Infrastructure and regional economic development in rural China

China Economic Review 15 (2004) 203 – 214 Infrastructure and regional economic development in rural China Shenggen FAN a,b,*, Xiaobo ZHANG a a International Food Policy Research Institute, Washington, DC, USA b Chinese Academy of Agricultural Sciences, Beijing, China Abstract Infrastructure affects rural development through many channels, such as improved agricultural productivity, increased rural nonfarm employment, and rural migration into urban sectors. However, the role of infrastructure has not been paid enough attention in the literature due to lack of reliable data on various infrastructure indicators. By using newly available detailed data on rural infrastructure from the Agricultural Census and other official sources, this paper uses a traditional source accounting approach to identify the specific role of rural infrastructure and other public capital in explaining productivity difference among regions, throwing new lights on how to allocate limited public resources for both growth and regional equity purposes. D 2004 Published by Elsevier Inc. JEL classification: H54; O47; O53; R11 Keywords: Chinese agriculture; Census; Infrastructure; Regional development 1. Introduction Rapid growth in Chinese agriculture after the reforms has triggered numerous studies to analyze the sources of the rapid growth. These studies include Fan (1990, 1991), Fan and Pardey (1997), Huang and Rozelle (1996), Lin (1992), McMillan, Whalley, and Zhu (1989), and Zhang and Carter (1997). Most of these studies attempted to analyze the impact of institutional changes and the increased use of inputs on production growth during the reform period from the end of the 1970s to the beginning of the 1990s. Fan and Pardey (1997) and Fan (2000) were among the first to point out that omitted variables, such as research and development (R&D) investment would bias the estimate of the effect of institutional change. To address this concern, they included a research stock * Corresponding author. International Food Policy Research Institute, 2033 K Street, NW, Washington, DC 20006, USA. Tel.: +1-202-862-5677; fax: +1-202-467-4439. E-mail address: [email protected] (S. Fan). 1043-951X/$ - see front matter D 2004 Published by Elsevier Inc. doi:10.1016/j.chieco.2004.03.001 204 S. Fan, X. Zhang / China Economic Review 15 (2004) 203–214 variable in the production function to account for the contribution of R&D investment to rapid production growth, in addition to the increased use of inputs and institutional changes. They found that by ignoring the R&D variable in the production function estimation, the effects of institutional change would be overestimated to a large extent. In addition to R&D investment, government investments in roads, electrification, education, and other public investment in rural areas may have also contributed to the rapid growth in agricultural production. It is highly likely that omitting these variables will bias the estimates of the production function for Chinese agriculture as well. One of the most important features in rural China is the rapid development of rural nonfarm economies since the economic reform in 1978. But very few have analyzed the sources of growth in this sector. The only exception is Fan, Zhang, and Robinson (2003), who decomposed the growth in the nonfarm sector into growth in labor and capital. But they failed to include public capital as an input in their source accounting, partly due to the lack of reliable public capital data. Associated with the rapid economic growth, regional disparity in productivity has also increased for China for the last two decades. The regional difference in productivity is a major determinant of income disparity, an increasing concern by policymakers and many scholars alike. The uneven regional development in nonfarm activities, particularly in the nonfarm sector, has been regarded as one major driving force behind the changes in rural regional inequality (Rozelle, 1994; Zhang & Fan, 2004). However, despite a large body of literature on the sources of growth, few studies have attempted to account for the sources of regional difference in productivity of both the agricultural and nonfarm sectors (one exception is Fan, 1990), and no studies have systematically assessed the roles of public investment in such differences in regional development. The motivation of this study is to include these public investment variables that are newly available from the Census to estimate the production