Evaluation summary and metrics: “Banning wildlife trade can boost demand for unregulated threatened species”
The authors gave a detailed response to each of the specific points raised by the evaluators. We present these interstitial responses below. (Recall, the overall evaluation metrics and predictions and and the evaluation manager’s summary can be found here).
The evaluations are positive overall. However, both evaluators suggest that more detail about the SDID method would help the reader understand the process and interpret results. In addition, the process engaged in by the authors to select spillover and control species would benefit from more clarity, and it would be good if the authors could clarify how they decided that spillover effects would not also affect the trade of control species.
We are grateful for the positive opinion from you and the reviewers. We have updated the description of the SDID method with reference to key literature. We have also replied to the comments concerning the species selection process and reported results of several sensitivity analyses to support our findings. Please see below the responses to the comments and relevant revisions.
Note from the Unjournal: As per our policies, the authors were not made aware of the evaluators identities of the response, even when, as in this case, the evaluator chose to sign their name.
Kubo et al assessed the possible impacts of wildlife trade bans on non-target species using an online auction dataset spanning 10-years. The authors demonstrated spillover effects in the form of increased trade volume involving closely related species. The spillover effects differed between the three broad groups studied, leading the authors to posit that spillover effects may differ as a function of demand for the banned species, as well as the availability of legal alternatives in the market. Overall, I thought that this is an interesting and topical paper that provides important support for anecdotes of unintended negative outcomes from trade bans. I was also intrigued by the authors’ application of synthetic difference-in-diifferences (SDID) which seemed a potentially powerful method for assessing the broad effects of policy decisions.
Despite my general appreciation of this work, I feel that the evidence supporting the authors’ overarching conclusion was not presented with sufficient clarity. This is because the modelling approach is fairly advanced, yet the details provided were too scant.
Thank you for your positive opinion and helpful comments on the manuscript. We have now updated the description of the method. Please see below for details.
The most important component of this study, in my opinion, lies in the authors’ selection of “spillover” and “control” species, as I expect this to be highly influential on SDID outcomes. For “spillover” species, I recommend the authors better justify their selection by explaining, from a buyer’s point-of-view, why these would be realistic alternatives. The authors provide strong justification for giant water bugs (i.e., same market name), but not for salamanders and freshwater fish. “Spillover” species for the latter two were close-relatives, which could be a reasonable choice if the authors cite evidence to establish the logic that underlie a potential buyer’s decision to choose phylogenetically close alternatives in the event of a ban. As these are likely to be kept as pets, perhaps other species traits (e.g., appearance, size) that may not necessarily be linked to phylogeny may be more important? To clarify, I do not believe that the authors’ approach is wrong. I do, however, suggest the authors better explain their selection process.
We appreciate your helpful comment. Since most studies exploring substitutability in wildlife trade have focused on traditional Chinese medicine (e.g., Lockwood Doughty et al. 2021; Moorhouse et al. 2022), it is not possible to conclude that phylogenetic closeness is the most important factor to judge substitutability of online wildlife (products) for consumers. However, we believe that a phylogenetic closeness is an objective index to describe similarities with the original banned species. That is, although other traits can be the indexes as you mentioned, this study supposes that the closer the distance, the greater the likelihood of substitution than the farther the distance. Also, as the Note S2 in the SI has explained, the mis-identification does not overestimate the spillover impacts. It means that, even from the conservative view, the results of this study show that there are spillover impacts of banning wildlife trade on the trade of non-banned species. We have updated the sentence in Discussion by following your suggestion; the text now reads as follows:
As previous studies exploring substitutability in wildlife trade have mainly focused on traditional Chinese medicine49, it is not possible to judge how well these criteria capture the similarities between banned and potential spillover species to consider substitutability. However, these procedures concerning criteria selection do not overestimate the spillover impacts even if there is misassignment concerning species in the SDID models (see Note S2 in the SI Appendix for details). (Ln 281-287)
Relatedly, “control” units were defined as “trades in the same categories as banned species, excluding potential spillover species” (Page 12, Paragraph 2). This is too vague for readers to follow and potentially replicate. I could not deduce what the term “categories” refer to. The identities of top control units were detailed in Fig S2 and Fig S4, but the texts were in Japanese (Fig S2) or too small to read (S4). From what I could tell, some of the control units were congeners of the banned salamander species and selected “spillovers”. I therefore wondered about how phylogenetic relatedness of “spillover” species were ranked and how the authors decided that spillover effects would not also affect the trade of “control” species.
