Industrialization Trends Essay
235 Services-led industrialization in India: Assessment and lessons Nirvikar Singh* 1. Introduction India has become one of the fastest growing economies in the world over the last two decades, arguably aided in this performance by economic reforms. One of the striking aspects of India’s recent growth has been the dynamism of the service sector, particularly information technology (IT) and IT enabled services (ITES), while, in contrast, manufacturing has been less robust.
The contribution of the service sector was particularly striking in the 1990s, which not only saw rapid growth (averaging over 6 per cent since 1992), but also a high contribution – over 60 per cent (Hansda, 2001) – from services. This growth trajectory, which has been termed “services-led” industrialization, or even a “services revolution” (Gordon and Gupta, 2004), seems to stand out from the previous experience of economic development, which followed the traditional path from agriculture to manufacturing, with services becoming important at a later stage.
In India, in contrast, there was a sharp increase in the share of services in GDP, from 37 per cent in 1980 to 49 per cent in 2002,1 while the share of manufacturing remained about the same, at 16 per cent (Kochhar et al. , 2006). Thus questions arise about India’s development pattern, including its nature, sustainability and replicability. This paper reviews the recent growth experience of India, identifies the major contributing factors to its pattern of development, and examines the prospects for further “services-led” industrialization in India.
The analysis will draw on theoretical models as well as case studies of the Indian experience. The structure of the paper is as follows. The next section provides a conceptual framework for the examination of the Indian experience. In particular, we discuss the nature of services, their distinction from products, and their categorization. We argue that the precise nature of the services being considered is important for any analysis of growth impacts, and that one therefore has to go beyond broad national income accounts categories to understand the role of services in industrialization.
Also article Purely Monopolistic Seller,
Section 3 provides a brief overview of India’s overall growth experience, followed by a more detailed examination of the contribution of the service sector to growth, and the relative performance of manufacturing and agriculture in Section 4. Section 5 * Department of Economics and Santa Cruz Centre for International Economics, University of California, Santa Cruz. 236 Industrial Development for the 21st Century examines the potential for spillovers from IT, ITES and other service sectors such as financial services, to the rest of the economy.
We draw on econometric work on productivity growth, as well as input-output analysis of linkages to understand these possible spillovers. It is possible, based on this evidence, to hypothesize that India’s manufacturing sector development may have been constrained, at least in part, by weaknesses in key service sectors such as transportation and electricity. In fact, other evidence suggests that telecommunications reforms were also critical for the rapid growth in IT and ITES. 2 In Section 6, we consider the particular role of international trade in services.
This deserves special treatment because of its growing importance, and because balance of payments constraints have often been major barriers to stable growth for developing countries. In Section 7 we discuss the consequences for employment of different growth paths, and the challenges of education and manpower training to support and sustain India’s development path. Section 8 briefly considers the social and environmental issues associated with “services-led” or alternative development paths.
The focus in this section is on regional inequality issues, as well as impacts that do not necessarily manifest themselves in the national income accounts and growth statistics. Section 9 relates the Indian experience as analyzed in this paper to some of the recent discussions of industrial policy, and development policy more generally. Section 10 offers a summary conclusion of the paper’s findings. 2. Conceptual framework Much of the literature on service-led growth begins with broad national income accounting categories.
However, it is important both to disaggregate services to identify important differences within that category, and stepping back from the accounting, to understand the key ways in which services differ from products, which themselves are classified as outputs of the manufacturing sector. A good starting point for analysis is a consideration of the precise nature of services vis-a-vis other kinds of productive activities. The differences will be seen to depend on the nature of market transactions as well as intrinsic characteristics of services vs. products, and both are subject to change as economies and technologies evolve.
The classical writers, such as Adam Smith and, earlier, the Physiocrats discussed services from the perspective of a distinction between productive and unproductive labour. Smith’s idea of “unproductive labour” included the servants of wealthy individuals and government employees, as well as the military, the clergy, lawyers, doctors, writers and musicians, all of whom are now included in the service sector. 3 Of course, modern economics, based on utility and demand theory, has jettisoned this view, and services, however defined, are counted as productive activities.
Various modern conceptual discussions of the distinct nature of services make similar observations. Hill (1977), for example, emphasizes the non-storability of services, which requires that services must be consumed as they are produced. Griliches (1992) defines services as anything Services-led industrialization in India 237 that is the result of labour that does not produce a tangible commodity. Lack of tangibility is fundamentally what leads to non-storability, and to non-transferability (Economic Council of Canada, 1991). However, intangibility must, in the digital age, be carefully interpreted.
Thus, software programs and various forms of digital electronic content have only limited tangibility, but are storable and transferable. The creation of these digital products may often be classified as a service, as also their delivery. The marketing literature pays considerable attention to the nature of services, and in this literature, the dimensions that distinguish services from products are summarized as follows: (1) intangibility, (2) heterogeneity, (3) simultaneity of production and consumption, and (4) perishability (e. g. , Parasuraman, Zeithani and Berry, 1985; Rust and Chung, 2005). Several more fundamental economic implications are drawn from these distinguishing characteristics. For example, in this schema, intangibility, perishability and simultaneity all work against maintaining inventories. Heterogeneity and simultaneity (where the consumer is involved in creating the service) makes it more difficult to achieve scale economies. All of these features make display and consumer assessment of characteristics harder to achieve. Of course, the use of the above four characteristics to distinguish services from products is subject to several caveats.
First, the differences are mostly of degree, especially in comparing services to certain kinds of products. 5 Second, and more fundamentally, the distinction between products and services is often based on the contractual nature of the market transaction. At the most fundamental level, every product either provides services directly, or is combined with other products to generate services. Thus, an automobile is purchased as a product, but its value is based on the stream of services that it will provide over time.
All such durable goods may be leased or rented, in which case the service is more explicitly contracted for than in the case of a product transaction. The more expensive the product (e. g. , an airplane vs. an automobile), the less possible or economically desirable it is for an individual to access the services through a purchase. In fact, contracting for services from specialist providers is often a way to achieve economies of scale in providing services (or in making the underlying products) that would not be possible otherwise.
