The theory of dynamic absorptive capacity, absorption capacity, or even absorptive capacity (ACAP) ( Garzón-Castrillón, 2016) is seen as a stream of Dynamic Capacities (DC) theory. It has received focus in recent decades by scholars in Brazil and abroad ( Apriliyantia & Alon, 2017; Guerra, Tondolo & Camargo, 2016; Ferreira & Ferreira, 2017; Zonatto, 2018).

ACAP is related to Penrose's ( 1959) theory of the firm. In the literature, the ACAP theme begins to be approached in greater depth by Cohen and Levinthal ( 1990). The authors refer to the ability to recognize external information, assimilate the implications of this knowledge in the firm, and apply that knowledge to develop innovative performance or generate innovations. In addition, the authors demonstrate the importance of a Research and Development (R&D) department, using the absorptive capacity to appropriate information and knowledge.

The approach proposed by Lane and Lubatkin ( 1998) is that ACAP is a process that involves three capacities that affect the ability to evaluate, assimilate and commercialize. In this sense, these authors accept that ACAP is formed by the abilities of identification, assimilation, and application of knowledge.

Zahra and George ( 2002) adopt a more procedural perspective on ACAP. They argue that effective internal knowledge sharing, and integration are critical to this capability. In this aspect, ACAP would be a multidimensional construct formed by a set of organizational routines and processes through which companies produce a dynamic organizational capacity. ACAP is the combination of four capabilities that influence the organization to create and develop the knowledge needed to build other organizational capabilities, which are the basis for a competitive advantage.

These capabilities are, at the same time, different and complementary in influencing organizational results. Thus, Zahra and George ( 2002) propose a set of organizational routines and processes with two distinct dimensions: potential absorptive capacity (PACAP), which includes acquisition and assimilation capacities; and realized absorptive capacity (RACAP), which includes transformation and application capabilities, which the authors connect to the concept of dynamic capabilities, where the organization’s essential resources originate from competitive advantages.

The approach proposed by Lane, Koka and Pathak ( 2006) reaffirms the multidimensional aspect of ACAP. However, the authors return to the three capacities of the seminal model by Cohen and Levinthal ( 1989). They separate the sequential learning processes as far as each process requires different organizational capabilities. Going further, Lane, Koka and Pathak ( 2006, p.856), when combining the original ideas of Cohen and Levinthal ( 1989, p.128), suggest the following more detailed definition of absorptive capacity:

Absorptive capacity is the company’s ability to use externally realized knowledge through three sequential processes: (1) recognize and understand as potentially valuable new knowledge that may come from outside the company, through exploratory learning; (2) assimilatevaluable new knowledge through transformative learning; and (3) using assimilated knowledge to create new knowledge and commercial outlets through exploration learning.

Lowik, Kraaijenbrink and Groen ( 2016) view ACAP as usually considered a multilevel capability. Its multilevel nature concerns the ACAP of each individual and the mechanisms of social integration at the organizational level. Furthermore, for these authors, the ACAP of a company is not simply the sum of the ACAP of its employees. Therefore, it is helpful to consider which aspects of the ACAP are distinctly organizational.

We adopted the concept of ACAP by Cohen and Levinthal ( 1990) to guide this study. It refers to the firm’s collective ability to recognize the value of new external knowledge, assimilate it and apply it. The adoption of this concept is justified given its adherence to the research objective. In addition, the study converges with the multilevel absorptive capacity perspective proposed by Lowik, Kraaijenbrink and Groen ( 2016). Ferasso and Grenier ( 2021) state that capabilities act at the internal level of the company and between companies that are associated for the co-development of innovative projects.

The study of Hussain et al. ( 2022) identify:

The links among antecedents, multi-dimensions of absorptive capacity and competitive advantage. Antecedents include prior related knowledge and a firm’s combinative and information technology capabilities. Multi-dimensions are PAC (acquisition and assimilation) and RAC (transformation and exploitation of knowledge) that are the main constructs in the framework. Competitive advantage is the outcome variable. The purpose of such amalgamation is to provide a holistic view for managers and ractitioners to successfully develop, maintain and enhance a firm’s.

