Quantifying Export Flows of Used Electronics: Advanced Methods to

Feb 17, 2014 - These agreements imply that Canada and Mexico can trade waste ... This material is available free of charge via the Internet at http://...
0 downloads 0 Views 1MB Size
Article pubs.acs.org/est

Quantifying Export Flows of Used Electronics: Advanced Methods to Resolve Used Goods within Trade Data Huabo Duan,* T. Reed Miller, Jeremy Gregory, and Randolph Kirchain Materials System Laboratory, Engineering Systems Division, Massachusetts Institute of Technology, Cambridge, Massachusetts 02139, United States S Supporting Information *

ABSTRACT: There is limited convincing quantitative data on the export of used electronics from the United States (U.S.). Thus, we advance a methodology to quantify the export flows of whole units of used electronics from the U.S. using detailed export trade data, and demonstrate the methodology using laptops. Since used electronics are not explicitly identified in export trade data, we hypothesize that exports with a low unit value below a used−new threshold specific to a destination world region are used. The importance of using the most disaggregated trade data set available when resolving used and new goods is illustrated. Two detailed U.S. export trade data sets were combined to arrive at quantities and unit values for each port, mode of transport, month, trade partner country, and trade code. We add rigor to the determination of the used−new threshold by utilizing both the Neighborhood valley-emphasis method (NVEM) and published sales prices. This analysis found that 748 to 1199 thousand units of used laptops were exported from the U.S. in 2010, of which 78− 81% are destined for non-OECD countries. Asia was found to be the largest destination of used laptop exports across all used− new threshold methods. Latin American and the Caribbean was the second largest recipient of these exports. North America and Europe also received used laptops from the U.S. Only a small fraction of used laptops was exported to Africa. However, these quantities are lower bound estimates because not all shipments of used laptops may be shipped using the proper laptop trade code. Still, this approach has the potential to give insight into the quantity and destinations of the exports if applied to all used electronics product types across a series of years.



INTRODUCTION The recycling of used electronics has grown tremendously over the last few decades. The management of used electronics in the United States (U.S.) has been the subject of extensive debate, which has revolved around two major points: disposal in U.S. landfills and exportation to developing countries.1,2 The U.S. federal government has taken a particular interest in addressing these issues. In 2011 an Interagency Task Force cochaired by the Council on Environmental Quality (CEQ), the Environmental Protection Agency (USEPA), and the General Services Administration (GSA) released The National Strategy for Electronics Stewardship Benchmarks to specify federal actions for ensuring electronic stewardship in the U.S.3 Recommendations focus on incentivizing design of greener electronics, ensuring the federal government leads by example in acquiring, managing, reusing, and recycling its electronics, as well as increasing domestic recycling efforts, reducing harm from U.S. exports of electronic waste, and improving safe handling of used electronics in developing countries. Furthermore, in January of 2012 the United States Trade Representative (USTR) requested that the United States International Trade Commission (USITC) conduct an investigation and prepare a report that describes U.S. exports of used electronic products.4 © 2014 American Chemical Society

The particular focus on obtaining more information on transboundary movements of used electronics from the U.S. to the rest of the world is because there is limited data on such activities due to the inherent challenges in collecting such data. These challenges include limited mechanisms for data collection, aggregate trade codes for new and used exports, lack of consistent definitions for categorizing and labeling used electronics and their components, minimal regulatory oversight, and limited agreement on the definitions of end uses (i.e., reuse vs recycling). In spite of these challenges, a characterization of the sources, destinations, and quantities of used electronics flows would be instrumental in the development of policies to monitor and control the movements of these products. The objective of this work is to advance such a methodology for exports in comparison with domestic flows. Some studies have estimated the quantity of generation and collection of used electronics at the global, regional, or country level.5−9 Generation refers to used electronics coming directly out of use or postuse storage destined for collection by a Received: Revised: Accepted: Published: 3263

September 30, 2013 February 14, 2014 February 16, 2014 February 17, 2014 dx.doi.org/10.1021/es404365z | Environ. Sci. Technol. 2014, 48, 3263−3271

