Concept and measurement of the absorptive capacity

In its original design, the AC is understood as the ability of an external company to recognize the value of new information to assimilate it and apply it for commercial purposes ( ^{Cohen & Levinthal, 1990} ). Subsequent refinements have led to a concept of AC with four components: acquire, assimilate, transform, and use the knowledge to obtain a dynamic organizational capacity ( ^{Zahra & George, 2002} ). In this sense, said authors define AC as a set of routines and processes. Furthermore, the four components are grouped into two big dimensions: potential capacity (acquisition and assimilation) and achieved capacity (transformation and use).

Regarding the way of measuring or quantifying the AC, scientific literature addresses several approaches. In some works, the AC has been measured through scales built ad hoc, through theoretical reviews, and statistical analyses. For example, scales have been created for the variables linked to the four theoretical dimensions: acquisition, assimilation, transformation and use ( ^{Camisón & Forés, 2010}; ^{Jansen, van den Bosch, & Volberda, 2005}; ^{Jiménez, García, & Molina, 2011} ). Some works even propose new dimensions or constructs, and design the corresponding scale to measure it. An example of this is ^{Lichtenthaler (2009}) , who elaborates a scale to measure the AC from three dimensions (explore, transform, and use) as well as six constructs, two for each dimension accordingly (recognize and assimilate; maintain and reactivate; transform and apply). Similarly, ^{Vega-Jurado, Gutiérrez-Gracia, and Fernández-de-Lucio (2008)} propose a scale to quantify a set of determinants or precedents of the AC classified into three dimensions: organizational knowledge, formalization, and social integration mechanisms.

On the other hand, there is a series of proposals to measure the AC through simpler, one-dimensional character ( proxy ) indicators, such as research and development expenses (R&D) ( ^{Rothaermel & Alexandre, 2009} ); expenses in R&D and experience of the employees ( ^{Grimpe & Sofka, 2009} ); as well as the expense and intensity in R&D ( ^{De Jong & Freel, 2010}).

Another approach has been to measure the AC through a combination of variables searching for a kind of index, i.e., internal expense in R&D, the presence of permanent R&D, the training of the R&D personnel, and the percentage of scientists and researchers of the organization ( ^{Escribano, Fosfuri, & Tribó, 2009} ). By the same token, ^{Murovec and Prodan (2009)} measured the AC through a bi-dimensional construct according to the source of information: scientific and market-based, used by the company. For their part, ^{Monge-González, Rodríguez-Álvarez, and Leiva (2015)} measure the AC for the micro, small and medium sized Costa Rican companies through an index created from variables that emerged from a survey, grouping said variables according to the determining groups proposed by ^{Vega-Jurado et al. (2008)} . In the methodological section, the index that was used for this article will be addressed in greater detail.

Performance and absorptive capacity of the companies

The AC has been linked to some differences in the performance of the companies, both directly and indirectly through other variables. This means that it operates as a mediator variable.

The main exponents of the AC have theoretically outlined a linear relation between it and the performance of the companies ( ^{Cohen & Levinthal, 1990}; ^{Zahra & George, 2002} ). In the most recent empirical works, the AC is associated with business performance, but from a perspective that sees the AC as a mediator variable; more concretely, between: the internationalization activities and the financial performance ( ^{Zahra & Hayton, 2008} ); the relationship learning and the innovative performance ( ^{Chen, Lin, & Chang, 2009} ); the mix of technology sources (new vs known technology) and the financial performance ( ^{Rothaermel & Alexandre, 2009} ); as well as between the strategic alliances and the innovative and financial performance ( ^{George, Zahra, Wheatley, & Khan, 2001} ).

