Did the Plan Sénégal Emergent Affect Cropping Decisions in the Senegal River Basin?
Round 1
Reviewer 1 Report
Although the paper addresses a relevant research question, the theoretical and empirical analysis have too many shortcomings to be published in Economies.
First, the theoretical model is based on the assumption that rural smallholders in developing countries are profit maximizers. This is a very strong assumption, which has been demonstrated since the 1960s that it does not hold. In reality, farmers are utility maximizers, rather than profit maximizers, as they cannot separate their production and consumption decisions if markets are imperfect. This is something the authors acknowledge themselves, by outlining land, labor and credit market imperfections as most important constraints. So the whole theoretical model the empirical analysis is based on, does not hold. In addition, there were some small errors in the profit maximizing model:
- Line 151: first ‘output’ should be ‘price’
- Equation 2: the subscript j is nowhere explained
Second, the data collection is not clearly described. Target households are located in villages that are characterized by rapid transformation while control households are located in slow transformation villages. The authors do not define what they exactly mean with rapid or slow transformation. They mention that the control villages were selected along the Senegal River but a closer look at the map in Figure 1 reveals that control villages 22, 23, 24, 25 and 26 are located further inland, in the department of Ranerou Ferlo. These villages are not representative at all to compare with villages along the Senegal river and I wonder how the authors matched these villages with the treatment villages.
Also, how many villages in total were selected? From the text I derive that there are 20 target villages, and 10 control villages. So I assume that 10 households per village were selected in the treatment group while there were 20 households selected per control village. Why was this not equal? In addition, the map shows that there were 40 villages in total. How can you explain this?
Could you also give more explanation about the two questionnaires that were used (one to measure rural transformation and one to measure employment in local enterprises)? Were these structured with different quantitative modules? Or did you use more open, semi-structured interviews?
Third, the empirical analysis has some serious flaws. The whole analysis and interpretation of the results is based on just one measure, i.e. the shares of area planted with different crops. Structural transformation is a complex process and entails several components, such as assets, livelihood strategies and upward income mobility / poverty reduction. You cannot draw conclusions and policy implications based on this measure alone. You need to calculate at least measures of total land, productivity, profitability, total farm revenues, etc. Also, specify whether you use average shares or shares of the average.
The test statistic that you use is way too complex. To test significant differences, you can just conduct a t-test or chi² test.
The analysis of the causal effect is not correct. You do not control for any observable characteristic of the farm-household (in terms of human/physical/natural capital, location, etc.). You say that you matched control villages with treatment villages but you do not show this. You need to present at least a table with descriptive statistics and test whether they differ significantly across the treatment and control groups. The fact that you have panel data is promising, but you do not use fixed effects or random effects to analyse the causal effect of the program. You create a difference in difference information statistic but you do not take into account that these villages may differ at the baseline, and that these differences explain the commercialization process. Also, the null hypothesis is that IΔ1 = IΔ2. If you find a p-value of 0.1303, then you fail to reject that IΔ1 is equal to IΔ2. Hence, it is assumed that they are similar to each other. This is not what you derive: you assume that they are different from each other, which is wrong.
Some other minor comments:
- Line 60: 'am' should be changed into 'an'
- Line 101: space in between 'onions' and 'but'
- Line 103: 'notional' should be changed into 'national'
- When the figure is printed in black and white, the difference between the control and treatment villages is not clear. Use gray-scale colors to make sure that the differences are also clear when printed in non-color.
Author Response
See attached file.
Author Response File: 
Author Response.pdf
Reviewer 2 Report
Comments on Senegal article to authors
The paper purports to measure the short-term impact of agricultural policies in Senegal in inducing greater specialization to commercial crops (especially rice) over the period 2014-17. It uses data from a household survey to compare changes in shares of area allocated to different crops among households that were in areas subject to programs and policies aimed at promoting commercial crops compared with matched households outside of these areas.
The following elements in the analysis and presentation make it difficult for the reader to follow the argument and become convinced of the results.
