Bottom up thinking points the way for REDD

More good stuff from CIFOR, this time a survey of 23 different pilot REDD+ projects from around the tropics. The variety of approaches on show just goes to show, again, the benefits of Bill Easterly’s ‘seekers’ over ‘planners’. At both national and international levels I fear there is not enough flexibility in how government officials expect REDD+ to be delivered. And while there is plenty of justified scepticism about the prospects for REDD+ itself, I reckon a lot of that would go away if the price for carbon climbed up to the $20-30 per tonne of carbon dioxide that many experts think is required to push the global economy into making the necessary changes to head off catastrophic climate change. Less faffing around in negotiations and a clearer regulatory landscape would no doubt help too.

Emerging from what?

One of the most illuminating insights in Ben Ramalingam’s Aid on the Edge of Chaos (see last week’s review) was on how well suited complexity science is to tackling issues of systemic risk in the global financial system (e.g. too much inter-connectedness amongst banks and other financial institutions). The reason being that there was a wealth of data, millions upon millions of individual transactions, just sitting there waiting to be analysed if only someone could bring the right toolset (and a powerful enough computer). Complexity science can take all that apparent randomness and help us tease out significant emergent patterns and behaviour.

I thought this was particularly illuminating because it perfectly illustrates one of the major challenges of bringing complexity science techniques to bear on development problems: for the most part we do not already have the data, and going out and collecting it is very expensive. Different analytical approaches no doubt differ in their data requirements, but I suspect that in many cases that chaos nerds have an even bigger problem in this respect than randomistas. In short without the huge morass of data there is too little random feedstock from which patterns can emerge.

If we combine that problem with one of the main challenges to RCT’s global domination – limited external validity when context is everything* – I am worried that complexity thinking may sometimes me the equivalent of the proverbial sledgehammer used to crack a nut. It may be that the nut is so hard to crack that nothing short of a sledgehammer will suffice to do the job, but the reality is that we cannot go round deploying chaos science sledgehammers everywhere, not least because I doubt there are enough capable chaosistas.

But there is another emergent pattern out there, of bloggers sounding really stupid when they write about things they don’t understand. So now maybe Ben and co can tell me how badly I am wrong …

* In chaotic systems this is represented in the extreme sensitivity to initial conditions, hence the joke about the butterfly flapping its wings in the rainforest triggering a thunderstorm on the other side of the world.

The power law and why Aid is SNAFU

According to Ben Ramalingam’s new book, Aid on the Edge of Chaos, emergent characteristics of complex systems (a category that covers most targets of international development aid) often follow a power law in which most results are clustered together but which are offset by a long tail, e.g. lots of mostly poor people and a few incredibly, stinking rich folk. This tail is fatter, and contains much more extreme elements than you would expect from a Normal distribution, so that ignoring it as a few outliers can be incredibly dangerous. Conversely, such as in the case of earthquakes, the tail commands all the attention over swiftly forgotten smaller events. Maybe the same could be said of success rates in Aid projects? Boosters focus on the Green Revolution and the eradication of small-pox ‘fat tail’, sceptics obsess about the vast majority of Aid projects and spending which appears to achieve very little. Both are right, and both are wrong, and both could probably do with reading Ramalingam’s book.

I hope the title does not put people off the book. It suggests to me an anarchic office environment with harassed over-worked managers, when in fact the chaos that results from too many aid projects is rather more slow moving, if no less SNAFU. Ramalingam does have a good justification for his choice of title, but you will have to get to the final few paragraphs to understand it.

The book comes in three parts, and is a mixed read. The first part is a well-written indictment of the many failures and hubris of the international aid system. It treads familiar ground for anyone who has read Ferguson, Easterly and others. I have yet to tire of reading such critiques partly, perhaps, because they fit well with my prior beliefs, but also because such tales of failure are often instructive, useful to remind oneself what not to do!

Some of the targets may be easy, but all the more deserving of criticism. Occasionally it over-reaches, e.g. in condemning the reliance of orthodox economics on the idealised Homo economicus without acknowledging the many useful findings it has produced. Perhaps better editing would have helped, since such lapses are an easy mistake to make when a polemicist’s blood is up, but they do not detract significantly from the argument.

The second part introduces the reader to the power law and other elements of complexity science, and struggles manfully against the reader’s presumed lack of familiarity with this difficult subject. Part of the problem it faces, I think, is that complexity science is as yet a very young discipline. Theory and understanding are still very much in development, hence appropriate analogies and clear explanations are not well established. This presents a barrier to comprehension of a school of thought that is conceptually difficult to grasp.

