Skip to main content

Advertisement

Log in

Trade, Resource Use and Pollution: A Synthesis

  • Published:
Environmental and Resource Economics Aims and scope Submit manuscript

Abstract

We develop a two-sector dynamic model of trade and the environment. Both resource-good and manufacturing sectors generate environmental burdens that can adversely affect the environment in a country over time. The resource-good sector is environmentally sensitive, with its productivity suffering in line with declines in the environmental stock. Countries can be categorized into two types: one in which the resource-good sector is more environmentally harmful than the manufacturing sector, and another in which the opposite is true. At every point in time, labor allocations and specialization patterns under free trade are dependent on country size, technology, household preferences, and environmental stocks. Trading countries’ types have implications for the dynamics of environmental stocks and the corresponding environmental and welfare consequences of trade. If two trading countries are of the same type, one country exports its dirtier goods and the other must export its cleaner goods; consequently, at least one country gains from trade. If the two countries are of different types, they can export their respective dirtier goods and may both lose from trade.

This is a preview of subscription content, log in via an institution to check access.

Access this article

Price excludes VAT (USA)
Tax calculation will be finalised during checkout.

Instant access to the full article PDF.

Fig. 1
Fig. 2
Fig. 3
Fig. 4
Fig. 5
Fig. 6
Fig. 7
Fig. 8

Similar content being viewed by others

Notes

  1. According to UNCTAD (2013), the share of South–South trade in total world trade increased from slightly less than 30 percent in 1995 to slightly more than 40 percent in 2012. Furthermore, the share and value added of manufacturing are usually much higher in a developing country’s exports to other developing countries than in its exports to developed countries.

  2. Managi et al. (2009) and Tsurumi and Managi (2012) empirically analyzed the sensitivity of results for the relationship between trade openness and environmental quality to differences between developing countries and OECD countries. Using panel data covering the sulfur dioxide (SO2) and carbon dioxide (CO2) emissions of 88 countries from 1973 to 2000, Managi et al. (2009) showed that trade openness increases SO2 and CO2 emissions in non-OECD countries but decreases them in OECD countries. Using data on the annual rate of deforestation for 142 countries from 1990 to 2003, Tsurumi and Managi (2012) found that an increase in trade openness increases deforestation for non-OECD countries while slowing down deforestation for OECD countries.

  3. The 2018 World Air Quality Report by IQAir AirVisual (available at https://www.airvisual.com/world-most-polluted-cities) indicates that Asian locations dominate among the 100 cities with the highest average PM2.5 (particulate matter of less than 2.5 microns in diameter) levels, with cities in India, China, Pakistan and Bangladesh occupying the top 50. The report also notes that although sources of pollution vary by region and city, common contributors include vehicle exhaust, open crop and biomass burning, industrial emissions and coal burning.

  4. Africa is rich in natural resources including mineral reserves, arable land, forests and wildlife. These natural resources are, however, under threat from human activities. For example, UNEP (2016) reported severe land degradation in sub-Saharan Africa caused by diverse factors such as unsustainable farming practices, unregulated conversions of forest and grassland to cropland, mining activities, and desertification. Nkonya et al. (2016) estimated the annual cost of land degradation at approximately 63 billion USD or 7% of sub-Saharan Africa’s total GDP.

  5. Lenzen et al. (2012) found that 30 percent of global species threats are due to international trade and pointed out that in many developed countries, the consumption of imported coffee, tea, sugar, textiles, fish, and other commodities leaves a biodiversity footprint that is larger abroad (i.e., in developing countries) than at home.

  6. 6 We characterize the environmental quality by a single index based on an idea similar to the Environmental Performance Index (EPI) constructed jointly by research centers at Yale University and Columbia University (https://epi.yale.edu). The 2020 EPI is constructed from 32 performance indicators across 11 categories, including air and water quality, biodiversity, forests, and fisheries.

  7. Besides slash-and-burn agriculture, sub-Saharan African countries are also well known for mining of non-renewable resources such as diamond, tungsten, and tin. Those mining activities are generally not affected by air and water pollution. Therefore, the environmentally sensitive sector that we consider relates to the production of renewable rather than non-renewable resources.

  8. In reality, manufacturing can be significantly affected by environmental issues. For example, poor air and water quality can damage the health of workers, and acid rain can corrode buildings and machinery, all making production more costly. The magnitude of these impacts varies depending on the amount and type of pollutants, as well as on the technologies and equipment in use. There are only two sectors in our model, and in order to formulate the difference among industries in vulnerability to environmental degradation, it is a reasonable simplification to leave one sector with its productivity independent of environmental changes.

  9. One may consider a case in which environmental degradation caused by pollution in the manufacturing sector does not influence the productivity of the resource-good sector. For example, air pollution in urban areas does not influence the productivity of marine fisheries in coastal areas. We rule out such possibilities by not considering industries that are spatially separated within a country.

  10. The Brander–Taylor model has been extended in a number of directions. Brander and Taylor (1997b) extended their original model by introducing countries that differ in their resource management regimes (open access in one country and optimally regulated in the other). Jinji (2007) analyzed the case in which two countries may differ in both their resource-management regime and relative resource abundance. By introducing land as another input and focusing on forest resources, Jinji (2006) examined a model with an endogenous carrying capacity of the resource. Copeland and Taylor (2009) developed a theory of resource management with an endogenous property rights regime by combining the original model with a simple model of moral hazard. Takarada et al. (2013) considered a model of fishery resources shared by two countries.

  11. Using a balanced panel of 732 municipalities in the Brazilian Amazon from 2000 to 2010, Faria and Almeida (2016) empirically found that as openness to trade in the Amazon increases, deforestation also increases, and that soybean and beef cattle production drives deforestation.

  12. There are a number of applications of the Copeland–Taylor model. Copeland and Taylor (1997) considered government policy that controls pollution. Focusing on transboundary pollution, Unteroberdoerster (2001), Benarroch and Thille (2001), and Suga (2007) reexamined the spatial separation result. In another direction, Kondoh (2006) and Beladi et al. (2000) applied the Copeland–Taylor model to international migration and international capital movement, respectively.

  13. The environmental stock S can be interpreted as a proxy that measures the overall capacity of the environment, similar to the EPI as mentioned in footnote 6. One may think that the level of S indicates the abundance of renewable resources such as fisheries and forests, the quality of ambient environment including air, water and land, and the assimilative/purifying capacity of the ecosystem to absorb and decompose wastes and pollutants.

  14. Clearly, \(\beta\) is endogenously determined in equilibrium at every point in time.

  15. This implies that if no economic activity occurs (\(E=0\)), the environmental stock at the steady state will be equal to K.

  16. As we assume the fixed technology for a given environmental stock, we do not consider the development of cleaner technologies, which can capture trade-induced technique effects. Therefore, we assume that E depends on the outputs, which capture the scale effect. Alternatively, E can be formulated so that it depends on inputs rather than outputs: \(E={\tilde{l}}_{f}L_{f}+{\tilde{l}}_{m}L_{m}\), where \({\tilde{l}}_{f}\) and \({\tilde{l}}_{m}\) are positive constants. The main findings of our analysis remains valid under this alternative specification.

  17. Note that our model includes the Brander–Taylor model (where \(l_{m}=0\) and \(l_{f}>0\)) and the Copeland–Taylor model (where \(l_{f}=0\) and \(l_{m}>0\)) as special cases. If \(l_{f}=l_{m}=0\), the model degenerates into a Ricardian model in the long run. To see this, note that the level of the environmental stock at the steady state is K, and thus productivity in the resource-good sector is fixed at \(A\left( K\right)\), yielding a typical Ricardian two-sector model.

  18. We do not introduce environment-related variables in the utility function to highlight the channels through production sectors. Introducing such variables only reinforces the main results.

  19. For example, the Brander–Taylor model assumes a logistic form of \(G\left( S\right)\) with the maximum greater than \(l_{m}a\); the Copeland–Taylor model assumes a linear, downward-sloping \(G\left( S\right)\) with \(G\left( 0\right) >l_{m}a\).

  20. See the Appendix for formal proofs of the lemma and propositions.

  21. This is because \(l_{f}A\left( S\right)\) measures the environmental burden of employing a unit of labor in the resource-good sector, with \(l_{m}a\) the counterpart for the manufacturing sector.

