District effects and sustainability: web-based text mining evidence on sustainability signaling in the Italian textile industry

The interest, which has grown in the last decades, in sustainability and/or twin and just transitions in industries and socio-technical systems around the world, has, of course, reverberated in a growing set of studies and research, trying to analyze and understand the organizational, territorial and institutional correlations of such transitions. Recent geopolitical conflicts, tragic wars and large-scale denialism have slowed down attention but have not stopped long-term investments, innovations and studies on sustainability. Within the objects of study, business clusters and industrial districts play a persistent and diffused role.

Districts and clusters have been increasingly recognized as key enablers of sustainability-oriented transformations, especially for small and medium-sized enterprises. Their dense relational structure facilitates the diffusion of environmental and social practices, while collective institutions and shared infrastructures reduce the costs of compliance with formal standards (Cainelli et al., 2015; De Marchi et al., 2013). Recent contributions further highlight how districts and clusters may act as driving forces of sustainability transitions by enabling coordinated adoption of certifications and green innovations at the local level (Hervás-Oliver et al., 2024; Sedita and Maghssudipour, 2024; Bellandi and Stark, 2025, Ermini et al., 2024). In addition, the evolution of districts and clusters toward more knowledge-intensive and innovation-driven systems reinforces their role in sustainability adoption. The integration of digital technologies and Industry 4.0 solutions enhances firms’ ability to monitor, measure and implement sustainability practices (Bellandi and De Propris, 2021), also strengthening the link between local embeddedness and formal certifications (Bettiol et al., 2021).

We defer to the Introduction of this S.I. for a broader discussion of the topic and key references to a related stream of past S.I., collected books and individual papers. Nonetheless, it is necessary here to recall a few points that help locate the specific contribution we propose in this paper:

District effect: we propose and comment on a quantitative-based analysis about the performance of firms operating either within or outside industrial districts, after specific delimitations of the units of investigation (see point b). This qualifies our paper as a (would-be) contribution to the stream of quantitative studies on the “district effect.” This stream emerged in Italy in the 1990s with first applications of econometric models to large business databases, finding that, controlling for their respective sectors (defined as aggregations of Nomenclature statistique des activités économiques dans la Communauté européenne (NACE) 2-levels divisions), Italian manufacturing small-to-medium sized firms tended to show higher performances on various dimensions (such as rates of returns, export intensity, skill rewards) if localized within industrial districts than outside them (Becattini and Musotti, 2003; de Blasio et al, 2009). Applications have thereafter extended to different countries and other measures of performance, such as innovation intensity, including “sustainability innovation” (Hervás-Oliver et al., 2024) as recalled above, that is, the field to which we apply our analysis. More recently, considering the high degree of heterogeneity of firms (and complementary factors), within and outside industrial districts (see point c), methods of comparison after matching (of similar firms across sectoral and size dimensions defined at a quite high degree of granularity) have started to be applied. Indeed, lack of attention to the impacts of heterogeneity might also explain divergent findings exposed in the literature (Hervás-Oliver et al., 2024). We precisely adopted a formal method of score matching, following a similar analysis proposed by Sedita and Maghssudipour (2024) in the field of sustainability innovation in Italian fashion firms, with some variations both in methodological details and in the object of analysis and research questions.

Units of investigation: the first investigations of the district effect concerned firms within and outside the industrial districts, as conceptually defined by Becattini (see Becattini and Musotti, 2003) and statistically identified by Istituto Nazionale di Statistica (ISTAT) (Sforzi, 2015). They are local systems, that is, travel-to-work areas including sets of contiguous municipalities (NUTS4), usually at or below the province/county level (NUTS3) with a manufacturing identity pivoting on a main sectoral specialization (usually at the NACE/ATECO 2 digits) and on a comparatively large employment in small and medium-sized enterprises (SMEs). When the data available for specific research at the level of local systems (or from municipalities, that is, NUTS4) have not allowed the adoption of this unit of investigation, it has been considered, as a possible proxy, the first upper level of available territorialized statistics, for example, in Italy, the provinces (NUTS3), with a distinction between provinces featuring a high presence of industrial districts and those that are not (Becattini and Coltorti, 2006). Following instead the cluster’s tradition of studies (Porter, 1998), the direct units of investigation are localized industries, that is, sets of interrelated business and complementary organizations operating in a delimited geographical area and showing a main sectoral specialization (Ketels, 2013). In the statistical analysis presented here, we adopt a combination of these two traditions, taking advantage of research done for Italy by Intesa San Paolo (2024, 2025). Specifically, we look at Italian provinces that both show a high or relatively high intensity of industrial districts and include the main Italian industrial districts with a textile specialization as defined by ISTAT; for simplicity, we refer to those areas as “textile districts”; finally, we distinguish and compare textile firms (and subcategories) within and outside those areas.

