Determinants of Related and Unrelated Export Diversification
Round 1
Reviewer 1 Report
Title of the paper:
Determinants of related and unrelated export diversification.
Summary of the paper:
The paper analyses the determinants of export diversification using a large sample of countries observed over the 1996-2011 period. The paper explores the role played by selected domestic factors − e.g. FDI, human capital, trade openness − as determinants of export variety in the manufacturing industry, also distinguishing between related and unrelated variety. The key result, obtained through dynamic GMM estimation techniques, is that domestic, country-level factors have differential effects according to the form of export variety considered. This suggests the relevance of exploring and disentangling the nature and type of exporting activity.
General Comment:
The paper contributes to a wide cross-country literature by analysing the determinants of country-level export diversification. In doing this, it presents an interesting contribution: it introduces the concepts of related and unrelated variety to capture and disentangle the nature of countries' export structure. However, the paper does not seem ready for publication in its current form. In particular, it is not well written in terms of phrasing, although it is easy to read. It also presents several typos, as well as spelling and grammar mistakes. The literature review seems to be adequate, and it is well focused on the key issue under investigation. However, it neither discusses a series of theoretical limitations characterising the variety measures employed, and which have been underlined by recent contributions, nor considers more recent approaches to capture the variety phenomenon. Moreover, it seems that the empirical specification is rather restrictive. In fact, it does not account for a series of factors which may help explaining the variety in country-level exporting. I have few suggestions which, I hope, could help the author(s) to improve the work.
Comments:
1. Although the paper is easy to read, it presents several typos and grammar mistakes. I would suggest to check the manuscript carefully, also in terms of phrasing.
2. Please, always include the year of publication when citing within the text.
3. Please, try to make sections and sub-sections more linked and fluent.
4. Please, check the notation of equations: for example, the variable for remoteness is indicated as "Remit", although − I suppose, being a geographical distance − it is a time-invariant variable.
5. I like the idea of introducing the distinction between related and unrelated variety in the analysis of export diversification. There is a large body of literature on (related) variety, and recent contributions have emphasised the limits of the "standard" measures of relatedness proposed by Frenken et al. (2007) and based on a simple association among statistical sub-sectors. I would therefore suggest to the author(s) to discuss these limitations and drawbacks in the paper.
6. I would suggest the author(s) to test alternative definition of related variety. I explain myself. Related variety as proposed by Frenken et al. (2007) captures cognitive-based measure of diversification. This could be relevant in the context of exporting, even though other forms of relatedness may matter, e.g. technological or market-based. I would suggest the author(s) to explore this way by proposing two alternative measures of related variety, according to data availability: (i) one measure capturing technological proximity, in line with Los (2000); (ii) a second measure capturing market-based proximity, in line with Cainelli and Iacobucci (2012) and/or Cainelli et al. (2016).
7. I would suggest the author(s) to enrich their set of explanatory variables by introducing measures of overall, related and unrelated variety which capture the level of industrial diversification in a country − e.g. using employment or value added data at sector level. In fact, it could be that a high level of export diversification is driven by a highly diversified industrial structure.
8. Please, justify the inclusion of a dynamic component in the empirical equation.
9. Please, try to justify the choice of Rotterdam, New York and Tokyo as three major markets − maybe in a footnote, using some statistics.
10. Please, include geographic and/or geo-political dummy variables in the empirical model.
11. Please, include dummy variables for membership to commercial agreements in the empirical model.
12. Please, specify the set of internal instruments used for each variable (e.g. time-lagged dependent variable, etc.). Why trade openness should be treated as pre-determined rather than endogenous? For example, it would suspect a strong reverse causality in this case. Please, justify your choices.
13. Please, check for multicollinearity (e.g. through a VIF).
14. Please, replicate your analysis for the sub-samples of developed vs. developing countries, following the graphical insights reported in the paper.
15. Please, report the estimated (average) elasticities of the variables interacted. Please, disentangle the estimated coefficients of the interaction terms: which is the moderation effect of human capital on the FDI-variety relationship? Either report graphs or report a table with elasticities calculated at selected values of the moderator.
