Forensic Basis Analysis of Apr 12-15 Gold Crash

With all the discussion of the crash of April 12-15, I thought it would be interesting to do a little forensic analysis. Instead of looking at price graphs, let’s look at a graph of the gold basis. Contrary to the currently popular assertion that the gold price crashed due to manipulation, the evidence points to something else. The question of whether gold is manipulated by the “naked” selling of futures, or whether other forces are responsible for gross price movements, is important to anyone who wants to trade the metal.

For a quick tutorial, the basis and cobasis are spreads between the spot price of the metal and the price of a futures contract. Basis is the profit one could make to carry the metal, which is to simultaneously buy the physical metal (at the ask) and sell a future (on the bid). Basis = Future(bid) – Spot(ask). Cobasis is the profit one could make to decarry, which is to sell physical and buy the future. Cobasis = Spot(bid) – Future(ask). Basis and cobasis are quoted as annualized percentages, so they can be compared to other yield opportunities in the market.

Carrying and decarrying are arbitrages. The arbitrager cares about spread and changes in spread, not about price or changes in price.

Here is a graph showing the long term basis and cobasis curves for Apr 11 before the crash, the 12th the first day of the crash, the 15th the second day, and the 16th the day after. Each line represents one full basis curve from present through the end of 2016. What we’re interested in, on this graph, is the change from one line to the next, which represents one day to the next.

gold longterm carry

First, let’s look at the cobasis because it is simpler here. We see volatility only for the June contract. From April 11 to 12, it rose a small amount, then for the remaining days it fell fairly sharply and was no longer in backwardation (backwardation is when the cobasis is positive, i.e. decarrying is profitable, i.e. the physical commodity is scarce). All farther months remained stable. There is no evidence here of massive dumping of futures or of anything else. The fact that the cobasis fell means that the offer on the future rose relative to the bid on spot. The offer on the future was robust during the crash. Make of that what you will.

The more interesting thing to look at on this graph is the basis, and the interesting part is mostly in what changed on April 15. We can hardly see the dark blue line for the 11th because the green line of the 12th is on top of it. This is normal and shows nothing untoward happened on the 12th.

But on the 15th, All Hell Broke Loose.

On the 15th, the basis was collapsing especially for contracts in 2014. Contracts in 2013 and 2015-2016 we affected but not anywhere near to the same degree. The falling basis means the bid on the futures contract dropped.

Whoever normally makes the bid on these distant contracts did not show up to the market that day. One might argue (indeed many gold commentators do argue) that there was massive naked selling of futures contracts. There are four reasons why I do not believe this allegation fits the evidence.

First, the basis is a very small number and a very sensitive indicator. Aside from Apr 15, it is in the vicinity of 0.4% annualized. That is roughly $6 per year (assuming a $1500 per ounce gold price). If the price of gold was smashed down by more than $200 via selling of futures, then the change in the basis would be much, much larger.

Second, as I wrote previously, the open interest did not change much during the time frame, and actually fell from the 12th to the 15th.

Third, by far the largest change in the basis was for the August 2014 contract. For the days in question, there were less than 1000 contracts open and there was minimal trading volume. This is not the contract that would offer leverage to change the gold price.

Finally, the cobasis did play along as it would have in case of futures dumping.

Now that we have discussed what the basis data does not show, let’s look at what it does. First, the cobasis falls on April 15 and the next day as well. It had been above zero—in backwardation—but it fell below zero. This indicates that physical metal came into the market. This pushed down the bid in the spot market. Since cobasis is Spot(bid) – Future(ask), the cobasis fell.

Second, the basis for futures contracts in 2014 (centered around August) dropped precipitously. Since basis is Future(bid) – Spot(ask), we know that the bid on the contracts in 2014 fell. While this could mean that there was concentrated selling, it does not make sense why it would be in such a distant contract with low open interest (and the low volume does not support this theory).

My conclusion is that on April 15, liquidity dried up. Whoever normally makes the bid on these distant contracts did not show up to the market that day. This would fit with a general picture of balance sheet stress and credit contraction (especially in Europe).

38 replies
  1. Mo says:

    I really appreciate your perspective. . .(tho it makes the brain hurt a little in the struggle to comprehend).

    We’re getting a golden waterfall this morning. What sources do you use to acquire the data necessary to assess the situation in the manner you did in this article, if you don’t mind sharing?

