China’s Technology Parks and Regional Economic Growth* Albert Guangzhou HU Assistant Professor Department of Economics National University of Singapore 10 Kent Ridge Crescent Singapore 119260 Tel: (65)68743957/Fax: (65)67752646 Email: ecshua@nus.edu.sg Prepared for The Fourth International Conference on the Chinese Economy: The Efficiency of China’s Economic Policy CERDI, Clermont-Ferrand, France October 23 – 24, 2003 Abstract China’s technology parks have spurred fast economic growth in the cities that host them. I examine whether this is merely in response to the policy incentives or there have been external economies from the concentration of high-technology firms as expected by policy makers. Using data on China’s 52 national technology parks from 1992 to 2000 and data on the metropolises that host them, I find results that are consistent with the neoclassical growth explanation - labor productivity across technology parks is converging and there is no evidence of geographical external economies. But host city’s foreign direct investment exhibits robust productivity spillover to technology parks. JEL classification: O3, O4 Keywords: Technology parks, localization and agglomeration, externalities, and China * I am grateful to NUS Academic Research Fund (R-122-000-058-112) for financial support and to Goh Minghe for capable research assistance. I. Introduction The clustering of high technology firms and the synergies it creates among various institutions in the cluster is a defining characteristic of Silicon Valley and Route 128 in the U.S. Observers have noted that such concentration of innovative firms and individuals in a region helps to create an entrepreneurial and innovative culture that breeds a continuous stream of innovations helped by information sharing and knowledge spillover, both across firms and between firms and academic institutions, often via informal channels (Saxenian, 1994). Although neither Silicon Valley nor Route 128 came into existence by design, it does not prevent countries around the world, developed or underdeveloped, from emulating the American success stories. Usually bounded regions or areas are designated as technology parks, science parks, or something similar to that effect. Policy incentives are usually offered to encourage new firms to be set up in the park. Better known examples of such parks include Cambridge, U.K., Sophia-Antipolis of France, Tsukuba in Japan, and Taiwan’s Tsinchu Technology Park1. The idea that geographical concentration generates externalities through localization and agglomeration dates back to Marshall (1920). Three forces drive the formation and growth of regional clustering of industries according to Marshall: information exchange or knowledge spillovers, the advantage of a thick labor market, and the backward and forward linkages that a large local market can foster. Voluminous studies have applied the Marshallian external economies to understand and explain industry concentration and the growth of cities2 (Henderson, 1986; Glaeser et al 1992; Krugman 1993; Black and Henderson, 1999). These forces have been responsible even more directly for the formation and growth of Silicon Valley and Route 128. Beginning in the early 1990s, technology parks have been established in 53 major Chinese metropolises under the Torch program, which is an important science and technology (S&T) initiative by the Chinese government. These parks resemble their counterparts from other parts of the world in aiming to build a concentration of hightechnology companies by granting policy incentives such as tax holidays. The main 1 2 Castells and Hall (1994) provide a detailed description of major technology parks around the world. For a comprehensive survey of the field see Fujita, Krugman, and Venables (2001). 1 objective is to expedite technology diffusion and create synergies among the academic and financial institutions and corporations within or near the park. The major metropolises, where most of these technology parks are located, seem to have the necessary ingredients to create localization and agglomeration externalities with their strong industrial base, advanced infrastructure, and rich technological resources3. However, the technology park initiative raises several concerns. First, can such policy lead to a concentration of high-technology firms? Second, it is unclear whether a concentration of firms driven by policy incentives rather than natural technological synergies can capture and generate benefits from external economies of localization and agglomeration, not to mention the innovative drive and entrepreneurial spirits that define Silicon Valley. Finally, whether these parks will be sustainable