Environmental Science & Technology

Article

processor or disposal, while collection refers to the subset that are collected by a processor for reuse, recycling, or export. It is generally assumed that exported electronics are a subset of those that are collected. The USEPA produced initial reports in 2007 and a revised report in 2011 which provided transparent, deterministic quantitative characterizations of generation and collection of used electronics, but the scope of the study excluded estimation of exports.10,11 A variety of approaches for quantitative characterization of transboundary flows of used electronics have been proposed and are described in a separate published report.12 A summary of the key approaches is included here. Enforcement data includes reporting from regulated items as well as customs seizure reports. The regulation of the export of cathode ray tubes (CRTs) by the USEPA allowed E-Scrap News to aggregate reported export allowances through information requests. These reported allowances are not audited and verified, and they are expected to be overestimates because firms’ seek high allowances at the beginning of a regulatory period to allow for operational flexibility.13 Customs seizure reports may provide useful anecdotal information, but are inherently nonrandom and are therefore inappropriate for extrapolation to a national scale. Handler surveys include inquiries of recyclers and collectors in both exporting and importing countries.14 U.S. recycling industry associations have conducted surveys on used electronics recycling. In particular, the Institute of Scrap Recycling Industries (ISRI) and the International Association of Electronics Recyclers (IAER, which was acquired by ISRI in 2009) released reports with U.S. survey results discussing export flows in 2003, 2006, and 2010.15 The Northeast Recycling Council (NERC) also analyzed the used electronics market, including export, in 200316 by surveying reuse facilities serving the Northeast. As mentioned above, in 2012 the USITC4,17 conducted a survey, which included questions about export, “of 5200 refurbishers, recyclers, brokers, information technology asset managers, and other handlers of used electronic products”.18 While surveys can provide numerous insights, there will always be questions about whether the survey population is representative of the target population as well as about the existence of bias in responses. Mass balance approaches quantify exports as the difference between estimates of generation and other (nonexport) modes of disposition. Such studies generally first estimate flows of generated used electronics, and then estimate all subsequent domestic flows (reuse, recycling, and landfill). Such estimates require data for many different aspects of the used electronics reverse supply chain. The most sophisticated mass balance estimate of U.S. exports to date was carried out by Kahhat and Williams (2012). Their study estimates U.S. exports of used computers and monitors based on detailed, nationally representative residential, and business surveys.19 As the authors note, mass balance estimates are an approximation of used electronics exports and, at present, do not delineate export destination. Trade data contains information on export or import flows of material or product streams. For many countries including the U.S., general exports are subdivided into domestic exports and re-exports; domestic exports originated in the exporting country while re-exports originated in another country. Official trade databases maintain the value, and in some cases quantity or weight, of goods imported and exported. The market demand and legal authority surrounding the export of used

electronics and their derivatives have been discussed in the Supporting Information (SI). It was found that normal trade is ongoing for commodity grade derivatives of used electronics. It is important to make clear that all methods which employ trade data to estimate used flows face at least one common challenge: false or otherwise inaccurate code reporting. Some used products are shipped using improper trade codes to avoid either tariffs, scrutiny, or legislation and one cannot know the extent of this practice without an extensive and costly verification effort in the ports. Thus, all trade data methodologies result in an underestimate of total used electronics export. Two methods have been described in the literature for quantifying used electronics flows specifically using trade data. First, used products can be differentiated directly if a harmonized commodity description and coding system (HS) code is assigned for used and/or secondhand electronics. Trade codes are harmonized globally at the six digit level, while countries have discretion to make further categorical distinctions at the 8 and 10 digit level. While some countries’ trade data differentiate between new and used electronic items using eight or ten digit codes, U.S. trade data currently does not, aggregating new and used under a single six digit code. Kahhat and Williams identified that Peruvian 20 shipment level trade data includes information on the characteristics of equipment (new, used, refurbished, or broken). Similarly, Yoshida and Kojima described that Japan has had HS codes for used electronics since 2008, including used televisions, refrigerators, washing machines, and air conditioners.21 A second quantification method which makes use of trade data infers the magnitude of used exports based on the relative unit value of shipments within a given trade code. Terazano, Yoshida, and colleagues first posed this concept, which we will refer to as the used−new unit value threshold method, and have applied it extensively to estimate the magnitude of various used product flows from Japan. Some used electronics and electrics exports were distinguished using the export unit value in Japan prior to 2008.21 Exported home appliances and end-of-life vehicles were distinguished between used and new by using an estimated unit value threshold derived from industry surveys in order to quantify several years of flows from Japan.6,7,21−23 The U.S.’ comparable set of industry surveys, the U.S. Census Bureau Current Industrial Reports (CIR), often withheld computer-related data “to avoid disclosing data of individual companies”. Further, the CIR program has been terminated for budgetary reasons. The USITC report mentioned earlier analyzed 2011 shipment-level export trade data by unit value, but did not explicitly infer which portion of exports were used, instead reporting the 10th, 25th, and 50th percentiles of electronics exports by unit value.17 These studies have clearly demonstrated that that there are interesting insights which can be extracted from trade data, a source of information which is both publically available and updated regularly. This study builds on this previous work in four important ways. First it applies the used−new unit value threshold method to a novel case context, the United States. Additionally, it explicitly explores three important methodological issues which have not been previously discussed in the literature. These are (a) the method to determine the value of the unit value threshold, (b) the impact of data granularity (in terms of temporal and spatial resolution) on the result, and (c) the significance of re-export flows. The challenge is that current studies have spent relatively little effort to determine the 3264

dx.doi.org/10.1021/es404365z | Environ. Sci. Technol. 2014, 48, 3263−3271

Environmental Science & Technology

Article

Figure 1. Approach for determination of used−new threshold (z).

