77636 Technology Adoption and the Investment Climate: Firm-Level Evidence for Eastern Europe and Central Asia Paulo G. Correa, Ana M. Fernandes, and Chris J. Uregian Survey data for 7,000 ï¬?rms in 28 countries in Eastern Europe and Central Asia are used to examine the correlates of technology adoption proxied by ISO certiï¬?cation and web use. Complementary inputs such as skilled labor, managerial capacity, research and development, ï¬?nance, and good infrastructure are shown to be important correlates of technology adoption. The link between market incentives and technology adoption is more nuanced. While stronger consumer pressure is signiï¬?cantly associ- ated with technology adoption, competitor pressure is not, suggesting that in develop- ing economies where many input markets are imperfect, it is primarily ï¬?rms with rents that are able to adopt new technology. Foreign-owned ï¬?rms exhibit signiï¬?cantly better technology adoption outcomes, but privatized ï¬?rms with domestic owners do not. JEL codes: F1, F2, O3 Differences in technology—deï¬?ned broadly to encompass all types of knowledge relevant to the production of goods and services—are an important determinant of differences in total factor productivity across countries (Prescott 1998; Hall and Jones 1999). While new technology is generated in only a few research and development (R&D)-intensive economies, the expansion in the volume of capital goods trade indicates the broader availability of technology embodied in new machinery and equipment (Eaton and Kortum 2001). Yet, when faced with similar technological alternatives, ï¬?rms in some countries choose less efï¬?cient technologies even when more efï¬?cient ones are available because barriers to new technology adoption (such as those linked to the regulatory framework) distort the relative payoffs in favor of suboptimal technologies (Parente and Prescott 1994). Accordingly, countries have widely divergent living standards not because Paulo G. Correa ( pcorrea@worldbank.org) is a senior economist in the Europe and Central Asia region of the World Bank. Ana M. Fernandes (corresponding author, afernandes@worldbank.org) is an economist in the Development Research Group of the World Bank. Chris Uregian (curegian@worldbank .org) is a consultant in the Europe and Central Asia region of the World Bank. The authors thank three anonymous referees and the journal editor for helpful comments that substantially improved the article, as well as Brett Coleman, Itzahk Goldberg, Smita Kuriakose, and Jean-Louis Racine for their suggestions. THE WORLD BANK ECONOMIC REVIEW, VOL. 24, NO. 1, pp. 121 –147 doi:10.1093/wber/lhp021 Advance Access Publication January 12, 2010 # The Author 2010. Published by Oxford University Press on behalf of the International Bank for Reconstruction and Development / THE WORLD BANK. All rights reserved. For permissions, please e-mail: journals.permissions@oxfordjournals.org 121 122 THE WORLD BANK ECONOMIC REVIEW they access different stocks of knowledge, but because of differences in constraints imposed on technology choices (Parente and Prescott 2005). This article uses a unique data set to explore how the investment climate affects a ï¬?rm’s technology choices. The World Bank Enterprise Surveys for ï¬?rms in the Eastern Europe and Central Asia region in 2002 and 2005 contain information on technology adoption and on aspects of the investment climate related to complementary inputs (skilled labor, ï¬?nance) and the market incen- tives facing ï¬?rms (competition, ownership) for a large number of transition economies. Studies on technology diffusion often take a macroeconomic trade-related focus, as illustrated by Zeira (1998) and Keller (2004). Most of the applied microeconomic studies reviewed by Hall and Khan (2003) estimate diffusion curves for a few technologies in a few countries, providing limited under- standing of the barriers to technology adoption (Caselli and Coleman 2001; Comin and Hobijn 2008). Microeconomic evidence on the determinants of technology adoption in developing countries is scarce because of data limit- ations. This article attempts to ï¬?ll in this gap by studying the technology choices of ï¬?rms in Europe and Central Asia after a decade of transition to a market economy. The results show that access to appropriate complementary inputs—skilled labor, managerial capacity, R&D, ï¬?nance, and to a lesser extent good infrastructure—are strongly positively associated with International Organization for Standardization (ISO) 9000 certiï¬?cation and web use by ï¬?rms. The relationship between market incentives and technology adoption is more nuanced. While consumer pressure is related to ISO certiï¬?cation and web use, competitor pressure is not. Fully foreign-owned ï¬?rms and joint ventures exhibit signiï¬?cantly better technology adoption outcomes, but privatization to domestic owners is not systematically associated with more frequent technology adoption. The results provide evidence of robust correlations but cannot be interpreted as causal relationships because of the nature of the data. The article is organized as follows. Section I describes the technology adop- tion measures. Section II presents the empirical strategy and methodological issues. Section III presents the results. Section IV offers some policy impli- cations of the ï¬?ndings. I . D ATA AND TE C H N O LO GY A D O P TI O N ME A S U R E S The Enterprise Surveys are conducted by the World Bank and the European Bank for Reconstruction and Development in 28 Europe and Central Asia countries.1 The samples include a cross section of 6,667 ï¬?rms in 2002 and 9,655 ï¬?rms in 2005, plus a panel sample of 1,446 ï¬?rms surveyed in both 2002 1. The surveys were formerly known as Business Environment and Enterprise Performance Surveys and are described at www.enterprisesurveys.org. Correa, Fernandes, and Uregian 123 and 2005.2 The surveys cover manufacturing and services sectors and are representative of the universe of ï¬?rms by sector and location within each country.3 Two proxies are used for the adoption of new technology: ISO 9000 certiï¬?cation and use of the web (email and Internet) for business operations. ISO 9000 is a set of internationally accepted standards and technical regu- lations on quality management systems in manufacturing and services ï¬?rms developed by the ISO (1998).4 ISO certiï¬?cation focuses on improving a ï¬?rm’s operating processes to enhance quality and efï¬?ciency (Benner and Veloso 2008).5 ISO certiï¬?cation requires detailed review and documentation of the routines underlying delivery of the ï¬?rm’s products and services. The routines are subject to improvements to rationalize processes and streamline interfaces between the ï¬?rm’s subunits. Adopting standardized best practices throughout the ï¬?rm ensures that the organizational processes are repeated, allowing for continued efï¬?ciency improvements. ISO certiï¬?cation is awarded based on a detailed review of a ï¬?rm’s processes, documentation that the processes comply with ISO quality system standards, and an audit by an accredited third party (Arora and Asundi 1999). After ISO certiï¬?cation is awarded, regular audits are conducted to ensure continuing compliance. There are advantages and limitations in choosing ISO certiï¬?cation to proxy for new technology adoption, but on balance ISO certiï¬?cation captures rel- evant aspects of new technology adoption for developing country ï¬?rms. First, ISO certiï¬?cation indicates adherence to consistent process standards; thus it represents adoption of advanced organizational technology by ï¬?rms, a key component of technological knowledge (Lipsey and Carlaw 2004). However, ISO certiï¬?cation is intended to minimize variations in quality; it is not awarded for product or service quality and does not capture product design improve- ments by ï¬?rms. Second, though ISO certiï¬?cation does not map into the adop- tion of easily identiï¬?able technologies, it is objective and comparable across ï¬?rms, sectors, and countries, which is crucial for this study of correlates of technology adoption across ï¬?rms. Third, the adoption of international stan- dards and technical regulations through ISO certiï¬?cation is a major channel for U.K. and U.S. ï¬?rms to acquire technological information and introduce product and process technology upgrading (Blind and others 2005; Corbett, Montes-Sancho, and Kirsh 2005). For developing country ï¬?rms not yet operat- ing at the world technological frontier, this channel is likely to be even more important. Fourth, ISO certiï¬?cation facilitates the entry of local ï¬?rms into global supply chain networks, which often brings the transfer of knowledge from technologically advanced buyers (Humphrey and Schmitz 2000). 2. Safavian and Sharma (2007) use this panel sample to study ï¬?rm access to ï¬?nance. 3. Correa, Fernandes, and Uregian (2008) show the country and sector composition of the samples. 4. The principle underlying ISO certiï¬?cation is that better deï¬?ned and documented processes lead to better output (Arora and Asundi 1999). 