  22. If \(G\left( S\right)\) happens to pass through the intersection of \(l_{f}A\left( S\right) L\) and \(l_{m}aL\), the interval \(\Gamma\) degenerates to a point \(\Gamma =\left\{ S_{c}\right\}\), and consequently \(l_{f}A\left( S\right) =l_{m}a\) holds for all \(S\in \Gamma\). We ignore this knife-edge case.

  23. 23 Roughly speaking, an environment (or resource)-abundant country is more likely to be of the BT type, and a labor-abundant country is more likely to be of the CT type. Assume a logistic form \(G\left( S\right) =rS\left( 1-S/K\right)\), and let \(A\left( S\right) =a_{f}S\), where r and \(a_{f}\) are parameters (r is proportional to \(G\left( S\right)\) and measures the service flow from the environmental stock). Some algebra yields that the country is of BT type if \(r/L>\left( l_{f}a_{f}\right) ^{2}K/\left( l_{f}a_{f}K-l_{m}a\right)\) and is of CT type if the inequality is reversed. Interpreting r/L as a measure of the relative service flow (of the environment versus labor) as in Brander and Taylor (1998), the condition says that a country with sufficient environmental abundance (i.e., sufficiently high r/L) is of BT type and a country with sufficient labor abundance (i.e., sufficiently low r/L) is of CT type.

  24. Southeast Asian countries such as Indonesia and Latin American countries such as Brazil are typical examples of BT countries. See Strutt and Anderson (2000) and Gumilang et al. (2011) for environmental impacts of trade liberalization in Indonesia. For environmental degradation in relation to agricultural production in Brazil, see Martinelli et al. (2010).

  25. China and India are typical examples of CT countries (these countries also have large populations). By contrast, small island countries produce handmade products using primitive technology, which can be considered an example of the coexistence of “low manufacturing productivity” and “clean manufacturing technology” and thus of BT type.

  26. To see this, consider a point on the long-run PPF where \(\beta =\beta _{1}\) holds. Then, the point can be expressed by \(\left( X_{m},X_{f}\right) =\left( \left( 1-\beta _{1}\right) aL,A\left( S_{1}\right) \beta _{1}L\right)\), where \(S_{1}=S_{\infty }\left( \beta _{1}\right)\). The level of the environmental stock required to produce at the point is \(S_{1}\). The short-run PPF corresponding to \(S_{1}\) can be obtained by substituting \(S_{1}\) for S in (10), which we can verify passes though the point \(\left( \left( 1-\beta _{1}\right) aL,A\left( S_{1}\right) \beta _{1}L\right)\). Figure 3 illustrates this argument.

  27. The case of \(v<1\) can be similarly discussed.

  28. Suppose that Home is much smaller than Foreign in effective size. Home’s response to trade can be well approximated by a small open economy: it completely specializes (\(\beta =1\) or 0) as long as \(v\ne 1\). By contrast, Foreign observes little change in its labor allocation after trade liberalization (\(\beta ^{*}\approx b\)).

  29. Intuitively, when people spend more on the resource good than on the manufactured good (\(q>1\)) and when Home and Foreign differ little in size, supply would not be able to match demand if no country were devoted to the production of the resource good.

  30. To save space and avoid redundancy, details are omitted here and can be found in the Online Appendix.

  31. For a simple exposition, the \(v=z/q\) and \(v=1\) curves appear as straight lines in the figure. In general, this is not the case since v is not necessarily a function of \(S^{*}/S\).

  32. Recall that \(S_{\infty }\left( \beta \right)\) denotes the relation from Home labor allocation \(\beta\) to the steady-state level of Home environmental stock S defined by (7) and (8), and \(S_{\infty }^{*}\left( \beta ^{*}\right)\) is the counterpart for Foreign.

  33. In either case, the \({\dot{S}}^{*}=0\) curve must consist of a negatively sloped part in the (f,d) region if Foreign is of the CT type. To see this, note that the two boundary points of the \({\dot{S}}^{*}=0\) curve in the (f,d) region (one with \(v=1\) and the other with \(v=z/q=\left( 1-b\right) z/b\)) satisfy that \(S^{*}=S_{\infty }^{*}\left( b-\left( 1-b\right) z\right)\) at \(v=1\) and \(S^{*}=S_{\infty }^{*}\left( 0\right)\) at \(v=z/q=\left( 1-b\right) z/b\) (which can be obtained by substituting the values of v corresponding to the boundary points into the expression of \(\beta ^{*}\) in the (f,d) region, namely \(\beta ^{*}=b-\left( 1-b\right) z/v\)). Since \(S_{\infty }^{*\prime }>0\) holds in the CT-type Foreign, \(S_{\infty }^{*}\left( b-\left( 1-b\right) z\right) >S_{\infty }^{*}\left( 0\right)\), and thus, it is impossible for the \({\dot{S}}^{*}=0\) curve to be entirely positively sloped in the (f,d) region. See Fig. 7 for examples.

  34. See Corollary S2 in the Online Appendix for a formal proof.

  35. The TOT and green effects can be specified as follows. Let \(\left( X_{m}^{A},X_{f}^{A}\right)\) denote the production (consumption) bundle at the autarkic steady state and \(\beta _{T}\), \(\left( X_{m}^{T},X_{f}^{T}\right)\), and \(\left( C_{m}^{T},C_{f}^{T}\right)\) denote, respectively, the labor allocation, production bundle, and consumption bundle at the trade steady state. The green effect arises from an income gap between trade and autarky (measured in the autarkic price):

    $$\begin{aligned} \Delta _{Green}=P_{A}X_{m}^{T}+X_{f}^{T}-\left( P_{A}X_{m}^{A}+X_{f}^{A}\right) =\left( A\left( S_{\infty }\left( \beta _{T}\right) \right) -A\left( S_{\infty }\left( b\right) \right) \right) \beta _{T}L, \end{aligned}$$

    where the second equality follows from \(P_{A}=A\left( S_{\infty }\left( b\right) \right) /a\), \(X_{m}^{T}=a\left( 1-\beta _{T}\right) L\), \(X_{f}^{T}=A\left( S_{\infty }\left( \beta _{T}\right) \right) \beta _{T}L\), \(X_{m}^{A}=a\left( 1-b\right) L\), and \(X_{f}^{A}=A\left( S_{\infty }\left( b\right) \right) bL\). This shows that the source of the green effect is productivity change in the resource-good sector from \(A\left( S_{\infty }\left( b\right) \right)\) to \(A\left( S_{\infty }\left( \beta _{T}\right) \right)\). The TOT effect, on the other hand, arises from the gap between expenditure and income under trade (also measured in the autarkic price):

    $$\begin{aligned} \Delta _{TOT}=P_{A}C_{m}^{T}+C_{f}^{T}-\left( P_{A}X_{m}^{T}+X_{f}^{T}\right) =\left( P_{T}-P_{A}\right) \left( X_{m}^{T}-C_{m}^{T}\right) , \end{aligned}$$

    where the second equality follows from the trade balance condition \(P_{T}C_{m}^{T}+C_{f}^{T}=\left( P_{T}X_{m}^{T}+X_{f}^{T}\right)\).

  36. These results mimic those in Brander and Taylor (1998) and Copeland and Taylor (1999). This is not surprising because the two-country trade in Brander and Taylor (1998) behaves similarly to the trade between two countries of the BT type in our model, whereas that in Copeland and Taylor (1999) is similar to the trade between two countries of the CT type.

  37. Our analytical results are consistent with empirical findings. Focusing on Southeast Asia, Coxhead (2007) documented a shift in comparative advantage during 2000 to 2004 from labor-intensive manufacturing to natural resource products in the poorer, resource-abundant countries of the region (Indonesia, Vietnam, Laos, Cambodia, and Myanmar), which was largely driven by China’s economic growth and global integration. Such a shift, together with the absence of adequate environmental regulations, accelerated resource depletion in those countries, casting a shadow over their sustainable futures. On the other hand, Wang et al. (2020) found that, from 2002 to 2015, China was consistently a net exporter of embodied SO2 emissions to many countries, including those in Latin America and Southeast Asia. The net export flow of SO2 emissions increased significantly between 2002 and 2007, after China’s accession to the WTO in 2001. Moreover, Wang et al. (2017) estimated that international exports in China contributed to a nearly 20 percent increase in the national population-weighted mean PM2.5 concentration in 2007.

  38. As discussed in footnote 23, BT and CT countries are more likely to be resource-abundant and labor-abundant countries, respectively. In the traditional Heckscher–Ohlin type model without environmental externalities, both trading countries gain by exporting goods that use abundant factors intensively. However, unlike the standard Heckscher–Ohlin model, our model exhibits nonconvexity in the economy’s production set because of environmental externalities on the production side. Proposition 7 indicates that the environmental externalities may result in the failure of the traditional gains-from-trade proposition.

