The increasingly dynamic, competitive, and rapidly changing environment (Teece et al., 1997) requires firms to learn to build new practices. When developing its absorptive capacity (Zahra & George, 2002), the firm has more significant possibilities to acquire external information, assimilate it, transform it, and use it (Behnam et al., 2018) by applying it in its processes. In addition to improving their operations, firms that use external knowledge can develop skills and practices (Zhang et al., 2015). Such improvement occurs because the knowledge base influences how the firm recognizes and uses new knowledge (Zahra & George, 2002).

Although the use of absorptive capacity to understand sustainability practices is relatively recent (Garay et al., 2017), studies indicate that ACAP facilitates the adoption of successful sustainability strategies (Liu et al., 2019) because firms need to combine information from various sources, usually external to them (Ferro et al., 2019). It is observed that the adoption of sustainability practices is affected, in some way, by access to external knowledge (Liu et al., 2019). In this sense, keeping in touch with engaging internet channels, magazines, and even newspapers dealing with sustainability can help acquire relevant information (Garay et al., 2017).

ACAP can assist firms in adopting sustainability practices (Dzhengiz & Niesten, 2020), and new learning and practices aimed at sustainability can act as facilitators to generate competitive advantage and achieve better performance (Aboelmaged & Hashem, 2019). Thus, it is challenging to imitate ACAP and sustainability initiatives since they are related to complex processes and tacit knowledge. Therefore, the competitive advantage generated by such factors will be difficult to be copied by competitors (Garay et al., 2017).

Small and medium-sized firms can, based on ACAP, boost their sustainable innovation practices, needing to continually improve their technological and organizational capabilities (Brasil et al., 2016; Widya-Hasuti et al., 2018). Firms are challenged to consider economic, social, and environmental sustainability, also based on the success achieved in sustainable innovation, which, in turn, is significantly influenced by ACAP (Widya-Hasuti et al., 2018). In this sense, sustainability itself has been considered a form of innovation (Hjalager, 1997), and using several external sources of information is relevant for firms to innovate (Chen et al., 2015).

Firms must develop and improve their ability to acquire, assimilate, transform, and use external knowledge (Zahra & George, 2002), as this ability impacts the development of their operations and practices (Delmas et al., 2011). Those with high ACAP, when adopting economic, social, and environmental sustainability practices, tend to attract advantages, such as cost reduction (Garay et al., 2017).

The economic, social, and environmental sustainability dimensions are still not fully considered, providing the possibility of deepening sustainability practices in these dimensions, based on knowledge, as one of the most critical resources for the firm (Segarra-Oña et al., 2016). The adoption of environmental (Delmas et al., 2011) economic and social strategies (Pinkse et al., 2010) will be facilitated the more the firm develops its ability to identify, assimilate, transform, and use specific environmental information (Riikkinen et al., 2017). Firms that seek to develop eco-efficient operations have, in the absorptive capacity, assistance to identify that knowledge that is most important (Effiong & Singhal, 2014).

The development of new practices that pay attention to sustainability aspects is related to the generation of ideas and innovative ways of carrying out organizational tasks. Knowing what the internal needs are, and keeping an eye on the external environment (Dzhengiz & Niesten, 2020) to identify important information, assimilating them to integrate with internal knowledge, allows the firm to generate sustainable operations (Albort-Morant, Leal-Rodríguez et al., 2018). The motivation of managers of small and medium-sized firms to adopt sustainability practices is linked to channels and information sources about sustainability, and the perceived usefulness of the information acquired (Garay et al., 2017).

TBL has become the most widely used theoretical understanding to explain sustainability in the business world (Milne & Gray, 2013), and researchers are using the concept of ACAP to explain sustainability strategies in firms (Liu et al., 2019). What still needs to be tested is the effect of ACAP on economic, social, and environmental sustainability practices. It is still necessary to analyze and deepen sustainability practices (Garay et al., 2017) to explain the influence of absorptive capacity on its adoption (Dzhengiz & Niesten, 2020).

Turning attention to economic sustainability practices, Garay et al. (2017) indicate that ACAP can be considered a predecessor, acting as a mechanism to explain practices aimed at economic sustainability. ACAP allows firms to reduce costs (Delmas et al., 2011) and obtain better results than their competitors (Garay et al., 2017). Thus, the first research hypothesis was formulated:

H1: Absorptive capacity positively and directly influences economic sustainability practices.