We have added the explanation of “categories” and updated the text around “categories”. The text reads as follows:
Traded categories, such as amphibian, were given by the auction website (i.e., https://auctions.yahoo.co.jp/list1/0-all.html). (Ln 265-266)
The control units were the trades in the same categories given by the auction website as those of the banned species, excluding those of the potential spillover species. (Ln 291-292)
Also, we have updated Fig S2 and S4 to enlarge the species names. As we have replied to the comment above (i.e., #1-2), the misidentification might underestimate but not overestimate the impacts of policy change (see the response to your comment #1-2 and Note S2 for details).
With my admittedly limited understanding of SDID, I am also wondering if the issues regarding “spillover” and “control” species selection could have been averted if the authors use an unrelated group of animals (e.g., turtles) to parameterise their synthetic controls, assuming this group was not subject to similar bans. This may also help overcome the potential issue of any spillover effects in the currently selected “control” units which could obfuscate the estimation of DiD values. If the appeal of SDID was the allowance for differences in trend between intervention and control groups before the ban, do control units need to be close relatives of the spillover species?
The SDID has unique advantages by combining attractive features of difference-in-differences (DID) and synthetic control (SC). The SDID reweights and matches pre-exposure trends to make the control time trend parallel to (policy) treated unit. Therefore, completely unrelated species are not likely to be controls because selecting unrelated species does not help to develop (artificial) parallel trends to explore causal impacts. For details, see Arkhangelsky et al. (2021).
I appreciate the novelty of applying SDID, but I am concerned that there is insufficient context to ease comprehension if this work were to be submitted to journals with a broader readership. I think the description of Eq. 1 as a method to solving the “minimisation problem” epitomises my concern. I could be in the minority, but “minimisation” is not a term I encounter frequently in my reading. Therefore, I did not initially understand why there was a “minimisation problem” that had to be solved, much less understand how to solve it. I suggest the authors provide a brief explanation about what SDID (or even DiD) achieves in simpler terms (e.g., assess the effects of interventions by comparing observed outcomes against predicted outcomes representing non-intervention).
I liked the figures presented in this paper. In particular, I appreciate the clean aesthetics of the plots presented here. However, figures depicting outcomes of SDID in the main text and the supplementary section can be difficult to decipher without additional details about the application of SDID (or even of DID and SC). Without prior knowledge, the captions and text do not provide sufficient information about what the readers should look out for in the plots on the left side of Figures 3, S3, and S5. For instance, the caption mentions “arrows” indicating estimated effects, but the arrows are difficult to see on the plot. Moreover, I recommend the authors include additional information about the vertical lines representing ban enforcement, as well as the significance of trend lines representing post-ban averages and the SDID synthetic control, respectively. This will make it easier to understand what the “estimates” in plots on the right of Figures 3, S3, and S5 signify. Relatedly, the captions specify that plots on the right of these figures represent “estimates concerning trade volumes of each taxon”. In my understanding, these should instead refer to the estimated spillover effects of the ban? If my interpretation is correct, the labelling of a 0 value for estimates (i.e., vertical broken line) as “Trade (n)” is quite confusing. I recommend the use of more precise descriptions in the plot and captions.
We have revised the paragraphs in 4.2 Identification strategy. To avoid ambiguity, we have used the equations with a brief explanation and with reference to key literature (Arkhangelsky et al. 2021; Kranz 2022).
We have also updated the explanation of each figure. In particular, we have deleted “Trade (n)” on the right column of all relevant figures and used “Estimated spillover effects of the ban” to avoid the confusion. Thanks for your suggestion.
I appreciate the concise nature of the paper. The authors did a good job of providing key information butI believe that there is some room for improvement. First, some context about the volume or relative importance of online auctions as a platform for trading in animals could help readers better understand the significance and applicability of findings to the wider wildlife trade.
Second, the authors provide additional information about the relevant policies in the methods section, but this information may be better placed in earlier parts of the text to avert confusion about focal species selection.
Third, I believe that the argumentation leading to the authors’ conceptualisation of spillover effects (Fig. 5) can be further developed. The authors argue that spillover effects may be diluted when more alternatives are available in the market, but they do not explain what “alternatives” mean in the context of the wildlife (e.g., pet) trade. The text (page 6) assumes that animals in the “freshwater taxon” were potential alternatives to the golden venus chub, while animals within the “salamander taxon” were potential alternatives to the Tokyo salamander. These assumptions imply that potential buyers are unlikely to consider taxonomically distant animals as alternatives to banned species, yet I am unaware of supporting studies/papers. I recommend the authors provide additional justification for this assumption, preferably by citing relevant literature.
