The World Bank and Climate Projects: A Matter of Definition

What counts as climate finance? Analysis of 2,554 projects financed by the World Bank between 2000 and 2022 showed not just a bias to mitigation but also a very wide range of activities, some of which looked at best tenuously related to climate, including teacher training in Guyana and civil service payment automation in Afghanistan.

In the last couple of years, patterns of behavior don’t appear to have changed that much. We replicated the scraping technique of the earlier paper to look at World Bank climate projects in FY23 and FY24. Of 925 projects that reported thematic areas, 825 were tagged as climate change-related. 751 projects were tagged as related to climate adaptation and 695 to mitigation.

Country-specific climate change spending, calculated by multiplying the share of the project allocated to climate and the value of the loan, totaled $55 billion, out of a project total of $153 billion for FY23 and FY24—about one-third of total spending. Of this, 64 percent was directed towards mitigation and 36 percent towards adaptation. The figure below shows that for every income group, mitigation spending exceeds that on adaptation. It is somewhat of a mystery why the World Bank claims to have spent over $3 billion in two years on mitigation in the poorest countries that account for less than one percent of global emissions.

And for climate, it still appears that almost everything counts in small amounts: the pattern of labeling a little bit of a lot of projects as climate-focused continues. Figure 2 shows that 630 climate projects were tagged with less than 10 percent of loan value being related to climate. The problems with unreliability and inadequate documentation also persist. The World Bank has launched a new version of the protocols for data exchange on its climate finance data (the API) that has dropped multiple data fields supported by the previous version. Ad hoc changes to the labeling of projects, without any warning to the user, have made the database difficult to use (as we observed when we ran the same analysis twice over the course of a month and got different results).

To be fair, and to repeat, it is utterly justifiable to have a broad sectoral definition of adaptation: economic development is a powerful adaptation strategy. But adaptation finance should be focused on the countries who need it (and deserve it) most: the poorest countries who have contributed the least to climate change but are some of the most exposed to its risks. Meanwhile, it is a legitimate question as to how much the Bank should be financing activities that are only justifiable on grounds of their climate mitigation impact: clients have made clear they don’t think the institution should be focused on that. And in all of this, it remains very hard to say how much, if any, of what is being labeled climate finance by the World Bank is “activities that wouldn’t happen absent a climate focus.”

Given all of that, World Bank financing justified on the grounds of climate change should be clearly new and additional (remember that phrase?), as part of a climate dedicated capital increase. That would help transparently meet new climate financing commitments while preserving existing finance for client priorities. Any resulting adaptation finance should be directed toward the poorest countries which deserve it most, and the mitigation finance linked to projects that have low-, no-, or negative-emission approaches as a primary focus.

Of course, two years after the initial study, the World Bank (and the world) is in a different place. Look at the average number of mentions of the word “climate” in World Bank President Ajay Banga’s official speeches: 3.33 per speech in 2024, 0.33 so far in 2025. That rhetorical shift points to the fact that chances for a climate-dedicated capital increase are extremely slim given the current policy stance of the World Bank’s largest shareholder. But a transparent, additional approach for climate finance should still be the long-term end-goal for the World Bank as much as it should be globally.

Addendum

A recent post by Hallegatte et al. on the World Bank’s “Development and a Changing Climate” site raises some concerns with this blog. It is worth starting with where we agree: it is fiscally and economically responsible to build infrastructure that responds to the risks of climate change, and we are very much in favor of the bus rapid transit schemes, higher rice yields, and electricity access that they mention as examples of climate friendly development projects. As we also suggested above, it is to be expected that there is frequently an overlap between adaptation and development —and surely there can be an overlap between projects that deliver mitigation and development benefits.

It is also worth noting that at times we think the concerns in the Hallegatte blog might be based on a misreading of what we said rather than a necessary disagreement. The World Bank blog suggests we ask on adaptation: “why isn’t all our adaptation finance directed to the poorest countries?” As you can see above, we argue instead “adaptation finance should be focused on the countries who need it (and deserve it) most: the poorest countries”—that the poorest countries are most at risk is something that is widely accepted by World Bank senior management, too. Again, suggest Hallegatte et al., “we spend almost nine times more on mitigation in middle- and high-income countries than in low-income countries … it is disingenuous to argue that we have a bias towards mitigation in low-income countries. We simply do not.” But the data supports the actual claim we make above that “for every income group, mitigation spending exceeds that on adaptation.” (Similarly, Hallegatte et al. argue that “Contrary to what CGD claims, only one fifth of our climate finance comes from ‘small interventions,’ defined as less than 10% in climate finance attribution.” We didn’t make any claim about the percentage of financing that comes from small interventions, merely pointed out that 630 out of 925 projects have low share tags.)

Hallegatte and colleagues are additionally concerned about methodology. They argue that “it is unclear how CGD has constructed its sample, which significantly differs from our data.” As mentioned above, the approach follows the one used in this paper, involving scraping data from project documents, and the code and data are available. That’s how the paper can cite data prior to 2011, to answer another concern.

And the paper’s approach was driven by (ongoing) issues around transparency in the Bank’s own climate reporting. According to the joint MDB methodology, the climate tag must be accompanied by a rationale for why it’s there (exactly the kind of rationales Hallegatte et al. provide for the examples in their blog). As the original paper found, all too often such a rationale is not included in the project documentation. The Bank’s climate tagging may be getting better but it is still a work in progress, and so it is impossible to verify the Bank’s Scorecard claims on emissions reductions or why projects increase climate resilience.

But perhaps the key issue is this: Hallegatte et al. suggest “recent criticisms from the CGD blog are misguided... why do we support mitigation in the poorest countries that contribute the least to global emissions? Because we finance impactful development projects that also happen to reduce emissions… These are not climate projects; they are development projects that positively impact jobs and people and are built on technologies that have low emissions.” That is, we’d suggest, the challenge, and that’s why our blog was titled “a matter of definition.” That the Bank can simultaneously claim to have provided $43 billion in climate finance while also claiming it doesn’t do climate projects, that it is impossible to know if or how different the Bank’s portfolio would look if the institution didn't have a climate finance target, is somewhat of a problem. We accept that the current definition and calculation of climate finance used by the World Bank allows for all of that, but it is not the transparent, additional approach that our blog calls for.