Industrial and Corporate Change Advance Access published January 3, 2008 Industrial and Corporate Change, pp. 1 of 36 doi:10.1093/icc/dtm038
ICT, skills, and organizational change: evidence from Italian manufacturing firms Paola Giuri, Salvatore Torrisi and Natalia Zinovyeva
This article examines the complementarity among information and communication technologies (ICT), skills, and organizational change from a panel of 680 Italian manufacturing firms during 1995–2003. By drawing on different statistical methods, we found evidence of complementarity between skills and organizational change, but did not find evidence of complementarity between ICT and skills. Moreover, our results show that the hypothesis of full complementarity among ICT, human capital, and organizational change does not apply to small and medium firms. Instead, we discovered that organizational change yields negative effects on the complementarity between ICT and human capital.
1. Introduction This article explores the relationships among the adoption of information and communication technologies (ICT), skills, and organizational change, and the implications of different adoption strategies for a firm’s productivity. Earlier studies in this line of research have primarily focused on the shifts in the demand for skilled labor associated with either technical change (i.e. skill-biased technical change or SBTC) or organizational change (i.e. skill-biased organizational change or SBOC). Recently, the empirical literature has begun to analyze the association between the demand for skilled labor and technical change in the broad context of workplace reorganization, thus looking at the productivity implications of both technical and organizational change (Bresnahan, 1999; Bresnahan et al., 2002; and a recent survey by Arvanitis, 2005). The literature has addressed the issue of complementarity by relying on data at different levels of aggregation: countries, industries, and firms. There are also several studies that have tried to test the SBTC hypothesis by focusing on the complementarity that occurs at the individual level when analyzing the relationship between computers and the human capital of computer users. Each level of analysis has its own advantages and drawbacks. While studies based on industry-level or countrylevel data do not capture important sources of variance across firms, studies centered
ß The Author 2007. Published by Oxford University Press on behalf of Associazione ICC. All rights reserved.
2 of 36
P. Giuri et al.
on individual workers have other important disadvantages. These works miss an important dimension of the complementarity story represented by organizational changes, which take place at the firm level. Recognizing the role of organizational change implies that one accounts for both direct and indirect effects of new technologies on labor demand. Direct effects arise from the adoption of skilled labor that is required by the use of new technologies (i.e. computer skills). Indirect effects are spurred by organizational co-inventions and product or services innovations that may produce an additional effect on the productivity of skilled labor. The association between ICT and skills adoption is then mediated by important organizational changes. The use of ICT increases the volume of data analysis and transactions within the firm and across firms. This in turn gives rise to modifications in the organization of the firm and calls for analytical and cognitive skills (e.g. marketing analysis and quality control data analysis capabilities). Moreover, the use of ICT may spur decentralization of authority and more flexible forms of division of labor such as teamwork, multi-tasking, job rotation, justin-time, and quality circles. Workers have to deal with greater autonomy, responsibility and uncertainty. This requires both cognitive skills and “people” skills that are important for interacting and communicating with colleagues, customers, and suppliers. To account for the multiple interactions among ICT, skills, and organizational change, Bresnahan (1999) has introduced the concept of organizational complementarity between ICT and highly skilled workers. This concept is in line with the theory of modern manufacturing, according to which new flexible technologies are part of a system or cluster of organizational changes (Milgrom and Roberts, 1990). The interdependence between concurrent inventions (new forms of work organization and human capital) imposes significant adjustment costs which vary across different ICT adopters. It is likely then that in the short term different firms will have different combinations of co-inventions. Firms that have managed to adapt their organization to ICT will enjoy significant productivity gains from ICT investments because of complementarity. The existing literature has mainly tested this hypothesis on large firms or in samples where large and smaller firms are pooled together; however, the organizational peculiarities of small and medium sized firms (SMEs) call for a deeper understanding of productivity gains of ICT investments, skills, and organizational change in this category of firms. SMEs typically adopt more flexible and simpler organizational structures than large firms. Higher levels of organizational efficiency in these firms can be achieved by small organizational changes and a more intensive use of skilled labor without necessarily adopting complex ICT solutions. Moreover, sophisticated ICT solutions like Enterprise Resource Planning (ERP), Customer Relationship Management (CRM), and Supply Chain Management (SCM) systems, may require additional
ICT, skills, and organizational change
3 of 36
organizational costs that in a SME probably will not be offset by substantial improvements in division of labor and coordination. One reason why ICT may yield disappointing productivity gains to SMEs is that many available technological solutions have been developed for the needs of large firms and do not account for the specific organizational characteristics of smaller firms (Levy and Powell, 2000). ICT solutions like ERP, for example, have intrinsic business models that do not meet the needs of SMEs (Olsen and Saetre, 2007). One may argue then if “SMEs really need to use such sophisticated electronic systems and whether there are alternative, less costly ways for them to integrate business processes electronically” (European Commission, 2003: 6). Therefore, simultaneous investments in ICT, skills, and organizational change in SMEs may lead to results that are in contrast with the organizational complementarity hypothesis because, compared with large firms, SMEs face greater difficulties in managing different inventions at the same time, finding and retaining high skill personnel and re-engineering their business processes to fully integrate ICT into their organization. Testing the hypothesis of organizational complementarity in SMEs is extremely important since a large share of the economy is comprised of SMEs. In this article we test the theory of complementarity between ICT, skills, and organizational change by examining 680 Italian manufacturing firms during the years 1995–2003. About 94% of firms in our sample are small and medium sized. We adopt two approaches for studying the complementarity. First, we analyze the correlations between the investments in ICT, skills, and organizational changes conditional upon various firm-specific characteristics. Second, we examine the productivity effects of pair-wise interactions between ICT, skills, and organizational change. We also analyze the productivity gains of the interaction between the three complements together. We use lagged values of independent variables to moderate the problem of simultaneity and estimate a production equation with fixed effects to deal with timeinvariant unobserved heterogeneity. This article contributes to the literature on the following grounds. First, unlike earlier studies that have mostly focused on large firms, we explore the issue of complementarity in a sample consisting primarily of small and medium firms. Second, we study the interactions among all three potential complements in a production function framework and provide novel empirical evidence on Italian manufacturing firms. Earlier studies on various European countries have mostly focused on either the SBTC hypothesis or the SBOC hypothesis (Caroli and Van Reenen, 2001), while only a few have addressed the issue of complementarity among ICT, skills, and organizational change (Arvanitis, 2005). Our findings do not provide any evidence in favor of the hypothesis of complementarity between ICT, skills, and organizational change (organizational complementarity). Instead, our results support the hypothesis of pair-wise complementarity between organizational change, or the use of modern workplace practices, and skills.
4 of 36
P. Giuri et al.
Finally, we do not find any significant complementarity between ICT and skills. These results are consistent with the hypothesis that complementarity among ICT, skills, and organizational change does not apply to SMEs. This article is organized as follows: Section 2 summarizes the main findings in the literature; Section 3 describes the methodology; Section 4 describes the data and variables used in the empirical analysis; Section 5 reports and discusses the results; Section 6 concludes.
2. Background literature and hypotheses Modern, flexible manufacturing systems require the adoption of a cluster of complementary inventions: new technologies, skills, and new forms of division of labor that depart from mass production systems and bureaucratic, centralized organization. As Milgrom and Roberts (1990, 1995) have noted, new technologies, skills, and workplace innovations can be viewed as strategic complements in a production function setting. Empirical works on the complementarity among these factors have initially followed two distinct directions: the test of the SBTC hypothesis and the test of the SBOC hypothesis. More recently, these two streams of the literature have converged on a more comprehensive approach that accounts for the productivity effects arising from the joint use of ICT investments, skills or human capital, and organizational change.
2.1 Complementarity between skills and technical change According to the SBTC hypothesis, technological change, and particularly the adoption of ICTs, increases the demand for skilled labor with respect to unskilled labor and leads to increasing wage inequality (Acemoglu, 1998; Autor et al., 1998; Machin and Van Reenen, 1998). The use of computer-based technologies induces an increasing demand for skilled labor relative to manual, unskilled workers. This hypothesis is supported by the fact that the demand for skills appears to increase within plants and industries, rather than being associated with labor relocation towards specific sectors (e.g. services) (Berman et al., 1994). However, SBTC is particularly significant in industries such as office machines, electrical machinery, printing, and publishing. Together they account for 40% of within-industry increase in the relative demand for skills in a sample of OECD countries (Berman et al., 1997).1 1 Available empirical evidence on Italy points out the rising importance of skilled workers associated with technological change (Erickson and Ichino, 1994; Casavola et al., 1996; Manacorda, 1996). Most of these studies rely on very rough measures of technical change and do not isolate the effect of ICT in particular. More recently, Bratti and Matteucci (2004) have tested the SBTC hypothesis by
ICT, skills, and organizational change
5 of 36
2.2 Complementarity between skills and organizational change Another implication of the economics of modern manufacturing is that organizational changes taking place at the firm level produce a further shock to the relative demand for skills. In brief, this is the SBOC hypothesis. The main argument of the SBOC hypothesis is that the adoption of new organizational systems based on decentralized decision-making and delayering calls for more skilled people. Caroli and Van Reenen (2001) compared the benefits and costs of new decentralized organizations. The benefits are represented by the reduction of the cost of information transfer and communication, a greater reactivity of firms to external changes, the reduction of the cost of monitoring activities, and the increase in job satisfaction due to job enrichment, i.e., a greater involvement in problem solving, higher information sharing, and participation in decision making. The costs of decentralization arise from higher risk of duplication of information, increased probability of mistakes due to a lower level of control, reduced returns to specialization, and reduced worker efficiency associated with greater stress. The SBOC theory predicts that skills raise the benefits and reduce the costs of decentralization of responsibility within organizations; therefore, skill-intensive firms that introduce organizational changes will have greater productivity gains than non skill-intensive firms. This is because skilled workers have a greater ability to handle information, communicate, and interact with other people; they also tend to be more autonomous and more satisfied with their work. Empirical evidence from different countries (especially France and the United States) lends support to the SBOC hypothesis (Black and Lynch, 2001; Caroli and van Reenen, 2001).