16. I would suggest to enrich your conclusions by underlining the main policy implications of your analysis. What policy makers should learn from your theoretical intuitions and empirical evidence?
References:
Cainelli G. and Iacobucci D. (2012) Agglomeration, related variety, and vertical integration, Economic Geography, 88(3), 255-277.
Cainelli G., Ganau R. and Iacobucci D. (2016) Do geographic concentration and vertically related variety foster firm productivity? Micro-evidence from Italy, Growth and Change, 47(2), 197-217.
Los B. (2000) The empirical performance of a new inter-industry technology spillover measure. In Saviotti P. P. and Nooteboom B. (Eds.) Technology and Knowledge. From the Firm to Innovation Systems, pp. 118-151. Cheltenham: Edward Elgar.
Comments for author File: 
Comments.pdf
Author Response
Reviewer 1
I would like to thank Reviewer 1 for the insightful comments. In response to the general comments, The limitations of the paper as well as the indices are presented at the very end of the conclusion. As far as the country specific export effects are concerned, those are likely to be absorbed by the country specific dummies. Below I will respond to the comments made by the Reviewer 1.
Comments:
1. Although the paper is easy to read, it presents several typos and grammar mistakes. I would suggest to check the manuscript carefully, also in terms of phrasing.
The paper has now been proofread by a professional proofreader.
2. Please, always include the year of publication when citing within the text.
I have rechecked the manuscript but I could not find any citation without year of publication.
3. Please, try to make sections and sub-sections more linked and fluent.
4. Please, check the notation of equations: for example, the variable for remoteness is indicated as "Remit", although − I suppose, being a geographical distance − it is a time-invariant variable.
Thank you for pointing it out. Error is corrected.
5. I like the idea of introducing the distinction between related and unrelated variety in the analysis of export diversification. There is a large body of literature on (related) variety, and recent contributions have emphasised the limits of the "standard" measures of relatedness proposed by Frenken et al. (2007) and based on a simple association among statistical sub-sectors. I would therefore suggest to the author(s) to discuss these limitations and drawbacks in the paper.
Limitations are already discussed at the end of the conclusion.
6. I would suggest the author(s) to test alternative definition of related variety. I explain myself. Related variety as proposed by Frenken et al. (2007) captures cognitive-based measure of diversification. This could be relevant in the context of exporting, even though other forms of relatedness may matter, e.g. technological or market-based. I would suggest the author(s) to explore this way by proposing two alternative measures of related variety, according to data availability: (i) one measure capturing technological proximity, in line with Los (2000); (ii) a second measure capturing market-based proximity, in line with Cainelli and Iacobucci (2012) and/or Cainelli et al. (2016).
Thank you for this interesting point. It is definitely an area to explore but I think it goes beyond the scope of the paper. Also one has to distinguish between relatedness and related variety. Relatedness as defined by Hidalgo et al. (2009) measures relatedness and unrelatedness within one indicator where higher values reflect higher relatedness and vice versa. Also, this measure does not necessarily reflect diversification. This is the reason why Frenken et al (2007)’s measures cannot be directly compared with other relatedness indicators. However, as I stated above, I really like the idea to compare technological proximity with market-based proximity which can be a different project. I thank the reviewer for that.
7. I would suggest the author(s) to enrich their set of explanatory variables by introducing measures of overall, related and unrelated variety which capture the level of industrial diversification in a country − e.g. using employment or value added data at sector level. In fact, it could be that a high level of export diversification is driven by a highly diversified industrial structure.
I agree that industrial classification can determine export diversification. Infact, in some papers, export diversification is used as a proxy for industrial diversification.
8. Please, justify the inclusion of a dynamic component in the empirical equation.
Dynamic component is necessary to account for the inter-temporal self-reinforcing effect of diversification. From the figures, it shows that there is high level of persistence in the dependent variables that could violate the assumptions related to the residuals.
9. Please, try to justify the choice of Rotterdam, New York and Tokyo as three major markets − maybe in a footnote, using some statistics.
Explanation is added on page 6.
10. Please, include geographic and/or geo-political dummy variables in the empirical model.
We have geographical distance as well as country specific dummies. I think these dummies absorb geographical effects in the models. Country specific dummies were not mentioned in the paper before which was a mistake. They are now mentioned.
