Effect of enterprise ownership and foreign competition on internet diffusion in the transition economies.

Clarke, George R.G.

INTRODUCTION

After several, decades of slow economic growth and modest improvements in productivity, growth accelerated in the United States in the mid- to late 1990s. Whereas output increased by only 2.8 percent per year and output per labour hour increased by only 1.0 percent per year between 1972 and 1995, they increased by 4.9 percent and 2.7 percent per year respectively between 1995 and 1999 (Gordon, 2000, p. 53). Although there is considerable uncertainty regarding the reason for the increase in growth, many observers attributed it to growing investment in information technology in general and to the Internet in particular. (1) Although the benefits of information technology are still in dispute, these changes led to considerable discussion about whether countries that failed to make similar investments would be left behind as growth in technologically more advanced economies accelerated. This concern was especially marked for low- and middle-income countries, where Internet access and the use of information technology is far less common.

The digital divide between the rich developed world and the poor developing world is visible even when comparing the mainly middle-income economies of Eastern Europe and Central Asia with high-income OECD countries. Over 25 percent of the inhabitants of high-income OECD countries had Internet access in 1999, compared to about 6-7 percent of people in Central Europe and the Baltics, and 1-2 percent of people in South Eastern Europe and the Commonwealth of Independent States.

The importance of foreign investment as a source of technological transfers suggests that encouraging foreign investors from developed countries to invest in developing countries might reduce the disparity between rich and poor countries. (2) In addition to increasing the use of information technology among enterprises that directly receive inflows of foreign investment, several mechanisms might encourage diffusion among domestically owned enterprises in the host economy. For example, workers and managers who leave foreign-owned enterprises to join existing domestic firms might encourage their new employers to copy the techniques used by foreign-owned enterprises (including more intensive use of information technology). Alternatively, domestic enterprises, including competitors and upstream and downstream firms, might simply observe and copy the foreign-owned enterprises' business techniques. Since the benefits associated with network industries are greater when coverage is higher, enterprises that use the Internet will generally have an incentive to encourage up- and downstream firms to adopt it. Further, although foreign-owned enterprises have strong incentives to prevent domestic competitors from copying their business models, some leakage, especially of generic knowledge such as use of information technology, seems inevitable. Finally, foreign-owned enterprises' demand for Internet services might encourage the formation of support companies (eg, web-hosting or web-design companies) that can then sell their services to other companies in the host country.

Using enterprise-level data from 21 low and middle income economies in Eastern Europe and Central Asia, this paper looks at whether foreign investment increases Internet access in host countries. First, it looks at whether foreign-owned firms appear to be more likely to have Internet access than their domestic counterparts. Second, it looks at whether domestically owned enterprises competing either with foreign-owned enterprises operating in the host country or with imports also appear more likely to have access to the Internet--something that might indicate diffusion due to foreign trade or investment. Finally, the paper looks at whether foreign direct investment (FDI) appears to increase Internet access for enterprises other than the foreign-owned firms and their direct competitors in the host country. In general, there appears to be strong evidence that foreign trade and investment encourage higher levels of Internet access throughout the host economy.

Although the recent discussion on the 'digital divide' between developing and developed countries makes the question of Internet access interesting in its own right, the topic is also of interest because of its relationship with more general questions about international transfers of technology between developing and developed countries. Over the past decade, a large literature has emerged looking at how enterprises in developing countries gain access to new technologies, often focusing on the role of foreign investment and trade. In general, although foreign investment appears to result in improved productivity in the enterprises that receive the investment, there is less evidence of broad spillovers to the economy as a whole. However, since most studies have focused on the effect of foreign investment on productivity, it is possible that the negative results regarding spillovers are due to the short-term pressure that foreign entry puts on domestic enterprises through product market competition, rather than a lack of technological transfers. (3) Since this study looks at the adoption of a new technology directly, it is a useful complement to the existing literature since it avoids the possibility that pecuniary externalities will obscure technological spillovers.

EFFECT OF FOREIGN INVESTMENT ON ACCESS TO TECHNOLOGY

Although R&D expenditures are low in developing and transition economies, enterprises in these countries might gain access to new technologies in other ways, including foreign direct investment, joint ventures with foreign firms, licensing, and imports of capital goods. (4) Of these methods, foreign ownership is often seen as one of the most effective ways for enterprises in developing and transition economies to gain access to new technologies. In addition to giving access to hard technological knowledge (eg, blueprints, product designs and machinery), foreign investment might also lead to transfers of generic knowledge (eg, improved management techniques or experience using information technology), which might be harder to transmit through methods such as licensing or imports of capital goods. Foreign investment might be especially effective in Eastern Europe and Central Asia due to their relatively large stock of skilled engineers and scientists and domestic enterprises' relative inexperience with modern marketing and management before the start of the transition.

Since it is hard to directly assess the effect of foreign investment on technology transfers, most studies have focused on the effect of foreign ownership on productivity. In general, there is strong evidence that foreign investment improves productivity in enterprises in developing and transition economies, with many recent studies finding that productivity is higher and productivity growth faster in foreign-owned enterprises in these countries. For example, in a recent study using panel data from Venezuela, Aitken and Harrison (1999) find that foreign ownership increases productivity in small, but not large, manufacturing plants, even after controlling for plant-specific effects. In contrast, Haddad and Harrison (1993) found that foreign-owned enterprises in Morocco were more productive than wholly domestically owned enterprises, but that their productivity grew more slowly. Since the start of the transition, many studies have looked at the effect of foreign ownership on productivity and productivity growth in Eastern Europe and Central Asia, generally finding that foreign-owned enterprises are more productive than other enterprises. (5)

Although it might not be surprising that foreign-owned enterprises are more efficient than other enterprises in developing and transition economies, foreign ownership might have broader benefits for the economy as a whole. In addition to affecting the technology, and productivity, of the recipient firm, foreign investment might have spillover benefits for other enterprises in the host country. Saggi (2000) lists several potential spillovers including:

1. 'Demonstration effects', where domestically owned enterprises are able to observe the technologies that the foreign-owned enterprise uses and the goods that it produces and can imitate the production processes or reverse engineer products, allowing the foreign-owned enterprises' technologies to spread throughout the economy.

2. Labour turnover, where domestic enterprises hire former employees of the foreign-owned enterprise gaining access to the foreign-owned enterprise's products or processes.

3. Vertical linkages, where foreign-owned enterprises transfer technologies or provide technical support to the enterprises that are their suppliers or customers or to whom they sub-contract work.

Saggi (2000) distinguishes between these 'pure' externalities and pecuniary externalities that result from the effect of foreign investment on market structure.

Although the theoretical possibility of spillovers to other enterprises is attractive, there is little empirical evidence to support the assertion that there are large spillovers associated with foreign investment. First, although some studies have round that the mechanisms that might transmit spillovers are common, others have round little evidence of them. (6) Second, several recent studies that have looked for evidence of spillovers by looking at the effect of foreign entry in a given sector on the productivity of domestically owned enterprises have failed to find strong results.

In the 1970s and 1980s, a large number of studies looked at industry-level data, generally finding that productivity and productivity growth were higher in sectors with significant foreign investment. (7) However, as pointed out in Aitken and Harrison (1999, p. 611), if foreign investors are attracted to sectors that are more productive, domestic firms in these sectors would appear more productive than in other sectors even if spillovers were not important. To try to control for self-selection into industries where domestic enterprises are more efficient, several recent studies have used firm-level data, generally finding little evidence to support the assertion that spillovers are important. In fact, several studies have found that foreign entry might actually harm the productivity of their domestically owned competitors. (8) For example, using data from Morocco in the 1980s, Haddad and Harrison (1993) round that productivity growth was slower for domestic firms in sectors with high foreign investment than it was for firms in other sectors, although the difference was not statistically significant. In addition, Aitken and Harrison (1999) round that foreign investment in a sector actually reduced productivity for domestically owned plants in Venezuela. In a similar analysis for the Czech Republic, Djankov and Hoekman (2000) also round that foreign investment reduced the productivity of wholly domestically owned enterprises. One plausible explanation for the negative effect on domestically owned enterprises might be that foreign entry affects market structure. Aitken and Harrison (1999, p. 607) note:

 If imperfectly competitive [domestic] firms face fixed costs of
 production, a foreign firm with lower marginal costs will have an
 incentive to increase production relative to its domestic
 competitor. In this environment, entering foreign firms producing
 for the local market can draw demand from domestic firms, causing
 them to cut production. The productivity of domestic firms would
 fall as they spread their fixed costs over a smaller market, forcing
 them back up their average cost curves. If the productivity decline
 from this demand effect is large enough, net domestic productivity
 can decline even if the multinational transfers technology.

