Three ratios dominate the search for equity value: pricetoearnings, pricetobook, and dividend yield. These are not the only measures, of course, but P/E, P/B and Yield are nearly universal dimensions in comparing stocks, defining style, and assessing value. But in spite of their wide acceptance, we find they don't all share in actual pricing effect.
An IllDefined Construct
Every market metric has limitations. All are imperfectly measured, and all need a context for interpretation. (Interest rates, for example, are obviously relevant in understanding P/E or yield.) But book value has some deeper problems. Book value is a residual value on the balance sheet; the difference between book assets (which isn't the actual value of assets) and book liabilities (which, in wholly different ways, is not the actual value of liabilities). In both its asset and liability components, book value embodies the cumulative effects of every assumption, estimation, or accounting quirk in the history of a company. Errors or omissions are accumulated virtually forever.
So it's amazing how much attention book value gets. Academic and professional research cites price/book ratio in valuations, style assessment, risk analysis and more. Portfolio managers cite book value routinely and often, as do advisors and the financial press.
What The Market Says
With all the attention, we should expect book value to have substantial bearing on market price. Relevant value metrics should have some reasonably systematic relationship to market prices. The relationship needn't be constant or universal (nothing meets that standard) but should have some generally consistent effect.
But while earnings and dividends are generally and consistently related to prices, book value is not. Consider the two correlation matrices in the frame below. The left matrix shows simple crosssectional correlations across the 30 stocks of the DJIA as of midNovember 2003. Earnings and dividends show the highest correlation, as we would expect. In the "Price" column, we see positive correlations with earnings (+.56) and with dividends (+.36), and a small negative correlation with book (.17). [A correlation of .17 across 30 observations is nonsignificant; essentially zero.]
DJIA Correlations 
DJIA "Partial" Correlations


EPS 
Divs  Book  Price  EPS  Divs  Book  Price  
EPS 
1.00 
+.66  .10  +.56 
EPS 
1.00  +.60  .11  +.45  
Divs 
1.00  +.07  +.36 
Divs 
1.00  +.18  .00  
Book  1.00  .17  Book  1.00  .14  
Price  1.00  Price  1.00 
A Closer Look
The rightside above shows "partial" correlations. These statistics remove shared effects among the variables. That is, each cell shows the correlation uniquely attributable to just that pair. The dividend effect has disappeared, because its raw relation to price (+.36) was actually just a reflection of its relation with EPS and EPS' relation with Price. The EPS correlation is also restated a bit lower for the same reason, but here the independent relation to price is strong enough to maintain a (significant) correlation of +.45. But the Book correlation was nonsignificant both in the "raw" correlation, and in the "partial."
Even when book and market values seem correlated at first, the partials analysis shows the relation as spurious. Just five years before the above analysis (November 1998), the raw DJI correlation of price:book was +.41, but dissolved into .05 after accounting for EPS. At yearend 1995, the net unique correlation across the DJ 30 was +.05 after EPS effect. Using a different stock sample ("Nifty Fifty") in 1998, we had the same result (net unique correlation of +.02), this time with dividends rather than EPS playing the primary mediating role.
Inference
We have just enough of these crosssectional pricing correlations to conclude that book value is generally not a pricing factor. We have found some few occasions when the price:book relationship seems temporarily robust (and in certain sectors like Utilities it is more than transitory). But those occasions have always proven transitory. The evidence suggests clearly (special cases aside), that "book" is an unreliable and overemphasized valuation metric.