On the imbalance between buy and sell volumes in the LBMA survey

In 2011 the LBMA surveyed its members as part of getting gold recognised as a zero weighted risk asset under BASEL 3. The survey was unique because it attempted to show actual gold market trading volumes, not just the net clearing figures the LBMA provides. The survey showed sales turnover of 5,593,473,000 ounces and purchase turnover of 5,350,183,000 (173,985 tonnes and 166,409 tonnes respectively).

Blogger FOFOA speculated that the difference of 7576 tonnes represents net sales and and increase in paper gold by bullion banks:

"From that LBMA survey, we can see that the LBMA had net sales in one quarter of 7,575 tonnes of paper gold. That’s a gross increase in the amount of paper gold in existence over only three months. 100:1 actually seems conservative in this light. That’s most likely FOREX use of gold as a hedge or a currency play. But even still, the BBs have to hedge their price exposure when selling that much paper gold. Without a hedge, that would be a 7,575 tonne naked short position for the BBs.

...

There’s no way they hedged all of that in the “gold” market (Comex/mining forwards/GLD). It’s simply not big enough to absorb that rate of flow without rising a lot faster than we saw it rise. So the BBs must be hedging this exposure the way they hedge net positions against other currencies in the FOREX market, simply using complex formulas and derivatives that look at correlations between different things."

FOFOA restated this idea in this interview: "The rate at which the banking system created "paper gold" was 11 times faster than real gold was being mined."

I believe this is an incorrect interpretation of the survey, as I commented: "The reason there is a difference between buying and selling volumes is because not all LBMA market participants were surveyed yet they divided in half all inter-bank volumes (so as not to double count)."

FOFOA responded here and we had some offline email discussions with FOFOA posting a response here. That is the background to my response to FOFOA below.

FOFOA: "Bron has a serious problem with my explanation"

My reaction is driven by a gut feeling that 7,576 tonnes paper creation hedged by some synthetic construction "using correlated asset derivatives" is just unrealistic. An oft made point about gold is that it is not correlated to other assets and indeed its correlation to them changes over time. As a result I am skeptical that one could construct an ongoing synthetic long gold position which would not blow up - just look at the problem JPM had with its CIO's synthetic credit portfolio.

The thesis is that 7,576 tonnes of paper gold was created in Q1 2011. The assumption is that this is not a one off and the outstanding paper gold position is much larger. I understand the FX and other markets are large but if the banks have been doing this for some quarters and not just Q1 2011, then we may be talking some significant positions in the "correlated assets", giving rise to the "whale" problem that JPM had/has.

FOFOA: “some reason why half of the banks sell more to clients while the other half buy more from clients, and then why the sellers (on average) reported and the buyers didn't”

Some reasons as follows. Except for the market makers (and perhaps only the largest ones), each bank would specialise in certain market segments/customer bases. For example, an Australian bank will have a competitive advantage in banking for Australian miners, so their bullion desks would be always buying (paper) gold from miners and selling it to other London banks. We take the miner’s physical and are always (net) selling, which we would layoff with another London bank. Similar would apply to those targeting the jewellery industry, scrap flows, etc etc. I therefore think it is more likely than not that each of the 56 banks would be either net buyers or sellers, with possibly only the 11 LBMA market makers having a more even balance given their central role in the web/network.

I think it is statistically impossible for any bank over a period like a quarter to have clients buying and selling exactly the same amount of ounces. The question is what is a reasonable buy/sell difference. Those with a wide and diverse range of clients (eg market makers) are more likely to have buy and sell volumes closer together whereas a bank who specialises in a market/region or customer base to have a bigger difference. This leads to the quote below.

FOFOA: "the 20 non-reporting banks on average would have had to be buying 11.3 more tonnes each day from their clients than they were selling to clients. Meanwhile the reporting banks would, on average, have to have been selling 6.3 more tonnes each of the 67 trading days to their clients than they were buying from clients. Hopefully you can see why this requires more than just a statistical explanation." [note: the LBMA survey was over 63 trading days]

I see those sort of tonnage volumes to be entirely reasonable considering the amount of paper trade that goes on. Perth Mint for example refines about a tonne or so of gold a day, so we banks would be reporting us as 63 tonnes of net sales. That is just physical and only one small part of the overall mine physical and paper trading. I'd like to do a bit more work on volumes in other markets to get a handle on this issue.

FOFOA “Each trade is assumed to be random as are the choice of non-reporting banks, otherwise we have an "unusual event" that needs explaining, but Bron is not arguing for that and neither am I.”

Trades may well be random in aggregate, but the choice of non-reporting banks is highly unlikely to be random IMO. As intuitivereason noted (July 2, 2012 5:32 AM) 56 is a small sample: "at 36 of 56, there is still a lot of statistical variation possible, let alone allowing for selection bias - there may well also be a greater tendency for those who were net buyers to not respond."

With the commercial sensitivities involved, it is highly likely that there is some skew in who did not report. The banks all compete against each other and given my comment above about each specialising, some of the banks who are net buyers may not have trusted the LBMA process so declined to report, particularly if they are relatively smaller in size. Keep in mind all of the market makers were included and they are likely net SELLERs, (that's where the COMEX short position comes from and would show up on this survey as “client” sales).

I think burningfiat's explanation (July 2, 2012 3:27 AM) that "which 20 LBMA members choose not to participate in the survey? Speculation: Mostly the one's with secretive giant dip buying clients?" should also be considered as a valid point.

FOFOA: "we have three possibilities, one of which is true"

This is too dogmatic, I think it is more likely that we have three possibilities, ALL of which are true. Why can only one be a factor?

There is also a fourth factor. While I agree that the banks don't have a huge naked short position, they would have some long or short position at the end of Q1 2011. We need a guesstimate as to what sort of internal risk limit would be applied - that is how many ounces are they allowed to go short or long.

I think to get to the real net paper creation you have to take the 7,576 tonne difference and first remove for the effect of any skew in who did/didn’t report and factor in the “divide by 2” effect on that skew. You then need to take off any likely net long/short position. What is left needs to be explained. I am going to refine my model over the weekend with the objective of trying to remove these other factors.

I think the LBMA survey has given us an insight into the usually opaque gold market. Understanding what this survey is really saying is a work in progress. Any ideas/comments welcome.