The cognitive structures of individuals are considered important antecedents of absorptive capacity ( Cohen & Levinthal, 1990; Lane, Koka & Pathak, 2006; Todorova & Durisin, 2007). Individual cognition refers to how individuals tend to process information and make decisions ( Fiske & Taylor, 2013).

Organizations are open social systems that process information from the external environment, which are used for decision-making ( Csaszar & Levinthal, 2016; McCarthy, 2003; Daskou, Yannopoulos & Koutoulas, 2012). Including formal and informal collection systems and data. This information about the external world is then filtered and processed for decision-making ( McCarthy, 2003). According to DeToni & Milan ( 2008) and Daskou, Yannopoulos & Koutoulas ( 2012), managers create belief systems or mental models, simplified representations of the information environment.

Mental models are cognitive elements built on an object, and the individual evaluates the image of the object as an impression and a set of attitudes and perceptions about the object ( Dichter, 1985; Martineau, 1958). This mental level of perception leaves an intellectual residue, providing a reason for pragmatic, functional, and logical evaluations.

Mental representations derived from mental models are fundamental in strategy because they allow managers to consider alternative strategies offline ( Gavetti & Levinthal, 2000). That is, mental representations allow managers to consider the merits of alternative strategies without the need to invest and execute the various options ( Csaszar & Levinthal, 2016).

For the authors, Badke-Schaub, Neumann, Lauche and Mohammed ( 2007), mental models reflect the ability to categorize what we know and organize knowledge. It starts from the idea that mental models are built by internal patterns of how the world works or simplifications. Lucas and Mai ( 2022) propose that, like any mental model, the influence a given mental model has on a worker depends on whether the worker holds that mental model in the first place and whether that mental model is activated.

Considering Karakaya and Yannopoulos ( 2010) mental models’ approach, managers receive, process, and disseminate information about their external environment. However, they often need to sift through enormous amounts of information to develop strategies and respond to environmental challenges. Complementing the idea, Zhao and Parry ( 2012) state that the network of meanings shapes mental models, which reflect on managers’ experiences, as well as inferences from the observed or communicated experiences of others. The way in which these experiences are perceived and interpreted depends, in part, on managerial beliefs and norms, as well as the firm’s strategies.

Lucas and Mai ( 2022) propose that these mental models are impactful because they prime action tendencies (preparation versus production) that differentially influence behavior during creative work.

Apriliyantia and Alon ( 2017) state that people are driven by the interactions of their behavior, cognition, and external environment. Through a cognitive process, the external environment can determine the organization’s behavior ( Bandura, 2001). According to Cognitive Theory, the organization relies on the absorptive capacity of its members ( Cohen & Levinthal, 1990) The members are influenced by their cognitive structures ( Volberda, Foss & Lyles, 2010). Moreover, these cognitive structures lead individuals to identify, choose, process knowledge, and determine their behavior and decision-making ( Bandura, 2001).

In the understanding of Wood and Mckelvie ( 2015), mental models can be understood as simplified images of the environment that support beliefs and judgments.

Martignoni, Menon and Siggelkow ( 2016) affirmed that mental models have two components: (i) the connection of representations that capture beliefs about mental model elements and (ii) performance (representations that capture the mapping from actions to performance results).

Given this, Godkin's ( 2008) position regarding mental models mentioned that these are structures from which individual interpretations of the environment derive, including organizational routines that determine perceptions of environmental reality. These individuals carry their mental models and share them with organizations, emphasizing the relationship between cognitive and organizational absorptive capacity ( Lane, Koka & Pathak, 2006).

In summary, it is argued that in the mental model’s approach, managers receive, process, and disseminate information in their environment (Day, 1994). Based on a bibliographic review associating mental models and absorptive capacity, the study that came closest to the purpose of this research was that of DeToni, Mioranza, Milan and Larentis ( 2014). The authors adopt the network perspective of mental models as exogenous variables associated with organizational performance.

