Trade Openness and CO2 Emissions: Evidence from Tunisia
Round 1
Reviewer 1 Report
Please refer to attached file.
Comments for author File: 
Comments.pdf
Author Response
Comments
1. The basic methodological approach is appropriate, and indeed somewhat similar exercises have
already been done for Tunisia. Thus, the authors’ main contributions are (i) newer data, and (ii)
the finding of an asymmetrical effect of trade openness. These are useful, but modest, additions
to our understanding of the issue.
Response:
First of all, we would like to convey our thanks to the reviewer. The provided comments are really helpful in improving the quality of this research paper.
The clarity for the use of asymmetrical effect of trade openness has been incorporated with help of three figures 1-3 and by adding description in the lines 69-90 and 333-228.
Comments
2. It would be helpful to have some more descriptive statistics before being presented with the
econometric results, perhaps in the form of a series of small graphs. The reader needs to get a
better “feel” for the problem that is being addressed.
Response:
Three figures 1-3 are added to display the trends of used variables and the observations in the trends and expected relationship are discussed in the lines 69-90.
Comments
3. Trade openness is not properly defined. It is “measured as percentage of GDP”, but it is not clear
what is being measured. Does it cover imports of goods and services? Or just merchandise? Or
imports plus exports? Is it sensitive to fluctuations in the exchange rate? It is not clear that this
is an exogenous variable, at least as measured here. Trade openness suggests a measure of
access (e.g. tariff rates, non-tariff barriers), while the measure used is simply the weight of trade
in GDP. Given that the potential asymmetrical effects are central to the storyline of this paper,
it is essential to explain this better.
Response:
Definitions of all variables (including trade openness) and justification of proxy of TO have been added in the lines 228-235.
Following [16,18] and other literature on trade opens and CO2 emissions relationship, we used the proxy of trade openness as percentage of sum of exports and imports of goods and services.
Comments
4. The paper appears to get the econometrics right, but is less successful in explaining things. Why
would we even expect an asymmetry in the effect of trade openness on CO2 emissions? Is there
a convincing explanation for the structural break in 1983? Is it really plausible that “the present
increasing economic growth of Tunisia is not harmful for environment”?
Response:
The justification for “asymmetry in the effect of trade openness on CO2 emissions” has been added in lines 69-78. The evidence of asymmetry from the trends of trade openness and CO2 emissions has been added in lines 81-90.
A justification for break year 1983 has been added in the lines 293-301.
We thank for your comment No. 5. Accordingly, new estimations showed that Tunisia in on first stage of U-shaped relationship and present increasing economic growth of Tunisia is harmful for environment. The wrong statement “the present increasing economic growth of Tunisia is not harmful for environment” has been corrected in lines 18-19, 318-321 and 359-361.
Comments
5. The problem with the finding that higher GDP is associated with lower CO2 emissions arises, I
think, because energy consumption is also included as a regressor. Conditional on energy
consumption, it may be plausible that higher GDP goes with lower CO2 emissions, but is this an
interesting conclusion? Surely GDP and energy consumption are positively related. It is not clear
that energy consumption should be included in the model.
Response:
This is really a very useful comments as using energy consumption is responsible for biased estimation of income and CO2 emission relationship which is also mentioned by Jaforullah and King [2017]. Our previous estimation was biased due to using energy consumption in the model. Therefore, we exclude the energy consumption from the model and also put its justification following this comment and Jaforullah and King [2017] in the lines (217-221). This time, we focus on the quadratic effect of income variable and trade openness variable in the model. Further following comment No. 6 of Reviewer 2, we estimate the effect of trade openness on the CO2 emissions in both linear and nonlinear ARDL models assuming both symmetry and asymmetry for comparison and robustness. New estimates are presented in table 3 in line 311 and interpretations and other all sections are changes according to the new results.
Specific Items
Comments
6. Line 80. Make it clear that what is being measured is “Granger causality”, but not necessarily
true causality.
Response:
Line 80 has been removed as per comment No. 4 of reviewer 2. Granger causality has been clearly mentioned in the Table 1 (summery of literature) in line 133 and other many places in lines 130, 134 and 156.
Comments
7. Lines 131-135. These equations have been included without proper explanation. Most of the
terms have not been defined. As a practical matter, it is not clear that most of them are needed:
when the results are discussed, the MZt and MSB and “MPT” measures seem to be ignored. If
they are important, explain them and use them; if not, leave them out.
