China vs. India: a microeconomic look at comparative macroeconomic performance

Taye Mengistae, Lixin Colin Xu, Bernard YeungMay 2006

Abstract

In comparable random samples of manufacturing businesses drawn from the two countries, Chinese establishments are found to have higher total factor productivity on the average than their Indian counterparts. Controlling for initial size, age, and line of industry, the average employment growth rate is higher for Indian establishments. Chinese plants grow faster in value added terms, nonetheless. This is mainly because the average net investment rate in fixed assets is higher in Chinese businesses. To a lesser extent, it is also because productivity grows faster on the average in Chinese plants. Partly because of this, the aggregate productivity growth rate that we compute industry by industry is larger for the Chinese sample. A second reason why the aggregate productivity growth rate is higher for the China sample is that allocative efficiency gains are larger in Chinese industry. By this we mean that market shares are reallocated from less productive plants to the more productive more rapidly (or steeply) in the Chinese sample. This is consistent with another finding: catch up effects and life cycle effects in productivity and growth, are stronger in the Chinese sample than in the Indian sample. Lastly, such key elements of the business climate as labor market flexibility and access to finance are major sources of the productivity and growth gaps between Chinese and Indian plants. If nothing else mattered, the average Chinese businesses would be more productive and would grow faster than its Indian counterparts on account of business climate differences between the two countries. This is not so much because business climate indicators are better in China than in India as because the marginal return to improvements in indicators is higher in China.

1

If [by the time China’s saving rates start to fall] India has completed the second generation reforms, built up its infrastructure and fully integrated itself into the world economy, we might find that the tortoise overtakes the hare… This race between the two Asian giants is set to be the most dramatic event of this century. (Deepak Lal in Business Standard, March 15, 2005)

1. Introduction

China’s and India’s are among the largest economies in the world today. They have also been among the fastest growing over the last two decades and a half. They both entered the 1980s at comparable levels of per capita income following three decades of growth-China at an average rate of 4.4 percent per annum, and India at a rate of 3.75 percent (Srinivasan, 2003). 1 Since then China’s economy has taken off to a state of unprecedented growth that averaged 10.1 percent per annum in the 1980s, 10.3 per cent per annum in the 1990s, and has yet to show any sign of slowing down. India’s GDP growth has also picked up to an averaged 5.6 per cent a year in the 1980s, 6 percent per annum in the 1990s’, and even higher since. Although India’s growth rate has been remarkably high by any standard, the sustained growth gap between the two countries has intrigued observers, especially given what seemed to be significant similarities in their initial conditions. According to Srinivasan (2003), India’s GDP per capita stood at 853 in 1990 international dollars in 1973 as compared to China’ 839. The divergence in growth rates since then has created a widening income gap in China’s favor, which stood at 3,117 dollars versus 1746 dollars by 1998 (Srinivansan, 2003). Figure 1 shows the evolution of the gap in Purchasing Power Parity terms computed from data in the World Bank’s World Development Indicators.

In this paper we analyze data from comparable samples of manufacturing businesses drawn from the two countries in order to help shade light on two complementary questions: Why is per capita income so much higher today in China than in India? And why is China’s GDP growing so much faster? One hypothesis is that China’s better performance on both scores reflects differences in the quality of institutions or in the immediate policy environment in which businesses operate. Another is that the contrast is partly a consequence of China’s earlier investments in superior physical infrastructure paying off. These are no doubt macro economic issues in the investigation of which the analysis of available national aggregate data has yet to be brought to bear. At the same time a key limitation of aggregate analysis has to be recognized in this particular context. This is that, at this stage, available time series are bound to be too short on key variables for problems of econometric identification to be resolved satisfactorily based solely on the observation of cross-country differences in national or sector aggregates. To this should be added what seems to be widespread skepticism about the comparability of China’s national account aggregates with India’s.2

1 The growth rate figures reported in the following lines are also from Table 3 of the same paper by Srinivasan.2 Although there is solid consensus that China’s economy has grown much faster than India’s since the early eighties, many suspect that China’s growth rates might have been significantly overstated relative to India’s due to discrepancies in national accounting conventions. See , for example, Srinivasan (2003) and Young (2003).

2

Part of the remedy to this should be the exploitation of sub national variation in economic performance and its determinants as an additional means of identification. An obvious instance or component of such a strategy of is the analysis of firm level data, which are regularly generated in both countries by a variety of agencies. The data on the analysis of which we report in this paper come from business surveys that the World Bank sponsored in the two countries in 2003. The India survey covered 1860 manufacturing establishments sampled from the country’s top 40 industrial cities and its major exporting industries. The China survey covered 2400 enterprises sampled from 18 cities and 5 regions, and a wider set of industries including most of those covered by the India survey. Both surveys include production, employment and investment data on each business on annual basis for the three years leading to the year of survey. This is in addition to data on the local business and policy environment of each establishment as of the survey year, including indicators of the quality of the financial, regulatory, infrastructural, and labor market settings in which it operated at the time of the survey.

What does information of this kind have to do with the (macro economic) questions of growth and development we just raised? Per capita income is probably used far more often as indicator of wellbeing than anything else, but one obvious interpretation of it is also as an index aggregate labor productivity, as is the case in, for example, Hall and Jones (1999). Thus the fact that it is presently twice in China of what it is India means that China’s labor productivity is at least higher than India’s. In general this should mean that output per worker is greater in the average Chinese firm than it is in its Indian counterpart, either because production is more capital intensive in the Chinese firm, or because total factor productivity is higher, or both. Likewise, China’s higher GDP growth rates should be reflected in faster growth of the average Chinese firm or faster allocative productivity gains in China’s industries.3 Like its aggregate analogue, growth at the firm level can only originate in one of two sources, namely, growth in factor inputs, and growth in their productivity. If the average Chinese firm is indeed growing faster than the average Indian firm, then it must be investing at a higher rate in physical or human capital, or its net job creation rate must be higher, or it must have greater total factor productivity growth.

Our analysis is focused on two issues. The first concerns whether or not the performances of Chinese and Indian firms differ significantly in terms of productivity and growth, as should be expected from the macro-economic performance contrast between the two countries. Secondly, assuming that such differences do exist, how far can they be attributed to differences in “business environment”? The first issue can be broken down into a series of subsidiary questions the answers to which describe the linkages between business climate and firm level determinants of aggregate productivity and growth. These include, first, whether or not there is significant productivity gap between Chinese and Indian firms as the per capita income gap between the two countries suggests. Secondly, do Chinese firms grow faster as should be expected from China’s better GDP growth performance? Third, assuming they do,

3 This is strictly true if the structure of production is the same in the two countries and the number of firms is fixed within each industry in each country, or the equilibrium size distribution of firms within each industry does not vary across countries. We are abstracting here from the aggregation problems that arise due to the failure of these assumptions in practice.

3

what are the proximate sources of their faster growth: is it that they invest at a higher rate, or that they are getting more efficient in factor use more quickly, or some combination of both?

Since the data we analyze here are entirely on manufacturing firms, our answers to these questions are most pertinent to the comparative performance of the manufacturing sectors of the two economies. However, given the weight of manufacturing in each economy, and given that China has done particularly well in this sector compared to India, knowledge of the factors behind the contrast between the performance of Chinese manufacturers and their Indian counterparts should help us better understand of the relative performance of the broader national economies. In the context of manufacturing, the projection of the performance indicators of the average firm to its aggregate analogues would be strictly valid only on two assumptions. One is that the structure of manufacturing production is the same between the two countries. The second is that the equilibrium size distribution of plants within individual industries is the same for both countries. Since we are working with samples of observations from the selected industries rather than census data on all sectors, we have no way of testing either of these assumptions. We have nonetheless sought to make our conclusion robust to the possible failure of the first assumption by confining our comparison of businesses to industries that are common to both countries.

The actual size distribution of businesses could vary between the two countries in any of the industries from which our data are drawn as result of policy induced distortions, or as a consequence of differences in the stages of industry evolution observed at the time of the surveys. This in turn should drive a wedge between the (sample) average firm’s performance we observe and the aggregate performance we ultimately care about-that is, between (sample) average firm level productivity and (aggregate) industry productivity, on the one hand, and between the average firm growth rate and the industry growth rate, on the other. In order to eliminate this distortion we compute mean firm level performance indicators conditional on firm size and firm age. By helping us to control for differences in catch up and life cycle effects stemming from differences in the stages of industry evolution between the two countries, this should help us get at true industry effects in performance gaps. In addition we computed market share weighted (sample) mean levels and growth rates of productivity in order to separate the dynamics of firm level productivity from intra-industry reallocation effects on aggregate productivity, which, together with the relative strength of catch up and life cycle effects, provide a picture of the comparative dynamism of industry in the two countries.

