Knowing Thy Neighbor: What Information Neighbors Have and How Best - - PowerPoint PPT Presentation

▶
knowing thy neighbor what information neighbors have and
SMART_READER_LITE
LIVE PREVIEW

Knowing Thy Neighbor: What Information Neighbors Have and How Best - - PowerPoint PPT Presentation

Apr 09, 2024 524 likes •986 views

Knowing Thy Neighbor: What Information Neighbors Have and How Best to Elicit it Reshmaan Hussam, Natalia Rigol, Benjamin N. Roth June 2, 2016 Hussam, Rigol and Roth June 2, 2016 1 / 22 Introduction Outline Introduction 1 The Experiment 2

slide-1
SLIDE 1

Knowing Thy Neighbor: What Information Neighbors Have and How Best to Elicit it

Reshmaan Hussam, Natalia Rigol, Benjamin N. Roth June 2, 2016

Hussam, Rigol and Roth June 2, 2016 1 / 22

slide-2
SLIDE 2

Introduction

Outline

1

Introduction

2

The Experiment

3

Results

4

Conclusion

Hussam, Rigol and Roth June 2, 2016 2 / 22

slide-3
SLIDE 3

Introduction

Motivation

1 Asymmetric information is a major obstacle to providing services and

support for the poor.

Hussam, Rigol and Roth June 2, 2016 3 / 22

slide-4
SLIDE 4

Introduction

Motivation

1 Asymmetric information is a major obstacle to providing services and

support for the poor.

Ideally would like to allocate capital to entrepreneurs with high marginal returns, credit to reliable borrowers, subsidies to the poor, etc.

Hussam, Rigol and Roth June 2, 2016 3 / 22

slide-5
SLIDE 5

Introduction

Motivation

1 Asymmetric information is a major obstacle to providing services and

support for the poor.

Ideally would like to allocate capital to entrepreneurs with high marginal returns, credit to reliable borrowers, subsidies to the poor, etc.

2 However, formal information about the intended targets of these

services is sparse or non existent.

Hussam, Rigol and Roth June 2, 2016 3 / 22

slide-6
SLIDE 6

Introduction

Motivation

1 Asymmetric information is a major obstacle to providing services and

support for the poor.

Ideally would like to allocate capital to entrepreneurs with high marginal returns, credit to reliable borrowers, subsidies to the poor, etc.

2 However, formal information about the intended targets of these

services is sparse or non existent.

3 Governments, NGOs, and MFIs screening recipients for aid and credit

may rely on community information. But...

Hussam, Rigol and Roth June 2, 2016 3 / 22

slide-7
SLIDE 7

Introduction

Motivation

1 Asymmetric information is a major obstacle to providing services and

support for the poor.

Ideally would like to allocate capital to entrepreneurs with high marginal returns, credit to reliable borrowers, subsidies to the poor, etc.

2 However, formal information about the intended targets of these

services is sparse or non existent.

3 Governments, NGOs, and MFIs screening recipients for aid and credit

may rely on community information. But...

What is the quality of the information held by community members?

Hussam, Rigol and Roth June 2, 2016 3 / 22

slide-8
SLIDE 8

Introduction

Motivation

1 Asymmetric information is a major obstacle to providing services and

support for the poor.

Ideally would like to allocate capital to entrepreneurs with high marginal returns, credit to reliable borrowers, subsidies to the poor, etc.

2 However, formal information about the intended targets of these

services is sparse or non existent.

3 Governments, NGOs, and MFIs screening recipients for aid and credit

may rely on community information. But...

What is the quality of the information held by community members? Should we be taking incenetives seriously?

Hussam, Rigol and Roth June 2, 2016 3 / 22

slide-9
SLIDE 9

Introduction

Highlights

1 Members of peri-urban communities have high quality information

about one another regarding, among other things, their entrepreneurial ability.

Hussam, Rigol and Roth June 2, 2016 4 / 22

slide-10
SLIDE 10

Introduction

Highlights

1 Members of peri-urban communities have high quality information

about one another regarding, among other things, their entrepreneurial ability.

2 Community members distort their reports when they’re being used to

inform real allocations (the distribution of grants).

Hussam, Rigol and Roth June 2, 2016 4 / 22

slide-11
SLIDE 11

Introduction

Highlights

1 Members of peri-urban communities have high quality information

about one another regarding, among other things, their entrepreneurial ability.

2 Community members distort their reports when they’re being used to

inform real allocations (the distribution of grants).

3 Simple techniques motivated by mechanism design are effective in

realigning incentives for truthfulness.

Hussam, Rigol and Roth June 2, 2016 4 / 22

slide-12
SLIDE 12

Introduction

Highlights

1 Members of peri-urban communities have high quality information

about one another regarding, among other things, their entrepreneurial ability.