2.3 Complementarity among ICT, skills, and organizational change ICTs are a general-purpose technology (GPT) or a new technological paradigm. As such they induce major changes in the system of production and institutional settings (Dosi, 1982; Freeman and Perez, 1988; Bresnahan and Trajtenberg, 1995; David and Wright, 1999; Aghion, 2002). Users and producers need a long time to experiment with these new technologies and to adapt their organizations to new systems of production. The supply of skills required by the new technologies takes time to materialize; the same is true for methods of production. This gives rise to disequilibrium in the labor market. The overall impact of GPTs then is not only direct, but takes place also through a series of secondary innovations (including using data from Italian manufacturing. They have analyzed the impact of ICT and R&D expenditures on the shares of production and non-production workers and found a strong, generalized effect for non-production workers only. These results show that new technologies are a substitute for unskilled workers, but do not support the hypothesis of complementarity between ICT and skills.
6 of 36
P. Giuri et al.
reorganization of workplace). Adjustment processes may be characterized by periods of productivity slowdown. The “genericness” of ICTs drives the focus of the literature on organizational complementarity between ICT and skilled labor (Bresnahan, 1999). In this perspective, technological and organizational change together call for more skilled labor. An important reason is that there is limited substitution of computers for human work. The substitution is apparent for simple tasks not requiring further processing, like record keeping or computation, which are normally carried out by less skilled people. For more complex and cognitive work, typically carried out by more highly skilled managers and professionals, the automation of tasks and substitutions by computers is more difficult. At the same time, standardization and automation of repetitive procedures allow for a greater centralization of data and information management. This creates new tasks, which involve more generalist, problem-solving roles implying more autonomy and responsibility. A greater amount of information produced and transmitted within the organization and across organizations also calls for additional cognitive, analytical skills (e.g. marketing analysis capabilities) and interactive or people skills. The intensity of the threefold complementarity between the adoption of ICT, related organizational changes, and skills depends on the type of computer use. Following a popular classification in the ICT industry, Bresnahan (1999) distinguishes among three types of computer use: (i) Organizational computing, such as corporate accounting and transaction processing systems, which affect various levels of corporate and departmental administrative processes—e.g., ERP and Material Requirements Planning (MRP); (ii) Scientific-technical computing (e.g. client-server technical applications like Computer-Aided Design/ComputerAided Engineering (CAD/CAE) and Computer-Aided Manufacturing (CAM), which is mostly targeted to specific departments; (iii) Individual productivity computing like word-processing. The use of computing, especially “organizational computing,” is mostly important in service-intensive activities, i.e., the service sector and the office departments of manufacturing firms. The increase in the relative demand for skilled people is therefore more likely to occur in these activities than in manufacturing ones. However, many manufacturing firms adopt applications like ERP, MRP, and database management systems (DBMS), which call for significant organizational changes and skilled labor. Furthermore, many manufacturing firms adopt new computerbased systems for supply chain andCRM, which also require specific cognitive and people skills (Clark, 2003). Since there are different adjustment costs and adjustment timing across the complements, the effects on productivity are likely to occur only in the long run (Brynjolfsson and Hitt, 2000a). The adjustment time is firm-specific and therefore in the short run we should expect cross-sectional differences in the adoption of complements across firms.
ICT, skills, and organizational change
7 of 36
Bresnahan et al. (2002) tested the hypothesis of organizational complementarity in US firms. These scholars posited that the adoption of ICT is more effective in organizations with more skilled people and with decentralized workplace organization. It is the cluster of complementary inventions then that drives the productivity gains rather than the single components. Gretton et al. (2002) also found evidence of complementarity between human capital and ICT, and between organizational change and human capital in Australia. The evidence on complementarity across European countries is less clear and widespread. For instance, Arvanitis (2005) found evidence only for the complementarity between ICT and human capital in the Swiss business services. The literature on complementarity between ICT, organizational change, and skills in other European countries confirms that these variables produce positive and significant marginal effects on labor productivity but have only limited interaction effects (Caroli and Van Reenen, 2001; Hempell, 2003; Bertschek and Kaiser, 2004). Recent studies on Italian data report similar findings. For example, Piva and Vivarelli (2004) examined a panel of 488 Italian manufacturing firms in the period 1989–1997 and found that organizational change has a significant marginal effect on the demand for skills. Their analysis, based on research and development (R&D) as a measure of technical change, however, does not support the hypothesis of complementarity between R&D and skills. Drawing on the same dataset, Piva et al. (2005) also found that the interaction of R&D and organizational change yields a significant effect on the share of white-collar workers (a proxy for skilled labor) and a negative effect on the share of blue-collar workers. This evidence appears to be in line with the hypothesis of organizational complementarity, but also suggests that the interactions among technical change, skills, and organizational change are complex and difficult to measure. Overall, these findings suggest that the relationships between ICT and organizational change are quite controversial.
2.4 Complementarity in large and small firms Existing empirical studies focus mainly on large firms (Bresnahan et al., 2002) or rely on samples where large firms are over-weighted (Caroli and Van Reenen, 2001) or pooled together with smaller firms (Bertschek and Kaiser, 2004). With few exceptions (Levy and Powell, 2000; Harland et al., 2007; Olsen and Saetre, 2007), available case studies on ICT adoption are also mostly centered on large firms (Murnane et al., 1999; Brynjolfsson and Hitt, 2000b). The potential differences between large firms and SMEs in the patterns of complementarity then remain largely unexplored. However, there are reasons to believe that the complementarity hypothesis may not apply to SMEs. Large firms probably have a greater demand for all the complements compared with smaller firms; therefore, they should benefit from the use of ICT to a larger extent than smaller firms. For instance, large organizations
8 of 36
P. Giuri et al.
face greater complexity and require more highly skilled labor (Fisher and Govindarajan, 1992). In large firms there is a greater amount of information to be processed and a greater number of documents, tasks, and people that have to be coordinated. In fact, several large organizations in the 1990s have introduced new forms of coordination, information sharing, and decision-making, and have invested in ICT. Like large firms, small ICT-intensive firms must invest in skills to make their technological investments effective. In addition, ICT increases the opportunities for communication and information sharing. This spurs a greater codification of procedural knowledge (procedural changes), which in turn calls for more skills. Skilled workers are important for planning and implementing new and more formalized procedures. But the joint investment in ICT and investment in skills may be particularly painful for small firms. In small firms, significant investments in ICT and skills are often induced by business relations with large firms which use electronic transaction systems such as CRM and SCM: “in some cases SMEs may not have the choice between participation and non-participation as large enterprises are conducting more and more transactions exclusively by electronic means” (European Commission, 2003: 20). Adaptation to ICT in such cases is particularly difficult since not only the internal organization has to be modified but also the interfaces with the external environment need to be re-organized to cope with different routines and standards. For example, in SCM, SMEs typically do not have enough skills and time to implement the internal applications needed to work with larger firms (Harland et al., 2007). Moreover, the investments required to adapt the organization to the external change might be perceived as a source of rigidity by the workers of smaller firms. SMEs examined by Harland et al. (2007), for instance, did not plan to change their business model to adapt to IT-enabled SCM because they consider the new model as being in contrast with the flexibility and adaptability asked for by their customers. Several SMEs then invest in basic ICT infrastructure such as computers and internet connection that require some skill upgrading but do not require significant organizational changes. Like investments in ICT, small firms’ investments in organizational changes are constrained by the limited scale and complexity of operations and by their greater flexibility. To some extent, this may result in a trade-off between organizational change and skills or ICT. For instance, the adoption of ICT may induce a small firm to automate relatively routine tasks (e.g. accounting or auditing), and consequently to reduce its demand for unskilled personnel. At the same time, however, the limited size of the organization does not allow the use of complex organizational computing, which would require investments in human capital containing cognitive, analytical, and people skills. Similarly, in a small firm the employment of skilled workers may co-exist with an informal, traditional division of labor. Moreover, the limited size
ICT, skills, and organizational change
9 of 36
may not call for formal modern organizational practices like job rotation or teamwork. The lack of attention to SMEs in the literature appears partly due to the fact that data on complementarity among skills, ICT investments, and organizational change in smaller firms are difficult to obtain. However, we believe that looking at SMEs is important. The use of ICT in these firms has increased over time as a consequence of declining quality-adjusted prices. There are reasons to believe then that the complementarity with organizational change and skills has set in motion a process of organizational adaptation that it is worthy of study. This may be especially applicable in countries like Italy, where small and medium sized firms account for a very large share of the economy. In this context it is interesting to see the productivity gains of those manufacturing firms that try to introduce both organizational and technical innovations.
2.5 Hypotheses Our goal is to test the hypotheses of complementarity among ICT, human capital, and organizational change in a sample mainly composed of SME Italian manufacturing firms. We start by testing the hypothesis that productivity of human capital increases with the adoption of new technologies. Unlike earlier studies that have used R&D expenditures or other proxies for technical change, we are interested in ICT investments because the diffusion of this technology, like other GPTs in the past, is believed to spur the adoption of complementary innovations. Our main measure of the stock of skilled labor is the percentage of employees with an upper-secondary or university education. Second, we test the hypothesis of complementarity between skills and the use of modern organizational practices or performing organizational change. Third, we test the organizational complementarity hypothesis, according to which the productivity premium of simultaneous investments in ICT and human capital is increased by organizational changes.
3. Methodology 3.1 Adoption approach Two main econometric approaches are typically used in the literature for testing the hypothesis of complementarity. The most popular one is the adoption approach. The simplest version of this approach relies on reduced-form estimations of the investments in one of the complements conditional upon the adoption of other complements, controlling for other observable characteristics of the adopter.
