Microfinance Participation in Thailand

Round 1

Reviewer 1 Report

Please, be careful and distinguish between the percentage increase (decrease) and "percentage point" increase, decrease. 

Most of the coefficient influence/interpretation is "percentage point" increase, decrease. 

Author Response

comment 1: Please, be careful and distinguish between the percentage increase (decrease) and "percentage point" increase, decrease.

 

Most of the coefficient influence/interpretation is "percentage point" increase, decrease.

 

Our Response – We thank the reviewer for his/her comments. We have revised the manuscript so that this is clear. Please see lines 566-567, lines 590-591, line 596, page 17; lines 605-606, lines 619-620, line 639, page 18; line 649, page 19; line 718, line 724, page 20; line 732, line 734, page 21; line 776, page 22; lines 827-828, lines 833-835, page23; line 870, line 879, page 24.

Author Response File: Author Response.docx

Reviewer 2 Report

The literature review must be implemented. Perhaps with some more recent references.

Despite this the work is good.

Author Response

comment 1: The literature review must be implemented. Perhaps with some more recent references. Despite this the work is good.

 

Our Response – We thank the reviewer for his/her suggestions. We have added more to the literature review (please see lines 255-263, page 6).

Reviewer 3 Report

This is a good paper and worth to be published in this journal.

Authors find a very interesting research topic, use appropriate methodology, and present the results clearly.

I don't see any need for significant modifications.

It might be better if the authors mentioned the summary of the main results briefly in the introduction to guide the readers.

Author Response

comment 1: This is a good paper and worth to be published in this journal. Authors find a very interesting research topic, use appropriate methodology, and present the results clearly. I don't see any need for significant modifications.

 

Our Response – We thank the reviewer for his/her kind comments.

 

Minor comment 2: It might be better if the authors mentioned the summary of the main results briefly in the introduction to guide the readers.

 

Our Response – We thank the reviewer for his/her suggestions. As requested, we have added a summary of the main results in the introduction to the article (please see lines 95-98, pages 2-3).

Reviewer 4 Report

I thank the authors for this wonderful research, which is indeed informative for micro-insurance program planning and implementation. In the study, the authors aim to identify the determinates that affect decision to participate in Village Funds, and Saving Groups for Production (SGPs) from a set of alternatives with the use of sophisticated econometric methods. I feel that the paper also has much to offer to the growing climate change adaptation literature, as microfinancing mechanisms are often a priority for assisting rural communities to adapt to growing risks of climate change.

General comments:

I have marked specific comments in the PDF of the manuscript, and below are my general comments:

1. After reading the text, I notice a high degree of repetition in the text. For instance, when one reads section 3 (literature review) and section 4.1 (the conceptual framework) there is significant overall and I do not understand why authors chose such a structure.

I would suggest reducing the non-necessary subsections and re-structuring the paper to reduce redundancy.

2. I would suggest text form line 392 to line 437 could be moved to Section 5: Results and discussion. Although it is left to deliberations by and among authors.
3. Why section 5.1: Estimation Strategies cannot be a part of Methods, and why is it included in Results and discussion?
4. In Section 5: Results and discussion, It would be worthy to add more points of discussion. For instance, the authors report that female-headed households participate more in MFI schemes. What is the characteristics of these female-headed households are they mostly educated or not? Such critical reasoning will help in understand and interpreting the results more comprehensively.
5. I would also suggest the authors mention about the limitations of the methods used in this study and what is the further scope of this study.
6. Some specific comments are marked in the PDF.

All the best!

Comments for author File: Comments.pdf

Author Response

Comments and Suggestions 1: I thank the authors for this wonderful research, which is indeed informative for micro-insurance program planning and implementation. In the study, the authors aim to identify the determinates that affect decision to participate in Village Funds, and Saving Groups for Production (SGPs) from a set of alternatives with the use of sophisticated econometric methods. I feel that the paper also has much to offer to the growing climate change adaptation literature, as microfinancing mechanisms are often a priority for assisting rural communities to adapt to growing risks of climate change.

 

Our Response – We thank the reviewer for his/her kind comments.

 

General comments

 

General comment 1: After reading the text, I notice a high degree of repetition in the text. For instance, when one reads section 3 (literature review) and section 4.1 (the conceptual framework) there is significant overall and I do not understand why authors chose such a structure.

I would suggest reducing the non-necessary subsections and re-structuring the paper to reduce redundancy.

 

Our Response – We thank the reviewer for his/her suggestion. We have revised the structure by moving section 4.1 to section 3 (literature review) (please see lines 178-238, pages 4- 5).

 

General comment 2: I would suggest text form line 392 to line 437 could be moved to Section 5: Results and discussion. Although it is left to deliberations by and among authors.

 

Our Response – We have moved text from line 392 to line 437 to Section 5 (please see lines 447-493, pages 11-12).

 

General comment 3: Why section 5.1: Estimation Strategies cannot be a part of Methods, and why is it included in Results and discussion?

 

Our Response – We have moved section 5.1: Estimation Strategies to the methods section (please see lines 400-434, page 10).

 

General comment 4: In Section 5: Results and discussion, It would be worthy to add more points of discussion. For instance, the authors report that female-headed households participate more in MFI schemes. What is the characteristics of these female-headed households are they mostly educated or not? Such critical reasoning will help in understand and interpreting the results more comprehensively.

 

Our Response – We have added more information (please see lines 665-669, page 19; lines 779-781, page 22; lines 882-884, page 24).

