How much does recycling reduce imports? Evidence from metallic raw materials

ABSTRACT

In countries with limited exhaustible natural resources, reducing imports of raw materials is increasingly viewed as a significant side benefit of waste recycling. Using a panel of 21 developed and developing countries from 1994 to 2008, we seek to measure the size of this benefit by estimating the impact of metal scrap recovery on imports of metallic raw materials. We address the endogeneity of metal recovery with exogenous country characteristics including population density and the level of education. We also develop a strategy for controlling for the price volatility in raw material markets. We find that domestic metal recovery is substituted for imports of secondary raw materials while leaving imports of primary raw materials unaffected. The overall effect is a 3.3% decrease in imports of metallic raw materials when metal recovery grows by 10%. Thus, waste recycling policies may have a sizeable impact on trade balance and on security of raw material supply.

KEYWORDS:

• Raw materials
• trade
• recycling
• metal

JEL CLASSIFICATION:

• F18
• L61
• Q53

Disclosure statement

No potential conflict of interest was reported by the authors.

Notes

1 See the review by the European Commission (2000) on the valuation of environmental externalities arising from waste disposal. See also Hu and Shy (2001) who document the health effects of waste incineration.

2 Steel scrap is an example of secondary raw metal. For secondary raw metals, there exist different grades depending on their quality and physical properties. Trade of steel scrap takes place on the London Metal Exchange platform.

3 Virgin raw materials consist in renewable or non-renewable resources that have never been used into production. Iron ore and wood log are examples of virgin raw materials.

4 In many cases, the economic cost of recycling is higher than the cost of waste disposal, particularly for household waste (EPA 1994; European Commission 2002). The net social benefit of recycling is thus a priori unclear. This question has been much less explored in literature. A recent study by Kinnaman, Shinkuma, and Yamamoto (2014) however shows that recycling rates are higher than the socially optimal rates in Japan.

5 This macroeconomic gain is for instance taken into account in Flachenecker, Bleischwitz, and Rentschler (2017)’s cost-benefit analysis.

6 The flagship initiative of a resource-efficient Europe was introduced in 2011 by the European Commission as part of an overall strategy to generate growth and jobs.

7 The other advantages frequently cited are green job creation and improvements of social standards. These benefits concern other policies. For instance, the prohibition of child labour, which is frequent in waste sorting in some developing countries (Beall 1997).

8 Agnolucci, Flachenecker, and Söderberg (2017) find that a 1% increase in the GDP growth rate causes the domestic material consumption growth rate to increase by 2.7% in Western European countries suggesting that policies have so far failed to significantly decouple GDP and material use.

9 We choose to analyze metallic raw materials because the availability of data on metal waste recovery is much higher than the availability of data on non-metal waste recovery.

10 This magnitude is the marginal effect for an average country of our sample.

11 Note that this functional form does not allow for the inclusion of a country that does not produce metal ore.

12 Relatedly, we know that log-linearizing a gravity model can be problematic. In particular, Santos Silva and Tenreyro (2006) show that the PPML estimator performs better than OLS in the presence of heteroscedasticity and with a dependent variable with many zeros. As we aggregate all imports, there is no zero in our sample. We have however run the PPML estimator to mitigate potential heteroscedasticity problems and found similar results.

13 Employing the GMM-IV estimator in the case identified above is equivalent to performing a Two Stage Least Square IV estimator. In our base specification, equation (2) is just identified because we assume two endogenous regressors and use two instruments.

14 Ideally, we would use waste policy stringency as instrumental variable. As far as we know, the existing proxies are available for a limited number of countries and years. Using such variable would dramatically reduce the number of observation and might generate sample selection. An alternative instrument for mining could be a country’s metal ore endowment because this clearly impacts economic agents’ decisions when it comes to extraction. Unfortunately, country-comparable data are not available.

15 Tariff are not a large concern given that tariff on metallic raw materials are low in a large majority of countries.

16 The estimated coefficient for recovery is -0.280 instead of -0.326 in the base model (see table 2).

17 We provide more details in section 4.2.

18 The list of the countries is available in Appendix Table A1.

19 See Appendix Table A2.

20 The output of the metal waste recovery industry is the variable that restricts our sample the most.

21 Based on imports in 2007. The actual figure is slightly lower because our calculation is based on 72 countries including Canada and the United States of America for which trade data are available. The excluded economies are unlikely to weigh much in total trade.