  2. petter_w says:

    The basis curve for 4/15 is a lot more erratic than on the other dates.
    I did my master thesis in computing error probabilites in communication systems through computer simulation. If you wanted to measure an error rate of say 10e-4 you would have to simulate 10e5 bits to count 10 errors to measure the error rate within a certain confidence interval (i.e. you are 90% sure assuming a normal distribution of the signal level around the mean that your result is within a range of +-50 %). If we would see graphs like 4/15 basis and especially for the june contract we would say that there is not enough data which in the case of financial markets is illiquidity.

    I am sure that when you calculate the basis/cobasis that you are measuring trades within certain time intervals and calculate mean values or that you match up values somehow. For example to get the basis for one day I would need to look at all the spot ask and match them up with future bids during a time interval (I would assume within seconds due to price fluctuations) and then average all the basis values over all the time intervals to get one basis value for a particular day.
    What stands out here is that the cobasis curve seems to behave a lot better than the basis curve, in other words it is smoother. Why is that? I assume you use the same methodology to calculate basis and cobasis.

    The only explanation I can come up with is that
    Future(bid) – Spot(ask) (basis) has less data than Spot(bid) – Future(ask) (cobasis), in other words there is more bidding than asking in the spot market and more asking than bidding in the futures market. I am not sure such an assymetri is possible, but if it is then it means the interest in cash gold (spot bid) is higher than the futures market (future bid) and vice versa. If so it means that the pysical market determines the price.

  3. Keith Weiner says:

    Mo: one could use any realtime quoting service. I use eSignal.

    petter: thanks for sharing some very interesting thoughts re statistical analysis. I believe the data is sufficient for both basis and cobasis (it’s the same number of samples, taken at the same times, etc. for each). There is a continuous bid and ask, btw; at any given time there is always a bid and an ask at least for gold which does not flash crash to $0.01 like stocks have done. I believe the bid truly dropped especially for the Aug 2014 contract, which means either that there was lots of selling of the contract on the bid or the party who normally sets the bid was not playing on April 15.

    • petter_w says:

      Thank you for the answer.
      The basis curve 4/15 is erratic also for prior to june/aug 2014. What is interesting is the asymmetry in the smoothness. With bid/ask in spot/future being continuous signals (which they have to be, someone will always want to buy/sell) then I would expect the basis and cobasis graphs to be equally smooth. Can you comment on this asymmetry in smoothness?

      • Keith Weiner says:

        The asymmetry is not due to desire or lack thereof to buy/sell but to the rate of change of the price at which they want to buy/sell.

        It could very well be that a holder of longer-dated futures needed to sell. Those contracts are thinly held and thinly traded even in normal times, and on April 15 it was not normal times. The upshot is that it would not have taken much selling to push down the bid significantly.

        • petter_w says:

          I guess that to calculate the basis for one day (get one data point) you will sample every few sec and get thousands of data points and make an average over all samples.

          A side note:If we restrict the discussion to 15/4.
          The future bid on the feb-14 and aug/oct-13 contracts seem roughly equal but dec-13 is higher.

          I guess a fundamental problem with calculating one data point from thousands of samples is that in unnormal times like 15/4 you can get strange results, if – in the case of the basis for arguments sake – future bids move more rapidly than ususual or is thinly traded.

          I guess another approach to the problem is to look at the basis and cobasis after the plunge. If gold splipped into backwardation again as your weekly basis letter says then wouldn’t that be a sign that physical metal was sold 12/4 and 15/4 but is then becoming scare again at the lower prices? Can you comment on that?
          And if my interpretation is right, is it possible that the physical metal that was sold was warehouse receipts or other promises to deliver metal that might or might not be there?

  4. monetary says:

    How fine grained is your data? The source of the “price manipulation through future selling” seems to be the following article, which states the key events took place in a span of 30 min each. If the futures selling “led” the spot selling, that would suggest something different than if the spot selling happened first.

    • Keith Weiner says:

      monetary: You may be interested in my previous articles on the crash of Apr 12-15 including the one where I look at open interest. Bottom line: open interest declined, it did not rise as one would expect in a scenario of creating hundreds of thousands of new contracts to dump on the market. Also, if demand for physical was strong and the suppression occurred only in futures, then there would have been a massive and rising backwardation.