 

General comment 5: I would also suggest the authors mention about the limitations of the methods used in this study and what is the further scope of this study.

 

Our Response – We have added in information about the study’s limitations in terms of its method and provided recommendations for future studies (please see lines 925-934, page 25).

 

General comment 6: Some specific comments are marked in the PDF.

 

Our Response – We have revised the journal article in line with the comments provided in the PDF (please see lines 26-31, lines 42-45, page 1; line 52, lines 62-66, page 2; lines 104-112, line 130, page 3; line 165, lines 178-194, page 4; lines 195-238, page 5).

 

 

 

We trust that these changes will meet with the satisfaction of the reviewers and editor alike. Thank you for your efforts on our behalf. We believe that the manuscript has been markedly improved by your attention to it.

 

Sincerely,

The Authors

Round 2

Reviewer 1 Report

I accepted your version.

Reviewer 4 Report

I thank the authors for revising the manuscript. Comments have been adequately addressed. The manuscript is publishable.

This manuscript is a resubmission of an earlier submission. The following is a list of the peer review reports and author responses from that submission.

Round 1

Reviewer 1 Report

Referee report:

Microfinance Participation in Thailand

 

The paper uses the Thai Socioeconomic Survey (2017) to identify factors affecting Thai households’ participation in microfinance programs.  Using the multinomial logit regressions authors conclude that Village Fund (VF) targets low-income rural households, older individuals with lower education, and female household. Moreover, their regressions show that VF participants are usually larger households. On the other hand, the competing MF program called Saving Groups for Production (SGPs) attracts economically active households with young household heads, well-educated, characterized by a higher household income.

While the paper tackles an interesting topic, I have several questions and suggestions, listed in random order below. Overall, there are certain issues in the empirical execution that needs explanation and maybe new estimation using a different set of alternatives, and/or method. At this moment conclusions could be drawn from the empirical analysis that potentially suffers from econometric flaws.

 

 

 

Main comments

• For the discussion of the MFI and their influence, the paper would strongly benefit by a graph/Table comparing characteristics of the three MFI groups in Thailand (SFIs, SPGs and VSs, and informal MFI). In terms of assets and participation and evolution over time. At least estimates and/or range.

 

• The literature review is too detailed and rather fragmented. I did not find a system by which the review of the empirical literature is organized. I would ask the authors to significantly change and rewrite this section. Maybe, putting leading and relevant papers into a table, columns countries and abbreviations of factors. In the table then show either size of the effect, or at least sign… Please, provide a better structure; say, each paragraph should address a group of factors (countries, and MFI).

 

• I am afraid that the set of used alternatives does not satisfy certain (independence) assumptions (also results presented in Table 4 are weird, negative statistics and marginal effects for category 4 are around 0, basically badly identified category). On page 9, 415-417: “In this study, households choose to participate in microfinance programs (or not to participate) based on their options; they have more than two mutually exclusive alternatives. The multinomial logit model is used to determine the factors that affect credit participation in Thailand (VFs and SGPs). The model is coded as four outcomes that affect microfinance credit participation (1 = non-VF and SGP borrowing; 2 = borrowing from VFs (only); 3 = borrowing from SGPs (only); 4 = borrowing from both VFs and SGPs).” Here I have several reservations and questions:

 

1. “1 = non-VF and SGP borrowing” … If I understand the authors' description, this could cover SGIs as well as informal MFIs. This means that this category corresponds to a very different group of MFI participants. It should be clarified which participants belong to this category (or if possible, to split the category into SGPs and informal MFIs).
2. Category 4 = borrowing from both VFs and SGPs) is a problematic category, it is not independent of the other choices… I would suggest in the main regression to exclude category 4 and analyze it separately. This suggestion is supported by estimated marginal effects (Table 5) for category 4 … which are all around 0, basically confirming that it is a badly identified category. Maybe it is worth to analyze it separately. What are the characteristics of those individuals that use both VFs and SGPs?
3. So, the survey covers all individuals using some MFI financing, right? It should be better spelled out.
4. In this respect, I do not understand the tests IIA with negative test statistics? I would read is as a problematic set of alternatives. It needs additional explanation.

 

• Some of the weak identification issues could be potentially solved/double-checked using simple logit or probit (or linear probability) models on several sub-sets. Simply i) analyzing those (0/1) who opted for VFs and SGPs (=1) and the other sources of MFI; on a subsample of VFs and SGPs: ii) analyze those SGPs who used also VP, iii) analyze that VPs who used also SGP, and iv) on a subsample of VP and SGP delete those using both sources and run logit or probit between these reduced categories.

 

• In all logit regression tables, please use up to three decimal points. There is no reason to use more digits. Variable education is likely coded; hence it would be better to use a set of dummies. Again, for the set of dummies, please, clearly indicate what is "omitted category" (i.e., the base category).

 

• For the interpretation, please do not overstate the effects based on the statistical significance of the coefficients. Please, look and use the size of the estimated marginal effects, i.e., the estimated probability increase/decrease caused by a unit change of the underlying variable.