5. An example of an operating process is a manufacturing assembly line, with discrete stages assigned to individuals and machines and a speciï¬?c product resulting at the end (Benner 1999). 124 THE WORLD BANK ECONOMIC REVIEW Web use is a proxy for a ï¬?rm’s use of information and communication technology in business operations. Information and communication technol- ogy, an advance that changed modes of production and operation, is con- sidered the preeminent general purpose technology of the last two decades. It is used pervasively across sectors, has prompted further innovation (Bresnahan and Trajtenberg 1995), and can have beneï¬?cial effects on productivity growth (Indjikian and Siegel 2005). There are also advantages and limitations in choosing web use to proxy for new technology adoption. First, web use captures the adoption of general-use information and communication technologies that allow ï¬?rms to process and transmit information faster, improve management and internal organization, and achieve efï¬?ciency gains. Second, web use is an objective measure, compar- able across ï¬?rms, sectors, and countries, making it well suited to a study of technology adoption correlates across ï¬?rms. One limitation of the web use measure is its inability to capture the adoption of important information and communication technologies applied to improve production (such as computer- aided manufacturing) or reduce coordination and transaction costs (such as local area networks). Another limitation is that it does not capture ï¬?rms’ inten- sity of usage. Three-quarters of ï¬?rms in Europe and Central Asia consider machinery and equipment acquisition as their main source of technological update, according to the 2005 survey.6 The technological advances conveyed by information and communication technology are embodied in new capital goods by nature: for example, for a ï¬?rm to use the web it needs to acquire compatible computers and telecommunication tools (Boucekkine, Del Rio, and Licandro 2003). Thus, the choice of web use as a proxy for technology adoption implicitly presumes the importance of capital-embodied technological change,7 while that is less obvious for ISO certiï¬?cation. The two proxies are the best available across ï¬?rms, sectors, and countries and enable examination of two dimensions of technology adoption not exploited before in a developing country context. However, they have limit- ations, and future research on the determinants of technology adoption should aim to collect cross-country ï¬?rm-level information on the intensity of usage, measured by total investments in high-technology capital or the percentage of 6. The other possible sources of technology updating were the hiring of key personnel/consultants with technological expertise; new license or turnkey operations from international sources; new license or turnkey operations from domestic sources; developed or adapted within the ï¬?rm; transferred from the parent company; developed in cooperation with customers; developed in cooperation with suppliers; obtained from business or industry associations; or obtained from universities or public institutions. The extent to which new technology is embodied in new capital equipment, the subject of long-standing debate, is not addressed here. 7. The importance of capital-embodied technological change for output growth is shown by Hulten (1992) and Long and Summers (1993), though Hulten (1992) also shows a role for disembodied technical change. Correa, Fernandes, and Uregian 125 employees using high-technology capital (such as ofï¬?ce, computing, and accounting machinery).8 In addition to ISO certiï¬?cation and web use, this study uses survey data on ï¬?rm ownership, size, workforce skills, managerial capacity, research and development (R&D) expenditures, access to credit, infrastructure failures faced, competition proxies, exports, and imported inputs (see appendix table A-1 for all variables used in the analysis and their deï¬?nitions). In 2002, 13.6 percent of ï¬?rms were ISO certiï¬?ed and 58.2 percent used the web (table 1). In 2005, 12.5 percent of ï¬?rms were ISO certiï¬?ed and 67.4 percent used the web. For both proxies and years, technology adoption is more prevalent in the EU-8 countries (deï¬?ned in appendix table A-1). There is sub- stantial heterogeneity in technology adoption across sectors (table 2). Firms in mining, quarrying, and construction are signiï¬?cantly more likely to be ISO cer- tiï¬?ed but no more likely than the average ï¬?rm to use the web.9 Firms in manu- facturing are signiï¬?cantly more likely to be ISO certiï¬?ed and use the web than are ï¬?rms in services (real estate being the exception). The ISO certiï¬?cation ï¬?nding is consistent with global experience of greater prevalence of certiï¬?cation in manufacturing, where quality signals matter for export competitiveness.10 In ordinary least squares regressions (not reported here), ISO certiï¬?cation and web use are strongly correlated with ï¬?rm performance in Europe and Central Asia, after controlling for sector ï¬?xed effects and GDP per capita (or country ï¬?xed effects) based on the 2002 or 2005 samples. ISO-certiï¬?ed ï¬?rms and web users exhibit signiï¬?cantly higher average value added per worker and faster sales growth and pay higher wages than ï¬?rms not adopting those tech- nologies, within sectors and countries. The data do not enable establishing causality for these estimated performance premia of technology adopters, but the strong correlation validates the use of ISO certiï¬?cation and web use as proxies for economically relevant technology upgrading by ï¬?rms. I I . E M P I R I CA L S T R AT E GY AND ME T H O D O LO G I CA L IS S U E S This section describes the empirical strategy, discusses the investment climate factors considered, and highlights the methodological issues associated with the empirical strategy. Empirical Strategy When deciding whether to adopt a new technology, ï¬?rms are assumed to make a proï¬?t-maximizing cost–beneï¬?t assessment of different alternatives. A ï¬?rm 8. Basant and others (2006) undertake a ï¬?rst effort in this direction for Brazilian and Indian ï¬?rms. 9. Adoption of ISO certiï¬?cation in mining and quarrying is likely associated with the resource-intensive exports of some transition economies, whereas in construction it may be explained by government requirements that contractors be ISO certiï¬?ed (Guler, Guillen, and Macpherson 2005). 10. In particular, EU directives requiring quality system registration have made ISO certiï¬?cation imperative for ï¬?rms in Europe and Central Asia aspiring to collaborate with EU ï¬?rms through supplier relationships (Guler, Guillen, and Macpherson 2005). 126 THE WORLD BANK ECONOMIC REVIEW T A B L E 1 . Summary Statistics (Percent Unless Otherwise Indicated) 2002 sample 2005 sample Panel sample Variable Mean Observations Mean Observations Mean Observations ISO-certiï¬?cation dummy 13.6 6,610 12.5 9,655 14.0 2,887 variable Web-use dummy variable 58.2 6,667 67.4 9,655 65.4 2,892 Share of skilled labor 18.0 6,572 17.0 9,519 20.0 2,634 Manager with college 70.2 6,611 — — — — education dummy variable Manager age (years) 44.6 6,610 — — — — R&D intensity 0.0 6,667 0.0 6,984 0.4 980 Access to ï¬?nance dummy 40.3 6,655 42.1 9,655 42.7 2,888 variable Number of days with power 11.0 6,656 9.7 9,645 15.9 2,881 outages Number of days with 5.8 6,653 1.7 9,635 3.2 2,868 unavailable telephone lines Dummy variable for market 65.7 6,667 — — — — share less than 5 percent Price – cost margin (percent) 18.9 5,656 22.7 8,460 20.4 2,045 Dummy variable for pressure 74.1 6,586 75.2 9,538 73.4 2,852 to innovate from competitors Dummy variable for pressure 71.7 6,560 72.2 9,466 69.2 2,838 to innovate from consumers Number of permanent 143 6,636 102 9,654 109 2,888 workers Dummy variable for 14.4 6,667 13.7 9,655 15.9 2,892 privatized ï¬?rm Dummy variable for private 71.7 6,667 77.6 9,655 69.8 2,892 ï¬?rm (from start-up) Dummy variable for fully 6.6 6,667 5.1 9,655 7.4 2,892 foreign-owned ï¬?rm Dummy variable for joint 9.5 6,667 6.8 9,655 6.4 2,892 venture Ownership share of largest 76.7 6,180 76.8 9,563 76.3 2,788 shareholder Export share 11.4 6,636 10.1 9,642 10.6 2,886 Imported inputs share 29.8 6,667 28.0 9,655 30.9 2,892 — is not available. Source: Authors’ analysis based on 2002 and 2005 Enterprise Surveys; see text for details. decides to adopt if the corresponding expected net beneï¬?ts (beneï¬?ts minus costs) are larger than those of the alternatives, including that of not adopting new technology. Let pà ijc be the net beneï¬?ts for ï¬?rm i in sector j and country c and Adoption be a dummy variable that equals 1 if ï¬?rm i adopts ISO Correa, Fernandes, and Uregian 127 T A B L E 2 . Technology Adoption across Sectors for Sample Firms in Europe and Central Asia (percent) ISO certiï¬?cation Web use Sector 2002 sample 2005 sample 2002 sample 2005 sample Region average 13.6 12.5 58.2 67.4 Mining and quarrying, energy 36.7** 14.3 70.0 75.0 related Mining and quarrying, not energy 21.3 19.4 62.5 70.1 related Manufacturing Food beverages and tobacco 23.6*** 16.2*** 54.1* 57.0*** Textiles 17.4 7.9*** 58.4 63.0** Leather 9.7 8.3 58.1 54.2* Wood 23.8* 7.1* 50.0 60.0 Pulp and paper 11.2 13.8 81.0*** 84.1*** Petroleum 8.3 16.7 91.7*** 100.0*** Chemicals 26.6** 36.6*** 78.8*** 88.1*** Rubber and plastics 24.5* 23.3** 62.3*** 88.9*** Nonmetallic minerals 17.2 20.3** 54.0 64.2 Metals 24.4*** 19.0*** 65.2* 77.2*** Machinery and equipment 33.8*** 24.9*** 79.9*** 82.1*** Electrical and optical equipment 34.9*** 30.4*** 81.1*** 88.7*** Transport equipment 42.3*** 42.6*** 86.8*** 87.2*** Other manufacturing 20.3* 11.6 64.9 70.7 Services Construction 17.4*** 15.9*** 56.7 69.9* Wholesale and retail trade 8.7*** 8.0*** 50.0*** 62.3*** Hotels and restaurants 6.2*** 7.1*** 42.2*** 53.0*** Transport, storage, and 10.0*** 11.0 71.8*** 78.5*** communications Real estate and business 10.9** 9.2*** 71.4*** 79.5*** activities Other services 5.1*** 4.5*** 50.1*** 53.7*** *Signiï¬?cant at the 10 percent level; **Signiï¬?cant at the 5 percent level; ***Signiï¬?cant at the 1 percent level. Source: Authors’ analysis based on 2002 and 2005 Enterprise Surveys; see text for details. certiï¬?cation or web use. Then:  1 if pà ijc . 