PAM 3130 Midterm

Quasi Experiments

"Natural" experiment: -changes in the economic environment that create nearly identical treatment and control groups for studying the effect of environmental change, allowing researchers to take advantage of randomization created by external forces

If Y is in logs and X is continuous, how to interpret coeff. B?

A one unit change in X generates a 100*B percent change in Y

Diff-in-diff

(treatment after - treatment before) - (control after - control before)

Instrumental variables analysis (?)

-If we are worried that our assignment to a treatment group isn't random, we find some third variable (the "instrument") that affects entry into the treatment group, but is not correlated with the outcome variable -Key assumption: The instrument does not affect the dependent variable independently from the likelihood of the independent variable occurring

Assumptions of the Standard Model

-Individuals motivated by expected utility maximization -Utility governed by selfish concerns -Individuals have time consistent preferences -Framing should not matter, only info -All income and assets are fungible

What does the standard model suggest about why status quo framing matters?

-Preferences are constant and you are facing the same choice -Transition costs could be high -May assume status quo is the status quo for a good reason

Prospect Theory: 3 main features

1) Evaluation of choices are made relative to a reference point 2) Diminishing sensitivity (risk-averse in gains, risk-loving in losses) 3) Loss Aversion

How to calculate the coefficient for loss aversion?

1) WTA/WTP 2) X/Y 50-50 gamble: measures whether you would accept a bet where you have a 50% chance of winning X and a 50% chance of losing Y (i.e., if coeff.=2, you would accept a bet where u have a 50% chance of winning $200 and a 50% chance of losing $100)

Gneezy and Rustichini: Daycare Experiment

10 schools: in 6, parents were charged each time they picked kids up late, then after 10 weeks the fine was removed Results: # parents picking up kids late incrased after the fine was implemented. Also was a permanent effect --> after getting rid of the fine, # of late pickups didn't decrease. Explanation: maybe parents view this as cheap babysitting --> crowds out their intrinsic motivation to pick up their kid on time.

Mellstrom and Johanneson (2008): Crowding out in Blood Donation - Was Titmuss Right?

2 Hypotheses, tested separately on men and women: 1) Supply of blood will be less with a payment (crowding out) 2) Supply of blood higher when monetary reward can be given to charity (image motivation) Overall Sample: No significant decrease in blood supply due to crowding out, and no significant increase in supply due to the charity option Women: Evidence of crowing out of intrinsic motivation, but this crowding out disappears when people can donate the reward to charity Men: No evidence of crowding out, and the charity option decreases blood donation Used a diff-in-diff to see differences between men and women

DellaVigna and Malmendier (2006): Paying Not to Go to the Gym

2 different gym contracts: -pay as you go: $10 per visit -monthly membership: $70/month Standard model says buy membership if you plan to go more than 7 times per month --> but avg. member only goes 4 times Could show both types of present bias: 1) Naive: over-optimism - didn't realize they were present biased so thought they'd go more 2) Sophisticated: commitment device - maybe without a contract they would have gone even less

Lacetera, Macis & Slonim (2010): Will There Be Blood?

2 studies: natural experiment with existing incentives, RCT varying incentives Question: do incentives increase or decrease blood donation? 5 groups: 1-3) Treatment: $5, 10 or 15 4) Received "surprise" incentive 5) No incentive Collected data on donations in treatment and control drives, before and after the experiment begins. Alternative explanations: supports standard model, but may not increase overall blood donation because people from other blood drives may just come to the one that they get paid at (other blood drives not in experiment had decreased attendance).

Ariely & Wertenbroch (2002): Procrastination

3 questions: 1) Are ppl willing to self-impose costly deadlines to overcome procrastination? 2) Are self-imposed deadlines effective in improving task performance? 3) When self-imposing deadlines, do people set them optimally for maximum performance enhancement? Method: Students have to proofread 3 texts due on certain dates. Paid $0.10 per correction, lose $1/day if late 3 groups: 1) even-spaced deadline: one due every 7 days 2) End deadline: 3 due after 21 days 3) self-imposed deadline Results: Evenly-spaced deadlines were most effective, then self-imposed, then end deadline. Evenly-spaced deadline seem to solve sophisticated present bias, end deadlines show naive present bias

If Y is in logs and X is in logs, how to interpret coeff. B?

A 100% change in X geenrates a 100*B percent change in Y

If Y is binary and X is in logs, how to interpret coeff. B?

A 100% change in X generates a 100*B percentage point change in the probability Y=1 occurs

If Y is continuous and X is ln(continuous X), how to interpret coeff. B?