and fulfill their expected goals depends in part on whether there are dynamic externalities from localization and agglomeration. Such externalities keep up the returns to investment and generate sustained growth. The Chinese technology park development experience provides a valuable window to examine these issues. The large number of technology parks established, the heterogeneity of the host cities, and the time that has lapsed since the parks were first established, provide an opportunity to identify external economies of regional concentration, if any, and other regional features contributing to the success of technology parks and therefore to evaluate the effectiveness of the policy of artificially forging regional concentration of technology-intensive firms. To the extent that the policy does lead to a concentration of high-technology firms and the concentration generates spillovers large enough to warrant the subsidies, the government has a proactive role to play in contributing to regional development and overall economic growth. Using data on 52 Chinese national technology parks from 1992 to 2000 and matching data on the metropolises that host these technology parks, we investigate three questions. First, are there static externalities from concentrating high-technology firms in the parks and locating them in a major metropolis? Second, do the external economies of localization and agglomeration accelerate the growth of the technology parks? In other words, are there dynamic externalities? Lastly, we want to examine two specific sources 3 DeBresson and Hu (1996) report some evidence of the clustering of innovation activities in China in a general context. 2 of spillover. Have the technology parks been able to reap any external benefits from the education and research resources the metropolises have to offer as would be expected for a technology park? Another source of technology spillover is foreign direct investment (FDI). We ask whether the technology parks have absorbed any technology spillovers that FDI flowing into the host cities may have generated through the potential linkages that foreign invested firms may foster with local firms. The next section describes the development of technology parks in China. Section 3 discusses methodological issues and the estimation results. The final section concludes with some remarks on the policy implications. II. The formation and growth of China’s technology parks 2.1. China’s Science and technology policy and technology parks The “Torch Program” is one of many science and technology policy initiatives the Chinese government has adopted. The main objective is to “mobilize the technological capabilities and resources of research institutes, higher education institutions, and large and medium-size enterprises to develop high and new-technology products, establish technology-oriented enterprises, and pave the way for the commercialization of innovations that will come out of major national science and technology programs.” (Yuan et al, 1992, p. 197). A critical ingredient of the Torch program is the establishment of science and technology parks, where most of the new and high- technology commercialization efforts are expected to take place. The first technology park approved by the central government was established in 1988 in Beijing. It centers on Zhongguancong, which plays host to some of China’s most prestigious universities including Beijing and Tsinghua universities and many of the research institutes of the Chinese Academy of Sciences. In 1991, another 26 technology parks were approved by the State Council, followed by yet another 25 in the following year. The establishment of the Yanglin Agricultural Technology Park (Shannxi Province) in 1997 brings the total number of national technology parks to 53. In the meantime, a number of technology parks have also been established by various levels of local government. We will focus on the national technology parks. 3 2.2 Host cities and technology parks: a first look Most of the parks are located in coastal provinces (35) with only five located in the western part of China. Many parks (22) are hosted by the provincial capital cities (22) or central-government supervised municipalities (4), where educational and innovative resources tend to concentrate. The rest of the host cities except Yanglin are also major metropolises with a strong industrial base that the technology parks may leverage on. Although it is not an explicit objective of the Chinese government’s technology park initiative, regional development is obviously a driving force behind the waves of technology park building. Almost all Eastern Chinese provinces have at least one