Thursday, February 08, 2007

Scenario Analysis or Sensitivity Analysis

This is another no-brainer concept that is given a cool and sophisticated name so that financial advisers and analysts can brag about their knowledge in investment.

For those first-timer to this blog, pls read the relevant posts that are linked here in order to understand what I am trying to say. Most likely this post will overwhelm you if you read it without the background knowledge.

Essentially what you do in scenario or sensitivity analysis is that you try to stress-test your assumptions and see if they work when the state of the world has change.

I guess the easy example here is SIA. In a previous post, I mentioned that SIA earned $1 EPS in 2005 and given that it is trading at $17, this means that its PER is 17x, or an earnings yield of 6% (i.e. you expect SIA to earn you 6% for every dollar you invest.)

Now we need to test how robust is the EPS of $1 (or the earnings yield of 6%), i.e. we want to know if SIA can actually deliver an EPS of $1 when the world has changed. Usually we will set up 3 scenarios.

1) The base case is where the state of the world is as today, goldilocks, not too hot not too cold. In this case, we assume that SIA will continue to deliver the EPS of $1. Hence its PER is 17x and its earnings yield is 6%.

2) Then we have the nightmare scenario where the world goes into recession, i.e. Sept 11 again or some war breaks out. Obviously we have to assume EPS drops to some very low level, maybe $0.10.

3) And finally we have the blue sky scenario where the world prosper and SIA lives happily ever after and EPS becomes an astronomical no, like $10.

After that, if you like doing math, you can acsribe probabilities to each of the state and calculated the expected EPS of SIA. For me, I am just interested in the nightmare scenario. I want to know if it happens, is SIA still cheap. Obviously, if the nightmare scenario comes true, SIA can only earn me 10c for every $17 of the stock, I might as well buy Singapore T-bills. Hence I will not buy SIA.

The fun part here is how to ascribe an EPS to the nightmare and blue sky scenario. What is the appropriate no.? Is 10c low enough for the nightmare scenario? In which case PER goes to 170x and earnings yield become 0.6%? Or is it 1c, then PER goes to 1700x. Woah, that’s better than Google at its peak (PER 400x or so I think).

I would say the success rate of this kind of analysis only gets better with experience. It also depends on your view of the world and your emotional state and personality. A conservative investor will think that EPS goes to zero and hence will not buy SIA, bcos he thinks SIA will simply drop like a rock if another catastrophe strikes. A greedy and bullish investor who have only seen bright and sunny days will think that SIA will fly no matter what happens. And he will say SIA is still very cheap at PER of 17x.

See also SWOT analysis
and Investment philosophy and process


  1. Hi 8percentpa,

    Thanks for your analogy for the blue sky and nightmare scenario. I understand what you are trying to say by using sensitivity analysis to forecast, but often times the future is murky and many companies are in danger of alling victim to "acts of god" which may wipe out a significant amount of their assets.

    Thus, for P/E projections, I would say industry and risk analysis is necessary in order to achieve a fair rate of success. A business operating in an earthquake belt is obviously more vulnerable to disasters than one operating in a safer environment; thus we might attribute a lower P/E to the former a higher one to the latter even if they are in the same industry.

    As for the SIA case, I would recommend taking an average P/E for the entire industry (airlines, in this case) and project it for SIA; but add a premium to the P/E as SIA is currently a market leader. This also goes for market laggards where the P/E should be discounted against the industry average.

  2. If you look at the inverse of PE which is earnings yield, taking a higher PE means that the market is willing to pay for lower yield (20x = 5% yield only) of course, we are usually using just a 1yr or 2yr forward PE, which excludes LT growth.

    But personally, I would not invest in stocks with PE more than 20x bcos it means that the stock has to grow tremendously in order to just validate its high PE.

    It is much safer to buy a cheap PE stock that has got growth. That way when the growth does come true, it's an added bonus.