I am doing a series of articles on the book “Rethinking Canadian Aid” (University of Ottawa Press, 2015), and now it’s time for “Chapter 6: Mimicry and Motives: Canadian Aid Allocation in Longitudinal Perspective” by Liam Swiss. As I start the critique of this section, I have to note three large disclosures.
First and foremost, Liam is a close friend. While it wouldn’t stop me from disagreeing with him on just about any subject (!), I respect his work quite a bit. In fact, it is the primary reason I’m reading the book — it includes a chapter from him.
Secondly, and I’ll lose a lot of credibility here, I quite like the chapter and it is the best one so far. That is quite unrelated to my first disclosure, and far more related to my third disclosure.
Third, I have a bias for numerical evidence and linear extrapolations from data. I am not a statistician, and am often quite suspicious of advanced statistical techniques when it leaves the academic world and approaches the policy world. What works in science, where a lot of variables can be held constant, seems to lose application when it gets warped and tortured into overall policy analysis and then re-engineered into a case-by-case application. I may be able to follow the gist of using wavelet applications to analyse non-linear epidemiological data, it doesn’t mean I want to use it for practical policy analysis where those pesky other variables that worked so well in theory tend not to remain constant. Which means I like data analysis that goes just “so far” and “no farther”. You’ll see later why I mention this bias before getting into the actual document.
…This chapter uses aid allocation patterns to do two things: (1) discern which other donor countries Canada’s aid allocation most closely resembles over time to identify which countries and, in turn, motives Canada may be emulating in its aid practices; and (2) examine several key factors underpinning the provision of aid to recipient countries on a dyadic basis to highlight the motives that drive Canadian aid relationships over time.
Such a simple statement, and yet the basis of any good discussion of aid “policy” should start here — separate from the high-principled phrases of NGOs or press releases, where does our money go and how does it compare internationally? Pages 104-106 are simple mind-maps with the recipients of Canadian aid listed — a network diagram with Canada as the hub and our recipients as the spokes — for 1960, 1985, and 2010. Forty-five years in two increments, and the diagrams are powerful. They show, almost unequivocally, just how dispersed Canadian aid policy has become. The diagrams increase in complexity and density, leaving 2010 looking very much like “everyone who asks”. I almost wish that similar diagrams had been developed back in the early 2000s for Minister Susan Whelan when she was arguing for country concentration, as they are pretty powerful (yet simple) representations of data. Pages 108-109 go further with the data, looking at the degree of similarity in lists of donor recipients between Canada and 28 other donors using Jaccard coefficients.
The goal of these two parts are simple — first, Swiss demonstrates that our aid recipient network is not static, it has changed dramatically between 1965 and 2010, and not through normal rhetoric, but with the data clearly represented. Second, Canada’s list of aid recipients is compared with the individual lists of aid recipients for 28 other countries to see if we are more like one country than another in our approach.
It used to be said, fairly commonly and perhaps even by many of the other authors in the book, that Canada has moved away from the donor darlings like the “altruistic/humanitarian” Nordics and are now like “crass, self-interested Americans”. Yet, the data upsets that apple cart fairly strongly, showing that a large number of the countries are basically starting to have pretty high similarities between programs. One could even argue that the foundations are laid for donor harmonization, as we have very similar aid recipient lists. And one might argue that’s a good thing. Yet, as much as I love the data and analysis, the little niggling voices at the back of my brain are screaming “but, but, but…”.
One caveat I have is whether similarity in aid lists itself, assumed to tell us something, really tells us anything at all. If 50 countries needed help, and some 10 donors help maybe 25 each, ideally there would be 5 donors in every country. The lists might average out to only 12-13 out of 25 would be similar, leaving you a maximum coefficient of about .500 (if I follow the Jaccard approach correctly). In a perfect distribution, .500 would be the max any one country SHOULD have with another, so if it goes above that, it may be that two countries are doing similar things, but it also means some recipients might be receiving little support. Country concentration and similarities may equally be a sign of a problem more than a solution, so I’m worried about the extrapolations from a comparison that starts off being ambiguous to analyse. Not a detraction from Swiss’ analysis, but a slight challenge to the base meaning of correlation (other than that two countries are potentially similar in reach).
