Technology Adoption and Absorption: The Case of Shanghai Firms

Loren Brandt and Susan Chun Zhu April 26, 2005

Abstract Using a unique rm-level survey dataset collected by one of the authors, this paper addresses two questions central to the literature on technology diusion: What factors inuence technology adoption, and how well do rms absorb newly acquired technology? We nd that given the same technical capacity, rms with better access to cheap bank credit are more likely to adopt larger technology projects and invest more in imported equipment from technologically-advanced countries. On the other hand, the return to technology investment diers signicantly across rms. In particular, rms with better access to cheap credit have signicantly lower project protability and capacity utilization. These results have important implications for the role of nancial development in technology diusion. (JEL Classication: O3)
Brandt: Department of Economics, University of Toronto, 150 St. George Street, Toronto, Ontario, M5S 3G7, Canada. Zhu (corresponding author): Department of Economics, Michigan State University, Marshall Hall, East Lansing, MI, 48824, USA. We are grateful to Steven Haider, Bart Hobijn, Rich Jensen, Raoul Minetti, Ximing Wu, seminar participants at McGill University, Michigan State University, and conference participants in the Mid-West International Economics Meeting 2002, the 9th EIIT conference, the Econometric Society North American Summer Meeting 2003, the NBER Summer Institute 2003, and the CEA Annual Meetings 2004. Brandt would like to acknowledge the support of the Shanghai Economic Research Center and the National Statistical Bureau for their help in carrying out the survey. He would also like to thank former colleague Yehuda Kotowitz for his important contribution to the project at an earlier date. Financial support for the survey was provided by the Canadian International Development Agency.

Introduction

Technology diusion is an essential component of technological progress and thus an important source of economic growth. A major channel through which technology diusion occurs is investment in technology-embodied machinery and equipment. As suggested by many macro-level studies (e.g., De Long and Summers 1991, Jones 1994, Eaton and Kortum 2001, Caselli and Wilson 2003), equipment investment is signicantly associated with economic growth for a wide range of countries. Given this strong link between equipment investment and economic development, promoting investment in technologyembodied equipment and facilitating successful absorption of newly acquired technology are central issues for developing countries. To address these issues, micro-level studies are necessary. However, most micro-level studies of these issues are in the context of developed countries, especially the United States (e.g., Dunne 1994, Rose and Joskow 1990, Cohen and Levin 1989). Detailed empirical studies on less developed countries at the rm level are scarce.1 Since less advanced countries dier in their institutions, technology, and endowments from advanced countries, we expect that the process of technology diusion would also dier. Yet not much is known empirically about the factors aecting technology adoption by rms in less advanced countries and the subsequent impact of these choices on rm performance. In this paper we examine technology adoption and absorption in China using a unique rm-level survey dataset collected by one of the authors. This survey covers 250 randomlyselected rms in Shanghai, the home of some of Chinas most technologically advanced rms. The survey provides us detailed information on each rms largest technology renovation project (jishu kaifa xiangmu) that was carried out between 1985 and 1992. These technology projects were for the express purpose of renovating or modernizing the
One notable exception is Vishwasrao and Bosshardt (2001). However, in their paper the issue of technology absorption is not addressed.
1

production operations of a rm. On average investment in equipment and new production facilities constituted about 80 percent of the total project expenditure. In particular, expenditure on imported equipment captured nearly 30 percent of the project budget. The survey also contains rich information on the rms, their market environment, and project performance. Our survey covers an important period of Chinas economic reform. In the early 1980s, China began a systematic eort to reform her industrial enterprises. Firms that were largely controlled by the state were given new autonomy, including their investment decisions in the technology projects addressed in this paper. Through the introduction of the dual-track system (Naughton 1995, Lau, Qian, and Roland 2000), the government relaxed controls over most product prices. Barriers to entry were also reduced in many sectors, which fostered more competition in product markets. However, economic reform in Chinas nancial sector seriously lagged behind. Lending by Chinas state-owned banks, which dominated the nancial sector, was largely administratively determined through the national credit plan (Lardy 1998, Brandt and Zhu 2000). Nearly 90 percent of all lending by the state-owned banks went to the state-owned rms at real interest rates that were well below the opportunity cost of investment. In this paper we will present strong evidence that distortions in the nancial sector generate distortions in investment behavior of those state-owned rms and further aect their performance. We rst develop a simple theoretical model to examine the link between access to bank loans, technology choices, and rm performance. As suggested by our model, rms that enjoy preferential access to cheap bank credit adopt larger projects, but have lower protability. This establishes the central hypothesis of the paper. To test this hypothesis, we exploit the fact that during the period of our study, rms diered greatly by ownership types in their ability to access cheap bank loans. In particular, bank loans were more important as a form of government subsidy to state-owned

rms than other types of rms. Therefore, the ability to access bank loans should aect state-owned rms much stronger than other ownership types. Our hypothesis is strongly supported by the data. We nd that state-owned rms with better access to bank loans adopted signicantly larger projects and invested more in imported equipment.2 The estimates suggest that for state-owned rms, a 1 percentage point increase in the fraction of project expenditure that is nanced through bank loans is associated with a 1.38 percent increase in project size and a 2.13 percent increase in the expenditure on imported equipment. Furthermore, state-owned rms with better access to bank loans have lower project protability and capacity utilization: a 1 percentage point increase in the share of project expenditure that is nanced through bank loans is related to a 2.30 percent decrease in project protability and a 0.26 percent reduction in project capacity utilization rates. In contrast, we nd no strong evidence that access to bank loans was aecting technology adoption and absorption for other ownership types. Our results suggest that although government support (mainly in the form of cheap bank loans) can encourage rms to adopt more advanced technology, the government may be unable to solve easily the problem of technology absorption. In fact, pushing enterprises to adopt more expensive and advanced technology can lower project performance given the weak capabilities of these rms. These results have important implications for the role of nancial development in technology diusion. Access to credit at preferential terms has been a key ingredient to the industrial strategies followed by Chinas Asian counterparts including Japan, South Korea and Taiwan. By all indications, the results of these policies have been mixed (Amsden 1989, Calder 1993, Katz 1998, Wade 1990). Our ndings highlight that distortions in the nancial sector likely generate mismatch between rms and technology, often resulting in
It is worth pointing out that during the period of our study, the lending by state-owned banks to state-owned rms was largely directed by the governments credit plan. State-owned banks also lacked the expertise of choosing good projects. Therefore, one can condently rule out the possibility that it is the banks that choose to nance larger and better projects.
2

under-utilized capacity with higher absorption costs and lower protability. Our results also provide empirical support for the literature that emphasizes the role of nancial development in economic growth (Gerschenkron 1962, Greenwood and Jovanovic 1990, Levine 1997, Rajan and Zingales 1998, Acemoglu, Aghion and Zilibotti 2002). As stressed in the literature, technology diusion is also likely to be inuenced by the technical capacity of rms. Since the survey provides rich information on rms, we are able to control for various rm attributes that may aect technology adoption and absorption. In particular, we nd that (1) rms with larger xed assets and higher output prior to the project adopt bigger projects; (2) more protable rms invest more in imported equipment while labor-intensive rms invest more in domestic equipment; (3) rms with higher physical capital intensity and protability prior to the project have higher project protability; and (4) project protability and capacity utilization increase over the length of the project. These results suggest that technical capacity and learning by doing also play roles in technology diusion. Thus, our results are also consistent with the studies that emphasize the importance of technical capacity in technology transfer (see Evenson and Westphal 1995). The paper is organized as follows. Section 2 describes the data, including an overview of the rms in our sample, their operations and market environment, and technology projects. Section 3 presents a simple theoretical model of rm investment decisions. Detailed empirical analysis is given in sections 4 and 5. Section 4 focuses on technology choices, i.e., project size, and investment in imported and domestic equipment. Section 5 analyzes project performance in terms of project protability and capacity utilization. Section 6 draws the conclusions.

