Upload
doantruc
View
230
Download
0
Embed Size (px)
Citation preview
Survey Design
Diane Steele, Juan Muñoz, Raka Banerjee, Talip Kilic
Module 1: Multi-topic Household Surveys
Poverty and Inequality Course
DECPI
March 7, 2012
Qualities of a good survey
Relevance Answer the policy questions of import
Quality Methodology Accuracy and reliability Adherence to international statistical guidelines
Timeliness Punctuality Production time Comparability over time
Sustainability Public dissemination Documentation
The total survey design
approach
Sampling
Data
management
Tools & Methods
Fieldwork
Or the integration of
the development of questionnaires and
methods,
sampling design,
the organization, training and
supervision of the field staff
and computerized quality controls of
fieldwork
to deliver analysts
and policy makers
reliable datasets on time
Survey design and implementation are
processes that require documentation
at all stages
Designing Household Survey
Questionnaires Diane Steele
Module 1: Multi-topic Household Surveys Poverty and Inequality Course
DECPI March 7, 2012
Questionnaire Design vs Sampling
Sampling
Complicated, in range of survey tasks –leave for specialist-
Has theory
Advantage: can tell if done right
Questionnaire design
No integrated theory, some techniques (cognitive testing)
Limited knowledge of impact of changes (LSMS IV need)
Cannot always tell when done right
Process is important
Process
• What are data needs?
• Can household survey meet needs?
• Is hhld survey best option?
Purpose of
Survey
Questionnaire Content
Questionnaire format
Organization of Field Work
Example: LSMS Surveys
Measure Welfare: Levels, Distribution, Causes
As there are multiple measures of interest:
– Consumption
– Income
– Wealth, Savings, Human Capital
– Objective measures
– Subjective measures
LSMS surveys use:
Multi-topic Household Questionnaire
Price Questionnaire
Example: LSMS Surveys
Measure Welfare: Levels, Distribution, Causes
Inform Policy Who Benefits/Uses Programs, public services
Impact of Programs
Use of public services
Outcomes policies designed to effect
Availability of Services to Household
Prices Charged
Quality of Services
LSMS surveys use: Household Questionnaire
Community Questionnaire
Facility Questionnaire
Example: LSMS Surveys
Measure Welfare: Levels, Distribution, Causes
Inform Policy
Determinants Why observed social outcomes occur
Household behavior
LSMS surveys use:
Household Questionnaire
Community Questionnaire
Facility Questionnaire
Starting Point
Discussion with policy makers
– What are key policies to measure
– What are other key issues
Starting Point
Discussion with policy makers
Discussion with Statistical Office
– Issues
– Feasibility
Starting Point
Discussion with policy makers
Discussion with Statistical Office
Create Data Users’ Group
Starting Point
Discussion with policy makers
Discussion with Statistical Office
Create Data Users’ Group
Existing Questionnaires
– Comparability
categories
variables
reference periods, age groups
– Panel
how to track households, individuals
Starting Point
Discussion with policy makers
Discussion with Statistical Office
Create Data Users’ Group
Existing Questionnaires
Qualitative research-
Sequential
Additional topics
Starting Point
Discussion with policy makers
Discussion with Statistical Office
Create Data Users’ Group
Existing Questionnaires
Qualitative research
“Increasing Policy Relevance of LSMS”
– Revision of experience
– Identify issues that hhld survey data can address
– Draft modules linked to policy issues
Starting Point
Discussion with policy makers
Discussion with Statistical Office
Create Data Users’ Group
Existing Questionnaires
Qualitative research
“Increasing Policy Relevance of LSMS”
Research
Concepts
Techniques
Designing for Quality
Multi-topic Questionnaire: Complexity
– Missing values
– Internal Consistencies
– Inaccuracies
– Omission of key issues by analysts
Quality Control Mechanisms
Explicit Questions
Explicit Questions
Example:
Language?
What is the highest
level and grade of
education that
[NAME] completed?
Education?
Quality Control Mechanisms
Explicit Questions
Pre-Coded
Pre-Coded Questions
2. Is [NAME]
......................
