Style Rotation: Value vs Growth

Value and growth are not opposing asset classes; they are relative exposures embedded in the same equity universe. A value-growth rotation overlay changes the portfolio's relative style weight when expected relative returns or risks are unusually asymmetric. That distinction matters: a long-only portfolio can remain fully invested while moving from an expensive duration-like equity profile toward cheaper, more cyclical companies.

The research challenge is severe. Style cycles can run for years, definitions shift with accounting and sector composition, and a signal that “explains” the last cycle is often a retrospective narrative. A useful model earns its place only if it is simple, delayed, diversified across evidence, and governed by strict exposure limits.

Define investable style legs

Start with sector-neutral long-short portfolios, not broad value and growth ETFs. Rank stocks within region and industry using composite descriptors, form value and growth legs, and neutralize market beta, size, and momentum. The resulting spread isolates the relative style return rather than an accidental energy-versus-technology allocation.

Signal group	Value-favorable reading	Why it can fail
Relative valuation	Wide value discount	Cheap firms may have impaired fundamentals
Real yields	Rising or high real yields	Equity-duration relationship is nonlinear
Inflation/growth	Reacceleration with inflation	Can become a sector proxy
Earnings revisions	Improving value revisions	Revisions are short-lived and crowded
Trend	Value spread recovering	Late after sharp reversals
Dispersion	Wide cross-sectional dispersion	May reflect untradeable distress

Relative valuation should be measured against history and within industries. A simple aggregate is the z-score of the valuation spread between the value and growth legs, using multiple metrics such as book-to-price, earnings yield, and free-cash-flow yield. Do not assume a wide spread must close on the next rebalance: valuation is a state variable, not a timing clock.

Combine signals conservatively

Signals naturally arrive at different frequencies. Real yields update daily, earnings revisions may update with estimates, and accounting data update quarterly. Lag each input to when it was observable, robustly normalize it, and cap its influence. Combining independent weak predictors is preferable to optimizing one threshold.

import pandas as pd

def bounded_score(x: pd.Series, window: int = 756) -> pd.Series:
    z = (x - x.rolling(window).mean()) / x.rolling(window).std()
    return z.clip(-2, 2) / 2

def rotation_signal(valuation, real_yield, revisions, trend):
    # Positive values favor value; weights should be chosen before evaluation.
    return (0.35 * bounded_score(valuation)
            + 0.20 * bounded_score(real_yield)
            + 0.25 * bounded_score(revisions)
            + 0.20 * bounded_score(trend))

Map the score to a small active style tilt, not an all-or-nothing switch. For example, allow an overlay between -20% and +20% of active risk, then volatility-scale it. Smooth target weights and include turnover costs. A style rotation strategy that changes its mind weekly is usually trading estimation noise.

Understand the hidden risk exposures

Growth stocks often have long-duration cash flows and are sensitive to discount-rate changes. Value can carry cyclical, financial, commodity, and distress exposure. During a banking shock, “value” may be a concentrated balance-sheet risk rather than a cheap multiple. Use a fundamental factor model to measure the portfolio's intended style tilt alongside unintended country, industry, leverage, and liquidity tilts.

Risk control	Implementation
Sector neutrality	Constrain active sector weights or hedge sector baskets
Beta neutrality	Re-estimate beta with robust, rolling data
Name concentration	Cap issuer and correlated group exposure
Transaction costs	Model spread, impact, taxes, and borrow
Drawdown governor	Reduce overlay risk after model failure, not after normal noise

The factor investing framework is useful here: rotation changes the allocation among factors and should not replace a diversified long-run factor program. Check exposures through performance attribution to distinguish factor selection from timing and implementation return.

Validate across regimes

Evaluate both the standalone spread and the incremental effect on the parent portfolio. Use rolling out-of-sample windows, conservative costs, and samples that include the technology bust, zero-rate period, inflation shock, and fast risk-on rebounds. Report hit rate, drawdown, turnover, correlation to equities, and conditional returns by rate and inflation regime. A positive full-sample t-statistic conceals whether the model is just one historical episode.

Stress scenarios should include an abrupt fall in real yields, a commodity price collapse, and a credit event. These reveal whether the overlay is actually style diversification or a levered macro trade.

Key takeaways

• Rotate a sector-neutral value-growth spread, not sector labels masquerading as style.
• Treat valuation as a slow conditional signal and combine it with revisions, rates, and trend.
• Constrain the overlay tightly; timing evidence is weaker than long-run factor evidence.
• Attribute results by style, sector, macro, and trading cost before changing the model.