Compared to environmental and economic sustainability practices, social sustainability has been little researched (Engert et al., 2016), perhaps because it is less tangible (Branco & Rodrigues, 2006). It appears that ACAP can generate social sustainability practices (Riikkinen et al., 2017). The absorptive capacity helps the firm turn its gaze and attention to issues involving society (Shahzad et al., 2020), enabling the implementation of practices aimed at this audience (Ling, 2013). In knowledge-intensive organizations, managers understand the role of ACAP in influencing social sustainability actions (Crilly & Sloan, 2012). Therefore, the second hypothesis is presented:

H2: Absorptive capacity positively and directly influences social sustainability practices.

Consumers’ increasing attention to environmental issues is leading firms to integrate environmental practices into their operations (Borland et al., 2016). Considering environmental practices as a matter of competitive advantage, firms are investing in the identification and development of capabilities focused on the area (Hesselbarth & Shaltegger, 2014). There are indications that the absorptive capacity generates environmental sustainability (Dzhengiz & Niesten, 2020). Firms that seek to develop sustainability practices focused on environmental issues can benefit from ACAP to identify the most important and useful knowledge (Delmas et al., 2011; Effiong & Singhal, 2014; Riikkinen et al., 2017). Thus, the third research hypothesis is:

H3: Absorptive capacity positively and directly influences environmental sustainability practices.

Small and medium-sized firms face challenges in adopting sustainability practices (Bruijn & Lulofs, 2000; Revell & Rutherfoord, 2003), and partnerships with other firms in a cluster can help overcome these barriers (Hunt, 2000; Revell & Rutherfoord, 2003). In addition, participation in some clusters is an incentive for organizational learning in small and medium-sized firms (Simpson et al., 2004), for the search for knowledge (Segarra-Oña et al., 2016), and participation in programs aimed at sustainability (Friedman & Miles, 2001).

When looking for sustainability concepts, the firm may have doubts about how to adopt the practices, what their scope is, and the relation with the cost, since the profit will not come only because the firm is doing good (Eccles & Serafeim, 2013). Knowing the sustainability practices that other firms adopt help develop sustainability strategically (Garlet et al., 2017). Participating more actively in a cluster or a network can strengthen the relation between absorptive capacity and firms’ adoption of practices (Tsai, 2001). It is noteworthy that different levels of participation in the cluster generate different results regarding access to knowledge, resources, development of ACAP, and the adoption of technological practices (Grimstad & Burgess, 2014; Lis & Rozkwitalska, 2020).

A firm that participates in a cluster and has a high ACAP will seek new knowledge/information, allowing interconnections between firms (Chandrashekar & Hillemane, 2018). The authors indicate that participating in a cluster makes it possible to move resources and establish partnerships. When considering the global economy, it can be understood that elements such as knowledge, motivations, and relations established in a geographic space or region make it possible to generate sustainable competitive advantages.

Clustering allows firms to train workers and suppliers and reduce business costs - the firms participating in a cluster share worldviews, resources, and knowledge (Di-Serio, 2007). The interactions between the firms’ internal resources and the locality resources allow them to be heterogeneous, given the contextual aspects of the resources and capacities (Hervas-Oliver & Albors-Garrigos, 2009) - understanding that the firms that make up a cluster form a network (Berg et al., 2001). The level of participation in the cluster will reflect their strength of competition and cooperation. In addition, it will impact their learning and the establishment of their practices.

Sustainability practices appear to be different when the analysis is carried out in the cultural cluster that comprises more than one country than when it is carried out only at the level of the same country (Miska et al., 2018). Thus, the relations established in the clusters seem to have greater importance than being located in a given country. In clusters, firms, because a collaboration network links them, explore new business opportunities, seeking to increase their economic performance, minimize environmental impacts, and create benefits for the local community (Bellantuono et al., 2017). The cluster is inserted in the place where its performance encompasses economic, social, and environmental issues, with challenges for managers to develop practices focused on these three issues (Larentis et al., 2013).

Economic, social, and environmental practices in the cluster context have been little investigated in the literature, reinforcing the need to measure how much the cluster environment contributes to the firm’s learning in generating sustainability (Garay et al., 2017; McLennan et al., 2016). A firm’s participation in the cluster is based on the business carried out, the strategies outlined, and the relations maintained with other firms in the cluster (Wilkinson & Young, 2002).

Firms that participate in a cluster can share experiences, processes, learning, and practices (Eisingerich et al., 2010) since clusters are like repositories of knowledge that can originate from external sources (McLennan et al., 2016). The network of firms established within the cluster enables sharing ideas, technologies, and knowledge, based on the interaction established between the participants (Eisingerich et al., 2010), facilitating the learning process (Young, 2010).