Given this challenge it is ironic that the book suffers from too much space given over to this part, presumably in some attempt to keep the book balanced between the three parts. Such space has to be used up somehow: Ramalingam has chosen to do so through regular diversions into the history of complexity science. It is laudable that Ramalingam wants to tip his hat to the giants in his field, but such diversions are not fully contextualised (since this is not a history of the development of complexity sciences) and thus not especially illuminating. They are also somewhat repetitive, and distracting from the main argument.

One criticism of the use of complexity theory in development is that it is good for telling us after the fact what went wrong, hence the long list of shame in part one. But it often seems less good at telling us what we should do instead. This is slightly unfair because for the biggest aid agencies with millions of dollars to spend, investing $50,000 (say) in a complexity assessment could save a lot of money from being wasted on a doomed project. (If only aid agency incentives worked that way …) Ramalingam makes this point, but then in part three goes further with a series of examples of where complexity thinking has been used positively to underpin some highly successful development programmes.

How you respond to these examples may depend upon your background. I loved the on-the-ground examples such as the Subak system for irrigation management in Bali, the ecosystem-based approach to tackling endemic malaria in parts of Kenya, and using positive deviance to find ways to reduce child malnutrition in Vietnam, but others left me wondering “So what?” Conversely the talk of chaotic patterns in epidemiology may leave anyone with a decent grounding in ecological population dynamics thinking “Well, duh!” Some of these examples could thus perhaps have done with a better connection back to the critiques of part one to highlight why the use of complexity science is important.

Pressures of time meant that I took much longer to finish reading the book than ideal, and it maybe that a more focused reading would have been easier on the brain, but by the end I found the book frustrating. The overall aims and structure of the book are clear, but in the detail Ramalingam often appears to lose sight of where he is going with too many digressions in what may be an attempt to humanise an extremely abstruse subject. Ultimately I think the book needed more on aid and aid projects, and how they can be improved by the introduction of complexity science, and less on complexity science itself and its practitioners.

All of which is a pity because the ideas contained within the book are incredibly important. Indeed, despite those flaws, I have little hesitation in recommending it to anyone working in development or in developing countries. Unfortunately I suspect the very deliberate (and quite correct) decision not to offer any panaceas will limit the book’s impact on how most aid agencies operate. The world will be poorer as a result.

Theorising pig flight

“Everyone these days (funders, bosses etc) seems to be demanding a Theory of Change (ToC), although when challenged, many have only the haziest notion of what they mean by it. It’s a great opportunity, but also a risk, if ToCs become so debased that they are no more than logframes on steroids.”

That was Duncan Green writing a couple of months back. I totally dig the turn of phrase, but (luckily!) have so far escaped any such experiences of being enslaved to a donor’s preconception of what a ToC should look like. On the other hand I do find logframes (or ‘lockframes’ in the memorable corruption) more than a bit tiresome such that I might be inclined to reverse the comparison, and describe logframes as theories of change on steroids.

If you are worried that you might fall into that class of people who “have only the haziest notion” of what a ToC is then you can go read Duncan’s blog post (plus an excellent selection of comments) or Google for a whole bunch of other informative web sites. But over in this little corner of cyberspace I quite like my ignorance of the more formal definitions. Not that the above sources are not useful, quite the opposite, but I prefer the ‘pornography test’, which is to say I believe I have a pretty good intuitive idea of what a theory of change looks like, and I reckon I know one when I see one.

To me a ToC is primarily just a reasoned explanation of how what one proposes to do will actually deliver the impact you expect to achieve. It can be summarised in nicely boxed flow diagrams and the like, but for me the real test of a ToC is that it can stand up to reasoned, sceptical argument.

Where, I believe, so many conservation and development projects go wrong in their design, is not in their use or lack of use any particular framework, but in just plain sloppy thinking and lack of self-criticism. Part of the problem in development, it seems to me, is that we are too often too nice to each other, and not inclined to criticise (constructively!). This challenge can be exacerbated when discussions cross cultural boundaries: local ownership and deference to regional social norms are important, but should not trump having a workable plan in the first place.

Conversely we are also often too formal. A single written proposal, however, well constructed is never as satisfying as being able to discuss and probe people’s plans in person. And who reads those tediously long project documents any way? The result: too many projects approved primarily on the basis of the executive summary, without real testing of assumptions. And that’s when those flashy graphics really come into their own: great for communicating the central thrust of an idea, useless at exposing logical fallacies.