Knowledge acquisition (tacit or explicit) derives from one of the primary practices recommended by managers for decision-making. Siachou, Daskou & Yannopoulos ( 2011) estimated that mental models play an essential role in the three capacities of absorptive capacity. Therefore, this study intends to understand the effect of incubated entrepreneurs’ mental models considering this set of variables identified in the literature on ACAP from a multilevel perspective. Based on the above, the following research hypothesis is presented:

H1: Entrepreneurs’ mental models positively and significantly influence organizational absorptive capacity.

Therefore, Table 1 presents a synthesis of the cognitive variables of the mental models pointed out by the literature review and their respective hypotheses.

Table 1 -
Cognitive variables of Mental Models and their relationships with Absorptive Capacity

This study adopted characteristics of descriptive research with a quantitative approach to the data. For sampling purposes, all companies incubated in the set of innovation and technology centers of the Community Universities of Rio Grande do Sul (Universidades Comunitárias do Rio Grande do Sul - COMUNG) and the Santa Catarina Association of Educational Foundations (Associação Catarinense das Fundações Educacionais (ACAFE) were considered.

COMUNG is a group of Community Institutions from Rio Grande do Sul which constitutes the Consortium of Gaucho Community Universities. The institutions that form the Consortium represent a network of education, science, and technology, which covers almost all municipalities in the state ( COMUNG, n.d.). ACAFE is formed by the Educational Foundations of the State of Santa Catarina. The educational foundations that support Higher Education Institutions came together around the constitution of a body with the task of planning, articulating, and coordinating integrated actions between them ( https://acafe.org.br/site/, retrieved in July 15, 2019). Therefore, COMUNG comprises fifteen universities, and ACAFE comprises eleven universities and five university centers.

Thus, the estimated population of companies in incubators and technology parks at COMUNG and ACAFE is 682, among those classified in the pre-incubated, incubated, and graduated phases. From this population, we defined the sample size. For greater accuracy in estimating the minimum sample size, the software G*Power® version 3.1.9.4 was used ( Faul, Erdfelder, Lang & Buchner 2007). The parameters adopted for a predictor were 80%, with a significance level of 5% (0.05) and a mean effect (f2) of 0.15 ( Hair, Risher, Sarstedt, & Ringle, 2019). Thus, using the “a priori” statistical power, the value of 55 responses was obtained; Nevertheless, based on twice that number for greater consistency of the model ( Ringle, Silva & Bido, 2014). Therefore, the obtained sample of 132 responses has statistical power to verify the hypotheses established in this study.

Entrepreneurs who have their companies incubated in the innovation and technology centers of COMUNG and ACAFE were considered research subjects. In this study, we have assumed that the entrepreneur’s characteristics impact the company’s management. Since a person’s behavior and decisions reflect their characteristics, considering the influences of the business environment with its unexpected events ( Bressan & Toledo, 2013).

The instrument for data collection was a closed questionnaire consisting of 51 questions with a Likert-type scale of seven points, with 1 for totally disagree and 7 for totally agree. It was structured with the help of the Google Forms® tool and sent to the e-mail list of incubated companies. The link was sent to the coordinators of the innovation and technology centers at Comung and the Acafe System and forwarded to the entrepreneurs.

As an independent variable, we have the set of six cognitive variables of the mental models: knowledge, creativity, relationship, strategy, experience, and risk management. For measurement purposes, the set of cognitive variables was elaborated from the literature, considering the possible positive influence on organizational absorptive capacity. Figure 1 presents the conceptual model and research hypotheses.

Figure 1 -
Conceptual theoretical model and research hypotheses
Source: Prepared by the author based on research data (2017)

We considered the three capacities (recognition, assimilation, and application) to measure the dependent variable ACAP, according to the approach proposed by Cohen and Levinthal ( 1989, 1990) and Lane, Koka and Pathak ( 2006). We used the following references: Cadiz, Sawyer and Griffith ( 2009), Cappellari ( 2017), and Enkel, Heil, Hengstler and Wirth ( 2017) to prepare the data collection instrument.