2
2.1

Data and Economic Background

The Survey

The survey was carried out by the Shanghai branch of the National Statistical Bureau on our behalf in the fall of 1995 and spring of 1996. Altogether, 250 rms in Shanghai were randomly selected and surveyed in ve sectors chemical, machinery, transportation, electrical equipment and machinery, and electronic and telecommunications equipment industries. In 1993 these ve sectors were the source of 40 percent of total industrial output in Shanghai, and 30 percent nationally. Survey forms were reviewed on two separate occasions and sent back to the rms for revision. After careful assessment, 2 rms were dropped from the sample because of unsatisfactory forms, leaving us with a total of 248 rms, 124 of which were state-owned enterprises, 74 collectively-owned enterprises, and 50 joint ventures.3 The survey itself was divided into four components. The rst part provides basic information on the rm, including xed assets, employment, prots, and output for the period 1987-93. The second part contains data on wages, labor composition by education, and other personnel information. The third part, which was lled out by either the general manager or factory manager of the rm, covers the rms operations and market environment. The nal part, which is the core of the survey, provides detailed information on each rms largest technology renovation project (jishu kaifa xiangmu) that was carried out between 1985 and 1992. These technology projects were for the express purpose of renovating or modernizing the production operations of a rm. Because rms were concerned about disclosing proprietary information, we limited ourselves only to the
3 Collectively-owned enterprises are rms that were established, owned and managed by lower levels of governments. In our sample, they include rms owned by the Shanghai government, as well as outlying counties and townships that formally made up the municipality of Shanghai. Joint ventures are usually formed by Chinese rms (state-owned or collectively-owned) and foreign partners. Provisions of multiyear tax holidays and expanded autonomy for joint ventures provided state-owned and collectively-owned rm managers with additional incentives to nd oshore investors and technology suppliers.

projects carried out before 1993. However, this provides us a minimum of two years of performance for each project.

2.2

Shanghai

Historically, Shanghai was the center of Chinese industry and the home of some of Chinas largest and best rms. In 1988, Shanghai rms were the source of 7.1 percent of the national gross value of industrial output and 12.3 percent of the national industry prots. On average Shanghai rms were larger and more capital intensive than rms in the rest of China. They also had higher labor productivity and protability (i.e., prots per xed assets). In addition, Shanghai has much higher concentration of state-owned enterprises. In 1988, the share of gross value of industrial output produced by state-sector rms was 56.8 percent nationally. In Shanghai, however, the percentage was considerably higher, 70.5 percent. During the period of our study, technology imports from advanced countries played an increasingly important role in modernizing Chinese rms. Between 1983 and 1992, Shanghai rms signed 2,512 technology import contracts, of which 1,175 had started production by 1992. Up through 1992 the cumulative value of these agreements (which presumably includes the value of capital equipment) was $US 2.5 billions, of which 23.9 percent was with rms from Japan, 20.5 percent was with Germany, and 15.4 percent was with the United States (Shanghai Tongji Nianjian, select years).

2.3

Firms Operations and Market Environment

Our survey covers an important period of Chinas economic reform. In the early 1980s China began a systematic eort to reform its industrial enterprises (Naughton 1995). Firms that were largely owned and controlled by the state were given new autonomy over production decisions and provided more powerful economic incentives. For example, they 6

were allowed to retain a portion of their prots and to provide bonuses to managers and workers. In the process, their objectives shifted from fullling output quotas or other plan targets to increasing prots. At the same time, the central government reduced barriers to entry in many sectors, and relaxed controls over prices of most products and inputs through the introduction of the dual-track system (Byrd 1991, Lau, Qian and Roland 2000). There has been considerable debate in the literature over the eect these reforms had on state-owned enterprise behavior and productivity vis-`-vis their non-state-owned a counterparts, especially up through the early-to-mid 1990s (Lardy, 1998). The general consensus now appears to be that while these reforms helped to alter the behavior of state-owned rms and improve their performance (Groves et. al. 1994, Li 1997, Shirley and Xu 2000), their protability and productivity growth continued to lag signicantly behind rms in the non-state sector (Jeerson and Rawski 1994). Soft budget constraints facing state-owned rms (i.e., rms know ex-ante that they will be able to default on their debt in the event of project losses) appear to be the likely cause. Over the period we are examining, Chinas nancial system was dominated by four state-owned banks, which held more than 80 percent of the assets of the nancial system. Lending by these banks was largely administratively determined through the national credit plan, and heavily biased in favor of state-owned rms. Up through the early 1990s, between 85-90 percent of all lending by the state-owned banks consistently went to the state-owned rms in the form of working capital and xed investment loans (Jinrong Nianjian, select years, Brandt and Zhu 2000). There was a signicant subsidy component implicit in lending. Real interest rates were low, and in fact occasionally negative (Lardy 1998). Soft budget constraints further reduced the eective borrowing costs facing these state-owned rms.4
The bias in bank lending towards state-owned rms and budget softness can be linked to their size and their responsibility for worker welfare, including housing, health services, retirement benets, etc.
4

The eects of industrial reform on rms are evident in our data. Table 1 summarizes self-reported information on rms objectives, product pricing, factors aecting protability, and barriers to investment. With respect to rm objectives and product market competition, the dierences between state-owned rms and rms in the non-state sector appear modest during the period of our study. Two-thirds of all types of rms reported that reducing production costs and achieving economies of scale through market expansion were top objectives, suggesting that improving rm protability is a goal common to all ownership types. In addition, more than half of state-owned and collectively-owned rms and 84 percent of joint ventures reported that the prices of their main products were determined by market forces. Regarding rm protability, the majority of all types of rms listed competition from domestic rms as an important factor. In contrast, less than one-third of rms considered government interference as a signicant factor. Hence, most rms considered their production decisions to be aected more directly by domestic market competition than by government intervention. Despite signicant changes in the product markets in the late 80s and early 90s, economic reform in the nancial sector largely lagged behind. This imposed serious constraints on rm investment and protability. Over 80 percent of rms considered availability of funds as an important factor aecting protability. For the majority of all types of rms, insucient internal funds and diculty in getting bank loans were major barriers to investment. However, as shown below, rms diered greatly by ownership in their ability to access external funds (mainly bank loans).

2.4

Firm Attributes

Table 2 provides summary statistics on rm attributes with respect to rm age, size, physical capital intensity, human capital, and rm protability. The left panel of the table
Although state-owned rms were inecient, the political costs of hardening budgets and cutting o funding during this period were still high.

displays information for the full sample of 248 rms. The right panel reports summary statistics for a smaller sample that only includes rms with information on employment, prots, output, and xed assets prior to the technology project.5 For state-owned and collectively-owned rms, the dierences between the two samples are small in general. However, joint ventures in the smaller sample appear to be less capital-intensive, and have lower prots per worker than those in the full sample. In the empirical analysis we mainly exploit the smaller sample in order to control for intrinsic rm attributes. As shown in Table 2, the data reveal considerable rm-level heterogeneity both within and across ownership groups. State-owned rms are substantially older and larger than either collectively-owned rms or joint ventures. On the other hand, joint ventures are signicantly more capital intensive than other rms. This is complemented by greater human capital in the joint ventures. Thus, joint ventures have higher rm protability than other rms, as expected.

2.5

Technology Projects

In this paper we focus on technology projects. For all types of rms, the primary purpose of technology projects was either to introduce new products or improve the quality of existing products. (New products here mean the products that are new to the rm but not necessarily new to the entire product market.) This process of renovating or modernizing production operations is a major component of a rms strategy for succeeding in intensied product-market competition. Table 3 reports project size, a breakdown of project expenditure, sources of project nancing, and project performance. For all types of rms, the investment carried out
The reduction of sample size arises from three sources. First, 9 state-owned rms, 15 collectivelyowned rms and 29 joint ventures were established at the same time as the project, and so we do not have the information on rms themselves prior to the project. Second, 6 state-owned rms, 5 collectively-owned rms and 1 joint venture had projects that started before 1987, but rms were only asked to report data retroactively up to 1987. Finally, 6 state-owned rms, 4 collectively-owned rms, and 4 joint ventures did not report information on rm prots, xed assets or output prior to the project.
5

as part of the technology projects represents a signicant portion of the xed productive assets. For state-owned rms, the median size of new investment is 31 percent of the rms assets at the time when the project begun, while it constitutes 50 percent for collectivelyowned rms and 41 percent for joint ventures. Clearly, the economic success of these rms is tied to the success of these projects, which helps justify our focus on these renovation projects. Further, on average, state-owned rms undertook substantially larger projects than joint ventures and especially collectively-owned rms. For all types of rms, equipment (including domestic and imported) constitutes between 57 percent and 79 percent of total project expenditure. State-owned rms and joint ventures, however, are much more likely to import equipment: 75 percent of state-owned rms and 63 percent of joint ventures imported equipment, compared to only 42 percent of collectively-owned rms that did. This behavior coincides with a larger average share of investment in imported equipment for the state-owned rms and joint ventures. As reported by the rms, the origin of imported equipment is mainly from technologicallyadvanced countries including the United States, Japan, and Germany. Most rms consider the availability of funds as a major factor aecting their investment and protability. However, rms dier signicantly in their ability to access external funds. As displayed in Table 3, 84 percent of state-owned rms used bank loans while 63 percent of collectively-owned rms and just a half of joint ventures obtained bank loans to nance their technology projects. On average, bank loans covered 46 percent of project expenditure for state-owned rms, but only 29 percent for collectively-owned rms and 36 percent for joint ventures. The enormous size dierence in these projects across ownership groups implies that a majority of the credit was going to state-owned rms. The bottom of Table 3 reports project performance in terms of project protability and capacity utilization rates. Project protability is measured by gross project prots relative to project size. By 1993, project protability was the lowest for state-owned rms