  39. In this respect, one could incorporate the pollution haven effect as well as strategic environmental/trade policy into the model. It would also be of great interest to introduce firm location decisions and evaluate whether the pollution haven hypothesis works to strengthen or weaken our results on the environmental consequences and welfare effects of trade. For the pollution haven argument, see, for example, Copeland and Taylor (2004) and, more recently, Cherniwchan et al. (2017).

  40. To see why, assume to the contrary that there exists \(\beta _{1}\in \left[ 0,1\right]\) such that \(S_{\infty }\left( \beta _{1}\right) \ne S_{c}\). Recalling that \({\dot{S}}=0\) holds for \(\beta =\beta _{1}\) and \(S=S_{c}\), this means that the stability condition is violated for \(\beta =\beta _{1}\) and \(S=S_{c}\): \(G^{\prime }\left( S_{c}\right) \ge l_{f}A^{\prime }\left( S_{c}\right) \beta _{1}L\); otherwise, we would rather have \(S_{\infty }\left( \beta _{1}\right) =S_{c}\). By contrast, the stability of \(\beta =\beta _{c}\) and \(S=S_{c}\) requires \(G^{\prime }\left( S_{c}\right) <l_{f}A^{\prime }\left( S_{c}\right) \beta _{c}L\), which together with \(G^{\prime }\left( S_{c}\right) \ge l_{f}A^{\prime }\left( S_{c}\right) \beta _{1}L\) gives \(\beta _{c}>\beta _{1}\). Hence, there exists \(\beta _{2}\in \left[ \beta _{1},\beta _{c}\right)\) such that \(G^{\prime }\left( S_{c}\right) =l_{f}A^{\prime }\left( S_{c}\right) \beta _{2}L\). This means that for any \(\beta \in \left( \beta _{2},1\right]\), the stability condition \(G^{\prime }\left( S_{c}\right) <l_{f}A^{\prime }\left( S_{c}\right) \beta L\), and thus, \(S_{\infty }\left( \beta \right) =S_{c}\) holds. For any \(\beta \in \left[ 0,\beta _{2}\right]\), however, the stability condition fails, and thus \(S_{\infty }\left( \beta \right) \ne S_{c}\). Then, it follows that \(\lim _{\beta \rightarrow \beta _{2}^{+}}S_{\infty }\left( \beta \right) =S_{c}\ne S_{\infty }\left( \beta _{2}\right)\), meaning that \(S_{\infty }\left( \beta \right)\) is discontinuous at \(\beta _{2}\). This contradicts the continuity of \(S_{\infty }\left( \beta \right)\).

  41. When both countries are of CT type, the environment in the country exporting the resource good is enhanced, making the country better at producing its exports. By contrast, the country exporting the manufactured good faces environmental deterioration and thus becomes better at manufacturing.

References

  • Beladi H, Chau NH, Khan MA (2000) North–South investment flows and optimal environmental policies. J Environ Econ Manag 40(3):275–296

    Article  Google Scholar 

  • Benarroch M, Thille H (2001) Transboundary pollution and the gains from trade. J Int Econ 55(1):139–159

    Article  Google Scholar 

  • Brander JA, Taylor MS (1997a) International trade and open-access renewable resources: the small open economy case. Can J Econ 30(3):526–552

    Article  Google Scholar 

  • Brander JA, Taylor MS (1997b) International trade between consumer and conservationist countries. Resour Energy Econ 19(4):267–297

  • Brander JA, Taylor MS (1998) Open access renewable resources: trade and trade policy in a two-country model. J Int Econ 44(2):181–209

  • Cherniwchan J, Copeland BR, Taylor MS (2017) Trade and the environment: new methods, measurements, and results. Annu Rev Econ 9(1):59–85

    Article  Google Scholar 

  • Chichilnisky G (1994) North–South trade and the global environment. Am Econ Rev 84(4):851–874

    Google Scholar 

  • Copeland BR, Taylor MS (1997) The trade-induced degradation hypothesis. Resour Energy Econ 19(4):321–344

  • Copeland BR, Taylor MS (1999) Trade, spatial separation, and the environment. J Int Econ 47(1):137–168

    Article  Google Scholar 

  • Copeland BR, Taylor MS (2004) Trade, growth, and the environment. J Econ Literature 42(1):7–71

    Article  Google Scholar 

  • Copeland BR, Taylor MS (2009) Trade, tragedy, and the commons. Am Econ Rev 99(3):725–749

    Article  Google Scholar 

  • Coxhead I (2007) A new resource curse? Impacts of China’s boom on comparative advantage and resource dependence in Southeast Asia. World Dev 35(7):1099–1119

    Article  Google Scholar 

  • Faria WR, Almeida AN (2016) Relationship between openness to trade and deforestation: empirical evidence from the Brazilian Amazon. Ecol Econ 121:85–97

  • Gumilang H, Mukhopadhyay K, Thomassin PJ (2011) Economic and environmental impacts of trade liberalization: the case of Indonesia. Econ Model 28(3):1030–1041

    Article  Google Scholar 

  • Jinji N (2006) International trade and terrestrial open-access renewable resources in a small open economy. Can J Econ 39(3):790–808

    Article  Google Scholar 

  • Jinji N (2007) International trade and renewable resources under asymmetries of resource abundance and resource management. Environ Resource Econ 37(4):621–642

    Article  Google Scholar 

  • Karp L, Sacheti S, Zhao J (2001) Common ground between free-traders and environmentalists. Int Econ Rev 42(3):617–648

    Article  Google Scholar 

  • Kondoh K (2006) Transboundary pollution and international migration. Rev Int Econ 14(2):248–260

    Article  Google Scholar 

  • Lenzen M, Moran D, Kanemoto K, Foran B, Lobefaro L, Geschke A (2012) International trade drives biodiversity threats in developing nations. Nature 486(7401):109–112

    Article  Google Scholar 

  • Managi S, Hibiki A, Tsurumi T (2009) Does trade openness improve environmental quality? J Environ Econ Manag 58(3):346–363

    Article  Google Scholar 

  • Martinelli LA, Naylor R, Vitousek PM, Moutinho P (2010) Agriculture in Brazil: impacts, costs, and opportunities for a sustainable future. Curr Opinion Environ Sustain 2(5–6):431–438

    Article  Google Scholar 

  • Nkonya E, Johnson T, Kwon HY, Kato E (2016) Economics of land degradation in sub-Saharan Africa. In: Nkonya E, Mirzabaev A, von Braun J (eds) Economics of land degradation and improvement – a global assessment for sustainable development. Springer, Cham

  • Panagariya A (1981) Variable returns to scale in production and patterns of specialization. Am Econ Rev 71(1):221–230

    Google Scholar 

  • Rus HA (2016) Renewable resources, pollution and trade. Rev Int Econ 24(2):364–391

    Article  Google Scholar 

  • Strutt A, Anderson K (2000) Will trade liberalization harm the environment? The case of Indonesia to 2020. Environ Resource Econ 17(3):203–232

    Article  Google Scholar 

  • Suga N (2007) An analysis of transboundary pollution and the gains from trade: reconsideration. Pac Econ Rev 12(2):225–235

    Article  Google Scholar 

  • Takarada Y, Dong W, Ogawa T (2013) Shared renewable resources: gains from trade and trade policy. Rev Int Econ 21(5):1032–1047

    Article  Google Scholar 

  • Tsurumi T, Managi S (2012) The effect of trade openness on deforestation: empirical analysis for 142 countries. Environ Econ Policy Stud 16(4):305–324

    Article  Google Scholar 

  • UNCTAD (2013) Trade and development report, 2013. United Nations, New York and Geneva

  • UNEP (2016) GEO-6 regional assessment for Africa. United Nations Environment Programme, Nairobi

  • Unteroberdoerster O (2001) Trade and transboundary pollution: spatial separation reconsidered. J Environ Econ Manag 41(3):269–285

    Article  Google Scholar 

  • Wang F, Li Y, Zhang W, He P, Jiang L, Cai B, Zhang J, Zhang P, Pan H, Jiang H (2020) China’s trade-off between economic benefits and sulfur dioxide emissions in changing global trade. Earth’s Future 8(1): e2019EF001354

