Exploring Secondary Markets to Improve Circularity: A comparative case study of photovoltaics and hard-disk drives

Each year renewable energy generation increases notably with solar panel installations, but these panels have a limited lifespan and will produce between 2 and 4 million metric tons of waste by 2040. Similarly, there are currently between 20 to 70 million hard-disk drives (HDDs) reaching end-of-life (EOL) annually. The circular economy (CE) strives to recycle and reuse materials that are rare and expensive to obtain, minimizing waste. However, studying the potential circularity of photovoltaics (PV) and HDDs requires various data, for instance, on the maturity of the secondhand markets. In this context, the objective of the present study is to identify the current state of secondhand PV and HDD markets. After conducting a literature review, an automated data collection process was set up for that purpose. The analysis of the literature and collected data assess the maturity of the secondhand PV modules and HDDs markets and highlight differences between them.


Introduction
The commercial market consists of countless product innovations as companies push the boundaries of existing technologies. The U.S. economy largely consists of linear flows of goods (i.e., so called "take make waste" approach) where products are manufactured, sold, and then landfilled [3]. A linear economy produces a large amount of waste as products reach their end-of-life (EOL) and no longer function as intended. As the solar panel market grows, the photovoltaics (PV) industry is expected to generate between 2-4 million tons of waste in the United States by 2040 [11]. With a growing data-storage demand [6], around 20 to 70 million EOL hard-disk drives (HDDs) are generated per year. Besides waste management issues, the growing demand of critical and precious metals used in these two technologies means they can be threatened by sudden supply restrictions [20]. These problems can be solved by changing to a circular economy which reinforces the concepts of reduce, reuse, and recycle (3R) [15]. A circular economy improves on the 3R methodology by adding refusing (e.g. material replacement) repairing, and repurposing to help narrow, close, and slow the waste generation loops [15]. In addition, this CE approach can lead to IOP Publishing doi: 10.1088/1757-899X/1196/1/012029 2 other waste reduction methods such as the sharing economy and product service systems that extend the life of both products and their components [15].
The circular economy requires strong communication and collaboration between the various actors of the supply-chain (extraction, manufacturing, distribution, and use) so that pathways exist to recover EOL products and their components. In practice this has led to the establishment of industrial eco-parks, where waste and co-products from one industry are utilized by another [2] [7]. Recent research, for instance, proposes to apply this concept to the PV industry [13]. In another research project, the CE concept of closed loops was applied to recover the rare earth elements (i.e., neodymium and dysprosium) from used HDDs and reuse them in new drives [10].
However, the current available data pertaining to the development of the CE is severely lacking. For example, strides in CE for PV have been made in the EU through work with Circusol [23] yet few data were made easily accessible. The issue is not only restricted to energy technologies: recent research highlights that the lack of data or ambiguity of current data is preventing the progression of the CE model [19]. The lack of data issue has been mentioned in a few CE studies, for instance, in a New Zealand study on electronic waste [1], an Italian study on construction waste [5], and a Trinidad and Tobago study on plastics [14]. However, if we get past this issue, the circular economy has the potential to greatly reduce waste and environmental impacts [3] and decrease economic costs, and increase employment opportunities for the economy at large [16].
To successfully transition to a circular economy, the current market and material flows needs to be understood. Several types of modeling may help study the CE transition, the choice of the model ultimately depending on the research question [24]. For instance, methods from industrial ecology such as material flow analysis or life cycle assessment can provide information on the sustainability of the CE transition [24]. Agent-based modeling (ABM), by contrast, can help study the CE transition in a different way by providing information on the effect of the intertwined decisions of multiple actors [24]. While developing an analytical model is an alternative solution, market behaviors are often erratic and do not align with other stakeholders' behaviors, thus documenting an ABM through observation may prove more reliable [25]. This approach, however, requires a large quantity of various data, such as recycling and manufacturing costs, prices of secondhand products, and the number and type of actors involved in the CE transition [24]. However, as already mentioned, information surrounding product repair and secondary markets is often non-existent. Despite the effort required to understand these unknown markets, they can be easily broken up into case studies and thoroughly examined.
The PV and the HDD markets are facing a growing need for material recovery and recycling. For instance, the developing PV market is growing about 9% annually worldwide, while the HDD companies are transitioning to a more circular model [10][11] [21]. Currently about 50 million HDDs reach their endof-life annually and need to be shredded or recycled [10]. Due to the recent development of the PV market and the long product life (about 30 years), only about 1.5 million panels are reaching their EOL, but around 32.4 million new panels are being installed each year, which brings an increasing focus on PV EOL management. While still small, the secondary HDD market is far more robust than the PV market and sells 1.5-3 million drives annually that do not include HDDs for internal reuse by large cloud storage companies like Google, Facebook, and Microsoft [21]. In addition to being larger than the PV market, the HDD market is far more developed, containing third party hard drive wiping and restoration companies and some secondary market regulations to hold a standard. Both PVs and HDDs contain precious metals or rare earth elements (REEs) that can be extracted for reuse if the module or disk drive cannot be reused as a unit [17]. For example, silver accounts for about half the cost of a PV module, yet is less than 1% of the module's weight, dissuading companies from pursuing recycling efforts for other PV components due to high costs and low benefit [11]. Several studies have acknowledged the existence of a secondhand PV market, yet lack detail to properly characterize the market. Tao and colleagues state that refurbished modules may be sold at 50% the price of new modules [22]. In addition, the IRENA-IEA 2016 report estimates that the secondhand module ratio might be closer to 70% of the original sales price, but does not provide much detail on how this number was inferred [11]. Regarding to HDDs, one of the International Electronics Manufacturing Initiative (iNEMI) projects reported prices for used drives [10] but this study was interested in the possibility of finding more data-points. Thus, this study aims to provide both a quantitative and qualitive analysis of both the secondhand PV and HDD markets. One of the main findings is the characterization of the price ratio between secondhand and firsthand PV modules and HDDs. In understanding the relationship between these markets, it is possible to document the underdeveloped state of the secondhand PV market (as compared to used electronics such as HDDs) and work to improve its traction within a sustainable economy.