10

but the highest for joint ventures. State-owned rms also had the lowest project capacity utilization rates.6 To summarize, we nd that (1) rms dier substantially in terms of size, age, human capital and physical capital intensity. These attributes capture a rms capability to absorb newly acquired technology; (2) rms dier greatly in the ability to access external funds. In particular, state-owned rms have better access to cheap bank credits than other rms; (3) compared to other ownership types, state-owned rms adopt much larger projects and invest more in imported equipment; and (4) state-owned rms appear to have poorer project performance than other rms. In the following we will present a theoretical framework in which the dierences in investment behavior and subsequent project performance are linked to the dierences in technical capacity and the ability to access bank loans.

Theoretical Framework

In this section we present a simple theoretical framework to examine rm investment decisions. We assume that all rms are prot-maximizers. This is a simplifying assumption. It is likely that state-owned rms have other objectives such as maximizing output or employment. However, based on the discussion in section 2.3, the industrial reforms in the late 80s and early 90s increased the weight state-owned rms put on improving prots compared to fullling output quotas or other planned targets. Firms dier in the capability to absorb new technology. Let i denote rm is technical capacity. A bigger i indicates that the rm has higher technical capacity. Firms also dier substantially in the ability to access funds. The major sources of project nancing come from bank loans and internal funds.7 Let rB be the interest rate charged by banks, and
The project performance of state-owned rms may be overstated due to the fact that on average state-owned rms started technology projects earlier than other rms. 7 Foreign investment covered 20 percent of project expenditures for joint ventures (see Table 3). For
6

11

rI the opportunity cost of internal funds. Because bank loans are a form of government subsidy to rms,8 we assume that rB < rI . This is contrary to the usual assumption that the cost of internal funds is lower than that of external funds due to adverse selection or moral hazard problems (e.g., Johnson, McMillan, and Woodru 2002). Since rB is lower than rI , a rm must rst seek bank loans before using their internal funds to nance their projects. How much credit a rm can obtain is determined by banks on the basis of rm characteristics (mainly ownership and size). That is, the rm cannot choose how much bank loans to obtain. Let bi be the fraction of project expenditure that is nanced through bank loans for rm i. A higher bi indicates that a rm has better access to bank loans. Then the eective interest rate for rm i is bi rB + (1 bi ) rI . Let ki be rm is project investment, and p the price of capital goods. Let R (ki ) be the expected project revenue in each period, and C (ki , i ) the expected cost of operating new technology in each period. The expected project prot in each period is R (ki ) C (ki , i ). Thus, the present value of expected project prots is [R (ki ) C (ki , i )] / bi rB + (1 bi ) rI . Note that the current capital stock is suppressed in R (ki ) and C (ki , i ). All variable inputs are optimally chosen. For simplicity, we also omit depreciation, taxes, and costs of adjusting the capital stock. In addition, we do not consider any strategic interaction between rms. Then the rms investment problem can be characterized as R (ki ) C (ki , i ) pki . bi rB + (1 bi ) rI

max
ki

In order to have an interior solution to this problem, we make the following assumptions: (1) R (ki ) is continuously dierentiable, strictly increasing, and strictly consimplicity, we treat foreign investment as equivalent to internal funds in terms of the opportunity cost of investment. 8 Bank loans here play the same role that government transfers do in the model of Shleifer and Vishy (1994). Moreover, although our model does not explicitly deal with soft budget constraints, our results should hold in their presence since soft budget constraints essentially lower the interest rate charged by banks (Dewatripont and Maskin 1995).

12

cave; (2) C (ki , i ) is continuously dierentiable, strictly increasing in ki , strictly decreasing in i , and strictly convex for any given i ; and (3) R (0) C (0, i ) = 0 and
ki 0

lim R (ki ) Cki (ki , i ) > 0. Assumptions (1) and (2) guarantee that the second-order

condition holds, i.e., R (ki ) Cki (ki , i ) < 0. The rst-order condition for the rms problem is

R (ki ) Cki (ki , i ) = bi rB + (1 bi ) rI p.

(1)

That is, rms choose the optimal investment ki i , bi , rB , rI , p by balancing the marginal

return to investment against the user cost of capital. Based on this rst-order condition, it is straightforward to derive the eect of technical capacity (i ) and the ability to access bank loans (bi ) on project size as follows:
p rB rI ki = bi R (ki ) Cki (ki , i ) Cki ,i (ki , i ) ki = . i R (ki ) Cki (ki , i )

(2) (3)

The implications of equations (2) and (3) are summarized in proposition 1. Proposition 1 (Technology Adoption)
(1) If rB < rI , then ki /bi > 0. In addition, ki /bi increases in rI rB . (2) If Cki ,i (ki , i ) < 0 (i.e., ki and i are complementary), then ki /i > 0.

That is, rms with better access to bank loans and higher technical capacity adopt larger projects. In addition, the eect of bank loans on project size is larger when the gap between rB and rI is bigger. This gap between rB and rI is largely determined by ownership types. Since subsidies are more important in loans to state-owned rms, the gap between rB and rI is larger for state-owned rms than other rms.9 Thus, we expect
The dierence between rB and rI would be reinforced by the dierence in the hardness of the budget constraints between state-owned rms and non-state-owned counterparts.
9

13

that the impact of bank loans on project investment would be the strongest for stateowned rms. We are also interested in the impact of i and bi on project performance. One performance measure is project protability which is a ratio of project prots to project size,
i.e., i [R (ki ) C (ki , i )] /pki . The ability to access bank loans (bi ) aects project protability via its impact on technology choice (ki ). Using the rst-order condition in

equation (1), it is straightforward to derive the eect of bi on i as follows:

[R (ki ) C (ki , i )] bi rB + (1 bi ) rI pki I i i ki = = r rB . 2 bi ki bi (ki ) R (ki ) Cki (ki , i )

(4)

The implication of equation (4) is summarized in proposition 2.10 Proposition 2 (Technology Absorption) If rB < rI , then i /bi < 0. In addition, i /bi decreases in rI rB . That is, given the same technical capacity, rms with better access to cheap bank loans have lower project protability. This is because rms with better access to cheap credit choose larger projects, which increases the costs of adopting and using new technology and thus reduces the rate of project prots. Technical capacity has two opposing eects on project protability. On the one hand, higher technical capacity directly reduces the cost of using new technology and thus raises protability. On the other hand, technical capacity indirectly aects project protability via its impact on project size: rms with higher technical capacity choose larger projects. This raises the costs of adoption and absorption and hence lowers protability. If the direct eect dominates, rms with higher technical capacity have higher project protability.
10

Under assumption (3) (i.e., R (0) C (0, i ) = 0 and lim R (ki ) Cki (ki , i ) > 0), the optimal
ki 0

investment ki must satisfy [R (ki ) C (ki , i )] bi rB + (1 bi ) rI pki > 0.

14

This completes our discussion about the theoretical model. This model provides a basis for the following empirical analysis.

Technology Adoption

In this section we focus on investment decisions by rms. As suggested by proposition 1, rms investment decisions are determined by technical capacity (i ) and the ability to access bank loans (bi ). Since the eect of bi on investment depends on the wedge between rB and rI , which is mainly determined by ownership types, we include the interaction term between bi and ownership types. (See equation (2)). Our empirical specication is

tech choicei = f (ownershipi , bi ownershipi , i , Zi ) + i

(5)

where i indexes rms; tech choicei represents project size; ownershipi indicates ownership types; bi is the fraction of project expenditure that is nanced through bank loans; i is a vector of variables measuring technical capacity including rm size, age, human capital, and physical capital intensity; Zi is a vector of other control variables including rm protability (proxy for the availability of internal funds), sector dummies, and time dummies indicating the year when the project started (proxy for investment opportunities facing all rms); and i is the error term. We only use pre-project rm attributes as controls.11 This is because most technology projects involve large investment in xed assets, which would have substantial eects on rm physical capital intensity, prots, and other rm performance after the project.
11 Data on worker education are only available for the year 1994. However, between 1992 and 1994, the average annual turnover rate in our sample of rms was below 1.4 percent (the median was below 0.8 percent) for managers and below 3.6 percent (the median was below 1.9 percent) for production workers. Thus, we expect that educational levels of workers were very similar before and after the project. In addition, our estimation reveals that worker education does not aect investment signicantly. Excluding worker education from our regressions has no signicant eect on our other results.

