How green is public spending? The Lombardy experiment in green budgeting

How much of Lombardy’s public spending contributes to the ecological transition? To answer this question, SDA Bocconi designed and piloted a green budgeting methodology tailored to the regional administrative context. The project, done in collaboration with the Lombardy Region and funded by the European Union through the Technical Support Instrument, trialed an operational tool capable of linking environmental goals, administrative measures, and spending decisions.

More than 80 regional managers and civil servants tested the methodology directly, while an additional 160 participated in internal dissemination activities.

The project also investigated the use of artificial intelligence to assist in classifying public expenditures, attracting the attention of European institutions and other regional administrations. While the experiment highlighted interesting opportunities for preliminary screening and consistency checks, it also revealed a significant limitation: final classification still requires expert judgment, a deep understanding of the administrative context, and adequate documentation of environmental objectives and impacts.

Balancing rigor and usability

Green budgeting aims to assess the extent to which public spending contributes to environmental objectives by classifying investments and programs according to their impact on climate, biodiversity, pollution, or the use of natural resources. From this perspective, the budget also becomes a space where the relationship between public resources, strategic priorities, and expected environmental impacts can be made more explicit. The European Commission, recognizing the international proliferation of this tool, has developed a Green Budgeting Reference Framework that is increasingly inspiring national and regional initiatives.

In some countries, such as France, the process is driven from the top down. In fact, since 2020, the central government has published a national green budget, subsequently extending it to local administrations. In other European contexts, such as the German state of Baden-Württemberg or Spain’s Basque Country, the process has emerged in a more decentralized, experimental way beginning at the local level, where authorities are advancing initiatives and gradually developing shared tools.

The project conducted with the Lombardy Region fits into this broader context, and starts with a specific question:

• Is it possible to develop a technically rigorous green budgeting methodology that public managers can actually use in their routine decision-making processes?

The SDA Bocconi research team developed a methodology that integrates environmental policy and managerial perspectives, focusing attention on the role of managers and civil servants who translate public programs, calls for proposals, and funding schemes into administrative measures and spending decisions. The project also ties into major European initiatives in the field of public sustainability, ranging from the Green Budgeting Reference Framework and OECD guidelines on green tagging to the principles of the EU Taxonomy and the Do No Significant Harm (DNSH) principle already applied in EU-funded programs.

A co-design effort

The project gradually evolved through a process of analysis, co-design, testing, training, and dialogue with other European administrations. The research team—Francesca Casalini, Daniela Preite, and Silvia Rota—first analyzed international best practices and then developed a methodology tailored to Lombardy’s administrative context.

The classification system distinguishes five categories of expenditure: “green,” when spending contributes positively to environmental objectives; “brown,” when it generates harmful effects; “mixed,” when it produces both positive and negative effects on different environmental dimensions; “neutral,” when it has no significant environmental impact; and “unclassifiable,” when not enough information is available. For example, with regard to heating, an incentive to replace polluting boilers with low-emission systems was classified as green; a measure compensating for rising fuel and energy costs in local public transportation was classified as brown, since it encourages the consumption of traditional energy sources without addressing transition; and a project which creates interurban bicycle paths was classified as mixed because, while it promotes sustainable mobility, it was not possible to rule out impacts on land use and biodiversity along the new route.

To guide managers in the assessment process, the researchers developed a decision tree: a structured, step-by-step sequence of questions that helps them classify the measures in question.

The method was initially tested on a sample of 15 legislative decrees establishing spending commitments. It was then applied to an additional 60 decrees, for a total of 75 classified measures across the testing phases. Eighty regional employees from seven different general directorates were asked to assess the robustness of the methodology,

At the same time, relevant administrative documents were analyzed using generative AI tools to explore their potential in facilitating the classification process. The comparison between human and AI classifications produced one of the most thought-provoking findings of the experiment: In 57% of cases, artificial intelligence reached the same classification as regional managers and civil servants, while in 37% of cases it was unable to classify the measures due to insufficient information. Indeed, much of the information needed to assess the environmental impact of the measures was not explicitly stated in the legislative decrees, but it was implicit knowledge for the managers and officials involved. AI does not make mistakes in these cases; it simply cannot access information that is not there. This finding suggests that AI tools can be useful for preliminary screening and consistency checks, while also confirming that final decisions still require expert skills and knowledge of the administrative context. Importantly, this lesson extends beyond green budgeting: The quality of administrative documentation is a fundamental prerequisite for any future use of digital and AI-based tools in public decision-making processes.

The other dimensions of sustainability

The model developed by SDA Bocconi shows that green budgeting can become a public governance tool only if it is treated as an organizational process. In the project described here, researchers were able to develop an operational methodology and test it on 75 spending commitment legislative decrees, onboarding managers and civil servants from different general directorates. Their work highlights a key lesson: To classify public spending robustly, environmental objectives, expected impacts, and safeguards must be made more explicit in administrative acts.

Future prospects are wide-ranging. So far, the methodology has been applied ex post, meaning to expenditures that had already been committed and reported. The next step will be extending it to budget planning, namely the phase in which future allocations of public resources are decided. Looking ahead, the method could also be integrated with environmental KPI systems capable not only of classifying an expenditure but also of measuring its real-world effects, for example in terms of emissions reductions, climate adaptation, or ecosystem quality.

Moreover, the same methodological architecture was designed to be extended to the other dimensions of sustainability (social and economic) paving the way for increasingly integrated public budgeting systems focused on the overall impacts of government policies.