      I do not look at high-frequency data. If the alleged manipulation occurs and is done within a fraction of a second, I would probably not see it.

      • monetary says:

        I’ve read it again, thanks. Does your data therefore effectively rule out a massive sale on either the spot or the futures side of the equation, because it would have dropped either basis or cobasis off the chart?

        We’ve got three possible scenarios here:
        a) Massive price drop in futures bringing down spot
        b) Massive price drop in spot bringing down futures
        c) Massive price drop in both simultaneously

        a) you seem to have ruled out fairly thoroughly, UNLESS the futures drop was fast enough that it triggered automatic bulk selling of spot, bringing down the spot price so it looks to your data as if the two moved together

        b) nobody’s really considered. What if someone decided to dump 400 tonnes of spot on the market? (Or, as the ZHs would assume, 400 tons of spot tungsten) The sustained open interest seems to rule out the price drop being the result of someone “not showing up to market.”

        c) ??? here be dragons. Maybe “something big happened behind the scenes,” consistent with the news that Cyprus might be selling their gold, except on a much bigger scale…

  5. mossmoon says:

    There is an obvious logical fallacy at the heart of KW’s basis critique of manipulation. He is claiming that an analysis of prices that may or may not be manipulated reveals the manipulation of those prices. This is textbook fallacy of self-reference.

    When I read it I am reminded of Krugman talking liquidity traps or whatever. It all rests on circular reasoning. (No insult intended.)

    The bottom line is that he cannot prove that physical metal determines price. He stubbornly ASSUMES that spot is set exclusively by an offer from physical. He cannot prove that and yet HE NEEDS TO PROVE THAT for basis analysis to be relevant. Fekete himself agreed that spot does not equal physical. If physical metal is not determining price, then spreads are completely irrelevant to manipulation. Period.

    • petter_w says:

      The problem with the spot price and the cash market is that delivery is rarely taken but that most contracts are settled in cash. The possibility to take delivery legitimizes the Comex/LBMA system. As we have seen in Cyprus and with MFG the rules can change overnight.
      I personally think the basis anylsis is valid, but I think it is naive to suggest that gold that is on someones balance sheet has not been leased out. At the end of the day it will be like musical chairs because there more than likely are duration mismatches in the gold trade.
      What I am not sure about is the definition of cash gold (see my above post). Does it include warehouse receipts or gold that has been leased out. As long as the market believes the physical gold is there everything is fine.

    • monetary says:

      I suggested in one of the other posts that the dataset also include a representative street-market price for gold bullion to rule out just this possibility. If the price of a Maple Leaf or Philharmoniker in your home town starts seriously diverging from the spot price (+ usual premium), that would be interesting to note. However, it’s not clear that this is happening at the moment.

      As long as you can buy physical gold to hold in your hand for near the spot price, then Keith’s analysis should hold. The question of whether all the gold the markets think exists actually does is academic — I think — provided spot is an accurate measurement of physical gold prices. If it turns out that the gold on which the spot price is based is just as hypothetical as the gold on which futures prices are based, then we would expect to see a “backwardation” of sorts on the physical market rising strongly.

      • JR says:


        I have it on good authority, from Bron at the Perth Mint (since he works ‘in gold’ I’ll believe him before others), that the vast majority of gold traded ‘wholesale’ is actually in the form unallocated, or, gold bank deposits.

        Unallocated is in fact the form in which the Perth Mint keeps its account ‘loco London’.

        Thus the ‘spot’ price quoted by Keith & others is unallocated gold. FWIW, just one more reason to be sceptical of claims of ‘foretelling’ by the basis. And no Keith, I’m not accusing you of such claims.

        Also FWIW, the Perth Mint has never had any issue with redeeming physical gold from London & at times it imports it by the pallet load.

        • JR says:

          I might add, my own personal preference for looking at ‘paper gold’ trading at a discount to physical – hardly a revolutionary idea – would be the spread between the bid for Perth (or other) Mint physical & the bid of the Mint for gold loco London.

          So far, no banana. Make of that what you will, but I call it an irrational, so unmeasurable, bid for $.

        • bronsuchecki says:

          The spot/cash price quoted on Bloomberg or Reuters is for unallocated gold, but that unallocated gold is covertable into physical. What is often not considered by many is that while bullion bank unallocated is fractional, there is still a significant physical market operating underneath that and which uses unallocated to settle.

