University sustainability rankings and sustainability performance: assessing the relationship

The ongoing debate on rankings in the higher education sector (Marginson and Van Der Wende, 2007) has recently been affected by the introduction of sustainability rankings (Kaiser et al., 2022). Sustainable development represents one of the greatest challenges currently facing society, and education has been recognized as a crucial element in addressing this issue (Lozano et al., 2015; Rau et al., 2018). Over the past years, rankings have become widely used tools for assessing and comparing university quality (Hazelkorn, 2014; Urdari et al., 2017). However, their impact and credibility remain subjects of extensive discussion (Kauppi, 2018).

Given the societal push towards sustainability and the expected contribution of Higher Education Institutions (HEIs), new rankings have emerged to measure their commitment to sustainable development (Gutiérrez-Mijares et al., 2023; Suwartha and Sari, 2013). The most prominent are the UI GreenMetric (Henceforth GM ranking), the Times Higher Education Impact ranking (THE-IR) and the QS Sustainability ranking (QS-SR). Like traditional rankings, sustainability rankings face criticism regarding their impact and reliability in measuring the quality of higher education (Lauder et al., 2015; Saisana et al., 2011). Moreover, the multifaceted nature of sustainability complicates its assessment (Adams et al., 2014; Böhringer and Jochem, 2007).

Despite criticism and new proposals for evaluating universities, rankings remain pervasive in the higher education sector, highlighting the need to examine both their validity and impact (Olcay and Bulu, 2017). While extensive literature exists on rankings (e.g. Marginson and Van Der Wende (2007)), sustainability rankings remain scarcely studied. Some researchers have addressed the assessment issue through a qualitative analysis of indicators (Bautista-Puig et al., 2022; Galleli et al., 2022; Suwartha and Sari, 2013), but studies specifically examining the effects of these rankings are limited (Atici et al., 2021; De La Poza et al., 2021; Sierra-García et al., 2024).

This study addresses this gap by investigating the relationship between sustainability performance and sustainability rankings. First, while prior literature has examined rankings’ validity (Kaiser et al., 2022) and their organizational effects (Rindova et al., 2018) as separate issues, we argue that these two dimensions are inherently correlated. Accordingly, we conceptualize the relationship between rankings and performance as a complex bidirectional process. Second, as rankings can be interpreted as evaluative tools that signal institutional quality, we posit that improvements in sustainability performance should be associated with corresponding changes in ranking outcomes, provided that such signals are not distortive (Espeland and Stevens, 1998; Spence, 1973). Third, rankings as external evaluative tools can shape organizational responses and incentivize universities to adapt their strategies (Oliver, 1991; Rindova et al., 2018). Drawing on organizational and institutional theories (DiMaggio and Powell, 1983), we argue that rankings create incentives to prioritize performance dimensions that are rewarded, even at the expense of other aspects of sustainability.

To investigate these mechanisms, we collected data on 75 Italian HEIs from 2010 to 2023 and analyzed the relationship using the GM ranking. We consider data regarding three main university missions (teaching, research and third mission). All data were obtained from publicly available sources, ensuring transparency and reproducibility. Unlike most existing studies that rely on ranking-collected data (Atici et al., 2021), we measure performance externally. This allows us to directly assess ranking validity while including both ranked and nonranked institutions. Finally, Italy is selected as an appropriate empirical setting for the study due to the early adoption of this ranking and the recent governmental emphasis on sustainability. To test our hypotheses, we used the Granger causality test and dynamic panel models.

The findings suggest that the GM ranking and sustainability performance stand in a bidirectional relationship. The GM ranking responds positively to improvements in performance, particularly those related to teaching and research, thereby supporting its validity. At the same time, our results indicate that rankings can significantly shape sustainability performance, encouraging improvements in dimensions that enhance the score while potentially neglecting others.

Empirical insights into the relationship between sustainability ranking and performance can greatly benefit key stakeholders, such as university managers and policymakers. Managers can leverage this knowledge to optimize their strategic planning and operational management (Rindova et al., 2018). Similarly, policymakers frequently use university rankings as a proxy for institutional performance due to the lack of information (Dill and Soo, 2005). A more detailed understanding of the rankings can facilitate the formulation of more informed and targeted policies.

The remainder of the paper is organized as follows. Section 2 presents a review of the literature on the integration of sustainability in HEIs and provides an overview of sustainability ranking systems. Section 3 outlines the theoretical framework and develops the corresponding hypotheses. Section 4 describes the research design, including data and methodology. Section 5 presents and discusses the results. Section 6 provides a detailed examination of the implications. Finally, Section 7 concludes the study and discusses its limitations.

Sustainability is a broad concept, and since its introduction, many definitions have emerged. The most cited definition from Brundtland (1987) describes sustainable development as meeting present needs without compromising the ability of future generations to meet their own. However, given the widespread integration of sustainability concepts across various research fields and the resulting proliferation of interpretations, establishing a single universally accepted definition remains challenging. What remains consistent across these definitions is the emphasis on the interconnection among the economy, environment and society, with the idea that these elements should support or reinforce one another in a reciprocal relationship (Elkington, 1998).

