The Impact of Special Economic Zones on Total Factor Productivity of Chinese Provinces: A panel analysis of trends in growth from 1974 to 2011

Ziwei Judy Hao 14.33 Spring 2013 1st draft

1

I. Introduction: A Special Economic Zone (SEZ) is a geographical region that has economic laws more liberal than a country’s regular economic laws. As a trade capacity development tool, SEZs promote rapid economic growth by using tax and business incentives to attract foreign investment and technology. By offering privileged terms, SEZs attract investment and foreign exchange, spur employment and boost the development of improved technologies and infrastructure. Today, there are approximately 3,000 SEZs operating in 120 countries. The People’s Republic of China has been considered as one of the world’s most successful in terms of SEZ development and contribution. China’s SEZs were introduced after Deng Xiaoping’s economic reforms were implemented in China in 1979. These SEZs were designated as areas where market­driven capitalist policies are implemented to entice foreign businesses to invest in China. The policies include investment in new infrastructure like office buildings and banks and preferential tax exemptions for foreign firms who want to invest in China. Some the most well known SEZ cities are Shengzhen, Shanghai, and Beijing. From 1979 to the present, the number of SEZ cities grew quickly from the original five cities in the southeast coast to an astonishing 118 cities. Along with SEZ development in the past 30 years, China has achieved phenomenal economic growth, an unprecedented development “miracle” in human history. Since the institution of its reforms and Open Door policy in 1978, China’s gross domestic product (GDP) has been growing at an average annual rate of more than 9%. While China has seen definite economic growth since the introduction of SEZs, the benefits of SEZ have yet to been evaluated on the province level. Many previous papers, such as the one by Wong (1987) , focuses on quantifying the movement of foreign direct investment (FDI) flowing into well known SEZ cities over the years. A few recent papers have looked into effect of SEZ status on economic productivity, particularly that of Wang (2010), which concluded that SEZ status increases total productivity factor (TFP) for a group of experimental SEZ towns. Following that paper, Xu and Chen (2011) uses invested capital values and productivity for each province from 1979 to 2005 to identifies spillover effects from SEZ induced FDI increases on neighboring cities. However, there is no concrete results on exploring productivity trends for each province before and after establishing an SEZ city.

2

Serving as a stepping stone towards this goal, this paper seeks to answer two major questions: (1) How large are the growth benefits of SEZs brought about to a provinces compared to the state of its productivity prior to the introduction of an SEZ? (2) How can numerically predict annual changes in productivity growth before, during and after when an SEZ is introduced to a province? While the goal for this knowledge faces difficulty in controlled research and data collection, the results will hopefully shed more light on macro­level impact of SEZs on the domestic economy. To shed light on the proposed questions, I’ve collected a unique dataset of economic statistics from 1973 to 2011 for 30 provinces. My results show that: 1. (to be added) 2. (to be added) The remainder of the paper is organized as follows: Section II outlines the background for this research paper from literature on SEZ related experiements. Section III presents the methodology for data collection and econometric framework for the regression. Section IV reports results and data analysis. The final section concludes and summarizes the findings.

II. Background A. Literature Review on History of SEZs and FDI growth in first four SEZs By 1970s, China was still a developing country coming out of three decades of isolation and strict revolutionary regime. In dire need of change, leadership turned its head towards opening the country to global contacts and influences. Deng Xiaoping, the chief architect behind China’s economic reforms in 1978 proposed in the Open Door Policy to experiment with free international trade to boost investment, ideas, and development in China. The easiest solution was to establish cities with relaxed trade, labor, and entrance policies to attract foreign direct investment (FDI), expand China’s productivity, and accelerate the introduction of new technology. In July of 1979, the government decided that Guangdong and Fujian provinces should be the first to conduct economic exchanges with other countries and implementing “special policies and flexible measures.” By August of 1980, cities of Shenzhen, Zhuhai, and Shantou in the Guangdong Province were designated as special economic zones (SEZs). Later in October, Xiamen in Fujian Province was

