2.1. Supply of recyclable waste

First, we describe the supply of recyclable waste in exporting countries. Consumers discard good X after consuming it. Once collected, some of the discarded good X enters landfills, and the rest is recycled.

The market for recyclable waste is assumed to be perfectly competitive. The marginal cost of collecting discarded goods and producing recyclable waste is determined by the recovery rate and the quantity of the final good consumed. We describe the supply of recyclable waste produced by collectors in country i asFootnote ⁶

(1) $$Bs_i =q_{b\comma i}^{\beta} \cdot \hbox{Re}_i \cdot x_i\comma \; \quad 0 \lt \beta \lt 1\comma \;$$

where q _{b, i} , Re _i and x _i denote the producer's price of recyclable waste collected in country i, the recovery rate of country i, and the consumption of X in country i, respectively.Footnote ⁷ The producer's price is the per-unit revenue received by waste collectors. We assume that the quantity of recovered material is less than that of material discarded by consumers.Footnote ⁸ When the marginal cost of collecting an additional unit of waste rises, β > 0.Footnote ⁹ Moreover, in general, a higher per capita GDP implies a higher recovery rate because consumers’ environmental consciousness increases with per capita GDP. Figures 1a and 1b provide the evidence for this relationship. They show that, for a given population, higher per capita GDP implies greater total consumption. Thus, higher per capita GDP leads to greater supply of recyclable waste. Although an increase in consumers’ environmental consciousness may decrease the generation of waste through changes in consumption behavior, we do not observe such a clear relationship at this point (see figure 2).

Figure 1a.

(a) GDP/capita and recycling rate: paper and cardboard (2005)

Notes: The sample number is 29. Due to limitation of data availability on the recycling rate, six samples are those of 2004 and two samples are those of 2003.

Figure 1b.

GDP/capita and recycling rate: glass (2005)

Notes: The sample number is 24. Due to limitation of data availability on the recycling rate, four samples are those of 2004 and two samples are those of 2003. Source: OECD Statistics. Environment: Waste, OECD Environmental Data Compendium 2006–2008; and UN, National Accounts Main Aggregates Database.

Figure 2.

Generation of municipal waste per capita and GDP (2005)

Notes: The sample number is 27. Due to limitation of data availability on the generation of waste, three samples are those of other years (2003, 2004 and 2006). Source: OECD Statistics. Environment: Waste, OECD Environmental Data Compendium 2006–2008; and UN, National Accounts Main Aggregates Database.

Collectors of recyclable waste need not differentiate the quality of waste shipped to one country or another. Thus, the producer's price of recyclable waste produced in country i and exported to country j is the same as that exported to country k.

2.2. Demand for recyclable waste

Recyclers in importing countries produce recycled materials from recyclable waste. We consider the following two-step determination for the supply of recycled materials.Footnote ¹⁰ First, one unit of input (Z _{B, j} ) for the production of recycled material is made by mixing imported and domestic recyclable wastes. In terms of recyclers, the quality of recyclable waste generated in any pair of countries generally differs because of variation in the quality of consumed goods, the manner of consumption and the means of discarding. Thus, different countries’ recyclable wastes are imperfect substitutes for each other.

The technology of mixing wastes to produce the input is described by

$$Z_{B\comma j} =\prod_i b_{i\, j}^\gamma \cdot b_{jj}^\gamma\comma \; \qquad i\in N\comma \; \quad \lpar N+1\rpar \gamma =1\comma \;$$

where b _ij denotes recyclable waste exported from country i to country j·b _jj for any j indicates the domestic recyclable waste.Footnote ¹¹ Inputs are determined such that the unit cost is minimized as follows:

$$Min\ \sum_i p_{b\comma ij} b_{ij} +p_{b\comma jj} \cdot b_{jj}\comma \; \ s.t.\ Z_{B\comma j} =1\comma \;$$

where p _{b, ij} denotes the price of imported recyclable waste from country i to country j.Footnote ¹² Solving the cost-minimization problem, the demand function is

(2) $$b_{ij} =p_{b\comma ij}^{-1} \cdot \bar{P}_{B\comma j}\comma \;$$

where $\bar {P}_{B\comma j} =\prod_{i} {p_{b\comma ij}^{\gamma} \ \lpar i\in N\rpar }$ , which is the price index of recyclable waste. The unit cost is given by $p_{ZB\comma j} =\lpar N+1\rpar \cdot \bar{P}_{B.j}$ .