In the context of transitioning to a sustainable society, higher education has been recognized as a key driver of sustainability transformation (Gutiérrez-Mijares et al., 2023; Miller et al., 2021). Velazquez et al. (2006) defined a sustainable university as an institution that integrates sustainability into its teaching, research, outreach and operations to minimize its negative environmental, economic and social impacts.

This has led universities to change their operations, redefine their missions and shift organizational cultures to accommodate sustainability-related concerns (Prior et al., 2024). Given the lack of consensus on the definition of sustainability, initiatives within the higher education sector are diverse (Gutiérrez-Mijares et al., 2023). A sustainable university can be defined in many ways, depending on the perspective taken (Lozano et al., 2015). The current literature highlights several key areas, including education, research, community engagement or outreach, campus operations and assessment or reporting practices (Fischer et al., 2015; Lozano, 2006).

With the growing focus on sustainability, universities and policymakers have started considering metrics to assess institutions based on their sustainability performance (Gutiérrez-Mijares et al., 2023). Nevertheless, sustainability measurement adds another layer of complexity to the already challenging task of evaluating HEIs (Lauder et al., 2015). Performance measurement serves various purposes, such as evaluation, communication, planning future actions and making comparisons (Findler et al., 2019). Accordingly, various assessment tools and measures have been developed to evaluate sustainability alongside traditional indicators of excellence (Adams et al., 2014; Otto et al., 2021).

In general, universities pursue three main missions: teaching, research and outreach – also known as third mission – all of which play a crucial role in the transition to sustainability (Stephens et al., 2008). First, HEIs have a social responsibility in societal development through the education of future generations. According to Emanuel and Adams (2011), the knowledge of sustainability is essential for the transition and universities can integrate this concept into the curriculum using three distinct methods: incorporating it into existing courses, offering specialized courses and establishing entire sustainable programs (Lukman and Glavič, 2007). In any case, university programs can substantially affect the perception and attitudes of students (Kagawa, 2007).

In addition to their educational mission, university research plays a key role in knowledge creation (Confraria et al., 2024) and is pivotal for sustainable development (Waas et al., 2010). Various bibliometric measures are used in research evaluation; among them, two simple yet informative indicators are the number of publications and the number of citations (Ellegaard and Wallin, 2015). The former can be considered a proxy for research production, while the latter is commonly used as an indicator of research impact (Moed, 2005; Waltman et al., 2011). Although highly debated, evidence suggests that citation counts are positively associated with research quality (Aksnes et al., 2019). Furthermore, despite the complexity and controversy surrounding bibliometric indicators, these measures become less susceptible to criticism the more they are aggregated (Aksnes et al., 2019).

Next, the area commonly defined as the third mission, or outreach, must be considered (Jongbloed et al., 2008). This mission encompasses activities aimed at transferring knowledge to society, including various forms of communication and social engagement (Di Berardino and Corsi, 2018). The connection with stakeholders outside academia is crucial for the sustainability transition and includes creating spinoffs and startups, generating intellectual property and participating in collaborative research, all of which support social and sustainability objectives (Trencher et al., 2014).

In addition, the campus environment plays a crucial role in sustainability transition due to the various complex activities taking place within it (Alshuwaikhat and Abubakar, 2008). A sustainable university campus should promote energy and resource conservation, waste reduction, efficient environmental management and equity and social justice (Lozano et al., 2015).

Given the advocacy for sustainable development, HEIs face growing pressure to adopt and demonstrate sustainable practices. Showcasing this commitment has become essential, driven by stakeholder demand (Olcay and Bulu, 2017) and government calls for accountability and transparency (Bleiklie, 1998). HEIs can achieve this either through direct communication of their activities (Dienes et al., 2016) or by relying on third-party assessments (Prior et al., 2024).

In this context, third-party assessments have gained momentum, largely due to their perceived legitimacy (Devinney et al., 2008). Among external indicators of university performance, rankings are increasingly valued for translating qualitative attributes into standardized, comparable metrics (Calderon, 2023; Lozano, 2006). As excellence in sustainability is viewed as a natural extension of universities’ excellence, there is a growing demand for global rankings that assess universities using sustainability criteria (Johnes, 2018), providing an alternative perspective for evaluating institutions (Calderon, 2023).

Within this context, sustainability rankings provide opportunities and a framework for building a green university (Suwartha and Sari, 2013). However, these rankings also present criticalities, including methodological limitations (Lauder et al., 2015), challenges in data collection (Ragazzi and Ghidini, 2017), the presence of biases and a lack of real commitment (Alberti et al., 2025), as well as difficulties in ensuring comparability (Calderon, 2023). Nonetheless, several scholars argue that, even with imperfections, sustainability rankings can be desirable tools to incentivize environmental progress (Atici et al., 2021; Galleli et al., 2022; Smolennikov et al., 2024).

