Where NZIF 2.0 Implementations Diverge

The framework is not the decision

A common assumption is that adopting the Net Zero Investment Framework 2.0 (NZIF 2.0) settles the hard questions. In practice it opens them. NZIF 2.0 defines the categories - "committed", "aligning", "aligned", "achieving" - and the criteria that sit behind them. It does not specify which climate dataset feeds those criteria, which temperature pathway the assessment runs against, how fair-share considerations enter, or how a country's rating should behave from one year to the next. Each of those is a separate decision, and each one shapes the output more than the framework itself does.

That is why two teams can both report that they implement NZIF 2.0 and produce ratings that disagree on the same country. The divergence does not come from the framework. It comes from the implementation choices layered underneath it - choices that are easy to make quickly and hard to unwind later.

Why these problems surface late

NZIF 2.0 is often read as a conceptual scoring framework. Teams underestimate that most of the effort lands in technical execution (assembling, cleaning, and reconciling data) rather than in the framework logic. The first version of a rating usually looks reasonable. The problems appear later: when a country jumps a category for no real reason, when emerging markets cluster at the bottom, when last year's ratings can no longer be compared to this year's. By then the architecture is in place and the cost of changing it is high.

The challenges below are the ones we see surface most often. They are not edge cases. They are the predictable consequences of decisions that felt minor when they were made.

Data access is not the free, open resource teams expect

An NZIF 2.0 rating draws on many different sources - emissions, targets, policy, transition pathways - and they do not sit in one place. Teams assume climate data is freely available and ready to use. Three things undercut that assumption: the relevant detail often lives in individual documents, data providers cover only parts of the picture, and commercial use is frequently restricted by licence.

Long-term strategies, which feed the Ambition criterion, are openly available through theUNFCCC, but as one document per country, each of which has to be read and assessed by hand. Aggregators such as ASCOR pre-process some of this, but reduce it to a net-zero target year and a pathway-compatibility flag; anything more (whether all greenhouse gases are covered, whether the target is legally binding) sends a team back to the primary documents. Other core sources carry hard licensing terms. CCPI requires a licence agreement, and only its Premium tier carries the absolute ratings needed. CAT data is public but requires a request for commercial use. Parts of ASCOR rest on PRIMAP emissions data, which is non-commercial only, a constraint a user drawing PRIMAP through ASCOR often is not aware of.

In practice this shows up as manual effort, hidden legal exposure, and delay. Licence negotiations can run to nine months and a poorly chosen aggregated source can corrupt a rating outright. The Net Zero Tracker database, for a long period, recorded only an announced Danish target that counted as not yet legally binding; the binding target disappeared from the database entirely, which is exactly the kind of gap that produces a wrong country assessment and, downstream, a wrong investment decision.

Climate sources are treated as interchangeable scores - they are not

The IIGCC points teams toward several sources for NZIF 2.0: ASCOR, CAT, CCPI. They rest on different methodological foundations and are not directly comparable. The differences concentrate in three places, and each one breaks comparability in a different way:

1. Different pathways

CAT and ASCOR both run against 1.5°C least-cost and fair-share pathways, but their fair-share methods differ sharply and produce different target paths. CAT blends several IPCC-relevant fair-share studies into a range; ASCOR relies on a single approach that does not yet sit within the IPCC fair-share literature. CCPI uses a less strict but Paris-compatible well-below-2°C pathway built on the common-but-differentiated-convergence approach. Mix these inside one rating and the pathway choice, not the country's climate performance, drives the result.

2. Different policy logic

CCPI assesses policy through local experts, weighing the whole policy mix in national context: differentiated, but carrying subjectivity. CAT models the expected emissions effect of existing policy, assuming adopted measures are fully implemented. ASCOR checks, through standardized indicators, whether specific instruments exist. Three sources, three answers to the question “is this country's policy any good?".

3. Different structure

CCPI is an aggregated, relative score, though its underlying indicator ratings are absolute and usable for NZIF. CAT gives an absolute country rating but assesses against both least-cost and fair-share paths, which can diverge enough that one country reads very differently depending on the path chosen. ASCOR isa database with no aggregate verdict, leaving selection and weighting to the user - and that selection is not always done consistently. Emerging markets and developing economies tend to score worse where indicators are simply unavailable, so the assessment rests on a thinner base than it appears to.

