Can temporary subsidies and agricultural extension build sustainable adoption?

|

This page in:

A fair number of governments in developing countries support agricultural subsidy programs.   One of the arguments for these subsidies is that there is some kind of market failure (information is often cited) that the subsidy is meant to overcome.    So, that means when the subsidy is removed (which is the politically hard part), we should see adoption sustained.    There isn’t much clear evidence on this, but two recent papers provide some insight.
  
This week I’ll talk about one of these papers, a fascinating “reverse randomized control trial,” orchestrated by Ram Fishman, Stephen Smith, Vida Bobic and Munshi Sulaiman.  Fishman and co. are working in Uganda, with the NGO BRAC, and looking at an agricultural program where the funding has dried up.  Instead of just shutting it down, BRAC and Fishman and co. randomly phase out certain components.    This lets us see what happens when these programs close down and, in particular, how sustainable the benefits are. 
 
Before getting to the end of the program let’s take a look at what it does.   There are two main components.    The first focuses on knowledge.   Here we have model farmers (MFs) who set up a demonstration plot and get a bunch of improved (high yielding variety) seeds to show others how they work.   The MFs get seeds for themselves, but also a small amount to give to each of the farmers they train.    The second component focuses on building a market channel.   Here we have community agricultural promoters (CAPs) who set up as input suppliers, selling the seeds which (during the program) they can buy from BRAC at a roughly 20 percent discount (and free delivery).  This is a program aimed at female farmers and both the CAPs and MFs are selected from among the pool of female farmers in the village (with the MFs also having a minimum land requirement so they could set up the demonstration plot).
 
Now comes the experiment.   As BRAC is winding down, Fishman and co. have them randomize villages into 3 versions of this rollback: a) rolling back support for the CAPs, b) rolling back support for the MFs, and c) continuing the program (the control, in this reverse case).  
 
At this point, you are probably saying to yourself:  When is nothing as good as something?   Answer: when your program has impacts that last beyond the program OR when it never had an impact in the first place.  So Fishman and co. have to convince us that the program actually did something in the first place.   They bring two arguments to bear on this. 
 
The first is an RCT of the program from elsewhere in Uganda (the southwest versus the east for this reverse experiment).  If you are going to get stuck on external validity, skip this paragraph.    If not, the RCT shows positive and significant impacts on a range of outcomes: number of acres cultivated, number of crops, purchase of improved seeds, purchase of improved seeds from BRAC, total production, whether the farmer received revenue from crop sales, and the amount of crop sales revenues.   The impact on improved seed use comes in at around 7 percentage points. 
 
Now, the version of the program that is the subject of this reverse experiment was not initially randomly assigned.    So, for the second part of the evidence of impact, Fishman and co. turn to propensity score matching.    There is a good discussion of potential selection issues in the paper and, in short, Fishman and co. do the best they can (and they also give OLS estimates for comparison). Again, there is a meaningful and significant boost in the probability of using improved seeds. There is some indication these seeds were purchased (large but not significant coefficients for the PSM estimates, and a large and significant effect in the OLS).  There were also boosts to crop diversification and fertilizer use.  Farmers also changed their techniques, becoming more likely to do crop rotation and line sowing.  
 
OK, so by both metrics the program seemed to have worked – in particular it got farmers using the improved, high yielding seeds.   So what happens when the program is pulled back?   Figure II from the paper gives us a summary when they pool both the CAP and MF rollback.
 
From Fishman, et. al.

 
What we are looking for here is no impact.  And indeed, for improved seed, and the purchase of improved seeds, we can see that the effects of the phaseout had no impact.   And the 95 percent confidence of these estimates do not overlap with the OLS or PSM estimates of impact (the red and gray boxes – note that these are negative to show us how far things would have to fall to undo the program).   For fertilizer use, things are less clear – here there is overlap, so Fishman and co can’t rule that these were not sustained.   For changes in practices, though, things again look good – these estimates are zero-ish and the confidence interval does not overlap with the program impact.  
 
Fishman and co. disaggregate this by the CAP rollback versus the MF rollback, and they also have 1 season and 3 season follow up data points.   In terms of the CAP versus the MF, the effects on improved seed use are negative (but still very small) for the CAP, but positive (and small and not statistically different from the CAP) for the MF rollback.   The results for quantities are fairly similar (although there is an interesting significant and positive effect for MF rollback after three seasons).     
 
What is particularly intriguing is the source of seeds over time.   After one season, all of the folks who were phased out of some aspect of the program are significantly (at 10 percent) less likely to buy seeds from the CAP and model farmer, but more likely (again at 10 percent, and small) to buy from other BRAC sources.   After three seasons:  folks are significantly more likely to buy from market sources (10 percent significance) or from BRAC and less likely to buy from the CAP or model farmer.   So, after a while, the market seems to be stepping up. 

So these results are consistent with the idea that a temporary subsidy (plus some extension) overcomes an information failure.  And, when the subsidy is pulled back, a fair number of folks will seek out the market to continue the practice.    One critical thing to note:  folks do continue to buy from BRAC, and BRAC took pains to make sure their seeds were of high quality.  So this is a market of a particular type.     In two weeks we’ll take a look at a paper which looks at fertilizer…with some surprising results.  
 
Countries

Authors

Markus Goldstein

Lead Economist, Africa Gender Innovation Lab and Chief Economists Office

Asif Dowla
July 13, 2017

I am intrigued by the first line. Developed countries continue to subsidize their agriculture: Farm subsidy in the USA and Common Agricultural Policy in EU. The Queen of England receives farm subsidy and there is a map of Manhattan that shows people who receive farm subsidy.
BRAC reputation is an important factor here. I don't think we talk enough about the role played by the implementation agencies and how this should be taken into account in extrapolating external validity of RCTs

Toba Omotilewa
August 12, 2017

Interesting findings from Fishman & co. I am currently working on a randomized study in Uganda (too) where I find a one-time subsidy on postharvest technology increases market participation.
Dr. Goldstein, it has been over two weeks and I am still waiting to see the second paper you promised to discuss. Please share the second article and your thoughts too.
Thanks!