by the HS code 847130. The analysis is for the U.S. in the year 2010. Used−New threshold estimation. The threshold values, z, were determined using three separate methods for comparison purposes. U.S. Export NVEM and China Export NVEM utilize the neighborhood valley-emphasis method (NVEM) algorithm 24 with U.S. export and Chinese export data, respectively. Chinese data is considered anticipating that the majority of exported goods are new, since China is a major manufacturer of electronics.25 The third method, Export Published Values Method (Export Pub.), takes advantage of published reference values for used goods, and applies the same threshold to all world regions. Further details on calculating thresholds using three methods are described in the SI. The NVEM approaches assume that there is a bimodal distribution in the export unit values for a given world region distinguishing used and new products. In some cases, there is multimodal distribution if used, typical new, and expensive new products exist within the shipments. In this case, we only differentiate between the used and typical new distributions. The threshold value z is the valley between the postulated used and new normal distributions underlying the bimodal distribution, as demonstrated in Figure 1. Due to price adaptation to different markets, the NVEM threshold approaches do not assume the same threshold applies to all export destination regions; it is assumed that the used− new threshold is consistent across a world region for a type of good. World regions were defined both by World Bank country income groups 26 and UN macro geographical region 27 for vessel, air, and land transport. Co (2007) found that “U.S. exporters do price discriminate across markets”, based on income level, English language, and to some extent changes in currency exchange rate.28 Baldwin and Harrigan (2007) analyze all 2005 U.S. export data and find that “distance has a very large positive effect on unit values”. They also found a negative relationship with export unit value and destination market size.29Export trade data inventory and processing: While all trade databases are derived from official statistics from the U.S. Census Bureau (Census), the data available for analysis are in many different forms with varying fields of information; some are free to the public and others must be purchased. A summary of the available data sets and the information they contain is provided in Table S2 in the SI. There are three steps required to process the trade data into a form that can be utilized to estimate quantities of used electronics exported: calculate

thresholds, and different thresholds have not been created for different export markets. In addition, whether the results are sensitive to the level of data aggregation is not considered. Reflecting on all of the methods that have been used to study export flows of used electronics, it is clear that they all have their limitations. However, given that this is a relatively new field of study, it would be ideal to analyze transboundary flows from many different methods in order to try and paint a picture of the situation from multiple perspectives. Comparing the outcomes from analyses using these different methodologies, such as this study’s results with the USITC survey results, is a way of bounding estimates on export flows until methods and data can be refined to improve these estimates.



MATERIALS AND METHODS Approaches. We have chosen to implement an advanced trade data approach, considering it to be a reasonable trade-off between effort required and quality of information gained. Rich U.S. trade databases are available inexpensively and the results are expected to give a sense of actual movement and direction and to offer a lower bound used export estimate. The overall approach is to utilize detailed, disaggregated trade data to distinguish the quantity of used electronics exports based on export unit value, and sum the used exports. The methodology is applicable for a wide variety of durable used goods exported from any country with detailed trade data sets, and can be used to track trends over a series of years. Since used and new products are not differentiated in the worldwide or U.S. trade codes, under the hypothesis that lowvalue products are used, we have advanced the process of determining used−new unit value thresholds in an efficient manner. These thresholds reflect the minimum export unit value for new products and the maximum export unit value for used products. They are estimated for each destination world region in three ways, two mathematical and another based on market prices. Ideally, the unit value of exports would be determined per shipment in order to avoid aggregation bias, and to take advantage of added data fields available at the shipment level. However, access to U.S. shipment-level export data with reported HS codes is restricted for confidentiality. We believe that it holds promise for creating a comprehensive assessment of the likely destinations and lower-bound estimate of the quantity of used electronics exported. Case Study. The methodological advancements are demonstrated here using the laptop, which is clearly defined 3265

dx.doi.org/10.1021/es404365z | Environ. Sci. Technol. 2014, 48, 3263−3271

Environmental Science & Technology

Article

disaggregated domestic export unit values, estimate used−new thresholds, and calculate the amount of used products exported. (1). Disaggregated Unit Value Calculations. Used and new electronics can be better differentiated with disaggregated data when approaching individual shipments; aggregation results in average values masking the true underlying unit value of individual exports. All Census-based export data sets contain the trade partner nation, n, and the value, v, either reported as free-on-board values (FOB) which exclude duties or shipping charges, or cost, insurance and freight values (CIF) which are inclusive of those costs. The UN Comtrade database 30 reports exports as FOB values and imports as CIF values; some detailed data sets such as SICEX31 offer both. When possible, FOB values were utilized for consistency. Some data sets contain quantity of goods, q, and/or weight w. The various levels of data aggregation are as follows. The trade flows, f, utilized were either: general export, fg, domestic export, fe, or total imports, f i; the classification general exports aggregates domestic exports and re-exports. We assume that exports of used electronics are only within the domestic export subset of general exports. All data sets utilized reported data at the monthly level m; UN Comtrade reports v and w at the monthly level, but q only at the annual level. Some data sets combined all transport modes, while others distinguished air, vessel and land transport (indexed on t). Regarding the regional aggregation of shipments, r, some data sets reported trade at the country level rc, some at the district (group of ports) level rd, some at the port level rp, and some at the shipment level rs. As mentioned previously, the ideal U.S. export trade data set of detailed shipment level reporting is restricted. While bill of lading export data is available for purchase from PIERS,32 it is for ocean freight only, and with approximate export codes; this is likely an incomplete snapshot of total export trade. Researchers may be able to access shipment level data after approval from Census Research Data Centers. Port-level data from USA Trade Online 33 contains w but lacks q and does not separate domestic exports from re-exports. Therefore, a method was developed to approximate port-level domestic export unit values and quantities using data sets with q and domestic exports separated from re-exports. Our model was developed to combine data sets at port-level and district-level in order to arrive at the most disaggregated unit value, u. Port-level quantity data are needed to calculate the port-level unit value (described below). Unfortunately, the data sets utilized do not contain this information for land shipments, so alternatives were sought for U.S. exports to Canada and Mexico. “Canada and the United States participate in a “data exchange”, in which the export statistics of each country are derived from the counterpart import data; therefore, there are no unexplained differences in their trade statistics. However, differences between the official trade statistics of the United States and Mexico, and Canada and Mexico are sizeable”.34 Therefore, port-level Canadian import data from STATCAN 35 is substituted for U.S. domestic export data to Canada. Quantity data is available via SICEX for U.S. exports to Mexico, and therefore U.S. domestic export data to Mexico at the district-level is utilized. Table 1 presents the data sets utilized, subscripts correspond to the number of the data set. For convenience, the symbols and terms that are summarized in SI Table S3. First, all data utilized was aggregated to the annual, all transport mode, partner country level to check for consistency across v, q, and w in comparison with UN Comtrade data.

