The Credit Accessibility and Adoption of New Agricultural Inputs Nexus: Assessing the Role of Financial Institutions in Sudan
, Abbas Ali Chandio 3, Maurice Osewe 1, Muhammad Abdullah 4 and Yueqing Ji 1,*
Calum Turvey
Round 1
Reviewer 1 Report
This paper investigates agricultural finance in Sudan. It is nicely written, based on primary data, and provides an extensive literature review.
The problem is that the authors are attempting to determine whether lending has an impact on input use AND the factors affecting loan-making decisions. Just in this setup the two issues are endogenous. From the demand side are increases in fertilizer and hybrids (and perhaps land holdings) a consequence of borrowing OR from the supply side do lenders look to improved seeds and fertilizers in order to make a loan? Causality is jointly determined.
The authors decide on a bivariate probit model separating the two banks. But it seems to me from the statistics that no farmers borrow from both banks so I am not sure what is gained in doing this. This also eliminates the possibility of dealing with the endogeneity issue since I am pretty sure there is no IV based bivariate probit model.
I strongly recommend that the authors reconsider the econometric model model and consider IV Probit or Logit.
Consider
Fertilizer = f(loan, and other variables to identify)
Seed = g(loan, and other variables to identify)
Loan – h(fertilizer, seed, and other variables affecting supply side to identify)
Fertilizer, Seed and Loan are endogenous. This is 2SLS setup. The first two address your question of whether access to credit increases fertilizer and seed use; the third addresses the questions whether lenders are more likely to lend to farmers with fertilizer or improved seed. It will help you in determining the causal relationship between the demand-driven and supply-driven questions you ask.
As you are also interested in whether there is a difference in the lender-borrower relationship between the two banks then run a model including both banks (maybe with a dummy variable for bank source) and then run two separate models for each bank. The marginal effects or coefficients will give you the answers you seek.
As you have two distinct regions you may also want to examine regional effects.
On this note, Table 3 has to be redone. The numbers and % are relative to the total sample. This table should only include those that received a loan from either bank and the percentages should relate to percent for ABS and FCB. For example the frequency distribution for age should sum to 100% for ABS and 100% for FCB, then you can discuss proportionality in terms of sample, and for variables you choose to include in regressions you can determine which were influential in affecting loan supply and/or loan demand.
In the financial inclusion literature there is a big difference between access and usage. Your loan variable is usage. You can perhaps use the distance variable (among others) to capture access. Just because an individual has access to credit does not mean he will use it.
Also the area variable is mixed and matched with hectares. Be consistent. I would convert to hectares as that would be easier for global readers.
I am sorry I have to raise the endogeneity issue, but it is important in these kinds of studies. Otherwise the problem is good, you write very well, the background literature review is good, I always appreciate efforts to collect primary data (just don’t be too confident about attributing results to all of Sudan), and find descriptive tables useful (and sometimes way more interesting than regressions).
Author Response
Reviewer 1
This paper investigates agricultural finance in Sudan. It is nicely written, based on primary data, and provides an extensive literature review.
Author response. We appreciate your thorough revision and insightful suggestions. And we thank you for the encouraging compliment to our work.
The problem is that the authors are attempting to determine whether lending has an impact on input use AND the factors affecting loan-making decisions. Just in this setup the two issues are endogenous. From the demand side are increases in fertilizer and hybrids (and perhaps land holdings) a consequence of borrowing OR from the supply side do lenders look to improved seeds and fertilizers in order to make a loan? Causality is jointly determined.
The authors decide on a bivariate probit model separating the two banks. But it seems to me from the statistics that no farmers borrow from both banks so I am not sure what is gained in doing this. This also eliminates the possibility of dealing with the endogeneity issue since I am pretty sure there is no IV based bivariate probit model.
I strongly recommend that the authors reconsider the econometric model and consider IV Probit or Logit.
