Recently, several gold commentators have stated that there is an anomaly in physical delivery numbers of the COMEX futures exchange in New York. After reviewing the numbers I see no fraud in the data.
Written by Jan Nieuwenhuijs, originally published at Voima Gold Insight.
When it comes to the price of gold, there is always much discussion in the gold space regarding the COMEX futures exchange. Because futures are traded with leverage, they have a significant impact on the gold price. Therefore, it’s always worth scrutinizing COMEX trading to be sure the gold price is a fair representation of supply and demand.
The other day I commented on an article by Robert Kientz titled “COMEX's Gold And Silver Futures Market Trade Data Not Adding Up” (I have been told Kientz’s work is based on an analysis by Kirian van Hest). Although I haven’t read everything by Kientz and van Hest, the gist of the initial article* I commented on was that the number of contracts that were delivered in early September did not match the change in open interest for this contract.
To make sense of the delivery data I’ve collected trading volume and open interest numbers for the September 2020 contract from Nick Laird at GoldChartsRUs.com. CME Group, which is the owner of the COMEX, only publishes granular trading volume and open interest statistics going back four days, so I asked data miner Nick Laird for data going back further. The only dissimilarity I could find is that CME Group discloses delivery data on one of their webpages with a one-day lag. Possibly, this leads to confusion. As an example, on the CME’s delivery notices webpage we can see that on September 9, 2020, deliveries accounted for 243 contracts (red box below).
However, on CME’s trading volume webpage, deliveries on September 9, 2020, accounted for 104 contracts (red box below).
The explanation can be found in the delivery notices report (exhibit 1). One can see that one day prior to September 9, 2020, exactly 104 contracts were delivered. This is because on the delivery notices report it reads in the top left “intent date.” So, the delivery notices report discloses intents for deliveries, which then are being executed the next day (and as such reported on the trading volume page). If I compare delivery numbers for each day of the September contract between the notices report and trading volume webpage, they are all the same, there is just a lag of one day in the latter.
If I make a table of daily open interest, change in open interest, deliveries, the difference between the change in open interest and deliveries (“OI change + deliveries”), and trading volume numbers, everything makes perfect sense. Have a look below:
What we can see, day by day, is that the difference between the change in open interest and deliveries and can be offset by trading volume. For example, on August 31, 2020, the number of contracts physically delivered accounted for 1928, and the change in open interest was -1888. The difference between the change in open interest and deliveries is 40 contracts, but can be explained by a trading volume of 158 contracts. Trading volume always opens new contracts (short and long) but how this affects the open interest depends on the positions of the traders before they entered the trades. Volume can increase or decrease the open interest.
My conclusion is that there is no evidence of fraud, because on every day in the table (exhibit 3) volume exceeds the difference between the change in open interest and deliveries (“OI change + deliveries”).
* The original article by Kientz on Seeking Alpha has been taken offline, but can still be viewed on FX Empire.