RESEARCH METHODS FESTIVAL 2012: BRINGING THE LAB TO THE FIELD ANANDI MANI, UNIVERSITY OF WARWICK &...
-
Upload
zion-hocking -
Category
Documents
-
view
218 -
download
1
Transcript of RESEARCH METHODS FESTIVAL 2012: BRINGING THE LAB TO THE FIELD ANANDI MANI, UNIVERSITY OF WARWICK &...
RESEARCH METHODS FESTIVAL 2012: BRINGING THE
LAB TO THE FIELDANANDI MANI, UNIVERSITY OF WARWICK & CAGE
Types of Experiments (Harrison-List JEL)
______AFE_________FFE_________NFE______________________________
Lab [field experiments] NE, PSM, IV, STR, etc.
Conventional lab experiment (Lab) employs a standard subject pool of students, an abstract
framing, and an imposed set of rules Artefactual field experiment (AFE)
same as a conventional lab experiment but with a non-standard subject pool
Framed field experiment (FFE) same as an artefactual field experiment but with field context in
the commodity, task, information, stakes, time frame, etc. Natural field experiment (NFE)
same as a framed field experiment but where the environment is the one that the subjects naturally undertake these tasks, such that the subjects do not know that they are in an experiment
Motivation
Development Economists have been doing Field Experiments using Randomized Control Trials (RCTs) for over a decade now, addressing a wide range of questions, e.g. Do Cameras in Schools improve Teacher Attendance &
Student Outcomes? Does Microfinance spur Business Investment among the
poor? Some of these RCTs come under criticism for a lack of
light on the Mechanisms underlying the observed findings (Deaton(2009))
Lab Experiments may help identify Potential Reasons for Certain Outcomes Observed in
Survey Data Mechanisms Underlying Some Field Experiment Findings,
which would help increase the External Validity
Advantages of Lab Experiments
Better Control: Lab Experiment Design makes it feasible to generate results ceteris paribus Testing alternative theoretical mechanisms Test Institutions (e.g. Auction formats)
Scope for Replication & Comparison across Cultural Settings
Cheaper Market Design Pilots
Outline of Talk
Applications:(A) GENDER DIFFERENCES IN PRODUCTIVITY & PAY Fact 1: Capital Returns (de Mel et al -- 2009): Lower returns of
Women owned firms associated with less supportive spouses Q: Could Intra-Household Decision-Making Play a role in this?
Fact 2: Women in the US earn 75% of what men do on the labor market – and education, experience, hours worked don’t explain more than 50% of it Q: Gender Differences in Competitive Behavior explain this gap?
(B) POVERTY & DECISION-MAKING Fact 3: Poor seem to make irrational decisions on Savings, Human
Capital Investment Q: Could Poverty Affect Stress Levels & Cognitive Ability?
(C) PITFALLS OF LAB EXPERIMENTS & POSSIBLE SOLUTIONS
Intra-HH DecisionsHow does decision- making work within the family?
Laws about Property Rightsand Inheritance
Entrepreneurship/Income Generation
Programs
Schemes to encourageHuman capital
investment
Does it amplify the inequities of market outcomes for its members, or does it mitigate them?
HH Decision-Making: Experiment Questions
Assuming that HH members do not share a common set of preferences…
Q1: Is HH decision-making efficient – i.e. do members maximize HH (Investment) returns ? OR …
Q2: If not, why do they sacrifice HH income? Is it …for economic reasons – for instance, greater
bargaining power/ control over HH resources (how much?)
…or for other socially influenced reasons? …and do they do this only when their spouse won’t
know?
Arguments for an Experimental Approach
One-phrase Summary of Survey Based Empirical Studies of HH resource allocation decisions: Can’t be sure!
