In this current knowledge-based economy, firms' productivity and competitive advantage are no longer based on physical and financial assets but on intangible assets. This has compelled knowledge-intensive firms to look for a more reliable source for higher productivity and competitive advantage by focusing on their intellectual capital, which cannot be easily imitated. As banks are classified as knowledge intensive, this study examines investment in intellectual capital by banks and examines how it has improved bank productivity measured in terms of asset turnover (ATO) and employee productivity (EP). Using a panel of 73 commercial banks in India for a 12-year period (2006e2017), the study found that some components of intellectual capital improves productivity, and others do not.
Hence, in order to ensure a valid statistical inference amid problems of heteroskedasticity and autocorrelation in our models, we relied on the Rogers (1993) clustered robust standard errors. The advantage of it produces heteroskedasticity consistent standard errors that are robust, which is appropriate for balanced panel data, as in our case.
id: 064de22f7438899794ea53c02ddef907 - page: 5
Checking for the normality of data is imperative in deciding which correlation matrix to apply. According to Brooks (2014), the normality assumption is also important for conducting single or joint hypothesis tests regarding the model parameters. The Shapiro Wilk test was applied, and since our data was not normally distributed, the Spearman correlation matrix is used to show the correlation among variables. Based on the suggestions of Kennedy (1985), a correlation coefcient of more than 0.8 shows the existence of multicollinearity, which is a serious problem. No evidence of high correlation between the explanatory variables was found (See Table S2, available online), except in the case of VAIC and HCE; however, this is still not a problem because the two are not included in the same equation in our study. We also conducted a multicollinearity analysis through the variance ination factor (VIF), and the results show no multicollinearity among our main dependent variables based on sugg
id: 4023a83531f584c7c68927f9d5c6add6 - page: 5
In panel data with series of more than 10 years, there is always the possibility of non-stationary shocks that will affect 5.3. Results of panel models
id: b5680e5457dbdbd3688fdd84e8e0e2d8 - page: 5
5.3.1. Panel regression results: IC components and EP Regression results on the contribution of various IC components are presented in Table 2, in which EP is the dependent variable. As per the results of the Hausman test, an FE model is presented in Table 2 for the full sample and public banks whereas private and foreign banks are analyzed based on the results of the RE model. According to our regression results, CEE is the only component of IC that has a positive and signicant coefcient on employee productivity for the full sample, which conrms H2a. This nding is in line with Bontis et al. (2015), whose studies conrmed a signicant association between CEE and employee productivity. However, both HCE and SCE are insignicant in inuencing productivity, hence both H1a and H3a are rejected. This therefore indicates that most banks in India their have still not realized the salient importance of 223 224 G.K. Oppong, J.K. Pattanayak / Borsa _Istanbul Review 19-3 (2019) 219e227
id: 6a477fbfd20e0276e96a9c3f3bfa8669 - page: 5