functions for both agricultural and nonagricultural economies in rural China and to decompose the sources of difference in productivity among regions. In particular, the specific role of infrastructure in explaining the difference in productivity among regions will be evaluated. There are two major advantages in using the Agricultural Census data. First, the census reports detailed infrastructure information at the country level, which is more disaggregate than the provincial level data commonly seen in the official statistical yearbooks. Second, the arable land area and labor force data are more accurately measured than the previous official sources (Ash & Edmonds, 1998; Smil, 1999). The paper is organized as follows: Section 2 reviews the regional distribution of public capital in rural China. Section 3 develops a conceptual framework and model for the purpose of our analysis. Section 4 describes the data and Section 5 discusses our estimated results. We conclude the paper, and point out the limitations of the current study and future research directions in the Section 6. 2. Regional dimension of rural infrastructure The Agricultural Census provides a unique opportunity to analyze the regional dimension of rural infrastructure in China. Table 1 presents the selected infrastructure indicators by 205 S. Fan, X. Zhang / China Economic Review 15 (2004) 203–214 Table 1 Regional difference in rural infrastructure (1996) Province Road density 2 National Beijing Tianjin Hebei Shanxi Inner Mongolia Liaoning Jilin Heilongjiang Shanghai Jiangsu Zhejiang Anhui Fujian Jiangxi Shandong Henan Hubei Hunan Guangdong Guangxi Hainan Chongqing Sichuan Guizhou Yunnan Tibet Shaanxi Gansu Qinghai Ningxia Xinjiang Electricity use Rural telephone km/10,000 km km/10,000 labor km/10,000 person kW/person Set/10,000 labor Set/10,000 people 1679 6310 5258 3021 3578 484 2985 2136 1200 17676 6863 3505 4905 4305 3529 6358 4382 4199 4633 3843 2287 28.7 48 27 18 40 64 31 48 54 36 19 15 20 35 29 21 15 31 30 25 24 51 22 25 29 51 339 37 42 72 32 90 18.4 28 17 11 24 42 21 29 35 26 13 10 13 21 19 14 9 20 20 14 15 29 16 17 18 32 199 23 26 46 19 52 260 709 844 252 309 150 375 184 177 1771 453 525 113 383 115 287 195 182 129 625 114 63 159 165 78 268 22 172 190 273 161 159 303 1024 625 222 205 229 502 286 388 2767 604 596 160 735 109 242 111 319 171 1258 97 163 121 88 54 108 65 101 81 247 104 172 283 933 555 207 183 199 487 266 345 2760 573 582 156 594 102 206 106 296 160 1222 91 158 111 74 42 96 54 95 71 244 103 166 2050 3172 2840 344 3210 1300 207 1082 277 When calculating road density, Hainan and Chongqing are included in Guangdong and Sichuan, respectively. Source: calculated from the Agricultural Census. province in 1996 when the census was conducted. First, we compare the newly available Census data with the official data which are published previously in various China Statistical Yearbooks by the State Statistical Bureau (SSB) or other government agencies. For road density, the Census reported 1679 km per 10,000 km2, which is 34% higher than the official data, released from the Ministry of Transportation. Therefore, the data from the Ministry of Transportation may have understated the road density in rural areas. With respect to rural telephone, the Census reported 283 sets per 10,000 rural residents, which is 43% higher than 197 sets reported by SSB Statistical Yearbook. For rural electricity consumption, the Census reported 260 kW per person for 1996, while the SSB Statistical Yearbook reported 200 kW per person in rural China, a 30% difference. 206 S. Fan, X. Zhang / China Economic Review 15 (2004) 203–214 In terms of the illiteracy rate, the Census data reported 14% for the rural population above the age of 7 years. This percentage is comparable to 11% in 1996 for agricultural labor reported by SSB’s Rural Statistical Yearbook, 1997. The higher rate for the general population than agricultural labor may be due to the fact that the general population may have higher illiteracy rate than total labor force. With respect to R&D spending and personnel, the data are not easily comparable. The Census reports such data only for the township level, while the official SSB or Ministry of Science and Technology reports the data above the county level. Nevertheless, the Census data provide unique and valuable information about science and technology at the lower level, which has never been reported before by other official sources. Table 2 Percentage of rural population with different education levels (1996) National Beijing