Introduction

As the 21^st century progresses, we can notice that the competitive level between organizations is increasing, which is due to current economic and demographic trends that derive from globalization and market opening based on technological advances (^{Martínez & Gallego, 2007}; ^{Ritter, 2008}). Thus, organizations need to increase their competitive value to excel in a rapidly-changing and challenging environment (^{Lip, 2005}; Martinez & Gallego, 2007). One way to increase the competitive advantage is through organizational learning (OL), as this allows adaptation to change and innovation. (^{Conde & Castañeda, 2014}; ^{López, Ahumada, Olivares & González, 2012}; Martínez & Gallego, 2007; ^{Osorio, 2003}; ^{Pomajambo, 2013}).

OL is a process that occurs within the organizations based on the interpretation of different knowledge allowing a transcendent learning (^{Inche, 2010}). ^{Del Río and Santisteban (2011)} mention that OL is experiential and should take into account the organization's stakeholders and management processes.

OL is important because it is crucial in economic development and increased productivity by obtaining intellectual capital, which together enable a competitive advantage (^{Armenteros, Guerrero, Noyola & Molina, 2012}; ^{García Zapata, 2005}; ^{Osorio, 2003}). Also, OL can occur at an individual, group, or organizational level; and it serves as a resource to develop the competencies expected by stakeholders (^{Riquelme, Cravero & Saavedra, 2008}). In addition, levels of OL also refer to the context in which the processes required for the company to learn are carried out; in other words, from the individual, group, or organizational approach different ideas are generated, which contribute to a more innovative and effective decision-making. (^{Del Rio & Santisteban, 2011}, ^{Garzón & Fisher, 2010}). On the other hand, it is important that organizations implement the conditions that enable the development of OL since such learning does not take place naturally (^{Alcover & Gil, 2002}; Garzón & Fisher, 2008; ^{Milian, Lugo & Cespón, 2008}).

In this regard, ^{Castañeda and Fernandez (2007)} attribute three conditions for OL to develop at all levels: organizational culture, training, and strategic clarity. In short, OL is important to generate competitive advantage, but it is necessary that organizations implement appropriate actions for its development, individually, as a group, and at an organizational level.

In that sense, OL measurement is relevant because it allows companies to obtain good results through the acquisition of new knowledge, which becomes a competitive advantage (^{Riquelme et al., 2008}). Measurement of OL should be implemented because it allows to generate measurable indicators of the intellectual capital within organizations and the extent of determining factors in the success of the business (^{Castellano, 2013}; ^{García Zapata, 2005}). In this regard, ^{Osorio (2003)} mentions that while intellectual capital is the set of intangible assets in an organization, it must be studied because it creates potential value in companies. Previous studies on OL agree that a company’s workers must have an added value through the development not only of knowledge and technical skills, but also of social and emotional ones that enable them to adapt to globalization (^{Calderón, Alvarez & Naranjo, 2011}; Calderón, Naranjo & Alvarez, 2010; Castellano, 2013; ^{Ospina, 2010}; ^{Pesca de Acosta, 2009}).

In this regard, in Latin America the measurement evidence of this construct has gained recent interest from its conceptual analysis (^{Castañeda, 2004}; ^{Del Rio & Santisteban, 2011}, ^{Garzón & Fisher, 2008}) to the development of measurement instruments (Castañeda & Fernandez, 2007). OL measurement, therefore, is relevant because it allows the measurement of intangibles present in an organization that work as impact value in the financial development of it (^{Villegas, Hernandez & Salazar, 2016}). To that end, the Organizational-Learning Levels and Conditions Scale was developed in Colombia (Castañeda & Fernandez, 2007) and has been used in studies in Colombia and Chile (^{Conde & Castañeda, 2014}; ^{Lopez et al., 2012}).

