1 Introduction

Recently, a particular type of skills, namely soft skills have gained relevance in both the academic and practitioner worlds. Contemporary management sciences representatives consider soft skills as one of the people factors that are key for achieving organizational development and effectiveness (Krawczyk-Sokolowska et al. 2019). Together with individual competencies, soft skills have been widely studied over the last decade (Dunbar et al. 2016; Fleming et al. 2009; Orr et al. 2011; Pang et al. 2019; Poon 2014; Stewart et al. 2016; Succi 2019; Wheeler 2016). Nevertheless, studies on soft skills seem to be characterized by theoretical dispersion and lack of a consensual definition (Lipman 2019). This blurriness translates, for instance, into some authors using soft skills and competencies as interchangeable terms. Chaffer and webb (2017) refer to generic skills as competencies. Wesley et al. (2017) use the term competence as a synonym of skill. Similarly, Fernandes et al. (2018) define Emotional competencies as skills.

According to Wats and Wats (2009), the terminology used to refer to soft skills varies from country to country. For example, some people call them key competencies, soft skills, generic skills, or employability skills in Australia. European authors call soft skills vital competencies, while in the United States, some equivalent terms are employability skills or workplace know-how. In the same line, Succi (2019) suggests there are different ways of referring to soft skills, such as life skills (WHO 1994), transversal skills, generic competencies and critical competencies (OECD 2001, 2012), and lifelong learning (European Union 2006).

The confusion between soft skills and competencies may be due to the fact that, for some scholars, the more general concept of skills is not clearly differentiated from competencies (Bamiatzi et al. 2015). In the soft skills literature, the confusion is surprising. Specially considering that in the seminal and most cited works there is a clear conceptual differentiation between skills and competencies (Boyatzis 1982; Spencer and Spencer 1993).

Methodologically, a similar landscape is found when examining competencies and soft skills literature. As expected in any field nourished by the social sciences, studies on competencies and soft skills also offer a wide variety of choices. Desirable or not, methodological pluralism pervades even within qualitative and quantitative approaches (Heesen et al. 2019). In order to assess the state of a theory or a concept, reviewers should consider methodological aspects of the extant literature. The state of prior knowledge and the methodological aspects of corresponding studies must be intertwined. Furthermore, internal consistency between the elements used by the research method of studies including data types, data collection techniques and analyses should be ensured (Edmondson and Mcmanus 2007). These methodological decisions follow the research design, defined as the rationale that leads the researcher in responding to the research question (Salkind 2010). Each of these methodological decisions affect the studies’ contributions to theories and concepts. According to Suddaby (2010) theoretical and conceptual developments can be hindered by lack of construct clarity. Weak concept operationalization leads to internal and external validity problems. In the behavioral sciences, most variables are indirectly measured and are then assessed with error (Wolf et al. 2013). Researchers should use adequate methods when developing and testing metrics, otherwise they will hamper the validity of their results. They need to assess the psychometric properties of the scales used to collect data for further hypothesis testing. For instance, a study that inquires if a leader's emotional intelligence affects a team's performance and fails to validate the corresponding scales would assume that the variables are measured correctly. Suppose that measurement errors associated with the items of the scale of emotional intelligence are incredibly high. This association suggests that the data gathered about the emotional intelligence of the leaders would be biased and, in turn, would jeopardize the conclusions of the study about the effect of that soft skills on team performance.

In view of the conceptual, theoretical and methodological pluralism of the business-related literature on the concept of soft skills, we conducted a systematic review. This methodology allows locating, assembling and evaluating relevant literature on the topic (Khan et al. 2003; Kisely et al. 2014). A systematic, as opposed to a regular review was employed to respond to these research questions. A systematic review adopts a replicable, scientific and transparent process to reduce researcher bias by monitoring and documenting decisions, procedures and conclusions. Its results provide the foundations for making conceptual, theoretical, and methodological contributions to the subject matter (Marabelli and Newell 2014). It is not a literature review in a traditional sense, but a self-contained research project guided by specific research questions (Denyer and Tranfield 2009). This review aims to provide new insights that stimulate future research and theory building for studying soft skills, by responding to three research questions:

  • Research question 1: How are soft skills and competencies conceptualized in the reviewed literature?

  • Research question 2: What are the main theories used in the study of soft skills and competencies?

  • Research question 3: Methodologically, what are the main characteristics of those studies?

This paper is structured as follows. In the first section we outline the method used in this review. We further explain why a systematic review instead of a narrative review was employed. We explain how, by adopting a replicable, scientific and transparent process, this paper complement prior narrative reviews. Subsequently, the systematic review protocol is explained. The results are then presented in the next section. Drawing on the results of this SR, we propose a theoretical model aimed at shedding light into the soft skills’ nomological network. By discussing the theoretical model, we posit that competencies and soft skills are different but related constructs and suggest where soft skills should be situated with respect to related constructs. In face of the blurriness surrounding the notion of soft skills, the discussion of the theoretical model also provides an alternative for the conceptualizacion of soft skills. In this way, we expect to bring new theoretical and methodological insights to business and management scholars interested in studying soft skills.

The contributions of this systematic review are varied. First, it contributes to the conceptual clarity of the field by pointing at the differences and relationship between competencies and soft skills. Second, this systematic review summarizes the methodological characteristics of extant studies as a means to outline those that need to be improved in order for the field to advance. Third, it identifies theoretical and methodological alternatives for the study of soft skills and offers insight as to how they may be used to advance the field.

2 Method

We conducted a systematic review to respond to the research questions. A systematic review is a methodology that establishes specific criteria for assessing and summarizing previous studies accurately and reliably to answer specific research questions. The systematic review allows minimizing research bias by using organized methods of locating, assembling and evaluating relevant literature on the topic (Khan et al. 2003; Kisely et al. 2014). Hence, the systematic review overcomes narrative reviews because of the use of a replicable, scientific and transparent process (Tranfield et al. 2003). In this sense, this systematic review contributes to the understanding of soft skills by complementing prior narrative reviews focused on some types of soft skills (e.g. Klein et al. 2006). For this SR, we followed Tranfield et al. (2003), who provided guidelines to adopt systematic review principles for studying management literature. The first stage consisted of developing a protocol review to document the steps to be taken. A Microsoft Excel Workbook was then used to establish the five steps of the systematic review (Denyer and Tranfield 2009). Table 1 presents a summary of the systematic review process.