 

 

 

Minor points

• Page 1, line 35-37: “The Gini index shows that income inequality in Thailand is the 35 highest in Southeast Asia (Bird et al. 2011). The index changes between 1988 and 2017 from 0.487 to 36 0.365, despite a declining poverty rate over the period (WB 2019).” The last sentence is worthy of additional explanation or an inside, why this is the case…

 

• Page 13, 541: “Note: *, **, *** indicate significance at 10%, 5%, and 1% levels, respectively.”.. Why is this note attached to Table 1, with the descriptive statistics? If the Table contains also a battery of say, mean tests, it should be mentioned… Besides, I did not find any significant marks in Table 1...I see that some of the tests of equality of means are performed in Table 2. I suggest removing Table 2 and in the text describing Table 1, make a note that for each of the characteristics presented in Table 1, authors rejected the null hypothesis of equality of the means on the level of 1%. Similarly for Table 3. It does not bring any extra information, the results of the tests (all giving the same results) can be summarized in the text. However, the authors should describe the testable hypothesis (i.e., which variables were used to test equality, which gives the number of degrees of freedom. Also, why they use this test and not say any test of homogeneity of the distribution).

 

 

Author Response

Reviewer #1’s comments:

 

Main comments

Main comment 1: For the discussion of the MFI and their influence, the paper would strongly benefit by a graph/Table comparing characteristics of the three MFI groups in Thailand (SFIs, SPGs and VSs, and informal MFI). In terms of assets and participation and evolution over time. At least estimates and/or range.

 

Our Response – We thanked the Reviewer for the suggestion. We added more information about VF and SGP (see lines 102-118, page 3 and lines 128-133, page 3).

 

We have tried to find more information to make graph/Table comparing characteristics of VF and SGP in terms of assets and participation and evolution over time. However, we were unable to find this information because the government does not currently have a database that contains this information.

 

Main comment 2: The literature review is too detailed and rather fragmented. I did not find a system by which the review of the empirical literature is organized. I would ask the authors to significantly change and rewrite this section. Maybe, putting leading and relevant papers into a table, columns countries and abbreviations of factors. In the table then show either size of the effect, or at least sign… Please, provide a better structure; say, each paragraph should address a group of factors (countries, and MFI).

 

Our Response – We thanked the Reviewer for the suggestion. We revised the literature review (see line 135, page 3 to line 188, page 4).

 

Main comment 3: I am afraid that the set of used alternatives does not satisfy certain (independence) assumptions (also results presented in Table 4 are weird, negative statistics and marginal effects for category 4 are around 0, basically badly identified category). On page 9, 415-417: “In this study, households choose to participate in microfinance programs (or not to participate) based on their options; they have more than two mutually exclusive alternatives. The multinomial logit model is used to determine the factors that affect credit participation in Thailand (VFs and SGPs). The model is coded as four outcomes that affect microfinance credit participation (1 = non-VF and SGP borrowing; 2 = borrowing from VFs (only); 3 = borrowing from SGPs (only); 4 = borrowing from both VFs and SGPs).” Here I have several reservations and questions:

 

Main comment 3.1: “1 = non-VF and SGP borrowing” … If I understand the authors' description, this could cover SGIs as well as informal MFIs. This means that this category corresponds to a very different group of MFI participants. It should be clarified which participants belong to this category (or if possible, to split the category into SGPs and informal MFIs).

 

Our Response – We thanked the Reviewer for the suggestion. We can explain that category 1 = non-VF and SGP borrowing means the households that do not borrow from both VFs and SGPs. Our study does not cover informal MFIs. This study is interested in VFs and SGPs, that are semiformal MFIs. We define this category as households who do not borrow from both VF and SGP.

 

To be clearer, we added more information about the categories (see line 314-318, page 7).

 

Main comment 3.2: Category 4 = borrowing from both VFs and SGPs) is a problematic category, it is not independent of the other choices… I would suggest in the main regression to exclude category 4 and analyze it separately. This suggestion is supported by estimated marginal effects (Table 5) for category 4 … which are all around 0, basically confirming that it is a badly identified category. Maybe it is worth to analyze it separately. What are the characteristics of those individuals that use both VFs and SGPs?

 

Our Response – We thanked the Reviewer for the comment and suggestion. We addressed the reviewer’s point by point as follows:

 

3.2.1 “Category 4 = borrowing from both VFs and SGPs) is a problematic category, it is not independent of the other choices… I would suggest in the main regression to exclude category 4 and analyze it separately. This suggestion is supported by estimated marginal effects (Table 5) for category 4 … which are all around 0, basically confirming that it is a badly identified category. Maybe it is worth to analyze it separately.”

 

Our Response –We tried to run multinomial logit model by excluding category 4, we found that the result is not different. The marginal effect is not different as well. When we checked IIA, the result violated this assumption.

 

3.2.2 “What are the characteristics of those individuals that use both VFs and SGPs?”

 

Our Response –The characteristics of households that take the loan from both VFs and SGPs are the average borrowers of 54.16 years old. The information shows that most of the both VF and SGP borrowers are farmers. Both VF and SGP borrowers have monthly household incomes about THB 25.17. Both VF and SGP borrowers have the highest average number of motorcycles (1.69 motorcycles).

 

Main comment 3.3: So, the survey covers all individuals using some MFI financing, right? It should be better spelled out.

 

Our Response – We thanked the Reviewer for the suggestion. The data shows households taking the loans. We tried to exclude category 4, and the result is not different. Moreover, the result violated the assumption of IIA.

 

Main comment 3.4: In this respect, I do not understand the tests IIA with negative test statistics? I would read is as a problematic set of alternatives. It needs additional explanation.

 

Our Response – We thanked the Reviewer for the comment and suggestion. We added more information (see lines 439, page 10 to 444, page 11).