MJ’s theory of porcine aerodynamics: flashy graphics may not stand up to serious scrutiny.

So I like donors who are prepared to get into a real conversation with their grantees, to get to know them and their plans a bit better. Such relationships can more easily support adaptive management, which in turn allows you to be a bit more relaxed about any flaws in the original proposal, because now you have a framework in which to manage deviations from the plan.

And how do you succeed with all those awkward discussions in which design flaws are impertinently probed? As one of the commenters on Duncan’s post put it: “the first order of business is to build TRUST.”

Map-a-doodle-do

“Not only does such a dodgy map give a false sense of confidence, it very explicitly prioritises certain knowledge types. In my field the typical example is expert engineers rather than communities with lived experiences.”

That was Shaz Jameson commenting on my piece a couple of weeks ago that worried about the perils inherent in mapping conservation problems rather than conservation solutions. (And before you say “But conservationists are always mapping their protected areas. Aren’t they solutions?” ask yourself whether they really are solutions, or just wannabe solutions?)

I think Shaz is bang on about the power and knowledge inequities that are inescapable in such a situation.

“A few, biophysical variables are singled out by engineering ‘experts’ and mapped with pretty colours. The map is spat out and then somewhat paraded as legitimate.
Harsh, I know. It’s easy to say that ‘oh it depends on what you want the map for’, but as you say, it very easily slips into false confidence.”

The criticism can seem quite harsh, and I equivocated for a while about blogging my thoughts, but Shaz’s comment has helped reinforce my convictions. I am not saying, and I suspect Shaz is not saying either, that we should just stop mapping the challenges we are faced. Mapping is such a crucial communication tool, that it would be absolutely nuts to do without it. But when it is so critical, and you have the capability to create a map while someone else does not, even if they have more at stake, it does imply all sorts of ethical obligations upon the would-be mappers.

Be careful where you point that map!

Once a map has been created, it can, indeed too often will be used for purposes other than for which it was intended. On many levels that is absolutely fine: it is the great thing about knowledge that what may be a trivial output for one person may be a critical piece of data for another whose purposes were never considered by the original knowledge creator. But, as anyone who has ever been misquoted knows, knowledge users can be casual in their use of their sources.

At least in any scientific endeavour most numeric results are these days reported with a standard error or confidence limits to indicate the degree of precision with which the result should be interpreted. But maps rarely come with such cautionary notes, and even where a diligent cartographer has noted the scale of data applicability, few people will understand the significance of such a statement, and probably even fewer pay it any attention.

Take, for example, that map of carbon and biodiversity values for Tanzania that I complained about previously. At first sight it looks pretty detailed, but closer inspection will reveal the biodiversity axis is in fact reported in huge hexagons that are several hundred km across, but these are somewhat obscured under the much more detailed biomass carbon data (with a scale of 5km per pixel). Worse, if you think for a bit about what the biodiversity index itself represents you will rapidly conclude that in itself it can only be an extremely approximate measure.

Again, as with the concerns over power relations, this does not necessarily mean you should not go about creating the map. But you should certainly try to follow best cartographic practice, and assume that people will not read the accompanying small print when they decide to utilise your work. Thus, as was done with that map of Tanzania, it is not unreasonable to overlay two sets of data of quite different scales, but you definitely should beware of users who might assume that the spatial accuracy on both layers is equally high. One possible solution: consider blurring boundaries so that there are no sharp boundaries for decision makers to latch on to with undeserved confidence.

(Note: it is clear from the text of the paper in which the Tanzania map was presented that the authors were well aware of various problems and limitations of data scale, although they do not explicitly discuss the different scaled data in their sample map, which is nonetheless presented as a sample tool for guiding REDD+ investments.)

I map therefore I conserve

Another criticism I have that merits further elucidation is over the motivation of the mappers themselves. Whether it is the implied “top down planning approach” (my earlier post) or the prioritisation of “certain knowledge types” (Shaz), I get worried about not just what is put in or left out, but why some maps are created at all (the prioritisation of certain activity types).

This criticism applies not so much to academic endeavours such as that which I have critiqued in this and my previous post. Instead it is directed at the conservation BINGOs and associated entities that, of late, have developed substantial units dedicated to mapping conservation problems and their own attempts to resolve them. Precisely because they are such great communication tools (and all successful BINGOs know the importance of good communications), I fear that the maps produced can give quite misleading impressions of competence and imply levels of understanding that do not exist on the ground.