The instrument underwent validation by two experts to investigate possible ambiguous or difficult-to-understand items, allowing the elimination or inclusion of new items and ensuring the effective structuring of the questions ( Devellis, 2016). The experts recommended significant adjustments concerning the questions’ verbal agreement since the company is considered the social body of the entrepreneur, that is, as an immediate reflection of the leader’s personality ( Bazzo, 2011).

The theoretical relationships investigated in the research were analyzed through Structural Equation Modeling (SEM) ( Ringle, Silva & Bido, 2014) to answer the established hypotheses. For analysis of the measurement model¹

1 Measurement model (also referred to as external model - outer model), which report the relationships between constructs and indicator variables (Hair et al., 2014). with the SEM technique, the SmartPLS® software version 3.2.8 was used. The use of Smart-PLS in SEM is based on a set of non-parametric evaluation criteria, employing procedures such as bootstrapping and blindfolding for the evaluation of the measurement model of the measures of the constructs (convergent and discriminant validity and internal and composite reliability) and of the structural model²

2 Structural model (also called inner model), which highlights the relationships (paths) between the constructs (Hair et al., 2014).(size and significance of path coefficients [path], predictive validity of the model [R²], and effect sizes [F²]) ( Hair, Hult, Ringle, & Sarstedt, 2013, 2013).

The first hypothesis to be evaluated deals with the influence of mental models on organizational absorptive capacity. The other hypotheses suggest the influence of cognitive variables: knowledge, creativity, relationship, strategy, experience, and risk management on organizational absorptive capacity.

Thus, two reflective models were estimated to meet this research’s hypotheses because of the impossibility of establishing hypotheses in which the indicators are related. Although the present hypotheses relationships between latent variables (VL), they must be understood as indicators when used to measure the second-order LV ( Garver & Mentzer, 1999). Thus, the first model includes second-order latent variables and intends to respond to hypothesis H1. The second model includes first-order latent variables (VL) and seeks to respond to hypotheses H2, H3, H4, H5, H6, and H7. Next, we present the analysis and discussion of the research findings.

A total of 132 companies belonging to the innovation centers of COMUNG and the ACAFE System took part in the study. We found that most of the participants were in the situation of incubated companies, making up 50% (65) of the sample of this study. At this stage, the company supports the enterprise’s development by promoting favorable conditions for its growth, followed by graduated companies with 25% (34).

After being strengthened for a certain period, companies entering the market may or may not maintain the link with the incubator. Pre-incubated companies are represented by 25% (32), a stage in which companies are encouraged to develop their business potential, benefiting from sponsorship to start the enterprise effectively.

As a segment of activity, the largest representativeness (45 companies) of the sample studied was Information Technology (IT), with the development of management software, hardware, and digital solutions. The industry sector also presented representation and health, communication, and services, among others mentioned.

The estimated model (H1) was evaluated for its convergent validity (Average Variance Extracted - AVE), reliability, and discriminant validity (Fornell-Lacker criterion and cross-loadings), recommended by Hair et al. ( 2013) for reflective models.

Table 2 -
Correlation matrix, discriminant validity (Fornell-Larcker criterion), AVE e CR – 1st estimated model

This validation process observed the factor loading coefficients for each first and second-order LV indicator. As parameters for the first estimated model, factor loadings >0.6 were maintained, according to Hair et al. ( 2013). For optimization and objectification purposes, ( Table 2) presents the correlation matrix. On its diagonal, the values refer to the discriminant validity (Fornell-Larcker criterion) and the AVE and CR values of the order of the first and second-order variables.

The AVE values of the APL, ASS, CONH, CRIAT, ESTRAT, EXPER GESTRISC, REC, and REL indicators reached values >0.50. For the CR of the first order, LV values >0.70 were obtained, with the variable with the highest value being APL, with 0.907. Also, the analyzed data presented statistical normality for the second order VL, ACAP, and MODEL. The reliability indicators were higher than 0.50 for AVE and >0.70 for CR.