 

Main comment 4: Some of the weak identification issues could be potentially solved/double-checked using simple logit or probit (or linear probability) models on several sub-sets. Simply i) analyzing those (0/1) who opted for VFs and SGPs (=1) and the other sources of MFI; on a subsample of VFs and SGPs: ii) analyze those SGPs who used also VP, iii) analyze that VPs who used also SGP, and iv) on a subsample of VP and SGP delete those using both sources and run logit or probit between these reduced categories.

 

Our Response – We thanked the Reviewer for the suggestion. We addressed the reviewer’s point by point as follows:

 

4.1 “Some of the weak identification issues could be potentially solved/double-checked using simple logit or probit (or linear probability) models on several sub-sets. Simply i) analyzing those (0/1) who opted for VFs and SGPs (=1) and the other sources of MFI; on a subsample of VFs and SGPs:”

 

Our Response –We analyzed the logit model, that is, those who borrow from VF, SGP and both VF and SGP (=1) and the others who do not borrow from VF and SGP.

 

Our result showed that VF and SGP participation is significantly explained by household head characteristics (age(+), female(+), education(+), married(+), single(-)), demographics (household size(+), dependency ratio(-), number of children(+), number of elderly people(-)), occupation (farmer(+), formal and informal worker(+)), income, expenditure, and assets (monthly expenditure on food and beverages(-), number of cars(-), number of motorcycles(+)), and other variables (central(+), north(+), northeast(+), south(+), rural households(+), difficulty obtaining an emergency loan(-)). Most of these variables are significant at 0.01 level except for south and car significant at 0.05 and 0.1 level, respectively.

 

The empirical results show that VFs and SGPs serve households in rural areas. Households with higher dependency ratios are less likely to borrow from VFs and SGPs. Most VF and SGP borrowers can access other loan sources when they need to obtain emergency loans. The programs also provides loans to elderly and higher-educated household heads. VFs and SGPs target women. Larger households are more likely to access VFs and SGPs. VFs and SGPs also grant loans to formal and informal workers.

 

Following this model, we can say that the results from our study with multinomial logit model give the result better than the logit model.

 

4.2 “ii) analyze those SGPs who used also VP, iii) analyze that VPs who used also SGP,”

 

Our Response –We cannot sort out the SGP borrowers who also used VF and VF borrowers who also used SGP. We just only know the borrowers in this group use both VF and SGP from the data set.

 

4.3 “and iv) on a subsample of VP and SGP delete those using both sources and run logit or probit between these reduced categories.”

 

Our Response –We analyzed the logit model as those who borrow both VF and SGP (=1) and the others who do not borrow from VF and SGP.

 

Our result showed that both VF and SGP participation is significantly explained by household head characteristics (age(+), female(+), education(+), married(+)), occupation (farmer(+), formal and informal worker(+)), income, expenditure, and assets (number of motorcycles(+)), and other variables (central(+), north(+), northeast(+), south(+)). Most these variables are significant at 0.01 level except married informal worker and motorcycles significant at 0.05 level.

 

Following this model, we can conclude that VF and SGP borrowers are high-educated and female household heads in rural areas. Households that own their own motorcycles have a higher probability of borrowing from both programs. Both VF and SGP borrowers are employed in a range of jobs: farmers, and in both formal and informal occupations. The marginal effect is still low and does not look better than our model.

 

Main comment 5: In all logit regression tables, please use up to three decimal points. There is no reason to use more digits. Variable education is likely coded; hence it would be better to use a set of dummies. Again, for the set of dummies, please, clearly indicate what is "omitted category" (i.e., the base category).

 

Our Response – We thanked the Reviewer for the suggestion. We addressed the reviewer’s point by point as follows:

 

5.1 “In all logit regression tables, please use up to three decimal points. There is no reason to use more digits.”

 

Our Response –We revised the values in Table 3 (see pages 13-14).

 

5.2 “Variable education is likely coded; hence it would be better to use a set of dummies.”

 

Our Response – We would like to explain that Variable education is not a dummy variable. This variable is years of education.

 

- We added a bracket to explain this variable in Table 3 (see pages 13-14).

 

5.3 “Again, for the set of dummies, please, clearly indicate what is "omitted category" (i.e., the base category).”

 

Our Response –We added some brackets to explain these variables in Table 3 (see pages 13-14).

 

Main comment 6: For the interpretation, please do not overstate the effects based on the statistical significance of the coefficients. Please, look and use the size of the estimated marginal effects, i.e., the estimated probability increase/decrease caused by a unit change of the underlying variable.

 

Our Response – We thanked the Reviewer for the suggestion. We revised accordingly (see line 483, 495, 505, 506, page 15, and line 515, 521, 538, 544 page 16, and line 554, 568, 575, 582, 586, page 17, and 596, 606, 623, page 18, and 634, 642, 653, 659, 667, 668, page 19, and 679, 684, 691, 695, 710, page 20, and 725, 733, 738, 747, 755, page 21, and 760, 762, 774, 780, 789, 797, page 22, and 805, page 23).

 

Minor points

 

Minor point 1: Page 1, line 35-37: “The Gini index shows that income inequality in Thailand is the highest in Southeast Asia (Bird et al. 2011). The index changes between 1988 and 2017 from 0.487 to 0.365, despite a declining poverty rate over the period (WB 2019).” The last sentence is worthy of additional explanation or an inside, why this is the case.

 

Our Response – We thanked the Reviewer for the comment and suggestion. We added more information (see lines 36-38, page 1).

