Latent Constructs in Social Science Research

A precept of conducting quantitative empirical research is the ability to capture numerical representations (data) of phenomena and test relationships among them. To do this, researchers need reliable and accurate measures of phenomena that cannot be measured directly. Such measures are latent variables in social science research.

Psychology, social psychology, sociology, and other social sciences study variables that cannot be observed. Unlike with physical attributes such as height, weight, and volume, researchers cannot measure variables such as extraversion, intelligence, and self-image directly. Thus, latent constructs require a different type of instrument to measure them.

Definition of a latent construct

Latent constructs are theoretical; they cannot be observed directly and therefore cannot be measured directly either. To measure a latent construct, researchers capture indicators that represent the underlying construct. The indicators are directly observable and are believed by the researcher to represent the variable that cannot be observed. Byrne (1998) states:

The researcher must operationally define the latent variable of interest in terms of behavior believed to represent it. As such, the unobserved variable is linked to one that is observable, thereby making its measurement possible. (p. 4)

Latent construct example

Suppose a researcher is interested in measuring subjects' degrees of extraversion with a survey. Extraversion is a latent construct because it cannot be measured directly. Therefore, the researcher must develop or use an instrument that includes items that capture behavior indicative of extraversion.

Extraversion is a trait that represents a person’s need for socializing; an extravert craves and seeks social situations. Consequently, appropriate items on the survey might include asking subjects to indicate how many social situations they engage in over a period or the degree to which they would be willing to attend a social function. Note that these observable variables (i.e., number of social situations and willingness to attend a function) are indicators of the underlying latent construct (i.e., extraversion).

Problems with latent constructs

Although indicators provide a way to measure constructs that cannot be measured directly, there is no way of knowing whether the underlying latent construct has been captured reliably or accurately. A number of statistical procedures can help with quantitative data, but they can only hint at whether the magnitude of some numbers is similar to the magnitude of others.

Many researchers develop items for surveys with the intent of measuring a latent construct but do not consider whether the variable is captured reliably and accurately; they give no attention to the convergent and discriminant validity of the instrument. The result is a confounding element in the research that is discovered when the researcher attempts to publish the study.

Before latent variables can be measured appropriately, testing is necessary to suggest that the indicators chosen to estimate the latent construct do just that. Simultaneously developing an instrument and using it in the same study is unwise because editors and reviewers then have two subjects to criticize in your manuscript instead of one.

If possible, stick with measures that are established well in the literature. If you cannot find a measure, perhaps you have discovered a gap in the literature that a validation study might fill. Always look for opportunities where barriers seem to exist.

Reference

Byrne, B. M. (1998). Structural equation modeling with LISREL, PRELIS. and SIMPLIS. Mahhaw, NJ: Lawrence Erlbaum Associates.