Table 1. Data Sets Utilized for U.S. Exports Calculations database

1. USA Trade Online

value

v1( fg,m,n,rp,t)

quantity weight

w1( fg,m,n,rp,t)

2. SICEX (U.S. Exports)

3. STATCAN (Canada Imports)

v2( fe,m,n,rd,t), v2( fg,m,n,rd,t) q2( fe,m,n,rd,t), q2( fg,m,n,rd,t) w2( fe,m,n,rd,t), w2( fg,m,n,rd,t)

v4( f i,m,n,rp,t) q4( f i,m,n,rp,t)

Minor issues were encountered with regards to inconsistencies in country classification (e.g., Sudan, Curacao) across data sets; trade with these countries was very small. The disaggregated U.S. domestic export unit value u was calculated at two levels of aggregation: district-level, and portlevel. As a reminder, at the port-level, Canadian import data was substituted for U.S. domestic export data, and district-level export data was used for exports to Mexico. The district-level U.S. domestic export unit value was calculated with SICEX data as shown in eq 1. u 2(fe , m , n , rd , t ) =

v2(fe , m , n , rd , t ) q2(fe , m , n , rd , t )

(1)

The domestic export district-level unit weight x2 (fe,m,n,rd,t) for each month, partner nation, port, and transport mode is found as shown in eq 2. To arrive at the approximate port-level unit value u1−2( fe,m,n,rp,t) for non-North American countries, the general export port-level value per weight is multiplied by the corresponding district-level unit weight in eq 3. x 2(fe , m , n , rd , t ) =

w2(fe , m , n , rd , t ) q2(fe , m , n , rd , t )

(2)

u1 − 2(fe , m , n , rp , t ) ≅

v1(fg , m , n , rp , t ) w1(fg , m , n , rp , t )

× x 2(fe , m , n , rd , t )

(3)

The approximate port-level quantity q1−2 ( fe,m,n,rp,t) essentially allocates a district’s domestic export quantity to a port based on the port’s share of the district’s general export weight for a given month and trade partner nation. In eq 4 the approximate port-level quantity is equivalent to the fraction of port-level general export weight out of district-level general export weight multiplied by the district-level domestic export quantity. q1 − 2(fe , m , n , rp , t ) ≅

w1(fg , m , n , rp , t ) w2(fg , m , n , rd , t )

× q2(fe , m , n , rd , t )

(4)

To calculate Canadian import unit values for trade with the U.S. as country of origin n, the value is simply divided by quantity for each month, port or district, and transport mode. Canadian import unit value is shown in eq 5. u3(fi , m , n , rp , t ) =

v3(fi , m , n , rp , t ) q3(fi , m , n , rp , t )

(5)

(2). Threshold Calculations. Thresholds were calculated at the port-level, for each world region and for vessel, air, and land transport. Since the data sets utilized largely report export 3266

dx.doi.org/10.1021/es404365z | Environ. Sci. Technol. 2014, 48, 3263−3271

Environmental Science & Technology

Article

values that do not include freight costs, it may seem superfluous to find different thresholds for transport modes. However, considerable differences in unit value distributions have been observed for this data set based on mode of transport, so it is important to investigate the impact of this issue on the outcomes of the analysis. The neighborhood valley-emphasis method (NVEM) was employed to determine the used−new threshold unit value z for U.S. Export NVEM and China Export NVEM approaches. Fan and Lei24 differentiate between modes in a distribution using NVEM, which they developed to find the threshold of a bimodal histogram of a grayscale image. The method finds the optimal threshold z, which simultaneously maximizes the variance between the modes (or classes) and minimizes the probability of the unit value bin u at and around the optimal threshold. By considering not only the probability at the potential threshold u but neighboring u as well, sporadic dips not corresponding to true valleys are not selected. The method for distinguishing thresholds using the NVEM as well as the calculated threshold unit values z for this analysis are described in the SI. When a particular world region and mode of transport had insufficient export quantity for analysis, appropriate related thresholds were substituted. To summarize, the laptop thresholds range across all world regions and modes of transport from $180 to $300 for U.S. Export NVEM and $175 to $300 for China Export NVEM. An example of the threshold range found by China Export NVEM is shown in Figure S7 and S8 in the SI, with approximate distributions superimposed on the histogram. (3). Calculate Amount of Used Electronics Exports. The sum of used U.S. domestic exports to each partner nation n at the district-level,qz2(fe,n), was found using eq 6. Similarly, the sum of used U.S. domestic exports to each partner nation n at the port-level qz1−2(fe,n), was found using eq 7 for all non-North American countries; qz3(f i,n) for Canada was found in eq 8. District-level calculations: z

q2z(fe , n) =



is only reported between two trade partners. The exported product may be subsequently re-exported. 4. The methodology is demonstrated with laptops only, which may not be representative of all used electronic products. In our other work, it has been demonstrated for other products.37 5. The methodology can only be used to track whole units and not scrap commodity streams from units disassembled in the U.S.. It is not possible to distinguish used electronic scrap within export data of scrap commodity streams.

