Discussion Paper
No. 2018-56 | July 17, 2018
Somdeep Chatterjee
Storage infrastructure and agricultural yield: evidence from a capital investment subsidy scheme

Abstract

In a developing economy, the availability of storage infrastructure is considered essential for two purposes; the reduction of post-harvest losses resulting in food shortage, and allowing for gains from inter-temporal trade due to potential arbitrage opportunities arising out of volatility in food grain prices. This paper provides empirical evidence on a lesser studied impact of storage infrastructure, viz, agricultural yield. The author exploits potentially exogenous variation generated by the intensity of access to a capital investment subsidy program for construction and renovation of rural godowns in India to identify causal effects of better storage on yield. He finds that the program led to an increase in rice yield by 0.3 tons per hectare, approximately a 20% increase compared to the baseline. A potential mediating channel for such an effect would be reduced storage costs facilitating better investments in productive inputs. As supportive evidence, the author finds that fertilizer consumption increased by 21% in response to the intervention.

JEL Classification:

Q12, Q18, O12, O13

Links

Cite As

[Please cite the corresponding journal article] Somdeep Chatterjee (2018). Storage infrastructure and agricultural yield: evidence from a capital investment subsidy scheme. Economics Discussion Papers, No 2018-56, Kiel Institute for the World Economy. http://www.economics-ejournal.org/economics/discussionpapers/2018-56


Comments and Questions



Anonymous - Referee Report 1
August 20, 2018 - 09:00
Reviewer comments and suggestions (i) Is the contribution of the paper potentially significant? Comments and suggestions: This manuscript partially significant as the data is older one (upto 2010), it’s of 8 years old data, therfore the scenario may be changed in recent years and the second observation is that, author has used simple regression analysis rather that improved Statistical and econometric tools for better results and impact assessment. The results obtained were interesting showing the effect of subsidy scheme on grains yield linked to storage infrastructure. Overall the study is adequate. (ii) Is the analysis correct?Comments and suggestions: The analysis is correct and adequate. The overall strength of the paper is, the results obtained were interesting and highlighting the significance of subsidy scheme linking to storage infrastructure. The falsification excersice found good and highlighting the key issues and finally robustness checks shows the concluding part of the manuscript. With respect to weaknesses, usage of data is older one and there are some methodological issues related to use of models.

Somdeep Chatterjee - Response to Reviewer Comments
August 20, 2018 - 09:51
At the outset, I would like to thank the reviewer for taking time out to review this paper and for providing feedback on the draft. Here are my point by point responses to the comments made by the reviewer. 1.a) The reviewer has expressed concerns that the data is dated and hence the scenario might have changed over recent times. I totally empathize with the view and agree with the reviewer's thoughts here but I would like to clarify that the data set used is one of the most comprehensive databases from ICRISAT and it provided me with data upto 2009-10. Also, since the program being studied is quite dated, I had chosen about 10 years of data before and after the policy to make the analysis robust. With data that is more recent the problem would be that it is far away from the policy year and the effects of the policy may have waned away by now. Also, several other changes may have happened in the interim, which would make the estimation spurious. The reviewer acknowledged that the study is overall adequate and results were interesting, I thank him/her for these comments. 1.b) The reviewer commented that the analysis is based on simple regression rather than more sophisticated tools. I do not necessarily agree with this observation because I do not use simple ordinary least squares (OLS) regressions but instead use a niche quasi-experimental design exploiting exogenous variation in intensity of the program and potential access. I use a triple-difference estimation design (differrence in diff-in-diffs) because simple OLS is likely to be biased because of endogeneity. The falsification exercise supports this specific estimation strategy. I think one of the contributions of this paper would be to introduce this novel way of estimating the effects of this program on grain production. However, acknowledging the concerns of the reviewer, if any specific suggestions are made about alternate statistical models, I will be happy to explore the potential of those. 2. The reviewer has commented that the analysis is correct and adequate, I am deeply obliged to him/her for these views. 3. The reviewer also acknowledged the contribution of the paper in terms of providing one of the first direct estimates of storage infrastructure on produce and has also appreciated the use of the falsification exercise and robustness checks. I thank him/her for these observations and encouraging words.