15

Unfortunately, this reduces our sample from 248 rms to 169 rms for reasons explained earlier (see footnote 5). In particular, about two-thirds of the joint ventures drop out of the sample, which seriously restricts our ability to analyze the behavior of joint ventures. On the other hand, as shown in Table 2, joint ventures look more similar to other rms in the small sample than they do in the full sample. Thus, using the small sample should work against nding signicantly dierent eects between joint ventures and other rms.

4.1

Project Size

Results on project size are reported in Table 4. Column 1 shows that state-owned rms have signicantly larger projects than collectively-owned rms even after controlling for sector and the year in which the project started. Column 2 adds controls for rm size xed productive assets, output, and employment prior to the project. The impact of rm size on innovation and technology diusion has been examined extensively (see Cohen and Levin 1989). All of the existing models predict that larger rms should adopt bigger projects.12 Column 2 shows that all three measures of rm size have a positive eect on project size. In particular, rms with more xed productive assets and higher output prior to the technology project adopt signicantly larger projects. The coecient on state-owned rm becomes insignicantly negative. A rms technical capacity is related to their physical capital intensity, human capital and the vintage of existing technology. Human capital is captured by worker education. The vintage of the rms existing technology is proxied by rm age. The estimates in column 3 show that physical capital intensity, worker education and rm age are not
A larger rm can better reap the benet of adopting a new technology due to economies of scale. A larger rm may also have more internal sources, be more diversied, and be better able to hedge against the risks associated with technology adoption. Moreover, the accelerator model in the investment literature suggests that higher sales or output lead to higher level of investment. The replacement model implies that rms with more capital assets have higher demand for investment due to the need of capital replacement.
12

16

signicantly correlated with project size.13 As will be shown in the next section, however, physical capital intensity is signicantly and positively correlated with expenditure on imported equipment but not on domestic equipment. The insignicant eect of human capital and technology vintage may be due to measurement errors. Worker education may fail to capture rm-specic human capital, which is also crucial for technology absorption. The use of rm age as a proxy for technology vintage ignores the fact that older rms may upgrade their technology. Since insucient internal funds and diculty in getting bank loans are the biggest barriers to rm investment (see Table 1), we are particularly interested in how the ability to access funds aects project size. In column 4, we include rm protability i.e., the ratio of rm prots to xed assets, as a proxy for the availability of internal funds. We nd no evidence that rm protability prior to the project has a signicant eect on project size when we also control for rm size and other rm attributes. In column 5 we add interactions between access to bank loans and rm ownership. Access to bank loans is measured by the fraction of project expenditure that is nanced through bank loans.14 Interestingly, the estimated coecient on the interaction between bank loans and state-owned rm suggests that a 1 percentage point increase in the share of bank loans is associated with a 1.38 percent increase in project size for state-owned rms. This provides support for our theoretical prediction that rms with better access to cheap bank loans adopt larger projects. At the same time, the estimated coecient on the interaction between bank loans and collectively-owned rm is positive, but it is statistically insignicant and its magnitude is also just a third of that for state-owned rms. This dierence between state-owned and
Since physical capital intensity is dened as a ratio of xed productive assets to employment, employment is excluded to avoid perfect collinearity. 14 A regression of access to bank loans on rm attributes reveals that larger rms in terms of the number of workers and xed assets prior to the project obtained relatively more bank loans to cover their project expenditure. On the other hand, access to bank loans does not appear to be related signicantly to rm age, sector, and rm protability and output prior to the project.
13

17

collectively-owned rms can be explained by the fact that bank loans are more important as a form of government subsidy to state-owned rms than to collectively-owned rms. That is, the wedge between the cost of bank loans and that of internal funds is much larger for state-owned rms than for collectively-owned rms. Therefore, the ability to access bank loans aects state-owned rms much more than collectively-owned rms. We also nd that for joint ventures, a 1 percentage point increase in the share of bank loans is related to a 1.50 percent decrease in project size. This is contrary to our theoretical prediction. One explanation is that the cost of bank loans for joint ventures may in fact be higher than the cost of internal funds. Alternatively, for joint ventures bank loans are mainly used to purchase domestic equipment, which is cheaper than imported equipment. (Unlike other rms, joint ventures can use foreign investment to nance their purchases of imported equipment.) So joint ventures that use more bank loans to nance their projects also have smaller projects. We will provide empirical evidence that supports the second interpretation. A potential concern is that our measure of bank loans may be endogenous; banks may decide to nance larger and better projects. However, according to the literature on credit rationing in China, lending by Chinas state-owned banks is more heavily inuenced by government policy than project protability. During the period of our study, state-owned banks often lacked incentives to select good projects over bad ones (Brandt and Li, 2003). They also lacked the expertise to choose good projects that were well matched with rm capability. We expect, therefore, the link between the quality of projects and the size of bank loans should be weak. In addition, due to data limitation, we do not have a more convincing measure of a rms ability to access bank loans. Hence, we interpret our result on the relationship between bank loans and project size as correlation rather than causality. We also experiment with other measures of bank loans. Column 6 reports the estimates

18

when bank loan is a dummy variable, which equals 1 if the rm obtained bank loans to nance the technology project. The results are very similar to the benchmark estimates in column 5. This implies that our results are not driven by a few rms that have very high or very low shares of bank loans. Finally, in all of the specications presented in Table 4, we control for rm sector and the year when the technology project started. (To save space, these results are not reported in the table.) Projects tend to be signicantly larger than average in the transportation industry.15 We also nd that the projects that were carried out after 1989 are signicantly smaller in size than those projects implemented before 1989. This result is consistent with the economic retrenchment beginning late in 1988. For the next several years, tight monetary policy was implemented in order to combat ination. Firm protability was also lower during this period. This shortage of funds tended to reduce project size.

4.2

Investment in Imported versus Domestic Equipment

Inspired by the literature on the relationship between importing equipment and economic growth (e.g., Eaton and Kortum 2001, Caselli and Wilson 2003), we also empirically examine the choice between imported and domestic equipment. In particular, we look at investment in imported and domestic equipment separately. Results are given in Table 5. Columns 1-2 report simple OLS estimates. A comparison between imported and domestic equipment reveals several interesting results. First, for state-owned rms, a 1 percentage point increase in the share of bank loans is associated with a 2.13 percent increase in the expenditure on imported equipment but just a 0.07 percent increase in the purchase of domestic equipment. For collectively-owned rms, we also nd that the estiMost of the projects in transportation were linked to the development of Shanghais emerging auto industry. The key venture was a joint venture between Shanghai Automotive Industrial Corporation and Germany Volkswagen, complemented by investment in new and existing parts suppliers.
15

19

mated coecient on bank loans is larger for imported equipment than that for domestic equipment, although the eect is statistically insignicant. In contrast, for joint ventures, a 1 percentage point increase in the share of bank loans is related to a 2.23 percent decrease of investment in imported equipment but a 1.83 percent increase in domestic equipment. Therefore, unlike state-owned and collectively-owned rms, bank loans extended to joint ventures appear to be tied to the purchase of domestic equipment. This dierence may be due to the fact that joint ventures can use foreign investment to purchase imported equipment. Second, more protable rms invest more in imported equipment. The estimate implies that a 1 percent increase in rm protability is associated with a 0.42 percent increase in expenditure on imported equipment. In contrast, the estimated coecient on rm profitability for domestic equipment is small and statistically insignicant. Firm protability indicates the availability of internal funds. Protability may also be linked to eciency of the rm. Thus, the result provides some evidence that better rms adopt more advanced technology. Third, more capital-intensive rms purchase less domestic equipment. Capital intensity represents the production technique of a rm. Compared to labor-intensive rms, capital-intensive rms are likely to equip their workers with more sophisticated machinery. Their workers may also have greater experience with more capital-intensive and technologically advanced production processes.16 Thus, more capital-intensive rms are likely to have lower demand for domestic equipment. Fourth, rms with more xed productive assets prior to the project spend relatively more on imported than domestic equipment. In particular, the estimates suggest that a 1 percent increase in xed assets is related to a 0.95 percent increase in expenditure on
The simple correlation between physical capital intensity and the percentage of workers with college degree or above is 0.23 (p-value = 0.002), while the simple correlation between physical capital intensity and the percentage of workers with junior high-school diploma is -0.15 (p-value = 0.05). This provides preliminary evidence that physical capital and skills are complementary.
16