The market for recycled material is perfectly competitive. Given the unit cost of input (p _ZB.j ), the producer's price of recycled material (q _{r, j} ) and the wage (w _j ), the profit of this sector is defined as

(3) $$\pi_{r\comma j} = q_{r\comma j} Rs_j - \lpar w_j l_{r\comma j} \lpar \bar{K}_{r\comma j}\rpar + p_{ZB\comma j}\rpar \cdot Rs_j^{\rho}\comma \; \qquad \rho \gt 1\comma \; \quad {l}_{r\comma j}^{\prime} \lt 0\comma \;$$

where Rs _j denotes the quantity of recycled material produced in country j. Moreover, l _{r, j} and $\bar{K}_{r\comma j}$ denote the labor input used for the production of recycled material and the amount of the specific factor, which is capital stock, in the production of the recycled material, respectively.Footnote ¹³ The condition ρ > 1 implies that marginal cost is increasing, and the profit (producer's surplus) is paid to the specific factor owners. As with suppliers of recyclable waste, recyclers need not differentiate between the quality of product shipped to one country versus another. Solving the profit maximization problem, the supply function is obtained as

(4) $$Rs_j =\left({q_{r\comma j} \over \omega_{r\comma j} \rho}\right)^{{1 \over \rho - 1}}\comma \;$$

where $\omega_{r\comma j} =w_{j} l_{r\comma j} \lpar \bar{K}_{r\comma j}\rpar + p_{ZB\comma j}$ .

From (2) and (4), the demand for recyclable waste collected in country i that is shipped for recycling to country j can be obtained as follows:

(5) $$Bd_{ij} = p_{b\comma ij}^{-1} \cdot \bar{P}_{B\comma j} \cdot Rs_j^{\rho}.$$

Moreover, we consider that some waste enters the landfills of the importing country. It is likely that higher per capita GDP is accompanied by greater environmental consciousness and a lower ratio of imported waste entering landfills. Thus, the demand for recyclable waste can be rewritten as

(6) $$Bd_{ij} = p_{b\comma ij}^{-1} \cdot \bar {P}_{B\comma j} \cdot Rs_j^{\rho} \cdot a_j \bar{E}_j^{\tau}\comma \; \quad a_j \gt 0\comma \; \ \tau \lt 0\comma \; \ a_j \bar{E}_j^{\tau} \gt 1\comma \;$$

where $\bar{E}_{j}$ denotes the per capita GDP of country j and $1/\lpar a_{j} \bar{E}_{j}^{\tau}\rpar $ is the ratio of recycled imported waste to total imported waste for the importer (country j). Finally, τ represents the income elasticity of this ratio.

3.1. Wages of importing countries

Based on (4) and (7) and given the price of recycled materials, we can see that there is an inverse relationship between per-unit variable cost (ω _r,j ) and the scale of the recycling sector. In general, a higher wage rate suggests a greater per capita quantity of capital stock (figure 3). Thus, the relationship between wage and per-unit variable cost depends on whether the recycling process is labor or capital intensive. If the recycling sector is labor intensive, then the higher the wage is in an importing country, the lower the capital rent is in that country, which in turn implies lower recyclable waste imports for any given price of recycled material. On the other hand, if the recycling sector is capital intensive, the higher the wage is in an importing country, the more recyclable waste it imports for any given price of recycled material. The same producer price prevails when recycled material can be traded freely between any two countries and when transportation costs are the same for any two countries.Footnote ¹⁵ In such a case, this cost factor determines the scale of the recycling sector.