Table 1 Soft skills systematic review steps and highlights
Full size table

After the formulation of the questions guiding the systematic review process, it was necessary to identify the search terms and the sources of information. Regarding search terms, this systematic review went beyond soft skills and included competencies as a search term. This decision was made considering that every construct is immersed in a given nomological network. Put differently, a construct is conceptually connected to other constructs (Cronbach and Meehl 1955; Suddaby 2010). As suggested before, there were reasons to believe that competencies and soft skills are connected. Again, some authors use them interchangeably (e.g. Spowart 2011). When conducting a systematic review, an election between breadth and depth needs to be made, and balance between those needs is recommended to achieve a good review (Fisch and Block 2018). As a result, we selected the general term soft skills instead of including the types of soft skills as in a previous narrative review (e.g. Klein et al. 2006). Overall, 119 articles were thoroughly reviewed in this SR. Regarding sources of information, we selected the Web of Science and Scopus databases. These are recognized as the two most important multidisciplinary bibliographic databases (Wang and Waltman 2016).

For the third step, namely study assessment and selection, we ran a first round of searches and found 709 journal articles. Web of Science and Scopus allowed us to apply the following inclusion criteria: search terms in the publication title, journal articles, language, timeframe, discipline fields (Table 1). The number of articles retrieved by the first search, the fact that the literature on soft skills cites works published as far back as 1973 (McClelland 1973), and the use of the term soft skills since 1990 (Moss and Tilly 1996) suggested that the soft skills research field was mature. When a research field is mature, quantitative research is called for (Edmondson and Mcmanus 2007). Hence, only quantitative and mixed studies were examined in this SR. The amount of studies resulting from the preliminary search and the maturity of the soft skills field also suggested that 2014–2019 was a sufficient date range for this SR. Since our interest was aimed at studies on soft skills in business and management, articles related to other academic fields, such as engineering or nursing, were excluded.

We registered the bibliographical metadata of each article and added an inclusion–exclusion column/cell for justifying the decision to include each article in the review. This first list of articles was called Core Contributions List 1. After assessing the quality of the studies and reading the abstracts, a second Core Contributions List was created. In this list, we only included articles related to the systematic review questions and those listed in the Journal Citation Report (JCR) and/or the Scimago Journal Rank (SJR) in 2018. Of note, the journals listed in these databases are recognized as relevant in their academic fields and as offering validated knowledge (Wang and Waltman 2016). List 2 contained 280 potentially relevant articles. Then, two members of the research team identified the research type and methodology of the 280 potentially relevant articles and agreed on including 226 empirical, quantitative and mixed studies (Core Contributions List 3). Given that decisions regarding inclusion and exclusion are often subjective (Tranfield et al. 2003) four researchers participated in the final inclusion discussion. They analyzed the 226 full-text articles by creating an extraction matrix to collect and code information from the 226 studies and decided on the final list of studies to be included in the review. Then, each author independently coded approximately 56 articles, and met every week to discuss the texts and enhance the consistency of the coding process. Additionally, to guarantee interrater reliability in the application of the inclusion criteria, a second researcher reviewed the articles as well. When discrepancies arose, a third reviewer provided an opinion to solve discrepancies. The final list included 119 articles. The main reasons that explain the exclusion of 107 studies were that those studies were related to competencies at a team, project, firm or organization-level, not at an individual level, or considered fields other than business and management.

In the fourth step of the SR, analysis and synthesis, we had already designed a matrix to extract and code information from the 119 included studies. The matrix included: conceptualizations, theories, hypothesis, sampling, participants (n), research designs, instruments or measurements, data analysis, study validity and reliability, and future research lines/gaps. Finally, step five, reporting the results, is presented in the following section.

2.1 Conceptualization of soft skills

One of the key challenges identified in the selected literature is that there are many definitions of competency. Furthermore, the terms “skills,” “expertise,” “acumen,” and “competency” are interrelated and often used interchangeably in the literature (Bamiatzi et al. 2015). Other authors define skills as components of competency. Therefore, before presenting a definition of soft skills, we found it relevant to analyze the concepts of competence, competency, and competencies. Competence is the generic capability of a professional, whereas a competent professional can accomplish their job assignments. On the other hand, competency is one of the components of the individual’s competence. Competencies are the plural of competency (Mulder 2015). A competent professional possesses a series of highly developed competencies. Some authors define the term competency as an underlying characteristic of a person, which results in effective and or superior performance in a job. It may be a motive, trait, skill, aspect of self-image, social role, or a body of knowledge (Boyatzis 1982; Spencer and Spencer 1993). It is important to bear in mind that following this definition of competency, used by the most-cited authors in the papers included in this systematic review [Boyatzis 1982 (30); Spencer and Spencer 1993 (18); McClelland 1973 (15); Mulder 2006, 2007, 2014, 2015 (11)], skills appear to be a component of that construct. Clarifying the relationship between soft skills and competencies is vital since, as we expected, some authors treat these terms interchangeably. We found that 37% (44) of the articles do not differentiate between softs skills and competencies. As for the other 63% (75) of the papers, we found that the authors refer directly or indirectly to soft skills. Nevertheless, in those studies, the authors do not define the term (Fig. 1).

In the view of the main authors cited to define competency, a skill is defined as the ability to perform a certain physical or mental task (Spencer and Spencer 1993) that is functionally related to attaining a performance goal (Boyatzis 1982). The exact origin of the classification of skills between hard and soft skills is less clear. However, within this systematic review it was possible to look at the work of Whitmore and Fry (1974). These authors defined soft skills as important job-related skills that involve little or no interaction with machines and can be applied in a variety of job contexts. This definition suggests that soft skills are generic skills, contrary to specific skills required for particular fields and disciplines (Albandea and Giret 2018). Conversely, hard skills refer to the tangible technical expertise and know-how needed for work (Levant et al. 2016).