The general point is that the boundary between products and services is often a function of market and economic conditions, rather than the intrinsic characteristics of what is being exchanged. Previous analysts of this issue do recognize that the practical application of these conceptual distinctions in classifying services incorporates the contractual nature of the transaction. For example, in the case of utilities, Melvin (1995) points out that they are typically classified as services in the national income accounts, although gas, electricity and water are tangible goods.
In these cases, the distribution systems are a critical aspect of enabling consumption of these goods, and the contractual relationship is primarily for delivery services using the distribution systems. 238 Industrial Development for the 21st Century One exception to this blurring of the boundary between products and services is the case of personal services. To take a simple example, a practitioner of traditional Indian medicine (Ayurveda) may make a diagnosis based only on an examination of the patient, consisting of visual inspection and pulse reading, without any tools whatsoever.
No product is involved in this service provision. Such pure personal services, without the use of any capital or tools, are increasingly rare: a modern doctor’s services typically incorporate the use of large quantities of expensive equipment. Other personal services, such as those of lawyers and accountants, also involve the use of physical capital in their delivery, just as in the case of the utility services discussed in the previous paragraph. One idea that we develop in this section is the increasingly “industrial” or “manufacturing-like” character of such personal services.
It is also worth noting that knowledge services, which have traditionally been personal services, are becoming more “industrialized” through the use of information technology: we also develop this point further in this section. By these characterizations, we mean the development of standardization, routinization, scalability and replicability for these services. Besides the marketing literature, one area where the conceptual distinction between products and services has been explored in detail is that of international trade, where the categorizations are important for shaping international trade policies and agreements.
Locational characteristics are central to discussions of services in international trade. For example, Bhagwati (1984) argues that services can be divided into two classes: those that require physical proximity of the user and provider; and those that do not. Services that require physical proximity can be further divided into three groups: a) Mobile provider and immobile user, e. g. , transporting labour to a construction site; b) Mobile user and immobile provider, e. g. , a patient going to a hospital; c) Mobile user and mobile provider, e. g. , students and professors meeting in a university for lectures.
A variation on the above classification is the basis for the World Trade Organization’s (WTO’s) categorization of services trade, which categorizes different “modes of supply” (Sampson and Snape, 1985; and Sapir and Winter, 1994 – see section 6 of this chapter for details). Digital technology has also reduced the need for contact or physical proximity in many cases – these changes are at the heart of the new tradability of various knowledge-based services. Bhagwati (1985) was an early observer of the trend toward removal of proximity requirements through technological change, i. e. the increased richness and lower cost of long-distance communications. Melvin (1989, 1990) also explores this trend, characterizing services either as activities that require the double coincidence of time and location, or as activities that overcome constraints of time (e. g. , storage services) and location (e. g. , transportation services). He describes Services-led industrialization in India 239 knowledge or information services as also overcoming a constraint – that of ignorance. Digital communications technology is, of course, behind the new wave of international trade in information or knowledge services.
In the international trade context, the skill content of services can be an important distinguishing feature (Delaunay and Gadrey, 1991; Economic Council of Canada, 1991; Melvin, 1990). Some kinds of low-skill services are susceptible to routinization and automation, whereas others are not (Dossani and Kenney, 2004). The former are easier to trade. In some cases, service activities can be unbundled or “splintered” (Bhagwati, 1984) to allow the offshore outsourcing of the lower skill components (e. g. , software testing in the overall process of software development). One trend that also seems to be emerging, however, is trade in services with higher skill content, such as financial analysis or patent preparation. After one has clarified some of the conceptual characteristics of services, and their distinction from products, it is important to consider how measurement and classification are done in practice: “rather than discussing definitions, it may be more useful to take an operational approach and to examine what are actually called services in the national accounts and related statistical ources” (Griliches, 1992, p. 6). The Central Statistical Office in India provides a standard breakdown of services into 15 sub-sectors for that country (Gordon and Gupta, 2004), as shown in Table 1. 7 Note that this Indian classification does not include utilities, which are often classified as services elsewhere. The salient point is the heterogeneity of the activities that are classified as services, in terms of capital intensity, skill content, need for physical contact, and importance of tangible products as part of the service. . India’s growth experience: overview Since our interest is in the role of services in India’s recent growth, it is useful to have some perspective on the nature and extent of that growth. As we noted in the introduction, in the 1980s India’s growth sped up from a moderate or slow rate prior to that decade and has since ranked among the highest in the world. There are numerous detailed reviews of India’s growth experience in the last couple of decades, and we shall be selective and relatively brief.
These reviews often discuss causal factors, albeit typically without formal econometric modelling, with the importance of economic reforms in the growth acceleration being the subject of some controversy. In other cases, econometric techniques have been used to identify structural breaks in the growth process, but without any causal modelling (e. g. , Wallack, 2003). In this section, in addition to discussing aggregate growth performance, we also examine studies that focus on the performance of the service sector.
Panagariya (2005) provides a very careful appraisal of India’s growth experience in the 1980s and 1990s, and of the ‘revisionist’ view, articulated by DeLong (2003), Rodrik (2003) and Rodrik and Subramanian (2004a), that economic policy reforms in the 1990s were not key to India’s growth 240 Industrial Development for the 21st Century performance in that decade. Those authors further argue that India’s growth surge is properly understood as beginning in the 1980s, before the 1991 and subsequent economic reforms.
Based on a careful examination of the data, Panagariya reaches three conclusions about India’s growth experience in the last 25 years: 1. Growth during the 1980s was inconsistent, with the last three years of that decade contributing 7. 6 per cent annual growth, without which growth in the 1980s was only marginally better than that of the previous three decades. 2. The high growth in the last three years of the 1980s was preceded or accompanied by significant economic reform, including trade and industrial policy liberalization. . Growth in the 1980s was fuelled by expansionary policies that entailed accumulation of a large external debt and contributed to an economic crisis. Panagariya’s final conclusion from his review of policy changes and growth performance is that it was “the 1991 market reforms and subsequent liberalizing policy changes that helped sustain growth”. Ultimately, the positive impact of economic policy reforms seems to have been accepted, even in the revisionist view.
