Laserfiche WebLink
User Guide to The National Citizen SurveyT" <br />Understanding Survey Research <br />Survey Sampling <br />We systematically select households from a geocoded United States Postal Service (USPS) address list <br />to ensure that only households located within the boundaries of a community are surveyed. Systematic <br />sampling is a procedure whereby a complete list of all eligible addresses is culled, selecting every Nth <br />one (a number that changes depending on the size of the population and the sample size to be selected) <br />until the appropriate number of addresses is sampled. Not only does NRC scientifically and randomly <br />sample households to participate in The NCS, but we also select, without bias, the household member to <br />participate. This methodology helps ensure that the attitudes expressed by our respondent sample <br />closely approximate the attitudes of all adult residents living in the community. Without controlling <br />who in the household participates, it is likely that results would be biased towards those who are more <br />sedentary and those without jobs (who may have different opinions about some services). <br />The Basic Service of The NCS includes mailing to randomly selected households. Though response rates <br />across the US have dipped in recent years, the response rate for most administrations of The NCS <br />ranges between 20% and 40 %, which yields between 30o and 48o completed surveys. <br />Margin of Error and Confidence Intervals <br />It is customary to describe the precision of estimates made from <br />Number of Margin <br />surveys by a "level of confidence" and accompanying "confidence <br />completed surveys of error <br />interval" (or margin of error). A traditional level of confidence, and <br />100 ±9.8% <br />the one used for The NCS, is 95 %. The 95% confidence interval can <br />300 ±5.7% <br />be any size and quantifies the sampling error or imprecision of the <br />400 ±4.9% <br />survey results because some residents' opinions are used to estimate <br />500 ±4.4% <br />all residents' opinions. The relationship between sample size and <br />750 ±3.6% <br />precision of estimates or margin of error (at the 95% confidence <br />level) is shown in the adjacent table. With a typical sample size for The <br />NCS, this means an estimated <br />margin of error at the 95% confidence level of plus or minus four to six <br />percentage points. <br />A 95% confidence interval indicates that for every loo random samples of the same number of <br />residents, 95 of the confidence intervals created will include the "true" population response. This theory <br />is applied in practice to mean that the "true" perspective of the target population lies within the <br />confidence interval created for a single survey. For example, if 75% of residents rate a service as <br />"excellent" or "good," then the 4% margin of error (for the 95% confidence interval) indicates that the <br />range of likely responses for the entire community is between 71% and 79 %. This source of uncertainty <br />is called sampling error. In addition to sampling error, other sources of error may affect any survey, <br />including the non - response of residents with opinions different from survey responders. Though <br />standardized on The NCS, on other surveys, differences in question wording, order, translation and <br />data entry, as examples, can lead to somewhat varying results. <br />For subgroups of responses, the margin of error increases because the sample size for the subgroup is <br />smaller. For subgroups of approximately loo respondents, the margin of error is plus or minus 10 <br />percentage points. <br />17 <br />