Soft skills have different definitions. Some scholars argue that soft skills are non-technical and do not rely on abstract reasoning, involving interpersonal and intrapersonal abilities to facilitate mastered performance in particular social contexts (Hurrell et al. 2013). Some authors refer to soft skills as pervasive or generic skills (Viviers et al. 2016; Keevy 2016). Others, see soft skills in terms of personality traits, preferences, motivations or other components of a competency, meaning that they are primarily cognitive in nature and are influenced by individuals’ intelligence (Semeijn et al. 2005; Wesley et al. 2017). Another group of scholars indicate that soft skills are, in fact, skills as they require learned behavior, trained, and based on individual predispositions (Balcar 2016; Albandea and Giret 2018).

It is noteworthy that, despite the widespread use of the term soft skills since 1990 (Moss and Tilly 1996), only 13 out of 119 (11%) articles provided a definition of soft skills. Table 2 exhibits the main authors cited to provide soft skills definitions in the papers analyzed in this SR.

Table 2 Definitions of Soft Skills
Full size table

The definitions of soft skills shown in Table 2 suggest that those skills have two main components: intrapersonal and interpersonal skills. We found that the concept started focusing solely on interpersonal skills (Rainsbury 2002), which may be why terms such as people and social skills became relevant when studying soft skills. After 2011 other authors complemented the definition referring to intrapersonal skills, explaining that soft skills comprise not only how to handle interactions with others but also the ability to manage oneself (Laker and Powell (2011).

Furthermore, the conceptualizations of soft skills displayed in Table 2 suggest that those skills have two main components: intrapersonal and interpersonal skills. However, given that sometimes soft skills are synonymous with core skills, personal skills, and people skills (Ibrahim et al. 2017) or social and emotional skills (OECD 2017), we analyzed whether the scholars using these alternative terms were studying both intrapersonal or interpersonal skills, despite not mentioning the precise concept of soft skills. Interestingly, the findings show that 29 out of 119 (25%) articles used within their studies a classification of skills and competencies related to intrapersonal and interpersonal skills. Table 3 (see “Appendix 1”) presents examples of these classifications. For instance, Deaconu et al. (2014) studied professional and transversal competencies. Dippenaar and Schaap (2017) referred to intrapersonal competency, interpersonal skills, adaptability, and general mood, while Khan (2018) studied delivery-related competencies, interpersonal competencies, and strategic competencies. Meanwhile, Shah and Prakash (2018) addressed strategic, analytical, personal, managerial, professional, and leadership skills.

Table 3 Classification of competencies and skills
Full size table

Finally, even if the 119 articles studied soft skills to some extent, 43 (36%) devoted their research to one particular skill. Leadership (Quintana et al. 2014; Michael Clark et al. 2016; Ren and Zhu 2017; Seidel et al. 2017; Ahmed and Anantatmula 2017; Wang et al. 2018; Shao 2018; Nijhuis et al. 2018; Wei et al. 2018), and emotional competencies (Delcourt et al. 2017; Fernandes et al. 2018; Gruicic and Benton 2015; Khalili 2016; Maqboo et al. 2017; Matute et al. 2018; Padilla-Meléndez et al. 2014; Tam et al. 2014; Tognazzo et al. 2017a, b) received increasing attention from scholars over the past five years. In the next section of this paper, we present a broader perspective of the findings indicating which are the theories behind the competency and soft skills constructs as defined in the papers included in this SR.

2.2 Theories used in the study of soft skills

The following results have been drawn considering an explicit declaration of the use of a theory in the selected studies by their authors. Surprisingly, 51 (43%) of the selected papers did not explicitly declare a theory for framing their studies while 68 (57%) papers did. In most cases, the papers used a single theory (82%). Overall, the papers use 50 theories, amongst which a majority of theories (34) were only used once. The Resource-Based View (RBV) and dynamic capabilities theories appear to be the most frequently used theories. Although the results show that the frequency of usage of most theories is low, it is worth emphasizing that some of the theories used in the papers are intimately related. In this sense, in light of the theoretical dispersion found in this SR, and to facilitate an understanding of the theoretical landscape in the soft skills field, we placed the theories in several groups: namely, resource-related theories, capital theories, learning theories, and leadership theories. We present the frequency of use of each theory and group in Fig. 2. We also provide a detailed list of studies and theories within each group in “Appendix 2”.

The first theoretical trend found in this systematic review is associated with RBV and related theories. There are three closely related theories yet different to some degree: RBV, dynamic capabilities, and knowledge-based view (KBV). Penrose RBV in 1959, and this theory has evolved ever since, resulting in the creation of KBV in the mid-90 s and the concept of dynamic capabilities towards the end of the decade (Fitri Ande et al. 2018). Noteworthy, the theories in this group refer to competency in general, as opposed to soft skills in particular. Hence, we posit that other theories focusing on soft skills instead of competencies would be more suitable to the field.

As defined by Barney (1991, 1995), RBV affirms that resources are a firm’s assets. These resources are knowledge, capabilities, processes, and valuable, rare, and difficult to imitate. When effectively managed, they become a source of competitive advantage for the firm (Fitri Ande et al. 2018). Under this theoretical umbrella, the manager’s competence is vital in building a firm’s competitive advantage (Barney 1991, 1995). Thus, the soft skills of the managerial force also become essential to achieve such an advantage. In this regard, RBV has been used to study the relationship between top management’s tangible competencies (Akhtar et al. 2018) and talent mindset competency (Luna-Arocas and Morley 2015) and performance. RBV has also inspired 360-degree management competency assessments (Liang et al. 2017). It has also been the basis to study the mediating role of employee competencies in the relationship between high-performing human resources practices and firm performance (Van Esch et al. 2018), and the impact of transformational leadership, entrepreneurial competence, and technical competence on firm performance (Ng and Kee (2018).

