BEHAVIORAL ECONOMICS 116
examples of Sunk Cost Fallacies
- Richard Thaler gave the following two hypothetical examples the sunk cost fallacy --> " A family pays $40 for tickets to a basketball game to be played 60 miles from their home. On the day of the game there is a snowstorm. They decide to go anyway, but note in passing that had the tickets been given to them, they would have stayed home" ---> "a man joins a tennis club and pays $300 yearly membership fee. After two weeks of playing he develops a tennis elbow. He continues to play (in pain) saying 'i don't want to waste the $300!"
Defining Behavioral Economics
- Behavioral Economists have identified systematic deviations from the rational model called behavioral anomalies - thus, unlike the rational choice model, there is no unifying theory of behavioral economics - so, behavioral economics defines itself in contrast to the rational choice model. - another way to think about it: behavioral economics focuses on how scarce decision resources are allocated
Subject Behavior in lab experiments I
- Experimenter Demand Effects: this refers to the concern that subjects may try to please (or maybe frustrate) the experimenter if they are able to guess what the hypothesis is - this is also why a "neutral context" is good
Testing Theory: Observational Data
- Limitations of using observations data to test theory -> observational data records transactions -> no opportunity for input from researcher to "experiment"
Transaction Utility
- Thaler's notion of "transaction Utility" is defined as: the utility one received for feeling they have received greater value in a transaction than what they have given away in paying for the good - one example: thrill from buying a TV that is half off - could be negative: pain after spending a lot of money on something
Review: Standard Assumption
- Utility is increasing in x1 and x2 - preferences are complete and transitive ---> complete: given any two possible consumption bundles (x1a,x2a) and (x1b,x2b), the consumer prefers bundle A, prefers bundle B or is indifferent the two. No possible pair of bundles exists for which the consumer has no preference - Transitive: given any three bundles, if the consumer prefers (x1a,x2a) to (x1b,x2b) and the consumer prefers (x1b,x1b) to (x1c,x2c) then the consumer cannot prefer (x1cx2c) to (x1a,x2a)
Transaction Utility and Three Behavioral Anomalies
- Utility is typically based on how choices affect your lifetime wellbeing. Behavioral economics has noted that people make choices based on individual transactions - Transaction Utility can explain three behavioral anomalies: 1) sunk cost fallacy 2) flat rate bias 3) reference dependent preferences
Limitations of the Neoclassical Model
- We Typically assume that people have well understood and stable preferences ---> but then what is the role of advertising? - People know the results of her choices ---> but do we even know the quality of an avocado until you get it home? - people will make the best possible choice ---> but we all have made choices we regret. - people are self-interested ---> have you ever done something for someone else with no expectation of anything in return?
science: Hypotheses
- a key aspect of science is that hypotheses can be falsified through observation - first formulate a hypothesis -> hypothesis: If A then B- "A is sufficient for B" -> Hypothesis: B is true only is A- "A is necessary for B" -> Hypothesis: B is true if and only is A- "A is necessary and sufficient for B" - hypotheses cannot be proven to be true... they can only be proven to be false
Review: The Canonical Rational Model
- an individual decision problem over two choices x1 and x2 maximizes: max(x1,x2)U(x1,x2) - subject to a constraint: p1x1 + p2x2 </ y - where U(-) is the utility from consuming x1 and x2 and p1 and p2 are the price of each good and y is the total budget.
how do we mimic the counterfactual?
- because there is no exact replica of a single person we look for a group of people that on average are the same as participants would have been without the program - say you just let teachers sign up for textbooks first-come-first-serve and compare them to teachers who didn't sign up.
