Worldwide convergence of productivity levels: recent empirical evidence.
DO LOW-PRODUCTIVITY countries have an advantage of backwardness that affords them rapid productivity growth? According to the convergence theory, low-productivity countries with adequate technological skills and market development can borrow cost-reducing technology from higher-productivity countries to achieve rapid productivity growth. Economic historians have noted that the initial gap in productivity with the industrial leader, in addition to rapid contemporaneous growth in technological knowledge, represents an opportunity for swift productivity gains. These are the necessary conditions for rapid growth.
This paper examines recent trends in output per worker in 109 countries to determine whether productivity converged in the 1960 to 1985 period. Literacy rates are used to group countries with similar potential for convergence. The approach used here builds on the methodology employed in Baumol (1986) and Kravis and Lipsey (1984), among others. Regressions of initial productivity levels on rates of change will illustrate whether the grouping of countries has an impact on the strength of the recorded convergence process and whether convergence was strongest among the industrialized countries that were initially well prepared to reap the benefits of technological progress that occurred in the United States after the war years.
STUDIES OF CONVERGENCE
Denison (1967), Abramovitz (1979, 1986), and Baumol (1986), among others, have noted that, within certain groups of countries at comparable stages of development, nations with low levels of relative labor productivity tend to experience the fastest rates of subsequent productivity growth, i.e., they tend to catch up with the lead productivity country over time. Abramovitz (1979) generalized by stating that within a group of ten industrialized countries "the less productive the country in 1950, the more rapidly its productivity rose."
The studies in the literature share a common foundation. The initial lag in productivity is described as a technological opportunity for the lower-productivity countries, assuming the existence of sufficient absorptive capacity and resources. A country can only reap the benefits of this backlog of technological innovations once it has overcome the major social and political barriers to industrialization, according to Gerschenkron (1962). He points out that even then the development of a backward country is likely to follow a path very different from that of a more advanced industrialized country.
A country's degree of backwardness is defined in Abramovitz's model by its initial productivity gap with the industrial leader. When old capital is replaced by new capital in the lead country, the technological frontier advances by the amount of new knowledge acquired since the old equipment was installed. The technological backward countries can increase their technological frontiers much further, because their plants and equipment are likely to be technologically older than their chronological age. This presents the opportunity to advance more -- in absolute terms -- than the lead country when replacing obsolete capital. The more technologically backward a country, the faster its potential rate of growth in labor productivity, which increases both with the capital/labor ratio and the level of technology embodied in a country's capital stock.
Determinants of Productivity Growth Rates
The pace at which the gap is closed depends on several factors, including market openness, labor mobility, and the investment environment that determines the rate of capital accumulation and technological change. Kendrick (1979, 1981) and Abramovitz (1986) spell out several factors, common to many productivity studies, that determine the actual speed of productivity growth. These include embodied and disembodied technological change that leads to productivity enhancing innovations, the exploitation of scale economies, the reallocation of labor and capital to their most productive employment (structural change), and capital accumulation, which increases the capital-labor ratio.
A country must have access to large markets if the leader's path of technological progress is scale dependent. A laggard country cannot reap the benefits of large-scale manufacturing innovations in the U.S., for example, if its markets are small and its trade is restricted. The reduction of market distortions, trade barriers, and transport costs in the early 1900s expanded the industrialized countries' markets and provided the change for them to improve resource allocation and productivity growth. Growing markets gave them the opportunity to take advantage of scale economies by modifying their production processes. Industries operating under increasing returns benefits by converting to scale-dependent production methods as national product increases and markets expand.
Government actions also have an impact on potential productivity growth rates. They affect economic decisionmaking through domestic policies, trade regulations, and fiscal and monetary policies. Government activities affect not only incentives to innovate and invest and the general efficiency of resource allocations, but also the ease of the adjustment process that comes hand-in-hand with the implementation of new technologies and processes. Furthermore, the issue of market structure is important, because many centrally planned economies have highly educated workforces, but they do not share the economic growth rates of free-market economies with comparable levels of education. Policy choices clearly have an important impact on a country's ability to achieve rapid productivity growth.
In addition to the discussion of the determinants of productivity growth rates, a portion of the literature on productivity analysis seeks to identify the capability for reaping the benefits of technological backwardness. The term "social capabilities" has been used to describe this concept of readiness; the term is impossible to define and measure precisely, but it is related to a country's market structure and stage of economic development. For the purpose of testing the convergence hypothesis, the best measure of social capabilities would be an indicator of the readiness of a follower country to incorporate new technologies and methods into its established manufacturing plants and practices in order to take advantage of its "opportunity of technological backwardness." For example, the number of trained engineers and managers in a country, or a measure of the educational attainment of the workforce, such as the number of high school or college graduates, could be used. However, these two variables omit the important reference to a country's market structure discussed earlier. Clearly, no one indicator has been identified yet that satisfactorily represents these concepts of readiness and ability.