1. The authors acknowledge in the introduction (lines 25 ff) that farmers often follow diversified livelihood strategies to help mitigate risk. This implies that risk mitigation, as opposed to single-year expected profit maximization, is an important part of the farmers’ strategies, as small farmers are risk adverse. Yet the theoretical model presented in section 3.1 of the paper is one of simple profit maximization. This seems a bit inconsistent with what is implied in the introduction of the paper. The argumentation needs to be clarified about why the simple profit maximization is an appropriate model, particularly for both the farmers is the less favorable areas (used as controls) as well as those in the more dynamic areas.
2. The exposition about the selection and matching of treatment and control households (starting on line 180 plus figure 1) is not clear.
a. Figure 1 shows 20 control villages (5 of which are completely off-river, which raises questions about whether they would be appropriate controls for on-river treatment villages). But the text says that there were 20 treatment villages but only 10 control villages. Which was the case?
b. The text further states that there were 20 sample households chosen from each of the control villages. Were there then only 10 households chosen from each of the treatment villages? This needs to be cleared up.
c. Line 192 says that treatment and control households were matched based on poverty and income activity levels. But to assess the impact of agricultural policies, are those the most relevant matching variables? Presumably, one would want to match them according to their potential to respond to more favorable incentives emanating from policies—i.e., variables such as access to additional production resources that are not the direct focus of policy (labor, land, soil fertility, existing irrigation facilities, etc.). If the treatment households have, for example, better access to irrigation infrastructure than the control households, then naturally more favorable rice production policies will elicit a stronger response from them than they would from control households. But that doesn’t logically imply that one would get a similar response if those policies were applied in the control villages. So attributing all the observed differences in differences just to policy is misleading.
3. The transition of discussion from the data displayed in Table 1 to the subsequent statistical analysis starting around line 206 is difficult to follow (at least for me). The conclusions from the statistical analysis seem to be in contradiction with the basic data in table 1. The paper would benefit from a clearer explanation with regard to the following points:
a. The discussion of the data in table 1 notes that the relative rate of change in shares of area devoted to rice was actually faster for the control households than for the treatment households. The authors then raise the question of whether the difference in relative rates between the two types of households is statistically significant.
b. The authors do not note the even more dramatic shift in the share of area devoted to another key commercial crop that is a focus of the agricultural policy—onions. While the share of area devoted to onions among the treatment households fell between 2014 and 2017 by 30% (from 6.5% to 4.5%), it quadrupled among the control group (from 3.9% to 15.6%). It seems odd that the authors don’t even mention this dramatic change and focus solely on rice.
c. On rice, the subsequent part of the paper goes through the derivation and calculation of an information statistic (and its distribution) to determine whether the rate of change of the area share going to rice is statistically different between the control and the treatment group. After a series of calculations (which the average reader probably won’t follow), the authors conclude that even though the data in table 1 show a faster relative rate of growth in rice share of area among the control households than the treatment households, the share is really growing significantly faster for the treatment households. It’s not clear to me how these two seemingly contradictory findings are reconciled.
d. Even more so, if one includes the big changes in onion shares (increasing rapidly for the control households and shrinking for the treatment households), it leaves one wondering how the authors conclude that the policies are leading to a faster (relative) commercialization shift among the treatment households. On this latter point, the statement on lines 291-292 that there has been an economically significant reallocation towards onions among the treatment households seems to be contradicted by the data in table 1.
e. Lines 292-293 state that “In 2017 the treatment group planted over 90% of its land to rice and onions, compared to less than 50% for the control group.” While this is true, one gets a very different picture when one regards the shift in shares between 2014 and 2017. In absolute terms, the shares devoted to rice and onions by the treatment group increased by 10.5% (from 81.4% to 91.9% -- a relative increase of 12.9%). Amongst the control households, the absolute increase in shares was a much larger 17.9% (from 31.1% to 49.0%--a relative increase of 57.6%).
f. Line 290. It’s not clear to me how the authors define an “economically significant” reallocation of land as distinct from a statistically significant reallocation. What is the standard used to determine whether the reallocation is economically significant?
