Quick Definition
Multi-Year Averaging — Multi Year Averaging is a methodology that uses financial data from multiple years (typically 3 5 years) to calculate profit level indicators for benchmarking purposes.
Our glossary is continuously expanding. For deep-dive guides on specific topics, explore our blog or contact us with suggestions.
Multi-Year Averaging is a methodology that uses financial data from multiple years (typically 3-5 years) to calculate profit level indicators for benchmarking purposes. By averaging results over several years, the analysis smooths out annual fluctuations caused by business cycles, one-time events, or economic volatility, producing more reliable arm's length ranges.
Multi-year analysis addresses the reality that independent companies' profitability varies year to year, and any single year may be unrepresentative of normal operating conditions.
The OECD Transfer Pricing Guidelines (2022) address multi-year data in Chapter III. The Guidelines note that it may be useful to consider data from years prior to the year of the transaction, as multiple year data can provide information about relevant business and product life cycles of comparables. Multi-year data should be used where they add value to the transfer pricing analysis.
US Treasury Regulations §1.482-1(f)(2)(iii)(D) similarly permit use of data from multiple years where such data would enhance the reliability of the analysis.
Common Multi-Year Approaches:
| Method | Description | Use Case |
|---|---|---|
| Simple Average | Equal weight to each year | When all years are equally representative |
| Weighted Average | Weight by revenue or other factor | When years have different significance |
| Rolling Average | Moving window of years | For ongoing compliance monitoring |
| Period Selection | Choose representative period | Exclude anomalous years |
Typical Timeframes:
| Jurisdiction | Common Practice |
|---|---|
| OECD (general) | 3-5 years |
| US | 3 years (multiple year analysis) |
| India | Single year or 3 years |
| Germany | Often 3 years |
Why 3 Years? Three years balances recency (recent data reflects current conditions) with stability (smooths single-year anomalies). Fewer years may be too volatile; more years may include outdated information.
Comparable Company: Industrial distributor with fluctuating results
| Year | Operating Margin |
|---|---|
| 2021 | 2.1% |
| 2022 | 4.8% |
| 2023 | 3.5% |
Single-Year vs. Multi-Year:
| Approach | Result | Commentary |
|---|---|---|
| 2023 Only | 3.5% | May not capture full cycle |
| Simple Average (3 years) | 3.47% | Smooths 2022 peak |
| Weighted Average (by revenue) | Varies | If 2023 revenue was largest, skews toward 3.5% |
For Benchmarking: Using 3-year averages for all comparables produces a more stable arm's length range than single-year data.
| Entity | Multi-Year Approach |
|---|---|
| Tested Party | Current year is primary focus; multi-year shows trends |
| Comparables | Multi-year averaging recommended for reliable benchmarks |
Most benchmarking studies use multi-year comparable data but test against the current year tested party result. Some practitioners also average the tested party's results, particularly for APAs or when demonstrating consistent compliance over time.
Multi-year is generally preferred because it smooths annual fluctuations and produces more reliable benchmarks. Single-year data may be appropriate if: (1) the industry is stable with low volatility, (2) recent structural changes make historical data irrelevant, or (3) local rules require single-year analysis.
Three to five years is standard practice. Three years is most common—balancing stability with recency. Five years may be appropriate for highly cyclical industries. More than five years risks including outdated data that doesn't reflect current market conditions.
Options: (1) exclude the comparable if data gaps are significant, (2) use available years only and document the limitation, (3) interpolate missing values if appropriate (less common). Missing year(s) in the middle of a 5-year series is more problematic than missing the oldest year.
Simple averaging (equal weights) is most common and easier to defend. Weighted averaging (by revenue, assets, or other factor) may be appropriate if recent years are more relevant or if size changes materially. Document your rationale for whichever approach you choose.
If a year includes one-time events (restructuring, pandemic impact, acquisition), you can: (1) exclude that year with documentation, (2) adjust for the one-time item if quantifiable, or (3) include it and let multi-year averaging naturally dilute the impact. Excluding years requires strong justification.
Yes—the IQR is calculated from the averaged PLIs, not from individual years. For each comparable, calculate the multi-year average PLI, then compute Q1, median, and Q3 from those averages. This produces one arm's length range for the analysis period.
Ideally no—use consistent periods for all comparables. However, if a comparable only has 3 years of data while others have 5, you may include it using available years. Document any inconsistencies and assess whether they affect reliability.