10 of 36
P. Giuri et al.
The underlying idea is that some particular inputs that can vary easily in the short term can be predicted by other firms’ fixed or quasi-fixed choice inputs. In this context, complementarity implies a positive correlation between the levels of adoption of the hypothesized complements. However, estimations of reduced-form equations may lead to a simultaneous equations type of bias due to the endogeneity of the choice variables. To handle the problem of endogeneity of regressors in reduced-form equations, Arora and Gambardella (1990) and Arora (1996) suggest a one-sided test for complementarity that utilizes the conditional correlations between the residuals of reduced-form regressions of hypothesized complements on observable exogenous variables. This test has also some limitations since the non-negativity of the variancecovariance matrix of residuals could be the result of unobserved exogenous variables that affect the endogenous inputs in a correlated way for reasons different from complementarity. To overcome this problem, Bagues (2004) has developed a test of complementarity that is based on the inter-temporal structure of residuals. This test appears to be robust to the existence of omitted variables that are not serially auto-correlated. But, like other dynamic panel techniques, this approach relies on the assumption of stability of the complementarity effect across time and is very demanding in terms of longitudinal data.2 In this article we first look at the correlations between residuals obtained from adoption regressions where the dependent variables are proxies for ICT, skills capital, and organizational changes, respectively. Our explanatory variables are firms’ sizes and a series of dummies, which account for geographical location and sector. A high degree of correlation between ICT, human capital, and organizational change would imply that empirical models that estimate the impact of only one of these variables on productivity might lead to biased results due to omission of other correlated choice variables.
3.2 Production function approach An alternative method for testing complementarity is the production function approach. The test of complementarity in this context is based on the t-test of the pair-wise interactions between the potential complements (Caroli and Van Reenen, 2001; Bresnahan et al., 2002). The limitations of this method are apparent when the number of choice variables is larger than two (Lokshin et al., 2004). Estimates of pairwise interaction effects ignore more complex linkages among
2
Athey and Stern (1998) review different econometric methods for testing theories of complementarity and provide an evaluation under alternative assumptions on the economic and statistical environment.
ICT, skills, and organizational change
11 of 36
more than two complements. At the same time, as it was argued by Ichniowski et al. (1997), putting all the interaction terms for the complements inside the regression could lead to confounded estimates due to collinearity among regressors.3 We use the production function approach for testing complementarity in the following way. We start considering the standard estimation model, which is used in various earlier works (Caroli and van Reenen, 2001; Bresnahan et al., 2002): logðSi Mi Þ ¼ f ðLi , Ki , Qci ; zi , ci Þ,
ð1Þ
where for each i firm Si stands for sales, Mi is the materials bill (the dependent variable then is the log of the value added), Li measures labor expenses, Ki is a measure of capital, and Qci is a measure of firm i’s choice of the hypothesized group of complements. Our measures of the potential complements are entered in levels along with their interactions with one another. The complements considered in this analysis are the log of ICT stock (ICTstockit), the share of skilled labor (SKILLSit) and proxies for organizational change (OCi.). We also use a set of proxies for “modern” work practices (such as teamwork, quality circles, workers mobility) as an alternative to OC.4 Controls including detailed information on technological sector and geographical location are captured by the term zi. The term ci represents firms’ unobserved heterogeneity. Capturing the influence of unobserved characteristics of the coefficients of interest is very important when performing estimations of firms’ production function, especially when using information about the firms in different technological and product sectors. If unobserved firm-specific factors influence both firms’ decision making and production processes, ignoring the presence of these factors could bias the results substantially.
3
Recent alternative approaches for studying complementarity in a production function context rely on the use of multiple inequality restrictions (Wolak, 1989; Lokshin et al., 2004). These studies approach the problem of collinearity by identifying different clusters of adoption practices and compare the productivity outcomes of alternative clusters (Ichniowski et al., 1997). Following these studies, the hypothesis of complementarity is tested by relying on the theory of supermodularity of objective function, which asserts the necessary conditions for two or more activities to be complements (Milgrom and Roberts, 1990). According to this theory two activities yi and yj are complements in the objective function f if the following inequality holds for all possible values of the other arguments of f (Athey and Stern, 1998): f ðyiH , yjH ,Þ f ðyiL , yjH ,Þ f ðyiH , yjL ,Þ f ðyiL , yjL ,Þ, where H and L stand for high and low adoption intensity, respectively, and f is a measure of performance. This condition posits that the marginal productivity of each activity increases with the adoption of the other. However, with this methodology we lose a lot of information on variation because continuous variables are transformed into dummies identifying the adoption of a complement. As a consequence, the precision of estimates may decrease compared with alternative approaches based on continuous regressors. 4
The exact definitions and the description of these variables are illustrated in the following section.
12 of 36
P. Giuri et al.
OLS estimations on cross-section data generally rule out the presence of the unobserved term ci, which may be correlated with the regressors and thus generate inconsistent estimates. Since our dataset contains firm-level observations over the period 1995–2003, we perform fixed-effect estimations, which allow controlling for firm-level time-invariant sources of unobserved heterogeneity. More precisely, the log linear model that we estimate is assumed to be of the following form: log VAit ¼ t þ log Lit þ log Kit þ Qcit þ zit þ ci þ uit ,
ð2Þ
where ci is a firm-specific effect which is assumed to remain constant over time. The fixed-effect estimation model yields consistent estimation of equation (2) allowing for correlation between ci and other independent variables. Alternatively, ci can be viewed as an individual random effect, under the assumption that it is orthogonal to the observed explanatory variables (and therefore is a part of the error term). A standard formulation of the fixed-effect model, the within-group transformation, exploits the non-variability of firms’ unobserved heterogeneity across time and differentiates it by subtracting time-averages from all the variables in the model (Wooldridge, 2002). A drawback of the fixed-effect estimator remarked by earlier works (Griliches and Mairesse, 1995; Brynjolfsson and Hitt, 2000a) is that it tends to magnify the impact of errors-in-variables, thus attenuating the effect of estimates (attenuation bias). In addition to performing fixed-effect estimations we exploit the time dimension to reduce the simultaneity problem. In order to do this, we take all explanatory variables in equation (2) with a 3-year lag both in OLS and fixed-effect regressions.
4. Data and variables 4.1 Data Our empirical analysis employs three sources of data. The main dataset is drawn from a survey of Italian manufacturing firms conducted by a leading Italian bank, Mediocredito Centrale (now Capitalia), in two waves covering the years 1995–1997 and 1998–2000. Each wave includes data on a sample of over 4000 Italian manufacturing firms with at least 10 employees, belonging to different sectors, geographical areas, and size classes. The dataset provides three types of data: (i) annual accounting data from 1989 to 2000; (ii) annual survey data on quantitative company characteristics (e.g. employment and investments) from 1995 to 2000; and (iii) qualitative variables such as firms’ group memberships, core sectors, innovation activities, and organizational changes for the periods 1995–1997 and 1998–2000. Some of these variables are observed only once every 3 years while others are available on an annual basis. The database is
ICT, skills, and organizational change
13 of 36
a statistically significant sample of the Italian manufacturing sector and was obtained by a stratified random selection process.5 Our analysis draws on a panel of 1302 manufacturing firms that participated in both waves of the survey. From this sample we excluded the 265 outliers with the following characteristics: (i) firms that during the surveyed period performed acquisitions or divestures that resulted in a significant variation of the number of employees (more than one standard deviation calculated for previous 3 years);6 (ii) firms not involved in acquisition or divestiture operations and reporting a variation of their average number of employees between 1995–1997 and 1998–2000 larger than three standard deviations; (iii) inconsistent observations such as firms that reported very low total investments and exceptionally high ICT investments in the same period. As mentioned above, the Mediocredito surveys provide information about firms’ activity for the 1995–1997 and 1998–2000 time periods. To perform a fixed-effect estimation of equation (2) and to lag our independent variables we need observations on performance at another point in time. We then matched the Mediocredito dataset with accounting data drawn from the Bureau Van Djik’s Amadeus dataset for the period 2000–2003. Amadeus data were not available for 190 firms, which were dropped from our sample. We finally eliminate all observations with missing values in both periods for the variables used in our analysis. We end-up with a final unbalanced panel of 965 observations covering 680 firms. Of these, 285 are present in both periods, 154 only in the first period, and 241 only in the second period. Unfortunately, the Mediocredito dataset provides limited information about firms’ organizational characteristics. For a subset of our sample we employ data from another survey of Italian manufacturing firms conducted in 2003 (ICT_Adoption survey, see Corsino et al., 2005), which includes more detailed information on the use of “modern” organization practices such as teamwork, quality circles, or job rotation. This survey provides data on a sample of 1014 firms selected with the same stratified random selection procedure adopted for the Mediocredito survey. We matched all three datasets and obtained a sample of 203 firms for which we have information on the use of several organizational practices during the period 1998–2002. 5
Sample size and composition have been obtained by the Neyman formula, which allows us to minimize the sample error. Data for firms with over 500 employees cover the population of manufacturing firms. The questionnaire on qualitative information contains sections on investments, R&D, internationalization and labor forces. For details on the survey design see the report of Capitalia (2002), available in Italian at the following website http://www.unicreditcapitalia.eu/it/Capitalia/Ricerche/Osservatorio/Indagini_Imprese_Manifatturiere/INDAG_ MANIFATT_RAPPORTO_8.pdf (December 2007). 6
Piva and Vivarelli (2004) also followed this approach with the data drawn from the same dataset.
