LBMA Country-of-Origin Analysis Report
Report Introduction and Analysis
In 2019, LBMA launched its gold and silver country-of-origin (CoO) dataset, which surveyed GDL refiners’ treatment of mine production and recycling. Importantly, GDL refiners were asked to identify the origin of their gold and silver intake, a methodology which has introduced greater transparency for the refining industry. In other words, this marked the first time that the wider precious metals market could start to understand, with some granularity, the relationship between producing and recycling locations on the one hand, and the main GDL refining centres for these two metals on the other. Fast forward to this year and the data release covering 2020 means that not only is three years’ worth of data now available and improved coverage, but the time period under review arguably takes on greater significance, given that it covers both the height of the pandemic and the exchange-for-physical (EFP) crisis.
Before delving into the two datasets, it is worth in particular understanding what “recycling” means to the GDL refiners, that completed the LBMA survey, compared with the definition used by MF. For the former, a good starting point is to look at their Sustainability and Responsible Sourcing Report, which incorporates the CoO analysis, including the methodology. In short, LBMA’s recycling includes the melt of GDL bars. Below is a summary of the MF approach to recycling:
- Includes: old jewellery and silverware that is sold back for cash and which is captured where the scrap is generated, not where it is refined
- Includes: for ethylene oxide (EO) catalysts, silver is captured where it is refined, not where the EO catalyst is installed
- Includes: recovery of gold/silver from end-of-life products outside of jewellery and silverware, which is captured where it is refined
- Excludes: old jewellery that is exchanged for new; production or process scrap from all areas of gold/silver fabrication
- Excludes: the melt of all bars, including Good Delivery Bars
Even though the different approaches to recycling can make it difficult to compare the two series for some markets, as the following analysis sets out it is revealing how many similarities do emerge, for both gold and silver, especially once it becomes clear how some of the data is classified by GDL refiners.
Turning to the CoO data and looking first at gold and the large-scale mining (LSM) data. This presents one of the closest matches between LBMA, of 2,467t in 2020, and MF at 2,814t. This suggests that 347t is refined by non-GDL operations, for example including (but not limited to) those based in the United Arab Emirates (UAE), mainland China or India. Interestingly, the LBMA total is up almost 9% on 2019, compared with a 2.5% drop for MF. The latter is mostly due to the impact of COVID, whereas the upside for LBMA might reflect improved coverage of their GDL refiners in the context of this questionnaire.
|Global Totals, 2020||LBMA (t)||Y/Y (%)||MF (t)||Y/Y (%)|
Mainland China offers a good example of the disparity between the two. LBMA reports a jump from 235t being refined from mining in 2019 to 317t a year later. This compares with a 3.9% y/y decline in the MF series for Chinese gold mine production, to 368t in 2020, a 10-year low. One complication may be a misallocation of some recycling, that was reported to LBMA, which in fact might include some mine supply. A comparison of the combined Chinese gold mining/recycling data for last year shows that 643t was reported to LBMA, against 558t for MF. As discussed in the MF focus box the 89t differential may partly reflect the different treatment of scrap, as the consultancy excludes both “exchanged” recycling and process scrap.
|China, 2020||LBMA (t)||MF (t)|
By contrast, the 2020 Russian LSM figures are within 8% of each other, at 305t (LBMA) and 332t (MF), although the respective year-on-year changes vary, at 4% and 1%. Perhaps unsurprisingly, LBMA data for 2020 shows that Russian gold mine production was almost entirely processed by local refiners. (As an aside, the vast majority of this metal was then shipped to the UK that year, as well as in 2021.) It is also possible that some of the gold, reported to LBMA as being recycling, may be mine-related given that the combined mining/scrap total of 356t almost matches MF’s figure of 363t.
|Russia, 2020||LBMA (t)||MF (t)|
Staying with gold mine supply, LBMA data for the Philippines points to another potential methodological difference with the reported data. First, the 2020 LSM (33t) and recycling (12t) figures appear too high. These compare with MF at 26t and 3t respectively. Breaking these down further, MF report 17t of LSM and 8t of ASM. While the latter is estimated it is revealing that all of the material from the Philippines that was treated by GDL refineries was labelled LSM. This could reflect how some of the material is captured and initially treated in the domestic market, as there may be occasions when some ASM might be purchased by a nearby LSM operation before being exported. The other issue concerns the high silver content of some LSM operations in the Philippines and whether some of the quantities have been reported on a gross or fine basis (which may apply for gold treated at Chinese GDL operations).
