Attribution Analysis

What Is Attribution Analysis?

Attribution analysis explains why an investment portfolio performed the way it did compared with a benchmark index, such as a broad stock market index. Rather than stopping at one performance number, it separates the overall gap, often called active return or excess return, into pieces that can be traced to particular choices. Fund managers, consultants, and institutional investors use it to see which decisions added value and which reduced results.

How It Works

The analysis compares the portfolio with its benchmark at two levels. The first is the mix of sectors, regions, or asset classes held. The second is how the individual holdings within each group performed. Each difference is assigned to an effect. A common way to measure a single sector uses these formulas:

Allocation effect = (portfolio weight − benchmark weight) × benchmark sector return

Selection effect = benchmark weight × (portfolio sector return − benchmark sector return)

Adding the effects across all sectors accounts for the total gap between the portfolio and the benchmark.

Main Components

• Allocation effect: the gain or loss from holding more or less of a sector, region, or asset class than the benchmark does.
• Security selection effect: the gain or loss from choosing particular securities within a group.
• interaction effect: the combined result of allocation and selection decisions made together.
• currency risk effect: for international portfolios, the impact of exchange-rate movements on returns.

Why It Matters and Its Limits

Attribution analysis helps investors understand the sources of returns, check whether a manager's stated strategy matches what was actually done, and compare results across periods. It does not predict future performance. Results depend on the benchmark chosen, the time period measured, and how holdings are grouped, so two reasonable analyses of the same portfolio can reach different conclusions. It also explains past results only. A strong attribution in one period does not show that the same decisions will produce similar results later.