Materials and Methods
This section introduces the approach of our project. Figure 1 summarizes the different steps undertaken to reach our research objectives which are to: x Contribute to the CE research by collecting data on the primary and secondary PV and HDD markets; x Identify the shortcomings of the secondhand PV market to improve the available knowledge about secondhand modules. There are various ways of collecting market data. A literature review or interview can be conducted, and more recently, web scraping has become a common practice [12]. In this study, several methods were used due to wide range of markets and availability of data. For both used PV modules and used HDDs, the types of data collected were the year of production of the product, its origin (manufacturing location), the model name, the product's capacity (Watt-peak (Wp) for PV modules and Terabytes (TB) for HDDs), and the unit price on the secondary market. For new PV modules and HDDs, only the average yearly price was collected as the primary focus of the study was on secondary markets.
The primary source for new PV prices were collected from the Wood Mackenzie 2019 report (Figure 1, step 1). Secondhand PV market prices were collected using the Beautiful Soup Python plugin, that automatically collected PV prices and other descriptors daily from Energybin.com during a period of 10 months (Figure 1, step 2 and Figure 2). Other secondhand data from Kinectsolar.com was scraped prior to this project and has been updated for this analysis. Finally, a third source of used PV modules was manually collected from GreatSolarPanels.com since the website design and limited inventory made writing a scraping script unnecessary. In total, 20 websites were looked at for collecting secondhand PV data, but only three contained secondhand modules for sale. The lack of websites selling secondhand modules was a challenge, since even some websites that had inventory were duplicates of websites we had already used. Moreover, the websites frequently changed the layouts of their webpages, which required updates of the web scraping code and may have caused some data points to be missed. When all collected data was compiled into a data frame, there were around 2,000 data points, but when duplicates were removed our IOP Publishing doi:10.1088/1757-899X/1196/1/012029 4 number dropped down to 117 data points (Figure 1, step 3). This is due in part to how our web scrapers collected data daily and the lack of module sales on the websites. Over the collection period, it was clear that most modules take a long time to sell (often a month or two). Finally, modules were often sold in batches, meaning that one data points represent between 2 to 2992 modules being sold (for a total of 13,367 modules). For example, in March 2021, 624 SUN-250 PV modules were for sale on the Energybin website, representing a single data point. Once organized, data were plotted in various way to analyze and find trends in the PV secondary market and to enable its comparison with the HDDs secondary market (Figure 1, step  4).
A similar method was used to collect data on HDDs secondary market. The New HDD market price-perterabyte (PPT) information was collected from Coughlin (2020) (Figure 1, step 1).The used HDD market prices were automatically collected from diskprices.com, a website which scrapes secondhand HDDs for sale on all amazon platforms internationally for the same duration as PV modules (Figure 1, step 2). Similarly to the PV data, duplicates were removed (Figure 1, step 3), leading to 17,781 data points. Because the year of the HDD model was not indicated, a sample of 127 HDDs were then selected to collect that missing information. In the HDD sub-sample, one data point is corresponding to one used HDD. In addition to collecting used HDD prices, third party EOL HDD companies were contacted. Re-Tek, a data-wiping and electronic refurbishment company provided an internal report, which was used in this study (Figure 1, step 1). Once all data were collected, plots were drawn to analyze trends and enable comparison with the PV modules secondary market (Figure 1, step 4).
For the analysis (Figure 1, step 4), once the data collection period had ended, the data was organized and cleaned up using the pandas Python plugin. The data was then converted into an identical format from each website and duplicates were removed in the data frame. Once each row of data was unique and homogenous, the seaborn and matplotlib plugins were used to generate various visuals for the analysis. Scatter plots and box plots were used as well as some basic linear regression plots. The violin plot function from seaborn was also applied to get a sense of the underlying distribution of the data. Results are presented in the next section.

