Revised Collins/Bosworth Growth Accounting Decompositions

Transcription

1 AERC Explaining n Economic Growth Project Revised Collins/Bosworth Growth Accounting Decompositions March 2003 Benno J. Ndulu* and Stephen A. O Connell** We provide revised growth accounting decompositions based on new data series provided by S. Collins and B. Bosworth, for the iod The enclosed tables replace Tables , pages 26-37, in Ndulu and O Connell (2000), Background Information on Economic Growth. We also briefly discuss new empirical work that addresses some important shortcomings in the empirical literature on human capital and economic growth. *Lead Economist, World Bank Resident Mission, Dar es Salaam and Research Associate, CSAE. **Professor of Economics, Swarthmore College, and Research Associate, CSAE. We are grateful to Susan Collins for access to the data.

2 4.(revised) Collins-Bosworth growth accounting decompositions. The following tables provide revised data corresponding to section 4 of B. Ndulu and S. O Connell Background Information on Economic Growth, April See Collins and Bosworth (1996) for details of the growth accounting methodology, which is also described briefly in O Connell and Ndulu (2000), s Growth Exience: A Focus on Sources of Growth. The O Connell and Ndulu pas can be accessed from O Connell s web page: Recall that the growth decomposition is based on an aggregate production function of the form (1) Y = AK ( L h), where Y is real GDP, A is total factor productivity, K is the physical capital stock, L is the labor force, and h is an index of labor quality (human capital ). In - terms this is (2) y = Ak 0.35 h 0.65 where y Y / L is real GDP and k K / L is the physical capital stock. Taking logs, we obtain an exact decomposition of growth in real GDP (bgn) into the contributions of physical capital (bck), human capital (bch), and the residual (bcr): (3) bgn( t) = ln y( t) ln y( t 1) = ln A( t) ln A( t 1) *[ln k( t) ln k( t 1)] [ln h( t) ln h( t 1)] = bcr + bck + bch. Tables 4.2 and 4.3 show the decomposition by halfdecade, for five regions (Table 4.2) and for individual countries in Sub-Saharan (Table 4.3). For countries included in the latter table, each country teams will receive an individual spreadsheets containing the annual data for their country. 4.1 Notes on the new data The new Collins/Bosworth growth accounting sample consists of 84 countries instead of 88. The table directly below shows the changes in coverage and regional composition relative to the original sample of 88 countries. 1

3 Table 4.0 Changes in composition of the Collins/Bosworth sample. Region Number of countries Old Data New Data Changes in region composition Sub-Saharan Latin America and Caribbean Congo, D. R. and Sudan excluded No changes. S. Asia 4 4 No changes. E. Asia and Pacific Middle East and N. Industrial countries 9 8 Myanmar excluded Malta excluded; Cyprus included (previously in Industrial country group). Cyprus excluded (moved to Middle East and N. group). TOTAL The new data also differ from the old data on a country-by-country basis, for various reasons. One is that the real GDP, real investment, and labor force are from the most recent World Development Indicators data, which means that they are not only extended to 2000 but also possibly revised relative to the previous series. Another is that the new capital stock uses a depreciation rate of 5% rather than the 4% used in Collins and Bosworth (1996); this means that the physical capital stock grows slightly more slowly in the new data, and the contribution of physical capital is therefore slightly smaller than in the old data. A final difference comes from the calculation of the education variable (see next section). 4.2 Revised educational attainment data and the role of human capital in growth regressions You will recall that O Connell and Ndulu (2000) omitted human capital variables from their pooled full specification, because the estimated coefficients were typically insignificant or of the wrong sign. As discussed by O Connell and 2

