Value pays in some sectors and not in others

Value is one of the best-known factors. Buy cheap stocks, sell expensive ones, and collect the premium Fama and French (1992) documented. But cheap stocks are not spread evenly across the market. Banks, energy and utilities trade at low prices relative to book value almost all the time. Tech and healthcare almost never do. A value portfolio sorted on the whole market is partly a bet on a few sectors.

So I asked a simpler question. Inside each sector, do cheap stocks beat expensive ones? I used 60 years of US data, 1966 to 2025, and the 12 Fama-French sectors. Inside each one I went long the cheapest third by book-to-market and short the most expensive third, with large and mid caps only and value weights.

Click to enlarge

The answer depends a lot on the sector. Inside financials the value spread earned 7.1% a year (t-stat 4.2). Inside energy 6.0%, inside utilities 3.8%. Inside business equipment and tech it lost 1.4% a year, and retail and consumer durables were also slightly negative. Only four of the 12 sectors have a t-stat above 2.

The pattern is fairly stable over time. Splitting the sample in two halves, the ranking of sectors holds up, with a rank correlation of 0.57 between 1966-1995 and 1996-2025. Energy earned the same premium in both halves, close to 6% a year. Tech was weak in the first half and clearly negative in the second. The level moves, financials went from 10.2% to 4.0%, but the sectors where value works tend to be the same.

A possible reading is that book value means different things in different businesses. In a bank or a utility, book equity is a decent measure of what the firm owns. In a software or pharma company, most of the assets are intangible and never reach the balance sheet, so a low price-to-book there is often a signal of a weak business, not of a cheap one. Recent work points the same way. Eisfeldt, Kim and Papanikolaou (2022) add intangible capital to book equity and adjust for accounting differences across industries, and their value factor earns more than the standard one, also in the recent years when value struggled. Golubov and Konstantinidi (2019) split market-to-book in two parts and find that the whole value premium comes from the part that compares price with fundamental value, not from book value itself. This test does not separate these explanations, but it suggests where to look.

Click to enlarge

There is a practical side to this. If value works so differently across sectors, part of the risk of a plain value portfolio comes from the sector bets it takes by accident. One way to remove them is to sort inside each sector and then add the pieces up. I did this for a standard value factor, built like Fama and French's HML with small and large stocks.

The premium is almost the same, 2.9% a year sector-neutral against 2.7% plain. The risk is not. Volatility falls from 12.2% to 8.7%, and the Sharpe ratio goes from 0.22 to 0.33. The two versions are highly correlated, 0.91, so this is the same factor with less noise. It does not fix the bad decade of value. From 2007 to 2025 the plain version lost 1.7% a year and the sector-neutral one lost 0.5%. Less damage, but no rescue.

For anyone building a systematic value strategy, the takeaway is to look inside the factor before using it. A single value premium hides sectors that pay and sectors that do not, and some of the volatility is a sector bet that the data suggest was never paid for. Sector-neutral is a cheap fix for the second problem. The first one is harder. Book-to-market is only one way to measure cheapness, and the best measure may not be the same in every sector. Earnings or cash flow yields could say more in a software company, where book value says little, while book value may still be the right anchor in a bank. Choosing the value measure sector by sector is a natural next step, and one that needs care, because it is also an easy way to overfit.

Method. JKP global factor data for the US (CRSP and Compustat), common stocks, 1966-2025, monthly rebalancing. Value is book equity over current market value. Sectors are the 12 Fama-French industries from SIC codes. Figure 1 sorts inside each sector into thirds, value-weighted, stocks above the NYSE 20th size percentile. Figure 2 follows HML, with the NYSE median splitting small and large stocks, thirds by book-to-market inside each size group, value-weighted, and the average of the small and large spreads. Price above $1.

What it does not say. These are gross returns before trading costs. The sector estimates are noisy, with 50 to 300 stocks per sector at a time, so the exact ranking will not repeat. Book-to-market is only one measure of value, and the results may differ with earnings or cash flow multiples, or outside the US.

References

• Eisfeldt, A. L., E. Kim, and D. Papanikolaou, 2022, Intangible value, Critical Finance Review 11, 299-332.
• Fama, E. F., and K. R. French, 1992, The cross-section of expected stock returns, Journal of Finance 47, 427-465.
• Golubov, A., and T. Konstantinidi, 2019, Where is the risk in value? Evidence from a market-to-book decomposition, Journal of Finance 74, 3135-3186.
• Jensen, T. I., B. Kelly, and L. H. Pedersen, 2023, Is there a replication crisis in finance?, Journal of Finance 78, 2465-2518.

Data. Jensen, Kelly and Pedersen global factor data, USA, 1966-2025, built from CRSP and Compustat (WRDS). Noax Capital calculations.

Research illustration, not investment advice. Simulated and past results do not indicate future results.