The failure mode that follows is data blending. Teams reach for multiple sources to raise country coverage and create methodological breaks in the process: mixed temperature scenarios inside one rating, combined fair-share approaches built on different foundations, one country standing as “aligned” and “not aligned” at once depending on the source. The underlying error is the belief that all climate data says the same thing. It does not.

Fair share does not travel with a single indicator

Countries sit at very different development levels and carry different responsibilities, so an NZIF-rating has to treat them fairly to compare them fairly. The trap is assuming fair share is automatically present in the data. It often is not, and where it is, it usually lives at the aggregate level, not in any one indicator.

The CCPI, for example, builds fair share into several components: its well-below-2°C pathways, its policy assessment, its target evaluation. Pull a single raw indicator - current emissions, say - out of that structure and the fair-share logic is gone, which produces systematically skewed results. Emerging markets score worse for reasons that have nothing to do with their climate effort. Countries on a legitimate development path read as failing because their emissions are rising. The ratings come out unbalanced across country groups, and allocation decisions quietly under-weight whole regions.

Ratings move when nothing real has changed

An NZIF-rating is meant to be absolute: to show whether a country is improving or slipping over time. That only works if the assessments stay comparable year to year, and several things conspire against that.

Some sources shift their own foundations: ASCOR's fair-share pathways changed in each of its last two versions, so a country's rating can move without its real position moving at all. Relative sources create the opposite problem: a country can improve in absolute terms and stay flat in a CCPI overall ranking because its peers improved too. And the NZIF implementation itself can drift, if the bar for each category is set low early and tightened later; once that happens, it is no longer clear whether a country improved or the ruler changed.

A related effect is volatility at the thresholds. Countries near a category boundary flip back and forth on small input changes. Denmark, for example, moved between "aligning" and "aligned" across 2022 to 2024 because a single sensitive indicator, or an annually recomputed pathway, tips them over a line. The CCPI's policy assessment is deliberately sensitive, which is right for a monitoring index and wrong for an investor rating that needs to be stable enough to act on. Reclassification adds a sharper version of the same problem: when Algeria moved up a World Bank income group from 2024 to 2025, stricter NZIF criteria suddenly applied, and a rating can drop from aligning to no commitment while the country's actual policy is unchanged. Investors have told us directly that they cannot anticipate these moves.

Too many countries land in too few categories

In practice, most countries cluster in "no commitment", "committed", and occasionally "aligning". The higher categories stay nearly empty. Several teams reached the same finding independently during development. Countries with genuinely different trajectories receive the same rating, which leaves investors unable to tell the steady improvers from those that just clear the floor.

A frequent cause is the scenario choice. 1.5°C remains the scientific and policy anchor, but on current trajectories it is barely reachable without overshoot, and once fair-share considerations are layered on, almost no country qualifies as "aligning" or better. A team that holds strictly to 1.5°C can end up with the bulk of its universe in committed and essentially nothing above it - a distribution that is scientifically defensible and close to useless for portfolio construction. A well-below-2°C pathway produces more realistic assessments and stays Paris-compatible, since the agreement itself targets well below 2°C while pursuing 1.5°C.

The thinness compounds at the engagement stage. A final NZIF-rating is heavily aggregated; for a dialogue with a sovereign debt issuer it rarely shows which specific measures would move the country up. Partial progress that does not cross a category boundary stays invisible, so engagement is hard to target and hard to measure. The aggregate rating is easy to communicate and, for this purpose, too coarse to act on.

Implementation is a chain of decisions, not a single choice

None of these challenges is a flaw in NZIF 2.0. They are the consequences of the decisions the framework leaves open: which data, which pathway, how fair share enters, how ratings behave over time, how finely countries are distinguished. Each decision interacts with the others, and a defensible choice in one place can undermine a choice made elsewhere - a stable pathway paired with a volatile indicator, a high-coverage data mix that breaks comparability.

The teams whose implementations hold up are not the ones that found a shortcut. They are the ones that treated these as connected decisions and made them deliberately, in sequence, with the trade-offs visible. What those decisions look like in practice - the choices the robust implementations share - is the subject of the companion note to this one.