«Der Open-Access-Publikationsserver der ZBW – Leibniz-Informationszentrum Wirtschaft The Open Access Publication Server of the ZBW – Leibniz ...»
In addition to a positive correlation between productivity and ICT, there are also correlations between the intensity of ICT and ﬁrm heterogeneity. If ﬁrms use ICT intensively, they tend to deviate more from the median sector productivity (i.e. evidence more heterogeneity). Firm heterogeneity is also correlated to other ﬁrm characteristics. For instance, innovation tend to be positively correlated with ﬁrm heterogeneity.
The sampled ﬁrms are classiﬁed into 14 industries. The banking and insurance sector was excluded from the analysis given the diﬃculties measuring sales. Firms with no obvious classiﬁcation were also excluded.
3.2 Empirical Strategy
As mentioned before, the measure of ﬁrm heterogeneity at the ﬁrm level is constructed as the absolute value of the deviation of a given ﬁrm labor productivity, with respect to the median labor productivity in its corresponding industry. Moreover, ICT intensity is captured by the percentage of employees that work mainly with a PC (PCW). A set of dichotomous variables showing the adoption of diﬀerent ICT applications such as enterprise resource planning systems (ERP), supply chain management (SCM) and customer relationship management software (CRM) is also available.
Additionally, the data includes information on the innovative inputs (i.e. R&D) and outputs (i.e. innovations introduced to the market) of the sampled ﬁrms. In the present analysis, the fraction of total employees that is engaged in R&D activities is used as a measure of the incentives to invest in R&D personnel (i.e. innovative activity). Additional general information at the ﬁrm and sectoral level is also included.
The empirical analysis follows a two stage strategy. In the ﬁrst stage the direct eﬀect of ICT on ﬁrm heterogeneity is estimated. This stage is implemented using OLS regressions, although the results are robust to diﬀerent methods. In the second stage, the analysis follows an instrumental variables approach, where ﬁrm heterogeneity is instrumented by diﬀerent measures of ICT use (providing a measure of ICT induced heterogeneity) and then related with the ﬁrms’ R&D incentives.
Equations (1) and (2) describe the general relationship between ICT, ﬁrm heterogeneity (H ) and innovation incentives (I ) and highlight the two stage instrumental variable strategy.
The key identiﬁcation assumption corresponds to the exclusion of ICT from equation (2).
As discussed in the introduction, ICT generates ﬁrm heterogeneity, which in turn aﬀect ﬁrm strategies. Therefore, it is assumed that ICT aﬀects innovation incentives only through their impact on ﬁrm heterogeneity. Diﬀerent statistical tests support this assumption.
Speciﬁcally, Hi,t is the measure of ﬁrm heterogeneity for ﬁrm i in t, which corresponds to the year 2006. ICTi,t−1 corresponds to the variables capturing the intensity of ICT use, as well as the adoption of ICT applications (i.e. ERP, SCM and CRM) with t − 1 corresponding to the year 2003. Ii,t represents ﬁrm i’s innovation incentives captured by the fraction of total employees engaged in R&D activities. And Xi,t−1 is a matrix with control variables such as the economic sector a ﬁrm belongs to (14 classes), the geographic location of the ﬁrm (east or west Germany), as well as ﬁrm’s age up to 2007, the presence of exporting activities in 2003, among others.
The analysis considers alternative functional forms for f (·) and g(·). In particular, the ﬁrst stage assumes a linear speciﬁcation for f (·) and the main results are presented in Table 4.
The results are robust to nonlinear considerations of f (·). In order to consider this possibility, a nonparametric analysis was performed. The analysis of the second stage also considered a linear speciﬁcation for g(·). However, given the censoring present in the R&D data available, a (nonlinear) Tobit model was also considered. In addition, in order to check the robustness of the results, an alternative semiparametric approach with a nonparametric ﬁrst stage was also implemented (see Appendix A and B). The results of the second stage are presented in Table 5. Alternative robustness tests considering diﬀerent samples (as it will be clear below), as well as diﬀerent types of information regarding innovation activities were also performed.
The analysis of the ﬁrst stage is presented in Table 4. In particular, the objective of the ﬁrst stage is to consider the relationship between the intensity of ICT use and ﬁrm heterogeneity, controlling for diﬀerent ﬁrm characteristics that might inﬂuence such relationship. Table 4 present the analysis by means of ordinary least squares regression, which amount to assume a liner functional form for f (·) in equation (1). The results presented in Table 4 are robust to nonlinear considerations of f (·). In order to consider this possibility, a nonparametric analysis was performed.
In particular, the speciﬁcation considered in column 1 of Table 4 estimates the direct impact of the intensity of ICT use in 2003 on the ﬁrm heterogeneity observed in 2006. The coeﬃcient on ICT shows no impact on the observed ﬁrm heterogeneity, suggesting no independent impact of the intensity of ICT use. In order to account for the potential persistence in ﬁrm heterogeneity derived by the persistence in productivity diﬀerences, column 2 includes a set of dummy variables that locate each ﬁrm into the corresponding quartile of its sector speciﬁc productivity distribution. As can be observed, the result of column 1 is maintained with the speciﬁcation presented in column 2.
However, as highlighted by the economic literature, the adoption of ICT implies reorganization at the ﬁrm level in order to exploit the beneﬁts of the ICT implementation. One dimension of potential complementary investments required to exploit the gains of ICT corresponds to the introduction of ICT (software) applications. By introducing speciﬁc ICT applications, a given ﬁrm might achieve a minimal ICT infrastructure needed to reap the beneﬁts of the investments in computers and software.