According to Teece et al. (1997), the dynamic capabilities theory proposes that a firm’s resource base needs to be constantly modified in a dynamic environment. Especially, considering that the firm exists within an open system that means it needs to interact with the environment. Under this approach, firms must constantly sense, seize and reconfigure their resources (Teece et al. 1997). We found that the dynamic capabilities approach has been used to establish the required competencies for the effective management of legislated business rehabilitation (Rajaram and Singh 2018), to confirm a positive effect of managerial competencies on engagement (Lara and Salas-Vallina 2017), to link management competence to the performance of small enterprises (Zacca and Dayan 2018), to test the relationship between talent mindset competency, job satisfaction and job performance (Luna-Arocas and Morley 2015), and to assess the impact of competences, namely transformational leadership, entrepreneurial and technical competence, on firm performance (Ng et al. 2016; Ng and Kee 2018).

Through a KBV lens, knowledge may reside within individuals and is considered to be the most important resource for developing a firm’s competitive advantage (Grant 1996). KBV proposes that knowledge is the most important resource and source of competitive advantage (Kogut and Zander 1992). There are several examples of studies on soft skills using this theoretical base. One study draws on KBV to understand supply chain management (SCM) competencies by splitting them into individual and organizational components and their impact on SCM performance (Flöthmann et al. 2018a). Another research study examined in this systematic review employed KBV to understand the competency requirements of supply chain planners and analysts (Flöthmann et al. 2018b).

The second group of theories identified in this systematic review are learning theories. Culture learning, individual learning and experiential learning are included in this group. Most of them appear to have a starting point in the work of Kolb (1984), which can be considered as the basis for the study of competencies as it provides a generic competency model (Yamazaki 2014). Other theories in this group are derived from the work of Bandura (1977, 1986), namely, social learning theory and social constructivist theory. The theories in this group hold great potential to investigate the process through which competencies, and particularly soft skills, develop. Employing these theoretical frameworks may aid in understanding the effect of training processes and educational programs in the development of soft skills.

Culture learning theory focuses on studying how a sojourner may learn or acquire the relevant social knowledge and skills of another culture (Ward et al. 2001) by means of contact with the locals (Van Bakel et al. 2014). This theory has been used by Van Bakel et al. (2014) to assess the impact of being in contact with local hosts on an individual’s intercultural competency. Individual learning theory proposes that learning is the outcome of an individual’s experience and motivation (Kolb 1984). This theory has been used by Bartel-Radic and Giannelloni (2017) to show that most personality traits do not determine cross-cultural knowledge. Experiential learning theories are based on the propositions of Kolb (1984). According to this theory, in integration with a competency-based approach (Boyatzis 1982), an individual’s emotional, social and cognitive competences can be shaped by their experiences and interactions with the environment. This theory was used by Padilla-Meléndez et al. (2014) to understand the effect of changes in emotional competences on students’ entrepreneurial intention, measuring it before and after an outdoor training activity (Kolb 1984; Boyatzis and Kolb 1991, 1995). Kolb’s (1984) work was also applied in a study conducted to explore the competency of host-country nationals in multinational enterprises (Yamazaki 2014).

Theories based on Bandura’s social learning theory propose that human behavior is substantially learned through vicarious models (Bandura 1986). Drawing on this theory and the work of Teven (2007), who proposed that there is a relationship of mutual influence between supervisor and subordinate, Steele and Plenty (2015) studied differences in communication competence, communication satisfaction and job satisfaction in the supervisor-subordinate relationship. This theory has also been used by Wang et al. (2014) to study the development of cross cultural competence in expatriates, considering that their new environment and social relationships provide opportunities for learning (Black and Mendenhall 1990).

Banduras’ (1986) social cognitive theory was used by Pérez-López et al. (2016) to determine that there is a direct relationship between resilience and entrepreneurial intention. It was also used by Ren and Zhu (2017) to understand why managerial leaders undertake self-development activities. The social constructivist theory was first introduced by Vygotsky (1978) who proposed that learning and knowledge are constructed through social interactions, and as such, are not the result of individual experiences but social processes. It suggests that learners will build their own understanding and knowledge of the world through their experiences and their reflections on those experiences (Harasim 2012). This theory has been used as the basis to study the impact of training opportunities on trainee perceptions of their competency levels by Chaffer and Webb (2017).

The third group of theories identified in this systematic review are leadership theories. These theories enable investigating the characteristics of leaders, in terms of both competencies and skills, and how they influence followers and team performance. Consequently, authors may increasingly use these theories in multilevel studies to understand the relationship between a leaders' soft skills and followers' performance. However, as we mention below, they are limited since they focus on the leader and not on the leadership process itself.

According to Yukl, Gordon, and Taber (2002), change orientation is a key dimension of an effective leader’s behavior. Yukl et al. (2002) referred to general leadership theories and in particular to the taxonomies of leadership behavior. This theory was used by Quintana et al. (2014) to study the effects of disclosing the specific leadership profile of graduates on the leadership behavior of actual leaders 5 years after graduation. This same author, along with others like Yukl (2008); Yukl et al. (2002), and Yukl and Lepsinger (2004) proposed the flexible leadership theory. According to this theory, only the behaviors of individuals occupying leadership positions have an influence on the determinants of organizational effectiveness. Furthermore, it is the role of organizations to admit only those individuals with competencies that are significant enough to achieve firm performance Tognazzo et al. (2017a, b). This was the theoretical lens employed by Tognazzo et al. (2017a, b) to identify the emotional intelligence competencies of top leaders that affect firm performance. The implicit leadership approach is also included in the leadership theories used in the soft skills studies included in this systematic review. This theory proposes that different raters may have different views of a manager’s behavior (Judge et al. 2002). According to that perspective, Semeijn et al. (2014) showed how different raters, including subordinates, peers, and supervisors, value competencies in managers in different ways.

