APT, CAPM, and the Index Model: How the Three Relate

Investors reading about risk-adjusted returns constantly encounter three overlapping frameworks that sound similar but rest on genuinely different logical foundations. Confusing them leads to misreading exactly what a fund's performance report, or a stock's fair-value estimate, is actually claiming.

By Karanvir Singh, founder of Quantix Invest

Three tools, three different foundations

The capital asset pricing model, the single-index model, and arbitrage pricing theory are frequently discussed as if they were interchangeable, since all three produce a formula that looks broadly similar, an expected return built from a risk-free rate plus one or more risk-premium terms. But they rest on entirely different logical foundations, were derived using different arguments, and answer subtly different questions. Understanding what distinguishes them is not academic hairsplitting; it directly affects how much weight you should put on a beta figure, an alpha figure, or a factor-model return decomposition when you see one in practice.

A useful way to keep the three straight is to ask, for each one, what kind of claim it is actually making. CAPM makes an equilibrium claim: this is what returns must look like if every investor behaves a certain way. The index model makes a measurement claim: this is what a regression of past returns against a chosen benchmark actually shows, with no assumption about why. APT makes a no-arbitrage claim: this is what returns must look like if no one can construct a riskless, self-financing profit from the pattern. All three produce a formula of similar shape, but only by keeping straight which kind of claim underlies a given number can you judge how much weight that number deserves.

CAPM: an equilibrium theory

CAPM is an equilibrium theory: it derives its pricing equation from the assumption that every rational, mean-variance-optimizing investor holds some combination of the risk-free asset and a single optimal risky portfolio, and that in equilibrium this optimal portfolio must be the entire market of investable assets, since if any asset were excluded, no one would hold it and its price would fall to zero. From this chain of reasoning, only market risk, measured by beta, should be priced: expected return = risk-free rate + beta × market risk premium. The model's honesty is also its weakness: it requires assumptions, homogeneous expectations, frictionless borrowing, a fully observable market portfolio, that do not hold exactly in any real market, which is why the model is criticized even as it remains widely used.

CAPM's real theoretical contribution is not the formula itself but the underlying separation result it rests on: that every investor's optimal risky holdings should be some scaled combination of the same single portfolio, differing only in how much of it they hold relative to the risk-free asset, not in its internal composition. This is a genuinely strong claim about portfolio structure, and it is what ultimately justifies treating market beta as the single relevant risk measure; if investors genuinely held meaningfully different risky portfolios from each other, based on different circumstances or constraints, the entire logical chain that produces the security market line would break down.

The index model: a practical statistical tool

The single-index model is not, strictly speaking, an economic theory of equilibrium at all; it is a practical statistical simplification, developed by William Sharpe as a computational shortcut for portfolio construction before CAPM was fully developed. It regresses a security's historical excess returns directly against the excess returns of a chosen market index, a broad domestic stock benchmark, most commonly, producing an estimated beta (the regression slope) and an alpha (the regression intercept), plus a residual term capturing everything the index alone does not explain. Crucially, the index model does not claim the chosen index is the true, all-encompassing market portfolio CAPM theorizes about; it simply uses the index as a convenient, observable proxy, sidestepping Roll's critique in practice even though it does not resolve it in theory.

Because the index model is fundamentally a statistical tool rather than an economic theory, its output includes information CAPM's pure theoretical version does not: an explicit intercept term, alpha, representing the average return the security earned beyond what its measured beta and the chosen index would predict, and an R-squared statistic, showing what fraction of the security's total return variance the index actually explains. A low R-squared is itself informative, signaling that whatever single index was chosen captures only a small part of what actually drives that security's returns, a warning sign that a single-factor read of its risk is likely to be unreliable.

APT: a no-arbitrage theory

Arbitrage pricing theory, by contrast, requires no assumption about investor preferences or a single market portfolio at all. It derives its pricing relationship purely from the requirement that no riskless arbitrage opportunity can persist, applied to well-diversified portfolios exposed to one or more common statistical factors. This makes APT considerably more flexible than CAPM, since it easily accommodates multiple factors, but that same flexibility means APT does not, by itself, specify which factors matter or how many there should be; that determination is left to empirical research and economic judgment applied separately from the theory itself.

This trade-off between the three tools is worth stating plainly. CAPM buys theoretical elegance and a clean economic story at the cost of untestable assumptions. The index model buys practical, immediately computable estimates at the cost of no underlying economic theory explaining why the resulting numbers should be trusted. APT buys flexibility and a weaker, more defensible set of assumptions at the cost of leaving the actual factor selection as an open, contestable empirical question. None of the three is simply better than the others in every respect; each is suited to a different kind of question.

