What Does SDG Monitoring Practice Tell Us? An Analysis of 120 Voluntary Local Reviews

Abstract

This paper examines how local governments monitor the Sustainable Development Goals (SDGs) through voluntary local reviews (VLRs), focusing on localised practices from 2016 to 2024. Given the crucial role of local authorities in SDG implementation, the research aims to analyse differences in monitoring approaches and indicator usage across various types of local governments worldwide. The analysis involved 120 VLRs, using both quantitative methods to assess the number of indicators per goal and semantic clustering techniques to explore thematic patterns. Results show that cities are the most active in SDG localisation, especially for goals aligned with urban governance, such as Goal 11 (Sustainable Cities) and Goal 4 (Quality Education). Some goals, such as Goal 15 (Life on Land) and Goal 17 (Partnerships), are less frequently monitored, reflecting challenges in translating national objectives to local contexts. Semantic clustering highlighted both strong alignments and gaps in SDG indicator usage, particularly for overlapping goals. The paper underscores the need for more tailored, context-specific indicators for local authorities to effectively monitor SDGs, while highlighting the leadership role that cities play in this process.

1. Introduction

The Sustainable Development Goals (SDGs) are a set of 17 interlinked global objectives established by the United Nations in 2015, designed to serve as a “blueprint to achieve a better and more sustainable future for all” by 2030 [1]. These goals encompass a broad spectrum of development issues, ranging from poverty eradication and gender equality to climate action and the fostering of peaceful, inclusive societies. Each goal is accompanied by specific targets and indicators to monitor progress. Their significance lies in their comprehensive and inclusive framework, which acknowledges the interconnectedness of social, economic, and environmental dimensions of development [2]. This holistic perspective ensures that progress in one area supports and reinforces progress in others, thereby fostering a more resilient and sustainable global community. The SDGs are also governed by the principle of ‘leaving no one behind’, a commitment to prioritise the needs of the most vulnerable and marginalised populations. Failure to achieve the SDGs could mean that existing inequalities among the world’s population persist, the natural environment further degrades, and phenomena of poverty and conflicts continue to exist [3,4,5,6,7]. On the other hand, the literature suggests that successful implementation of the SDGs could result in a more equitable, prosperous, and sustainable world [8]. The SDGs thus represent not only an imperative but also a strategy for ensuring long-term global stability and prosperity. They do so by touching on root causes of systemic issues and to an extent offer a pathway to a future where all people have the opportunity to thrive [9].

1.1. Reporting on the SDGs

Reporting on the SDGs is done via voluntary reviews at various government levels, including local, regional, and national. These voluntary national reviews (VNRs) and voluntary local reviews (VLRs) serve as tools for assessing progress, identifying challenges, and sharing best practices among peers on SDG monitoring and implementation. When conducting such reviews, governments demonstrate transparency and accountability, as they expose the current status of SDG performance (often exposing to the public and the world cases of which they are not necessarily proud) [10]. This willingness to report on the SDGs despite potential deficiencies points to a commitment to sustainable development [11]. For local, regional, and national governments, the act of reporting is not just a formal procedure, but a tool for self-critique and eventually improvement. These voluntary reviews are also central in driving better policy initiatives, as continuous monitoring and reporting on the SDGs enable the issuing authorities to adapt and refine their plans, initiatives, and strategies in order to be more impactful and achieve the desired outcomes [12,13]. Participation in such reviews also means that governments conducting them contribute to the global repository of knowledge and experience on monitoring and achieving the SDGs [14]. This collective experience is central for advancing the SDGs, as it can allow the exchange of ideas, solutions, and impactful practices. Above all, though, it stresses the fact that sustainable development is not an individual but a shared responsibility, requiring efforts at all levels of governance.

1.2. Localised Monitoring of the SDGs

Monitoring the SDGs at the local level is crucial, because it is at this level that meaningful progress toward achieving the SDGs can be observed. Local governments are on the front line of implementation, transposing international and national policies into actions that directly impact communities [15]. By focusing on localised monitoring (herein understood as ‘the process of monitoring strategies at the local level for achieving global, national, and subnational sustainable development goals and targets’ [16]), they ensure that the unique challenges and opportunities of each area are addressed, adopting solutions that are tailored to their local needs and specificities [17,18,19]. One of the key reasons that localised monitoring is essential is that it prevents the masking of inequalities that can occur when data are aggregated at higher levels. National averages can often hide significant disparities within territories [20]. For example, while a country might show overall progress in reducing poverty, specific regions or cities may still face deprivation and inequalities [21]. Furthermore, with SDG localisation, where people can feel more engaged with the goal of transforming society toward sustainability. Citizens are more likely to participate in and support initiatives that directly affect their lives [22,23]. Engaged communities are vital for the success of the SDGs, as they can drive bottom-up change and hold local governments accountable. According to the OECD, over 65% of the SDG targets require local government action and community engagement, highlighting the critical role of local entities in achieving these global objectives [24].