The first global ranking focused on sustainability is the GM ranking, conducted by the University of Indonesia and published annually since 2010 (Galleli et al., 2022). GM includes six categories of indicators: Settings and Infrastructure; Energy and Climate Change; Waste; Water; Transportation; and Education and Research. GM ranking has successfully disseminated the topic of sustainability and gained substantial popularity in recent years (Boiocchi et al., 2024), as reflected in the increasing number of participating institutions. However, like any ranking, it has drawbacks, including concerns about the validity and reliability of the data, because the ranking is based on a self-administered questionnaire with data provided by the universities themselves (Dill and Soo, 2005).

The success of GM and the growing importance of sustainability have also contributed to the development of other rankings, such as the Times Higher Education Impact ranking (THE-IR) and the QS Sustainability ranking (QS-SR). The THE-IR, published annually since 2019, measures universities’ contributions to the SDGs using research, institutional and policy indicators. The QS-SR, first issued in 2022, highlights how universities address key environmental, social and governance (ESG) challenges.

Rankings were originally developed as information products aimed at measuring and signaling university performance (Rindova et al., 2018; Spence, 1973). Within this framework, rankings represent quality assurance tools, helping to reduce information asymmetry between universities and their stakeholders (Chatterji and Toffel, 2010). This signaling function is particularly important in areas such as higher education, where quality assessment can be complex or subjective (Rindova et al., 2018). This transparency enables more informed decision-making by students, families, funding bodies and policymakers (Nassa and Arora, 2021).

However, rankings are not merely passive reflections of performance. A growing body of literature suggests that they also shape organizational behavior, influencing outcomes such as research quality, student demand, technology transfer and financial resources (Bastedo and Bowman, 2011; Horstschräer, 2012; Sine et al., 2003). This dual role implies that rankings may act both as evaluative mechanisms and as drivers of change (Siler and Larivière, 2022). Building on this, we argue that the relationship between rankings and sustainability performance is inherently bidirectional: rankings are expected to reflect underlying performance, while at the same time influencing it. Accordingly, we introduce the following hypotheses:

A key assumption underlying the use of rankings is that they provide a credible signal of organizational quality. Drawing on signaling theory (Spence, 1973), a signal is informative only if it accurately reflects the underlying attributes it intends to convey (Connelly et al., 2011). In this sense, university rankings should mirror actual improvements in institutional performance.

Nevertheless, this assumption has been widely debated. Rankings have faced substantial criticism (Marginson and Van Der Wende, 2007). Scholars have questioned the validity and reliability of the indicators, aggregation models and data sources used (Dill and Soo, 2005; Hamann and Ringel, 2023). Concerns have also been raised about manipulation strategies, including performance enhancement through self-reported data or the strategic inflation of participation and input measures (Hazelkorn, 2014). Finally, De Witte and Hudrlikova (2013) and Barron (2017) argue that rankings frequently reflect reputation and visibility rather than genuine institutional quality, favoring well-established universities.

However, despite these criticisms, rankings remain widely used within organizations and by external audiences. Internally, they provide performance benchmarks that can support strategic decision-making and resource allocation (Nõmm and Randma-Liiv, 2012). Externally, they respond to the growing demand for accountability by informing stakeholders about institutional activities (Costa and Pesci, 2016).

Given their widespread adoption and intended signaling function, we argue that improvements in sustainability performance should be reflected in ranking positions. Accordingly, rankings can be interpreted as informative signals of underlying performance, even though part of the underlying information is based on self-reported data provided by universities. Importantly, our study does not seek to test signaling theory per se, which already provides the conceptual foundation for interpreting rankings as signals. Rather, we build on this framework to empirically examine whether these signals are aligned with independently measured performance:

Beyond reflecting performance, rankings can actively influence organizational behavior. Their growing prominence has increased the pressure on universities to improve their positions, thereby shaping strategic decisions (Kaidesoja, 2022; Shore and Wright, 2015).

From an institutional perspective, rankings act as external pressures that organizations are incentivized to respond to. Drawing on the theory of institutional isomorphism (DiMaggio and Powell, 1983), organizations tend to become more similar when exposed to such pressures through three mechanisms: coercive, mimetic and normative. University rankings operate through all three of these mechanisms: coercive mechanisms arise when rankings are linked to funding allocation or regulatory frameworks (Marginson, 2007); mimetic mechanisms occur when institutions imitate higher-ranked universities perceived as successful models (Grewal et al., 2008); and normative mechanisms reflect professional and academic expectations within the higher education community (Hazelkorn, 2008).

A key mechanism underlying this response is reactivity, defined as behavioral change in response to being measured and evaluated (Espeland and Sauder, 2007). Because ranking positions are highly visible, universities are incentivized to improve them, often by adapting their internal processes and resource allocation (Fowles et al., 2016; Tóth et al., 2024). As part of this reactive process, universities tend to prioritize activities that are explicitly captured by ranking methodologies. This often translates into increased investments in areas known to improve ranking scores (Hazelkorn, 2014). As a consequence, rankings do not merely evaluate performance but actively shape it, influencing how organizations define and pursue success (Devinney et al., 2008).