Tianjin Hebei Shanxi Inner Mongolia Liaoning Jilin Heilongjiang Shanghai Jiangsu Zhejiang Anhui Fujian Jiangxi Shandong Henan Hubei Hunan Guangdong Guangxi Hainan Chongqing Sichuan Guizhou Yunnan Tibet Shaanxi Gansu Qinghai Ningxia Xinjiang Illiterate and semi-illiterate Primary school Junior middle school Senior middle school Special secondary school College and above 14.01 6.28 7.15 10.03 8.81 17.26 6.31 8.27 8.67 13.83 12.53 13.85 16.32 7.01 11.67 9.80 13.57 14.63 9.65 7.93 10.42 17.73 11.87 15.67 29.95 28.09 75.71 16.35 35.57 46.04 31.16 15.73 42.15 21.04 37.54 39.11 35.74 39.66 41.29 45.47 43.54 28.39 36.94 44.26 42.52 51.53 48.90 40.65 33.41 40.97 43.76 42.87 47.98 35.60 52.03 48.60 45.47 49.11 22.93 35.21 34.40 31.82 33.37 53.69 38.04 59.08 48.30 43.87 48.41 36.54 47.22 40.75 41.95 49.38 42.49 36.30 37.18 35.83 34.29 43.09 46.26 38.21 39.42 41.39 35.77 38.76 32.44 32.42 22.23 20.25 1.21 40.88 24.12 19.10 29.68 25.24 5.07 10.67 6.21 6.45 6.20 5.66 4.08 4.56 4.77 6.37 7.12 5.07 3.28 4.86 4.51 5.63 6.13 5.49 6.37 6.81 5.00 7.20 3.20 2.88 1.76 2.12 0.09 6.90 5.25 2.77 4.70 3.78 0.57 2.23 0.64 0.44 0.62 0.70 0.79 0.77 0.84 1.44 0.63 0.37 0.55 0.61 0.52 0.66 0.49 0.60 0.65 0.75 0.71 0.59 0.37 0.35 0.50 0.38 0.05 0.50 0.51 0.21 0.86 1.18 0.16 0.70 0.16 0.09 0.22 0.18 0.30 0.19 0.24 0.60 0.29 0.15 0.14 0.16 0.12 0.15 0.14 0.10 0.15 0.24 0.12 0.12 0.10 0.08 0.08 0.06 0.01 0.15 0.15 0.06 0.22 0.39 Source: calculated from the Agricultural Census. 207 S. Fan, X. Zhang / China Economic Review 15 (2004) 203–214 The regional data reveal that the infrastructure development is highly correlated with the economic development level. Road density measured as the length of rural town roads per 10,000 km2 has very large regional variation. If we exclude Beijing, Shanghai, and Tianjin in our analysis, Jiangsu has the highest road density, and Shandong has the second. Not surprisingly, Inner Mongolia, Tibet, Qinghai, and Xinjiang have the lowest road densities among all provinces. However, if we use the length of roads per rural resident, it is the western provinces or regions that per capita length of roads are the highest due to the relatively lower population density. In terms of rural electricity, again, it is the coastal region that has the highest per capita consumption. For example, Guangdong, Jiangsu, and Zhejiang have more than 400 kW Table 3 Science and technology personnel and expenses (1996) National Beijing Tianjin Hebei Shanxi Inner Mongolia Liaoning Jilin Heilongjiang Shanghai Jiangsu Zhejiang Anhui Fujian Jiangxi Shandong Henan Hubei Hunan Guangdong Guangxi Hainan Chongqing Sichuan Guizhou Yunnan Tibet Shaanxi Gansu Qinghai Ningxia Xinjiang Number of S&T personnel (per 10,000 rural labor) Number of S&T personnel (per 10,000 rural residents) S&T spending in Yuan (per rural labor) S&T spending in Yuan (per rural residents) 90.89 200.16 253.67 147.87 81.39 344.24 88.93 126.34 166.80 150.28 81.50 57.80 54.56 92.89 47.77 92.89 123.86 84.61 79.70 80.93 66.82 143.37 37.32 64.45 27.10 35.05 44.86 104.83 104.42 68.40 48.42 345.98 58.41 115.69 155.73 90.82 48.94 224.39 58.88 77.55 107.64 107.68 55.64 39.10 35.04 55.40 30.90 61.33 79.42 54.11 52.38 45.99 42.16 82.08 26.49 44.86 17.12 22.32 26.41 64.08 64.93 43.59 28.99 199.60 0.81 0.90 0.38 0.33 0.47 0.24 1.38 0.43 0.40 5.93 1.12 1.11 0.46 1.62 0.22 1.26 0.25 0.77 0.88 2.76 0.38 1.10 0.39 0.91 0.43 0.61 0.02 0.15 0.07 0.00 0.14 0.24 0.52 0.52 0.23 0.20 0.28 0.16 0.92 0.26 0.26 4.25 0.77 0.75 0.30 0.97 0.14 0.83 0.16 0.49 0.58 1.57 0.24 0.63 0.28 0.64 0.27 0.39 0.01 0.09 0.04 0.00 0.08 0.14 Source: calculated from the Agricultural Census. 208 S. Fan, X. Zhang / China Economic Review 15 (2004) 203–214 per person per year, while in Inner Mongolai, Tibet, Xinjiang, Guizhou, and surprisingly some central provinces, such as Anhui, Jiangxi, and Guangxi, per capita electricity consumption is less then 200 kW in 1996. The difference in rural telephone possession is the largest among all types of rural infrastructure. In Guangdong, Jiangsu, Zhejiang, and Fujian, for every 10,000 residents, there are more than 500 telephone sets. But in Gansu, Tibet, Guizhou, Sichuan, and Guangxi, less than 100 sets are possessed for every 10,000 rural residents. The education data reveals that in the western region, the Census reported much higher illiteracy rate than the official SSB Rural Statistical Yearbook, 1997 (Table 2). For example, in Tibet and Qinghai, the Census recorded 76% and 46%, compared to 61% and 34% reported by Rural Statistical Yearbook, respectively. The