 

Minor point 2: Page 13, 541: “Note: *, **, *** indicate significance at 10%, 5%, and 1% levels, respectively.” Why is this note attached to Table 1, with the descriptive statistics? If the Table contains also a battery of say, mean tests, it should be mentioned… Besides, I did not find any significant marks in Table 1...I see that some of the tests of equality of means are performed in Table 2. I suggest removing Table 2 and in the text describing Table 1, make a note that for each of the characteristics presented in Table 1, authors rejected the null hypothesis of equality of the means on the level of 1%. Similarly for Table 3. It does not bring any extra information, the results of the tests (all giving the same results) can be summarized in the text. However, the authors should describe the testable hypothesis (i.e., which variables were used to test equality, which gives the number of degrees of freedom. Also, why they use this test and not say any test of homogeneity of the distribution).

 

Our Response – We thanked the Reviewer for the comments and suggestions. We addressed the reviewer’s point by point as follows:

 

2.1 “Page 13, 541: “Note: *, **, *** indicate significance at 10%, 5%, and 1% levels, respectively.” Why is this note attached to Table 1, with the descriptive statistics? If the Table contains also a battery of say, mean tests, it should be mentioned… Besides, I did not find any significant marks in Table 1.”

 

Our Response –We removed “Note: *, **, *** indicate significance at 10%, 5%, and 1% levels, respectively,” in Table 1 (see page 12).

 

2.2 “I suggest removing Table 2 and in the text describing Table 1, make a note that for each of the characteristics presented in Table 1, authors rejected the null hypothesis of equality of the means on the level of 1%.”

 

Our Response –We removed Table 2 and make a note in Table 1 (see line 462-463, page 12). We added more information in this test about null hypothesis and why we use this test (see lines 427-429, page 10).

 

2.3 “Similarly for Table 3. It does not bring any extra information, the results of the tests (all giving the same results) can be summarized in the text.”

 

Our Response – We removed Table 3. We added more information and summarized this test (see lines 434-436, page 10).

 

2.4 “However, the authors should describe the testable hypothesis (i.e., which variables were used to test equality, which gives the number of degrees of freedom. Also, why they use this test”

 

Our Response – We thanked the Reviewer for the suggestion. We addressed these in comments 2.2 and 2.3.

 

2.5 “and not say any test of homogeneity of the distribution).”

 

Our Response – We thanked the Reviewer for the suggestion. As the literature review, El-Habil (2012)[1] states that multinomial logit model does not require homogeneity.

 

[1] (El-Habil 2012) El-Habil, Abdalla M. 2012. An Application on Multinomial Logistic Regression Model. Pakistan Journal of Statistics and Operation Research 8: 271-91.

Author Response File: Author Response.docx

Reviewer 2 Report

The paper provides a logit analysis of household survey data which examines participation in village funds and savings groups for production in Thailand.  The key issues the paper needs to address are as follows:

• the motivation for the paper is very very weak - after reading the paper I have no idea why I should be interested in the findings - they seem to replicate many findings elsewhere and add little if anything to the literature - indeed two previous studies seem to have covered the same ground - so if the findings are the same what is the added value here?
• the paper is long and discusses each finding in some detail while seeming to present them both as cause and effect - so presenting some circularity of argument at times  - eg if VFs target women then is it of interest to find that FHHHs are more likely to borrow from them? so what is really at issue here? yes it tells us that the policy is to an extent successful  - if it is not policy and policy is to lend to HoH then what is going on and what do we draw from that [ I note that the proportion of FHHHs is very similar in VF and SGP as in all respondents].  
• in general the paper needs to be clearer that logits do not show causality they only show associations
• the lit review is poorly written and does not consolidate the findings from existing studies into a set of issues that motivate a new research question which can be answered by this study.  It simply lists many studies that have similar methodologies and find similar kinds of findings about participation.  so why is another such study needed?
• the methodology section seems to start with more lit review related findings  - the aspects of credit studies that are more complex are ultimately not dealt with in this study
• data- appears to be at household leve and yet some of the discussion appears to refer to individual level factors.  this requires care - does VF and SGP participation specify household heads only?  what of households with other productive earners? how would this show up (or not) in household level data?  what are the intra-household dynamics of borrowing and is the survey at all attuned to this?
• Methodology - implies that the model needs to be collectively exhaustive - this is clearly not the case as borrowing from other parts of the financial sector is not modelled

 

Points of detail:

• line 575 and in other sections  - the result does not tell us that female households "borrow more" it tells us that they are more likley to borrow as correctly construed in the rest of the para.
• findings such as age are rather less interesting if this is true of all sources of credit as is probably the case.
• discussion of economic activity - presumably the base variable is no income source?  this doesn't seem to be assessed - it is unsrprising that those without an income source can't borrow
• there is no discussion of multi-collinearity among variables such as income and expenditure
• line 643 - the paper gets into a circular argument - it suggests that non-borrowers are likely to be less financially stressed if they are borrowing, while borrowing is intended to reduce financial stress
• the issue here is the need to explore and understand relationships between household lifecycle stage and likelihood of borrowing - clearly hhs with children and higher dependency rates are more likely to need to generate higher incomes.  Yet higher dependency rates seem to be associated with lower likelihoods of borrowing. 

Author Response

Reviewer #2’s comments:

Main comments

 

Main comment 1: the motivation for the paper is very weak - after reading the paper I have no idea why I should be interested in the findings - they seem to replicate many findings elsewhere and add little if anything to the literature - indeed two previous studies seem to have covered the same ground - so if the findings are the same what is the added value here?