The transformational leadership theory is also included in the fourth group. This theory proposes that leaders need to develop a deeper connection with their subordinates (Bass and Riggio 2006). That premise was used by Knight and Paterson (2018) to study the key behavioral competencies that leaders in the field of corporate sustainability require, and to propose an assessment model. The last item in this group of theories on leadership is the great man approach. According to this theory, leadership competency models are facilitators for grooming project managers to become leaders (Hollenbeck et al. 2006). The great man approach was used by Ahmad, Kausar, and Azhar (2015), and Ahmed and Anantatmula (2017) to examine the effectiveness of HR professionals, including willingness as an important competency.

The fourth group of theories in this systematic review is made up of capital-related theories. Although the theories in this group are rooted in RBV, we placed it in a different group to facilitate understanding the theoretical trends in this systematic review. According to Adner and Helfat (2003), the human and social capital of managers are components of their dynamic capabilities, being the third their managerial cognition. Human capital theory states that a firm’s investment in human capital contributes to increased productivity and is therefore critical to a firm’s success (Hitt et al. 2001). Human capital has been proposed as the missing link in the dynamic capabilities theory for creating a competitive advantage. Human Capital consists of an individual’s knowledge, skills and experience (Baird 2014), and is gaining importance in an increasingly dynamic work scenario (Hennekam 2016). Research on human capital has focused on the role of knowledge, skills, experience and/or education as the link between individual and organizational outcomes and human capital (Ployhart and Moliterno 2011). This theory has been used as the theoretical framework for several studies included in this SR. A good example is the study by Hennekam (2016) wherein the author inquired into the influence of the motivation, integrity and social skills on both intrinsic and extrinsic career success. Social capital is another capital-related theory and it states that an individual’s social capital allows them to acquire implicit knowledge which in turn contribute to refining existing knowledge. It has been defined as the ability possessed by individuals to benefit from their social structures, networks and memberships (Honig and Davidsson 2000) and to identify who knows valuable information and how accessible it is (Borgatti and Cross 2003). In this regard, social capital has been considered to be key to entrepreneurial success (Lans et al. 2015) and has been used to propose a separation between an individual’s ambidextrous and their behavioral competencies (Lee and Lee 2016), and to study how and to what extent social competence influences social capital amongst students with latent entrepreneurial ambitions (Lans et al. 2015).

Finally, although not frequently used, we were able to identify one more theory: contingency theory. Contingency theory was used by Shao (2018) to investigate the moderating effect played by a program’s context in the relationship between leadership competences and program success. According to contingency theory, organizational processes and structure need to fit with their context to achieve improved performance (Lawrance and Lorsch 1969; Schoonhoven 1981). Contingency theory was also used by Amedu and Dulewicz (2018) to study the relationship between CEO power, competencies, and firm performance.

In summary, each group of theories has the potential to contribute to the study of soft skills. The first group of theories, those related to the resource-based view, treat competencies and soft skills as resources that impact organizational performance. In this regard, they help to understand the contribution of employees’ competencies and soft skills to firm-level performance, but do not focus on individual antecedents and outcomes of these variables that by definition are individual level factors. The second group, learning theories, includes those theoretical approaches that are related to individual learning. These theories seem to have a better fit to the study of soft skills’ antecedents and consequences. They also encompass the interaction of individual factors with context as part of the learning processes. The third group of approaches, leadership theories, used in the study of competencies and soft skills is composed of leadership theories. Traditionally, these theories focus on the traits of leaders and how those traits, in this case competencies and soft skills, affect performance at various levels. Yet, soft skills are individual traits and are present in individuals across all organizational levels. Put differently, soft skills and their effect on work settings go beyond leadership and managerial positions. The last group of theoretical approaches, capital-related theories, identified as contributing to the study of competencies and soft skills includes human and social capital theories. They are useful to understand how a firm’s investment in such capital may influence organizational level-outcomes. With regards to contingency theory, some insights about its potential for the study of soft skills are presented at the end of the results section, where we propose a soft skills theoretical model.

2.3 Methods used in the study of soft skills

Regarding the research design of the papers included in this SR, the results indicate that most studies adopted a cross-sectional design (69%). 39 papers directly declared to have used a cross-sectional design, while 43 of the sample studies did not declare the use of that type of research design. We had to review the data collection procedures (data collection at one time point) of those 43 studies to arrive at this conclusion. The cross-sectional design is pervasive in the studies belonging to each of the four groups of theories we identify in this systematic review. Most of the soft skills papers based on capital-related theories adopt a cross-sectional design (80%). Similarly, 71% of the studies that draw on RBV and 60% of research using leadership theories use this research design. The percentage of soft skills papers adopting the cross-sectional design decreases but is still significant in soft skills studies related to learning theories (55%). Besides the cross-sectional design, soft-skills studies adopt a variety of research designs. Figure 3 portrays the frequencies of research designs in the reviewed papers.

Of note, studies claiming to investigate the antecedents or consequences of soft skills or competencies used cross-sectional designs (e.g. Ahmed and Anantatmula 2017; Gorenak and Ferjan 2015; Gruden and Stare 2018; Hennekam 2016; Marzagão and Carvalho 2016; Wang et al. 2014). Now, cross-sectional designs need to be optimized to shed light on causal relationships. These optimizations include, for instance, incorporating time and different sources into measurement strategies (Spector 2019). As indicated below, few studies adopting the cross-sectional design used different data sources.

Of the studies included in this review, 51 (43%) were conducted in Europe; mainly in Spain (11), Netherlands (7), Italy (6), Slovenia (6) and the United Kingdom (4). A significant proportion (26%) of the 119 studies examined here used samples from Asia. For the most part, the participants in these 31 studies were from India (6), China (4), Malaysia (4) and Pakistan (4). The remaining studies collected data in the Americas (13), 9 of which were conducted in USA and 4 in Latin America. Studies conducted in Africa accounted for 7% of the articles (8), 7 of which were carried out in South Africa. Only 3 studies (3%) used samples from Oceania. Five studies analyzed data from different regions.