4 Soludo in their July 2001 memo to case study teams 1, weak formance of human capital variables has been a general problem in the growth literature, particularly when high-frequency data are used (e.g., annual data or halfdecade averages). Cohen and Soto (2001) have since developed a new dataset on educational attainment that they claim is less affected by measurement error than the widely-used Barro-Lee (1993) dataset. Cohen and Soto (2001) and Soto (2002) use these data in panel growth regressions (using 10-year averages) and find that a careful treatment of collinearity and endogeneity allows one to recover statistically significant, correctly-signed, and empirically plausible effects of human capital investment on growth. They estimate a direct return to schooling of 7-10 cent, which is very close to the roughly 10 cent that is characteristic of the microeconomic literature. They conclude from this that externalities to schooling which would raise the social or economy-wide return to schooling above the private, microeconomic return do not appear to be present. However, they show that if one calculates the full or long-run effect of an increase in educational attainment as the effect that would prevail if the ratio of physical capital to GDP were allowed to return to its original value following an increase in human capital a calculation that therefore adds an indirect effect via additional accumulation of physical capital then one gets a long-run impact of education on income that is somewhat larger, on the order of cent for each additional average year of education. As described in Collins and Bosworth (1996), the Collins/Bosworth growth accounting decomposition uses a human capital variable that is constructed by applying a 7% return to years of educational attainment. The 1996 calculations used the Barro-Lee (1993) data on educational attainment. The new calculations use an average of the revised Barro-Lee data (Barro and Lee (2000)) and the new Cohen and Soto (2001) data, where possible. Using the average is appealing, because as long as the two series are unbiased and the correlation between their measurement errors is not too high and positive, the average of the two series should have smaller measurement error than either series taken alone. 1 Available on the website given in section 1 above. 3

5 Table 4.1(revised) Definitions of variables. Variable Definition Units Source bgn growth in real GDP. cent Primarily World Bank, World Development Indicators, with labor force for developing countries measured by economically active population. bck contribution of physical capital = 0.35*(growth of physical capital ). Growth in physical capital generated by applying petual inventory method to national-accounts investment data, using initial 1950 capital stocks, with a depreciation rate of cent Initial capital stocks from Nehru and Dhareshwar (1993); investment mainly from World Bank, World Development Indicators. bch contribution of education = 0.65*(growth in labor quality index). The labor quality index imputes a return of 7 cent to an additional year of average schooling attainment in the adult population. See Collins and Bosworth (1996) for details. cent Schooling attainment calculated as average of Barro- Lee (2000) and Cohen-Soto (2001) data, where both are available, or by available series if only one is available. bcr TFP residual = bgn bck bch. cent Calculated as indicated. 4

6 Region Sub- Saharan * Table (revised) Growth of Real GDP Worker, bgn Marginal Distribution by Region and (balanced sample) Latin America & Caribbean E. Asia & Pacific * Middle East & N. Industrial Countries * S. Asia *The regional composition differs from original table. See table 4.0. Notes: Each cell of the table shows the mean (in bold) for the corresponding region and halfdecade. The standard deviation is directly below the mean, and the number of observations in that cell is directly below the standard deviation. The bottom right-hand cell gives the mean, standard deviation, and number of observations for the full sample. 5

7 Region Table (revised) Contribution of Growth in Physical Capital Worker, bck Marginal Distribution by Region and (balanced sample) Sub- Saharan * Latin America & Caribbean E. Asia & Pacific * Middle East & N. Industrial Countries * S. Asia *The regional composition differs from original table. See table 4.0. Notes: Each cell of the table shows the mean (in bold) for the corresponding region and halfdecade. The standard deviation is directly below the mean, and the number of observations in that cell is directly below the standard deviation. The bottom right-hand cell gives the mean, standard deviation, and number of observations for the full sample. 6

8 Region Table (revised) Contribution of Growth in Education Worker, bch Marginal Distribution by Region and (balanced sample) Sub- Saharan * Latin America & Caribbean E. Asia & Pacific * Middle East & N. Industrial Countries * S. Asia *The regional composition differs from original table. See table 4.0. Notes: Each cell of the table shows the mean (in bold) for the corresponding region and halfdecade. The standard deviation is directly below the mean, and the number of observations in that cell is directly below the standard deviation. The bottom right-hand cell gives the mean, standard deviation, and number of observations for the full sample. 7