Stakes in Lab Experiments
- how do you make incentives "real"? -> pay subjects cash money that depends on decisions - payments must be big enough to be taken seriously, both the average and marginal payoffs -> making searching for the best strategy worth the effort - but you still have to give people a fair chance to falsify the theory --> cannot make the predicted choice too much better than rest - ideally: pay on average about two or three times the prevailing wage in market
Testing Theory: in the lab
- how to estimate demand curve? -> run a lab experiment: give participants a $10 budget, and randomly assign them to different groups * group 1: offer opportunity to purchase good 1 with p1=$1, p2= $1 after having endowed the participants with $10 * group 2: a second treatment could increase p1 while holding the budget and the price of good 2 constant
Interpreting Results
- if consumption of good 1 increased, rational choice model would be rejected - if consumption decreased we would fail to reject the rational model
selection bias
- if we compare outcomes for those with and without the program the difference will have two parts: i)that caused by the program ii) that caused by underlying differences between participants and nonparticipants - our estimate of impact will not, on average, be equal to the tru impact of the program unless ii) is zero and there is no selection bias
Quick Primer on Casual Revolution
- imagine you are a development economist and you want to know the casual impact of an intervention: providing textbooks to school children in Kenya? - what do you need to know to answer that question ideally: what would have happened to those exact children in the parallel universe where they never received textbooks? - what is that called> the counterfactual - do we observe the true counterfactual> no - can we observe a valid counterfactual> yes
implementing Experiments
- implementing both lab and field experiments requires careful consideration of a number of issues - focusing on considerations that are specific to lab experiments here
Place of Behavioral Economics
- important part of describing human behavior (i.e. descriptive economics) - also gained a great deal of traction in settings where people want to apply economics normatively
Lab experiments in economics
- in 2002 the nobel prize in economics went to two pioneers of experimental methods, vernon smith and danny kahneman -> smith sowed economics models can work -> kahneman showed important situations where economics models don't work - in 2018 prize went to richard thaler
Buffets
- in the rational model should the price of a buffet determine how much you should eat? - Thaler theorized that because of transaction utility, the price of a pizza buffet would affect how much pizza people consume
issues with all experiments
- incentives must always be believed --> economists never lie to subject --> this is so we can be as sure as possible that subjects do not doubt what we tell them - subjects' welfare must be protected ---> History: Milgrom study and zimbardo study * psychological harm _---> now : any research must be approved by an institutional review board in advance
example 6: internet auctions
- research finding (Dodonova and Khoroshilov) - internet auction often have buy-now prices. if you are willing to pay that amount, you can have the product immediately instead of waiting to see the outcome the auction - finding: higher "buy now" prices in online auctions increased how much people bid - possible explanations: "buy now" price creates a reference point
example 4: telecommunications
- research finding (Kridel, Lehman, Weisman) > setting: Dallas in the late 1980s > long distance plans: flat-rate v linear pricing for long distance calling > 76% of customers who had the flat-rate plan would have saved money by switching to the linear plan
Implications of Context Dependent preferences
- still possible to manipulate reference points >list price for item on sale > restaurants may put a very expensive item on the menu - consumers make mistakes.
example 5: gym membership
-research finding (della vigna & malmendier) > analyzing 8000 gym records at a gym that had either a $70 monthly fee, or a $10 per visit fee > average member attended 4.3 times per month > could these gym members save money by quitting their gym membership?
steps to conduct a (typical) lab experiment
1) design study and raise funding (a few thousand dollars) 2) get IRB approval 3) Recruit college students to come to the "lab" on campus 4) often run several sessions over weeks 5) sometimes might have same subjects come back for a second session a couple months later 6) analyze data and write papers
steps to conduct a field experiment (measuring impact of a program)
1) find a program implementer (i.e. NGO or government) (optional but most common approach) 2) design study and raise funding (tens of thousands to hundreds of thousands of dollars) 3) get approval from IRB 4) collect baseline data (often household surveys) 5) implement program or study (typically takes months sometimes years) 6) collect data on outcomes (often years later) 7) analyze data and write papers
solving the model
derivative U(x1,x2)/U(x1,x2)= p1/p2 - where -p1/p2 is the slope of the budget constraint and the left hand side is the slope of the indifference curve
Uses of Behavioral Economics
- behavioral economics may be used therapeutically to help people make better decisions --> you might notice you systematically make decisions that are not in your best interest - alternatively, firms may use behavioral economics strategically to exploit consumer biases and increase profits --> convey product information differently to improve profits if customers don't understand of information - or behavioral economics may just give academics a richer understanding of the world - or used by governments to improve policy design and make citizens better off
Theory and Reference-dependence
- can simply add a reference point to a utility function > make transaction utility contingent on how much above or below the price it is >problem: no predictive power - challange: hard to come up with a theory of how people select reference points in a particular setting
Example 2: The concorde paradox and public policy
- concorde project was originally started by the French and British governments in 1962 * supposed to cost $480 million * by 1964 the projects cost $900 million * eventually cost six times what was predicted - then airlines didn't want to buy them so only 14 were ever used
Thaler's model of transaction utility
- consumer problem: U(x1,x2,z(x1,p0))= u(x1,x2) + z(x1,p0) - subject to : p0x1 + p2x2 </ y - where z represents the perception of a good deal