In order to consider this possibility, the speciﬁcation presented in column 3 includes speciﬁc ICT applications adopted by the sampled ﬁrms. In particular, EPR and SCM systems described previously are included. The hypothesis behind this speciﬁcation states that the impact of ICT on ﬁrm heterogeneity does not only depend on the presence of ICT equipment, but also on the way such infrastructure is used. However, as the results shown in columns 1 and 2, the coeﬃcients on ICT intensity and ICT applications show no independent impact on ICT on ﬁrm heterogeneity.
Column 4 extends the speciﬁcations presented in columns 1-3 to consider the interactions term between ICT and ERP. It analyzes whether the impact of ICT is associated with complementarities between the diﬀerent ICT components. The results show positive and signiﬁcant coeﬃcients for the interaction term (coeﬀ.: 0.19, std. error: 0.10), suggesting an important complementarity between the diﬀerent components of the ICT infrastructure adopted at the ﬁrm level. That is, the impact of the intensity of ICT on ﬁrm heterogeneity is conditional on the presence of ERP systems. In other words, there is a critical infrastructure needed in order for ICT to diﬀerentiate a ﬁrm with respect to his competitors (i.e. induce ﬁrm heterogeneity).
Note that if ICT is not used intensively, the marginal eﬀect of the ERP applications on ﬁrm heterogeneity is negative because the independent impact of ERP is negative (coeﬀ:
-0.11, std. error: 0.05).
Column 5 considers and alternative hypothesis to the one presented in column 4. In particular, it investigates the existence of complementarities in the introduction of ICT, but on diﬀerent ICT applications. It considers the interaction term between PCW and SCM. As can be clear from the results presented in the table, the complementarity argument does not hold for this type of application. Moreover, additional results not reported in Table 4 show that the outcome of column 4 also extends to the consideration of the interaction term between PCW and CRM.
This result is not surprising at least for two reasons. First, the ERP application is a generic general purpose software in comparison to the SCM and CRM applications. In consequence, the impact of SCM and CRM might be related with particular activities of the sampled ﬁrms that are not captured with the available data. These activities might be related with the ﬁrm speciﬁc relations with their suppliers or with costumers for the case of SCM and CRM, respectively. Second, SCM and CRM applications tend to be adopted by ﬁrms after a basic ICT infrastructure in general, and ERP systems in particular, are adopted successfully.
However, the data permit a testable hypothesis of the previous argument. That is, if there is any complementarity between ICT components that is related with their characteristics and/or timing of adoption, then the interaction terms to be considered simultaneously should be PCW and ERP on the one hand, and ERP and SCM (or CRM) on the other hand.
In this manner, a regression analysis can capture the role of the intensity of ICT use as a determinant to adopt ERP system and, subsequently, provided that ERP systems were adopted successfully, the impact of SCM or CRM should be conditional on the adoption of ERP. This hypothesis is tested in column 6.
The result presented in column 6 shows that, indeed, the impact of ICT not only depends on the introduction of ICT applications, but the complementarity between ICT components is a determinant of ﬁrm heterogeneity. This is highlighted by the signiﬁcant coeﬃcient of the interaction terms of PCW and ERP, and ERP and SCM (coeﬀ: 0.19, std. error: 0.10 and coeﬀ: 0.13, std. error: 0.07, respectively). Moreover, the observed complementarity is consistent with the previous argument suggesting that such complementarities depend on the characteristics of the considered applications (e.g. timing of adoption). Note again that if ICT is not used intensively, the marginal eﬀect of the particular ICT applications on ﬁrm heterogeneity is negative.
In sum, the results of Table 4 show that ICT aﬀects ﬁrm heterogeneity only when ICT is used intensively and jointly with particular ICT applications. These results are robust to nonparametric speciﬁcations. If ICT impacts productivity positively and induces heterogeneity, it can be argued that ICT represent a source of volatility that stimulates the process of creative destruction. If this is the case, ﬁrm strategies should react accordingly, specially strategies that can provide a competitive advantage such as R&D initiatives. This is analyzed in Table 5 for a subsample of ﬁrms that reported ICT consulting activities in 2003. This selection was performed in order to consider ﬁrms for which ICT infrastructure is important. Table 6 shows that the value and signiﬁcance level of the coeﬃcients using the full sample are very close to the values reported in Table 5. However, the speciﬁcation tests are inconclusive.
Table 5 presents the main results of the analysis and performs the two stage approach taken into account the speciﬁcation presented in column 4 of Table 4. Column 1 presents the benchmark case considering an OLS speciﬁcation. Following a instrumental variable approach, column 2 considers the impact of ﬁrm heterogeneity on R&D incentives where ICT variables act as instruments. Column 2 shows a positive impact of ICT induced heterogeneity on the fraction of employees that perform R&D (coeﬀ: 0.06, std. error: 0.03). The reported p-values of the endogeneity and overidentiﬁcation tests (0.10 and 0.83, respectively) suggest the validity of the IV approach and the exclusion restriction for the considered ﬁrms. In order to verify the robustness of this result, column 3 and 4 consider a Tobit and IV Tobit speciﬁcation due to the censoring present in the R&D information. Interestingly, the results are maintained, including the required speciﬁcation tests.