22. Where did [NAME] go toobtain medical care?
Male............1Female........2
Own
home...................1
Provider's
home...................2
Health
post...................3
Public Health
clinic.................4
Private Health Clinic..5
Public
Hospital...............6
Private
Hospital...............7
Other..................8
Quality Control Mechanisms
Explicit Questions
Pre-coded
Explicit Skip Patterns
Explicit Skip Patterns
4. Do youcurrentlysmokecigarettes?
5. Have youever smokedcigarettes?
6. At what agedid you startsmoking?
7. In the past30 days did youconsumealcoholicbeverages?
Yes............1 >>Q6No..............2
Yes.............1No...............2 >>Q7
Age in years Yes...............1No.................2 >>Sect. 5
Quality Control Mechanisms
Explicit Questions
Pre-coded
Explicit Skip Patterns
Direct Informants
Quality of data
Respondent burden
Different attitudes
Modules
Roster
Parents of Hhld members
Housing, Utilities
Education
Health
Labor and Other Income
Privatization
Migration
Fertility
Credit
Expenditures /
Consumption
Agriculture
Non-Agr. Businesses
Other Income Hhld
Anthropometrics
Modules Head Individual Best nformed
Roster
Parents of Hhld members
Housing, Utilities
Education
Health
Labor and Other Income
Privatization
Migration
Fertility
Credit
Expenditures /
Consumption
Agriculture
Non-Agr. Businesses
Other Income Hhld
Anthropometrics
Quality Control Mechanisms
Explicit Questions
Pre-coded
Explicit Skip Patterns
Direct Informants
Sensitive Issues Last – Fertility
– Savings and debt
– Income
Quality Control Mechanisms
Explicit Questions
Pre-coded
Explicit Skip Patterns
Direct Informants
Sensitive Issues Last
Packaging
Quality Control Mechanisms
Explicit Questions
Pre-coded
Explicit Skip Patterns
Direct Informants
Sensitive Issues Last
Packaging
Two-round format
Two Round Format
– Breaks up interview
– Allows entry and checking of first half data
– Allows for corrections with respondents
– Reference periods
Quality Control Mechanisms
Explicit Questions
Pre-coded
Explicit Skip Patterns
Direct Informants
Sensitive Issues Last
Packaging
Two-round format
Small Sample
Quality Control Mechanisms
Explicit Questions
Pre-coded
Explicit Skip Patterns
Direct Informants
Sensitive Issues Last
Packaging
Two-round format
Small Sample
Training
Pilot or Field Testing of
Questionnaires
Ensure Questionnaires capable of collecting
required information:
Is the full scope of information needed being
collected?
Covering Credit?
Roster
Parents of Hhld members
Housing, Utilities
Education
Health
Labor and Other Income
Privatization
Migration
Fertility
Credit
Expenditures /
Consumption
Agriculture
Non-Agr. Businesses
Other Income Hhld
Anthropometrics Individual loans:
formal-informal,
unmet demand
Mortgages
Durable
goods
Food
Purchases
Supplier
credit
Supplier
credit
Pilot or Field Testing of
Questionnaires
Ensure Questionnaires capable of collecting required information:
Is the full scope of information needed being collected?
Does every section respond to policy needs?
Is the information collected in different sections internally consistent?
Specific Sections
Is the section inclusive?
Ex. Does the questionnaire include all
prevalent
– activities
– living arrangements
– sources of income
– consumption items
Individual Questions
Clear in all circumstances?
Lead to unambiguous responses?
Alternative interpretations?
All responses anticipated (pre-coded)?
Are skip patterns accurate/complete?
Pilot Test: How to
Involve both field staff and analysts
Test all instruments
In lab, in field in all conceivable situations
Ensure enough responses to each section
to test well
Document Process
Testing Employment Module
Self-employed farmers
Self-employed businesspersons
Employees
Unpaid family workers
Unemployed
Homemakers
Males, females, children
Testing Housing Module
Homeowners
Renters
Squatters
Multi-family Units
Testing Agricultural Module
Subsistence Farmers
Cash-crop Farmers
Landowners
Hhlds raising Livestock
Absentee owners
Revisions
Leave plenty of time to revise
– affects questionnaire
– training
– data entry
Test new revisions
– Take multiple versions to field
DOCUMENTATION CHECKLIST
• QUESTIONNAIRE DESIGN
– Questionnaire
– Pilot test report
– Interviewer manual
– Supervisor manual
– Other manuals