Localised monitoring also allows for more effective and responsive governance. Local governments can quickly identify and address emerging issues, adapting policies and programs to meet changing needs [25]. For instance, cities and regions that monitor air quality can implement immediate measures to reduce pollution and protect public health, contributing to SDG 3 (Good Health and Well-Being) and SDG 13 (Climate Action) [26]. Numerically, localised monitoring can reveal specific insights that drive targeted interventions. For example, data from local education systems can show dropout rates and literacy levels within different neighbourhoods, guiding resources to areas most in need and helping achieve SDG 4 (Quality Education) [27,28]. Additionally, urban areas that track waste management practices can identify the effectiveness of recycling programs, informing efforts to reach SDG 12 (Responsible Consumption and Production) [29]. These targeted actions, grounded in detailed local data, significantly advance overall progress toward the SDGs.

1.3. Objectives

Building on the context and significance of the SDGs and the necessity of localised monitoring, this research aims to provide a comprehensive analysis of how the SDGs are monitored at the local level through VLRs. The objectives of this research are multifaceted: to examine SDG reporting in local monitoring, as well as the indicators employed by local governments worldwide to monitor SDG progress; to investigate differences in monitoring practices across different types of local governments; and to conduct a semantic analysis of indicators to determine if they align with the 17 SDG goals or if new areas of focus emerge. Understanding these factors helps reveal how local contexts can shape the interpretation and measurement of the SDGs, highlighting regional priorities and challenges. To guide this research, the following key questions are formulated.

• Are all SDGs equally monitored, or are some prioritised over others? If so, which ones, and why?
• How do local governments utilise indicators to monitor SDGs? Do local indicator frameworks mirror institutional (UN, EU) SDG indicators in terms of size?
• What variations exist across different world regions and types of local governments?
• Does the semantic analysis of VLRs indicators align with the 17 SDGs, or do new clusters of interest emerge?

To answer our research questions, we analysed VLRs published by local and regional governments from around the world. These publicly accessible documents provide a wealth of data on how places monitor and report on SDG progress, offering a unique perspective into localised practices and priorities. The methodology involved compiling VLRs from various local and regional governments, analysing the number of indicators used per SDG, investigating variations in monitoring practices among local authorities and differences across geographical areas, and conducting a semantic analysis of the indicators.

The remainder of the paper is structured as follows. Section 2 provides information on the data sources, the selection criteria, the data collection process, and the analysis methods used; Section 3 presents the findings obtained through the different types of analysis performed and reflects on the results. Finally, Section 4 highlights the main elements for reflection, potential for further studies, and further work.

2. Materials and Methods

This section outlines the data sources, selection criteria, data collection process, and analysis methods used in this research. By detailing these steps, we aim to provide a transparent and replicable approach for understanding how local governments monitor the SDGs through VLRs.

2.1. Data Sources

The primary data source for this research is online VLRs published between 2016 and 2024. VLRs are reports prepared by subnational governments to monitor and showcase their progress toward achieving the SDGs. As of the time of writing of this paper, a total of 325 VLRs have been published worldwide (see Appendix A for a list of VLRs from the authors’ own collection and Figure 1 for a visual representation).

Figure 1. VLRs published worldwide per type of issuing authority, as reported therein (authors’ own elaboration).

The data on the number of VLRs published per year reveal a clear and significant upward trend (Figure 2—left). In 2016, only two VLRs were published, marking the beginning of what would become a rapidly expanding practice. The following year, 2017, saw a fourfold increase to eight VLRs. This upward momentum continued steadily, with 15 VLRs published in 2018 and 30 in 2019. The substantial rise in 2020, with 48 VLRs, indicates a growing recognition of the importance of local-level SDG monitoring. The most dramatic increases occurred in 2021 and 2022, with the number of VLRs reaching 80 and 81, respectively. This surge reflects a significant mobilisation of local governments, likely influenced by global advocacy and the realisation of the vital role that local actions play in achieving global sustainability targets. The slight decline to 51 VLRs in 2023 might indicate a stabilisation phase, where the initial wave of reporting had covered many early adopters. As of the data available for 2024, 10 VLRs have been published. Given that this number only accounts for part of the year (June 2024), it suggests that the practice of issuing VLRs remains robust and ongoing.