Thank you for your helpful suggestion. First, we have updated the introduction to include some examples of online wildlife trade while keeping the narrative and conciseness. The text now reads as follows
The recent rise in internet usage and the move of online trade can fuel wildlife trades through online e-commerce platforms and social media, increasing the risk of illegal and/or unsustainable trade in a wide variety of wildlife, for example, fish, reptiles, and mammals 7, 8, 9, 10. (Ln 21-24)
Second, we appreciate your suggestion; however, we have refrained from providing details of (relevant) policies in the introduction to make the narrative simple. Instead of that, we hope our modification of the approach which was noted above (Response to your comment #1-1) has resolved the challenge of species selection.
Third, to the best of our knowledge, there have been no empirical studies working on alternatives in the context of pet trade; however, as replied to your comment above (#1-2) some studies have discussed substitutability in the context of traditional Chinese medicine (e.g., Lockwood Doughty et al. 2021; Moorhouse et al. 2022), which might support that wildlife with more similarities may be more likely to become a substitute/alternative products. See our responses to your comment #1-2 for further explanation.
Finally, there were several instances of imprecise or unclear writing. I list these below, along with some suggestions for the authors’ consideration:
Page 2: “It activated the underground market” suggests that underground markets only came into existence when CITES regulations came into effect. Perhaps consider revising to “These regulations coincided with a growing underground market”.
We have carefully read the literature and confirmed the narrative. To highlight the risk of unintended/negative consequences based on the previous findings, we have remained this text here.
Page 2: “Even a few empirical studies have focused on introducing trade ban policies on banned species” is a confusing sentence. Consider revising to “A small number of empirical studies focus on quantifying the effects of trade bans, but the focus was on species that were the targets of the ban”.
By considering your comments and to make the narrative more straightforward, we have updated the paragraph. The text now reads:
However, most studies investigating the effect of trade ban policies have focused solely on banned species12, while the scarcity of comprehensive datasets on wildlife trade have resulted in limited policy evaluation studies20. (Ln 34-36)
Page 7: Two sentences about exotic species trade and native species policies in developed countries were quite confusing to read. I recommend editing the sentences to “An increase in exotic species trade can increase overexploitation risk in source countries and lead to population declines unless appropriate management is implemented. Developing source countries may struggle to cope with the additional management needs as they often struggle to implement robust natural resource governance”.
We have updated the sentence according to your suggestion. The text now reads as follows:
An increase in exotic species trade can increase overexploitation risk in source countries and lead to population declines unless appropriate management is implemented27. Source countries, often developing countries, may face difficulties in meeting additional management needs because of the trade ban, as they often struggle to implement robust natural resource governance4. (Ln 121-125)
4) Page 9: “evidence regarding cross-country spillovers” seems to be a very serious issue but no citations were provided to help readers learn more about it. I recommend citing the relevant sources.
Sorry for this confusion. This sentence is supported by our findings; to the best of our knowledge, there have been no studies working on spillovers caused by wildlife trade ban regulations. We have replaced evidence with findings to avoid the confusion.
5) Page 9: “We suggest the development of a database comprising banned and non-banned species” is a vague statement that may cover all known species. I recommend the authors be more specific, perhaps by narrowing the statement down to species known to be in the trade.
Thank you for this comment. To narrow it down, we have noted traded/tradable species here (Ln 181). However, we believe (ideally) it is worth monitoring all known species; therefore, applying modern technologies such as machine learning would be essential (see Ln 166-178 in Discussion for example).
In conclusion, I believe that this is a very promising study with an important, policy-relevant message. However, the paper needs to be revised for clarity. In particular, additional details about the study’s modelling approach will help improve reader comprehension and strengthen the authors’ argument about the significance of spillover effects from trade bans.
We appreciate your helpful comments and positive evaluation. We have now updated the details of modeling parts with the following key references. See the 4.2 Identification strategy section for details.
A generally well-written/reasonably well-argued paper addressing an important implication of the wildlife trade - the indirect, and often incidental effect of trade bans on the sale of species that may be sought in markets by buyers as alternatives. The paper argues that bans on the trade of species of conservation concern has spillover effects into the trade of closely related species to meet market demand - the premise is straightforward and often talked about but there have not been that many studies to my knowledge that explicitly tests such as hypothesis. Well done to the authors for putting this together.