Allocation decisions directly observed real-time No need to infer decisions from data reported ex-post Survey responses may “adjusted to fit” local social and
cultural norms(Bertrand-Mullainathan(2001)), whereas... Actions speak better than words
Focus on Investment rather than Consumption Decisions No scope for effects arising from possible
substitutions outside the experiment
Experiment Location & Sample
Anantapur district, Andhra Pradesh (2nd most drought prone)
300 couples, from 32 villages -- all wives members of Self-Help Groups (SHGs) run by a single NGO
Promised Participation Fee (Rs.50 – about 62p), roughly equal to daily wages, with scope to make more based on their performance
Experiment Protocol
3-4 villages participating daily (10 day experiment)
Participating couples from each village brought in to NGO location
Separate waiting area for men and women Three couples taken to six separate rooms,
where.. Experiment explained and options presented
by a coordinator Data recorded by two independent data entry
staff Separate waiting areas for male and female
participants who completed the experiment Individual payment to participants upon
completion Participants taken back to village when all
payments completed
p 11
Experiment Tasks
No Tradeoff
Investor Tradeoff between higher HH income and own control over it
o4 Investment Decisions for each Spouse, individually presented in random orderoTask: Allocate Rs.50 (seed money) across two Investment options Blue and Red oEfficient Investment Allocation = Rs. 50 in Blue
Investment Means
INVESTOR CONTROL OVER INCOME
ALL MENWOMEN
N=502N=250 N=252
(1) (2) (3) Fixed Share 44.95 42.2 47.68
(11.42)(13.24
) (8.44) Low Control 36.63 34.94 38.32
(19.27)(18.33
) (20.06) Medium Control 38.30 35.62 40.98
(18.77)(18.85
) (18.35) High Control 43.37 41.36 45.36
(14.93)(15.92
) (13.63)
Overall Mean Investment - (across 4 decisions) 40.81 38.53 43.09
(16.75)(17.03
) (16.18)
p 12Efficient Investment: Rs.50 in Blue option Both Men and Women are Inefficient.
They’re willing to sacrifice HH income, to gain more control over it for themselves.
But Men don’t maximize HH returns even if their share of control is Fixed! WHY??
Fraction of Efficient HHs
A third of men are inefficient even when their share of control is Fixed – i.e. they undercut their own income (and their wife’s) rather than maximize HH returns
“Irrationality” not explained by low education/confusion, lack of experience with financial decisions, longer term effects on bargaining power within HH
Why are Men inefficient under Fixed Shares
Men don’t like it when their wife’s share exceeds theirs
When her share>50% They are willing to
undercut their own income..
to ensure she does not earn too much more than themselves.
30 40 50 60 700.00
5.00
10.00
15.00
20.00
25.00
30.00
35.00
40.00
45.00
50.0047.68 48.48 47.39
43.5246.59
42.40 41.1943.80
37.50
32.32
Wife's Invt. (mean)Husband's Invt.(Mean)
Wives of “Spiteful” Husbands are more inefficient in other three decisions, where Control over HH income depends upon investment allocation. CONCLUSION: Consistent with de Mel et al(2009) finding, Less cooperative spouses => Lower Productivity on Women’s Businesses
Gender Earnings Gap: Motivation
POTENTIAL SOURCES OF GENDER GAP IN LABOR MARKET OUTCOMES Occupation choice Experience & Continuity in labor market participation Discrimination Psychological factors
Stereotype threat Claude Steele (1997): Additional anxiety causes choking under
pressure when performing a task Ambady et al (1999), Psychological Science
Self-Confidence
Competitive Behavior
Performance under Competition
Gender Differences: Gneezy-Niederle-Rustichini(2003, QJE)-Summary
Lab Experiment conducted in Israel with students from Technion Participants Task: Solving Mazes on a computer Studied Participants’ performance under three payment schemes
(a) Non-competitive (Piece rate compensation) (b) Competitive (Winner-take-all tournament) (c) Random pay setting (One person in Group of 6 is paid, rest are not)
Main findings: Men’s performance improves considerably going from (a) to (b), whereas
women’s performance does not change Women’s performance is much worse when their tournament group
includes men than when it has only women
Do Women prefer to Compete less?
Niederle-Vesterlund( Aug 2007, QJE) Women may choose lower powered jobs for multiple reasons:
Responsibility & Time demands of such jobs, given family considerations Discrimination may discourage attempts to obtain these jobs Competitive Pressure of such jobs?
Experiments allow choice of tasks with similar time demands, where innate abilities do not differ among men and women, and discrimination is ruled out
Theories (about why women shy away from high-profile jobs): They may Dislike Competition Lack Confidence, relative to men Be Risk Averse Have Feedback Aversion (They’re more discouraged by negative feedback).
Experimental Design makes it possible to distinguish among various channels
Experiment Details
Lab experiment with students at University of Pittsburgh, groups
Task: Addition of sets of five 2 digit-numbers, for five minutes
Information to Participants: Only on own absolute performance, no information on others’ performance. Information provided real time, as task is performed.
Studied Payment Scheme Choice of Men vs. Women: Piece-rate vs. Winner-take all (Competitive) scheme, given information above.
4 participants per group, two male and two female (20 groups)
Experiment Design Task 1: Piece rate (PR) of $0.5 per correctly solved addition Task 2: Tournament (T; winner take all) rate of $2 per correctly solved addition
At a 25% chance of tournament win, both payment schemes generate the same expected payoff.
Tournament payoff is in per task terms to avoid guesswork about what would be a high enough fixed payment to induce tournament entry among high performers
Task 3: First choose payment scheme (PR or T) and then do addition task Participants evaluated against others’ performance in Task2 –why?
