The Planner’s Use of Information, Ch.
2
“Survey Methods for Planners” Nancy Nishikawa
The survey is a way of creating an area-specific, customized
database.
- A “sample
survey” allows planners to generalize findings from a relatively small
number of respondents to a larger population.
- Sources
of error
- “Intrusive
measures”
- Operationalization of
complex social issues
- Procedure
- Define
central concepts of study
- Choose
indicators to assess these phenomena
- Construct
questions, for which the response will better describe, evaluate, or
provide understanding for each element
- Cross-validate
survey results against other measures and findings
Survey Research Objectives
Begin by asking questions, rather than identifying specific
data to be obtained.
·
Reason for conducting survey
·
Kind of information needed
o When
to consider alternatives
§
Inappropriate for obtaining accurate information
about sequence of historical events
§
Not good for tapping the flow of activities at
individual or group level
§
Ineffective for assessing involvement in illicit
activities
o Factors
difficult to control
§
Level of interest (can affect response rate)
§
Contamination effects (experience with prior
surveys)
§
Relationship with respondents (some groups more
responsive because of authority issues)
§
Sense of security and privacy
·
Importance of types of information
o Profile
data (characteristics of survey population)
o Environmental
data (circumstances in which respondents live)
o Behavioral
data (relevant social behavior)
o Psychological
data (opinions, preferences, attitudes, awareness, motives, expectations)
·
Level of accuracy needed
·
Users of information
·
Consider the alternatives
o Observation
§
Observations must be fully recorded (don’t rely
on memory)
§
Distinguish between observation of actual
occurrences and interpretation of those occurrences
o Key-Informant
Interview
§
Qualifications of key informants
§
Potential sources of bias from each key
informant
§
Particularly useful where there is a
communication gap
o Group
interviews
§
Focus groups
§
Brainstorming
§
Nominal group technique
§
Delphi process
Administrative Costs of
Surveys
- Interview
costs are usually largest item in budget
- One-third
of interviewer’s time is direct interviewing
- Typical
cost $50-120 per completed interview
- Telephone
surveys are typically less expensive ($15-40
- Mail
surveys are least expensive
- Time
budgets also often underestimated
Design of Surveys
- Cross-Sectional
Surveys
- Unweighted
- Weighted (oversample certain subgroups)
- Contrasting
samples (sampling from groups already known to show substantial
differences)
- Longitudinal
Surveys
- Before-and-after
Study (baseline data can be difficult to obtain)
- Trend
Analysis (periodic data collection)
- Panel
Analysis (trend analysis from the same group of respondents)
Sampling the Population to
Be Surveyed
- Census
(“sampling” everyone in a target population)
- Probability
Sampling
- Findings
of sample survey accurately relate only to population from which sample
was drawn
- Determining
target population:
- Spatial
area
- Respondent
selection
- Bias
(sampling error)—discrepancy between distribution of characteristics in
sample compared to population
- Sampling
Frame
- Sources: lists, registers, maps
- Criteria:
- Cover
entire population
- Complete
(exclude no one)
- Avoid
duplication (include no one twice)
- Accurate
(up to date)
- Available
(must have access to information)
- Types
of error from sampling
- Sampling
error (some characteristic part of population missing from sample)
- Coverage
error (some areas from population missing from sample)
- Measurement
error (response does not
accurately reflect respondent)
- Nonresponse error (systematic difference between
respondents and nonrespondents)
- Types
of sampling frames
- Systematic
sample (randomly select first
unit, then choose every nth
member)
- Simple
random sample (each individual selected by random number)
- Stratified
sample (population divided by stratum, individuals randomly sampled from
strata)
- Cluster
sample (population divided into clusters, random sample of clusters
completely sampled)
- Nonprobability samples
- Availability
sample (survey people passing a specific location)
- Convenience
sample (survey people already assembled in a group)
- Purposive/judgmental
sample (survey based on familiarity with population)
- Quota
sampling (availability sample based on matrix of population
characteristics)
- Volunteer
sampling (self-selected group of respondents)
- Nonprobability Case Sampling
- Matching/contrasting
cases
- Typical
cases
- Critical
cases
- Snowball
technique
- Sampling
Issues
- Population
with greater similarity can be represented with smaller sample
- Stratified
sampling calls for fewest number of cases
- The
more categories needed for each analysis (“cross-tabulations”), the
larger the sample needed
- Rule
of thumb: no fewer than 20 cases
per cell
- Confidence
levels: measure of confidence that
estimate from sample correctly describes population
Methods for Gathering
Survey Data
- Face-to-face
Personal Interview
- Increases
chance that individual will participate
- Increases
likelihood of complete and accurate answers
- Decreases
likelihood of skipped questions
- Permits
“probing” (clarifying) of responses
- Permits
control over sequencing of questions
- Increases
accessibility to isolated populations
- Permits
use of visual aids
- But has high costs, requires
skilled interviewers, and can introduce interviewer bias
- Telephone
Interview
- Greater
economy
- Permits
screening respondents for desired characteristics
- Can
avoid some physical risks
- CATI
permits
- Programmed
skip patterns & preset follow-ups
- Random
presentation of order of responses
- Compressing
data collection and data entry tasks
- Random-digit
dialing overcomes some sampling bias issues
- But difficult to establish social
rapport, problem of hang-ups, and questions must be shorter and less
demanding.
