Science and Technology Takeoff in Historical Perspective

Transcription

1 Science and Technology Takeoff in Historical Perspective Gao Jian Tsinghua University Gary H. Jefferson Brandeis University January 3, 2005 Draft: for review and comment only

2 1. Introduction Economists agree that the sustained growth of living standards requires on-going technological progress. There is also recognition that an important driver of technological change is research and development spending. Cross country statistics, however, show that R&D intensity, i.e., R&D expenditure as a share of GDP, varies widely across counties. There exists a clear tendency for R&D intensity to be higher in high income countries than in low income countries. In this paper we explore the systematic differences in R&D intensities that are exhibited in cross-country data. In particular, we focus on the tendency for economies, as their income levels rise above a particular threshold to systematically and often abruptly increase their levels of R&D spending as a share of total GDP. We pay particular attention to the phenomenon in which rising R&D intensity occurs over a relatively short period of time and then levels off. We characterize this phenomenon of an abrupt onetime increase in R&D intensity as science and technology (S&T) takeoff. This paper investigates both the statistical regularities of S&T takeoff and the underlying economic conditions that potentially explain this phenomenon of takeoff in which R&D spending rising more rapidly than GDP. 2. Cross-country data The Human Development Report, 2001 (UNDP) reports on ratios of R&D expenditure to GDP for 70 countries using data that falls in the time frame of The scatter plot in Figure 1 shows the ratio of R&D spending /GDP in relation to the log of GDP per capita. One pattern that emerges from Figure 1 is the tendency for richer countries to exhibit ratios of R&D to GDP that are higher than those of lower-income countries. Among the 21 OECD economies for which data are reported, the average R&D/GDP ratio stood at 2.0 percent. Among the other 49 countries, the ratio stood at 0.7 percent. Among the non-oecd economies, just two countries Israel and South Korea reported R&D intensities in the 1990s equal to or greater than two percent. In Table 1 we report estimates of the statistical relationship between the ratio of R&D/GDP and the log of income per capita. Column (1) in Table 1 show that estimates, which use the full sample of 70 enterprises, show a positive and robust relationship between R&D intensity and the log of income per capita. In order to test for non-linearity in the relationship between R&D intensity and per capita income, we implement a piecewise regression in which we define low-income countries with levels of per capita income less than $10,000, high-income countries with incomes greater than $25,500, and middle-income countries with incomes lying between $10,000 and $25,500. The key result from columns (2), (3), and (4) in Table 1 is the tendency for the subsample of middle income countries to exhibit a significantly larger

3 and more robust impact of income per capita on R&D intensity than income per capita exerts on R&D intensity for lower and higher income ranges. It is this middle income range associated with the steep relationship between GDP per capita and R&D intensity that represents the takeoff stage of the S&T transition. These results indicate that the pattern of R&D intensification follows an S shaped logistic curve. We observe that among the 13 countries with R&D intensities equal to two percent or greater, six have large populations, i.e. more than 45 million in Another way of framing this observation is that all but one of the large countries (i.e. Italy) report high R&D intensities equal to or greater than two percent. Eight of the nine OECD countries with R&D intensities less than two percent are all smaller countries, i.e. with populations in 1999 with less than 40 million. To test the significance of population, we include a dummy for countries with populations 45 million. Column (5) of Table 1 shows that the estimate of the dummy is positive and significant at the 15 percent level. If we omit Sweden, an extreme outlier, whose R&D intensity is 3.8 percent, the estimate of the size dummy becomes highly statistically significant. We discuss the potential role of size in the S&T takeoff in Section 4. In the next section, we examine the experience of specific countries that have made the transition from low to high R&D intensity. 3. The historical evidence of S&T takeoff Table 2 documents patterns of rising R&D intensity in the six of the seven largest OECD countries (Italy is omitted). For five of these countries the U.S., Germany, France, Japan, and S. Korea - we are able to document patterns of a rapid ascent in R&D intensity in which R&D spending rose substantially faster than GDP. When data became available for the U.K. in 1960, its level of R&D intensity had already exceeded 2 percent. For each of the other five large OECD countries, the ascent occurred within a relatively short period. For the U.S., the first country for which we can document the rise in R&D intensity from one to two percent, the transition required 10 years. For South Korea, the most recent large OECD country to make the transition, the ascent occurred within a span of just five years. Japan s transition required the longest period; it spanned 19 years. Other OECD countries that had achieved or exceeded the two percent R&D levels prior to 2000 include the Netherlands, Sweden, and Switzerland. The other smaller OECD economies all registered R&D intensities in the range of one to two percent. Non- OECD countries/regions that exceeded the two percent level include Israel and Taiwan. Counties whose R&D intensities by 2000 fell in the range of one to two percent include Belarus, the Czech Republic, Singapore, and Slovenia. In 2000, China s R&D intensity rose to one percent; by 2003, it had risen to 1.3 percent. 4. The causes of takeoff What factors cause R&D spending to outpace GDP growth? Why should countries exhibit relatively abrupt increases in R&D intensity whereby their R&D/GDP ratios hit

