Five Decades of Productivity and Efficiency Changes in World Agriculture (1969–2013)

Round 1

Reviewer 1 Report

• This could be a good paper if expanded significantly. The subject is relevant, the information is well organized, and the methodology seems alright. The literature review is brief but adequate. Yet, the paper lacks two fundamental things. First, an appropriate description of the concepts used. Although the authors reference O’Donell and other works, the paper does not explain which are the differences between technical change, technical efficiency change, scale efficiency change, mix efficiency change, residual mix-efficiency… They should also explain how those concepts apply to the agricultural sector by providing real examples. Second, the paper lacks a deep discussion on the meaning and policy implications of its findings. It is a well-organized paper, but it lacks an interpretation that truly descent to the actual reality (or realities) of world agriculture.

Some other (minor) issues:

• The authors begin by saying that agricultural development and growth will remain one of the topmost priority agenda in the development arena. This needs justification, because is not that clear.
• It is said that “the incidence of global food price hike that occurred during the end of the last decade was fuelled by declining agricultural productivity” … yet the referenced paper does not support that claim.
• The positive effects of increased agricultural productivity on the process of economic development and on the fight against poverty are repeatedly mentioned. However, no mention is made to the issue of prices (i.e. the fact that TFP gains in agriculture usually translate into lower prices rather than into higher incomes). This is an old but still ongoing debate (see, for instance, N. Konig “The failure of agrarian capitalism”), so perhaps it deserves a brief mention… mostly because great TFP gains in agriculture have varied policy implications. In fact, the final section (“conclusions and policy implications”) should be expanded significantly.
• The range of the y axis on figure 1 should be around 0.6 – 1.5 (there are no values above 1.5).

Author Response

Changes made following the comments of the Reviewer 1:

• This could be a good paper if expanded significantly. The subject is relevant, the information is well organized, and the methodology seems alright. The literature review is brief but adequate. Yet, the paper lacks two fundamental things. First, an appropriate description of the concepts used. Although the authors reference O’Donell and other works, the paper does not explain which are the differences between technical change, technical efficiency change, scale efficiency change, mix efficiency change, residual mix-efficiency… They should also explain how those concepts apply to the agricultural sector by providing real examples.

REPLY: We have described the concepts more elaborately in the methodology section citing Chris O'Donnell’s paper (line# 157 to 189 in page#4 & 5). We have used additional figure (Figure 1 in page#5) and equations (8 & 9 in page#4) to describe those.

• Second, the paper lacks a deep discussion on the meaning and policy implications of its findings. It is a well-organized paper, but it lacks an interpretation that truly descent to the actual reality (or realities) of world agriculture.

REPLY: We have restructured the section and proposed four specific policies (line# 393 to 413 in page#18).

• The authors begin by saying that agricultural development and growth will remain one of the topmost priority agenda in the development arena. This needs justification, because is not that clear.
  REPLY: We have modified the sentence mentioning importance of agriculture in food security, employment and rural development.
• It is said that “the incidence of global food price hike that occurred during the end of the last decade was fuelled by declining agricultural productivity” … yet the referenced paper does not support that claim.
  REPLY: We thank the reviewer for thorough review. We have corrected this.
• The positive effects of increased agricultural productivity on the process of economic development and on the fight against poverty are repeatedly mentioned. However, no mention is made to the issue of prices (i.e. the fact that TFP gains in agriculture usually translate into lower prices rather than into higher incomes). This is an old but still ongoing debate (see, for instance, N. Konig “The failure of agrarian capitalism”), so perhaps it deserves a brief mention… mostly because great TFP gains in agriculture have varied policy implications. In fact, the final section (“conclusions and policy implications”) should be expanded significantly.

REPLY: Thanks for mentioning this important dimension. In the Introduction, we have added additional literature (e.g. Alston et al. 1995, Irz et al. 2001; Koning, 2002) mentioning the effect on price and farm income (please refer to line#53 to 63 in page#2). We have also addressed the issue in the Conclusions and Policy Implications section (fourth policy, line# 411 to 413 in page#18).

• The range of the y axis on figure 1 should be around 0.6 – 1.5 (there are no values above 1.5).
  REPLY: We did scaling of all the graphs.

In summary, we have conducted a major revision in order to accommodate most of your comments to the best of our abilities. Therefore, we now believe that the revised version is suitable for publication.

Reviewer 2 Report

The paper investigates an interesting and important topic. The authors apply a recently introduced method and make some contributions to the literature examining total factor productivity (TFP) at macro level, especially in the field of global TFP growth examination.