To highlight our main results, we find that output per worker is higher for the China sample than for India sample. This is in part because the average Chinese plant is more capital intensive. It is partly because total factor productivity is higher in for the China sample. The average Chinese establishment is also about the same age as its Indian counterpart, but much larger by all three measures of scale, that is , sales revenue, fixed assets and employment. This is consistent with a second set of results, namely, that output and fixed assets growth rates are higher for the Chinese sample than for the Indian sample, while employment growth rates are higher for the Indian

4

sample.4 Third, of the two sources of the growth advantage of Chinese sample, higher rate of investment is by far the more important. It accounts for more than four times the growth advantage explained by faster TFP growth. China’s faster productivity growth at the firm level has meant that the growth rate of aggregate (or industry level) productivity has been higher for Chinese sample. This effect of on the growth rate of aggregate productivity has been reinforced by allocative efficiency being higher in the China sample. Consistent with this catch up and life cycle effects are found to be stronger in the Chinese sample.

To investigate how far differences in business environment could explain the first three sets of results we estimate various firm performance equations. The main business climate influences in the TFP gap between Chinese and Indian firms are differences labor market flexibility, in access to finance, and in levels of skill and technology. Differences in access to finance and in skills and technology are also powerful influences in the growth performance gap between the two groups. This finding is consistent with results of other cross-country firm level studies based on the World Bank’s investment climate surveys including Dollar et al. (2005), Eifert et al. (2005)…. A novelty of our estimation strategy compared to existing work is that we allow for the possibility that the marginal effects of individual elements of business climate could vary between the two countries even if all business climate indicators had assumed the same values in both countries. . It turns out that while the better performance of Chinese firms in our sample is partly on account of “better business environment”, this less because China’s business climate indicators are better than India’s than because the marginal effect of a better business climate on firm productivity or on firm growth is higher in China.

The rest of the paper is organized as follows. We lay out the empirical framework of our analysis in the next section. We discuss our data in Section 3 along with the econometric issues arising from them. Our findings are reported in detail in Section 4. We conclude in Section 5.

2. Empirical Framework

In order to address the questions we have posed we have carried out three distinct analytic tasks. The first of these concerns the measurement of relative performance of Chinese and Indian firms. The second involves accounting for observed performance gaps between the two groups of firms, in terms of their proximate causes (or components). The third task is one of explaining the gaps in the sense of identifying their ultimate causes, of which one set, we hypothesize, is business environment.

Measuring and accounting for performance gaps

Our basic measures of performance are plant level productivity and the plant level rate of output growth. The reason we have chosen these particular indicators is that they are the micro-economic analogues of the two main indicators of aggregate

4 If the growth pattern observed across the two underlying populations has been going on for some time, it is plausible that the World Bank surveys capture cross-sections in which the average Chinese firm is larger and more capital intensive.

5

performance, namely, per capita income and the rate of GDP growth. Knowledge o the factors behind the relative productivity and growth performance Chinese and Indian producers should help us understand as to why per capita income and the GDP growth rate are higher in China than in India.

Our approach to accounting for performance gaps is also strictly analogues to established practices in aggregate growth accounting and accounting for cross-country aggregate productivity gaps. Although the latter is relatively uncommon, it is levels analogue of growth accounting and, as shown in Hall and Jones (1999), no less illuminating a tool of analysis of international inequality. In adapting it to our setting, a natural plant level analogue for per capita income is output per worker. The problem with this particular productivity measure is that it is a function of factor proportions, which, depending on relative factor prices could vary between Chinese firms and Indian firms even when they produce an identical product. As a rule output per worker should be higher where capital input per worker is also higher. Since there is nothing inherently good or bad about greater capital intensity from the point of view of efficiency in resource use at the firm level, however, a meaningful comparison of labor productivity between groups of firms should make allowances for possible differences in factor proportions. For this reason we use as our productivity measure total factor productivity (TFP), rather than output per worker.

As already pointed out, the fact that Chinese firms are growing faster on the average than Indian firms in our dataset poses the question of whether this is because they are investing at a higher rate, or because they are getting ever more productive. This was the very question that Young (1994, 1995) raised in the context of the exceptional growth performance of what were then known as the “Newly Industrialized Countries” of Hong Kong, Singapore, South Korea, and Taiwan. The answer Young provided then was that, contrary to what then seemed to be the conventional wisdom, faster accumulation and “static neoclassical gains from sectoral reallocation”, rather than rapid TFP growth accounted for “the lion’s share of the East Asian growth miracle.” Young (2003) draws more or less the same conclusion about the growth performance of the Chinese economy over the period 1978-1998: in China too, TFP growth was a far less important source of GDP growth than accumulation and gains from the reallocation of manpower and capital between sectors. Our results basically confirm this latter finding based on firm level data in as far as investment in fixed assets turns out to be far more important than TFP growth as a source of the growth advantage of Chinese firms over Indian firms in our sample.5

Firm level productivity, allocative efficiency gains, and aggregate productivity growth

That said higher productivity growth remains to be one source of the growth advantage of Chinese firms over their Indian counterparts. And while we lack the data to test Young’s other hypothesis that inter-sector reallocation of manpower and

5 The view that China’s faster growth rate largely reflects its higher investment rates more than anything else appears to command consensus. See, for example, Srinivasan (2003) and Martin and Manole (2004). The only contested point here seems to be whether or not productivity growth has played a significant role in the growth gap between the two countries (Srinivasan, 2003).

6

capital has been a significant source of China’s GDP growth, we do find that intra industry allocation of market share from less productive firms to the more productive has been a more important source of productivity growth in the China sample than in the India sample our dataset. We draw this conclusion based on an Olley-Pakes cross-sectional decomposition of industry level productivity growth for each sample.6 The decomposition expresses the average productivity, , of a given industry in year as the weighted mean of establishment level productivities, , with establishment market share, , as weights where indexes establishments. The decomposition can alternatively be written as

(1) , where

letters with upper bars represent unweighted industry means of variables. In other words the industry-level average productivity is the sum of the (unweighted) average of establishment level productivity and the sample covariance between establishment productivity and market share. A positive covariance term implies that more productive firms have higher market shares. Considering changes over time, this means that it is not necessary that increases for average industry productivity to grow. can increase even in the absence of significant changes in as a result of the reallocation of market share in favor of more productive firms, which measures of industry deregulation or market liberalization are often found to lead to. In practice actual change in industry productivity is often a result of a bit of both, and one objective of our analysis has been to assess the relative weight of the two elements as potential sources of the productivity growth gaps we observed between our China and India samples.

Explaining performance gaps: the role of business climate

Output per worker is higher in the average Chinese firm than in its Indian counterpart for two reasons. One is because Chinese plants have more capital per worker. The other is because their total factor productivity is higher. Likewise Chinese firms grow faster on the average in terms of output partly because they invest at a higher rate and partly because of their faster TFP growth. But why are TFP and its rate of growth higher for the average Chinese plant? And why do Chinese firms invest at higher rates? A popular hypothesis is that part of the answer lies in China’s allegedly superior physical infrastructure. A second common hypothesis is that other aspects of the business environment significantly differ between the two countries, and have on balance influenced economic outcomes in China’s favor. Differences seem to be particularly pronounced between the two sets of firms in terms of access to finance, labor market flexibility, the predictability of the regulatory environment and the level of skills and technology. In order to test these hypotheses we have estimated performance equation on each country dataset whereby the performance, , of firm , is assumed to depend on a vector of business climate variables, , a vector of firm level controls and industry

6 This is set out in Olley and Pakes (1996), and is discussed in Foster, Haltiwanger, and Krizan (2001) in the context of the wider literature on productivity decomposition.

7

characteristics, , and an iid random error term, summing up a set of unobservable influences assumed to be orthogonal to firm characteristics and business environment. The performance equations we have estimated are of the form (2)

where indexes China or India, are constants, and is random error terms assumed to be iid and orthogonal to and

. It is possible, but not necessarily the case that the coefficients of the two versions of equation (2) are identical. In particular, it could be that , meaning that the various elements of business environment we just listed would have the same marginal effect in China as India, or improvements in business climate would have the same marginal “rate of return” in both countries. In that case the average performance gap between the two countries on account of differences in business climate would simply be the differences in business climate indicators scaled up by the common marginal rate of return. However, it is also possible that marginal rates of return are different for the two countries. We have therefore chosen to treat the equality of and as a testable proposition rather than an assumption of our analysis. Because we find that the two sets of coefficients are in fact different, we conclude that there would always be performance gaps between Chinese and Indian firms for business climate reasons even if even if all business climate indictors assumed identical values in the two countries. Let be the amount by which the mean performance of indicator of Chinese firms exceeds that of Indian firms on account of differences with respect of business environment indicator . Then we have

(3) ,

where is the coefficient of the kth business climate indicator in the performance equation and is the mean value of in the indicated country. This is an Oaxaca-Blinder decomposition of the performance gap into the “endowment effect” of the fact that has different values in China and India, and the “rate of return effect” of the fact that the marginal effect of would be different in China from what it is in India even when the indicator assumes the same values in the two counties.7

The effects of firm and industry characteristics on the performance gaps can likewise be decomposed into “endowment” and “rate of return” effects.8

7 The Oaxaca-Blinder decomposition was introduced to the literature by Oaxaca (1973) and Blinder (1973) in the early 1970’s as technique for accounting for wage gaps between labor market groups and sectors. As a purely statistical technique it clearly has a much wider relevance than the analysis of labor market earnings.8 Let and , so that we can write the performance equation (2) as

. The Oaxaca –Blinder decomposition of the total gap between the mean performance indicator of Chinese firms and Indian firms would be given by

.