2 Community members distort their reports when they’re being used to

inform real allocations (the distribution of grants).

3 Simple techniques motivated by mechanism design are effective in

realigning incentives for truthfulness.

Monetary incentives

Hussam, Rigol and Roth June 2, 2016 4 / 22

slide-13
SLIDE 13

Introduction

Highlights

1 Members of peri-urban communities have high quality information

about one another regarding, among other things, their entrepreneurial ability.

2 Community members distort their reports when they’re being used to

inform real allocations (the distribution of grants).

3 Simple techniques motivated by mechanism design are effective in

realigning incentives for truthfulness.

Monetary incentives Privacy

Hussam, Rigol and Roth June 2, 2016 4 / 22

slide-14
SLIDE 14

Introduction

Highlights

1 Members of peri-urban communities have high quality information

about one another regarding, among other things, their entrepreneurial ability.

2 Community members distort their reports when they’re being used to

inform real allocations (the distribution of grants).

3 Simple techniques motivated by mechanism design are effective in

realigning incentives for truthfulness.

Monetary incentives Privacy Cross reporting

Hussam, Rigol and Roth June 2, 2016 4 / 22

slide-15
SLIDE 15

Introduction

Highlights

1 Members of peri-urban communities have high quality information

about one another regarding, among other things, their entrepreneurial ability.

2 Community members distort their reports when they’re being used to

inform real allocations (the distribution of grants).

3 Simple techniques motivated by mechanism design are effective in

realigning incentives for truthfulness.

Monetary incentives Privacy Cross reporting Zero sum elicitation

Hussam, Rigol and Roth June 2, 2016 4 / 22

slide-16
SLIDE 16

The Experiment

Outline

1

Introduction

2

The Experiment

3

Results

4

Conclusion

Hussam, Rigol and Roth June 2, 2016 5 / 22

slide-17
SLIDE 17

The Experiment

The Sample

Conducted census of all business owners in 9 peri-urban communities around Amravati, India

Hussam, Rigol and Roth June 2, 2016 6 / 22

slide-18
SLIDE 18

The Experiment

The Sample

Conducted census of all business owners in 9 peri-urban communities around Amravati, India 1576 households had a non-farm business with capital < $1000 and no paid, permenant employees

Hussam, Rigol and Roth June 2, 2016 6 / 22

slide-19
SLIDE 19

The Experiment

The Sample

Conducted census of all business owners in 9 peri-urban communities around Amravati, India 1576 households had a non-farm business with capital < $1000 and no paid, permenant employees 1380 participated in our study

Hussam, Rigol and Roth June 2, 2016 6 / 22

slide-20
SLIDE 20

The Experiment

Timeline

Census: September 2015 Recruitment: October 2015 Baseline survey: December 2015 - April 2016 Elicitation exercise: February 2016 - May 2016 Followup: Ongoing

Hussam, Rigol and Roth June 2, 2016 7 / 22

slide-21
SLIDE 21

The Experiment

Elicitation Questions

We ask questions about Marginal Return to Capital Income Profits Assets Medical Expenses Work Hours Digit Span Likelihood to Repay a Loan Business Ability Deservingness of the Grant

Hussam, Rigol and Roth June 2, 2016 8 / 22

slide-22
SLIDE 22

The Experiment

Elicitation Exercise

We asked respondents answers to above questions during baseline (before they had knowledge of the elicitation exercise) We then clubbed the entrepreneurs into groups of 4-6 based on geographic proximity and invited them to a central hall to rank one another along above dimensions. Groups and individuals randomized into the following treatments

Hussam, Rigol and Roth June 2, 2016 9 / 22

slide-23
SLIDE 23

The Experiment

Experimental Design

Hussam, Rigol and Roth June 2, 2016 10 / 22

slide-24
SLIDE 24

The Experiment

Non-Random Design Features

Relative Rankings and Quintiles

Hussam, Rigol and Roth June 2, 2016 11 / 22

slide-25
SLIDE 25

The Experiment

Non-Random Design Features

Relative Rankings and Quintiles “Cross reporting”

Hussam, Rigol and Roth June 2, 2016 11 / 22

slide-26
SLIDE 26

The Experiment

Non-Random Design Features

Relative Rankings and Quintiles “Cross reporting” Peer Prediction...

Hussam, Rigol and Roth June 2, 2016 11 / 22

slide-27
SLIDE 27

The Experiment

Elicitation in the Field

Hussam, Rigol and Roth June 2, 2016 12 / 22

slide-28
SLIDE 28

The Experiment

Elicitation in the Field

Hussam, Rigol and Roth June 2, 2016 13 / 22

slide-29
SLIDE 29

Results

Outline

1

Introduction

2

The Experiment

3

Results

4

Conclusion

Hussam, Rigol and Roth June 2, 2016 14 / 22

slide-30
SLIDE 30

Results

Main Specifications

How much people know:

Outcomei = α0 + α1Rankik + Xi + γc + ǫik for person i, cluster c, ranker k.

