Systematic investing
Quantitative equity
Quantitative equity managers choose stocks by rules: rank thousands of companies on measurable characteristics – value, momentum, quality, low risk and their own proprietary signals – and hold the best and avoid or short the worst. It is slower than statistical arbitrage, holds positions for weeks to months, and can run far larger amounts of money.
At a glance
At a glance
- Makes money from
- Returns from factors and signals that predict stock performance over weeks to months
- Holds positions for
- Weeks to months
- Trades
- Shares worldwide, with futures and swaps to hedge or gain exposure
- Biggest risk
- Long factor droughts – value lagged for much of 2018 to 2020 – and crowding
- Who does it
- Systematic managers such as AQR, Two Sigma and Cubist (Point72), alongside large asset managers
- 8.3% – a year, the US momentum premium (1927–2013)
- 4.7% – a year, the US value premium (1927–2013)
- −55% – value’s drawdown (2007 to 2020)
- 163% – rise in the losing stocks (March–May 2009, the momentum crash)
- $1.5tn – in smart-beta funds (Morningstar, end of 2022)
- ~0.05 – a good information coefficient (per stock, per period)
The business
Most quantitative equity is asset management: pension funds, endowments and other investors pay a fee to have money invested systematically. Some products are long-only “smart beta” funds tilted towards factors, with low fees and very large assets; others are market-neutral long–short portfolios sold as a return stream that does not depend on the market, with higher fees; and multi-manager hedge funds run dozens of quant teams side by side.
The factors come from decades of academic research: small companies and cheap ones (by book value or earnings) tended to beat the market in long samples, and so did stocks that had recently done well (momentum), profitable ones (quality) and less volatile ones (low risk). Whether each premium persists, and why – compensation for risk, investors’ behaviour, or limits on arbitrage – is still argued over.
The classic factor premia
From data to trades
The factors every quant knows
| Factor | Buys | Sells | Where it comes from |
|---|---|---|---|
| Value | Cheap stocks: high book, earnings or cash flow for the price | Expensive stocks | Fama and French, 1992 |
| Momentum | Last year’s winners, skipping the latest month | Last year’s losers | Jegadeesh and Titman, 1993 |
| Size | Small companies | Large companies | Banz, 1981 |
| Quality | Profitable, stable, low-debt companies | The opposite | Novy-Marx, 2013; Asness et al. |
| Low risk | Low-volatility or low-beta stocks | High-volatility stocks | Frazzini and Pedersen, 2014 |
A trade, step by step
A quant equity manager does not trade one stock on a hunch; it turns scores for thousands of stocks into forecasts, and forecasts into a portfolio. Grinold’s rule turns a standardised score into an expected return: forecast = volatility × information coefficient × score.
Worked example
From scores to a portfolio
A stock scores +1.2 standard deviations on value, +0.6 on momentum and +0.3 on quality. Its residual volatility is 25% a year, and the combined signal has an information coefficient of 0.05. The manager holds 500 stocks.
- Combined score0.70
- Forecast+0.875% a year
- Ideal information ratio1.12
- With constraints (transfer coefficient 0.5)0.56
- At 4% tracking error≈ 2.2% a year
A forecast under 1% a year for any one stock – tiny next to its 25% volatility – adds up across 500 stocks to a respectable return over the benchmark. Constraints such as no short selling in a long-only fund throw away about half of it, which is why long–short versions exist.
Long–short books hedge the market, but equal money on each side is not the same as no market exposure. If the longs have a beta of 0.8 and the shorts 1.2, a $100 million book on each side is net short $40 million of market. Managers neutralise beta, sectors and factors they do not mean to hold, in a risk model, before the optimiser trades.
A hedged portfolio against the market
The maths that matters
Factor models. A stock’s or a fund’s return over cash is explained by its exposures to factor returns – the market, value, size, momentum and others: . Whatever the factors do not explain is alpha – and many managers’ alpha disappears once the right factors are included.
Worked example
How much of a fund’s return is skill?
A fund returned 12% in a year when cash paid 4%, the market 10% and the value factor 3%. Its market beta is 1.1 and its value loading 0.4.
- Cash4%4.0%
- Market6.6%
- Value1.2%
- Explained11.8%
- Alpha0.2%
The fund beat the market by 2 points, but almost all of it came from more market risk and a tilt to cheap stocks – exposures an investor could buy for a few basis points in an index or smart-beta fund. Only 0.2% was skill, before fees.
The fund’s 12%, taken apart
Portfolio construction. The optimiser maximises forecast return minus a penalty for risk and costs, , subject to limits on positions, sectors and turnover. The covariance matrix of thousands of stocks cannot be estimated directly – there are more pairs than data points – so it comes from a factor risk model.
Testing honestly. Data must be point-in-time (as it was known then, not as later restated), the universe must include companies that later went bust, and costs must be realistic. The bar for a new factor is high: see the overfitting discussion under statistical arbitrage.
Where the money comes from
- Management fees: the bulk of the industry’s revenue, from large long-only and smart-beta mandates – a scale business.
- Performance fees: from market-neutral and multi-strategy hedge funds, where returns uncorrelated with the market are the product.
- Proprietary signals: the edge over plain factors – alternative data, text, faster or cleaner versions of known ideas, and better combinations.
- Efficient implementation: lower costs and better risk control turn the same signals into higher returns after costs.
Factor returns are lumpy. AQR, one of the largest quant managers, saw its assets fall by about half from their 2018 peak of $226 billion through the long underperformance of value, before value’s strong rebound from late 2020. Investors find it hard to stay with a strategy through a multi-year drought – which may itself be part of why the premia exist.
