- You know the earnings – they are predictable (hence the need to find companies with predictable earnings)
- The predictability of the earnings allows you to, at a particular time point, know/fix one of the variables (among two: E and P/E).
- You don’t know when the booms and busts will happen
- But you know that there will be booms and busts
- Hence there will be mean reversion in the P/Es
- Hence Buffett’s method can be used to predict selling prices of non-cyclical companies
- Earnings depend on the cycles
- Cycles are not predictable, length of cycles varies each time, you don’t know when the booms and busts will happen
- Hence earnings are not predictable (with reasonable accuracy)
- Timing of P/Es are also not predictable, as they depend on the booms and busts
- Earnings and P/Es have -ve correlation. When markets are booming, earnings are high hence P/Es are low. When markets are bottoming, earnings are low hence P/Es are high.
- You know there will be booms and busts
- You know that the average P/E will materialise for sure, within the next 10 years. But this info is useless without predictable earnings. You need to know/fix one of the variables (among two: E and P/E), at a particular point in time in the future.
- Hence prices of cyclical companies are not predictable. Buffett’s method won’t exactly work here.
- Similarly, Christopher Browne’s method of using metrics (e.g. P/B, P/CF, etc.) of past M&A, LBO transactions won’t exactly work as well, because it will need projections of future B, CF, etc.
- However that doesn’t mean that cyclical companies must be put into the “too hard” bin.
- You will still be able to apply Buffett’s and Browne’s methods if you use very conservative estimates for your projections, e.g.
- Project earnings 10 years later as equal to this current cycle’s lowest earnings, and use the lowest P/E.
- Project earnings 10 years later as equal to this current cycle’s average earnings (no growth), and use the average P/E as usual.
- If you use very conservative estimates, and still able to get a good IRR, then it can be a good investment. In fact, this will then be a much better steal as compared to investing in non-cyclical companies, where the estimates are not as conservative.
- Another potential method is to calculate earnings that are averaged out across one cycle (e.g. 10 years), and calculate historical P/E ratios using the “10-year average earnings”. Buffett’s method may still be applied using the newly calculated average P/E ratio and the projected “10-year average earnings” 10 years later.