Heterogeneity: the literature has acknowledged for a long time the necessary heterogeneity of business forms and related organizations within single districts or clusters and across them (Hervas-Oliver and Boix, 2022; Hervás-Oliver et al., 2024). And, of course, in comparisons of firms within and outside districts or clusters the question of heterogeneity should not be ignored, because performances depend not only on the localization of firms but also on other characteristics of structure and strategy and a homogenous or random distribution of such other characteristics cannot be presumed. Indeed, it is a defining feature that the sectoral specializations of Italian industrial districts show a proportion of SMEs higher than the national average in the same sectors and a lower comparative presence of large firms (or plants) (Becattini and Coltorti, 2006). On the other hand, micro firms tend to be more diffuse outside industrial districts. Our database excludes micro firms and the method of matching considers both the size of firms (small, medium-sized or large) and subsectors within the textile specialization.

Under such premises, our research points out the assessment of district effects in the field of sustainability, that is, more precisely, a comparison in measures of sustainability within and outside Italian “textile districts” at the end of the first quarter of the century (2025). In what follows, we start with a section where two connected research questions are translated into hypotheses to be tested (Section 2); a subsequent illustration concerns the methodology and the data set (Section 3); after that, the usual sequence of results (Section 4), discussion (Section 5) and conclusions (Section 6) follows.

We look at a stream of studies that try to get measures of sustainability performances (by firms in different contexts) from business communication on sustainability (Bettiol et al., 2024; Sedita and Maghssudipour, 2024). The recent diffusion of communication via business websites and the increasing use of methods for scraping and analyzing their contents allow researchers to retrieve large amounts of data on these subjects from firms’ websites.

Communication on sustainability may have different goals and contents. The firms usually aim to signify to stakeholders their commitment to one or more dimensions of sustainability. Sometimes communication is associated with more concrete signals, such as certificates and/or social reporting and/or governance adaptations (e.g. a B. Corp. status). Overstating sustainability efforts (greenwashing) or under-communication of actual practices depend on various incentives and the institutional context (Bansal et al., 2015).

The first research question (RQ1) concerns, therefore, a confirmation, on a broad sample of textiles firms, of what is suggested by previous research on fashion firms in Italy (Bettiol et al., 2024; Sedita and Maghssudipour, 2024), that is, that communication on sustainability tends to be stronger within districts than outside (a district effect). Indeed, it is not only a question of confirmation, because, among the ATECO/NACE 2 digit sectors of fashion, that of textiles is the only one with a large dominance of business to business (B2B) markets for their products, and, of course, communication on sustainability could be weaker in general or in any case different, from that of B2C sectors trying to push images and beliefs directly on the final consumers (Bettiol et al., 2024). May such district effects also appear when focusing precisely on a B2B sector? On the supply side, district firms tend in any case to benefit from a larger local availability of specific business services and this may extend to services supporting sustainability practices and communication. On the demand side, communication on sustainability is a tool of nonprice competition on quality in which leading firms of Italian textile districts are also specialized, in consideration of the extended responsibility of their B2C customers. Finally, sustainability disclosure varies substantially across dimensions: environmental communication tends to be more standardized, whereas social disclosure is more context-dependent and shaped by firm-specific stakeholder relationships (Ioannou and Serafeim, 2015; Christensen et al., 2021). On this basis and referring to the qualifications recalled in the Introduction, we propose three connected hypotheses of investigation, H1 more straightforward, H2a and H2b more tentative:

