how to analyse outcomes and their negation?

[Simon Toubeau]

Hi everyone,

My co-author and I are writing a paper on policy coordination (and its absence) in multi-level states responses to Covid. We have 13 countries in 4 phases over 2020 (=52 country obs).

We have three outcomes grounded in two concepts of shared crisis responsibility or regional crisis responsibility. So it's a cube with four corners. But there are only three outcomes: centralization (0,0), and non-coordination (0,1), and coordination (1,1). [One of the corners is logically possible, but does not exist- the central government coordinates with regions that have no self-government]

Each outcome is a negation of the other. Eithe the central government assumes all responsibility, or it leaves the regions to go their own way, or it coordinates with regional governments.

Question: How do you think we should we analyse the outcome and the negation of the outcome when there are two (rather than one) categories that are antithetical to the outcome?

This has affected the coverage/relevance of our solutions because there are always many more cases that are not included in the outcome (or included in the non-outcome) than are included, and we have this for each outcome…

In the same vein, when countries have a weak membership score to a category (e.g. coordination), this could suggest that they have a strong membership to two other categories, either centralisation (e.g. France) or non-coordination (e.g. USA).

All the articles I have read for inspiration (e.g. Vis, Emenegger, Schneider et al.) have a single dimension for the outcome and its negation ( e.g. Job security regulation and its negation, goes from more or tighter regulation/protection, to less or looser regulation/protection).

I have two solutions in mind:

1) Either we collapse the two dimensions of the outcome into a single dimension. This would range from 0 (centralisation) to 2 (coordination).

But this would overlook the fact that centralisation and coordination are both centralised approaches, since the central government is involved, but in the latter regions are influential.

2) In the analysis, we construct three outcomes and their singular negation, where we look at 1) coordination vs non-coordination, 2) centralization vs non-coordination, 3) coordination vs. centralisation.

- We would be taking out certain case observations depending on the outcome and its negation being investigated. e.g. cases of centralization are taken out when looking at the coordination vs non-coordination.

- The calibration of each case would thus be unique to each outcome and it's singular negation.

Any thoughts/idea/comments/feedback would be gratefully received!

[Adrian Dușa]

Hello Simon,

Just a few quick thoughts.

If the outcomes are mutually exclusive, then it doesn't really matter if centralisation and coordination are both centralised approaches. That is a theoretical distinction that QCA is unaware of.

If there are cases that are found in both (as centralised approaches) then they would not be mutually exclusive. If they are, it doesn't matter.

I believe the most beneficial and simple thing would be to create a multi-value outcome (1, 2 and 3, or 0, 1 and 2 that works just the same) and then create separate analyses for each, using outcome = Y[1]

(for instance, where Y is the outcome and 1 is the value to be explained)

The negation of that would be outcome = ~Y[1]

(if I'm not mistaken)

For three values in the outcome, you'd have three analyses for the presence and three for their negation, where the negation of value 1 is everything having values 0 and 2 (or 1 and 3, depending how you coded).

I hope it helps,

Adrian

--
You received this message because you are subscribed to the Google Groups "QCA with R" group.
To unsubscribe from this group and stop receiving emails from it, send an email to qcawithr+u...@googlegroups.com.
To view this discussion on the web visit https://groups.google.com/d/msgid/qcawithr/c5cfe187-7597-4447-b0be-848bd2e36e54n%40googlegroups.com.

[Patrick A. Mello]

Hi Simon,

 

Thanks for your question. 

 

In my view, this is outside the scope of the "QCA with R" list because it's essentially a research design question. That also makes it difficult to answer from the outside. You are probably best situated to say what the most suitable conceptualization of your outcome(s) is and what you aim to achieve with the QCA design.

 

That said, your proposed solution of collapsing the two dimensions into a single outcome sounds quite reasonable to me. I would try to reduce the complexity of the research design, also because this makes it more amenable to interpretation once you get your QCA results. It might also be a good option to explore the trajectories of individual countries after the QCA, be it case studies or case illustrations. 

 

Best,

Patrick

 

dr. Patrick A. Mello 
Assistant Professor of International Security, Department of Political Science           

and Public Administration, Faculty of Social Sciences 

p.a....@vu.nl | Research Profile | www.patrickmello.com | 
De Boelelaan 1105, 1081 HV Amsterdam | www.vu.nl | 

--