Response from Banerjee and Duflo

Abhijit and Esther respond in the comments section of yesterday’s South Sudan post. Lest you miss the response, an excellent excerpt:

Our recommendations were based on a simple idea, very much rooted in politics: An effectively implemented redistributive policy is a very good way to give a new state a clearer identity in the minds of the voters. This can create ownership and start a virtuous cycle where the majority has a stake in fighting against the take over the state by one group. It also creates a basis for developing state capacity; focusing on doing one or two things well is the best way to give the state’s agents credibility and build their skills.

Read the full comment or, as they suggest, read their book for the full argument why, as Abhijit writes, the primacy of politics is way over-sold. I am intrigued.

Yes, I confess: I have not yet read the book (it’s on my Kindle awaiting my June vacation), and I will reconsider and (who knows?) even recant after the reading. I hope they are right, but am quite cautious.

There were many other excellent comments on the post. One good point made more than once: the government of South Sudan knows better than anyone how to corral warlords and (no surprise) are way ahead of me. Another good point, complementary to Banerjee and Duflo, is that redistributive spending may be the best way to spend oil dollars (and the only truly legitimate one).

Fair enough, but I’m curious to see what the evidence on redistribution programs  (including randomized trials) has to say about state attitudes and national identity. My impression so far: the political impacts of cash grant and community development programs have been meager. Here’s one new paper, on Sierra Leone, that makes me pessimistic. I have two experiments–from Liberia and northern Uganda–with eerily similar early results.

Now, a second disclaimer: I have never been to South Sudan. Think of this as commentary Palin-style: I’ve gazed across the border (in this case, from Uganda), and them’s all the qualifications I need. And you should take me with just as much seriousness.

Even so, it may help to explain my vantage point: several years in post-conflict African countries watching justice and security sacrificed for premature social programs, a growing worry that large scale social programs weaken rather than strengthen a fragile state, and a doleful skepticism that national identity is fostered through redistributive aid.

Lant Pritchett and Michael Woolcock make the argument much better than I do. (And Lant, I understand, is spending some serious time in South Sudan these days.)

In the meantime, the Banerjee Duflo book is here. For those who read faster than I do, comments welcome.

12 thoughts on “Response from Banerjee and Duflo

  1. “I’m curious to see what the evidence on redistribution programs (including randomized trials) has to say about state attitudes and national identity.”

    Not exactly what you are asking for, but pretty close, can be found in the first round of results from Afghanistan’s CDD, the National Solidarity Programme. Main website here:

    This is, I believe, the best CDD evaluation that’s been done and the one that people point to in arguing for CDD in post-conflict (or ongoing conflict) countries.

    Overall results on the relevant attitudes are mixed. From the interim report:

    “NSP results in an improvement in how male villagers perceive a wide array of government and nongovernment actors, including the president of Afghanistan, provincial governor, district governor, central government officials, government judges, NGO workers, and even western soldiers. Despite having a more favorable view of these individuals and institutions, NSP does not render male villagers more willing to accept the authority of the central government over local crimes, does not increase knowledge among men or women of parliamentary representatives, and has only a small, if any, impact on attitudes toward government taxation of village income-earners.”

  2. You can read my review of the book here I think they overdo the siren song of the technocrat – ‘just forget the politics and focus on incremental changes’, but I sympathise with their frustration with some of the excessive doom of the political economy people.

  3. I’m happy to see this conversation take the turn to what I think is the most significant part of the B&D view–really engaging on the theory of change rather than endlessly quibbling about the finer details of randomizations.

    I think most people who critique B&Ds preference for micro-incrementalism fail to truly understand how much they believe it. The suggestion is that they aim for big policy changes to embrace the micro. But my read of them, from writing and from conversation, is that they reject even this. They really do believe that the most fruitful way forward to is to ignore the politics entirely and get this principal and that medical facility manager to do something which is entirely in their power, slightly differently. The improvements from such changes aggregate and multiply much faster and more completely than conventional wisdom accepts.

    I’m persuaded but not convinced and hope that more of the discussion focuses on this issue than on the RCTs for a while.

    I have a review of Poor Economics to this effect coming in Stanford Social Innovation Review, and an interview of Abhijit and Esther that covers some of this ground will be published on Philanthropy Action shortly.

  4. I seem to recall evidence from Brazil of positive effects of Bolsa Família on support for the ruling party vote – which could translate to support for the very idea of the state.

  5. Really? Don’t folk like (from memory) Levitsky argue that that kind of thing is actually just support for populism a la Perón or Chávez? Real support for the state is when people support the very institutions rather than just the party or the personality. And only support for the institutions gives you stability.

    Though from what I’ve seen Bolsa Familia is thought to have been a roaring success, quite possibly very fairly.

  6. The problem with Abhijit’s suggestion is its lack of pragmatism. If a “universal cash grant” were to be mandated how high could it ever be? At these levels is it enough to bootstrap anything of relevance?

  7. It seems to me that CCTs will translate in support for the ruling party only in the short term. This effect probably vanishes in the long run as opposition parties commit to these transfers and voters’ preferences shift because they got richer. I am not sure that someone has documented that but I clearly see this happening in many places in Latin America.

    Indeed you should note that populist rulers of Latin America have not historically relied on CCTs and other transfers and policies that have clear and transparent assignment rules. I believe they haven’t done so precisely because these policies do not create a long term clientelist connection between the ruling politician and its voters.

    Rather populist rulers in Latin America have historically relied in manipulating wages and prices as a way to transfer resources towards the majority of its voters by increasing minimum wages at very fast rates, by reducing utilities and food prices and by creating trade barriers. These policies are discretionary and create distortions and inflationary pressures. The distortions and inflationary pressures that these policies create basically mean that the effects of these policies will be reversed unless the populist ruler keeps reinforcing them. So there a combination with discretionary assignment and possibility of reversal that creates a long term clientelist connection between the populist ruler and its voters.

    Just to make my case empirically. Your hypothesis means that there will be a positive correlation between CCT popularity and populist practices. But what we see in Latin America is clearly the opposite: persistence of populist practices is actually higher in countries such as Argentina and Venezuela that don’t have widespread CCT programs as Brazil and Mexico do.

    While I’m not sure whether CCTs and other social policies that Banerjee and Duflo advocate will translate in more support to the newly created state of South Sudan, I do believe that making a connection of these policies with the ‘bad politics’ we often see in Latin America and in other places of the underdeveloped world is a big mistake!

  8. You can avoid extreme poverty and provide helpful insurance with small amounts of money. Probably oil revenues are enough for the government to pay for such a grant.

  9. I wouldn’t use evidence from Brazil as a guide as to whether such program would result in better political prospects for the ruling party in South Sudan.
    As a Brazilian, I’d say it’s a completely different political and economical landscape (maybe even culturally), leaving comparisons such as these really void in predicting power.

  10. I think Banerjee was talking about a one time grant. This would have to be quite large if it is to permanently help someone out of extreme poverty. Unless you are talking of recurring payments.

  11. I agree with you. If it is a one time grant, than it has to be quite large. I was thinking about recurring payments.

  12. Some numbers:

    The population of South Sudan is about 10 million. I estimate the mean annual income of about $300 per capita. To bootstrap substantially I estimate a one-time payment of at least one year’s income.

    Thus Banerjee’s scheme needs a corpus of about $3 billion at least if not more. That’s roughly 5% of their GDP and 25% of the Sudanese government’s yearly budgeted expenses.

    Seems clearly pie in the sky.