This behavior is further reinforced by external stakeholders, who rely on rankings to evaluate institutions (Martins, 2005), as well as by competitive dynamics, where top-ranked universities set benchmarks that others seek to emulate (Jones, 2017). In addition, from a resource-based perspective, higher-ranked universities may gain access to additional resources, further strengthening incentives to invest in ranking-relevant dimensions (Rindova et al., 2018; Waddock, 2008).

However, the prioritization of ranking-related dimensions may come at a cost. Given resource constraints, increased investment in measured areas may lead to the neglect of nonmeasured activities. This dynamic is consistent with the idea that performance metrics can induce selective optimization, encouraging organizations to focus on what is measured rather than on overall performance (Heinrich, 2002; Rindova et al., 2018). In this context, universities may improve indicators included in rankings while overlooking broader or less visible aspects.

Finally, it should be acknowledged that such reactive behavior is not universal. Universities may resist rankings, selectively engage with them or attempt to influence evaluation criteria (Clementino and Perkins, 2021; Greenwood et al., 2011; Pollock et al., 2018). Moreover, rankings can also foster genuine improvements by enhancing accountability and transparency, encouraging the development of monitoring and reporting systems that support broader performance gains (Clementino and Perkins, 2021; Hamed et al., 2025).

Nevertheless, even in the presence of these positive effects, we argue that the combination of external pressure, resource constraints and measurement focus suggests that ranking-induced improvements are likely to be unevenly distributed across performance dimensions. This reasoning leads to the formulation of the following hypothesis:

Figure 1 summarizes the relationships under investigation with the hypotheses formulated and their expected directions and signs.

This research focuses on the Italian higher education system. Italian universities provide a relevant case for several reasons. First, they are involved in sustainable development, as evidenced by their participation in the Network for Sustainable Development (RUS). This network was established in 2015 and now includes almost all universities. Second, recently, the National Recovery and Resilience Plan (PNRR) has provided funding to promote sustainable development in universities, making this case interesting for policymakers (Sacchi et al., 2023). Finally, since the first edition of the GM ranking, several Italian universities have participated, and this number continues to grow, showcasing their commitment to sustainability.

Italian universities are distributed across the national territory and, regardless of the issuing institution, university degrees have the same legal value throughout the country (Cattaneo et al., 2017). In line with the objectives of the study, we exclude special universities and distance-learning institutions due to differences in their student populations and institutional mission ^[1] (see e.g. Beine et al., 2020; Cattaneo et al., 2017; Triventi and Trivellato, 2012). These institutions do not offer traditional degree programs, often have only PhD-level student populations, and in several cases do not engage in teaching or research activities in a structured way. In addition, three private institutions were excluded from the analysis due to the unavailability of the requisite data ^[2].

The final sample comprises 75 public and private universities observed between 2010 and 2023. Despite the exclusion of some institutions, the sample remains highly representative ^[3], encompassing the largest student population within the system.

We selected the GreenMetric ranking for our analysis because it was the first established sustainability ranking and has had a long-term impact on universities, ensuring a representative data set, while also covering all university missions. The GM data was sourced from its dedicated website ^[4]. The variable Ranking is included in all regressions as a continuous variable, measured as the inverse of the ranking position. This transformation allows for a more intuitive interpretation of the estimated coefficients, as higher values of the variable correspond to better ranking positions. In addition, the measure is rescaled by multiplying by 100 to avoid excessively small coefficients.

To evaluate sustainability performance, we acknowledged its multifaceted nature and the multiple missions of universities. For each institution, we collected data on sustainability-related academic programs, research activities and third mission initiatives. These variables reflect the three core missions of universities – teaching, research and outreach – as defined by Lozano et al. (2015). Importantly, the data were collected from external and verifiable sources, including Scopus, Orbit, university websites and spinoff registries. This approach ensures both objectivity and reproducibility, in contrast with self-reported surveys. Notably, some of these variables, such as the number of sustainability-related courses, publications and spinoffs, are also included in the GM ranking, but in that case, they are self-reported by universities via a questionnaire. Our approach provides an independent validation of these components, while also incorporating additional variables that are not captured in the GreenMetric methodology. Table 1 reports the variables, along with their measures, data sources and summary information.

In our analysis, we included several control variables. Specifically, we controlled for the number of students to account for size, the proportion of STEM faculty as a proxy for scientific orientation, and tuition fees to reflect the market model of universities. We also controlled for inclusion in the THE ranking to capture the excellence or reputation of a university (Jacqmin, 2021). In addition, we accounted for the impact of policies affecting the higher education system by including controls for two events: the establishment of RUS in 2015 and the PNRR in 2021, which provides funding for sustainable initiatives. Definitions of control variables are provided in Table 2.

We estimated our models using a balanced panel data set of 1,050 observations over 14 years. First, we use Granger causality tests to explore the temporal direction of the relationship between sustainability performance and university ranking. This test allows us to test H1a and H1b. Second, we estimate dynamic panel models to account for both persistence over time and potential endogeneity and to estimate the magnitude of these effects. These models enable us to test H2, H3a and H3b.