|Philippines, 2020||LBMA (t)||MF (t)|
Keeping with East Asia, two other markets are worth highlighting, Indonesia and Thailand. In the former the gap between LBMA and MF for LSM mine output, of 41t and 66t, respectively, largely reflects the 29t of gold extracted from by-product copper production (at Grasberg and Batu Hijau) in 2020. The CoO for this gold may be captured where the concentrate is smelted (this concept is explored in more detail in the silver analysis below). The other noteworthy figure is the LBMA recycled gold of 112t for Indonesia, which far exceeds MF’s estimate of 29t. This could mean that some of the recycling includes ASM, which MF indicates at 35t in 2020, although it is worth noting that, in keeping with all MF ASM figures this is estimated and so may be subject to revision.
|Indonesia, 2020||LBMA (t)||MF (t)|
With regards to Thailand, even though modest in scale, it is difficult to explain the difference in the mine production data for LBMA (close to 10t) and MF (just 1t). The recycling figures also appear very different; 219t reported by GDL refiners to LBMA in 2020, against MF’s estimate of 76t. However, an important share of the 219t reflects the way in which the Thai retail gold investment sector works. Short-term trading is commonplace, which sees investors buy and sell physical gold on a frequent basis. As a result, gross investment will far exceed net demand. 2020 was also an anomaly, with net disinvestment surging, basis MF figures, to 88t in 2020. Some investors also took the opportunity to enter the market, and so it is quite conceivable that the 219t reported by GDL refineries can be reconciled with MF’s net liquidations of 88t and our 76t of gold recycling.
|Thailand, 2020||LBMA (t)||MF (t)|
Returning to ASM, it is perhaps surprising that, in total only 16t was reported to LBMA as being processed by GDL refiners. This compares with MF’s 2020 estimate of 660t for global ASM output. One reason for this was touched on above, which, although slim, could involve some ASGM being co-mingled with doré from small LSMs before it is delivered to a refinery. Most important though is the fact that an important share of ASM is treated by non-GDL entities, for example in the UAE or India. However, the release of LBMA’s ASM Report at the 2022 LBMA/LPPM Conference in Lisbon, which sets out how GDL refiners can potentially increase their exposure to ASGM, could see ASM processed by GDL operations start to rise.
Gold Mine Production: A Comparison of LBMA and MF Data, Selected Markets
|Tonnes||LBMA LSM||MF LSM||LBMA ASM||MF ASM|
Source: LBMA, Metals Focus
At first sight, the silver data reported to LBMA appears more difficult to compare to MF’s statistics. In particular, the processing of 1,000oz bars may inflate some recycling figures; it can also impact the reported mine production data. LBMA and MF silver recycling figures can also vary. For example, as discussed in the introduction, MF include the change-out of spent ethylene oxide (EO) catalysts, with the silver often recovered by non-GDL operations.
Understanding the CoO also throws up some challenges. For example, when a silver-containing base metal concentrate is shipped to a smelter, the outturn in the form of a silver-bearing doré may understandably (given co-mingling) be reported by the GDL refiner as the COO, rather than where the base metals concentrate was actually mined.
For some GDL operations with base metal smelters it may also be difficult to differentiate between silver (and other precious metals) that are recovered from mining concentrate as opposed to industrial scrap. As a result, it may be more instructive to look at combined silver recycling and mine production for LBMA and MF. For the former, this generates a total of 30,406t in 2019, falling to 27,107t a year later. With regards to MF, the year-on-year change is less pronounced, with the combined total dropping from 30,600t in 2019 to 29,451t in 2020. Even so, it is encouraging that the figures broadly match.
|Global Totals, 2020||LBMA (t)||Y/Y (%)||MF (t)||Y/Y (%)|
Despite apparent challenges with reviewing production and scrap data it is still worth looking at some of the individual figures. LBMA data suggests that silver mine production rose from 16,176t in 2018 to 19,479t in 2019, before falling back a year later to 17,355t. By contrast, the three-year MF series shows: 2018, 26,444t; 2019, 26,001t, 2020, 24,293t. Leaving aside variations in the global totals, the trends are not that dissimilar. The year-on-year gain for LBMA in 2019 reflects improved data coverage, while both LBMA and MF in 2020 highlight the impact of COVID on silver mine supply.
