Sample Size for Community Needs & Resource Assessment Surveys
When community leaders go searching for answers, they often decide
that a survey is the best way to collect information. Before conducting
a community survey there are many issues to resolve. For example,
who is to be surveyed? How will the sample be selected? Will your
survey be mailed or conducted by telephone? How many people should
be surveyed? It is tempting to survey everyone, but when the population
is large, surveying every person is difficult, time consuming and
expensive. It is also unnecessary. You don't have to eat an entire
cheesecake to know how it tastes.
A small piece is enough. Like the cheesecake, sampling procedures
allow one to survey a portion of the population and obtain data
that can be considered nearly as reliable as if the whole population
had been surveyed.
Determining the best sample size is not a simple process. The following
factors influence sample size decisions:
- Size - The larger the population, the larger the sample
- Number of sub-groups - Many surveys target different
sub-populations. For example, to separately analyze the results
from surveys completed by the elderly, the sample size must meet
the requirements for a specific level of accuracy for the general
survey and for the subgroup of the elderly. To provide meaningful
information on subgroups usually requires a larger sample size.
- Level of accuracy - Most people work for an accuracy
level of .95 with a .05 confidence interval. This means that you
are 95% certain the results are not off by more than 5%, plus
or minus. The level of accuracy is directly proportional to the
sample size. As the required standard of accuracy increases, so
does the size of the required sample.
- Cost - Fixed costs do not vary with sample size.
These include the cost to design the survey, analyze the results
and write a final report. Fixed costs usually represent a major
portion of the budget. Variable costs increase with increases
in sample size. Variable costs include printing, collecting and
- Time - Time constraints may restrict the number of persons
Is there a Magic Number?
Statisticians use various formulas to calculate sample size. The
following chart provides a guideline to help determine the number
of completed surveys needed for a representative sample with a 95%
confidence level (plus or minus 5%). These numbers represent a good
rule of thumb and can be used for defining sample size for community
surveys where respondents fall into one of two subgroups and there
is equal (50/50) probability of a respondent falling into one of
the two subgroups (e.g., male/female).
However, you may want to survey more than two sub-groups, such
as different school districts, age categories or income levels.
To ensure a representative number of observations in a sub-group,
determine the actual size of that subgroup and use Figure 1 to estimate
a sample size for that subgroup.
Poor response rates are a serious problem in any survey. A needs
assessment survey with a low return rate may not accurately represent
the population. It is important to contact members of the original
sample and request that they complete the survey. Use post card
reminders, letters and phone calls to encourage responses. Send
new surveys. The cost of reminder notes and letters must be incorporated
into the cost of your survey a priori. One cannot compensate for
a low return rate by starting with a larger sample or adding to
the original sample once it becomes evident that the return rate
is inadequate. A low response rate from a large sample is as unacceptable
as a low response rate from a small sample.
For more information about community needs assessment surveys refer
to the Laboratory for Community and Economic Development (LCED)
publication, Conducting Needs Assessment Studies in Your Community,
or ask your local unit office of the University of Illinois Cooperative
Extension Service to contact LCED.
- Babbie, E.R. (1990) Survey Research Methods. 2nd. Revised
edition. Belmont, CA: Wadsworth Publishing Company.
- Dillman, D.A. (1978). Mail and Telephone Surveys: The Total
Design Method. New York: John Wiley.
- Salant, P. And D.A. Dillman (1994). How to Conduct Your Own
Survey. New York: John Wiley.