For example, while Rodrik and Subramanian (2004b) – based on their earlier analysis (2004a) – use a growth accounting methodology to evaluate and project India’s growth performance which de-emphasizes economic liberalization policies, they do not dismiss such policies. They focus on “meta-institutions” such as democracy and the rule of law, as well as conventional economic inputs such as human and physical capital, and productivity growth. In emphasizing these fundamentals,8 they highlight the infrastructure and human capital built up under the pre-liberalization policy regime.
At another level, however, their assumptions about the impacts of policy are not that different from those of Panagariya, since they state that “policy liberalisation will progressively erode the licence-quota-permit raj as a source of corruption and patronage that has had such a corrosive effect on public institutions. ” In addition to this indirect effect, they also attribute productivity growth directly to reforms that removed the “shackles on the private sector. ” The case of IT illustrates the positive impact of the 1990s reforms, with telecommunications reform and liberalization leading the way for that sector.
One of the best-known, most successful participants in this industry has documented the positive impact of liberalization, including general steps to ease the conduct of business (Murthy, 2004). While Panagariya shows that average growth rates are very sensitive to how one divides the period under consideration, one might argue that the data itself should be allowed to reveal when a growth acceleration occurred for India, rather than being determined subjectively. Time series methods to test for structural breaks are designed to achieve this. Wallack’s (2003) study
Services-led industrialization in India 241 is the most comprehensive along these lines, and examines time series data for 1951-2001. 9 The tests are applied to aggregate GDP data as well as to sectoral data. The latter, in particular, is of interest for understanding the recent pattern of economic development in India. Since these are pure time series estimations, there are no structural explanatory variables, but one can plausibly compare the identified break dates with what we know independently about policy changes or other macroeconomic events.
The results are summarized in Table 2 and provide some plausible connections between growth and economic environment changes (such as low interest rates, an investment surge, or the IT boom). As Wallack herself emphasizes, the results are not as robust as one would like, since other years often have F-statistics close to the maximum values. 10 However, there are two important conclusions. First, there is evidence for a break in the aggregate GDP growth rate in 1980 (and in 1987 for GNP).
Second, there is little evidence that this was caused by a break in the growth of any sector – instead, as T. N. Srinivasan has emphasized, much of the growth rate change probably came from a shift in activity from slower-growing to faster-growing sectors (Srinivasan, 2003). 11 The results of the statistical tests suggest some exceptions to this observation (Table 2). One is the breaks in financial, real estate and business services in 1974 and 1980. Wallack suggests these were the result of low interest rates in 1974, and an investment boom beginning in 1980. 2 There is also a break in trade, transport, storage and communication services in 1992: Wallack relates this to telecommunications reform and the growth of IT, but those factors may have come into play somewhat later. Clearly, this empirical approach needs to be explored further to understand the behaviour of the Indian economy. In addition to the time series evidence for structural breaks in some service sectors, there are more general empirical observations on service sector growth in India. For example, the share of services in gross capital formation declined from 57. per cent in 1950-51 to 48. 8 per cent a decade later, and further to 43. 7 per cent in 1970-71. Thereafter, however, the share has not changed that much, hovering around 40 per cent. One can make conjectures about changes in capital and labour productivity in the service sector (Hansda, 2002), but since the composition of the sector has changed substantially over time, and these sub-sectors have different capital intensities, it is difficult to reach any definite conclusion from the data as presented (Hansda, 2002, Tables 7 through 9).
Hansda also suggests that the terms of trade moved against the service sector in the latter part of the 1990s, with finance, insurance, real estate and business services and trade, transport, storage, and communication together contributing to that decline. Liberalization, competition and technological progress may all have contributed to this development. 13 A disaggregated view of service sector growth in India, using the CSO classification as described in Table 1 (Gordon and Gupta, 2004), gives a 242 Industrial Development for the 21st Century stronger indication of growth acceleration than the formal time series analysis.
Gordon and Gupta note that 12 of 15 service sub-sectors in India grew faster than GDP over the 50 year period beginning in 1951, but the growth acceleration in the 1990s, which was responsible for India becoming an outlier in the service sector share (as compared to typical developing countries at that income level), was the strongest in business services, communication, and banking services, followed by hotels and restaurants and community services (Gordon and Gupta, Table 5). These five sub-sectors together accounted for the entire acceleration in services growth in the 1990s.
They assert that growth in other sub-sectors in the 1990s was broadly similar to, or lower than, that in the previous decades. 14 This pattern of growth is considered further in Sections 4 and 5 of this chapter, in the context of service sector contributions and linkages in the economy. Two other aspects of the performance of India’s service sector are noteworthy. First, following an international trend, FDI inflows into India have moved increasingly away from manufacturing, and towards services sector, even as they have increased overall.
These flows have been heavily concentrated in telecommunications and financial services (Banga, 2005, Figure 7). Second, the structure of indirect taxation in India, with historically light or zero taxation of services due to the nature of constitutional tax provision, has meant that the recent growth of the service sector has been associated with an overall tax-GDP ratio below the norm for countries at similar income levels. 15 While services crossed 50 per cent as a share of GDP in the 1990s, they contributed only about 10 per cent of total tax revenue (Hansda, 2002, Table 11).
This situation is changing, with moves to change constitutional tax assignments and bring services broadly into the indirect tax base. In addition to analyses of aggregate and sectoral growth, summarized above, a recent review of India’s pattern of development (Kochhar et al. , 2006) focuses on skill intensity, and suggests that the nature of the economic policies followed by India from independence onward created a pattern of specialization that was already somewhat distinct from the typical developing country at comparable income levels.
This, in turn, shaped the country’s development pattern in the post-1980 reform period. In particular, India’s manufacturing sector was more diversified, more skill-intensive, and less (unskilled) labour-intensive than average. In fact, it is suggested that the data indicates that India’s service sector was smaller than normal up to the early 1980s. 16 Thus, while the last two decades have seen a continuation of unusually skill-intensive growth, the rapid growth since the 1980s in services (still chiefly skill-intensive, however) represents a departure from the past.