A 100% change in X generates a B unit change in Y

Randomized Control Trial -what is it? -what is the key assumption?

A group of randomly selected individuals receives a treatment, and another randomly selected group does not (treatment + control group) Key assumption: Results of those in the treatment and control groups would have been the same if not for the policy intervention

If Y is binary and X is continuous, how to interpret coeff. B? (ex. If education is cont. and earnings are binary, e.g. =1 if earn over $50,000, what does B=.07 mean?)

A one unit change in X generates a 100*B percentage point change in the probability Y=1 occurs Ex: B= .01 means an extra year of education is associated with a 7 percentage point increase in the probablity of earning over $50,000

If Y is continuous, and X is continuous, how to interpret coeff. B?

A one unit change in X generates a B unit change in Y

Regression Discontinuity Design -what is it? -key assumption?

An analysis that relies on a strict cut-off criterion for eligibility for the treatment in order to approximate experimental design (ex. looking at whether being a national merit scholar increases the likelihood of getting into Cornell, where there is a strict cutoff for merit scholar) -Key assumption: individuals can't sort around the cut-off easily

How to calculate if a coefficient B is significant? (at the 95% confidence level)

CI = B+-(2*S.E.) If 0 falls within the CI, then the coefficient is not significant

Esteves-Sorenson and Perretti (2012): Defaults and TV

Collected minute-by-minute data on Italian TV viewers - Question: When do people watch the news? Men: more likely to watch the news if preceded by soccer Women: More likely to watch if preceded by a romantic series Even though there is a very low switching cost, people still won't switch the TV off/on

Difference between RCT and diff-in-diff assumptions

Diff-in-Diff says the trends in the results would be the same in treatment and control group were it not for intervention, while the RCT says the actual results would be the same --> RCT assumption includes the diff-in-diff assumption - it's more expansive

Gneezy & Rustichini (2000): Pay Enough or Don't Pay At All (2 experiments) 1) Provide rewards for an IQ test? 2) Provide rewards for giving to charity -Alternative explanations?

Experiment 1: Question: does providing a reward for each correct answer produce better results on an IQ test? 4 groups: 1) $1 per problem 2) 33c per problem 3) 3 c per problem 4) No incentive Results: in the treatments in which money was offered, a larger amount yielded a higher performance (pretty much, not exactly). However, subjects offered a very small monetary incentives performed more poorly than those offered no compensation. Alternative explanation: Incomplete contracts - if no payment, participation consists of answering the questions, making them "due". If paid: decide how much effort to exert based on wage Experiment 2: Question: Does providing a reward for collecting money for charity crowd out intrinsic motivation to do so? 3 groups: 1) No reward 2) Get 1% of money you raise 3) Get 10% of money you raise Results: amount collected fell from group 1 to group 2, and then rose in group 3 but still wasn't as high as in group 1

Endowment Effect

Extra utility gain from "having" an object (rather than "getting" an object) - individuals value an object more when they own it. Similar to loss aversion - the disutility from giving something up is more than the utility from receiving something. Says WTA>WTP

Do participants respond differently when receiving compensation in the form of money vs. gifts?

For low payments, people prefer gifts more than money, but pretty much he same for higher value payments

Describe what a value function looks like -What are the three things that a value function shows? How?

Gains/losses on X, Value on Y S-shape, much steeper in losses than in gains (an additional gain has a lower additional value than a loss does) 3 things: 1) People use a reference point to evaluate utility --> reference point = origin 2) People are loss averse --> there are different slopes in the gian vs. loss 3) People have diminishing sensitivity --> people are risk-averse in gains, risk-loving in losses

Soman & Cheema (2011): Savings

How does a visual effect and partitioning money affect savings? No picture: partitioning increases saving With picture: partitioning increases savings more Diff-in-diff: partitionXpicture shows the increased difference between partitioned and non-partitioned when pictures were added

investment vs. indulgence in present bias

If an activity is an investment (PiNow<0 and PiLater>0), doer does activity less than planner wants if an activity is an indulgence (PiNow>0 and PiLater<0), doer does activity more than planner wants)

Benartzi and Thaler (2004): Save More Tomorow: Using Behavioral Economics to Increase Employee Saving

Implements program where 3% of funds go to retirement whenever you get a raise --> creates a default contribution rate (also type of commitment device to solve present bias --> also addresses loss aversion of not wanting to give up current money) Results: people in the program on avg. quadrupled their savings rate -An example of libertarian paternalism: a soft nudge

Titmuss (1970): The Gift Relationship: From Human Blood to Social Policy - what is wrong with this study?