Figure 2. VLRs published worldwide by year of publication (left) and language (right) (authors’ own elaboration).

The data on VLRs’ publishing languages highlight the diversity and global reach of these reports (Figure 2—right). The 325 VLRs have been published in various languages, reflecting the international commitment to the SDGs and the importance of making these documents accessible to different linguistic audiences. English is the predominant language for VLR publications, with 163 reports solely in English. Spanish follows as the second-most common language, with 75 VLRs published. French and Portuguese also feature prominently, with 23 and 16 VLRs published, respectively, highlighting the active participation of francophone and lusophone regions in the SDG process. Interestingly, several VLRs are published in multiple languages, demonstrating efforts to reach diverse audiences, for example, combinations like Arabic and English (3 reports) or English and German (6 reports), and English, Spanish, and Portuguese (1 report). Other languages represented include Catalan (3 reports), Italian (3 reports), German (5 reports), Norwegian (5 reports), Finnish (1 report), and Turkish (1 report).

The data on the types of authorities issuing VLRs provide insights into the diverse range of local and regional governments engaged in SDG monitoring and reporting (Figure 3—left). Cities are the most prolific issuers of VLRs, with 213 reports. This number stresses the crucial role of urban centres in driving sustainable development. Cities, often hubs of economic activity and innovation, face unique challenges and opportunities that make their contributions to the SDGs particularly impactful. Regions follow as the second-most common type of issuing authority, with 30 VLRs. Regional governments often oversee larger geographical areas than cities and can coordinate SDG efforts across multiple municipalities, enhancing coherence and effectiveness in sustainable development initiatives. Provinces and states also play important roles, with 22 and 14 VLRs, respectively. Counties, municipalities, and metropolitan cities further contribute to the diversity of reporting authorities, with 13, 11, and 7 VLRs, respectively. It is important to note that this classification of issuing authorities does not originate from the authors, but directly from the VLRs themselves. The way each local or regional government reports its type does not necessarily adhere to standardised classifications.

Figure 3. VLRs published worldwide by issuing authority (left) and continent (right) (authors’ own elaboration).

The distribution of VLRs across different continents highlights the global commitment to the SDGs (Figure 3—right). Europe leads with the highest number of VLRs, totalling 135. Europe’s active participation can be attributed to the region’s well-established governance structures and a strong tradition of sustainability initiatives. South America follows with 68 VLRs. Asia has published 60 VLRs, demonstrating significant involvement from several countries across the continent. North America has 39 VLRs, showcasing the continent’s active role in SDG monitoring and reporting. Africa has issued 20 VLRs. While this number is smaller compared to other regions, it represents a growing awareness and effort to address sustainable development challenges across the continent. Oceania has published three VLRs, reflecting the region’s emerging participation in the SDG process.

Analysis of the type of information included in VLRs reveals the varying approaches that local and regional governments take to report on their progress toward the SDGs (Figure 4). Out of the total VLRs published, 231 (71%) contain both quantitative and qualitative information. This category includes a mixture of reported data on SDG indicators, complemented by qualitative descriptions of the local monitoring process. There are also 24 (7%) VLRs that focus solely on quantitative data. These reports provide specific measurements and figures related to SDG indicators, offering a clear and concise view of the progress made, but lack the qualitative context that explains the underlying factors and interpretations of these numbers. On the other hand, 70 (22%) VLRs are solely qualitative: these reports do not include specific data regarding SDG indicators, but offer qualitative information and interpretations on the monitoring efforts by the local and regional governments. These narratives can provide valuable insights into the strategies, challenges, and successes from a descriptive perspective, but they do not offer the hard data necessary for rigorous statistical analysis.

Figure 4. VLRs published worldwide by type of information included (authors’ own elaboration).

Given that VLRs vary widely in terms of content, format, and language, and in order to be able to extract consistent information and ensure a robust and manageable indicator set, we restricted our analysis to VLRs fulfilling the following criteria:

• Language: only VLRs written in English were included, narrowing the sample to 190 reports.
• Content: VLRs that provide non-indicator or monitoring-related qualitative information were excluded from the analysis (N = 70), due to the scope of this paper on monitoring practices.

As a result, out of the 325 VLRs published at the time of writing, this analysis focuses on the 120 that are in English and contain quantitative and/or qualitative information on monitoring the SDGs at a local level through concrete indicators. This sample includes 42 VLRs from Europe, 40 from Asia, 14 from North America, 13 from Africa, 8 from South America, and 3 from Oceania. Moreover, the sample encompasses different categories of local authorities, with a predominance of cities (70.2% of VLRs in the sample), followed by regions (9.6%), municipalities (7.4%), counties (4.3%), districts (3.2%), states (3.2%), and provinces (2.1%).