The first study to report evidence of validity and reliability of the ENCAO is the construction and validation of it (^{Castañeda & Fernandez, 2007}). In this study, the instrument was applied to 845 participants (61% men), from different socioeconomic and educational levels. A previous pilot study was conducted with students from a private university in Colombia applying the expert judge criterion which revealed content-related evidence of validity. Also, an analysis of the main components with Varimax rotation was conducted, this explained 59.51% of the variance and extracted six components for the 28 items that were drafted: (a) Group Learning (7 items), (b) Culture (5 items), (c) Training (7 items), (d) Organizational Learning (5 items), (e) Individual Learning (2 items), and (f) Strategic Clarity (2 items). With regard to reliability, the authors report a .92 Cronbach's alpha for the total scale, without reporting this coefficient for the six dimensions extracted. In a subsequent main component analysis performed in the same study, the authors report the same solution of six dimensions without considering seven items that showed .35 factorial cross-loadings in more than one component, with a final version of 21 items as a result.

In a further study, ^{Conde and Castañeda (2014)} sought to get evidence of validity and reliability of the ENCAO adding one more condition (Support) to the original instrument to adapt it to a university academic context. The sample consisted of 111 investigations qualified by research system in Colombia (COLCIENCIAS). Data were collected from 82 research groups segmented by categories A= 3 groups, B= 17 groups, C= 22 groups, and D= 40 groups (where A top category and D minimum category; COLCIENCIAS, 2015). A peer review was performed with 6 experts exploring clarity, relevance, and grammar, shortening the original 28-item instrument to 23 items which are graded by a 5-level Likert-type scale (0 = never, 4 = Very often). The reliability of each of the subscales was calculated using Cronbach's alpha obtaining .56 for Individual Learning (IL); .85 for Group Learning (GL); .80 for Organizational Learning (OL); .70 for Organizational Learning Culture (OLC); .70 for Training; .60 for Organizational Support; .85 for Clarity, and .79 for the global scale. This study exhibits two alpha coefficients below the acceptable minimum (α < .70, ^{Prieto & Delgado, 2010}); additionally, no report on the analysis of the internal structure of the ENCAO in the researchers' sample is evidenced.

Finally, ^{Castañeda (2015)} conducted a study whose methodological basis was the validation of an instrument on the conditions of OL (Castañeda, 2010; Castañeda & Fernandez, 2007) aiming to explore the relevance of these conditions from the modification of the original instrument. To do this, he used a convenience sample of different companies (N = 613; 58% men), divided into three groups: public (n = 304), commercial (n = 201), and advertising (n = 108). The instrument achieved content validity through a pilot study with 36 graduate students and expert opinion.

The instrument proposed in the latter study (^{Castañeda, 2015}) consists of 28 items and 4 condition components: Organizational Learning Culture (OLC, items 16 to 19), Training (items 20-23), Strategic Clarity (items 5- 9), and Organizational Support (items 24-28). In addition, the Organizational Learning Scale consists of 10 items measuring the following levels of learning: Individual (IL) with 4 items; Group (GL) with 3 items, and Organizational (OL) with 3 items. To detect evidence of validity, a factorial analysis of main components with Varimax rotation was performed. Therefore, a significant factorial loading was considered when it was greater than .50 and it was decided to remove item 1 of Strategic clarity, items 3 and 4 of Training, and item 5 of Organizational Support. Furthermore, a .89 KMO was obtained, Bartlett's test p < .001 and a cumulative explained variance of 69.5% was obtained. The author reports a global mean reliability with a .92 Cronbach's alpha. Regarding the components: OLC.81; Training .78; Strategic Clarity .81, Organizational Support .81; IL.56; GL .87, and OL .80.

According to previous research mentioned it can be expressed that the instrument to be used in the study has had variations in the distribution of components in different contexts, which is evidence of good practice in measuring by on the researchers' part because evidence of validity and reliability adapted to different contexts are reported (^{Muñiz, Elosua & Hambleton, 2013}), although it reflects also the lack of a defined factor structure. The studies mentioned also show analysis procedures not currently methodologically recommended regarding the analysis of its psychometric properties. For example, the use of the package known as Little Jiffy (main component analysis, Kaiser criterion, and Varimax rotation) to analyze the internal structure of an instrument (^{Lloret-Segura, Ferreres-Traver, Hernandez, & Thomas-Miguel, 2014}), and the report of a total alpha when it is not a one-dimensional instrument. Moreover, the review of psychometric history of the ENCAO reflects that the learning conditions and levels gain consensus in a theoretical framework, and not necessarily on a practical level. While it is true that all scales above present conditions and levels, these vary according to the population and study objectives (^{Castañeda, 2004}; ^{Conde & Castañeda, 2014}; Castañeda & Fernandez, 2007). For this reason, it is advisable and important to continue reporting psychometric evidence in different contexts in order to assess the consistency of the instrument as well as its dimensionality.

