Hi all,
What quantitative factors would you use when selecting mining stocks?
When I backtest for non-OTC mining stocks, there are on average 150 any one point in time. I’m hoping find some excess returns with a few simple filters that leave us with enough stocks in the sample.
-I had trouble getting anything earnings related to work (which makes sense for mining)
-Short interest appears to work well
Any other ideas?
Maybe something like operating cashflow, or some EBIT multiple. I don’t have any expertise or experience looking into mining stocks, but I’d think you’d want a way to evaluate their expected reserves, and the going price of whatever they’re selling (for example, price of coal, iron ore). Efficiency ratios would probably matter alot (of the conference calls I read they compare themselves by how efficient their mines are vs. peers), but not sure of how that would be reflected quantitatively because they’re so subject to commodity prices, but maybe a ratio of operating cashflow, or EBIT. I’d also probably look at debt levels. Alot of these miners are up to their neck in debt. Anyhow, just some thoughts. Again, i have no special insight into the sector at all.
The challenge here is that the numbers that drive mining stocks are not the ones picked up by financial databases; reserves, production, production yields, costs per unit, price per unit, etc. But because stocks do what they do without asking what’s convenient for folks like us, we’re not off the hook from having to address them.
Since we can’t do what we must directly, I think we have to do the next best thing; do it indirectly. That means tapping into and riding the coat-tails of those who do it. So I’d focus heavily on analyst-type metrics ad well as momentum and technical analysis, possibly supplemented with broader foundational fundamentals, such as longer term ROE or ROA trends (the theory being that skillful management is skillful management and if you demonstrate its presence, the important numbers wind up falling into line).
You can probably work on biotech and pharma the same way.
So this is where I could come in and say how great statistics is. But rather, I would like to say how important the above is.
When I do backtests I am not really doing statistics or finding a “statistic.” Usually I am finding a “parameter” from that period. In other words, I am testing all of the stocks with that rank and not just taking a sample. Maybe this is good. I do not have to find a standard error to estimate the parameter. I just have found it for that population. 1999-now for P123. Maybe I am doing a little 'descriptive statistics."
What I have to do is what Marc just did. Justify why I think I am measuring things that might give me a similar parameter for the population going forward.
Out-of-sample might be a little different. There might be some real statistics to do here. One could also argue that the parameter I just measured can be used as a statistic for the larger population that includes the future. I am okay with this and I will probably continue to do some of this myself. Just realize this has nothing to do with a “random sample.” Perhaps, this could be taken as a best case scenario. If the backtest does not do exceptionally well with this optimistic assumption (that that parameter serves as a proxy sample of the future) the port should not be funded. More a possible red flag than a green light.
Just my $0.02 and my evolving take on this.
-Jim
Try investing cash flow. Marc’s right that we don’t really have access to that crunchy industry-specific stuff, but I believe that exploration cash expenses are going to be in the investing section regardless of the specific lines. In fact, I think they just generally put it under CapEx, so give that a shot too.
I would also advise looking at EBIT or EBITDA in this case. Given that you’re looking at an industry that lives and dies on its assets, I suspect that depreciation will be distortive in this case.
Oh, and look at the “Other” lines. When the companies themselves report industry specific things, CompuStat needs to put it somewhere in their standardization process, so a bunch of it ends up in Other. (It’s why I advise using those lines as general “quality” metrics.) I know for a fact that if you look at Oil Exploration companies, they all have huge Other lines, which is where CompuStat dumps the oil exploration expenses. I suspect that mining might be the same way, though I don’t know for sure.
Mining and petroleum production use a lot of the same accounting conventions, such as units-of-production to compute DD&A, and deciding between full-cost and/or successful efforts for capitalizing costs to PP&E. You can exploit knowledge of the accounting to infer some interesting things about project economics. Surprisingly, GAAP metrics are much more predictive of future equity returns than industry-specific metrics (see: S&P Capital IQ’s White Paper, Drilling for Alpha: http://marketintelligence.spglobal.com/our-thinking/ideas/drilling-for-alpha-in-the-oil-gas-industry).
But one also should account for commodity price volatility and the company’s revenues and cost sensitivities. Commodity producers are immensely susceptible to commodities prices. They are price-takers. On the other hand, consumer goods, services, and other asset portfolios have the capability to pull price levers through intangible value creation (i.e., branding). Metals and petroleum go through immense cycles, causing immense rifts between accounting metrics and long-term economic prospects. There are simple ways to deal with this uncertainty, but they are beyond elementary modelling.
To be a little more specific, I recommend looking into modeling annuities.
The primary variables being: realized selling price per unit, units of production, decline curve for units of production, cost per unit, fixed overhead, financial costs, price and cost inflation, time value (i.e., discount rate), capital intensity (e.g, replacement cost, per unit maintenance CapEx, etc…)
From these, then infer secondary inputs, vis-a-vis economic limits: project life (in years), economic resource potential, and residual resource potential.
From these key variables, one can infer some key metrics which one should expect to be important in assessing economic resource potential.
For example:
EBITDA ~ [units of production] *([realized selling price per unit]-[cost per unit]) - [fixed overhead]
EBIT ~ EBITDA - ([per unit maintenance CapEx][units of production][%decline rate])
[Project Life] ~ Ln(([EBITDA] + [fixed overhead]) / ([fixed overhead] + [financial costs])) / [%decline rate]
…et cetera.
Thanks for all of the great replies everyone. Looks like I’ve got a bit of work ahead of me! 
pdem: what exactly do you mean by “other” lines?