Another crucial component outlined in the research plan is the process of selecting the appropriate research methodologies. This

process may include the following:

  • Developing a clearly defined methodology to determine what information is needed for a particular research regardless of topic, which encourages a more systematic design and builds confidence that the findings are comprehensive and robust. Examples of common research methodologies include:
    • Descriptive and exploratory research.
    • Quasi-experimental designs using found, constructed or statistically matched comparison groups.
    • Randomized “clinical” or “community” trials (RCT), reducing the likelihood of making biased comparisons by using random assignment of individual cases or community units.
    • Systematic Reviews or Meta-Analysis.
    • Action Research.


Although a description of all possible research approaches and designs that can be used in child welfare research far exceeds the scope of this guidance, this chapter presents five research approaches that can be relatively readily compared.1 The first three types of studies are non-experimental and the latter two are experimental.


Exploratory studies identify questions and characteristics that might help to explain a phenomenon of interest about which little is known. Such a phenomenon might include understanding a population that is new to the child welfare agency or a problem that has not been previously encountered.


Descriptive studies endeavor to clarify basic characteristics of people, services or other events that can help explain something about who is being served and what is happening in the process. These studies help to group populations or processes together in meaningful ways so that their dimensions can be understood and responsive alterations in services considered.


Correlational designs investigate the relationship between two variables or attempt to understand the strength and direction of relationships. For example, a correlational study can help to tell how the age of a child at the time of adoption is related to the number of post-placement services that their family requests in the following year. This is a starting point for determining a relationship but does not necessarily demonstrate a causal link (e.g., the expectations of the adoptive parents or the educational readiness of the child).


Quasi-experimental designs are those that increase the level of certainty that a change in a service or program component causes a hypothesized outcome of interest. Quasi-experimental designs help to overcome some limitations of RCTs because some conditions (e.g., placement into kinship care or for adoption) cannot be randomly assigned as can participation in other services (e.g., conventional vs. wrap around services). In these cases, a quasi-experimental design that uses matched comparisons or propensity score matching may be necessary to try to minimize the selection bias (e.g., that less difficult children are more likely to be accepted by kin or to be adopted than are children with more complex and intense problems).


Experimental designs, also known as randomized clinical trials, are designs that manipulate the provision of a service or living condition so that program participants have equivalent chances of receiving any of the conditions (few experimental designs in child welfare have more than three conditions). An ethical necessity of these designs is that there is no compelling evidence that any of the conditions have significantly better outcomes than any other. These designs are favored by policy makers because they provide the least ambiguous results of all the ways of studying program outcomes. Unfortunately, they are also among the most

expensive and difficult to conduct and are not feasible in all research situations.

Change & Innovation Agency
FEi Systems
PCG Human Services
Governing Magazine