Consider
Fertilizer = f (loan, and other variables to identify)
Seed = g (loan, and other variables to identify)
Loan – h (fertilizer, seed, and other variables affecting supply side to identify)
Fertilizer, Seed and Loan are endogenous. This is 2SLS setup. The first two address your question of whether access to credit increases fertilizer and seed use; the third addresses the questions whether lenders are more likely to lend to farmers with fertilizer or improved seed. It will help you in determining the causal relationship between the demand-driven and supply-driven questions you ask.
As you are also interested in whether there is a difference in the lender-borrower relationship between the two banks then run a model including both banks (maybe with a dummy variable for bank source) and then run two separate models for each bank. The marginal effects or coefficients will give you the answers you seek.
As you have two distinct regions you may also want to examine regional effects.
Author response. We appreciate your detailed and wonderful suggestions for the empirical models. On one hand, Farmers’ demand for input use may cause the demand for credit (or loan-getting). On the other hand, the banks’ supply of credit hence famers’ loan-getting will help them to purchasing input. Thus, farmers’ loan-getting and input use behaviors are indeed endogenous. And a lot of papers have discussed these issues.
This paper uses a very different approach to answers these questions: 1) what kinds of farmers can get credit from the banks, or what kinds of farmers would be chosen as borrowers by the banks in Sudan? 2) Is the farmers who are more likely to adopting new inputs also the ones preferred by banks? Or whether farmers’ credit demand induced by new inputs use has been well satisfied by the banks.
1) what kinds of farmers can get credit from the banks, or what kinds of farmers will be chosen as borrowers by the banks in Sudan? 2）Is the farmers preferred by banks also the ones who are more likely to adopting new input. If the answer is YES, then the banks do a good job in the sense of helping the adoption of new input.
To do this job, we analysis the factors determining whether a farmer has been or will be chosen as a borrower by the banks and the factors determining a farmer whether use new inputs simultaneously. Then, we further investigate the correlation between the possibility of being chosen and new input use. In this approach, we want to see the nexus of credit availability and adoption of new inputs through other factors. Such as the area of cultivated land is important for new input adoption, if it is also an important factor for farmers being chosen as borrower by the banks, then the nexus will be increased and the banks do a good job in choosing the “right” farmers.
We also use the IV approach you suggest to investigate the correlation directly. The adoption of chemical fertilizers and improved varieties are instrumented by the ratio of adopters for other observations in the same village. In this approach, we want to know whether farmers’ credit demand induced by the chance of using new input actually has been satisfied by the banks.
We further compare the nexus for the sample more near Agricultural Bank of Sudan (ABS) and the sample more near Farmer’s Commercial Bank (FCB) to know which bank do a better job.
On this note, Table 3 has to be redone. The numbers and % are relative to the total sample. This table should only include those that received a loan from either bank and the percentages should relate to percent for ABS and FCB. For example, the frequency distribution for age should sum to 100% for ABS and 100% for FCB, then you can discuss proportionality in terms of sample, and for variables you choose to include in regressions you can determine which were influential in affecting loan supply and/or loan demand.
Author response. Thank you for your suggestion. We have restructured the table based or your suggestion. Kindly refer to table (3).
In the financial inclusion literature, there is a big difference between access and usage. Your loan variable is usage. You can perhaps use the distance variable (among others) to capture access. Just because an individual has access to credit does not mean he will use it.
Author response. We appreciate your view on this, in fact we first ask the farmers whether he/she obtain credit from bank, and once he/she answer “NO”, we asked them whether you could obtain credit if they have applied for. By doing this, we define accessibility as whether they can get credit from banks if they need and apply for it. Credit accessibility is more exogenous than credit usage.
Also, the area variable is mixed and matched with hectares. Be consistent. I would convert to hectares as that would be easier for global readers.
Author response. Thank you for your suggestion. The land size has been converted to the hectares. Kindly refer to line 294 and 295, also refer to table (1) and table (2). In addition to line 376, 380 and 382.
I am sorry I have to raise the endogeneity issue, but it is important in these kinds of studies. Otherwise, the problem is good, you write very well, the background literature review is good, I always appreciate efforts to collect primary data (just don’t be too confident about attributing results to all of Sudan), and find descriptive tables useful (and sometimes way more interesting than regressions).