gap in the education level between the eastern and western regions may have been higher than previously believed. The Census data on science and technology personnel and spending uncovers a striking phenomenon (Table 3). It is the western region, for example Xinjiang, and Inner Mongolia, that have the highest ratios of science and technology personnel to rural population or labor. But in terms of science and technology spending, the region has the lowest. This implies that the science and technology personnel in less developed areas experience a severe shortage of operation funds compared to their eastern cohorts. In summary, the Census data reveal a higher level of rural infrastructure development than previously thought. But it also uncovers a larger regional difference not only in the development of rural infrastructure, but also in the development of education and science and technology. This may explain why the western region has lagged behind despite rapid economic growth for the nation as a whole. 3. Conceptual framework and model There have been numerous studies on the estimation of production functions for both agricultural and nonfarm sectors. One significant feature in these previous studies is the use of a single-equation approach. There are at least two disadvantages to this approach. First, many production determinants are generated from the same economic process. In other words, these variables are also endogenous variables, and ignoring this characteristic leads to biased estimates of the production functions. Second, certain economic variables affect the rural economy through multiple channels. For example, improved rural infrastructure will not only contribute growth in agricultural production, but also affect nonfarm production. It is very difficult to capture these different effects in a singleequation approach. This study uses a simultaneous equations model to estimate the effects of rural infrastructure on both farm and nonfarm production. AY ¼ f ðLAND; AGLABOR; FERT; MACH; IR; RD; SCHY; ROADS; RTRÞ; ð1Þ NAY ¼ f ðRILABOR; ELEC; SCHY; ROADS; RTRÞ: ð2Þ S. Fan, X. Zhang / China Economic Review 15 (2004) 203–214 209 Eq. (1) models the agricultural production function. The dependent variable is gross agricultural output value (AY). Land (LAND), labor (AGLABOR), fertilizer (FERT), machinery (MACH) are included as conventional inputs. We include the following variables in the equation to capture the impact of technology, infrastructure and education on agricultural production: percentage of irrigated area in total cropped area (IR); number of agricultural researchers and extension staff (RD), road density (ROADS), number of rural telephone sets per thousand rural residents (RTR), and average years of schooling for population over the age of 7 years (SCHY).1 Eq. (2) is a production function for nonagricultural activities in rural areas. The dependent variable is gross value of the township and village enterprises (NAY). Labor input used in the nonfarm sector (RILABOR), infrastructure, and the labor education level are independent variables included in the function.2 The electricity consumption (ELEC) is used to proxy for fixed and current capital used in the nonfarm sector. Following Fan (1991), Fan and Pardey (1997), and Lin (1992), we use the traditional Cobb –Douglas form for both agricultural and nonfarm equations. In this form, the coefficients of independent variables are simply their elasticities with respect to the dependent variable. Regional dummies are also added to capture the impact of other factors that are not included in the equations. To account for the sources of difference in productivity, we choose labor productivity in our analysis. Labor productivity is one of the most important indicators in economic development and is one of the major determinants of rural income. Following Fan (1990) and Hayami and Ruttan (1985), we use the following accounting formula: X DPi X D Xi DY
Y L ¼ bi ai