          So currently spot unallocated = spot physical.

          • JR says:

            I have trouble seeing where COMEX even fits into this equation. What percentage of the wholesale gold market runs through US gold futures?

        • monetary says:

          Thanks. It looks like BullionVault has a good explanation of allocated vs unallocated:

          What I don’t fully understand is the following… if unallocated gold represents debt and not possession, what’s to prevent the debtor from taking out a “double mortgage” and selling unallocated gold to more than one person?

          If that’s possible, and the vast majority of spot gold on the market is unallocated, then we really do have no idea at all how much physical gold backs the whole thing.

          If that’s so, then there’s also a wonderful way of manipulating the spot price and therefore the basis without ever having to part with “fizz.”

          Anyone remember “The Producers”?

          What would happen if someone simply decreased the “fraction” of real gold backing the fractional buillion bank unallocated reserves?

  6. bronsuchecki says:

    JR re COMEX, probably most of it I think apart from maybe large institutions and others who deal direct with bullion banks and want the flexibility only the OTC market can give.

    • JR says:

      Qualitative limit Bron, QUAL-itative limit. There is no quantitative limit to the monetary system.

      Also Bron, perhaps “unallocated = spot physical” is not the best way to describe it?

      How about utilsing the computer terminology –

      unallocated := spot physical

      As in an assignation. Hate to break it to you but if you consider unallocated equal to spot physical, you disagree with Fekete & agree with Mises about a promise to redeem gold being the same as gold.

      Unallocated is not physical, so my above point stands. It might be money good, but it is not money.

  7. Keith Weiner says:

    I will just add reiterate one comment. Anyone who thinks that there is a “paper” price which is much lower than the “real physical” price please let me know where one can sell real physical gold at the “unmanipulated” price of $1600, $1900, or the $30,000 that gold is “supposed” to be without the manipulation.

    If this is the “obvious” fallacy in my theory, then it should be pretty easy to tell me where there are buyers who will pay the real price.

    If not, then what you have to admit is obvious is that there is (today) one price for gold. The spread between futures and physical metal is tiny–so small that other than bullion banks and a few followers of the basis theory, no one else seems aware of it.

    If paper gold and physical gold trade at virtually the identical price…

    • petter_w says:

      I think the speculation or conspiracy theory revolves around gold that has been leased out but is still carried on the books as gold in the inventory.
      There is also a basic trust issue. Given MFG, Madoff, Cyprus, the mortgage fraud, IndyMac, tons of missing cash in Iraq, the slow return of the German gold and an endless list of shenanigans and that gold is real money it is not unreasonable to entertain the idea that manipulation is going on.
      In Vietnam you could find people that are willing to pay $1600 or more for an ounce since the premiums there are 20 % according to a chart from FeketeResearch.
      $1900 and specially $30000 are harder.

  8. Keith Weiner says:

    Also, thank you to everyone here for scrutinizing the basis theory and the evidence I’ve been presenting.

    I would also encourage you to demand the same for the manipulation conspiracy theory. Otherwise this could look like the debate over creationism.

    • mossmoon says:

      I am not saying there are two prices. I am saying the physical price is a derivative of the paper price, not the other way around. But as I referred to above, this will just go round and round.

      The evidence for manipulation lies outside of the basis analysis. That’s all I’m saying.

      For example, what is point of leasing a bar of silver? There are basically two things you can do with a bar of silver. Use it in an industrial application or sell it to someone else. Why would anyone (legitimately) rent a bar of silver? You don’t rent a bar of silver like a butcher doesn’t rent a slab of meat. You buy it. Lease means sell. Billions of ounces of silver that do not exist have been sold as real silver through these fraudulent contracts. These are contracts you call “cash” metal. But they are not metal. There is enough of this paper metal sold to infer that it seriously distorts true price discovery. These contracts are not theoretical, they are real and they are fraudulent. Whether or not they arose out of a “conspiracy” is irrelevant to me.

      Personally I think the basis analysis and your insights are fascinating enough without trying to debunk conspiracy theories.

      • monetary says:

        I made this point earlier: as long as physical tracks spot prices, it’s academic. At the moment, it does: the premium on a 1 oz coin in my area has held steady at 2% +/- 1% since the crash.