Kochhar et al. examine the performance of the Indian economy in the 1980s and 1990s, starting from these apparently unusual 1980 preconditions. GDP per worker and total factor productivity (TFP) both grew more rapidly in these two decades, as compared to the 1970s. Cross-country level regressions for the year 2000 suggest that India’s services share of GDP was Services-led industrialization in India 243 significantly higher than the norm in that year (by 3. 8 percentage points), but only when a country size variable was included.
On the other hand, services employment was always significantly lower than the norm, and by a large magnitude, of 17 percentage points. This result is consistent, then, with the kind of skill-biased development observed in India pre-1980. Regressions using the change in the share of services also indicated that India was a positive outlier during the period 1981-2000 (Kochhar et al. , Table 5). According to this analysis, the 1980s and 1990s also saw an accentuation of the skill bias in India’s manufacturing, including whatever is measured of the informal sector.
A conjecture is that incomplete labour market reforms and continued skewed education sector spending might be underlying factors behind the observed pattern of growth, in which case it might be altered by policy changes. On the other hand, hysteresis in development paths might accentuate the observed biases in the future. State-level sectoral growth data seems to support the latter possibility: states such as Tamil Nadu and Maharashtra, according to this analysis, appear to be behaving more like advanced skill-intensive countries in their growth pattern.
The concerns raised by Kochhar et al. about India’s somewhat skewed pattern of development, and its implications for income and regional inequality, will be taken up in more detail in later sections, particularly in the context of employment, education and training (Section 7), social issues (Section 8), and broader lessons for development policy (Section 9). Here, we offer one word of caution on their analysis. Historical patterns of development and cross-country comparisons are an important component of empirical analysis. 7 Nevertheless, these comparisons may miss certain kinds of innovations that are not captured in the data. In particular, the development of modern business process outsourcing represents a rather new form of organizing economic activity, and an opportunity at many skill levels, that may not have been possible in the past. This goes back to our conceptual discussion in the previous section. To offer an imperfect analogy, Japan in 1950 was not viewed as a likely candidate for building a world class automobile industry. The industry was both capital-intensive and technologically advanced.
Nevertheless, Japan not only succeeded, but there were arguably positive spillovers from this effort to other engineering-intensive manufacturing. The imperfections in this analogy include substantial differences in initial conditions and social heterogeneity, but the caveat on extrapolating from the past should be kept in mind. 4. Understanding service sector growth and its impacts The discussion in the last section has suggested that India’s more rapid growth in the 1980s and, especially, the 1990s has been the result of policy changes as well as initial conditions.
The pattern of growth has been skewed toward the service sector, as well as toward skill-intensive activities. As noted in the introduction, India’s service sector was responsible for over 60 per cent 244 Industrial Development for the 21st Century of its GDP growth in the 1990s, well above the sector’s contribution in previous decades. A small number of service sub-sectors seems to have led the overall contribution of services to GDP growth. In fact, the five fast-growing sub-sectors in the 1990s accounted for effectively all of the more rapid services growth.
In this section, we begin with the empirical question of what factors can explain observed service sector and sub-sector growth in India. One possible answer is increased specialization, or splintering (Kravis et al. , 1983; Bhagwati, 1984), which, if it occurs across sectors, would alter the aggregate accounting. More precisely, if services components of manufacturing activity such as accounting, research and development, or logistics are splintered off and outsourced to other firms, they will be accounted for as service sector contributions to GDP, rather than being subsumed in manufacturing value added.
Using input-output coefficients constructed by Sastry et al. (2003), Gordon and Gupta (2004) provide one estimate of this effect, based on changes in those coefficients. The input coefficients of services in agriculture and manufacturing increased enough in the 1980s to add 0. 5 percentage points to service sector growth in that decade. 18 Using input-output coefficients constructed from 1998-99 data, we are able to repeat this calculation for the 1990s, obtaining essentially no contribution of splintering in the 1990s, if measured in this manner. 9 However, this data and methodology does not permit an analysis of the extent to which cross-country splintering, which became more important in the 1990s, e. g. , through offshore outsourcing of business services, would explain the observed patterns of service sector growth. In the latter case, there is a real shift in economic activity to India, whereas domestic splintering is more of an accounting change. Even in the latter case, however, when specialization is associated with efficiency improvements, purely domestic splintering may reflect a positive economic change.
Along with increased specialization, another basic reason for an increasing share of services in GDP as an economy grows is a higher than average income elasticity of demand. 20 Thus, the share of services in private final consumption expenditure grew faster in the two decades beginning 1980 than did services as a share of GDP (Hansda, 2002, Table 5). Aside from a big jump in the share of the sub-category “operation of personal transport equipment” in the 1980s, the shares of different kinds of final consumption services were relatively stable in this period (Hansda, 2002, Table 6).
Gordon and Gupta (2004, Table 8) estimate that final consumption of services grew at a similar rate to overall services in the 1990s, but slower in the 1980s, which would be consistent with our calculation (and their argument) that splintering through specialization was a less significant contributing factor to services growth in the latter decade. Finally, the accelerated growth in services in the 1990s, along with a decline in their relative prices,21 is not consistent with a purely demand driven explanation of services growth. A third possible factor in explaining service sector performance in India
Services-led industrialization in India 245 is the role of policy liberalization. Based on an industry case study (Murthy, 2004), we have noted the importance of liberalization in general, and telecommunications reform in particular, as a factor in the growth of IT and ITES. Gordon and Gupta (2004) make essentially the same point using patterns in the data. In the absence of any good measure of liberalization, they note that the communications sub-sector grew rapidly in the 1990s coincident with deregulation, that the private sector share in the service sector grew in this period, and that FDI was positively correlated with services growth.
This period also saw a rapid increase in service sector exports, which may have been aided by domestic liberalization: the international aspects of India’s service sector growth are taken up in Section 6. Some further quantification of the contribution of the various factors affecting services growth (splintering, high income elasticity, exports and policy reforms) in the 1990s is provided by Gordon and Gupta through a basic growth accounting exercise. They first estimate a pre-1990 trend growth rate of about 1. 5 per cent for service sub-sectors that accelerated in the 1990s.