Results
Despite having primary data for PV and HDD prices and ratios, additional market data gives a more holistic view of the PV market. One of the biggest differences between the PV and HDD market is the average useful lifetime of the products (Table 1). PVs have a useful life that is five times longer than HDDs. This  (Table 1).  The secondhand PV market represents 0.05% of total annual sales, which is two factors of magnitude smaller than the secondhand HDD market (4.2% of total sales). In addition, in looking at the data from a waste perspective, the secondhand PV market absorbs about 1.0% of the waste, while the secondhand HDD market absorbs 6% of the waste generated annually. So, while there is room for waste reduction efforts in both markets, the secondhand HDD market absorbs six times more waste than the secondhand PV market. The annual used sales were estimated from our primary data collection and the literature [8] [18].
Next, we present an analysis of the primary data collected. First, Table 2 presents the summary statistics for both the PV and HDD secondary markets. The table shows that most of the secondhand modules were manufactured in 2013, while the oldest module for sale was from 2000. Given their shorter lifetime, used HDDs were, in average, manufactured in 2017.  Figure 3-a shows that there is a small trend of increasing used/new price ratios as the years progress, but overall, the secondhand PV module prices tend to vary more in recent years (Figure 3-a). Figure 3-a also shows the points by website, revealing that no similarities exist between each website's pricing structure. In Figure 3-b we compared the average used price-per-watt (PPW) with the new PPW obtained from the Wood Mackenzie Report. The blue horizontal line was used to represent the average used PPW across the years since projected linear regression model did not fit the data set well. There is no relationship between the average secondhand PPW and the average PPW for new modules, showing a lack of price structure in the secondhand PV market.   Secondhand PV modules are typically sold at less than the new module price (one outlier at 114% of new PV price) ( Table 2); HDDs have greater variability ( Figure 5-a). This can be explained since websites list secondhand HDDs with product specifications (e.g., Xbox, Mac, PC etc.) but HDDs are not unique in function and are cross compatible in most cases. This type of market manipulation or price inflation also exists in the PV market (when a specific model is highly demanded due to its compatibility with existing PV projects but rare) but did not appear in our dataset. Finally, Figures 5-b) and 5-c) show the PPW and PPT for all PV modules and HDDs.

Conclusion
In understanding the relationship of both the new and used solar panels and HDD markets, this study documented the underdeveloped state of the used solar panel market. This can improve the traction of secondary markets in the sustainable economy. This joint study with the HDD market provides valuable insights on how the PV market can evolve with more regulations, refurbishment companies, and general engagement of stakeholders. Overall, the used PV market is underdeveloped and needs further continuous research to understand what factors (e.g., improved warranties/regulations for used modules) could boost its traction in a circular economy. More research would also lead to better estimations regarding the secondary markets' sizes as new data are available (Table 1). Future research could compare other emerging secondary markets to the mature secondary market for used cars to provide insights in how such emerging secondary markets could be best developed to contribute to a more circular economy.