Response:
Parameters of equations are defined this time and further clarifications have been added in the lines 244-249. In table 2, stationary has been mentioned in the all statistics individually by putting * and description of all statistics are added in the lines 277-283.
Comments
8. Line 156. The AIC has a tendency to overfit. Ng and Perron use a modified version; would that be
useful here? How many lags were used eventually?
Response:
This is absolutely right that AIC has tendency of over fitting. But, other lag selection criteria have also limitations. AIC is generally considered as a better criterion in case of small sample case as our time sample is small. So, we utilize it.
Lags are mentioned in table 3, second row, line 311.
Ng-Perron test is justifies in lines 249-252.
Comments
9. Table 2: It’s enough to show the variables, parameters, and p-values. The S.E. and t-statistics are
not very useful, which is why they are not in fact mentioned in the text; they could be omitted.
Response:
Table 2 is now table 3. In table 3, we omit the SE and t-values as per comment. This practice is also useful because reviewer 2 ask to add the linear ARDL results also. So, visibility of results is more clear and smart now.
Comments
10. Throughout the paper there are significant grammatical problems. Here is one example:
“This high response of CO2 emissions to the energy consumption is corroborated the
fact of the increasing proportion of Tunisia fossil fuels energy consumption from 80% to
89% in the 1975 to 2014 [11], during our investigated period.”
Response:
Grammar and language problems are corrected through the article this time.
Author Response File: 
Author Response.docx
Reviewer 2 Report
This research deals with the asymmetrical impacts of trade openness on CO2 emissions in Tunisia over the period of 1976-2014.
I would like to see the following improvements before making a concrete decision on the manuscript:
1. The authors seem to be using secondary referencing rather than original references, for example when they try to explain PHH they refer to reference of Mahood et al (2019) which is a wrong practice. All concepts and hypotheses should be based on primary references and in chronological order so that the researchers are acknowledged appropriately. To this extent, the impact of trade on CO2 emissions are introduced to empirical research by Halicioglu, F. (2009) “An econometric study….” Energy Policy, Vol.37, pp.1156-116 should be incorporated into the literature review section.
2. The authors should avoid using multiple references such as (4-7). No more than 3 references in per line are appropriate.
3. Page 2 lines of 46-50 should be revised as these sentences give the impression that the authors estimate the impact of Tunisia-EU trade on CO2 emissions which is no true.
4. There should be a separate and extended literature review section and a summary table of the previous empirical works in chronological order.
5. The econometric methodology should be revised since the authors provide unit root testing formulas without explaining the parameters in them.
6. The authors should also estimate the model of ARDL and compare and contrast with to the NARDL model.
7. There should be a separate data section and explaining the units of measurements of data and their source.
8. The authors should provide a footnote that they are willing to share their raw data set in Excel with those who wish to replicate the results of this study.
9. The authors use the critical values of reference (25) without any justification for it.
10. Table 10 contains Wald test results without critical values of p- values.
11. Lagged error correction term is above minus 1, explain and interpret it appropriately.
12. There are number of grammatical and syntax errors such as using percentage and % at the same or, the expression of we may… Thus, the entire manuscript requires a professional editing.
Author Response
This research deals with the asymmetrical impacts of trade openness on CO2 emissions in Tunisia over the period of 1976-2014.
I would like to see the following improvements before making a concrete decision on the manuscript:
Comments
1. The authors seem to be using secondary referencing rather than original references, for example when they try to explain PHH they refer to reference of Mahmood et al (2019) which is a wrong practice. All concepts and hypotheses should be based on primary references and in chronological order so that the researchers are acknowledged appropriately. To this extent, the impact of trade on CO2 emissions are introduced to empirical research by Halicioglu, F. (2009) “An econometric study….” Energy Policy, Vol.37, pp.1156-116 should be incorporated into the literature review section.
Response:
First of all, we would like to convey our thanks to the reviewer. The provided comments are really helpful in improving the quality of this research paper.
This comment has been cared and confirmed in the throughout referencing. Particularly, PHH is cited with reference of Copeland and Taylor (1994) in line 53. Further, the study of Halicioglu (2009) has been incorporated in lines 125-130, in table 1 line 133 and reference in lines 412-413.
Comments
2. The authors should avoid using multiple references such as (4-7). No more than 3 references in per line are appropriate.
Response:
This comment has been cared in throughout the article. Particularly, 3 references are set in line 42.
Comments
3. Page 2 lines of 46-50 should be revised as these sentences give the impression that the authors estimate the impact of Tunisia-EU trade on CO2 emissions which is no true.