Nonetheless, the technological borrowing that leads to convergence is most likely to occur among countries with relatively comparable levels of social capabilities. Abramovitz (1979) points out that a group of countries' technological and social "abilities" do not have to be identical, but that a country's relative level of social capabilities seems to be the only factor that inhibits the potential for productivity-enhancing technology transfer. A country's initial level of backwardness did not happen by chance. Gerschenkron (1962) also emphasizes that countries face not only great opportunities by virtue of their backwardness, but that they also face greater obstacles to industrialization the greater their relative backwardness and the longer it persists. The lack of progress itself hampers the successful exploitation of the opportunities that the technological backlog offers. Therefore, tests for catchup should be among groups of countries with relatively comparable levels of social capabilities.
The problem of implementing this concept has been the subject of much discussion in the literature. The assumption that the free-market industrialized countries have roughly the same "social capabilities" has led most researchers to study them as a group, starting with Kuznets in 1956. Abramovitz (1986) justifies this choice by noting informally that they have roughly similar education levels, market structure and organization, openness to competition and free trade, experience with financial institutions, and management of large corporations.
Limits to Catch Up
The convergence theory implies that the catchup process is self-limiting. Because the rapid growth potential of lagging countries depends in part on the extent of their gap in technological knowledge with the leader, productivity growth slows as the gap narrows. It becomes more difficult for the follower to advance technologically, because original research expends more resources than adapting borrowed technology. In addition, Baumol and McLennan (1985) point out that, as the gap narrows, it is harder for the follower to identify technology worth borrowing. Furthermore, Abramovitz (1986) argues that during periods of slow economic growth it is more difficult for an industry to advance technologically. The technological gap has narrowed considerably in the postwar period, leading Abramovitz (1986) to conclude that the catchup process will contribute less to industrialized countries' growth in the future, barring rapid growth in the lead country.
In fact, the leader now has the opportunity to borrow from the old laggards in certain fields, and may even switch positions on the aggregate productivity ladder -- although the catchup hypothesis does not anticipate changes in position. Over time (and under different exchange rate conversions) the lead country may change -- a dynamic the theory does not address. The United States surpassed Britain as the productivity leader at the turn of the century, and Abramovitz (1986) asserts that Japan is likely to overtake the United States as leader. Furthermore, Baumol and McLennan (1985) show that by 1981 Japanese productivity levels in manufacturing had almost completely converged to those in the United States. In addition, they point out that in some industries, such as steel and automobiles, the Japanese have already surpassed productivity levels in the United States.
Baumol (1986) likens the innovation-sharing process that narrows gaps in productivity levels to the sharing of an international public good. The public good in this case in any process, innovation or policy that increases productivity in the production of a certain commodity. Only countries that produce or trade in that product can take advantage of the innovation. Baumol describes the industrialized countries that produce and consume similar goods as in a constant race to obtain and implement new cost-reducing technologies. The less-developed countries that do not manufacture these common products are left out of the catchup process.
Results of Previous Studies
Several studies support the convergence hypothesis within certain groups of countries. Kravis and Lipsey (1984) provided some evidence of convergence in the industrial countries and the low-income countries from 1960 to 1979. They found little support for the catchup hypothesis in the world as a whole. Abramovitz (1986) found that labor-productivity measures moved closer together as a group from 1970 to 1979 in the group of industrialized countries. He concluded that large productivity gaps provide the potential for rapid growth leading to convergence.
Baumol (1986) presented a model consistent with that of Abramovitz. He concluded that there has not been a long-run downward trend in U.S. productivity growth. Instead, there has been a short-run falloff in growth following the rapid postwar expansion. In his view, the post-depression era presented a backlog of ideals waiting to be implemented. This was a catchup period within the U.S. economy that spread to the other industrialized countries through the diffusion of technology.
Baumol (1986) used data compiled by Summers and Heston (1984) to examine seventy two countries that he split into productivity "clubs:" (1) the free-market industrialized countries; (2) the centrally planned economies; and (3) the middle-income free-market economies. While each group seemed to have followed the catchup pattern somewhat, his results indicate little evidence of convergence amont the groups. A fourth group of the poorest, least-developed countries did not share this rapid growth from 1950 to 1980, according to Baumol. He reasoned that their poor productivity performance may have been due in part to their limited product mix and education levels. Because the poorest low-income LDCs do not produce and trade high-technology goods, they cannot take advantage of production innovations in those sectors.
