is 25%” –when spin = 60, mean guess for “% African is 45%”
Jacowitz and Kahneman : Is the Mississippi River more or less than 70 miles long? How long is it? (median response 300 miles) Is the Mississippi River more or less than 2000 miles long? How long is it? (median response 1500 miles) Was the telephone invented before or after 1850? When was it invented? (median response 1870) Was the telephone invented before or after 1920? When was it invented? (median response 1900)
Prospect Theory Subjects who have already been given $1000 are subsequently asked to choose either a certain reward of $500 (84%) or a 50% chance of earning $1000 (16%) A different sample of subjects are given $2000, and asked to choose either a certain loss of $500(31%) or a 50% chance of losing $1000 (69%) a certain reward of $1500 a 50% chance of earning $1000 and a 50% chance of earning $2000
1980’s Endowment effect (Thaler) –“Mugs,” markets, and the passage to economics. Experiments Anomalies Column (Thaler) Behavioral finance Not much formal modeling
1990’s Formalization –Fairness, reciprocity, and social preferences –Intertemporal choice –Learning –Behavioral Game Theory –JDM biases-Quasi Bayesian approaches Self serving bias, Confirmatory bias, Overconfidence Field evidence Acceptance of behavioral economics in the profession
2000+ Clark Medal: Matthew Rabin Nobel Prizes: –George Akerlof (2001) –Daniel Kahneman (2002) –Robert Shiller (2013) Interventions, policy, “nudges” Behavioral IO, development, public finance Behavioral economics starts to feel like normal science (maybe it’s time to sell?)
What will probably be the key growth areas in the coming decades? Theory Field experiments/natural experiments Structural estimation of behavioral models Policy Biosocial science
Outline Introductions Definition of Behavioral Economics Methodology Seven properties Thumbnail history (for more details look at slides)