Sample size plays an important role in the power of traditional statistical tests (Cohen 1988). In more complex analysis such as SEM, sample size requirements go beyond power and should consider aspects such as bias or error (Wolf et al. 2013). Although Montecarlo simulations have demonstrated that sample size requirements depend on aspects such as complexity (Wolf et al. 2013), a commonly accepted rule of thumb for SEM is a 200 observation minimum (Kline 2011). This systematic review on soft skills found 17 (14%) studies with sample sizes greater than 1000 observations, whereas 59 analyzed data from over 200 observations. As mentioned below, most of these studies implemented traditional statistical tests despite having a large enough sample size and other conditions to apply more suitable analyses. Of note, half (8) of the studies with more than 1000 cases used secondary data, and 5 of these tested the psychometric properties of the measures. The rest of the studies used samples ranging from 18 to 652 participants (M = 220.46; SD = 154.07). Nearly 24% of all the studies analyzed used student samples. This percentage goes up to 50% in studies with sample sizes of over 1000 observations.

As for the metrics used by the studies included in this SR, the following are the frequencies in terms of raters: Most studies (90; 75%) only used self-reported metrics for participant traits (e.g. soft skills). Out of the 82 (69%) papers that adopted a cross-sectional design, 63 (53% of the total) used only self-reported scales. The authors of 24 (20%) administered self-reported metrics but included variables rated by a different profile of participants or used additional secondary data in their studies. The remaining articles (5) that did not use self-reporting, gathered data from external raters. In the latter case, some studies used the same external rater (e.g. Poitras et al. 2015), while others used different external raters (e.g. Zacca and Dayan 2018). It is important to bear in mind that the use of only one rater or the same measurement method for all variables in the study (e.g. only self-reported metrics) affects the validation of relationships between soft skills and concepts that are part of their nomological network. Only 20 (17% of the total) of the 90 studies using self-reported measures applied procedural or statistical remedies to common method bias (Podsakoff et al. 2003). This low percentage suggests that there is still much to do against known menaces to the external validity of studies, such as common method bias in the soft skills field.

Regarding the number of items used to operationalize the variables, most studies (96; 81%) used at least one multiple item metric. In this SR, when authors aggregated multiple items into composite variables, the decision was made to codify the papers as reports of studies using multiple item metrics, even if their authors did not report the use of that type of scale (e.g. Balcar 2016). Surprisingly, authors adopt different perspectives of the same metric. This is the case in some studies using the data previously collected for the REFLEX project. Whereas some authors used the corresponding items as single item metrics (Vila et al. 2014), other scholars used the items as indicators of latent variables (Quintana et al. 2014). As we emphasize below, if authors select multiple-item scales for a given study, suitable analyses should be used to test the hypotheses and the psychometric properties of the measures. The results of studies using multiple-item measures are not conclusive if authors do not provide evidence of the psychometric properties of the scales. This kind of omission is identifiable in several studies reviewed in this systematic review.

The fact that 81% of the studies used at least one multiple-item measure shows a tendency to use multiple item scales in the literature. The preeminence of multiple-item scales is constant in the groups of theories we identify in this systematic review (RVB 64%, capital-related theories 100%, leadership 80%, and learning 90%). We expected this as most variables studied in the behavioral sciences are assessed indirectly (Wolf et al. 2013). Surprisingly, one of the findings presented below indicates that not all studies using multiple item scales reported the psychometric properties of the metrics or used suitable analytical techniques. Failing to assess the psychometric properties of this type of scale goes against the validity of the studies. Whereas 76 studies out of the total studies reviewed reported measurement validity, the remaining 43 (25%) studies failed to assess either content or construct validity. Overall, 35 of those 42 papers did not assess scale reliability either. Conversely, 28 out of the 42 studies that failed to test validity evaluated construct reliability. However, these are different psychometric properties and reliability does not account for construct validity. Around 11% (14) of the sampled papers did not test any psychometric property of the metrics prior to testing their hypotheses, although 9 of them operationalized their variables using multiple item scales. Lack of validity and reliability tests while administering multiple items scales hampers study validity, as no evidence is provided regarding how and to what extent respondents understood the items proposed in a similar manner. Put differently, using multiple items scales and not evaluating validity and metric reliability leads to questions regarding the quality of information gathered and analyzed in the studies.

When according to theory and conceptualizations, variables are latent and multiple item measures are applied in research, scale validity and reliability are directly related to hypothesis testing. In such cases, prior and simultaneous analysis of psychometric properties should be conducted. In this systematic review on soft skills, we found that this was not the norm. Overall, 34 Studies used SEM to test hypotheses and only 1 adopted Rasch analysis to test the relationship of the scales’ items with latent variables. Only 1 study conducted a path analysis SEM, excluding the measurement model from the specified model (Mesly 2015). Four studies used CFA—a member of the SEM family—with the aim of validating skills or competencies, scales were included as using SEM to test hypotheses. Other authors used CFA to scale validation but did not employ SEM to validate the hypotheses of the studies. Therefore, 70% of the sample studies (84) used first generation analysis (e.g. regressions, ANOVA, t-tests) to validate their hypotheses even though half (42) of them tested construct validity and 54 evaluated scale reliability. This indicates that the variables analyzed in those studies where measured through the use of multiple-item scales. Now, traditional techniques such as regressions and variance analyses often imply the use of composite scores leaving aside measurement error (Marsh et al. 2014). This has negative implications for theory and concept development, and specifically, validation of the relationship of soft skills to related constructs (i.e. nomological networks). For instance, in the context of standard regression analyses, ignoring the independent variables’ measurement error affects the power of tests (Hair et al. 1999). It is worth mentioning that, despite accomplishing traditional sample requirements for conducting SEM and hence including error when testing their hypotheses, 36 out of the 59 studies with samples of more than 200 cases selected first generation analyses. Moreover, 22 of these 36 studies, while having the necessary sample size for SEM reported reliability, reported using multiple item metrics and therefore the suitableness of that analytical approach instead of first generation analyses. Now, both SEM and regression analysis have limitations in terms of their estimation of causal effects. Aspects such as reverse and reciprocal causation and omitted variables, can hinder causal inferences. Approaches borrowed from econometricians can help draw this type of inference. Recently, Maydeu-Olivares et al. (2020) proposed the instrumental variable approach to enhance causal inference in psychology. Instrumental variables have two conditions: they are correlated with the model’s exogenous variable and are not correlated with the disturbance of the outcome variable. In this SR, we found that only 1 study implemented this approach to draw soft skills-related causal inferences soft skills (Leoni 2014). In the next section of this SR, we present suggestions for future research related to the studies included in the review. It is worth mentioning that, besides the paper by Leoni (2014), no article mentioned instrumental variables as an alternative to enhance their causal inferences.