That is why the aim of this study is to analyze the psychometric properties of the ENCAO, specifically by reporting validity evidence related to the internal structure and external evidence of validity by association with another variable (job satisfaction); as well as evidence of reliability by the internal consistency method with Cronbach's alpha. Importantly, this research is of instrumental type, i.e., it seeks to analyze the psychometric properties of psychological measurement instruments (^{Ato, López & Benavente, 2013}), in this case within the organizational setting.

Method

Results

Next, the results of the data analysis are presented according to the evidence of validity (internal structure and relationships with other variables) and reliability (internal consistency) obtained.

In the first place, the internal structure of the ENCAO (^{Castañeda & Fernández, 2007}) was analyzed. Initially, the homogeneity of the items was analyzed by corrected item-test correlations (r _itc), for which values greater than .20 were retained (^{Kline, 1986}); the range of the r _itc of the ENCAO items was between .30 and .72; that is why all the items were maintained for the factorial analysis. In addition, the statistics of central tendency and dispersion of the items were calculated (See Table 1) where no violations of the assumption of univariate normality were identified, since the skewness and kurtosis indices were lower than +/- 1.5 (^{Forero, Maydeu-Olivares & Gallardo-Pujol, 2009}). However, in the ^{Mardia (1970)} analysis for skewness and multivariate kurtosis, we found a skewness coefficient of 51.34, df= 1771, p = 1.0, and a kurtosis coefficient of 572.76, p <.001, demonstrating the absence of a multivariate normal distribution of the data; for this reason the method of extraction of Unweighted Least Squares (ULS) and the matrix of polychoric correlations were used, which turn out to be an option more in line with the ordinal level of the items and against the violation of the multivariate normality assumption (Lloret -Segura et al., 2014).

Table 1 Factorial loadings of the items of the Organizational-Learning Levels and Conditions Scale (ENCAO). 

Note: OTC = Organizational Training and Culture; OL = Organizational Learning; GL = Group Learning; ISL = Individual Strategic Learning.

Prior to the application of the EFA, a .92 measure of sample adequacy (KMO) was found, which is considered excellent (^{Ferrando & Anguiano-Carrasco, 2010}). In addition, the Bartlett Sphericity Test obtained a statistically significant score, x² (210) = 3213.80, p <.001, which means that there is a sufficient degree of relationship between the items of the instrument to be able to analyze them factorially.

The results of the extraction analysis of Unweighted Least Squares on the polychoric matrix by the criterion of eigenvalues greater than one, suggested extracting four factors that explained 60.33% of the total variance for the 21 items according to the suggestion of the original instrument (^{Castañeda & Fernández, 2007}). Likewise, the estimates of the communalities extracted for the items of the scale were obtained (See Table 1). And, regarding this, it is important to analyze the communalities since the low values (h ² <.20, ^{Child, 2006}; h ² <.30, ^{Costello & Osborne, 2005}) can generate a significant alteration in the results of the factorial analysis (^{Velicer & Fava, 1998}). In general, the results showed moderate communalities in most of the items (mean h ² = .53).

To obtain an interpretable solution, an oblique Promin rotation was performed, assuming that the extracted factors are correlated (See Table 1). The first component is the so-called Organizational Training and Culture (OTC), in which 2 items of the Culture dimension have been adjusted (items 19 and 21); the 4 items of the Training dimension (items 24, 26, 27, and 28), and one item of the Organizational Learning dimension (item 10), presenting factorial loadings between .36 and .88, which are considered as acceptable factorial loadings (> .30) (McDonald as quoted in ^{Ferrando & Anguiano-Carrasco, 2010}) . In the second factor, called Organizational Learning (OL), 3 items of the original dimension were grouped (items 12, 13, and 14), and 3 items of the Culture dimension (items 15, 17, and 18) presenting factorial loadings between .31 and .89. The third factor is called Group Learning (GL), in which the 4 items of the original dimension (items 7, 6, 8, and 9) have been adjusted with factor loadings of .58 to .91. The fourth factor called Individual Strategic Learning (ISL) presents an adjustment of the items of the Strategic Clarity dimension (items 3 and 4) and of the Individual Learning dimension (items 1 and 2) with .32 and .86 factor loadings. It is worth mentioning that item 18 presented a cross-loading in the OTC and OL factors.

