16 June 2009

Metal Accounting I

This article discusses the issues associated with keeping track of precious metals. I call it metal accounting because the point is to ensure that ounce debits (ie assets) always equals ounce credits (ie liabilities). This should be of interest to anyone holding unallocated metal because the extent that your "custodian" doesn't have control over their metal activities is the extent that your holding is not backed and thus the custodian is exposed to precious metal prices. If this exposure is excessive, and the price rises, they go bankrupt.

I put custodian in inverted commas because unallocated metal, even if backed 100% by physical, is not the same as a true custodial service, commonly referred to as allocated metal. With allocated, you hold title to the physical metal and the storer is just a safekeeper of your metal. It is off the balance sheet of the storer and control of it is a very simple process: run listing of how many bars your are holding for a client, do a count of the bars in the vault, the two should equal.

Unallocated metal, on the other hand, is on the balance sheet of the storer. This is why it is so important that debits equal credits, from an ounce point of view. At first you may think that the controls around keeping track of allocated should apply to unallocated - if you owe 100oz to clients, then you should have 100oz of physical gold on site. What this article hopefully reveals is that it is not necessarily that simple and an appreciation of the need for stronger controls.

The Golden Table

Let me start with an imperfect analogy for the manufacture of precious metal products: making a wood table. Looking at your plans, you go down to the hardware store and buy some wood and nails, say it costs $100 all up. It is not likely that you will get the exact lengths you need, so some sawing is involved. Whack a few nails in and you have your table.

If I asked you what the table cost, you look at me strangely and say $100 and wonder why I was so stupid. Your answer, however, has made one assumption: that your "by-products" of the table making process are worthless. What are these by-products? They are the wood offcuts and sawdust and your assumption is most likely correct.

Now consider that you are making the same table out of gold. Lets assume the same $100 purchasing cost for the raw gold (it would have to be a really small table) and same process - you have to cut up the gold planks. When I asked what the cost was, would you still say $100? Of course not, you aren't going to sweep up the golddust and throw it and the gold offcuts in the bin like you would with the wood. You would melt them down and sell them to a refinery and the money you would get back would reduce the initial cost of $100. It is like the hardware store giving you a refund for the wood offcuts and sawdust.

This is what makes precious metal manufacture different from normal manufacture and is a function of the high value of precious metals and the fact that you can melt the by-products and reuse them without any or much loss of "utility". My first exposure to this was when I looked at a stocktake count summary and saw a line called "sweeps". It was literally the amount of gold after refining from the sweepings from the factory floor. That plus the fact that the counts were done down to 1/1000th of an ounce that was my first indication that this minting business was just a little bit different.

The existence of by-products introduces our first complication in precious metal control - estimations. To help illustrate the issue, let us first complicate our gold table process. As your local hardware store doesn't sell gold planks, you have to buy standard size gold bars from your local refinery. You therefore have to melt them and pour them into a mould for the legs of your table.

To melt and pour gold, you have to heat it to above its melting point. The reason for this is that gold cools very quickly and if it is just at its melting point it will go solid before you can finishing pouring it. However, this creates a problem because when something is above its melting point (but not yet at its boiling point), some of the liquid is evaporating. Now you might think how much gold would really evaporate and I don't know the technical answer to that. But what I do know is that it must be enough because above any gold furnace I've seen there is a hood that sucks in the fumes, taking it to a "scrubber" that collects the gold particles. However much gold is evaporated, it must be worth enough to go to all that trouble. Consider also that the crucibles in which the gold bars are melted also, over time, absorb amounts of gold.


So how do you do a precious metals stocktake? First step is working out your "theoretical" or book inventory. Say you received 100oz of raw gold and recorded shipments of 90oz of coins. 100 minus 90 equals 10oz. Second step seems simple enough, go around and count all the physical gold and it should add up to 10oz. Easy.