Flat-rate pricing scheme
- consumers now maximize utility, but account for the choosing between flat rate or the linear pricing option in the budget constraint - subject to maxx1,x2U(x1,x2) - where f=1 indicates a fixed price plan and f=0 indicated a linear pricing plan
Lab Experiments in Economics
- decisions have real stakes: earn money based on decision - not allowed to use deception: not allowed to lie to or deceive participants
Don't Observe demand curve
- demand and supply jointly determine price and quantity -> you don't observe enough information to falsify the hypothesis that price of good 1 increasing causes to go down - if you are researcher, you can't raise or lower price of a good - even if you can raise and lower the price, you cant hold supply constant in the real world -> other companies sell same or similar products
Behavioral Economics and Experiments
- describe human behavior that deviates from the neoclassical model - must show such behavior exists - how? often through experiments
External Validity in Lab Experiments
- does what we learn in the lab generalize outside the lab? - we must believe there is some correspondence to the world but how precise that is cannot be known without more study - sometimes we think a "precise representation" to the real world is important (e.g. measuring risk aversion or forming regulatory policy) and sometimes it isn/t (e.g. testing whether game theory works)
Field Experiments in Behavioral Economics
- evaluate the impact of an intervention (i.e. program) that has some relationship to behavioral economics -> intervention is designed to mitigate impact of a behavioral bias * i.e. help people have more self-control -> intervention that takes advantage of behavioral biases to improve outcomes * tell people about a norm, i.e. that everyone else is paying their taxes -> test a theory -> occasionally, actually try to influence behavioural biases * i.e. teach children patience
example 5: beverage demand and references
- experiment (Thaler): note this is an economics experiment that relies on a hypothetical (not real stakes) but from 1980s - subjects are told: you are at the beach. your friend asks how much you would pay for a beer of a specific brand (to go for drinking on the beach) > treatment 1: from a run-down grocery store > treatment 2: from a fancy resort hotel - your friend will but iy fro you if the price is less than or equal to your stated willingness to pay - results: respondents will pay $2.60 for beer from the resort versus $1.50 from beer from the grocery
Example 3: Pizza Buffets and Pricing
- experiment (just and wansink) -> design: 66 individuals at an all you can eat pizza restaurant were given either * treatment 1: a free drink coupon * treatment 2: coupon for 50 % off and a free drink coupon -> results: those offered the half price pizza ate 25% less than those who paid full prize * in both treatment those who reported the pizza as less tasty are more: more bad pizza to get your money's worth? - managers: charging less for buffets may reduce consumption and this potentially increase profit per consumer --> depending on how by how much consumption is reduced - consumers: using behavioral economics "therapeutically" - focus on how much you are enjoying the pizza not on how much you paid for it
Example 1: Theater Tickets and Pricing Programs
- experiments (Arkes and blumer): --> Design: Randomly offered different prices to the first 60 people who were in line to pay full price for season tickets to the campus theatre ---> Treatments: * full price $15 per ticket * some received a $2 per ticket discount * others received a $7 per ticket discount ---> results: * fall semester: those who paid full price attended more plays in the fall semester * Spring semester, all ticket holders were attending plays at the same rate
What are field Experiments?
- experiments influences what decisions are being made or in what context - people make decisions that they would in their everyday life -> may or may not know that they are part of a field experiment - decisions are made in a natural environment - some things cannot be measured/controlled
Behavioral Economics and Firms
- firms as profit maximizers: do firms as economics actors make behavioral optimization mistakes? - firms face completive pressures, so firms that make systematic mistakes exit the market --> utility maximization allows for differences in taste but profit maximizations does not allow for differences in the measurement of profit - that said, firm behavior often accounts for consumers behavioral preferences and biases ---> sometimes they try to act in the best interest of their employees - this places a heavier burden on behavioral economists to prove the existence of behavior that is inconsistent with profit maximizations by firms
Behavioral Anomaly #2: Flat rate bias
- flat rate: a pricing option where price does not depend on quantity or quantity is fixed *monthly bus pass - flat rate bias: a biased preference towards choosing flate rate pricing option
Rational Explanations for the flat-rate bias
- individuals may prefer the flat rate in case of high use "emergencies" and therefore choose a flat-rate plan for its option value > uncertainty and risk aversion (we will take a lot more about choice under uncertainty later in the course) -individuals may not know expected usage >miravete conducts a survey of users in kentucky and people chose a phone plan consistent with expected use > systematically over estimate usage, however *many mistakes are temporary and people do in fact switch once lower usage is recognized > 6 % provide suggestive evidence of flat rate bias
Explanations for the Flat-Rate Bias
- individuals prefer to decouple the payment from the consumption or payment decoupling > avoid negative transaction utility (dissociating the payment from the actual consumption) such as watching taxi meter go up. what about uber? - individuals may have a distance for linear pricing: cognitive cost of each doing CBA for each phone call - individuals may have a problem with self-control (they anticipate using the high flat gym rate as a way to make themselves go to the gym)
Behavioral Anaomaly #3: context dependent preferences
- individuals use a reference point when making decisions - for example, a hamburger that normally costs $5, feels like a good deal if you were offered the same hamburger for $3. but would feel like a rip off it normally costs $2 - the reference point is an example of the context of a choice problem
Internal vs External validity
- internal validity: did you get the correct answer with the parameters of your study? - external validity: does what you learned generalize to outside the study?
