Benchmark

A benchmark is a reference point chosen in advance against which the performance of a portfolio or of an individual strategy is measured. The return on an investment, taken on its own, says nothing: the profit may have come from picking the right securities, or simply from the fact that the whole market was moving up. A benchmark separates these two sources: it shows what the investor would have earned without making any decisions other than one — to stay in the chosen asset class.

How a benchmark is chosen

A benchmark should replicate the composition and risk of whatever it is being compared with. A portfolio holding a broad range of Russian equities is compared with an index of the equity market, a portfolio of government bonds with an index of comparable duration, and a mixed portfolio with a composite benchmark in which the weights of the asset classes mirror the portfolio's own asset allocation. The benchmark is fixed before the period begins and is not changed along the way; otherwise the comparison loses its meaning. The second requirement is replicability: the benchmark must be something the investor could actually buy and hold without making active decisions, that is, a pure buy-and-hold strategy. The difference between the portfolio's return and the benchmark's over the same period is the contribution of the investor's own actions — from security selection to rebalancing and attempts to time the market.

Example: index-based benchmarks

Exchange indices are the most common basis for a benchmark, and their daily moves show the base from which a portfolio's performance is measured:

If a portfolio consists of equities and a broad-market index has been chosen as the benchmark, then that is the index it must be compared with — not whichever index happened to rise the least over the period.

Where a benchmark misleads

The main distortion is a benchmark chosen after the result is already known. Among the many indices and periods available, there is almost always one against which the portfolio looks strong; when testing a strategy on historical data this is the same error as look-ahead bias in a backtest. The second trap is technical: a price index does not include dividends, whereas the portfolio receives them, so comparing a portfolio with a price index mechanically overstates its performance. For an accurate measurement, the total return version of the index is used, in which payouts are reinvested. Finally, a benchmark measures only underperformance or outperformance and says nothing about risk: beating an index by concentrating in a single stock and beating it with comparable diversification are results of a very different nature, even though the difference from the benchmark is the same in both cases.