14 of 36
P. Giuri et al.
4.2 Variables and descriptive statistics This Section describes the variables used in our estimations. Table 1 summarizes the definitions of the variables. Table 2 reports some descriptive statistics of the main variables for the pooled sample of 965 observations. For the variables drawn from the ICT_Adoption survey the number of observations are 203. Table 1 Description of variables Variable
Description
Data source
Value added in 2000 and 2003; thousands of euros, real
Amadeus
Dependent variable VA
values (base 1995) Main regressors ICT stock
ICT stock of the firm in 1997 and 2000 computed by
Mediocredito
deflating firms’ ICT investments by using an extrapola-
Survey
tion of Gordon’s (1990) deflator for computers; thousands of euros SKILL ORG_CHANGE
Share of employees with upper secondary education or
Mediocredito
an university degree in 1997 and 2000
Survey
Dummy variable equal to 1 if the firm performed
Mediocredito
organizational changes induced by process or product
Survey
innovations in 1995–2000 MULTIF_TEAM
Dummy equal to 1 if the firm uses intensively multifunctional teams
ICT_Adoption Survey
SELFMAN_TEAM
Dummy equal to 1 if the firm uses intensively self-
ICT_Adoption
managed teams
Survey
Dummy equal to 1 if team work is an important
ICT_Adoption
promotion criteria within the firm
Survey
Dummy equal to 1 if the firm uses intensively
ICT_Adoption
incentives schemes linked to team performance
Survey
Dummy equal to 1 if the firm uses intensively quality
ICT_Adoption
circles linked to team performance
Survey
Dummy equal to 1 if information sharing with workers is
ICT_Adoption
an important practice
Survey
Dummy equal to 1 if the exchange of employees with
ICT_Adoption
supplier is an important practice
Survey
Dummy equal to 1 if the firm acquired new plants
ICT_Adoption
between 1998 and 2002
Survey
TEAM_ PROM TEAM_ PERF QUALITY_CIRCLE INFO_SHARING EXCH_SUPPLIER NEWPLANT
(continued)
ICT, skills, and organizational change
15 of 36
Table 1 Continued Variable
Description
Data source
Other variables and controls NonICT_stock
Book value of physical capital stock (property, plants and
Mediocredito
equipment) in 1997 and 2000 deflated by the GDP at
Survey
the estimated age of the capital stock; thousands of euros LaborExp
Labor expenses in 1997 and 2000 reported in the firm’s balance sheet; thousands of euros, real values (base
Mediocredito Survey
1995) INNOVATION YEAR
Dummy variable equal to 1 if the firm performed process
Mediocredito
or product innovations in 1995–2000.
Survey
2 dummies for the time period
Mediocredito Survey
ATECO PAVITT REGION MACRO_REGION SMALL FIRM MEDIUM FIRM LARGE FIRM
Dummies for the sector of activity of the firm, identified
Mediocredito
according to the ATECO 2-digit classification
Survey
Dummies for the 4 sectors of activity of the firm,
Mediocredito
identified according to the Pavitt classification
Survey
Dummies for the 20 Italian regions in which the firm is
Mediocredito
located
Survey
Dummies for the 4 Italian macro-regions in which the
Mediocredito
firm is located
Survey
Dummy variable equal to 1 if the firm has 10–50
Mediocredito
employees
Survey
Dummy variable equal to 1 if the firm has 51–250
Mediocredito
employees
Survey
Dummy variable equal to 1 if the firm has more than 250
Mediocredito
employees
Survey
4.2.1 Value added The dependent variable in our estimations is the firms’ log of value added in 2003 and 2000. As noted before, this information was drawn from the Bureau Van Djik’s Amadeus dataset because it is not available from the Mediocredito surveys used in this work. 4.2.2 ICT The Mediocredito survey reports the total ICT investments of the firm in the 3-year period covered by each survey. Since the survey provides the total ICT investments
16 of 36
P. Giuri et al.
Table 2 Descriptive statistics for the main variables N
Mean
Std. Dev.
Min
Max
VA
965
4109.795
14093.800
55.3
309985.3
LaborExp
965
2134.319
5137.700
99.8
77892.6
NonICTstock
965
3941.960
11582.790
0
217803.2
ICTstock
965
221.540
762.279
0
10520.7
SKILL
965
0.353
0.228
0
1
ORG_CHANGE
965
0.540
0.499
0
1
MULTIF_TEAM
203
0.571
0.496
0
1
SELFMAN_TEAM TEAM_ PROM
203 203
0.547 0.611
0.499 0.489
0 0
1 1
TEAM_ PERF
203
0.537
0.500
0
1
QUALITY_CIRCLE
203
0.621
0.486
0
1
INFO_SHARING
203
0.719
0.450
0
1
EXCH_SUPPLIER
203
0.310
0.464
0
1
NEWPLANT
203
0.616
0.488
0
1
INNOVATION
965
0.880
0.325
0
1
TRADITIONAL SCALE_INTENSIVE
965 965
0.517 0.176
0.500 0.381
0 0
1 1
SPECIALISED_SUPPLIER
965
0.283
0.451
0
1
HIGH-TECH
965
0.024
0.153
0
1
NORTH-EAST
965
0.428
0.495
0
1
NORTH-WEST
965
0.291
0.455
0
1
CENTRE
965
0.181
0.386
0
1
SOUTH
965
0.099
0.299
0
1
EMPLOY SMALL FIRM
965 965
77.953 0.668
161.466 0.471
11 0
2301 1
MEDIUM FIRM
965
0.267
0.443
0
1
LARGE FIRM
965
0.064
0.245
0
1
for 3 years, to obtain ICT stocks we deflated the value of ICT investments by using an extrapolation of Gordon’s (1990) deflator for computers with price change ¼ 19.3% per year (Bresnahan et al., 2002). A possible limitation of this measure is that it relies on a backward interpolation of past ICT investment flows. We also constructed a variable that approximates the value of the non-ICT stock of the firm. This variable is included in the production function estimations as a proxy for the capital input. Drawing on the methodology discussed by Hall (1990) we first estimated the age of the capital stock by calculating the 3-year average value of the ratio of firms’ net assets to the annual amortization, assuming a constant depreciation rate. The estimated value of the capital stock was deflated by the
ICT, skills, and organizational change
17 of 36
implicit price deflator of GDP at the calculated average age (base year ¼ 1995). To obtain a proxy of the non-ICT stock we subtracted the value of ICT stock from the total capital stock. 4.2.3 Skills To measure the level of skills in the firms’ workforce we use data on the composition of the workforce by level of education: university education, high school education, and secondary school education or lower. We build the variable SKILL that represents the percentage of employees with a university or a high school diploma at the end of each period covered by the Mediocredito surveys, i.e., 1997 and 2000. Table 2 shows that on average in our sample 35.3% of the employees had an upper secondary school diploma or a university degree. Overall, the share of employees with university education is only 3.7%. These shares are below the average of the Italian population aged 25–64 with at least an upper secondary education in 2000 (45.2%) or a tertiary education (9.4%) (Eurostat, 2007; OECD, 2007). These differences, however, are not surprising considering that skilled people are mostly employed in large firms in the service sector and public administration. Generally, larger organizations have a greater demand for skilled people and offer better career opportunities and higher wages (see, e.g., Belfield, 1999). Employment opportunities for educated people in smaller firms, especially in the manufacturing sector, are limited.
4.3 Organizational change and modern organization practices The Mediocredito survey contains two proxies for organizational change (i) a variable equal to 1 if the firm has undertaken some organizational change associated to product innovations, and (ii) a variable equal to 1 if the firm has undertaken some organizational change associated to process innovations, in the 3-year period covered by the survey. Since these measures of organizational change are by construction associated to product and process innovations, respectively, our estimations may be biased. This is especially the case of organizational change associated to process innovations, which might include the adoption of ICT. To better understand the relation between this type of organizational change and the fact that the firm undertakes product or process innovations, we first analyzed the correlation between these measures of organizational change and product or process innovations in the same period. We found that few firms that performed innovations in one of the two periods (1995–1997 or 1998–2000) have implemented related organizational changes in the same period. This suggests, in line with Bresnahan et al. (2002), that organizational restructuring may not be simultaneous with the decisions about ICT and human capital. Therefore, we opted for a dummy variable (ORG_CHANGE) that is equal
18 of 36
P. Giuri et al.
to 1 if the firm has implemented some organizational changes related to product or process innovation in the period 1995–2000. We also construct the variable INNOVATION, which is equal to 1 if the firm performed either a product or a process innovation in the period 1995–2000. Overall, Table 2 shows that in 54% of cases firms have undertaken some organizational change while around 88% of firms performed either product or process innovations. The Pearson coefficient of the correlation between the two variables is 0.247 and is statistically significant. This indicates that even if most firms in our sample perform either process or product innovation activity, our measure of organizational change is not independent from innovation.7 Therefore, to identify the effect of organizational change on productivity above and beyond the effect of innovation we include the variable INNOVATION in all our estimations. However, even controlling for innovation, a clear limitation of our proxy for organizational change is that it does not measure other types of organizational changes unrelated to product or process innovation, like the adoption of modern labor organization practices. Our variable ORG_CHANGE is then a proxy that distinguishes firms that have adapted their organization to technological innovation from firms whose organization is less adaptive. It is worth noting that in our empirical analysis we look at the complementarity between ICT and skills in separate sub-samples, representing firms that have introduced organizational changes and firms which have not, respectively. Unfortunately, the Mediocredito dataset does not provide more detailed information about the nature of organizational changes, such as the adoption of new work practices like de-layering, teamwork, and job rotation.8 The lack of detailed information on firm-level organizational change has also led earlier studies to resort to dummies for the occurrence of organizational change (Caroli et al., 2001; Piva et al., 2005) or to the level of product turnover as a measure of “creative destruction” within the organization (Thesmar and Thoenig, 2000). In order to check the robustness of our results with respect to the definition of organizational change, we employ a set of additional variables drawn from the ICT_Adoption Survey for a subset of our sample. A first group of variables describes the use of “modern” organizational practices. The firms were provided with a list of organizational practices containing descriptions of different aspects of teamwork, task flexibility, information sharing, use of non-conventional incentive mechanisms, and exchange of employees with suppliers. The respondents were asked to evaluate the importance of each particular organizational practice on a Likert scale from 1 (not important) to 5 (very important). From these responses we created a set of 7
The authors are grateful to one of the two anonymous reviewers for highlighting this weakness in our measure of organizational change. 8
See Caroli and van Reenen (2001), Bresnahan et al. (2002), Greenan (2003), Bertschek and Kaiser (2004).