Secondly, I fear that comparing simple aid lists / indices is almost meaningless. Canada, for example, maintained $50-200K “Canada Funds” in some developing countries. Since the projects were “developmental” in nature, and the recipients were ODA-eligible, the countries make Canada’s aid lists. But no one at CIDA would consider a “Canada Fund” to be an aid program, and Canada isn’t the only one that did these small projects. Equally, most aid managers would distinguish between a “full development program” (say $30M+), an “aid program” (say $10M+), a project fund (say $1M-10M), and Canada Funds (say less than $1M). Most aid managers would also separate out humanitarian assistance and some hardcore types would also remove anything delivered through a multilateral channel. In essence, triaging the list into “true development programs” where Canada has gone “all in” for investing vs. a bunch of countries where it has spent some money, but not fully committed, vs. a much smaller list by weight where we threw some bilateral money at it (potentially for political relations reasons — I’ll come back to this later when discussing diaspora levels) rather than true development engagement.
Here’s my fear with a simple coefficient of similarity using a basic index of recipients, with no triaging. If Canada gives $100M to China, $5M to India, and $1M to Haiti, and Denmark gives $1M to China, $90M to India, and $9M to Haiti, wouldn’t the coefficient look pretty positive since the two countries would both have the same recipient list? Yet the programs are extremely different in composition.
By a similar token that Swiss compensates for later, Canada has a very slow project approval engine compared to other donors. This temporal delay could mean for example that you shouldn’t compare Denmark 1965 to Canada 1965 but rather to Canada 1966 — Denmark might have be on the ground in ’65 but Canada might not have been disbursing until 1966. I think any recipient should be eliminated from the analysis of lists unless it has been a recipient for at least 2 years so that it would mean everyone was fully up and running for the comparison years.
I think there is some great deeper analyses to be done in that data, and despite my bias up front, I’d love to torture the data some more. Generally my goal would be to force the data lists into more homogenous categories to ensure that the measure of similarity is comparing lists of apples to lists of apples, and isn’t generating false coefficients that are being hidden by a laundry list of other fruit on the list. And, as a small foreshadowing to the next section, the triaging might also aid in extrapolating motives for aid patterns.
Having the most in common with the UK and the US in its early years as a donor, Canada then began to more closely parallel the like-minded group of donors, before again following a path where its aid allocation matched most closely the US and the UK. This preliminary analysis suggests that, rather than strike a maverick path of its own and allocate aid along a uniquely Canadian set of criteria, Canada has been a mimic over the years.
As mentioned above, many writers assume we’re more crass now and that we were more humanitarian previously, like the “like-minded” group. Except no evidence is ever presented to show that the “like-minded” groups were in fact humanitarian-minded other than their press releases and policy statements, nor that the Americans and Brits have actively been self-interested for other than policy statements. And this conclusion that Canada used to be similar to “like-minded” but are now similar to “US and Brits” is a HUGE marker for the next section. Remember that conclusion as I’ll come back to it.
After Swiss deals with concentration and similarities, pages 110-117 start to turn the attention towards attempting to detect motives for aid from aid disbursements. The basic premise is simple — if we give aid to more countries where we trade than where we don’t, perhaps trade is influencing our choice of aid recipients. I love the premise, and while it is notoriously difficult to track motives from spending (regardless of the entire discipline of political economy that I regularly find lacking), the approach is one that I think goes “just far enough, but not too far” in crunching the data.
As possible variables to determine what might drive aid, Swiss uses many of the same variables that CIDA’s internal analysis used back in the early 2000s. Tables upon tables of data were generated to help inform “country concentration” discussions. That’s not secret, it was all unclassified and often pulled from public sources. GDP per capita was a key factor, as it would be in any development policy discussion of “need” — everyone uses it. Some people substitute Purchasing Power Parity, trying to account for exchange rate differences but GDP per capita is standard development fare and a viable potential factor.