The motivation and research question are clear, the description of methodology is also appropriate.

I see only two weaknesses of the paper.

First, the authors examine many different countries, so technological heterogeneity is certainly an important issue, but the authors didn’t consider this. There are a lot of methods to consider technological heterogeneity and there are many application in the context of agricultural efficiency/productivity analysis; E.g. cluster analysis ((Baráth-Fertő, 2017); latent class models (Alvarez-del Corral, 2010; Sauer-Paul, 2013) or random parameter models (Cechura et al., 2017; Baráth et al., 2020). I suggest, at least to mention in the paper the issue of technological heterogeneity (e.g as a limitation or future direction of the paper) and these methods and extend your references with the suggested papers.

Second, although the authors making some comparison with results of other papers, I would suggest to extend a bit the discussion with some more comparison , eg the results for the USA from the O’ Donnel paper (O'Donnell, 2012); the results for Europe with (Baráth-Fertő, 2017) etc.

References

Alvarez, Antonio and Julio del Corral "Identifying different technologies using a latent class model: extensive versus intensive dairy farms." European Review of Agricultural Economics, Vol. 37, (2010) pp. 231-250.

Baráth, L. and I. Fertő "Productivity and Convergence in European Agriculture." Journal of Agricultural Economics, Vol. 68, (2017) pp. 228-248.

Baráth, Lajos, Imre Fertő and Heinrich Hockmann "Technological differences, theoretical consistency, and technical efficiency: The case of Hungarian crop-producing farms." Sustainability, Vol. 12, (2020) pp. 1147.

Cechura, Lukas, Aaron Grau, Heinrich Hockmann, Inna Levkovych and Zdenka Kroupova "Catching up or falling behind in European agriculture: The case of milk production." Journal of Agricultural Economics, Vol. 68, (2017) pp. 206-227.

O'Donnell, Christopher J "Nonparametric estimates of the components of productivity and profitability change in US agriculture." American Journal of Agricultural Economics, Vol. 94, (2012) pp. 873-890.

Sauer, Johannes and Catherine J Morrison Paul "The empirical identification of heterogeneous technologies and technical change." Applied Economics, Vol. 45, (2013) pp. 1461-1479.

Author Response

Changes made following the comments of the Reviewer 2:

The paper investigates an interesting and important topic. The authors apply a recently introduced method and make some contributions to the literature examining total factor productivity (TFP) at macro level, especially in the field of global TFP growth examination.

The motivation and research question are clear, the description of methodology is also appropriate.

I see only two weaknesses of the paper.

• First, the authors examine many different countries, so technological heterogeneity is certainly an important issue, but the authors didn’t consider this. There are a lot of methods to consider technological heterogeneity and there are many application in the context of agricultural efficiency/productivity analysis; E.g. cluster analysis ((Baráth-Fertő, 2017); latent class models (Alvarez-del Corral, 2010; Sauer-Paul, 2013) or random parameter models (Cechura et al., 2017; Baráth et al., 2020). I suggest, at least to mention in the paper the issue of technological heterogeneity (e.g as a limitation or future direction of the paper) and these methods and extend your references with the suggested papers.

REPLY: At the end of the methodology section we have addressed the issue of technological heterogeneity citing the appropriate literature (line#222 to 237, page#7). Due do data unavailability we could not address the issue in our analysis. But we have highlighted importance of future research in this direction.

• Second, although the authors making some comparison with results of other papers, I would suggest to extend a bit the discussion with some more comparison , eg the results for the USA from the O’ Donnel paper (O'Donnell, 2012); the results for Europe with (Baráth-Fertő, 2017) etc.

REPLY: We did the comparisons with the literature in the result section (line# 345 to 347 and line#356 to 358).  

• In summary, we have conducted a major revision in order to accommodate most of your comments to the best of our abilities. Therefore, we now believe that the revised version is suitable for publication.

Reviewer 3 Report

This paper proposes to use Färe-Primont index to analyze world agriculture productivity and its components over a period of 45 years. Once data has been prepared, the implementation of this approach is facilitated by availability of Dpin software. The most delicate part of the work then consists in presenting the results, given the large number of values obtained.

The presentation of the methodology used poses no problems. The addition of figures illustrating what is actually measured (distances to different production frontiers) would have made it easier to understand. Although important, DEA programs could be put in a technical appendix. They add nothing to the understanding of calculated measures. Authors must precise what is the meaning of a TFP measure of 0.20, of technical efficiency of 0.90, and so on.

