Catalog

Looking back

This section presents stylized facts about structural change in developing Asia over the past 35 years. It views the region's transformation through: movements in the composition of output and employment; the speed and breakdown of labor productivity growth; the pace of technological change; and developing patterns of specialization and diversification. These multiple changes are linked in subtle ways. Differences and changes in labor productivity provide incentives for resources to shift across sectors. Productivity growth, in turn, is linked to the underlying pace of technological progress and upgrading, but also to the mix of output and the creation of new activities, reflected in emerging patterns of specialization and diversification.

With regard to the data, those for industry and manufacturing are generally much better than for services or agriculture. Also, because of variable availability of data, country samples and time periods sometimes differ. This is seen perhaps most clearly for the Central Asian republics: since they were not independent states 35 years ago, information on their experiences is limited. Small economies in the Pacific and in other places are ill-served by data, too. Throughout, incomes are measured at market exchange rates in constant prices, using the World Development Indicators of the World Bank.

Movements of output and employment shares

Figures 3.1.1, 3.1.2, and 3.1.3 summarize graphically movement of output and employment across agriculture, industry, and services in developing Asia over the period 1970-2004, as per capita incomes change. Developing Asia's experience is set against the background of broad international patterns over the same period.

These data reveal a number of interesting features.

Most immediately, evolving patterns of specialization in developing Asia generally conform to wider international patterns of structural differentiation and change over the same period. But developing Asia's patterns depart from wider global averages in two ways. First, highincome countries in developing Asia tend to have smaller agricultural output and employment shares than high-income countries elsewhere. This is largely a function of differences in geography and agro-climatic conditions. Second, and perhaps more interestingly, developing Asia tends to be more industrialized than other parts of the global economy for given levels of per capita income. This is particularly true at lower levels of per capita income. But developing Asia also has a number of countries that have low industrial shares for their income levels (low, middle, and high). This reflects the presence of countries where industrialization has stalled or been retarded; the microeconomies of the Pacific that have virtually no industry but mid-level incomes; and the highly advanced service economy of Hong Kong, China.

Looking at the "cross-section dynamics," the data broadly confirm that agricultural output and employment shares tend to be smaller at higher per capita incomes, while the shares of industry and services tend to be larger. The rate at which agriculture shares taper off with

Finally, by comparing the evolution of shares for different countries at matching income levels, it becomes clear that with the passage of time, the speed of structural change has accelerated. This point is obvious when the experiences of the fast-growing economies of East Asia are compared with those of rich industrialized countries. East Asia compressed into the space of little more than a generation changes that took well over a century for older "industrialized" countries. Late starters have the advantage that they can copy those ahead and advance at a quicker pace. More recent comparisons suggest that this acceleration has continued. For example, higher industry shares are now being seen at lower per capita incomes than before.

These stylized facts, generated from a cross-country panel, are not necessarily a good guide to the evolution of economic structure in any particular country. The experience of developing Asia shows enormous variation both across countries and over time. As there are so many factors that could influence the pace and direction of structural change, explaining why some countries change quickly while others do not, and why they lean in a particular direction, requires in-depth study at a country level.

Dimensions of labor productivity growth

Differences in labor productivity (as well as returns to capital) across sectors are important catalysts of structural change. Aggregate labor productivity movements for selected countries in developing Asia are shown in Figure 3.1.4 and are compared with labor productivity for OECD, which approximates the productivity frontier. Aggregate labor productivity movements reflect the confluence of many factors as well as all the background conditions ("social capabilities") that influence them. As resources are reallocated across sectors, aggregate productivity changes occur. But changes in aggregate productivity will also depend on how productivity evolves at the sector level, i.e., on what products are produced and how they are produced.

