Sector-Neutral Value

Overview

Value investing as a systematic strategy is simple to state: buy the stocks that are cheap relative to what the business actually earns or owns, sell the ones that are expensive. The oldest and most studied factor in equities.

The naive version has a fatal flaw. If you rank every stock in the market on price to book, you do not end up with a portfolio of cheap companies. You end up with a portfolio of banks, utilities and carmakers, because those industries are structurally low price-to-book, and a portfolio of software companies on the short side, because software is structurally high price-to-book. You have not built a value strategy. You have built a bet that old-economy sectors will beat new-economy sectors, which is a completely different and much riskier proposition.

Sector-neutral value fixes this. You compare each bank to other banks, each software company to other software companies, and you buy the cheap ones inside each neighbourhood. The result is a cleaner, lower-volatility expression of the same idea.

Thesis (why the edge exists)

Two competing explanations, and the truth is probably a mix.

The risk explanation: cheap companies are cheap because they are genuinely riskier. They carry more debt, they have less flexible cost structures, and they get hurt worse in a recession. The extra return is payment for holding that risk. If this is right, value's premium is real and permanent but you will earn it by suffering at exactly the wrong moments.

The behavioural explanation: investors extrapolate. A company has had a few bad years, so people assume the bad years continue forever and mark it down too far. A company has had a few great years, so people assume the great years continue and pay too much. Reality mean-reverts, and the gap closes. If this is right, the premium persists as long as humans keep being human.

Either way, the sector-neutral version isolates the effect you actually want and strips out the industry composition bet you did not ask for.

Strategy logic

• Choose metrics. No single value ratio works everywhere. Book to price is the classic but has aged badly for asset-light companies. Earnings yield captures profitability-adjusted cheapness. Free cash flow yield is harder to manipulate with accounting choices. Use all three.
• Clean them. Real fundamental data has garbage in it. Winsorise the extremes and handle negatives sensibly, since a negative earnings yield is not "infinitely expensive", it is a different situation entirely.
• Standardise within sector. For each metric, compute the z-score using only the stocks in the same sector. A bank with a book to price of 1.2 might be the cheapest bank on the board, while a software company at 1.2 would be a distressed disaster.
• Blend. Average the three within-sector z-scores into one composite score.
• Rank and trade. Long the cheapest quintile, short the most expensive, rebalance on a schedule.

Parameters (knobs)

• Metric set: book to price alone (classic, most decayed), or a composite of book, earnings, cash flow and sales. Composites are more robust and less likely to be an artifact of one accounting quirk.
• Sector definition: GICS sector (11 buckets, coarse) or GICS industry group (24 buckets, finer). Finer is generally better but leaves you with thin buckets in a small universe.
• Reporting lag: 3 months is standard and safe, 6 months is ultra-conservative. Anything shorter risks using numbers before they were public.
• Rebalance frequency: quarterly is the natural pace since fundamentals only update quarterly anyway. Monthly rebalancing mostly just adds turnover.
• Distress screen: on or off. Turning it on kills a chunk of the raw value premium but removes most of the catastrophic single-name blowups.

Portfolio construction

Equal weight inside quintiles is standard for research. For live money, weight by liquidity so the book is actually tradeable.

Hold the sector weights of your benchmark. That is the entire point of the exercise. If the benchmark is 12 percent energy, your long book should be roughly 12 percent energy, filled with the cheapest energy names rather than energy in aggregate.

Cap single-name weight. Value books have a nasty habit of concentrating in whatever is currently being left for dead, and one accounting fraud can eat a year of returns.

Consider a light quality overlay even in a pure value strategy. Requiring positive gross profit or positive operating cash flow removes a lot of the names that are cheap for the excellent reason that they are dying.

Costs, capacity and turnover

Value is one of the friendlier factors on cost. Fundamentals move slowly, so the portfolio does too. Quarterly rebalancing on quintiles typically produces turnover in the 40 to 80 percent per year range, which is an order of magnitude gentler than short-horizon momentum.

Capacity is high. Cheap large caps are liquid, widely held and easy to trade. This is one of the few equity factors you can genuinely run billions in.

The costs that bite are not trading costs, they are the costs of being wrong for a long time. A value strategy that underperforms for five straight years will lose its investors long before it earns its premium, and that is a real cost even if it never shows up in a slippage report.

Backtest design checklist

• Point-in-time fundamentals. This is the number one source of fake value backtests. Restated financials leak the future. If your data vendor gives you "as-restated" numbers, your backtest is fiction.
• Reporting lag. A company's fiscal year ends in December but the numbers are published in March. Using December's book value on 1 January is time travel.
• Survivorship and delistings. Value portfolios hold companies that go bankrupt. If your database quietly drops them, your returns are wildly overstated.
• Negative earnings handling. Decide explicitly what to do with loss-making companies. Ranking them as "most expensive" is one defensible choice; excluding them is another. Silently generating a divide-by-zero is not.
• Financials and REITs. Book value means something completely different for a bank. Either treat them as their own sector with their own metrics, or exclude them and be honest that you did.
• The lost decade. Test 2010 to 2020 specifically. If your value strategy looks great in that window, you have almost certainly introduced a bug or a look-ahead.

Common failure modes

• Value traps. Cheap because it is dying. Retailers being eaten by e-commerce, print media, coal. The metric says buy, the business says run.
• The intangibles problem. Modern companies invest in R&D and brand, which get expensed rather than capitalised, so book value understates their real assets. Book to price systematically mislabels them as expensive. This is a large part of why classic value struggled through the 2010s.
• Long droughts. Value can underperform for a decade. Not a quarter, not a year, a decade. You must decide in advance whether you can actually sit through that.
• Crowding on the short side. Everyone shorts the same expensive names, and short squeezes in those names are violent.
• Accounting manipulation. Earnings can be managed. Cash flow is harder to fake, which is why the cash flow metric earns its place in the composite.

Variants

• Value plus quality. Screen the cheap names for profitability and low accruals. This is the single most useful improvement and is roughly what "quality value" funds do.
• Value plus momentum. The two factors are negatively correlated, so blending them produces a much smoother combined return than either alone. Buy cheap stocks that have stopped falling.
• Intangibles-adjusted book value. Capitalise R&D and advertising spend to build an adjusted book value. Fixes the biggest structural flaw in classic value.
• Enterprise-value based metrics. EV/EBITDA and EV/sales handle leverage differences better than equity-only ratios.
• Deep value. Only the cheapest 5 percent, higher expected return, far higher volatility and far more traps.

Our notes and suggestions

The lesson to internalise here is that sector neutrality is not a refinement, it is the difference between a strategy and an accident. Run the same backtest with and without sector neutralisation and look at the sector weights over time in the naive version. You will find entire years where the "value strategy" was simply short technology, and you will understand instantly why the naive version's drawdowns are so brutal.

Then go and check what fraction of your value premium survives once you also require the company to be profitable. That number, honestly computed, is the real strategy.