0 ð1Þ Adoptionijc ¼ 0 Otherwise The unobserved pà ijc is allowed to depend linearly on two sets of investment climate factors—access to complementary inputs, Inpijc , and market incentives Incijc —as well as ï¬?rm controls, Xijc , sector ï¬?xed effects, Ij , and GDP per capita, GDPpcc : ð2Þ pà ijc ¼ aInpijc þ bIncijc þ gXijc þ Ij þ GDPpcc þ 1ijc ; 128 THE WORLD BANK ECONOMIC REVIEW where 1ijc represents unobserved ï¬?rm characteristics influencing the adoption decision. The probability of adopting new technology for ï¬?rm i is given by: ð3Þ Prðadoptionijc ¼ 1Þ ¼ Prð1ijc . ÀaInpijc À bIncijc À gXijc À Ij À GDPpcc : . Complementary Inputs Two sets of investment climate factors can influence the probability of adopt- ing new technology—complementary inputs and market incentives (discussed below). A ï¬?rm’s access to complementary inputs can affect both the adjustment costs following the adoption of new technology and the beneï¬?ts derived from adoption. Both theoretical models and empirical evidence show how poor labor skills can delay a ï¬?rm’s technology adoption, whether because of an inability to operate advanced equipment—generally skill-biased—or because learning is technology-speciï¬?c and retraining workers is costly (Navaretti, Soloanga, and Takacs 2001; Berdugo, Sadik, and Sussman 2003; Alesina and Zeira 2006). The lack of highly qualiï¬?ed managerial capacity can constrain a ï¬?rm’s adoption of new technology by reducing its acquisition of information on available technological solutions for its needs and could increase its adjust- ment costs. Inadequate R&D investments by a ï¬?rm also hamper technology adoption possibilities since such investments are often made to develop its capabilities to assimilate and exploit external knowledge (Cohen and Levinthal 1989). Although capital goods are considered sound collateral in developed countries, failure to regulate movable collateral in countries in Europe and Central Asia might constrain ï¬?rms’ access to credit, restricting technology adoption. Depending on the gap between existing and new technologies and associated adjustment costs, complementary physical investments might be needed, making access to credit a decisive input for technological update. Access to credit is even more crucial because new technologies such as infor- mation and communication technology are frequently embodied in capital goods. Moreover, productivity gains from information and communication technologies seem to depend on complementary investments and organizational changes (Bresnahan, Brynjolfsson, and Hitt 2002). The availability and quality of physical infrastructure—especially electricity and telecommunications services—can be decisive in a ï¬?rm’s decision to adopt new technology, particularly information and communication technologies. Arnold, Mattoo, and Narcisco (2008) show that difï¬?culty obtaining adequate infrastructure services can constrain ï¬?rm performance. The empirical speciï¬?cations include all those measures as proxies for access to complementary inputs. Correa, Fernandes, and Uregian 129 Market Incentives Strong product market competition from domestic and foreign rivals and demand from consumers likely encourage incumbent ï¬?rms to invest in new and more productive technologies—product upgrades and cost reductions—rather than to spend on rent-seeking activities (Baumol 1990; Aghion and Schankerman 2004).11 Moreover, the entry of new competitors may foster innovation in incumbent ï¬?rms as they try to beat their competitors and survive (Aghion and others 2001). However, according to Schumpeter (1942), ï¬?rms facing a lower entry threat are best placed to innovate since innovation requires the expectation of tempor- ary monopoly rents. Strong product market competition can be detrimental to ï¬?rms by reducing the rents that successful innovators can appropriate after introducing an innovation to the market. Aghion and others (2009) show that stronger competition has a negative effect on incumbent ï¬?rms’ innovation incentives (though only in industries far behind the world technology frontier). These arguments can be applied to new technology adoption by ï¬?rms in devel- oping countries: investments in technologies or processes that are widespread in developed countries but that are new locally can appear too risky if ï¬?rms do not believe they will subsequently enjoy sufï¬?ciently high rents. Private ownership and control establish proï¬?t maximization as the ï¬?rm’s objective and should in principle foster adopting new technologies with ensuing productivity gains (Brown, Earle, and Telegdy 2007). Foreign owner- ship in particular, whether full or partial, exposes ï¬?rms to global best practice technology and management techniques (Djankov and Hoekman 2000). The empirical speciï¬?cations include these measures as proxies for market incentives. Methodology Equation (3) is estimated by maximum likelihood ( probit), assuming that the residual, 1ijc, is normally distributed. The probit speciï¬?cations control for sector ï¬?xed effects since differences in technology, product demand, and com- petition across sectors can influence ï¬?rms’ incentives to adopt new technology (Cohen and Levin 1989). GDP per capita is included in the speciï¬?cations to account for country heterogeneity in technology adoption not captured by differences in complementary inputs or market incentives.12 Moreover, stan- dard errors are adjusted for clustering at the country level to account for 11. Accordingly, Comin and Hobijn (2004) argue that trade openness is important to a country’s speed of adoption of advanced technologies because openness introduces pressures from foreign competition, reducing incumbents’ payoff from lobbying the government to deter the adoption of advanced technologies. 12. Country ï¬?xed effects cannot be used because of their collinearity with the location-speciï¬?c infrastructure index. 130 THE WORLD BANK ECONOMIC REVIEW possible correlations in technology adoption decisions across ï¬?rms within a country. The main results use probit estimation for ISO-certiï¬?cation and web-use regressions based on a cross section of ï¬?rms in Europe and Central Asia in 2002 or 2005. The results identify systematic correlations between complemen- tary inputs or market incentives and ISO certiï¬?cation or web use, but the esti- mated effects cannot be interpreted as causal. The effects of some investment climate factors may reflect omitted manage- rial ability or other unobservable ï¬?rm characteristics or suffer from reverse causality.13 The strategy to address this problem is threefold. First, the speciï¬?- cations include ï¬?rm controls—share of total output exported and share of total intermediate inputs imported—to minimize the risk that the estimates suffer from an omitted variables bias. The ï¬?rm controls considered capture the ï¬?rm’s access to international knowledge through its involvement in international trade—shown in previous research to be a determinant of technology transfer and adoption (Keller 2004; Almeida and Fernandes 2008). However, the ï¬?nd- ings based on the cross sections of ï¬?rms in 2002 and 2005 could still be driven partly by unobservables. Second, to compute the infrastructure index, ï¬?rm responses to the questions on electricity and telecommunications (detailed in the appendix) are averaged at the location level to correct for potential endogeneity of ï¬?rm perceptions about the quality of infrastructure with respect to technology adoption decisions. Third, the speciï¬?cation is estimated based on the panel sample using probit with random effects estimation. This approach shows how changes in access to complementary inputs and market incentives between 2002 and 2005 relate to ISO certiï¬?cation and web use by ï¬?rms, controlling for unobserved ï¬?rm hetero- geneity. A disadvantage of this approach is that probit with random effects esti- mation requires assuming that the unobserved ï¬?rm effects are uncorrelated with the regressors. While this assumption is veriï¬?ed when the unobserved ï¬?rm effects capture unexpected production breakdowns suffered by ï¬?rms during 2002–05, it might not be veriï¬?ed if those effects capture managerial risk aver- sion. Given the short panel dimension of the data—only two years of data per ï¬?rm—and the much smaller size of the panel relative to the cross-sectional samples, other estimation methods could not be used for the panel regressions.14 Thus, the magnitude of the panel results should be viewed with 13. For example, ï¬?rms with more able managers are more likely to adopt new technology but are also more likely to hire more skilled workers. Thus, the effect of skilled labor in the technology adoption regressions may to some extent reflect omitted managerial ability. 