What are index re-basing and definition changes, and why do they matter for CME?

Question

AcadiFi · Accepted Answer

Re-basing and definition changes are two distinct but related data pitfalls that can silently corrupt your analysis. Here's the distinction:

## Index Re-Basing

**What it is:** Changing the base period of an index — the point at which the index equals 100 (or some reference value). The underlying data doesn't change, but the scale does.

**Example:** Statistics Denmark announced that beginning January 2016, the Danish CPI would use 2015 as the new base year (August 2015 = 100). The old series was discontinued, and a new series was computed retrospectively back to 1980.

**The Danger:** If you inadvertently splice data from two different base periods — say, old-base CPI values before 2016 and new-base values after — you create an artificial jump or gap in the series. Inflation calculations across the splice point will be meaningless.

```mermaid
flowchart LR
    A["Old CPI Base
(e.g., 1990=100)
Value: 142.5"] -->|"Jan 2016
Re-base"| B["New CPI Base
(2015=100)
Value: 100.3"]
    A -->|"Wrong: splice
raw values"| C["Artificial drop
142.5 → 100.3
Fake deflation!"]
    A -->|"Correct: convert
to same base"| D["Consistent series
Smooth transition"]
    style C fill:#8b0000,stroke:#c9a84c,color:#e5e7eb
    style D fill:#006400,stroke:#c9a84c,color:#e5e7eb
```

**How to handle:** Always verify that all data points in a time series use the same base period. When a re-base occurs, either use the provider's retrospectively computed series (as Statistics Denmark did back to 1980) or convert old values to the new base yourself using the linking coefficient.

## Definition and Methodology Changes

**What it is:** When the statistical agency changes *how* an index is calculated — not just its scale, but its actual content or methodology.

**CPI Examples:**
- **1983 (US CPI-U):** The Bureau of Labor Statistics changed how owner-occupied housing was treated. Before 1983, the CPI included actual house prices; after 1983, it used 'owners' equivalent rent' — an imputed rental value. This change significantly affected the measured inflation rate.
- **1991 (US CPI-U):** BLS introduced regression-based product quality adjustments (hedonic pricing). When a computer doubles in speed for the same price, the CPI now records this as a *price decline* because the quality improved. This reduced measured inflation relative to the old methodology.

**Why This Matters for CME:**

1. **Historical comparisons break:** If you compute 'average CPI inflation from 1970 to 2020,' the pre-1983 and post-1983 portions are measuring fundamentally different things. The average conflates two incompatible measurement regimes.

2. **Model calibration errors:** A model calibrated to pre-methodology-change data may not work on post-change data (and vice versa). The relationship between CPI inflation and bond yields, for example, may appear to shift when the real change is in the CPI definition.

3. **Cross-country comparisons:** Different countries use different CPI methodologies. Comparing US CPI (with hedonic adjustments) to a country that doesn't use hedonic pricing is an apples-to-oranges comparison.

**Practical Rule:** When using any long time series, always check for methodology changes within the sample period. If a change occurred, decide whether to use only post-change data (shorter but consistent) or apply an adjustment factor to make pre-change data comparable.

**Exam Tip:** Vignettes may describe an analyst using a decades-long data series without acknowledging a methodology change. You should identify this as a source of error and recommend either shortening the sample or adjusting for the break.

For more on data quality challenges in CME, explore our article on forecasting pitfalls.

What are index re-basing and definition changes, and why do they matter for CME?

Master Level III with our CFA Course

Related Questions

Related Articles

Practice Questions