This is because, quite simply, conservation is a human endeavour: responding to problems created by people with solutions that must, necessarily, be implemented by people. Biodiversity and other physical measures are primarily just indicators of how well we are doing. (Although not without other uses in guiding management decision making.)

What worries me is that faced with exceedingly difficult, sometimes almost intractable, and nearly always highly complex conservation problems, mapping becomes a displacement activity. The motivations may be honourable: what better way to start to grapple with such complexities than by mapping them? However, even with the necessary skilled staff, producing such maps is not easy: data availability and quality can pose significant challenges. Significant that is, but not insuperable to a well-resourced institution like a good BINGO. So effort is piled into the mapping initiative, and everybody feels good because they can see progress is being made: by the end of the project they’ll have some top-notch PowerPoint slides. If only the action on the ground were half as good …

Alas dagger-in-the-back politics and brutal economic forces cannot be tamed by a GIS program.

The false confidence conferred by a dodgy map

Last Year Morten Jerven called into question the quality of statistics produced by (African) developing countries. In his musings on the political fallout from his publication, Professor Jerven summed up the situation as “governance by ignorance.” It also is clear that many stakeholders have a similar view but are reluctant to say so publicly for fear of the ‘anti-neo-colonialist’ backlash that Prof Jerven experienced.

Although I am slightly shocked that a statistic with as high a profile as GDP is so poorly computed, I really shouldn’t be surprised. Some of the government estimates for forest cover that the FAO collates each year in its annual Forest Resources Assessment are known to be extremely shaky; in some cases they are reportedly based on data years out of date, in others on little more than expert guesses. As with GDP, again many stakeholders are aware of this, and, in the case of donors, have money to throw at the problem.

Unfortunately this fits with a wider pattern in conservation: that we are getting better and better at identifying and measuring the biodiversity we are losing, but not much better at halting those losses. This is not to say that new knowledge is a bad thing; it is a vanishingly rare occurrence when an addition to our total body of knowledge does not increase the public good. But, as many researchers are all too aware, new knowledge can easily be misinterpreted or, worse, abused.

A paper last year from Gardner et al. (A framework for integrating biodiversity concerns into national REDD+ programmes) showed how biodiversity could be incorporated into REDD+ planning, and illustrated this with the map reproduced below that overlays biodiversity and carbon values for the case of Tanzania.

As I understand it, the map is a fair reflection of the current state of knowledge, and no-one should infer any particular agenda on the part of the paper authors. As a simplified guide to national decision making it looks eminently useful.

The trouble is, from what I gathered on a recent visit to the country, that this map may not be a good guide as to where are the best opportunities for effective REDD+ projects. The map highlights the Eastern Arc mountains as an area of “high opportunity (strong positive correlation in carbon and biodiversity values)” for REDD+ intervention. Most of these forests are, in theory, already protected in national parks and forest reserves, but which are threatened by encroachment and illegal resource extraction. A long developed strategy of Joint Forest Management (JFM) is intended to help resolve this, by giving local communities a share of forest revenue, except that this has been held up by the failure of the Tanzanian Government to agree a benefit sharing mechanism for JFM. I.e. potential REDD+ project developers would be advised to steer well clear of JFM in the Eastern Arc Mountains until the benefit sharing mechanism has been agreed and tested in practice.

Such policy issues cannot easily be represented on maps, and I would not expect an overlay to do so. However, theoretical exercises like this can be dangerous in how they may give policy makers and donors the illusion of agency: invest money in the sweet spots and best return on investment will be achieved. This omits the critical step that one needs credible potential solutions before investing money in actual projects. However, the reality of national planning in developing countries, often donor supported, and which this paper purports to assist, is that high level decisions may too easily be made without all the necessary information. Maps such as these are therefore potentially dangerous in implying a higher level of decision-directing knowledge than in fact exists.

(A second criticism is that such mapping exercises can sometimes almost imply a terra nullius – no-one’s land – attitude in which the wishes of existing inhabitants and forest users are irrelevant.)

In conclusion I would question the wisdom of a top down planning approach at all. Rather than prioritising problems I suggest we might be better off prioritising solutions. In which case the value of such maps are rather less than might be first supposed.

