Population of Inference

The population of inference refers to the population (or universe) to which the results from a sample survey are meant to generalize. Surveys are used to study characteristics of, and make generalizations about, populations. There are different terms that are used to describe the population, but the most commonly used is the target population, which is a finite set of elements to be studied. However, the term population of inference (or inferential population) is used more often during the conceptualization stage of research studies and surveys.

The target population is specified with the content of the study elements (e.g. general population, students), the units to which the elements belong (e.g. household, school classroom), the geographic boundaries (e.g. country, state), and the time periods (e.g. month, year). In contrast, the population of inference is rather loosely defined in that the specificities of these design elements are not addressed in detail. However, one may argue that the population of inference is more explicitly specified than the population of interest.

To illustrate how these terms are related, consider the following example. A research project may proceed as follows: (a) the population of interest is defined as the general population of the United States; (b) the population of inference for the survey is conceptualized to fulfill the research intent by refining the time scope to the year 2008; (c) the target population is operationa-lized as the noninstitutional civilian persons residing in the conterminous United States between March 1, 2008, and September 30, 2008, who speak English or Spanish. Notice that the definition of the population becomes more crystallized, the breadth of the study population becomes narrower, and the population itself becomes more operationalizable through this process. Of course, these three “populations” may coincide in some cases, for instance, when studying members of certain associations or employees of certain corporations.

Note also that the term universe is also often used as a synonym of population. The main difference between the universe in statistics and the population considered in surveys is that the former contains a hypothetical infinite set of elements generated by a theoretical statistical model, while the latter is more tangible, as it is a finite set of elements existing in real life. Because surveys are operationalized on the concept of finite populations, the term universe is not frequently used in survey research and statistics literature.

Although the conceptualization of each survey is carried out for one specific population of inference and the data collection is conducted on one target population, there may be multiple populations of inference when collected survey data are used. For instance, the [Page 592]Current Population Survey (CPS) in the United States is conducted to provide officiai statistics for the employment and economic situation of the U.S. population. However, individual studies using the CPS data make inferences not only about the general population but also its various subgroups, such as female workers, senior citizens ages 65 and older, households living under the U.S. federal poverty threshold, non-Hispanic whites, rural residents, and so on.