This study looks at whether domestically owned enterprises that competed with foreign enterprises were more likely to adopt a new technology (ie, access to the Internet), not at the effect of foreign entry on domestic productivity. This allows us to identify whether foreign investment encourages the adoption of new technologies, without being concerned about negative effects on market structure.

When considering the results from the paper, it is important to keep two important points in mind. A first point is that even if enterprises competing with foreign-owned firms are more likely to adopt the new technology (ie, access to the Internet) than enterprises competing with domestically owned firms, this would not rule out the possibility that foreign entry has a negative impact on the productivity of domestic enterprises. Even if domestically owned enterprises competing with foreign enterprises are more likely to adopt the new technology than other domestic enterprises, this does not necessarily imply that they are able to use if effectively to improve productivity. (9) Consequently, it might have little impact on overall productivity. Further, even if adoption does raise productivity, it would still be possible that negative pecuniary externalities might outweigh any positive spillovers from the adoption of the new technology.

A second point is that the paper looks at the adoption of a specific technology with some characteristics that make adoption and diffusion easier than they would be for other technologies. One important difference between Internet access and other technologies is that the human capital that enterprises need to gain access to the Internet is not highly firm specific. This will make it easier for Internet use to spread through the economy than for firm or industry-specific technologies that require people with very specialised skill sets and mean that it is plausible that 'demonstration effects' might be important for the entire economy, not just for enterprises that are direct competitors. For example, firms will often be able to outsource their Information Technology (IT) needs, reducing the costs associated with adoption. In practice, access to support services for IT does not appear to be an especially severe problem in many low and middle-income countries. For example, 82 percent of enterprises in Peru reported that business support services were available in the area of Information Technology. Of these, 81 percent reported that services were affordable and, of these, 73 percent reported that they used them. Most users (96 percent) reported that service quality was either good or very good. (10)

One additional factor that might make diffusion of Internet use faster and easier than the diffusion of other technologies is that some basic, but useful, processes (eg, e-mail or web use) can be adopted easily. This appears to be have been the case in many low and middle-income countries at the end of the 1990s (ie, when the data used in this paper were collected). Using data from a 1998 survey of International Finance Corporation (IFC) client companies in 16 low- and middle-income countries, Daly and Miller (1998) round that although many enterprises reported using the Internet to send e-mail to other enterprises (74 percent of industrial enterprises) and to conduct Internet searches (60 percent of industrial enterprises), few reported using it in more sophisticated ways. For example, only nine of 113 enterprises reported that the Internet was even slightly useful for conducting on-line financial transactions. Further, Daly and Miller's (1998) results probably overestimate Internet use in low- and middle-income countries, especially for more sophisticated purposes. They note that users were probably more likely to respond to their survey than non-users and that IFC client companies were probably more technologically sophisticated than non-clients. Overall, these results suggest that initial decisions to start using Internet-related services do not require high levels of technical or financial sophistication, making diffusion relatively easy.

EMPIRICAL RESULTS

Data

The main source of data used in this paper is the World Business Environment Survey (WBES), a cross-sectional survey of industrial and service enterprises conducted in mid-1999 by the World Bank and several other agencies. (11) The main purpose of the WBES is to identify perceived constraints on enterprise performance and growth in developing and transition economies. The survey, therefore, has a large number of questions on how taxation, regulation, the performance of the financial sector, the institutional environment and corruption affect business operations. In contrast, the survey includes little information on enterprise characteristics or performance; although some information on assets, sales, broad sector of operations, ownership, employees, and enterprise growth was collected, detailed balance sheet information and profit and loss statements were not collected from participating enterprises. Further, although the WBES asked similar questions in the 80 countries, there were some differences between regions. For the purpose of this study, the most important difference was that questions on Internet access were asked only in Eastern Europe and Central Asia. (12)

In Eastern Europe and Central Asia, about 33 percent of enterprises in the WBES sample reported having access to the Internet (see Table 1). (13) However, this varied greatly between countries. Enterprises in Slovenia were most likely to report having access to the Internet (84.8 percent), while enterprises in Azerbaijan were least likely to report having access (7.8 percent). In general, enterprises in the CIS were less likely to report having access to the Internet than in any other region (see Figure 1). To control for country difference that might affect Internet access, either a set of country dummies or a set of country control variables are included in the analysis. The country-level control variables include telephone lines per 100 inhabitants, to control for development of the telecommunications sector, per capita income, urban population, and size of the country (see Table 1).

The main variables of interest are related to the enterprise's interactions with foreign enterprises. These include whether the enterprise has any foreign ownership (see Table 1), the overall level of FDI and imports into the country (see Table 1), and whether its main competitors were either foreign-owned enterprises producing in the home market or imports (see Table 2). Since most foreign investment in these countries is from the industrialised economies, where Internet access is more common than in Eastern Europe and Central Asia, it seems plausible that foreign-owned enterprises will be more likely to have access to the Internet than domestically owned enterprises. (14)

The information on the enterprises' competitors comes from a question that enterprises were asked about the main source of competition they faced in domestic markets. If there were substantial demonstration or labour turnover effects, enterprises facing competition from foreign-owned enterprises should be more likely to adopt similar technologies to foreign competitors than other enterprises. Further, if demonstration effects require direct observation or only occur when domestically owned companies hire former employees of their foreign competitors, then the effect of competition from foreign-owned local enterprises should be greater than the effect of competition from imports. Finally, if spillovers from foreign ownership are large, then foreign direct investment in other sectors of the economy might affect enterprises that are not direct competitors. Consequently, measures of total FDI and imports are also included in the analysis with country-level controls. (15)

In addition to providing information on Internet access, the survey also provided additional information on the enterprise's performance (see Table 1), the enterprise's largest shareholder, how many competitors the enterprise faced in domestic markets, how many full-time employees the enterprise had and the enterprise's sector of operations (see Table 2). (16) These are included in the analysis to control for enterprise-specific factors that might affect whether the company has Internet access. Since Internet access might affect enterprise performance rather than performance affecting Internet access, the analysis is conducted both with and without these variables.

Econometric model

The probability that enterprise i in country j has access to the Internet is assumed to be a function of a vector of enterprise characteristics ([X.sub.ij]) and country characteristics ([Z.sub.j]). The enterprise characteristics include ownership, sector of operations, size, how the enterprise was established, competition faced by the enterprise, and, in some specifications, enterprise performance. The country characteristics include per capita income, openness to trade and investment, telephone coverage, population and urban population. The probability of enterprise i having access to the Internet is

Prob(Internet [Access.sub.ij]) = [phi] ([alpha] + [beta][X.sub.ij] + [gamma][Z.sub.j])

where [phi](*) is the standard normal distribution and ([alpha], [beta], [gamma]) is the vector of coefficients. The model is estimated using standard maximum likelihood estimation. All estimated models in Table 3 include dummies indicating sector of operations and size of the enterprise (see Table 2 for categories). Results from the model are shown in Table 5.