Common in public policy
- john mccain on why we should explore mars... "theres too much invested there. There's infrastructure that's very expensive and very extensive" - question: how is government decision making different from individual decision making? - politicians may believe that public will turn on them if they abandon the project -> it may be seen as their failure - may be rational for the politician to continue the project using constituent money to help them politically - so it may be constituents who are subject to sunk cost fallacy
Dealing with "error" in experiments
- lack of understanding/learning in the study -> maybe subjects don't fully understand decisions -> maybe be asked to do a practice trade - errors from boredom -> this is why we try to make stakes high enough - errors from "homemade" decision rules ->sometimes these homemade rules are what you want to study (e.g. altruism) -> sometimes these rules may interfere and you want to suppress them (context about decisions may allow people to fall in mental habits)
Rational Choice Model and Homo Economicus
- need for a simple model drove classical economists to abstract from real behavior by making additional assumptions - the neoclassical model or Homo economicus has been a powerful tool in creating a unifying general baseline theory
Rejecting economic models
- no economic models are perfect - all can be falsified by just the right experiment - when are we satisfied that a model does "well enough" and when do we discard it? - the best way to reject a model is to find another model that predicts behavior much better and is easily replicated
does that mean we need to make experiments more "realistic"
- not necessarily - giving lab experiment a natural contect can bias choices -> a "neutral context" may free subjects to think about issues in a more careful way - we need to control the bits of reality that we may want to study, but keeping the other parts of reality out allows us to have the most powerful tests
How Much should we care?
- one view: Maybe it doesn't really matter than the standard assumptions are often violated ---> the purpose of a model is to be an abstraction - how much we should care depends on the application of the model - if occam's razor suggests the behavioral model is the best model ---> occam's razor says that the simplest explanation is the right one - if we care about the motivation of the decision maker
What are Lab Experiments?
- people (i.e. subjects or participants) are aware they are in an experiment - people are in an artificial environment and are asked to make choices that they would not make absent the experiment - typical subject populations are college students who are invited to a lab to participate in an experiment - experimenter has (nearly) complete control of everything important - lab-in-the-field experiments (Artefactual) field : a lab experiment that uses subjects that are not college students
flat rate bias
- predictions: > no consumer should chose flat-rate pricing if it would be more cost-effective to pay on a per-use basis > no consumer should consume more under linear pricing
Assumptions of the Neoclassical Model
- preferences are logical and complete: Utility is increasing in X and preferences are transitive and complete - Full Information: with full understanding of the world they live in - No limits on cognition: the cognitive ability to identify the best possible choices - no inconsistency from decision-making to action: The complete ability to execute their intended actions - self-interest: Decision makers act as if they are interested only in their own well-being
selection
- programs are started in specific places and at specific times for a reason, they are selected - people choose or select to participate or not participate - this selection process means that participants and non participants are on average different not just because of the program - non participants are often not a good counterfactual -> because people select when to take up a program, participants prior to the program may not be good counterfactual
what is so great about RCTs?
- randomly assign some students to get textbooks and assign the rest to the control group -> this is called a randomized control trial (RTC) or often in economics a field experiment - then both groups should be statistically identical on observable and unobservable characterisitcs
what model is fundamental to (micro)economics? what does that model do?
- rational choice model describes how people make decisions.
Behavioral Economics and Policy
- rational individuals in an efficient market cannot be helped by government policy - in recognizing behavioral economics, however: --> government make recognize people's behavioral biases, and design policies to mitigate their impact - or take advantage of people's behavioral biases to encourage them to behave in a way that optimizes social welfare
Why use students as subject?
