Causal Description and Explanation for Social Improvement and Equity

Background: In January 2016, the American Academy of Social Work and Social Welfare launched the Grand Challenges, an ambitious social agenda to improve individual and family well-being, strengthen communities, and create a more just society (Barth,  Gilmore, Flynn, Fraser, & Brekke, 2014). The translation of research evidence into improved and more equitable outcomes is central to the Grand Challenges, illuminating a path for social work researchers to increase the generation of rigorous, relevant evidence to guide policy and practice to positive impact (Hawkins et al., 2015). 

Purpose: In response to the Grand Challenges a thorough compendium of evidence is needed pertaining to causal description and explanation (Shadish, Cook, & Campbell, 2002).

Causal description involves identifying the causal relation between two variables, often a randomly assigned treatment and observed outcome. Researchers use comparisons of treated and control group outcomes to evidence a causal treatment effect and make claims for treatment efficacy. However, estimates of treatment effects do not provide a thorough explanation of the full set of causal factors that potentially combine and interact to produce observed differences in outcomes. That is, causal description does not provide good evidence regarding for whom and under what conditions interventions are most promising. 

Causal explanation, on the other hand, is precisely concerned with identifying the factors that depict how, for whom, and under what conditions effects are observed. Techniques for causal explanation range from exploratory to confirmatory and include theorizing, qualitative analysis, quantitative description, formal tests of mediation and moderation, and latent variable methods Latent variable techniques such as finite mixture modeling have proven especially useful for identifying potential subgroups within samples and the set of unique factors across subgroups.

Method: We reviewed two families of design and analysis techniques that, when combined, form a compendium of evidence consisting of causal description and explanation. The first family of techniques is randomized experimentation, which has a long history in the medical and social sciences. Despite its long history, misunderstandings persist about the conduct of and reporting of findings from randomized experiments (Deaton & Cartwright, 2016; Schulz, Altman, & Moher, 2010). Therefore, we think it advantageous to return to the basics of experimental design, conduct, and analysis to enhance that understanding with recently published articles that support methodological choices currently facing many social work researchers. Then, we move onto the second family of techniques, which are the relatively more modern techniques for exploring potential subgroups and subgroup differences through finite mixture modeling. Best practice recommendations for these methods continue to evolve as new methodological work emerges in the literature. Consequently, we have attempted to synthesize the most recent recommendations for best practice in mixture modeling.

Results: We provide a table of references that will inform choices when using these techniques. This table is intended to serve as a useful guide for thoughtful design and analysis, and ultimately, to improve knowledge and capacity to respond to the Grand Challenges.  With knowledge and capacity we can generate the policies and practices that yield more equitable opportunities and outcomes.