Anonymous - Invited Reader Comment
August 27, 2018 - 08:23
Subject: Invited reviewer comments on discussion paper no. 2018-56 I am thankful for the opportunity to review the paper entitled “Storage infrastructure and agricultural yield: evidence from a capital investment subsidy scheme” by Dr. Somdeep Chatterjee. I request you to find my assessment below. 1. The author has submitted a well-researched and a well-written piece on a very relevant topic. I heartily congratulate the author for his efforts. 2. The author aims to link improvements in agricultural yields and increases in fertilizer consumption to a capital investment subsidy scheme for the construction (or renovation) of rural godowns in India. This basic premise appears problematic because improvements in storage infrastructure alone cannot cause improvements in yields and the author does his best to disentangle the myriads of complexities. 3. The problem is glaring if one looks at the figure of the falsification exercise. The scheme was introduced in 2001-2002. Given the knowledge that capital intensive investments involve several processes and take some time to be operative (selection of area with proper utilities, taking a loan, buying land, making a scientific building plan, applying for subsidy, etc.), one would never expect yields to take-off from 2002 itself but after a lag of a few years. 4. It is plausible that the author has missed alternative explanations for the improvements in agricultural yields and increases in fertilizer consumption. The regression results give a vital clue – the impact is strongly seen for rice but not for wheat. 5. In 1989, the Indian Council of Agricultural Research (ICAR) launched a special project on creating domestic capacity in hybrid rice (HYV) production (MOA 2010). Between 1994 to 2009, 43 hybrids were released for commercial cultivation. Table 2 (MOA 2010, 24) shows that four major varieties of hybrid rice was released in 2000-01 that led to a major increase in seed production in 2000-02 (MOA 2010, 32). Since these seeds are costlier, they are more likely to be adopted in districts with better financing options. 6. Notably, in 2000-01 as well, the System of Rice Intensification (SRI) was also being adopted by Indian farmers (Sharma 2014). Both HYV and SRI together might explain why there was a jump in rice production post-2002 and also why the 95% confidence intervals are so large. 7. It is therefore suggested that the author look at the impact on other crops like maize, pulses, etc. to support his hypotheses. If it is available in the dataset, the author can choose to control for district-wise storage capacity and godown capacity utilization. These controls may narrow down the confidence intervals. 8. There are minor spelling errors that might not been picked up by the Latex editor. I could find words like dyanmics, prorgram, large.and, incorproate, and ie in the paper. 9. A tonne (metric ton) is not equal to a ton (US). An US ton is equal to 2,000 U.S. pounds, whereas a British tonne is equal to 1,000 kilograms that converts to 2,204.6 pounds. In this context, tonnes (and not tons) is the correct usage. References:MOA (2010) “Guidelines for seed production of hybrid rice”, Department of Agriculture and Cooperation, Ministry of Agriculture, Government of India. Accessed on 26-08-2018 from http://vikaspedia.in/agriculture/crop-production/package-of-practices/hybrid-rice-seed-production/view Rita Sharma (2014) “More rice from less water”, The Hindu. Accessed on 26-08-2018 from https://www.thehindu.com/opinion/op-ed/more-rice-from-less-water/article6183223.ece