4 one percent and then proceed in relatively short periods of time to double? We identify several factors that are associated with these rising R&D intensities. These factors can be broadly classified as structural changes on the demand side and behavioral changes within the firm sector on the supply side. While these changes on the demand and supply sides are not separable, i.e. they interact through mutually reinforcing channels, we can document and analyze these changes individually. Factors on the demand side consist of shifting patterns of consumption and production that are associated with rising incomes. On the supply side, we observe patterns of firm behavior in which R&D intensifies through two channels. First, over time firms which had not previously conducted R&D adopt R&D operations and, second, firms that had been R&D performers significantly intensify their R&D operations. Also on the supply side we identify the accumulation of complementarities to R&D that tend to be associated with rising wealth and living standards. While rising demand for more technologically sophisticated goods can be satisfied by imports, we note that market size, which allows for achieving scale economies, can motivate investment in the creation of production platforms that respond to demand. For this reason, it may be that country size is important to achieving symmetry between the demand and supply dimensions of takeoff. Finally, we investigate the possibility that S&T takeoffs may exhibit coordination across countries due to spillover effects. During , China s R&D ratio rose from 0.6 percent to one percent. We document these demand and supply patterns of using Chinese data from this period. Shifting consumption and production patterns. As living standards rise, the composition of goods and services shifts from those that are low in technological content to goods and services that are more technologically intensive. Automobiles substitute for bicycles; consumer electronics become ubiquitous, medical services and the equipment that supports them become more sophisticated. This pattern of technology intensification that accompanies rising living standards mirrors Engel s Law. As incomes rise, not only is the income elasticity of demand for non-agricultural goods greater than one, but also goods with low-technology content (e.g. bicycles, handicrafts, rudimentary medical care) are inferior goods. The implication is that the income elasticity of demand for hightechnology goods is high. Table 3 illustrates the contribution of three of these industries to the intensification of R&D in China s economy. The Table shows, for example, that among China s large and medium size industrial (LME) enterprises in the electronic and telecommunication sectors, during the latter half of the 1990s, the ratio of R&D spending to value added accelerated by 250 percent to reach nearly 7.5 percent. The overall impact on Chinese industry of the growth of R&D intensity in this sector was magnified by the fact that its market share in total industry sales doubled over this period. Table 3 shows that this association between rising R&D intensity and growing market share is also exhibited by China s equipment and machinery sector and the instrumentation sector.