Econometric results

Foreign shareholdings and largest shareholder

The coefficient on a dummy variable indicating that the enterprise has some foreign shareholders is positive and statistically significant (see Table 5, column 1). This suggests that enterprises in Eastern Europe and Central Asia that are at least partially foreign-owned are more likely to have access to the Internet than similar enterprises without foreign ownership. The results are similar whether country-level control variables or country dummies are used to control for country differences (see Table 5, columns 1 and 2). After controlling for whether an enterprise has any foreign ownership, enterprises with foreign companies as their largest shareholder do not appear any more likely to have access to the Internet than enterprises where the foreign owner is only a minority shareholder. (17) However, if the dummy variable indicating any foreign ownership is dropped, the dummy indicating that the largest shareholder is foreign becomes statistically significant and large (see Table 3, columns 5 and 4). (18) One concern regarding these results is that foreign enterprises might be especially likely to enter sectors and countries where Internet access is particularly common. Although the other control variables (eg, sector dummies, country controls and country dummies, and the dummies indicating whether the enterprise is competing with foreign or domestic enterprises) control for this to some degree, this remains a concern. (19)

Foreign ownership has a large effect on the probability that the enterprise has Internet access. Whereas a state-owned enterprise without any foreign shareholders has a 24.4 percent chance of having Internet access (see Table 4), a foreign-owned enterprise with a foreign company as its largest shareholders is twice as likely to have access to the Internet (48.8 percent). A state-owned company with some foreign ownership (ie, a company where the government is the largest shareholder but where a foreign company has a minority stake) has a 46.8 percent chance of having access to the Internet (see Table 4).

In addition, insider-owned enterprises generally appear to be less likely to have access to the Internet than other enterprises. The coefficient on the dummy variable indicating employee ownership is negative and statistically significant whether country controls or country dummies are included in the analysis, indicating that employee-owned enterprises are less likely to have Internet access that state-owned enterprises. Furthermore, the coefficient on employee ownership is smaller than the coefficient on other private ownership, suggesting that employee-owned enterprises are also less likely to have Internet access than outsider-owned private domestic enterprises. (20) In contrast, the null hypothesis that the coefficients for manager and employee-owned enterprises are equal cannot be rejected. (21)

The coefficient on the dummy variable indicating that the enterprises' managers are the largest shareholders is also negative, suggesting that manager-owned enterprises are less likely to have Internet access than state-owned enterprises. However, the coefficient is statistically insignificant when country controls are included in the analysis. Manager-owned enterprises also appear to be less likely to have Internet access than 'other' private enterprises. (22) Based upon the point estimates of the coefficients in column 1 of Table 3, manager-owned enterprises have a 17.1 percent chance of having Internet access; employee-owned enterprises have a 17.5 percent chance; while similar state-owned enterprises have a 24.4 percent chance (see Table 4).

Competition from foreign-owned enterprises

Enterprises who saw either foreign-owned enterprises producing domestically or imports as their main competition were more likely to have Internet access than enterprises that saw domestically owned enterprises as their main competition. In both cases, the effect is quite large. A (state-owned) enterprise whose main competition is foreign-owned enterprises producing domestically has a 34.5 percent chance of having Internet access, an enterprise whose main competition is imports has a 34.9 percent chance, whereas an enterprise whose main competition is domestically owned enterprises has only a 24.4 percent chance (see Table 4).

The result for competition with foreign-owned enterprises producing domestically is consistent with the hypothesis that demonstration or labour turnover effects affect enterprises' decisions to adopt new technologies (access to the Internet in this case). However, the coefficient on the dummy variable indicating that imports are the enterprise's main competition is similar in size to the coefficient indicating that foreign-owned domestic enterprises are the main competition. (23) If demonstration effects were important either because of direct observation of foreign-owned enterprises' operations or because domestically owned enterprises hire workers from foreign-owned plants, the coefficient on the dummy variable indicating competition with foreign-owned enterprises producing in the country should be larger than the coefficient indicating competition with imports. Taken together, these results suggest that although openness to trade and investment increase the likelihood that domestically owned competitors have Internet access, foreign investment is no more effective than trade in this respect.

One final question is whether the difference between foreign enterprises and domestic enterprises is greater in areas where enterprises are competing with foreign-owned enterprises than in areas where they are competing with domestic enterprises. To test whether this is the case, interaction terms between foreign ownership and competition are included in the base analysis (see Table 6, columns 1 and 2). The coefficients on the interaction terms are statistically insignificant indicating that foreign enterprises are equally more likely to have Internet access than similar domestic enterprises whether they are in sectors where their main competition is other foreign enterprises or they are in sectors where their main competition is domestic enterprises (see Figure 2).

[FIGURE 2 OMITTED]

Enterprise origins and competition

Enterprises that were established as either joint ventures or as private enterprises (ie, de novo private enterprises) were more likely to have access to the Internet than similar enterprises that either remained state-owned or had been privatised. The difference is quite large, with de novo enterprises having a 38.0 percent chance of having Internet access, joint ventures having a 66.0 percent chance, while state-owned or privatised enterprises have only 24.4 percent and 27.1 percent probabilities respectively. Finally, enterprises with no effective competition were generally more likely to have Internet access than enterprises with either one to three competitors or enterprises with more than three competitors (see Table 3). However, this result is not highly robust. When variables indicating enterprise performance are included in the analysis (see columns 5 and 6), the coefficient drops in both size and significance level. One plausible reason for this finding might be that enterprises facing little effective competition perform better, giving them the funds needed to invest in new technologies, such as Internet access.

Country-level measures of openness

In addition to the enterprise-level variables discussed above, the analysis also includes some country-level variables. Since these variables become collinear with the dummies once the country dummies are added, they are dropped when country dummies are included (see Table 3, columns 2, 4 and 6). The coefficient on foreign direct investment is statistically insignificant suggesting that FDI does not have a large effect on the probability that enterprises other than the enterprise the foreign company invests in (and the enterprise's competitors) have Internet access.

One concern is that the result for FDI might be affected by the inclusion of the oil-producing economies of Central Asia. FDI in two of these economies in particular, Azerbaijan and Kazakhstan, has been far higher than in any other country in the CIS since the start of transition. (24) However, this investment has almost exclusively flowed to the off sector and it is possible that spillovers to the rest of the economy from FDI in this sector are smaller than the spillovers from other FDI. (25) The results omitting the oil-producing economies of Central Asia are consistent with this hypothesis. Once these economies are omitted, the coefficient on FDI increases in magnitude and becomes statistically significant at a 1 percent level (see Table 6, columns 3 and 4). (26) The point estimate of the elasticity on FDI increases to 0.21 when these economies are omitted.

One puzzling result is that openness to trade at the country level appears to have a negative impact on Internet access; the coefficient on imports as a share of GDP is statistically significant and negative (see Table 3). The point estimate of the coefficient suggests that a 1 percent increase in imports decreases the probability that domestically owned enterprises have access to the Internet by 0.55 percent (see Table 5). Given that enterprises competing with imports appear more likely to have access to the Internet than enterprises competing with domestic enterprises, it might be surprising that the broader spillover effects from imports are negative.

To investigate this further, we divide imports into imports of raw materials (ie, imports of fuel, minerals, ores, and food) and imports of manufactured goods. (27) When this is done, the coefficient on imports of raw materials is negative and statistically significant, while the coefficient on imports of manufactured goods is positive, although statistically insignificant (see column 5 in Table 6). This suggests that imports of raw materials and imports of manufactured goods have different spillover effects on Internet access. One plausible explanation for this is whereas, on average, 72 percent of imports of manufactured goods in the countries in this sample come from high-income countries, only 31 percent of raw material imports come from these countries. Given that Internet access is significantly higher in high-income countries than in low- and middle-income countries, imports from high-income countries might have positive spillover effects, while imports from other low- and middle- income countries do not. Consistent with interpretation, when imports are split into imports from high-income countries and imports from low- and middle-income countries, the coefficient on imports from high-income countries is statistically significant and positive and the coefficient on imports from low-income countries is statistically significant and negative (see column 6 in Table 6). Unfortunately, the questions on the enterprises' main competitors do not ask whether the competitors are from high-income or low-income countries. Because the question on the enterprises' main competitors pools enterprises competing with imports from both low- and high-income countries, the impact of competition from imports from high-income countries might be greater than the estimated coefficient suggests. (28)

Country controls

The other country controls are also significant at, at least, a 5 percent level throughout the analysis. In general, enterprises in countries with higher per capita income, with larger urban populations and smaller countries appear more likely to have access to the Internet. In addition, enterprises in countries with more developed telecommunications systems appear to be more likely to have access to the Internet. A 1 percent increase in the number of mainlines per 100 inhabitants increases the probability that an enterprise has access to the Internet by 0.5 percent. This result is consistent with results from a country-level analysis in Dasgupta et al. (2000), which suggest that that cross-country differences in Internet use reflect the number of fixed mainlines per capita in a country. Including country dummies does not appear to either affect the enterprise-level results or to greatly increase the explanatory power of the analysis--the pseudo-[R.sup.2] is similar whether country dummies or country controls are included (see Tables 3 and 6). (29)

Enterprise performance

As a final set of control variables, some additional indicators of enterprise performance are also included in the analysis, including employment and sales growth--in general, better performing enterprises should contract less than worse performing enterprises--and percent of sales to the government. Since there is a large literature showing that foreign-owned enterprises in the transition economies generally perform better than domestically owned enterprises along a variety of dimensions, foreign-owned enterprises might be more likely to have access to the Internet simply because their stronger performance gives them better access to investment resources. (30) Similarly, employee-owned enterprises, which appear to perform worse than other enterprises, might have fewer resources for investment. (31)

In general, better performing enterprises appear to be more likely to have access to the Internet than worse performing enterprises (see Table 3), perhaps because they have more resources available for investment in new technologies. However, this has virtually no effect on other results. Most notably, the coefficients on foreign- and insider-ownership are virtually unchanged and remain highly significant even after these performance measures are added to the analysis. This suggests that better (worse) performance is not the only reason for the higher (lower) levels of access to the Internet for foreign- (employee-) owned enterprises.