A second research question (RQ2) concerns the consistency between density of communication and signs of real commitment. The same contributions referred to above show that the district effect seems to extend to consistency or alignment. Here, together with a comparison on consistency performance, we try to delineate and compare also the relative presence of conditions of (seemingly) weak consistency. An extended quantitative analysis on these combinations is, per se, quite novel, as far as we know. Furthermore, it provides further suggestions on the strength and nature of the district effect in this field. We maintain that industrial districts provide a distinctive context also in this respect, given that trust-based interactions and reputational mechanisms tend to reduce opportunistic greenwashing (Hervás-Oliver et al., 2024; Bellandi and Stark, 2025). However, while certifications serve as key signals for external stakeholders, communication strategies may remain more selective, considering that, in local transactions, dense communication on the website as a signaling device can be replaced by dense local networks of social and business relations (Bettiol et al., 2021). This can be translated into a specific hypothesis to be tested:

H3. Textile firms located in textile districts exhibit a higher alignment between sustainability practices and communication, and, in firms with a lower alignment, a greater prevalence of under-communication compared to nondistrict textile firms.

Our initial sample included 10479 Italian firms selected from the AIDA database (available at: Link to the cited article.), based on their ATECO code. An ATECO code is the official classification system, maintained by ISTAT. The ATECO system is hierarchical and provides a detailed classification down to six digits (available at: Link to the cited article.). The sample was created by selecting firms within ATECO Division 13 – Textile Manufacturing, which includes four (three-digits) subsectors: 13.1, preparation and spinning of textile fibers; 13.2, weaving; 13.3, finishing of textiles, clothing articles and similar activities; and 13.9 manufacturing of other textiles. AIDA collects information on all the firms operating in Italy with a legal form of limited company (therefore excluding a great part of micro firms) and provides each company’s tax identification number and website (when available), along with comprehensive financial and corporate information, covering up to the ten most recent fiscal years. From the initial sample, we filtered out firms without a website and those inactive (Bankrupted/Ceased/In liquidation), reaching a sample of 2,757 active firms with a URL on AIDA.

To identify textile firms located in textile districts, we relied on the reports produced by Intesa Sanpaolo (2024, 2025). These analyses, based on the proprietary Intesa Sanpaolo Integrated Database, identify the Italian industrial districts focusing on their main sectoral specializations based on clusters of SMEs and classify provinces according to their level of district intensity. Such metric captures territories characterized by strong sectoral specialization, high export orientation, specialized labor markets and integrated supply chains.

By intersecting provinces with high or medium district intensity with the main Italian textile districts (Biella Textile District, the Prato Textile and Apparel District, the Gallarate Apparel-Textile District, the Val Seriana Textile and Apparel District and the Como Silk-Textile District), we identified five relevant provinces: Biella, Prato, Varese, Bergamo and Como. We call them, for simplicity, “textile districts.” Finally, we classified all textile firms located within these areas as (textile) district firms.

Textual data were collected through large-scale web scraping of firms’ official websites, whose URLs were retrieved from AIDA. The identification of website pages to be scraped followed a two-stage procedure:

1. comprehensive extraction of all hyperlinks available on each website; and

2. semantic analysis and classification of relevant links into the sections Home, About, Sustainability and Certifications using a large language model.

Textual content from these sections, typically containing firms’ self-representations and sustainability narratives, was retrieved. The scraping process was iterative to maximize coverage and quality, with the final extraction in November 2025. Because sustainability certifications are often displayed as logos, we complemented text searches with an image-based identification using embeddings. Collected data were merged with the initial data set using tax identification numbers. Firms with inaccessible websites or missing content were excluded, yielding a final corpus of N = 2568 firms, of which N = 963 (37.5%) are in a textile district. Smaller firms were overrepresented among those without a valid website and the sample excluded most micro firms.

A custom dictionary was developed to capture sustainability-related content. The research team iteratively refined it to align with established sustainability dimensions. The initial set of terms was derived from recognized frameworks (e.g. environmental, social, governance (ESG) standards, global reporting initiative guidelines) and prior literature on sustainability disclosure, then adapted to the context of Italian textile firms. Technical terms specific to textile production were drawn from academic studies (Bressanelli et al., 2022; Saccani et al., 2023) and practitioner-oriented technical documentation (Menci et al., 2007; Miraglino et al., 2007).

