Cap-weighting has two biases, not one.

Cap-weighted benchmarks guide trillions, yet the weighting rule itself has escaped scientific scrutiny. Read on the TSX 60, the rule embeds two structural biases at once — and neither of them is behavioural.

Reinventing Benchmark Construction: Structural Bias in Market-Capitalization Weighting — Evidence from the TSX 60 · Mukul Pal · SSRN 7021439

Benchmark construction is the least examined trillion-dollar machine in finance. Factor research fills journals; behavioural finance won its Nobel; but the arithmetic that decides what an index is — multiply price by shares, rank, weight — has mostly been treated as settled plumbing. This paper treats it as the hypothesis, and tests it on Canada’s TSX 60.

The finding: market-capitalization weighting does not embed a single structural bias but two, operating simultaneously and in opposite corners of the index. Winner bias recursively reinforces capital toward already-dominant constituents. Laggard suppression systematically starves the bottom of the distribution — exactly where the statistics of reversion say the opportunity is richest.

The mechanics

The winner loop is closed and self-feeding: a rising price raises a constituent’s weight, and every portfolio benchmarked to the index buys more of it in proportion — momentum rewarded as if it were merit. The laggard side is quieter but just as structural: as weight decays toward the bottom, the rule allocates less and less to the names statistically most likely to revert, until index maintenance removes them entirely. Both effects are properties of the construction — distinct from investor psychology, and distinct from the factor tilts of Smart Beta, which re-weight constituents without ever questioning the frame.

To demonstrate that the biases are real and separable, the paper restates the benchmark with the Winner–Middle–Laggard (WML) methodology: constituents ranked relatively, each bin given a defined statistical role. The point of the exercise is not a new factor. It is evidence that the benchmark’s behaviour changes when — and only when — the construction rule changes.

The size of the giveaway

What does winner bias actually cost? Draw the weight curve. A cap-weighted benchmark stacks its weight on the largest names — the curve falls off a cliff, and the index’s fate rides on a handful of stocks. The redesigned curve distributes weight by information instead of size. Flattening that curve is not cosmetic: it changes which risks the portfolio is paid to carry, and it removes the mechanical obligation to buy more of whatever just went up.

The redesign is testable because it is narrow. Same universe, same liquidity screens, same rebalancing calendar — only the weighting rule changes. Whatever difference shows up in the outcome is attributable to design, not selection. That controlled experiment is the paper’s contribution: benchmark construction, treated as a variable rather than an inheritance.

What it means for portfolios

Every mandate measured against a cap-weighted index inherits its structure: the concentration, the bought momentum, the missing reversion. For an allocator, this means the “passive” default is an active structural bet, taken silently. For an active manager, it reframes the job — the benchmark you are asked to beat is also the design flaw you can beat it with.