2.2. Data Collection Process

The VLR-related information collection process involved a general review of all published VLRs, in order to record the necessary data in a structured database. The information recorded for each VLR included: its official title (as it appears in the published document); a hyperlink to the document source; the name of the locality for which the VLR was developed, as stated in the report; the country and continent classified according to the official UN definitions; the year and language of publication; the name and type (e.g., city, region, county, or municipality) of the local or regional government that issued the VLR, as stated in the report; and the type of information—author’s classification as qualitative (monitoring practices not necessarily reported, no data on SDG indicators, only narrative interpretation), quantitative (only data on SDG indicators with no narrative interpretation), or both (a mix of data and narrative).

The collection process involved a thorough review of the 120 identified VLRs to systematically extract relevant information on monitoring indicators. This was achieved by reading the VLRs and recording the necessary data per indicator in a structured database. For each indicator reported in the VLRs, the full indicator name as reported in the VLR, the associated SDG, and the target were also reported. Where this association was not provided, it was curated by the authors (e.g., associating neonatal mortality rate with SDG 3, Target 3.2). In total, 10,506 indicators were registered in the dataset.

2.3. Analysis Methods

With the collected data, we performed several analyses to understand the practices and patterns of SDG monitoring at the local level.

• Quantitative analysis: This paper’s quantitative analysis offers a comprehensive view of how local and regional governments monitor the SDGs using various indicators. We focused on two main aspects: variations in reporting practices across local governments and the number of indicators used per SDG. Our analysis included 120 VLRs and 10,506 indicators, categorised by SDG and target. Descriptive statistics, including mean, median, and percentile ranks, were used to identify central tendencies, distribution patterns, and any skewness in indicator usage. We visualised these distributions to highlight over- or under-reported SDGs. Additionally, we examined differences in indicator usage by local government type and across EU geographical regions.

• Semantic analysis and clustering: This was conducted on the 10,506 indicators included in the database to identify patterns and clusters based on indicators denomination, determining if they align with the 17 SDG goals or reveal new areas of interest for the local context. For this analysis, we employed the k-means clustering method of vector quantisation, widely used for partitioning observations into clusters to minimise the within-cluster sum of squares (WCSS) [30,31]. We opted for k-means clustering due to its simplicity, interpretability, and effectiveness in handling large datasets. The method is widely used for partitioning data into clusters that minimise within-cluster variance, aligning with our goal of grouping semantically related indicators into clusters corresponding to the SDGs. Additionally, the alignment of k-means with predefined clusters (17 SDGs) made it particularly suitable for this task. Each observation—in our case, the name of an indicator—was treated as a d-dimensional real vector. K-means clustering aims to partition these n observations into k (≤n) sets S = {S1, S2,…, Sk} to minimise WCSS. In our study, observations equal the names of indicators, and we transformed these indicator names into numerical vectors using the Sentence BERT (SBERT) encoder, which converts the full name of each indicator into an array of 384 values, regardless of the length of the name, resulting in a matrix of 10,506 indicators by 384 values for each name [32]. We selected the SBERT model due to its proven capability in transforming textual data into meaningful numerical vectors. It has been effectively applied in various semantic analysis and clustering tasks, especially within sustainability research contexts. This method aligns with our study’s requirements to maintain high dimensionality (d = 384) and facilitate consistent clustering across a large, diverse set of indicators [33,34]. To initialise the 17 clusters corresponding to the 17 SDGs, we used the Forgy initialisation method, which randomly selects k observations from the dataset as the initial means, spreading the initial means out across the data [35,36]. The clustering process was refined using Lloyd’s algorithm, which iteratively improves the assignment of observations (indicators) to the 17 clusters (k) [37]. Each observation is assigned to the cluster with the nearest mean, determined by the least squared Euclidean distance, while the means (centroids) of the clusters are recalculated based on the observations assigned to each cluster [38]. The objective function in k-means is the WCSS, which decreases after each iteration, ensuring a nonnegative monotonically decreasing sequence, and guarantees convergence of Lloyd’s algorithm. By setting the number of clusters to equal the number of SDGs, we aimed to observe if the resulting clusters aligned with the existing SDGs, revealing any overlaps and discrepancies in the indicators used for localised SDGs monitoring.