  • Wang H, Zhang Y, Zhao H, Xi L, Zhang Y, Zhu W, Nielsen CP, Li X, Zhang Q, Bi J, McElroy MB (2017) Trade-driven relocation of air pollution and health impacts in China. Nat Commun 8(1):1–7

    Google Scholar 

Download references

Acknowledgments

We thank two anonymous reviewers and Co-Editor, Carolyn Fischer, for valuable comments to improve the paper. We are grateful to Eric Bond, Ngo Van Long, and Yasuhiro Takarada for insightful discussions. We also thank Kazumi Asako, Brian Copeland, Taiji Furusawa, Jota Ishikawa, David Jinkins, Hayato Kato, Yan Ma, Carol McAusland, Hiroshi Mukunoki, Jun-ichi Nakamura, Takeshi Ogawa, Hiroshi Ohta, Gianmarco Ottaviano, Thomas Sampson, Yoichi Sugita, Katsuhisa Uchiyama, Morihiro Yomogida, and seminar and conference participants at the University of Bari, University of British Columbia, Copenhagen Business School, Hitotsubashi University, Kobe University, London School of Economics, McGill University, Development Bank of Japan (DBJ) Research Institute of Capital Formation, Xiamen University, Asia Pacific Trade Seminars, Australasian Trade Workshop, Japanese Society of International Economics Meeting, IEFS Japan Annual Meeting, Summer Workshop of Economic Theory, and World Congress of the Econometric Society for helpful comments. All remaining errors are ours. We gratefully acknowledge the Japan Society for the Promotion of Science (JSPS) for financial support (Nos. 16H03617 and 20H01492). We also thank the Hitotsubashi Institute for Advanced Study (HIAS) for generous support.

Author information

Authors and Affiliations

Authors

Corresponding author

Correspondence to Akihiko Yanase.

Additional information

Publisher's Note

Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.

Supplementary Information

Below is the link to the electronic supplementary material.

Supplementary file 1 (pdf 652 KB)

Appendix

Appendix

1.1 A.1 Proof of Lemma 1

Proof of Part (i)

Assume to the contrary that the sign of \(\left( l_{f}A\left( S\right) -l_{m}a\right)\) changes as S varies within \(\Gamma\). From the continuity of \(A\left( \cdot \right)\), there exists \(S_{c}\in \Gamma\) such that \(l_{f}A\left( S_{c}\right) =l_{m}a\). Since \(S_{c}\in \Gamma\), there exists \(\beta _{c}\in \left[ 0,1\right]\) such that \(S_{\infty }\left( \beta _{c}\right) =S_{c}\), from which it follows, according to the definition of \(S_{\infty }\left( \cdot \right)\), that

$$\begin{aligned} G\left( S_{c}\right) =\left( l_{f}A\left( S_{c}\right) \beta _{c}+l_{m}a\left( 1-\beta _{c}\right) \right) L=l_{m}aL. \end{aligned}$$

This means that \({\dot{S}}=0\) holds for any \(\beta \in \left[ 0,1\right]\) as long as \(S=S_{c}\), which together with the stability condition and the continuity of \(S_{\infty }\left( \beta \right)\) further implies that \(S_{\infty }\left( \beta \right) =S_{c}\) for all \(\beta \in \left[ 0,1\right]\).Footnote 40 Hence, \(\Gamma =\left\{ S_{c}\right\}\), and thus \(l_{f}A\left( S\right) -l_{m}a=0\) holds for all \(S\in \Gamma\). This leads to a contradiction.

Proof of Part (ii)

The continuity of \(S_{\infty }\left( \beta \right)\) implies that \(\Gamma\) can be written as a closed interval:

$$\begin{aligned} \Gamma =\left[ \gamma _{\min },\gamma _{\max }\right] , \end{aligned}$$

where \(\gamma _{\min }\) and \(\gamma _{\max }\) are the left and right endpoints of the interval. It then suffices to show that S cannot escape \(\Gamma\) through the two endpoints during dynamic transitions: \({\dot{S}}\ge 0\) for any \(\beta \in \left[ 0,1\right]\) if \(S=\gamma _{\min }\), and \({\dot{S}}\le 0\) for any \(\beta \in \left[ 0,1\right]\) if \(S=\gamma _{\max }\). In what follows, we prove only the former: \({\dot{S}}\ge 0\) for any \(\beta \in \left[ 0,1\right]\) if \(S=\gamma _{\min }\). The latter can be proven similarly, and thus the proof is omitted here.

According to part (i), we proceed by considering two cases.

Case 1: \(l_{f}A\left( S\right) =l_{m}a\) for all \(S\in \Gamma\). It then follows that

$$\begin{aligned} {\dot{S}}=G\left( S\right) -l_{m}aL, \end{aligned}$$

which gives \(S_{\infty }\left( \beta \right) =\gamma _{\min }=\gamma _{\max }\) for any \(\beta \in \left[ 0,1\right]\). This implies that once the economy reaches the steady state, it remains there regardless of the labor allocation.

Case 2: \(l_{f}A\left( S\right) \ne l_{m}a\) for all \(S\in \Gamma\). Assume to the contrary that there exists \(\beta _{1}\in \left[ 0,1\right]\) such that \({\dot{S}}<0\) if \(S=\gamma _{\min }\). Let \(S_{1}=S_{\infty }\left( \beta _{1}\right)\). Then \({\dot{S}}=0\) and \(\partial {\dot{S}}/\partial S<0\) holds for \(\beta =\beta _{1}\) and \(S=S_{1}\). Hence, for sufficiently small \(\delta >0\), \({\dot{S}}>0\) holds for \(\beta =\beta _{1}\) and \(S=S_{1}-\delta\). Recalling that we have assumed that \({\dot{S}}<0\) holds for \(\beta =\beta _{1}\) and \(S=\gamma _{\min }\), there exists \(S_{2}\in \left( \gamma _{\min },S_{1}\right)\) such that \({\dot{S}}=0\) holds if \(\beta =\beta _{1}\) and \(S=S_{2}\). That is,

$$\begin{aligned} G\left( S_{2}\right) =\left( l_{f}A\left( S_{2}\right) \beta _{1}+l_{m}a\left( 1-\beta _{1}\right) \right) L. \end{aligned}$$

On the other hand, since \(S_{2}\in \Gamma\) and \(S_{2}\ne S_{1}\), there exists \(\beta _{2}\in \left[ 0,1\right]\) such that \(\beta _{2}\ne \beta _{1}\) and \(S_{2}=S_{\infty }\left( \beta _{2}\right)\):

$$\begin{aligned} G\left( S_{2}\right) =\left( l_{f}A\left( S_{2}\right) \beta _{2}+l_{m}a\left( 1-\beta _{2}\right) \right) L. \end{aligned}$$

Recalling that in this case \(l_{f}A\left( S\right) \ne l_{m}a\) for all \(S\in \Gamma\), we have \(l_{f}A\left( S_{2}\right) \ne l_{m}a\), which together with the two equations above yields \(\beta _{1}=\beta _{2}\), leading to a contradiction.

1.2 A.2 Proof of Proposition 1

According to part (i) of Lemma 1, one of the following holds: \(l_{f}A\left( S\right) >l_{m}a\) for all \(S\in \Gamma\), or \(l_{f}A\left( S\right) <l_{m}a\) for all \(S\in \Gamma\), or \(l_{f}A\left( S\right) =l_{m}a\) for all \(S\in \Gamma\). Countries satisfying the first are of BT type; countries satisfying the second are of CT type. The third knife-edge case is ignored.

1.3 A.3 Proof of Proposition 2

Proof of Part (i)

Recall that \(S_{\infty }^{\prime }\left( \beta \right) <0\) in BT-type countries and \(S_{\infty }^{\prime }\left( \beta \right) >0\) in CT-type countries. It then follows from (13) that \(\text {SMRT}<\text {MRT}\) in BT-type countries and \(\text {SMRT}>\text {MRT}\) in CT-type countries.

Proof of Part (ii)

Note that a straight line connecting \(\left( aL,0\right)\) with another point on the long-run PPF is actually the short-run PPF that corresponds to the environmental stock required for sustainable production at that point. Since the slope (absolute value) of this short-run PPF is equal to the marginal rate of transformation (from manufacturing to resource goods), the result follows immediately from part (i).