Box 3.1.1 explains concepts of productivity convergence and catch-up. Two broad classes of country can be identified in Figure 3.1.4: those that are catching up or converging on the OECD frontier, and those that are making little headway in closing the gap. Among the catching-up countries themselves there are those that have progressed quickly and have substantially closed the gap and those where the gap is closing but is still wide. The NIEs have substantially closed the gap, though the Korea and Taipei,China still trail a little.

Relative and level measures can produce different pictures. Take Malaysia, the economy with the highest level of labor productivity outside OECD and the NIEs. Between 1980–1984 and 2000–2004 Malaysia's relative productivity improved from 16% of the OECD average to 21%. Malaysia is indeed "catching up. "But over the same period, the level productivity gap with OECD widened, from $28,823 in 1980–1984 to $36,904 in 2000–2004. Once Malaysia's relative productivity gets to about one third of the OECD average, the level gap will, though, start to close.

In a number of other countries too, including PRC, India, and Thailand, catch-up is occurring as level differences in productivity become wider. But again, if the current trajectories continue into the future, level gaps must eventually close.

But in some countries convergence is not occurring. Over the sample period, the Kyrgyz Republic and the Philippines fall into this category, and have lost ground in relative as well as level terms. In the postcrisis period, Indonesia, which had been converging, begins to fall further behind. Pakistan, too, has made little headway in closing labor productivity gaps. Unless these trends are reversed, level gaps in productivity levels will widen indefinitely. There is no evidence in these data that countries that start the period with lower initial levels of labor productivity catch up fastest.

Comparing aggregate labor productivity growth over time and across countries gives some broad clues as to how countries are faring, but for a more refined understanding it is necessary to drill beneath the aggregate numbers to see what is happening at the level of individual sectors (and the manufacturing subsector). Figures 3.1.5, 3.1.6, and 3.1.7 present comparable data for labor productivity gaps in industry, manufacturing, and services.

Trends in industrial labor productivity correlate quite closely with the aggregate picture but also show up some important differences. In particular, industrial catch-up for the PRC is proceeding much faster than it is for India. Within the NIEs, Singapore's industry now matches OECD productivity levels. Industrial productivity gaps for the ASEAN-4 countries are generally much smaller than the aggregate productivity gaps, and for Malaysia and Thailand are converging with the frontier. Again in Indonesia and the Philippines, industrial productivity gaps have widened over the sample period. Earlier gains by Indonesia fall away at the start of the 1990s. For the remaining countries, the relative industrial productivity gap has narrowed in Azerbaijan, Pakistan, and Viet Nam. After a relapse and a widening of the gap, the Kyrgyz Republic closed the gap a little between 2000 and 2004.

For manufacturing, the story for the NIEs barely changes. They have caught up steadily with OECD and gaps are now small, with Singapore in fact showing higher labor productivity levels than the OECD average. In the case of the PRC and India, the labor productivity gap is less for India than for the PRC, the reverse of what was observed for industry, but the PRC is catching up with India. Though manufacturing labor productivity gaps between India and OECD have been cut, catch-up has decelerated. Among the ASEAN-4, the gap is least for Malaysia and largest for Indonesia. The gap for the Philippines widens over the sample period. In Indonesia, manufacturing labor productivity stagnates in the postcrisis years, and the gap begins to widen. In Pakistan, too, there is evidence of stagnating labor productivity. The gaps are wide for other countries, but are closing in relative terms.

The story is more complicated in services and the data are possibly less reliable, given the well-known difficulties of measuring services output, and hence labor productivity. In OECD, the data suggest that services labor productivity growth is slower than in industry and manufacturing. In most of developing Asia, services productivity is also lower than in industry or manufacturing. Perhaps this reflects a high incidence of underemployment or disguised unemployment in the services sector as well as underlying technological conditions.