3

added to the list. These four SEZs were encouraged to test out pragmatic and innovative economic policies, serving as experimental ground for policies that, if proven effective, would be implemented more widely across the country. The emphasis on linkages with other highly developed, capitalist countries, such as the U.S., especially through liberalization of foreign investment and trade relations. These SEZs were deliberately select for their far southeastern coastal location, far from the center political influence in Beijing, minimizing potential risks from any major political troubles or changes happen during their experimental period. More specifically, the original four zones were sited in coastal areas with a long history of easy access to the outside world. Shenzhen, especially, had the greatest strategic location advantage because it is situated across a narrow river from Hong Kong, a paradigm of capitalism and modern methods of managing for China to imitate. At the time of its inauguration as an SEZ, Shenzhen was a small fishing village of 30,000 population encompassing no more than 3 square km of dilapidated buildings and poor roads along the coast (Xu and Chen, 2008). Because of these characteristics, Shenzhen has been the target of many major studies on the effect of new SEZ policies on its economic growth. Wong (1987) conducted 10 year panel study the change in FDI inflow to the first four SEZs since their founding. When the four opened foreign trade and investment, the SEZs saw an almost immediate impact. Within one year, the four zones accounted for 59.8% of the country’s total FDI inflow, with Shenzhen accounting for the majority share at 50.6% and the other three roughly 3% each (Wong 177) . By the end of 1985, realized FDI in the four SEZs totaled US$1.17 billion, roughly 21% of the national total (Wong 178). Wong’s study concluded that the combination of favorable policies and the right mix of production factors in the SEZs resulted in high rates of economic growth unprecedented in China. Seeing the initial achievements of the first group of SEZs, the central government granted SEZ status to 14 other coastal cities in 1984. Then from 1985 to 1988, the central government further included more coastal cities into the category. In 1990, the Pudong New Zone in Shanghai as well as more cities in the Yangzi River Valley opened its economy to foreign investment. From 1992 to 1994, the State Council has named a number of border cities and all the capital cities of inland provinces and autonomous regions as SEZs. This was the largest spread of the SEZ initiation process since 1984. The

4

later openings gradually subsided until all the provinces have received SEZ to 2007.

B. Literature review on Total Factor Productivity changes in certain SEZ cities Wang (2010) study is one of the first papers that calculates TFP for major SEZ cities. The first part of the paper looks at FDI inflow of the SEZ cities from 1979 to 2005. Then, it estimate impact of FDI changes by specifying an aggregate production function as the following: Y= ALB1KB2 A = B[1 + q(F/Y)]Fa where Y = real output, L = labor input, K = stock of domestic capital, F = stock of FDI, and A = total factor productivity level. This specification permits total factor productivity (TFP), A, to be determined endogenously by the stock of FDI as well as exogenous influences represented by the residual productivity factor, B. Not only does it provide the basis for assumptions in calculating TFP from labor and capital statistics, the results saw increases total factor productivity growth rate by 0.6% on average for SEZ cities.

C. Literature review on Spillover Effect from SEZs Domestic capital investment in towards production the initial 4 SEZs exceeded the foreign investment sent there by almost five times in amount. A major source of investment in production was due to the limitless supply of cheap labor and constant scale­up in production through technological innovation. For instance, by 1984, Shenzhen’s production expanded sixfold, Zhuhai’s 3 times, Shantou’s 1.4 times, and Xiamen’s 1.6 times, all in comparison to the country’s 1.5 times (Xu and Chen 18). A standard way to measure productivity based on labor and capital is by calculating the total productivity factor for each area. The mathematical model will be further explained in the methods section. Mitze and Özyurt’s 2011 paper looks at total production in Chinese provinces between 1978 and 2010 from openness to trade and FDI inflow, which can be said is directly related to the establishment of SEZs. The study, which uses “the inclusion of spatial effects into a dynamic error correction modelling framework” (Mitze and Özyurt 8) shows that, in addition to domestic investment intensity and infrastructure use, trade openness and inward FDI exert a positive impact on total 5

production. One particular results is that the geography has a strong indirect influence on productivity. A region is surrounded by high­productive regions with a good infrastructure and open trade (an SEZ) shows productivity growth rate. The magnitude of these impacts varies between coastal and interior provinces to fully exploit the benefits from such spillovers. This is an indicator that the trends studied across different time panels will need to account for spillover effect.

D. Approach based on literature reviews By combining the techniques used from the discussed literature, this paper will empirically examine the effect of SEZ introduction on TFP growth under a cross­province panel data framework: 1. Construct a measure for the change in total factor productivity (TFP) of each province from 1973 to 2011, using statistics and assumptions on provincial capital and labor. 2. Evaluate the impact of introducing new SEZ(s) each year, from the first establishment in 1979 to 2005, on each province’s TFP. I attempt to isolate SEZ’s effect on the TFP growth from other potential factors by looking at growth trends before and after a year when a large number of provinces receive their first SEZ, focusing on before and after 1984 as well as before and after 1992. 3. Develop a regression that relates annual TFP growth on the provincial level to years before, during, and after the an SEZ is established in each province. In addition, a set of control variables is introduced to capture provincial differences in income, labor quality, population density, investments, infrastructure, and spillover effect. The results provide more insight on how SEZs can transform a nation beyond the piecemeal results from past papers. III. Methodology and Data My province­level economic data set spans 30 provinces from mainland China from 1973 to 2011 (see table for data description). Data from Hongkong, Macao, Tibet, and Taiwan are not included due to the obvious differences in the political and legal systems between these areas and the other provinces.