- students are not a representative sample, but are more homogeneous than the population - lab experimentalists argue that this is an advantage -> personal differences among people can create unexpected randomness in the data -> a homogeneous subject population can reduce randomness, again adding to both power and goodness of fit * rationale for only including men in drug trials * but that is no longer allowed * field experimentalists increase sample size to increase power - other option: raise more money and get more people in your study
Rational Reasons for context dependent preferences
- the context may convey information > list price may be accurate and useful information > maybe expect hamburgers to be better an a more expensive restaurant - contect creates value >consumer beer in a nice atmosphere
counterfactual
- the counterfactual is what would have happened if program or policy had no been implemented - we never directly observe the true counterfactual ->we can never see the same person with and without the program at the same time - we have to mimic the counterfactual
Subject behavior in lab experiments II
- the effect of being observed: sometimes being observed can itself influence behavior and create incorrect inferences - this is why we try to control studies so that subjects are anonymous and choices are confidential - the best is when neither the other subjects not the experimenter can tell which participant made what decisions. so called "double blind or double anonymous"
lab experiments and economic science
- the history of behavioral economics is inextricably linked to experiment economics (particularly lab experiments) - behavioral economics grew to prominence along with the idea that theory can and should be tested -> often descriptive: aimed at offering people choices and simply observing how they respond - began to be taken seriously in the 1980s and picked up steam in the 1990s and 2000s -> the idea of experimenting in economics is an old one. experiments at the RAND corp tested the "nash equilibrium"
Measuring Impact and the Casual Revolution
- the impact of a program/policy is the difference in outcomes caused by the program - it is the difference between what happened and what would have happened without the program
The Principal of Cost Benefit
- the principal of cost benefit can be defined as: -> a choice should be made if the benefits outweigh the costs -> if multiple choices are mutually exclusive, the choice with the highest net-benefit (benefits minus costs) should be taken - the application of this principal is known as cost benefit analysis (CBA) and is central to all economics -> the central idea of equating marginal cost and marginal benefit (as in supply and demand) can be thought of as an application of CBA -> so long as adding one more unit of consumption has a higher (marginal) benefit than the (marginal) cost of adding one unit of consumption one should continue to consume more
The Role of behavioral economics
- the rational model is the first theory behavioral economists apply when describing individual behavior - only if the rational model becomes impractical or inaccurate do we seek alternative explanations
What is an Experiment?
- the social scientist "controls" the environment in which the observations are made - "control" means you manufacture the "if" part and measure the information you need to falsify your hypothesis -> to test "is A then B" *Formulate hypothesis: if i make A happen b won't * Make sure A is true as well as you can * then see if B appears - if B appears, then hypothesis is falsified
Behavioral Anomaly #1: Sunk Cost Fallacy
- the sunk cost fallacy occurs when -> an individual tries to recover sunk cost by continuing an activity for which there is a negative return -> i.e. chasing bad money with good
Modeling the Sunk Cost Fallacy
- to test for rationality individuals choice of x1* should depend only on their remaining budget y -> if the effect of the budget is small relative to total consumption, then x1* should be independent of wealth - a behavioral model could insert the level of fixed costs p0 as a parameter of the optimal x1* -> consumption will go up if fixed costs go up: dx1*/dp0>0 - however to understand why we can consider a model of transaction utility. specifically thaler calls the utility of getting a good deal, acquisition utility
Experimental & Behavioral Economics
- traditionally, the vast majority of economics research was either: -> theory -> empirical research using observational data - broadly economists do three things with observational data -> study the impact of programs and policy changes -> test theory -> descriptive work (i.e. document the economy)
Problem
- two good consumption problem: the demand curve tells us that the quantity of good 1 demanded depends -> negatively on the price of good 1 -> positively on the price of good 2 -> in some unspecified way on income - need to estimate the demand curve: i.e. the impact of price on quatity
Economics science and the causal revolution
- two sets of researchers in economics that have pushed empirical economics methods forward - (Behavioral) theorists pushing towards "economics science" and testing theory in lab experiments and eventually field experiments - applied microeconomics conducting randomized controlled trials (i.e field experiments) as part of a "casual Revolution" to measure the impact of economic programs -> RCTs are considered the gold standard in identifying casual impacts across the sciences and social sciences
Rationality and Demand Curves
- we could also solve the above problem to obtain two demand functions: --> x1*(p1,p2,y) and x2*(p1,p2,y): the amounts of good 1 and good 2 that will make the consumer as well off as he can be given prices and budget - this model implies a set of relationships between prices and quantities
Rational Explanations for the sunk-cost fallacy
- what are some rational explanations for sunk-cost behavior? -> firms or politicians may care about signaling, i.e. what will people think if i quit? -> people recieve extra utility from finishing * if this is a firm, maybe a managers will consider positive morale effects of finishing * rational, but not typical profit-maximizing assumption
Behavioral Macroeconomics
-People's individual deviations from the rational model in their decision can aggregate up across the economy and affect - behavioral finance: how people's behavioral preferences and biases affect financial markets
What is Economics?
The allocation of scarce resources to provide for unlimited human wants.