selected number from 0 to 9. Use the last digit of the
social security numbers listed on the roster that match
that number, and you will have as random a selection as
you most likely will need. If the sample size is too large
(with 10 possible numbers your sample size will be
roughly 10 percent of your total population), then
randomly select another number and match it with social
security numbers that have the same digit in the second
from the end position. Never use the beginning digits of
the social security number because you will end up with
a bias in your survey of respondents who come from the
same geographic area.
WRITING THE QUESTIONS
Social scientist Paul Lazarsfeld said, People do not
answer what you say (or write), but what they think you
mean. Without making the average respondent sound
like the Gilligans Island routine of answering questions
literally for laughs, when you write survey questions,
rememberoften times no one will be around to explain
just what they mean.
There are three standards for good survey question
writing. The first is precision. Does the question ask
exactly what you want it to ask? If you request
information on activities and say recently, do you
know what everybodys definition of recently is? Can
there be any other way to understand the question than
the way you have intended the respondent to read it?
The second standard is relevance to the respondent.
Does the survey ask for opinions on something that the
respondent might have no possible thoughts on? An
example of this would be asking Third Class
Boatswains Mates about F-14 maneuvers. It obviously
is not relevant to them.
The final standard involves personal assumptions.
Do any of your questions assume the respondent will
understand what is meant by a certain phrase or
question? A classic example of this has been written in
scores of research methods books. A British lady asks
another British lady if she was alone last night. The
second lady says with confidence, yes. However, the
house was full of servants. The lady who answered the
question was obviously assuming the questioner wanted
to know if she was with anyone that counted in social
circles. Assumptions are just as dangerous in surveys as
they are in news stories.
A good method of ridding a survey of assumption
problems and other unintended bias is to gather together
some willing souls and have them complete the survey.
Check the answers to the questions and ask the
respondents if they had any problems understanding the
questions. Then listen to their answers. Do not browbeat
them, or call them idiots for not understanding what you
wrote. If they did not get it, how many in the sample you
take will misunderstand as well?
Structurally, the following are the three primary
types of survey questions:
The yes-no question is simple and is used for
data that is censuslike. Demographics is usually
an either-or question. Yes-no categories include
male-female, married-single, and so forth.
The multiple choice question is by far the most
common type of question format used. It gives
the respondent some alternatives beyond the
confining yes-no choice. It is as easy to tabulate
as the yes-no question and provides insight into
intensities of attitudes.
The open-ended question is next to impossible
to tabulate and for this reason, it should be used
sparingly. However, if there is a topic that is
suspected of harboring deep emotions, perhaps
an open-ended question is best. This type of
question lets the respondent express his exact
thoughts. It is also a good indicator of potential
trouble spots your office might not be aware of
at all. There is nothing wrong with leaving a large
blank space at the end of your survey (or the back
side) to record any opinions the respondents
would like to share. You will not be able to
tabulate the responses for statistical reliability,
but they will make excellent informal data.
Another way of measuring the intensity of a
particular belief or attitude, other than the open-ended
question, involves using a scale. The scale most
frequently used is the Likert scale, which is named after
the researcher most associated with its development.
The Likert scale will reveal your audiences perception
of various topics. Moreover, a Likert scale question
makes a statement and categorizes the responses from