Firms, small and medium-sized ones, are interested in participating in a cluster, as they see a greater possibility of obtaining an advantage, such as economies of scale (Gardiner & Scott, 2014). Participating in a cluster can lead small and medium-sized firms to access external knowledge and information (Granek & Hassanali, 2006). Clustering can be a more straightforward path to achieving sustainability (McLennan et al., 2016) and having a greater possibility of developing sustainability practices.

In general, putting economic, social, and environmental sustainability into practice is still a challenge for firms. However, the cluster environment, which makes it possible to learn from other firms, can facilitate sustainability practices. There is a possibility to learn from the experiences and relations established through partnerships (Giuliani & Bell, 2005). The environment where the firm is installed, and the relations established can enhance the learning processes to generate sustainability practices. Therefore, the H4 hypothesis arises:

H4: The level of firm participation in a cluster moderates the relation between ACAP and sustainability practices. Higher levels of participation in the cluster will have a more significant effect of ACAP on economic, social, and environmental sustainability practices than lower levels of participation.

The primary data were collected between January and March 2020, in small and medium-sized firms participating in any of the 11 clusters in the state of Santa Catarina, Brazil (Begnini & Carvalho, 2021). The clusters are spread across the territory of the state of Santa Catarina and refer to the following economic activities and municipalities: textile products (Blumenau - 1 and Brusque - 2); wood products (Caçador); food products (Chapecó); electrical machines, appliances, and materials (Jaraguá do Sul); metallurgy and manufacture of machinery and equipment (Joinville); manufacturing of transport equipment, except motor vehicles (Navegantes); leather, travel items, and footwear (São João Batista); computer equipment, electronic and optical products (São José); and non-metallic, mobile mineral products (Criciúma).

Data were collected using a questionnaire with Likert-type scales with a seven-point scale. The elaboration of the research instrument questions sought to meet the criteria established by Fink (2003) considering that after each analysis the indicated improvements were made. After being prepared, the questions were sent for analysis by students in the area. Subsequently, the instrument was sent to specialists in the area and, afterward, presented and discussed in the research group and at the ANPAD Annual Meeting (EnANPAD).

The next step was to apply the questionnaire to six entrepreneurs followed by an informal conversation. The survey instrument was then inserted into the Survey Monkey platform and a pre-test was carried out with 41 responses from managers of firms participating in the Chapecó food cluster. With these data in hand, some statistical tests were carried out and, given the results, the data collection was carried out. The questions that make up the questionnaire are presented in Annex 1.

Eleven questionnaires were built, one for each cluster, and the link of each questionnaire was forwarded to the managers of the firms. The firms’ contacts came from two sources: (a) from a database formed by the researcher; and (b) from a firm specialized in databases that, based on the National Classification of Economic Activities (Classificação Nacional de Atividades Econômicas - CNAE), extracts the contacts of firms from each of the eleven municipalities, according to the main activity of each cluster. Contact was made with 5,718 small and medium-sized firms participating in any of the indicated clusters. In total, 456 responses were received, of which 417 proved to be valid and formed the sample.

Altogether, 456 firms sent the answered instrument. When analyzing missing data, five responses were excluded because they were incomplete. Univariate outliers were sought through data standardization and identification of cases with scores greater than 3 (Field, 2011; Hair et al., 2009) and multivariate outliers by calculating the Mahalanobis distance divided by degrees of freedom, considering those with a value greater than 3.5 (Field, 2011; Hair et al., 2009), 16 observations being excluded. The reliability test was performed using Cronbach’s alpha (AC) and inter-item and item-total correlations.

In these cases, the AC was always greater than 0.7, and no variable was excluded (Field, 2011; Hair et al., 2005). Observations were found to be independent (Hair, Black et al., 2014). It was observed that the variables are linear because the covariance values were different from zero (Cooper & Schindler, 2016; Hair, Black et al., 2014). As for normality, it was noted that the values of asymmetry and kurtosis did not exceed 2 and 7, respectively (Finney & DiStefano, 2013). Although it is possible to notice that there is multicollinearity in some variables, the values of the variance inflation factor (VIF) remained below 10, which is acceptable (Field, 2011; Hair et al., 2009; Kline, 2015). Eighteen observations were also excluded because large firms answered them. In the end, the sample was composed of 417 firms.

There are three endogenous constructs: economic sustainability practices, social sustainability practices, and environmental sustainability practices. It refers to the adoption of sustainability practices by small and medium-sized firms that participate in a cluster. The statements used to measure each of the three constructs were prepared based on the field’s literature. The constructs identified as reflective were measured by five statements each. Sustainability practices have been used as endogenous constructs in some studies (Fairfield et al., 2011; Padilha et al., 2020; Riikkinen et al., 2017).