9 Region Sub- Saharan * Table (revised) Estimated Residual, bcr Marginal Distribution by Region and (balanced sample) Latin America & Caribbean E. Asia & Pacific * Middle East & N. Industrial Countries * S. Asia *The regional composition differs from original table. See table 4.0. Notes: Each cell of the table shows the mean (in bold) for the corresponding region and halfdecade. The standard deviation is directly below the mean, and the number of observations in that cell is directly below the standard deviation. The bottom right-hand cell gives the mean, standard deviation, and number of observations for the full sample. 8

10 Table 4.3 (revised) Growth accounting decompositions. Country and Growth in Real GDP Physical capital Contribution of: Education Residual bgn bck bch bcr Angola Benin Botswana Burkina Faso Burundi Cameroon C.A.R. Chad Congo, D. R N.A. N.A. N.A. N.A N.A. N.A. N.A. N.A N.A. N.A. N.A. N.A N.A. N.A. N.A. N.A N.A. N.A. N.A. N.A N.A. N.A. N.A. N.A N.A. N.A. N.A. N.A. N.A. N.A. N.A. N.A. Congo, Rep. 9

11 Table 4.3(revised) continued Country and Growth in Real GDP Physical capital Contribution of: Education Residual bgn bck bch bcr Cote d Ivoire Ethiopia Gabon Ghana Guinea 10

12 Table 4.3(revised) continued Country and Growth in Real GDP Physical capital Contribution of: Education Residual bgn bck bch bcr Kenya Madagascar Malawi Mali

13 Table 4.3(revised) continued Country and Growth in Real GDP Physical capital Contribution of: Education Residual bgn bck bch bcr Mauritius Mozambique Namibia Niger Nigeria

14 Table 4.3(revised) continued Country and Growth in Real GDP Physical capital Contribution of: Education Residual bgn bck bch bcr Rwanda Senegal Sierra Leone South

15 Table 4.3(revised) continued Country and Growth in Real GDP Physical capital Contribution of: Education Residual bgn bck bch bcr Sudan N.A. N.A. N.A. N.A N.A. N.A. N.A. N.A N.A. N.A. N.A. N.A N.A. N.A. N.A. N.A N.A. N.A. N.A. N.A N.A. N.A. N.A. N.A N.A. N.A. N.A. N.A N.A. N.A. N.A. N.A. N.A. N.A. N.A. N.A. Tanzania Togo Uganda Zambia

16 Table 4.3(revised) continued Country and Growth in Real GDP Physical capital Contribution of: Education Residual bgn bck bch bcr Zimbabwe Source: Data provided by Susan Collins. 5

17 References Barro, R. and J.-W. Lee (1993), International comparisons of educational attainment, Journal of Monetary Economics 32(3), Barro, R. and J.-W. Lee (2000), International data on educational attainment: updates and implications, CID Working Pa No. 42, Harvard University. Cohen, D. and M. Soto (2001), Growth and human capital: good data, good results, Technical Pa No. 179, OECD Development Centre, Paris. Collins, Susan M. and Barry P. Bosworth (1996), Economic Growth in East Asia: Accumulation versus Assimilation, Brookings Pas on Economic Activity 2: Nehru, V. and A. Dhareshwar (1993), A New Database on Physical Capital Stock: Sources, Methodology and Results, Revista de Analisisis Economico 8(1): Ndulu, B. and S. O Connell (2000), Background information on economic growth, Swarthmore College. O Connell, S. and B. Ndulu (2000), s growth exience: a focus on sources of growth, Swarthmore College. Soto, M. (2002), Rediscovering education in growth regressions, Technical Pa No. 202, OECD Development Centre. 6