- Mail-in
Questionnaire
- Cost-effective
- Dillman’s (2000) “Tailored Design Method”
- Warm-up
(pre-notice letter)
- Sales
pitch (detailed, persuasive cover letter)
- Thank-you
postcards (a few days to a week later—also serves as a reminder)
- Replacement
questionnaires to non-respondents (two weeks later)
- Phone/special
delivery final contact to nonrespondents (week
later)
- Web-based
Questionnaires
- Even
more cost-effective
- Permits
greater control over survey experience
- Can
include visual aids
- But
- respondents
will have different computer hardware, screen configurations, browsers,
and transmission speeds
- problem
of representativeness of respondents
- coverage
error is an issue, although online user population is looking more and
more like general population
- indiscriminate
access (persons outside target population accessing questionnaire) is
also a problem
- competence
to design online questionnaires may not be readily available
Designing the Questionnaire
- Constructing
questionnaire items
- Open-ended
questions (but creates problem of post-coding the responses)
- Closed
questions (multiple answers create coding problems)
- Useful
making distinctions of degree
- Rating
scale (like semantic differential—no order implied)
- Likert scale (ordinal scale)
- Numerical
scale (ordered scale with equal intervals)
- General
guidelines
- Questions
should be straightforward and unambiguous—avoid opportunities for
misinterpretation
- Avoid
double-barreled questions
- Avoid
expertise error—questions must be within expertise of intended responses
- Avoid
slang or jargon (such terms are subject to misinterpretation, or convey
no meaning)
- Ordering
questionnaire items
- Funnel
sequence vs. inverted funnel
- Arrange
questions so relationship to overall purpose of study makes sense to
respondents
- Pay
particular attention to skip patterns
Evaluating the Survey
Instrument
- Pilot
study—preliminary study, to see what is there
- Pretest—determine
which alternative procedure to use (debrief respondents)
- How
do respondents react?
- Able
to follow instructions?
- Completed
in expected time?
- Choices
in closed questions adequate?
- Trial
run—evaluate operational plan as a whole prior to final run (including
training personnel)
Project
1. Consider the MRAP
Proposal to the EPA, and presume your task is #3 (Interrelationships Between the Basin’s
Diverse “Cultures” and Basin Policy Decisions).
Design a survey to provide data for this stage of the process.
2.
Suppose you were evaluating an in-house staff development program,
“Women in Leadership,” like the Midwest
Women’s Leadership Institute. Devise
a survey to assess the impact of the program on the participants, the
organization, and the community, with the data to be collected 5 years after
the participants completed the program.
3.
Suppose you were asked to assess the city’s new neighborhood housing
program to address the issue of student housing. You have access to the consultant’s
report that led to the program. Design
a community survey to measure their satisfaction with the new program.
Additional Readings (other than those
at end of chapter)
Bearden, Williiam O. & Richard
G. Netemeyer.
1999. Handbook of Marketing Scales.
Newbury Park, CA:
Sage Publications.
Davis, James A. 1971. Elementary
Survey Analysis. Englewood Cliffs, NJ: Prentice-Hall, 1971.
Hyman, Herbert.
1955. Survey Design and Analysis.
NY: The Free Press.
Miller, Delbert C.
& Neil Salkind. 2002. Handbook of Research Design and Social
Measurement, 6th Ed. Newbury Park, CA: Sage Publications.
Nard, Peter M.
2006. Doing Survey Research: A Guide
to Quantitative Methods. Boston, MA: Pearson Education Inc.
Rosenfeld, Paul, and others.
1993. Improving Organizational Surveys.
Newbury Park, CA: Sage
Publications.
© 2006 A.J.Filipovitch
Revised 3 October 2009