A little more room should be given to the discussion of the choices made regarding the definition of production technology. Why such a choice when other studies favor other definitions of agricultural technology?

The agricultural production function involves eight output and six input variables. The definition of agricultural production technology at such a disaggregated level raises the following questions:

• The way capital is defined seems unusual, especially if classical rules defined by OECD are used (see OECD, 2001). Indeed, authors consider machinery, or, more precisely, ``Total horse power of all the agricultural machinery. ‘’ This choice therefore deserves to be explained and justified in relation to existing literature.
• The total number of output and input variables, 14, is quite large. DEA type estimators are nonparametric estimators, and therefore subject to curse of dimensionality (see Dariao and Simar, 2007), which means they become very inaccurate except for sample sizes much larger than that used in the article (104 countries x 45 years). What are the consequences on the indices calculated using the approach proposed in O'Donnell's various papers? In other words, how precisely are the indices calculated?
• Such extensive disaggregation of agricultural production generates, by definition, zero values for certain outputs in certain countries. Does this have an impact on the calculated measures?

The presentation of results is quite confusing. On the one hand, an editing effort is more than necessary to correctly number the figures and their reference in the text. On the other hand, these figures are difficult to read. Consider, for instance, Figure 1. This figure conveys information on quantities that have not been defined before in the text: dTech, dOse… Time must then be spent in interpreting the evolution of each of these quantities. Then, reading of the following figures will be easier.

Miscellaneous:

• Line 122: [33]).
• Line 153: unreadable notation
• Numbering of figures. Where are Figures 2,3,5, and 7?
• Numbering of Tables in Appendix: T able A1 instead of Appendix Table 1 and so on.

References

Daraio, C., Simar, L. (2007). Advanced Robust and Nonparametric Methods in Efficiency Analysis. Springer.

OECD (2000). Measuring Capital. OECD Publications Services, Paris, France.

Author Response

Changes made following the comments of the Reviewer 3

• This paper proposes to use Färe-Primont index to analyze world agriculture productivity and its components over a period of 45 years. Once data has been prepared, the implementation of this approach is facilitated by availability of Dpin software. The most delicate part of the work then consists in presenting the results, given the large number of values obtained.

The presentation of the methodology used poses no problems. The addition of figures illustrating what is actually measured (distances to different production frontiers) would have made it easier to understand. Although important, DEA programs could be put in a technical appendix. They add nothing to the understanding of calculated measures. Authors must precise what is the meaning of a TFP measure of 0.20, of technical efficiency of 0.90, and so on.

REPLY: We have added a figure (Figure 1 in page#5) so that the readers have a better understanding.

We have shifted the DEA section to appendix.

We have also tried to explain the values of the efficiency measures more elaborately (line# 250 to 264 in page#7).

• A little more room should be given to the discussion of the choices made regarding the definition of production technology. Why such a choice when other studies favour other definitions of agricultural technology?

REPLY: Thank you for your comment. The production technology in this FPI approach do not need any specification of the technology, which is an advantage. We have added an additional sentence in the text: “In other words, it does not need specification of any functional form of the underlying production technology, e.g., Cobb-Douglas or a more flexible translog, which is essential in a parametric approach.”  

• The agricultural production function involves eight output and six input variables. The definition of agricultural production technology at such a disaggregated level raises the following questions:
• The way capital is defined seems unusual, especially if classical rules defined by OECD are used (see OECD, 2001). Indeed, authors consider machinery, or, more precisely, ``Total horse power of all the agricultural machinery. ‘’ This choice therefore deserves to be explained and justified in relation to existing literature.

REPLY: We did not imply machinery as capital stock. It is another form of input used for farm power requirements. One can think of this as the level of mechanization at best. Capital stock estimation will be a huge task and actually may not reflect any true values. Avila and Evenson (2010) constructed two indexes named the Invention-Innovation Capital and Technology Mastery Capital. The Invention-Innovation Capital Index was constructed based on two indicators: agricultural scientists per unit of cropland and R&D as a percentage of GDP, to measure the adaptive invention and innovative capacity. But these information are not available for a number of countries and also do not cover the time period covered in our study.

Reference:

Avila, A.F.D.; Evenson, R.E. Total Factor Productivity Growth in Agriculture: The Role of Technological Capital. In Handbook of Agricultural Economics; Pingali, P.L., Evenson, R.E., Eds.; Academic Press: Burlington, The Netherlands, 2010; pp. 3769–3822.