11. Please, include dummy variables for membership to commercial agreements in the empirical model.
While this is a very interesting suggestion, there are a number of trade agreements that countries subscribe to most of them are regional. Since the dataset used in this study includes 130 countries from different regions, we may have to include several dummies which will result is loss of degree of freedom.
12. Please, specify the set of internal instruments used for each variable (e.g. time-lagged dependent variable, etc.). Why trade openness should be treated as pre-determined rather than endogenous? For example, it would suspect a strong reverse causality in this case. Please, justify your choices.
Previous research doesn’t show a clear relationship between trade openness and diversification. There are a number of reasons, both technical and conceptual, behind this finding. On technical side, since trade openness includes imports as well, increase in imports would increase the overall trade openness index ceteris paribus which may or may not have causal implications for diversification. Moreover, the reason for assuming this variable to be predetermined is that while it is easier to expect a causal relationship between openness and diversification, it is less trivial to expect diversification to cause trade openness.
13. Please, check for multicollinearity (e.g. through a VIF).
Thank you for this comment. An additional row is added under the results table to represent VIF.
14. Please, replicate your analysis for the sub-samples of developed vs. developing countries, following the graphical insights reported in the paper.
Thank you for this comment. Since GMM methodology is suited for large N, N=130 is already at the lower boundary or “large” sample size. If we split the sample, N will be smaller and that would affect the validity of results.
15. Please, report the estimated (average) elasticities of the variables interacted. Please, disentangle the estimated coefficients of the interaction terms: which is the moderation effect of human capital on the FDI-variety relationship? Either report graphs or report a table with elasticities calculated at selected values of the moderator.
Thank you for the suggestion. Figures 4 and 5 are added to show the interaction effect.
16. I would suggest to enrich your conclusions by underlining the main policy implications of your analysis. What policy makers should learn from your theoretical intuitions and empirical evidence?
Thank you for the suggestion. I have added policy implications in the conclusions.
Thank you once again.
Warm regards
Author.
Author Response File: Author Response.pdf
Reviewer 2 Report
Determinants of Related and Unrelated Export Diversification
Economies (Manuscript–214022)
Comments to the Author(s):
The paper entitled "Determinants of Related and Unrelated Export Diversification” examines the determinants of the export diversification in 130 countries. The data covers the period from 1996 to 2011.
I think that the author(s) fail to provide very details of the main problem and the mainstream research approaches; thus, there is no contribution in the paper. In other words, the author(s) do not provide a clear cue to convince readers to understand the importance of this employed study.
The comments to follow are structured according to the structure of the paper:
Abstract
1) The abstract should be shortened. There are also typos in the abstract.
Introduction
2) The author(s) presents an introduction that explains content describing arguments in the literature. However, I feel that the author did not clearly specify a research problem or at least contextualize the issue under consideration.
3) I would have liked to have seen the first paragraph of the introduction explicitly stating the research problem and then followed by the second paragraph that explicitly states the research question.
4) I suppose the key issue is the need for a motivation behind this research. Why is it important? Furthermore, in the introduction, the reader also needs to know how the paper empirically answers the research question, and what the main findings are.
5) I also feel that the introduction can be condensed. Therefore, more time can be spent on, detailing the research question, motivating for the research, explaining how the paper seeks to address the research question and the contributions of the research. In short, in the first paragraph, the author(s) should clearly state the aim of the study as well as the main research question (the main hypothesis) of the study. The current argument on the oil-exporting countries in the first paragraph looks unrelated to the rest of the introduction.
6) It is not clear whether the authors mention about the market diversification or product diversification in the introduction.
7) It is unnecessary to explain the regional aspects and consequences of the export diversification. Since the empirical analysis considers the cross-country cases there is no need to focus on the within country papers/aspects.
Literature Review
8) I think that it is important to keep linking or referring to the research objective so that the reader can get a sense of where this research fits and ultimately where it contributes to the field. This issue is missing in the paper; and therefore, the literature review section is weak.