The math, worked through twice

Suppose a stock has an index-model beta of 1.2, estimated by regressing its returns against a broad equity benchmark, with an assumed market premium of 5% and a risk-free rate of 4%. The index-model-based estimate of required return, following the same functional form as CAPM, is 4% + 1.2 × 5% = 4% + 6% = 10%. Under a multifactor APT approach applied to the same stock, suppose it also shows meaningful sensitivity to a credit-spread factor, with a factor beta of 0.3 and a factor premium of 1.5%, adding 0.3 × 1.5% = 0.45% to the estimate, for a combined APT-based expected return of 10% + 0.45% = 10.45%. The index model, using one factor, produced 10%; the multifactor APT approach, using two, produced 10.45%, a difference of 0.45 percentage points that traces entirely to the credit-spread exposure the single-index approach could not see.

Now suppose a second stock has an identical index-model beta of 1.2, giving the same 10% single-factor estimate, but its credit-spread factor beta is -0.4 instead of positive, reflecting a defensive balance sheet that tends to benefit when credit spreads widen. Its APT-based estimate becomes 10% + (-0.4 × 1.5%) = 10% - 0.6% = 9.4%. Two stocks with identical single-factor betas, and therefore identical index-model estimates, produce meaningfully different APT-based estimates, 10.45% versus 9.4%, a full percentage point apart, purely because of their opposite credit-spread exposures, a distinction the single-index model has no way to express.

When the three models agree, and when they don't

In the special case where only one systematic factor genuinely drives returns, and where that factor is well proxied by the chosen market index, all three models converge to essentially the same pricing equation: CAPM's theoretical justification, the index model's practical estimate, and APT's single-factor version all point to the same number. The models diverge in exactly the situations that matter most in practice: when a security's return is genuinely driven by more than one systematic force, when the available market index is a poor proxy for the true market portfolio, or when an investor is trying to determine whether an apparent CAPM-alpha reflects real skill or simply an omitted factor exposure the single-index regression could not capture.

Sector concentration is a particularly common trigger for divergence between the models in practice. A fund heavily concentrated in energy stocks, for instance, might show a market beta near 1.0 against a broad domestic equity index, producing a benign-looking single-factor risk estimate, while a multifactor decomposition reveals a large, undiversified exposure to oil price movements that the broad index simply does not track closely enough to reveal. An investor relying on the single-index beta alone would systematically underestimate this fund's true risk concentration, precisely the blind spot a multifactor APT-style analysis is built to expose.

Reading a performance report with all three in mind

When a fund report shows a positive "alpha," the first useful question is which underlying model produced that number. An alpha computed against a single broad index is really an index-model alpha, and it can be entirely explained away, sometimes completely, by a multifactor decomposition showing the fund simply held more small-cap or value exposure than the single index captures. An alpha that survives a full multifactor decomposition is a meaningfully stronger claim of genuine skill than one measured against a single index alone, and the gap between the two figures is itself informative about how much of the fund's reported outperformance is really just an unrewarded, uncompensated factor tilt dressed up as skill.

Actionable breakdown

• Identify which model underlies any risk-adjusted figure you are shown.
  • Single-factor reporting (CAPM- or index-model-style) is the simplest and least reliable.
  • Multifactor reporting (APT-style) is more rigorous but harder to compare across providers.
• Check which index was used as the market proxy.
  • A poorly matched index skews the beta estimate meaningfully.
• Treat a single-factor alpha claim as provisional.
  • Ask whether it survives a size, value, and momentum decomposition.
• Prefer multifactor comparisons when evaluating funds with different style tilts.
  • A CAPM-only comparison can badly mislead across style categories.

Common pitfalls

A frequent error is treating CAPM's theoretical market portfolio as literally identical to whatever stock index a report happens to use, when the index model deliberately sidesteps that equivalence claim and CAPM's theoretical version has never actually been directly tested, per Roll's critique. A second pitfall is assuming a high single-index alpha proves genuine manager skill, when it frequently reflects nothing more than an omitted, easily replicable factor exposure. A third pitfall is overlooking that all three models can disagree meaningfully, sometimes sharply, during sudden factor rotations, periods when small-cap, value, or momentum exposures swing hard in one direction, which is precisely when a single-factor read of a portfolio's risk is least trustworthy.

The bottom line

CAPM, the index model, and APT are progressively more flexible tools for pricing risk, built on progressively different foundations and different kinds of claims, and knowing which one underlies a stated risk-adjusted figure tells you how much genuine confidence that figure actually deserves.

Related reading: arbitrage pricing theory, the core CAPM formula and mechanism, the single-index model, factor investing, building a multifactor APT model.

All articles · The deep guides