4. A more minor point: The paper sets out as one of its objectives to test whether growth in staple food crops can serve as strong an impetus to poverty reduction and economic growth as an emphasis on export crops. It sets this up as a major debate. But if the staple food crop is an import substitute, for which demand is growing quickly, as is the case for rice in Senegal, is the distinction between the export crop and the import substitute really that great? Both are generating foreign exchange (either earned or saved) and if both are oriented towards a market with growing demand, what is the big conceptual difference between the two? I suggest de-emphasizing this point, as it seems to be a bit of a false distinction.
5. Some minor typographical errors:
a. Line 118. Insert the word “so” before “that”.
b. Line 190. Insert the word “among” before “51”.
c. Line 206. Subscripts are inverted on what should read S1i
d. Line 225. “receive” at end of line should be “receives”. [1/3 is a singular noun].
Author Response
See attached document
Author Response File: Author Response.pdf
Round 2
Reviewer 2 Report
This revised version of the paper is much improved and addresses most of my prior comments. Moving some of the material on methods to an appendix also helps with the flow of the article.
Remaining points/comments:
1. While the reframing of the conceptual approach as a household model (rather than as a simple profit maximization) is very helpful, the specification of the two cropping activities as rice and onions seems to imply that the major tradeoff in labor and land allocation is between these two crops. But the focus of the article is on commercialization of agriculture, so it seems to me that it would be better to specify three agricultural activities—rice, onions, and a subsistence crop, such as sorghum. That would put the emphasis in the model on the tradeoffs occurring in resource allocation between commercial crops and subsistence crops. Alternatively (but perhaps less satisfactory), one could have an aggregate “commercial agriculture activity” (some amalgam of rice and onions) vs. a subsistence activity.
2. In lines 299ff, the authors explain that the information test statistic examines whether the control group is changing its land allocation faster than that of the treatment group—i.e., that the groups are converging. The test not rejecting the null hypothesis, one concludes that the control group is not changing faster than the treatment group. That explanation appears clear. But is this finding consistent with the statement on lines 333-334 that the treatment group is commercializing faster than the control group? A reader might be excused in concluding from the earlier discussion of the test statistic, the fact that the test did not show that the control group was commercializing faster than the treatment group does not prove the converse—that the treatment group is commercializing faster than the control group. Some clarification is needed on this point.
3. Lines 350-351 state that the results show that farm consolidation is not necessary to promote commercialization of farming. True, but the proponents of consolidation can argue that there is nothing in these findings to indicate that commercialization might not proceed faster if consolidation takes place. The latter point is, I think, the gist of their argument. They don’t argue that no commercialization will take place on small farms, only that it will be slower than in the context of larger farms.
4. While the revised version’s explanation of sampling methods is much clearer than the earlier version, a couple of comments remain:
a. Line 225 still says that only 10 control villages were selected, while elsewhere the article (e.g., line 232, Figure 1 and Appendix A) states that 20 control villages were chosen.
b. Lines 228-230 state that control villages selected from off-river locations were “not that different” from the treatment villages, all of which were on-river. “Not that different” is a vague statement. Perhaps a short appendix table could be added comparing the off-river and on-river villages to convince the reader that “not that different” has a more concrete meaning.
5. In table 2, it is striking that the correlation between rice and onion production in the control group shifts from positive (complementary) in 2014 to sharply negative (competitive) in 2017—much more negative than with the control group. Might this indicate that in the absence of the supporting services provided to the treatment group over the period 2015-17, farmers in the control villages, as they commercialized, had to make difficult tradeoffs about where to allocate resources—including fertilizer—tradeoffs that were not so severe for the treatment villages due to the services provided to them (e.g., expanded access to credit and production inputs)? This implication would seem to reinforce the authors’ conclusions about the positive impact of the programs and projects they discuss.
6. A few minor typos need correcting:
a. Line 249: “adds” should be “add” (as the preceding noun – “shares” – is plural).
b. Line 289 – remove the gaps between x and t in “Nex t”
c. Line 329 – add space between “a” and “significant”
Author Response
See attached file
Author Response File: Author Response.pdf