3. Results and Discussion

In order to understand how the process of localising SDGs has taken place through subnational monitoring exercises, this section analyses reporting patterns and frequent practices observed across the VLRs reviewed. Firstly, we identify the main goals monitored locally, the average number of indicators used, as well as goals left behind. In doing so, we established a comparison with the UN standard framework and the 2024 EU monitoring framework, in order to identify differences and similarities in local and regional monitoring. Secondly, we proceeded with a similar analysis by type of subnational authority to grasp similarities, differences, and possible patterns.

3.1. Quantitative Analysis

3.1.1. Goal Coverage (% of VLRs Reporting Indicators per Goal)

Local reporting and monitoring activities feature broad differences in addressing the different goals (Figure 5). Fourteen of the SDGs are reported in at least 60% of VLRs reviewed, with four of them being covered by over 80% of the VLRs (Goals 3, 4, 8 and 11), whereas Life on Land (Goal 15), Partnership for the Goals (Goal 17) and Life below Water (Goal 14) register major shortfalls, appearing in 58%, 47%, and 27% of the VLRs reviewed.

Figure 5. Distribution of VLRs per goal reported (authors’ own elaboration).

This unequal coverage of the goals is the result of the combination of local priorities, constraints, and data availability. As an example, among the most reported goals we encountered that were well established in official statistics concerned education, economic growth, and health. Similar coverage reached for Goal 11 testifies to a remarkable effort from local authorities in this regard. However, the analysed sample is strongly represented by city reporting, and local authorities have a clear interest and probably good data sources to monitor sustainable cities and communities (Goal 11). Additionally, the indicators and targets of Goal 11 can be more easily anchored to local contexts when not already established on a local basis compared to goals such as Life on Land (Goal 15) or Partnership for the Goals (Goal 17), for which targets and indicators have a strong national and supranational focus and therefore require translation into local contexts. Finally, certain goals are designed for specific geographical conditions and might therefore be excluded from reporting, e.g., Goal 15 (Life on Land), with limited appearances in local reporting.

3.1.2. Number of Indicators to Monitor the SDGs

The number of indicators used in local monitoring practices varies broadly across territories (from a minimum of 6 used by Viken County in its 2021 VLR to a maximum of 497 reported by São Paolo in its 2021 report). On average though, the analysed VLRs used 87 indicators to monitor the SDGs at local level.

A similar variety of frameworks and practices can be observed per goal, with an average number of indicators/goal of approximately 14 in the UN framework and 8 indicators/goal in the Eurostat 2024 indicator set (Figure 6). In the latter, large differences in the number of indicators used to monitor are avoided, in order to keep balance between goals. This is not the case for the reviewed VLRs, where the distribution of the number of indicators shows an unbalanced representation per goal, ranging from 11 or more indicators for goals most frequently informed in VLRs (such as 11, 4, 3 and 8) to 4 indicators or fewer for rarely reported goals (such as 17 and 13). This points to important reporting challenges in local SDG monitoring. On average, goals such as 17, 16, and 3 present a significant lack of indicators compared to the UN proposal. As previously mentioned, this could result from local priorities and circumstances (i.e., data availability), as well as being linked to the national/supranational focus of the goal. Moreover, in many countries, Goal 3 is a policy field of national or regional competence: although it appears to be a priority in local level reporting with regard to the percentage of VLRs covering it (Figure 5), most policies would not fall within the field of action of local government. In the UN framework, Goals 16 and 17 are informed through indicators that are either clearly addressing the national level or cover themes that are out of standard official statistics fields. As a consequence, to monitor these goals, local authorities would need to work on their initial definition, translate it in terms that can be meaningful at the local level, and identify the related indicators. This exercise requires an advanced understanding, maturity, and engagement with the goals [39].

Figure 6. Average number of indicators reported per goal observed in VLRs reviewed. The green triangles and red lines represent the suggested average EUROSTAT and UN number of indicators per goal, respectively (authors’ own elaboration).

The average number of indicators used to monitor SDGs locally varies according to data or information availability (Figure 7): the higher the average number of indicators, the more variability in the number of indicators observed across the VLRs. If we consider a higher number of indicators as a sign of monitoring engagement or commitment, we can identify, per goal, the median number of indicators used by the VLRs to set a threshold or minimum number of indicators used in those VLRs reporting exhaustively in terms of number of indicators. Despite significant outliers, we can conclude that better reporting practices use a minimum number of indicators, ranging from 2 for Goal 14 to 11 for Goal 11. However, for most of the goals (13 of 17), this figure ranges from two to five.

Figure 7. Box plot of number of indicators per goal (in different colours) in VLRs reviewed.