In an interesting piece of work, ^{Wales et al. (2013)} find a curvilinear relation (in an inverted U-shape) between the AC and the financial performance, reaching the maximum level of financial performance when the AC has an intermediate level. Furthermore, its effect was moderated by another variable denominated entrepreneurial orientation. To reach these results, said authors carried out an investigation between 285 small and medium sized Swedish enterprises (SMEs). In their reasoning, the authors assert that this curvilinear relation can be explained through the incremental growth of the costs associated with the increase of the AC in the enterprises. In other words, from a certain moment, the marginal cost is higher than the marginal benefit of increasing the AC.

Factors affecting the absorptive capacity

Seen from another perspective, the AC as such is influenced by different variables in the business sector. That is, the focus is not on the relation of the AC with the performance, but rather on what factors influence the AC of the organizations.

Within this approach, ^{Vega-Jurado et al. (2008)} propose that the AC is determined by three types of variables: organizational knowledge, formalization, and social integration mechanisms. In their proposal, said authors argue that these three factors combine with the relevant external knowledge, resulting in a specific level of AC. On the other hand, ^{Flor, Oltra, and García (2011)} find that the AC varies according to the type of strategy used by the enterprise. For this, the ^{Miles and Snow typology (1978)} is used, finding differences between the absorptive capacity (differentiated between potential and accomplished) and the types of strategy of the enterprises being studied (exploration, defender, analyzer, and reactive). For his part, ^{Jansen et al. (2005)} link the AC with the organizational coordination, systems, and socialization capacities; whereas ^{Liao, Fei, and Chen (2007)} link it to the interaction that the employees have with their own knowledge and with that of their colleagues in the company (knowledge sharing).

Multinational companies in the local economy and the absorptive capacity of local companies

A relatively unexplored line and in which the interest of our work lies, is whether the AC of the local companies can be positively affected by the hiring of former MNC employees.

As we know, when an MNC operates in a certain country it can become a source of knowledge for the local business community. This knowledge is transferred from the MNC to the local companies in two ways: through knowledge transfer mechanisms (commercial agreements, alliances, exchange of better practices, access to technology, etc.) as well as knowledge spillovers ( ^{Spencer, 2008} ). The difference is that, the MNC does not receive any type of compensation for the knowledge spillovers, turning this into a positive external factor for the host economy ( ^{Görg & Strobl, 2005}; ^{Javorcik, 2004}; ^{Smeets, 2008}; ^{Zhang et al., 2010} ).

This theoretical relation between the knowledge spillovers from the MNC to the local companies has been empirically studied with different results. Some works ( ^{Blomström, 1986}; ^{Buckley, Clegg, & Wang, 2007} ; ^{Tian, 2007}; ^{Wei & Liu, 2006} ) found positive results (for example, in terms of productivity growth in the companies of the host country) whereas others show negative results ( ^{Aitken & Harrison, 1999}; ^{Feinberg & Majumdar, 2001} ).

Looking to understand this diversity of results, ^{Smeets (2008)} indicates the importance of fully understanding and modeling the way in which the channels through which said knowledge spillovers can occur operate. Citing ^{Görg and Strobl (2005)} , they indicate that in several cases the researchers only worry about finding a statistical relation (positive or negative) between the performance of the local companies and the existence of MNC in the host country, which implicitly supposes that the knowledge spillovers happen within a “black box”.

A recent study by ^{Farole and Winkler (2014)} presents an interesting conceptual framework based on a previous work by ^{Paus and Gallagher (2008)} to explain the dynamic of knowledge spillovers from the MNC to local companies. This framework identifies the factors that act as mediators for the existence of knowledge spillovers associated with the operation of MNC in a host country.

According to these authors, the determining elements for the appearance of knowledge spillovers can be grouped into three categories. In the first place, the characteristics of multinational corporations , such as the reasons for the establishment of operations in the host country, the production and global supply strategies, input models, as well as the time spent in the host country. Secondly, the capacity of national companies to absorb the knowledge and the technology from the multinational corporations ( absorptive capacity ) through the three channels described by ^{Saggi (2002)}. Finally, the factors and the institutional framework of the host country, such as the regulation of the labor market, copyright, the access to financing, and the infrastructure for learning and innovation. An important aspect of this conceptual framework is that the set of factors that act as mediators for the spillover of knowledge from multinationals to local companies is dynamic and not stationary. Its composition may change with the passage of time.