8

3. Data and econometric issues

Survey data sources

Although the World Bank Surveys from which our data are drawn were quite similar in sample design and survey instrument, they were planned and executed independently and had significant differences in both respects. The India survey covered 1860 manufacturing establishments, sampled from the top 40 industrial cities in the country. The forty cities were selected from 12 of the largest 15 states by picking the largest 3 or 4 industrial centers from each state. These 12 states were Andhra Pradesh, Delhi, Gujarat, Karnataka, Kerala, Haryana, Maharashtra, Madhya Pradesh, Punjab, Tamil Nadu, Uttar Pradesh, and West Bengal. Between them the 12 states account for well over 90 percent of India’s industrial GDP. The 3 or 4 cities covered in each state also accounted for the bulk of manufacturing outputs of their respective states. In each city, samples were drawn from the main exporting or import competing manufacturing industries, namely: food processing, textiles, garments and leather goods, chemicals and pharmaceuticals, household electronics, electrical equipment and parts, auto and parts, metallurgical products and tools. The total sample was allocated between states in proportion to state’s share in the national employment total of the eight industries. Each state’s allocation was then drawn by systematic sampling from a consolidated list of registered firms employing at least 10 workers and belonging to one of the eight industries. The list was restricted to the selected 3 to 4 industrial cities within each state and sorted by ascending employment size. The systematic sampling rule set an establishment’s probability of selection proportional to the establishment’s number of employees. In the China survey enterprises were sampled from 18 cities considered to be representative of five regions. The cities were: (1) Benxi, Dalian, Changchun, and Haerbin, from the Northeast Region; (2) Hangzhou, Wenzhou, Shenzhen, and Jiangmen, from the Coastal Region; (3) Nanchang, Zhenzhou, Wuhan, and Changsha, from the Central Region; (4) Nanning, Guiyang, Chongqing, and Kunming from the Southwest Region; and (5) Xi’an and Langzhou.from the Northwest Region. Each of these cities was allotted a sample size of either 100 or 150 firms. These were randomly drawn from an electronic database of firms according to several criteria. Unlike the India survey, the China survey covered firms from manufacturing as well as service industries. As in the India survey, the sampling frame in China was also restricted to businesses the employment size of which was above a minimum cut off point. In the Chinese case the cut off employment level was set at 20 workers for manufacturing industries and 15 employees for service industries.

Firm characteristics: industry, size, and age profiles

To ensure comparability, our analysis is confined to manufacturing firms drawn from industries that were covered both by the India survey and the China survey. As a result we have excluded textiles producers from the India sample and all service sector establishments and producers of transport equipment from the China sample. The businesses on which we actually analyze data are consequently 1565 firms from China and 1735 firms in India. The distribution of these by industry is shown in Table 1. Summary statistics of the variable used in our analysis are given in Table 2.

9

The average business age is about 16 years in both samples. However, the median age is significantly lower for the China sample. The average Chinese business is also several times larger than its Indian counterpart by all three measures, namely, number of employees, book value of fixed assets, and annual value added. Both surveys provide information on all these and other financial indicators for a period of 2 or 3 years for most establishments. We have therefore been able to measure the growth performance of most businesses in the dataset in terms of all three measures of size. We have also been able to estimate for most businesses total factor productivity in levels as well as its annual growth for a period of up to two years.

Measuring growth and productivity

As one of our two primary performance indicators, the rate of growth of output is measured here as the log difference in annual value added. Likewise the employment growth rate is the log difference in the annual average number of employees. The rate of growth of capital stock is the log difference between the book value of fixed assets at the end of a fiscal year from that at the beginning of the year. We define the growth rate of productivity as the annual log difference in total factor productivity.

We measure total factor productivity itself as the amount by which the actual output of a plant or establishment exceeds or falls short of a counterfactual output on a reference technology given its actual inputs. We estimate this technology for each of the seven industries listed in Table 1 using our sample observations based on the assumption that, in each industry, output is Cobb-Douglas in capital services and labor input. We define output as annual value added, in constant US dollars. We proxy capital services by the constant dollar book value of fixed assets at the end of the fiscal year, . Labor input is measured by the average number of employees during a fiscal year, . We define the reference technology as

(4)

where is the TFP term, and and are constant factor share parameters. We have estimated (log) TFP as , where and are estimates of and obtained by fitting (4) to our data. A major concern, in obtaining the estimates and

is that is a state variable that influences input choices: inherently more productive firms could employ more resources. Consistent estimates of the factor share parameters cannot therefore be obtained by applying OLS to (4).9 To address this problem we have used the Levinsohn- Petrin estimator (Levinsohn and Petrin, 2003), which is in fact the only feasible solution to the problem of input endogeneity given our data. This is essentially a two step procedure in the fist step of which we use material inputs as proxy for exogenous variation in productivity to consistently estimate the share of labor (as the non-state input). The estimate so obtained is then used to consistently estimate the share

9 The main references on this problem and proposed solutions are Olley and Pakes (1996), Blundell and Bond (2000), and Levinsohn and Petrin (2003). See also Ackerberg and Caves (2003) for a critique of the Levinsohn –Petrin estimator we have used here, and Wooldridge (2005), for its interpretation in a system equation framework.

10

of capital by minimizing the excess of capital over the share of all other inputs at the mean productivity predicted in the first step.10

One drawback of our using the Levinsohn-Petrin estimator to address the endogeneity problem is that we have to use value added as our output measure since material inputs is the only non-state input that we can use to control for unobservables.11 The problem with using value added as our measure of output is that it is a computed variable rather than one that is reported in company records. As it turns out, reported sales and purchase figures imply negative value added in many cases. It seems that the natural thing to do is not to consider these as valid entries. This could lead to bias in our estimates of China-India performance gaps since it would seem that we would have to drop more Indian firms than Chinese firms from the survey samples. In order to avoid the bias this would lead to in our estimates of the China-India performance gaps we have decided to drop 8 percent of firms on the lowest end of the distribution of value added per employees from each country sample in estimating the production function.12

Gauging business climate

There have been two primary considerations behind our selection of the business climate variables used as regressors in our performance equations. One is the list of factors that the policy literature identifies to be significant determinants of economic performance in the institutional setting of business operations in either country. The second is the availability of comparable indicators at the micro level in both countries. The quality of physical infrastructure, access to finance, labor market flexibility, predictability of government regulation, and levels of technology or workforce skills are all considered to be important influences in economic performance either in China or in India. We also happen to have at least one common proxy in both the China and the India samples of our data.

China is reported to have invested far more than India in physical infrastructure almost since the 1980s. The share of investment in this particular sector is believed to have averaged 15 to 20 percent of GDP since the mid 1990s as compared to India’s less than 7 percent of GDP, which China’s investment at about 8 times India’s in absolute terms (Ahya and Xie, 2004). At the same time, infrastructure is often cited as one of the key bottlenecks to growth in India. 13 Within the category of infrastructure, the blame has particularly been focused on the problem of expensive and unreliability of power supply to industry (World Bank, 2004). We therefore use as our proxy for the quality of infrastructure the proportion of annual sales that businesses report in surveys to have lost due to power outages. On the average firms in the Indian sample report about 9 percent in

10 Alternative approaches to the problem of endogeneity of inputs are proposed in Olley and Pakes (1996) and Blundell and Bond (2000). Neither of these is feasible with our data, though.11 We also estimated a production function with gross output as the dependent variable and capital, labor and materials as three distinct inputs. With this approach we cannot deal with the problem of endogeneity of input choices since material input is used as regressors. TFP on this approach is simply estimated as the fixed effects residual (including the fixed effects). The correlation between the Levinsohn-Petrin TFP and the fixed effects TFP is nonetheless quite high, with a correlation coefficient of 0.66. 12 The 8 % city cut off point ensures that all Indian firms with positive value added are included, but means that some Chinese firms had to be dropped even when they had positive value added.13 See, for example, Pinto, Zahir and Pang (2006).

11

lost sales due to power outages, against 2 percent of loss in sales as the average for the China sample (Table 2).