Although performance might affect Internet access, Internet access might also affect enterprise performance, introducing the possibility of reverse causation when the performance variables are included. Therefore, the analysis is conducted both with and without these variables (see columns 1 and 2 and columns 5 and 6 in Table 3 respectively). In practice, the main results are virtually identical whether these performance indicators are included in the analysis or not.

CONCLUSIONS

The results from this study support the assertion that foreign investment increases Internet access for enterprises in low and middle-income countries in Eastern Europe and Central Asia. The strongest result is that Internet access is more common among enterprises that are partly foreign-owned than it is among enterprises that are fully domestically owned. The effect of foreign ownership appears large--enterprises that are partly foreign-owned are almost twice as likely to have access to the Internet as state-owned and privately owned enterprises with no foreign ownership. Further, the correlation between foreign ownership and Internet access does not seem to be simply because foreign-owned enterprises tend to outperform other enterprises in the transition economies, giving them easier access to financing. The correlation remains statistically significant even after including variables to control for enterprise performance and indicators of the level of competition that the enterprise faces in domestic markets.

The results also suggest that foreign investment has positive spillovers for other domestically owned enterprises with respect to Internet access. In particular, enterprises that compete with either foreign-owned domestic enterprises or imports are more likely to have Internet access. Since competition with imports and foreign-owned domestic enterprises both appear to increase the likelihood, this suggests that proximity is not very important. Although past studies (eg, Aitken and Harrison, 1999) have found that competition from foreign-owned firms reduces the productivity of their domestic competitors, the negative effect of foreign entry on the productivity of domestic competitors is thought to be due to foreign entry affecting market structure. Since this study does not address the question of the size, or even existence, of benefits related to Internet access, it is unclear whether positive technological spillovers round in this study would outweigh pecuniary externalities.

Finally, Internet access appears more common in countries that are more open in terms of foreign investment and trade. In particular, enterprises are more likely to have Internet access in countries that import more from high-income countries. In contrast, imports from low- and middle-income countries are negatively correlated with Internet access. In addition, Internet access in more common in countries with higher levels of FDI even after controlling for other factors (eg, urbanisation, per capita income and telecommunications infrastructure) that might also affect Internet access. It is important to note that the result for FDI holds only after the oil-exporting economies of Central Asia are excluded from the analysis. This strongly suggests that FDI does not always increase the likelihood that a domestic enterprise will have Internet access--investment in a single (extractive) sector might not have the same beneficial effect as other types of investment.

Other factors also affect Internet access. Employee-owned enterprises are less likely to have access to the Internet than outsider-owned private enterprises or state-owned enterprises. This holds when country dummies and performance measures are included in the analysis, suggesting that it is not due to employee ownership being more common in countries where Internet access is restricted or to employee-owned enterprises finding it harder to finance new investment. Finally, enterprises in countries with better telephone systems are more likely to have Internet access even after controlling for income and urbanisation. This result is consistent with results from a country-level study by Dasgupta et al. (2000), which suggests that the number of mainlines per capita explains most of the gap between developed and developing countries with regards to Internet connectivity. This stresses that steps that would improve the performance of providers of fixed-line telephone services (eg, privatising state-owned fixed line monopolies) would increase Internet access.

The presence of positive spillovers from foreign investment suggests that it might be appropriate for governments to take steps to encourage foreign direct investment. However, although there is some evidence that investment in information technology has improved the productivity of enterprises in the US, there is very little evidence on the how great the effect of Internet access or investment in information technology is on firm performance in developing or transition economies. (32) Although the lack of evidence regarding the effect of Internet access on firm performance in the transition economies argues against taking dramatic steps to encourage foreign investment, it does give added weight to arguments for improving the business environment. For example, there is strong evidence that corruption, which is a serious problem in many transition economies, discourages foreign investment and slows economic growth. (33) Since reducing corruption and taking other steps to improve the business environment would both encourage foreign investment and improve the functioning of the domestic economy, they would benefit the domestic economy even if Internet access had little short-term impact on productivity or growth.

Acknowledgements

The data used in this paper are from the World Business Environment Survey (WBES) [c] 2000 The World Bank Group. I thank an anonymous referee, L. Colin Xu, Ricardo Martin and participants at the 2003 Transformation, Integration and Globalization Economic Research (TIGER) conference 'The 'New Economy' and Post-Socialist Transition' for comments on earlier drafts. I also thank Luke Haggarty and Andrew Stone for their help with the data. Responsibility for all errors, omissions, and opinions rests solely with the author. All findings, interpretations, and conclusions expressed in this paper are entirely those of the author and do not necessarily represent the views of the World Bank, its Executive Directors, or the countries they represent.

Table 1: Means and standard deviations of variables

Variable Source Mean Standard
 deviation

Enterprise characteristics
 Does enterprise have access to WBES 0.33 0.47
 the Internet? (1--yes, 0--no)

 Does any foreign company have WBES 0.08 0.27
 a financial stake in your
 organisation? (1--yes, 0--no)

 Percentage change in employment WBES 6.52 60.39
 between 1996 and 1999

 Percentage change in sales between WBES 13.43 67.33
 1996 and 1999

 Percent of sales accounted for by WBES 16.93 25.50
 state sector

Country control variables
 Net incoming foreign direct investment WDI 4.54 5.11
 in 1998 (share of GDP)

 Imports of goods and services in 1998 WDI 46.68 17.91
 (share of GDP)

 Main telephone lines per 100 inhabitants ITU 22.09 10.61
 in 1999

 Urban Population (share of total) in 1998 WDI 61.69 12.33

 Per capita GDP in 1998 (PPP, international WDI 5.91 3.18
 dollars, 000s)

 Population in 1998 (natural log) WDI 16.41 1.38

Note: For source variables, WBES implies that data come
from the World Business Environment Survey (WBES) [c]
2000 The World Bank Group. WDI implies that data come
from World Bank, 2001. World Development Indicators.
World Bank, Washington DC. ITU implies that data come
from International Telecommunication Union, 2000.
World Telecommunication Indicators Database. International
Telecommunication Union, Geneva, Switzerland.