        If, however, it’s possible to create “spot” gold out of thin air, then that represents a possible explanation for the price crash. It also means that all the numbers we’re dealing with (including the price of physical gold at your corner gold shop) are not necessarily driven by the forces we believe them to be determined by… kind of like making scientific observations from “A Bug’s Life” versus the “National Geographic Channel.”

    • JR says:

      The manipulation conspiracy theory is bunk, I’ve not seen any actual evidence – heresay & speculation is not evidence – that it occurs.

      Irrational behaviour leads to all sorts of ridiculous explanations, because no one knows what will happen. Not even me.

  9. Keith Weiner says:

    If paper is manipulated and physical is in demand then what would motivate sellers of physical to lower their ask? This is not a “round and round” but a pointed rhetorical question. Why would owners of real metal part with it at a price that paper manipulators seek to create? Or another way of looking at it is: who is dumping his physical metal at $1350-$1450 today and why is he doing it?

    This is why I keep asking where can Is sell a real bar of gold at $1900 or $1600?

    By the way, silver is money, unlike a slab of meat.

    • monetary says:

      What defines the value of gold in dollar terms? The price other people are willing to pay. If 90% of the gold market is selling near spot and you decide to sell, you probably will sell near spot too — unless you have a really good reason to believe it’s worth more and the price is likely to rise.

      This is not true for cultures where gold has mystical / traditional value, which is why in India and China we’ve seen precisely what would be expected if the “spot” price dropped below the “real” price… massive buying and shortages.

      Since the investment market sees gold primarily in terms of its monetary value, that’s not happening. The “manipulated” price IS THE REAL PRICE, because everyone accepts it as “real.”

      The key distinction is that everyone accepts it as real while believing it to be the product of free market forces. (compared with e.g the yen)

      If the latter is not true and the price is not just the product of free market forces, there is in gold an “unknown unknown”… something that most investors don’t know they don’t know.

      While we don’t know exactly what the magnitude of this quantity is, we can make some guesses as to its direction and effect if it were to exist. If the gold price is being manipulated by leasing out the same bar two or more times, then the spot price is being artificially suppressed and the “unmanipulated” price would be significantly higher.

      Now, as you point out, that “unmanipulated” price is a mythical, theoretical quantity that only becomes real if the manipulation is discovered or for another reason fails. The old adage, “the market can stay irrational longer than you can stay solvent” suggests a belief in its existence is a poor reason for investment in and of itself… unless, of course, you have reason to believe the “curtain” will be lifted. Hence the interest in premiums of physical gold over spot, which would act as an indicator.

      • monetary says:

        To follow up: there’s a report over at Bloomberg (search for “TD Securities”, I can’t post the URL here) that the used gold supply has dropped to a five year low, while Perth Mint sales are at a five year high… and Indian imports are expected to surge 47 percent in 2Q 2013 just to meet demand.

        As to price vs physical connection, worth quoting: “Physical demand or the scrap market is not the main indicator for prices as the paper market dwarfs both these markets.”

        In other words, the paper gold market is so much larger that if manipulation were taking place, we would not see “gold bars at $1900.”

        • petter_w says:

          monetary: The drop might have been influenced by selling to achieve a certain objective but as long as physical gold is bought and sold at the spot price (or very close) the physical market determines the price.
          At the end of the day the physical market determines the price. It is outside the scope of a basis analysis to determine duration mismatches due to gold that has been leased out etc. The paper gold market can best be described as gold that has been re-hypoticated umpteen times. Where the rubber meets the road is when a lot of players want physical. Another way to look at the gold market is that it is leveraged. It is well documented that central banks have been dishoarding gold and they might have an interest in keeping it down. But regardless – the physical market has to determine the price. The basis will explode and the cobasis will collapse when people no longer trust the phoney credit system and want gold or when the whole crazy credit based system simply collapses under its own weight.

    • virgule says:

      I’ll venture an answer here. Please don’t shoot at me, I’ve been reading KW for a long time, but I do not claim to be an expert in anything – I’m just an individual with opened eyes, and trying to apply common sense.

