Chained vs Fixed-Weight Price Indexes: Substitution Bias

A chained vs fixed-weight price index differs in how often it reweights the basket of goods and services. Chained indexes (like the U.S. Consumer Price Index, now primarily chained CPI) update the consumption basket annually or quarterly to reflect what households actually buy when prices shift. Fixed-weight indexes lock in a single basket at the index’s base year and never change it. The result: fixed-weight indexes systematically overstate inflation because they ignore the human tendency to buy less of goods that have become relatively expensive—a distortion called substitution bias.

Why Substitution Bias Matters

When the price of beef rises sharply, families typically buy less beef and more chicken instead. A fixed-weight index assumes the household still purchases the same quantity of beef it did in year one, year five, and year ten. That’s unrealistic. Over time, the gap between a fixed-weight index and an index that accounts for actual substitution can be substantial—often 0.2 to 0.5 percentage points per year, depending on the volatility of relative prices.

To see the mechanics: imagine an index tracking only two goods. In the base year (year 1), a household buys 100 eggs at $2 each and 10 loaves of bread at $3 each. The basket costs $200 + $30 = $230. In year 5, eggs cost $3 and bread costs $3.50. A fixed-weight index multiplies: 100 eggs at $3 plus 10 loaves at $3.50 = $300 + $35 = $335. Inflation = ($335 − $230) / $230 ≈ 46%. But the household, now facing dearer eggs, may have switched to buying 70 eggs and 15 loaves. Their actual spending: (70 × $3) + (15 × $3.50) = $210 + $52.50 = $262.50—only about 14% more. The fixed-weight method overstates their cost of living by capturing a phantom rise based on consumption patterns that no longer exist.

How Chained Indexes Work

Chained indexes address this by updating the reference basket regularly. Under a one-year chaining approach, the weights reflect the typical consumption basket in the previous year. When the calendar flips, the new year’s index relinks to the prior year at a ratio of 1.0, then applies weights from that second year. This process repeats annually.

The U.S. Bureau of Labor Statistics moved the official CPI to chained methodology in recent years (the Chained CPI, or C-CPI-U, has been published as an alternative since 2002 and is increasingly favored for policy). The Chained CPI consistently runs below the traditional fixed-weight CPI, typically by 0.2–0.4 percentage points per year. Over a decade, this compounds—a fixed-weight index that reads 30% may correspond to a chained index closer to 25%.

When Fixed-Weight Indexes Make Sense

Fixed-weight indexes are not obsolete. They serve specific roles:

• Historical consistency. A fixed-weight index published at the same frequency for decades maintains a transparent, unchanging methodology. Analysts can compare inflation in 1985 to 2015 without worrying about the weighting scheme shifting.
• Simplicity. Computing a fixed-weight index requires no new survey of consumer habits each year. Once the basket is set, the math is straightforward.
• Policy lag. Governments sometimes want a measure that reflects the “lived experience” at a historical moment, not a constantly moving target. Congressional appropriations, tax brackets, and benefit formulas sometimes reference fixed-weight inflation to avoid giving the index-maker too much discretion.

The trade-off is accuracy. A fixed-weight index will overstate the true increase in the cost of living for a household that adjusts its purchases when relative prices move.

The Size of the Bias

Research by the Federal Reserve and academic economists has quantified substitution bias empirically. In the U.S., the cumulative spread between fixed-weight and chained CPI measures has sometimes exceeded 8–10 percentage points over a 15-year window. For a Social Security beneficiary or a bond investor whose income adjusts with CPI, the choice between fixed and chained can translate to meaningful money.

A small example: if fixed-weight CPI records 2.5% inflation and chained CPI records 2.0%, a pension adjustment based on fixed-weight would deliver an extra $0.50 per $100 of benefit—which, over time, compounds significantly for retirees.

Substitution at Different Levels

Substitution bias operates at multiple tiers. When the price of apples rises relative to oranges, people substitute among fruits (lower-level substitution). When fruit becomes expensive relative to vegetables, the basket may shift away from fruit altogether (higher-level substitution). Older fixed-weight indexes only account for substitution within a category (e.g., which fruit) but miss the broader category shifts. Chained indexes, particularly those with multiple tiers of weighting (as the CPI now uses), capture both.

Practical Use: Which Index to Use

For most forward-looking cost-of-living analysis, chained indexes are preferable. They better reflect what people can actually afford to buy as prices shift. For historical price research—comparing the nominal price of a 1960 car to a 2010 car in real terms—a consistent fixed-weight index may be more useful because it holds the weighting methodology constant across time.

Policy makers increasingly favor chained measures. The U.S. Federal Reserve targets inflation primarily through chained core personal consumption expenditures (PCE). Some governments have switched to geometric-mean chaining, which updates weights even more frequently and can further reduce substitution bias.

See also