Risk and what goes wrong
The risks are factor droughts that last years, sudden crashes in factors that have worked, and crowding – the same positions in many funds. A risk model measures exposures to known factors; it cannot measure exposure to what other funds hold.
The factors are published · 1992–1993
Fama and French show that size and value explain much of the cross-section of US returns, and Jegadeesh and Titman document momentum – foundations for the industry that followed.
The quant quake · August 2007
Crowded long–short quant portfolios fall together for days with no news, then mostly recover. See statistical arbitrage.
The momentum crash · March–May 2009
As markets rebounded from the financial crisis, the past year’s losers – mostly beaten-down, high-risk stocks – rose 163% in three months, and momentum portfolios short them lost heavily.
The quant winter · 2018–2020
Value fell about 55% from its 2007 peak by mid-2020, the worst drawdown in its recorded history, and many quant funds shrank.
The vaccine rotation · 9 November 2020
News of an effective vaccine sent beaten-down value stocks soaring and pandemic winners falling, giving momentum one of its worst days on record.
Technology and data
- Point-in-time databases of prices, accounts, estimates and corporate actions across tens of thousands of companies and decades.
- Alternative data: card transactions, web traffic, job postings, satellite images, and text from filings, news and earnings calls, increasingly read with language models.
- Factor risk models and optimisers that handle thousands of stocks and many constraints.
- Research platforms for fast, honest backtests, and execution systems that trade large portfolios with little impact.
Who works here
Quant researcher
Finds and tests signals, builds risk and cost models, and improves how signals are combined – mostly in Python on large datasets, with a lot of statistical scepticism.
Portfolio manager
Owns a strategy or a book: decides which signals go in, how much risk to take and how to respond to drawdowns, and explains the results to investors.
Data engineer or data scientist
Finds, buys, cleans and maps new datasets to companies, point in time – a large share of the work at data-driven firms.
Quant developer
Builds the research platform, the production pipeline and the trading systems that run the portfolios every day.
In interviews for this desk
Research interviews lean on statistics and regression, finance theory (CAPM, factor models), Python data work and research judgement – above all, how you would avoid overfitting a backtest.
A stock’s combined score is 0.70, its residual volatility 25% and your signal’s IC 0.05. What return do you forecast?
Grinold’s rule: a year. Forecasts are small relative to volatility because the signal is weak; the portfolio, not the stock, is what is meant to work.
A fund made 12% when cash paid 4%, the market 10% and the value factor 3%. Its beta is 1.1 and its value loading 0.4. What was its alpha?
is explained by cash and factors, so alpha is 0.2%. Beating the market is not the same as skill.
Your long–short book has $100m long at beta 0.8 and $100m short at beta 1.2. What is your market exposure?
: net short $40m of market despite being dollar-neutral. To be beta-neutral, scale the short side to m, or hedge with index futures.
Why does momentum crash, and when?
After a market fall, past losers are mostly high-beta stocks, so the momentum portfolio is short the market. When the market rebounds sharply – as in 2009 – the losers rally hardest. Daniel and Moskowitz show crashes cluster in rebounds after bear markets, and that scaling momentum by its recent volatility helps.
How would you test whether a new value signal adds anything?
Point-in-time data with delisted companies included; a fixed rule decided before looking at results; costs; regress its returns on the known factors to see whether alpha remains; test out of sample and in other countries; and demand a t-statistic of about 3, not 2.
Your signal has IC 0.05 on 500 stocks, but constraints give a transfer coefficient of 0.5. What information ratio do you expect?
; at 4% tracking error, about 2.2% a year over the benchmark. Breadth is overstated if the stocks’ residuals are correlated.
Why might the value premium have weakened?
Candidate reasons: book value misses intangible assets such as software and brands; very low interest rates favoured long-duration growth companies; and the premium was published and widely traded. None is settled – and value’s rebound from late 2020 showed how risky it is to declare a factor dead.
Firm guides: AQR · Two Sigma · Point72 and Cubist
Glossary
- Factor
- A characteristic shared by many stocks that explains their returns.
- Alpha
- Return not explained by exposure to factors.
- Beta
- Sensitivity to the market or to a factor.
- Smart beta
- Long-only funds that tilt systematically towards factors, at low fees.
- Information ratio
- Return over a benchmark divided by its volatility (the tracking error).
- Tracking error
- The volatility of a portfolio’s return relative to its benchmark.
- Transfer coefficient
- How fully constraints let a portfolio follow its forecasts, from 0 to 1.
- Risk model
- A model of stocks’ exposures to factors, used to estimate portfolio risk.
- Point-in-time data
- Data as it was known on each date, without later corrections.
- Survivorship bias
- Testing only on companies that still exist, flattering the results.
- Factor crowding
- Many funds holding the same factor tilts, so they unwind together.
Further reading
- Common Risk Factors in the Returns on Stocks and Bonds – Eugene Fama and Kenneth French, 1993. Journal of Financial Economics
- Returns to Buying Winners and Selling Losers – Narasimhan Jegadeesh and Sheridan Titman, 1993. Journal of Finance
- Value and Momentum Everywhere – Clifford Asness, Tobias Moskowitz and Lasse Heje Pedersen, 2013. Journal of Finance
- Momentum Crashes – Kent Daniel and Tobias Moskowitz, 2016. Journal of Financial Economics
- Active Portfolio Management – Richard Grinold and Ronald Kahn, 1999
- Kenneth R. French Data Library – Kenneth French, 2026. Free factor returns going back to 1926