Constructing a causal model from scratch may be difficult. Especially if answering your research question involves theories from multiple disciplines, it is unlikely that you can rely on causal models from previous research.

If you use existing data sources, and cannot gather additional data, start with drawing a causal model including the variables available in your data. Also if you collect your own data, drawing a causal model connecting the variables that you will collect data on is a helpful tool to get your thoughts in order.

To construct a comprehensive causal model, it is often helpful to start working from the question: “Who does what, when, how and why?” This question consists of three parts. For each of these parts you can design a model:

  • The “Who does What?” part can be displayed in an actor model, that identifies the relevant actors in a field.
  • The “What happens When and How?” part can be displayed in a process model, that identifies chronological phases in the process that you are trying to understand.
  • The “Why?” part is displayed in a causal model.

Actor Models

Actor models are useful because they identify different groups of actors that are important in a field. An example of an actor model is displayed below. The figure shows three groups of actors: donors, intermediary organizations, and recipients. Donors form the supply side of contributions; nonprofit organizations represent recipients and form the demand side of contributions. It is important not only to identify who acts, but also what these actors do. When drawing an actor model, ask yourself: Who does What?

This model contains no positive or negative signs because some characteristics of the actors have positive relationships and others have negative relationships with characteristics of other actors. Also it is difficult to quantify the importance of the actors. The importance can be determined in various respects that cannot easily be displayed as causal influences.

Process Models

Process models are useful to establish the temporal order of variables. An example of a process model is displayed in the figure below. The model simplifies the actor model somewhat, showing only two of the actors (donors and nonprofit organizations) from the figure above in separate rows.

Process models typically include too many variables for a single research project. One could design a complicated causal model for each of the phases. The process model above contains many interesting research problems. One example is: “How do potential donors decide what amount to donated to nonprofit organizations?” This research problem concerns only one phase of the process (Phase 5) and could be the question to be answered in a bachelor or master thesis. A second research problem concerns two phases (Phase 1 and 2): ; “How do nonprofit organizations design programs to address needs of recipients?” This question is too broad for a bachelor thesis; it could be a question to be answered in a master thesis. A third research problem, “How do fundraising campaigns of nonprofit organizations express needs of recipients and how do perceptions of these needs by potential donors influence charitable giving?” concerns three phases (Phase 3, 4 and 5). This question deserves a PhD dissertation.

Typologies, mind maps and conceptual models

In addition to actor models and process models there are three other forms of models: typologies (or idealtypes), mind maps and conceptual models. If you are writing an empirical paper or thesis in which you are testing hypotheses my advice is to minimize the use of typologies, mind maps and conceptual models. Typologies and idealtypes may be useful heuristic shortcuts when you want to classify actors or phenomena. They hinge upon the systematic co-occurrence of characteristics, some of which may be clustered.

The causal model is by far the best tool to display hypotheses. The other models include too much information, are too general or abstract, or include the wrong kind of information – the kind that cannot be used to formulate a hypothesis. Certainly do not include mind maps or conceptual models in your theory section. Some of the other models may be useful when you have trouble to come up with hypotheses. But they tend to get confusing when you include arrows that reflect causal influences.

In truly exploratory research it is good to start with a mind map, in which relevant concepts are logically ordered from a core to an increasingly distant periphery. Mind maps may be useful when you are working in a group and you need a common understanding of the logical order of concepts. However, a concept is not a hypothesis about a relationship between two variables. Conceptual models suffer from the same problem as mind maps: arrows in conceptual models reflect theoretical linkages, but not causal influences.

From a cluster model to a causal model

A common problem in the construction of causal models is that clusters of variables appear that summarize groups of variables: you find yourself putting ‘differences’, ‘individual characteristics’, ‘social groups’ or ‘external factors’ in boxes and drawing arrows between them. The problem of such a model is that it encompasses too many variables to be tested. It does not give you much direction in your analyses. Anything goes. If your research is purely exploratory, that is fine. But in the more frequent case, your cluster model is a result of not choosing the most relevant variables. The challenge now is to get to a more limited set of variables. Which options are possible within the clusters that you have identified? Which differences, characteristics, groups or factors seem most interesting? When you put these words in the model, force yourself to specify which one is most important. This does not mean you have to limit yourself to just one forever. You must kill your darlings – but you can save them for later.