Response:
New lines 43-49 have cared this issue this time.
Comments
4. There should be a separate and extended literature review section and a summary table of the previous empirical works in chronological order.
Response:
A separate section No. 2 (Literature Review) has been added in lines 114-209 and introduction section has also improved and extended with also caring comment No. 2 of Reviewer 1.
Table 1 of literature summery has been added in line 133.
Literature Review is extended by adding twelve further studies on the topic along with previous quoted studies.
New references are added in lines 395-396, 408-425 and 429-435.
Comments
5. The econometric methodology should be revised since the authors provide unit root testing formulas without explaining the parameters in them.
Response:
Unit root testing equations are more elaborated by defining each parameter in the lines 244-249.
Comments
6. The authors should also estimate the model of ARDL and compare and contrast with to the NARDL model.
Response:
The estimation of linear ARDL has been incorporated along with nonlinear ARDL estimates in table 3 in the line 311. Further, model of both linear and nonlinear ARDL are modified as per comment No. 5 of Reviewer 1. Further, interpretations and all other sections of paper have also been revised in this regard.
Comments
7. There should be a separate data section and explaining the units of measurements of data and their source.
Response:
Data has been properly defined in terms of equations in the lines 228-233 and source is mentioned in lines 234-235.
Comments
8. The authors should provide a footnote that they are willing to share their raw data set in Excel with those who wish to replicate the results of this study.
Response:
A note is added in line 371.
Comments
9. The authors use the critical values of reference (25) without any justification for it.
Response:
A justification of Kripfganz and Schneider [25] (now [37]) has been added in lines 305-308.
Comments
10. Table 2 contains Wald test results without critical values of p- values.
Response:
p-values < 0.01 in () along with chi-square value of Wald test in table 3 are corroborating that Wald test rejects the null hypothesis of symmetry and asymmetrical effects are proved.
Comments
11. Lagged error correction term is above minus 1, explain and interpret it appropriately.
Response:
The interpretation of lagged error correction term is extended in lines 340-343.
Comments
12. There are number of grammatical and syntax errors such as using percentage and % at the same or, the expression of we may… Thus, the entire manuscript requires a professional editing.
Response:
Grammatical and syntax errors have been cared throughout the article. Language of all article has also been improved this time.
Author Response File: Author Response.docx
Round 2
Reviewer 1 Report
This paper is much improved over the first version, and the econometric section looks solid and is relatively clear.
However, the paper is not yet ready for publication.
1. The English still needs serious improvement, and the authors are encouraged to work with someone with strong writing skills in English to polish it. For example, here are lines 43-47:
EKC hypothesis testing becomes a workhorse in the environment literature after the initiatives of trade liberalization in the world since 1980s. Tunisia also started the trade liberalization policies since 1980s to foster the economic growth [7]. This liberalization fosters her trade with the world and also with her geographical closed trading partners of European Union (EU). The most of Tunisian exports are comprised of manufactured items.
A better version would read something like this:
The EKC hypothesis has been a workhorse of the environmental literature since trade liberalization became more widespread in the 1980s. Tunisia also introduced trade liberalization in the 1980s to foster economic growth [7]. This liberalization helped expand its trade with the world at large, and with the geographically close trading partners of the European Union (EU). Most Tunisian exports consist of manufactured items.
2. Line 90: The claim is made that the asymmetry “is evident from the figures”. This is not clear (to me at least), and one of the two models finds no evidence of it (but I do like Figures 1-3). More generally, the argument for asymmetry is theoretically weak. Lines 69-78 are supposed to make this case, but the logic does not hold up. One is left with the impression that the authors (initially) found an empirical effect, and then wrote an article around that. But even the evidence of asymmetry is fragile, so the focus of the article has now shifted more toward a discussion of the EKC hypothesis. That’s OK, and indeed I would recommend changing the title, and focus, of the article to reflect this evolution. Make a clearer and stronger theoretical case for why we should expect to see an asymmetric effect of trade “openness” on CO2 emissions, and use the theory to predict the direction of that effect. The data can then be used to test the theory. In doing so, don’t forget that about a fifth of trade is in services (mainly tourism), and the link between this and CO2 emissions may be very different from that between manufacturing and CO2 emissions.
3. The literature summary in Table 1 is useful. Be careful not to use the term “causality” where it should be “Granger causality”. There are some oddities: Chebbi et al. claim to find bidirectional causality between trade openness and CO2 emissions, but how are we to interpret these findings? Surely higher CO2 emissions do not lead to more trade openness, at least not in any meaningful sense. My point is this: many of these time series exercises lead to results that are difficult to interpret in a sensible way.