 

Our Response – We thanked the Reviewer for the comment. As discussed in the literature review section, studies have shown that, in Thailand, most low income and poor people can access financial services from community-based MFIs, such as VFs, and SGPs (Microfinance Services Ltd. 2013[1]; Suwaruchiporn 2016[2]). However, no studies have investigated the determinants affecting household participation in both VFs and SGPs in Thailand. It is important to understand the key factors that affect households participate in VFs and SGPs at the same time. Previous researchers have focused on the participation in VFs (Fongthong and Suriyan 2014[3]; Menkhoff and Rungruxsirivornn 2011[4]), but the participation in both VFs and SGPs remains under-researched. Therefore, this paper investigates the factors that affect Thai households to participate in VFs and SGPs. The contribution of this study is to improve credit access and the implementation of credit policies for rural households. We added more information (see lines 65-73, page 2).

 

Main comment 2: The paper is long and discusses each finding in some detail while seeming to present them both as cause and effect - so presenting some circularity of argument at times  - eg if VFs target women then is it of interest to find that FHHHs are more likely to borrow from them? so what is really at issue here? yes it tells us that the policy is to an extent successful  - if it is not policy and policy is to lend to HoH then what is going on and what do we draw from that [ I note that the proportion of FHHHs is very similar in VF and SGP as in all respondents]. in general the paper needs to be clearer that logits do not show causality they only show associations

 

Our Response – We thanked the Reviewer for the comment and the suggestion. We addressed the reviewer’s point by point as follows:

 

2.1 “The paper is long and discusses each finding in some detail while seeming to present them both as cause and effect - so presenting some circularity of argument at times” 

 

Our Response –We revised some parts to be clearer (see lines 484-492, page 15 and line 647-649, page 19)

 

2.2 “- eg if VFs target women then is it of interest to find that FHHHs are more likely to borrow from them? so what is really at issue here? yes it tells us that the policy is to an extent successful  - if it is not policy and policy is to lend to HoH then what is going on and what do we draw from that [ I note that the proportion of FHHHs is very similar in VF and SGP as in all respondents].”

 

Our Response –Our study shows that a female household head is a significant positive predictor of SGP participation at the 10% level. SGPs’ involve gathering people with different status in the village to help each other to solve their investment problems (Luxchaigul 2014[5]). Therefore, the main target is not for women. This result implies that many microfinance programs encourage women to borrow because they present a lower credit risk (Cull et al. 2018[6]; Fongthong and Suriya 2014[7]). In addition, Ouattara et al. (2020)[8] explain that women are more involved in food crop production such as vegetable, with a quick rate of return and ready markets for their produce. This then increases their profit margins leading to reduce credit default on women. Moreover, Khandker (2005)[9] found that serving the loan for women can have stronger impacts on households than serving for men. We can draw that SGPs are also pro women and this loan can help women investing in their families.

 

“2.3 in general the paper needs to be clearer that logits do not show causality they only show associations”

 

Our Response –We agree with the reviewer that multinomial logit model just only shows association not causality. We rewrite some part to make it clearer (see lines 484-492, page 15 and lines 647-649, page 19).

 

Main comment 3: the lit review is poorly written and does not consolidate the findings from existing studies into a set of issues that motivate a new research question which can be answered by this study. It simply lists many studies that have similar methodologies and find similar kinds of findings about participation.  so why is another such study needed?

 

Our Response – We thanked the Reviewer for the comment and suggestion. We revised the literature review (see line 135, page 3 to line 188, page 4).

 

Main comment 4: the methodology section seems to start with more lit review related findings - the aspects of credit studies that are more complex are ultimately not dealt with in this study

 

Our Response – We thanked the Reviewer for the suggestion. We show the aspects of credit because we would like to explain that some determinants may play other roles in explaining credit participation. Scholars investigating poor households have noted that the above determinants may play other roles in explaining credit participation. These factors may drive credit demand factors rather than the components of creditworthiness. This means that physical endowments (e.g., assets and land ownership) and human endowments (e.g., education) have a negative impact on credit participation (Doan et al. 2010)[10]. We provide the example via Khandker (2005)[11] that shows the group-lending microfinance.

 

Main comment 5: data- appears to be at household level and yet some of the discussion appears to refer to individual level factors.  this requires care - does VF and SGP participation specify household heads only?  what of households with other productive earners? how would this show up (or not) in household level data?  what are the intra-household dynamics of borrowing and is the survey at all attuned to this?

 

Our Response – We thanked the Reviewer for the comments and suggestions. We addressed the reviewer’s point by point as follows:

 

5.1 “data- appears to be at household level and yet some of the discussion appears to refer to individual level factors.  this requires care”

 

Our Response –We revised some parts to make it clearer (see lines 481-482, page 15 and lines 505-506, page 15).

 

5.2 “does VF and SGP participation specify household heads only?”

 

Our Response –VF and SGP do not specify household heads only. Everyone can access to these programs if they are eligible. However, we use the household head data to represent the household because the information only shows the household debt.

 

5.3 “what of households with other productive earners?”

 

Our Response –The information just shows the household income and household members that are working. We can identify the members who are working and those who are not working. Thus we can find the dependency ratio from this.

 

5.4 how would this show up (or not) in household level data?

 

Our Response –. Household level data shows what household members do. The data shows each member in a household.

 

5.5 “what are the intra-household dynamics of borrowing and is the survey at all attuned to this?”