Not for the first time, and why we might foul it up again

Owen Barder is keeping a nice list of major public figures’ claims that “For the first time ever, we have a real opportunity to end extreme poverty within a generation.” in the words of World Bank president, Jim Kim, the latest to so pronounce. The list goes right back to Woodrow Wilson addressing the League of Nations in 1919. Obviously previous generations have not lived up to such lofty aspirations. Why not? And why should we be any different? I present some wild speculation …

Back in 1919, perhaps for the first time, the western powers could say they truly knew most of the world. The major features had been well mapped, and many distant peoples had been ‘civilized’ (aka colonised by racist imperialists). The sun never set on the British Empire and the industrial revolution had made some Americans fabulously wealthy. There were a lot of poor people in the world, but not so many, and Westerners had a surfeit of confidence as to their capacity to achieve great things. Moreover, in a world before the widespread existence of welfare states, it is possible they were not aiming that high.

What happened? Two major changes. Developing countries won their independence from the colonial yoke. This hugely increased their welfare in one important dimension (political freedom), but possibly impeded progress on technocratic goals such as raising average incomes due, in part, to the need to first concentrate on building the capacity of those new states. One signal success, nonetheless, does stand out: the drastic decline in infant and maternal mortality rates. So while economic development was stalling in many countries, populations were exploding. Suddenly it became a lot harder to eliminate poverty.

Fast forward to 2013 and we have passed an important inflexion point: now the number of desperately poor people in the world is declining in absolute terms, and not just in China. The zero goals some people are suggesting should follow the Millennium Development Goals when they expire in 2015 appear tantalisingly in reach.

So why might we fail again? What new issue might once again expose our hubris? I give you two words: climate change.

Aid project selection & implementation

Some great quotes in a new working paper proposing a different approach to M&E by Lant Prichett et al. My eye was particularly caught by these two from the conclusion.

“The reality of the project selection process, inside government organizations and between government organizations, tends to be an adversarial process of choosing among projects, which puts project advocates in the position of making much stronger claims for project benefits than can be supported, and being more specific than they would like to be.”

I’m relatively relaxed about the tendency to make over-ambitious claims of expected project impact since everyone does it, and is thus likely to fairly well factored into how projects are viewed. The problem of over-specificity in design is, I think, a bigger problem since it leads to significant wasted effort during the project proposal stage developing ridiculously over-detailed action plans and budgets. Most donors like to think they are flexible when it comes to plan and budget changes mid-grant, but the simple requirement to obtain approval is a deterrent to project managers and a source of risk: what if they do not approve the changes?

The issue of over-specified designs has other implications for implementation too:

“Organizations like the World Bank perpetually over-emphasize, over-reward, and over-fund ex ante project design over implementation. This is because in the standard model, implementation is just faithful execution of what has already been designed, whereby the thinking is done up front and the implementation is just legwork. However, de facto many successful project designs are discovered when project implementers are given the flexibility to learn, explore and experiment.”

As I wrote before: good strategies need good implementation. If the implication – that big donors like the World Bank already know this basic fact – is correct then it really makes me question the whole competitive grant awarding process that dominates NGO involvement in conservation and development. Donors could save everyone a lot of trouble by awarding grants on much shorter project outlines combined with a good track record of delivery (which needs to be much more robustly assessed). Good NGOs would be strongly incentivised to deliver good outcomes since otherwise they would lose their future funding. An entry level system would still allow new players to prove themselves, and also those fallen stars to re-establish themselves.

I will blog again tomorrow on the core proposal of the paper when I’ve had longer to digest it.

Hat tip: the Blattman

The importance of choosing the right tool for the job

Duh! I mean obvious or what? Except, as is well known, people equipped with only hammers often see too many problems as solvable purely by use of a hammer. International donor engagement with Community-Based Natural Resource Management (CBNRM) has long struck me a case in point.

The international donors come armed with their hammer, which is money to support the government in the recipient country. This is primarily what big international donors do. They work often through diplomatic or pseudo-diplomatic channels. Aid should be channelled to the host government; anything else could be construed as meddling in someone else’s country. This is a pretty good hammer for things that the host country government is already good at, e.g. building new schools all over the country: the more money you pour in, the more new schools get built.

CBNRM has been just one of the sexiest concepts in conservation for about 30 years now. Even Payments for Ecosystem Services (PES), including REDD, hasn’t fully displaced it, just been added to the mix, since rich folk much prefer to pay those poster children of poverty porn – poor rural communities of farmers – for protecting their local environment than a bunch of shapeless bureaucrats. So plenty of donor investment in conservation in recent decades has focused on CBNRM. Except that, as I pointed out previously, the same government officials who are charged with enforcing local environmental laws do not really make the most sensible agents of change when it comes to facilitating the development of CBNRM projects.