20

imported equipment and a 0.60 percent increase in expenditure on domestic equipment. The statistical signicance is also higher for imported equipment than that for domestic equipment. Because 61 rms did not purchase any imported equipment and 43 rms did not invest in domestic equipment, we also use the left-censored Tobit model to take into account the problem of corner solutions. Results are displayed in columns 3-4. The Tobit estimates have similar signs and signicance levels as the OLS estimates, and the estimated marginal eects are also quite similar to the OLS estimates. For example, the estimate in column 3 implies that for state-owned enterprises, the marginal eect of access to bank loans on expected investment in imported equipment is 1.74 (evaluated at the mean values of the covariates). This marginal eect is slightly lower than the OLS estimate of 2.13 in column 1. The estimated marginal eect of rm protability on the purchase of imported equipment is 0.49, which is slightly higher than the OLS estimate of 0.42. Thus, the Tobit estimates in columns 3-4 conrm our previous results. In column 5 we examine the share of expenditure on imported equipment. Unlike total expenditure on imported equipment, access to bank loans is not signicantly correlated with the share of expenditure on imported equipment for state-owned rms. That is, bank loans have an eect on the size but not the composition of technology investment. On the other hand, more capital-intensive and more protable rms have signicantly higher share of imported equipment. This is consistent with the results in columns 1-4 that protable rms invest more in imported equipment while capital-intensive rms spend less on domestic equipment. Finally, in all specications we control for sector and the year in which the technology projects started. (To save space, the results are not reported in the table.) Interestingly, we nd that investment in domestic equipment appears to be more volatile and more inuenced by macroeconomic uctuations. In contrast, investment in imported equipment

21

is fairly stable over time. Furthermore, transportation, electrical equipment and machinery, and electronic and telecommunications equipment industries have signicantly higher investment in imported equipment than the chemical industry. However, investment in domestic equipment does not appear to vary substantially across sectors. To summarize, we nd that although the economic reform pushed rms to face more competition in product markets, the government still exerted inuences on rm investment decisions through project nancing. In particular, easy access to cheap bank credit encouraged state-owned rms to adopt bigger projects and invest more in imported equipment. However, access to bank credits was neutral with respect to decisions of collectively-owned rms. At the same time, we also nd that rms with more xed assets and higher output prior to the project adopt bigger projects; more protable rms invest more in imported equipment while labor-intensive rms invest more in domestic equipment. Thus, technical capacity also plays a role in aecting decisions about technology investment.

Technology Absorption

To evaluate project performance, we mainly exploit two measures project protability (i.e., a ratio of gross project prots to project size) and capacity utilization rates. Our measure of project protability corresponds to the i in the theoretical model. Note that both performance measures will be a product of the technical capabilities of the rm, and the commercial soundness of the project. Empirically, it is often hard to separate the technological side from the commercial side. The survey provides us a panel of data on project gross prots and capacity utilization rates between 1987 and 1995. Since rms started their technology projects in dierent years, the panel is unbalanced. For 15 percent of the rms, we have 4 years of data on project performance and for 23 percent of the rms we have 5 years or more. This allows us to evaluate project performance over a 5-year horizon. 22

5.1

Project Protability

As suggested by proposition 2 in section 3, project protability is linked to the ability to access bank loans. The eect of bank loans works through their impact on technology choices. The discussion in section 3 also suggests that technical capacity has two opposing eects on project protability. Thus, the empirical specication is as follows:

it = ownershipi 1 + bi ownershipi 2 + i 3 + Zit 4 + it

(6)

where i indexes rms; t indexes years; it is project protability and measured by the ratio of project gross prots to project size; Zit include rm protability prior to the project (proxy for internal funds), and a vector of dummy variables controlling for sector, year and duration of the project; and it is the error term. The denitions of ownershipi , bi and i are the same as those in equation (5). Our main interest lies in the reduced-form eect of a rms ability to access bank loans on project protability, 2 . There is one caveat about our measure of project protability. For the projects that produce new products, project prots are well dened.17 However, some of the technology projects involve products that are related to the existing products produced by the rm. In this case it becomes dicult to measure accurately project prots. Thus, our measure of project protability is likely to have measurement error. However, if the measurement error of project protability is not systematically correlated with our measures of rm technical capacity and the ability to access bank loans, the measurement error in project protability should not bias our estimates since it can be absorbed by the error term in the regression.18
In our sample, 59 percent of state-owned rms, 58 percent of collectively-owned rms, and 69 percent of joint ventures have technology projects that involve new products. 18 We also estimate equation (6) controlling for whether the project involves new products. We nd no evidence that projects involving new products have signicantly higher or lower protability than the projects that involve products related to existing ones.
17

23

Pooled OLS estimates are displayed in columns 1-2 of Table 6. In column 1, the estimated coecient on the interaction between bank loans and state-owned rm is 2.30 with a t-statistic of 3.56, indicating that for state-owned rms a 1 percentage point increase in the share of expenditure that is nanced through bank loans is associated with a 2.30 percent decrease in project protability. This result lends support for the theoretical prediction that given technical capacity, rms with better access to cheap bank credits have lower project protability. At the same time, for collectively-owned rms the partial correlation between the ability to access bank loans and project protability is small and statistically insignificant. This result is consistent with the previous nding that access to bank loans is not signicantly correlated with project size for collectively-owned rms. These results further suggest that for collectively-owned rms, the cost of bank loans is close to that of internal funds so that access to bank loans should not have any signicant eect on investment and thus project performance. Furthermore, for joint ventures the access to bank loans is weakly and positively related to project protability. Therefore, we nd that unlike state-owned rms, bank loans do not appear to be a form of subsidy to either collectively-owned rms or joint ventures. Column 1 also shows that more protable and more capital-intensive rms have higher project protability. Note that rm protability may not only capture the availability of internal sources prior to the project, but also be related to managerial eectiveness. At the same time, the estimates of the coecients on project length indicate that project protability improves over the life of the project. In particular, project protability is approximately 1.24 times higher for projects with 5 (or above) years of operation than for start-up projects. This increase in protability is also statistically signicant. This result indicates that rms may improve their production eciency through learning by doing. The result may also be driven by increasing market share of a projects products.

24

In order to examine whether a higher share of imported equipment increases the cost of absorption and thus lowers project protability, in column 2 we add the share of expenditure on imported equipment.19 We nd that a 1 percentage point increase in the share of imported equipment in fact raises project protability by 0.77 percent, although this eect is not statistically signicant. The positive eect of investment in imported equipment reects the fact that although imported equipment may increase the diculty in absorbing technology and raise the demand for more expensive imported intermediate inputs, imported equipment may also allow a rm to produce better products and achieve larger market shares.20 This latter eect increases sales and prots, thus osetting the negative eect of higher absorption costs associated with imported equipment. Even though we have controlled for various rm attributes in columns 1 and 2, it is still possible that some unobserved rm eects are inuencing project protability. Since the share of bank loans and other rm attributes in our specication are time invariant, we use random-eects estimation. In this estimation, the unobserved rm eects are assumed to be uncorrelated with the ability to access bank loans and other rm attributes. Estimates are given in columns 3-4. The random-eect estimates are fairly similar to those from the pooled OLS regressions. The only major dierence is that the share of imported equipment becomes more signicantly correlated with project protability. To deal with the possible correlation between unobserved rm eects and other rm attributes, we also experiment with the Hausman-Taylor estimator.21 In this specication, we allow for the possibility that the unobserved rm eects are correlated with rm protability, size, physical capital intensity, and especially the ability to access bank
We also estimate equation (6) including the interaction between ownership and the share of expenditure on imported equipment. However, the estimated coecients do not dier signicantly across ownership types. Thus, we only report the results without the interaction. 20 Our data suggest that the expenditure share of imported equipment is positively correlated with the share of imported intermediate inputs after controlling for ownership and sector. On the other hand, there is a positive correlation between the expenditure share of imported equipment and the market share of project products (0.16 with a p-value of 0.05). 21 See Wooldridge (2002) pages 325-28.
19

25

loans. Results are displayed in columns 5-6. Most estimates become much less precise. However, as shown in column 5, our key results still hold qualitatively. In particular, state-owned rms with better access to bank loans have lower project protability. In contrast, for both collectively-owned rms and joint ventures, the estimated coecients on bank loans are positive. The eect of bank loans on project protability also diers signicantly between state-owned rms and joint ventures. (The statistic of the t-test is 4.46 with a p-value of 0.03.) In column 6, although the coecient on bank loans turns positive for state-owned rms, it is smaller than that for collectively-owned rms and much smaller than that for joint ventures.