XiL þ Pi0 i L 0 L 0 ð3Þ We use the average productivity at the national level ( Y/L)0 as our base for comparison, that is, we try to explain the difference in productivity between each region and the national average.3 In Eq. (3), labor productivity difference is explained by the difference in the use of conventional inputs Xi, such as labor, land, fertilizer, and machinery, all measured on a per labor basis, and the difference in rural infrastructure, education, and science and technology capacity, denoted by Pi. If we divide every term on the right-hand side by the productivity difference (on the left-hand side), then the difference in productivity can be explained by the right-hand side variables in terms of percentages. 1 The electricity variable is excluded mainly because it is highly correlated with road and telephone variables. 2 Ideally, the capital variable should also be included in the function. But there is no such data available at the county level. 3 This decomposition implicitly assumes a constant return to scale, that is, Sai = 1. This assumption is not too realistic, as evidenced by Fan (1991) and Zhang and Carter (1997). 210 S. Fan, X. Zhang / China Economic Review 15 (2004) 203–214 4. Data explanations Our analysis is based on the county level. Most of infrastructure, education, and technology variables are available in the Agricultural Census. However, Agricultural Census does not report detailed information on agricultural and nonfarm output. Input uses are also not available. Therefore, in this analysis, we combine the Census data with the data from other SSB sources, such as China Statistical Yearbooks and China’s Rural Statistical Yearbooks. Agricultural output—agricultural output is measured as gross agricultural production value. The data is taken from the SSB official statistical source. Nonfarm output—nonfarm output is measured as gross output value of township and village enterprises. The sources of the data are official SSB and Ministry of Agriculture publications. Agricultural labor—agricultural labor is measured in stock terms as the number of persons engaged in agricultural production at the end of each year. They are taken from the Census. Nonfarm labor—nonfarm labor is measured as number of employees in the township and village enterprises reported by the Agricultural Census. Land—land is total arable land used for agricultural production. The data is taken form the Census. Machinery—machinery input is measured as horsepower of machinery used in agricultural production. Because the Census does not report horsepower of machinery, we use the data from the SSB Statistical Yearbook. Irrigation—irrigation services used in agriculture are proxied by the ratio of irrigated area. Because the published Census data do not report irrigated areas by county, we use the data from official sources of SSB and Ministry of Agriculture. Fertilizer—it is measured as pure nutrients of chemical fertilizer. The data are taken from official sources of SSB and Ministry of Agriculture. Roads—the length of township roads is reported by the Census. We divided the road length by the geographic areas to obtain the road density variable for our analysis. Rural telephone—number of rural telephone sets is available from the Census. We use the number of telephone sets per 10,000 rural residents as our telephone variable. Education—for the education variable, we use the percentage of population with different education levels to calculate the average years of schooling as our education variable, assuming 0 year for a person who is illiterate and semiilliterate, 5 years with primary school education, 8 years with junior high-school education, 12 years with high-school education, 13 years with professional school education, and 16 years with college and above education. The Agricultural Census reports the percentages of population with different education levels who are above the age of 7. Electricity consumption—electricity consumption in the nonfarm and agricultural sectors are reported by various issues of China Rural Statistical Yearbooks. S. Fan, X. Zhang / China Economic Review 15 (2004) 203–214 211 Science and technology—we use the number of science and technology personnel per 10,000 rural labor at the township level to represent the capacity of science and technology. The data are taken from the Census. 5. Results Table 4 presents the estimated results of production functions for agriculture and nonfarm economies. Only 15 provinces or regions reported county level data in recent SSB provincial publications on the Agricultural Census. They are the following: Beijing, Tianjin, Shanxi, Heilongjiang, Shanghai, Jiangsu, Zhejiang, Fujian, Jiangxi, Shandong, Hunan, Sichuan, Tibet, Shaanxi, and Ningxia. Although they cover roughly half of the provinces, the number of observations covers only 45% of the total number of counties. Therefore, the sample we used in our regression may not represent the whole of China. Most coefficients in both agricultural and nonfarm production functions are statistically significant. The coefficients for conventional