In the empirical specification, sustainability performance is denoted by $Y$⁠, while ranking is denoted by $X$⁠. However, the two sets of variables are used interchangeably as dependent and independent variables in different model specifications. In detail, to test H1a and H2, we use ranking as the dependent variable and sustainability performance as the independent variable, whereas for H1b, H3a and H3b, we reverse the specification. In all regressions, we included the control variables described in Table 2 (denoted as C), lagged by one period.

We tested the causal relationship between GM ranking and universities’ sustainability performance using the Granger causality test (Granger, 1969). This approach involves creating a regression model that incorporates previous values of one variable to determine if they can help predict the value of another variable. To apply the Granger test, we first tested the stationarity of our variables using the Levin-Lin-Chu test (Levin et al., 2002). If the level variable is nonstationary, the series is differenced and year dummies are included to capture trend variations. Equation (1) presents the general regression model. However, two different specifications are used depending on the nature of the dependent variable: a Tobit panel model when ranking is the dependent variable, given its left-truncated nature and an OLS fixed-effects panel model when sustainability performance is the dependent variable ^[5]. The existence of bidirectional relationships is assessed based on the statistical significance of the lagged explanatory variables in the respective specifications.

In these tests, we included the variables lagged by three periods. We selected a three-year lag structure based on improved model fit (AIC and BIC) and the rule of thumb for dynamic panels (Time ≥ 5 + 3*lags, Thornton and Batten, 1985), balancing empirical performance with sample size constraints:

To analyze the relationship between ranking and sustainability performance, we use dynamic model specifications. This allows us to assess the direction and magnitude of the effect of each variable on the other by examining the estimated coefficients in both specifications. We apply the GMM technique, which is appropriate for data sets with many cross-sections and short time series (Arellano and Bover, 1995). This method uses the lags of the variables as instrumental variables and removes cross-sectional fixed effects. In this research, two-step system GMM was applied, which is asymptotically more effective. Equation (2) presents the dynamic panel regression model, while equation (3) presents the GMM difference model, which also captures all unobservable fixed effects:

Table 3 reports the descriptive statistics for the sustainability performance metrics analyzed. The variables associated with courses, publications, citations, patents and ranking are available from 2010 to 2023, resulting in 1,050 observations. Conversely, information regarding spinoff activities extends only up to 2019. Each variable exhibits a distinct range and set of values, reflecting the different aspects and metrics of sustainability performance analyzed.

Table 4 presents the outcomes of the panel unit root tests, which aim to reject the null hypothesis that a unit root exists. The results indicate that most variables in levels are nonstationary. Nevertheless, using first differences addresses this problem for all variables, except for ranking. Given the significance of this variable in our analysis, we opted to include year dummies to capture the trend.

Table 5 shows the results of the Granger test conducted to examine the causal effect of sustainability performance on ranking position. The results suggest that the number of sustainability courses (Model 1), sustainability publications (Model 2), citations (Model 3) and green patents (Model 5) have a positive causal effect on ranking position, while no evidence was found for sustainability spinoffs (Model 4). These results support H1a, indicating that an enhancement in sustainability performance is associated with a causal improvement in ranking position. The lack of empirical evidence regarding spinoffs can be attributed to the inherent challenges in measuring sustainable spinoffs – which, unlike green patents, lack a specific identification code – and to the limited observation period.

Table 6 presents the results of the Granger test, which examines the causality direction of ranking on performance. In this regression, we also include the variables related to policies (RUS and PNRR). The results show that ranking exerts a causal effect on the number of sustainability publications (Model 2), citations (Model 3) and green patents (Model 5), while no evidence of an effect is found for sustainability courses (Model 1) and spinoffs (Model 4). The causal effect of ranking on sustainability performance allows us to partially accept H1b.

Overall, these results confirm a bidirectional causal relationship between sustainability ranking and sustainability performance, whereby each influences the other. Results are also confirmed by the robustness checks [see  Appendix – Tables A1 and A2 for the causal direction from performance to ranking (H1a), and Tables A3 and A4 for the opposite causal direction (H1b)]. Furthermore, despite being based on the GM ranking, our results can be generalized. Every ranking, even if based on different measures, follows a similar underlying mechanism (Galleli et al., 2022). Moreover, because Italian universities exhibit similar positions across rankings, their behavior in GM is likely to be reflected in other rankings as well.

Table 7 presents the results of the dynamic panel model with ranking as the dependent variable. These models account for endogeneity, allowing interpretation of the coefficients. Model 1 considers the number of sustainability courses as a metric of sustainable performance, whose impact on the ranking is positive and statistically significant (0.045, p-value < 5%). Models 2 and 3, focusing respectively on sustainability publications and citations, show a positive impact on the ranking (0.979, p-value < 5% and 0.015, p-value < 5%, respectively). Models 4 (sustainability spinoffs) and 5 (green patents) report no significant effect on the ranking.