RESULTS Figure 2 presents a comparison of laptop trade data aggregation at various levels. The uppermost chart shows the quantity of

D

12

∑ ∑ ∑ ∑ q2(fe , m , n , rd , t ) u 2 = 0 m = 0 rd

t

(6)

Port-level calculations: z

q1z− 2(fe , n) =

12

P

∑ ∑ ∑ ∑ q1 − 2(fe , m , n , rp , t ) u1 − 2 = 0 m = 0 rp z

q3z(fi , n) =

12

t

P

∑ ∑ ∑ ∑ q3(fi , m , n , rp , t ) u3 = 0 m = 0

(7)

p

t

Figure 2. U.S. domestic laptop exports screened by unit value ($25 interval) with various level data in 2010 (dashed lines represent the of range of export pub threshold): (a) annual, country-level; (b) monthly, district-level; (c) monthly and port level.

(8)

As a reminder, when aggregating U.S. domestic export quantity sums across North American world regions at the port-level, Canadian port-level import data and Mexican district-level data was substituted for port-level exports to those countries for reasons explained above. Limitations. There are several shortcomings associated with using trade data from the Census as well as general constraints for the methodology proposed in this study, including: 1. Intentional or accidental trade code misclassification of products resulting in an underestimate of actual trade.20 2. General trade data reporting errors.36 3. The first export destination country is known, but not necessarily the final destination country, since trade data

domestic exports aggregated at the annual, country-level. The next chart disaggregates the data by month and districts (groups of ports). The last chart further disaggregates the data by approximate port-level data. The dashed lines indicated a used−new threshold range for laptops of $200−$250 based on the Export Pub. Threshold method, derived from market data. The trend shows that with more disaggregated data, the histogram is less concentrated around its mean, and more lowvalue (and high-value) exports can be distinguished. Of course, we would expect this trend to continue if the data was further disaggregated by individual shipments, and thus it is a reasonable approximation. 3267

dx.doi.org/10.1021/es404365z | Environ. Sci. Technol. 2014, 48, 3263−3271

Environmental Science & Technology

Article

Figure 3. Quantities of used laptops exported by threshold method, World Bank Country Income Group, UN World Macro Region Classification. Error bars represent minimum and maximum values.

($180−-$185). It is expected that the Hong Kong estimate from China Export NVEM is a special case related to China’s proximity and trade relations with Hong Kong. Used electronics have been shipped to all of the investigated world regions, but the majority of used laptops have gone to Asian and Latin American and Caribbean (LAC) countries that are within the high income or upper middle income categories. It is important to remember, however, that we cannot be certain that these are the final destinations for the exported products. Indeed, some of the destinations may simply be temporary ports for re-export activities. An investigation of the likelihood of re-export from the top 10 destination countries is discussed in SI (section 8); several top destination countries are probably re-exporters while others are likely the final destination. This study also compares the quantity of exported used laptops by mode of transport for each used−new threshold method. Across all methods based on average values, the fraction of used laptops shipped by air surprisingly ranges from 44% to 51%, followed by 41% to 50% for vessel exports, and 6% to 12% for land exports. This reaffirms the original intuition that the use of the bill of lading data to represent all exports despite the lack of air or land export data would exclude a significant portion of exports.

Figure 3 presents the results of the application of the three used−new thresholds to the prepared U.S. domestic export trade data, with the estimates broken down by World Bank country income group and UN world macro region classification. The overall results using the three used−new thresholds are comparable, estimating that on average 871, 1158, and 896 thousand used laptops were exported in 2010 using the U.S. Export NVEM, China Export NVEM, and Export Pub. Method respectively. The U.S. Export NVEM estimated 865−878 thousand exported laptops were used, approximately 24% of the total laptop exports from the U.S. in year 2010. The greatest discrepancy between threshold methods in these results is observed for the high income Asian destination countries, likely due to China’s different trade partnerships with other Asian countries as compared to those of the U.S.. As a reminder, the threshold values, z, were determined using separate methods for comparison purposes. U.S. Export NVEM and China Export NVEM utilize the neighborhood valley-emphasis method (NVEM) algorithm for each destination world region with U.S. export and Chinese export data, respectively. Here, China Export NVEM arrives at a higher threshold range ($295−$305) for exports from China to High-Income Asia, particularly Hong Kong, than does U.S. Export NVEM for exports from the U.S. to the same region 3268

dx.doi.org/10.1021/es404365z | Environ. Sci. Technol. 2014, 48, 3263−3271

Environmental Science & Technology

Article

Figure 4. Top 10 destination countries of used domestic laptop exports. Error bars represent minimum and maximum values.