In US: blood donations are compensated In UK: blood donations are voluntary Simply compared quantity and quality of blood was lower in the US than in the UK (lower quantity could be due to crowding out of intrinsic motivation, lower quality due to adverse selection) Solow attacked this study bc he said there needed to be a diff-in-diff: needed to compare the trends of blood donations in the US and UK to see if there were other things affecting blood donations (i.e., US and UK are very different countries) --> should test a before an after of not compensated and compensated within the US and UK

Pallais (2013): Small Differences that Matter: Mistakes in Applying to College

Question: Is there default behavior in the number of ACT scores sent to colleges when they number of free scores rises from 3 to 4? Results: Number of people sending 4 repots rose after the policy change.

Hastings & Washington (2010) - The First of the Month Effect

Question: do people who receive food stamps spend more at the beginning of the month due to present bias? Result: Higher expenditure at the beginning of the month Standard model says you would spend same amount over the course of the whole month

What does the paper written by Homonoff address?

Title: Can Small incentives have large effects? The impact of taxes vs. bonuses on disposable bag use Question: Do customers respond more to a 5-cent tax than a 5-cent bonus? Results: Those who had a 5-cent tax had a much larger increase in reusable bag use than those who received a bonus Standard theory: says consumers will have the same response regardless of tax/bonus Used a diff-in-diff: difference in bag use from before tax to after tax between 2 counties (experimental and control counties)

Robustness Check for default behavior

Maybe people view the default as being the status quo for a reason - they view it as advice and that they should follow suit because it must be the best choice if everyone is doing.

Royer, Stehr &Sydnor (2014): Gym Attendance

Offered incentive to join a gym, and then after a month of this incentive period, half randomly selected to have a self-funded commitment contract, and the other to not have anything. In the self-funded commitment contract, people pledged to use the gym over the next 2 months (and put down a certain amount of money) - they would be refunded with the money if they kept their word, and if they didn't, their money was given to charity. Significant increase in attendance with the commitment device.

Johnson and Goldstein (2003): Do Defaults Save Lives?

Online survey telling participants that just moved to a new state to decide if they would like to be donors Randomized to 3 groups: 1) Opt-in 2) Opt-out 3) Neutral Those in opt-out have much higher consent rates to being a donor

Default Behavior

People have a tendency to stick with their default, even if there are very low transition costs

Meier and Sprenger (2010): Present-Biased Preferences and Credit Card Borrowing

People who show present bias have more credit card debt - more likely to spend more today

Patterson (2014): Internet Distraction

Performance on online courses increased significantly after time-limits on internet time were implemented.

Commitment device

Present-biased individuals may use commitment devices to align planner and doer selves

Deci (1971): SOMA game

Psych students came to the lab in which they have to make different configurations and played a game 3 sessions: 1) No compensation 2) Treatment group paid $1 per configuration 3) No compensation Slight increase in seconds spent on the puzzle from session 1 to 2, and then a decrease from 2 to 3 Results: Even though extrinsic crowded out intrinsic, it showed that extrinsic rewards was the most effective session

Heyman & Ariely (2004): Market vs. Social Norms

Question 1: Is the relationship b/w compensation level and effort different in social vs. money markets? --> in a money market, effort increases with payment --> in a social market,effort will be at a high level and insensitive to payment Question 2: Will including both monetary payments and signals of social exchanges cause individuals to perceive an exchange as a money-market exchange (if both social and money market, will money market have a stronger effect?) --> stronger focus on money reward vs. non-money reward For low payment: gifts motivate people more than money, monetized gifts produce about same effort as money For high payment: gifts , monetized gifts, and money all produce about same effort (effort increases in money/monetized gifts groups)

Haggag and Paci (2014): Default Tips -Type of design? -Assumption made? -Results?

Question: Do default tip amounts in taxis have an impact on tip amount? -One taxi company changed its default depending on the fare amount --under $15 fare: default buttons are $2, 3 and 4 --over $15: defaults were 20%, 25, 30 Design: Regression Discontinuity Design Assumption: all ride characteristics that affect tips vary smoothly with the base tip Results: Sudden jump in tips right at $15 - discontinuity shows that people are affected by defaults because in other companies without this default tipping, they have a smooth line

Madrian and Shea (2001): The Power of Suggestion: Inertia in 401(k) Participation and Savings Behavior

Question: Does framing of retirement plan affect participation rate? Opt-in vs. opt-out 401(k) plan Econometric model: Quasi-experimental RCT - assumption is RCT: same # of people would have participated in 401(k) were it not for the policy change Coefficient: COHORT=1 if NEW, =0 if WINDOW --> coeff. showed that automatic enrollment increased the probability of participation by 50.4 percentage points compared to those who were immediately eligible but not automatically enrolled

Lacetera and Macis (2010): Do all material incentives for pro-social activities backfire?

