Insurance Stocks: Shooting Down An Overwrought Valuation Metric

Mega-kudos to Merrill Lynch’s Ed Spehar. In a June 28 note, he makes a point about valuing insurers that should be blindingly obvious, but for some reason is not: insurance analysts’ (and, increasingly, bank analysts) favorite stockpicking strategy, recommending companies whose ROEs are high relative to their price-to-book values, is just a “fancy way to say ‘buy low P/E stocks, sell high P/E stocks.’” He’s right.:

The oft-cited price-to-book versus ROE regression equation for life insurers is simply another way of looking at P/E ratios. Users of this equation are regressing “price divided by book value” against “earnings divided by book value (ROE)”, and essentially concluding that low PE stocks (those that trade below the regression line) are more attractive than high PE stocks (those that trade above the regression line). . . .

From 2002 through 2006, life insurance stock portfolios based on a price-to-book versus ROE regression investment strategy (rebalanced annually) would have generated returns very similar to life insurance stock portfolios based on a forward P/E investment strategy. Approximately 80% to 90% of the names in the most and least attractive portfolios are the same under both strategies, which supports the notion that both strategies are very similar. [emphasis added]

The reason that ROE-vs.-price-to-book and low-P/E strategies spit out virtually identical sets of names, by the way, is that ROE-vs-price-to-book is exactly the same expression, mathematically, as P/E is. Additionally, the very high r-squareds that are cited as justification for the strength of the model are also meaningless. They simply show the industry has a very low P/E dispersion.

Let’s compare three hypothetical companies, and you’ll get the idea:

OK? So the group’s ROE-vs.-price-to-book regression would look like this:

. . . while it’s price-to-earnings regression would look like this:

Suspiciously similar! Note, in particular that the r-squareds are exactly the same.

Anyway, those ROE-vs-price-to-book regression charts, which seem to have become standard issue in sell-side reports on insurers, don’t say as much as their fancy-looking numbers seem to imply. After all, no self-respecting analyst would publish a chart that shows the dispersion of P/E ratios within his industry. He'd think it would be overly simplistic--and he'd be right.

Points to Spehar for pointing all this out. His report ought to be distributed to every insurance analyst (and research director) on Wall Street.

Comments

[FIG trader:]

Spehar's argument (in his 6/28 piece) that using a DCF (intrinsic value) model would have generated higher returns (versus p/e or regressed ROE-price/book) is a bit specious. Over longer periods of time (he selected 2002 to 2006), using DCF (i.e or FCF_free cash flow or distributable dividends from insurance subs) and earnings should be the same, the difference is only a time lag between stat and GAAP earnings, for the most part. He is a card carrying member of the sellside set and one in good standing. He well knows that analysts will trot out any nonsense to justify their price targets (aggressive or too conservative), regardless of what conviction they have in the underlying assumptions. Its conviction on the stock (and management) that drives the recommendations and gets analysts to tweak the estimates and projections north or south. Who would naively believe that analysts have the know how (absent management guidance) to make detailed Statutory (i.e cash earnings) assumptions, given limited quarterly availability of data? Why he would attempt to make a point where none needs to be made is beyond me. The reason some people (I for one) like to focus on ROE/price book has more to do with potential risk/reward than trying to figure out which "supposed strategy" works better over time. The variability in book value for insurers from quarter to quarter is minimal, while a big earnings miss distorts the whole p/e measuring mechanism, without affecting the price/book materially (it does affect that quarters ROE which is why smoothed ROEs over numerous quarters are better) Lets take Ed Spehar's only ""sell" rating UNM, a rating I believe he has consistently maintained through thick and thin (from mid teens to 11 and all the way up thropugh to the current 26). It has had a low price book (for years it traded at a discount), with a modestly rising ROE, which he assumed couldn't be increased. The idea for analysts was to identify the transition point for this company, and the beginning of the ascent in the forward ROE or EPS. If you did the price/book would follow higher. Those that did were rewarded. Or lets take another favorite stock with a high price/book,uh......FMD comes to mind first. Its not that the DCF, P/E or Price/book strategies are better. Its that nobody knows where 2008/2009/2010 ROEs will be, but there is a unusually high probability that the ROE would be lower, and that the price/book would follow. Hence my January short call (historic was the way I put it)

2007 Jul 05 05:19 PM