ICT, skills, and organizational change
19 of 36
dummy variables that take value 1 when the score assigned by the respondent to a given organizational practice is equal or larger than 3. This indicates that the use of the given organizational practice is important within the firm. Specifically, we construct the following variables: – MULTIF_TEAM: dummy equal to 1 if the firm uses multi-functional teams – SELFMAN_TEAM: dummy equal to 1 if the firm uses self-managed teams – TEAM_PROM: dummy equal to 1 if teamwork is an important promotion criteria within the firm – TEAM_PERF: dummy equal to 1 if the firm uses incentives schemes linked to team performance – QUALITY_CIRCLE: dummy equal to 1 if the firm uses quality circles linked to team performance – INFOSHARING_WORKERS: dummy equal to 1 if information sharing with workers is an important practice – WORKER_EXCHANGE_SUPPLIER: dummy equal to 1 if the exchange of employees with supplier is an important practice. We also employ a variable that accounts for organizational changes that occurred 5 years before the survey. Precisely, we use the variable NEWPLANT, which is a dummy equal to 1 if between 1998 and 2002 the firm acquired new plants. The descriptive statistics in Table 2 indicate that the different organizational practices are quite diffuse in our sample. Around 30% of firms exchange employees with suppliers and more than half of the firms share information with employees and make use of teamwork, different incentive mechanisms, and quality circles. We should highlight that the above variables on organizational practices are only available for a cross-section of 203 firms. This limits the possibility of controlling for unobserved heterogeneity and therefore this dataset can only be used for robustness checks.
4.4 Other variables and controls We measure the labor input in the production function estimations by the log of labor expenses (LABOR_EXP) in 1997 and 2000. We also use the following control variables: – YEAR: dummies for the time period (2000 and 2003) – ATECO: Dummies for the sector of activity of the firm, identified according to the ATECO 2-digit classification – PAVITT: Dummies for the four sectors of activity of the firm, identified according to the Pavitt classification. Table 2 reports that only 2.4% of firms operate in high tech sectors, while most firms belong to traditional and specialized suppliers sectors (about 51 and 28%, respectively) – REGION:Dummies for the 20 Italian regions in which the firm is located
20 of 36
P. Giuri et al.
– MACRO_REGION: Dummies for the four Italian macro-regions in which the firm is located (North-East, North-West, Centre, South). Table 2 shows the composition of the sample in terms of firm size: 66.8% of the firms are small, 26.7% are medium, and only 6.4% are large.9 Compared to previous studies on the complementarity between ICT, skills, and workplace organization, this sample clearly provides a better focus on SMEs.
5. Results 5.1 Correlations among residuals Table 3 reports the Pearson’s correlation between residuals of the reduced-form OLS regressions of the potential complements (ICT, SKILL, and organizational variables) against a set of controls, including the log of the number of employees and dummies for geographical regions, 2-digit ATECO sectors, and time dummies. A positive correlation between residuals provides evidence consistent with the hypothesis of pair-wise complementarity. The main results are the following. ICT is positively correlated with the share of skilled workers and with ORG_CHANGE. However, since organizational change is related to innovation activities, we cannot unambiguously interpret the correlation between ICT and organizational change without controlling for the presence of innovation in the regression. This will be done in our productivity analysis. Moreover, ICT does not appear to be significantly correlated with several modern organization practices. We also find that skills are positively correlated with organizational change and with the other measures of firms’ organizational practices, but the coefficients are not significant. Finally, the correlations between residuals of non-ICT stock and organizational variables are not significant. Non-ICT stock is only significantly correlated with the acquisition of new plants.10 Overall, these findings are in line with earlier works showing that the relationships among skills, organizational change, and technical change are specific to ICT rather than being a general pattern of technical change (see e.g., Bresnahan et al., 2002). 9 It is worth noting that, according to the Italian Statitical Institute (ISTAT), 98% of Italian firms have between 1 and 49 employees and about 56% of firms are located in North Italy. Our sample over-represents medium and large firms located in Northern regions. At the same time, we decided not to adopt probability weighted estimation techniques because in about 37% of strata identified by Mediocredito Centrale, we have at most one observation in our final sample. 10
Capital goods are a typical channel of embodied technical progress especially in traditional sectors.
Table 3 Correlations between residuals in OLS estimations of ICT, skills, and organizational practices (1)
(2)
(1) (2)
Log(ICT stock) 1 Log (Non ICT stock) 0.087***
(3)
SKILL
0.190*** 0.030
(4)
ORG_CHANGE
0.191***
(5)
MULTIF_TEAM
0.073
(6)
SELFMAN_TEAM
(7) (8)
(3)
(4)
(5)
(6)
(7)
(8)
(9)
(10)
(11)
(12)
1 1 0.024
1
0.047
0.094
0.027
0.152**
0.049
0.065
0.124*
0.597*** 1
TEAM_ PROM
0.078
0.046
0.029
0.027
0.451*** 0.512***
1
TEAM_ PERF
0.089
0.111
0.034 0.027
0.408*** 0.477***
0.541*** 1
(9) QUALITY_CIRCLE (10) INFO_SHARING
0.119* 0.071
0.010 0.013
0.060 0.043 0.047 0.001
0.760*** 0.704*** 0.336*** 0.370***
0.481*** 0.492*** 1 0.410*** 0.318*** 0.401*** 1
(11) EXCH_SUPPLIER
0.100
0.087
0.095
(12) NEWPLANT
0.071
0.130*
0.015 0.000
0.043
1
0.342*** 0.340*** -0.028
0.034
0.266*** 0.374*** 0.384*** 0.346*** 1 0.095
0.037
0.109
0.073
0.091 1
Notes. (1) *P50.10, **P50.05, ***P50.01. The number of observations is 965 in (1)–(4) and 203 in (5)–(12). (2) Pearson’s correlation coefficients between residuals obtained from OLS regressions of ICT stock, Non-ICT stock, SKILL, and various organizational practices on the log number of employees and dummy variables for region, sector (2 digits) and year.
ICT, skills, and organizational change
0.005
21 of 36
22 of 36
P. Giuri et al.
It is also worth noting that the various measures of organizational practices are positively and significantly correlated with each other, suggesting that modern firms adopt clusters of different organizational practices. It is important to recall that the correlation analysis above represents only a first descriptive check of complementarity, since it does not account for the possibility of unobserved exogenous factors unrelated to complementarity or a common measurement error that may affect the hypothesized variables in a correlated way.
5.2 Estimations of the production function We perform three sets of estimations to study the complementarity in the context of the production function. It is worth noting that the production function approach aims at capturing differences across firms that use different combinations of complements. As mentioned before, differences in adaptation costs, information barriers, and timing will result in different short-term combinations of complements, which in turn will be transferred into different productivity levels. We first run OLS estimations of the log of value added in the pooled sample of 965 observations. We perform four regressions models in which we include alternatively the interaction terms between ICT and skills, skills and organizational change, ICT and organizational change, and all three pair-wise interactions together with the interaction term among the three complements (models 1–4 in Table 4). We also estimate two models including the interaction term between ICT and skills in the sub-samples of firms that did and did not perform organizational changes, respectively (models 5 and 6 in Table 4). As expected, the main production inputs—non-ICT capital and labor—appear to be the most significant components of the production output in all regression models. Both indicators of the adoption of new technologies—ICT and innovation— are also significantly correlated with firms’ output. In contrast, skilled labor and organizational change do not seem to yield a significant productivity premium to the average firm in our sample. It is important to emphasize that since our measure of organizational change is linked to the firms’ innovation activity by definition, we control for the presence of innovations in all regressions.11 The test of pair-wise complementarity between ICT, skills, and organizational changes produces the following results. As shown by the non-significant coefficient for the interaction term in model 1 of Table 4, we do not find evidence of complementarity between ICT and skills. Instead, we find evidence of complementarity between organizational changes and skilled labor. On average, organizational change has no strong direct effect; however, the productivity of skilled labor increases significantly with the introduction of 11
We also run regressions that do not include interaction terms between the complements and the results for our measures of ICT, skills, and organizational change remain very similar.
Table 4 OLS estimations. Dependent variable: log of value added (1) ICT_SKILL Log(LaborExp) Log(NonICTstock) Log(ICTstock) SKILL
(3) ICT_ORG
(4) ICT_SKILL_ORG
0.935*** (0.022) 0.934*** (0.021) 0.938*** (0.022) 0.942*** (0.022) 0.922*** (0.036) 0.036*** (0.011) 0.035*** (0.011) 0.035*** (0.011) 0.035*** (0.011) 0.061*** (0.016) 0.030** (0.014) 0.025*** (0.009) 0.117 (0.151)
0.035** (0.011)
0.085 (0.097)
0.011 (0.033)
ORG_CHANGE
0.091* (0.054)
SKILL*ORG_CHANGE
0.293** (0.126)
ICT*ORG_CHANGE
0.075 (0.063)
0.034** (0.023)
0.042** (0.019)
0.261 (0.212)
0.218 (0.220)
0.657*** (0.215)
0.043 (0.053)
0.014 (0.054)
0.095** (0.044)
0.184 (0.113) 0.848*** (0.298)
0.018 (0.015)
0.019 (0.028) 0.045 (0.055)
0.256*** (0.092)
0.032 (0.404)
0.216 (0.354)
0.115** (0.045)
0.111** (0.046)
0.108** (0.046)
0.127 (0.069) 0.109** (0.046)
constant
0.709** (0.304)
0.756** (0.298)
0.712** (0.299)
0.709** (0.308)
Adj. R2
965 0.848
965 0.849
0.960*** (0.028) 0.017 (0.014)
0.022** (0.022)
ICT*SKILL*ORG_CHANGE INNOVATION N
(5) (6) ICT_SKILL No ORG_CHANGE ICT_SKILL ORG_CHANGE
965 0.848
965 0.85
444 0.81
521 0.873
Note. *P50.10, **P50.05, ***P50.01. Standard errors are shown in parentheses. All regressions include dummies for YEAR, REGION, and ATECO classification.