Ran a survey hypothetically asking whether receiving cash vs. a git would 1) increase likelihood to donate 2) stay the same 3) decrease likelihood 4) cease any donation Results: Most people said that the diff between gifts and cash wouldn't affect their decision. Some offered cash said they would not donate anymore.

Difference-in-Difference -what is it? -what is the key assumption?

Shows difference between the changes in outcome (results) for the treatment group and for the control group -Key assumption: parallel trends assumption - the trends of the results over time would have been the same in the control and treatment group were it not for the intervention

Sophisticated vs. Naive Present Bias

Sophisticated: I know I have a problem Naive: I don't think I have a problem

Exponential Discounting vs. Present Bias model (quasi-hyperbolic model)

Standard model: Allows individuals to prefer consumption now more, but preferences are time consistent (individuals have the same preferences about future plans at different points in time) - can discount the future using a discount factor - the discount factor b/w any 2 periods is independent of when the utility is evaluated (that is, time consistent preferences) Present bias: adds another parameter, beta, to account for time inconsistent preferences - beta < 1 discounts all periods that are not today by beta ---> today is special: from today to tomorrow, you discount the future by betaXdelta, but between every other day, you only discount by delta. Doer does activity more than planner wants if PiLater<0 Doer does activity less than planner wants if PiLater>0

Treatment-on-the-treated effect

TOT=(beta1)/(takeup rate)

Kessler and Roth (2014): Don't Take 'No' For An Answer: An Experiment With Actual Organ Donor Registrations What econometric model is used?

Tested whether switching from opt-in frame to a neutral frame (active choice) would affect registration rates. Also looked at whether individuals are more likely to donate organs if deceased had actively selected "no" rather than not opted-in Results: No effect on consent rates. Individuals were more likely to support donating organs of a deceased who did opt-in rather than one who said "no" in the active choice frame. Econometric model: diff-in-diff: coefficient represents the difference in registration rates between the active choice and opt-in policy compared to other states who maintained the opt-in policy.

Ariely, Bracha, Meier (2009): Doing Good or Doing Well?

Tests Image Motivation: Does the desire for social approval make people act more prosocially in the public sphere? Experiment: People asked to donate to charity in private and in public, also with and without monetary incentives Results: Without any private monetary incentives, subjects donate more in the public condition than in the private one. Monetary incentives did not increase the effort in the public condition but they did in the private condition --> supports hypothesis that monetary incentives crowd out the image motivation to behave prosocially Uses diff-in-diff to estimate how monetary incentives crowd out image motivation: private incentive X public --> coeff. shows difference in charity from before the incentive to after the incentive in the public compared to the private

Carroll, Choi, Laibson, Madrian, and Metrick (2009): Optimal Defaults and Active Decisions

Tests effects of active choice vs. opt-in retirement plan (there is no default) Results: participation rate increases by 28 percentage points from opt-in to active choice

Gine, Karlan & Zinman (2010): Put Your Money Where Your Butt Is

Tests the effectiveness of a voluntary commitment device to quit smoking (could voluntarily sign up for CARES) Randomly assigned to one of 4 groups: 1A/1B: CARES (with/without collection) 2: Cue Cards 3: Control In column 3: B=0.057: subjects assigned to the CARES treatment group are a significant 5.7 percentage points more likely to pass the urine test than the control group In Column1: B=the effect of being offered the opportunity to open a CARES account (i.e. assigned to the CARES group), not whether you actually opened the account - only 11% of the CARES group opens an account

Libertarian Paternalism

Thaler and Sunstein's idea that policymakers should be choice architects: Libertarian: people should be free to do what they like Paternalism: policymakers should steer people's choices in a way that makes choosers better off, as judged by themselves Choice architects: people who organize the context in which people make decisions - altering people's behavior in a predictable way without changing economic incentives/forbidding any options

Intrinsic vs. Extrinsic motivation: Standard model vs Behavioral model

The idea that extrinsic motivations (i.e., a reward) can crowd out intrinsic motivations (simple pleasure from doing something) Standard model: offering additional financial incentive can only increase likelihood of doing something --> effort will increase monotonically with the size of the reward Behavioral: extrinsic motivation can crowd out intrinsic, which could decrease likelihood of doing something

If Y is in logs and X is binary, how to interpret coeff. B?