technology park. While major metropolises such as Beijing and Shanghai clearly have substantial resources their technology parks can tap into, it is not clear if some of the other cities are capable of building and sustaining a technology park4. The technology parks offer various policy incentives to encourage investment and new firm formation in the parks5. For example, new firms are exempt from corporate income tax for two years. License is waived for the import of materials and parts used in producing goods for export. A firm’s revenue from technology transfer is only taxable beyond the first 300 thousand yuan. Intangible assets such as intellectual property can be factored into a company’s registered capital. Although these policy incentives create a more profitable business environment than firms in the rest of the city can enjoy, some Chinese scholars argue that they are less favorable than those offered to similar firms in a special economic zone (Mou, 1999). [Insert Table 1 here] Table 1 summarizes key information on the technology parks6. Three variables demonstrate the rapid growth of the technology parks: labor productivity, firm formation, 4 Despite the different endowment of host cities, most technology parks adopt the strategy of developing the electronics and information technologies (MOST, 2001). This sector accounts for 71 percent of total sales in the Beijing Technology Park and 61 percent in Shenyang. 5 Firms are required to have their technology and products certified by a government agency to be high and new- technology before they can be set up in a technology park (MOST, 2001). One criterion is that such firms have to spend at least three percent of sales on research and development. Such high-technology status test is to be repeated every year, failing which would disqualify a firm from enjoying the various policy incentives provided by the government. 6 The data on technology parks are from MOST (2001). Various issues of the China Urban Statistical Yearbook (NSB, various issues) provide the data on matching cities. 4 and technology parks’ share of the host city’s industrial output. Based on national averages, labor productivity in the technology parks had grown by three times from 1992 to 2000. New firm formation also jumped dramatically over the same period. Technology parks have also become an important component of the host city’s industrial activities. The average share of the host city’s industrial output contributed by technology parks rose from 2 percent to 31 percent in eight years. Table 1 also presents information on the largest five and the fast growing five technology parks. Beijing tops the size list with Shanghai closely behind. One notable feature of the Beijing park is the extremely large number of firms, indicative of a much more entrepreneurial environment than elsewhere. The largest parks also tend to have a high concentration of tertiary institutions and FDI. Two of the fastest growing technology parks, Mianyang and Zibo, are located in relatively small cities. Two of the fastest growing technology parks, Jilin and Nanjing, have made it to the largest list. 2.3 The growth of host cities and technology parks There are at least two reasons for which we would expect technology parks to grow faster than their host cities. First, the policy incentives discussed above increase the return to any investment made in the technology park than in the rest of the city. Second, if the technology parks are to leverage on and capture the external economies the host city has to offer and could not internalize, they should be growing even faster. Figure 1 plots the growth rate of industrial output of technology parks and their host cities. All parks (excluding four parks which did not report data for 1992) have out grown their host cities. Even the slowest growing Ulumuqi Technology Park has been growing at an annual rate of 12.9 percent, 1.2 percent faster than the growth of the city’s industrial output. The growth of technology parks is not independent of that of the host cities. The positive sloped trend line implies that technology parks in high-growth cities also tend to grow faster than those in low-growth cities. The positive relationship is also confirmed by regressing the growth rate of technology parks against that of host cities, which yields a statistically significant coefficient of 0.93, providing ground to suspect some sort of linkages between technology parks and their host cities. [Insert Figure 1 here] 5 The two reasons for higher technology park growth represent two alternative theories to explain regional growth. The first one is a neoclassical growth theory explanation, which would predict that eventually