My Growth vs. Value Experiment

Barry Gitarts submits : My issue with value investing has always been that it is based on rearview assessment. Most studies favor value over growth mainly because they have been done in hindsight with data flawed by survey bias.

I decided to compile a small list of growth and value companies by sector, and will revisit their performance at a later date. This way we can avoid the flaws that come with testing in hindsight.

TECHNOLOGY
GOOG (P/E 70) vs. MSFT (P/E 18)

BEVERAGES
HANS (P/E 56) vs. KO (P/E 20)

BUILDING PRODUCTS
VMC (P/E 20) vs. LPX (P/E 5)

RETAIL
SHLD (P/E 23.5) vs. WMT (P/E 17.6)

PHARMA
KG (P/E 40) vs. PFE (P/E 14)

INVESTMENT FIRMS
SEIC (P/E 24) vs. GS (P/E 8.6)

COMPUTER HARDWARE
HPQ (P/E 24.5) vs. IBM (P/E 14.6)

TELECOM
Q (P/E -35 (negative EPS) vs. VZ (12.8)

Related: Choosing the Right ETF: Growth Versus Value in Asset Allocation.

Comments

[NO DooDahs]

I think you should probably start with a definition of "growth" and "value" that is more complex than just relative PE ratios.

Then perhaps entertain the notion that sectors might not conveniently have both "growth" and "value" in them. My (limited) experience with value screening suggests that whole sectors fall into disfavor pretty much simultaneously, making all the stocks in a sector "values" - and vice versa, e.g. there are no "value biotechs."

[WC Whiner]

I concur. This is not an "experiment" but an active investing game in which your picks must have high multiples and must be paired with comparable non-picks that have low multiples. It could be fun, but it will tell us approximately nothing, since stockpicking skill over such a tiny list will inevitably be swamped by the effects of luck.

I tend not to attribute the value premium entirely to survivorship bias, since the effect survived the introduction of inclusive databases just fine. Rather, I see two effects in play. One, "value" is a proxy for risk. Higher-risk positions have higher expected returns. Two, in a probabilistic world, any basket of cheaper names will by random chance contain a higher proportion of names below true value than a basket of expensive names. Rinse and repeat the rolls of the dice and the braindead value screener will as a result over time add a tiny bit of value.

It's my prejudice that the risk effect predominates, but I have not tried to measure it. I would, however, be very surprised if both my effects were null and if you rigorously could back up your blanket analysis.