ICT, skills, and organizational change
ICT*SKILL
(2) SKILL_ORG
23 of 36
24 of 36
P. Giuri et al.
organizational change (model 2). We do not find evidence of complementarity between ICT and organization (model 3). In model 4 we check for evidence of threefold complementarity among ICT, skills, and organizational change by including all pair-wise interaction terms and an interaction term for the three complements. Interestingly, the interaction term ICT*SKILL*ORG_CHANGE has a negative sign and the coefficient is almost significant, while all pair wise interactions keep the same sign and significance levels. This result would suggest that while on average there are no complementary effects arising from the adoption of ICT and skilled labor, when organizational changes are also implemented within the firm, there are negative short-term effects on productivity. To better understand this result we estimate the regression model 1 for the sub-samples of firms that did not implement organizational changes (model 5) and for firms that did implement some organizational change, respectively (model 6). The results are quite revealing, as they show that only in the sample of firms that performed some organizational change the coefficient for skilled labor is positive and highly significant. This is consistent with the results of models 2 and 4 showing that on average the productivity of highly educated employees is higher in firms that perform organizational changes. Moreover, we confirm the findings of model 4, as the interaction term between ICT and skills in model 6 is now negative and significant. The sign of this interaction is not significant in the sample of firms that did not perform organizational changes. These findings do not support the organizational complementarity hypothesis, as the joint investment in ICT, qualified employees, and organizational change have negative effects on firms’ productivity. In the second set of regressions (Table 5) we perform fixed effect estimations of the same six models discussed above. As already described in Section 4.1, because of missing observations for some regressors in one of the two periods, the number of firms in the panel drops to 285. The main findings of the OLS regressions described in Table 4 are confirmed when we run fixed-effect estimations.12 Consistently with other studies (Griliches 12
To account for a potentially important source of human capital accumulation, we also employed in unreported regressions a variable measuring the percentage of employees that take part in training courses. As several studies have pointed out, hiring and dismissing involve substantial adjustment costs, which are particularly high in the case of high-skilled workers (Hamermesh and Pfann, 1996; Hempell, 2003). Given that much of the knowledge related to the internal processes and organization is firm-specific, firms could find it easier to train their current employees rather than hiring new employees. In our sample, the share of workers participating in courses is only 3.8%. We first run OLS estimations and find some evidence of complementarity between ICT and courses that might support our hypothesis. A possible interpretation of this result is that courses may be specifically devoted to the acquisition of ICT skills, and this is quite reasonable in the period covered in our analysis, which was characterized by the ICT boom. However, this effect is very weak as it disappears in fixed effect estimations.
Table 5 Fixed effect estimations. Dependent variable: log of value added ICT_SKILL
SKILL_ORG
ICT_ORG
ICT_SKILL_ORG
ICT_SKILL No ORG_CHANGE ICT_SKILL ORG_CHANGE
(1)
(2)
(3)
(4)
(5)
(6)
Log(LaborExp)
0.573*** (0.128)
0.598*** (0.128) 0.574*** (0.128) 0.644*** (0.130) 0.622*** (0.216)
0.612*** (0.170)
Log(NonICTstock) Log(ICTstock)
0.043** (0.021) 0.041** (0.021) 0.042** (0.021) 0.044** (0.021) 0.071** (0.031) 0.05** (0.021) 0.058*** (0.015) 0.042* (0.021) 0.063* (0.034) 0.071** (0.035)
0.02 (0.030) 0.033 (0.027)
SKILL ICT*SKILL
0.14 (0.218)
0.108 (0.130)
0.019 (0.046)
SKILL*ORG_CHANGE
–
–
–
–
0.291* (0.172)
ICT*ORG_CHANGE ICT*SKILL*ORG_CHANGE INNOVATION Constant N Number of firms in panel Adj. R2
0.275 (0.303)
0.814** (0.340)
0.049 (0.076)
0.058 (0.076)
0.129** (0.065)
– – 1.129** (0.448)
-0.025 (0.028)
0.027 (0.042)
–
–
–
0.184* (0.099) –
–
–
–
–
3.844*** (0.883)
– –
3.684*** (0.876) 3.871*** (0.869) 3.329*** (0.899) 3.695*** (1.474)
– – 3.417*** (1.188)
965
965
965
965
444
521
285 1.929
285 1.9
285 1.916
285 1.87
121 2.164
164 1.679
25 of 36
Note. *P50.10, **P50.05, ***P50.01. Standard errors are shown in parentheses. All regressions include dummies for YEAR, REGION, and ATECO classification.
ICT, skills, and organizational change
ORG_CHANGE
0.276 (0.304)
26 of 36
P. Giuri et al.
and Mairesse, 1995), in fixed-effect estimations the coefficients of the main inputs fade away with respect to OLS estimations.13 We opted for fixed-effect estimations rather than the random effect model because the Hausman test rejects the orthogonality hypothesis in our sample in all model specifications. We also find that the coefficients on SKILL*ORG_CHANGE and the three-term interaction have the same signs as in the OLS regressions and maintain their significance levels. We finally carry out a set of regressions that exploit the more detailed information about the firms’ use of modern organizational practices (Table 6). For the subsample of 203 firms for which these data are available we estimate by OLS the log of value added in the absence and in the presence of various organizational practices, respectively (models 2–15). We also run the same regression with a proxy of organizational change—the acquisition of new plants (models 16 and 17). As a reference point, in column 1 of Table 6 we report the OLS estimation of the complete sample of 203 firms. Overall, Table 6 confirms our previous findings about the presence of SBOC and the negative complementarity between skills and ICT in the presence of modern organizational practices. In most models, the productivity premium of highly skilled workers is positive and significant when firms intensively use self-managed teams, incentive schemes linked to team performance, quality circles, and exchange of workers with suppliers. In these cases the coefficient on the interaction between ICT and skills is negative and significant, thus rejecting the hypothesis of organizational complementarity. In other cases such as multi-functional teams, where teamwork is a criteria for career advancement and information sharing with employees, the coefficients for SKILL and ICT* SKILL have the same signs as those discussed, albeit at a lower level of significance. Finally, the previous results are confirmed when a dummy for organizational changes related to the acquisition of new plants is used (models 16 and 17). Although these findings are based on a relatively small cross-section of firms that does not allow for panel data estimation, they provide further evidence on the impact
13
The negative coefficient on non ICT-capital stock in the fixed effect estimations requires some explanation. One reason is probably the attenuation bias discussed in Section 3.2. Another plausible explanation is self-selection induced by exit behavior, which implies a downward bias in the capital coefficient. As Olley and Pakes (1997) have argued, firms’ exit decisions are made, at least in part, on the basis of their expected future productivity. If profits are increasing in capital stocks whereas productivity decreases with capital stocks then survivors will continue to produce at decreasing levels of productivity (“firms with larger capital stocks can expect larger future returns for any given level of productivity, and hence will continue in operation at lower [productivity] realizations”) (p. 1274). As a consequence, the capital coefficient will be downward biased if we do not account for self-selection induced by exit behavior.
Table 6 OLS estimations. Dependent variable: log of value added TOTAL SAMPLE
MULTIF_TEAM
MULTIF_TEAM
SELFMAN_TEAM
SELFMAN_TEAM
TEAM_PROM
TEAM_PROM
TEAM_PERF
TEAM_PERF
–
No
Yes
No
Yes
No
Yes
No
Yes
(1)
(2)
(3)
(4)
(5)
(6)
(7)
(8)
(9)
Log(LaborExpenses) 0.960*** (0.061) 0.815*** (0.098) 0.973*** (0.084) 0.838*** (0.113) 0.908*** (0.073) 0.817*** (0.080) 1.023*** (0.097) 0.811*** (0.101)
1.018*** (0.081)
0.028 (0.024)
0.037 (0.035)
0.032 (0.033)
0.005 (0.041)
0.063** (0.028)
0.031 (0.026)
0.007 (0.045)
0.029 (0.043)
0.025 (0.030)
Log(ICTstock)
0.045 (0.052)
0.072 (0.085)
0.06 (0.079)
0.027 (0.101)
0.107* (0.061)
0.089 (0.071)
0.047 (0.077)
0.064 (0.087)
0.117* (0.068)
0.146 (0.784)
1.515** (0.601)
SKILL ICT*SKILL Cons N Adj R2
0.645 (0.461)
0.068 (0.845)
1.054 (0.649)
0.016 (0.920)
0.990* (0.527)
0.174 (0.615)
1.051 (0.685)
0.111 (0.096)
0.017 (0.209)
0.180 (0.133)
0.013 (0.210)
0.188* (0.109)
0.044 (0.131)
0.169 (0.140)
0.462 (0.372)
0.868 (0.541)
0.499 (0.555)
1.139* (0.613)
0.564 (0.469)
1.283* (0.659)
0.092 (0.525)
0.069 (0.169) 0.255** (0.121) 1.292** (0.616)
0.421 (0.490)
203
87
116
92
111
79
124
94
109
0.785
0.736
0.801
0.653
0.871
0.81
0.769
0.666
0.855
(continued)
ICT, skills, and organizational change
Log(NonICTstock)
27 of 36
28 of 36
Log(LaborExpenses)
QUALITY CIRCLE
QUALITY CIRCLE
EXCH_SUPPLIER
EXCH_SUPPLIER
INFO_SHARING
INFO_SHARING
NEWPLANT
NEWPLANT
No
Yes
No
Yes
No
Yes
No
Yes
(10)
(11)
(12)
(13)
(14)
(15)
(16)
(17)
1.011*** (0.066)
0.837*** (0.122)
0.970*** (0.075)
0.918*** (0.074)
1.032*** (0.101)
0.854*** (0.104)
0.992*** (0.075)
0.864*** (0.124)
Log(NonICTstock)
0.035 (0.039)
0.030 (0.032)
0.040 (0.026)
0.024 (0.053)
0.095** (0.038)
0.009 (0.032)
0.014 (0.046)
0.038 (0.028)
Log(ICTstock)
0.031 (0.107)
0.088 (0.063)
-0.026 (0.065)
0.146* (0.078)
0.021 (0.098)
0.047 (0.062)
0.040 (0.109)
0.072 (0.059)
SKILL ICT*SKILL cons N Adj R2
0.534 (0.990)
1.151** (0.552)
0.227 (0.581)
2.022*** (0.682)
0.896 (0.863)
0.924* (0.552)
-0.214 (1.032)
1.119** (0.505)
0.117 (0.227)
0.184 (0.111)
0.102 (0.121)
0.429*** (0.145)
0.104 (0.180)
0.160 (0.114)
0.139 (0.225)
0.218** (0.103)
1.029 (0.748)
0.171 (0.453)
0.746* (0.431)
0.193 (0.657)
0.518 (0.583)
0.563 (0.476)
1.178 (0.715)
0.139 (0.420)
77
126
140
63
57
146
78
125
0.575
0.845
0.788
0.85
0.808
0.789
0.71
0.845
Note. *P50.10, **P50.05, ***P50.01. Standard errors are shown in parentheses. All regressions include: dummies for the regions with a large number of observations in the sample (Lombardia and Veneto) and for the four Italian macro-areas for the other regions, and dummies for the sectors with a large number of observations (machinery, textile, metal) and for Pavitt sectors in all other cases. We also performed regressions including only the dummies for PAVITT and MACRO-AREA; results do not change significantly.