The dictionary comprises three sections: environmental sustainability, social sustainability and sustainability certifications. Each section includes keywords (Italian/English), conceptual labels grouping synonyms and broader thematic areas. The final dictionary contains 111 environmental, 57 social and 64 certification labels, organized in 13 thematic areas. Table 1 shows an extract from the dictionary.

Sustainability communication was measured using natural language processing (NLP) techniques applied to the textual corpus. Specifically, for each company, three indicators were constructed by aggregating information extracted from its website content according to the three sections of our dictionary: environmental sustainability index (ESI), social sustainability index (SSI) and certification count index (CCI) (see Section A.1 in Supplementary Material A). CCI was computed as the simple count of sustainability certification occurrences (i.e. the number of certifications mentioned on the company’s website). ESI and SSI, instead, were calculated by summing the binary term frequency/inverse document frequency (TF-IDF) weights (Aizawa, 2003) of all environmental- and social-sustainability-related terms appearing on the website. In this way, companies with a larger number of unique or uncommon sustainability-related terms on their website obtained higher ESI and SSI values. The use of TF-IDF related scores is consistent with recent studies (Sedita and Maghssudipour, 2024).

In this study, we maintain (recall H1, H2a, H2b and H3 in Section 2 above) that industrial districts represent distinctive socio-economic environments impacting on firms’ strategies and narratives and thus we test whether belonging to a district supports higher values of ESI, SSI and CCI. Addressing potential confounding due to heterogeneity of firm characteristics (see the Introduction) is essential. To this end, we relied on a propensity score matching (PSM) methodology (see Section A.3 in Supplementary Material A). Coherently with Sedita and Maghssudipour (2024), district membership is treated as a structural and time-invariant attribute and matching with firms outside districts is conditional on this. In line with standard practice (Rosenbaum and Rubin, 1983; Stuart, 2010), matching is implemented by using a nearest-neighbors method, whereby each district firm is paired with non-district firms displaying the highest degree of similarity in terms of other observable characteristics; specifically, in our study, the 2023 revenues and the number of employees as covariates and controlling for the ATECO industry codes at different levels of granularity (two, three and six-digit). Business size variables are well-established determinants of sustainability performance and communication (Caliendo and Kopeinig, 2008; Barzotto and Mariotti, 2018). To assess robustness and balance the trade-off between match quality and sample size, we implemented alternative matching ratios (1:1, 1:2, 1:3). Overall, this strategy ensures well-balanced samples of district and nondistrict firms and provides a robust basis for estimating the district effect on sustainability communication.

After we matched each textile district firm with similar textile nondistrict firms, we measured the effect of localization in a textile district by calculating the average treatment effect on the treated (ATT, τ) (see Section A.4 in Supplementary Material A). This allows comparing the average value of CCI, ESI and SSI for district firms with the average value for the similar nondistrict firms they were matched with. Thus, τ indicates, on average, how much higher (or lower) textile district firms score on these sustainability indicators because of their localization within a textile district. To assess the robustness of the estimated treatment effects to potential unobserved heterogeneity, a sensitivity analysis based on Rosenbaum bounds was conducted (Rosenbaum, 2002).

The keyword extraction from firms’ websites provides an initial overview of how sustainability is communicated across the sample. Among the 2,568 firms analyzed, 53.7% show at least one environmental keyword, although the overall intensity of environmental communication remains modest, with firms reporting fewer than two such keywords on average (M = 1.86, SD = 2.647). Similarly, 47.3% of firms mention at least one social keyword and the average presence of these keywords is even more limited (M = 0.627, SD = 0.77), with lower variability across firms. When distinguishing between district and non-district firms, the prevalence of sustainability communication is consistently higher among the former. District firms more frequently refer to environmental (57.11%) and social (50.15%) content than non-district firms (whose corresponding percentages are 51.6% and 45.5%, respectively). Furthermore, only 30% of firms report at least one sustainability certification, suggesting that formal, externally verified commitments are considerably less common than more generic sustainability claims. Table 2 shows the descriptive statistics of ESI, SSI, CCI and of the covariates used in the analysis.

