Standard Model

\[\max_c\mathbb{E}\sum_{t=0}^T\delta^{t}u(c_{t}) \]subject to budget constraints

Behavioral models

Potentially one or more of:

• Imperfectly maximized
• Regions of concavity and convexity
• Non-Bayesian
• Present biased
• Loss-aversion
• Other-regardingness

• Knows objective
• \(u(\cdot)\) concave
• \(\mathbb{E}\) from Bayes rule
• Exponential discounting
• \(c\) in levels not relative
• Only own outcome

Standard Approach:

• Derive predictions for how individuals respond to price and income changes

  • Predictions may be ambiguous as they depend on the precise nature of preference

  • The functional form for utility, or quantitative values of preference parameters, are allowed to take on whatever values help fit the data.

• The model does impose many restrictions on behavior, due to assumptions about rationality, concavity, rational updating of beliefs, etc...

• It is still quite vague, though, because preferences are a black box and act as a degree of freedom.

Behavioral approach

• Directly measure standard forms of preferences (e.g., degree of concavity of utility), beliefs.

  • Once preference variation is observed, this becomes a valuable new source of predictions.

  • Preference measures can rank people, but also allow quantitative calibration of parameters, or functional form restrictions. 

• Measure and formalize non-standard forms of preferences, non-Bayesian belief formation

  • e.g., allow for altruism, self-control problems, bounded rationality in optimizing, reference dependent utility

• One way to organize topics in behavioral economics is according to which piece of the standard model is being modified, i.e., optimization, expectations, preferences, et.. (see Rabin, 2002)

Experimental Economics

• Defined by a methodology; includes both lab and field
• By randomly varying \(X\) if we observe \(Y\) the regression \[Y=\beta_0+\beta_1 X + \epsilon\] is well specified if \(X\perp u\), and can illustrate a causal effect of \(X\) on \(Y\)
• Advantages are clear identification of causation
• Disadvantages are in understanding the scope of the effect
  • Why experimental methods often pair well with theory

• Provides a type of evidence that other methods cannot
  • Need a strong argument why it is better than using observational data
• Generate new ideas, or new ways of thinking about things
  • The quantitative results normally taken with a grain of salt
  • Want to show qualitative connections
  • Though some do estimate structural parameters using the lab
•  Qualitative insights can be very revealing
  •  For example show that in some settings people consistently update beliefs in the opposite direction of Bayes rule
• Support one model over an other, qualitatively, in a controlled setting
• Suggests new models of decision making, and new things to look for in observational data.

Good Lab Experiments

Example lab experiment

• The experiment gives people the option to vote about giving money to different causes, at a financial cost
• Measures strength of preferences, tests whether people with stronger preferences are more likely to pay the costs of voting

• Paper proposes a model in which people with stronger preferences about a political issue are more likely to turn out to vote
  • Because they are willing to incur the cost of going to the polls
• The model implies that policies which make voting more difficult, will lead to only people with extreme preferences voting, and greater political polarization

Example lab experiment

Uninteresting lab experiments

A type of uninteresting lab experiment:

• You give people some relatively simple financial incentives corresponding to a model
• You test whether they adhere to the predictions of equilibrium in the model
• Either they match the predictions, in which case referees see that subjects like money, respond to incentives, can add and subtract
• Or they don't match the predictions, in which case referees argue that subjects didn't put that much effort into understanding the rules and incentives

Uninteresting lab experiments

• You need to have an ex ante hunch that people will deviate from the model due to some interesting mechanism, like non-standard preferences, or some systematic bias in reasoning
• Or claim that incentives are relatively complex (in a realistic way) and it is not obvious that people will respond to incentives (e.g., strategies in dynamic games).

Example lab experiment

• Subject knows there are two possible states, Heads or Tails, which correspond to different urn compositions

• There are two stages to the experiment, get paid on green ball

  1. They get a draw from the left urn in the first stage

  2. They get to choose which urn to draw from in the second stage

Heads

Tails

Left Urn

Right Urn

Example lab experiment

• Bayes rule says:
  • stay left in stage 2 if they get a red in stage 1
  • switch to right in stage 2, if they get a green in stage 1.
• Instead, many people stick with left, if they got a green ball from left in the first stage.
• Suggests "reinforcement learning" rather than Bayesian updating 

Left Urn

Right Urn

Experimental Faculty

David Huffman

Example research topics:

• Overconfidence and biased memory
• Complexity and workplace incentives

Colin Sullivan

Example research topics:

• Discrimination in labor markets
• Market design: organ donation

Lise Vesterlund

Example research topics:

• Gender differences in labor market outcomes
• Charitable giving

Stephanie Wang

Example research topics:

• Measurement of beliefs, nature of human belief updating
• Psychology of poverty

Alistair Wilson

Example research topics:

• Assessing behavior in market design
• Equilibrium selection in repeated games

• Our department has a state of the art lab for economics experiments, the Pittsburgh Experimental Economics Lab (PEEL)
• Requires training and IRB permission to run studies there

• The department also hosts the BEDI:

  • Started by Lise Vesterlund

  • Goal to encourage and facilitate behavioral research

  • Particular focus on connecting researchers to practitioners (firms, public sector organizations)

• Funding: Experiments involve costs for incentives etc.

  • Both the PEEL lab and BEDI have pots of money for graduate student experiments

  • Students develop their ideas, and when approved by faculty, can get funding

• Jobs:

  • There are academic job ads that specifically call for behavioral or experimental

  • Many others, esp. applied micro, theory, where behavioral people can apply

Considerations for grad students interested in behavioral/experimental

• Information Pooling in the Household (Priyoma Mustafi)

• Self-serving beliefs about others' effort and preferences for redistribution (Bea Sanhueza)
• Excuse-based procrastination (Marissa Lepper)
• Lasting effects of temporary affirmative action (Neeraja Gupta)
• Hidden cost of affirmative action (Mallory Avery)
• Optimal goal setting for procrastinators (Yiming Liu)

Example student research projects

• Improving understanding of morality

  • Moral wiggle room, excuses

  • Lying costs

  • Image concerns

• ​Improving understanding of sophistication

  •  Are people aware of their own biases?

  • Can people use one bias strategically to offset another?

  • What is complexity?

• Attention and salience has been a big topic recently

• Using psychological factors to increase the power of incentives, or substitute for incentives (recognition, respect, etc.)

• Where do preference differences come from?

• How does culture impact economic behavior?

A few example topics