These results provide partial support for H2, which posits that improvements in sustainability performance are associated with higher ranking positions. The findings indicate that advancements in education and research are reflected in ranking scores; in particular, sustainability-related publications significantly influence rankings, consistent with patterns observed in traditional ranking systems (Vernon et al., 2018). Moreover, the number of citations received by sustainability papers is also impacted, despite not being explicitly included in the data set. By relying on externally collected data, we are able to assess whether rankings capture underlying improvements in sustainability performance, thereby providing evidence of their external validity.

Table 8 presents the results of the dynamic panel model with sustainability performance as dependent variables. Model 1 shows that ranking positively affects sustainability courses (0.311, p-value < 10%). Similarly, in Model 2, GM has a positive effect on sustainability publications (12.49, p-value < 10%), while Model 4 finds no significant determinants for sustainability spinoffs. While the results for sustainability spinoffs are not statistically significant, this may partly reflect the restricted timeframe available for this variable. The remaining models, however, provide support for H3a. By contrast, Model 3 shows a strong negative effect on citations (−847.6, p-value < 5%) and Model 5 reveals a negative effect of ranking on green patents (−0.443, p-value < 10%). These findings corroborate H3b^[6].

Interestingly, when considering the impact of sustainability-related policies, the RUS variable shows a positive effect on sustainability courses, publications and citations. This may reflect the fact that the network primarily focuses on the educational and research missions of universities, which are more institutionalized and easier to coordinate across institutions, while paying comparatively less attention to third mission activities, which encompass a broader and more heterogeneous range of initiatives. The PNRR is positively associated with sustainability courses, suggesting that it may have incentivized the creation of new academic programs. Given the recent introduction of PNRR funding, teaching-related outcomes may be the most immediately responsive dimension. It is also worth noting that these policy variables are closely correlated with time trends and may therefore absorb part of broader temporal effects, capturing changes that occurred over the same period beyond the specific policy intervention itself.

Overall, these results suggest that rankings can prompt organizations to adopt reactive conformance measures aimed at improving performance in the areas explicitly included in the ranking, especially when such measures are relatively straightforward (Espeland and Sauder, 2007; Fowles et al., 2016). Concurrently, a higher ranking position is associated with a reduction in sustainability performance in dimensions not captured by the ranking itself. While we are not able to disentangle the relative contribution of the different underlying drivers, it is nonetheless useful to discuss some plausible mechanisms that may help interpret these findings.

One possible explanation relates to the so-called gaming effect (Johnes, 2018): the pressure to improve ranking positions may push researchers to focus on “mainstream” publications that increase productivity, but are less likely to have a real impact and attract citations (Ter Bogt and Scapens, 2012). Similarly, patents are resource-intensive and universities may prefer to reallocate scarce resources to activities with more immediate visibility in the rankings. However, other dynamics may also be at play. Participation in rankings can enhance organizational accountability, making universities aware of their performance (Hamed et al., 2025). In response, institutions may find it easier to prioritize improvements in measured indicators, while progress in other areas may require longer time horizons. Differences across disciplines may further shape these strategic responses, and the effects on sustainability outcomes may take time to materialize. Overall, these findings highlight the need to critically reflect on which indicators are included in rankings (Dill and Soo, 2005), and more broadly on how to assess the impact of universities on sustainability.

To explore the broader implications of sustainability beyond the variables used to test our hypotheses, we conduct a performance analysis to examine whether rankings exert a wider influence on sustainability measures. This section provides supplementary evidence related to H3b; however, it considers only ranked institutions. Our goal is to assess the impact of the ranking not only on sustainability measures related to the three core missions of universities but also on broader aspects of sustainability, including infrastructure, energy, waste production and recycling, water usage and transportation. To achieve this, we relied only on data provided by GM, which collects sustainability-related information from universities. We collected the scores assigned by the GM organization to the following areas: Setting and Infrastructure, Energy and Climate Change, Waste, Water and Transportation over the 2014–2023 period. Table 9 provides the definitions and sources of these variables.

To test the impact of the ranking on these performance measures, we first address potential sample selection issues: because we consider only universities included in the ranking, as the measures are retrieved from it, we implement a two-stage estimation procedure following Heckman (1979). In the first stage, we estimate a probit model that captures the probability of a university being included in the GM ranking by considering the whole sample. This step allows us to model the selection process and to obtain an inverse Mills ratio used to correct for potential selection bias in the second stage. Identification is ensured through an exclusion restriction, whereby at least one variable is included in the selection equation but excluded from the outcome equation. In the second stage, we estimate an OLS regression on the subsample of universities included in the ranking, where the dependent variable is the GreenMetric score in the different areas outlined in Table 9. This specification includes the full set of control variables (excluding the exclusion restriction) and incorporates fixed effects. The inverse Mills ratio derived from the first stage is added as an additional regressor to control for potential selection bias.

Table 10 presents the regression results obtained using the Heckman correction model. In the first stage, we calculate the probability of being ranked in GM, introducing a variable called “imitative behavior.” This variable represents the proportion of universities ranked within the same NUTS 2 region as the university considered in the same year ^[7]. This variable serves as our exclusion variable and has a significant positive impact on the probability of being included in the ranking, confirming its validity. This stage allows us to calculate the IMR, which is included in the second stage to correct for potential bias.