14. Conditional ï¬?xed effects logit estimation is an alternative method for panel regressions with a binary-dependent variable (Maddala 1987), but the estimation relies on ï¬?rms’ changing status in the dependent variable, which would imply the use of a fraction of an already small sample, possibly resulting in biased estimates. Correa, Fernandes, and Uregian 131 caution and taken only as an indication of the patterns of correlation once ï¬?rm heterogeneity is imperfectly accounted for. A ï¬?nal concern is that a ï¬?rm’s access to complementary inputs and market incentives (such as foreign ownership) could influence the ISO’s decision to award certiï¬?cation, despite the fact that none of the variables is explicitly among the certiï¬?cation eligibility criteria.15 Therefore, a positive coefï¬?cient on skilled labor, for example, would capture not only the positive effect of that input in the ï¬?rm’s decision to adopt ISO certiï¬?cation but also the influence such input may have played in the ISO’s decision to award certiï¬?cation. II I. R ES ULT S This section presents the results of estimating several variants of equation (3). Complementary Inputs Firms employing a larger share of skilled labor ( professionals) are signiï¬?cantly more likely to be ISO certiï¬?ed and to use the web (table 3). All else equal, an increase in a ï¬?rm’s skilled labor share by one standard deviation (22.1 percent) would be associated with a 1.5 percent increase in the frequency of ISO certiï¬?- cation (regression 1) and an 8.3 percent increase in the frequency of web use (regression 4). The results for the panel sample conï¬?rm the importance of skills by showing that ï¬?rms that increased their skilled labor share between 2002 and 2005 are signiï¬?cantly more likely to become ISO certiï¬?ed (regression 3) or to use the web (regression 6). These results are consistent with evidence in Guler, Guillen, and Macpherson (2002) on the importance of professionals with a technical background for ISO certiï¬?cation, because such professionals can more easily deal with the technical aspects of quality standards. Managerial education is also strongly positively associated with ISO certiï¬?- cation and web use. Firms run by managers with a college or postgraduate degree in 2002 are almost 4 percent more likely to be ISO certiï¬?ed and 23 percent more likely to use the web, all else held constant. While ISO-certiï¬?ed ï¬?rms are more likely to be run by older managers, web use is more frequent in ï¬?rms run by younger managers.16 Firms with higher R&D intensity are also signiï¬?cantly more likely to be ISO certiï¬?ed and to use the web. A one standard deviation higher R&D intensity (5.6 percent) is associated with a 2.4 percent increase in the frequency of web use in 2002, all else held constant. The effects of R&D intensity on ISO certiï¬?cation are positive and weaker for the panel sample but strong for web use. These ï¬?ndings show the importance of complementary investments in skills and R&D for technology adoption in ï¬?rms in Europe and Central Asia and 15. The authors thank an anonymous referee for highlighting this possibility. 16. Managerial characteristics are not available in the 2005 survey, and so they are also excluded from the panel regressions. T A B L E 3 . Correlates of Technology Adoption for ISO-Certiï¬?cation and Web-Use-Dependent Variables 132 ISO-certiï¬?cation dummy variable Web-use dummy variable 2002 sample 2005 sample Panel sample 2002 sample 2005 sample Panel sample Variable (1) (2) (3) (4) (5) (6) Complementary inputs Share of skilled 0.065 (0.009)*** 0.061 (0.000)*** 0.139 (0.000)*** 0.374 (0.000)*** 0.319 (0.000)*** 0.109 (0.002)*** labor Manager with 0.038 (0.000)*** 0.233 (0.000)*** college education dummy variable Manager age 0.001 (0.006)*** 2 0.003 (0.080)* R&D intensity 0.168 (0.013)** 0.547 0.000)*** 0.211 (0.214) 0.436 (0.080)* 1.430 (0.019)** 0.185 (0.001)*** Access to ï¬?nance 0.046 (0.000)*** 0.038 (0.000)*** 0.049 (0.003)*** 0.110 (0.000)*** 0.140 (0.000)*** 0.042 (0.000)*** THE WORLD BANK ECONOMIC REVIEW dummy variable Infrastructure 2 0.011 (0.039)** -0.005 (0.127) 2 0.010 (0.125) 0.061 (0.007)*** 0.056 (0.000)*** 2 0.010 (0.000)*** indexa Market incentives Dummy variable 2 0.049 (0.000)*** 2 0.104 (0.000)*** for market share less than 5 percent Price – cost margin 0.003 (0.895) 0.030 (0.823) 2 0.003 (0.969) 0.010 (0.118) (percent) Dummy variable 2 0.002 (0.870) 0.013 (0.242) 0.023 (0.338) 2 0.003 (0.884) 0.039 (0.003)*** 0.015 (0.606) for pressure to innovate from competitors Dummy variable 0.019 (0.046)** 0.019 (0.055)* 0.022 (0.308) 0.046 (0.003)*** 0.013 (0.336) 0.015 (0.296) for pressure to innovate from consumers Dummy variable 0.067 (0.000)*** 0.082 (0.000)*** 0.091 (0.001)*** 0.170 (0.000)*** 0.195 (0.000)*** 0.073 (0.000)*** for medium-size ï¬?rms Dummy variable 0.101 (0.000)*** 0.160 (0.000)*** 0.214 (0.000)*** 0.249 (0.000)*** 0.240 (0.000)*** 0.213 (0.000)*** for large ï¬?rms Dummy variable 0.020 (0.247) 2 0.028 (0.176) 2 0.031 (0.338) 2 0.006 (0.865) 2 0.058 (0.205) 2 0.023 (0.738) for privatized ï¬?rm Dummy variable 0.008 (0.603) 2 0.042 (0.083)* 2 0.010 (0.709) 0.097 (0.003)*** 2 0.006 (0.857) 2 0.009 (0.193) for private ï¬?rm (from start-up) Dummy variable 0.068 (0.000)*** 0.036 (0.068)* 0.032 (0.376) 0.242 (0.000)*** 0.126 (0.001)*** 0.026 (0.002)*** for fully foreign-owned ï¬?rm Dummy variable 0.047 (0.026)** 0.026 (0.082)* 0.142 (0.001)*** 0.159 (0.000)*** 0.104 (0.007)*** 0.121 (0.071)* for joint venture Ownership share 2 0.002 (0.921) 2 0.032 (0.068)* 2 0.007 (0.942) 2 0.071 (0.065)* 2 0.130 (0.000)*** 2 0.002 (0.000)*** of largest shareholder Controls Export share 0.078 (0.002)*** 0.061 (0.001)*** 0.095 0.075)* 0.256 (0.000)*** 0.332 (0.000)*** 0.318 (0.001)*** Imported inputs 0.024 (0.087)* 0.044 (0.000)*** 0.028 (0.221) 0.282 (0.000)*** 0.273 (0.000)*** 0.269 (0.000)*** share GDP per capita 0.028 (0.004)*** 0.027 (0.000)*** 0.017 (0.264) 0.160 (0.000)*** 0.140 (0.000)*** 0.126 (0.000)*** Number of 5,589 7,968 1,906 5,625 7,965 1,916 observations *Signiï¬?cant at the 10 percent level; **Signiï¬?cant at the 5 percent level; ***Signiï¬?cant at the 1 percent level. Note: Marginal effects at mean values from probit regressions are shown. Numbers in parentheses are p-values corresponding to robust standard errors clustered by country. All regressions include sector ï¬?xed effects. See appendix table A-1 for variable deï¬?nitions. Correa, Fernandes, and Uregian a. Higher values indicate better infrastructure. Source: Authors’ analysis based on 2002 and 2005 Enterprise Surveys; see text for details. 133 134 THE WORLD BANK ECONOMIC REVIEW support the role of R&D in helping ï¬?rms develop absorptive capacities for external knowledge (Cohen and Levinthal 1989). R&D activities are also likely to have spillovers into managerial activities, as ï¬?rms learn about their techno- logical bottlenecks and possible solutions. Access to ï¬?nance is strongly associated with ISO certiï¬?cation and web use. In 2005 a ï¬?rm with a bank loan is 4.2 percent more likely to be ISO certiï¬?ed and 15.8 percent more likely to use the web, all else held constant (see table 3). Firms gaining access to ï¬?nance between 2002 and 2005 are signiï¬?- cantly more likely to be ISO certiï¬?ed or to use the web. Access to ï¬?nance can be crucial for decisions on adopting new technologies and on making the complementary investments needed to absorb and efï¬?ciently use the technol- ogies. For web use, the importance of ï¬?nance relates to the way technology is embodied in new capital goods. For ISO certiï¬?cation, the ï¬?ndings likely reflect the substantial costs related to the search for information on ISO standards, technical assistance for process improvement and the adoption of standards, and the fees to apply for the certiï¬?cation (Guler, Guillen, and Macpherson 2002). The ï¬?ndings provide evidence of an important microchannel through which ï¬?nance may affect growth in ï¬?rms in Europe and Central Asia, by increasing technology adoption. Better infrastructure (indicated by a higher value for the infrastructure index) is signiï¬?cantly positively associated with web use (see table 3). Since the index captures the quality of the telecommunications network in the ï¬?rm’s location, this result reflects its importance for the efï¬?cient use of general- purpose information and communication technology by the ï¬?rm. However, better infrastructure is also negatively associated with ISO certiï¬?cation—signiï¬?