The movement of X from 0 to 1 produces a 100*B percent change in Y

If Y is binary and X is binary, how to interpret coeff. B?

The movement of X from 0 to 1 produces a 100*B percentage point change in the probability Y=1 occurs

If Y is continuous and X is binary, how to interpret coeff. B?

The movement of X from 0 to 1 produces a B unit change in Y

Coefficient of Loss Aversion

The ratio of the amount you would need to win and the amount you could potentially lose (or could be willingness to accept (WTA)/willingness to pay (WTP)

Field (2009): Education debt burden and career choice

Title: Education debt burden and career choice, evidence from a financial aid experiment at NYU law school Experiment: top law schools offered tuition subsidies to students who work for non-profits after law school - can enter a lottery to receive a tuition subsidy in 1 of 2 ways: -Loss: received tuition waivers, pay NYU back if don't enter public sector -Gain: pay tuition upfront, NYU repays student if enter public sector Question: Does the framing of these subsidies influence the likelihood of working in the public sector? Results: Students in the loss group are much more likely to enter public interest law after graduation

Fryer, Levitt, List & Sadoff: Enhancing the Efficacy of Teacher Incentives

Title: Enhancing the Efficacy of teacher incentives through loss aversion: a field experiment Question: does the framing of rewards for teachers affect test scores? Gain: teachers receive a bonus at the end of the year Loss: teachers receive a bonus at the beginning of the year Control: no incentive Results: those in the loss group produced higher percentile ranks in their students Robustness Checks: Did teachers use money to buy supplies that could raise test scores? Attrition? Cheating?

Pope & Schweitzer: Is Tiger Woods Loss Averse?

Title: Is Tiger Woods Loss Averse? Question: Are golfers more or less successful when they are putting for par vs. other types of putts? Results: those putting for par rather than birdie were more likely to make the hole - "par" is a reference point, and putting below this point is experiencing a loss Standard model: says that those putting for birdie and par would produce same results

Allen, Dechow, Pope & Wu (2014): Reference Dependent Marathon Runners

Title: Reference-Dependent Preferences, Evident from Marathon Runners Question: do runners bunch around round-numbered finishing times? Results: the distribution of marathon times spikes right before every hour mark (especially before 4 hours, which is a common time) Behavioral theory: the round numbers could be a reference point (or a finishing time goal) and if you finish after this round number, you're experiencing a loss - the desire to finish under the reference point shows prospect theory bc a) you evaluate your choices based on a reference point and b) you avoid finishing below the reference point bc of loss aversion Standard model: says that the distribution would be smooth - there would be no bunching around round numbers

Pope & Simonsohn (2011): Round Numbers As Goals

Title: Round Numbers as Goals Question: Are students (specifically juniors) more likely to retake the SATs if they get a score just below a round number? Results: In the distribution of FINAL SAT scores for juniors and seniors, there are little jumps around round numbers, showing that these people retook the SAT to reach a higher (reference point) score, or a round score.

Levitt, List, Neckermann, and Sadoff: The Behavioralist Goes to School

Title: The Behavioralist Goes to School Question: Does the framing of financial rewards for exam performance (given to the student) increase test scores? Method: Students were given cash if their score improved since the last test. 3 groups: Gain: given $ after test Loss: $ given at beginning, then taken away if score didn't improve Control: no incentive Results: Both gain and loss students significantly increase effort on tests, loss treatment slightly more effective Robustness checks: did students use money to buy helpful supplies? Did they cheat? Do they usually not try very hard because there are no stakes for the test?

Ashraf, Karlan, and Yin (2006): Savings

Treatment group offered "Save, Earn, Enjoy Deposits" (SEED) accounts -More likely to open an account if showed present bias in lab questions

Difference between a utility function and a value function

Utility function: evaluation of choices are made over final states of wealth Value function: evaluation of choices are made relative to a reference point

Overjustification Effect - what study can we see this in?

When extrinsic rewards are offered, individuals focus on them more than the intrinsic rewards, so once the rewards are removed, individuals continue to focus on the extrinsic rewards and therefore reduce effort below the level exerted before the extrinsic reward was offered. --Can be seen in Deci's psych game study: initially, # of moves in the game increased with extrinsic reward, but then once the reward was taken away, # moves decreased

Samuelson and Zeckhauser (1988): Status Quo Bias in Decision Making

You inherit from your uncle: 1) Large sum of money 2) High-risk securities 3) Low-risk securities You can choose to invest in a low-risk or high-risk company - those who inherited high risk securities are more likely to pick the high-risk company (status quo bias)