Table 2: Distribution of enterprises in sample

What is biggest competitive threat to enterprise?
(omitted category is domestic enterprises)
 Foreign firms producing in domestic markets (not imports) 7.4%
 Legal and illegal imports 11.0%

Who is the largest shareholder in enterprise?
(omitted category is government)
 A foreign company 3.4%
 Enterprise's managers 2.9%
 Enterprise's employees 11.0%
 Other private (individuals, families, domestic 65.6%
companies, banks or investment funds)

How was enterprise established? (omitted category is
state-owned, including subsidiaries and privatised
state-owned)
 Private from time of start up (no state-owned predecessor) 53.3%
 Joint venture with foreign and domestic partners 1.3%

How many competitors does enterprise's major product line
face in domestic markets? (omitted category is more than three)
 Between one and three 9.9%
 No competitors 12.6%

How many full-time employees and casual staff in total
work for this company? (omitted category is over 500)
 Less than nine 26.5%
 Between 10 and 49? 20.0%
 Between 50 and 99? 16.0%
 Between 100 and 199? 13.7%
 Between 200 and 499? 15.4%

What is enterprise's main area of activity?
(omitted category is 'other')
 Farming, fishing or forestry 13.5%
 Mining or quarrying 0.8%
 Manufacturing 29.7%
 Building or construction 8.8%
 Power generation 0.4%
 Wholesale trade 12.5%
 Retail trade 14.4%
 Transportation 6.1%
 Financial services 1.6%
 Personal services 5.3%
 Business services 4.9%

Data source: World Business Environment Survey (WBES)
[c] 2000 The World Bank Group

Table 3: Effect of ownership on probability
of enterprise having Internet access

Estimation method Probit
Dependent variable Enterprise has access to Internet
Number of observations 2999 2999
Sector dummies Yes Yes
Size of enterprise dummies Yes Yes
Country dummies No Yes

Foreign shareholding
 Any foreign shareholding 0.6125 *** 0.6361 ***
 (4.67) (4.69)
Ownership
 Largest shareholder--foreign 0.0518 0.0404
 (0.24) (0.18)
 Largest shareholder--managers -0.2581 -0.3436 *
 (-1.49) (-1.95)
 Largest shareholder--employees -0.2398 ** -0.3049 **
 (-2.04) (-2.51)
 Largest shareholder--other private 0.0823 -0.0051
 (0.86) (-0.05)

Competition from foreigners
 Main competition--imports 0.3054 *** 0.3118 ***
 (3.50) (3.47)
 Main competition--foreign-owned 0.2944 *** 0.2608 **
 domestic enterprises (-2.92) (-2.52)

Enterprise-level controls
 Firm established as private enterprise 0.3043 *** 0.3220 ***
 (3.92) (4.01)
 Firm established as joint venture 0.4944 ** 0.5024 **
 (2.07) (2.00)
 Between one and three competitors -0.0141 -0.0554
 (-0.14) (-0.53)
 No competitors 0.1709 ** 0.1441 *
 (2.10) (1.73)

Country-level measures of openness
 Foreign direct investment (% of GDP) 0.0063
 (0.85)
 Imports (% of GDP) -0.0119 ***
 (-4.42)

Country controls
 Number of telephone lines per 100 0.0228 ***
 inhabitants (4.00)
 Urban population (percent of population) 0.0100 **
 (2.44)
 Per capita GDP (OOOs of US$) 0.0826 ***
 (5.40)
 Population (natural log) -0.1713 ***
 (-4.33)

Enterprise-level performance
 Employment growth (over last 3 years)

 Sales growth (over last 3 years)

 Sales to government (% of sales)

Pseudo-[R.sup.2] 0.25 0.28

Estimation method Probit
Dependent variable Enterprise has access to Internet
Number of observations 3006 3006
Sector dummies Yes Yes
Size of enterprise dummies Yes Yes
Country dummies No Yes

Foreign shareholding
 Any foreign shareholding

Ownership
 Largest shareholder--foreign 0.6497 *** 0.6496 ***
 (3.66) (3.58)
 Largest shareholder--managers -0.2041 -0.2853
 (-1.19) (-1.63)
 Largest shareholder--employees -0.2304 ** -0.2950 **
 (-1.97) (-2.43)
 Largest shareholder--other private 0.1174 0.0322
 (1.23) (0.33)

Competition from foreigners
 Main competition--imports 0.3200 *** 0.3265 ***
 (3.69) (3.66)
 Main competition--foreign-owned 0.3036 *** 0.2693 ***
 domestic enterprises (-3.05) (-2.64)

Enterprise-level controls
 Firm established as private enterprise 0.3127 *** 0.3259 ***
 (4.06) (4.10)
 Firm established as joint venture 0.6982 *** 0.7150 ***
 (3.04) (2.97)
 Between one and three competitors 0.0059 -0.0309
 (0.06) (-0.30)
 No competitors 0.1814 ** 0.1536 **
 (2.24) (1.86)

Country-level measures of openness
 Foreign direct investment (% of GDP) 0.0051
 (0.69)
 Imports (% of GDP) -0.0119 ***
 (-4.48)

Country controls
 Number of telephone lines per 100 0.0226 ***
 inhabitants (3.99)
 Urban population (percent of population) 0.0096 **
 (2.37)
 Per capita GDP (OOOs of US$) 0.0814 ***
 (5.33)
 Population (natural log) -0.1793 ***
 (-4.59)

Enterprise-level performance
 Employment growth (over last 3 years)

 Sales growth (over last 3 years)

 Sales to government (% of sales)

Pseudo-[R.sup.2] 0.25 0.27

Estimation method Probit
Dependent variable Enterprise has access to Internet
Number of observations 2798 2798
Sector dummies Yes Yes
Size of enterprise dummies Yes Yes
Country dummies No Yes

Foreign shareholding
 Any foreign shareholding 0.5810 *** 0.6265 ***
 (4.10) (4.28)
Ownership
 Largest shareholder--foreign -0.0101 -0.0183
 (-0.04) (-0.08)
 Largest shareholder--managers -0.2139 -0.3097 *
 (-1.17) (-1.66)
 Largest shareholder--employees -0.2811 ** -0.3228 ***
 (-2.31) (-2.56)
 Largest shareholder--other private 0.0504 -0.0305
 (0.50) (-0.29)

Competition from foreigners
 Main competition--imports 0.3309 *** 0.3378 ***
 (3.62) (3.59)
 Main competition--foreign-owned 0.2370 ** 0.2059 *
 domestic enterprises (2.28) (1.93)

Enterprise-level controls
 Firm established as private enterprise 0.2156 *** 0.2445 ***
 (2.62) (2.88)
 Firm established as joint venture 0.4666 * 0.5184 **
 (1.85) (1.97)
 Between one and three competitors 0.0285 -0.0317
 (0.27) (-0.29)
 No competitors 0.1516 ** 0.1290
 (1.77) (1.48)

Country-level measures of openness
 Foreign direct investment (% of GDP) 0.0062
 (0.80)
 Imports (% of GDP) -0.0121 ***
 (-4.33)

Country controls
 Number of telephone lines per 100 0.0208 ***
 inhabitants (3.46)
 Urban population (percent of population) 0.0113 ***
 (2.65)
 Per capita GDP (000s of US$) 0.0815 ***
 (5.10)
 Population (natural log) -0.1900 ***
 (-4.56)

Enterprise-level performance
 Employment growth (over last 3 years) 0.0023 *** 0.0022 ***
 (3.96) (3.78)
 Sales growth (over last 3 years) 0.0017 *** 0.0017 ***
 (3.61) (3.70)
 Sales to government (% of sales) 0.0009 *** 0.0017
 (0.71) (1.33)
Pseudo-[R.sup.2] 0.27 0.29

Note: t-Statistics in parentheses

*** Significant at 1 percent level.

** Significant at 5 percent level.

* Significant at 10 percent level.

Data source: The World Business Environment Survey (WBES)
[c] 2000 The World Bank Group Omitted categories are state-owned
enterprises (as largest shareholders) and enterprises established
as state-owned enterprises (origin). The pseudo [R.sup.2] is
1-([L.sub.1/[L.sub.0]), where [L.sub.0] is the log-likelihood
of a model with only a constant and [L.sub.1] is the
log-likelihood of the estimated model (Judge et al., 1988).

Table 4: Effect of dummy variables on probability of having access
to the internet

 Probability of
 having Internet
 access (%)

Base enterprise 24.4

Foreign shareholding
 Any foreign shareholding 46.8

Ownership
 Largest shareholder--foreign (a) 48.8
 Largest shareholder--managers 17.1
 Largest shareholder--employees 17.5
 Largest shareholder--other private 27.1

Competition from foreigners
 Main competition--imports 34.9
 Main competition--foreign-owned domestic
 enterprises 34.5

Enterprise-level controls
 Firm established as private enterprise (b) 38.0
 Firm established as joint venture between foreign
 and domestic enterprises (c) 66.0
 Between one and three competitors 24.0
 No competitors 30.1

Note: Probabilities are calculated setting all continuous variables to
their respective means and using coefficients from Table 3, column (1).
The base enterprise is a state-owned enterprise, whose main competition
comes from other domestically owned enterprises, with more than three
competitors for its main product line, with between 50 and 100 workers
(median size), in the manufacturing sector (most common sector). All
other enterprises are the same as the base type with changes as noted
in the title column.