To calculate reliability, the internal consistency method was used by calculating Cronbach's alpha coefficient. Next, the Cronbach alpha coefficients of the four subscales were reported along with their respective confidence intervals (^{Fischer, 1992}). OTC obtained a Cronbach's alpha equal to .84 (CI, 95% = .81, .87) and the corrected item-total correlation range was .53 to .68, being acceptable values (^{Pérez & Medrano, 2010}). OL obtained a Cronbach's alpha equal to .83 (CI, 95% = .80, .86) and the corrected total correlation range was .51 to .70. For GL, a Cronbach's alpha equal to .80 (CI, 95% = .76, .83) was obtained and the corrected total correlation range was .60 to .65. Finally, ISL obtained a Cronbach's alpha equal to .60 (CI, 95% = .53, .66) and the corrected total correlation range was .33 to .41.

To confirm the structure found in the EFA, we proceeded with a CFA considering the estimation method for Robust Maximum Likelihood and considering the ordinal level of the items by analyzing the matrix of polychoric correlations between the items. In this analysis, the adjustment of 3 measurement models was evaluated: the structure of 6 factors related to the 28 items of the first version of the ENCAO (M₁, ^{Castañeda & Fernández, 2007}), the structure of 6 factors related to 21 items of the final version of the ENCAO (M₂, Castañeda & Fernández, 2007) and the structure of 4 factors related to 21 items (M₃) of the EFA result previously conducted in this research. For this analysis, the total sample used for the EFA was divided into two random subsamples of 194 participants each, as currently recommended (^{Lloret Segura et al., 2014}). The first subsample was used to evaluate the fit of the M₁ and M₂ models, while the second subsample was used to evaluate the M₃ fit.

As shown in Table 2, according to the results of the AFC and to the evaluation of the adjustment indices, the model of four factors related to 21 items (M₃) presented the best fit; although it is worth mentioning that the model of 6 factors related to 21 items (M₂) also presented an acceptable adjustment.

Table 2 Index of adjustment of the Organizational-Learning Levels and Conditions Scale (ENCAO). 

Note: M₁ = 6 factors related (28 items); M₂ = 6 factors related (21 items); M = 4 factors related (21 items).

Then, a normality analysis was performed through the Kolmogorov-Smirnov (KS) test, taking as reference the sample size (N> 50). It was found that none of the factors presents a normal distribution (p <.05) and that is why it was determined to use the nonparametric Spearman test for the convergent validity analysis. Thus, the analysis of correlations between factors was carried out in order to analyze the degree of intensity between the existing relationships between factors of the proposed model and the Job Satisfaction scale (ESL, ^{Boluarte & Merino, 2015}) (See Table 3).

Table 3 Correlations and Descriptive Statistics of the Organizational-Learning Levels and Conditions Scale (ENCAO). 

Note. ^** p < .01.

As for the descriptive analysis of the factors, whose scores were calculated from the sum of items divided by the number of items of the factor, it is observed that GL reaches the highest score versus the other subscales (See Table 3), although, broadly speaking, in the 4 dimensions there is a tendency toward high scores (Minimum = 1 and Maximum = 5). Regarding the relationships between the factors of the proposed model, it was found that all are positive and statistically significant (p <.05) with a moderate magnitude (> .30) (^{Hirzel & Guisan, 2002}; ^{Taylor, 1990}). On the other hand, the relationships between the factors of the proposed model and the job satisfaction factor are positive and statistically significant, taking into account that the most direct association occurs between job satisfaction and organizational training (r_s = .46, p <.001) and organizational learning (r_s = .46, p <.001), while the lowest is with individual learning (r_s = .31 p <.001). In general, all the correlations account for the convergence of the ENCAO dimensions and its relation with the job satisfaction variable according to what is expected.