Absorptive capacity was the exogenous reflexive construct. It used the ACAP formed by four dimensions (Zahra & George, 2002), the construct being measured by an assertion by dimension. ACAP has been used as an exogenous construct in some studies (Aboelmaged & Hashem, 2019; Albort-Morant, Leal-Rodríguez et al., 2018; Riikkinen et al., 2017).

The level of firm’s participation in a cluster was the reflective moderator construct. The four statements used to measure it were elaborated based on the existing literature. Moderation was carried out through multi-group analysis. First, a new variable was formed with the observations’ averages and then the average of the new variable. Of the total, 205 firms remained with values above the average. They were classified with a high level of participation in the cluster. Two hundred and twelve firms remained with the value below the average, classified with a low level of participation in the cluster. Cluster participation was used as a moderating variable (Frazão, 2018).

As control variables, firm size and age were used. These are variables widely used as controls (Zou et al., 2018). The control variables, size, and age were tested to identify whether they influenced endogenous constructs. It is possible to affirm, with 95% confidence, that neither the age group nor the firms’ size influences the adoption of economic, social, and environmental sustainability practices.

The collected data were tabulated in Microsoft Excel spreadsheets and treated with SPSS and AMOS software. Of the 417 firms that made up the sample, 32.85% were between 10 and 19 years old, 27.09 between 20 and 29 years old, 20.62% over 30 years old, and 19.44% up to nine years old. Most of the answers originated from the clusters in the municipalities of Chapecó, Joinville, and Blumenau with 98 (food products), 93 (metallurgy and manufacture of machines and equipment), and 93 (textile products) answers respectively. Subsequently, São José (43 - computer equipment, electronic and optical products), Brusque (24 - textile), Jaraguá do Sul (20 - electrical machinery, equipment, and materials), Criciúma (16 - non-metallic mineral products), Caçador (12 - wood products), São Bento do Sul (10 - furniture), São João Batista (7 - leather, travel items, and footwear), and Navegantes (1 - transport equipment, except motor vehicles).

After addressing issues related to missing data, outliers, reliability, independence of observations, linearity, normality, and absence of multicollinearity, the next step was to organize the analysis model employing SEM, using AMOS (Awang, 2012). Usually, the initial model needs to be changed to arrive at an acceptable and reliable adjustment (Hair, Black et al., 2014). To assess the model, the use of X² associated with degrees of freedom, the CFI, and the RMSEA (Hair et al., 2005) is sufficient. In addition to these, it was decided to include the GFI. Table 1 shows the values for the initial and final models, after the necessary adjustments, and the reference values.

Table 1
Adjustment index, initial and final model values, and reference value

Note. X² = 375.58; df = 114; p = 0.000.

In comparison with the initial model, the end has undergone some changes:

1. (a) The variable PSECO3 was excluded;

2. (b) The PSAMB1 variable was excluded;

3. (c) Covariance between the error terms of PSSOC1 and PSSOC2 was introduced;

4. (d) Covariance between the error terms of PSSOC2 e PSSOC3 was inserted.

The exclusions occurred because the factorial loads presented were below 0.50 and 0.60. Covariance was included because when analyzing the research instrument’s assertions, it was noticed that they maintain a relation with each other. For convergent validity, calculations were performed for extracted average variance (AVE), whose values must be greater than 0.50, of composite reliability (CC), which must be greater than 0.70. AVE and CC’s values were not calculated for the cluster’s constructive level (NPC). They were used as moderator multi-groups and did not participate in the model (Awang, 2012; Hair et al., 2009). Cronbach’s alpha (AC) was also calculated, whose values must be greater than 0.70. These are expressed in Table 2.

As for the discriminant validity, the correlation loads between the constructs were used. Diagonally, the correlation loads were replaced by the values of the AVE root. Such values need to be greater than the correlation values between the constructs (Malhotra, 2019). Table 3 shows the results.

Table 2
Convergent validity

Table 3
Discriminant validity

It can be seen that the model presented adjustment, reliability, and validity. It was observed that all the factor loads between the construct and the latent variables were more significant than 0.60, meeting the expectations (Hair, Hult et al., 2014; Kline, 2015; Malhotra, 2019). Table 4 shows the hypothesis test, where the relations between the exogenous construct and the endogenous constructs were tested.

Table 4
Hypothesis test

Such results support the hypotheses H1, H2, and H3, indicating that absorptive capacity generates economic, social, and environmental sustainability practices. Figure 1 shows the path coefficients’ values and the factor loads between the observed variables and the constructs.