Among the NIEs, Hong Kong, China now has levels of services labor productivity higher than OECD. Taipei,China is very close to the OECD frontier, but Korea seems to lag a long way behind and the gap is not closing quickly. The fragmentary data for the PRC and India suggest that services labor productivity is lower in the PRC than in India. Catch-up with OECD is glacial. In the ASEAN-4, Malaysia is the only country that appears to be catching up, but has seen stagnation over the last decade. Earlier gains by Thailand would appear to have been partially given up. And it is difficult to detect any evidence of convergence for Indonesia (from 1990 on) and the Philippines. Among the remaining countries, only Pakistan has made headway.

The links between aggregate labor productivity growth and the sector components shown in Figures 3.1.5 to 3.1.7 are shown in Figure 3.1.8. However, sector contributions to the total depend not just on by how much their own productivity grows but also on their share in total output. More complicated decompositions also take into account shifts in employment across sectors.

It is immediately clear that the contribution of agriculture to aggregate labor productivity growth has been uniformly small. This is due both to the comparatively low output share of agriculture and to small labor productivity gains. For the Kyrgyz Republic, the data measure contributions to a fall in aggregate labor productivity (see Figure 3.1.4). The services contribution (which is a negative number) actually represents an improvement in its productivity. After the dissolution of the former Soviet Union, employment in the industry and services sectors of the Kyrgyz Republic contracted and many workers moved back to the farm. In the PRC, Indonesia, Korea, Malaysia, Thailand, and Viet Nam, industrial productivity growth dominates aggregate advances over the respective sample periods. But in Hong Kong, China; India; Kyrgyz Republic (where it is the only positive component); Pakistan; Philippines; Singapore; and Taipei,China, services make the largest contribution. This is because services have a large share in output in these economies, dilating the impact of modest gains in labor services productivity.

Labor productivity growth can be further broken down into within-sector and between-sector components. As workers move out of agriculture and into higher (labor) productivity activities in industry and services, aggregate productivity is lifted. This shift effect is commonly referred to as "Baumol's structural bonus. "Box 3.1.2 explains how to measure the bonus, and Figure 3.1.9 shows the breakdown of productivity growth into the bonus and within-sector productivity growth.

As seen in the figure, for most countries the contribution of within-sector labor productivity growth to aggregate labor productivity growth dominates the bonus that occurs as employment is reallocated from agriculture to industry and services. Yet the latter is by no means insignificant, accounting for more than 20% of the aggregate gains in productivity in six countries (the Kyrgyz Republic effect is negative). In Thailand, the structural reallocation effect (i.e., the bonus) outweighs within-sector productivity gains. As there is still a large reservoir of workers in agriculture in many of Asia's developing countries, and as Baumol's structural bonus is still largely untapped, it represents a potentially large source of future productivity gains.

Baumol's structural bonus is made up of the contributions of migration from agriculture to industry and from agriculture to services. This is shown in Figure 3.1.10. In developing Asia, the transfer of workers from agriculture to services has provided the largest gains. This is an important finding that helps explain the dynamics of employment in labor-surplus economies. In many countries of developing Asia, agriculture, not industry, has supplied abundant labor to services. Had the transfer of workers been from agriculture to industry, the structural bonus would have been larger.

In some countries, industry appears to contribute negatively to aggregate productivity growth through the reallocation effect. This reflects the movement of workers out of industry, most probably to services. The shrinking employment share in Hong Kong, China reflects the maturation of the economy. For the PRC, it reflects a base period (1987) when that economy still had a large number of workers employed by inefficient industrial state-owned enterprises (SOEs), who subsequently lost their jobs as these SOEs were downscaled or closed. The negative contribution of services in Singapore is an artifact of a calculation that divides a positive number for the services bonus by a total structural bonus that is negative (see the equation for Figure 3.1.9 in Box 3.1.2). A negative reallocation effect occurs in Singapore because of a falling share of industrial employment.