The paper provides a timely case study of the nature and broader consequences of trade bans in the context of the wildlife trade, and why these consequences need to be more closely looked at after their implementation. Data presentation and analytical framework based on the application of SDID appears sound - and the authors have also gone on to conduct sensitivity analyses.
I would recommend the publication of this paper with some minor revisions, including tightening the language at many parts of the paper for clarity and coherence, and also more caveats for the (public) dataset used.
Principal claim - trade bans have spillover effects into the sale of species not specifically targeted by the ban per se. The paper tests this hypothesis and found what I would consider to be reasonable evidence/support (although the volume of the species sold are relatively small in my view). That said, I haven’t seen many studies that have investigate the causality of policy changes on trade of specific species, so I find it interesting to see this being demonstrated here.
My confidence on the claims made - 70-75%%.
Analytical approach is sound and novel (this is the first time i have seen the use of SDID to this sort of analyses), but i would recommend more explicit recognition of the limitations that would come with such a dataset (do you think there is leakage, sale of the banned species through alternative markets). Blanket bans can drive several types of outcomes in the trade of wildlife, and in many parts of the world where governance is weak, there is bought to be leakage into the black market (so what is reported formally may not fully capture the scale of trade) - this would need to be made clever.
Ideally, it would be good to explore such patterns for a large suite of species (and species that are trader in high volumes) but i appreciate that this may not always be realistically possible.
We appreciate your positive opinion and helpful feedback. For our responses to your comments, please see below.
P2: More background to the online wildlife trade should be given in the intro - for context setting. Suggest to provide examples of species and species groups popular in the online trade. I find that the intro currently reads very generically, and not particularly informative at this stage.
We appreciate your comment. We have added some examples of online wildlife trade while keeping our narratives. The text now reads:
The recent rise in internet usage and the move of online trade can fuel wildlife trades through online e-commerce platforms and social media, increasing the risk of illegal and/or unsustainable trade in a wide variety of wildlife, for example, fish, reptiles, and mammals 7, 8, 9, 10. (Ln 21-24).
P6: Interested to see how you derived these numbers for the alternative taxa to be traded. Please provide citations. Also hard to define what is ‘subsitutable’ - in the eyes of buyers, although one reasonable position is to provide lists of closely related species.
We apologize for this confusion. We have identified the spillover impacts in each taxon which the banned species belongs to. We also note that selecting three species might be arbitrary; however, the procedure is objective and reproductive by three steps (see Methods for details). In particular, since application of Google Trends and other social media usage reveal human demand and interest in the context of wildlife trade and others (e.g., Ladle et al. 2016 ; Di Minin et al. 2018 ; Padhi & Pati 2017 ), we believe that the usage of Google Trends in this study is more directly associated with spillover mechanisms, and as noted in the Method section, if no candidate species were available after considering the two criteria for each banned species, the three phylogenetically closest species in the market were selected using the phylogeny with reference to iNaturalist (https://www.inaturalist.org/) and OneZoom (https://www.onezoom.org/). Please see the Methods section for details. Furthermore, we have also conducted several sensitivity analyses, for example, focusing on a species in each taxon instead of three species (Table S4 in the SI).
P8: Sounds more like you are providing policy and management recommendations, than ‘implications’. Lots of recommended steps provided here - do make sure they are well substantiated- and backed by sources
We have carefully updated the paragraph with references according to the latest findings. The text now reads as follows:
The findings have several important insights into consequences. First, demand caused by the trade ban can reach non-banned threatened species in countries where the ban is implemented. For example, a substitute for the giant water bugs Cybister chinensis is designated as vulnerable in the Japanese Red List and is extinct in five prefectures26. This effect potentially harms conservation of many legally tradable threatened species currently sold online by driving up demand for the alternatives, which can activate wild harvest. This concern is not limited to the country implementing trade bans. Spillover effects are not necessarily limited to native species; they are also applicable to exotic species, such as giant water bugs, from other countries. An increase in exotic species trade can increase overexploitation risk and lead to population decline unless appropriate management is implemented27. Trade regulations regarding native species in developed countries can burden developing countries with additional management needs as these countries often struggle to implement robust natural resource governance4. In addition to overexploitation risk, the imports and breeding of exotic species stimulated by the policy change raise concerns about the potential impacts of invasive species if exotic pet species are (un)intentionally released27, 28. Moreover, the more often exotic species in closely related taxa to the banned species are imported, the higher the risk of disease outbreaks that could affect the banned species29. Considering these concerns, we highlight the possibility of the existence of a feedback loop in which trade bans may accelerate biodiversity conservation threats. Trade bans can stimulate demand for non-banned species, triggering new regulations, and stimulate demand for other non-banned species, thus creating a self-perpetuating loop. Indeed, the commercial trade of Cybister chinensis, for which we detected the positive spillover effects from the big water bug in this study, has been prohibited since January 2023. As these processes can occur with other feedback challenges like Anthropogenic Allee Effect13, these multiple feedback loops may accelerate human-induced extinction of both banned and non-banned species6, 30. (Ln 113-137)
P8: Has there been any examples where a species has been substituted in a formal/management-driven way?