mainland for transboundary movements of recyclable wastes.38 We mentioned that the first export destination country is known, but not necessarily the final destination country, since trade data is only reported between two trade-partners. However, an estimate of the probability that the first destination country of an export is not its final destination is given by the ratio of re-exports to imports where available (see section S8 Re-export evaluation, SI).Using this approach, it was found that most of the laptops (new and used combined) that Hong Kong imported from the world have been re-exported to China (∼35%), India (∼10%), and other neighboring Eastern Asian countries. UAE is the third most important re-export center in the world after Hong Kong and Singapore,39 Lebanon is also an important transshipment point for goods coming from or going to a variety of Arab countries (see detailed analysis in SI). Other studies have documented that both UAE and Lebanon do not typically reuse or recycle used items domestically, but instead re-export them to surrounding countries such as Pakistan.23 All of this information serves as a reminder that several of the Asian destinations reported in this study are likely to be transshipment points for re-export. Based on the analysis in this paper, Latin America and the Caribbean (LAC) is the second most common destination region for used laptops. LAC is not the largest U.S. regional trade partner, but historically is one of the fastest growing ones.40 The major destinations include Argentina, Chile, and Bolivia. LAC has a rapidly increasing Internet usage rate along with fast growing computer sales.41 Like many other developing countries, both new and used electronics are imported to meet domestic consumption demand. There is a clear example in Peru where the main purpose of imported used computers from the U.S. in Peru is expected to be for reuse.20 Another study reporting on used electronics generation in Chile released in 2007 42 showed that the sales prices for refurbished computers were in the range of US$30−US$240. The cheaper equipment usually stems from local auctions, whereas the refurbished equipment is often imported from the U.S.. Developed countries, such as the United Kingdom (UK), which is one of the major international trade centers in Europe,43 also received used laptops in 2010 according to the analysis presented here. Some of the valuable materials used in electronics are most efficiently recycled in large smelters only located in developed countries in Europe, North America, or Asia which would explain why the laptops would be sent to developed countries. Another explanation could be that transfer of used electronics requires notification and consent under the Basel Convention, but is more simplified between OECD

Figure 4 compares the quantity of used laptops exported to the top ten destination countries for each used−new threshold method, which happens to be the same across all methods. These countries account for 71%, 78%, and 74% of the total used laptop exports for U.S. Export NVEM, China Export NVEM, and Export Pub. Method, respectively. The countries were ordered by U.S. Export NVEM.



DISCUSSION After a review of relevant approaches, the used−new unit value threshold method of analyzing trade data was selected and methods were advanced to estimate the quantity of used electronics export from the U.S. in 2010 to world regions. Three kinds of used−new thresholds were developed, the importance of creating different thresholds for different world regions was discussed, and the need to use disaggregated trade data was demonstrated. Overall, the trade data method presented in this study is successful in providing lower bound estimates of used laptop export quantities. In addition, the methodology has also recently been applied to other electronics, including TVs, monitors, desktops and mobile phones.37 Analysis using this method provides insight into the destinations of these exports, as well as the re-export destinations of a subset after import into the destination country. While the total quantity of used laptop exports reported here is most likely a lower bound, the proportions of exports to world regions is likely accurate (see justification in section S8 Re-export evaluation, SI). These proportions may be skewed if import restrictions in some countries result in more incidents of intentional misclassification for exports to specific countries. Unintentional export reporting errors are likely to be distributed proportionally. Other studies have demonstrated that export flows can shift from year to year. As such, it would be inappropriate to generalize based on the analysis of the one year of data (as presented here), nevertheless, discussion of these results makes clear the types of insights which are possible from the used−new unit value threshold method. The analysis presented here suggests that Asian countries were the most common destinations for used laptops exported from the U.S. in 2010. The major destinations include Lebanon, Hong Kong, United Arab Emirates (UAE), and China. Hong Kong, which is a special administrative region of China, is a duty-free port and serves as the leading transit port for international distribution in the Asia region. It is particularly active as a transit port for goods bound for China from Japan, the U.S., and Europe, and also functions as the window to the 3269