Proof of Part (iii)

Use (13) and \(\beta =1-X_{m}/aL\) to obtain

$$\begin{aligned} T_{\infty }^{\prime \prime }=\frac{d}{dX_{m}}\left( \frac{dT_{\infty }\left( X_{m}\right) }{dX_{m}}\right) =\frac{d\beta }{dX_{m}}\frac{d}{d\beta }\left( \frac{dT_{\infty }\left( X_{m}\right) }{dX_{m}}\right) =\frac{\Delta }{a^{2}L}, \end{aligned}$$

where \(\Delta \equiv \beta A^{\prime }S_{\infty }^{\prime \prime }+\beta A^{\prime \prime }S_{\infty }^{\prime 2}+2A^{\prime }S_{\infty }^{\prime }\). It follows from (9) that

$$\begin{aligned} \beta A^{\prime }S_{\infty }^{\prime \prime }=\beta A^{\prime }L\frac{\varPhi _{1}L}{\varPhi _{2}^{2}}\left( 2l_{f}A^{\prime }+\frac{\varPhi _{1}}{\varPhi _{2}}\left( l_{f}A^{\prime \prime }\beta L-G^{\prime \prime }\right) \right) , \end{aligned}$$

where \(\varPhi _{1}\equiv l_{f}A-l_{m}a\), \(\varPhi _{2}\equiv G^{\prime }-l_{f}A^{\prime }\beta L\). Similarly, we have \(\beta A^{\prime \prime }S_{\infty }^{\prime 2}=\left( \varPhi _{1}L/\varPhi _{2}^{2}\right) \varPhi _{1}A^{\prime \prime }\beta L\) and \(2A^{\prime }S_{\infty }^{\prime }=\left( \varPhi _{1}L/\varPhi _{2}^{2}\right) 2\varPhi _{2}A^{\prime }\). These together yield

$$\begin{aligned} \Delta =\frac{\varPhi _{1}L}{\varPhi _{2}^{2}}A^{\prime 2}\left( 2\frac{G^{\prime }}{A^{\prime }}-\frac{\varPhi _{1}}{\varPhi _{2}}\beta L\frac{d}{dS}\left( \frac{G^{\prime }}{A^{\prime }}\right) \right) . \end{aligned}$$

It then follows that

$$\begin{aligned} T_{\infty }^{\prime \prime }=\frac{\varPhi _{1}A^{\prime 2}}{a^{2}\varPhi _{2}^{2}}\left( 2\frac{G^{\prime }}{A^{\prime }}-\frac{\varPhi _{1}}{\varPhi _{2}}\beta L\frac{d}{dS}\left( \frac{G^{\prime }}{A^{\prime }}\right) \right) . \end{aligned}$$

Noting that \(\beta \rightarrow 0\) as \(X_{m}\rightarrow aL\), we have

$$\begin{aligned} \lim _{X_{m}\rightarrow aL}T_{\infty }^{\prime \prime }=\frac{2\varPhi _{1}A^{\prime }G^{\prime }}{a^{2}\varPhi _{2}^{2}}. \end{aligned}$$

Stability requires that \(G^{\prime }-l_{f}A^{\prime }\beta L<0\) hold for all \(\beta \in \left[ 0,1\right]\), which implies \(\lim _{X_{m}\rightarrow aL}G^{\prime }<0\) (since \(\beta \rightarrow 0\) as \(X_{m}\rightarrow aL\) and \(G^{\prime }-l_{f}A^{\prime }\beta L\rightarrow G^{\prime }\) as \(\beta \rightarrow 0\)). This means that \(T_{\infty }^{\prime \prime }\) has the opposite sign to \(\varPhi _{1}\) as \(X_{m}\rightarrow aL\), which is negative in BT-type countries but positive in CT-type countries.

1.4 A.4 Proof of Proposition 3

We derive the conditions for the respective world specialization patterns to arise, taking the environmental stocks S and \(S^{*}\) as given. As byproducts, we obtain the labor allocations \(\beta\) and \(\beta ^{*}\) and the world relative price P. Note that some results in Table 2 further use the spending ratio definition \(q\equiv b/\left( 1-b\right)\).

Pattern (f,d)

If pattern (f,d) arises, Home produces only the resource good, \(X_{f}=A\left( S\right) L\), and Foreign produces both goods, \(X_{f}^{*}=A^{*}\left( S^{*}\right) \beta ^{*}L^{*}\) and \(X_{m}^{*}=a^{*}\left( 1-\beta ^{*}\right) L^{*}\). The world relative price of the manufactured good with respect to the resource good satisfies

$$\begin{aligned} P=\frac{A^{*}\left( S^{*}\right) }{a^{*}}, \end{aligned}$$
(29)

which depends only on Foreign’s condition. In pattern (f,d), the resource good is produced in both countries, and thus the wage in Home is \(w=A\left( S\right)\) and that in Foreign is \(w^{*}=A^{*}\left( S^{*}\right)\). Since the manufactured good is supplied only by Foreign, the world market-clearing condition requires, given the same preference in the two countries, \(\left( 1-b\right) \left( wL+w^{*}L^{*}\right) =PX_{m}^{*}\). This gives Foreign’s labor allocation:

$$\begin{aligned} \beta ^{*}=b-\left( 1-b\right) \frac{z}{v}, \end{aligned}$$
(30)

where, as defined in (19) and (20),

$$\begin{aligned} v\equiv \frac{A^{*}\left( S^{*}\right) a}{A\left( S\right) a^{*}},\quad z\equiv \frac{aL}{a^{*}L^{*}}. \end{aligned}$$

Since Foreign produces both goods, we have \(\beta ^{*}>0\), which together with (30) yields \(v>\left( 1-b\right) z/b\). Recalling that a necessary condition for Foreign to export the manufactured good is \(v\le 1\), the necessary condition for pattern (f,d) to arise is therefore

$$\begin{aligned} \frac{1-b}{b}z<v\le 1. \end{aligned}$$
(31)

A sufficient condition can be obtained by excluding \(v=1\):

$$\begin{aligned} \frac{1-b}{b}z<v<1. \end{aligned}$$
(32)

Pattern (f,m)

In pattern (f,m), Home produces only the resource good, \(X_{f}=A\left( S\right) L\), and Foreign produces only the manufactured good, \(X_{m}^{*}=a^{*}L^{*}\). The wage in Home is \(w=A\left( S\right)\), whereas that in Foreign is \(w^{*}=a^{*}P\). Since both countries completely specialize, the world relative price of the manufactured good, P, is determined such that world supply equals world demand, \(b\left( wL+w^{*}L^{*}\right) =X_{f}\), which gives

$$\begin{aligned} P=\frac{\left( 1-b\right) A\left( S\right) L}{ba^{*}L^{*}}. \end{aligned}$$
(33)

For the two countries to completely specialize and Home to export the resource good, the necessary condition is \(A^{*}\left( S^{*}\right) /a^{*}\le P\le A\left( S\right) /a\), which yields

$$\begin{aligned} v\le \frac{1-b}{b}z\le 1. \end{aligned}$$
(34)

A sufficient condition can be obtained by excluding \(v=1\):

$$\begin{aligned} v\le \frac{1-b}{b}z\le 1 \text{ and } v\ne 1. \end{aligned}$$
(35)

Pattern (d,m)

In pattern (d,m), Home produces both goods, \(X_{f}=A\left( S\right) \beta L\) and \(X_{m}=a(1-\beta )L\), and Foreign produces only the manufactured good, \(X_{m}^{*}=a^{*}L^{*}\). The world relative price of the manufactured good depends only on Home’s condition:

$$\begin{aligned} P=\frac{A\left( S\right) }{a}. \end{aligned}$$
(36)

The wage in Home is \(w=A\left( S\right)\), whereas that in Foreign is \(w^{*}=a^{*}P=A\left( S\right) a^{*}/a\). Since the resource good is supplied only by Home, the market-clearing condition requires \(b\left( wL+w^{*}L^{*}\right) =X_{f}\). This gives Home’s labor allocation:

$$\begin{aligned} \beta =b\left( 1+\frac{1}{z}\right) . \end{aligned}$$
(37)

Since Home produces both goods, we have \(\beta <1\), and therefore using (37), \(\left( 1-b\right) z/b>1\). Recalling that a necessary condition for Home to export the resource good is \(v\le 1\), the necessary condition for pattern (d,m) is therefore

$$\begin{aligned} v\le 1<\frac{1-b}{b}z. \end{aligned}$$
(38)

A sufficient condition follows by excluding \(v=1\):

$$\begin{aligned} v<1<\frac{1-b}{b}z. \end{aligned}$$
(39)

Pattern (d,f)