22 In the International Standard Industrial Classification of All Economic Activities (ISIC) 3.1, basic metals manufacturing is classified under Division 27 while metal recovery is classified under Class 3710. The mapping between ISIC classes of the industrial production and the HS codes of the trade data is our own. We simply consider that secondary raw metals identified in Appendix Table A2 correspond to Class 3710.

23 We collect the quantity in terms of metal content of Aluminum, Antimony, Chrome, Cobalt, Copper, Gold, Iron, Lead, Molybdenum, Nickel, Silver, Tin, Titanium, Tungsten, and Zinc because metal content per gram of ores differs from a mine to another.

24 The mining industry is identified under division 13 in ISIC Rev. 3.1. We simply consider that virgin raw metals identified in Appendix Table A2 correspond to division 13.

25 The inclusion of year dummies in the specification does not solve the problem of prices since we cannot assume that the price of secondary raw metals is the same in every country of our sample. Also, we do not use quantity as our dependent variable since it gives a higher importance to heavy raw metals. This would generate an inconsistency between the dependent variable and the recovery variable.

26 We use the Paasche rather than the Laspeyres formula because the latter is not appropriate to deflate output at current prices (IMF 2004).

27 For instance a trade flow of 750 USD is reported as 1,000 USD.

28 Recovery rates that are the share of total import value used to calculate the prices are available upon request.

29 These are ad valorem tariffs. We do not use trade-weighted average tariffs for different reasons. First, the weights use import values which makes the tariff variable endogenous by construction. Also, weighted average tariffs underestimate the level of trade barriers in comparison to simple average tariff. In particular, prohibitive tariffs that totally block trade are not included because their weight is zero (UNCTAD and WTO 2012). In contrast, a prohibitive tariff will be included in a simple average.

30 Austria, Czech Republic, Finland, Hungary, Italy, Norway, Poland, Slovakia, Spain, and Sweden.

31 The OECD and the European Commission organise the production of statistics on industrials activities that are comparable across member states to conduct comparative studies for instance.

32 We do not claim that we do not get an estimate of the true parameter for the entire population of countries.

33 Ideally, we would estimate our model with first-difference. Unfortunately, this method is much costlier in terms of data than the fixed effects approach. Using first-difference GMM would mean to lose 19% of our observations. Considering the size of our sample, first differencing is not only reducing the statistical power dramatically but can also introduce a bias by selecting countries with a larger amount of data.

34 The precise formula is marginal effect = elasticity * (mean of imports / mean of recovery). The numerical application gives the effect of an additional USD produced in waste recovery that is equal to -0.326 x (1.98 billion USD / 1.24 billion USD) = -0.52 USD of fewer imports of metallic raw materials.

35 We believe that it is not a big concern as education is likely affect productivity with a considerable time lag. In a fixed effect estimation, this lagged effect should not affect the estimate significantly.

36 EPA (1994) reported on the recycling operating and maintenance cost and landfill tipping fees for 23 U.S. communities from various U.S. states. In 1990, the average recycling operating and maintenance cost was 101.5 USD per ton and the average tipping fee was 49.5 USD per ton. Recycling operating and maintenance costs were higher than landfill tipping fees in 74% of communities. More recently, the European Commission (2002) reported for Austria that landfill costs ranged from 63 to 111 euro per ton, incineration costs from 111 to 340 euro per ton, and recycling costs from 50 to 493 euro per ton. These costs do not include revenues from energy production and/or material production.

37 Based on author calculations from the UN Comtrade Database and products selected in Appendix Table A2.

 

Additional information

Funding

Both authors gratefully acknowledge financial support from the French Environmental Agency [APR Déchets et Société, grant number: 1310 C 0006]. Matthieu Glachant also acknowledges support through the ANR funded SINCERE project [Grant Number ANR-14-ORAR-0001-01].