6

A. Calculate TFP growth Following Wang (2010), I set gross domestic output as a constant return to scale function of capital (K) and human labor capital (H*L) inputs that changes over time t due to technological progress for each province i: Y i, t = F(K i, t, H i, t * L i, t ) The data values for capital (K) is the total capital per province each year. The data value for human capital is represented by the average years of schooling (H) times employment (L). Because the data from the Statistical Yearbooks don’t include that of average years of schooling but offers the number of graduates per three given levels of schooling, I can use the perpetual inventory method to estimate average years of schooling. Initially proposed by Barro and Lee (2001), this method has been widely applied to measure the average years of schooling for studies conducted on China (e.g. Démurger, 2001). I follow specifically the method set by Démurger (2001) and measure years of schooling accumulated at three broad levels: primary, secondary, and higher education. The calculation is carried out in two steps. First, I calculate the human capital accumulated at three schooling levels using the perpetual inventory method specified: E1, i, t = Pri i, t + Sec i, t + High i, t E 2, i, t = Sec i, t + High i, t E 3, i, t = High i, t where Ej, i, t is the number of students who have completed at least level j of schooling in province i at time t and j denotes the level of schooling, specifically, j=1 indicating primary (Pri) education, j=2 secondary (Sec) education, and j=3 higher (Hi) education. Next, I assigned length in years of respective schooling cycles for each level of schooling. Following Demurger (2001), I assign the weights for primary, secondary, and higher schoolings at 5, 10 and 14.5 years, respectively. I divide the weighted sum of Ej, i, t by the total population (Pop i, t), we obtain the aggregate stock of human capital per capita, specified as Schooling i, t= (5 E1, i, t + 10 E 2, i, t ＋14.5 E 3, i, t ) / Pop i, t Lastly, I find human labor capital as Schooling i, t multiplied by the Li, t from the respective year and province. 7

Returning to the equation, Y i, t = F(K i, t , H i, t * L i, t ), I differentiated it and divided it by output (Y) to get:

Fi represents the partial derivative of F with respect to i (for capital K or human labor capital HL). The two terms in parentheses represent the share of each factor out of total factor payments. Total factor productivity growth, dY/Y, represents the proportional change in output that would have occurred in the absence of any input changes. It is calculated as a residual item by subtracting from output growth the percentage change in capital and labor: , where θk , the provincial capital share. I use the national capital share (0.4) as a proxy for the provincial capital share, as suggested by Wang (2010).

B. Analyzing trends in TFP growth over time

Table 1: Year versus number of news provinces that established its first SEZ Year 1979 1980 1981 1982 1983 1984 1985 1986 1987 1988 1989 1st SEZ Province 0 1 1 0 0 6 0 1 0 2 0 Year 1990 1991 1992 1993 1994 1995 1996 1997 1998 1999 2000 1st SEZ Province 0 3 10 3 2 0 0 0 0 0 1

Image 1. The spread of SEZ movement through municipalities from 1980 ­ 1992 8

From Table 1 and Image 1 above, we see that there are two periods that peak in the number of provinces establishing their first SEZs. In 1984, 6 new provinces with average distance to the coast of 15 miles from the southeast coast joined the SEZ movement. These provinces are Shandong, Jiangsu, Zhejiang, Hebei, Liaoning, and Tianjin. In 1992, 10 new provinces in central and northern areas (average distance of 426 miles to the nearest coast) joined. These provinces are Jiangxi , Sichuan, Shanxi, Jilin, Guizhou, Anhui, Henan, Hunan, Shaanxi, Guangxi, Yunnan, Ningxia, and Inner Mongolia. While the immediate years before and after 1984 did not have any new provinces initiating SEZ, the ones around 1992 did (3 in 1991 and 3 in 1993). For analysis, I will group the 1983­1985 provinces as peak group 1 and 1991­1993 provinces as peak group 2. I find the average TFP growth from 1974 to 2011 for 6 categories: provinces with first SEZ established before 1983, during 1983­1985, after 1985, before 1991, during 1991­1993, and after 1993. I will graph and compare the TFP growth trend line from before 1983, during 1983­1985, and after 1985 in one graph over time with each other and then do the same for before 1991, during 1991­1993, and after 1993. By graphing the changes in TFP growth trends for provinces that established their first SEZ before and after these 9

two peak periods, I can isolate and measure the effect of SEZ establishment on TFP growth.

