What's Different About High-Growth, Momentum Stocks?

Tonight on Mad Money (video above), Jim Cramer argued that growth or momentum stocks are in a unique class and have to be judged differently from other stocks. For example, Apple (AAPL), Chipotle (CMG) and Netflix (NFLX) are typically downgraded by analysts because their P/E ratios are to high, that is, they are too expensive.

Cramer's argument is that for these momentum stocks, what really matters is that the smart money (hedge funds and mutual funds) want to hold growth stocks. Regardless of price, you can ride the momentum of these stocks as long as you are willing to take profits after a good run.

What interested me about momentum stocks is their properties when estimated with the Random Stock Walker models. What I have found is that these stocks all have significant unit roots (growth dynamics) and their attractors are also significantly related to (being driven by) growth in the U.S. economy.
The bootstrap forecast using the USL20 model shows strong growth potential but also shows a strong downside (the lower 98% prediction interval).
Chipotle also has strong growth potential but does not have the same downside risk. The same is true for Apple.
The Random Stock Walker models suggest that Netflix, at least, should be viewed cautiously. Too bad! Netflix is a product I like although many analysts think it has a difficult future ahead given potential competitors.

ARM Holdings (ARMH) is also mentioned in the Mad Money video (above). It is different from the other momentum stocks Cramer mentions (AAPL, CMG and NFLX). I'll talk about ARM Holdings in a future post.

THEORY and METHOD There are some interesting issues here. It is very difficult, because of the unit roots, to clearly determine whether these stocks are random walks (P[t] = 1 P[t-1] + V), business-as-usual (pure momentum, P[t] = a P[t-1] + V, a gt 1) or stocks that are linked to growth in the U.S. economy (P[t] = a P[t-1] + b S[t-1] + V, a lt 1). The models look very similar using the Akaike Information Criteria (AIC). It is necessary to look at the bootstrap confidence intervals for coefficients in the US index models. All the stocks are significantly related to growth in the U.S. economy and the unit roots are removed from the model when the state variables for the U.S. economy are included.

Notice also that Cramer talks about normal stocks being "driven by news." In terms of the Random Stock Walker models, news is the variability term, V, in P[t] = P[t-1] + V.