In the second stage, we evaluate the impact of ranking position on sustainability variables at large. The results show a positive impact of GM ranking on these sustainability scores and thus performance. The inclusion of year fixed effects, along with their growing coefficients and impact, confirms that Italian universities entering the ranking are more likely to improve their performance and, consequently, their ranking position.

Control variables suggest that larger universities are better equipped to implement efficiency mechanisms, positively impacting energy, climate change and water consumption. In addition, universities that are also ranked in the THE ranking perform better in these areas, indicating that experienced universities or those with dedicated ranking offices are more effective at collecting relevant data.

Overall, these results show that rankings impact sustainability and can be interpreted in different ways. First, universities that are ranked tend to improve their sustainability performance, thereby confirming isomorphism theory (DiMaggio and Powell, 1983). Second, universities may become more accountable by improving their ability to collect and communicate data, as well as by becoming more aware of their performance (Dal Molin et al., 2017). Finally, ranked universities are more likely to focus on improving the performance metrics included in the rankings, reflecting a reactive response (Espeland and Sauder, 2007). Although not directly linked to a specific hypothesis, these results serve as supplementary evidence that corroborates the impact of rankings in enhancing performance included in the ranking measures.

As outlined in the Introduction, this study offers practical implications for university managers, policy implications for policymakers and external stakeholders, and theoretical implications for future research in this field.

Organizations perceive a favorable ranking position as an indication of a legitimate and credible third-party assessment that can influence stakeholder perceptions (Devinney et al., 2008). In this sense, university managers may view rankings as a form of feedback or an external source of pressure (Andrades et al., 2024; Rindova et al., 2018).

Rankings can also impact an institution’s accountability by driving the collection of more performance data, influencing both the organizational structure and institutional priorities (Hazelkorn, 2008). As a result of the proliferation of rankings, many universities have established formal internal mechanisms to review institutional rankings and their own institution’s position (Hazelkorn, 2008). Because top-ranked universities are externally rewarded by stakeholders, this legitimizes their sustainability activities and influences the strategies of all universities (DiMaggio and Powell, 1983; Jones, 2017).

Recognizing the causal link between sustainability performance and ranking can support strategy development, especially given the public sector’s recent emphasis on performance measurement (Adams et al., 2014; Söderlind and Geschwind, 2019). However, managerial decisions may be driven more by the desire to improve ranking positions (Locke, 2014) rather than by the objective of achieving strategic goals (O’Connell and Saunders, 2013). Finally, our research also highlights potential ranking biases, which may lead some universities to opt out of participation (Clementino and Perkins, 2021).

Beyond university managers, policymakers frequently use university rankings as a proxy for performance to address a lack of information (Dill and Soo, 2005). Because governments are key stakeholders for HEIs, these rankings can significantly influence funding allocation and grant distribution (Hazelkorn, 2014). In addition, governments could use rankings as an alternative to direct regulation to encourage universities to enhance efficiency and performance (Dill and Soo, 2005). This research can illuminate the true sensitivity of rankings to actual performance.

Our findings empirically confirm that if stakeholders and policymakers use rankings as a basis for their choices, it may result in suboptimal decisions (Saisana et al., 2011). Furthermore, if policymakers use university rankings to determine grants, they may inadvertently incentivize universities to manipulate rankings rather than pursue genuine sustainability (Dill and Soo, 2005).

These considerations also open up a reflection regarding the possibility of a public intervention in the development and distribution of rankings (Dill and Soo, 2005). One possible approach is to introduce an independent audit system to certify the data submitted for ranking, which enhances reliability, credibility and comparability across institutions. However, several challenges must be considered. First, the cost of implementing such audits may be too high, particularly for smaller institutions. Second, given the global distribution of participating universities, the feasibility of a centralized international audit system is a major concern.

A more viable alternative would be to involve national quality assurance agencies. For instance, in Italy, ANVUR (the National Agency for the Evaluation of Universities and Research Institutes) is responsible for assessing the quality of universities and research activities through panels of internal and external experts. In this context, ANVUR has already undertaken evaluation activities that incorporate sustainability-related dimensions in selected case studies. However, these sustainability dimensions remain broad in scope and do not always include direct or standardized measures. Therefore, this experience could be further extended and systematized into a more formalized and consistent audit framework for sustainability data. In the first phase, ANVUR could involve selected sustainability experts to define and establish indicators, minimum disclosure requirements and auditing standards. Subsequently, a pilot phase could be implemented on a limited sample of universities to test the feasibility of data collection and the effectiveness of the proposed indicators. Pilot cases could also help contain initial implementation costs. Following the pilot phase, the system could be gradually scaled up. ANVUR could conduct periodic audits (e.g. every two–three years), combining document-based verification with random on-site inspections. This approach would support the establishment of dedicated personnel responsible for sustainability-related data collection and reporting within universities. In addition, to address the global nature of rankings, coordination among national agencies could be promoted through the establishment of international networks, enabling mutual recognition of audit standards and the alignment of measurement approaches. This could enhance both universities’ efforts to comply with ANVUR indicators and the reliability of the data reported to international rankings. As a third step, once the maturity and effectiveness of the indicators have been confirmed, a funding-linked incentive mechanism could be introduced, whereby a portion of public funding is allocated based on performance. This could help promote a virtuous cycle, encouraging universities to further invest in sustainability.

