In this study, we investigate the (in)direct relationship between learning orientation and firm performance. The study is guided by the DCs framework. We collected data from 182 MSMEs operating in TPs in Poland. We used two methods (PAPI, CAWI) in our quantitative empirical research.
For the analysis of empirical data, we used the methods of description and statistical inference. The values obtained by means of Cronbach’s alpha values showed very good reliability of questionnaire. We have assumed that the coefficient deciding whether a tool is reliable should be at least 0.70. The results of the Kolmogorov-Smirnov tests indicate grounds for assuming that the variables are not normally distributed. We present the results of the Kolmogorov–Smirnov tests and Cronbach’s alpha coefficients in Table 1.
In the next step, we applied the correlation analysis between the variables by using the rho-Spearman coefficient. We present the results of correlations among the analysed variables in Table 2. The analysis of data included in Table 2 indicated weak or very weak correlations among the variables in individual configurations. LO positively correlates with FP (rs = 0.197; p < 0.01). This means that the increase in LO is accompanied, on average, by a small increase in FP.
There is also a positive, although very weak (rs = 0.151) correlation between MD and LO, which was statistically significant (p < 0.05). This means that the increase in MD is accompanied by, on average, a slight increase in LO.
At the same time, the results of the correlation analysis indicated a weak but positive correlation between one of the dimensions of MD, i.e. speed of change in technology and competition and LO (rs = 0.0.236; p < 0.01). Relationships between the two remaining dimensions of MD were not statistically significant.
In addition, the aforementioned MD dimension also positively correlates with FP. The correlation between the MD dimension called speed of change in technology and competition and FP is positive, weak and statistically significant (rs = 0.181; p < 0.05). This means that the increase in the speed of change in technology and competition is accompanied by, on average, a slight increase in FP. Relationships between the two remaining dimensions of MD and FP were not statistically significant.
Correlation analysis encourages deeper recognition and understanding of LO-FP relationship in the context of MD.
We used linear regression models in order to verify the hypotheses, which allowed for a global assessment of relationships among all analysed variables.
The values of coefficients obtained for permanent effects in this model inform about how much the expected value of explanatory variable changes along with the unitary growth of a given predictor. The explanatory variable (predictor) is a variable in a statistical model (as well as in an econometric model) on the basis of which the response variable is calculated. In Model 1 there is one explanatory variable (LO); while in Model 2 there are two explanatory variables (LO, MD). The response variable is FP. The statistical significance of these coefficients was verified by a test based on the t statistics. For all the mentioned tests, p<0.05 indicated the statistical significance of the analysed relationships.
The assessment of the impact of LO on FP is dictated by the H.1 hypothesis verification.
While the assessment of the impact of dynamism of the market in which enterprises operate in explaining the impact of LO on FP is dictated by the H.2 hypothesis verification.
H.1: Learning orientation is positively related to firm performance.
H.2: Market dynamism moderates the learning orientation-firm performance relationship; the positive effect of learning orientation on firm performance is likely to be stronger under high market dynamism than under low market dynamism.
The results of testing the H1 and H2 hypotheses are presented in Table 3.
We estimated Models 1 and 2 in Table 3 by using the Akaike Information Criteria (AIC). The AIC for both models was similar, i.e. 568.28 for the first model and 571.12 for the second one. AIC levels for both models indicated acceptable matching levels. The lower the AIC value, the better the predictive values of the model. The model coefficient is a parameter determined by its most likely value. The confidence interval of the model coefficient indicates in which range its less probable but possible values may be. It also has a diagnostic value. If the value of the regression coefficient contains “0”, the coefficient has no substantive value for the model. Model 1 explained 13.5% of the data variation (R2 = 0.135), while Model 2 explained 14.0% of the data variation (R2 = 0.140), which is slightly more than Model 1. The analysis of the models presented in Table 3 leads to several findings. In the first model, only LO was positively related to FP and only slightly explained the variability of the dependent variable. It has a small but statistically significant impact on FP (coefficient: 0.38; p=0.00). The linear regression model (Model 1) confirms the thesis about the positive impact of LO on FP. It may be assumed that an increase in the assessment of LO by one point, with no change in the other parameters of the model, would result in an increase in average FP by 0.38. This model explains 13.5% of the data variability (R2 = 0.135). Secondly, the linear regression model (Model 2) did not confirm the thesis about the moderating role of MD on the LO-FP relationship. None of the predictors showed statistical significance (p<0.05) in Model 2. What is more, taking the MD variable into account affects the quality of the model, and MD itself adopts negative prediction indicators, which means that better FP in responding to changes in the level of MD deteriorates the overall FP. However, the research has not confirmed whether MD - a higher-order construct built of three first-order constructs, i.e. the speed of changes in technology and competition, unpredictability of changes in technology and competition, uncertainty of customer behaviour - increases the importance of LO for increasing FP, and thus achieving a competitive advantage. Thirdly, the control variables were insignificant in both models. This means that the control variables in the form of enterprise size do not have a statistically significant effect on the dependent variable. Therefore, the introduction of two control variables and a moderating variable reduced the impact of LO on FP to a statistically insignificant level.
The results of the study show that firm performance benefits from LO-related behaviours. Learning orientation is an important stimulant of firm performance, while market dynamism has not been classified as a moderator of the learning orientation-firm performance relationship.