Breaking this down further, primary silver production was the hardest hit in 2020, exceeding the decline in by-product silver output. On a regional basis, the most sizeable losses were concentrated in South America. Interestingly, this is difficult to discern from LBMA data for the Americas (which includes North America) as their regional total falls sharply from 13,785t in 2019 to just 9,700t a year later. (By contrast, MF data points to a 11% year-on-year drop, leaving silver production for the Americas at 13,496t.) There are examples across the region of an apparent collapse in silver production as reported to LBMA, which makes it worthwhile in exploring the data in a little more detail.
Silver Recycling: A Comparison of LBMA and MF Data, Selected Markets
|Recycling, 2020||LBMA (t)||MF (t)|
As touched on above, the apparent discrepancy between the two production datasets may lie in where base metal concentrates are refined and how this is categorised. For example, Japan reports 2,039t of recycled silver, and although the CoO also denotes Japan it is in fact a sizeable importer, especially of copper concentrates, and particularly from the Americas and Australia. Staying with East Asia, LBMA data shows 1,882t of recycling originating from China/Hong Kong, which compares with 784t shown by MF for 2020. The mainland is also a sizeable importer of base metal-bearing concentrates, also from the Americas and Australia. Other key base metal smelter locations (that refine silver) are Germany, Belgium, Sweden and Canada, although of these three only Germany appears in LBMA data as a sizeable recycler in 2020, totalling 1,678t, against MF’s estimate of 297t for that year.
Although a somewhat crude approach, combining LBMA recycling volumes for Japan, China, Hong Kong, Germany, Belgium, Sweden and Canada for 2020, less MF’s silver scrap supply estimates, gives a total of 4,483t. If we also look at some of the key differences between LBMA and MF silver mine production data, we arrive at the following:
Silver Mine production: A Comparison of LBMA and MF, Selected Markets
|Tonnes||LBMA Mine Production||MF Primary/Gold By-Product||MF Base Metals By-Product|
Source: LBMA, Metals Focus
Even though there are still some important discrepancies, this does suggest that silver recovered as a by-product from base metal mines is being defined as a secondary feed, with the CoO often listed where the base metal concentrate is smelted. At a global basis, it is also worth putting into context how important each area of silver mine production is. According to MF, the relevant shares for 2020 were: primary silver mines, 27%; as a by-product from gold, 15%; from copper mines, 26%, from lead/zinc, 31%, with other sources at 1%.
Moving away from silver mine production, there appear to be few material differences between the two recycling series. One that does stand out is the US (as shown in the table above), but this largely reflects the recovery of silver from spent EO catalysts, which falls outside LBMA’s data coverage. Another is India where LBMA reports 166t of silver recycling against 495t for MF in 2020. Even though the country features three GDL refiners, the total refinery count in India is much higher; out of a total of 43, some 16 would treat silver. More important, old silverware and jewellery is usually processed in-house by manufacturers and/or retailers, and so there is often little need for this to be specifically sent to a GLD or non-GLD refiner.
At the global level, if we subtract the estimated recovery of silver from base metals concentrates from LBMA total silver recycling, this leaves 5,266t for 2020. This appears to compare favourably with MF’s total of 5,142t, that includes silver recovered from spent EOs (much of which would not be treated by GDL refiners). MF also includes silver treated by non-GDL, particularly in India and across much of the Middle East, although the latter is a small silverware/jewellery fabricator and consumer. That said, it is perhaps not surprising that LBMA series is higher, as this covers total silver recycling, and so will include production (or process) scrap which is not included in MF’s recycling datasets. Overall though, given the estimated nature of MF’s silver recycling and the inherent challenges for both organisations in accurately capturing this data it is noteworthy how much these two series, at the global level at least, appear to match.