as diminishing returns set in, the marginal products of capital will be equalized and output per worker will converge across technology parks. The second theory is that the Marshallian external economies result in increasing returns so that the gap in output per worker between technology parks is likely to widen over time. The two forces are opposing each other. To further explore the driving forces behind of the growth of technology parks, in Figure 2, we plot the growth rate of labor productivity (industrial output per worker) against the initial level (1992) of it for technology parks. There is a clear trend of convergence – technology parks with a lower level of productivity tend to have higher labor productivity growth. Using 1993 as the starting year or 1999 as the ending year obtained the same trend. This seems to be consistent with the neoclassical growth mechanism - as capital accumulation slows down, labor productivity growth decelerates. The policy implication seems to be that the policy incentives are the main driving force behind the growth of technology parks and dominate external economies arising from agglomeration and localization, if they exist at all. A crude test of the existence of the external economies from locating in the vicinity of a large urban area is to relate the rate of labor productivity growth in the technology park to its host city’s initial size. In other words, we are asking whether the supposed dynamic externalities inhabiting large cities lead to higher growth rate of technology parks residing in them. Figure 3 is a graphical representation of such a test. In contrast with Figure 2, labor productivity of technology parks in larger cities does not necessarily grow slower than that in smaller cities. If anything, there seems to be a positive albeit insignificant relationship between the rate of labor productivity growth in a technology park and the initial size of the host city. [Insert Figures 2 and 3 here] 6 III. Localization, agglomeration, and technology parks 3.1 The empirical methodology To identify the scale economies that may arise from establishing technology parks and estimate its magnitude, we estimate a dual factor usage equation similar to that in Henderson (1986). We assume that the production technology of a firm in a technology park can be characterized by the following production function: Y = A(G ) f ( K , L, M ) (1) The function f(•) exhibits constant returns to scale (CRS). Economies of scale (G) that are due to localization and agglomeration enter the production function shifter A(•) in a Hicks neutral way. Given these assumptions and if the firms within the park are relatively homogeneous, we can aggregate the firm production function to a technology park production function. Having specified its production technology, we can write down the firm’s unit cost function, which is the dual representation of its production technology: c= c( PK , PL , PM ) A(G ) (2) By Shephard’s lemma and a first-order Taylor series expansion about P=1, we can transform (2) into an estimable equation: log(Y / L) = C 0 + log A(G ) + α K PK + α L PL + α M PM (3) Of the factor prices, we only have data on the price of labor, the average wage rate. For the other two prices, we will use proxies to control for inter-region variation in them. Our main interest focuses on the production technology shifter, A(G), which we assume to depend on knowledge spillover, both within the technology park and between the technology park and the host city; access to specialized labor forces, particularly science and technology personnel of the host city; and backward and forward linkages created between firms in the technology park and firms in the technology park and the host city. Therefore A(G) is determined by both pure knowledge spillover and pecuniary spillovers. Since foreign direct investment is expected to generate productivity spillover to local firms, the amount of FDI a host city receives also enters G. We therefore specify A(G) as follows: A(G ) = A(1 / L, N ,UNIV , SHR3, FDI ) 7 (4) where L, the employment of technology parks, is a proxy for forces of localization or concentration of high-technology firms. Holding the overall size of the city constant, larger technology parks are more likely to breed knowledge spillovers between firms within the park. The total host city population, N, indicates the extent to which technology parks may have access to a large market and foster linkages with related industries in the city, i.e., the external economies from urbanization. Given the expected connection between technology parks and the local universities, we use UNIV, the local university enrollment, as a proxy for the pool