In pattern (d,f), we have \(X_{f}=A\left( S\right) \beta L\), \(X_{m}=a\left( 1-\beta \right) L\), and \(X_{f}^{*}=A^{*}\left( S^{*}\right) L^{*}\). The world relative price of the manufactured good is given by (36). The wages are \(w=A\left( S\right)\) and \(w^{*}=A^{*}\left( S^{*}\right)\). The market-clearing condition requires \(\left( 1-b\right) \left( wL+w^{*}L^{*}\right) =PX_{m}\), which gives Home’s labor allocation:

$$\begin{aligned} \beta =b-\left( 1-b\right) \frac{v}{z}. \end{aligned}$$
(40)

It follows from \(\beta >0\) that \(v<bz/\left( 1-b\right)\), which together with \(v\ge 1\) gives the necessary condition for pattern (d,f):

$$\begin{aligned} 1\le v<\frac{b}{1-b}z. \end{aligned}$$
(41)

Again, excluding \(v=1\) yields a sufficient condition:

$$\begin{aligned} 1<v<\frac{b}{1-b}z. \end{aligned}$$
(42)

Pattern (m,f)

In pattern (m,f), we have \(X_{m}=aL\) and \(X_{f}^{*}=A^{*}\left( S^{*}\right) L^{*}\). The wages are \(w=aP\) and \(w^{*}=A^{*}\left( S^{*}\right)\). The world relative price of the manufactured good is determined by the market-clearing condition \(b\left( wL+w^{*}L^{*}\right) =X_{f}\), which gives

$$\begin{aligned} P=\frac{\left( 1-b\right) A^{*}\left( S^{*}\right) L^{*}}{baL}. \end{aligned}$$
(43)

For the two countries to completely specialize and Foreign to export the resource good, the necessary condition is \(A\left( S\right) /a\le P\le A^{*}\left( S^{*}\right) /a^{*}\), which yields

$$\begin{aligned} 1\le \frac{b}{1-b}z\le v. \end{aligned}$$
(44)

A sufficient condition can be obtained by excluding \(v=1\):

$$\begin{aligned} 1\le \frac{b}{1-b}z\le v \text{ and } v\ne 1. \end{aligned}$$
(45)

Pattern (m,d)

In pattern (m,d), we have \(X_{m}=aL\), \(X_{f}^{*}=A^{*}\left( S^{*}\right) \beta ^{*}L^{*}\), and \(X_{m}^{*}=a^{*}\left( 1-\beta ^{*}\right) L^{*}\). The world relative price of the manufactured good is given by (29). The wages are \(w=aP\) and \(w^{*}=A^{*}\left( S^{*}\right)\). The market-clearing condition requires \(b\left( wL+w^{*}L^{*}\right) =X_{f}^{*}\), which gives Foreign’s labor allocation:

$$\begin{aligned} \beta ^{*}=b\left( 1+z\right) . \end{aligned}$$
(46)

It follows from \(\beta ^{*}<1\) that \(bz/\left( 1-b\right) <1\), which together with \(v\ge 1\) gives the necessary condition for pattern (m,d):

$$\begin{aligned} \frac{b}{1-b}z<1\le v. \end{aligned}$$
(47)

A sufficient condition follows by excluding \(v=1\):

$$\begin{aligned} \frac{b}{1-b}z<1<v. \end{aligned}$$
(48)

Pattern (d,d)

In pattern (d,d), both countries produce both goods. A necessary condition for pattern (d,d) to arise is that \(v=1\). The wages are \(w=A\left( S\right)\) and \(w^{*}=A^{*}\left( S^{*}\right)\). The market-clearing condition requires \(b\left( wL+w^{*}L^{*}\right) =A\left( S\right) \beta L+A^{*}\left( S^{*}\right) \beta ^{*}L^{*}\), which gives (22), namely,

$$\begin{aligned} \beta z+\beta ^{*}=b\left( z+1\right) . \end{aligned}$$

With (22) as the only constraint on the two variables, \(\beta\) and \(\beta ^{*}\) are indeterminate (and thus other patterns may arise).

1.5 A.5 Proof of Proposition 4

Let \(\beta _{T}\) and \(\beta _{T}^{*}\) denote the labor allocations in Home and Foreign at the trade steady state. Supposing that the trade pattern is Home exporting the resource good and Foreign exporting the manufactured good at the trade steady state, we have

$$\begin{aligned} \beta _{T}>b,\quad \beta _{T}^{*}<b. \end{aligned}$$

Consider two countries of the same type. If both countries are the BT type,

$$\begin{aligned} S_{\infty }\left( \beta _{T}\right) <S_{\infty }\left( b\right) ,\quad S_{\infty }^{*}\left( \beta _{T}^{*}\right) >S_{\infty }^{*}\left( b\right) . \end{aligned}$$

If both countries are the CT type, then

$$\begin{aligned} S_{\infty }\left( \beta _{T}\right) >S_{\infty }\left( b\right) ,\quad S_{\infty }^{*}\left( \beta _{T}^{*}\right) <S_{\infty }^{*}\left( b\right) . \end{aligned}$$

In either case, the environmental stock in one country increases and that in the other decreases at the trade steady state (vis-a-vis those at the autarkic steady state).

Consider now two countries of different types. If Home is of BT type (thus Foreign of CT type),

$$\begin{aligned} S_{\infty }\left( \beta _{T}\right)<S_{\infty }\left( b\right) ,\quad S_{\infty }^{*}\left( \beta _{T}^{*}\right) <S_{\infty }^{*}\left( b\right) . \end{aligned}$$

If Home is of CT type (thus Foreign of BT type), then the opposite holds:

$$\begin{aligned} S_{\infty }\left( \beta _{T}\right)>S_{\infty }\left( b\right) ,\quad S_{\infty }^{*}\left( \beta _{T}^{*}\right) >S_{\infty }^{*}\left( b\right) . \end{aligned}$$

In either case, the environmental stocks in the two countries change in the same direction at the trade steady state (vis-a-vis those at the autarkic steady state).

From similar arguments, the same conclusion holds if the trade pattern is reversed.

1.6 A.6 Proof of Proposition 5

We derive welfare impacts of trade on each country by comparing the trade steady-state utility level (\(V_{T}\) for Home and \(V_{T}^{*}\) for Foreign) with autarky (\(V_{A}\) for Home and \(V_{A}^{*}\) for Foreign). It is helpful to proceed in two cases categorized by the trade pattern.

Home Exports the Resource Good

Suppose that Home exports the resource good at the trade steady state and thus Foreign exports the manufactured good. We consider the welfare impacts of trade first on Foreign and then on Home.

Foreign either remains diversified (\(\beta _{T}^{*}>0\)) or specializes in manufacturing (\(\beta _{T}^{*}=0\)) at the trade steady state. We show that in either case Foreign gains from trade.

If Foreign remains diversified, the world relative price at the trade steady state can be expressed by, according to Table 2,

$$\begin{aligned} P_{T}=\frac{A^{*}\left( S_{\infty }^{*}\left( \beta _{T}^{*}\right) \right) }{a^{*}}. \end{aligned}$$

Since Foreign exports the manufactured good (\(\beta _{T}^{*}<b\)) and is of BT type, we have

$$\begin{aligned} P_{T}>P_{A}^{*}=\frac{A^{*}\left( S_{\infty }^{*}\left( b\right) \right) }{a^{*}}. \end{aligned}$$
(49)

Similar to (16), Foreign’s utility level at the autarkic steady state can be expressed by

$$\begin{aligned} V_{A}^{*}&=B+\ln L^{*}+b\ln A^{*}\left( S_{\infty }^{*}\left( b\right) \right) +\left( 1-b\right) \ln a^{*}\nonumber \\&=B+\ln L^{*}+\ln a^{*}+b\ln P_{A}^{*}, \end{aligned}$$
(50)

which together with (26) yields

$$\begin{aligned} V_{T}^{*}-V_{A}^{*}=b\ln \frac{P_{T}}{P_{A}^{*}}. \end{aligned}$$
(51)

It follows from (49) that \(V_{T}^{*}>V_{A}^{*}\) holds (i.e., Foreign gains from trade).

If Foreign specializes in manufacturing, it holds that

$$\begin{aligned} P_{T}\ge \frac{A^{*}\left( S_{\infty }^{*}\left( 0\right) \right) }{a^{*}}, \end{aligned}$$
(52)

from which we also obtain (49). Since (51) still holds in this case, we have, again, that \(V_{T}^{*}>V_{A}^{*}\) (i.e., Foreign gains from trade).