C. Econometric Regression for TFP Growth TFP Growthit = α SEZdummyit + δ i+ δ i*(1 ­ 1973) + γ t + λ TotalSEZit+ B1*Git + B2*Wageit + B3*GIPdit + B4*Popit + B5*Roadit + B6*Schoolit + B7 Employmentit +B8*Kit+ B9*FDIit + Δε it The key variable is SEZdummyit, which indicates the SEZ presences in a province i at time t and is set to an integer value ∈ [0, 1] . SEZdummy = 1, if the province establishes an SEZ and SEZdummy = 0, if otherwise. TotalSEZit ,the number of SEZs beyond the first one in province i at time t, will control for spillover effect. I use δ i as the province fixed effect to control for time invariant province characteristics, such as differences in geographical location and culture, that could potentially influence the timing of SEZ granting. To control for common macroeconomic shocks, such as many different reforms carried out in China and financial downturns, which might influence the growth rate of TFP, I further add year fixed effect, γ t. (t ­ 1973) is the trend in growth starting from 1974, which marks the beginning of my sample. Other added regional variables that may affect TFP growth include: (1) G is the annual growth in real per capita GDP (G = GDPt/GDPt­1 – 1) which standardizes growth effects; (2) Wage, the annual growth in average worker and staff wage, which proxies the labor price; (3) GIPd, the annual growth in regional gross industrial output value of domestic enterprises, to measure regional productivity; (4) Pop, annual growth in population which captures the potential market size; (5) Road, the annual growth in road length which captures the infrastructure effect; (6) School, the annual growth in the average years of schooling which measures labor quality; (7) Employment, the number of workers and staff employed to measure labor market size; (8) K, the annual growth in gross capital; and lastly (9) FDI, the annual growth in FDI for each region, which controls for foreign investments in TFP. Please see Table 2 for data sources. The expected signs of the coefficients are as follows: Coefficients α λ B1 B2 B3 B4 B5 B6 B7 B8 B9

10

Expected Sign

+

­

+

+

+

+

+

­

+

­

+

The main coefficients to focus on are α and γ . First, α , the coefficient for SEZDummy , is expected to be significantly positive because SEZ presence is expected boost provincial TFP as it did for the cities observed in Wang’s paper (2010). γ , the coefficient of TotalSEZ, should be negative because increasing the number of SEZs in a province will result in diminishing return to scale once the boost from spillover effect of SEZs is accounted.

Table 2: Variables and Data Description (Source: China Statistical Yearbook) Variable TFP Growth GDP K Employment School Pop. Wage Road GIPd FDI Data Description Calculated from GDP growth, capital growth , employment, and average years of schooling Gross Domestic Product (100 million yuan) Gross regional capital (100 million yuan) Number of staff and workers employed (10000 persons) Average years of schooling (year). Calculated from primary, secondary, higher education graduation statistics, and population Regional population (10000 persons) Average Wages of Workers and Staff (Yuan) Average length of highway roads (km) Regional gross industrial output value of domestic enterprises (100 million yuan) Foreign direct investment (100 million yuan)

IV. Results and Analysis A. TFP Growth Calculation and Trends

11

TFP growth is constructed from percent growth values of GDP, gross capital, average years of schooling, population, and employment levels. Displayed in table 3 below are the descriptive statistics of all the variables used. Percent growth for each variable is calculated by Growth(Var) = Vart/Vart­1 – 1.

Table 3: Descriptive Statistics of the Variables % Growth Variables TFP G Capital School Employment Pop. Wage GIPd Road Length FDI # Obs. 1140 1140 1140 1140 1140 1140 1140 1140 1140 1140 Mean 0.0714 0.1527 0.2088 ­0.0074 0.0221 0.0118 0.1224 0.1711 0.0525 0.2088 Std Dev 0.0785 0.0819 0.1951 0.0523 0.044 0.0153 0.0784 0.1678 0.1682 0.1951 Minimum ­0.3329 ­0.1524 ­0.3866 ­0.4652 ­0.4378 ­0.2476 ­0.2100 ­0.6627 ­1.000 ­0.3866 Medium 0.0730 0.1452 0.1927 ­0.0115 0.0218 0.0112 0.1228 0.1677 0.0189 0.1927 Maximum 0.3385 0.5344 1.9038 0.5269 0.4654 0.1898 0.6502 0.9423 1.9736 1.9038

The TFP growth values are then grouped based on their respective province and SEZ induction into the previously mentioned time categories: before 1983, during 1983­1985, after 1985, before 1991, during 1991­1993, and after 1993. For each category, the TFP values are then averaged across the grouped provinces and graphed in two separate sets from 1974 to 2010 (see Graph 1 and Graph 2).