3.1.2 Baumol's Structural Bonus and the decomposition of productivity growth Following Chenery et al. (1986), the economywide growth rate of labor productivity can be decomposed into two parts: one, the sum of the growth rates of labor productivity within sectors (weighted by the sector's share in output); two, the effect of labor reallocation between sectors of different productivity, calculated as the sum of the changes in the employment shares of the sectors (industry and services) receiving employment moving out of agriculture multiplied by the differential in labor productivity with respect to agriculture. That is: where , ,, and. L is labor, Q is output, ´ denotes end-of-period values, 0 start-of-period values, ˆ time rates of change, and the suffixes sectors (A = agriculture, I = industry, S = services). The effect of the transfer of labor on productivity is what Baumol et al. (1985; 1989) call the "structural bonus. "Backward economies with a large pool of employment in low-productivity activities (normally agriculture) experience a bonus from structural change. This occurs because the transfer of labor from low- to high-productivity activities automatically increases the productivity level of the economy (i.e., a composition effect). This happens even if this transfer of resources is mainly a shift from agriculture to services (where productivity might not be significantly higher). However, as the logistic pattern of structural change drives resources toward services, and given that productivity growth in this sector is usually slower than in industry, countries eventually experience a "structural burden. "That is, as the share of labor in services increases, the aggregate rate of growth of the economy decreases.

Technological upgrading

Summing up, there is strong evidence that manufacturing in several economies (especially Korea; Malaysia; Singapore; Taipei,China; and to a lesser extent Thailand) in developing Asia has undergone important transformations and shifted output to more technology- and scale-intensive subsectors. In some other economies (the PRC and India, for example), the shift to more technology- and scale-intensive subsectors is taking place more slowly, but has started at a very low income base. In yet other countries, the evidence is lacking.

Patterns of specialization and diversification

Having linked the movement of output and employment to productivity gains and changes in technology, this subsection asks how these have been reflected in evolving patterns of specialization. The theory of comparative advantage predicts that as countries open up to trade, they will specialize in those activities that use intensively those factors that are in abundant supply.

Figure 3.1.13 graphs an index of output specialization against per capita income. Lower values indicate greater diversification. At the UNIDO 3-digit level, country experiences appear to vary widely. Increasing diversification (not specialization) is apparent as incomes rise at low levels in Bangladesh, India, Indonesia, Nepal, Pakistan, and Thailand. There is no economy that becomes more specialized within comparable low income ranges. Increasing specialization is only detected at higher income levels in Korea; Malaysia; Singapore; and Taipei,China.

Compared to the PRC at similar income levels, India has a more specialized pattern of manufacturing output, and is marginally more technologically sophisticated (Figure 3.1.11). Kochhar et al. (2006) have also shown that India has a more skill-based and capital-intensive pattern of production than the PRC.

Some differences appear when specialization and diversification are viewed through the optic of employment rather than output. Employment measures for both the PRC and India indicate a trend toward diversification. In terms of employment, Thailand exhibits increasing specialization rather than diversification. While the trend toward specialization remains in Malaysia and Singapore, the index is static in Korea, meandering around a stable average.

Imbs and Wacziarg (2003), present evidence that suggests that, at low levels of per capita income, economies tend to diversify and subsequently, as their income rises, they then specialize (i.e., show lower diversification). Graphically, this would be represented as a U-shape. Rodrik (2006) has interpreted Imbs and Wacziarg (2003)'s findings as suggesting that whatever is driving economic development, it is not comparative advantage. Individual country experiences in developing Asia do not directly fit the U-shaped pattern of specialization for value added (or employment) suggested by Imbs and Wacziarg (2003). But this is not surprising as data are being viewed over a comparatively short time frame. But if the data for all countries are combined (Figure 3.1.14), a distinctive U-shaped pattern emerges.

incomes in Figure 3.1.14 are calculated at market exchange rates, not purchasing power parity prices, and this may explain why the turning point is observed at a much lower level of income than in Imbs and Wacziarg. Given that principles of comparative advantage do not chime readily with developing Asia's experience, the structure of Asia's exports is now examined more closely.