(a) If the largest shareholder is foreign, the dummy indicating any
foreign shareholder is also set to '1'.

(b) If the firm is established as private, the dummy indicating that
the largest shareholder is (other) private (i.e., not state-owned)
is also set to '1'.

(c) If the firm is a joint venture between foreign and domestic, the
dummy indicating some foreign shareholding is set to '1'.

Table 5: Elasticities of the probability of having Internet access
with respect to continuous variables

Variable Elasticity

Country-level measures of openness
 Net incoming foreign direct investment in 1998
 (share of GDP) 0.03
 Imports of goods and services in 1998 (share of GDP) -0.55 ***

Country control variables
 Main telephone Lines per 100 inhabitants in 1999 0.50 ***
 Urban Population (share of total) in 1998 0.62 **
 Per capita GDP in 1998 (PPP, international dollars,
 000s). 0.49 ***
 Population in 1998 (natural log) -0.17 ***

Enterprise-level performance
 Percentage change in employment between 1996 and 1999. 0.04 ***
 Percentage change in sales between 1996 and 1999. 0.03 ***
 Percent of sates accounted for by state sector. 0.03 ***

*** Significant at 1 percent level. ** Significant at 5 percent level.
* Significant at 10 percent level. Note: Probabilities are calculated
setting all continuous variables to their respective means and using
coefficients from Table 3, column (1). The base enterprise is a
state-owned enterprise, whose main competition comes from other
domestically owned enterprises, with more than three competitors
for its main product line, with between 50 and 100 workers
(median size), in the manufacturing sector (most common sector).

Table 6: Effect of ownership on probability of enterprise having
Internet access

Estimation method
Dependent variable
Sample All All

Number of observations 2999 2999
Sector dummies Yes Yes
Size of enterprise dummies Yes Yes
Country dummies No Yes

Foreign shareholding
 Any foreign shareholding 0.6280 *** 0.6579 ***
 (4.26) (4.31)

Interaction term
 Foreign companies facing competition 0.0937 0.0932
 from foreign-owned companies (0.32) (0.31)
 Foreign companies facing competition -0.1543 -0.1799
 from imports (-0.56) (-0.64)

Ownership
 Largest shareholder--foreign 0.0416 0.0268
 (0.19) (0.12)
 Largest shareholder--managers -0.2620 -0.3487**
 (-1.51) (-1.98)
 Largest shareholder--employees -0.2407 ** -0.3061 **
 (-2.05) (-2.52)
 Largest shareholder--other private 0.0811 -0.0065
 (0.84) (-0.07)

Competition from foreigners
 Main competition--imports 0.3220 *** 0.3312 ***
 (3.52) (3.52)
 Main competition--foreign-owned 0.2785 *** 0.2441 **
 domestic enterprises (2.57) (2.20)

Enterprise-level controls
 Firm established as private 0.3036 *** 0.3211 ***
 enterprise (3.90) (4.00)
 Firm established as joint venture 0.4961 ** 0.5035 **
 (2.06) (1.99)
 Between one and three competitors -0.0150 -0.0565
 (-0.15) (-0.54)
 No competitors 0.1715 ** 0.1447 *
 (2.10) (1.73)

Country-level measures of openness
 Foreign direct investment (% of GDP) 0.0061
 (0.83)
 Imports (% of GDP) -0.0118 ***
 (-4.39)
 Manufacturing imports (% of GDP)

 Raw material imports (% of GDP)

 Imports from low- and middle-income
 countries (% of GDP)
 Imports from high-income countries
 (% of GDP)

Country controls
 Number of telephone Lines per 100 0.0227 ***
 inhabitants (3.97)
 Urban population 0.0099 ***
 (percent of population) (2.42)
 Per capita GDP (OOOs of US$) 0.0830 ***
 (5.41)
 Population (natural Log) -0.1705 ***
 (-4.30)
Pseudo-[R.sup.2] 0.25 0.28

 Probit
 Enterprise has access
 to Internet

Estimation method Oil Oil
Dependent variable Exporters Exporters
Sample Omitted Omitted

Number of observations 2638 2638
Sector dummies Yes Yes
Size of enterprise dummies Yes Yes
Country dummies No Yes

Foreign shareholding
 Any foreign shareholding 0.7015 *** 0.7256 ***
 (4.95) (4.95)

Interaction term
 Foreign companies facing competition
 from foreign-owned companies
 Foreign companies facing competition
 from imports

Ownership
 Largest shareholder--foreign -0.1514 -0.1592
 (-0.65) (-0.66)
 Largest shareholder--managers -0.2946 * -0.3684 **
 (-1.67) (-2.04)
 Largest shareholder--employees -0.2571 ** -0.3304 **
 (-2.06) (-2.55)
 Largest shareholder--other private 0.0644 -0.0252
 (0.62) (-0.23)

Competition from foreigners
 Main competition--imports 0.3200 *** 0.2952 ***
 (3.42) (3.09)
 Main competition--foreign-owned 0.2426 ** 0.2193 **
 domestic enterprises (2.32) (2.04)

Enterprise-level controls
 Firm established as private 0.3030 *** 0.3230 ***
 enterprise (3.72) (3.83)
 Firm established as joint venture 0.4593 * 0.5013 *
 (1.72) (1.78)
 Between one and three competitors 0.0442 -0.0156
 (0.41) (-0.14)
 No competitors 0.1184 0.0937
 (1.38) (1.06)

Country-level measures of openness
 Foreign direct investment (% of GDP) 0.0463 ***
 (3.54)
 Imports (% of GDP) -0.0125 ***
 (-4.57)
 Manufacturing imports (% of GDP)

 Raw material imports (% of GDP)

 Imports from low- and middle-income
 countries (% of GDP)
 Imports from high-income countries
 (% of GDP)

Country controls
 Number of telephone Lines per 100 0.0193 ***
 inhabitants (3.20)
 Urban population 0.0063
 (percent of population) (1.38)
 Per capita GDP (OOOs of US$) 0.0990 ***
 (6.11)
 Population (natural Log) -0.1189 ***
 (-2.76)
Pseudo-[R.sup.2] 0.25 0.27

Estimation method
Dependent variable
Sample All All

Number of observations 2636 6636
Sector dummies Yes Yes
Size of enterprise dummies Yes Yes
Country dummies No No

Foreign shareholding
 Any foreign shareholding 0.6019 *** 0.5546 ***
 (4.24) (3.87)

Interaction term
 Foreign companies facing competition
 from foreign-owned companies
 Foreign companies facing competition
 from imports

Ownership
 Largest shareholder--foreign 0.0259 0.0334
 (0.11) (0.14)
 Largest shareholder--managers -0.2248 -0.2415
 (-1.25) (-1.34)
 Largest shareholder--employees -0.1828 -0.2107 *
 (-1.47) (-1.68)
 Largest shareholder--other private 0.1350 0.0895
 (1.32) (0.87)

Competition from foreigners
 Main competition--imports 0.2743 *** 0.2415 **
 (2.94) (2.57)
 Main competition--foreign-owned 0.3004 *** 0.2775 ***
 domestic enterprises (2.84) (2.61)

Enterprise-level controls
 Firm established as private 0.2585 *** 0.2578 ***
 enterprise (3.11) (3.08)
 Firm established as joint venture 0.4246 0.4469
 (1.54) (1.59)
 Between one and three competitors -0.1085 -0.0965
 (-0.96) (-0.85)
 No competitors 0.1825 ** 0.1688 *
 (2.06) (1.89)

Country-level measures of openness
 Foreign direct investment (% of GDP) -0.0057 -0.0052
 (-0.57) (-0.63)
 Imports (% of GDP)

 Manufacturing imports (% of GDP) 0.0042
 (0.66)
 Raw material imports (% of GDP) -0.0191 *
 (-1.85)
 Imports from low- and middle-income -0.0164 ***
 countries (% of GDP) (-4.51)
 Imports from high-income countries 0.0133 ***
 (% of GDP) (3.07)

Country controls
 Number of telephone Lines per 100 0.0269 *** 0.0227 ***
 inhabitants (3.88) (3.30)
 Urban population 0.0039 0.0026
 (percent of population) (0.64) (0.43)
 Per capita GDP (OOOs of US$) 0.0475 * 0.0341 *
 (1.75) (-1.92)
 Population (natural Log) -0.0607 -0.0189
 (-0.99) (-0.31)
Pseudo-[R.sup.2] 0.25 0.26

Data source: The World Business Environment Survey (WBES)
(c) 2000 The World Bank Group

See Table 3 for notes to this table.