The literature suggests that access to finance might be more of a growth bottleneck in China than in India. Although China’s higher investment rate is claimed to have been facilitated by low interest bank loans in the 1980s and the 1990’s, the financial system has been plagued recently by extremely high rates of non-performing loans and very low rates of return on bank assets compared to the Indian banking system (Ahya and Xie, 2004).14 Possibly because the credit restraining measures that Chinese authorities are reported to have taken in response are taking effect, our indicators suggest that Chinese firms have poorer access to finance than their Indian counterparts. One of these indicators is whether or not a business has an active bank overdraft facility. Since firms could substitute trade credit when faced with tight bank credit, we use this along side a second indicator, which is the proportion of inputs purchased on credit. The proportion of firms that have bank overdraft facility in China is 29% as compared to 56% in India (Table 2). Only 12 percent of input purchases are also on credit in China, as opposed to 66 % in India.

Deepak Lal’s reference to ‘second generation reforms’ in the opening quote almost certainly includes making India’s labor market more flexible than it is today. Although there are outstanding labor market reform issues in China as well, reforms that took place in the mid 1980s and mid 1990s are believed by many to have made the Chinese labor market more flexible than India’s in terms of the ease with which firms can adjust staffing levels to product market and technological developments. (Ahya and Xie, 2004). In India one of the key provisions of the existing labor code requires businesses employing more than 100 workers to seek the permission of the state government for closure or the retrenchment of workers, which permission, critics point out, is rarely granted (Sachs et al., 1999). This is believed to have added significantly to duration of insolvency procedures in the country. It is also claimed to force firms to maintain suboptimal scale of operation. Related items of the labor law include the ‘service-rules’ provisions of the Industrial Employment Act of 1946 and the provisions of the Contract Labor (Abolition and Regulation) Act of 1970. The Industrial Employment Act provides for the definition of job content, employee status and area of work by state law or by collective agreement, after which changes would not be made without getting the consent of all workers.15 Zagha (1999) points out that this has always made it difficult for businesses ‘to shift workers not only between plants and locations, but also between different jobs in the same plant.’ As a way out of such restrictions businesses may resort to contract workers, as per the provision of yet another law, namely, the Contract Labor Act. This law gives state governments the right to abolish contract labor in any industry in any part of the state. In states where recourse to contract labor has been more restricted as a result, keeping employment below the threshold level of 100 employees or contracting out jobs has been the only way of maintaining flexibility in the allocation of manpower.

14 See also Deutsche Bank Research (2005).15 This too applies to establishments with more than 100 employees, but Zagha (1999) notes that some states have made the provisions mandatory to firms with 50 or more workers while other states have abolished the employment size limit altogether.

12

The immediate consequence of China’s labor market reforms of the mid 1980s and mid 1990’s has been to increase the proportion of workers on temporary contracts (Ahya and Xie, 2005). As well, variations in the strictness of the enforcement of the labor law in India seem to be highly correlated with the same proportion. We therefore use the proportion of non-permanent employees in the workforce as one of our indicators of labor market flexibility. A complementary indicator on which we have observations on both the China and the India samples is the overstaffing ratio reported by managers at the time of the survey. This is the proportion of current employees that managers could lay off without reducing output. While overstaffing of this kind could be a result of voluntary labor hoarding, the India survey gives indications that restrictive labor laws are part of the list of reasons behind the phenomenon. The values of both indicators in Table 2 are consistent with the Chinese labor market being more flexible. Although the difference between the mean proportions of non-permanent workers is not statistically significant, the median proportion in China is more than twice that of India. Both the mean and median overstaffing ratios are also significantly larger in the Indian sample.

In both countries recent growth has benefited from rapid expansion of exports. Martin and Manole (2004) note that in both countries exports have progressively shifted to more skill intensive and more high tech manufactures, this being more so in China, where the growth rate of exports has also been a great deal faster. As they point out this suggests that there must have been significant increase in the availability or utilization of skills and technology. However, the question of how far difference in this respect explains the performance gap between the two economies has yet to be addressed explicitly. The picture that emerges from a comparison of conventional indicators of availability between the two countries is rather mixed. China clearly has the advantage on adult literacy and school enrollment rates (including those for tertiary education), but India is also believed to have more qualified engineers and other categories (Deutsche Bank Research, 2005).16 The indicator of firm level skill levels that we use in our performance equations is the proportion of workers that regularly use computers on their jobs. On this measure Chinese firms appear to have a slight edge over their Indian counterparts. On the average 22 percent of workers in a Chinese business use computers regularly on the job as compared to 17 percent in an Indian establishment (Table 2).

Our last indicator of business climate relates to the predictability of government regulation of industry to maintain environmental, safety, health, and labor standards. Many of these standards are enforced through inspection visits by government officials. While the standards are probably not much different from what is enforced routinely in developed economies, individual government officers seem to have far more discretion in enforcement in the developing world. In many cases inspection visits are arbitrary or too frequent, and are viewed by business people as a form of veiled demand for bribes, as the price of avoiding future visits. The price is sometimes worth paying to avoid disruption to production plans or save valuable staff time, including that of senior management. Our proxy for the predictability of regulation is the local coefficient of variation in the reported frequency of inspection visits per year. Both the mean and the median of this indicator are not statistically different between the China and the India samples (Table 2).

16 India’s adult literacy rate stood at 68% against 95% for China in 2003 (Deutsche Bank Research, 2005). The tertiary enrollment rates for the same year were 11% and 13% for India and China respectively.

13

Endogeneity of business climate indicators to firm performance

Unlike our measures of performance on which we typically have 2 to 3 annual observations for each firm, all seven of our indicators of business environment are observed only for the year of the survey. We have therefore estimated equation (2) on a cross-section by pooling the performance indicators over time and assuming that the business environment indicators are constant over the three year period leading up to the survey. This means that business environment indicators included in the equation would very likely be correlated with the error term if these are measured at the firm level. This is because the error term will probably include unobservables that drive both the performance of a firm and the business environment indicators it reports. It is possible, for example, that inherently more productive firms cope better with frequent power outages through the use of more flexible processes and production schedules. They could consequently lose less in potential output than other firms. Similarly, our indicators of labor market flexibility could be correlated with business growth or productivity if, for example, labor regulation is a more binding constraint on more innovative firms, which may have less scope for manipulating the share of non-permanent employees in their labor force. Likewise with the other indicators: inherently high growth firms could be more attractive to potential lenders; or could rely more on information technology; or better control the behavior of government inspectors; and so on.

To alleviate the bias that these instances of endogeneity would lead to in the estimation of performance equations by OLS we measure all business climate indicators as city averages of firm level observations. This would be equivalent to the use of city dummies as instruments for the indicators, and should remove the bias, if we can assume that the location decision of firms is exogenous to performance. This would be a reasonable assumption if either location decisions are irreversible once made, or that there are no unobservables that influence both the performance of firms and their choice of location. Otherwise the estimation of the performance equation would be biased. Because the vast majority of firms are small and medium sized we think the assumption of irreversible location is a reasonable one. Just in case it is not we have run a robustness check by running the regression on the subset of small firms in each country sample, for which the scope for cross-city mobility would seem to be far more limited.17 [[We need to do the robustness check]]

4. Estimation Results

Comparing performance

Summary statistics of the two primary measures of performance, namely total factor productivity (in log) and annual value added growth are given in Table 2. Also reported in the same table are summary statistics of the rates of the three sources of value added growth, namely, employment growth, capital stock growth and growth of total factor productivity. The most noticeable element of the summary is probably the huge TFP premium of the China sample. The log difference in mean TFP is 1.23 in favor of the 17 This is the approach used in Dollar et al. (2005).

14

China sample, meaning that the average TFP for the China sample is more than three times that of the India sample. This together with the higher capital intensity of Chinese firms that we read the employment and capital stock rows of the table makes value added per worker larger in the China sample than it is for Indian firms.

Chinese firms also grow much faster in terms of output, the average growth rate in value added being 12 per cent per annum in their case as compared to 5 percent per annum for the India sample. This is in spite of the fact that the average employment growth rate is significantly higher for the Indian firms at an average of 4 percent per annum as compared to 2 percent per annum for the Chinese sample. This, however, is more than made up for by the higher growth rate both of capital stock and total factor productivity in the China sample to result in the higher growth rate of output of Chinese firms. Of these two, growth in capital stock is by far the more important reason why the output growth rate is so much higher. The average rate of growth of net fixed assets for Chinese firms is 11 per cent per annum against a rate of -1.0 percent per annum for the India sample. On the other hand rates of TFP growth are comparable between the two samples at 3 percent per annum and 4 percent per annum for the India and China respectively. To facilitate computation of the relative weight of these gaps in the growth rate of output, we have regressed the growth rate of value added on the rates of growth of employment, fixed assets and total factor productivity for each country in Table 3. Given that the underlying specification is in fact an accounting identity, nothing should be read into the goodness of fit of the regression or the fact that the coefficient estimates are almost identical across the two country samples. The regression nonetheless shows that differences in factor growth account for at least 74 per cent of the gap in the rate of growth of output between Chinese and Indian firm. TFP growth accounts for at most 26 percent of the same gap. This means that Chinese firms are growing faster than Indian firms mainly because they are investing in plant and equipment at higher rate. However, the 26 percent contribution of TFP growth for the gap in output growth is fairly large. Since we are talking about a 7 percentage point gap in output growth rate, this means that Chinese firms are probably growing by 1.8 percentage points a year faster than their Indian counterparts on account of their higher rate of TFP growth.