Distance between capital cities is the technical way of asking if countries give more to their neighbours and in their own backyards or if they have a truly “global” view of aid. For example, Australia in the early 2000s very clearly decided to concentrate on its backyard — Asia and Africa. They cut bilateral aid to most non-Asian countries; they reviewed the spending patterns of UN funds and programs and if they didn’t have at least 50% (as I recall) of their funding in Australian aid recipients (their own test of similarity), they cut all core funding to that fund and only gave project funding for countries on the Australian bilateral list; they replicated that hard-nosed approach to funding Australian NGOs, focusing on those whose geographic priorities aligned with the new aid list. With Foreign Affairs arguing constantly that Canada is both an “Americas” and a “Asian-Pacific” country, it’s a fair question if our aid program matches those claims.
Other obvious factors to include are trade; humanitarian need (although I would eliminate that factor, as per my list above — if you’re looking to see if self-interest influences development aid more than humanitarian principles, then including humanitarian assistance might mask the regular influence elsewhere); and total population in recipient countries (to account for the skewing by large countries like China, Brazil, India).
What may be truly innovative in approach to me though are the indicators included to capture good governance (the Polity IV score — it has always been difficult in the past to find a reliable and somewhat universal indicator other than sub-elements of the UNDP Human Development Index); intrastate conflict (an interesting element on its own, but perhaps a little weak as a potential proxy for security interests, could be adjusted with adding military deployments perhaps); and Total Recipient ODA volumes (as an potential indicator of similarity and/or mimicry).
I still have all the same concerns as early on — triaging the lists, establishing thresholds, eliminating humanitarian, etc. Or at least running the numbers multiple times on sub-datasets. But I also would love to see some other considerations.
In terms of comparisons with funding sources (country donors), I’d love to see a similar analysis that showed degree of similarity (like the first half of the paper) with multilateral organizations like UNDP, UNICEF, World Bank, and Regional Development Banks (perhaps we are following our multilateral masters) or analysis for various membership clubs. At the donor level, I’d like to see separate groupings for funders of UN Specialized Agencies (an assessed contribution that goes with UN membership), G7 members, Nordics, and major English-speaking members of DAC (US/Canada/UK/Ireland/Australia/NZ). On the recipient side, I would love a sub-dataset for Commonwealth, Francophonie, Small Island Developing States, HIPC, OAS, APEC, and the LLDCs. Maybe grouped coefficients for Asia, Africa, and the Americas, and maybe even a separate analysis for Central and Eastern Europe given that it started at DFAIT and moved over to CIDA because of the complete lack of infrastructure and capacity for DFAIT to manage projects rather than words and people. (As an aside, that’s not a slam against DFAIT, most departments with a heavy policy focus show the same schism when they try to manage project funding).
So, I’m in. Sure, I want “more More MORE!” when it comes to the sub-datasets, but that would be enough for Swiss to write a whole book (hmm, Swiss, if you’re not busy for the next 10 years, could you get on that please?), and way beyond the scope of this paper. I love the dataset that goes just far enough without torturing it beyond recognition, and I want more, but let’s see what he found.
Comparing Canada to the other donors reveals two interesting conclusions about the factors that contribute to Canadian provision of aid to a country. First, the only consistent factor over time appears to be a country’s level of economic development. The poorer a country, the more aid it will receive. This holds for both Canada and the majority of the donors in my sample. In this sense, Canada is following the pack and providing aid along the lines of helping those most in economic need. Likewise, Canada resembles other donors in terms of what appears not to matter: trade, democracy, disaster, and conflict — none of these factors are robustly associated with multiple donors’ aid allocation over time.
Excuse me for a moment while I take that paragraph, blow it up, and mail it to every academic, every NGO, the peer review secretariat at the OECD, and well, just about everyone I know who has argued about the “self-interested”, evil Canada that doesn’t do development for developmental or humanitarian reasons. I may even want to refer to it in EVERY OTHER SECTION I REVIEW, because Swiss just analysed his way to a ferocious “booyah” for most of the literature in the area.
