Consultative Group on International Agricultural Research CGR-22 ~~ CGIARCG=2 z~~;KUr I^TA ]Fk Study Paper Number 22 The International Agricultural Research Centers Their Impact on Spending for National Agricultural Research and Extension Robert E. Evenson FILE COPY 0006 The International Agricultural Research Centers CGIAR Study Papers No. 1 Technological Innovation in Agriculture: The Political Economy of Its Rate and Bias No. 2 Modern Varieties, International Agricultural Research, and the Poor No. 3 Plant Genetic Resources: The Impact of the International Agricultural Research Centers No. 4 Costa Rica and the CGIAR Centers: A Study of Their Collaboration in Agricultural Research No. 5 Guatemala and the CGIAR Centers: A Study of Their Collaboration in Agricultural Research No. 6 Zimbabwe and the CGIAR Centers: A Study of Their Collaboration in Agricultural Research No. 7 Nepal and the CGIAR Centers: A Study of Their Collaboration in Agricultural Research No. 8 Bangladesh and the CGIAR Centers: A Study of Their Collaboration in Agricultural Research No. 9 Brazil and the CGIAR Centers: A Study of Their Collaboration in Agricultural Research No. 10 Indonesia and the CGIAR Centers: A Study of Their Collaboration in Agricultural Research No. 11 Ecuador and the CGIAR Centers: A Study of Their Collaboration in Agricultural Research No. 12 Peru and the CGIAR Centers: A Study of Their Collaboration in Agricultural Research No. 13 Syria and the CGIAR Centers: A Study of Their Collaboration in Agricultural Research No. 14 Cuba and the CGIAR Centers: A Study of Their Collaboration in Agricultural Research No. 15 Philippines and the CGIAR Centers: A Study of Their Collaboration in Agricultural Research No. 16 Thailand and the CGIAR Centers: A Study of Their Collaboration in Agricultural Research No. 17 Gender-Related Impacts and the Work of the International Agricultural Research Centers No. 18 India and the International Crops Research Institute for the Semi-Arid Tropics: A Study of Their Collaboration in Agricultural Research No. 19 Burma and the CGIAR Centers: A Study of Their Collaboration in Agricultural Research No. 20 Chile and the CGIAR Centers: A Study of Their Collaboration in Agricultural Research No. 21 The Impact of Agricultural Research in Tropical Africa: A Study of the Collaboration between the International and National Research Systems Consultative Group on International Agricultural Research CGIAR Study Paper Number 22 The International Agricultural Research Centers Their Impact on Spending for National Agricultural Research and Extension Robert E. Evenson The World Bank Washington, D.C. Copyright (© 1987 The Intemational Bank for Reconstruction and Development/THE WORLD BANK 1818 H Street, N.W. Washington, D.C. 20433, U.S.A. First printing April 1987 All rights reserved Manufactured in the United States of America At its annual meeting in November 1983 the Consultative Group on International Agricultural Research (CGIAR) commissioned a wide-ranging impact study of the results of the activities of the international agricultural research organizations under its sponsorship. An Advisory Committee was appointed to oversee the study and to present the principal findings at the annual meeetings of the CGIAR in October 1985. The impact study director was given responsibility for preparing the main report and commissioning a series of papers on particular research issues and on the work of the centers in selected countries. This paper is one of that series. The judgments expressed herein are those of the author(s). They do not necessarily reflect the views of the World Bank, of affiliated organizations, including the CGIAR Secretariat, of the international agricultural research centers supported by the CGIAR, of the donors to the CGIAR, or of any individual acting on their behalf. Staff of many national and international organizations provided valued information, but neither they nor their institutions are responsible for the views expressed in this paper. Neither are the views necessarily consistent with those expressed in the main and summary reports, and they should not be attributed to the Advisory Committee or the study director. This paper has been prepared and published informally in order to share the information with the least possible delay. Robert E. Evenson, a specialist in agricultural development, is professor of economics at Yale University. Library of Congress Cataloging-in-Publication Data Evenson, Robert E. (Robert Eugene), 1934- The international agricultural research centers. (CGIAR study paper, ISSN 0257-3148 ; no. 22) Bibliography: p. 1. Agriculture--Research--International cooperation. 