Figure 1
Complete path diagram, with standardized estimates.

It appears ACAP explains about 30% of the data variance in PSECO, 24% of the variance in PSAMB, and 39% of the variance in PSSOC. The greater the effect size, the greater the degree to which the phenomenon under study is manifested (Hair, Black et al., 2014). Following the parameters of Cohen (1998) considering the r², the effect size was large (0.30) in the ACAP-PSECO ratio, medium (0.24) in the ACAP-PSAMB ratio, and large (0.39) in the ACAP ratio - PSSOC

Moderation is an effect that occurs when there is the inclusion of a third variable that influences the relation between the endogenous and exogenous constructs (Awang, 2012; Hair et al., 2005). The moderating effect can be verified by a metric or categorical variable, which divides the sample into two or more subgroups (Hair, Black et al., 2014; Vieira, 2009). In this case, the objective is to verify how the structural model fits into different groups (Vieira, 2009). Before considering the moderating variable, the exogenous construct’s effect must exist and be significant on the endogenous (Awang, 2012). It is then necessary to test whether the exogenous construct’s effect on the endogenous changes according to the condition of the moderator (Hair, Black et al., 2014).

After forming a new variable, which is the average of the four variables of participation in the cluster ((CL1CRESC + CL2PARCE + CL3INOVA + CL4 PR OD) / 4), its average was established, which was 4.57. Below the average, the level 1 (low) group was constituted with 212 observations, and above the average, the level 2 (high) group was formed with 205 firms.

The free structural model was estimated with the level 1 group, which presented an X² of 276.574, with 114 degrees of freedom. The restricted model had an X² of 302.997, with 117 degrees of freedom. The low level of participation showed an X² variation of 26.423 and a variation of degrees of freedom of 3; Using the chi-square distribution formula with ‘n’ degrees of freedom, the significance of the difference between group level 1 and ‘n’ [level 2] was calculated, which was p-value 0.000. As a result, the presented p-value < 0.005 shows that the models are different and moderation occurs (Hair, Black et al., 2014).

Afterward, the free and restricted structural model for the level 2 (high) group was estimated. The X² variation was 22.050, and the degrees of freedom were 3. While in the free model, the value of X² was 274.740, with 114 degrees of freedom, in the restricted model, it was 296.790, with 117 degrees of freedom. When calculating the significance of the difference between the models, the value was p-value 0.000. With a p-value < 0.005, it can be said that there is a significant difference between the models, and moderation occurs (Hair, Black et al., 2014), as shown in Table 5.

Table 5
Free and restricted model for low level and high level of moderation

The adjustment indexes, in this case, X²/df, CFI, GFI, and RMSEA, were compared between the free model and the restricted model. If the restricted model has adjustments as reasonable as those of the free model, then moderation cannot be sustained (Hair et al., 2009). If the free model presents better adjustments than the restricted model, then moderation is sustained (Awang, 2021; Hair et al., 2009).

For group 1 - low level of participation -, the free model presented the results 2.426 (X²/df), 0.869 (GFI), 0.912 (CFI), and 0.082 (RMSEA). The restricted model had 2.590 for X²/df, 0.855 for GFI, 0.899 for CFI, and 0.087 for RMSEA. For the high level of participation, the free model’s adjustment indexes were 2.410 for X²/df, 0.857 for GFI, 0.920 for CFI, and 0.083 for RMSEA. Moreover, the restricted model presented 2.537 for X²/df, 0.846 for GFI, 0.911 for CFI, and 0.087 for RMSEA.

It is noticed that both at the low and at the high level, the restricted model presented adjustment results not as good as the free model, supporting the existence of moderation (Hair et al., 2009). As the results indicate moderation, we started to explore how the effect varies between groups, analyzing the β values in each relation. The values are shown in Table 6.

Table 6
Values of β coefficients of multi-group analysis

It was decided to present both the standardized beta and the non-standardized beta (Vieira, 2009), and that the analysis would be on the non-standardized (Kline, 2015). In the relation between ACAP and PSECO, the path coefficient was 0.62 for high level and 0.46 for low level. In the relation between ACAP and PSAMB, the high level showed a beta of 0.57 and the low level a beta of 0.38. In the third relation between ACAP and PSSOC, the beta for the low level was 0.49, and for the high level it was 0.73. The high level of participation in the cluster allows ACAP to significantly affect economic, social, and environmental sustainability practices, supporting H4, as shown in Table 7.

Table 7
Result of hypothesis H4

* Corresponding author