Somdeep Chatterjee - Response to Invited Reader Comments
August 28, 2018 - 09:02 | Author's Homepage
Thank You so much for reading through my paper and providing with very relevant and useful feedback. Here are my point by point responses to your observations: 1. The author has submitted a well-researched and a well-written piece on a very relevant topic. I heartily congratulate the author for his efforts. Response: Thanks for your appreciation. Its always a pleasure to get words of encouragement from a fellow academic. 2. The author aims to link improvements in agricultural yields and increases in fertilizer consumption to a capital investment subsidy scheme for the construction (or renovation) of rural godowns in India. This basic premise appears problematic because improvements in storage infrastructure alone cannot cause improvements in yields and the author does his best to disentangle the myriads of complexities. Response: I totally agree with your comment. However, that is precisely why the estimates coming out of my empirical strategy should be treated as "Intent-to-treat" reduced form effects of the program and not estimates of elasticities of yield in response to access to godowns. 3. The problem is glaring if one looks at the figure of the falsification exercise. The scheme was introduced in 2001-2002. Given the knowledge that capital intensive investments involve several processes and take some time to be operative (selection of area with proper utilities, taking a loan, buying land, making a scientific building plan, applying for subsidy, etc.), one would never expect yields to take-off from 2002 itself but after a lag of a few years. Response: I do not necessarily agree that the falsification backfires, if that is what you imply. The falsification essentially suggests that pre-existing differences among the identified cross-sections are statistically indistinguishable from zero. If the announcement about the program came in 2001-02, even if it is a capital intensive scheme, to me it is unsurprising that productivity goes up soon after. This can be explained in terms of forward looking agents updating their expectations. Consider the conceptual setting where a farmer is at the beginning of the cropping season and comes to know about this program. Even if the "upgradation" of godowns is a capital intensive activity, the reason for counter-factual lower yields would be the fear of post-harvest losses! Now, the farmer is "insured" against the possibility of post-harvest losses because by the time the harvest is completed, it is not unlikely that storage capacities would have been upgraded. As a result, yields beginning 2002-03 are likely to go up! It should go up even further in 2003-04, as per the argument put forth by you, which is also what I find. Therefore, I do not think the falsification is actually concerning, I would rather think it is reassuring. 4. It is plausible that the author has missed alternative explanations for the improvements in agricultural yields and increases in fertilizer consumption. The regression results give a vital clue – the impact is strongly seen for rice but not for wheat. Response: True, but if there has to be any alternate explanation, it must be the case that something other than this happened to rice production AND fertilizer consumption in exactly the districts with more banks (as per my definition) in states with more go downs (as per the definition) and in 2001-02 onwards. If any of these 3 is not true, then what I am estimating cannot be the effect of that hypothetical something else. I do not know of any alternate policy that would have affected productivity and fertilizer use exactly along these three dimensions, which makes the triple difference framework that I use much more compelling. 5. In 1989, the Indian Council of Agricultural Research (ICAR) launched a special project on creating domestic capacity in hybrid rice (HYV) production (MOA 2010). Between 1994 to 2009, 43 hybrids were released for commercial cultivation. Table 2 (MOA 2010, 24) shows that four major varieties of hybrid rice was released in 2000-01 that led to a major increase in seed production in 2000-02 (MOA 2010, 32). Since these seeds are costlier, they are more likely to be adopted in districts with better financing options. Response: Thanks for pointing out this potential confounder. I will add a discussion on this in the next draft of the paper. However, even with this story, I do not necessarily think that the effects of GBY are engulfed entirely. Firstly, even if better financing options are correlated with HYV variety releases, for it to be an empirical issue, it has to be true that these seeds were adopted more in states with more GBY godowns! It is not obvious to me why this has to be true. If not, then the empirical strategy I use still holds good. Even if this were true, it is not obvious why it should affect fertilizer use. As you said, these seeds are costlier and hence the household budget constraint would be tightened further, that makes it less plausible to think of this affecting investment in fertilizers. My story on the contrary is that of a cost reduction. Post harvest losses coming down essentially makes the budget constraint favorable for the farmer and hence he can now invest in fertilizers etc, leading to an increase in yield. 6. Notably, in 2000-01 as well, the System of Rice Intensification (SRI) was also being adopted by Indian farmers (Sharma 2014). Both HYV and SRI together might explain why there was a jump in rice production post-2002 and also why the 95% confidence intervals are so large. Response: As above, SRI needed to have affected banked districts in more godown states differentially for it to be a confounder. It is again not obvious why it would affect these cross sections differentially rather than uniformly. 7. It is therefore suggested that the author look at the impact on other crops like maize, pulses, etc. to support his hypotheses. If it is available in the dataset, the author can choose to control for district-wise storage capacity and godown capacity utilization. These controls may narrow down the confidence intervals. Response: I did try with other crops like pulses, the effects survive but with clustered standard errors, they lose on some precision as standard errors blow up. I do not think district wise storage capacity at the point of the policy can be used as it might subsume the district fixed effects. Same goes for godown capacity utilization. Intertemporal storage capacity has actually been accounted for in terms of the 3rd difference using state GBY intensitiy. 8. There are minor spelling errors that might not been picked up by the Latex editor. I could find words like dyanmics, prorgram, large.and, incorporate, and ie in the paper. Response: Thanks a lot for this. I will run a thorough spellcheck on the next draft. 9. A tonne (metric ton) is not equal to a ton (US). An US ton is equal to 2,000 U.S. pounds, whereas a British tonne is equal to 1,000 kilograms that converts to 2,204.6 pounds. In this context, tonnes (and not tons) is the correct usage. Response: I will correct for this in the next draft. Thanks again for these inputs.