5.2

Capacity Utilization

Another related measure of project performance is capacity utilization rates.22 We estimate equation (6) with i replaced by project capacity utilization rates. As shown in Table 7, the results are largely consistent with those in Table 6. We nd that state-owned rms with better access to cheap bank loans have signicantly lower capacity utilization rates. However, we nd no evidence that bank loans are signicantly associated with capacity utilization for collectively-owned rms and joint ventures. In addition, capacity utilization rates increase over the life of the projects. On the other hand, higher shares of imported equipment do not appear to reduce capacity utilization. Table 7 also shows that after controlling for other rm attributes and access to bank loans, state-owned rms have signicantly lower capacity utilization rates than collectively-owned rms. For example, in column 1 the coecient on state-owned rm is 0.18 (t-statistic = 2.86). On the other hand, rms with more xed assets prior to the project have signicantly higher capacity utilization. To summarize, we nd that state-owned rms with better access to cheap bank credits
The simple correlation between project protability and capacity utilization rates is 0.17 with a p-value of 0.001.
22

26

have relatively lower project protability and capacity utilization rates. This indicates that cheap bank credits induce state-owned rms to adopt too large projects given their technical capacity. Those bigger projects increase the costs of adopting and absorbing technology and thus reduce project protability and capacity utilization. On the other hand, rms which are more protable prior to the project experience higher project profitability. Firms with more xed assets prior to the project have signicantly higher project capacity utilization rates. Finally, project protability and capacity utilization improve with the length of the project.

Conclusions

Using a unique rm-level survey dataset collected by one of the authors, this paper addressed two questions central to the literature on technology diusion: what factors inuence technology adoption, and how well do rms absorb newly acquired technology? We focused on technology renovation projects, which were a key component of the modernization of Chinese rms during the last two decades. Fixed investment was largely organized around these projects. Between the mid-80s and early 90s, industrial reforms had still not progressed to the point at which investment decisions were fully decentralized to rms. Largely reecting the continued state control over the banking system, we nd that access to loans from Chinas state-owned banks was critical to mobilizing the resources required to carry out technology projects. With better access to bank loans than other rms, state-owned rms carried out signicantly larger projects and imported more technologically advanced equipment. On the other hand, we nd that project performance diers signicantly across rms. State-owned rms with better access to bank loans realized signicantly lower return to technology investment. Our explanation for this is that state-owned rms selected projects 27

that were too large and too technologically-advanced to absorb and operate eciently. Our empirical results have important implications for the role of nancial development in technology diusion and economic growth. In particular, distortions in the nancial sector likely generate a mismatch between technology and rms, resulting in under-utilized capacity with higher absorption costs and lower protability.

28

References
[1] Acemoglu, Daron, Philippe Aghion, and Fabrizio Zilibotti (2002), Distance to Frontier, Selection, and Economic Growth, NBER working paper #9066. [2] Amsden, Alice (1989), Asias New Giant, New York: Oxford University Press. [3] Brandt, Loren and Hongbin Li (2003), Bank Discrimination in Transition Economies: Ideology, Information, or Incentives? Journal of Comparative Economics, 387-413. [4] Brandt, Loren and Xiaodong Zhu (2000), Redistribution in a Decentralizing Economy: Growth and Ination in China under Reform, Journal of Political Economy, 422-439. [5] Byrd, William A. (1991), The Market Mechanism and Economic Reforms in China, Armonk, New York: M.E. Sharpe. [6] Calder, Kent (1993), Strategic Capitalism: Private Business and Public Purpose in Japanese Industrial Finance, Princeton, N.J.: Princeton University Press. [7] Caselli, Francesco and Daniel Wilson (2003), Importing Technology, NBER working paper #9928. [8] Cohen, Wesley M. and Richard C. Levin (1989), Empirical Studies of Innovation and Market Structure, in Richard Schmalensee and Robert D. Willig, eds., Handbook of Industrial Organization, Vol. II, Elsevier Science B.V., Amsterdam, the Netherlands. [9] De Long, J. Bradford and Lawrence H. Summers (1991), Equipment Investment and Economic Growth, Quarterly Journal of Economics, 445-502. [10] Dewatripont, M. and E. Maskin (1995), Credit and Eciency in Centralized and Decentralized Economies, Review of Economic Studies, 541-55. [11] Dunne, Timothy (1994), Plant Age and Technology Use in U.S. Manufacturing Industries, Rand Journal of Economics, 488-99. [12] Eaton, Jonathan and Samuel Kortum (2001), Trade in Capital Goods, European Economic Review, 1195-1235. [13] Evenson, Robert E. and Larry E. Westphal (1995), Technological Change and Technology Strategy, in Jere Bharman and T.T. Srinivasan, eds., Handbook of Development Economics, Vol. IIIA, Elsevier Science B.V., Amsterdam, the Netherlands. [14] Gerschenkron, Alexander (1962), Economic Backwardness in Historical Perspective, Harvard University Press, Cambridge MA. [15] Greenwood, Jeremy and Boyan Jovanovic (1990), Financial Development, Growth, and the Distribution of Income, Journal of Political Economy, 1076-1107. 29

[16] Groves, Theodore, Yongmiao Hong, John McMillan, and Barry Naughton (1994), Autonomy and Incentives in Chinese State Enterprises, Quarterly Journal of Economics, 183-209. [17] Hannan, Timothy H. and John M. McDowell (1984), The Determinants of Technology Adoption: The Case of the Banking Firm, Rand Journal of Economics, 328-35. [18] Jeerson, Gary H. and Thomas G. Rawski (1994), Enterprise Reform in Chinese Industry, Journal of Economic Perspective, 47-70. [19] Johnson, Simon, John McMillan and Christopher Woodru (2002), Property Rights and Finance, American Economic Review, 1335-56. [20] Jones, Charles (1994), Economic Growth and the Relative Price of Capital, Journal of Monetary Economics, 359-82. [21] Katz, Richard (1998), Japan: The System that Sourced, New York: M.E. Sharpe. [22] Lardy, Nicholas (1998), Chinas Unnished Revolution, Brookings Institution. [23] Lau, Lawrence, Yingyi Qian, and Gerard Roland (2000), Reform without Losers: An Interpretation of Chinas Dual-Track Approach to Transition, Journal of Political Economy, 120-43. [24] Levin, Sharon G., Stanford L. Levin, and John B. Meisel (1987), A Dynamic Analysis of the Adoption of a New Technology: The Case of Optical Scanners, Review of Economics and Statistics, 12-7. [25] Levine, Ross (1997), Financial Development and Economic Growth: Views and Agenda, Journal of Economic Literature, 688-726. [26] Li, Wei. (1997), The Impact of the Chinese Reform on the Performance of Chinese State-owned Enterprises, 1980-1989, Journal of Political Economy, 1080-1106. [27] Naughton, Barry (1995), Growing Out of Plan, Cambridge University Press. [28] Rajan, Raghuram G. and Luigi Zingales (1998), Financial Dependence and Growth, American Economic Review, 559-86. [29] Rose, Nancy L. and Paul L. Joskow (1990), The Diusion of New Technologies: Evidence from the Electric Utility Industry, Rand Journal of Economics, 354-72. [30] Shirley, Mary and Lixin Xu (2001), Empirical Eects of Performance Contracts: Evidence from China, Journal of Law, Economics and Organization, 168-200. [31] Shleifer, Andrei and Robert W. Vishny (1994), Politicians and Firms, Quarterly Journal of Economics, 995-1025. [32] Vishwasrao, Sharmila and William Bosshardt (2001), Foreign Ownership and Technology Adoption: Evidence from Indian Firms, Journal of Development Economics, 367-87. 30

[33] Wade, Robert (1990), Governing the Market: Economic Theory and the Role of Government in East Asia Industrialization, Princeton, N.J.: Princeton University Press. [34] Wooldridge, Jerey M. (2002), Econometric Analysis of Cross Section and Panel Data, Cambridge, MA: The MIT Press.