Baumol and Wolff (1988) reinforced earlier findings in response to important criticism by DeLong (1988). DeLong pointed out that studies of convergence are flawed if they test for catchup in groups of countries that have already exhibited rapid productivity gains. As Baumol and Wolff acknowledged, the group must be selected on the basis of each country's initial potential to achieve rapid productivity gains.
DATA AND METHODOLOGY
The analysis in this paper relies on purchasing-power-parity (PPP) adjusted output per capita data compiled by Summers and Heston (1988) and labor force data from the International Labour Office (ILO, 1985). From these data, productivity measures were developed for 109 market countries in 1960 and 1985, adjusted for PPP (see Zagardo  for detailed information on data and methodology.) To minimize the prolonged impact of post-WWII reconstruction, 1960 was chosen as the base year in this study. The end-year is extended from previous studies to 1985 because that was the last year for which consistent, comparable data were available for which consistent, comparable data were available for a wide range of countries at the time this study was undertaken. Furthermore, these years are the approximate peaks in the business cycle. (1)
The Summers and Heston (1988) estimates of GDP per capita in 1980 dollars are free from the biases of exchange rate conversions and therefore give a more accurate measure of relative country output per worker. The PPP adjustment to the real GDP per capita data reduces the apparent disparity in income between lower- and high-income countries caused by exchange rate conversions.
To minimize the impact of measurement discrepancies on worldwide comparisons of productivity, estimates of "economically active persons" were assembled from the ILO (1985) to represent labor inputs. The ILO has developed "a consistent set of data based on uniform concepts, methods and classification schemes for all countries." Nevertheless, certain weaknesses remain (see Zagardo  for a discussion of the limitations).
These data were combined to develop the real GDP per worker and the growth rate figures for 1960 and 1985. Evidence in Zagardo (1990) confirms the Kravis and Lipsey (1984) argument that measured growth rates in per capita real GDP are systematically lower than corresponding per worker measures for the poorest countries; per worker measures are a better indicator of productivity changes.
Literacy Rates -- A Working Indicator
of Social Capabilities
It is not a simple matter to define the construct a measure that reflects the appropriate level of social capabilities needed to participate in convergence. Base-period illiteracy rates, reported by the United Nations (1967, 1977) are used here to represent capabilities, or labor quality. (2) This approach avoids the causality problems associated with using end-period rather than base-period measures to indicate social capabilities. (3) Base-period literacy rates may also be indicators of other measures of market structure, educational capability, prosperity, and policy choices. In addition, they may be more appropriate for less developed countries than for industrialized countries.
These data were sorted by decreasing illiteracy rates for the available year closest to 1960. (4) They range from 99 percent iiliterate for Niger to 2 percent illiterate for the United States and Japan. Rates were not reported for fifteen industrialized countries, but, for the purposes of this analysis, it was assumed that these countries had illiteracy rates of less than 10 percent in 1960. These literacy rates provide the basis for selecting convergence test groups, because the countries in each group are assumed to be in comparable positions on the scale of readiness for further economic development.
Table 1 shows that the four convergence test groups (and two subgroups) contain countries in four ranges of illiteracy rates: group 1, 76-100 percent illiterate; group 2, 50-75 percent illiterate; group 3, 26-49 percent illiterate; and group 4, 0-25 percent illiterate. Group 4 is further split between those countries with illiteracy rates above and below 10 percent, i.e., group 4(a), 10-25 percent illiterate; and group 4(b), 0-9 percent illiterate (including the fifteen industrialized countries without repored illiteracy rates).
The estimated model in Table 1 consists of a regression of the average annual growth rate in labor productivity from 1960 to 1985 (PTYGROW), on a country's level of PPP-adjusted GDP per worker in 1960 (PTY1960). In one-sided tests of coefficient significance, the results confirm the discussion and results in Baumol and Abramovitz that convergence would be demonstrated in the group of industrialized countries $(4(b)$). These countries share a high
<TABLE DATA OMITTED>
degree of "social readiness," measured here in terms of high literacy rates: more than 90 percent. Convergence is also observed for those countries with above-average literacy rates (groups 3 and 4). Furthermore, as expected, catchup is not exhibited for the world as a whole. Two-sided tests of coefficient significance reject the existence of divergence in the case of positive coefficient estimates. Note also that the explanatory power of the model is much lower for the lowest-literacy groups (groups 1 and 2). Clearly there are other factors not included in this model, such as the rates of savings and investment, technology growth and growth in the capital:labor ratio, and policy choices that are important determinants of productivity growth rates. (5)
For the ninety-four countries with reported literacy rates, a literacy variable was added to the original Baumol regression. The estimated model becomes a regression of PTYGROW on PTY1960 and LIT1960 ( a country's illiteracy rate around 1960). An inverse relationship would indicate that, other things being equal, higher literacy rates contribute to faster productivity growth, given the initial level of output per worker.