Now along comes a World Bank report critiquing participatory approaches to development. It includes a substantial section on CBNRM, which is certainly pleasing to see. The main conclusion is that the majority of CBNRM projects do not appear very successful; depending on how they interact with existing policies and laws they may even be counter-productive, helping the rich at the expense of the poor. Now I am familiar to some extent with some of the CBNRM projects they are talking about (or citing others talking about), and I noticed one thing in common: they were all classic international donor funded efforts, working through local government.

So yeah, I’m not particularly surprised by Messrs Mansuri and Rao’s findings, but I’m not convinced that it greatly undermines the theory of CBNRM so much as the practice in the context of international development aid. (Which one presumes was very much Mansuri and Rao’s starting point, given their employer, and is supported by their associated blog post. All of which is not to say the report is not worth a read.) Major donors, of course, employ some very bright people, some of whom have come to realise the same thing. Alas, while they would choose to fund us – a small NGO – if they could, the only thing their employer ever gives them is a metaphorical hammer.

Public v private, principles v practice

Over at Duncan Green’s Poverty to Power blog there’s a bit of a ding-dong (parts 1, 2 and 3) between Justin Sandefur and Kevin Watkins over whether donor money for education in developing countries should be invested in state or private schools. Both have data and research findings to support their positions, and both make the point that a bit of both should have a place in any sensible education donor’s portfolio. So far so good, so why the all the fuss and strident opinions? One doesn’t have to be a psychoanalyst to suspect that attitudes to private education in our own countries – a fraught subject if ever there was one – are spilling over into the international development sphere.

If that is the case then I should at least declare my own loyalties; I was lucky enough to go to a private school, and very possibly would not be working in conservation and development today if I had not had that chance. But except for the economists and the religiously inspired, development is a decidedly left-of-centre vocation, and so the balance of commenters and voters in Duncan’s poll appear to be against donor funding for private schools.

My only experience of education systems is as a consumer and employer of other consumers, but you do not need any special insights to agree with the primary assertion made by both that there is a big problem of education quality in many developing countries, and a significant issue of access in at least some countries, e.g Pakistan. The essential quandary is how to practically improve the quality of education without compromising on principles of access and fairness. Private schools by definition cost people money and so are likely to be less accessible to poorer people.

But there are lots of essential public goods for which we are accustomed to paying for. Many, e.g. power, water and telecoms, are tightly regulated as a result, and rightly so. They do not, however, come for free. Yet access to information on market prices via a mobile telephone might potentially be more useful and important to a poor farmer’s children than a questionable education would be. Food is an even more fundamental requirement for life than education, and yet just about everybody has to pay for what they eat.

So just because education is a vitally important public good, I do not buy the argument that it needs to be provided directly by the state. As with, say, generating electricity, you may find that private operators are able to do so both better and more efficiently. On the other hand, given the situation in most countries that the state is the dominant provider of education, and that this generally embodies principles of universal access (at least in theory), I am quite open to the argument that funds for education might best be spent improving that system …

… if such funding can actually lead to significant and cost-effective improvements. Lots of newly-built but empty classrooms (paid for by donors) coupled with miserable educational achievement in rural areas around here even when there are teachers present suggests there can be a flaw in this assumption. So if better results can be achieved by supporting low cost private schools then I am all for them. The same ‘if’, of course, applies, and outside of American tea partiers I imagine there are few people who would support private schooling as a matter of principle; the evidence base if their favour needs to be there. (In the absence of sufficient evidence either way funding some pilot projects would be reasonable.)

Ultimately, in my view, it is better to educate some people well, rather than everyone badly. For all the problems with elites and elitism in developing countries, without an elite you do not have anyone to run your country. (Something, it seems, South Sudan is struggling with.) The principle of universal access only becomes a concern when the thing being accessed is actually worth something, or at least when we are able to fix the problems so that it is worth something. As stated many times before donor funds are sometimes quite good at delivering concrete outputs (more classrooms) but much less frequently successful at delivering less tangible outcomes (better educational achievement).

Is opposition within the development sector to private schooling based upon prejudices founded in developed country politics or actual realities in developing countries? I have no expertise to say one way or the other, but I do find similar biases in tropical conservation, in which many so-called experts insist on the gold standard of protected areas despite plentiful evidence that they often do not work very well on the ground. (This is not the same as saying they do not work at all.) As with education, rich country principles seem to occupy the moral high ground over practical solutions that might actually suit poor countries better. That is a shame for both the people affected and the wildlife that might otherwise be conserved.