In order to see how the firm level TFP growth rate gap itself translate to industry terms we have computed the Olley-Pakes decomposition of industry level TFP for two consecutive years. The results are reported in Table 4. We see right at the end of the table that aggregate industry productivity increased by 3.1 percentage point for the Chinese sample over the two year period as compared to 2.2 percentage points of growth for the India sample. This is a substantial albeit not very large gap, and reflects mainly the greater allocative efficiency gains in the Chinese sample. While firm level TFP growth has been a more important source of growth in aggregate productivity than allocative efficiency gains in both the China and the India samples, allocative efficiency gains led to productivity growth by a full percentage point in the China sample, while aggregate productivity in fact fell by half a percentage point in the India sample due to allocative efficiency losses. Estimated performance equations

15

Our estimation results of performance equations are reported in Table 5. The performance measure for the equation estimated in the first column is the log of total factor productivity. In column 2 we estimate an output growth equation with the rate of growth of annual value added as the dependent variable. The other columns refer to the employment growth, the fixed assets growth and the TFP growth equations in that order. Obviously the key columns are the first two, columns 3 to 5 being in a sense extensions of the output growth regression of the second column. While column 1 and 2 relate to productivity and output growth as distinct measures of performance, the performance indicators of columns 3 through to 5 relate to the constituent elements of output growth. In view of the fact that the average firm in the Chinese sample is much larger than its Indian counterpart, we include employment size among our firm level controls all five equations. We also include firm age and sector dummies to control for lifecycle effect in performance. While age, size and line of industry should be no less outcomes of business environment than productivity or growth, our focus here is on effects conditional on the existing industry, size and age distribution of firms.

There are two striking patterns in Table 5. One is that, on the whole, the marginal effects of business climate seem to be stronger in the China sample. Secondly age and size effects also tend to be larger for the China sample. In column1, for example, total factor productivity increases rapidly with a businesses scale of operations in both samples, but the effect is stronger in the China sample by as much 15 percent. Perhaps more importantly, there is no evidence of age or lifecycle effects in the India sample, while there is a strong and inverse effect in the China sample. In the China sample younger firms tend to be more productive. This is important since age can proxy for entry cohort effects. Age could also capture passive learning effects such as learning by doing. To the extent there is passive learning, older firms should be more productive, at least up to a point in the life cycle. If younger firms are nonetheless more productive as is the case in the China sample, new entry cohorts must be more productive than incumbents to a point that more than makes up for their relative “inexperience”. This in itself would be a source of growth in aggregate (industry) productivity, which we are setting aside by including age among the controls in the productivity equation.

A measure of the contrast in the marginal effects on productivity of business climate between the two samples is that labor market flexibility is the only statistically significant influence in the Indian sample, while skills and access to finance are what matter in the Chinese sample. The only business climate indicator that has statistically significant coefficient in the India equation is the proportion of non-permanent workers. This, however, is not among the statistically significant indicators for the China sample, which are the proportion of computer users, the share of inputs purchased on credit, and the proportion of businesses with bank overdraft facilities. Total factor productivity is higher in India where the proportion of non-permanent workers is higher, which we interpret as evidence of positive association between labor market flexibility and productivity. For the China sample, the coefficients of the proportion of computer users is positive and statistically significant as are the coefficients of both indicators of access to finance. We interpret the first as evidence that total factor productivity is higher where skill or

16

technology levels are higher. Productivity is also higher where access to finance is greater.18

Turning to the value added growth equation of column 2 (of Table 5), we see that, although life cycle effects seem to be similar across the two samples, the effects of business environment again differ sharply. In both samples, younger firms grow faster in revenue terms, as should be expected. The business climate variables with statistically significant influences are also the same in the two samples, these being skills and access to finance. Of these two access to finance has a positive and statistically significant effect on output growth in as far as it is captured by access to bank overdraft facilities. The contrast is in two ways. First, the coefficient of this indicator is substantially larger in the China sample. Secondly, the coefficient of our skills and technology indicator is negative in the growth equation for China. Our interpretation of this is that relatively high-tech or more skill intensive firms grow faster than low tech businesses in India, while the reverse is the case in China. This in turn suggests that, conditional on the current structure of industry, skills and technology are a constraint to growth in India, but not so in China. This reverses the pattern observed in connection with the effect of skills and technology on productivity, whereby total factor productivity is higher on the average in high-tech or more skills intensive firms in China but not so in India.

Moving to the determinants of the constituent elements of output growth in columns 3 through 5 of Table 5, we note again that the broad pattern where both life cycle effects and the effects of businesses climate differs between the two countries applies here as well. For example, we see in column 3 that larger firms have lower job growth rates in both countries, but the same effects is three times as strong in China as it is in India, suggesting that small firms are more important in creating jobs in China than in India. Employment growth is also slower in older firms in both countries. This pattern too is more pronounced in China than in India, again suggesting a stronger creative destruction process in China. A case of the contrasting effects of business climate between the two countries in the context employment growth is that the effect of skills and technology seems to be much more pronounced in India than in China. On the other hand labor market flexibility, while being important in both, is a stronger influence on employment growth in China than in India. In both samples the employment growth rate is higher where the proportion of non-permanent workers is higher, but the estimated coefficient is several times larger for the China sample. A third case of contrast in the effect of business climate between the two samples is that employment growth is faster where there is more access to bank finance or trade credit, but again the effect is stronger in China. This is not surprising since Chinese banks are well-known in lending to favor large and state-owned enterprises (Cull and Xu, 2000, 2003; Huang, 2003; Brandt and Li, 2003), while job creation rates are faster in small firms.19 Fourth, while there is no association between our indicator of the predictability of regulation and employment growth rates in the Indian sample, greater unpredictability seems to reduce the job creation rate in the China sample.

18 The positive association between productivity and the proportion of inputs purchased on credit is consistently with the observation about the substitution of informal financing for formal financing in making Chinese firms functioning (Allen, Qian and Qian, 2005).19 Interestingly, there is evidence that informal finance (i.e., trade credit) does help job creation, but the association is not statistically significant (t=1.50)

17

In column 4 of Table 5 we see the contrast in the patterns of growth in capital stock between the two samples. First, in the India sample, net investment rates are higher in larger (in employment terms) and more skill intensive or high tech firms, suggesting that capital formation is probably driven more by firms in the upper ends of the size distribution and technology spectrum. This, however, does not seem to be the case in the China sample, where net investment rate is (randomly) uniform across the size spectrum. The rate is also actually lower for more high tech or more skill intensive firms. Secondly, while there are no age effects in capital growth in the India sample, net investment rates are higher in younger firms in the China sample. Third, the effect of business environment on capital growth rates is sharply different between the two samples. One aspect of this is the difference in the association between skills or technology and investment rates that we already refered to. A second is that greater access to finance is associated with greater assets growth in the China sample while there is no such association in the India sample. A third aspect is that net investment rates is highly correlated with both of our indicators of labor market flexibility in the China sample, while there is no such correlation in the India sample. In the China sample the rate of capital growth decreases both with the overstaffing rate and with the proportion of non-permanent workers.

The pattern of TFP growth we see in column 6 is far more similar across the two samples than is the case with the growth patterns in employment and capital stock. For example in both country samples, productivity growth is faster in younger firms. Access to finance is also positively associated with TFP growth in both samples. However, even here there is an important difference in that higher skill and technology levels are associated with faster productivity growth in the India sample while this is not the case for Chinese firms.

Our discussion of the results of Table 5 so far has been intended to identify the major influences on firm performance in the two country samples. We now move to the question of how far these influences explain the performance gaps observed between the samples. In order to address this question, we present in Table 6 the Oaxaca-Blinder decomposition of the effect of business climate variables on productivity and output growth, but confining ourselves to influences that are statistically significant in at least one of the two country samples. Apart from enabling us to disentangle “endowment effects” from “rate of return effects”, this will help us assess the weight of each aspect of business climate relative to other covariates of performance. We present in Table 7 the full decomposition for the entire range of regressors (including firm level controls) and covering effects on employment growth and assets growth as well.