3.1.3. Indicator Use by Issuing Authority

On average, the analysed monitoring frameworks feature 87 indicators to report on SDG localisation; however, clear differences can be spotted in relation to different types of local authorities. As depicted in Figure 8, cities tend to use a larger number of indicators compared to other types of local authorities, and the mean is significantly higher (=100), indicating that some cities might be using a very large number of indicators, which skews the average up (median = 75). Regions appear to use fewer indicators on average and register a lower range of variation in the number of indicators than cities. States, on the other hand, appear to have a mean (=69) and median (=72) that are very close, indicating a more symmetric distribution of the number of indicators used across goals. Provinces and counties show a similar tendency. Districts use a moderate number of indicators, with the mean being slightly higher than the median (55 vs. 48). Finally, municipalities feature the fewest number of indicators on average (32).

Figure 8. Box plot of number of indicators per issuing authority (in different colours) in VLRs reviewed.

If the number of indicators reflects engagement and commitment in SDG reporting and monitoring, cities are clearly leading, followed by regions and states, though significantly lagging behind. Provinces, counties, and municipalities remain in the lower range of reporting. Our analysis confirms that cities are the most active and experienced local authorities, with 18% having published two or more VLRs.

3.2. Semantic Analysis and Clustering

To analyse patterns in the 10,506 indicators from our database, we applied semantic analysis and clustering techniques to explore how well these indicators align with the 17 SDGs. Using k-means clustering, we aimed to identify clusters based on indicator names and uncover potential new areas of interest specific to local contexts. This approach allowed us to investigate overlaps, gaps, and discrepancies in localised SDG monitoring.

Figure 9 depicts the 17 resulting clusters and the spatial distribution of the 10,506 dots/indicators. The indicators are positioned in regard to the meaning of the full sentence forming them and grouped around the 17 themes that appear as more prominent. The closer the meaning of two indicators is, the closer the dots that represent them will be. As observed, the clusters positioned towards the centre of the figure are denser, being mostly populated by points located very closely one to another (e.g., clusters 6, 1, 13, 16), while the clusters positioned in the periphery of the image include also more scattered points (e.g., clusters 12, 9, 11, 15). This indicates that the clusters positioned towards the centre of the image are made up of indicators that are semantically very closely related, with a very similar meaning or touching upon topics that are strongly connected, while those that are at the periphery include also less closely related indicators—addressing topics that are further apart.

Figure 9. Cloud of 17 clusters and their centroids based on semantic analysis of SDG indicators (authors’ own elaboration).

Figure 10 displays the 20 most frequent words appearing per cluster in the indicator names, the number of indicators that each cluster represents, and the 3 most frequent words per each cluster. The 17 clusters vary significantly in size, ranging from 126 indicators in cluster 13 (tree/forest/woodland) to 1202 indicators in cluster 1 (education/student/school), with an average of 618 indicators per cluster. As each cluster is semantically correlated, certain words appear in multiple clusters due to their relevance to different topics. For example, the word ‘women’ is prominent in both cluster 5 (gender equality) and cluster 9 (violence-related issues).

Figure 10. The 17 clusters and corresponding word cloud, showing the 20 most represented words (authors’ own elaboration).

The combined observation of Figure 9 and Figure 10 allows some further considerations. Cluster 13, though the smallest, appears to be highly consistent in terms of content (as formed by dots that are positioned very closely ones to another), while clusters 2 and 12, despite being large, are less cohesive, as they also include relative distant dots (therefore, indicators with relatively disconnected meanings/content). Cluster 1 is both the largest and one of the most consistent.

A comparison between clusters and goals reveals four cases: (1) a 1-to-1 correspondence from a single cluster to a single goal, (2) a 2-to-1 correspondence from two clusters to a single goal, (3) a 1-to-2 correspondence from a single cluster to two different goals, and (4) no direct correspondence between clusters and goals.

• In the first case, cluster 1 aligns with Goal 4 (Quality Education), cluster 4 with Goal 1 (No Poverty), cluster 5 with Goal 5 (Gender Equality), and cluster 13 with Goal 15 (Life on Land).
• In the second case, two clusters inform one goal: Clusters 6 and 10 for Goal 3 (Good Health and Well-Being), clusters 7 and 15 for Goal 6 (Clean Water and Sanitation), clusters 9 and 11 for Goal 16 (Peace, Justice, and Strong Institutions), and clusters 12 and 14 for Goal 8 (Decent Work and Economic Growth).
• In the third case, cluster 2 informs both Goal 7 (Affordable and Clean Energy) and Goal 13 (Climate Action).
• The fourth case includes clusters 3, 8, 16, and 17, which show no direct correspondence with seven goals (2—Zero Hunger; 9—Industry, Innovation, and Infrastructure; 10—Reduced Inequalities; 11—Sustainable Cities and Communities; 12—Responsible Consumption and Production; 14—Life below Water; and 17—Partnerships for the Goals). Goals like 11 (Sustainable Cities) and 12 (Responsible Consumption and Production) are overlapping, spanning multiple clusters, while Goals 2 (Zero Hunger) and 9 (Industry, Innovation, and Infrastructure) also exhibit transversal characteristics. Goals 10 (Reduced Inequalities), 14 (Life below Water), and 17 (Partnerships for the Goals) do not feature prominently in the clusters, likely due to their focus areas or low representation in the dataset.