- cantly so in 2002. The panel sample results show that in locations where infra- structure improved between 2002 and 2005, the frequency of web use increased but that of ISO certiï¬?cation decreased. One possible explanation for this counterintuitive sign in the ISO-certiï¬?cation regressions is that the effect of infrastructure on ISO certiï¬?ca- tion operates through other variables also included in the regressions. Another is that the infrastructure index does not account for the costs of remoteness and the risk of losses in transit (which could be proxied by the quality of the road infrastructure), which would be the most important infrastructure-related potential determinants of ISO certiï¬?cation. Nevertheless, the infrastructure index does exhibit sensible values, substantially higher for EU-8 and Southeastern European countries plus Turkey (with the exception of Albania) than for countries in the Commonwealth of Independent States, whose infra- structure is worse. The ï¬?ndings on the effects of access to complementary inputs on technology adoption are not driven by any speciï¬?c subgroup of countries (table 4). The importance of skills, R&D, and ï¬?nance for ISO certiï¬?cation and web use is conï¬?rmed in all country subgroups. Better infrastructure has a positive effect on web use and a (weak) negative effect on ISO certiï¬?cation in all country T A B L E 4 . Correlates of Technology Adoption across Country Subgroups for ISO-Certiï¬?cation and Web-Use-Dependent Variables EU-8 countries Commonwealth of Independent States countries Southeastern European countries and Turkey ISO-certiï¬?cation dummy ISO-certiï¬?cation dummy ISO-certiï¬?cation dummy variable Web-use dummy variable variable Web-use dummy variable variable Web-use dummy variable 2002 sample 2005 sample 2002 sample 2005 sample 2002 sample 2005 sample 2002 sample 2005 sample 2002 sample 2005 sample 2002 sample 2005 sample Variable (1) (2) (3) (4) (5) (6) (7) (8) (9) (10) (11) (12) Complementary inputs Share of skilled 0.107 0.082 0.232 0.147 2 0.025 0.045 0.470 0.345 0.123 0.078 0.230 0.390 labor (0.007)*** (0.008)*** (0.000)*** (0.002)*** (0.294) (0.022)** (0.000)*** (0.000)*** (0.000)*** (0.001)*** (0.004)*** (0.000)*** Manager with 0.033 0.141 0.015 0.203 0.025 0.272 college (0.000)*** (0.000)*** (0.171) (0.000)*** (0.273) (0.000)*** education dummy variable Manager age 0.003 2 0.001 0 2 0.006 0.001 0.001 (0.000)*** (0.457) (0.542) (0.000)*** (0.105) (0.662) R&D intensity 0.145 0.635 0.271 3.701 0.102 0.495 0.737 1.226 0.297 0.530 0.031 1.336 (0.303) (0.103) (0.000)*** (0.000)*** (0.166) (0.029)** (0.106) (0.148) (0.049)** (0.023)** (0.933) (0.081)* Access to ï¬?nance 0.057 0.047 0.098 0.053 0.037 0.016 0.069 0.161 0.038 0.059 0.111 0.146 dummy (0.000)*** (0.001)*** (0.000)*** (0.000)*** (0.011)** (0.154) (0.028)** (0.000)*** (0.000)*** (0.000)*** (0.002)*** (0.000)*** variable Infrastructure 2 0.027 2 0.031 2 0.058 0.022 2 0.013 2 0.003 0.080 0.093 2 0.002 2 0.001 0.105 0.049 indexa (0.579) (0.006)*** (0.302) (0.004)*** (0.003)*** (0.659) (0.002)*** (0.000)*** (0.788) (0.829) (0.035)** (0.000)*** Market incentives Dummy variable 2 0.082 2 0.077 2 0.029 2 0.076 2 0.049 2 0.080 for market (0.001)*** (0.000)*** (0.055)* (0.064)* (0.000)*** (0.077)* share less than 5 percent Price– cost margin 2 0.075 2 0.053 0.045 0.192 0.003 2 0.062 (percent) (0.214) (0.048)** (0.328) (0.197) (0.940) (0.360) Dummy variable 2 0.017 0.064 0.003 0.022 0.005 2 0.010 0.003 0.046 0.013 0.017 2 0.024 0.019 for pressure to (0.210) (0.002)*** (0.863) (0.004)*** (0.640) (0.379) (0.878) (0.012)** (0.535) (0.358) (0.561) (0.456) innovate from competitors Correa, Fernandes, and Uregian Dummy variable 2 0.004 2 0.002 0.024 0.005 0.024 0.021 0.040 0.007 0.023 0.029 0.042 0.030 for pressure to (0.800) (0.916) (0.172) (0.687) (0.087)* (0.155) (0.062)* (0.712) (0.076)* (0.032)** (0.333) (0.119) innovate from consumers 135 (Continued ) TABLE 4. Continued 136 EU-8 countries Commonwealth of Independent States countries Southeastern European countries and Turkey ISO-certiï¬?cation dummy ISO-certiï¬?cation dummy ISO-certiï¬?cation dummy variable Web-use dummy variable variable Web-use dummy variable variable Web-use dummy variable 2002 sample 2005 sample 2002 sample 2005 sample 2002 sample 2005 sample 2002 sample 2005 sample 2002 sample 2005 sample 2002 sample 2005 sample Variable (1) (2) (3) (4) (5) (6) (7) (8) (9) (10) (11) (12) Dummy variable 0.125 0.111 0.105 0.087 0.051 0.042 0.213 0.234 0.042 0.112 0.127 0.213 for (0.000)*** (0.003)*** (0.000)*** (0.000)*** (0.017)** (0.000)*** (0.000)*** (0.000)*** (0.000)*** (0.000)*** (0.000)*** (0.000)*** medium-size ï¬?rms Dummy variable 0.232 0.225 0.152 0.079 0.043 0.083 0.310 0.341 0.047 0.246 0.227 0.248 for large ï¬?rms (0.000)*** (0.000)*** (0.000)*** (0.000)*** (0.136) (0.000)*** (0.000)*** (0.000)*** (0.069)* (0.000)*** (0.000)*** (0.000)*** Dummy variable 0.046 2 0.060 2 0.040 2 0.008 2 0.017 2 0.013 0.007 2 0.087 0.037 2 0.011 0.016 0.061 for privatized (0.094)* (0.027)** (0.206) (0.794) (0.346) (0.748) (0.881) (0.268) (0.450) (0.746) (0.844) (0.293) ï¬?rm Dummy variable 0.008 2 0.125 2 0.004 2 0.035 0.011 2 0.006 0.157 0.026 2 0.011 2 0.027 0.026 0.014 for private ï¬?rm (0.727) (0.000)*** (0.848) (0.001)*** (0.666) (0.874) (0.000)*** (0.686) (0.721) (0.562) (0.497) (0.811) (from start-up) Dummy variable 0.046 0.038 0.064 0.062 0.123 0.036 0.406 0.152 0.054 0.041 0.229 0.089 THE WORLD BANK ECONOMIC REVIEW for fully (0.074)* (0.351) (0.135) (0.063)* (0.000)*** (0.182) (0.000)*** (0.029)** (0.161) (0.395) (0.003)*** (0.299) foreign-owned ï¬?rms Dummy variable 0.019 0.047 0.077 0.065 0.048 0.019 0.205 0.076 0.046 0.017 0.103 0.133 for joint (0.589) (0.100)* (0.078)* (0.037)** (0.116) (0.355) (0.000)*** (0.258) (0.147) (0.607) (0.039)** (0.079)* venture Ownership share of 0.031 2 0.066 2 0.069 2 0.104 2 0.046 2 0.013 2 0.106 2 0.120 0.017 2 0.049 2 0.017 2 0.147 largest (0.355) (0.041)** (0.040)** (0.000)*** (0.133) (0.536) (0.140) (0.012)** (0.671) (0.082)* (0.811) (0.000)*** shareholder Controls Export share 0.054 0.063 0.145 0.215 0.063 0.050 0.289 0.447 0.106 0.064 0.248 0.215 (0.001)*** (0.001)*** (0.002)*** (0.000)*** (0.162) (0.027)** (0.000)*** (0.000)*** (0.040)** (0.373) (0.006)*** (0.001)*** Imported inputs 0.009 0.036 0.149 0.083 0.038 0.043 0.311 0.341 0.024 0.058 0.217 0.270 share (0.660) (0.016)** (0.000)*** (0.000)*** (0.028)** (0.010)*** (0.000)*** (0.000)*** (0.467) (0.000)*** (0.000)*** (0.000)*** GDP per capita 0.082 0.030 0.168 0.080 0.046 0.010 0.139 0.153 0.019 0.014 2 0.092 0.022 (0.000)*** (0.136) (0.000)*** (0.066)* (0.000)*** (0.130) (0.001)*** (0.000)*** (0.317) (0.324) (0.409) (0.710) Number of 1,702 2,490 1,721 2,465 2,391 3,428 2,404 3,441 1,486 2,023 1,497 2,008 observations *Signiï¬?cant at the 10 percent level; **Signiï¬?cant at the 5 percent level; ***Signiï¬?cant at the 1 percent level. Note: Marginal effects at mean values from probit regressions are shown. Numbers in parentheses are p-values corresponding to robust standard errors clustered by country. All regressions include sector ï¬?xed effects. See appendix table A1 for variable deï¬?nitions. a. Higher values indicate better infrastructure. Source: Authors’ analysis based on 2002 and 2005 Enterprise Surveys; see text for details. Correa, Fernandes, and Uregian 137 subgroups. Finally, estimates of complementary inputs do not seem to suffer from multicollinearity: the coefï¬?cients for the 2002, 2005, and panel samples are close to those estimated using equation (3) but excluding the market incen- tive proxies and ï¬?rm controls (see table 3). Market Incentives Firms with smaller market shares in 2002 are signiï¬?cantly less likely to be ISO certiï¬?ed or to use the web (see table 3). Firms with lower price–cost margins in 2005 are also less likely to be ISO certiï¬?ed or to use the web, though the effects are weak.17 The panel sample results indicate that ï¬?rms reducing their price–cost margins between 2002 and 2005 are less likely to become ISO cer- tiï¬?ed or to use the web—signiï¬?cantly so for web use (see table 3, regressions 3 and 6). The pressure to innovate from competitors is only weakly positively corre- lated with technology adoption, with the exception of a signiï¬?cant correlation with web use in 2005. However, stronger pressure to innovate from consumers signiï¬?cantly increases the frequency of becoming ISO certiï¬?ed and using the web. The panel sample results show that ï¬?rms facing increased pressure from competitors or increased pressure from consumers to innovate between 2002 and 2005 are more likely to become ISO certiï¬?ed and to use the web. Because of the difï¬?culty in measuring market competition, the robustness of the results are checked by replacing the dummy variables for the pressure to innovate from consumers and competitors with two competition measures pro- posed by Carlin, Schaffer, and Seabright (2004): elasticity of demand and number of domestic competitors in a ï¬?rm’s main product. Firms facing more elastic demand (whose customers would react to a price increase by buying the product from competitors) are less likely to be ISO certiï¬?ed or to use the web (table 5, regressions 1, 2, 5, and 6). There is no difference in ISO certiï¬?cation or web use between ï¬?rms facing no domestic competitors in their main pro- ducts and ï¬?rms facing one to three competitors or more than four competitors (see table 5, regressions 3, 4, 7, and 8). Larger ï¬?rms are signiï¬?cantly more likely to adopt new technology. Within a sector in 2002, large ï¬?rms (more than 250 workers) were about 25 percent more likely to use the web than were small ï¬?rms (fewer than 50 workers; see table 3, regression 4). It could be that the economies of scale from which large ï¬?rms beneï¬?t are associated with higher productivity and a higher return to technology adoption, allowing large ï¬?rms to operate with a more efï¬?cient div- ision of labor and creating better conditions for technological upgrade. 