of local academic resources and technological capabilities that technology parks may tap into, through new firm formation and technology licensing. In addition, the share of local employment in services industries (SHR3) is used to capture how good the city is in providing services that are complementary to technology park business activities. Finally, to the extent that FDI generates spillover to local firms, we include FDI the host city receives as another determinant of a firm’s overall level of technology. In estimating (3), lacking of data on the prices of capital and intermediary materials forces us to make assumptions and use proxies. We assume that the price of capital is relatively uniform across different cities. Given that most of the cities are large cities and a banking system dominated by state-owned banks, this may not be an unrealistic assumption. We use the share of the city’s employment in primary industries (SHR1) as a proxy for a technology park’s access to intermediary materials and therefore their prices. 3.2 Static external economies We first treat the sample as a large cross section and estimate equation (3) to examine the existence of static external economies. The results are presented in Table 2. Total city population is in log, whereas technology park employment enters as one over the number of employees. This is done in part to reduce the multicolinearity between N and L. If there are external economies of concentration of high-technology parks, labor productivity should be higher the larger the technology park’s overall employment, and therefore a negative coefficient of the variable, 1/L. Likewise, if a technology park benefits from situating in a large metropolis, with its large product and labor market, 8 academic and technological resources, and potential linkages that can be forged between the park and the city, labor productivity should be higher in technology parks in larger metropolises than in smaller ones. This suggests a positive sign of log (N). [Insert Table 2 here] Table 2 clearly shows evidence for neither sort of external economies. The effect of specialization is largely nonexistent. A larger technology park holding the size of the city constant does not have any impact on labor productivity in the park. And opposite to the expected sign, the effect of large metropolises is robustly negative. The estimate in column (1) implies that a one percent increase in city population reduces labor productivity in the technology park by 60 yuan. An issue that may confound the interpretation of the OLS coefficients is that both labor productivity in the technology park and the size of the city and the technology park can potentially be subject to certain common characteristics that are specific to the city. If that is the case, the external economies estimates will be biased. We re-estimate column (1) using a fixed-effect estimator and report the results in column (2). Since we only have one year’s data for the two employment share variables, we have to drop them in the fixed-effect estimation. Despite controlling for the region specific effects, there is still no evidence of localization and urbanization externalities. The coefficient of city population turns positive but statistically insignificant. Before concluding that there are no external economies in large cities for technology parks, we examine whether some of the amenities large cities have to offer exert any external benefits on technology parks despite the negative overall result. In column (3), we include UNIV to examine whether there is any spillover from host city academic and technological resources to high-technology parks. The coefficient estimate reveals that a one-percent increase in local college student population boosts labor productivity in technology parks by 118 yuan. In the meantime, the negative externality of urbanization as represented by the coefficient of N, becomes even more robust. The last specific source of spillover that we want to examine is FDI flowing into the host cities, some of which enters the technology parks seeking policy incentives. The impact of FDI carries both statistical and economic significance. For a one-percent increase of FDI flowing into a host city, labor productivity in the city’s technology park 9 is increased by 140 yuan based on the estimate in column (4). The negative coefficient of city size becomes even larger with the inclusion of the FDI variable. A potential problem with the specification in column (4) is the possibility that host city FDI is subject to the same productivity shock as is labor productivity in the technology park. We adopt the instrumental variables (IV) approach to tackle the