 

Our Response – Our finding shows that female household heads are more likely to borrow from VFs and SGPs than male household heads. We find that the positive number of children coefficient at the 10% level indicates that if households have more children, they have a 0.50% higher probability of VF participation than other households, all other factors being constant. This implies that these households take the loans to spend for their children. In addition, during the off-farm season, most farm households engage in non-farm activities, such as petty trading and driving a motorcycle taxi (Coleman 2006)[12]. This implies that they provide a way to earn additional income (Fongthong and Suriya 2014)[13].

 

The household survey does not show directly about intra-household dynamics of borrowing. This survey does not show how to allocate the capital intra- household. It provides information about household income and expenditure.

 

Main comment 6: Methodology - implies that the model needs to be collectively exhaustive - this is clearly not the case as borrowing from other parts of the financial sector is not modelled

 

Our Response – We thanked the Reviewer for the suggestion. As discussed in the methodology section, the alternatives are collectively exhaustive. This feature means that all the possible alternatives are included in the choice set. Our study focuses on VFs and SGPs, therefore all the possible alternatives are included in the choice set mean for the case of VFs and SGPs.

 

We added more information (see lines 260-261, page 6).

 

Points of detail:

Points of detail 1: line 575 and in other sections - the result does not tell us that female households "borrow more" it tells us that they are more likely to borrow as correctly construed in the rest of the para.

 

Our Response –We have revised it (see lines 481-482, page 15).

 

Points of detail 2: findings such as age are rather less interesting if this is true of all sources of credit as is probably the case.

 

Our Response –. Our finding shows that age of the household head from all group is significant. However, just only age of the household head, that borrowing from both VF and SGP, the marginal effect is too small.

 

Points of detail 3: discussion of economic activity - presumably the base variable is no income source?  this doesn't seem to be assessed - it is unsurprising that those without an income source can't borrow

 

Our Response – We agreed with the reviewer comments.

 

We revised accordingly (see lines 532-533, page 16 and see lines 810-811, page 23).

 

Points of detail 4: there is no discussion of multi-collinearity among variables such as income and expenditure

 

Our Response –This study conducts a variance inflation factor (VIF) to test for multicollinearity. The result shows that VIF of all variables are low and not above 10. The mean VIF for all variable is 2.5. Therefore, there are no multicollinearity problem.

 

We added this information (see lines 415-417, page 10 and see lines 431-433, page 10).

 

Points of detail 5: line 643 - the paper gets into a circular argument - it suggests that non-borrowers are likely to be less financially stressed if they are borrowing, while borrowing is intended to reduce financial stress

 

Our Response –. We revised it accordingly (see lines 532-533, page 16).

 

Points of detail 6: the issue here is the need to explore and understand relationships between household lifecycle stage and likelihood of borrowing - clearly hhs with children and higher dependency rates are more likely to need to generate higher incomes.  Yet higher dependency rates seem to be associated with lower likelihoods of borrowing.

 

Our Response – We agreed with the reviewer’s suggestion. Our results find that households with higher dependency rate and with elderly are less likely to borrow from all categories. Interestingly, we found that households with higher children are more likely to borrow from VFs.

 

We added more information from the previous study. For example, Takahashi et al. (2010)[14], evaluated the impact of microcredit programs in Indonesia, explain that the poor benefit more from microcredit participation via investment in their children’s schooling. Money spent on education helps to break the poverty vicious circle. Adjei et al. (2009)[15] evaluated the role of Ghanaian microfinance programs in asset-building and poverty reduction and find that borrowers often use the loans to improve their children’s education (see line 549, page 16 to line 552, page 17).

 

We trust that these changes will meet with the satisfaction of the reviewers and editor alike. Thank you for your efforts on our behalf. We believe the manuscript has been markedly improved by your attention to it.

 

Sincerely,

The Authors

 

 

[1] (Microfinance Services Ltd. 2013) Microfinance Services Ltd. 2013. Thailand Financial Inclusion Synthesis Assessment Report, Kingdom of Thailand: Development of a Strategic Framework for Financial Inclusion. Asian Development Bank. Available online: https://www.adb.org/sites/default/files/project-document/81802/45128-001-tacr-01-0.pdf (accessed on 29 February 2020).

[2] (Suwaruchiporn 2016) Suwaruchiporn, Panida. 2016. Financial Inclusion in Thailand [PowerPoint], Available online: http://www.fsa.go.jp/en/glopac/20160519-1.pdf (accessed on 29 February 2020).

[3] (Fongthong and Suriya 2014) Fongthong, Siwaporn and Komsan Suriya. 2014. Determinants of Borrowers of the Village and Urban Community Fund in Thailand. CMUJ of Social Sciences and Humanities 1: 21-37.

[4] (Menkhoff and Rungruxsirivorn 2011) Menkhoff, Lukas, and Ornsiri Rungruxsirivorn. 2011. Do Village Funds Improve Access to Finance?: Evidence from Thailand. World Development 39: 110-22.

[5] (Luxchaigul 2014) Luxchaigul, Nat. 2014. The Effective of Sustainable Development of the Saving for Production Groups in Northeast of Thailand. Environment Management and Sustainable Development 3: 168-80.

[6] (Cull et al. 2018) Cull, Robert, Asli Demirgüç-Kunt, and Jonathan Morduch. 2018. The Microfinance Business Model: Enduring Subsidy and Modest Profit. The World Bank Economic Review 32: 221-44.

[7] (Fongthong and Suriya 2014) Fongthong, Siwaporn and Komsan Suriya. 2014. Determinants of Borrowers of the Village and Urban Community Fund in Thailand. CMUJ of Social Sciences and Humanities 1: 21-37.

