1. While equity issues have been at the forefront of controversy surrounding the Green Revolution, many other criticisms have been levelled, for example, with regard to increasing financial burdens on the part of farmers as a result of purchasing the Green Revolution inputs, land degradation, increased susceptibility of harvests to pests, reduction of biodiversity or displacement of labour (Alston et al., 2006 Alston, J. M., Dehmer, S. and Pardey, P. G. 2006. “International initiatives in agricultural R & D: the changing fortunes of the CGIAR”. In Agricultural R & D in the Developing World: Too Little, Too Late?, Edited by: Pardey, Philip G., Alston, Julian M. and Piggott, Roley R. 313–360. Washington, DC: International Food Policy Research Institute.  [Google Scholar]).

2. In 2000, roughly half of the population in the developing world derived their livelihood from the agricultural sector (FAO, 2008 FAO. 2008. “FAOSTAT database. Food and Agriculture Organization”.  [Google Scholar]).

3. Recently, a large number of articles have been written as a result of improved data provided by the Food and Agriculture Organization (FAO). This more recent literature has commonly used a Malmquist index or Data Envelopment Analysis (DEA) approach (for example, Nin et al., 2003 Nin, A., Arndt, C. and Preckel, P. V. 2003. Is agricultural productivity in developing countries really shrinking? New evidence using a modified nonparametric approach. Journal of Development Economics, 71(2): 395–415. [Crossref] , [Google Scholar]; Coelli and Rao, 2003 Coelli, T. J. and Rao, D. P. 2003. “Total factor productivity growth in agriculture: a malmquist index analysis of 93 countries, 1980–2000. CEPA Working Papers Series WP022003”.  [Google Scholar]; Fulginiti et al., 2004 Fulginiti, L. E., Perrin, R. K. and Yu, B. 2004. Institutions and agricultural productivity in sub-Saharan Africa. Agricultural Economic, 31(2–3): 169–180.  [Google Scholar]). Relative to this literature, our parametric approach is useful in that it allows us to perform tests for statistical significance.

4. A recent paper, Larson, Butzer and Mundlak (2008 Larson, D. F., Butzer, R. and Mundlak, Y. 2008. “Heterogeneous technology and panel data: the case of the agricultural production function. World Bank Policy Research Working Paper 4536”.  [Google Scholar]), shares and addresses the issue of endogenous technology. Larson, Butzer and Mundlak allow variables commonly associated with TFP to influence output both directly and indirectly, by estimating the elasticity of inputs by decomposing the sum of squares of panel data into its orthogonal components. Although they employ a different methodology, use a different capital stock series and use a sample that consists of developed as well as developing countries, they too find an important role for capital in explaining differences in agricultural productivity across countries.

5. In 1965, the four least productive countries in terms of labour productivity were Yemen, Saudi Arabia, Burkina Faso and Zambia. In 2000, the least productive countries were Ethiopia, Gambia, Mozambique, and Yemen. The most productive countries in terms of labour productivity in 1965 were Chile, Mongolia, Argentina and Uruguay, and in 2000, they were Libya, Uruguay, Argentina and Lebanon.

6. We are extremely grateful to Professor Evenson for providing us with access to this dataset.

7. The data set covers land planted to 11 major crop seed varieties. These 11 crops are: wheat, rice, maize, sorghum, millet, barley, groundnuts, beans, lentils, cassava and potatoes. Our study therefore quantifies the extent to which high-yielding seed varieties for food crops, rather than for export crops, have contributed to increasing agricultural labour productivity in the developing world.

8. For the first term in the expression substituting x = y−a provides: . The second expression is similarly obtained by using a = y−x.

9. We are very grateful to Luisito Bertinelli for providing us with the rainfall data. This data was used in a recent paper by Bertinelli and co-authors, Salvador Barrios and Eric Strobl, on the contribution of rainfall to Africa's growth tragedy (Barrios et al., 2008).

10. In Online Appendix F we use recent data on distortions to agricultural incentives (Anderson and Masters, 2009 Anderson, K. and Masters, W. A. 2009. Distortions to Agricultural Incentives in Africa, Washington: World Bank. [Crossref] , [Google Scholar]) and find that roughly 40 per cent of TFP variation stems from cross-country variation of agricultural pricing policy.

11. We compare the contributions of inputs with the most recent studies using the standard approach, given that earlier studies used much smaller samples, given a lack of data for many countries, while Craig, Pardey and Roseboom, 1997 Craig, B. J., Pardey, P. G. and Roseboom, J. 1997. International productivity patterns: accounting for input quality, infrastructure, and research. American Journal of Agricultural Economics, 79(4): 1064–1076. [Crossref], [Web of Science ®] , [Google Scholar] and Wiebe et al., 2003 use large samples containing countries from the same regions as the countries in our sample, and covering a similar time period (1961–1990/2000).

12. Although not employing an accounting approach, a number of previous studies have drawn attention to the productivity contributions of capital per worker in agriculture, such as Gutierrez and Gutierrez (2003 Gutierrez, L. and Gutierrez, M. M. 2003. International R&D spillovers and productivity growth in the agricultural sector. A panel cointegration approach. European Review of Agricultural Economics, 30(3): 281–303. [Crossref], [Web of Science ®] , [Google Scholar]) and Larson, Butzer and Mundlak (2008). The majority of past studies have taken agricultural capital to be just tractors. Our measure of capital is broader in that we include harvesters-threshers and milking machines. Removing such other types of capital in our analysis maintains the contribution of capital per worker at roughly 30 per cent.

13. Two inputs often discussed in the agricultural economics literature are human capital and livestock. Although we do not have ideal data for including these inputs in our main analysis, in Online Appendix E we note the contributions of these variables for explaining labour productivity variation for our data.

14. In Online Appendix D we perform sensitivity analysis on the value of θ _T by allowing θ _T to vary between 1.1 and 2. We find that the main results of the paper are qualitatively robust to such a wide range of this productivity parameter.