This is assigned to splintering and high income elasticity, as relatively constant factors. The residual is about 1. 75 percentage points of service sector growth. Growth due to exports is estimated at about 0. 5 percentage point. 22 The remaining residual, of 1. 25 percentage points, is assigned to policy liberalization and technological progress. These estimates are somewhat rough and ready, but can provide a basis for examining more detailed econometric work on the contributions of and to service sector growth.
A set of sub-sector specific and pooled time series regressions to estimate the factors determining services growth are consistent with the conclusion of a positive impact of reform on services growth (Gordon and Gupta, 2004, Tables 9 and 10). There also appears to be a positive impact of industrial growth on services growth, as one might expect. 23 A complementary question to that of the impacts of industrial growth on services growth concerns the impact of services on manufacturing production and productivity (Banga and Goldar, 2004).
These authors perform a sources-of-growth analysis, in which services are included as an input to manufacturing. 24 They first use panel data for 148 three-digit level industries covering 18 years, 1980-81 to 1997-98, to estimate a production function, and based on this econometric estimation, the sources of growth analysis is carried out. It is estimated that, although service inputs contributed little to the production of the registered manufacturing sector during the 1980s (only 1 per cent of output growth), the contribution of services increased substantially in the 1990s (to about 25 per cent of output growth).
This, in turn, implies that excluding services inputs overstates the extent of manufacturing TFP growth in the 1990s. In fact, when the authors regress their TFP index on a set of explanatory variables that includes the ratio of services input to employment, a statistically significant positive relationship is found between services input and industrial productivity, though the estimated elasticity is 246 Industrial Development for the 21st Century not large. These results suggest that the increasing use of services in manufacturing in the 1990s favourably affected TFP. 5 The results of this analysis also suggest that trade reforms played an important role in increasing the use of services in the manufacturing sector. It is not clear, however, to what extent this increased use reflects splintering, or other changes in the organization of manufacturing. 26 Finally, Bosworth, Collins and Virmani (2007) examine the role of services as part of a detailed, up-to-date growth accounting for India. A major result is that TFP growth in the service sector has been very high (almost 3 per cent per year) in recent decades, while the contribution of capital is relatively low.
Since half this sector’s growth has come in ‘traditional’ sub-sectors such as trade, transportation, and community and personal services, where great productivity improvements might not be expected, they argue that price increases in services may be underestimated, leading to an overestimate of real growth in this sector. If correct, this view has serious implications for overall Indian growth measures. However, an alternative explanation, discussed later in this section, may lie in reductions in transaction costs associated with policy reforms.
Combined, the empirical results summarized above suggest that services have not only played a positive, direct role in India’s economic growth in the last decades, but may have also helped spur manufacturing growth. While the latter impact was not visible in aggregate data for the 1990s (barring a two-year boom in 1994-96), more recent data suggests stronger manufacturing growth (Ministry of Finance, 2006). Furthermore, since the growth of the services sector has been more robust than that of manufacturing, a relatively clear picture emerges of services driving the Indian growth process.
There are serious concerns about skill intensity and employment, which we consider in Section 7, but the central conceptual question from a growth perspective is whether the empirical experience can be related to an underlying theoretical understanding. The traditional development paradigm of “agriculture to manufacturing to services” (e. g. , Kuznets, 1959; Kaldor, 1966; Pack and Westphal, 1986) not only makes manufacturing central to the transition to modernity, but also necessary for sustained productivity growth, since an increasingly important service sector is viewed as a symptom of a mature economy and slower growth (Baumol, 1967).
Therefore, in addition to the above empirical studies, we also consider the conceptual role of services, and information technology in particular, in reducing transaction costs and improving productivity in manufacturing. In fact, a recent major empirical study of the United States (Triplett and Bosworth, 2004) concludes that services sector productivity increases were responsible for the strong growth in the US in the latter half of the 1990s, with information and communication technologies (ICTs) playing a significant role in this productivity effect.
One aspect of this impact is the possibility that technological change, including the use of ICTs in particular, has Services-led industrialization in India 247 made some services more routinized, and therefore more like manufacturing. For example, a large-scale call centre is run more like a factory, reaping economies of scale, rather than a traditional services organization, which would have traditionally provided customized or differentiated services at a relatively small scale. While there is some scepticism about the long-run impacts of ICTs (e. . , Joshi, 2004), there seems to be a well understood case for focusing on ICTs, providing both services and products, in the growth process. For example, Dasgupta and Singh (2005a, footnote 8) state that, “IT [information technology], however, produces both new products and processes. Nevertheless, from the Kaldorian perspective, what is important is whether IT, be it product, service or process, and manufacturing are subject to increasing or decreasing returns to scale, to dynamic economies of scale and to spill over effects for the rest of economy. While the macroeconomic data may not answer this question, microeconomic studies of developed countries directly establish the positive impacts of IT use on productivity (e. g. , Bresnahan, Brynjolfsson and Hitt, 2002; OECD, 2004). Another key aspect of ICTs’ development impacts comes through reductions in transaction costs. This term can be interpreted broadly, to include various kinds of costs of communication, transportation, and other aspects of exchange, such as costs of searching for options, negotiating contracts and enforcing agreements. 7 Singh (2004a) constructs an illustrative model to show precisely how transaction cost reductions can increase short run efficiency, as well as boost the long run steady state of the economy. The simplest and most basic aspect of transaction costs is that they result in resources’ being used up in exchange, and drive a wedge between the supply price and the demand price in a competitive market. If transaction costs are high enough, the relevant market will not exist, in the sense that the quantity exchanged is zero.
If the transaction cost consists purely of resources used up in the process of exchange, then the welfare loss from its existence consists of the usual Harberger deadweight loss triangle, plus the transaction cost rectangle, which can be substantially larger. If the transaction cost is actually a payment to an intermediary, or a tax imposed by the government, then the welfare loss is just the Harberger triangle. Romer (1994), in an influential paper, argues that even if the rectangles represent redistribution, and not a resource cost, there is a welfare cost that goes well beyond the usual Harberger triangles.
