While making some improvements to the Moneyterms explanation of the PEG ratio, I started thinking about ways of addressing the weaknesses of this ratio. The main problem with it is its short term nature, which makes shares that have mere enjoyed spikes in their growth, look like bargains.

The obvious solution is to take a look at long term earnings. The obvious solution is to use the average of several years earnings. This is a bit trickier to do than for the long term PE, because we need to be a lot more forward looking, because the PEG is used to select growth stocks.

The best, and conceptually simplest, alternative is to replace the recent growth rate with the geometric mean of growth forecasts for several years to come. The formula is:

PE ÷ (EPS_n ÷ EPS₀)^1/n

where PE is the PE ratio
EPS₀ is the latest historical EPS
EPS_n is the EPS n years further out.

Of course we we have enough years of forecasts to make this useful, we might as well as look at a forward PE in the last year for which we have a forecast, or get down to real fundamentals and use a DCF. Finally, we may not have more than one or two years reliable forecasts. So what do we do then?

One solution is to use a weighted average, using the same technique as for exponential smoothing. The advantages of this are:

1. It favours the numbers further out, which are the most important for a growth stock picking.
2. It gives some weight to earlier and historical numbers, so inconsistent performers are penalised.
3. Exponential smoothing has been shown to be effective for predicting trends in other areas of finance such as risk forecasting.

To apply this we use the formula:

Exponentially smoothed PEG = EPS ÷ s
s_t = α(g_t) + (1 – α)S_{t – 1}, and
s₀ = g₀

where s is the averaged growth, and
g is EPS growth
s without a subscript is s in the same year as the EPS number.

There is no obviously correct value for α. Certainly less than 0.5 to avoid being too sensitive the last year, but low enough to give a fairly sharp fall in weighting. Something around 0.25 is probably right.

We now have a modified smoothed PEG ratio which looks at both forecasts and consistency of growth, does not need forecasts for many years out, which is reasonably simple (if a little tedious if you do it by hand) to calculate. It is not perfect, no valuation ratio is, but it is a huge improvement on the the standard PEG ratio.