To give a sense of how powerful an influence business climate is on the observed performance differential between the two samples, the log TFP gap between them would be 1.39 if productivity depended solely on business environment as captured by our indicators (Table 6). This is greater than the raw (actual) sample mean difference of 1.23. As our discussion of column 1 of Table 5 showed, this is the combined effect of four element of business climate, namely, skills and technology, labor market flexibility, access to bank finance and access to trade credit. The mean indicators of skills and technology and labor market flexibility are both higher for the Chinese sample,

18

suggesting that the Chinese business environment is better on those counts. On the other hand, the means of the finance indicators are both higher for the Indian sample, suggesting the Indian firms have better access to finance than their counterparts. The combined effect of the four indicators on productivity is nonetheless in favor of higher productivity of the Chinese sample. This is not because skills and labor market flexibility are more powerful influences on productivity than access to finance.20 The first two columns of Table 6 show that it is rather because that the marginal effect of each of the four elements is stronger in China than in India. Although Indian firms are more productive than their Chinese counterparts because of the “endowment effect” of their better access to finance, the “rate of return” effect of better access finance is higher in China to a degree that makes the overall productivity effect of access to finance higher for Chinese firms.

We see from the third and fourth columns of Table 6 that labor market flexibility does not account for much of the (output) growth gap between the average Chinese firm and the average Indian firm in our sample. A possible reason for this is the fact that its effects on physical capital formation tend to cancel out the effects on employment growth. On the other hand differences in access to finance and levels of skills and technology together would generate an annual out growth gap of 9.5% in favor of the average Chinese firm, if they were the only sources of variation in value added growth. This is higher than the raw average growth gap of 7% we observe in the data, and goes to show how powerful an influence business environment is in growth performance as well. However, this time too the aggregate business climate impact masks two patterns. One is that the growth rate of Indian firms is actually higher on account of one of the two aspects of business climate that matter here, which is skills and technology. The overall outcome is consequently due to growth being faster on account of the other business climate element, that is, access finance. The second is that the Indian average growth rate is smaller on account of the aspect of business climate the mean indicator of which is actually higher for the Indian sample, which is access to finance. Ditto with the average Chinese growth rate, which is smaller on account of the business climate element which is better for the Chinese sample, which is level of skills and technology. The reason for this is that the “endowment effect” of China’s better indicator of skills and technology is undermined by a negative “rate of return effect”-that is by the fact that, unlike the case with India relatively higher skills and technology in China are associated smaller growth swamped by the higher rate of return of better access to finance to China. On the other hand, the positive “endowment effect” of India’s better access to finance is swamped by the higher “rate of return” to better access to finance in China.

The relative importance of firm level controls can likewise be read from the full decomposition given in Table 7. For example we see from the TFP panel of the decomposition (Panel A) that the fact that Chinese firms are larger than their Indian counterparts accounts for 59 percent of the explained component of the TFP gap between the two samples, the unexplained part being the component due to unobserved country effects. Panel C also shows that the fact that Chinese firms are much larger on the average in employment terms is the single most important factor explaining why their employment growth rate is smaller than that of the Indian sample, dwarfing all other

20 On the contrary, access to finance is the more powerful influence in each country sample than skills and technology and labor market flexibility combined.

19

influences including elements of business environment. Panel D shows that unobserved country effects account for the higher net investment rate of Chinese firms far more than do all our explanatory variables combined.

5. Conclusion

This paper has analyzed data on samples of manufacturing plants drawn from China and India with the aim of shading light on the reasons why the Chinese economy has done so much better in terms of per capita income and GDP growth. Consistent with the gap in per capita income, the average Chinese plant in our sample has significantly higher labor productivity, in part because it is more capital intensive and in part because its total factor productivity is higher than that of its Indian counterpart. Chinese plants are also larger. This is partly because Chinese firms grow a great deal faster on the average in terms of revenue and fixed assets. Higher rates of investment in physical capital are the main reason for the higher growth rate of output in Chinese firms. A less important, but significant source of their growth advantage has also been their faster rate of TFP growth. Faster firm level productivity growth has in turn helped make aggregate industry productivity growth rates higher for the Chinese sample than for the India sample. There is also a second source of the greater industry level productivity growth of Chinese firms, which are the greater allocative efficiency gains of the Chinese sample.

The finding that allocative efficiency gains are greater in the Chinese sample is consistent with a second result: that catch up (or inverse size) effects and life cycle (or age) effects are generally stronger in the Chinese sample than in the Indian sample. In particular it is worth noting that there is no evidence of age effects in firm level productivity in the India sample, while younger firms are more productive in the Chinese sample. Employment growth rates are higher in smaller and younger firms in the two samples, but again both these effects happen to be stronger for the China sample. Also, while there are no age effects in the rate of net investment in the India sample, the rate is higher for younger firms in the China sample.

Together size and age effects account for a large proportion of the performance gaps between the Chinese sample and the India sample. Since the age and size distribution of firms very much depends on the institutional and policy environment in which firms function, an assessment of the role of what we have collectively termed business climate here, should probably involve the modeling the determination of both distributions. Our analysis shows that, even setting aside its effects on firm age and firm size, business climate accounts for a large part of the performance gap that we observe between the China and India samples. Labor market flexibility, access to finance, and the availability of skills and technology are the main element of business climate contributing to the TFP gap between the two groups of firms. Differences in access to finance and skill and technology levels also account for a large proportion of the growth gap. On the whole, we can say that business climate differences are one of the factors behind the performance gaps we have documented to exist between Chinese firms and Indian firms. It is important to note that this is not so much because business climate indicators are better in China on the average as because the marginal effect of improvement in business climate is larger than in India. To put it another way, even if China and India had an identical set of values for our indices of business climate, the average Chinese firm would have better

20

performance indicators than its Indian counterpart because of the influence of business environment on performance.

We should close with a serious qualifier to these conclusions. This is that they are based on purely cross-sectional variation in business climate indicators. It should be clear from our discussion of the identification problems this poses, and from our robustness checks, that our results are unlikely to be driven entirely by reverse causation-that is by the possible tendency of inherently more productive or faster growing firms self-selecting into localities where the business climate happens to be better. This of course is no substitute for formal selectivity bias correction techniques that would be possible with better data. However, at this stage the data we have analyzed in the paper seem to be the best available set of observations yet of co-variation between micro-economic performance and business environment.

References

Ahya, C., and A. Xie, 2004. ‘India and China: A Special Economic Analysis,’ Morgan Stanley Equity Research, July.

Allen, Franklin, Jun Qian, Meijun Qian. 2005. "Law, Finance and Economic Growth in China," Journal of Financial Economics 77 (1), 57-116.

Blinder, A. S. 1973. “Wage Discrimination: Reduced Form and Structural Estimates,” Journal of Human Resources, 8, pp. 436-455.

Blundell, R. and S. Bond. 2000. ‘GMM Estimation with Persistent Panel Data: An Application to Production Functions,’ Econometric Reviews, 19, pp. 321-340.

Brandt, Loren, and Hongbin Li. "Bank Discrimination in Transition Economies: Ideology, Information or Incentives," Journal of Comparative Economics 31 (3), 387-413.

Cull, Robert, and Lixin Colin Xu. 2000. "Bureaucrats, State Banks, and the Efficiency of Credit Allocation: The Experience of Chinese State-Owned Enterprises," Journal of Comparative Economics, 28, 1-31.

Cull, Robert, and Lixin Colin Xu. 2003. “Who Gets Credit? The Behavior of Bureaucrats and State Banks in Allocating Credit to Chinese SOEs,” Journal of Development Economics 71, 533-559.

Deutsche Bank. 2005. ‘China and India Chart Book,’ Deutsche Bank Research, October.

Dollar, D., M. Hallward-Driemeier, and T. Mengistae. 2005. ‘Investment Climate and Firm Performance in Developing Economies,’ Economic Development and Cultural Change, 54(1), pp. 1-31.

Foster, Lucia, John Haltiwanger, and C.J. Krizan. "Aggregate Productivity Growth:Lessons from Microeconomic Evidence." In Hulten, Charles, Edwin R.

21

Dean, and Michael J. Harper, eds., New Development in Productivity Analysis. NBER studies in Income and Wealth, Vol. 63.

Hall, Robert., and Charles Jones. 1999. ‘Why Do Some Countries Produce So Much More Output Per Worker Than Others?’ Quarterly Journal of Economics, 114 (February), 83-116.

Huang, Yasheng. Selling China.

Lal, Deepak. 2005. ‘India vs. China,’ Business Standard, March 15, 2005.

Levine, Ross. 1997. "Financial Development and Economic Growth: Views and Agenda," Journal of Economic Literature 35 (2), 688-726.

Levinsohn, J. and A. Petrin. 2003. ‘Estimating Production Functions Using Inputs to Control for Unobservables,’ Review of Economic Studies, 70, pp. 317-341.

Martin, W., and V. Manole. 2004. ‘China’s Emergence as the Workshop of the World,” Stanford Center for International Development, Working Paper No. 216. June.

Oaxaca, R.L. 1973. “Male-Female Wage Differentials in Urban Labor Markets,” International Economic Review, 14, pp. 693-709.