The basic idea is that if there is a fixed cost of producing a good, transaction costs can reduce operating profits to the extent that some goods are no longer produced, leading to a larger loss of welfare than would otherwise be the case. From a development perspective, the main implication of domestic transaction costs, i. e. , their effect on the equilibrium number of intermediate goods, and hence on output and welfare, must be sought in a dynamic model. Romer hints at this, but the first formal model is that of Singh (2004a), building on the analysis of Ciccone and Matsuyama (1996). This uses a dynamic 48 Industrial Development for the 21st Century monopolistic competition model, in which an economy that inherits a small range of differentiated intermediate inputs can be trapped into a lower stage of development. In particular, the model may have multiple steady state equilibria, and the starting point of the economy can determine which steady state is approached. Singh (2004a) shows that if transaction costs determine the initial number of varieties of the intermediate good, then reducing transaction costs can jump start the economy, moving it to a path that leads to a higher steady state.
In addition, transaction costs will affect the long-run equilibrium number of varieties, and hence the level of the steady state. Transaction costs can reduce the long-run steady state level of the economy, and even arrest the process of development. 28 The use of ICTs, by reducing transaction costs, can mitigate both these effects. 29 Note that, although Singh (2004a) emphasizes ICTs, improved transportation and distribution can also bring down transaction costs.
The connection to the empirical work on the growth of the Indian service sector reviewed earlier in this section should be clear: the fast growing services have included some sub-sectors that are critical determinants of major aspects of transaction costs for the Indian economy. From this perspective, one might also view infrastructure constraints in transportation as one manifestation of high transaction costs. 30 In addition to affecting transaction costs, certain kinds of services may impact the rate of innovation, so that their growth is critical in determining development.
The obvious perspective to take in this case is that of the endogenous growth literature, which emphasizes the role of innovation (e. g. , Grossman and Helpman, 1991; Romer, 1990). In this context, the critical aspect of services growth is their knowledge intensity. The comprehensive study of Dahlman and Utz (2005) analyzes the development of a “knowledge economy” in India. These authors emphasize R&D quite broadly, and range over issues of governance institutions and education, as well as the national system of innovation.
Implicitly their treatment of education partially deals with the concerns about skill intensity expressed by Kochhar et al. (2006), by highlighting the importance of skilled labour in development: we postpone details until Section 7. Their perspective converges with the transaction cost approach in emphasizing the need to improve the ICT infrastructure, which will in turn speed and enhance the acquisition, creation, dissemination, and use of knowledge. Singh (2003) provides a formal analysis of how improving ICTs can enhance growth.
His growth model, designed to capture the special role of ICTs in economic growth as an enabler of efficient communication and storage of information, is an extension of the recombinant growth model of Weitzman (1998). The central idea of this approach is that new ideas are formed through combinations of old ideas. A key property of this formulation is that the increase in the number of ideas is faster than geometric growth (Weitzman, 1998, Lemma, p. 338). The rate at which potential ideas are converted into new ones depends on a “success rate”, which is a function
Services-led industrialization in India 249 of the current level of resources spent per potential new idea, i. e. , the level of R&D. In Weitzman’s model, all ideas are the same, and the actual number of new ideas is given by the number of potential new ideas multiplied by the success rate. Singh (2003) modifies Weitzman’s model to allow the stock of ICT knowledge to independently affect this success rate, so that ICT gives the growth process an extra ‘kick,’ beyond that which comes from recombinant growth in general. 5. Linkages and spillovers
The analysis of the previous section can be extended and complemented by explicitly considering linkages in an input-output framework. The goal is to quantify the extent to which services, manufacturing and agriculture have spillover effects on each other. A single year’s input-output data provides this information in a static context, whereas examining changes in coefficients over time can provide some insight into structural changes in the economy. For example, Sastry et al. (2003)31 construct input coefficient matrices for four years: 1968-69, 1979-80, 1989-90 and 1993-94 (see Table 3).
The basic observation is that, over this period, agricultural production became more industry- and services-intensive, whereas industrial production became less agriculture-intensive and more services-intensive. For the production of services, the main change was greater service intensity. Some of these changes took place more uniformly through this time period, but the greatest changes in the Indian economy’s production structure, by this measure, came in the 1980s and 1990s. These observations are, of course, consistent with the analyses presented earlier. 2 The input-output approach is applied at a much more disaggregated level by Hansda (2001). His goal is to address the larger question of the sustainability of services-led growth for the Indian economy. He uses input-output transactions tables for 1993-94, disaggregated at the level of 115 activities (22 in agriculture, 80 in industry, and 13 in services), as well as at the level of 10 broad sectors. At these disaggregated levels, Hansda’s analysis suggests that the Indian economy is quite services-intensive with industry being the most services-intensive sector.
Table 4 (adapted from Hansda, 2001, Table 5, and partially updated to 1998-99), summarizes the pattern of above average intensity for sectors falling into the three overall categories of agriculture, industry and services. The main conclusion, on the services intensity of industrial activities, is validated by the bold numbers in Table 4, indicating a large fraction of industrial activities with above-average services intensity. This conclusion is in line with the Sastry et al. aggregated data for 1993-94, and our calculations for 1998-99 (Table 3).
One can also examine the data at an intermediate level of aggregation, as in Table 5 (adapted from Hansda, 2001, Table 6 and partially updated). The numbers here are the direct (for 1993-94 and 1998-99) and indirect (for 1993-94) sectoral intensities, corresponding to the numbers reported by Sastry et al. , but 250 Industrial Development for the 21st Century at a level of disaggregation that allows a greater understanding of the pattern of linkages of the different services sub-sectors. Note that the columns are further aggregated at the level of the three sectors – agriculture, industry and services.
The different aggregation gives intensities somewhat at variance with the Sastry et al. calculations, but the conclusion of the importance of services still emerges clearly. Manufacturing, construction and “electricity, gas and water supply” are all services intensive. On the other hand, among services activities, “transport, storage and communication” are together quite industry intensive, while “personal, social and other services,” and “trade, hotels and restaurants” are also somewhat industry intensive. As one might expect, the level of linkages to and from agriculture, as indicated by these intensities, is quite low.
