field experiments

40
Field Experiments

Upload: toki

Post on 25-Feb-2016

42 views

Category:

Documents


4 download

DESCRIPTION

Field Experiments. Typical Format. Researcher manipulates something in the real world, exposing randomly-assigned groups of people to different treatments - PowerPoint PPT Presentation

TRANSCRIPT

Page 1: Field Experiments

Field Experiments

Page 2: Field Experiments

Typical Format

• Researcher manipulates something in the real world, exposing randomly-assigned groups of people to different treatments

• Possibly the most fruitful area of research in social sciences, but also greatest ethical challenges since the experiments can be so powerful and have real-world effects

Page 3: Field Experiments

Field Experiments in Practice• Most medical clinical trials are field

experiments• Since the experiment takes place outside of

the laboratory, researchers cannot completely control randomization

• Some subjects selected for treatment may refuse, others may switch from control to treatment or from treatment to control– Example: Dartmouth Spine Study

Page 4: Field Experiments

As Treated vs Intent to Treat

• As treated (AT) analysis compares subjects given the treatment they received.– Does not consider treatment group subject

assigned, only which they received• Intent to Treat (ITT) compares subjects in the

groups to which they were randomized, regardless of whether they received or adhered to the allocated treatment.– Required in medical trials

Page 5: Field Experiments

Intent to Treat: Adherence

• Problem 1: Adherence: Subjects who did not adhere to treatment or switched to control should be included– Analyze data in group to which subject allocated,

not what they actually received

Page 6: Field Experiments

Intent to Treat: Loss to Follow-Up

• Problem 2: Loss to Follow-up: Subjects who did not complete trial or could not be contacted to provide data should be included– Cannot be deleted from analysis since they are

not missing due to random factors– Missing data problem

Page 7: Field Experiments

Applications in Social Sciences

• Many field experiments manipulate the information that people receive

• Information is a less powerful and shorter-term manipulation, making the ethical issues less significant

Page 8: Field Experiments

The Effects of Choice and Enhanced Personal Responsibility for the Aged:

A Field Experiment in an Institutional Setting

Ellen Langer and Judith RodinJournal of Personality and Social Psychology

1976, 34(2)

Page 9: Field Experiments

Research Question

Objective helplessness as well as feelings of helplessness and hopelessness may contribute to psychological withdrawal, physical disease, and death.

Objective control and feelings of mastery may contribute to physical health and personal efficacy.

Page 10: Field Experiments

Study Design

• Conducted in nursing home for generally upper income residents

• Four floors in the home.• Two floors selected for study based on

similarity in the residents’ physical and psychological health and prior socioeconomic status.

Page 11: Field Experiments

Randomization?

• Residents assigned to floors based on availability

• Residents on two different floors exposed to different “treatments”

Page 12: Field Experiments

Treatment: A Meeting on Lounge of Each Floor

Both floors held meeting directed by the same staff member, who read from two different scripts

Script 1: Responsibility-Induced Message[Treatment]

Script 2: No Responsibility [Control]

Page 13: Field Experiments

Treatment Group: Responsibility Message

“Many of you don't realize the influence you have over your own lives here. Take a minute to think of the decisions you can and should be making.”

“You should be deciding how you want your rooms to be arranged—whether you want it to be as it is or whether you want the staff to help you rearrange the furniture. It’s your life and you can make of it whatever you want.”

Page 14: Field Experiments

Treatment Group (continued)

“If you are unsatisfied with anything here, you have the influence to change it. It's your responsibility to make your complaints known, to tell us what you would like to change, to tell us what you would like.”

Page 15: Field Experiments

Treatment Group (continued)

“Also, I wanted to take this opportunity to give you each a present from the Arden House. The plants are yours to keep and take care of as you'd like.” [A box of small plants was passed around, and patients were given two decisions to make: first, whether or not they wanted a plant at all, and second, to choose which one they wanted. All residents did select a plant.]

Page 16: Field Experiments

Treatment Group (continued)

“One last thing, I wanted to tell you that we're showing a movie two nights next week, Thursday and Friday. You should decide which night you’d like to go, if you choose to see it at all.”

Page 17: Field Experiments

Control Group

“I was surprised to learn that many of you don't know about the things that are available to you; that many of you don’t realize all you're allowed to do here.”