Olley, S. and A. Pakes. 1996. ‘The Dynamics of Productivity in the Telecommunications Equipment Industry,’ Econometrica, 64(6), pp. 1263-1297.

Pinto, B., F. Zahir, and G. Pang. 2006. ‘From Rising Debt to Rising Growth in India: Microeconomic Dominance?’ Processed, World Bank, Washington, D.C., February.

Sachs, J., A. Varshney, and N. Bajpai. 1999. India in the Era of Economic Reforms. New Delhi: Oxford University Press.

Srinivasan, T. N., 2003. ‘China and India: Growth and Poverty, 1980-2000.’Standford Center for International Development, Working Paper No. 182. September.

Young, Alwyn. 1994. ‘Lessons from East Asian NICS: A contrarian view,” European Economic Review, 38, pp. 964-973.

_______1995. ‘The Tyranny of Numbers: Confronting the Statistical Realities of the East Asian Growth Experience,” Quarterly Journal of Economics, August, pp. 641-680._______2003. “Gold into Base Metals: Productivity Growth in the People’ Republic of China during the Reform Period,” Journal of Political Economy, 111(6), pp. 1220-1261.

Wooldridge, J. M. 2005. ‘On Estimating Firm-Level Production Functions Using Proxy Variables To Control for Unobservables,” Processed, Department of Economics, Michigan State University, July.

22

World Bank. 2004. India: Investment Climate Assessment 2004-Improving Manufacturing Competitiveness. Finance and Private Sector Development Unit, South Asia Region, Washington, D.C.

Zhaga, R. 1999. ‘Labor and India’s Economic Reforms,’ in Sachs and et al. (eds.), India in the Era of Economic Reforms.

23

Table 1: Industry distribution of firms to which the data refer

  China India

  Number % Number %

Garments and leather products 352.0 22.5 357.0 20.6Household electronics 63.0 4.0 137.0 7.9Electrical equipment and parts 461.0 29.5 163.0 9.4Auto and parts 358.0 22.9 265.0 15.3Food processing 71.0 4.5 216.0 12.5Chemicals and pharmaceuticals 102.0 6.5 421.0 24.3Metallurgical products and tools 158.0 10.1 176.0 10.1Total 1565.0 100.0 1735.0 100.0

24

Table 2. Summary Statistics.India China China-India

mean median S.D. mean median S.D. mean dif median dif

china 0.00 0.00 0.00 1.00 1.00 0.00

age 15.80 12.00 16.56 15.74 9.00 14.13 -0.06 -3

V (in mil) 1.01 0.05 7.57 6.87 0.76 36.30 5.86 0.71

K (in mil) 1.04 0.04 17.10 10.20 0.86 48.90 9.16 0.82

L 86.98 18.00 285.70 456.95 146.00 1066.30 369.97 128

tfpLP_W -0.63 -0.57 1.43 0.60 0.58 1.46 1.23 1.15

tfpLPG_W 0.03 0.01 0.50 0.04 0.05 0.51 0.01 0.04

Vgrow_W 0.05 0.01 0.54 0.12 0.08 1.08 0.07 0.07

Kgrow_W -0.01 -0.03 0.27 0.11 0.03 0.33 0.12 0.06

Lgrow_W 0.04 0.00 0.14 0.02 0.00 0.25 -0.02 0

Mlossppower 0.09 0.09 0.05 0.02 0.02 0.01 -0.07 -0.07

MsL_computer 0.17 0.18 0.05 0.22 0.21 0.09 0.05 0.03

MsL_nperm 0.43 0.20 0.84 0.44 0.43 0.13 0.01 0.23

Moverman 0.11 0.10 0.06 0.06 0.05 0.04 -0.05 -0.05

Moverdraftt 0.56 0.58 0.20 0.29 0.26 0.10 -0.27 -0.32

MtradecreditS 0.66 0.67 0.08 0.12 0.11 0.07 -0.54 -0.56

CVinspectt 1.56 1.49 0.72 1.52 1.48 0.27 -0.04 -0.01

25

Table 2A. Correlation Matrix of Key Variables

Vgrow_W Kgrow_W Lgrow_W tfpLPG_W tfpLP_W lnL1 lnage

Mlosspowe

r MsL_computer MsL_nPerm Moverman

Moverdraf

t MtradecreditS

Vgrow_W 1

Kgrow_W 0.0673 1

Lgrow_W 0.2053 0.1209 1

tfpLPG_W 0.9574 -0.1347 0.0806 1

tfpLP_W 0.2049 0.0933 0.0385 0.1866 1

lnL1 0.035 0.1309 -0.1619 0.0316 0.6402 1

lnage -0.0881 -0.0682 -0.1987 -0.0611 0.0442 0.1976 1

Mlosspower -0.0205 -0.1568 0.0166 0.0061 -0.2881 -0.4499 -0.0013 1

MsL_computer 0.07 0.0967 0.0426 0.0503 0.1785 0.272 -0.1086 -0.1988 1

MsL_nperm 0.0109 0.0199 0.0207 0.0065 0.0915 0.1263 -0.011 -0.0753 0.0244 1

Moverman -0.053 -0.0595 -0.0208 -0.038 -0.1893 -0.2238 0.0839 0.1362 -0.37 -0.1144 1

Moverdraft 0.0191 -0.0954 0.0604 0.031 -0.1564 -0.2189 0.104 0.4273 -0.1467 0.2273 0.3019 1

MtradecreditS -0.0219 -0.1966 0.0447 0.0089 -0.3665 -0.553 0.0243 0.641 -0.2798 -0.0429 0.3163 0.602 1

CVinspect -0.0275 0.0021 -0.0125 -0.0287 -0.0044 0.0102 0.0816 0.0381 -0.1283 0.1556 -0.0901 0.1363 -0.0369

26

Table 2B. Correlation Matrix of Key Variables: India and China

India Vgrow_W Kgrow_W Lgrow_W tfpLPG_W tfpLP_W lnL1 lnage

Mlosspowe

r MsL_computer MsL_nperm Moverman

Moverdraf

t MtradecreditS

Kgrow_W 0.0202 1

Lgrow_W 0.179 0.1306 1

tfpLPG_W 0.9684 -0.1701 0.0843 1

tfpLP_W 0.1194 0.009 -0.0537 0.1172 1

lnL1 -0.004 0.0662 -0.1401 -0.0037 0.541 1

lnage -0.0842 0.0459 -0.1411 -0.0784 0.1833 0.2885 1

Mlosspower 0.0117 -0.032 -0.0312 0.0198 -0.0048 -0.0984 -0.0706 1

MsL_computer 0.1073 0.0371 0.12 0.0944 -0.1117 -0.1064 -0.098 0.175 1

MsL_nperm 0.01 0.0313 0.0143 0.0061 0.1189 0.2133 -0.005 -0.1085 0.0021 1

Moverman -0.0248 0.0384 0.0199 -0.0287 0.0529 0.0618 0.0387 -0.266 -0.2331 -0.0861 1

Moverdraft 0.0434 0.0224 0.0369 0.04 0.1232 0.247 0.1564 -0.0144 -0.0743 0.2978 0.1469 1

MtradecreditS 0.0389 -0.0028 0.074 0.0383 -0.034 -0.0395 -0.0607 -0.1529 0.1285 -0.1843 -0.1803 0.0654 1

CVinspect -0.0349 0.0157 -0.0129 -0.038 0.05 0.0727 0.0831 -0.0002 -0.0607 0.1585 -0.1495 0.1661 -0.2002

China: Vgrow_W Kgrow_W Lgrow_W tfpLPG_W tfpLP_W lnL1 lnage Mlossp~r MsL_co~r MsL_np~m Moverman Moverd~t Mtrade~S

Kgrow_W 0.0928 1

Lgrow_W 0.2327 0.1332 1

tfpLPG_W 0.9479 -0.115 0.0833 1

tfpLP_W 0.2908 0.0158 0.1222 0.2834 1

lnL1 0.0327 -0.0275 -0.2014 0.0714 0.5542 1

lnage -0.0887 -0.1438 -0.2439 -0.0441 -0.0434 0.2682 1

Mlosspower 0.0039 0.0042 0.0352 -0.0047 -0.1059 -0.023 0.022 1

MsL_computer 0.0372 0.023 0.0375 0.0302 0.1392 0.195 -0.1047 -0.1345 1

MsL_nperm 0.034 -0.003 0.1044 0.0203 0.1312 0.0788 -0.0657 0.2458 0.17 1

Moverman -0.0656 0.0046 -0.0881 -0.0549 -0.1766 -0.0714 0.1202 0.1336 -0.3521 -0.5805 1

Moverdraft 0.0795 0.0665 0.0876 0.0539 0.107 0.1749 -0.0048 0.3779 0.3024 0.3676 -0.0378 1