While theoretically appealing, this mechanism could also present some challenges. First, sustainability is a broad concept, and universities may define their own approaches (Böhringer and Jochem, 2007). Second, while this could incentivize positive institutional change, it also introduces the risk of unintended consequences (Andrades et al., 2024). For example, institutions might engage in “greenwashing” practices, focusing on what can be measured while disregarding other equally important sustainability efforts (Alberti et al., 2025; Espeland and Sauder, 2007). To mitigate these effects, it would be advisable to avoid aggregating all sustainability performance into a single composite ranking, because disaggregated indicators would allow more detailed evaluations (Daraio and Bonaccorsi, 2017; Moggi, 2023). Nevertheless, simplification pressures from media and public stakeholders often increase the demand for overall rankings. In addition, high-ranking institutions are interested in preserving these hierarchies. In light of this, we strongly recommend that policymakers exercise caution when using rankings as decision-making tools (Saisana et al., 2011; Soh, 2012). At the same time, we suggest that governments and policymakers promote the systematic collection of reliable, audited sustainability data to inform evidence-based policies and institutional strategies.

Beyond policymakers, this research can also benefit other stakeholders. Despite the value placed on rankings for their ability to synthesize information, this study demonstrates that the data presented in such rankings do not necessarily reflect reality. Conversely, institutions that appear to be more sustainable may merely be engaging in ranking-driven competition.

Our findings contribute to the extant literature by representing the first empirical observation of the causal relationships between sustainability ranking and sustainability performance within universities. To the best of our knowledge, only Atici et al. (2021) and De La Poza et al. (2021) have studied the impact of sustainability rankings on academic performance, while Sierra-García et al. (2024) identified the contribution of university quality to sustainability rankings. Our research adds to the body of knowledge regarding rankings and their impact on HEIs, recommending that future studies examine and potentially include this dual relationship. In addition, we aim to contribute to the ongoing debate about sustainability performance and its assessment.

This paper tested three sets of hypotheses regarding the relationship between sustainability ranking and sustainability performance. First, the results show a bidirectional relationship between sustainability ranking and sustainability performance. Second, the results confirm the validity of GM Ranking in reflecting actual performance, especially for sustainability courses and publications. Third, GM ranking impacts the sustainability performance of universities. It positively affects the variables included in the ranking measure, while negatively affecting the others.

These results support the idea that rankings both reflect and influence performance. However, their influence is different among performance: they react to ranking by increasing measures that can boost the score while neglecting the others (Heinrich, 2002). This may be an example of reactive behavior in response to isomorphism pressure whereby universities seek to align themselves with top-ranked institutions (DiMaggio and Powell, 1983; Espeland and Sauder, 2007; Rindova et al., 2018). However, it may also suggest that universities become accountable in those dimensions, and other sustainability impacts take longer to be effective. Future research could further disentangle the underlying mechanisms by investigating how different allocation strategies and organizational responses affect the relationship between rankings and the multidimensional nature of sustainability performance. However, despite its limitations, this study represents an important first step in uncovering these dynamics.

Although this study can make a valuable contribution to the field, some limitations and future directions should be acknowledged. From a performance perspective, while online-retrieved data are verifiable, they may omit some important features and characteristics of universities. The advent of official or internationally shared metrics will facilitate comparability. In terms of evaluation, we opted to use indicator-based measures, despite the existence of alternative approaches, such as account and narrative assessments; further research could explore this avenue. GM ranking is used, although different indicators may demonstrate varying degrees of effectiveness. Once sufficient data has been collected on other sustainability rankings, it will be possible to examine this relationship, assess the robustness of our findings and explore potential differences in institutional behavior. Finally, our study focuses on a single country, and future research may expand this perspective.

Tables A1 and A2 report the results on the causal impact of sustainability performance on ranking using two alternative specifications. Table A1 presents estimates based on an OLS model instead of a Tobit model, while Table A2 excludes first differences and includes university fixed effects. The findings confirm that sustainability performance has a causal impact on ranking.

Tables A3 and A4 also test the causal impact of ranking on sustainability performance using two alternative specifications. Table A3 removes first differences and includes university fixed effects, whereas Table A4 examines this relationship while excluding the policy variables (RUS and PNRR). These findings corroborate the causal impact of ranking on sustainability performance and, together with the previous tables, support the existence of a bidirectional relationship between ranking and performance.

Table A5 shows a test for the citations and sustainability patents due to make a sensitivity analysis around the negative result. Column (1) and column (2) test different lag of the ranking on the number of citations obtained by sustainability papers. The results remain robust even if considering the ranking variables lagged respectively of two and three periods. Column (3) shows a falsification test for patents by considering as dependent variables nonsustainable patents. The fact that ranking in this case is not significant confirms the validity of our results.