      I was at Hua Seng Heng in Bangkok when gold shops re-opened after a 3-day holiday, right after the April crash. I’ve never seen a crowd like this, with queues 20-deep throughout the entire length of the long counter, and back-flowing into the street. It was 2 PM and they’d been trading for over 4 hours already. I like to observe the dealers while they ply their trade, and that day was particularly cool as blokes in their 20’s would get out of a taxi with a rucksack, drop aprox. THB 3M in cash on the counter (USD 100K) and walk away with 2 x 1kg bars several times per hour (plus 100’s of people buying gold jewelry by weight). Shelves were half-empty – but trolleys were moving around to replenish the shelves. I noticed the floor supervisor was a lot more active on his phones than usual, and the manual controls seemed both much tighter than usual, yet more staffed than usual, so commerce was not slowed down (everything is manual there, zero computers – even when they sell USD 10 million in a day!).

      I asked myself why on earth would Hua Seng Heng sell all this gold at the paper price, with all these people queuing to buy physical? Anyone with a sound mind would be bumping up the price, isn’t it?

      My belief is that the supervisor was monitoring the flow very carefully, and placing buy orders on futures market at the exact same pace, to ensure the sustainability of this trading operation. ie exactly what KW was stating above: gold dealers care about the spread, not the price, and Hua Seng Heng was doing exactly that.

      This of course assumes/implies HSS is confident it can replenish its stock at that new price – and they do seem to be, as I had no problems to purchase USD 150K of gold in small 2.5 Oz bars (5 baht) later during the day (I was concerned and had to wait 30 minutes, but I did get them).

      My conclusion is that as long as the dealer is confident that he will have delivery from the exchange, on the future contracts he is placing to cover his physical sales, he has no reason to not follow the exchange price.

      So the real question to me, is not so much why or where do we buy at that price, but how long can the exchanges continue deliver physical at that price? It’s a question of rate of exhaustion.

      • Paul Fillion says:

        I have been in Thailand for 15 years, not a gold expert.

        However, I can answer your question as to why they didn’t bump the price to take advantage of strong demand.

        They couldn’t.

        They belong to the Gold Traders Association, which publishes prices which all members are required to post and adhere to.

        There are four prices posted, buying and selling jewelry and buying and selling bullion. The prices can change many times during a day.

        If one dealer broke away and tried raising prices unilaterally, they would have no customers, as everybody would go elsewhere.

  10. cayadopi says:

    Any chance the buyer who normally shows up to buy the distant contracts, was an Authorized Participant (“AP”) in GLD or SLV and instead showed up to bid on GLD or SLV in order to retires shares purchased after the huge price drop April 2013 and take delivery of physical?

    (As in “AP” wants physical now, or “AP” needed to deliver physical on a near term short futures position, or private direct sale to another party, etc.)

    Thinking out loud – in trying to understand clues in your work versus clues in other reports,

    1. BigBoy buys physical, sells futures (covered at that point).

    2. Instead of storing physical, BigBoy put it to work and became an “AP” – buy GLD paper, sell physical to the ETF. Then sell at higher price GLD paper to the public (who doesn’t understand the “AP” aspect.)

    3. Enter monkey wrench, BigBuyer wants delivery on that futures contract in #1. BigBoy is now sorta naked, but sorta not, since the “AP” has his physical stored over at the ETF and only the “AP” can remove the physical. All BigBoy needs to do is buy back GLD shares to access his physical.

    4. “AP” takes advantage of a huge price drop in April 12-15 and doesn’t show up to bid on the Jun-Oct 2014 futures, and uses the cash instead to purchase GLD shares NY.

    5. BigBoy retire the GLD shares in London for physical, then deliver the physical for whichever month there is notice of a physical delivery to BigBuyer, taking advantage of another arbitrage situation.

    Could the problem have been simply a near term futures contract requiring physical delivery, with an “AP” having stored the physical over in the GLD vault? Coupled with the balance sheet stress issue being a lack of enough cash to buy GLD at market prices on April 11th, or that buying at market on April 11th would have caused large losses for the “AP”?

    When NOVICE-HERE read the reports that huge amounts of physical are/were being delivered, followed by your insights/analysis and the chart in this article, I couldn’t help but jump to this line of thinking.

    Hope you let me know if I’m totally off-base in connecting the dots this way so I don’t continue down a dead end rabbit hole in something I’m pondering.

Leave a Reply

Want to join the discussion?
Feel free to contribute!

Leave a Reply

This site uses Akismet to reduce spam. Learn how your comment data is processed.