A direct correspondence between one or two of the 17 clusters and one or two of the 17 SDGs was established for 10 goals only. These were mostly goals with clearly defined objects (e.g., hunger, education, economy, water, energy, gender equality, climate change), which are perceived as such by local authorities and therefore reflected in practice through the use of indicators that align very well with these goals. Such insights could inform the custodian institutions and the ongoing discussion on how to foster localisation practices and where to find gaps.

Among the seven goals not included in the direct correspondence, we find overlapping goals that encompass diverse topics, thus being at the same time complex to apprehend in terms of contents and scopes and difficult to implement as they might not align well with the structure of local authorities. For example, Goal 11 (Sustainable Cities and Communities)—highly represented due to the large proportion of city-based VLRs in the analysed sample—is scattered across multiple clusters, possibly because many indicators focus on more localised actions or specific policy areas rather than the broader ‘city’ scale. Thus, when VLRs focus on smaller territorial scales, indicators for Goal 11 might need to be made more context-specific, addressing neighbourhoods or smaller localities. Similarly, the high proportion of cities in the sample may explain the results for Goal 2 (Zero Hunger), given the challenges of conducting agricultural activities within city boundaries. Semantic analysis helps in revealing the challenges that local administrations might face when monitoring overlapping goals, as well as a potential need for internal reorganisation, or for strengthening internal collaboration to accelerate the progress in specific areas. Goal 12 (Sustainable Consumption and Production), which spans tourism, waste, and circularity, requires cross-departmental coordination to collect data. In contrast, goals more aligned with specific local departments (e.g., health, education) may also show clearer alignment with specific clusters.

Finally, the case of Goal 15 (Life on Land) and its corresponding cluster 13 are worth analysing, as this adds nuance to the findings of Section 3.1.1: despite figuring among those less covered by VLRs (58% of VLRs), Goal 15 is clearly identified in a dedicated cluster in the semantic analysis, although the smallest one. In other words, when Goal 15 is monitored at local level, this is done in a consistent way (as also shown by the proximity of points in Figure 9), bringing recognition to the relevance of this goal as a stand-alone topic at local level. Moreover, the proximity with cluster 16 could indicate an increased relevance of Goal 15’s sphere (Life on Land).

Among the 17 clusters identified through semantic analysis, 4 are not directly aligned with specific SDGs. For these clusters, a deeper analysis was performed using word clouds representing the 50 most frequent terms in each cluster and a review of all indicators within each cluster.

Cluster 3, the second largest with 969 indicators, centres on the concept of service provision, extending beyond specific services to residents. This cluster can be classified as functional, in contrast to the thematic clusters (clusters 8, 16, 17) discussed later. Cluster 17, comprising 667 indicators, focuses clearly on transport, and cluster 8, with 359 indicators, on waste. Two interpretations arise from this: either a need for dedicated SDGs on transport and waste, or the conclusion that local authorities are already proficient in managing these areas, given the extensive range of indicators and the capacity to monitor them consistently (which could be linked to the timeframe of local mandates, allowing effective policy implementation in these areas). However, cluster 8 highlights that waste is primarily monitored from the perspective of its collection and recycling, with little attention to reducing waste generation or promoting circular economy strategies. While the topic of waste is well covered, the indicators do not yet reflect the desired global direction, particularly concerning sustainability. This suggests the need for more forward-looking monitoring frameworks at the local level, potentially dividing indicators into two subsets: one for established indicators and another for areas requiring urgent development and innovation. Cluster 16, with 630 indicators, tells a different story: it focuses on green spaces, including parks, productive areas, and natural environments, with a clear emphasis on land protection and management. The most frequent terms (biodiversity, conservation, and protection) resonate with global environmental concerns and align with recent European policy developments, such as the Nature Restoration Law. This consistency suggests that a dedicated SDG on land management for environmental preservation could be designed, transcending landscape divisions and acknowledging the overlapping importance of this issue.