5 R&D intensification at the firm level. Table 4 shows a synchronous rise in the intensities of firm-level R&D operations resulting from both the adoption of new R&D operations and learning-by-doing in established R&D centers. Table 4 shows during 1995 to 2001 increasing rates of adoption of R&D activity by China s industrial LMEs. During this period, the proportion of LMEs performing R&D rose from 20.2 to 29.1 percent, a rise of nearly one-half. Over the same period, the number of high performing LMEs with R&D to value added ratios in excess of 10 percent grew from 620 to 1,553. Not surprisingly, the intensification of patterns of consumption on the demand size are reflected on the production and supply side. Increasing returns and complementarities to R&D. Arguably, by itself spending on R&D creates low and stagnant returns. Through the interaction of R&D spending with R&D complements, returns to R&D rise thus motivating the reallocation of further rounds of resources to R&D activity. Such complements include public investments, such as spending on basic education, higher education, research institutes and basic research, and science parks. Complements also arise from private investment in training, R&D networks, and imported technology. All of these serve to increase the variety and supply of complementarities to R&D that serve to raise the efficiency and economic returns to firm-level R&D. In their study of the R&D capabilities and returns to R&D in five Chinese cities and Seoul, Korea, Jefferson and Zhong (2004) demonstrate the close association of a range of R&D complements, i.e., both public and private goods, and the returns to R&D. Market size. The link between the rising demand for technologically sophisticated good and local R&D intensification is not inevitable. Countries may import technologically sophisticated goods. In larger countries, however, where opportunities exits for establishing production platforms that can achieve scale economies, we might expect a stronger commercial motive to set up production in close proximity to evolving demand. Markets size may explain why S&T takeoff has occurred in virtually all of the largest OECD economies but not in certain smaller OECD countries (e.g. Norway, Australia, Belgium, Austrialia, and New Zealand) while it has occurred in all of the largest OECD economies. 1 China is one example of an economy in which size is critical. While other developing countries may import technologically sophisticated electronic and telecommunications equipment and other high-tech goods, including electrical machinery, instrumentation, and automobiles, multinationals are clamoring to set up production for these R&D intensive industries in close proximity to China s burgeoning consumer markets. International spillovers. Figure 2 shows several patterns of coordinated rise in R&D intensities. The most sustained of these is that of Japan, Germany, and France where the takeoffs begin in the period During the late 1960s through the 1 The desire to serve large and fast growing consumer markets creates a premium for the establishment of production centers that can benefit from learning-by-doing and learning-by-using in close proximity to burgeoning demand.

6 mid-1970s, these countries and the U.S. for which the transition was completed earlier, experienced a coordinated dip in their R&D intensities, which later in the late 1970s resumed their upward momentum over the following decade. 5. Conclusions

7 Figure 1 Variable (1) All observations Table 1 RD/GDP versus log(gdp/capita) (2) Low income < $10,000 (3) Middle income $10,000 < x < $25,500 (4) High income > $25,500 (5) High incom >$25,500 w large countr dummy (1.33) (1.62) Constant (6.33) (-2.10) (3.49) (1.40) ln(gdp/capita) (9.03) (3.40) (3.70) (1.70) Population > n.a. n.a. n.a. n.a million (1.63) Adj. R-square (observations) (70) (46) (13) (9) (9)

8 country Dates: years achieved 1%/ and 2% U.S. 1951/1960 U.K. W. Germany Table 2. Historical documentation of S&T takeoff or 1947/1956 Source (title and page number) Gray and Peterson: Economic Development of the United States, Richard D.lrwin.Inc.1974, p R&D Christoph- Friedrich von Braun: The Innovation War, Carl Hanser Verlag, 1998, p. 29; R&D France R&D Japan Italy Has not achieved S&T takeoff Science and Engineering Indicators 2000, NSF, p S. Korea OECD %-2.1% Estimated weighted average of 12 EC countries. Source: Chris Freeman and Luc Soete, The Economics of Industrial Innovation (3rd Edition), MIT Press, 1997, p. 300.

9 3 2 1 US Japan Sout h Kor ea UK Fr ance Ger many Chi na Br azi l Figure 2. Historic R&D/GDP (GNP) in 8 Countries

10 Table 3. The role of changing industry composition (LME database) Industry R&D/VA 1995 (%) R&D/VA 2000 (%) R&D/VA 2000/1995 Sales share 2000/1995 Elec. and telecom. equipment Elec. equip. and machinery Instruments and meters Total industry Source: Jefferson, Bai, Guan, and Yu (2003). Note: VA = value added. Table 4. Distribution of R&D/VA (r), all enterprises (%) Year 0 0<r 1 1<r 2 2<r 4 4<r 6 6<r 10 r> % increase (45.6) Source: NBS-LME data set. Note: The figures in parentheses represent the share of total LMEs that fall within each percentile range of R&D/intensity.

11 References Chris Freeman and Luc Soete, The Economics of Industrial Innovation (3th Edition), MIT Press, Christoph- Friedrich von Braun: The Innovation War, Carl Hanser Verlag, 1998 Gray and Peterson: Economic Development of the United States, Richard D.lrwin.Inc.1974 Rostow Science and Engineering Indicators 2000, NSF

12 Table 1a. Statistical Significance of Regressing RD/GDP on ln(gdp/capita) income t-stat t-prob < < < < < < < < < < < This table shows that for countries with income per capita < $4,000 there is no statistically significant relationship between R&D intensity and income per capita.