17. The 2002 survey also provides information on price–cost margins. The regressions including that measure instead of market share show that ï¬?rms with lower price–cost margins in 2002 are less likely to be ISO certiï¬?ed or to use the web. Since the 2005 survey provides information only on price– cost margins, that variable is used in the panel regressions. T A B L E 5 . Correlates of Technology Adoption—Alternative Competition Measures for ISO-Certiï¬?cation and 138 Web-Use-Dependent Variables ISO-certiï¬?cation dummy variable Web-use dummy variable 2002 sample 2005 sample 2002 sample 2005 sample 2002 sample 2005 sample 2002 sample 2005 sample Variables (1) (2) (3) (4) (5) (6) (7) (8) Complementary inputs Share of skilled labor 0.068 0.063 0.064 0.074 0.372 0.332 0.371 0.181 (0.004)*** (0.000)*** (0.008)*** (0.000)*** (0.000)*** (0.000)*** (0.000)*** (0.000)*** Manager with college 0.037 0.037 0.227 0.228 education dummy (0.000)*** (0.000)*** (0.000)*** (0.000)*** variable Manager age 0.001 0.001 2 0.003 2 0.003 (0.003)*** (0.006)*** (0.098)* (0.092)* R&D intensity 0.159 0.534 0.161 1.007 0.426 1.391 0.428 2.596 THE WORLD BANK ECONOMIC REVIEW (0.021)** (0.000)*** (0.017)** (0.000)*** (0.085)* (0.020)** (0.083)* (0.000)*** Access to ï¬?nance 0.048 0.040 0.048 0.039 0.114 0.145 0.113 0.091 dummy variable (0.000)*** (0.000)*** (0.000)*** (0.002)*** (0.000)*** (0.000)*** (0.000)*** (0.000)*** Infrastructure indexa 2 0.010 2 0.006 2 0.009 2 0.009 0.062 0.057 0.062 0.026 (0.066)* (0.146) (0.072)* (0.127) (0.007)*** (0.000)*** (0.007)*** (0.000)*** Market incentives Price – cost margin 0.011 2 0.028 2 0.027 2 0.054 (percent) (0.701) (0.397) (0.668) (0.146) Dummy variable for 2 0.008 2 0.013 0.013 2 0.002 elasticity of demand: (0.479) (0.193) (0.410) (0.914) if prices increased by 10 percent customers would still buy from ï¬?rm but slightly lower quantities Dummy variable for 2 0.013 2 0.009 2 0.005 2 0.073 elasticity of demand: (0.286) (0.370) (0.826) (0.001)*** if prices increased by 10 percent customers would still buy from ï¬?rm but much lower quantities Dummy variable for 2 0.029 2 0.012 2 0.026 2 0.092 elasticity of demand: (0.029)** (0.216) (0.248) (0.000)*** if prices increased by 10 percent customers would buy from competitors Dummy variable for 0.028 0.001 0.124 2 0.007 ï¬?rm facing 1 – 3 (0.489) (0.978) (0.091)* (0.868) competitors in domestic market Dummy variable for 2 0.011 2 0.005 0.109 2 0.005 ï¬?rm facing four or (0.770) (0.911) (0.174) (0.890) more competitors in domestic market Dummy for 0.065 0.081 0.065 0.111 0.173 0.196 0.174 0.089 medium-size ï¬?rms (0.000)*** (0.000)*** (0.000)*** (0.000)*** (0.000)*** (0.000)*** (0.000)*** (0.000)*** Dummy variable for 0.098 0.153 0.096 0.194 0.249 0.230 0.250 0.101 large ï¬?rms (0.000)*** (0.000)*** (0.000)*** (0.000)*** (0.000)*** (0.000)*** (0.000)*** (0.000)*** Dummy variable for 0.027 2 0.029 0.027 2 0.036 2 0.001 2 0.041 2 0.007 2 0.018 privatized ï¬?rm (0.121) (0.156) (0.122) (0.286) (0.984) (0.343) (0.838) (0.612) Dummy variable for 0.014 2 0.043 0.014 2 0.054 0.100 0.006 0.096 0.016 private ï¬?rm (0.396) (0.072)* (0.347) (0.124) (0.002)*** (0.860) (0.005)*** (0.666) (from start-up) Dummy variable for 0.071 0.038 0.066 0.047 0.23 0.134 0.236 0.076 Correa, Fernandes, and Uregian fully foreign owned (0.000)*** (0.056)* (0.000)*** (0.088)* (0.000)*** (0.000)*** (0.000)*** (0.003)*** (Continued ) 139 140 TABLE 5. Continued ISO-certiï¬?cation dummy variable Web-use dummy variable 2002 sample 2005 sample 2002 sample 2005 sample 2002 sample 2005 sample 2002 sample 2005 sample Variables (1) (2) (3) (4) (5) (6) (7) (8) Dummy variable for 0.048 0.026 0.047 0.019 0.158 0.101 0.157 0.036 joint venture (0.017)** (0.084)* (0.025)** (0.444) (0.000)*** (0.008)*** (0.000)*** (0.120) Ownership share of 2 0.005 2 0.035 2 0.005 2 0.061 2 0.078 2 0.133 2 0.075 2 0.063 largest shareholder (0.777) (0.033)** (0.793) (0.012)** (0.035)** (0.000)*** (0.045)** (0.015)** Dummy variable for 2 0.044 2 0.041 2 0.101 2 0.104 market share less than (0.001)*** (0.002)*** (0.000)*** (0.000)*** 5 percent Controls Export share 0.072 0.060 0.072 0.071 0.264 0.360 0.263 0.160 (0.005)*** (0.002)*** (0.004)*** (0.024)** (0.000)*** (0.000)*** (0.000)*** (0.000)*** THE WORLD BANK ECONOMIC REVIEW Imported inputs share 0.025 0.045 0.024 0.063 0.283 0.277 0.284 0.108 (0.082)* (0.000)*** (0.096)* (0.000)*** (0.000)*** (0.000)*** (0.000)*** (0.000)*** GDP per capita 0.028 0.030 0.028 0.045 0.160 0.147 0.162 0.076 (0.002)*** (0.000)*** (0.003)*** (0.000)*** (0.000)*** (0.000)*** (0.000)*** (0.000)*** Number of observations 5,650 8,076 5,666 4,146 5,687 8,073 5,705 4,143 *Signiï¬?cant at the 10 percent level; **Signiï¬?cant at the 5 percent level; ***Signiï¬?cant at the 1 percent level. Note: Marginal effects at mean values from probit regressions are shown. Numbers in parentheses are p-values corresponding to robust standard errors clustered by country. All regressions include sector ï¬?xed effects. See appendix table A1 for variable deï¬?nitions. a. Higher values indicate better infrastructure. Source: Authors’ analysis based on 2002 and 2005 Enterprise Surveys; see text for details. Correa, Fernandes, and Uregian 141 Taken together, the ï¬?ndings on market shares, price–cost margins, competition, and size suggest that for ï¬?rms in Europe and Central Asia, con- centration is more conducive to technology adoption than is competition. This is consistent with the argument by Carlin, Schaffer, and Seabright (2004) that ï¬?rms in transition economies face resource constraints that make rents impor- tant in ï¬?nancing technology adoption. The ï¬?ndings here also support Schumpeter’s (1942) argument that innovation is costly and that ï¬?rms facing ï¬?nancial market imperfections often ï¬?nance technology adoption from retained earnings. These results are found despite the importance of access to external ï¬?nance for technology adoption identiï¬?ed earlier. Finally, that market incen- tives for technology adoption originate more from consumers than from com- petitors is likely to reflect the pressure from large ï¬?rms ( possibly multinationals) on smaller suppliers to upgrade, accompanied by knowledge transfer and training. For ï¬?rms in Europe and Central Asia, private ownership generally has no signiï¬?cant effect on technology adoption, except for web use in 2002 (see table 3). Moreover, privatization does not seem to be a vehicle for technology adoption: privatized ï¬?rms are no more likely than state-owned ï¬?rms to be ISO certiï¬?ed or to use the web. One possible interpretation for this ï¬?nding is that private and privatized ï¬?rms have an advantage in technology adoption because of wider access to complementary inputs, but once those inputs are controlled for, the private and privatized dummy variables have no additional effect. And state-owned ï¬?rms—especially large ones with access to ï¬?nance—may initially lag behind private ï¬?rms technologically, but they eventually catch up. To probe the privatization result further, the privatized dummy variable is decomposed into a dummy variable for ï¬?rms privatized less than three years ago and one for ï¬?rms privatized more than three years ago. Firms privatized less than three years ago are signiï¬?cantly less likely to be ISO certiï¬?ed or to use the web, perhaps because newly privatized ï¬?rms have other investment priori- ties, such as replacing old equipment (Goldberg and others 2008). There is a strong positive association between foreign ownership—full or through a joint venture—and new technology adoption (see table 3). Foreign-owned ï¬?rms are embedded in international networks, requiring fre- quent use of communications technology, and they compete in global markets, requiring the use of state of the art technology through internationally recog- nized technical