The need for additional SDGs has been debated both in academic and grey literature, as well as among stakeholders and think-tanks. A broad number of suggestions have been provided for an 18th SDG, including topics such as animal health, welfare and rights [40]; democracy, development, and international assistance [41]; communication [42]; culture [43]; social entrepreneurship [44]; space environment [45]; digital connections [46]; and housing [47] (to name a few). Most of those suggestions are intended to serve advocacy purposes, bringing attention to one specific issue or fields or highlighting under-explored nexuses (e.g., SDGs and environmental justice, animal and planet health), and come from global perspectives.

In this respect, the present research could contribute to the discussion from a quite different—and possibly complementary—angle, as it brings insights from the bottom up on which topics/issues are addressed by the monitoring practices at local level and which are missing, while looking at all SDGs in a systematic manner. The results presented herein might therefore prove useful not only in informing local policies but also in complementing the strategic reflections taking place at the global level on the assessment of the UN monitoring framework by providing a snapshot of what is reflected in the reality of the practice on the ground.

4. Conclusions

This research provides a comprehensive analysis of how SDGs are monitored at the local level through voluntary local reviews (VLRs). The analysis highlights the significant role cities play in SDG localisation, with varying levels of engagement and maturity depending on the type of local authority. Cities emerge as the most active participants, particularly in areas directly aligned with local governance mandates, such as Goal 11 (Sustainable Cities) and Goal 4 (Quality Education). In contrast, SDGs like Goal 15 (Life on Land) and Goal 17 (Partnerships for the Goals) are less frequently monitored, reflecting both local priorities and the inherent challenges in translating national or supranational objectives into local contexts.

The twofold approach of this research, combining descriptive statistics and semantic analysis, enriches our understanding of local SDG monitoring. While the descriptive analysis provides insights into the number of indicators used and the types of goals monitored, the semantic clustering offers deeper insights into the meaning and interlinkages between goals. For instance, the strong alignment of Goal 4 with Cluster 1 in the semantic analysis suggests that this goal is not only widely monitored but also well supported by local data. This points to its successful localisation, likely due to the well-established nature of education statistics and the clear alignment between its targets and local governance. However, the uneven distribution of indicators across different SDGs reveals a challenge. Goals such as 12 (Responsible Consumption and Production) and 11 (Sustainable Cities) are often fragmented across multiple clusters, indicating that local governments face difficulties in monitoring overlapping themes. This suggests the need for more targeted approaches that incorporate context-specific indicators, especially in urban areas where these goals are most relevant.

The predominance of urban authorities in our dataset (70.2% of VLRs analysed) highlights that SDG localisation is primarily an urban affair. Cities, with their articulated governance structures and capacity for data collection, are better positioned to monitor SDGs comprehensively. As a result, the findings of this research are largely reflective of urban contexts, which may explain the low representation of goals related to rural areas, such as Goal 2 (Zero Hunger).

A key takeaway from this study is that the more experienced local authorities tend to use a higher number of indicators. Cities with multiple VLRs often demonstrate a more mature and nuanced understanding of SDG monitoring. However, local monitoring frameworks generally feature fewer indicators and are differently distributed across goals compared to supranational frameworks like those of the UN and EU. Disparities underscore the need for tailored, locally relevant indicators that align with both global targets and local realities.

The results also suggest future research directions. Firstly, a similar quantitative analysis could be conducted focusing on targets, trying to identify under-reported ones. Secondly, expanding the analysis to include all published VLRs, regardless of language, would provide a broader understanding of SDG localisation practices. Thirdly, exploring the evolution of European VLRs in relation to EU policy developments could offer valuable insights into the interplay between regional policies and local monitoring. Additionally, focused studies could cover specific geographical contexts that face similar challenges (e.g., small islands, coastal areas, etc.) to analyse if and how monitoring frameworks are adapted. Moreover, further semantic analysis could be applied to overlapping goals to better understand their complex nature and the synergies or trade-offs they may present at the local level. Additionally, future research could explore the optimal number of clusters using unsupervised methods, such as the elbow method or silhouette analysis. This approach could provide valuable insights into synergies and interdependencies between SDG goals, offering a deeper understanding of overlapping themes and interlinkages in localised SDG monitoring frameworks. Such an analysis would complement the findings of this study by further uncovering areas where SDG goals intersect or overlap. Lastly, this study does not approach the governance sphere that underpins and enable data production; to improve the understanding of local monitoring practices. It would be highly relevant to also analyse the governance settings and the relevant strategies put in place by local authorities for SDG monitoring.