(1) Although some formal analyses have supported the assertion that investment in information technology increased labour productivity in the 1990s, others have found only modest effects. For example, Oliner and Sichel (2000) find that 0.45 percentage points of a roughly 1 percentage point increase in labour productivity in the non farm business sector could be attributed to investment in information technology. In contrast to results in Oliner and Sichel (2000), which suggested widespread benefits from investment in information technology, Gordon (2000) round that the gains were concentrated in computer hardware manufacturing and that there was no increase in productivity outside of durable manufacturing. Oliner and Sichel (2000, p. 19) attribute the difference in results to Gordon's (2000) treatment of cyclical effects. In a survey of firm-level evidence, Brynjolfsson and Hitt (2000) argue that the firm-level evidence suggests that information technology started affecting productivity in the early 1990s. Although several studies have round that investment in IT has improved procductivity in the US, the direct impact of e-commerce is thought to be small even in the United States. For example, Oliner and Sichel (2000) estimate that e-commerce has increased multifactor productivity growth in the US by considerably less than 0.1 percent per year. Since e-commerce has almost certainly had a greater impact in the US than it has had in middle and low-income eeonomies, the impact in the developing and transition economies is likely to be very small.

(2) For example, Sachs (2000) proposes FDI as a way of increasing access to technology (although not just information technology) in developing countries. Blomstrom and Kokko (1996), Barba Navaretti and Tarr (2000), and Saggi (2000) provide recent reviews of the literature on the effect of foreign investment and trade on the diffusion of technology in developing countries.

(3) Aitken and Harrison (1999, p. 607) suggest that entry by foreign-owned enterprises that are more efficient than domestic enterprises might cause a short-term drop in the efficiency of domestic enterprises if it reduces demand for their products, stopping them from achieving economies of scale.

(4) Research and development (R&D) expenditures are far lower in developing and transition economies than in developed countries, both in absolute per capita terms and as a share of GDP. For example, R&D expenditures accounted for about 2.4 percent of GNP in high-income OECD countries in 1996, but only 0.8 percent of GNP in the transition economies of Europe and Central Asia, similar to the level for other low- and middle-income economies. Data are from World Bank (2001), World Development Indicators. In 1994, the last year for which data were available, R&D expenditures accounted for about 0.84 percent of GNP in low- and middle-income countries.

(5) Djankov and Murrell (2002) presents a meta-analysis synthesising results from 23 studies that look at the effect of ownership on various measures of performance (ie, not just productivity) in the transition economies. They find that, overall, foreign-owned enterprises appear to perform better than, or as well as, all other ownership types in the transition economies.

(6) For example, although Pack (1997) finds a large amount of labour turnover between foreign multinationals and domestic enterprises in Taiwan, Gershenberg (1987) finds only limited turnover in Kenya.

(7) Saggi (2000), Haddad and Harrison (1993), and Barba Navaretti and Tarr (2000) provide brief surveys of this literature. Studies include Caves (1974), Globerman (1979), Blomstrom and Persson (1983), Blomstrom (1986), and Blomstrom and Wolff (1994).

(8) Other enterprise-level studies have found evidence of positive productivity spillovers. For example, Blomstrom and Sjoholm (1999) find positive spillovers on labour productivity of domestic firms from both majority and minority foreign investment in Indonesia in 1991.

(9) For example, domestic enterprises might be able to use new technologies productively only if they have sufficient levels of human capital. Consistent with this, Borensztein et al. (1998) find that FDI is more productive than domestic investment only when countries have a minimum threshold of human capital.

(10) Data come from the 2002 Investment Climate Survey for Peru, which was conducted by the World Bank in collaboration with Andean Development Corporation. The survey is described in World Bank (2003).

(11) The survey of the transition economies was conducted in collaboration with the European Bank for Reconstruction and Development. Hellman et al. (2000) and European Bank for Reconstruction and Development (1999) provide more complete descriptions of the survey.

(12) The countries in the sample were: (Commonwealth of Independent States) Armenia, Azerbaijan, Belarus, Georgia, Kazakhstan, the Kyrgyz Republic, Moldova, Russia, Ukraine and Uzbekistan; (Early Applicants to the EU) Czech Republic, Estonia, Hungary, Poland and Slovenia; (Other Central Europe and the Baltics) Lithuania and the Slovak Republic; (Southeastern Europe) Albania, Bulgaria, Croatia, and Romania. Questions on Internet access were also asked in Cambodia, Thailand, Turkey, and the West Bank. However, since these additional countries are less comparable with the economies of Eastern Europe and Central Asia than these economies are with each other, and because other control variables were not available for the additional countries, they are omitted from the analysis.

(13) The survey question reads '[d]oes your company have Internet access?' Although it seems that most managers answering the survey would interpret this to mean 'does the company have Internet access on the premises', it seems plausible that the manager might interpret it to include other methods of access (eg, whether he or she has access to the Internet at home).

(14) The most important countries were Germany, the United States, the United Kingdom, France and Austria. Only nine of the 268 foreign enterprises were from Russia.

(15) These measures are omitted when country dummies are included since they are collinear with them.

(16) The WBES provided categorical information on number of employees, not the actual number.

(17) The null hypotheses that majority foreign-owned enterprises are only as likely to have Internet access as employee-owned, manager-owned, state-owned and 'other' privately owned enterprises with some foreign ownership can be tested by comparing the coefficient on majority foreign-owned enterprises with the coefficients on the dummies for the other classes of largest shareholder. The null hypotheses cannot be rejected at conventional significance levels for any of the ownership categories. In the regressions with country controls, the [chi square][1] statistics are 1.53 (P-value=0.21), 1.68 (P-value =.19), 0.02 (P-value = 0.88) and 0.06 (P-value = 0.81) for manager-owned, employee-owned, other private domestic, and state-owned enterprises respectively. Results are similar for the regressions with country dummies.

(18) Tests of the null hypothesis that the coefficient on majority foreign-owned enterprises is different from the other coefficients reject the null hypothesis at conventional significance levels once the dummy variable indicating any foreign ownership is dropped. In the regressions with country controls, the [chi square][1] statistics are 14.94 (P-value=0.00), 22.35 (P-value=0.00), 10.32 (P-value=0.00) and 13.41 (P-value=0.00) for manager-owned, employee-owned, other private domestic, and state-owned enterprises respectively. Results are similar for the regressions with country dummies.

(19) Although it would be possible to control for endogeneity using a two stage estimation method, it is difficult to find instruments that are likely to be correlated with foreign ownership but uncorrelated with Internet access.

(20) When country controls are included, a hypothesis test rejects the null hypothesis that the coefficients on employee-owned and 'other' private enterprises are equal at conventional significance levels ([chi square][1] = 11.53, P-value = 0.00), Results are similar when country dummies are included.

(21) When country controls are included, the null hypothesis that the coefficients on manager owned enterprises and employee-owned enterprises are equal cannot be rejected at conventional significance levels ([chi square][1] = 0.02. P-value = 0.88). Results are similar when country dummies are included.

(22) When country controls are included, a hypothesis test of whether the coefficients are equal for manager owned enterprises and other private domestic enterprises rejects the null hypothesis that the coefficients are equal ([chi square][1]=4.19, P-value = 0.04). Results are similar when country dummies are included.

(23) We are unable to reject the null hypothesis that the coefficients are equal at conventional significance levels when either country controls or country dummies are included in the analysis.

(24) Between 1993 and 1998, there was $509 of FDI per capita in Azerbaijan and $431 per capita in Kazakhstan. In comparison, there was less than $130 per capita over the same period in the CIS and less than $100 per capita in most of the other economies. The other oil exporting countries in Central Asia have received far less FDI, $179 per capita in Turkmenistan and $31 per capita in Uzbekistan. Russia has also received far less FDI--$84 per capita over the same period. Data are from European Bank for Reconstruction and Development (2000).