Surprisingly, the same is true for “financing, insurance and real estate. ” The results for 1998-99, for direct intensities only, indicate relatively little change in magnitudes of agriculture and manufacturing intensities, as well as for their rankings, which capture relative intensities. Services intensities seem to have changed somewhat more from 1993-94 to 1998-99. In particular, “transport, storage and communications,” “financing, insurance and real estate,” and “personal and social services” all became more services intensive in absolute and relative magnitude.
The level of linkages to financial and real estate services remained relatively low, however. Given the expansion of this sub-sector in the 1990s, and policy statements about creating an international financial hub in Mumbai, it would be important to investigate how these linkages have changed since 1998-99. 33 One can also calculate backward and forward linkages more formally, using a procedure due to Rasmussen (1956). Backward linkages refer to stimulus from a sector to its input-providing sectors, while forward linkages capture stimuli in the opposite direction, i. e. , downstream in production.
Linkage measures are constructed as indices. An index value greater than one indicates a greater than average linkage. 34 Applying this method to 1998-99 data,35 we find that 43 out of the total of 115 activities – 6 of 22 agricultural activities, 28 of 80 industrial activities and 9 of 13 services activities – had relative backward index values above one. In fact, 8 of the 9 service activities exceeded the threshold by substantial margins. Therefore, we may conclude that services activities had proportionately the largest inducing effect on the rest in terms of backward linkage.
The results for the forward linkage indices are qualitatively similar, in terms of the relative importance of services: 19 of 115 activities – 5 of 22 agricultural activities, 5 of 80 industrial activities and 9 of 13 services activities – had high forward linkage indices. Hansda’s aggregated results for 1993-94 (his Table 7) are quite similar to these disaggregated indicators, finding strong backward linkages for “personal, social & other services” and “transport, storage & communication” in particular.
Forward linkages from services are weaker at this level of aggregation, emphasizing the need to consider as disaggregated a view of the economy as possible. Services-led industrialization in India 251 Two additional pieces of analysis are possible. Hansda (2001) calculates the coefficients of variation for the linkage indices. These coefficients measure the evenness across sub-sectors of a particular sector’s purchases (backward linkages) or sales (forward linkages). The low proportion of services activities with low backward coefficients of variation suggests that the backward linkages from services tend to be concentrated, i. . , their purchases are from a relatively small segment of the economy. Hansda also finds that the forward linkages from services are more evenly spread. Second, given conceptual criticisms of the Rasmussen indices of linkages (e. g. , Claus, 2002), he constructs an alternative index of vertical integration, based on the work of Heimler (1991). The results are that 14 out of the 115 activities have index values higher than the average, with 7 of those activities in the services sector, 6 in industry and 1 in agriculture. This supports the view that the services sector has the largest multiplier effect on the rest of the economy.
One way to interpret the results from the disaggregated input-output analysis is in keeping with the idea that India is a high transaction cost economy (Singh, 2004a). Transportation, trade and communications in India, all sectors which have high impacts on transaction costs, have all been markedly inefficient, held back by inefficiencies in government and by barriers to private participation. The 1980s and 1990s saw this situation begin to change, and a plausible interpretation of Hansda’s results is that, through linkages, this change has stimulated the entire economy to some extent.
One must bear in mind that this kind of input-output analysis is static, and there is no growth mechanism postulated. Nevertheless, one can argue that, to the extent that growth is driven by innovation, innovations in the services sector are likely to have positive implications for the growth of the rest of the economy. We can explore this hypothesis more explicitly by selectively perturbing input-output coefficients and tracing out the implications. The theoretical basis for the following exercise is developed in an important paper by Majumdar and Ossella (1999).
They show that in an input-output economy, under some further conditions,36 the long-run optimal growth factor (i. e. , the growth rate plus one) is given by g = ( )1/ , where is the discount factor for the representative consumer, is the parameter of the constant elasticity one-period utility function (equivalent to the coefficient of relative risk aversion), and is the inverse of the largest eigenvalue of the input-output matrix. Efficiency gains in sector j, such as would result from any kind of innovation, are modelled as proportional reductions in all the input requirements of that sector, i. . , the elements of column j of the input-output matrix. The maximal eigenvalue decreases in this case, implying an increase in the optimal growth rate of the economy. Since the other two parameters are unchanged, the ranking of growth rates is equivalent to the ranking of the ’s across all the individual sector perturbations. Hence, the sectors for which exogenous efficiency improvements would have the greatest growth impacts can be clearly identified. These sectors are termed “leading sectors” for the economy. 37 252 Industrial Development for the 21st Century
Majumdar and Ossella, using data from 1989, consider both 2 and 5 per cent reductions in input requirements, and identify the five leading sectors of the Indian economy for that year in each case as, in order of growth rate impacts, (1) electricity, gas and water supply, (2) iron and steel, (3) paper and paper products, (4) other chemicals, and (5) other manufacturing. Using = 0. 9,38 the growth factor for a 5 per cent reduction in input requirements in the electricity, gas and water supply sector, relative to the original growth factor, is 1. 081. To translate this into a growth rate impact, assume that the initial optimal growth rate was 6 per cent – the efficiency gain would change the optimal growth rate to 6. 9 per cent. We carry out the analysis along the lines of Majumdar and Ossella, using input-output data for 1998-99. We have 115 sectors instead of the 60 sectors used in 1989. Nevertheless, major sectoral classifications are quite similar. We only calculate the impacts of 5 per cent efficiency improvements, but consider the 10 leading sectors of the economy.
The results are presented in Table 6. Electricity, gas and water supply remains the most important leading sector, and its growth impact is estimated to be even higher than for the earlier data. Several heavy industry sectors feature in the top 10, paralleling the earlier results. The new feature of our results, however, is the prominence of services. Four services sub-sectors are in the list, and their nature and presence strongly bears out the transaction cost interpretation provided previously in this section.