P. Giuri et al.
Table 6 Continued
ICT, skills, and organizational change
29 of 36
of organizational changes and adoption of modern organizational practices, which is consistent with the results on the full sample. Overall, our results are also in line with previous evidence on the absence of complementarity between technical change and skilled labor, and on the presence of SBOC (Piva and Vivarelli, 2004; Piva et al., 2005). In addition, we find evidence that rejects the hypothesis of organizational complementarity. Why do organizational changes have such a negative effect on the productivity gains from the joint adoption of ICT and human capital? In general, the costs of the simultaneous adoption of ICT and skilled labor tend to increase because different dimensions of the organization are involved in the process of change, and this may result in higher coordination costs. This escalation of organizational costs calls our attention to a “classic” tension between stability and modification of organizational routines. As Nelson and Winter have noted, “the routinization of activities in an organization constitutes the most important form of storage of the organization’s specific operational knowledge” (Nelson and Winter, 1982: 99) and the basis for building the organizational capabilities. However, routines can become a source of organizational inertia when the firm has to solve unexpected problems or to undertake simultaneously different avenues of change such as ICT, skills, and organizational change. Large innovative firms possess strong meta-routines or dynamic capabilities that help to anticipate the change. However, even large, successful firms often fail to adapt rapidly to disruptive innovations, i.e., changes that make existing routines and capabilities obsolete (Tushman and Anderson, 1986; Henderson and Clark, 1990; Christensen and Rosenbloom, 1995). Even more puzzling is the case of firms that undertake continuous changes in their organizational routines that result in negative performance outcomes. A case in point is constituted by Lockheed, whose policies of continuous internal labor mobility and other related innovative strategies adopted in the production of the L-1011 TriStar contributed to reduced labor productivity of 40–50% per year (Benkard, 2000). However, we believe that the negative effect of organizational changes on the ICTskills complementarity that we find in our estimations is largely due to structural difficulties of SMEs in dealing with the complexity of multiple co-inventions. Recall that our sample is mainly composed of small and medium firms; large firms represent only about 6% of the total sample. Given the small number of observations for large firms, we cannot make reliable comparisons among firms in different size classes; however, in unreported regressions we replicated the analysis reported in this article excluding large firms from the sample and all results are confirmed. Thus our results show that in SMEs there is limited scope for productivity gains associated with the simultaneous adoption of different complementary strategies. Even when they succeed in adopting the full cluster of co-conventions, SMEs reveal a limited ability to manage a complex set of co-inventions. As mentioned before, the timing of the multidimensional innovation process may account in part for our findings. Our results suggest that inertia and
30 of 36
P. Giuri et al.
a limited absorptive capacity call for an incremental, gradual adoption of multiple co-inventions as opposed to a strategy of contemporaneous adoption. This is not in contrast with the theory of organizational complementarity, which does not make any assumption about the sequence of adoption of different co-inventions. However, the shortness of our panel does not allow for studying the inter-temporal adoption of complementary strategies and their long-run productivity effects. A full test of the organizational complementarity hypothesis in a dynamic setting will be carried out in our future research. A longer panel will allow for a better understanding of whether or not firms that gradually introduce innovations have the possibility to learn by experimenting with different dimensions of organizational change without completely disrupting existing routines.
6. Conclusions This article provides new empirical evidence on the relationships among ICT, skills, and organizational change in SMEs. Overall, our results provide support to the hypothesis of complementarity between skills and organizational change, and do not provide support to the hypothesis of complementarity between ICT and skills. We also find that organizational change yields negative effects on the complementarity between ICT and human capital. In line with earlier works, we analyzed complementarity in the context of the production function by estimating the marginal effects of the hypothesized complements, their pair-wise interactions, and threefold interactions. Our findings are not consistent with the complementarity theory by Milgrom and Roberts (1990) and the hypothesis of organizational complementarity between ICT and skills developed by Bresnahan (1999) and Bresnahan et al. (2002). According to the organizational complementarity hypothesis, only the full combination of co-inventions should generate significant productivity gains at the firm level, whereas intermediate combinations should not produce significant benefits to the firm. In contrast, we found a negative and significant effect of the adoption of the three co-inventions, which points to substitution rather than complementarity. Our results are consistent with the hypothesis that small firms experience a tradeoff between organizational change and ICT. This is true for several reasons. For instance, the limited size of the firms might preclude the use of complex organizational computing. At the same time, given the simplicity of SMEs’ organizational structures, an intensive use of ICTs associated with skilled people and new organizational practices might unnecessarily overburden the educated labor. Despite the richness of the employed datasets, our findings are limited by some data shortcomings worth mentioning. First, as in earlier studies, we are able to measure accurately the nature of workplace organization that the theory indicates as a complement to ICT and skills only for a cross-section of firms in our sample.
ICT, skills, and organizational change
31 of 36
Second, the study of complementarity in the context of the production function implies that we analyze the effects of different strategies on the best level of efficiency that is potentially attainable by the firm with a given production technology. However, the productivity effects of a new cluster of co-inventions take a long time to materialize. Even if a firm adopted all the complements (ICT, skills, and organizational co-inventions) it may have to learn how to manage the new production system before it would be able to fully enjoy the productivity gains arising from the new production systems. Because of their greater difficulties in implementing a cluster of co-inventions as compared with large firms, the average SME will probably take a longer time to benefit from the productivity gains promised by ICT. With panel data over a longer time span, short-term differences in adaptation costs should evaporate. Moreover, with a longer panel the productivity effects of transitions from one combination of complements to another could be analyzed more deeply. The availability of longer panel data in future research will allow for the employing of dynamic panel techniques to better account for the potential lagged productivity effects of the joint use of complementary inputs, and at the same time, to control for non-serially correlated time-variant heterogeneity. Third, our analysis is centered on SMEs. The limited number of large firms in our sample does not allow us to analyze more carefully complementarity in firms of different size. This is an issue that should be taken into account in future research. Finally, our sample is mainly composed of firms in traditional sectors. One may wonder whether this contributes to the lack of complementarity observed. However, this conclusion would be at odds with the difficulties experienced by high-tech firms like Lockheed that adopt different co-inventions. In contrast, we are aware of firms in traditional sectors such as Zara (from Spain) that appear to be successful at managing modern manufacturing systems and new technologies (McAfee et al., 2004). Although our analysis is far from being definitive, our results suggest that the test for the complementarity hypothesis requires a deeper understanding of the multiplicity of interactions between the hypothesized complements in different organizational environments. Finally, further analysis should be done to explain the drivers of complementarity by looking at more detailed data on firm-specific characteristics (Cassiman and Veugelers, 2002).
Acknowledgement We thank Sergio Lugaresi of Capitalia Bank for providing us with the data of two Mediocredito Surveys of Italian manufacturing firms. We also thank two anonymous referees and Elettra Agliardi, Ashish Arora, Manuel F. Bagu¨e´s, Bruno Cassiman,
32 of 36
P. Giuri et al.
Giovanni Dosi, Alfonso Gambardella, Keld Laursen, Mario Morroni, Reinhilde Veugelers, and participants in seminars held in London, Manchester, Milan, and Pisa for valuable comments to earlier drafts of this article. All errors are ours. We acknowledge financial support from the European Commission IHP Grant N. HPSE-CT-2002-00146.
Addresses for correspondence Paola Giuri, Laboratory of Economics and Management, Sant’Anna School of Advanced Studies, piazza Martiri della Liberta`, 33, 56127 Pisa, Italy. e-mail:
[email protected] Salvatore Torrisi, Department of Management, University of Bologna, via Capo di Lucca 34, 20126 Bologna, Italy. e-mail:
[email protected] Natalia Zinovyeva, Laboratory of Economics and Management, Sant’Anna School of Advanced Studies; BETA, Universite´ Louis Pasteur, Strasbourg, France. e-mail:
[email protected]
References Acemoglu, D. (1998), ‘Why do new technologies complement skills? Directed technical change and wage inequality,’ Quarterly Journal of Economics, 113, 1055–1090. Aghion, P. (2002), ‘Schumpeterian growth theory and the dynamics of income inequality,’ Econometrica, 70, 855–882. Arora, A. (1996), ‘Testing for complementarities in reduced-form regressions: a note,’ Economic Letters, 50, 51–55. Arora, A. and A. Gambardella (1990), ‘Complementarity and external linkages: the strategies of the large firms in biotechnology,’ Journal of Industrial Economics, 38, 361–379. Arvanitis, S. (2005), ‘Computerization, new workplace organization, skilled labor and firm productivity: evidence for the Swiss business sector,’ Economics of Innovation and New Technology, 14, 225–249. Athey, S. and S. Stern (1998), ‘An empirical framework for testing theories about complementarity in organizational design,’ NBER Working Papers 6600. Autor, D. H., F. L. Katz and A. B. Krueger (1998), ‘Computing inequality: have computers changed the labor market?,’ Quarterly Journal of Economics, 113, 1169–1213. Bagues, M. F. (2004), Complementarity in Innovation Strategies: Evidence from Pharmaceutical Dynamic Panel Data, Chapter from the unpublished PhD dissertation, Universitat Pompeu Fabra. Belfield, C. R. (1999), ‘The behaviour of graduates in the SME labour market: evidence and perceptions,’ Small Business Economics, 12, 249–259.
