Why Relative Climate Scores Break Net-Zero Tracking

When the score drops but the portfolio improves

A recurring pattern surfaces once institutional investors begin attaching net-zero targets to a sovereign climate score: the portfolio appears to deteriorate even as the underlying countries improve. A Eurozone government bond book can post a lower aggregate climate score from one year to the next while every issuer it holds has, in absolute terms, advanced on its climate indicators. Nothing in the portfolio has gone wrong. The problem sits in the metric.

This is not a data-quality issue, and it is rarely visible at the point of data selection. It becomes visible only in implementation, specifically, the moment a relative score is asked to do a job it was never constructed to do: track progress consistently over time.

The core insight

Most widely used sovereign climate indices are relative by construction. Scores are typically built through min-max normalization of individual indicators within each annual vintage, followed by weighting and aggregation. Within a single year, this produces clean internal comparability, every country sits on a common 0-to-100 scale. Across years, it does not.

The reason is structural. Under min-max normalization, the best- and worst-performing countries define the boundaries of the scale, and every other country is positioned between those two extremes. The scale itself therefore moves each year with the global distribution. Three consequences follow directly, and each one undermines time-series interpretation:

• A portfolio's aggregate score can decline even if every country it holds improves, provided that countries outside the portfolio improve more.
• The score is sensitive to the extremes. A single outlier entering or leaving the rated universe shifts the min or max, and with it every score in between.
• Observed year-on-year movement reflects relative repositioning within the global set, not absolute change in a country's climate indicators.

The same logic propagates into rating categories. When alignment is assessed using qualitative bands, for example, treating the top rating tiers as "aligned" or "aligning" with net zero, those bands are usually derived from the quantiles of the underlying relative score. A quantile-based rating is, by definition, a fixed-proportion classification: roughly the same share of countries falls into the top tiers every year, more or less regardless of how much real-world climate ambition has moved. An investor tracking the share of the portfolio rated "aligned" is therefore tracking a near-constant, nudged only by relative reshuffling among countries.

Why this matters in practice

Two implementation contexts make the problem visible, and both appear repeatedly across sovereign fixed-income mandates.

The first is score-based progress tracking. An investor establishes a baseline portfolio score in a given year and commits to year-on-year improvement. The intent is sound: a clear, quantified climate objective tied to the existing benchmark. But because the score is recalibrated annually against a shifting global set, the year-on-year delta is partly noise. A "decline" may simply mean that non-held countries improved faster; an "improvement" may reflect outlier movement at the boundaries rather than anything happening inside the portfolio. The target is real; the measurement against it is unstable.

The second is alignment-share targeting. Here the investor sets interim and long-term targets for the proportion of the portfolio rated as aligned or aligning with net zero, for example, a defined alignment percentage by 2030 and full alignment by 2050. Because the rating bands are quantile-derived, the aligned share is structurally resistant to upward movement: the number of countries in the top tiers stays broadly fixed, so the portfolio's alignment percentage drifts with relative repositioning rather than tracking genuine progress. An unintended year-on-year decline can appear that has nothing to do with any deterioration in real climate ambition.

In both cases the investor has done nothing wrong. The methodology simply is not designed to validate this type of long-term claim.

Four ways to restore time-consistency

No single adjustment works for every scenario, as the correct approach depends on whether the goal is to ensure measurement integrity, build benchmarks, or allocate capital. Four approaches are worth exploring:

1. Normalized ranks instead of relative scores. Converting each annual score into a normalized rank (i.e. scaling a country's position within the rated universe to a 0-to-1 value) removes much of the distortion. It is insensitive to outliers at the boundaries and robust to countries entering or leaving coverage. Ranks answer a cleaner question: is the portfolio shifting toward better-positioned countries, independent of how the annual scale happens to be scaled?
2. Portfolio score relative to the global mean. Expressing the portfolio-weighted score as a ratio to the global average score each year produces a benchmark that is largely independent of weighting scheme, eligibility constraints, and shifting score boundaries. A ratio above one indicates the portfolio sits above the global average; a rising ratio indicates improving relative standing. Usefully, this measure can register improvement through engagement with issuers already held - without reallocating capital - because it reflects the portfolio's position against a moving global field rather than its absolute score. For greater market realism, the global mean can be replaced by a global sovereign bond index, which restricts the comparison to investable, eligibility-screened issuers.
3. Active reallocation through climate tilting. Measurement integrity does not, on its own, move capital. Where the objective is genuine alignment in the spirit of Article 2.1(c) of the Paris Agreement, the weighting logic of the benchmark itself has to change, systematically increasing exposure to better-performing issuers and reducing it for weaker ones, within the relevant eligibility set. The year-on-year portfolio score will still fluctuate because of the metric's relative nature, but the allocation is doing the alignment work regardless of what the headline score reports in any given year.
4. An absolute score. The structural fix is to anchor scores to a fixed reference scale rather than re-scaling annually. An absolute metric would enable genuinely time-consistent country-level assessment and, if applied retrospectively, consistent backtesting. This is the cleanest answer to the tracking problem, though it is a methodological undertaking rather than a configuration change, and one the broader index-development community is only beginning to approach.

Key implications

For sovereign fixed-income and reporting teams, three points carry over directly:

First, the choice between a relative and an absolute metric is not a technical footnote, it determines whether a net-zero target is measurable at all. A target expressed against a relative score is a target measured with a moving ruler.

Second, the rating-versus-score distinction does not resolve the problem. Quantile-derived ratings inherit the relativity of the scores beneath them, and in some respects make it less visible.

Third, measurement and allocation are separable decisions. Normalized ranks and global-mean ratios fix the interpretation of progress; tilting and absolute scores change what the portfolio actually does or what the scale actually measures. A coherent net-zero implementation usually needs both halves addressed deliberately, not one standing in for the other.

Closing perspective

The deeper point is that sovereign climate alignment is a construction and tracking challenge, not a data-acquisition one. The data may be perfectly sound; the failure mode lives in how a relative metric interacts with a longitudinal target inside a benchmark-constrained portfolio. The two halves - fixing how progress is measured, and changing what the portfolio actually does - have to be addressed deliberately.

Integrating climate scores directly into index weighting logic, and pairing relative country scores with an absolute, framework-aligned tracking measure, is how that separation gets resolved in practice. The objective in both cases is the same: to let a net-zero target be tracked against something that actually holds still.