“We want your rooms to be as nice as they can be, and we've tried to make them that way for you. We want you to be happy here.”

Page 18: Field Experiments

Control Group

“If you have any complaints or suggestions about anything, let a nurse know what they are. Let us know how we can best help you. You should feel that you have free access to anyone on the staff, and we will do the best we can to provide individualized attention and time for you.”

Page 19: Field Experiments

Control Group

“Also, I wanted to take this opportunity to give you each a present from the Arden House. The plants are yours to keep. The nurses will water and care for them for you.”

[The nurse walked around with a box of plants and each patient was handed one.]

Page 20: Field Experiments

Control Group

“One last thing, I wanted to tell you that we're showing a movie next week on Thursday and Friday. We'll let you know later which day you're scheduled to see it.

Page 21: Field Experiments

Dependent Variables

Pre-post within and between subjects designQuestionnaire given to each subject one week before and three weeks after meetings”How much control do you have over your life?

(0=none, 8=total)How happy are you?How active do you feel?

Alertness rated by research assistant

Page 22: Field Experiments

Dependent Variables

• Nurses rated each resident on– Happiness– Alertness– Dependence– Sociability– Activity– Eating and Sleeping Habits

• Nurses’ ratings blind (did not know whether resident was in treatment or control group)

Page 23: Field Experiments

Dependent Variables

• Residents could enter competition to guess number of jelly beans in a jar

• Tenth day after treatment measured how far residents moved in their wheelchairs during one day

• Note: Many dependent variables measured– Often guarantees at least something turns up

significant

Page 24: Field Experiments

Results

Page 25: Field Experiments

Behavioral Results

• Responsibility group– More likely to attend movie– More likely to enter jelly-bean guessing contest

• No evidence of more wheelchair movement by responsibility group (floors too clean to measure movement?)

Page 26: Field Experiments

Conclusions

• 71% of control group residents more debilitated over 3-week study

• 93% of treatment group showed improvement

• Having personal control over one’s environment and taking care of a plant made subjects happier, healthier, more active

Page 27: Field Experiments

Concerns?

• Would be interesting to see in a 2x2 design whether taking care of the plant had an effect separate from message received

Page 28: Field Experiments

The Effects of Canvassing, Telephone Calls, and Direct Mail on Voter Turnout: A Field

Experiment

ALAN S. GERBER and DONALD P. GREENAmerican Political Science Review

Vol. 94(3): 2000

Page 29: Field Experiments

Field Experiment on Voter Turnout

• Conducted in city of New Haven, Connecticut• Test effect of encouraging people to vote by– canvassing (appearing at a person’s house), – telephone call– direct mail (up to three)

• 2 x 2 x 4 factorial design• Canvassing x telephone call x direct mail

Page 30: Field Experiments
Page 31: Field Experiments
Page 32: Field Experiments
Page 33: Field Experiments

Problem

The probability that a given person in the experiment votes is:Y=a + b1X1 + b2X2 + e

Y=1 if subject votesX1=1 if subject is difficult to contact

X2=1 if subject is actually contacted, 0 otherwise

Cannot assume b1=0

Creates correlation between X1 and e

Page 34: Field Experiments

Solution: Instrumental Variables

• Use the intent to treat as an instrumental variable– Uncorrelated with regression error due to random

assignment to intent to treat group– Correlated with endogenous variable since being

contacted during experiment is a function of being in the intent to treat goup

Page 35: Field Experiments
Page 36: Field Experiments

Field Experiments of Political Participation (Green, Gerber, Shachar)• Contact a random sample of registered voters

either by mail or face-to-face in 1998– We know whether the subjects voted in 1996

• Encourage people to vote• Check polling data to see if they voted in the

next election (1998 and 1999)

Page 37: Field Experiments
Page 38: Field Experiments

Results

Page 39: Field Experiments

Natural Experiment (Lassen)

• Question: Do more informed people vote, or do people vote become more informed?– What is the causal relationship between

information and voting?• Research Design: Find a natural experiment in

Copenhagen, Denmark, where 4 out of 15 districts carried out a pilot project on decentralization (PCD) before a vote on whether to decentralize

Page 40: Field Experiments