9) In the literature review, the author(s) should explicitly answer the question "which problem still requires to be solved in the literature?" Indeed, it is difficult to understand contributions to the existing literature. The author(s) explain many papers in the literature, and this paper should clearly state the main contribution.
10) I recommend the authors should read Aleksandra Parteka’s papers for the determinants of export diversification. They can also look at the paper of Gozgor and Can (2017), which is published in the Review of Development Economics for a potential reverse causality issue for the export diversification.
11) Section 2.2 needs to be revised since it includes some wrong arguments (e.g., a cited sentence to Aghion et al. (1998) and Lucas (1988)). A similar issue can be observed in Section 2.3 in a sentence that refers to Krugman and Venables (1990).
12) Why is the FDI an important determinant of the export diversification? This issue is missing in the literature review. In addition, the effect of the human capital on economic growth is elaborated in the literature review of the paper. Instead of this, the authors should look at the effects of human capital on diversification of the export basket.
Data, Explanatory Variables, and Methodology
13) This section details the regression specification and the data employed. I think the selection of the explanatory variables looks like arbitrary. The author(s) should clearly introduce of the motivation behind the proposed models.
14) In addition, why the authors do not implement a nonlinear unit root test when they consider a nonlinear effect of income on export diversification?
15) In addition, why the empirical analysis starts at 1996? There are the IMF data for the previous periods (until 1962) at the following link:
https://www.imf.org/external/np/res/dfidimf/diversification.htm
16) Section 4.2 should go to Section 3.
17) The authors said that they have used the system GMM estimations but they have cited the papers of the difference GMM estimations.
18) The authors should cite the paper of Roodman (2009), published in the Stata Journal, for collapsing the instruments.
19) It is not clear whether the log of the index of human capital is considered in the regressions.
Results and Discussion
20) The coefficients of the explanatory variables (e.g., the per capita GDP and the population) differ slightly from the previous literature and the nuances need to be discussed. I think it is also worth referring back to the literature and how the results relate to previous studies. Are the results consistent or not?
21) Table 7 indicates that the empirical analysis ends in 2013, but the text indicates that it ends in 2011.
22) I think it would be better to divide the countries as rich and poor countries and then run the regressions. It would be nice to see whether the determinants of the export diversification changes in terms of poor and rich economies. This could be a nice robustness check.
Conclusion
23) The author(s) should also reorganize the conclusion section: First, it is needed to be referred possible policy implications or inferences related to the obtained results. Second, the author(s) should discuss whether the obtained results in this paper are in line with the previous literature (which papers are obtained similar results and which papers are not). Third, the author(s) should discuss possible directions for a further research.
General
24) Some technical writing skills should also be improved. For example, the paper of Abramovitz (1986) is missing in the references.
25) Finally, this paper needs a professional proofreading. Therefore, the authors should check the whole text, carefully.
Author Response
Dear Reviewer 2,
I would like to thank you for the
insightful comments. Most of them certainly improved the paper. I
incorporated your comments wherever I could and justified my stance
wherever I could not.
Thank you once again.
Warm regards
Author.
Reviewer 2
I would like to thank Reviewer 2 for the insightful comments. Below I will respond to the comments made by the Reviewer 2.
============================
I think that the author(s) fail to provide very details of the main problem and the mainstream research approaches; thus, there is no contribution in the paper. In other words, the author(s) do not provide a clear cue to convince readers to understand the importance of this employed study.
Thank you for the comment. I respectfully disagree with the reviewer. The main contribution of the paper lies in the fact that it distinguishes among related, unrelated and overall variety. In addition to the introduction, there is a whole section “Type of Diversification” dedicated toward the importance of differentiating between related and unrelated variety.
Abstract
1) The abstract should be shortened. There are also typos in the abstract.
Thank you for the suggestion. Abstract is shortened.
Introduction
2) The author(s) presents an introduction that explains content describing arguments in the literature. However, I feel that the author did not clearly specify a research problem or at least contextualize the issue under consideration.
3) I would have liked to have seen the first paragraph of the introduction explicitly stating the research problem and then followed by the second paragraph that explicitly states the research question.