(25) For example, in 1998, there was $129 of FDI per capita in Azerbaijan. However, there was only $24 per capita outside of the oil sector. Excluding investment in the oil sector, FDI in Azerbaijan was similar to the level in other CIS economies for that year. Data are from International Monetary Fund (2000).

(26) Most of the other results of interest do not appear to be affected by this change. The only changes are that the coefficient on the dummy indicating that the enterprise has no competitors in its main market becomes statistically insignificant and the coefficient on urban population becomes insignificant when the country controls (rather than country dummies) are included in the analysis,

(27) Data come from the United Nations Statistical Division Commodity Trade Database. Data were not available for Armenia, Georgia, and Uzbekistan.

(28) One related point is that few of the enterprises in the sample produce food or raw materials (ie, most might be expected to be competing with imports from developed countries); only 0.8 percent of enterprises are in mining or quarrying and only 13.5 percent of enterprises are in farming, fishing or forestry (see Table 2). Excluding these enterprises from the sample does not affect the main results. That is, the coefficient on foreign ownership and the coefficients on competition from foreign-owned enterprises and imports remain positive and statistically significant at a 5 percent level or lower. Results are available from author upon request.

(29) The pseudo-[R.sup.2] is 1-([L.sub.1]/[L.sub.o]), where [L.sub.o] is the log likelihood of a model with only a constant and [L.sub.1] is the log-likelihood of the estimated model (Judge et al., 1988).

(30) See footnote 5. Better performing enterprises might both have better access to capital markets and have higher retained earnings. Given the underdeveloped nature of the banking systems and capital markets in these countries, retained earnings are a vital source of resources for investment in the transition economies.

(31) The meta-analysis in Djankov and Murrell (2002) indicates that ownership by foreign enterprises and individuals, ownership by investment funds, ownership by managers, and concentrated individual ownership was more effective than employee-ownership at improving enterprise performance.

(32) One study that looks at the effect of the Internet on firm performance in transition economies, Clarke (2001), finds that export growth is faster for industrial enterprises in transition economies with Internet access than for non-connected firms even after controlling for self-selection bias.

(33) Mauro (1995) shows that corruption has a large and statistically significant effect on economic growth. In addition, several recent papers have round that corruption is negatively correlated with foreign direct investment. Wei (1999), who uses FDI data from 45 developing and developed countries from 12 OECD countries, finds that corruption in the host country has a statistically significant effect on foreign direct investment. The effect is quite large--a one-point increase in corruption (on a five-point scale) would decrease foreign direct investment by about 16 percent. Similarly, Gastanga et al. (1998) also rind that corruption reduces foreign direct investment in a sample of 45 less-developed countries.

REFERENCES

Aitken, B and Harrison, AE. 1999: Do domestic firms benefit from direct foreign investment? Evidence from Venezuela. American Economic Review 89(3): 605-618.

Barba Navaretti, G and Tarr, DG. 2000: International knowledge flows and economic performance: A review of the evidence. World Bank Economic Review 13(1): 1-15.

Blomstrom, M. 1986: Foreign investment and productive efficiency: The case of Mexico. Journal of Industrial Economics 35(1): 97-110.

Blomstrom, M. and Persson, H. 1983: Foreign investment and spillover efficiency in an underdeveloped economy: evidence from the Mexican manufacturing industry. World Development 11(6): 493-510.

Blomstrom, M and Kokko, A. (1996). The impact of foreign investment on host countries: A review of the empirical evidence. Policy Research Working Paper 1745, World Bank: Washington DC.

Blomstrom, M and Sjoholm, F. 1999: Technology transfer and spillovers: does local participation with multinationals matter? European Economic Review 43(4-6): 915-923.

Blomstrom, M and Wolff, EN. 1994: Multinational corporations and productivity convergence in Mexico. In: Baumol, W, Nelson, R and Wolff, E (eds). Convergence of Productivity: Cross-National Studies and Historical Evidence. Oxford University Press: Oxford, UK.

Borensztein, E, De Gregorio, J and Lee, JW (1998). How does foreign direct investment affect economic growth? Journal of International Economics 45(1): 115-135.

Brynjolfsson, E and Hitt, LM. 2000: Beyond computation: information technology, organizational transformation and business performance. Journal of Economic Perspectives 14(4): 23-48.

Caves, RE. 1974: Multinational firms, competition and productivity in host-country markets. Economica 38(149): 1-27.

Clarke, GRG. 2001: Does internet connectivity affect export performance? Evidence from the transition economies. Mimeo, World Bank: Washington, DC.

Daly, J and Miller, RR. 1998. Corporations' use of the internet in developing countries. IFC Discussion paper #35, International Finance Corporation, World Bank: Washington, DC.

Dasgupta, S, Lall, S and Wheeler, D. 2000: Policy reform, economic growth and the digital divide: An econometric analysis. Policy Research Working Paper #2567, World Bank: Washington, DC.

Djankov, S and Hoekman, B. 2000: Foreign investment and productivity growth in Czech enterprises. World Bank Economic Review 14(1): 49-64.

Djankov, S and Murrell, P. 2002: Enterprise restructuring in transition: a quantitative survey. Journal of Economic Literature 40(3): 739-792.

European Bank for Reconstruction and Development. 1999: Transition Report 1999: Ten Years of Transition. European Bank for Reconstruction and Development, London, UK.

European Bank for Reconstruction and Development. 2000: Transition Report 2000: Employment, Skills, and Transition. European Bank for Reconstruction and Development, London, UK.

Gastanga, VM, Nugent, JB and Pashamova, B. 1998: Host country reforms and FDI inflows: How much difference do they make? World Development 26(7): 1299-1314.

Gershenberg, I. 1987: The training and spread of managerial know-how, a comparative analysis of multinational and other firms in Kenya. World Development 15(7): 931-939.

Globerman, S. 1979: Foreign direct investment and 'spillover' efficiency benefits in Canadian manufacturing industries. Canadian Journal of Economics 12(1): 42-56.

Gordon, RJ. 2000: Does the 'new economy' measure up to the great inventions of the past. Journal of Economic Perspectives 14(4): 49-74.

Haddad, M and Harrison, A. 1993: Are there positive spillovers from direct foreign investment? Journal of Development Economics 42(1): 51-74.

Hellman, J, Jones, G, Kaufmann, D and Schankerman, M. 2000: Measuring governance and state capture: The role of bureaucrats and firms in shaping the business environment. European Bank for Reconstruction and Development Working Paper #51, European Bank for Reconstruction and Development: London, UK.

International Monetary Fund. 2000. Azerbaijan Republic: Recent economic developments and selected issues. International Monetary Fund: Washington, DC.

Judge, GG, Griffiths, WE, Hill, C, Lutkepohl, H and Lee, T-C. 1988: The Theory and Practice of Econometrics, 2nd Edition. John Wiley and Sons: New York, NY.

Mauro, P. 1995: Corruption and growth. Quarterly Journal of Economics 110(3): 681-712.

Oliner, SD and Sichel, DE. 2000: The resurgence of growth in the late 1990s: is information technology the story? Journal of Economic Perspectives 14(4): 3-22.

Pack, H. 1997: The role of exports in Asian development In: Birdsall, N and Jaspersen, F (eds). Pathways to Growth: Comparing East Asia and Latin America. Johns Hopkins University Press: Baltimore, MD. pp. 227-251.

Sachs, J. 2000: A new map of the world. The Economist, June 22, 2000.

Saggi, K. 2000: Trade, foreign direct investment and international technology transfer: A survey. Policy Research Working Paper #2349, World Bank: Washington, DC.

Wei, S-J. 1999: How taxing is corruption on international investors? Review of Economics and Statistics 81(4): 1-12.

World Bank. 2003: Peru: Microeconomic constraints to growth study. The World Bank: Washington, DC.

GEORGE R.G. CLARKE

Development Research Group, MSN MC3-307, The World Bank, 1818 H Street, NW, Washington, DC 20433, USA. E-mail,: gclarke@worldbank.org