• The total number of output and input variables, 14, is quite large. DEA type estimators are nonparametric estimators, and therefore subject to curse of dimensionality (see Dariao and Simar, 2007), which means they become very inaccurate except for sample sizes much larger than that used in the article (104 countries x 45 years). What are the consequences on the indices calculated using the approach proposed in O'Donnell's various papers? In other words, how precisely are the indices calculated?

REPLY: We thank the reviewer for raising this critical methodological issue. The general rule of thumb for dimensionality is max  (Bogetoft and Otto, 2010) and in our case with 8 outputs and 6 inputs, the requirement of . But since our  is 104 for each year, and with a panel of 45 years, k is actually 4680, we are confident that the curse of dimensionality does not affect our estimates (line#436 to 439 in page#19).  

• Such extensive disaggregation of agricultural production generates, by definition, zero values for certain outputs in certain countries. Does this have an impact on the calculated measures?

REPLY: Since we are dealing with large dataset, some manipulation was required which are explained elaborately in the appendix (line# 441 to 462 in page # 20 & 21). We have also mentioned this in the main text so that the readers do not miss it (line# 245 to 247 in page#7).

• The presentation of results is quite confusing. On the one hand, an editing effort is more than necessary to correctly number the figures and their reference in the text. On the other hand, these figures are difficult to read. Consider, for instance, Figure 1. This figure conveys information on quantities that have not been defined before in the text: dTech, dOse… Time must then be spent in interpreting the evolution of each of these quantities. Then, reading of the following figures will be easier.

REPLY: We are sorry that we made some mistakes here that created the confusions. We have corrected the figure numbering. We have removed dTECH and other short forms from the figures since they are not mentioned in the text.

Miscellaneous:

• Line 122: [33]).
• Line 153: unreadable notation
• Numbering of figures. Where are Figures 2,3,5, and 7?
• Numbering of Tables in Appendix: T able A1 instead of Appendix Table 1 and so on.

REPLY: We did necessary correction based on the above four observations. 

• In summary, we have conducted a major revision in order to accommodate most of your comments to the best of our abilities. Therefore, we now believe that the revised version is suitable for publication.

Round 2

Reviewer 1 Report

The authors have done a very good job and I think the paper is ready for publication.

Author Response

The authors have done a very good job and I think the paper is ready for publication.

REPLY: Thank you very much for your complements and decision. Much appreciated.

Reviewer 3 Report

Referee report on manuscript 793216: ``Five Decades of Productivity and Efficiency Changes in Word Agriculture’’

Minor comments and typos:

• First paragraph of introduction is too long. It should be divided according to the principle: Only one idea in one paragraph.
• For instance, what is the link between the presentation of the literature on productivity measurement and what is developed previously, in line 60?
• Line 23: “The i-th firm FPI score in period t…”
• Line 134: Link between the following equations and the previous text?
• Possible repetitions between the paragraphs from line 135 to line 172.
• Equations defining OTE up to RME are given two times: in text (Eqs. (6) to (9)) and Figure 1. Maybe, the best place for these equations is in Figure 1.
• Line 182: Use of DPIN 3.0 could be mentioned in a footnote.
• Lines 184 and following: More details should be given about outputs and inputs in text, not in appendix. Authors must add a section on data including a presentation of outputs and inputs, a motivation of their choices (for instance why they are using machinery instead of capital), and a description of their sample.
• Line 224: “These argues…”
• Line 226: What are GR technologies?
• Line 236: “The author included 120 countries in the analysis and …”
• Last paragraph in page 6 and first ones in page 7: All works mentioned there could be summarized in a Table giving their main characteristics: years studied, number of countries in the sample, chosen method for measuring TFP, measured TFP rate of growth…
• Page 8: An in-depth analysis of the potential determinants of differences between regions of the world would have been very interesting
• Conclusion: Are there some limitations to the paper? Nothing is said about the environmental aspects of agricultural production or about the sustainability of the growth rates of TFP.

Authors did not clarify the issue of the curse of dimensionality. Even their sample is large, their DEA estimates will be quite imprecise with so many outputs and inputs.

Author Response

Thank you very much for your second round of comments. We confirm that we have addressed all of your major points in this revised version including a reanalysis of the data with reduced number of inputs and outputs to demonstrate robustness and stability of our original 8 output 6 input model. Details are presented below against each point:

• First paragraph of introduction is too long. It should be divided according to the principle: Only one idea in one paragraph.

