Mini-post: Are risk matrices “better than nothing?”

Risk matrices have many limitations but are they better than nothing?

Ron Gantt recently penned an article exploring how injury metrics may not be adequate safety performance measures (link below).

This got me thinking about similar arguments I’ve heard justifying other tools or methods, like risk matrices. A recent study highlighted a range of important structural and usage limitations with matrices. These include: risk acceptance inconsistency, range compression, centering bias, category-definition bias, ranking reversal and more (summary of the study + more on matrices below).

Importantly, they argued that some of these structural flaws were inherent to the design of the matrix and can’t be corrected.

Now, whether you agree or not with this sentiment or believe matrices can or can’t be beneficial (which isn’t what I’m arguing here), the authors highlight that describing the structural flaws and arbitrary rankings from matrices “does not require [anybody] to provide another method in their place … [the] arbitrariness of [some risk matrix results] is not conditional on whether or not other alternatives exist” (p63) **.

I think this is an important point: the validity and reliability of a matrix, like anything really, doesn’t depend on there being no other option, nor is it necessarily “better than nothing”.

[** They’re not suggesting there’s no alternative way to assess risk and risk decision aids etc, there’s clearly lots, but more around a lack of a direct and accepted replacement for the matrix.]