31

Table 1. Firm's Self-assessment about Operations and Market Environment

State-owned Firms A. Firm Objectives (%) Reducing production costs Achieving economies of scale through market expansion Entering export market Achieving economies of scope through product diversification B. Product Pricing (%) Market forces determine price Supervisory agency sets reference price Producer's association sets minimum price Supervisory agency sets mandatory price C. Factors Affecting Firm Profitability (%) Competition from domestic firms Availability of funds Inability to collect debt from clients Government interference D. Barriers to Investment (%) Insufficient internal funds Difficulty in getting bank loans Unprofitable investment opportunities Restrictions by supervisory agency Number of Firms 85 74 54 51 124 86 76 39 41 74 62 72 52 34 50 84 80 73 27 89 82 68 31 74 80 66 32 56 23 11 11 53 20 16 11 84 8 8 0 85 62 54 52 85 65 38 58 78 70 68 26 Collectivelyowned Firms Joint Ventures

Notes : This table reports the percentage of firms that select the items in the first column. For example, '62' in panel A under 'stateowned firms' indicates that 62 percent of state-owned firms consider 'achieving economies of scale through market expansion' as a top firm objective.

Table 2. Summary Statistics on Firm Attributes

Full Sample State-owned CollectivelyFirms owned Firms 66 19 15 30 70 20.2 (56.0) 425 (502) 61.9 (83.6) 40 (23) 28 (16) 15 (12) 0.5 (0.4) 8.1 (10.6) 124 10.6 (12.0) 74 32 (21) 0.6 (0.5) 27.0 (29.7) 50 30.1 (37.4) 1708 (1984) 21.0 (16.9) 42 (20) 36 (17) 23 (10) 0.4 (0.4) 7.5 (7.8) 103 28.3 (35.7) 1565 (1858) 22.7 (20.5) 43 (20) 35 (17) 23 (10) 0.4 (0.4) 27 (17) 59 (27) 23.2 (24.9) 264 (299) 4.5 (5.3) 26 36 38 69 22 9 30 42 28 4.7 (4.6) 308 (340) 20.7 (15.2) 59 (26) 27 (17) 14 (11) 0.5 (0.4) 11.1 (13.3) 50 Joint Venture State-owned Firms Collectivelyowned Firms Smaller Sample Joint Ventures

Period Established (%) Pre-1978 1979-89 1990-93

37 63 9.2 (8.5) 436 (502) 34.9 (34.2) 43 (26) 31 (17) 27 (20) 0.7 (0.5) 22.5 (35.2) 16

Firm Size Fixed Productive Assets (in million RMB)

Employment

Physical Capital Intensity Fixed Productive Assets Per Worker (in 1,000 RMB)

Human Capital (%) Junior High-School or Below

Senior High-School

College or Above

Firm Profitability Profits Per Fixed Productive Assets

Profits Per Worker (in 1,000 RMB)

Number of Firms

Notes : The smaller sample includes the firms which have information on firm employment, fixed assets, profits and output prior to the technology project. Firm size, physical capital intensity, and firm profitability are for 1993. Human capital is for 1994. Standard deviations are in parentheses.

Table 3. Summary Statistics on Technology Projects

State-owned Firms mean >0a Project Size Absolute size (in million RMB) Project size relative to fixed productive assetsb (%) Project Expenditure Breakdownc (%) Imported equipment Domestic equipment Construction of new facility Prototype development Sources of Project Financingc (%) Bank loans Internal funds Foreign investment Project Performanced Project gross profits relative to project size Project capacity utilization rate (%) Number of Firms 0.96 (1.56) 85.07 (16.07) 103 0.98 (1.29) 87.31 (11.44) 50 1.23 (1.90) 85.63 (16.86) 16 46 (32) 43 (31) 4 (14) 84 92 8 29 (31) 61 (34) 3 (14) 63 98 4 36 (44) 43 (40) 20 (34) 50 81 31 35 (32) 30 (29) 15 (19) 13 (20) 75 75 60 60 21 (30) 36 (30) 16 (21) 20 (24) 42 80 56 70 42 (43) 37 (40) 7 (13) 11 (23) 63 56 31 38 10.5 (20.8) 31 1.9 (2.4) 50 3.1 (4.6) 41 Collectively-owned Firms mean >0a

Joint Ventures mean >0a

Notes : This table reports summary statistics for the smaller sample that includes firms with information on firm employment, fixed assets, profits and output prior to the technology project. Standard deviations are in parentheses. a ) In the column of '>0', we report the percentage of firms for which the variable is greater than zero. For example, for collectively-owned firms, '42' in the row of 'imported equipment' indicates that 42 percent of collectively-owned firms invested in imported equipment. b ) The median values are reported. c ) Project expenditure also includes training and one-time technology transfer fees. Sources of project financing also come from equipment suppliers and others. Since these items are very small, they are not reported in the table. d ) Project profits and capacity utilization rates are for 1993.

Table 4. Project Size

(1) State-owned Firm Joint Venture Bank Loan*State-owned Firm Bank Loan*Collectively-owned Firm Bank Loan*Joint Venture Log(Firm Profits / Fixed Assets)0 Log(Firm Fixed Assets)0 Log(Firm Output)0 Log(Firm Employment)0 Log(Firm Fixed Assets / Employment)0 % of Senior High-school Graduates % of College Graduates Firm Established in the Period 1978-89 Firm Established in the 1990s R2 0.19 0.47 0.30 (1.91) 0.32 (2.08) 0.21 (1.18) -0.25 (-1.30) 0.00 (0.06) 0.01 (0.47) -0.08 (-0.28) 0.14 (0.38) 0.47 -0.25 (-1.29) 0.00 (0.01) 0.01 (0.45) -0.09 (-0.30) 0.14 (0.36) 0.47 -0.26 (-1.36) -0.00 (-0.50) 0.01 (0.68) -0.01 (-0.02) 0.32 (0.90) 0.52 -0.29 (-1.59) -0.00 (-0.15) 0.01 (0.62) 0.01 (0.02) 0.37 (1.04) 0.55 0.49 (2.55) 0.33 (2.07) 0.03 (0.28) 0.52 (2.17) 0.31 (1.64) 0.86 (2.97) 0.44 (1.03) (2) -0.26 (-0.94) -0.14 (-0.43) (3) -0.29 (-0.96) -0.25 (-0.71) (4) -0.29 (-0.97) -0.25 (-0.72) Benchmark (5) -0.32 (-1.05) -0.38 (-1.10) 1.38 (2.52) 0.40 (0.54) -1.50 (-2.45) 0.02 (0.18) 0.53 (2.27) 0.28 (1.55) (6) -0.47 (-1.60) -0.56 (-1.55) 1.68 (4.36) 0.07 (0.18) -1.05 (-1.90) -0.09 (-0.89) 0.44 (1.95) 0.42 (2.48)

Notes : This table reports estimates of equation (5). The dependent variable is the logarithm of project expenditure. All specifications include controls for firm sector and the year when the technology project started. There are 169 observations. The omitted category for ownership is collectively-owned firms. 'Bank loan' in column 5 is the fraction of project expenditure that is financed through bank loans, while in column 6 it is a dummy variable which equals 1 if the firm used bank loans to finance the project. Firm profits, fixed assets, output and employment are prior to the project. '% of senior high-school graduates' and '% of college graduates' represent the percentage of workers who have senior high-school diploma and college degree or above, respectively. The data on worker's education are for 1994. The omitted category is workers with a junior high-school diploma or below. The omitted category for firm age is firms which established before 1978. In parentheses are t -statistics. Standard errors are robust to heteroscedasticity.