No EVIDENCE of those other factors influencing aid? Even when you account for temporal lags and weighting factors? The ONLY factor that shows a correlation with our aid policy is developmental need? We didn’t skew it to our trade partners? We didn’t skew it to our political or geographic neighbours? And the evidence is testable, reviewable, open to challenge, and most importantly, understandable?
Now that’s the kind of analysis this public sector employee can embrace. Yeah, as I said above, I want to test that conclusion even further with smaller, more niggling datasets. I want to see if it holds if you throw in some other sub-groupings. But it’s pretty powerful at the macro level. Bomb-shell evidence and nicely done.
And yet, even I have three small doubts that some other factors may be hiding in the data, and I’m not sure they can be completely ruled out yet.
First, various influences may not be statistically significant enough to register in the dataset, but even I would not argue they are non-existent and never play a part. If trade influenced 1-2 countries only, I don’t know if it would trigger a high enough result on the macro index.
Second, taking a page from political economy circles, it isn’t always the overall spending that would show the influence but rather incremental spending. For example, if Canada had a new trade agreement with a South American country to whom we normally gave $10M a year in aid, and we boosted that in the years before or after by even $3M a year, that wouldn’t likely show up as a blip because our trade flows might not change much until 10 years after the trade deal. I’m not sure even reviewing incremental spending would capture that type of influence, but it worries me slightly that total aid flows to a country might mask micro-influences.
Third, there is a hidden variable that frequently has reared its head in country concentration discussions, which is a domestic political dimension. How many diaspora are there in Canada for a given recipient? Put differently, and somewhat bluntly, if Canada wants to cut aid to China, there are a heck of a lot of Chinese-Canadian ties (some of them well-represented in Parliament) who are going to complain long and loud about those cuts. So, while a given Minister or Prime Minister might want to concentrate, there is a political cost-benefit analysis to be done — is cutting a small amount of money to a country, perhaps an amount that is a rounding error on the overall aid budget, worth the flak that the PM is going to receive from diaspora in the country? It’s not pretty, but it is an operational reality when you’re trying to build support for your aid policy. It is likely a micro-influence, and I think it would only potentially register in a triaged list for small programs (under $10M), but it IS a consideration at the political level.
Page 116 and page 117 are where Swiss loses me. After the huge “booyah” above, I was pumped reading the final conclusions. But he sticks with the same paradigm that is prevalent in the literature and while his data isn’t necessarily enough to kill the rest of the literature completely, I have a problem making it fit.
Here’s my issue based on the original paradigm, re-worded to fit my interpretation:
- Canada used to be like the “like-minded group”;
- The like-minded group is believed to be more humanitarian;
- Canada is now like US and the UK;
- The US and the UK are believed to be more self-interested.
ERGO –> Canada used to be more humanitarian and is now more self-interested
But the evidence shows that conclusion to be false. Canada’s motives have not changed — we didn’t become less humanitarian and more self-interested. Which means something in the premises of #1 to #4 must be wrong. Since Swiss proves that #1 and 3 are right, then #2 and/or #4 must be wrong.
At first blush, I want to shout another giant “booyah”. His analysis even proves it — “none of these factors are robustly associated with multiple donors’ aid allocation over time.”
In a perfect world, these two conclusions together should:
- be delivered accompanied by a loud thundercrack that shakes the heavens;
- cause a few academics and NGO heads to quietly resign, citing personal reasons;
- send some journalists rushing to old-style telephone booths to file stories; and,
- cause a few Canadian chests to swell that we’re doing it right.
I know none of that will happen, but it’s the academic equivalent of a stand-up triple in baseball.
Could it mean too that perhaps the differences have nothing to do with self-interest or humanitarian values (philosophy of approach) and are more likely simply a different way of doing aid (disbursement approach)? Or even just that we’ve got a different list of recipients, based solely on need and reflecting more of a division of labour?
I know, I know, that goes way beyond Swiss’ evidence, and I’m reaching past the point where I normally want analyses to go. But it strikes me that the article retreats to the supposed safety of the existing paradigm, after drastically weakening the foundations of that same paradigm. I’m looking forward to seeing if any of the other sections completely demolish the remaining supports.
In short, I loved the article (shhh, don’t tell Liam I said that). And I want more.