The results in Table 2 support the hypothesis that, within certain country groups, literacy rates are highly significant indicators of a country's potential to share in rapid productivity growth. In particular, literacy rates are significant for the whole sample, plus groupw 1, 3 and 4(b). Only in the whole sample and in the most literate group of countries [4(b)], however, is the coefficient for base-period productivity also negative and significant, supporting convergence. Note also that the model is generally strongest for the most literate groups of countries.
Chow test are employed to discern differences in the impact of the independent variables in the subgroups and in the worldwide sample. The Chow test results reported in Table 1 provide support for the hypothesis that the impact of base-period productivity levels will be different in different groups of countries, because the coefficient estimate for the worldwide sample is insignificant and the Chow test rejects coefficient equality. This finding supports the significant convergence results reported for groups 3, 4, and 4(b).
Alternatively, in the second model of the ninety-four countries with reported literacy rates, Table 2 shows that higher literacy rates have a uniform effect on productivity growth rates across groups. Chow test results cannot reject coefficient equality across the study groups in this model, suggesting that the strong coefficient of literacy in the worldwide sample is also representative of the impact in the subgroups.
This article tests for convergence of aggregate output per worker for 109 countries from 1960 to 1985. It finds convergence among the world's most industrialized countries, in addition to a group with
<TABLE DATA OMITTED>
slightly above-average literacy rates. These test groups were created on the basis of literacy rates circa 1960, rather than base-period productivity levels. The tests also confirm that literacy rates are a significant independent variable in two groups and for the entire sample. In fact, convergence is documented for the whole sample only when literacy rates are included as an independent variable. Literacy rates, therefore, act as proxies for other ommitted variables and lend some additional explanatory power to convergence model. Furthermore, Chow test results indicate that literacy rates have a more uniform impact across country groups than base-period productivity levels.
This paper documents the occurrence of catchup in certain country groups, leaving an explanation of the casual factors to future work. The important factors are sure to include a policy variable, the role of capital investment (Wolff, 1991), and other variables not considered here.
One obvious area requiring more study is the idea of social capabilities. (6) Once work on this concept has started, new topics could be examined. For example, analysis of potential convergence in the group of East European countries could reveal future growth paths in these countries. The problem of measuring social capabilities is central to the interpretation of results. The East European nations have a structure different from that of other countries with similar literacy rates or income levels. There is a need to identify what other variables, in addition to literacy rates, signal that a country is prepared, or "socially capable," of participating in convergence. This measure could indicate whether these countries are at a stage of development and political structure at which they can reap the benefits of market liberalization. It could also help to identify the country to which the East European countries are likely to converge. Will Germany, or some less advanced country be the technological leader, providing the most appropriate technological contributions? Under certain conditions, Dowrick and Nguyen (1989) argue that the catchup parameter is stable over several different time periods. Using their model, the future growth path of a group of countries could be estimated once the lead country is identified.
(1) More recent evidence indicates that, for several less-developed countries, 1988 may be a more appropriate end year -- 1985 was a low point in the business cycle in some LDCs. The bias here may reduce the observed strength of the convergence relationship noted for these countries. When examining such a broad range of countries, it is unlikely that a single date will satisfy the requirement that every country is at the peak of its business cycle at the time of measurement.
(2) Illiteracy rates are defined as the percent of the population over ten years of age who would not, with understanding, read and write a sentence about everyday life. Rank based on illiteracy rate was determined by decreasing level of illiteracy. In the case of two or more countries with identical illiteracy rates, the rank was based on increasing levels of 1960 RGDP per worker, computed as described in the text. Of the 13 countries that record negative growth in real GDP per worker, only two, Guyana and Chile, are among those in the top half in terms of literacy. More than half, seven, are concentrated among the twenty-five least literate countries. This observation provides preliminary support for the use of base-period literacy rates as an indicator of the ability to participate in growth.
(3) Abramovitz (1979, 1986) and Olson (1982) note that social capabilities grow over time, leading to an increase in the amount and level of technological knowledge that low-productivity countries can borrow.