2.4 Theoretical and conceptual soft skills model

Based on the results of this SR, we propose a theoretical model to help understand how soft skills relate to other constructs such as competencies, abilities, personality and behavior. As mentioned before, soft skills scholars insist on studying antecedents and consequences of soft skills. For the sake of parsimony, the model depicted in Fig. 1 presents only some of the individual antecedents and consequences of soft skills in general terms.

Personality refers to an inheritable and stable set of individual traits. From the theoretical and empirical points of views, personality traits can be considered as exogenous variables in different psychological fields (Bollmann et al. 2019; Maydeu-Olivares et al. 2020). Hence, we posit that personality is the exogenous variable of the model where soft skills are encompassed within a broader concept of competencies. Research suggests that personality traits significantly contribute to competencies assessments made by psychologists (Heinsman et al. 2007). A previous meta-analysis showed that personality, combined with ability tests (i.e. verbal and numerical reasoning), predicts some supervisor rated data (Bartram 2005). Thus, the theoretical model proposed as a result of this systematic review includes ability as a second predictor of competencies. The same meta-analysis also showed that competencies predict some areas of job performance (Bartram 2005). The results of the meta-analysis are consistent with Spencer and Spencer’s (1993) flow model, one of the more cited frameworks in soft skills studies. Spencer and Spencer’s (1993) model posits that competencies mediate the effect of individual characteristics overall job performance.

The set of predictors and outcomes of competencies represented in the theoretical model is inspired in previous conceptual models and research results. Yet, as mentioned in this SR, confusion exists on where soft skills should be located with regard to competencies and other factors of their nomological network. As soft skills can be considered components of competencies, they would have the same set of predictors and outcomes as competencies. Following (cite) the definition of competencies, in the theoretical model, soft skills are depicted as parallel factors with respect to knowledge and motivation. Furthermore, the model proposes that there are two general types of soft skills.

The findings of this systematic review suggest that soft skills can be divided in two main types: Intra and interpersonal soft skills (Collet et al. 2015; Kechagias 2011; Laker and Powell 2011). Intrapersonal soft skills relate to abilities to manage oneself and include traits such as independence and self-actualisation (Dippenaar and Schaap 2017). Interpersonal soft skills are talents useful when relating to others. Examples of them are communication, cultural competence, conflict resolution, and facilitation (Warner 2020). As each soft skills is composed by a number of traits, soft skills should be conceptualized as latent variables. In turn, the types of soft skills, intra and interpersonal, can be thought of as second order factors or latent variables encompassing the corresponding group of latent soft skills variables. Put differently, soft skills are multidimensional constructs.

Organizational level variables are also included in the model in the name of context. Since in soft skills research the individual is the pertinent unit of analysis and mediators would represent the properties of the person (Baron and Kenny 1986), contextual variables are represented in the model as moderators instead of mediators. For instance, in experimental studies inquiring into the effectiveness of HR practices vs the absence of such treatment, context (e.g. participation in a training program) should be treated as a moderator. Clearly, the relationships between contextual-group level variables and individual aspects such as competencies and soft skills should be understood from a multilevel perspective.

To test specific parts of this soft skills competencies model, including moderation-mediation paths, we suggest using what back in the day Baron and Kenny (1986) named multi-item latent methods, also known as SEM. The use of SEM is implied in the model by depicting latent variables as ellipses. Neither the items and their associated loading and error, nor disturbances of the endogenous are represented in order to simplify the graphical representation of the model. Yet, as pointed out before, it is important to bear in mind that, in the behavioral sciences most variables are indirectly assessed and this implies measurement error (Wolf et al. 2013). However, context and employee behavior can also be represented as squares if measured as single item/observed variables (e.g. participation in soft skills training) (Fig. 4).

Some of the theories already used in previous studies on soft skills were the basis to map the relationships with other individual factors and contextual variables in the theoretical model. Learning theories and the contingency theory seemed to be more useful to develop hypotheses regarding the moderating effects of contextual variables on soft skills development and effects. For instance, these theories can help understand how context in the form of organizational practices and educational programs interact with soft skills to affect job performance. While traditional leadership theories have been criticized because they focus on the influence of leaders on followers and not on leadership processes per se (Uhl-Bien et al. 2014), learning theories focus on the development of individual variables (i.e. knowledge, learning and skills) and also allow explaining how context in terms of leadership processes—not only leaders’ characteristics- may foster or hinder those processes. Put differently, learning and contingency theories have the potential of supporting all the relationships portrayed in the model, while leadership, capital and RBV theories would only support the moderating effects displayed in the model. It is important to bear in mind that RBV and KBV focus on people as organizational resources and allow studying the relationship between managerial competencies and firm performance (see Table 4). Since soft skills are commonly defined as individual traits that may be present in employees working at different organizational levels, the model proposed in this systematic review portrays soft skills, their antecedents and consequences as individual variables and includes organizational resources and managers’ skills as contextual factors. In this way, the model can be applied to study managerial and non-managerial soft skills across the different levels of the organization.