2. Agriculture--Research--Economic aspects. 3. Agriculture--Research--Government policy. 4. Consultative Group on International Agricultural Research. 5. Agricultural extension work--Economic aspects. 6. Agricultural productivity. I. Title. II. Series: Study paper (Cdnsultative Group on International Agricultural Research) ; no. 22. S540.I56E94 1987 338.1'6 87-8257 ISBN 0-8213-0879-3 v Abstract Many CGIAR Centers have been in place for a number of years. Important changes, both in the development of CGIAR Centers and in national research and extension programs, have taken place over this period. Two important questions about the impact of the CGIAR Centers are addressed in this paper. The first is whether the existence of the CGIAR Centers has influenced the size and character of national research and extension programs. The second is whether the research in the CGIAR Centers and the national systems has had any impact on crop productivity. The study estimates that the CGIAR Centers have had a positive impact on investment in national research programs in each of the crops for which CGIAR crop research programs exist except cassava. Estimates for livestock and horticultural crop research programs show a significant positive CGIAR impact as well. National extension spending is also stimulated by CGIAR programs. These estimates are based on an econometric specification that takes into account the impact of several economic development aid initiatives in addition to the activities of the CGIAR impacts. The study estimates that CGIAR Center programs have had significant impacts on crop productivity for maize, millets, sorghum, rice, wheat, beans, cassava and potatoes in all the regions studied. National research programs have had a positive impact on crop productivity in most of these crops as well. In addition, national extension programs have been productive in some crops. These estimates are based on crop production data in 25 countries. vi Acknowledgments I wish to acknowledge the research assistance of M. Ann Judd and the secretarial assistance of Leila Adams of the Economic Growth Center at Yale University. Robert Herdt, Grant Scobie and Carl Pray provided valuable comments. vii Contents 1 A Descriptive Summary of National and International Program Development 1 2 Specifying the Determinants of Investment in Research and Extension 11 3 Econometric Estimates: Investment Analysis 19 3.1 Aid Determinants - Two-Period Data 19 3.2 Research and Extension Determinants - Two-Period Data 25 3.3 Annual Data Analysis 30 4 Policy Implications of Investment Analysis 35 5 Impact of Investment on Productivity 41 5.1 Specification of the Productivity Relationship 41 5.2 Productivity Impact Estimates 45 6 Policy Implications of Productivity Analysis 55 Notes 59 Appendix Table 1: Agricultural Research Expenditures and Worker Years, by Region 61 Appendix Table 2: Agricultural Extension Expenditures and Worker Yearst by Region 69 References 73 1 1 A Descriptive Summary of National and International Program Development National investment in agricultural research and extension programs has grown at an impressive rate in the past 25 years.1 Tables 11 and 2 summarize this investment; detailed national data are presented in Appendix Tables 1 and 2. It may be seen that, in 1980 constant dollars, research spending in developing countries increased from 1959 to 1980 by a multiple of 5.8 in Latin America, 6.9 in Asia, and 3.6 in Africa. The comparable spending multiples for extension investment were 6.4 in Latin America, 3.5 for Asia, and 2.2 for Africa. Scientist-year (SY) multiples were lower than spending multiples (6.0 for Latin America, 4.1 for Asia, 4.2 for Africa), reflecting rising real costs per SY. (For extension workers the multiples were 6.8 for Latin America, 1.8 for Asia, 2.9 for Africa). Table 3 shows how research and extension "spending intensities," i.e., spending as a percent of the domestic value of agricultural product (G.D.P.) has changed from 1959 to 1980. These data show that in 1959 the