Home either remains diversified (\(\beta _{T}<1\)) or specializes in the resource good (\(\beta _{T}=1\)) at the trade steady state. We show that Home necessarily gains when remaining diversified and may gain or lose when specialized.

If Home remains diversified, we have, according to Table 2,

$$\begin{aligned} P_{T}=\frac{A\left( S_{\infty }\left( \beta _{T}\right) \right) }{a}. \end{aligned}$$
(53)

Since Home exports the resource good (\(\beta _{T}>b\)) and is of BT type,

$$\begin{aligned} P_{T}<\frac{A\left( S_{\infty }\left( b\right) \right) }{a}=P_{A}. \end{aligned}$$
(54)

It follows from (16) and (25) that

$$\begin{aligned} V_{T}-V_{A}=b\ln \frac{P_{T}}{P_{A}}, \end{aligned}$$
(55)

which together with (54) yields \(V_{T}<V_{A}\) (i.e., Home loses from trade).

If Home specializes in the resource good at the trade steady state, then

$$\begin{aligned} P_{T}\le \frac{A\left( S_{\infty }\left( 1\right) \right) }{a}. \end{aligned}$$
(56)

Noting that \(A\left( S_{\infty }\left( 1\right) \right) /a<P_{A}\) (since Home is of BT type), we have that \(P_{T}<P_{A}\). It follows from (16) and (25) that

$$\begin{aligned} V_{T}-V_{A}=\ln \frac{A\left( S_{\infty }\left( 1\right) \right) }{A\left( S_{\infty }\left( b\right) \right) }-\left( 1-b\right) \ln \frac{P_{T}}{P_{A}}, \end{aligned}$$
(57)

the sign of which is indefinite given \(P_{T}<P_{A}\) and \(S_{\infty }\left( 1\right) <S_{\infty }\left( b\right)\). Letting \(V_{T}=V_{A}\) in (57) yields the threshold price:

$$\begin{aligned} P_{1}^{\prime \prime }=\frac{A\left( S_{\infty }\left( 1\right) \right) }{a}\left( \frac{A\left( S_{\infty }\left( 1\right) \right) }{A\left( S_{\infty }\left( b\right) \right) }\right) ^{\frac{b}{1-b}}, \end{aligned}$$
(58)

such that \(V_{T}>V_{A}\) (i.e., Home gains) if \(P_{T}<P_{1}^{\prime \prime }\) and \(V_{T}<V_{A}\) (i.e., Home loses) if \(P_{T}>P_{1}^{\prime \prime }\).

Foreign Exports the Resource Good

Now suppose that Foreign exports the resource good at the trade steady state. Similar arguments as above apply.

Specifically, Home (exporting the manufactured good) necessarily gains from trade. Foreign (exporting the resource good) loses if it remains diversified at the trade steady state. If Foreign specializes, it follows from (50) and (26) that

$$\begin{aligned} V_{T}^{*}-V_{A}^{*}=\ln \frac{A^{*}\left( S_{\infty }^{*}\left( 1\right) \right) }{A^{*}\left( S_{\infty }^{*}\left( b\right) \right) }-\left( 1-b\right) \ln \frac{P_{T}}{P_{A}^{*}}. \end{aligned}$$
(59)

Letting \(V_{T}^{*}=V_{A}^{*}\) gives the threshold price:

$$\begin{aligned} P_{2}^{\prime \prime }=\frac{A^{*}\left( S_{\infty }^{*}\left( 1\right) \right) }{a^{*}}\left( \frac{A^{*}\left( S_{\infty }^{*}\left( 1\right) \right) }{A^{*}\left( S_{\infty }^{*}\left( b\right) \right) }\right) ^{\frac{b}{1-b}}, \end{aligned}$$
(60)

such that \(V_{T}^{*}>V_{A}^{*}\) (i.e., Foreign gains) if \(P_{T}<P_{2}^{\prime \prime }\), and \(V_{T}^{*}<V_{A}^{*}\) (i.e., Foreign loses) if \(P_{T}>P_{2}^{\prime \prime }\).

Combining (58) and (60) gives (27).

For a given world specialization pattern at the trade steady state, the welfare impacts of trade on Home and Foreign can be obtained by applying the relevant results derived above.

1.7 A.7 Proof of Proposition 6

We examine the welfare impacts of trade by comparing the trade steady-state utility level (\(V_{T}\) for Home and \(V_{T}^{*}\) for Foreign) with autarky (\(V_{A}\) for Home and \(V_{A}^{*}\) for Foreign). We proceed in two cases as follows.

Home Exports the Resource Good

Suppose that Home exports the resource good at the trade steady state and thus Foreign exports the manufactured good. We consider the welfare impacts of trade first on Home and then on Foreign.

Home either remains diversified (\(\beta _{T}<1\)) or specializes in the resource good (\(\beta _{T}=1\)) at the trade steady state. We show that in either case, Home gains from trade.

If Home remains diversified, we have (53) and (55). Since Home exports the resource good (\(\beta _{T}>b\)) and is of CT type, \(S_{\infty }\left( \beta _{T}\right) >S_{\infty }\left( b\right)\) holds. This implies, by (53), \(P_{T}>P_{A}\), and thus by (55), \(V_{T}>V_{A}\) (i.e., Home gains from trade).

If Home specializes in the resource good, we have (56) and (57). Note that

$$\begin{aligned} \ln \frac{A\left( S_{\infty }\left( 1\right) \right) }{A\left( S_{\infty }\left( b\right) \right) }>\left( 1-b\right) \ln \frac{A\left( S_{\infty }\left( 1\right) \right) }{A\left( S_{\infty }\left( b\right) \right) }\ge \left( 1-b\right) \ln \frac{P_{T}}{P_{A}}, \end{aligned}$$

where the first inequality comes from \(S_{\infty }\left( 1\right) >S_{\infty }\left( b\right)\) (since Home is of CT type), and the second inequality from (56). This together with (57) also gives \(V_{T}>V_{A}\) (i.e., Home gains from trade).

Foreign either remains diversified (\(\beta _{T}^{*}>0\)) or specializes in manufacturing (\(\beta _{T}^{*}=0\)) at the trade steady state. We show that Foreign necessarily loses when remaining diversified and may gain or lose when specialized.

If Foreign remains diversified, (51) holds. Since Foreign exports the manufactured good (\(\beta _{T}^{*}<b\)) and is of CT type, we have

$$\begin{aligned} P_{T}=\frac{A^{*}\left( S_{\infty }^{*}\left( \beta _{T}^{*}\right) \right) }{a^{*}}<\frac{A^{*}\left( S_{\infty }^{*}\left( b\right) \right) }{a^{*}}=P_{A}^{*}, \end{aligned}$$
(61)

which together with (51) gives \(V_{T}^{*}<V_{A}^{*}\) (i.e., Foreign loses from trade).

If Foreign specializes in manufacturing, (51) holds, and

$$\begin{aligned} P_{T}\ge \frac{A^{*}\left( S_{\infty }^{*}\left( 0\right) \right) }{a^{*}}. \end{aligned}$$
(62)

Noting that \(A^{*}\left( S_{\infty }^{*}\left( 0\right) \right) /a^{*}<P_{A}^{*}\), the sign of (51) remains ambiguous. We further show that Foreign’s gains from trade depend on Home’s specialization pattern: Foreign gains if Home remains diversified and may gain or lose if Home specializes. The details are as follows. (i) If Home remains diversified at the trade steady state, we have \(S_{\infty }\left( \beta _{T}\right) >S_{\infty }\left( b\right)\), and thus by (53), \(P_{T}>P_{A}\). Note that the trade pattern between two countries of CT type is self-reinforcing in that the trade pattern remains unchanged during a dynamic transition (as long as there are no exogenous shocks).Footnote 41 This self-reinforcing feature ensures that if Home exports the resource good at the trade steady state, it also exports the resource good immediately after trade is opened, which requires

$$\begin{aligned} P_{A}\ge P_{A}^{*}. \end{aligned}$$

It then follows that \(P_{T}>P_{A}^{*}\) and thus \(V_{T}^{*}>V_{A}^{*}\) (i.e., Foreign gains from trade). (ii) If Home specializes in the resource good at the trade steady state, it requires \(P_{T}\le A\left( S_{\infty }\left( 1\right) \right) /a\) as given in (56). On the other hand, we have \(A\left( S_{\infty }\left( 1\right) \right) /a>P_{A}\ge P_{A}^{*}\) (the first inequality comes from Home being of CT type). Thus, the sign of (51) remains ambiguous. Letting \(V_{T}^{*}=V_{A}^{*}\) in (51) yields the threshold price:

$$\begin{aligned} P_{1}^{\prime \prime \prime }=\frac{A^{*}\left( S_{\infty }^{*}\left( b\right) \right) }{a^{*}}=P_{A}^{*}, \end{aligned}$$
(63)

such that \(V_{T}^{*}>V_{A}^{*}\) (i.e., Foreign gains) if \(P_{T}>P_{1}^{\prime \prime \prime }\) and \(V_{T}^{*}<V_{A}^{*}\) (i.e., Foreign loses) if \(P_{T}<P_{1}^{\prime \prime \prime }\).