Discussion

The aim of this study was to analyze the psychometric properties of the Organizational Learning Levels and conditions Scale (ENCAO; ^{Castañeda & Fernandez, 2007}), to warrant its use in the business setting in Lima Peru.

The final version has four factors with reliability coefficients between .60 and .84, which is an indicator of the reliability of scores in measuring the OL construct. This factorial model fits in the sample of private sector workers in the hydrocarbons sector despite the reduction of factors comprising it. The exploratory factor analysis was performed based on the polychoric correlations matrix and through the unweighted least squares analysis criterion determining factor reduction from the original model to four and with the 21 items recommended in the study by ^{Castañeda and Fernández (2007)}.

Importantly, they obtained a Cronbach's alpha between .79 and .98 in most previous studies based on the original instrument. (^{Lopez et al., 2012}; ^{Castañeda, 2015}; ^{Conde & Castañeda, 2014}), which suggests that, despite the change in the distribution of factors, these remain highly acceptable. Although low coefficients were also observed in some theoretical dimensions analyzed (Castañeda, 2015). Differences in reliability coefficients found in this study as in those mentioned above are due to the particular characteristics of each sample, as well as to the modifications undergone by the ENCAO in various studies (Castañeda, 2015; Conde & Castañeda , 2014) and to the formation of factors with a low number of items (2 items, for example), to the absence of an analysis of the internal structure prior to calculation of the alpha coefficients (Castañeda, 2015).

In the proposed model, Organizational Training and Culture (OTC) is the one explaining more variance (42.90%). This result differs from previous studies where it is evident that Group Learning (GL) explains an 80.91% index of the variable study (^{Conde & Castañeda, 2014}). On the other hand, the results found in this study are consistent with the original validation (Castañeda & Fernandez, 2007), where Organizational Learning Culture was the factor that explained a higher percentage of variance. Further, ^{López et al. (2012)} also found that the Learning Culture factor explains 64.73% of the variance. Which is consistent with the proposed model because OTC is understood as the development of shared cognitive skills in collaborators thus allowing to achieve the organizational goals and these levels to occur at different levels (^{Pucci, 2004}). In addition, the fact that both conditions of OL have come together in this sample reflects the fact that consolidation of the culture of an organization stems from a gradual training process of workers at all levels (^{Rodriguez, 2009}; ^{Sánchez, Tejero, Yurrebaso & Lanero, 2006}).

Furthermore, the confirmatory factor analysis shows that the structure of four related factors (21 items) shows the best fit to the data than the first version of the original instrument (6 factors: 28 items). That said, the proposed model allows to evaluate: (a) OTC described above; (b) Organizational Learning (OL), understood as the way workers attribute meaning to their work experiences from the social context developed in the organization (^{Martinez & Gallego, 2007}); (c) Group Learning (GL) as the acquisition of knowledge collectively where interpretation is shared among employees and in turn, promotes integration between them (^{McAnally-Salas & Sandoval, 2007}); and (d) Individual Strategic Learning (ISL) which refers to individual mental models created and integrated collectively into the organization for alignment with the organizational goals and strategies. The latter factor matches the original theoretical proposal (^{Castañeda & Fernandez, 2007}) where it was assumed that the four items comprising it explore individual aspects of organizational learning.

This four-factor model also constitutes an instrumental advantage as it allows to study OL from more specific constructs that enable companies to identify needs among the collaborator-organization link to establish action plans so as to ensure the competitive advantage by that organization. Similarly, it considers the three levels of OL proposed by ^{Castañeda and Fernandez (2007)}. Also, the four-factor model validated in this study takes the simpler structure criterion that exceeds three items per factor, which is a robust approach when the factorial analysis is developed in measuring instruments (^{Fleming & Merino, 2005}; Lloret -Segura et al., 2014). However, it should be noted that as the model with six factors presented an acceptable fit, it would be required to expand the number of items for the IL and CE dimensions in order to assess this model's fit more precisely and compare it with the structure found int his study, which was also a recommendation posed by Castañeda and Fernández (207).

Finally, it should be emphasized that the convergent validity analysis corroborates the relationship between the four factors of organizational learning with job satisfaction, showing a highly-significant and direct relationship relative to the sample under study. Additionally, these correlations have been from moderate to high indicating that the associations found have been relevant.