Figure 3.

Correlation between the wage and capital stock (1990)

Source: Peen World Tables and Laborsta (ILO).

However, when recycled material cannot be traded freely, wages can affect the scale of the recycling sector via another channel.Footnote ¹⁶ Because recycled material is an intermediate good, producers of final goods tend to procure recycled materials from domestic recyclers. Thus, a larger final goods industry in country j suggests greater demand for material from recyclers in country j. Accordingly, the market scale of recycled material produced in country j is larger. It usually holds that the scale of a country's industrial sector parallels its quantity of capital stock. Ceteris paribus, the higher a country's wages, the more material it demands, including recycled materials.

Consequently, wages can affect the trade volume of recyclable waste positively or negatively. If a lower wage country imports more recyclable waste, it implies that early-stage developing countries import a large quantity of recyclable waste. Recycling in these countries might be labor intensive, and negative externalities of importing waste from developed countries will likely overshadow gains from trade. If a higher wage country imports more recyclable waste, it implies that import volumes will increase along with industry expansion. In this case, significant gains from trade are likely because trade in recyclable waste supplies industries in these countries with input materials.

3.2. The market scale: population and GDP of importing countries

For a given per capita GDP, larger populations have greater demand for final goods. In this sense, population measures the market scale of final goods. GDP itself can also be a measure of market scale because a greater GDP implies a higher per capita GDP or/and a greater population size.

When final goods can be traded freely, consumers might be indifferent between domestic and foreign final goods. In that case, final demand in country j does not affect the scale of that country's final goods industry. In other words, consumers and final goods producers face the same integrated world market. However, if final goods cannot be traded freely, the GDP or population will positively affect the scale of the recycling sector and the demand for recyclable waste because consumers will tend to purchase domestically produced final goods. In the latter case, a larger market scale leads to a higher price of recycled materials and, accordingly, larger demand for recyclable waste (see equation (4)).

3.3. Per capita GDP of importing countries

As per capita GDP increases, people usually become more environmentally conscious, which is likely to increase the waste recovery rate. If so, imports of recyclable waste decrease because domestic and foreign recyclable wastes become substitutes. Moreover, recyclable waste imported into developing countries is sometimes dumped in landfills. Environmentally conscious populations disapprove of landfill dumping, which again suggests that imports of recyclable waste will decrease.Footnote ¹⁷ On the other hand, for a given population, per capita GDP can be a measure of the market scale. Similar to the case of GDP and population, a higher per capita GDP may lead to larger demand for recyclable waste.

In summary, the two effects are at odds with each other. However, when focusing on developing countries, their waste recovery rates are usually low. Therefore, the domestic supply of recyclable waste cannot meet increased demand. Thus, in the case of developing countries, assuming all other variables are fixed, the demand for imported recyclable waste is likely to become greater as per capita GDP increases.

3.4. Variables of exporting countries

The variables of exporting countries also affect trade volume. It follows from (7) and the assumption of β (0 < β< 1) that the quantity of final goods consumed and the recovery rate of exporting country j positively affect the export supply of recyclable waste. Given that other variables are fixed, an increase in per capita GDP increases demand for materials and, accordingly, for recyclable waste. However, for developed countries, waste recovery rates are generally high. Therefore, increases in the supply of recyclable waste are likely to exceed increases in demand. In such cases, the exporting country's per capita GDP positively affects its trade volume in recyclable waste.

A similar result is obtained for the exporting countries’ populations and GDP. A greater market scale leads to a greater demand for final goods and a greater supply of recyclable waste. If the latter effect dominates the former effect, the greater is the population or GDP of an exporting country, the more recyclable waste it exports.