4. The conclusions may need to be moderated a bit. Line 365 argues: the “increasing openness of [the] Tunisian economy in terms of trade is deteriorating the environment”. The results only show that Tunisian CO2 emissions may be increased; but perhaps emissions elsewhere are now lower, in which case world CO2 emissions may be essentially unaffected.
5. The recommendations may make sense, but they don’t arise from the article, and so should probably be left out.
Author Response
Reviewer 1
This paper is much improved over the first version, and the econometric section looks solid and is relatively clear.
However, the paper is not yet ready for publication.
Comments
1. The English still needs serious improvement, and the authors are encouraged to work with someone with strong writing skills in English to polish it. For example, here are lines 43-47:
EKC hypothesis testing becomes a workhorse in the environment literature after the initiatives of trade liberalization in the world since 1980s. Tunisia also started the trade liberalization policies since 1980s to foster the economic growth [7]. This liberalization fosters her trade with the world and also with her geographical closed trading partners of European Union (EU). The most of Tunisian exports are comprised of manufactured items.
A better version would read something like this:
The EKC hypothesis has been a workhorse of the environmental literature since trade liberalization became more widespread in the 1980s. Tunisia also introduced trade liberalization in the 1980s to foster economic growth [7]. This liberalization helped expand its trade with the world at large, and with the geographically close trading partners of the European Union (EU). Most Tunisian exports consist of manufactured items.
Response:
Thank you for your very valuable comments which helped a lot in polishing the present submitted version.
The recommended lines (now 42-46) are modified as per guidance.
Comments
2. Line 90: The claim is made that the asymmetry “is evident from the figures”. This is not clear (to me at least), and one of the two models finds no evidence of it (but I do like Figures 1-3). More generally, the argument for asymmetry is theoretically weak. Lines 69-78 are supposed to make this case, but the logic does not hold up. One is left with the impression that the authors (initially) found an empirical effect, and then wrote an article around that. But even the evidence of asymmetry is fragile, so the focus of the article has now shifted more toward a discussion of the EKC hypothesis. That’s OK, and indeed I would recommend changing the title, and focus, of the article to reflect this evolution. Make a clearer and stronger theoretical case for why we should expect to see an asymmetric effect of trade “openness” on CO2 emissions, and use the theory to predict the direction of that effect. The data can then be used to test the theory. In doing so, don’t forget that about a fifth of trade is in services (mainly tourism), and the link between this and CO2 emissions may be very different from that between manufacturing and CO2 emissions.
Response:
The theoretical arguments in favor of asymmetry have been added in the introduction section lines 68-80 and methodology lines 265-266 with solid references. And unclear sentences are removed this time.
Title has also been changed in lines 2-3.
Comments
3. The literature summary in Table 1 is useful. Be careful not to use the term “causality” where it should be “Granger causality”. There are some oddities: Chebbi et al. claim to find bidirectional causality between trade openness and CO2 emissions, but how are we to interpret these findings? Surely higher CO2 emissions do not lead to more trade openness, at least not in any meaningful sense. My point is this: many of these time series exercises lead to results that are difficult to interpret in a sensible way.
Response:
Table 1 and section related to literature review have been cared to write Granger causality where applicable.
Chebbi et al. (2011) found long run bidirectional causality reported in their paper in table 6. But, they did
not focus on the bidirectional causality result in their discussions. So, we replace the results of
Chebbi et al. (2011) in our table 1 and adjusted literature review accordingly.
Bidirectional causality is not our result. However, the objection of reviewer is valid. So, we moderated
the literature review in the light of this comment.
Comments
4. The conclusions may need to be moderated a bit. Line 365 argues: the “increasing openness of [the] Tunisian economy in terms of trade is deteriorating the environment”. The results only show that Tunisian CO2 emissions may be increased; but perhaps emissions elsewhere are now lower, in which case world CO2 emissions may be essentially unaffected.
Response:
This sentence is unclear so removed and last three sentences of conclusion section are also moderated in
this regard in lines 370-374.
Comments
5. The recommendations may make sense, but they don’t arise from the article, and so should probably
be left out.
Response:
Recommendations are removed from abstract and conclusion.
Author Response File: Author Response.docx
Reviewer 2 Report
This version of the manuscript is an improvement over the previous one. However, I am still concerned with some issues.
The authors should avoid using the term of "We may use, we may apply, etc."