These patterns are consistent with the distribution of sustainability indicators. Environmental communication, captured by the ESI, shows relatively low average values and large variability (M = 4.289, SD = 7.643), this is because most firms disclose limited environmental information, while a smaller subset reports substantially higher levels. Social communication, captured by SSI, appears more limited and homogeneous (M = 1.052, SD = 1.997). Certification activity, measured by CCI, remains relatively low on average and highly dispersed (M = 0.899, SD = 1.997), confirming that only a minority of firms adopt formalized sustainability standards.

A comparison between district and nondistrict firms reveals systematically higher levels of sustainability engagement among the former. District firms exhibit higher average values across all indicators, suggesting, on average, a more developed approach to both sustainability communication and certification. Similar differences appear in revenues and the number of employees, whose levels are on average higher in district firms.

Overall, sustainability communication and certification practices remain highly uneven. Many firms show minimal engagement and only a minority drive the upper end of the distribution. As anticipated above, PSM is employed to control for the impact of heterogeneity in the structural characteristics of the firms and to more accurately assess the district effect. Indeed, we see in Table 3, considering correlations calculated with a nonparametric Spearman’s ρ rank, that all sustainability indicators are positively associated with firm size, with correlations ranging from weak (ρ ≤ 0.3) to moderate (0.3 < ρ ≤ 0.5). Certifications show the strongest size dependence, while social sustainability communication displays the weakest association.

Covariate balance tests indicate that the matching procedure substantially improves comparability between treated (district) and control (nondistrict) firms. Prematching differences in key variables (revenues and number of employees) are reduced to below conventional thresholds after matching. The two-digit ATECO specification achieves the best trade-off between balance and sample size, ensuring full common support, while the three-digit specification yields similar results with a smaller matched sample and is used as a robustness check. More granular matching (six-digit) does not improve balance and significantly reduces the sample size; therefore, it is not retained. Overall, the satisfactory postmatching balance supports a causal interpretation of the estimated treatment effects under the standard selection-on-observables assumption. Detailed results are reported in Section A.2 of Supplementary Material A and Table B.I in Supplementary Material B.

In this study, treated units are textile firms located in textile districts, while control units are textile firms operating outside textile districts (see definitions in Sections 1 and 2 above). Accordingly, a positive and statistically significant ATT τ can be interpreted as a district effect.

Figure 1 reports the estimated values of τ for CCI, ESI and SSI across alternative matching specifications, varying both the matching ratio (1:1, 1:2) and the level of industry granularity (ATECO two- and three-digit). Results using 1:3 matching (not reported for brevity and available in Table B.II in Supplementary Material B) are fully consistent with those obtained under the 1:2 specification. In the figure, the diameter of each circle is proportional to τ, and the color indicates the significance level at which τ can be considered greater than zero. Overall, the results show a clear district effect on CCI, while evidence for ESI and SSI is more heterogeneous and sensitive to specification choices. Table B.II (Supplementary Material B) shows in detail the estimated values of τ, standard errors, p-values and number of treated and control firms for CCI, ESI and SSI across all matching specifications.

Starting with the CCI, the estimated τ is positive, large in magnitude and highly statistically significant across all specifications. In the baseline specification (ATECO two-digit, 1:1 matching), τ amounts to 0.426 (p < 0.01) and remains stable when increasing the number of matched controls to 1:2. Moving to more restrictive matching criteria (from ATECO two-digits to three-digits), the estimated effect decreases moderately but remains meaningful and statistically significant. This consistent pattern indicates that district firms are significantly more certified than their matched counterparts.