Thank you for the comment. Eventhough the concepts of related and unrelated variety are now 10 years old, they haven’t yet become well-known. Therefore, a preamble on the importance of the concepts being studied and background story is necessary. However, I have made some modifications to fulfill this requirements as much as possible.
4) I suppose the key issue is the need for a motivation behind this research. Why is it important? Furthermore, in the introduction, the reader also needs to know how the paper empirically answers the research question, and what the main findings are.
The whole intent of the introduction was to show the need for the study and it does explains main findings in the second last paragraph. However, I made an attempt to make the question a bit more clear.
5) I also feel that the introduction can be condensed. Therefore, more time can be spent on, detailing the research question, motivating for the research, explaining how the paper seeks to address the research question and the contributions of the research. In short, in the first paragraph, the author(s) should clearly state the aim of the study as well as the main research question (the main hypothesis) of the study. The
current argument on the oil-exporting countries in the first paragraph looks unrelated to the rest of the introduction.
The argument of oil-exporting countries is important because it shows how a shock is one sector significantly slowed down their exports which is an argument in favor of diversification, especially unrelated diversification. The size of the overall introduction is now condensed.
6) It is not clear whether the authors mention about the market diversification or product diversification in the introduction.
Thank you for this point. Introduction now clearly states that.
7) It is unnecessary to explain the regional aspects and consequences of the export diversification. Since the empirical analysis considers the cross-country cases there is no need to focus on the within country papers/aspects.
Thank you for this point. I partially agree to this point. The start of the literature review section clearly states the point of incomparability. However, while cross-sectional and country papers are different in context, they are not mutually exclusive. Country papers can use the findings of cross-country studies and vice versa. The findings of country case studies may not necessarily have the power of external validity but their findings are important and should be mentioned in the review section explicitly mentioning their limitations.
Literature Review
8) I think that it is important to keep linking or referring to the research objective so that the reader can get a sense of where this research fits and ultimately where it contributes to the field. This issue is missing in the paper; and therefore, the literature review section is weak.
9) In the literature review, the author(s) should explicitly answer the question "which problem still requires to be solved in the literature?" Indeed, it is difficult to understand contributions to the existing literature. The author(s) explain many papers in the literature, and this paper should clearly state the main contribution.
Thank you for this point. Main contribution of the paper is the distinction between related, unrelated and overall variety. It is stated in abstract and introduction.
10) I recommend the authors should read Aleksandra Parteka’s papers for the determinants of export diversification. They can also look at the paper of Gozgor and Can (2017), which is published in the Review of Development Economics for a potential reverse causality issue for the export diversification.
The issue of reverse causality is mentioned multiple times in the paper and GMM methodology was used precisely to address this issue. The papers of Parteka were already duly cited in the methodology section for the same point.
11) Section 2.2 needs to be revised since it includes some wrong arguments (e.g., a cited sentence to Aghion et al. (1998) and Lucas (1988)). A similar issue can be observed in Section 2.3 in a sentence that refers to Krugman and Venables (1990).
I am unable to understand the problem here. The cited sentences are not quotations and they only reflect the findings of the cited papers. I could not find “wrong” arguments in there.
12) Why is the FDI an important determinant of the export diversification? This issue is missing in the literature review. In addition, the effect of the human capital on economic growth is elaborated in the literature review of the paper. Instead of this, the authors should look at the effects of human capital on diversification of the export basket.
Section 2.3 is dedicated to FDI. Moreover, human capital has been rarely used as a determinant of diversification which is why, in order to support the use of human capital variable, human capital and growth literature is referred to. Also, the link between development and human capital is developed through entrepreneurship and innovation this is how I develop the relationship between human capital and diversification. This is explained in section 2.2.
Data, Explanatory Variables, and Methodology
13) This section details the regression specification and the data employed. I think the selection of the explanatory variables looks like arbitrary. The author(s) should clearly introduce of the motivation behind the proposed models.
It is clearly mentioned in the econometric model section that the explanatory variables are taken from the following studies: Agosin et al., (2012), Cadot et al., (2013) and Parteka & Tamberi, (2013). Moreover, literature review section is arranged precisely to inform that the selection of variables is motivated through literature review. The justification for the inclusion of each variable is described under its respective heading.
