Standard Definitions

Importance of Outcome Rates

Outcome rates of sample surveys and censuses are important to understand because they are indicators of potential nonresponse effects. Nonresponse and its effects are one of several types of potential nonran-dom errors in surveys (others include coverage error and measurement error). Nonresponse bias is the extent to which the representativeness of a sample is compromised if there are nonnegligible differences between nonresponders (i.e. those who were in the originally drawn sample but who did not complete an interview) and respondents (those who did complete an interview).

A sample is a set of elements, or cases—in social science research, they usually are people or households—drawn from some population. The population may be adults in the United States, likely voters in an election, newspaper readers in a metropolitan area, college students on a campus, or some other identifiable group. The researcher measures the characteristics of these cases, usually with an instrument called a questionnaire, and in the subsequent statistical analysis infers the statistics from the sample to the population. Error bias in the sample can hamper the ability of the researcher to infer accurately the sample statistics to the population.

Just as doctors can get indicators of a patient's health by checking key components of blood chemistry, researchers can get hints at sample validity by carefully performing certain checks on the sample. One, for example, is comparing sample statistics to known population parameters: the sample's percentage of men and women, say, to U.S. Census percentages. A second is to compare the sample's findings to other, similar samples done with similar populations at about the same time. Outcome rates also can be indicators of sample validity, but they should never be used as the sole judge of a sample's quality.

...