For the ESI, τ is positive in all specifications, but in the baseline 1:1 match, the effect is not statistically distinguishable from zero (τ = 0.602, p > 0.10). This likely reflects high variance, which reduces statistical power. Increasing the matching ratio to 1:2 expands the pool of comparable controls, lowers estimator variance and thereby increases power. As a result, the estimated effects become larger and statistically significant (τ ≈ 0.8, p < 0.05), even though the underlying point estimates remain broadly stable. A similar result appears with stricter industry constraints (though matching at the ATECO three-digit levels reduces the availability of suitable controls and thus lowers power, increasing the matching ratio compensates for this loss). Overall, the results suggest that district localization is associated with higher environmental-related textual communication, although detecting this effect requires sufficient statistical power due to inherent noise and high variability of text-based measures.

Finally, for the SSI, the results are weaker and more sensitive to specification choices. Under ATECO two-digit matching, the estimated effects are small and not statistically significant. However, moving to the more granular three-digit classification increases both the magnitude and significance of τ (reaching approximately 0.20–0.22), particularly under 1:2 matching. This pattern suggests the presence of within-industry heterogeneity. Social sustainability communication is likely more context-specific and varies across textile subsectors, so coarser classifications may obscure meaningful differences. In contrast, finer industry matching improves comparability, making the district effect more detectable.

Sensitivity analysis based on Rosenbaum bounds (reported in Supplementary Material A, Section A.5.) indicates that the estimated effects are robust to moderate levels of hidden bias for CCI and more sensitive for ESI.

To examine the relationship between firms’ substantive sustainability practices (proxied by CCI) and their sustainability communication strategies (captured by the sum of the ESI and SSI) in district and nondistrict firms, we use heatmaps, shown in Figure 2.

The analysis is conducted by splitting district firms (left panel) and nondistrict ones (right panel). The horizontal axis reports the CCI, the vertical axis the combined sustainability communication index. Each cell represents a group of firms sharing values of the two indicators within the same specified range; cell color reflects the share of firms (%) in the group; this percentage is also displayed in each cell.

The results show a markedly skewed distribution of firms toward the lowest levels of both certification intensity (CCI) and communication (ESI + SSI), but with important differences between district and nondistrict firms. In both panels, many firms show both values equal to zero. Notably, these firms are 251 in districts (26.1%) and 454 outside districts (28.3%), accounting for 27.45% of the total sample. Moreover, in both panels, the highest concentration of firms is in the bottom-left corner, particularly in the (CCI = 0 or 1–2 and ESI + SSI = 1–20) cells, confirming that a large share of firms exhibits minimal certification activity and very limited sustainability communication. However, the concentration at the lowest cells is more pronounced among nondistrict firms, suggesting that a large majority of firms remain weakly engaged in sustainability, both in terms of formal certifications and public disclosure.

District firms display a more dispersed pattern. Although the number of firms in the upper-right region (high CCI and high ESI + SSI) remains limited, district firms exhibit a slightly thicker upper tail, indicating a greater presence of firms more engaged in sustainability-related initiatives. Rather than reflecting a simple increase in either certifications or disclosure alone, this result suggests that district contexts facilitate the joint adoption of both dimensions.

An interesting pattern emerges also for misaligned strategies. In fact, the heatmaps reveal the presence of firms characterized by a mismatch between certification intensity and communication. Observations in the upper-left area (low CCI and high ESI + SSI) may be indicative of greenwashing, where firms engage in relatively high levels of sustainability communication without a corresponding level of certified practices. In this area, the two heatmaps show similar and low values, suggesting that overstating sustainability performance through communication alone is not systematically driven by district membership. Conversely, firms located in the lower-right area (high CCI and low ESI + SSI) adopt substantive sustainability practices but communicate them only to a limited extent (under-communication). This area is more populated within districts. This asymmetry suggests that district firms are more likely to undertake substantive sustainability actions without proportionally communicating with them.

Overall, the results support the view that industrial districts act as coordination mechanisms that not only influence the level of sustainability engagement but also shape the alignment between actions and communication.

This study provides novel evidence on how industrial districts shape firms’ sustainability strategies, distinguishing between substantive practices (certifications) and environmental and social communication. The results offer support for the proposed hypotheses, highlighting some important aspects.