Eliminates effects of beliefs about others’ choice on decision 10,000 (feasible) groups made with replacement from the data, avg. across 100 trials to
determine individual success probability in tournament.
Task 4: Choose payment scheme (PR or T) for (previous) Task 1; No new task To separate the preferences for competition from other factors such as risk aversion
& feedback aversion , on tournament entry decision Ask participants to guess their rank in task 1 and task 2 in their group of four
To measure effects of self-confidence on tournament entry and performance
Main Findings
Men and Women are equally good at Addition Task under Piece Rate and Tournament. Despite this,
being a woman reduces probability of selecting Tournament payment scheme in Task 3 by 38% Not explained by individual performance in previous
rounds (T1,T2) or current round (T3) itself. For women, total expected cost of under-entry is
much larger than cost of over-entry; for men it’s the reverse
Despite accounting for differences in Self-confidence, being female still reduces Tournament entry probability by 27.8%
Taking the Lab Design to the Field
Potential Concerns with the above 2 Experiments: Experiment 1: Performance could be influenced by Task specific
differences in ability (men have advantage in spatial ability and arm-throwing capacity) -- so mazes may not be to women’s advantage.
Experiment 2: Women’s observed Preferences for Competition may be due to being socialized to believe they are worse competitors than men – or that their behavior should be “ladylike” (less aggressive) ?
Gneezy-Leonard-List (2007) address both these concerns – How? Socialization: Repeat similar experiment design in one Matrilineal &
Matrilocal tribe and one Patriarchal tribe Task: Task unfamiliar to people in both tribes
Lab-in-the-Field: Design
Maasai (Tanzania): Patriarchal “Men treat us like donkeys”
Maasai woman (Hodgson (2001))
Khasi (NW India): Matrilineal “We are sick of playing the
roles of breeding bulls and baby-sitters” Khasi man (Ahmed (1994)
Subjects in 2 groups, randomly paired with 1person from other group (paired subject identity/sex not known) Task: Throw tennis ball into bucket 3 metres away (10 chances per subject) Payment Scheme: X per “success” irrespective of paired subject performance OR 3X per “success” if own performance better than paired person X=Rs.20 in India; X = 500 shillings in Tanzania
Maasai
Maasai men choose to compete at twice the rate that women do
Similar to findings in Western settings
Maasai vs. Khasi
Khasi women choose to compete at twice the rate that men do
And even at a rate slightly higher than Maasai men
Authors’ Conclusion: Any number of subtle influences on children or adults can cause differences in attitudes to competition -- even if the behavior is broadly framed by genetic endowment
Poverty, Stress & Cognitive Capacity
USING
SUGARCANE
HARVESTS
TO UNDERSTAND
THE PSYCHOLOGY OF POVERTY
Poverty, Cognitive Capacity & Decisions-1
A fundamental assumption of Economics is the Scarcity of Resources… Yet the Rational Model assumes that Mental
Capacity is Infinite ! But Decision-Making Takes Mental Effort,
and its Tiring!
Question: Does the State of Being Poor affect Cognitive Capacity? (Mani-Mullainathan-Shafir)
Sugarcane Harvests
Long Cycle Crop – about 11 months Farmers are down to the wire a few weeks before
Harvest Receive Lump sum Returns a few weeks after Harvest
Sugar Mills assign Cutting Dates to individual farmers, hence farmers don’t have control over when their Income arrives
Methodology: Compare Individual Farmers’ before vs. after Harvest on Measures of: Stress: Blood Pressure, Heart Rate (Round 1-- 2009) Cognitive Capacity & Attention: IQ(Raven’s) tests, Stroop
tests (Round 2 – 2011)
Summary of Findings
Main Findings: Poverty in the Pre-Harvest Period Raises Stress Levels Lowers IQ & Cognitive Capacity
Comments: These findings are not driven by Adverse Nutritional changes pre-harvest Learning Effects post-harvest (for IQ tests)
Potential Pitfalls – and some Solutions
Lack of Anonymity May elicit more pro-social behavior when observed Solution: Double blind experiments, Outcome measure unclear
Context and Framing Label “Wall Street” game vs. “Community” game affects play Solution: Neutral wording; Collect Background data on subjects
Self-Selection in Participants Biased sample -- Could be a problem in all Field Experiments Solution: Conduct experiment in different settings
Low Stakes may elicit non-serious behavior Solutions: Vary stakes, Treat results as lower/upper bound, Use suitable
subjects Relevance of Lab decisions to “real” behavior?
Track correlation b/w the two (e.g. Karlan(2005) AER) – Trust game outcome and Repayment of Microfinance Loan a year later