[8] (Ouattara et al. 2020) Ouattara, N’Banan, Xiong Xueping, BI, Trazié Bertrand Athanase Youan, Lacina Traoré, J. K. Ahiakpa, and Odountan Ambaliou Olounlade. 2020. Determinants of Smallholder Farmers’ Access to Microfinance Credits: A Case Study in Sassandra-Marahoué District, Côte d’Ivoire. Agricultural Finance Review ahead-of-print (ahead-of-print).

[9] (Khandker 2005) Khandker, S. R. 2005. Microfinance and Poverty: Evidence Using Panel Data from Bangladesh. World Bank Economic Review 19: 263-86.

[10] (Doan et al. 2010) Doan, Tinh Thanh, John Gibson, and Mark J. Holmes. 2010. What Determines Credit Participation and Credit Constraints of the Poor Inperi-Urban Areas, Vietnam?. Munich Personal RePEc Archive, Paper No. 27509.

[11] (Khandker 2005) Khandker, S. R. 2005. Microfinance and Poverty: Evidence Using Panel Data from Bangladesh. World Bank Economic Review 19: 263-86.

[12] (Coleman 2006) Coleman, Brett E. 2006. Microfinance in Northeast Thailand: Who Benefits and How Much?. World Development 34: 1612-38.

[13] (Fongthong and Suriya 2014) Fongthong, Siwaporn and Komsan Suriya. 2014. Determinants of Borrowers of the Village and Urban Community Fund in Thailand. CMUJ of Social Sciences and Humanities 1: 21-37.

[14] (Takahashi et al. 2010) Takahashi, Kazushi, Takayuki Higashikata, and Kazunari Tsukada. 2010. The Short‐term Poverty Impact of Small‐scale, Collateral‐free Microcredit in Indonesia: A Matching Estimator Approach. The Developing Economies 48: 128-55.

[15] (Adjei et al. 2009) Adjei, J. K., T. Arun, and F. Hossain 2009. The Role of Microfinance in Asset Building and Poverty Reduction: The Case of Sinapi Aba Trust of Ghana. Manchester: Brooks World Poverty Institute.

Author Response File: Author Response.docx

Round 2

Reviewer 1 Report

The paper has been improved, yet I would recommend minor revisions, that should devote the authors' attention to the differences between statistical significance and economic significance.

Several marginal effects are equal to a small fraction of the percent. It is not likely economically significant and authors should clearly indicate it in their interpretations, what variables were economically significant.

Also, there is a clear difference between VP and SGP borrowers. It would be relevant and interesting to analyze and explain the behavior/coefficients of the category of borrowers using both channels. Is it closer to VP or SGP?

Overall, it should not be hard for the authors to respond and to implement my comments.

Cheers,

Jan

Author Response

Dear Reviewer 1,

Please find the attached files.

Regards,

Authors

Author Response File: Author Response.docx

Reviewer 2 Report

Review of revised manuscript SGs in Thailand

 

The revised version appears not to have addressed the key comments in the original review:

 

• The motivation for the paper remains weak. The sentence has been added  “The contribution of this study is to improve credit access and the implementation of credit policies for rural households.”.  Sadly this is misunderstand what the purpose of an academic paper is.  The paper might do this if it was part of a policy discussion about these issues, but the motivation in an academic context has to be about furthering our understanding of how these credit schemes work.  The paper still does not clearly add to our understanding. As before the discussion of existing literature fails to identify a gap in our understanding which the paper is seeking to fill (and is able to do with methodological rigour and effectiveness).
• There is still no discussion in the paper of the issues of causality vs correlation and indeed the circularity of the issues around targeting women and then finding that FHHHs are more likely to access the services remains.
• Issues around household level data collection and individual borrowing have not been addressed. The issues around dependency ratios might be an issue here – a clear understanding is needed of how VF and SGP memerbship relates to household headship, and whether and how the data collection does or does not capture borrowing that is not undertaken by the household head. 
• I conclude that the paper has not materially addressed the underlying problems of the original paper in terms of motivation, additionality or methodological understanding (the problem of circularity). This is highlighted again by the additional sentence in the conclusion regarding the fact that VFs CANNOT help households with a high dependency ratio.  The fact that they are not used by these households does not mean that they CANNOT if the programme design was changed.  It means that households with high dependency ratios are not receiving credit from them.  There may be a range of reasons including the fact that village committees skew allocation away from these households.  These may indeed be a result of their view of their ability to pay.   Again there is some circularity and lack of precision in the understanding of what the conclusions might in fact mean and how the system is working.

 

 

Line 260/1 – SGP and VFs are not the only sources of borrowing, therefore they are not exhaustive.  There is an informal sector and a formal sector too.   If people are borrowing from other sources then this will affect their demand from these sources.

Line 492 – the result for women accessing the services indeed appears to be circular – women are perceived as more realiable borrowers and therefore tend to be targeted.  This does not mean that the conclusion that “Our finding draws that VF is successful in responding to the needs for credit of women” is valid!  It simply means women are targeted.  Morevoer since these are FHHHs then it clearly not the case that women who might be wives have their credit needs met.  Additionally, this fails to understanding the extensive debates around targeting women with microfinance.  See eg Goetz and Sen Gupta, World Development 1996.

Author Response

Dear Reviewer 2,

Please find attached files.

Regards,

Authors

Author Response File: Author Response.docx