Table 5. Investment in Imported Equipment versus Domestic Equipment

OLSa imported (1) State-owned Firm Joint Venture Bank Loan*State-owned Firm Bank Loan*Collectively-owned Firm Bank Loan*Joint Venture Log(Firm Profits / Fixed Assets)0 Log(Firm Fixed Assets)0 Log(Firm Output)0 Log(Firm Fixed Assets / Employment)0 % of Senior High-school Graduates % of College Graduates Firm Established in the Period 1978-89 Firm Established in the 1990s R2 Log Likelihood 0.43 (0.86) -0.28 (-0.37) 2.13 (2.99) 0.20 (0.19) -2.23 (-1.56) 0.42 (2.34) 0.95 (2.58) -0.05 (-0.16) -0.01 (-0.04) -0.00 (-0.30) 0.01 (0.78) -0.35 (-0.72) 0.53 (0.90) 0.45 domestic (2) -0.51 (-1.10) -0.31 (-0.44) 0.07 (0.10) -0.28 (-0.26) 1.83 (1.38) 0.07 (0.43) 0.60 (1.76) 0.12 (0.44) -0.73 (-2.43) 0.01 (0.74) -0.00 (-0.22) -0.15 (-0.30) 0.13 (0.21) 0.31 -309.90 -334.20 -88.95 imported (3) 0.59 (0.77) -0.27 (-0.24) 2.23 (2.33) 0.66 (0.36) -3.04 (-1.39) 0.62 (2.37) 1.53 (2.64) -0.24 (-0.52) 0.11 (0.29) 0.00 (0.04) 0.03 (1.13) -0.56 (-0.82) 0.79 (0.95) Tobitb domestic (4) -0.79 (-1.62) -0.73 (-0.75) -0.22 (-0.21) -0.06 (-0.04) 2.87 (1.60) 0.12 (0.56) 0.69 (1.72) 0.15 (0.49) -0.98 (-2.62) 0.01 (1.17) -0.01 (-0.42) -0.22 (-0.39) 0.01 (0.01) imported (5) 0.05 (0.50) 0.02 (0.14) 0.17 (1.30) 0.21 (0.93) -0.37 (-1.09) 0.07 (2.32) 0.10 (1.32) -0.01 (-0.18) 0.14 (2.47) 0.00 (0.01) 0.00 (1.58) -0.10 (-1.13) 0.02 (0.14)

Notes : This table examines investment in imported and domestic equipment. All specifications include controls for firm age, sector and the year when the technology project started. There are 169 observations. In parentheses are t statistics. Standard errors are robust to heteroscedasticity. See the notes to Table 4 for more detail about other variables. a ) The dependent variables in columns 1-2 are the logarithm of expenditure on imported equipment plus 1, and the logarithm of expenditure on domestic equipment plus 1, respectively. b ) Columns 3-5 report the left-censored Tobit estimates. 61 firms did not import equipment and 43 firms did not purchase domestic equipment. The dependent variables in columns 3-4 are the same as those in columns 1-2 correspondingly. The dependent variable in column 5 is the share of expenditure on imported equipment.

Table 6. Project Profitability

Pooled OLS (1) (2) State-owned Firm Joint Venture Bank Loan*State-owned Firm Bank Loan*Collectively-owned Firm Bank Loan*Joint Venture Log(Firm Profits / Fixed Assets)0 Log(Firm Fixed Assets)0 Log(Firm Output)0 Log(Firm Fixed Assets / Employment)0 2nd Year of Project 3rd Year of Project 4th Year of Project 5th Year of Project Expenditure Share of Imported Equipment R2 Wald
2

Random-Effects (3) (4) 0.06 (0.16) -0.44 (-0.83) -2.31 (-4.30) -0.18 (-0.22) 0.58 (0.54) 0.19 (1.45) -0.15 (-0.56) -0.14 (-0.68) 0.57 (2.49) 0.28 (3.54) 0.43 (2.63) 0.55 (2.33) 0.91 (2.68) 0.01 (0.03) -0.50 (-0.96) -2.51 (-4.67) -0.21 (-0.26) 0.81 (0.75) 0.16 (1.17) -0.18 (-0.66) -0.15 (-0.72) 0.46 (2.00) 0.28 (3.55) 0.43 (2.66) 0.55 (2.35) 0.92 (2.72) 0.97 (2.14) 0.23 0.26

Hausman-Taylor (5) (6) 0.12 (0.13) 0.02 (0.02) -3.41 (-0.43) 4.93 (1.05) 13.22 (1.00) 0.12 (0.12) -0.79 (-0.75) 0.00 (0.13) 1.74 (1.29) 0.27 (1.79) 0.38 (1.36) 0.46 (1.11) 0.75 (1.21) 0.91 (0.50) 1.66 (0.49) 3.38 (0.21) 6.95 (0.84) 23.73 (0.92) -0.20 (-0.11) -1.52 (-0.76) -0.00 (-0.08) 3.38 (1.09) 0.23 (1.30) 0.30 (0.89) 0.35 (0.69) 0.57 (0.76) -6.64 (-0.78)

-0.11 (-0.32) -0.28 (-0.50) -2.30 (-3.56) -0.22 (-0.33) 0.94 (0.76) 0.20 (1.51) -0.01 (-0.04) -0.18 (-0.95) 0.48 (1.91) 0.33 (3.97) 0.44 (1.91) 0.49 (1.58) 1.24 (2.99)

0.15 (0.43) -0.35 (-0.67) -2.50 (-3.73) -0.21 (-0.31) 1.06 (0.92) 0.16 (1.23) -0.03 (-0.13) -0.18 (-0.95) 0.40 (1.56) 0.32 (3.91) 0.44 (1.92) 0.48 (1.52) 1.24 (3.05) 0.77 (1.37)

0.24

0.27

180.10

145.10

Notes : This table reports the estimates of equation (6). The dependent variable is the logarithm of project gross profits relative to project size. All specifications include controls for human capital, firm age, sector and year. '2nd year of project' is a dummy variable which equals to 1 if the data are for a project in its 2nd year. The omitted category for the duration of project is the project's 1st year. See the notes to Table 4 for more detail about other variables. There are 398 observations. t -statistics are in parentheses. In columns 1-2, standard errors are robust to heteroscedasticity and serial correlation.

Table 7. Project Capacity Utilization

Pooled OLS (1) (2) State-owned Firm Joint Venture Bank Loan*State-owned Firm Bank Loan*Collectively-owned Firm Bank Loan*Joint Venture Log(Firm Profits / Fixed Assets)0 Log(Firm Fixed Assets)0 Log(Firm Output)0 Log(Firm Fixed Assets / Employment)0 2nd Year of Project 3rd Year of Project 4th Year of Project 5th Year of Project Expenditure Share of Imported Equipment R2 Wald
2

Random-Effects (3) (4) -0.17 (-3.01) -0.07 (-0.88) -0.20 (-2.33) 0.02 (0.17) 0.07 (0.38) 0.03 (1.52) 0.10 (2.25) -0.06 (-1.68) -0.01 (-0.19) 0.07 (3.59) 0.12 (3.20) 0.13 (2.58) 0.16 (2.35) -0.17 (-3.01) -0.09 (-1.06) -0.21 (-2.51) 0.03 (0.23) 0.10 (0.59) 0.03 (1.21) 0.09 (2.04) -0.05 (-1.54) -0.02 (-0.41) 0.07 (3.59) 0.12 (3.20) 0.13 (2.57) 0.16 (2.35) 0.10 (1.35) 0.28 0.29

Hausman-Taylor (5) (6) -0.39 (-0.75) -0.49 (-0.39) -1.79 (-0.29) 0.66 (0.27) -0.18 (-0.04) 0.19 (0.20) 0.27 (0.31) 0.00 (0.58) 0.05 (0.04) 0.11 (2.82) 0.21 (2.82) 0.26 (2.36) 0.34 (2.06) -0.40 (-0.76) -0.54 (-0.41) -1.50 (-0.23) 0.63 (0.26) -0.27 (-0.06) 0.13 (0.12) 0.24 (0.27) 0.00 (0.62) 0.03 (0.03) 0.11 (2.86) 0.21 (2.86) 0.26 (2.40) 0.35 (2.10) 0.42 (0.21)

-0.18 (-2.86) -0.09 (-0.76) -0.26 (-2.10) -0.01 (-0.14) 0.10 (0.50) 0.04 (1.66) 0.14 (1.94) -0.09 (-1.64) -0.01 (-0.33) 0.07 (3.23) 0.09 (2.54) 0.11 (2.15) 0.19 (3.75)

-0.18 (-2.92) -0.10 (-0.85) -0.28 (-2.33) -0.00 (-0.04) 0.13 (0.58) 0.04 (1.34) 0.13 (1.66) -0.08 (-1.44) -0.02 (-0.52) 0.07 (3.23) 0.09 (2.63) 0.11 (2.15) 0.19 (3.69) 0.08 (0.85)

0.29

0.30

137.32

141.67

Notes : This table examines project capacity utilization. The dependent variable is the logarithm of project capacity utilization rates. All specifications include controls for human capital, firm age, sector and year. See the notes to Tables 4 and 6 for more detail about other variables. There are 346 observations. t -statistics are in parentheses. In columns 1-2 standard errors are robust to heteroscedasticity and serial correlation.