At Causal Map, we do not endorse coding the strength of causal links. You can't really do it in the Causal Map app (see Coding with and using link metadata).

Qualitative impact evaluation is less interested in the strength of effects

Three types of objections to coding causal strength.#

Objection 1: Variable Construction#

Coding strength requires a massive amount of construction work: it involves thinking about the area of interest in terms of variables. This requires modelling specific entities that go up or down, or show differences in number. Constructing variables like this does allow for capturing and calculating correlations. But this construction process is often difficult and does not fit well with how people actually speak in most situations.

If different people are talking about poverty and wealth, employment and unemployment, to be sure you can try to squeeze this all into a shared model with just a couple of variables like say household income and household employment status. But that is a massive abstraction.

• So we favour bare propositions over variables.

Objection 2: Translating to Numbers#

• The second objection concerns the difficulty of translating all relationships into actual numbers.

Challenges with Standardization and Polarity#

• If general rules are used, people usually standardize the variables (e.g., ranging from zero to one). • Standardization is difficult for factors like a country’s population, especially when numbers may increase exponentially over time. • Such changes in magnitude occur even in quantitative sciences, often requiring arbitrary decisions about log transformations. • A more significant problem involves absences, negatives, and polarities. • Example of a strong positive link: If greater anger leads to greater shouting, this connection can be viewed as a "powerful transmission cable" with a high causal coefficient. • Example of a weak link: If anger ranges from zero to one but shouting remains low (e.g., 0.2), the connection has a very low coefficient of transmission. • This type of modelling becomes difficult when negative numbers are introduced. • Example: If high temperatures cause crop failures, a drop in temperature might see harvest go up. • However, extreme cold temperatures also cause crop failures. • It is difficult to model this complexity using a single, bipolar variable.

Objection 3: Aggregation Difficulties#

• The third objection involves the difficulty of aggregating information from multiple sources, assuming such numbers existed. • This third objection is irrelevant in approaches like participatory systems mapping, where a final number for each link is already agreed upon. In this case you could say there is only one source.