inputs in the agricultural production function, such as those for labor, land, fertilizer, and machinery, are in the same ranges of other studies (Fan, 1991; Fan & Pardey 1997; Zhang & Carter 1997). The labor and electricity variables (as a proxy for both fixed and current capitals) are also statistically significant in the nonfarm production function. One notable feature is that the coefficients for infrastructure and education variables are more significant in the nonfarm production equation than those in the agricultural production function. The fitness of both equations is exceptionally good with R2 of .865 for the agricultural production function, and .813 for the nonfarm production function, despite the fact that cross-sectional data are used. The road variable in the nonfarm sector is insignificant due to its high correlation with the telephone variable; therefore, we drop it in the final estimation. Table 5 presents the results of accounting. The numbers in parentheses are the difference in labor productivity level between each region and the national average. By Table 4 Estimation of the equation system Labor Land Fertilizer Machinery (or electricity) Research Irrigation Roads Years of schooling Telephone R2 Agricultural output Nonfarm output 0.262 (3.14)* 0.228 (8.47)* 0.150 (4.55)* 0.115 (6.34)* 0.104 (3.42)* 0.260 (9.48)* 0.032 (2.25)* 0.275 (1.81)* 0.056 (6.41)* .865 0.510 (16.84)* 0.480 (15.89)* 0.792 (1.94)* 0.119 (6.51)* .813 Regional dummies are added to capture the provincial fixed effect, but the coefficients are not reported here. Total number of observations is 1104. * Statistically significant at the 5% level. 212 S. Fan, X. Zhang / China Economic Review 15 (2004) 203–214 Table 5 Accounting for the sources of labor productivity difference among regions Agriculture Eastern Central Nonfarm Western Eastern Total rural Central Productivity 100.00 100.00 100.00 100.00 100.00 (47.06) (2.35) (  35.29) (38.08) (17.22) Land  0.96 230.35 15.36 Fertilizer 7.45 46.27 10.63 Machinery 0.72 121.43 9.87  11.46 3.62 Irrigation  3.44 103.20 7.45 S&T  0.08 4.58 0.16 Roads  2.16  30.85  4.25 0.00 0.00 Telephone 20.50  210.14 10.66 53.83  61.02 Education 1.57 21.04 6.35 5.68 8.41 Residual 76.39  185.88 43.78 51.95 148.99 Western Eastern Central Western 100.00 100.00 100.00 100.00 (  47.35) (41.32) (10.82) (  40.08)  0.35 99.08 9.26 2.69 19.90 6.41  5.92  7.07 54.29 3.60  1.24 44.39 4.50  0.03 1.97 0.10  0.00  0.78  13.27  2.56 16.88 41.82  125.16 13.13 13.84 4.20 13.84 9.32 75.20 60.76 4.95 56.25 assuming this difference as 100%, we can explain the productivity difference in terms of the percentages by various factors shown in the rest of the rows in the table.4 The sources of difference in agricultural labor productivity vary sharply among regions. The higher labor productivity in the eastern region is primarily explained by higher fertilizer use, better infrastructure and the residual, which accounts for other missing variables. This residual is particularly large, implying that other factors, rather than those included in the equation, may have played an even bigger role in explaining its higher productivity. For the central region, higher productivity is mainly explained by more use of land per labor together with more fertilizer, machinery, and irrigation use. In the western region, the lower productivity is due to lower land use per labor (and therefore lower fertilizer use), poorer infrastructure and human capitals, and more limited science and technology capacity. The residual that has not been accounted by the variables included is also quite large, indicating other factors may have also contributed to lower productivity in the region. For labor productivity in the nonfarm economy, roads and telephone together explained more than 60% of the difference between the regional and the national average in the eastern region. For the western region, nearly 40% of the productivity difference (lower than the national average) can be attributed to the physical infrastructure and lower education level. Large residual in the accounting for nonfarm productivity indicates that many other factors may also play a very important role in the nonfarm economy. For the overall rural economy (aggregation of both agricultural and nonfarm economies), public capital, such as roads, telecommunication, and education, explained about 45% of the higher productivity in the eastern region. In the western region, lower public capital accounted for 26% of the lower productivity. In the central region, however, because its productivity is very close to the national level, it is not obvious how public capital has affected its productivity difference when compared to the national average. 