In the first analysis for the discriminant validity, cross-loadings were considered. It is possible to verify that the indicators of the first and second-order LVs have higher values relative to the corresponding ones, which ensures their discriminant validity. For the analyzed constructs in which the criteria were met, we inferred that the assertions are effectively observable indicators proposed in this research.

In addition to this analysis, Hair et al. ( 2009; 2013) emphasize using a second criterion, the Fornell-Larker, where the values on the main diagonal must be higher than the values in the respective rows and columns (correlations), evidencing the presence of discriminant validity.

Thus, the correlations between the VL, in which it is possible to verify that they are smaller than the square root of the AVE (indicators have a stronger relationship with its VL than with the other VL), confirming the discriminant validity ( Table 2). After conducting the measurement step of the first reflective model, we proceed to the analysis of the structural model and the realization of the H1 hypothesis test.

The analysis of the structural model seeks to verify the statistical validity and confirm the adequacy of the measurement model, attesting to the significance of therelationships between the study constructs and the confirmation of the hypothesis. Table 3 shows the results of the path, t-value, p-value F² and R² tests.

The level of significance, the t-value of the VL, is more significant than 1.96, demonstrating that the model is adequate and with a significance level of 5%. For structural relationships of H1, the f² of 1.515 was >0.35. The reference value of 0.02 is considered minor, the value of 0.15 is considered medium, and the value of 0.35 is considered significant ( Hair et al., 2013). Next, we present the tests for the second estimated model.

Table 3 -
Structural Model Results – 1st Estimated Model

Tests were performed for CVA for this >0.50 and CR >0.70, as Hair et al. ( 2009, 2013) recommended. We eliminated indicators with factor loadings >0.6 ( Hair et al., 2013). Continuing, Table 4 presents the correlation matrix of the second estimated model, where, on the diagonal, are the values referring to the discriminant validity (Fornell-Larcker), as well as the values of AVE and CR.

As seen for the first order LV, all indicators reached values >0.50 for AVE and >0.70 for CR. The existence of discriminant validity (cross-loadings) in the second model, considering the Fornell-Larcker criterion, where the values of the square root of the AVE are higher than the values of the existing correlation between the cognitive mental models’ variables and ACAP, we infer the discriminant validity. Next, we present the analysis of the structural model, as well as the tests of hypotheses H2, H3, H4, H5, H6, and H7.

Table 4 -
Correlation matrix, discriminant validity (Fornell-Larcker criterion), AVE and CR - 2nd estimated model

As seen for the first order LV, all indicators reached values >0.50 for AVE and >0.70 for CR. The existence of discriminant validity (cross-loadings) in the second model, considering the Fornell-Larcker criterion, where the values of the square root of the AVE are higher than the values of the existing correlation between the cognitive mental models’ variables and ACAP, we infer the discriminant validity. Next, we present the analysis of the structural model, as well as the tests of hypotheses H2, H3, H4, H5, H6, and H7.

The results of the second structural model are represented in Table 5, where it is possible to verify the VL’s effect in the ACAP construct.

Regarding the evaluation of the effect size (f²), the results revealed that the relationships of VL CONH, CRIAT, and GESTRISC with ACAP are of medium and large effect, considering that Hair et al. ( 2013) indicate a value of 0.15 as a mediumeffect and a value of 0.35 as a significant effect. Therefore, the results suggest that the cognitive variables of the incubated entrepreneurs’ mental models significantly predict organizational ACAP. Furthermore, we infer that the hypotheses H2, H3, and H7 proved adequate for the model, as they presented a t-value greater than 1.96, with a significance of 5%. After concluding this analysis stage of the two estimated models, we discuss the results.

Table 5 -
Structural model results – second estimated model

E-mail: lisianehermes@upf.brE-mail: josausen@unijui.edu.brE-mail: gcfconsultoria@gmail.comE-mail: rossetto@univali.br