(4) Zimbabwe was dropped from the sample due to lack of consistent illiteracy data. Seven oil-exporting countries (Bahrain, Kuwait, Saudi Arabia, Iraq, Iran, Venezuela, and the United Arab Emirates) were omitted from the study because their measures of growth in output per worker undoubtedly do not represent true productivity gains. This leaves a large sample of 109 countries with which to test the convergence theory. A detailed list of the 109 countries is available from the author on request at Virginia Transportation Research Council, Box 3817, University Station, Charlottesville, VA, 22903.
(5) Also available from the author is a table that shows the extent to which each country has closed its relative productivity gap with the United States from 1960 to 1985.
(6) In a recent book, Baumol, Blackman and Wolff (1989) find levels of secondary education to be significant indicators of social capabilities.
Abramovitz, Moses. "Rapid Growth Potential and its Realization: The Experience of Capitalist Countries in the Postwar Period." In Economic Growth and Resources, pp. 1-30. Edited by E. Malinvaud. London: Macmillan, 1979.
Abramovitz, Moses. "Catching Up, Forging Ahead, and Falling Behind." The Journal of Economic History vol. XLVI, no. 2 (June 1986):385-406.
Baumol, William J. "Productivity Growth, Convergence and Welfare: What the Long-Run Data Show." The American Economic Review 76 (December 1986):1072-85.
Baumol, William and McLennan, Kenneth eds. "U.S. Productivity Performance and Its Implications." In Productivity Growth and U.S. Competitiveness, pp. 3-28. Oxford: Oxford University Press, 1985.
Baumol, William and Wolff, Edward N. "Productivity Growth, Convergence and Welfare: Reply." The American Economic Review 78 (December 1988):1155-59.
Baumol, William J.; Blackman, Sue Anne Batey; and Wolff, Edward N. Productivity and American Leadership: the Long View. Cambridge: MIT Press, 1989.
DeLong, J. Bradford. "Productivity Growth, Convergence, and Welfare: Comment." The American Economic Review 78 (December 1988):1138-1154.
Denison, Edward F. Why Growth Rates Differ: Postwar Experience in Nine Western Countries. Washington, D.C.: Brookings, 1967.
Dowrick, Steve, and Nguyen, Duc-Tho. "OECD Comparative Economic Growth 1950-85: Cath-Up and Convergence." The American Economic Review 79 (December 1989):1010-1030.
Gerschenkron, Alexander. Economic Backwardness in Historical Perspective. Cambridge: Harvard University Press, 1962.
International Labour Organization. Economically Active Population: 1950-2025. 3rd ed. 5 vols. Geneva, 1985.
Kendrick, John. "Productivity Trends and the Recent Slowdown: Historical Perspective, Causal Factors, and Policy Options." In Essays in Contemporary Economic Problems, pp. 17-113. Edited by William Fellner. Washington, D.C.: AEI, 1979.
Kendrick, John. "International Comparisons of Recent Productivity Trends." In Essays in Contemporary Economic Problems: Demand, Productivity, and Population, pp. 125-170. Edited by William Fellner. Washington, D.C.: AEI, 1981.
Kravis, Irving B. and Robert E. Lipsey. "The Diffusion of Economic Growth in the World Economy, 1950-80." In International Comparisons of Productivity and Causes of the Slowdown, pp. 109-151. Edited by John Kendrick. Cambridge: Ballinger with American Enterprise Institute, 1984.
Kuznets, Simon. "Quantitative Aspects of the Economic Growth of Nations: I Levels and Variability of Growth." Economic Development and Cultural Change 5 (October 1956).
Olson, Mancur. The Rise and Decline of Nations: Economic Growth, Stagflation and Social Rigidities. New Haven: 1982.
Summers, Robert and Heston, Alan. "A New Set of International Comparisons of Real Product and Price Levels: Estimates for 130 Countries, 1950-1985." Review of Income and Wealth 34 (March 1988):1-25.
United Nations. Statistical Yearbook. New York, 1967.
United Nations. Compendium of Social Statistics. New York, 1977.
Wolff, Edward N. "Capital Formation and Productivity Convergence Over the Long Term." The American Economic Review 83 (June 1991):565-579.
Zagardo, Janice Turtora. A New Look at International Comparisons of Productivity from 1960 to 1985. Ph.D. Dissertation, Graduate School of Arts and Sciences, George Washington University, September, 1990.
|Printer friendly Cite/link Email Feedback|
|Author:||Zagardo, Janice Turtora|
|Date:||Oct 1, 1991|
|Previous Article:||Economic growth in the 1990s.|
|Next Article:||Productivity trends in selected industries.|