simultaneity bias and use the number of doctors in a host city in the year 1991 – one year before our regression sample starts – as the instrument for host city FDI. The rationale is that the 1991 number of doctors should be correlated with the city’s infrastructure and therefore the amount of FDI it receives, but should be uncorrelated with any productivity shock that materialized during the sample period. The IV estimates in column (5) are quite similar to those of OLS. If anything, the impact of FDI becomes even stronger. We use the Hausman test to gauge the endogeneity of FDI in the equation estimated here. The test yielded a Chi square statistic of -0.01, which is far from allowing us to reject the hypothesis that FDI is exogenous in our equation. Beijing’s technology park is the oldest and largest park. It in many ways resembles a “real” high- technology park with all the right elements. Given its uniqueness, we exclude Beijing from the sample, replicate column (4) and report the results in column (6). The difference between columns (4) and (6) is minimal. The city wage rate is positive and statistically significant through out the six cases. The expected positive sign indicates that firms tend to substitute labor for other inputs as wage rate increases. The robust and positive coefficient of wage rate seems to be consistent with efficient input usage and a dynamic labor market at least in the technology parks. The city’s employment share of primary industries, SHR1, which is a proxy for the differences in the prices of intermediate inputs, also turns out to be positive and significant in all cases. Since we would expect intermediate inputs to be cheaper when they are in relative abundant supply (SHR1 higher), the positive sign is also consistent with profit maximizing behavior. The services share, SHR3, on the other hand is insignificant. Table 2 also reports estimates of the year dummies. The results in column (1) show a clear pattern of declining marginal profit of labor (labor productivity minus the wage rate) over time. Controlling for university enrollment leads to an even larger rate of 10 decline. However, the declining trend completely disappears after the FDI variable is included. The declining trend may well be driven by the convergence of labor productivity across technology parks. Since our data also indicates that FDI also shows a pattern of convergence, it is not surprising that the declining trend disappears when FDI is included. This is also affirmed in column (2) where region specific factors are controlled for and the declining trend turns into an increasing trend. 3.3 Dynamic externalities To examine whether there are dynamic externalities from localization and urbanization for technology parks, we estimate an equation where the dependent variable is the growth rate of labor productivity between 1992 and 2000. The independent variables include the 1992 labor productivity and technology shifter, i.e., Log(Y/L)1992 and Log(A(G))1992, and the growth rate of wage over the period. In other words, we are investigating whether localization and agglomeration have any long term impact on labor productivity. The results are reported in Table 3. [Insert Table 3 here] The convergence of labor productivity across technology parks that we observed in Figure 2 is again reflected in the negative and robust coefficient of Log(Y/L)1992 – the higher the starting level of labor productivity, the lower the rate of growth. The localization effect is now of the right sign but statistically insignificant. There is no evidence of either technology parks in larger cities growing faster than those in smaller cities. Turning to the specific sources of externalities, we find that university enrollment does not generate any long term impact on productivity growth despite the evidence of static externalities in Table 2. This seems to suggest that access to rich local academic and technological resources gives a competitive edge to a technology park, but such advantage is unlikely to be sustainable. On the other hand, FDI seems to be able to generate both static and dynamic externalities. A one-percent increase in the amount of FDI a host city receives increases the growth rate of labor productivity in the city’s technology park by 1.5 percent over the eight-year period. Another interesting impact of FDI is on the speed of convergence. Although ours is not a formal test of convergence (Barro and Sala-i-Martin, 1992), 11 controlling for FDI raises the coefficient of Log(Y/L)1992, which implies a higher rate of convergence. A possible explanation is that the dynamic externalities generate by FDI retards the convergence process. Finally, the coefficient of