ICT, skills, and organizational change
33 of 36
Benkard, C. L. (2000), ‘Learning and forgetting: the dynamics of aircraft production,’ American Economic Review, 90, 1034–1054. Berman, E., J. Bound and Z. Griliches (1994), ‘Changes in the demand for skilled labor within U.S. manufacturing: evidence from the annual survey of manufactures,’ Quarterly Journal of Economics, 109, 367–397. Berman, E., J. Bound and S. Manchin (1997), ‘Implication of skill-biased technological change: international evidence,’ NBER Working Paper 6166. Bertschek, I. and U. Kaiser (2004), ‘Productivity effects of organizational change: microeconometric evidence,’ Management Science, 50, 394–404. Black, S. E. and L. M. Lynch (2001), ‘How to compete: the impact of workplace practices and information technology on productivity,’ The Review of Economics and Statistics, MIT Press, 83, 434–445. Bratti, M. and N. Matteucci (2004), ‘Is there skill-biased technological change in Italian manufacturing? Evidence from firm-level data,’ Quaderni di ricerca, No. 202, Dipartimento di Economia, Universita` Politecnica delle Marche. Bresnahan, T. (1999), ‘Computerisation and wage dispersion: an analytical reinterpretation,’ Economic Journal, 109, 390–415. Bresnahan, T. and M. Trajtenberg (1995), ‘General purpose technologies: engines of growth?,’ Journal of Econometrics, 65, 83–108. Bresnahan, T., E. Brynjolfsson and L. Hitt (2002), ‘Information technology, workplace organization and the demand for skilled labor: firm-level evidence,’ Quarterly Journal of Economics, 117, 339–376. Brynjolfsson, E. and L. Hitt (2000a), ‘Computing productivity: firm-level evidence,’ Working Paper, MIT Sloan School. Brynjolfsson, E. and L. Hitt (2000b), ‘Beyond computation: information technology, organizational transformation, and business performance,’ Journal of Economic Perspectives, 14, 23–48. Caroli, E. and J. Van Reenen (2001), ‘Skill-biased organizational change? Evidence from a panel of British and French establishments,’ Quarterly Journal of Economics, 116, 1449–1492. Caroli, E., N. Greenan and G. Guellec (2001), ‘Organizational change and skill accumulation,’ Industrial and Corporate Change, 10, 479–504. Casavola, P., A. Gavosto and P. Sestito (1996), ‘Technical progress and wage dispersion in Italy: evidence from data,’ Annales d’economie et statistique, n. 41–42, 387–412. Cassiman, B. and R. Veugelers (2002), ‘Complementarity in innovation strategy: internal R&D, external technology acquisition, and cooperation in R&D’, CEPR Discussion Paper No. 3284, Katholieke Universiteit Leuven. Christensen, C. and R. Rosenbloom (1995), ‘Explaining the attacker’s advantage: technological paradigms, organizational dynamics and the value network,’ Research Policy, 24, 233–257. Clark, P. (2003), Organizational Innovations. Sage Publications: London.
34 of 36
P. Giuri et al.
Corsino, M., P. Giuri and S. Torrisi (2005), ‘L’adozione di software proprietario e open source nelle imprese italiane. Indagine su un campione di imprese manifatturiere italiane,’ Research report, Laboratory of Economics and Management, Sant’Anna School of Advanced Studies, Pisa, February. David, P. and G. Wright (1999), ‘Early twentieth century productivity growth dynamics: an inquiry into the economic history of ‘our ignorance’,’ SIEPR Discussion Paper No. 98-3, Stanford Institute of Economic Policy Research, Stanford University, Stanford CA. Dosi, G. (1982), ‘Technological paradigms and technological trajectories: a suggested interpretation of the determinants and directions of technical change,’ Research Policy, 11, 147–162. Erickson, C. L. and A. C. Ichino (1994), ‘Wage differentials in Italy: market force, institutions and inflation,’ in R. Freeman and L. F. Katz (eds), Differences and Changes in the Wage Structure. Chicago University Press and NBER: Chicago, pp. 105–127. European Commission (2003), ‘Adapting e-business policies in a changing environment: the lessons of the Go Digital initiative and the challenges ahead,’ Communication from the Commission to the Council, the European Parliament, the European Economic and Social Committee and the Committee of the Regions, Commission of the European Communities, COM(2003) 148 Final, Brussels 27 March. Eurostat (2007), http://epp.eurostat.ec.europa.eu. Fisher, J. and V. Govindarajan (1992), ‘Profit center manager compensation: an examination f market, political and human capital factors,’ Strategic Management Journal, 13, 205–217. Freeman, C. and C. Perez (1988), ‘Structural crises of adjustment: business cycles and investment behaviour,’ in G. Dosi, C. Freeman, R. Nelson, G. Silverberg and L. Soete (eds), Technical Change and Economic Theory. Pinter: London, pp. 38–66. Gordon, R. (1990), Measurement of Durable Goods Prices. University of Chicago Press (for NBER): Chicago. Greenan, N. (2003), ‘Organisational change, technology, employment and skills: an empirical study of French manufacturing,’ Cambridge Journal of Economics, 27, 287–316. Gretton, P., J. Gali and D. Parham (2002), ‘Uptake and impacts of ICT in the Australian economy: evidence from aggregate, sectoral and firm levels,’ Paper presented in The OECD Workshop on ICT and business performance, Paris, December 9. Griliches, Z. and J. Mairesse (1995), ‘Production functions: the search for identification,’ NBER Working Paper Series 5067, National Bureau of Economic Research, Inc. Hall, B. H. (1990), ‘The manufacturing sector master file: 1959–1987,’ NBER Working Paper No. 3366. Hamermesh, D. S. and G. A. Pfann (1996), ‘Adjustment costs in factor demand,’ Journal of Economic Literature, 34, 1264–1292. Harland, C. M., N. D. Caldwell, P. Powell and J. Zheng (2007), ‘Barriers to supply chain information integration: SMEs adrift of eLands,’ Journal of Operations Management, 25, 1234–1254.
ICT, skills, and organizational change
35 of 36
Hempell, T. (2003), ‘Do computers call for training? Firm-level evidence on complementarities between ICT and human capital investments,’ ZEW Discussion Paper No. 03-20, Mannheim. Henderson, R. M. and K. B. Clark (1990), ‘Architectural innovation: the reconfiguration of existing product technologies and the failure of established firms,’ Administrative Science Quarterly, 35, 9–30. Ichniowski, C., C. Shaw and G. Prennushi (1997), ‘The effects of human resource management practices on productivity: a study of steel finishing lines,’ American Economic Review, 87, 291–313. Levy, M. and P. Powell (2000), ‘Information systems strategy for small and medium sized enterprises: an organizational perspective,’ Journal of Strategic Information Systems, 9, 63–84. Lokshin, B., M. Carree and R. Belderbos (2004), Testing for complementarity and substitutability in case of multiple practices, Universiteit Maastricht, Mimeo. Machin, S. and J. Van Reenen (1998), ‘Technology and changes in skills structure: evidence from seven OECD countries,’ The Quarterly Journal of Economics, 113, 1215–1244. Manacorda, M. (1996), Dualismo territoriale e differenziali di reddito. Un’analisi dell’andamento della dispersione dei redditi individuali nel Nord e Sud d’Italia, Mimeo. McAfee, A., A. Sjoman, and V. Dessain (2004), Zara: IT for fast fashion. Harvard Business School Case No. 9-604-081, Harvard University Press: Cambridge, MA. Milgrom, P. and J. Roberts (1990), ‘The economics of modern manufacturing,’ American Economic Review, 80, 511–528. Milgrom, P. and J. Roberts (1995), ‘Complementarities and fit. Strategy, structure, and organizational change in modern manufacturing,’ Journal of Accounting and Economics, 19, 179–208. Murnane, R., F. Levy and D. Autor (1999), Technological change, computers and skill demands: evidence from the back office operations of a large bank, Mimeo, NBER Economic Research Labour Workshop, June. Nelson, R. and S. Winter (1982), An Evolutionary Theory of Economic Change. Harvard University Press: Cambridge, MA. OECD (2007), OECD Statistics, http://stats.oecd.org/wbos/default.aspx?DatasetCode¼RPERS. Olley, G. S. and A. Pakes (1997), ‘The dynamics of productivity in the telecommunications equipment industry,’ Econometrica, 64, 1263–1297. Olsen, K. A. and P. Saetre (2007), ‘IT for niche companies: is and ERP system the solution?,’ Information Systems Journal, 17, 37–58. Piva, M. and M. Vivarelli (2004), ‘The determinants of the skill bias in Italy: R&D, organization or globalization?,’ Economics of Innovation and New Technology, 13, 329–347. Piva, M., E. Santarelli and M. Vivarelli (2005), ‘The skill bias effect of technological and organisational change: evidence and policy implications,’ Research Policy, 34, 141–157.
36 of 36
P. Giuri et al.
Thesmar, D. and M. Thoenig (2000), ‘Creative destruction and firm organization choice: a new look into the growth inequality relationship,’ Quarterly Journal of Economics, 115, 1201–1237. Tushman, M. L. and P. Anderson (1986), ‘Technological discontinuities and organisational environments,’ Administrative Science Quarterly, 31, 439–465. Wooldridge, J. M. (2002), Econometric Analysis of Cross Section and Panel Data. Cambridge: MIT Press, Mass. Wolak, F. (1989), ‘Testing inequality constraints in linear econometric models,’ Journal of Econometrics, 41, 205–235.