The unit root test of Ng and Perron still needs to be explained further. For example, what does MSB, MZ, stand for? To me, this structural test should be complemented with another one to obtain more robust results.
In page 8, the alternative hypothesis of the null should be written out.
The authors should modify the sentence of " countries do not care about the pollution..."
In my previous report, comment 11 was stating that the lagged error correction coefficient obtained from the NARDL estimation was -1.4695 but in the revised version it is reported as -0.9132, how is that possible?
The authors should provide a footnote that they are willing to share their raw data set in Excel with those wish to replicate the results of this study.
I think the entire manuscript needs professional editing and polishing.
Author Response
Reviewer 2
This version of the manuscript is an improvement over the previous one. However, I am still concerned with some issues.
Comments
The authors should avoid using the term of "We may use, we may apply, etc."
Response:
Thank you for your very valuable comments which helped a lot in polishing the present submitted version.
We care this concern throughout the write-up. This time, article does not have any single sentence starting with “we may”.
Comments
The unit root test of Ng and Perron still needs to be explained further. For example, what does MSB, MZ, stand for?
Response:
A brief description for the MZa, MSB, MZt and MPT are added in lines 253-255 for clarity. We remain very brief because our focus is not to proof the superiority of modified versions over the Za, Sargent-Bhargava (SB), Zt and PT tests. So, we are just mentioning what does MZa, MSB, MZt and MPT stand for.
Comments
To me, this structural test should be complemented with another one to obtain more robust results.
Response:
Chow test is applied and added to verify the break point. Result is reported in lines 299-301.
Comments
In page 8, the alternative hypothesis of the null should be written out.
Response:
Alternative hypothesis is added in the line 278. Further, null hypothesis is also corrected from the wrong version of in line 277.
Please see in MS word file
Comments
The authors should modify the sentence of “countries do not care about the pollution..."
Response:
This sentence is replaced with a new one in lines 33-34.
Comments
In my previous report, comment 11 was stating that the lagged error correction coefficient obtained from the NARDL estimation was -1.4695 but in the revised version it is reported as -0.9132, how is that possible?
Response:
[Please note that both reviewers’ comments and our responses are attached as a MSWord file to both reviewers, once the response was uploaded after first round of review]
In first round review, reviewer 1 asked to remove energy consumption variable from the model. The comments and responses are (as it is):
““““Reviewer 1
Comment
5. The problem with the finding that higher GDP is associated with lower CO2 emissions arises, I
think, because energy consumption is also included as a regressor. Conditional on energy
consumption, it may be plausible that higher GDP goes with lower CO2 emissions, but is this an
interesting conclusion? Surely GDP and energy consumption are positively related. It is not clear
that energy consumption should be included in the model.
Response:
This is really a very useful comments as using energy consumption is responsible for biased estimation of income and CO2 emission relationship which is also mentioned by Jaforullah and King [2017]. Our previous estimation was biased due to using energy consumption in the model. Therefore, we exclude the energy consumption from the model and also put its justification following this comment and Jaforullah and King [2017] in the lines (217-221). This time, we focus on the quadratic effect of income variable and trade openness variable in the model. Further following comment No. 6 of Reviewer 2, we estimate the effect of trade openness on the CO2 emissions in both linear and nonlinear ARDL models assuming both symmetry and asymmetry for comparison and robustness. New estimates are presented in table 3 in line 311 and interpretations and other all sections are changes according to the new results.””””
In our new estimations of new models, the ECT values of two models are also new because models were changed and the ECT values were less than one. Further, removing energy consumption variable the estimations are more robust now. That why, we did not explain the reasons of “ ECT value more than one” in the response of the first round comments and simply interpreted the speed of convergence.
Comments
The authors should provide a footnote that they are willing to share their raw data set in Excel with those wish to replicate the results of this study.
Response:
Foot note is added in line 238 and page 7.
Comments
I think the entire manuscript needs professional editing and polishing.
Response:
Editing service is utilized from our English expert colleague and write-up of the paper is much improved this time.
Author Response File: Author Response.docx
Round 3
Reviewer 2 Report
The revised version of the manuscript has been improved substantially except the fact that the authors should write out the null hypothesis after equation 12 as follows: at least one of the parameters are different than zero.
Author Response
Comment:
The revised version of the manuscript has been improved substantially except the fact that the authors should write out the null hypothesis after equation 12 as follows: at least one of the parameters are different than zero.
Response: The suggested statement has been added in the lines 278-279.