the log of wage rate remains positive and significant. Compared with the estimates in Table 2, labor demand seems to be more elastic in the dynamic setting, perhaps in part due to the longer time firms have to adjust to productivity shocks and wage increases. IV. Concluding remarks and policy implications We set out to evaluate the rationale and effectiveness of an important science and technology policy of the Chinese government, i.e., the establishment of high and newtechnology parks in 53 Chinese metropolises. Two main research questions were raised in the beginning. First, are there external economies from localization and urbanization that can justify the policy incentives provided to the technology parks? Second, what are the determinants of the growth of technology parks? Behind our empirical exercise are two theories of regional economic growth. In the neoclassical growth framework, diminishing returns lead to convergence of income across regions. Policy intervention cannot change long run economic growth. Alternatively, localization and agglomeration can potentially generate external economies that have both static and dynamic impact on long run growth. Using data on the 52 technology parks from 1992 to 2000 and the matching host city information, we estimate an equation for labor productivity that is derived from the dual factor share equation. We set out to search for external economies of localization and agglomeration to China’s technology parks both statically and dynamically. Although there was evidence for positive spillover from the host city’s tertiary education institutions, we did not any evidence that would indicate firms in technology parks benefit from concentrating in a certain locus or being close to a large metropolis. If anything, there is some static inefficiency from locating in large cities. We found that there was considerable spillover to firms in the technology parks from the FDI that the host cities received. A one-percent increase in host city FDI raises 12 labor productivity in the technology parks by 157 yuan. The spillover also produces dynamic effect. A host city that received one-percent more FDI in 1992 saw labor productivity in its technology park increase by 1.5 percent more from 1992 to 2000. Our analysis indicated a robust trend of convergence of labor productivity across technology parks. Technology parks with a lower starting point saw their productivity grow much faster. This is consistent with the prediction of neoclassical theory of economic growth. The Chinese government’s technology park initiative has spurred faster growth in the cities that host them. This is largely achieved through policy incentives that raise the private return to investment. The technology parks have not been able to generate external economies from concentration that will ensure sustainable growth, as they were expected to. However, the technology parks have been successful in leveraging on linkages forged with FDI. One lesson from this study is that the technology park initiative and policies to attract FDI are strongly complementary policy instruments. The lack of evidence of geographical external economies in the growth and development of China’s technology parks certainly deserves more scrutiny with better data on the regional economies and the technology parks. The dynamic externalities may also take time to materialize. 13 Reference Barro, Robert J.and Sala-i-Martin, Xavier. “Convergence.” Journal of Political Economy. 100(2): 223-251. April 1992. Black, Duncan and Henderson, J. Vernon. “A Theory of Urban Growth.” Journal of Political Economy. 107(2): 252-284. 1999 Castells, Manuel and Hall, Peter. Technopoles of the world : the making of twenty-firstcentury industrial complexes. London: Routledge. 1994 DeBresson, Chris, and Hu, Xiaoping. “The Localisation of Clusters of Innovative Activity in Italy, France and China.” In Innovation, Patents and Technological Strategies. Paris: OECD. 1996. Fujita, Masahisa; Krugman, Paul R., and Venables, Anthony J. The Spatial Economy: Cities, Regions and International Trade. Cambridge, Mass.: MIT Press. 2001 Glaeser, Edward L; Kallal, Hedi D.; Scheinkman, Jose A.; and Shleifer, Andre. “Growth in Cities.” Journal of Political Economy. 100: 1126-52. December 1992. Henderson, J. Vernon. “Efficiency of Resource Usage and City Size.” Journal of Urban Economics. 19: 47-70, January 1986. Krugman, Paul R. “On the Number and Location of Cities.” European Economic Review. 37: 293-98. April 1993 Marshall, Alfred. Principles of Economics: An Introductory Volume. 8th edition. London: Macmillan, 1920. MOST (Ministry of Science and Technology of China). Annual Report of Development of High-Tech Industry Development Zone: 1991-2000 (. Beijing: Science and Technology Documents Press (Kexue Jishu Wenxian Chubanshe). 