Author response. Thank you so much for this praise, and we appreciate your opinion and had tried to avoid the endogeneity issue by avoiding directly regress credit usage on the adoption of new input.
Reviewer 2 Report
1) In 93 addition, this paper also attempts to give answers of following research questions: 1) How 94 do banks choose rural borrowers for credit? 2) which bank does a better job of adopting 95 technology diffusion?
This is confusing because now the unit of measurement is different. You switched from households to banks. Do you have data from banks?
2) This study was done in two regions, North Kordofan State and Al Jazera Scheme in 189 Sudan (see Figure 1).
What is the motivation for these two regions?
3) Where: ?1∗ and ?2∗ denote latent observed variables for access to credit from the 258 two banks
The notation is confusing. Earlier in the section you talk about y1 being the outcome variable and y2 being the treatment variable.
4) 30.9% of credit beneficiaries acquired their credit from 274 ABS, while 19% obtained their credit from FCB.
Can you provide information about the number of locations for ABS and FCB in the study regions? Does ABS have more locations than FCB?
5) Men make up 98% of farmers.
How are the gender relations in Sudan? Is it customary to approach the male of the household as opposed to the female? Are females not also active in the farm operation?
6) The average level of formal schooling for farmers is nine years demonstrat-282 ing a significant formal education regarding the government policies, banking transac-283 tions, and new technologies.
You call nine years significant? Is is typical in Sudan to learn about banking transactions in school?
7) However, the results indicate that 45.30% of 297 rural households are in farmer group associations, and 28% have membership in cooper-298 ative societies in their villages.
What is the difference between associations and co-operatives?
8) About 96% of the rural household head is in charge of 278 agricultural production.
So you surveyed individuals who are not farmers? Should those individuals not be excluded from your analysis?
9) Concern-389 ing the adoption of chemical fertilizers, results show that ABS lenders are more likely to use chem-390 ical fertilizers. This result implies that ABS contributed to diffusion technology adoption among 391 farmers.
No. Credit access is the outcome variable, not the predictor. So the relationship is reversed. Farmers who use chemical fertlizers are more likely to lend from ABS.
10) Access to Credit and Adoption of Agricultural Technologies 2 Nexus: Assessing the Role of Financial Institutions in Sudan
The manuscript title is very misleading. You do not run a model with technology adoption as an outcome variable. Instead, it is a predictor of credit access. Also, your predictors are farm characteristics and farm operator characteristics, not bank characteristics.
Author Response
Reviewer 2
1) In 93 additions, this paper also attempts to give answers of following research questions: 1) How 94 do banks choose rural borrowers for credit? 2) which bank does a better job of adopting 95 technology diffusion?
This is confusing because now the unit of measurement is different. You switched from households to banks. Do you have data from banks?
Author response: Thank you for your observation. No, we do not have data from bank. And we have rephrased the first question to be clearer (What kinds of farmers can get credit from the banks, or what kinds of farmers would be more likely chosen as borrowers by the banks in Sudan?) And the second research question rephrased to (Is the farmers who are more likely to adopting new inputs also the ones preferred by banks? Or whether farmers’ credit demand induced by new inputs use has been well satisfied by the banks.). kindly refer to line 101 – 105 of the revised text.
2) This study was done in two regions, North Kordofan State and Al Jazera Scheme in 189 Sudan (see Figure 1).
What is the motivation for these two regions?
Author response: Because credit is not widely available among Sudanese rural farmers, we were forced to concentrate our efforts in the areas with the best access to credit. We chose the Al Jazera scheme because it is the most important area of agricultural irrigation, and the government places a high value on this region due to its significant contributions to the irrigated agricultural sector. As a result, farmers have good access to credit.
Similarly, North Kordofan state is critical as a rain-fed agricultural area, and farmers have good access to financing, so we chose this area because of its importance in the rain-fed agricultural sector. In fact, there are some areas where farmers have access to credit, but due to financial constraints, we initially focused on these two areas.