MtradecreditS 0.0811 0.0203 0.0336 0.0683 0.2008 0.2019 -0.0914 -0.4859 0.3323 -0.0316 -0.3134 0.237 1

CVinspect -0.0164 -0.0111 -0.0254 -0.0143 -0.1265 -0.079 0.1116 0.568 -0.3579 0.1529 0.0456 0.0983 -0.3589

27

Table 5. Firm characteristics, business environment, and firm performance(1) (2) (3) (4) (5)tfpLP_W Vgrow_W Lgrow_W Kgrow_W tfpLPG_W

china dummy 0.558 0.422 0.390 0.453 0.111(0.88) (1.19) (3.79)*** (3.98)*** (0.45)

lnL1 0.632 0.009 -0.011 0.016 0.007(28.73)*** (0.79) (3.31)*** (2.21)** (0.62)

lnage 0.048 -0.073 -0.031 0.007 -0.054(0.95) (2.44)** (3.52)*** (0.64) (2.39)**

Mlosspower 0.873 -0.174 -0.111 -0.046 -0.157(0.57) (0.55) (1.43) (0.41) (0.53)

MsL_computer -0.914 1.133 0.321 0.293 0.993(1.07) (3.55)*** (2.98)*** (2.12)** (3.74)***

MsL_nperm 0.048 -0.010 0.007 0.006 -0.008(1.71)* (1.09) (2.13)** (1.50) (1.06)

Moverman 0.815 0.151 0.108 0.218 -0.220(0.96) (0.69) (0.98) (1.49) (1.03)

Moverdraft -0.035 0.194 0.054 -0.022 0.173(0.12) (2.81)*** (2.24)** (0.87) (2.87)***

MtradecreditS -0.416 -0.091 0.122 -0.009 -0.020(0.86) (0.41) (1.50) (0.10) (0.10)

CVinspect 0.002 -0.011 0.007 0.007 -0.026(0.03) (0.68) (1.59) (0.93) (1.51)

C_Mlosspower 6.512 6.186 1.779 -1.732 -0.046(1.25) (1.20) (1.58) (1.35) (0.02)

C_MsL_computer 2.756 -1.474 -0.359 -0.330 -1.226(3.03)*** (2.56)** (2.60)** (1.91)* (3.70)***

C_MsL_nperm 0.373 -0.504 0.082 -0.193 -0.149(1.66) (1.37) (1.51) (2.57)** (1.10)

C_Moverman -1.878 -1.283 -0.483 -0.515 -0.665(1.15) (1.46) (1.75)* (1.85)* (1.42)

C_Moverdraft 0.664 0.378 0.226 0.370 0.118(1.58) (1.57) (2.44)** (3.31)*** (0.65)

C_MtradecreditS 1.965 0.027 -0.027 -0.181 0.099(2.91)*** (0.05) (0.19) (1.09) (0.33)

C_CVinspect -0.168 0.113 -0.073 0.015 0.017(1.08) (0.87) (2.31)** (0.42) (0.25)

C_lnL1 -0.092 -0.004 -0.036 -0.015 0.019(2.24)** (0.13) (3.46)*** (1.75)* (1.10)

C_lnage -0.442 -0.089 -0.024 -0.077 0.014(6.35)*** (1.50) (2.00)** (5.62)*** (0.51)

Observations 2821 2887 3114 3014 2668R-squared 0.57 0.02 0.13 0.06 0.02

Note. We control for industry dummies. We allow for clustering at the city-level to account for within-city correlation of the error term.

28

Table 3. How well do capital, labor and TFP growth explain value added growth?(1) China (2) IndiaVgrow_W Vgrow_W

Lgrow_W 0.293 0.288(11.63)*** (11.88)***

Kgrow_W 0.320 0.367(20.70)*** (15.91)***

tfpLPG_W 1.025 1.033(84.51)*** (103.99)***

Constant -0.003 -0.001(1.30) (0.46)

Observations 2634 2221R-squared 0.95 0.97

remarkably similar in terms of elasticity

Table XX. The relative importance of factor growth *[ ]

+ ( - )*%explained

constant -0.003 -0.001 0.002 4.423Lgrow 0.288 0.293 -0.02 0.02 -0.006 -12.517Kgrow 0.367 0.32 0.12 0.11 0.039 85.958Pgrow 1.033 1.025 0.01 0.04 0.010 22.136

0.045 100

29

Table 7. Accounting for India-China differences

Panel A. ,of which, total being accounted for our explanatory variables: 1.217

Panel B. ,of which, total being accounted for our explanatory variables:

0.048

(1)+(2) (1)+(2)

Cby_china 45.9 45.9 875.9 875.9

Mlosspower -4.7 10.9 6.2 23.8 262.2 286

MsL_comput

er -3.7 49 45.3 114.5 -661.3 -546.8

MsL_nperm 0.1 13.6 13.7 -0.3 -463.4 -463.7

Moverman -3.1 -9.4 -12.5 -14.3 -162 -176.3

Moverdraft 0.8 16.1 16.9 -108.5 230.8 122.3

Mtradecredit

S 18.4 19.3 37.7 101.9 6.6 108.5

CVinspect 0 -21.1 -21.1 0.9 358.8 359.7

lnL1 97.3 -38.3 59 33.4 -39.9 -6.5

lnage -0.1 -90.9 -91 4 -463.1 -459.1

100 100

Panel C. ,of which, total being accounted for our explanatory variables: -0.016

Panel D. ,of which, total being accounted for our explanatory variables:

0.119(1)+(2

) (1)+(2)

Cby_china -2484.9

-

2484.

9 380.5 380.5

Mlosspower -46.7 -231.2 -277.9 2.6 -29.7 -27.1

MsL_comput

er -99.4 493.8 394.4 12 -59.9 -47.9

MsL_nperm -0.6 -230.4 -231 0.1 -71.9 -71.8

Moverman 31.3 187.1 218.4 -8.4 -26.3 -34.7

Moverdraft 93.1 -423.8 -330.7 4.9 91.5 96.4

Mtradecredit

S 419 20.8 439.8 4.2 -18.2 -14

CVinspect 1.8 704.8 706.6 -0.2 18.9 18.7

lnL1 135.1 1155.6

1290.

7 24.9 -63.5 -38.6

lnage -5.3 379.8 374.5 -0.1 -161.3 -161.4

99.9 100.1

30

Table 4. Aggregate Productivity Decomposition

India China

Year Industry tP simpleavg tP simple

avg

2001"garments & leather

prod." 7.87 7.718 0.152 8.7 8.582 0.118

2002"garments & leather

prod." 8.204 7.818 0.385 8.754 8.616 0.1382001 "electrical eqp./prts" 6.715 6.617 0.098 8.224 8.102 0.122

2002 "electrical eqp./prts" 6.741 6.65 0.091 8.409 8.128 0.281

2001 "Hhd. electronics" 7.962 7.678 0.284 9.798 9.397 0.401

2002 "Hhd. electronics" 7.691 7.661 0.030 9.431 9.369 0.062

2001 "auto & parts" 7.532 7.51 0.023 9.302 9.231 0.071

2002 "auto & parts" 7.544 7.528 0.016 9.451 9.348 0.104

2001 "Food processing" 7.664 7.596 0.069 8.72 8.639 0.081

2002 "Food processing" 7.885 7.602 0.283 8.774 8.704 0.071

2001 "Chem. & pharma" 7.129 6.809 0.320 7.913 7.798 0.115

2002 "Chem. & pharma" 6.894 6.84 0.053 8.192 7.795 0.397

2001"metallergical products

& tools" 8.642 8.637 0.006 9.88 9.689 0.191

2002"metallergical products

and tools" 8.71 8.652 0.058 9.744 9.628 0.116

aggregating across industries:

2001 Annual Aggregate 7.645 7.509 0.136 8.934 8.777 0.157

2002 Annual Aggregate 7.667 7.536 0.131 8.965 8.798 0.167

Table 6 : Oaxaca-Blinder Decomposition of the effect of business climate on TFP and output growth

Based on cols. 1 and 2 of Table 5

  Effect on total factor productivity Effect on growth rate of annual value added Gap at china' coeff India's at coeff. Diff

Gap at china' coeff

India's at coeff. Diff  

 ("Endowment gap")

("Rate of return gap")

("Endowment gap") ("Rate of return gap")

Skills/technology:  proportion of of computer users 0.09 0.47 0.04 -0.06Labor mart flexibility:  prop of temporary workers 0.00 0.16 overstaffing ratio  Access to Finance:  proportion with bank overdraft facility -0.17 0.37 -0.21 0.32ratio inputs on credit -0.84 1.30Total -0.91 2.30 -0.17 0.26  Total of the two components 1.39   0.10    

31

China and India: GDP per capita (PPP), 1980-2001

0

100

200

300

400

500

600

700

800

900

1000

1980

1982

1984

1986

1988

1990

1992

1994

1996

1998

2000

1980

=100

CHN IND

32