dx.doi.org/10.1021/es404365z | Environ. Sci. Technol. 2014, 48, 3263−3271

Environmental Science & Technology countries.5 Lastly, as shown in the SI, in 2010 the UK may have re-exported up to 20% of their imported laptops. Based on the analysis in this paper, the U.S.−North American trade partners (Canada and Mexico) were the other most common world regions for used laptop importers from the U.S.. The U.S. has had bilateral waste regulation agreements with Canada and Mexico since 1986. These agreements imply that Canada and Mexico can trade waste with the U.S. using USEPA shipment notifications. An analysis of the notifications to the USEPA for the shipment of broken CRTs in the year of 2010 and 2011 found that 56% and 24% of exported CRTs scrap was shipped to Canada and Mexico, respectively. The U.S. only exported a small fraction of its used laptops to Africa in 2010 according to this analysis. While used electronics exported to developed countries are often recycled, the same items that are exported to African countries are primarily intended for reuse, and European countries have been considered the primary exporters to Africa.44 An analysis of containers of used electronics imported into Nigeria was conducted by monitoring shipment manifests and providing shipping information for about 176 containers.44 Results revealed that almost 60% of the containers of used electronic came in from the UK. More than 75% of all containers came from Europe, approximately 15% from Asia, 5% from African ports (mainly Morocco), and 5% from North America. Comparable results have also been obtained by Amoyaw et al.,45 who analyzed the mass flow of used electronics in Ghana. Almost all electronics products in Ghana are imported mainly from developed countries, of which 84% are from Europe; only 8% are from the U.S. and Canada. To summarize, the used−new unit value threshold approach has the potential to give insight into both the quantity and destinations of the exports if applied to all used electronics product types across a series of years. This study has specifically shown that there are a number of systematic ways to identify threshold values, the importance of creating different thresholds for different world regions, and the need to use disaggregated trade data. Furthermore, this study discusses the importance of differentiating exports and re-exports as well as proposes a systematic method to infer likelihood of re-export from a destination country. Based on experiences gathering and analyzing trade data for this study, it is recommended that the U.S. government create “new Schedule B numbers to distinguish between new and used electronics in U.S. export data” as mentioned by the U.S. Government’s Interagency Task Force on Electronics Stewardship (2011), and to consider enhancing and extending the reporting requirements of the CRT Rule under RCRA to all used electronics, thereby enabling more transparent data collection.



ACKNOWLEDGMENTS



REFERENCES

This study was supported by the Solving the E-Waste Problem (StEP) initiative with a grant from the USEPA. Assistance from Jason Linnell at the National Center for Electronics Recycling on this effort is greatly appreciated.

(1) Kahhat, R.; Kim, J.; Xu, M.; Allenby, B.; Williams, E.; Zhang, P. Exploring e-waste management systems in the United States. Resour., Conserv. Recycl. 2008, 52 (7), 955−964. (2) SVTC.; BAN. Exporting Harm: The High-Tech Trashing of Asia; The Basel Action Network (BAN) and Silicon Valley Toxics Coalition (SVTC): Seattle, WA, 2002. (3) CEQ.; EPA.; GSA. National Strategy For Electronics Stewardship Benchmarks. Council on Environmental Quality (CEQ), Environmental Protection Agency (EPA) and Geological Society of America (GSA). 2011. (4) USITC. Used Electronic Products: An Examination of US Exports; Investigation No. 332-528; USITC: Washington, D.C. USA, 2012. (5) Schluep, M.; Wasswa, J.; Kreissler, B.; Nicholson, S. E-waste generation and management in Uganda. Proceedings of the 19th Waste Management Conference of the IWMSA (WasteCon2008), Durban, South Africa, 2008. (6) Terazono, A. Generation and material flow of e-waste in east Asia. In Regional Workshop on E-Waste Identification toward the Prevention of Illegal Transboundary Movemetn for Hazardours Waste and Other Wastes in Asia; BCRC: Beijing, China, 2008. (7) Terazono, A.; Murakami, S.; Abe, N.; Inanc, B.; Moriguchi, Y.; Sakai, S.-i.; Kojima, M.; Yoshida, A.; Li, J.; Yang, J.; Wong, M. H.; Jain, A.; Kim, I.-S.; Peralta, G. L.; Lin, C.-C.; Mungcharoen, T.; Williams, E. Current status and research on e-waste issues in Asia. J. Mater. Cycles Waste Manag. 2006, 8 (1), 1−12. (8) Yang, J.; Lu, B.; Xu, C. Weee flow and mitigating measures in china. Waste Manage. 2008, 28 (9), 1589−1597. (9) Yu, J.; Williams, E.; Ju, M.; Yang, Y. Forecasting global generation of obsolete personal computers. Environ. Sci. Technol. 2010, 44 (9), 3232−3237. (10) USEPA. Electronics Waste Management in the United States: Through 2009, E.530-R-11- 002; Prepared by ICF International for the U.S. Environmental Protection Agency, 2011. (11) USEPA. Electronics Waste Management in the United States (Approach 1); U.S. Environmental Protection Agency (USEPA): Washington, DC, US, 2008. (12) Miller, T. R.; Gregory, J.; Duan, H.; Kirchain, R. Characterizing Transboundary Flows of Used Electronics: Summary Report; Massachusetts Institute of Technology: Cambridge, MA, 2012. (13) Thomas, J. A look through the leaded glass. E-Scrap News 2013, (June 2013), 10−19. (14) Miller, T. R. Quantitative Characterization of Transboundary Flows of Used Electronics: A Case Study of the United States; Massachusetts Institute of Technology: Cambridge, MA, 2012. (15) Daoud, D. Report inside the US Electronics Recycling Industry; IDC, 2011. (16) NERC. Used Electronics Market Study Survey Analysis; Northeast Recycling Council, Inc. (NERC), 2003. (17) USITC. Used Electronic Products: An Examination of U.S.: United States International Trade Commission: Washington, DC, 2013. http://www.usitc.gov/publications/332/pub4379.pdf (18) USITC. U.S. Exports of used electronic products valued at $1.5 billion in 2011, says USITC; U.S. International Trade Commission, 2013. (19) Kahhat, R.; Williams, E. Materials flow analysis of e-waste: Domestic flows and exports of used computers from the United States. Resour. Conserv. Recyl. 2012, 67 (0), 67−74. (20) Kahhat, R.; Williams, E. Product or waste? Importation and endof-life processing of computers in Peru. Environ. Sci. Technol. 2009, 43 (15), 6010−6016.