standards. However, the technology adoption advantage of foreign-owned ï¬?rms is strong only in the less advanced transition economies (see table 4). The proxy for concentrated ownership is found to be generally negatively associated with technology adoption (see table 3). This ï¬?nding is counterintui- tive to the extent that ï¬?rms with better corporate control—for which owner- ship concentration is a proxy—are expected to have more incentives to adopt new technology to maximize proï¬?ts than are ï¬?rms with more diffuse owner- ship and more limited exercise of control. It could be that the negative 142 THE WORLD BANK ECONOMIC REVIEW association between technology adoption and ï¬?rms with high ownership concentration is explained by how privatization programs operated in many Europe and Central Asia countries. While some shares were distributed to the general population, the government retained direct or indirect control over a large proportion of the shares.18 More research is needed on the role of owner- ship control for technology adoption, exploiting panel data on ï¬?rms under- going important control changes. Overall, the ï¬?ndings on ownership suggest that foreign-owned ï¬?rms have signiï¬?cantly better technology adoption outcomes but that privatization to domestic owners brings no additional beneï¬?ts. The market incentive estimates do not seem to suffer from multicollinearity since the coefï¬?cients for the 2002, 2005, and panel samples are close to those in table 3 in the variant estimate of equation (3) that excludes the complementary input proxies and ï¬?rm controls. Other Results The coefï¬?cients on export shares and imported input shares reported in tables 3–5 show that plants engaged in these international activities are signiï¬?- cantly more likely to be ISO certiï¬?ed or to use the web, mirroring the ï¬?ndings of Criscuolo, Haskel, and Slaughter (forthcoming) for innovation. Exporters and importers are likely to learn about new technologies through their inter- action with foreign buyers and suppliers. These ï¬?ndings are strong for all country groups (see table 4). The regressions reported in tables 3–5 also show a positive and signiï¬?cant effect of GDP per capita on ISO certiï¬?cation and web use. While GDP per capita accounts for an array of differences across countries, a particularly rel- evant interpretation here is that countries in Europe and Central Asia with higher GDP per capita have better protection of property rights, contract enfor- cement, and regulatory regimes, which are all potentially important correlates of technology adoption that are difï¬?cult to measure at the ï¬?rm level. I V. C O N C L U S I O N S The international diffusion of technology presents an opportunity for develop- ing countries lagging behind the world technological frontier to reduce their income gap with developed countries. But ï¬?rst there is a need to understand why, when facing similar technological alternatives, different ï¬?rms in different countries make different technology adoption choices. The ï¬?ndings presented here on correlations between investment climate factors and technology adoption for ï¬?rms in 28 Europe and Central Asia countries have implications for policy reforms aimed at enabling faster and wider adoption of new technologies by private sector ï¬?rms in Europe and Central Asia. Firms with better access to complementary inputs (skilled labor, 18. The authors thank an anonymous referee for suggesting this interpretation. Correa, Fernandes, and Uregian 143 managerial skills R&D, ï¬?nance, and to a lesser extent good infrastructure) are more likely to be ISO certiï¬?ed or to use the web. The relationship between market incentives and ISO certiï¬?cation or web use is more nuanced. While pressures from consumers generate demand for ï¬?rms to adopt new technology, pressures from competitors do not. This ï¬?nding may be counterintuitive in a developed country setting but is consistent with pre- vious studies ï¬?nding that most ï¬?rms in Europe and Central Asia face substan- tial resource constraints ( particularly ï¬?nancial resources) so that only ï¬?rms with rents are able to ï¬?nance technology adoption. Accordingly, larger ï¬?rms are signiï¬?cantly more likely to adopt new technology. The ï¬?ndings also suggest that privatization is not necessarily a vehicle for technology adoption in Europe and Central Asia but that foreign ownership is strongly associated with better technology adoption outcomes, most likely through the access it provides to advanced foreign technology and management techniques. The broad policy implication of the ï¬?ndings is that increasing technology adoption by ï¬?rms operating under severe resource constraints requires invest- ment climate reforms that increase the availability of complementary inputs to investments in technology adoption, while also improving market incentives. The study identiï¬?es several robust correlates of technology adoption by examining two novel dimensions of technology in a developing country context. However, these technology adoption proxies have important limit- ations. A fruitful next step in data collection and research efforts would be to examine the determinants of the intensity of new technology adoption by ï¬?rms. FUNDING Support from the governments of Norway, Sweden and the UK through the Multidonor Trust Fund for Trade and Development is gratefully acknowledged. APPENDIX T A B L E A - 1 Variables and their Deï¬?nitions Variable name Deï¬?nition ISO-certiï¬?cation dummy variable Dummy variable equal to 1 if the ï¬?rm obtained a new quality accreditation (ISO 9000) in the three years prior to the survey Web-use dummy variable Dummy variable equal to 1 if the ï¬?rm uses email and the internet regularly in its interactions with clients and suppliers (Continued ) 144 THE WORLD BANK ECONOMIC REVIEW TABLE A-1 Continued Variable name Deï¬?nition Size categories Small ¼ fewer than 50 permanent workers; medium ¼ 50 – 249 workers; large ¼ 250 or more workers Manager with college education or Equal to 1 if general manager’s highest level of more dummy variable education is a university degree or a postgraduate degree Manager age Age of the ï¬?rm’s general manager Share of skilled labor Percentage of the ï¬?rm’s current permanent full-time workers that are professionals (accountants, engineers, scientists) R&D intensity R&D expenditures (including wages and salaries of R&D personnel, materials, and R&D-related education and training costs) as percentage of total ï¬?rm sales Access to ï¬?nance dummy variable Equal to 1 if the ï¬?rm has a bank loan or overdraft Infrastructure index First principal component derived from factor analysis of the negative of the average number of days with power outages or surges from the public grid in the ï¬?rm’s city and the negative of the average number of days with unavailable mainline telephone service in the ï¬?rm’s city in the year before the survey Dummy variable for market share Equal to 1 if the ï¬?rm’s percentage of total market sales equal to less than 5 percent is less than 5 percent (available only in the 2002 survey) Price – cost margin (percent) Percentage margin by which sales price of the ï¬?rm’s main product or service line in the domestic market exceeds its operating costs (materials inputs costs plus wages costs but not overhead and depreciation) Dummy variable for pressure to Equal to 1 if the ï¬?rm ranks pressure from domestic or innovate from competitors foreign competitors as being fairly important or very important for the ï¬?rm’s decisions about developing new products or services and markets Dummy variable for pressure to Equal to 1 if the ï¬?rm ranks pressure from customers as innovate from consumers being fairly important or very important for the ï¬?rm’s decisions about developing new products or services and markets Elasticity of demand faced by ï¬?rm What would happen to demand if the ï¬?rm were to raise the prices of its main product or services line 10 percent above the current level in the domestic market (after allowing for inflation), assuming that competitors maintained their current prices: customers would continue to buy from the ï¬?rm in the same quantities as now, customers would continue to buy from the ï¬?rm but at slightly lower quantities, customers would continue to buy from the ï¬?rm but at much lower quantities, many customers would buy from competitors instead (Continued ) Correa, Fernandes, and Uregian 145 TABLE A-1 Continued Variable name Deï¬?nition Number of competitors Number of competitors for the ï¬?rm’s main product in the domestic market: none, one to three, or four or more Dummy variable for privatized ï¬?rm Equal to 1 if the ï¬?rm was established through privatization of a state-owned ï¬?rm Dummy variable for private ï¬?rm (from Equal to 1 if the ï¬?rm was private from start-up (no start-up) state-owned predecessor) Dummy variable for fully foreign Equal to 1 if 100 percent of the ï¬?rm’s capital is owned owned by foreigners Dummy variable for joint venture Equal to 1 if any (but less than 100 percent) of the ï¬?rm’s capital is owned by foreigners Ownership share of largest shareholder Percentage of the ï¬?rm’s equity owned by the largest shareholder Export share Percentage of ï¬?rm output exported directly or indirectly Imported input share Percentage of intermediate inputs used by the ï¬?rm that are imported directly or indirectly GDP per capita (log) Values in constant 2000 dollars for 1995 (from World Bank various years) EU-8 countries Countries that joined the European Union in 2004: Czech Republic, Estonia, Hungary, Latvia, Lithuania, Poland, Slovak Republic, and Slovenia CIS countries Members of the Commonwealth of Independent States; all former Soviet republics except Estonia, Latvia, and Lithuania Southeastern European countries and Albania, Bosnia and Herzegovina, Bulgaria, Croatia, Turkey FYR Macedonia, Romania, Serbia and Montenegro,a and Turkey a. Data are for the period before separation. Source: Enterprise Surveys in 2002 or in 2005 unless otherwise stated. REFERENCES Aghion, Philippe, and Mark Schankerman. 