In short, both the work by ^{Farole and Winkler (2014)} as well as other previous works ( ^{Lim, 2001}; ^{Paus & Gallagher, 2008} ) indicate the importance of taking into consideration the fact that the learning relation between local companies and MNC depends, among other things, on the AC of the former. In other words, the AC influences the use that the local company gives to the knowledge obtained from the MNC. Our study considers this aspect, but it tries to answer a question that has yet to be answered: how does the hiring of former employees of multinational corporations influence the AC of local companies? The answer to this question is a fundamental element for the understanding of the dynamic of knowledge between the MNC and the local business conglomerate.

Sample

The data used comes from the survey that originated the second national study of the MSMEs carried out by the Observatorio Costarricense de las Mipymes (Omipymes) (^{Gómez, 2012} ). In this study, 117 micro, small, and medium sized Costa Rican¹ companies were interviewed via telephone through a probabilistic sample layered by the size and productive sector of the company (agriculture and fisheries; commerce; services; and industry). The maximum sampling error in the estimation of proportions, with a level of confidence of 95%, is of ±2.9 percentage points.

Sources of secondary and variable data

In this study, the variable of interest was the AC. For this, we used the absorptive capacity index (ACI) estimated by ^{Monge-González et al. (2015)} for micro, small and medium sized Costa Rican enterprises, such as the dependent variable. Said ACI was elaborated based on the approach of ^{Vega-Jurado et al. (2008)} , proposing an index comprised by three factors: organizational knowledge, formalization, and social integration mechanisms. The other variables included in the model are the following: (1) whether the local MSME has employees that have previously worked with a multinational company established in Costa Rica; and (2) the number of employees of the local MSME. Similarly, the sector in which the MSMEs operated was included in the modeling. The analysis of the variables was done through the model explained below.

Statistical analysis of data

To explain the behavior of the ACI, an equation that attempts to measure the effect of the hiring of former MNC employees by local companies on the absorptive capacity of the latter is created.

(1)

In the previous model: the i sub-index indicates the i th company and varies from 1 to 1167; exL _i is a variable that takes the value of 1 if the i th has former MNC employees and a value of zero otherwise; L _i is the number of employees of the ith company; and S _ij are three variables that take the value of 1 if the company belongs to the j th sector and a value of zero otherwise. The j = 1 sub-index indicates the service sector, j = 2 commerce, j = 3 manufacturing industry and j = 4 the agriculture and fisheries sector, which, given that it is not taken into consideration in the model, indicates that it is the reference sector. Random disruption is represented by ε _i , which is considered independently, identically, normally distributed, with a zero median and constant variance.

In Eq. (1) the interest coefficient is β ₁ , which measures the effect on the AC of the local company when hiring former employees of MNC. In general terms, it measures how many percentage points the index of absorptive capacity of the local company increases or decreases by when hiring former employees of MNC. Where the premise is that the knowledge previously acquired by these workers while working at an MNC can be used in their new positions in the local company and, therefore, positively affecting the performance of the latter.

Eq. (1) is estimated through Ordinary Least Squares (OLS). It is worth mentioning that no more control variables were included, given that for the construction of the CI several characteristics of the local companies were taken into consideration, which, were they to be included in the model, would invalidate it as they would be present on both sides of the equation.

Several statistical tests were done to verify that the assumptions of the regression model with OLS were fulfilled.

For each one of the models estimated, the assumptions were verified graphically. Histograms of the residuals were obtained to verify the normality of the assumptions, as well as the graph of normal probability. The graph of residuals against assumptions was used to verify the independence of errors, the constant variance, and the zero median. This led us to conclude that the estimates of the parameters of the models, as well as their standard errors, have all the characteristics provided by the ordinary least squares.