First, the findings provide strong support for H1. District firms exhibit systematically higher levels of certification intensity (proxied by CCI), with positive and robust treatment effects across all matching specifications. This result is consistent with the literature on clusters and districts, which emphasizes how embedded firms benefit from knowledge proximity, shared norms and peer monitoring, which lower the costs of adopting formal standards (Porter, 1998; Becattini and Coltorti, 2006; Hervas-Oliver and Boix, 2022). In line with more recent contributions, our results confirm that such contexts facilitate the diffusion of sustainability practices and certifications, specifically among SMEs embedded in dense sectoral networks (De Marchi et al., 2013; Cainelli et al., 2015).

Second, the results provide support for H2a and H2b, as well. Environmental sustainability communication (proxied by ESI) is positively associated with district localization (H2a), even if the statistical significance of this relationship depends on estimation efficiency and is not consistently robust across all specifications. Social sustainability communication (proxied by SSI) shows a significant district effect, but only when more granular industry classifications (three-digit ATECO) are employed. This result highlights the critical role of within-industry heterogeneity and confirms that coarse industry controls may obscure relevant variation. It is also consistent with prior literature showing that social disclosure is less standardized and more context-dependent than environmental reporting (Ioannou and Serafeim, 2015; Christensen et al., 2021).

Third, the analysis of alignment between sustainability practices and communication also provides support for H3. District firms are more likely to exhibit aligned strategies, but often under-communicate their sustainability endeavor, while no substantial differences emerge in terms of greenwashing. This suggests a tendency among district firms toward forms of under-communication, despite the presence of substantive sustainability actions, which might signal greenhushing. Greenhushing refers to the practice of downplaying or avoiding communication about sustainability initiatives to reduce exposure to external scrutiny, either from stakeholders with very rigid expectations regarding sustainability claims (Bettiol et al., 2024) or from audiences skeptical of sustainability itself. In our analysis of textile companies’ websites, we identified firms that are certified and therefore presumably engaged in sustainability practices, but that provide very limited communication about such initiatives. However, this pattern should not necessarily be interpreted as a deliberate concealment strategy. In many cases, it may simply reflect a generally low propensity toward communication. Several of these firms are small companies with limited resources for communication activities, often resulting in rather basic websites. Moreover, in B2B markets, the distance from final consumers reduces the need for strong sustainability narratives. This tendency is likely reinforced within industrial districts, where firms are embedded in long-standing local supply-chain relationships and trust-based interactions, making formal communication less central than informal reputation mechanisms.

This study contributes to the literature on industrial districts and sustainability by disentangling two dimensions that are often implicitly conflated: substantive sustainability practices and their public communication to explore how territorial embeddedness shapes firms’ sustainability strategies.

From a theoretical perspective, the findings offer three main contributions. First, they confirm the solidity of an extension of the notion of the district effect beyond traditional performance measures such as productivity and innovation, to include how firms structure and signal sustainability, even in B2B sectors. Second, the results suggest that in district contexts, together with a comparatively (with respect to nondistrict contexts) larger set of firms leading aligned practices and communication on sustainability, there are sets of firms that “do” more than they “say.” This is consistent with the emerging literature on selective disclosure of sustainability, highlighting the limits of disclosure-based measures and the potential divergence between symbolic and substantive actions. Third, and relatedly, the study contributes to the same literature by showing that to invest in sustainability practices and communicate weakly on this is not necessarily a strategic or opportunistic behavior, the so-called greenhushing, but it can be an outcome of relational governance systems, where trust, repeated interactions and informal reputation partially substitute for formal signaling.

These insights have important implications for how we interpret ESG indicators and sustainability metrics. In network-based production systems such as clusters and districts, communication-based measures risk introducing a systematic measurement bias, underestimating actual sustainability efforts, especially in B2B sectors. Such concern echoes a growing body of literature questioning the validity and comparability of ESG indicators across institutional contexts (Christensen et al., 2021; Grewal et al., 2019). Our results suggest that part of this bias may be structurally linked to the territorial and relational embeddedness of firms.