4.1. Empirical specification: a commodity-specific gravity model

Gravity models have empirically succeeded in explaining trade flows. Numerous studies employ them, and several studies have theoretically justified the use of gravity equations (see Anderson, Reference Anderson1979; Bergstrand, Reference Bergstrand1985, Reference Bergstrand1989, Reference Bergstrand1990, Anderson and Wincoop, Reference Anderson and van Wincoop2003; Baldwin and Taglioni, Reference Baldwin and Taglioni2006; Chaney, Reference Chaney2008; Helpman et al., Reference Helpman, Melitz and Rubinstein2008). According to (7), it is appropriate to base our estimation on a gravity model methodology. The predicted signs of the independent variables are those explained in the previous section. This study applies the Poisson pseudo-maximum likelihood (PPML) model following Tenreyro (Reference Tenreyro2007). Silva and Tenreyro (Reference Silva and Tenreyro2006) identify several problems associated with log-linearization in the gravity model. They argue that the expected value of error also depends on the regressors if the variance of error depends on the regressors (heteroscedasticity). Silva and Tenreyro (Reference Silva and Tenreyro2006) also argue that pairs of countries with zero bilateral exports need to be omitted from the sample as a result of the logarithmic transformation. This requirement can create additional bias. In our estimation, we use the Poisson model.

Our empirical commodity-specific gravity model of waste and scrap is as follows:

(8) $$\eqalign{B_{IJ} &=A\cdot GDP_I^{\alpha 1} GDP_J^{\alpha 2} N_I^{\alpha 3} N_J^{\alpha 4} RAW_I^{\alpha 5} RAW_J^{\alpha 6} W_I^{\alpha 7} W_J^{\alpha 8} C_{IJ}^{\alpha 9}\cr &\quad\times \exp \lsqb \alpha_{10} BORDER_{IJ} +\alpha_{11} APEC+\alpha _{12} EU+\alpha_{13} V_{IJ}\rsqb \varepsilon_{IJ}\comma \; }$$

where

• B _IJ  = value (dollars) of country i’s commodity (waste and scrap) imported by country j Footnote ¹⁸

• GDP _I  = per capita GDP of exporting country i

• GDP _J  = per capita GDP of importing country j

• N _I  = population of exporting country i

• N _J  = population of importing country j

• RAW _I  = total imports of raw materials in exporting country i

• RAW _J  = total imports of raw materials in importing country j

• W _I  = manufacturing wages in exporting country i

• W _J  = manufacturing wages in importing country j

• C _IJ  = shortest distance between country i’s commercial centers and country j’s import point

• BORDER _IJ  = a border dummy (=1) if countries i and j share a border (=0 otherwise) APEC = a dummy variable (=1) for intra-Asia–Pacific Economic Cooperation (APEC) flows and (=0 otherwise) EU = a dummy variable (=1) for intra-EU flows (=0 otherwise)

• V _IJ  = real exchange-rate volatility

• ɛ _IJ  = the error term We also conduct another estimation equation: we take GDP

We also conduct another estimation equation: we take GDP as an independent variable instead of per capita GDP and population. For our purpose, it is important to distinguish per capita GDP and population because the former also represents environmental consciousness. However, GDP is also able to measure the market scale simply and directly. Thus, we adopt both types of estimations.

We note that an alternative method considering multilateral resistance of each country to the world (see Anderson and Wincoop, Reference Anderson and van Wincoop2003) is not applicable in our case because the use of exporter-time dummy and importer-time dummy makes country specific variables in each year such as per capita GDP excluded from the specification and testing our main hypothesis non-testable in this study.

4.2. Data

We obtained bilateral export data (constant US$) from the Global Trade Atlas of GTI, Inc. with alternative Harmonized System codes (HS codes). Population and real per capita GDP (constant US$) are, wherever possible, from the Penn World Table. Where these data are unavailable, we use World Development Indicators and the International Monetary Fund's International Financial Statistics. Finally, we specify dummies for landlocked, borders and distance using the US Central Intelligence Agency's World Factbook. Wage data are from LABORSTA, a database of labor statistics compiled by the International Labour Office Bureau of Statistics. This particular gravity data set is the most comprehensive that we know of. The sample period is 17 years, 1995–2011, for all commodity products in our panel data set.