OK, lets say there are 5 x 1oz finished coins on the shelves and 3 ounces of "offcuts". But what about the gold in the sweeps, embedded in the crucibles, in the scrubbers? This is where one has to estimate the gold that is onsite, but not measurable - for example you don't want to crush up and refine your perfectly good crucibles just because it happens to be a stocktake date. Introducing estimations, however, introduces room for human error. This is minimised by keeping historical records of the usual gold recovery from spent crucibles, scrubbers etc, so that there is a reasonable basis or justification for the estimated "onsite but not measurable" gold. Lets say this is worked out to be 1 ounce.

We are still missing 1 ounce. At this point consider that not all "recovery" controls are 100% effective. Scrubbers still let some gold evaporated gold out, for example. I've only described a few of the many recovery type controls in a precious metal factory, there are many more and over high volumes of manufacture bits of gold can be lost. The use of the word "loss" is often interpreted as "theft" but it is more accurately described as a "production" loss. It is a sort of known unknown. But this is not really fair, because production managers, again from historical stocktakes, know that there is a certain ratio of production losses to volume manufactured, which enables them to calculate and expected production loss.

Lets say in our example that the production loss ratio is 1% (our production manager would get fired if that was an actual loss). The estimate loss is therefore 0.950oz (1% of 95 coins made). This leaves us with a stocktake result of 5+3+1+0.95 = 9.950oz against theoretical or book inventory of 10oz. What happened to the 0.050oz? In a precious metals stocktake this is the key question.

First thing that is looked at is the accuracy of the count of measurable/countable physical gold. Second the production manager reviews the by-product estimations. If these two look OK, then third is to consider the effectiveness of the recovery controls. For example, maybe there was a hole in the ducting to the scrubbers and thus more gold was lost to evaporation. If controls are OK then it only leaves two possibilities:

1. Your production loss ratio is not correct. Maybe for every 95 coins made you lose 1oz?
2. Maybe your production loss ratio is correct. Therefore, someone in the factory has managed to secrete 0.001oz out every week over the past year.

The problem is that it is not easy to answer the question, because the stocktake result relies on estimations. This is why mints have one other "recovery" control - physical metal detection of staff as they leave the factory!

The above discussion is simplified, of course. There are many more processes involved in a refinery or mint and many more opportunities for production losses, necessitating many more controls. This is where accurate historical records and experienced staff come in to keep account of precious metal.

There is one thing we have missed. In our example, we only have 9oz in physical metal. Whether the 1oz is 0.95oz of production losses and 0.05oz of theft, or 1oz of production loss, doesn't change the fact that we have lost 1oz. If the 10oz book inventory was funded/acquired from clients holding unallocated with us, then we only have 9oz of physical against 10oz of liabilities.

This is why this article was titled Metal Accounting I. In Metal Accounting II, we will discuss how the 1oz loss is dealt with and introduce yet more opportunities for gold to get lost.


  1. Hi Bron,

    "This is why this article was titled Metal Accounting I. In Metal Accounting II, we will discuss how the 1oz loss is dealt with and introduce yet more opportunities for gold to get lost."
    ...I can not find the n.II. any plans to finish this one?

  2. No plans at this time, as it is a bit technical and fair bit of work involved to write it up.

    Will get to it one day.

  3. Hi Bron
    I discovered your blog recently and it is unique in the perspective it gives into the PM world. I think there is one area where you might be able to shed some light, and that is the price discovery process for gold and silver. Recently there was an article on ZH that showed that Comex is entirely a paper game. 99% of gold contracts on the Comex are settled in GLD paper. The question remains - how does the price discovery for the actual physical metal happen? If an Indian dealer wants to buy gold, what price discovery process do the seller and the buyer go through? If you could write a post on this subject I am sure it would be eagerly read by many.
    Thanks for the great blog.
    Ken Kam

  4. Hi
    Where is the link for Metal Accounting II?

  5. Yes I never got around to part 2 of this, maybe one day when I find the time.