I am proposing to research the causes of humanitarian intervention because I would like to find out why some states/IGOs intervene militarily in humanitarian crises, in order to help my reader better understand why some humanitarian crises result in military intervention while others do not.[1]My research question would be more precisely articulated for large-n analysis as:
What explains the variation in states/IGOs responses to humanitarian crises?
In this instance, my dependent variable would be states/IGOs’ responses to humanitarian crises. In a preliminary sense, I have considered quantifying and operationalizing “response to humanitarian crisis” as ordinal numbers of 1-5 ranked as: (1) no response, (2) humanitarian aid delivery, (3) naming/shaming in UN resolutions, (4) sanctions, and (5) military intervention/peacekeeping. Had my research question focused on only one of these response types, it would have observed the variance in the scale of, say military intervention or humanitarian aid delivery, rather than the overall variance in the extent of response. I am attempting to study the latter. Binder does something similar to what I seek to accomplish when he defined his dependent variable as “strength of UN response”, though it lacked an interval number/unit of measurement.[2]As I am consulting various datasets, I am also searching for ways to operationalize my dependent variable into interval numbers. These contemplations are still early at this stage and will naturally require the use of datasets.
Moreover, it appears I will have to compile data from various datasets to assemble my own. This is because these datasets rarely include data exclusive to my cases (humanitarian crises). Some, like the Correlates of War Project’s “Militarized Interstate Disputes” dataset observe instances of military disputes (including those in response to humanitarian crises as well as those that are not) as cases.[3]The coverage of this dataset is first a limitation not in terms of breadth -as it stretches from 1816-2010-, but in the difficulty of “cleaning” the set to include only military disputes responding to humanitarian crises. Additionally, another concern about this sources’ limitation is its isolation of militarized responses and neglect of other types of responses which I must examine in my research. Though this dataset is very expansive within the realm of militarized responses (including anything from “threat to use force” to “declaration of war” as such responses), it still ignores the dimensions I listed above: the delivery of aid, naming/shaming, and sanctions.[4]Nevertheless, this source still provides helpful data including both the type/degree of militarized response, hostility levels, and number of fatalities. This dataset will also be supplemented by several others including The Correlates of War Projects’ “National Material Capabilities” dataset which includes figures for one of my independent variables – countervailing power –measured in part via military expenditure, military personnel, population size, etc.[5]The coverage and limitations of both mentioned datasets are identical.
Specifically, I would clean the first dataset by essentially “adding” a new column with a nominal indicator describing whether the instance of military intervention was in response to a humanitarian crisis or not (a definition which is of course a methodological decision of case selection which must in turn be justified). I would then exclusively use data from the remaining cases indicated as relating to humanitarian crises. As these instances would still only be cases of humanitarian crises that have provoked military intervention, my compilation of a dataset would be far from finished as I would have to do the same to other datasets I find that will have to consult, including those dealing broadly with the delivery of humanitarian aid, sanctions, etc.
[1]Booth, Wayne; Colomb, Gregory; Williams, Joseph and Fitzgerald, Willam. The Craft of Research, 4thEdition. (Chicago: University of Chicago Press, 2016), 54.
[2]Binder, Martin. 2015. “Paths to Intervention.” Journal of Peace Research 52 (6): 716.
[3]Correlates of War Project; Palmer, Glenn; D’Orazio, Vito; Kenwick, Michael R.; and McManus, Roseanne W. McManus; “Militarized Interstate Disputes (v4.3).” 2019.http://www.correlatesofwar.org/data-sets/MIDs. Accessed October 9, 2019.
[4]Ibid.
[5]Correlates of War Project; Sing, David J.; Bremer, Stuart. “National Material Capabilities (v5.0).” 2019.http://www.correlatesofwar.org/data-sets/national-material-capabilities. Accessed October 9, 2019.
Hi Mohammad!
You map out your question and datasets really well here, and I appreciate that your thought process is clearly expressed. I am also thinking about ways my dependent variable could be operationalized and measured at the interval level when no one specific dataset exists for it. I think it would be interesting to look at proxy interval measurements (something my mentor had recommended I do). You could find interval measurements for indicators like each of the five responses (like how much foreign investment there is or aid money spent there was after a crisis) or other similar topics and use those to make claims about and analyze your specific puzzle.
I’m interested in seeing what cases you decide to use (there are obviously many and it’s still early on in the research process) but I think once you have a solid list you can also gain interval measurements on a case by case basis. It would be interesting to see what information/data is or is not available for each and why. Individual information gaps or variances could possibly explain the larger variation in responses that you’re aiming to study.
Wow, that’s actually incredibly helpful! Would, say, the proxy interval measurements be combined into a single total measurement on “outcome” or would the measurements simply be standalone in a way, and then the analysis from there just guide it? Not sure if I’m being clear but I think it’s a very productive idea you suggested! Finally, as I’m thinking of cases as you mentioned, I’d be interested if you could elaborate on what you mean by “Individual information gaps… could possibly explain the larger variation in responses that you’re aiming to study”. Thanks a lot!
Mohammad — you are off to a good start here in considering what the Correlates of War Project could offer you in terms of data. Remember that we have a range of conflict-related databases (the Uppsala Conflict Data Program is another good source) from which you could collect this information. As you think about operationalizing your DV it is worth being a bit more precise in terms of what it is about “response” (quite a vague concept) that you want to capture. It is the amount of resources committed? The time committed? Some other aspect? Making the concept of “response” more precise will help you better locate targeted data. Remember, as well, that we really want a DV that is captured with an interval/ratio indicator for this methodology (note, as well, that the scale that you propose would really be an nominal scale since there is no inherent value ranking between the categories). How could you think about developing (or drawing from existing data) an interval/ratio indicator for your DV?