At the firm level, a couple of main takeaways emerge. First, even when justified by the local context, under-investment in communication on sustainability goes together with over-embeddedness in the same context, because formal disclosure remains a key evaluation tool for sustainability-oriented investors and global stakeholders (Grewal et al., 2019; Ioannou and Serafeim, 2015). Specifically, enhancing the alignment between practices and communication could represent a broader source of competitive advantage in nonlocal value chains where certifications and transparency increasingly act as entry requirements (De Marchi et al., 2013; Bettiol et al., 2021). Second, we saw that, within our core set of textile firms in Italy, that is, those with a legal status of limited company and with an active website, only a minority show strong signs of alignment in formalized practices (namely, certifications) and focused web communication on sustainability. Of course, to evaluate correctly the meaning of this seemingly small activation, cross-sectoral and cross-national comparisons would be necessary. However, probably, the fact that a large part of Italian textile firms does not show clear sustainability strategies cannot be explained only by the B2B nature of the sector. It is surely also a question of the prevalence of smaller-sized firms within the same core. It could also be a question of an evolving path, where activation on the frontier of sustainability practices and communication in B2B sectors featuring quality nonprice competition is a quite recent and nonconsolidated approach and more structured firms are leading the way to activation of broader sets of less structured firms with a lower level of readiness.

In that case, coming back to the system-level, given that in our analysis such leading firms appear to be comparatively more present in districts than outside, and that, as acknowledged by the literature, processes of diffusion of competitive practices and strategies tend to be stronger within districts than outside, it might be expected that the district effect evidenced in our analysis will not evaporate soon, confirming that industrial districts and their clusters of SMEs might effectively play an important role in sustainability transitions in fashion systems (Sedita and Maghssudipour, 2024; Bellandi and Stark, 2025).

From a policy perspective, the extension of the district effect, which our analysis confirms, suggests that industrial districts might be considered privileged arenas for testing and promoting policies of support to sustainability practices and communication models, even in manufacturing sectors such as textiles, where firms are not directly exposed to final consumers. At the same time, as recalled just above, the limited number of strongly active firms, especially among the smaller ones, raises an important question: whether this reflects an evolutionary process, currently driven by leading firms embedded in favorable district contexts (Bettiol et al., 2024) or rather a stagnating pattern in which only a restricted group of firms is keeping pace with sustainability transitions, that is, a strong form of heterogeneity (Hervás-Oliver et al., 2024). Policy interventions should therefore take care of the worst scenario and not only reward frontrunners, but also strengthen system-based, multi-actor and multi-level initiatives for a more direct and broader activation of smaller firms in practices and communication under a perspective of sustainability transitions in the fashion systems, such as provision of collective services for certification, training and technical assistance, platforms of interaction between manufacturing firms and knowledge intensive services, included universities and strategies of orchestration by sustainability-oriented lead firms with smaller suppliers (De Marchi et al., 2013; Bellandi and De Propris, 2021; Ermini et al., 2024). Furthermore, it is confirmed that a “one-size-fits-all” approach appears suboptimal when applied not only to firms of different sizes but also precisely to their localization in different contexts. Policies that rely exclusively on standardized disclosure requirements may fail to capture the diversity of governance mechanisms across contexts. For example, firms operating in industrial districts can rely on informal and locally embedded channels of accountability, which are not easily observable through formal reporting (Bettiol et al., 2021). This implies the need for more context-sensitive policy frameworks, combining standardized metrics with complementary tools capable of capturing relational and network-based forms of coordination. At the same time, recent regulatory trends pushing toward greater transparency (e.g. EU sustainability reporting frameworks) should consider the risk of penalizing firms embedded in contexts where disclosure is not the primary signaling device for some of their typical business components (Flammer, 2021).

Finally, this study opens several opportunities for future research. Further work could investigate the micro-foundations of greenhushing, exploring how firm-level characteristics, supply chain positioning and local institutional arrangements interact in shaping disclosure choices. Expanding the analysis across sectors and countries would help assess the generalizability of the results and disentangle district effects from broader patterns and go deeper into the relation between district effects and the power and role of large firms. Moreover, future studies could develop integrated measurement frameworks that combine certifications, communication and performance outcomes, providing a more comprehensive assessment of firms’ sustainability behavior.

The authors warmly thank Contesti srl (www.contesti.info/it/), for conducting the web scraping and providing the scraped text.

The supplementary material for this article can be found online.