ASSOCIATED CONTENT

* Supporting Information S

Additional information as noted in the text. This material is available free of charge via the Internet at http://pubs.acs.org.





Article

AUTHOR INFORMATION

Corresponding Author

*Phone: +1-617-253 6467; fax: +1-617-258 7471; e-mail: [email protected]. Notes

The authors declare no competing financial interest. 3270

dx.doi.org/10.1021/es404365z | Environ. Sci. Technol. 2014, 48, 3263−3271

Environmental Science & Technology

Article

(44) SBC. Where are WEEE in Africa? Secretariat of the Basel Convention. December, 2011. (45) Amoyaw-Osei, Y.; Agyekum, O.; Pwamang, J.; Mueller, E.; Fasko, R.; Schluep, M. Ghana e-waste country assessment. SBC. 2011. http://www.basel.int/Portals/4/Basel%20Convention/docs/eWaste/ E-wasteAssessment Ghana.pdf.

(21) Yoshida, A.; Kojima, M. Transboundary movement of recyclable resources: Current management system and practices in Japan. In The Workshop of the Asian Network for Prevention of Illegal Transboundary Movement of Hazardous Wastes, Tokyo, Japan, 2008. (22) Terazono, A., Yoshida, A. Current international flows of electronic waste, future tasks, and possible solutions. In E-Waste Management; Hieronymi, K., Kahhat, R., Williams, E., Eds.; Routledge, 2012; pp 137−163. (23) Fuse, M.; Nakajima, K.; Yagita, H. Outflow of resources from Japan focusing on end-of-life vehicles. Mater. Trans. 2007, 48 (9), 2436−2444. (24) Fan, J. L.; Lei, B. A modified valley-emphasis method for automatic thresholding. Pattern Recognit. Lett. 2011, 33 (12), 703−708. (25) Kaplinsky, R.; Santos-Paulino, A. A disaggregated analysis of EU imports: The implications for the study of patterns of trade and technology. Cambridge J. Econ. 2006, 30 (4), 587−611. (26) World Bank Country and lending groups. http://data. worldbank.org/about/country-classifications/country-and-lendinggroups (accessed 21 July 2012), (27) UN Statistics Division Composition of macro geographical (continental) regions, geographical sub-regions, and selected economic and other groupings. http://unstats.un.org/ unsd/methods/m49/ m49regin.htm (accessed 20 July 2012), (28) Co, C. Y. Factors that account for the large variations in US export prices. Rev. World Econ. 2007, 143 (3), 557−582. (29) Baldwin, R.; Harrigan, J. Zeros, Quality and Space: Trade Theory and Trade Evidence; National Bureau of Economic Research: Cambridge, MA, US, 2007. (30) UN Statistics Division UN comtrade: United Nations Commodity Trade Statistics Database. http://comtrade.un.org/db/ default.aspx. (31) Quintero Hermanos LTD. Sicex. http://www.sicex.com/. (32) PIERS. PIERS database of U.S. International trade. http://www. piers.com/data. (33) U.S. Census Bureau, USA Trade Data Online. 2012. (34) Mozes, S.; Oberg, D. U.S.-Canada Data Exchange; Statistics Canada and U.S. Census Bureau, 2002. (35) Statistics Canada. Population by year, by province and territory Statistics Canada. 2012. (36) Ferrantino, M. J.; Wang, Z. Accounting for discrepancies in bilateral trade: The case of China, Kong Kong, and the United States. China Econ. Rev. 2008, 19 (3), 502−520. (37) Duan, H.; Miller, T. R.; Gregory, J.; Kirchain, R. Quantitative Characterization of Transboundary Flows of Used Electronics: Analysis of Generation, Collection, and Export in the United States. December, 2013. Materials Systems Laboratory, MIT. http://www.step-initiative.org/tl_ files/step/_documents/MITNCER%20US%20Used%20Electronics%20Flows%20Report%20%20December%202013.pdf. (38) Kojima, M.; Yoshida, A. Hong KongThe transit-port for recyclable wastes & secondhand goods. Int.Trade Recycl. Resour. Asia 2005, 29, 53. (39) Parsa, A.; Keivani, R.; Sim, L. L.; Ong, S. E.; Younis, B. Emerging global cities: Comparison of Singapore and the cities of the United Arab Emirates. Real Estate Issues 2002, 27 (3/4), 95−101. (40) Hornbeck, J. F. US-Latin American Trade: Recent Trends and Policy Issues; Congressional Research Service, Library of Congress, 2011. (41) Boeni, H.; Silva, U.; Ott, D. E-waste recycling in Latin America: Overview, challenges and potential. Global Symposium on Recycling, Waste Treatment and Clean Technology, Cancun, Mexico, 2008; pp 665-673. (42) Steubing, B. E-Waste Generation in Chile. Master Thesis. Swiss Federal Institute of Technology Lausanne (EPFL): Lausanne, Switzeland, 2007. (43) Kelemen, R. D.; Vogel, D. Trading places: The role of the United States and the European Union in international environmental politics. Comp. Political Stud. 2010, 43 (4), 427−456. 3271

dx.doi.org/10.1021/es404365z | Environ. Sci. Technol. 2014, 48, 3263−3271