Table 4 Summary of the most frequently used theories
Full size table

3 Discussion

In this paper, we present a systematic review aimed at understanding the current state of soft skills conceptualization and research. The research questions that guided this systematic review were: (a) how are soft skills conceptualized in the reviewed literature? (b) what are the main theories used in the study of soft skills?, and (c) methodologically, what are the main characteristics of those studies?. As a result of this systematic review, we found that theoretical dispersion and different conceptualizations characterize soft skills studies. This blurriness hinders the development of a promising field. A clear-cut and common conceptualization of main concepts through the use of adequate and sound theories and methods, fosters conceptual and theoretical development (Suddaby 2010). As a result of this systematic review, we provide a new conceptualization of soft skills. With this conceptualization, we also aim to differentiate soft skills from competencies and provide a common ground for future studies inquiring the relationship of the former with their antecedents, moderators, and consequences. By detecting methodological flaws in previous studies, our paper also shows opportunities to improve the operationalization of soft skills.

Our first research question allowed us to draw several conclusions. In this systematic review, we found that soft skills are components of a higher-order construct (i.e., Competencies). We were also able to identify two types of soft skills: intrapersonal and interpersonal skills, personal skills, and people skills. The development of competencies, including soft skills, seems to be necessary to improve job performance. The development of soft skills depends on the interaction between individual innate traits (i.e., personality and abilities) and contextual factors. In organizational settings, the interaction between soft skills and training programs appears to be vital to enhancing employee job performance.

Regarding our second research question, we found that the field could benefit from an explicit use of theories. In fact, we identified a large percentage of the studies that do not explicitly report the use of a theoretical framework (e.g. Ghani et al. 2018; Lakshminarayanan et al. 2016; Pujol-Jover et al., 2015). More than expanding the body of previous literature or creating a conceptual model, research should contribute to theory. If researchers do not state what theory they are using to propose their hypotheses, their studies lack any clear theoretical contribution.

We were also able to identify four main groups of theories used by scholars studying soft skills. Future soft skills studies willing to use the RBV theory should bear in mind that, despite being recognized as one of the theories that truly aims to create a strategic approach to talent management (Lewis and Heckman 2006, p. 145), it has been also criticized. We believe that the knowledge-based view is useful mostly for the study of hard skills, or the role they play in the development of soft skills, while a dynamic capabilities approach may have potential to be used in studies relating to up-skilling and re-skilling that individuals are in so much need of for improving their employability in a dynamic business scenario. We suggest that future studies focusing on the relationship between training programs aiming at developing competencies and firm performance may benefit from the use of human capital theory when attempting to understand the effect of the firms’ investment in programs aiming at improving individual-level soft skills. With regard to learning theories, we believe that such an approach would be of great value to undertake studies dealing with the effect of educational processes on the development of competencies, such as that carried out by Levant et al. (2016). In this sense, these theories may be also used to study the impact of organizational practices, especially those relating to training and career development, on individuals’ soft skills. Learning theories have the potential to enhance the understanding of the relationship between soft skills, its antecedents and consequences, and how contextual variables intervene in those relationships. After systematically reviewing the literature, we found that contingency theory appears to be underestimated. The influence of contextual variables as moderators requires further study. Together with learning theories, contingency theory could add to the understanding of soft skills at the individual level, while allowing also to advance knowledge about their relationship with organizational level outcomes.

Finally, our third research question led us to propose the following directions in terms of methodology. First, efforts should be made to increase sample sizes so psychometric properties of soft skills scales can be tested. Construct validity and reliability tests are important as by definition soft skills are closely related to the behavioral sciences. Generally speaking, soft skills are conceptualized as human-individual traits and need to be assessed using multi item metrics. Consequently, soft skills should be treated as latent variables and, thus, the psychometric properties of assessments should be proven prior to and during hypothesis testing. Failing to do so would hamper developments in SS theory and practice.

More soft skills studies using samples from different regions/countries will help to validate theories and constructs. Before applying soft skills theories and practices in other regions/countries, research needs to establish whether previous research findings are suitable for understanding and managing employee skills in non US/European countries. Future studies also need to control for profession, job and industrial sector. Multicultural studies are also needed to disentangle the factor structure of types and specific soft skills.

To tackle methodological menaces to soft skills theory and concept development such as CMB and other endogeneity issues, future studies on soft skills must put greater care into research design. Of note, only one research study out of the 119 studies adopted the instrumental variables approach to cope with the endogeneity of the specified model (Leoni 2014). This approach, taken from economics research, is being used in other business and management areas, as well as in psychology (Maydeu-Olivares et al. 2020). We believe that personality can also be used as an instrumental variable helping to detect endogeneity issues in soft skills models (Bollmann et al. 2019; Maydeu-Olivares et al. 2020).

According to the findings of this systematic review and our theoretical model, the types of soft skills -intra and interpersonal- are second order factors comprising specific skills. Analytical approaches such as Bi-factor models could help to test such factor structures. Researchers should also consider studying the antecedents and consequences by means of Experimental and longitudinal designs. Specifically, experimental designs will help to assess the impact of educational and organizational practices directed to the development of such skills. While longitudinal studies will provide evidence of how soft skills develop in time, and what are the antecedents and consequences of soft skills trajectories.

3.1 Limitations and future studies

This systematic review did not attempt to quantitatively test findings about the relationship of soft skills with related constructs such as competencies. Future studies willing to review prior soft skills quantitative findings should adopt a meta-analytic approach. Such reviews will help quantitatively validating the soft skills’ nomological network and typologies posited in this systematic review. In particular, future meta-analyses should aim to test some of the antecedents and consequences of soft skills. Our systematic review suggests what the main antecedents and some of the consequences of soft skills are. Meta-analyses can also allow us to inquire more profoundly about soft skills' intrapersonal and interpersonal components. Based on our findings, we propose a division of soft skills into these two dimensions. However, this was not the focus of our work, and we suggest that future meta-analyses should try to confirm this part of the soft skills nomological network. Such quantitative evidence can be provided, for example, by verifying if the antecedents and moderators of soft skills exert a different effect on intrapersonal and interpersonal skills and testing if these components have differentiated effects on various outcomes.

Fig. 1
figure 1

Definition of competency: most cited authors

Full size image
Fig. 2
figure 2

Groups of theories and frequency of use

Full size image
Fig. 3
figure 3

Frequency of research designs in soft-skills studies

Full size image
Fig. 4
figure 4

Theoretical and conceptual soft skills model

Full size image