2001 MOU, Baozhu, eds. The Theory and Practice of China’s New and High- Technology Industry Development Zone (Zhongguo Gao Xin Jishu Chanye Kaifaqu Lilun yu Shijian). Beijing: China Price Press (Zhongguo Wujia Chubanshe). 1999 National Bureau of Statistics, China Urban Statistical Yearbook, various issues. Saxenian, AnnaLee. Regional Advantage: Culture and Competition in Silicon Valley and Route 128. Cambridge, Mass.: Harvard University Press. 1994 Yuan, G.Z. and Gao, J.S. Programs and Plans for the Development of Science and Technology of China, Beijing, China: National Defense Industry Press. 1992. 14 Figure 1. Growth rates of industrial output of technology parks and host cities: 1992-2000 0.3 Growth rate of host cities 0.25 0.2 0.15 0.1 0.05 0 0 0.1 0.2 0.3 0.4 0.5 0.6 Growth rate of technology parks 0.7 0.8 0.9 1 Figure 2. Growth rate of labor productivity in technology parks: 1992-2000 0.5 0.4 Growth rate of (Y/L) 0.3 0.2 0.1 0 2.5 3 3.5 4 4.5 5 5.5 6 6.5 -0.1 -0.2 Log of 1992 (Y/L) Figure 3. Growth rate of labor productivity in technology park and city size Growth rate of (Y/L) in technology parks: 1992-2000 0.5 0.4 0.3 0.2 0.1 0 14 15 16 17 -0.1 -0.2 Log of 1992 Y of host cities 18 19 20 Table 1. Basic information on technology parks Labor productivity Number of firms University enrollment 1992 2000 1992 2000 1992 2000 National 82 326 232 784 31733 70525 Average Largest five in 2000 Beijing 41 312 1510 6181 143784 221580 Shanghai 93 745 73 438 119130 224747 Jilin 28 607 27 393 12358 22000 Shenzhen 147 717 40 122 3730 14123 Nanjing 94 861 32 144 80487 216875 Fastest growing five Jilin 28 607 27 393 12358 22000 Mianyang 33 449 5 72 6412 24590 Zibo 21 251 6 110 3742 17039 Changchun 20 218 117 578 55635 127754 Nanjing 94 861 32 144 80487 216875 Note: Labor productivity is in thousand yuan and FDI is in thousand US dollars. FDI 1992 21590 Share of city's GVIO 2000 1992 2000 48090 0.02 0.31 52712 176892 1690 71539 13209 237810 302728 5114 196100 57271 0.02 0.02 0.003 0.01 0.004 0.38 0.12 0.52 0.18 0.36 1690 2552 9873 2581 13209 5114 3896 8666 12438 57271 0.003 0.01 0.01 0.02 0.004 0.52 0.75 0.14 0.32 0.36 Table 2. Technology parks and external economies: the static perspective (1) (2) (3) (4) (5) OLS FE OLS OLS IV Constant -8.931* -0.287 -10.538* -6.584* -6.130* (0.868) (2.387) (0.976) (1.104) (1.343) 1/L 29.656 57.837 56.272 66.143*** 67.420*** (38.989) (40.830) (38.061) (37.437) (37.246) log (N) -0.060** 0.095 -0.209* -0.247* -0.251* (0.028) (0.213) (0.045) (0.045) (0.045) log (wage) 1.607* 0.488*** 1.778* 1.221* 1.156* (0.106) (0.266) (0.116) (0.137) (0.175) SHR1 0.016* 0.017* 0.018* 0.018* (0.004) (0.004) (0.004) (0.004) SHR3 0.005 -0.0003 -0.002 -0.002 (0.004) (0.004) (0.004) (0.004) Log (UNIV) 0.118* 0.070** 0.065*** (0.032) (0.033) (0.038) Log (FDI) 0.140* 0.157* (0.022) (0.041) Year 1993 -0.061 0.175*** -0.101 -0.095 -0.095 (0.118) (0.100) (0.117) (0.112) (0.111) Year 1994 -0.312** 0.278*** -0.400* -0.163 -0.134 (0.133) (0.166) (0.135) (0.132) (0.139) Year 1995 -0.306** 0.459** -0.410* -0.111 -0.075 (0.139) (0.205) (0.140) (0.138) (0.149) Year 1996 -0.402* 0.488** -0.539* -0.190 -0.148 (0.155) (0.233) (0.157) (0.159) (0.171) Year 1997 -0.299** 0.671* -0.449* -0.048 -0.001 (0.153) (0.253) (0.158) (0.160) (0.174) Year 1998 -0.425* 0.679** -0.597* -0.128 -0.073 (0.159) (0.284) (0.164) (0.172) (0.191) Year 1999 -0.525* 0.699** -0.733* -0.199 -0.137 (0.174) (0.314) (0.182) (0.199) (0.212) Year 2000 -0.489* 0.868** -0.740* -0.113 -0.040 (0.171) (0.346) (0.185) (0.202) (0.235) Observations 459 459 452 446 446 2 R 0.58 0.434 0.59 0.62 0.62 Note: Dependent variable is Log(Y/L). * 1% significance level ** 5% significance level *** 10% signficance level (6) OLS -6.736* (1.141) 64.140*** (37.154) -0.233* (0.047) 1.235* (0.140) 0.018* (0.004) -0.001 (0.004) 0.066** (0.034) 0.139* (0.023) -0.104 (0.114) -0.171 (0.135) -0.125 (0.140) -0.206 (0.161) -0.067 (0.161) -0.146 (0.174) -0.221 (0.201) -0.132 (0.204) 437 0.62 Table 3. Technology parks and external economies: the dynamic perspective (1) (2) Constant 0.345* 0.363* (0.093) (0.099) log (Y/L)1992 -0.093* -0.093* (0.015) (0.015) 1/L1992 -3.251 -6.099 (3.772) (5.791) log(N)1992 -0.011 -0.005 (0.012) (0.014) Log (wage2000/wage1992) 0.219* 0.239* (0.056) (0.070) Log (UNIV)1992 -0.007 (0.009) Log (FDI)1992 Number of obs 2 R 47 47 0.65 0.65 Note: Dependent variable is log(Y/L)2000 - log(Y/L)1992. * 1% significance level ** 5% significance level *** 10% signficance level (3) 0.391* (0.098) -0.113* (0.016) -2.218 (4.757) -0.018 (0.012) 0.209* (0.066) -0.005 (0.008) 0.015* (0.006) 47 0.65