14) In addition, why the authors do not implement a nonlinear unit root test when they consider a nonlinear effect of income on export diversification?
Time series is not long enough to consider cointegration.
15) In addition, why the empirical analysis starts at 1996? There are the IMF data for the previous periods (until 1962) at the following link:
https://www.imf.org/external/np/res/dfidimf/diversification.htm
Thank you for this source. The above mentioned source presents ready-made indices of diversification that cannot be disaggregated into related and unrelated variety.
16) Section 4.2 should go to Section 3.
Thank you for the comment. Trends are explained in section 4 because section 4 explains the construction these indices. Otherwise trends will appear first then their construction which is counter intuitive.
17) The authors said that they have used the system GMM estimations but they have cited the papers of the difference GMM estimations.
This is so because existing studies use difference GMM. System GMM was used following the findings of Roodman (2009).
18) The authors should cite the paper of Roodman (2009), published in the Stata Journal, for collapsing the instruments.
Thank you for pointing this out. Correction is made.
19) It is not clear whether the log of the index of human capital is considered in the regressions.
Yes. It is in log as explained in the Econometric model.
Results and Discussion
20) The coefficients of the explanatory variables (e.g., the per capita GDP and the population) differ slightly from the previous literature and the nuances need to be discussed. I think it is also worth referring back to the literature and how the results relate to previous studies. Are the results consistent or not?
The signs and significance of the coefficients is already compared in the paper for consistency. The differences may also arise due to the disaggregation of the types of diversification which is the main contribution of the paper.
21) Table 7 indicates that the empirical analysis ends in 2013, but the text indicates that it ends in 2011.
I think the reviewer means Table 9. The reason for this is that variety variables are available upto 2013 but some other variables are not. Which is why overall analysis is restricted to 2011.
22) I think it would be better to divide the countries as rich and poor countries and then run the regressions. It would be nice to see whether the determinants of the export diversification changes in terms of poor and rich economies. This could be a nice robustness check.
As explained in the response to Reviewer 1, dividing the sample would reduce number of countries per sample which will violate the precondition to use GMM.
Conclusion
23) The author(s) should also reorganize the conclusion section: First, it is needed to be referred possible policy implications or inferences related to the obtained results. Second, the author(s) should discuss whether the obtained results in this paper are in line with the previous literature (which papers are obtained similar results and which papers are not). Third, the author(s) should discuss possible directions for a further research.
Thank you for this point. Conclusion is modified to address this problem.
General
24) Some technical writing skills should also be improved. For example, the paper of Abramovitz (1986) is missing in the references.
Thank you for pointing this out. Error is corrected.
25) Finally, this paper needs a professional proofreading. Therefore, the authors should check the whole text, carefully.
The text is now proofread by a professional.
Author Response File: Author Response.pdf
Round 2
Reviewer 1 Report
General comment
This revised version of the paper sounds better than the original one. However, unfortunately, some of the key suggestions I made to the author(s) have been entirely overlooked, and almost none of the empirical requests have been considered. In particular, those concerning robustness tests (i) using alternative related variety measures and (ii) focusing on sub-samples of countries (developed vs. developing). I will encourage, therefore, the author(s) to work more on their analysis in order to make their results more robust and of higher interest for readers.
Specific comments
(1) Please, check again the notation (not only in the equations): the variable for remoteness is indicated as "Remit" at line 383 (page 11), as well as in the tables.
(2) The figures you report in the manuscript clearly show differences in the trends for developed vs. developing countries. Please, test for this form of heterogeneity. If you do not feel confident by splitting the sample into two sub-samples, then add also three-way interaction terms which consider FDI, openness and human capital. Then, report plots with estimated marginal effects or elasticies.
(3) Please, add confidence intervals in the plots showing the estimated effects.
Author Response
Dear Reviewer 1,
Thank you for your comments. My response to your comments is included below in red.