As an alternative to using disaggregated input-output data, it is also possible to undertake a dynamic analysis of sectoral growth and linkages within the Indian economy through reduced-form time series modelling. Sastry et al. (2002) add a (non-oil) export sector to agriculture, industry and services, and estimate this four equation specification for annual data from 1981-82 to 1999-2000. Linkages from agriculture and industry to services, and agriculture and services to industry are incorporated by including those sectors’ outputs as causal variables.
On the other hand, agricultural growth is assumed to be determined independently. Their empirical results for the industrial sector come from a straightforward levels regression, and indicate that a 1 per cent rise in services (agriculture) would stimulate industrial output by 0. 40 (0. 25) percentage point. The services sector presents several additional challenges in terms of assumptions required for empirical analysis. Sastry et al. focus only on construction, “trade, hotel and restaurants” and “transport, storage and communication” as endogenous ervices, subject to linkages, leaving finance, and “community, social and personal services” as exogenous to the system. 39 The results for the entire sample period, with the equation estimated in growth rate terms, suggest that a percentage point increase in the growth rate of services would lead to a 0. 32 percentage point increase in the growth rate of industry. Given the substantial changes in the input-output coefficients over this period, and tests that reject parameter stability, they re-estimate the Services-led industrialization in India 253 quation for the 1990s only. The impact of services growth now increases to 0. 49 percentage points, and they argue that this is consistent with the 199394 input-output tables. In fact, the match is less good than the authors suggest, since the coefficient for agriculture is much higher in the time series specification than in the input-output table. Nevertheless, there is broad consistency from this approach concerning the importance of services. The input-output linkage analysis is limited by its focus on what are essentially pecuniary externalities.
As noted by Dasgupta and Singh (2005a), the Kaldorian hypothesis of the mechanisms of growth through structural change emphasizes technological externalities or spillovers, rather than pecuniary externalities. In this view, technological externalities are what made manufacturing special in the past, and now explain why ICTs (whether acting through products or services or both) are a special driver of growth. 40 This perspective is also consistent with the analysis of Singh (2003) with respect to the special role of ICTs in the innovation process, as discussed in Section 4.
In this context, Singh (2005) also surveys fieldwork and case studies which suggest that there are spillovers from ICTs in rural areas of India. In particular, the introduction of a range of rural ICT-based services not only reduces transaction costs, but also provides spillovers through knowledge acquisition and information access. It has also been argued (e. g. , Singh, 2003) that there were substantial spillovers from the IT services industry in India to IT enabled services (ITES), as well as to knowledge intensive sectors in general.
These included specific organizational expertise, customer knowledge, and general reputational effects. 41 In turn, there are linkages and spillovers from ITES to sub-sectors such as construction and transportation, not only through demand stimulus but also by creating a different set of requirements and expectations. 42 6. International trade in services Theoretical considerations and empirical evidence both suggest a positive role for trade liberalization in India’s growth since the 1980s.
Theories of comparative advantage and inter-industry trade (based on product differentiation and economies of scale) do not necessarily distinguish between agriculture, manufacturing and services as likely export sectors for a developing country, though developing nations are almost invariably those with a relative abundance of unskilled labour. Since much of India’s export growth has come in skill-intensive services,43 the international aspects of India’s “services-led” growth bear examination from empirical and theoretical perspectives.
Panagariya (2005) summarizes India’s experience with external liberalization, starting in the late 1970s. He documents the acceleration in India’s growth rate in the 1980s, and ties this improvement to the external liberalization that took place. Several studies (e. g. , Joshi and Little, 1994; Chand and Sen, 2002) formally establish a positive linkage between opening the economy and productivity improvements in manufacturing: this supports the conven- 254 Industrial Development for the 21st Century ional view connecting trade liberalization and development through improvements in manufacturing. Many developing countries have had to struggle with maintaining balance of payments equilibrium. In India’s case, balance of payments crises in the 1970s and before tended to shape a policy of keeping the Indian economy substantially closed to international trade and investment. In this respect, one of the most important aspects of services sector (particularly IT and ITES) growth in India has been its favourable international trade and balance of payments consequences.
Hansda (2002) provides some basic documentation of the change in the role of services in India’s international trade. For example, India’s share in world export of commercial services doubled from 0. 6 per cent in 1990 to 1. 2 per cent in 2000, while its share in global merchandise exports went up only marginally, from 0. 5 per cent to 0. 7 per cent, in the same period. 44 Data going back to the 1950s (Hansda, 2002, Table 10) show that, while India has run a persistent deficit in merchandise trade, there has been a consistent surplus in trade in services.
This surplus, only 10 per cent of the merchandise trade deficit in 1990-91, grew to onethird of that deficit by 2001-02. In aggregate, the ratio of services exports to merchandise exports increased from 25 per cent to 45 per cent in that period. Much of this increase came in the category of miscellaneous services, which include IT services and ITES such as business process outsourcing (BPO) and call centres. 45 Hansda also notes the higher positive balance in travel services, and suggests this may be related to a more favourable exchange rate, which was also a consequence of liberalization.
Travel connected to IT services and ITES may also have contributed to this increase. Numerous authors (e. g. , Bhagwati, 1985; Gordon and Gupta, 2004; Dasgupta and Singh, 2005a) note the increased ability to deliver services over long distances at a reasonable cost as a result of technological progress in ICTs, which has led to increased worldwide trade in services. Gordon and Gupta estimate that, in India, exports of services (in U. S. dollars) grew by an average of 15 per cent a year in the 1990s, compared with 9 per cent annually in the 1980s. Using the Sastry et al. nput-output matrix, and making some assumptions about input usage for services exports relative to services overall, they estimate the contribution of exports to annual average services growth to be about 0. 2 percentage points in the 1980s, and 0. 6 percentage points in the 1990s. In their regression analysis, they find that exports of services were highly significant for explaining overall growth of business services, which includes IT services and ITES – seemingly an unsurprising result. On the other hand, the growth of other kinds of services did not show much sensitivity to exports.
We have suggested that, in many respects, the recent services growth in India comes from outsourcing by developed country firms that has characteristics of manufacturing, namely scale and routinization. Even in general, trade in services can be analytically treated as similar to trade in goods. This Services-led industrialization in India 255 includes trade based on comparative adva