12

Graph 1: TFP growth trends before, during, and after 1983­1985

Graph 2: TFP growth trends before, during, and after 1991­1993

13

In Graph 1, the general trend movement (up and down turns) for all three categories are very similar for most parts. While the after and during lines are more closely matched in growth values, the before trendline experiences greater magnitude of changes, potentially due to the variation from the only two provincial data sets that fall in this category (Guangdong and Fujian). Prior to 1983, the before trendline is higher in its average TFP growth than the other two categories, particularly spiking upward when the first SEZs were founded in 1980. But reaching 1983, the before trendline surprisingly falls below the other two trend lines and does not rise above them until 1995. One possible explanation for

14

this situation is that the introduction of SEZs to other provinces along the southeast coast diverted investment and innovation to those locations. These new areas may have promised greater returns to investment and development because of their recent opening, which is indicated by the large upwards spike for the during trendline around 1983. Moving to Graph 2, the general trend movement (up and down turns) for all three categories are, again, very similar. While the after and during lines are more closely matched in growth values, the before trendline still experiences greater magnitude of changes. This before 1991 category depends on data from 11 provinces, which is much greater than the data size for the before 1982 category. An explanation for the occurrence is that the provinces before 1982 are mostly coastal areas that are more sensitive to foreign economic factors because they are usually the first to receive resources being on the coast as well as first to experience downward shocks. Before 1993, the before trendline is higher in its average TFP growth than the other two categories, particularly spiking upward when the first SEZs were founded in 1980 and then again in 1990 (possibly because of the opening of the Shanghai Stock exchange in 1990). By 1983, the before trendline merges with the other two trendlines, with the during and after trendlines seeing sudden increase. However, by 1994, when the majority of the new SEZ provinces were introduced, the before trendline shoots up from .1019 to its highest at TFP growth of .225. The during trend line also rises from around .1021 to .1701, which isn’t nearly as high as the before trend line. The after trendline follows the during trendline closely in its changes. After 1994, the before trendline falls below the other two trendlines again and not rising above them until 2007. Why did this pattern occur? Looking carefully at the locations of the 1993­1995 provinces (Image 1), the new provinces that receive SEZs in 1992 are mostly in central Asia, further from the coast. But because the resources, such as new technology and foreign investments, for these new provinces must still go through the ports, the before category provinces (mostly coastal) receive a major yet short boost in TFP growth in 1994. Immediately afterwards in 1995, the resources from the coasts are pushed towards the central areas, thus explaining the rise in during and after trendline, and fall in the before trendline. In addition, we see that the after trend lines for both graphs rise and fall with the other two trend lines, potentially following the trends due to spillover effect from SEZ provinces .

15

Image 1. Geographical spread of new SEZ provinces from 1980 to 1992.

B. TFP Growth Fixed Effect Model Revisiting the least squared fixed­effect linear model set up in the methods, Table 4 reports the least squares estimation results for the previously stated regression with growth of TFP from 1974 to 2010 as the dependent variable. The null hypothesis is H0: B= 0 and the alternative hypothesis is H0: B= 0. The t­statistic tests whether H0 is rejected or not. If it is, then the coefficient is significantly different from zero and has either a significantly positive or negative effect on the GDP growth rate.

Table 6: Least Squares Fixed Effect Regression of Growth of TFP
Coeff α γ B1 B2 B3 B4 B5 B6 B7 B8 B9

16

Expected Model

+
0.487

­
0.086 (.0178) *

+
0.1902 (.109)* *

+
0.901 (.768)* *

+
0.856 (.412)** *

+
0.24 (0.20 )

+
0.261 (0.21 )

+
­0.934 (0.67)* *

+
0.208 2 (.150 )*

+
­0.21 1 (.091) **

+
.227 0.325 **

STD ERR

(.347)* *

Notes:Robust, Standard error in parentheses. significant at *10% level; ** 5% level; *** 1% level.

The level of TFP grows with the presence of an SEZ. The coefficient is significantly positive at p