Anonymous - Referee Report 2
August 29, 2018 - 08:02
Report MS No. 2735Storage Infrastructure and Agricultural Yield: Evidence from a Capital Investment Subsidy Scheme (i) Is the contribution of the paper potentially significant? The paper is hypothetically significant in the sense that it gives the results that GBY program led to increase in rice yield and fertlizer consumption which is really appreciable. Moreover it links storage infrastructure under GBY subsidy scheme to increase the agricultural yield. But there are certain issues which need to be addressed as far as the research is concerned. First that the scheme was implemented in 2001-02 and it is taken into study in 2018 with data upto 2010 only. The mean rice yield for the apportioned data used in the analysis is around 1.77 tonnes per hectare which I think is less and needs to be checked. Moreover the author should mention the name of the districts taken for the study. (ii) Is the analysis correct? Author has used only regression analysis to find the results and for that matter analysis is correct. One observation which I think is very important is that it is not storage infrastructure which increase fertilizer consumption but farmer’s willingness and returns from recommended doses of fertilizer also plays an important role in fertilizer consumption. Author has explained research paper theoretically more than analytically. There is possibility of statistical analysis for finding the results in monetary terms.

Somdeep Chatterjee - Response to Reviewer 2 Comments
August 29, 2018 - 10:26 | Author's Homepage
I thank the reviewer for providing useful feedback and comments on my paper. Here are my point-wise responses. Is the contribution of the paper potentially significant? The paper is hypothetically significant in the sense that it gives the results that GBY program led to increase in rice yield and fertlizer consumption which is really appreciable. Moreover it links storage infrastructure under GBY subsidy scheme to increase the agricultural yield. But there are certain issues which need to be addressed as far as the research is concerned. First that the scheme was implemented in 2001-02 and it is taken into study in 2018 with data upto 2010 only. The mean rice yield for the apportioned data used in the analysis is around 1.77 tonnes per hectare which I think is less and needs to be checked. Moreover the author should mention the name of the districts taken for the study. Response: I thank you for your observations. However, I do not quite understand the phrase "hypothetically significant". I would have thought, even in a subjective judgment paradigm, that the answer is rather binary in terms of a yes or no. The comment about data updation is very similar to the one raised by reviewer 1 above and the response of mine is also the same. While I sympathise with the concerns, it is in the best interests of the empirical exercise that the data remains "tight" around the policy year which is why choosing similar number of pre- and post-policy years works out. Also, the mean rice yield is based on calculations for the given years in given districts. I have taken all districts in the 15 major states of India. Once again, I would like to reiterate that the empirical design relies on the fact that other contemporaneous policies would not affect the relevant cross sections (the kind of concerns raised by the invited reader above), which makes it all the more imperative to focus on years closer to the policy.(ii) Is the analysis correct? Author has used only regression analysis to find the results and for that matter analysis is correct. One observation which I think is very important is that it is not storage infrastructure which increase fertilizer consumption but farmer’s willingness and returns from recommended doses of fertilizer also plays an important role in fertilizer consumption. Author has explained research paper theoretically more than analytically. There is possibility of statistical analysis for finding the results in monetary terms. Response: I thank you for acknowledging the accuracy of the regression specifications. I also totally agree with you that there can be several other Channels that motivate fertilizer consumption. The empirical framework I use is a quasi-experimental "intent-to-treat" setup. As a result, the reduced form regressions would simply point out a broad causal relation between two variables and not rule out potentially other mediating channels. Precisely out of such concerns, would one not want to use a 2sls setting because that requires making strong assumptions on the exclusion restrictions, much of which would have to be controversial. I disagree that the research is theoretical, in fact it is very highly applied empirical work with just a little bit of theoretical explanations in the conceptual framework. I think a quasi-experimental design like this is a sophisticated statistical (or rather econometric) application to the data at hand, so I do not understand what additional statistical analysis you are suggesting. However, as all academics, having an open mind, I am all for exploring other methods if you have any specific in mind, especially if those appear more compelling than the triple difference strategy I employ. I also do not quite understand what results in monetary terms would imply. Thanks again for taking time out to read the paper. I really appreciate your feedback and discourse.