In our future study, we will cover others regions in order to give more information to support our conclusion. Kindly refer to line 199 to 207 of the revised text.
3) Where: ?1∗ and ?2∗ denote latent observed variables for access to credit from the 258 two banks
The notation is confusing. Earlier in the section you talk about y1 being the outcome variable and y2 being the treatment variable.
Author response: Thank you for this observation. We have changed the econometric model based on the suggestion of another reviewer. And the original model has been given up.
4) 30.9% of credit beneficiaries acquired their credit from 274 ABS, while 19% obtained their credit from FCB.
Can you provide information about the number of locations for ABS and FCB in the study regions? Does ABS have more locations than FCB?
Author response: Thank you for this observation. Yes, this is due to the fact that ABS has more locations than FCB, with 11 ABS locations distributed throughout Al Jazera state. While in North Kordofan state, there are 7 locations distributed to the localities' centers. In contrast, Al Jazera state has six FCB locations while North Kordofan state has only two. Kindly refer to line 281 to 284 of the revised text.
5) Men make up 98% of farmers.
How are the gender relations in Sudan? Is it customary to approach the male of the household as opposed to the female? Are females not also active in the farm operation?
Authors response: Thank you for the observation. Gender roles are highly patriarchal and rigidly defined in Sudan. The men are viewed as the main income earners, while women are seen as the homemakers. Husbands are expected to provide economically for their wives and children throughout their lives. And female is active and participate in agricultural activities but their participations is under their husband’s, father’s or brother’s control. Therefore, in Sudan, the head of the household is entrusted with carrying out agricultural operations with the help of family members, including the wife or daughters.
6) The average level of formal schooling for farmers is nine years demonstrat-282 ing a significant formal education regarding the government policies, banking transac-283 tions, and new technologies.
You call nine years significant? Is typical in Sudan to learn about banking transactions in school?
Author response: No, it is not, but what we mean by this point is that farmers have a good level of education that makes them understand bank transactions such as opening an account and other transactions, just as if there is an agricultural policy by the state aimed at developing agriculture, their level of education will enable them to understand well especially since the nine years of schooling in Sudan is an intermediate stage.
7) However, the results indicate that 45.30% of 297 rural households are in farmer group associations, and 28% have membership in cooper-298 ative societies in their villages.
What is the difference between associations and co-operatives?
Author response: Thank you for the observation. Cooperative societies are service associations and enjoy their ability to distribute inputs and obtain financing and marketing under the control of small farmers residing in rural areas. Also, cooperative societies take legal status and may be registered under the Cooperative Companies Law. Whereas, the Farmer associations is a demand union that includes a group of farmers in an area and aims to demand facilitation of the obstacles facing small farmers, and it does not take the legal status as cooperative societies.
8) About 96% of the rural household head is in charge of 278 agricultural production.
So, you surveyed individuals who are not farmers? Should those individuals not be excluded from your analysis?
Author response: No, all surveyed samples are farmers. Because there is more than one person in the family members who participate in the agricultural operations, so the question was about who is in charge of agricultural production among the family members. Because it could be the father, sons, or some else.
9) Concern-389 ing the adoption of chemical fertilizers, results show that ABS lenders are more likely to use chem-390 ical fertilizers. This result implies that ABS contributed to diffusion technology adoption among 391 farmers.
No. Credit access is the outcome variable, not the predictor. So, the relationship is reversed. Farmers who use chemical fertilizers are more likely to lend from ABS.
Author response: Thank you for the correction. Although we have rewritten the empirical results and abstract, your correction is noted.
10) Access to Credit and Adoption of Agricultural Technologies 2 Nexus: Assessing the Role of Financial Institutions in Sudan
The manuscript title is very misleading. You do not run a model with technology adoption as an outcome variable. Instead, it is a predictor of credit access. Also, your predictors are farm characteristics and farm operator characteristics, not bank characteristics.
Author response: Thank you for this observation. We have run a model with technology adoption. Kindly refer to table 6 of the revised text.
Round 2
Reviewer 2 Report
No new comments.
Author Response
Thank you for your valuable comments