Foreign Exports the Resource Good

Now suppose that Foreign exports the resource good at the trade steady state and thus Home exports the manufactured good. Similar arguments as above apply.

Specifically, Foreign (exporting the resource good) necessarily gains from trade. Home (exporting the manufactured good) loses if it remains diversified. If Home specializes, we have (55) and

$$\begin{aligned} P_{T}\ge \frac{A\left( S_{\infty }\left( 0\right) \right) }{a}. \end{aligned}$$

Noting that \(A\left( S_{\infty }\left( 0\right) \right) /a<P_{A}\) (since Home is of CT type), the sign of (55) is ambiguous. We further show that Home’s gains from trade depend on Foreign’s specialization pattern: Home gains if Foreign stays diversified and may gain or lose if Foreign specializes. The details are as follows. (i) If Foreign stays diversified, we have

$$\begin{aligned} P_{T}=\frac{A^{*}\left( S_{\infty }^{*}\left( \beta _{T}^{*}\right) \right) }{a^{*}}>\frac{A^{*}\left( S_{\infty }^{*}\left( b\right) \right) }{a^{*}}=P_{A}^{*} \end{aligned}$$

since Foreign exports the resource good (\(\beta _{T}^{*}>b\)) and is of CT type. The self-reinforcing trade pattern (which arises between two countries of CT type) ensures that

$$\begin{aligned} P_{A}^{*}\ge P_{A}, \end{aligned}$$

which gives \(P_{T}>P_{A}\), and thus \(V_{T}>V_{A}\) (i.e., Home gains from trade). (ii) If Foreign specializes in manufacturing, \(P_{T}\ge A^{*}\left( S_{\infty }^{*}\left( 0\right) \right) /a^{*}\) is required as given in (52). On the other hand, we have \(A^{*}\left( S_{\infty }^{*}\left( 0\right) \right) /a^{*}<P_{A}^{*}\) (since Foreign is of CT type). However, with these constraints, the sign of (55) remains ambiguous. Letting \(V_{T}=V_{A}\) in (55) yields the threshold price:

$$\begin{aligned} P_{2}^{\prime \prime \prime }=\frac{A\left( S_{\infty }\left( b\right) \right) }{a}=P_{A}, \end{aligned}$$
(64)

such that \(V_{T}>V_{A}\) (i.e., Home gains) if \(P_{T}>P_{2}^{\prime \prime \prime }\) and \(V_{T}<V_{A}\) (i.e., Home loses) if \(P_{T}<P_{2}^{\prime \prime \prime }\).

Combining (63) obtained in the first case and (64) in the second case gives (28).

For a specific world specialization pattern, the welfare impacts of trade on the two countries can be obtained by applying the relevant results derived above. Note that, however, there exists no (stable) pattern (d,d) at the trade steady state when both countries are of CT type. To see this, suppose that \(P_{A}>P_{A}^{*}\) such that Home exports the resource good and Foreign exports the manufactured good right after trade opening. Due to the self-reinforcing feature of the trade pattern as discussed above, the same trade pattern prevails at the trade steady state, which requires \(P_{T}>P_{A}\) if Home remains diversified and \(P_{T}<P_{A}^{*}\) if Foreign remains diversified. Therefore, if both Home and Foreign remain diversified at the trade steady state, \(P_{T}>P_{A}>P_{A}^{*}>P_{T}\) holds and leads to a contradiction. Moreover, as a result of intended policies or exogenous shocks, it is possible that there exists a trade steady state with pattern (d,d) between two CT-type countries. Such a trade steady state is, however, also unstable because of the self-reinforcing feature mentioned above (see the Online Appendix for a formal proof).

1.8 A.8 Proof of Proposition 7

We consider first what happens if Home is of BT type (and thus Foreign is of CT type since the two countries are of different types), and then we consider what happens if Foreign is of BT type.

Home is of BT Type

As for part (i), Home (the BT-type country) exports the manufactured good (\(\beta _{T}<b\)) at the trade steady state, and Foreign (the CT-type country) exports the resource good (\(\beta _{T}^{*}>b\)). Home either remains diversified (\(\beta _{T}>0\)) or specializes (\(\beta _{T}=0\)). If Home remains diversified,

$$\begin{aligned} P_{T}=\frac{A\left( S_{\infty }\left( \beta _{T}\right) \right) }{a}>P_{A}, \end{aligned}$$

where the inequality comes from \(\beta _{T}<b\) and the fact that Home is of BT type. If Home specializes,

$$\begin{aligned} P_{T}\ge \frac{A\left( S_{\infty }\left( 0\right) \right) }{a}>P_{A}. \end{aligned}$$

In either case, \(P_{T}>P_{A}\) holds. Since Home exports the manufactured good, we have that (55) holds, which yields \(V_{T}>V_{A}\). That is, Home (the BT-type country that exports the manufactured good) gains from trade in the long run.

Similarly, Foreign either remains diversified (\(\beta _{T}^{*}<1\)) or specializes in the production of the resource good (\(\beta _{T}^{*}=1\)). If Foreign remains diversified,

$$\begin{aligned} P_{T}=\frac{A^{*}\left( S_{\infty }^{*}\left( \beta _{T}^{*}\right) \right) }{a^{*}}>P_{A}^{*}, \end{aligned}$$

where the inequality comes from the fact that \(\beta _{T}^{*}>b\) and that Foreign is of CT type. If Foreign stays diversified, the difference in Foreign’s steady-state utility levels under trade and autarky can be expressed by (51), which together with \(P_{T}>P_{A}^{*}\) yields \(V_{T}^{*}>V_{A}^{*}\). If Foreign specializes,

$$\begin{aligned} P_{T}\le \frac{A^{*}\left( S_{\infty }^{*}\left( 1\right) \right) }{a^{*}}. \end{aligned}$$
(65)

The difference between the utility level at the trade steady state and that in autarky can be expressed by (59). Note that

$$\begin{aligned} \ln \frac{A^{*}\left( S_{\infty }^{*}\left( 1\right) \right) }{A^{*}\left( S_{\infty }^{*}\left( b\right) \right) }>\left( 1-b\right) \ln \frac{A^{*}\left( S_{\infty }^{*}\left( 1\right) \right) }{A^{*}\left( S_{\infty }^{*}\left( b\right) \right) }\ge \left( 1-b\right) \ln \frac{P_{T}}{P_{A}^{*}}, \end{aligned}$$

where the first inequality comes from the fact that Foreign is of CT type and thus \(S_{\infty }^{*}\left( 1\right) >S_{\infty }^{*}\left( b\right)\); the second inequality comes from (65). This together with (59) yields, again, that \(V_{T}^{*}>V_{A}^{*}\). In either case, Foreign (the CT-type country that exports the resource good) gains from trade.

As for part (ii), Home exports the resource good, and Foreign exports the manufactured good. As discussed in the proof of Proposition 5, Home necessarily loses from trade if it remains diversified and may gain or lose if it specializes. It loses if \(P_{T}>P^{\prime \prime }\). As discussed in the proof of Proposition 6, Foreign necessarily loses if remaining diversified and may gain or lose if specializing. It loses if \(P_{T}<P^{\prime \prime \prime }\).

Foreign is of BT Type

The arguments above remain true when switching Home (and Home-related variables) with Foreign (and Foreign-related variables).

Rights and permissions

Reprints and permissions

About this article

Check for updates. Verify currency and authenticity via CrossMark

Cite this article

Li, G., Yanase, A. Trade, Resource Use and Pollution: A Synthesis. Environ Resource Econ 83, 861–901 (2022). https://doi.org/10.1007/s10640-022-00713-x

Download citation

  • Accepted:

  • Published:

  • Issue Date:

  • DOI: https://doi.org/10.1007/s10640-022-00713-x

Keywords

JEL Classification

Navigation