Nonexperimental Research in Quantitative Research

Quantitative research for which the researcher does not manipulate an independent variable or randomly assign participants to treatment groups. Terms such as survey, correlational, descriptive, causal-comparative, ex post facto, predictive and explanatory, passive observation, natural experiments, meta-analysis, and cross-sectional and longitudinal describe the members of this family of research in the social sciences. A useful new 3 × 3 classification system classifies nonexperi-mental quantitative research along two dimensions: (1) primary research objective and (2) time. On the dimension of primary research objective, studies can be descriptive, predictive, or explanatory; on the dimension of time, research can be cross-sectional, longitudinal, or retrospective. Thus, it is possible to have nine different types of nonexperimental quantitative research, each mutually exclusive.

A major limitation of nonexperimental research in quantitative research is that it is very difficult to provide evidence of causality. Indeed, some researchers claim that it is not possible to establish causality from this type of research, but this view is problematic. It is important to understand that experimental research is the most powerful means of establishing causality. However, under the proper conditions, one can provide tentative, preliminary evidence of causality with nonexperimental research. Certain procedures, such as matching or statis tically holding the extraneous variable constant (e.g., partial correlations, analysis of covariance), can be used to control for the threat of potentially confounding extraneous variables identified by the researcher, thereby supporting a possible causality argument. In addition, causal modeling or structural equation modeling, another statistical technique, tests data against theoret ical models for evidence of causality. Overall, with a substantial, consistent body of nonexperimental evidence as well as strong theoretical support, one at least can suggest plausible cause and effect. For an example of such research, one has only to think back to the decades of nonexperimental research suggesting the link between smoking and cancer before smoking was declared a causal agent of lung cancer.

Remember that while experimental research may be the gold standard in scientific research, it is not always feasible or ethical to conduct such research. Even as we might prefer experiments for establishing cause and effect, in educational research, we must often use nonexperimental research to address our important research questions. Researchers must be certain that the research questions guide the study and then employ the most powerful research methods available to answer these questions. For more information, see Johnson (2001) and Johnson and Christenson (2008).