Export complexity and diversification

Linked to the idea that comparative advantage may not be a particularly good predictor of how output structures evolve, Hausmann et al. (2005a) have argued that specialization patterns are partly indeterminate and may be shaped by idiosyncratic and country-specific elements. Specifically, there would appear to be a strong relationship between the level of a country's income and the sophistication or complexity of its "export package". Does the experience of developing Asia fit with these ideas?

The sophistication or complexity of a country's export basket is associated with the income or productivity characteristics of countries around the world that export similar goods. So if a country's export basket has a high share of goods that rich countries specialize in, it attracts a high score. Conversely, export baskets overweight in goods that poor countries specialize in attract a low score. Measurement issues are explained in Box 3.1.4.

Figure 3.1.15 graphs the country scores. Unsurprisingly, the results show that the NIEs have the highest scores, followed by Malaysia, Thailand, PRC, Philippines, India, and Indonesia. Moreover, the scores of all these countries have increased over the years, indicating an increasing level of complexity or sophistication in their export basket. For Bangladesh the trend is flat, and for Mongolia, Pakistan, and countries in Central Asia the index trends down.

This index of the complexity or sophistication of the export basket depends on how the underlying components are changing: whether individual exports are becoming more or less sophisticated over time, and on how the composition of a country's export basket shifts. Between 1986 and 2004, a clear pattern emerges of export diversification in all countries, but there are distinct differences across economies (Table 3.1.2). In the high-income economies, such as Hong Kong, China; Korea; and Singapore, the shift toward diversification as measured by the fall in the share of the top 10 exports in the total is modest. But the structure of exports in these countries was already quite diversified in the base period (1986). There is only modest diversification, too, in Bangladesh, Pakistan, Philippines, and Sri Lanka. But in these countries, the structure of exports is comparatively specialized. By comparison, PRC, Malaysia, Thailand, as well as India and Indonesia, show much greater diversification over the period, converging on the levels seen in higher-income countries. The Indonesian data in the early 1980s were probably influenced by oil-price shocks. To some degree, these patterns shadow the trend seen in the manufacturing output data (Figure 3.1.13 above).

Summary

Developing Asia's experience of change is complex. The evidence presented in this section indicates that there have been multiple transformations, some more obviously linked to productivity growth and economic catch-up with rich countries than others.

Essentially, output and employment in developing Asia have moved as per capita incomes have risen in much the same way as in other parts of the world. But on balance, developing Asia is a bit more industrialized and industrialization has begun at lower income levels than in other regions. Output shifts are much more advanced than employment shifts. Viewed through the lens of output, Asia is a services and industrial economy; through the lens of employment, it is still an agrarian economy and, increasingly, a services economy. In some countries, employment has shifted from agriculture directly to services, bypassing industry.

Developing Asia's performance on productivity growth is mixed. Some countries have come close to bridging the gap with the OECD frontier, others have made considerable progress, and yet others are now showing promise. But there are also countries where productivity is stagnant and where the gaps with OECD-and with other countries in developing Asia-are getting wider. At a sector level, gaps are biggest in services and least in manufacturing. Advances in productivity have been largest in industry and least in agriculture. The reallocation of workers from agriculture to industry and services has indeed provided a "structural bonus," but to date it has been modest.

The countries that have been most successful in closing productivity gaps are those where manufacturing industry displays evidence of increasing technological sophistication. Increasing diversity rather than specialization appears to be associated with growth in productivity at low- and middle-income levels. Those economies that have been most successful in closing the gap with the OECD frontier have (other than Hong Kong, China) progressively specialized within manufacturing. There is also evidence of greater complexity and diversity in the export baskets of those countries that are furthest advanced in catch-up.

Critically, the analysis of this section suggests that developing Asia has enormous potential for catch-up growth. Agriculture still has a large reservoir of workers, and a large untapped "structural bonus" remains in play, which can boost productivity growth. In the next section, these ideas are taken up in the context of the challenges ahead.