Wednesday, February 6, 2013
Further comment on the CS writeup
Someone's literally screaming at me to get off my ass and do my errands today, but I need to say one more thing about my comments on the CS writeup.
While CS does also believe in mean reversion, they essentially make a supply-demand argument for expecting gold price to decline in future, and they do assert any decline should be slow.
Therein lies the problem.
If you were to assert there is a major supply-demand change coming that should immediately alter the price in a strong fashion (e.g., Branson starts mining asteroids for gold in 2014), then it's valid to leave your argument at the level of first approximation. We know something catastrophic will happen, so whether gold goes down 30% or 50% matters little.
It's the same as with a gold project valuation: if your theory is that Eike Batista is about to buy Guyana Goldfields, you don't need to do a detailed calc to determine a buyout price; the first approximation says if you buy GUY at $3.38 it'll make you good money and you don't have to quibble over pennies. Who cares: your target price be wrong anyway.
But if you instead say, for example, we should revalue GUY based on diesel prices escalating 10% annually, gold being in a slow downward trend of 3% per year, and taxes being increased by 2% at some point in the next ten years? OK, now a first approximation of the target price is swamped out by these individual factors. You need to do a second approximation, which involves looking under the hood at all empirical data.
Credit Suisse makes an argument that gold should slowly dwindle in price, based on fear trade receding, fed funds rate, expected inflation, mean reversion, an increase in junior miner hedging, decreased grade and increased capex/opex being "unimportant", global IP, American housing prices for some reason, competition for Chinese investment dollars, competitive currency devaluation... and nothing to do with India.
The problem is that if you add all these factors together and only expect the sum to result in a slow decrease in gold price, and you don't have data and a price model to determine by how much, you're really truly just making shit up. This is because if that sum results in a slow decrease, you really need to quantify each item to determine whether that increase can swamp out other important things like net Asian per-capita wealth generation, or whether the failure of one item to come to pass can alone change your predicted trend in the gold price.
Credit Suisse's writeup is really only a first-approximation argument for a slow decline in the price of gold. They don't provide any data, they don't provide a theoretical model or formula anywhere, and they are utterly ignoring some major inputs.
It's like, say, Northland doing up a mine plan that forgets to include labour, trucking equipment, and diesel costs, uses only the average ore grade, and assumes 100% recovery.
In this case?
Well, for example, if India buys 25% of world gold, and if India's buying is a function of per-capita wealth generation, then ceteris paribus an 8% annual growth in per capita Indian wealth means in five years they make up ~30% of annual world gold demand. Yet CS ignores India entirely.
If Chinese gold demand is a function of per-capita wealth generation, then 8% annual growth in per capita Chinese wealth needs to be sopped up entirely by competing asset classes to keep Chinese gold demand constant. If the other asset classes can sop up 120% of annual Chinese wealth growth, that reduces Chinese gold demand; if they can only sop up 80% of wealth, Chinese gold demand still increases.
Also: "fear trade receding", "expected inflation", "future increase in junior miner hedging" and the like are all things that a professional analyst at a world-class house like Credit Suisse should be able to model and quantify. Just throwing these factors in, without either any proof that they'll happen or any quantification of the size of their effect, is like answering an engineering question with "because Jesus loves us." It's not even an argument, it's just a bald unsupported assertion.
If this was a proper scientific document? Y'know, something you could rely on at all, something that you'd be justified in using to inform your own investment thesis?
They would investigate, model and quantify all inputs, including Chinese and Indian retail demand and their growth rates, detailed future gold production models, some sort of explanation for how Central Bank buying is affected by "fear", a quantification of how much "fear" will decrease, some detailed model of how much the gold price will suffer from Chinese asset class competition, a model to determine future increases in miner hedging and how it affects gold futures contract demand and how that can affect the physical price, and a model for future US real rates and how they correlate to gold.
And an explanation, any explanation, for that stupid bullshit about US housing prices.
They don't really need to show their work, but if they want someone with more than a highschool diploma, an MBA, or a post-doctoral fellowship at LSE to take them seriously then they probably should. Certainly not showing your work impacts reputation very negatively.
Each model should ideally come with error bars. I'm sorry, but that lesson on cumulative error in first-year physics had an effect on me.
A sensitivity analysis of the results would also be nice. At the very least it would demonstrate an acceptance that maybe some of your asserted future negative price pressures don't come to pass.
I know all this is a lot of work. And it means less time for coke and whores and reading the Daily Mail.
But fuck, guys! While your analysis was much more well-written than the childish hack job at Business Insider, it still does reek of amateurishness.