Predicting Molybdenum Production
March 5, 2011 2 Comments
Going forward, the all-important price of mol will depend sensitively on the competition between supply and demand. For a price inelastic commodity like molybdenum, a relatively small demand-supply imbalance can cause huge price swings. Molybdenum is only a small component of overall steel costs, so manufacturers will pay what it takes to get their mol fix. Likewise, nobody’s hoping for a molybdenum engagement ring. If there’s more than enough mol to go around (as was the case in 1980, when steel-rolling improvements caused the usage of molybdenum in pipelines to crater) then the price collapses. I’m trying to build a quantitative model of this phenomenon.
There’s a very useful investor’s presentation on the Thompson Creek Metals website that contains the following .ppt slide:
The slide shows how world mol demand has increased over the past fifty years, and projects demand to 5 years and 10 years forward, assuming the average year-to-year production increase of 4%. The resulting expectation is that there’ll be 600 million pounds of production in 2015, and 740 million pounds in 2021. Given the difficulty in bringing new projects on line, those numbers seem like likely to support a robust mol price. But can we get confidence limits on these predictions?
I sampled the fifty one production numbers from the above graph with a one-sigma measurement accuracy of ~ ±5 million pounds per year, the resulting series, is:
(86,80,92,95,111,130,132,141,160,180,
170,173,175,181,160,180,200,220,225,
230,240,200,130,210,220,210,200,215,
250,245,220,215,210,200,245,250,300,
305,300,320,300,305,330,350,360,400,
410,470,470,400,480)
I differenced the above sequence to get set of yearly percentage changes in demand. Assuming that these moves are serially uncorrelated on a year-to-year basis (and remembering, of course, that when you assume, you make an ass out of u and me) one can build foward trajectories for demand over the next ten years. Ten random example trajectories look like this:
After ten thousand ten-year trials, the average 2021 molybdenum production is 683±272 million pounds. The resulting distribution looks quite nicely log-normal. Central limit theorem in action:
So 740 million pounds in 2021 looks a little optimistic, but it’s within 1-sigma of expectations. Seems reasonable that you put your best foot forward if you’re doing an investor presentation…
About Greg Laughlin
Nice. One thing I have been thinking about is how can the Mol relationship be related to oil prices and the overall strength of the economy. It seems like there hasn’t really been a clear directional message in TC as a response to this whole oil spike other than generally drifting around with the market. Also, the site “http://www.assetcorrelation.com/” seems to believe that the correlation between USL and TC has a very similar shape to USL and SPY, which is down. Indicating that as the price of oil rises we are entering a new regime where the correlation between USL and SPY inverts from positive to negative. Even the relationship between XLE and USL is down – which is also sort of interesting because you would think these companies will make more money with higher oil prices. However the relationship between XLE and TC remains strong. Although it certainly seems fine to describe the demand based on empirical behavior as a stochastic process, it would be interesting and useful (for hedging purposes) if this process could be described as a function of other prices. For example Demand_mol(Price_mol, USL, SPY, population, …). Potentially this could be approached by assuming they are linear and finding the correlations, but apparently the behavior is non-linear and some of these relationships can go the other way. Maybe this could be described with a simple analytic model built from your Mol intuition with parameters fit based on previous price behavior.
Thanks for the comment. The working hypothesis is that the response of mol demand (and primary mol producers) to the overall economy is understandable, and that one might be able to obtain real alpha by building a model for mol, and then hedging out the underlying exposure to the economy using, say SP futures.
http://www.assetcorrelation.com/ is very cool, BTW. Here are two sample diagrams ginned up on the site that give some insight:
(1) A correlation matrix with TC and some usual suspects. The time horizon for the correlation coefficients is 5 years, incorporating in a single number the great mol run-up of ’06-’08, the financial crisis, and the recent action. Notice that the all-in molybdenum bugs have done pretty well over the last five years — an observation, *not* a prediction or a recommendation:
(2) Evolution of correlation between TC and GXC (a randomly chosen China ETF). The degree of correlation here has been steadily decreasing over the 2-month lifetime-to-date of molybdos.com: