Transcript Lecture 2

STATISTICS FOR MANAGERS
LECTURE 2:
SURVEY DESIGN
COMPONENTS OF SURVEYS
1.
2.
3.
4.
Sampling
Methods of data collection
Question design
Interviewing
1. SAMPLING

How well a sample represents a
population depend on:
1.
2.
3.
Sample frame
Specific selection procedure
Sample size
1.1. SAMPLE FRAME

Sample of people that has a chance
to be selected, given the sampling
approach. Classes:
Sampling is done from a more or less
complete list of individuals of the
population
 Sampling is done from a list of people
that go somewhere or do something
 Sampling is done in two or more stages

1.1. SAMPLE FRAME

Characteristic of the sample frame
to be evaluated:

Comprehensiveness: a sample can be
representative only of the sample
frame. Most sampling leave out a few
people. Examples: household based
sample exclude people in dormitories,
prison and nursing homes.
1.1. SAMPLE FRAME

Characteristic (cont.)

Probability of selection.It is not
necessary that a sampling scheme give
every member in the sample the same
chance of selection. However, we
should be able to find out the
probability of selection for each
individual selected. Example.
1.1. SAMPLE FRAME

Characteristic (cont.)

Efficiency. In some cases sampling
frames include units that are not
among those wanted to sample.
Examples: elderly using general
household and eliminating with no old
people; random dialing. Only issue:
cost.
1.2. SELECTION OF UNITS

One stage sampling:
Simple random sampling
 Systematic sampling
 Stratified sampling


Multistage sampling: no way to get
at the population directly. Link
population with groups.
1.2. SELECTION OF UNITS

Simple random sampling: members
of a population are selected one at a
time, independent of one another
and without replacement. Unless the
list is short and all unit are
prenumbered drawing from a simple
random sample is very laborious.
1.2. SELECTION OF UNITS

Systematic sampling. The
researcher determines the number
of entries in the list and the number
of elements to be selected. For
instance list of 8500 people and 100
required. 1 out of 85 should be
selected. The process start with a
random number.
1.2. SELECTION OF UNITS

Systematic sampling (cont.). Then
we take any 85th person in the list.
Problem: if the list is ordered
following some criteria or has a
recurring pattern that will differently
affect the sample depending on the
random start. Examples: couples
club were the male partner is always
listed first: only male or female.
1.2. SELECTION OF UNITS

Stratified sampling. Generally little is
known about the characteristics of
the individuals in the population
before the sample. However, we
may know some characteristics.
Usually stratification involves some
regional variable.
MULTISTAGE

Examples:
Sampling students from schools
 Area probability sampling
 Random digit dialing


Respondent selection: who should
be interviewed in the household?
MULTISTAGE
If information is easy to report any
adult who is home can answer the
question.
 If information is specialized the
researcher wants to interview the
household member most
knowledgeable.

MULTISTAGE
Group work: Why are registered
unemployed many less than survey
unemployed in Spain?
 Data

1.3. SAMPLE SIZE

Sample size and sampling errors will
be covered in the next lecture.
2. DATA COLLECTION

Choice of data collection mode
depend on:
Sampling
 Type of population
 Available staff
 Response rate
 Question form

2. DATA COLLECTION

Methods
Personal interview
 Telephone interview
 Self-administrated
 Mail
 Drop off and pick latter
 Internet

2.1 Personal interview

Advantages
There are some sample design that can
be implemented best by personal
interview (area probability samples)
 Most effective way of enlisting
cooperation for most populations
 Interviewer can answer respondents
questions, explain complex questions,
etc.

2.1 Personal interview

Advantages (cont)
Multimethod data collection (including
visual cues) are feasible.
 Confidence building is possible
(including reassurances for sensitive
material)
 Probably longer survey instruments are
possible in person than in any other
mode

2.1 Personal interview

Disadvantages
It is likely to be more costly than the
alternatives.
 A trained staff of interviewers that is
geographically near the sample is
needed
 Total data collection period is likely to
be longer than telephone procedures

2.1 Personal interview

Disadvantages (cont)

Some samples (high rise building, high
crime areas, elites, students) maybe
more accesible by some other mode
2.2 Telephone interview

Advantages
Lower unit cost than personal
interviews
 Random digit dialing (RDD)
 Better access to certain population
 Shorter data collection periods
 Interviewer administration (better than
mail or internet)

2.2 Telephone interview

Advantages (cont)
Interviewer staffing and management
easier than personal interviews: less
staff, not necessary to be near the
sample, supervision and quality control
potentially better
 Likely better response rate from a list
sample than mail

2.2 Telephone interview

Disadvantages
Sampling limitations (omitting those
without telephone).
 Non response higher than personal
interview.
 Questionnaire or measure constraints.
 Less appropriate for personal or
sensitive questions.

2.3 Self-administrated

Advantages
Easy presenting questions requiring
visual aid.
 Asking questions with long and complex
response categories.
 Asking batteries of similar questions.
 Respondent do not have to share
answer with interviewer.

2.3 Self-administrated

Disadvantages





Specially careful questionnaire design is
needed.
Open questions usually not good idea.
Good reading and writing skills by respondent
are needed.
The interviewer cannot exercise quality
control
Cannot control who answer the questions
2.4 Mail

Advantages
Low cost
 Minimal staff and facilities
 Provide access to widely dispersed
samples and people difficult to reach by
phone or in person.
 Respondent have time to give a
thoughtful answer, to look up records,
etc.

2.4 Mail

Disadvantages
Ineffectiveness of mail as way of
enlisting cooperation
 Disadvantages of not having
interviewer involved in data collections
 Need for good mailing addresses for
sample.

2.5 Drop off-pick up later

Advantages
The interviewer can explain the study,
answer questions and designate a
household respondent.
 Response rates tend to be like those of
personal interviews.
 More opportunity for thoughtful answer
and check out the records.
 Does not requires trained interw. staff

2.5 Drop off-pick up later

Disadvantages
Cost about as much as personal
interviews.
 Field staff is required (although less
trained that for personal interviews).

2.6 Internet surveys

Advantages
Low unit cost of data collection
 Potential high speed of return
 All the advantages of self-administrated
instruments.
 Al the advantages of computer assisted
instruments
 Time for checking out records, etc.

2.6 Internet surveys

Disadvantages
Limited to samples of internet users
 Need for good addresses
 Challenges of enlisting cooperation
 Various disadvantages of not having an
interviewer involved in data collection.

Exercises
How is the Census conducted?
 Other special samples:

Trial juries
 Measures of TV shares
 Measures of effectiveness of ads.

3. Question design
Decide what you want to find out;
this is the most important step in
writing a questionnaire.
 Always test your questions before
taking the survey.
 Keep it simple and clear
 Use specific questions instead of
general ones, if possible.

3. Question design II
Relate your questions to the concept
of interest.
 Decide whether to use open or
closed questions.
 Report the actual question asked.
 Avoid questions that prompt or
motivate the respondent to say what
you would like to hear.

3. Question design III
Use forced-choice rather than agreedisagree questions.
 Ask only one concept in each
question.
 Pay attention to the question order
effect.

3. Question design IV

Specific issues:
Contingent valuation
 Data on the registry of accounting.
 Trust questions and experiments.