We chose five waste and scrap commodities: (1) waste, parings and scrap of polymers of ethylene; (2) waste, parings and scrap of polymers of vinyl chloride; (3) waste, parings and scrap of polymers of other plastics; (4) ferrous waste and scrap, remelting scrap ingots of iron or steel; (5) copper waste and scrap. The HS codes appear in table 2. We chose these wastes because they fit the objective of this paper: they have large markets in many countries, and international markets for them exist.Footnote ¹⁹ The number of countries in our panel sample varies by the waste recycled because of data availability, ranging from 93 to 160 countries. Table 2 indicates the sample size for each waste.

Table 2.

Harmonized system codes of waste and scrap

As discussed in section 3, an increase in the market for final goods may influence the demand and supply of recyclable waste and its corresponding trade flows. Based on our theoretical analysis, we expect that the signs of these independent variables should be positive for waste exporters and importers.

The manufacturing wages of waste-importing countries are important in this study. The coefficient for this variable should be negative if less advanced countries with lower wages import more recyclable waste, and it should be positive if they import less. We could have used variables such as capital stock and the scale of the industrial sector to represent the scales of the recycling and final goods sectors (or the degree of development). We chose wages over those alternatives because of data availability.

Raw materials may be substitutes for recycled materials for any given output of final goods. Thus, we take it as an independent variable to control the estimation. The remaining variables include distance, dummies for APEC and EU membership, and real exchange rate volatility. As in conventional gravity estimation, the expected signs of the estimated coefficients for these variables are negative, positive, positive and negative, respectively.

4.3. Empirical results

Table 3 provides the estimated results. We adopt the PPML model following Tenreyro (Reference Tenreyro2007) to control sample selection and heteroscedasticity. The PPML estimator allows for the inclusion of zero trade observations. Exporter fixed effect and importer fixed effect are included in our analysis.

Table 3.

Empirical estimates of gravity models

Notes: z-statistics in parentheses.*, ** and*** denote significance at the 0.10, 0.05 and 0.01% level, respectively.

Most estimated parameters have the expected signs and are statistically significant. The results are similar to those of the previous studies that analyzed trade flows using gravity models. Coefficients of population and GDP for exporting and importing countries are positive. Almost all coefficients are statistically significant at 1 per cent confidence.

Coefficients of per capita GDP for exporting and importing countries are also positive. As discussed in section 3, whether the demand effect dominates the environmental consciousness effect is important in determining trade flows. The positive coefficient for waste-importing countries suggests that their demand for final goods strongly influences trade in recyclable waste. However, some of the coefficients for importing countries are insignificant, which also suggests that the environmental consciousness effect can be effective. In particular, some developed countries import few recyclable wastes.

Coefficients for manufacturing wages in the importing country are positive. Almost all of the coefficients are significant at 1 per cent confidence. In general, economic growth and industry expansion lead to higher wages. Because this result can be applied to the group of developing countries, we have shown that the more developed a developing country, the more recyclable waste it imports.

We note that the implications of these positive signs are different from the implications of ordinary gravity estimations for trade in final goods. In the case of trade in recyclable waste, the most serious problem is that the recycling process is separate from other production processes and located in the least developed countries, if it exists. This is a type of pollution haven problem. However, we find no evidence for a pollution haven for wastes and recycling. Our results suggest that recycling sectors are likely to be located along with the production process of final goods.

Our model also analyzes the effect of regional trade agreements by including variables representing trade flows for APEC and EU. We use the distance variable to approximate transportation costs, and we retain the adjacency dummy variable in the empirical model, as there is more trade between countries sharing common borders.

We estimate the effect of exchange rate volatility following Cho et al. (Reference Cho, Sheldon and McCorriston2002) and find that it has a negative and significant effect on the flows of most of the commodities examined. This finding is consistent with the findings of Cho et al. (Reference Cho, Sheldon and McCorriston2002), who suggested that exchange rate volatility impairs trade flows in sectoral trade. However, they also found that the negative effect is commodity specific.