For instance, what is the link between the presentation of the literature on productivity measurement and what is developed previously, in line 60?

REPLY: Thank you for this response. We have now shortened the paragraphs in the introduction section. Prior to discussing productivity analysis (i.e., line 61 onward), we have tried to mention the traditional path for agricultural development which was founded on the theory of comparative advantage aggressively promoted by the World Bank since second world was. We did not want to delve too much into this but wanted to touch on that to accommodate a comment from Referee#1. Therefore, it may seem out of place as we moved quickly to our main theme, i.e., productivity analysis.

• Line 223: “The i-th firm FPI score in period t…”

REPLY: The sentence has been changed to The i-th firm’s FPI score in period t…

• Line 134: Link between the following equations and the previous text?

REPLY:  Thank you for your response. I believe you were referring to series of equations on efficiency measures. After rearranging everything in response to your comments about repetition of equations and definitions, this is now resolved.

• Possible repetitions between the paragraphs from line 135 to line 172.

REPLY: We have checked the part and cannot find any repetition. Actually, some of the efficiency measures are in this section reiterated to enhance clarification. Also we have removed the texts regarding OSME and (Eq 7), since it is not used anywhere.

• Equations defining OTE up to RME are given two times: in text (Eqs. (6) to (9)) and Figure 1. Maybe, the best place for these equations is in Figure 1.

REPLY: Sorry for the repetition. We have removed the equations from the text and referred these to Figure 1.

• Line 182: Use of DPIN 3.0 could be mentioned in a footnote.

REPLY: MDPI does not allow footnotes, hence we are unable to carry out this suggestion.

• Lines 184 and following: More details should be given about outputs and inputs in text, not in appendix. Authors must add a section on data including a presentation of outputs and inputs, a motivation of their choices (for instance why they are using machinery instead of capital), and a description of their sample.

REPLY: Following the suggestion, the appendix table A1 and manipulation technique are brought in the main text (line#184 to 207).

• Line 224: “These argues…”

REPLY: Changed to “These argue….” (line#245)

• Line 226: What are GR technologies?

REPLY: We have provided a brief explanation in bracket Green Revolution technologies (i.e., a combination of high yielding varieties of rice/wheat/maize, inorganic fertilizers with supplementary irrigation and drainage controls) line# 247

• Line 236: “The author included 120 countries in the analysis and …”

REPLY: Sorry for the mistake and thanks for the thorough review. We have replaced the sentence as - The author included 26 Latin American and Caribbean … (Line#258)

• Last paragraph in page 6 and first ones in page 7: All works mentioned there could be summarized in a Table giving their main characteristics: years studied, number of countries in the sample, chosen method for measuring TFP, measured TFP rate of growth…

REPLY: Following the suggestion we have inserted a table (Table 2 in line # 269) citing the important literature in the format you have suggested.

• Page 8: An in-depth analysis of the potential determinants of differences between regions of the world would have been very interesting

Conclusion: Are there some limitations to the paper? Nothing is said about the environmental aspects of agricultural production or about the sustainability of the growth rates of TFP.

REPLY: We have now admitted limitation of our work and both of these issues are addressed in lines # 391 to 395. We have also cited the work of Herdt and Lynam (1992) to support the statement.

• Authors did not clarify the issue of the curse of dimensionality. Even their sample is large, their DEA estimates will be quite imprecise with so many outputs and inputs.

REPLY: Thank you for reminding us this topic. This time, in addition to explaining that curse of dimensionality is not an issue with literature evidence, we actually reanalyzed the data with reduced number of outputs and inputs just to prove that the estimates we have presented are not going to be imprecise. We reduced the outputs to 5 and inputs to 5, i.e., 10 instead of 14 variables. We have converted the livestock input into equivalent HP and added it to the machinery input. And all four crops previously expressed in values (fruits, vegetables, cash and oilseeds) are aggregated into one cash crop output. The new estimated TFP is almost similar to that of the original one with very little variation. We explained all this including presented a few Figure A1 in Appendix (Line # 411 to 419). Therefore, we are confident that since we are using large sample, the curse of dimensionality does not pose any problem in our analysis. Therefore, we have retained our original 14 variable matrix, i.e., 8 outputs and 6 inputs in our paper, which is one of our main contributions to the literature.

In summary, we have addressed all of your major points in this revised version, and strongly believe that the paper is now suitable for publication.