The spillover species included water bugs imported from other countries such as Brazil; however, to the best of our knowledge, there have been no species in a formal/management-driven way.
P8: How do you recommend that the monitoring be done, and how many taxa can you effectively monitor, to determine the nature and direction of these market shifts?
As we discussed (Ln 166-178), we believe that the latest computer science technology can be utilized to identify traded species and cover a variety of regulated and unregulated species in multiple taxa. With the database, the latest statistical and/or prediction models can help identify the determinants and predict future market shifts. For details, please see the following studies as examples (Hino et al. 2018, Stringham et al. 2021, Kulkarni & Di Minin 2023).
P8: Can cut away the usual discussion about how biodiversity conservation is afflicted by the lack of funds. Its well known, and does not add a lot to your discussion.
We have carefully read the sentence and improved the narrative to avoid the usual discussion although we have kept the text relevant to the lack of funds because it is worth highlighting here that budget constraint has limited monitoring effort. The text now reads:
Specifically, our findings insist that the monitoring should be extended to unregulated threatened species. However, as demonstrated by the case of Japan, where the conservation budget has increased only marginally while the number of protected species has been increasing38, it is essential to enhance monitoring efficiency and management capacity as conservation practices have faced substantial financial shortages39. There are several ways to improve monitoring efficiency and management capacity. (Ln 168-178)
P9: Do you have a good reasoning to want to pursue collaboration beyond CITES? Could CITES provide the umbrella for these collaborations? I find the last bits of the discussion to be rather general, and not much of a value-add.
Thank you for your comments. This might be also relevant to the above narrative. To develop the new framework to collaborate with countries beyond CITES, we need further effort, which seems not feasible under the current resource constraints. Since we acknowledge the current framework, we have kept this narrative here. (Ln 179-185)
P11: What steps did you take to manually check and confirm the species’ names?
Two or three people checked the same data (i.e., spreadsheet) to confirm whether the extracted species name was correct. If they had any conflicts concerning the names and/or there were any ambiguity, we checked html files and/or description of the wildlife products.
P1: ‘Regulations on the harvest and use of natural esosurces’
Response: We have updated it by following your suggestion.
P1: ‘knee-jerk’ probably captures what you mean more clearly.
Response: We do not fully understand your comment here but we believe the relevant words have been deleted and the updated paragraph is now satisfactory.
Response: We have deleted it to follow the updated findings and narratives.
P1: What kind of modern technologies? Vague.
Response: We have replaced it with machine learning technologies according to the main in our manuscript.
P2: Not clear what you mean by ‘distribution’ - of the species afflicted? Please re-word
Response: We have replaced it with species distributions.
P6: I think ‘show’ is a better word.
Response: We have replaced it with “show”.
P6: Side or incidental effects
Response: We have replaced it with “side”.
P6: ‘has important policy implications’
Response: We have replaced it with “show”.
Figure 5: left panel vs right panel
Response: We have deleted it.
P7: be specific - harms conservation by driving up demand (for the alternatives) - and increased wild harvest
Response: Thanks for the suggestion; we have updated the sentence as follows:
This effect potentially harms conservation of many legally tradable threatened species currently sold online by driving up demand for the alternatives, which can activate wild harvest. (Ln 116-118)
P7: accentuate threats to biodiversity - the trade bans can also effectively undermine the conservation of species and species groups
Response: We have replaced it with “accelerate threats to biodiversity conservation”.
P8: accelerate declines of species
Response: We have replaced it with “population decline”.
Response: We have replaced it with “demand”.
P10: increased volumes of harvest for the trade
Response: We have kept our words according to the policy in Japan.
P10: What is the reasoning why these three species were chosen?
Response: We do not provide whole reasonings; but, as noted in the manuscript, the government picked them as representative of threatened species in semi-natural ecosystems by considering their ecological status and characteristics (Ln 210-213).
P11: each taxon banned
Response: We have updated the whole paragraph in 4.2 Identification strategy. We hope the updated text is now satisfactory.