===================================================================
This revised version of the paper sounds better than the original
one. However, unfortunately, some of the key suggestions I made to the
author(s) have been entirely overlooked, and almost none of the
empirical requests have been considered. In particular, those concerning
robustness tests (i) using alternative related variety measures
I appreciate the comment, however, as I mentioned in my response to the previous version of the referee report, the suggested comparison cannot be done because the indices that are suggested to compare against, measure relatedness (high values shows higher relatedness and lower value suggests lower relatedness or higher unrelatedness) which is very different from the concepts of related variety and unrelated variety.
and (ii) focusing on sub-samples of countries (developed vs. developing). I will encourage, therefore, the author(s) to work more on their analysis in order to make their results more robust and of higher interest for readers.
Thank you for this comment as well. As I mentioned before, System GMM is designed for Large N and Small T. Large in this case means atleast more than 100 (some authors believe that 100 is too small for the use of GMM). If I split the sample, the number of cross-sections per analysis will go below 100 which will violate the preconditions of using GMM. In principle, I can do this exercise if Editor thinks it is necessary but the results will be unreliable because of fewer than recommended cross sections.
Specific comments
(1) Please, check again the notation (not only in the equations): the variable for remoteness is indicated as "Remit" at line 383 (page 11), as well as in the tables.
Thank you again. I corrected those now.
(2) The figures you report in the manuscript clearly show differences in the trends for developed vs. developing countries. Please, test for this form of heterogeneity. If you do not feel confident by splitting the sample into two sub-samples, then add also three-way interaction terms which consider FDI, openness and human capital. Then, report plots with estimated marginal effects or elasticies.
I am not sure how three-way interactions would add to the analysis. I find three-way interactions very hard to interpret because it is quite difficult to figure out where the effect is coming from. regarding the developing, developed country issue, please see my response above.
(3) Please, add confidence intervals in the plots showing the estimated effects.
Confidence intervals are included now in the figures.
Reviewer 2 Report
I have looked at the reply letter for my comments and suggestions and the revised version of the paper entitled: "Determinants of Related and Unrelated Export Diversification".
I still think that the selection of the explanatory variables, the starting date of the empirical analysis, and the selection of the econometric methodology are arbitrary.
Author Response
Dear Reviewer 2,
Thank you for your second round of comments. Please find my response to your comments below in Red:
================================================================
I still think that the selection of the explanatory variables, the starting date of the empirical analysis, and the selection of the econometric methodology are arbitrary.
1) Reviewer 2 had a concern in the first report that explanatory variables are selected arbitrarily. I tried to convince the reviewer that the whole literature review section is designed to justify why each of the variable is included.
2) Reviewer 2 generously indicated that IMF database has observations starting from 1960 that could increase the time-span of the analysis. I replied to this comment by stating that the suggested database does not provide three-digit exports from 1960, rather they provide diversification indices that cannot be disentangled into related and unrelated variety which is why we are bound to start from 1995. This does not mean that the choice of start of the analysis was arbitrary.
3) Reviewer 2 himself/herself pointed towards the concern of endogeniety in the estimations which is why I used GMM in the first place. All three variety measures have a clear persistence pattern shown in the plots which is why a dynamic model is needed that could also account for endogeneity. Cointegration analysis could not be used because time series is not long enough. The choice of econometric methodology is far from arbitrary.
Round 3
Reviewer 1 Report
I perfectly understand the problems related to sample size in GMM estimation. That is way I suggested to include three-way interaction terms with the developed/developing country dummy variable. The interpretation would be straightforward, as it can be derived by simply plotting the marginal effects (or the elasticities) for the two groups of countries. For example, you plot the marginal effect of FDI at different levels of human capital (Figure 4 in the paper) for the two groups of countries.
I would like to see this test. Otherwise the figures showing different patterns between the two group of countries would be unrelated with the rest of the paper.
Author Response
Dear Reviewer 1,
Thank you for clarifying your point about three way interaction. I estimated six additional regressions with a dummy for developing countries=1 and its three-way interaction with the FDI*HC and Opn*HC. All six interactions are insignificant as well as their main effects. Please find attached. I hope it clarifies the point that the estimation results do not vary for developing and developed countries.
Author Response File: Author Response.pdf
Reviewer 2 Report
I have no further suggestions.
Author Response
Thank you for your previous comments. It improved the paper alot.