Hussam, Rigol and Roth June 2, 2016 15 / 22

slide-31
SLIDE 31

Results

Main Specifications

How much people know:

Outcomei = α0 + α1Rankik + Xi + γc + ǫik for person i, cluster c, ranker k.

The effect of our treatments:

Outcomei = α0 + α1Rankik +

n βnRankik × Treatmentn + Xi + γc +

  • n δnTreatmentn + ǫik

Hussam, Rigol and Roth June 2, 2016 15 / 22

slide-32
SLIDE 32

Results

Main Specifications

How much people know:

Outcomei = α0 + α1Rankik + Xi + γc + ǫik for person i, cluster c, ranker k.

The effect of our treatments:

Outcomei = α0 + α1Rankik +

n βnRankik × Treatmentn + Xi + γc +

  • n δnTreatmentn + ǫik

Who respondents lie about:

Rankik = α0 + α1connectionik +

n βnconnectionik × Treatmentn +

Xi + γc +

n δnTreatmentn + ǫik

Hussam, Rigol and Roth June 2, 2016 15 / 22

slide-33
SLIDE 33

Results

Main Specifications

How much people know:

Outcomei = α0 + α1Rankik + Xi + γc + ǫik for person i, cluster c, ranker k.

The effect of our treatments:

Outcomei = α0 + α1Rankik +

n βnRankik × Treatmentn + Xi + γc +

  • n δnTreatmentn + ǫik

Who respondents lie about:

Rankik = α0 + α1connectionik +

n βnconnectionik × Treatmentn +

Xi + γc +

n δnTreatmentn + ǫik

Standard errors all clustered at the group level

Hussam, Rigol and Roth June 2, 2016 15 / 22

slide-34
SLIDE 34

Results

Respondents Have Valuable Information: Average Reports

Hussam, Rigol and Roth June 2, 2016 16 / 22

slide-35
SLIDE 35

Results

Respondents Have Valuable Information: Residual Information

Hussam, Rigol and Roth June 2, 2016 17 / 22

slide-36
SLIDE 36

Results

Respondents Lie When Reports Count

Hussam, Rigol and Roth June 2, 2016 18 / 22

slide-37
SLIDE 37

Results

Monetary Incentives and Public Reporting

Hussam, Rigol and Roth June 2, 2016 19 / 22

slide-38
SLIDE 38

Results

Who Respondents Lie For: Cross Reporting

Hussam, Rigol and Roth June 2, 2016 20 / 22

slide-39
SLIDE 39

Conclusion

Outline

1

Introduction

2

The Experiment

3

Results

4

Conclusion

Hussam, Rigol and Roth June 2, 2016 21 / 22

slide-40
SLIDE 40

Conclusion

Conclusion

People have valuable information about one another across a variety

  • f domains.

Hussam, Rigol and Roth June 2, 2016 22 / 22

slide-41
SLIDE 41

Conclusion

Conclusion

People have valuable information about one another across a variety

  • f domains.

Remains useful after controling for observables.

Hussam, Rigol and Roth June 2, 2016 22 / 22

slide-42
SLIDE 42

Conclusion

Conclusion

People have valuable information about one another across a variety

  • f domains.

Remains useful after controling for observables.

People lie to us when reports are used for the distribution of resources.

Hussam, Rigol and Roth June 2, 2016 22 / 22

slide-43
SLIDE 43

Conclusion

Conclusion

People have valuable information about one another across a variety

  • f domains.

Remains useful after controling for observables.

People lie to us when reports are used for the distribution of resources.

People lie in favor of themselves, their family members, and their close friends.

Hussam, Rigol and Roth June 2, 2016 22 / 22

slide-44
SLIDE 44

Conclusion

Conclusion

People have valuable information about one another across a variety

  • f domains.

Remains useful after controling for observables.

People lie to us when reports are used for the distribution of resources.

People lie in favor of themselves, their family members, and their close friends.

We expand the development economist’s toolbox for realigning incentives for truthfulness with a variety of simple mechanisms.

Hussam, Rigol and Roth June 2, 2016 22 / 22

slide-45
SLIDE 45

Conclusion

Conclusion

People have valuable information about one another across a variety

  • f domains.

Remains useful after controling for observables.

People lie to us when reports are used for the distribution of resources.

People lie in favor of themselves, their family members, and their close friends.

We expand the development economist’s toolbox for realigning incentives for truthfulness with a variety of simple mechanisms.

Peer Prediction Public Reporting Zero Sum Elicitation Cross Reporting

Hussam, Rigol and Roth June 2, 2016 22 / 22