2004. “On the Welfare Effects and Political Economy of Competition-Enhancing Policies.â€? Economic Journal 114(498):800– 24. Aghion, Philippe, Christopher Harris, Peter Howitt, and John Vickers. 2001. “Competition, Imitation and Growth with Step-by-Step Innovation.â€? Review of Economic Studies 68 (3):467– 92. Aghion, Philippe, Richard Blundell, Rachel Grifï¬?th, Peter Howitt, and Susanne Prantl. 2009. “The Effects of Entry on Incumbent Innovation and Productivity.â€? Review of Economics and Statistics 91(1):20–32. Alesina, Alberto, and Joseph Zeira. 2006. Technology and Labor Regulations. NBER Working Paper 12581. Cambridge, Mass.: National Bureau of Economic Research. Almeida, Rita, and Ana M. Fernandes. 2008. “Openness and Technological Innovations in Developing Countries: Evidence from Firm-Level Surveys.â€? Journal of Development Studies 44(5):701–27. Arnold, Jens, Aaditya Mattoo, and Gaia Narciso. 2008. “Services Inputs and Firm Productivity in Sub-Saharan Africa: Evidence from Firm-Level Data.â€? Journal of African Economies 17(4):578–99. Arora, Arora, and Jai Asundi. 1999. Quality Certiï¬?cation and the Economics of Contract Software Development: A Study of the Indian Software Service Companies. NBER Working Paper 7260. Cambridge, Mass.: National Bureau of Economic Research. 146 THE WORLD BANK ECONOMIC REVIEW Basant, Rakesh, Simon Commander, Rupert Harrison, and Naercio Menezes-Filho. 2006. ICT Adoption and Productivity in Developing Countries: New Firm Level Evidence from Brazil and India. IZA Discussion Paper 2294. Institute for the Study of Labor, Bonn, Germany. Baumol, William. 1990. “Entrepreneurship: Productive, Unproductive, and Destructive.â€? Journal of Political Economy 98(5):893–921. Benner, Mary. 1999. “Process Management and Organizational Adaptation to Technological Change: Evidence from a Simulation Study.â€? New York: Columbia University, Graduate School of Business. Benner, Mary, and Francisco Veloso. 2008. “ISO 9000 Practices and Financial Performance: A Technology Coherence Perspective.â€? Journal of Operations Management 26(3): 611– 29. Berdugo, Byniamin, Jacques Sadik, and Nathan Sussman. 2003. “Delays in Technology Adoption, Appropriate Human Capital, Natural Resources, and Growth.â€? Jerusalem: Hebrew University, Department of Economics. Blind, Knut, Andre Jungmittag, Paul Temple, Chris Spencer, Peter Swann, and Robert Witt. 2005. The Empirical Economics of Standards. DTI Economics Paper 12. London, U.K: U.K. Department of Trade and Industry. Boucekkine, Raouf, Fernando Del Rio, and Omar Licandro. 2003. “Embodied Technical Change, Learning by Doing, and Productivity Slowdown.â€? Scandinavian Journal of Economics 105(1): 87 –97. Bresnahan, Timothy, and Manuel Trajtenberg. 1995. “General Purpose Technologies ‘Engines of Growth’?" Journal of Econometrics 65(1):83–108. Bresnahan, Timothy, Erik Brynjolfsson, and Lorin Hitt. 2002. “Information Technology, Workplace Organization, and the Demand for Skilled Labor: Firm-Level Evidence.â€? Quarterly Journal of Economics 117(1):339 –76. Brown, J. D., John Earle, and Amos Telegdy. 2006. “The Productivity Effects of Privatization: Longitudinal Estimates from Hungary, Romania, Russia, and Ukraine.â€? Journal of Political Economy 114(1):61–99. Carlin, Wendy, Mark Schaffer, and Paul Seabright. 2004. “A Minimum of Rivalry: Evidence from Transition Economies on the Importance of Competition for Innovation and Growth. “Contributions article. The B&E Journal of Economic Analysis and Policy 3(1). Caselli, Francesco, and Wilbur Coleman II. 2001. “Cross-Country Technology Diffusion: The Case of Computers.â€? American Economic Review 91(2):328–35. Cohen, Wesley, and Richard Levin. 1989. “Empirical Studies of Innovation and Market Structure.â€? In Handbook of Industrial Organization Vol 2, ed. R. Schmalensee, and R. Willig. Amsterdam: North-Holland. Cohen, Wesley, and Daniel Levinthal. 1989. “Innovation and Learning: The Two Faces of R&D.â€? Economic Journal 99(397):569 –96. Comin, Diego, and Bart Hobijn. 2004. “Cross-Country Technology Adoption: Making the Theories Face the Facts.â€? Journal of Monetary Economics 51(1):39–83. ———. 2008. An Exploration of Technology Diffusion. Harvard Business School Working Paper 08-093. Cambridge, Mass: Harvard Business School. Corbett, Charles, Maria Montes-Sancho, and David Kirsch. 2005. “The Financial Impact of ISO 9000 Certiï¬?cation in the United States: An Empirical Analysis.â€? Management Science 51(7):1046– 59. Correa, Paulo, Ana M. Fernandes, and Chris Uregian. 2008. Technology Adoption and the Investment Climate: Firm-Level Evidence for Eastern Europe and Central Asia. Policy Research Working Paper 4707. World Bank, Development Research Group, Trade Team, Washington, DC. Criscuolo, Chiara, Jonathan Haskel, and Matthew Slaughter. Forthcoming. “Global Engagement and the Innovation Activities of Firms.â€? International Industrial Organization Journal. Djankov, Simeon, and Bernard Hoekman. 2000. “Foreign Investment and Productivity Growth in Czech Enterprises.â€? World Bank Economic Review 14(1):49– 64. Correa, Fernandes, and Uregian 147 Eaton, Jonathan, and Samuel Kortum. 2001. “Trade in Capital Goods.â€? European Economic Review 45(7):1195–235. Goldberg, Itzhak, Lee Branstetter, John Goddard, and Smita Kuriakose. 2008. Globalization and Technology Absorption in Europe and Central Asia: The Role of Trade, FDI, and Cross-border Knowledge Flows. Working Paper 150. Central Asia Region, World Bank, Washington, DC. Guler, Isin, Mauro Guillen, and John Macpherson. 2002. “Global Competition, Institutions, and the Diffusion of Organizational Practices: The International Spread of ISO 9000 Quality Certiï¬?cates.â€? Administrative Science Quarterly 47(2):207 – 32. Hall, Robert, and Charles Jones. 1999. “Why Do Some Countries Produce So Much More Output per Worker than Others?â€? Quarterly Journal of Economics 114(1):83– 116. Hall, Bronwyn, and Beethika Khan. 2003. Adoption of New Technology. NBER Working Paper 9730. Cambridge, Mass.: National Bureau of Economic Research. Humphrey, John, and Hubert Schmitz. 2000. Governance and Upgrading: Linking Industrial Cluster and Global Value Chain Research. IDS Working Paper 120. Falmer, Brighton, U.K.: Institute of Development Studies, University of Sussex. Indjikian, Rouben, and Donald Siegel. 2005. “The Impact of Investment in IT on Economic Performance: Implications for Developing Countries.â€? World Development 33(5):681– 700. ISO (International Organization for Standardization). 1998. Selection and Use of ISO 9000. Geneva: International Organization for Standardization. Keller, Wolfgang. 2004. “International Technology Diffusion.â€? Journal of Economic Literature 42(3):752– 82. Lipsey, Richard, and Kenneth Carlaw. 2004. “Total Factor Productivity and the Measurement of Technological Change.â€? Canadian Journal of Economics 37(4):1118– 50. Maddala, G.S. (1987). “Limited Dependent Variable Models Using Panel Data.â€? Journal of Human Resources 22(3):307–38. Navaretti, Giorgio, Isidro Soloaga, and Wendy Takacs. 2001. “Vintage Technologies and Skill Constraints: Evidence from U.S. Exports of New and Used Machines.â€? World Bank Economic Review 14(1):91–109. Parente, Stephen, and Edward Prescott. 1994. “Barriers to Technology Adoption and Development.â€? Journal of Political Economy 102(2):298 – 321. ———. 2005. “A Uniï¬?ed Theory of the Evolution of International Income Levels.â€? In P. Aghion, and S. Durlauf eds., Handbook of Economic Growth. Vol. 1. Amsterdam: North-Holland. Prescott, Edward. 1998. “Lawrence R. Klein Lecture 1997: Needed: A Theory of Total Factor Productivity.â€? International Economic Review 39(3):525–51. Safavian, Mehnaz, and Siddarth Sharma. 2007. “When Do Creditor Rights Work?â€? Journal of Comparative Economics 35(3):484– 508. Schumpeter, Joseph. 1942. Capitalism, Socialism, and Democracy. New York: Harper. World Bank. Various years. World Development Indicators. Washington, DC: World Bank. Zeira, Joseph. 1998. “Workers, Machines, and Economic Growth.â€? Quarterly Journal of Economics 113(4):1091 –117.