4 Because the development level in the Central region is close to the national average, the absolute difference in labor productivity is rather small. However, the decomposition analysis is based on relative percentage terms. Therefore, the results for the Central regions could be very sensitive. S. Fan, X. Zhang / China Economic Review 15 (2004) 203–214 213 6. Conclusions The 1996 Agricultural Census provides a unique dataset to analyze various issues on rural development in China. In particular, it provides very detailed data on rural infrastructure, education, and science and technology. This paper is an early attempt to use this data set. Partly due to the limited access, the data we have is not complete, covering only 45% of the country. We will pursue more detailed and more thorough analyses once we have a complete data for all counties. Despite the crudeness of the data and model we used, the results do shed new lights. First, rural infrastructure and education play a more important role in explaining the difference in rural nonfarm productivity than agricultural productivity. Because the rural nonfarm economy is a major determinant of rural income, investing more in rural infrastructure is key to an increase in overall income of the rural population. Second, the lower productivity in the western region is explained by its lower level of rural infrastructure, education, and science and technology. Therefore, improving both the level and efficiency of public capital in the west is a must to narrow its difference in productivity with other regions. This research merely serves as a touchstone for future research. One of the urgent future research topics is to search different policy options to mobilize resources to support public good provisions for the less developed western region. Under the current fiscal decentralization scheme, financing infrastructure in regions with a small nonfarm sector faces a great challenge. Lack of local revenues is one of the major causes of underinvestment in the less developed western region. Acknowledgements The funding from FAO to the first author for his travel to Beijing to participate in the International Seminar on Chinese Census Results September 19 – 22, 2000, Beijing, and the funding from ACIAR for all the authors in data collection, data compiling, and paper preparation are acknowledged. References Ash, R. F., & Edmonds, R. L. (1998). China’s land resources, environment and agricultural production. China Quarterly, 156, 836 – 879. Fan, S. (1990). Regional productivity growth in China’s agriculture. Boulder, CO: Westview Press. Fan, S. (1991). Effects of technological change and institutional reform on production growth in Chinese agriculture. American Journal of Agricultural Economics, 73(2), 266 – 275. Fan, S. (2000). Research investment and the economic returns to Chinese agricultural research. Journal of Productivity Analysis, 14(92), 163 – 180. Fan, S., & Pardey, P. (1997). Research, productivity, and output growth in Chinese agriculture. Journal of Development Economics, 53, 115 – 137. Fan, S., Zhang, X., & Robinson, S. (2003). Structural change and economic growth. Review of Development Economics, 7(3), 360 – 377. Hayami, Y., & Ruttan, V. (1985). Agricultural development: An international perspective. Baltimore, MD: John Hopkins University Press. 214 S. Fan, X. Zhang / China Economic Review 15 (2004) 203–214 Huang, J., & Rozelle, S. (1996). Environmental stress and grain yields in China. American Journal of Agricultural Economics, 77, 853 – 864. Lin, J. Y. (1992). Rural reforms and agricultural growth in China. American Economic Review, 82(1), 34 – 51. McMillan, J., Whalley, J., & Zhu, L. (1989). The impact of China’s economic reforms on agricultural productivity growth. Journal of Political Economy, 97, 781 – 807. Rozelle, S. (1994). Rural industrialization and increasing inequality: Emerging patterns in China’s reforming economy. Journal of Comparative Economics, 19(3), 362 – 391. Smil, V. (1999). China’s agricultural land. China Quarterly, 158, 414 – 429. Zhang, B., & Carter, C. A. (1997). Reforms, the weather, and productivity growth in China’s grain sector. American Journal of Agricultural Economics, 79, 1266 – 1277. Zhang, X., & Fan, S. (2004). Public investment and regional inequality in rural China. Agricultural Economics, 13(2), 89 – 100.