low-income and middle-income developing countries were approximately twice as spending intensive for extension as for research.2 The reverse was true for the industrialized countries. The rapid growth in spending intensities for research from 1959 to 1980 combined with little or no growth in extension intensities in the 1970s, produced roughly equal spending intensities for research and extension in most developing countries. Table 4 provides comparable data for "worker intensities" (i.e. ratios of workers to G.D.P). For research, the same general pattern reflected in spending intensities is reflected in the workers intensities. Because spending per SY is lower in developing countries, they fare better by this measure. The difference between the low-income and industrialized countries is much reduced. 2 Table 1 Agricultural Research Expenditures and Workers EXPENDITURES WORKERS (000 Constant 1980 US$) (Scientist-Years) 1959 1970 1980 1959 1970 1980 REGION/SUBREGION Western Europe 274,984 918,634 1,489,588 6,251 12,547 19,540 Ntorthern Europe 94,718 230,135 409,527 1,818 4,409 8,027 Central Europe 141,054 563,334 871,233 2,888 5,721 8,827 Southern Europe 39,212 125,165 208,828 1,545 2,417 2,696 Eastern. Europe and USSR 568,284 1,282,212 1,492,783 17,701 43,709 51,614 Easterni Europe 195,896 436,094 553,400 5,701 16,009 20,220 iUSSR 372,388 86,,) 3_8 939,383 12,000 27,700 31,394 North America and Oceania 760,466 1,485,043 1,722,390 8,449 11,683 13,607 North America 668,889 1,221,006 1,335,584 6,690 8,575 10,305 Oceania 91,577 264,037 386,806 1,759 3,113 3,302 Latin America 79,556 216,018 462,631 1,425 4,880 8,534 Termperate South America 31,088 57,119 80,247 364 1,022 1,527 Tropical S,outh America 34,792 J.28,958 269,443 570 2,698 4,840 Caribbear. and Central America 13,676 29,941 112,941 491 1,160 2,167 Africa 119,149 251,572 424,757 1,919 3,849 8,088 North Africa 20,789 49,703 62,037 590 1,1.22 2,340 West Africa 44,333 91,899 205,737 412 952 2,466 East Af;ica 12,740 49,218 75,156 221 684 1,632 Southetn. Africa 41,287 60,752 81,827 696 1,091 1 ,650J Asia 261,114 1,205,116 1,797,894 1.1,418 31,837 46,656 West Asia. 24,427 70,676 125,465 457 1,606 2,329 SOULh Asia 32,024 72,573 190,931 1,433 2,569 5,691 Southeast Asia 9,028 37,405 103,249 441 1,692 4,102 East Azia 141,469 521,971 734,694 7,837 13,720 17,262 China 54,166 502,491 643,555 1,250 1.2,250 17,272 WORLD TOTAIL 2,063,553 5,358,595 7,390,043 4,]163 108,510 148,039 Sources: Boyce, J. K. and R. E. Evenson, National and International Agricultural Research and Extension Programs. (New York: Tha Agricultural Development Council, 1°75); and M. Ann Judd, James K. Boyce, and Robert E. Evenson, "Investing ;n Agricultural Supply" (Discussion Paper N4o. 442, Yale University, Economic Growth Center, 1383). 3 Table 2 Agricultural Extension Expenditures and Workers EXPENDITURES WORKERS (000 Constant 1980 US$) (Scientist-Year) 1959 1970 1980 1959 1970 1980 REGION /STBREC ION Western Europe 234,016 457,675 514,305 15,988 24,388 27,881 Northern Europe 11.2,983 187,144 201,366 4,793 5,638 6,241 Central Europe 103,082 199,191 236,834 7,865 13,046 14,421 Southern Europe 17,950 71,340 76,105 3,330 5,704 7,219 Eastern Europe and USSR 367,329 562,935 750,301 29,000 43,000 55,000 Eastern Europe 126,624 191,460 278,149 9,340 15,749 21,546 USSR 240,705 371,475 472,152 19,660 27,251 33,454 North America and Oceania 383,358 601,950 760,155 13,530 15,113 14,966 North America 332,892 511,883 634,201 11,500 12,550 12,235 Oceania 50,466 90,067 125,954 2,080 2,563 2,731 Latin America 61,451 205,971 396,944 3,353 10,782 22,835 Temperate South tnerica 5,741. 44,242 44,379 205 1,056 1,292 Tropical South America 47,296 136,943 294,654 2,369 7,591 16,038 Caribbean and Central America 8,414 24,786 57,913 779 2,135 5,535 Africa 237,883 481,096 514,671 28,700 58,700 79,875 North Africa 84,634I 176,498 172,910 7,500 14,750 22,453 West Africa 53,600 181,324 204,982 9,000 22,000 29,478 East Africa 39,496 86,096 106,030 9,000 18,750 24:,2'11 Southern Africa 60,153 37,178 30,749 3,200 3,200 3,-733 Asia 143,876 412,937 507,113 86,900 142,500 148,780 West Asia 28,211 97,315 119,780 7,000 18,800 16,535 South Asia 56,422 87,727 82,194 57,000 74,000 80,958 Southeast Asia 19,747 55,441 63,959 9,500 30,500 33,987 East Asia 39,496 172,454 241,180 13,400 19,200 17,300 China n.a. n.a. n.a. n.a. n.a. n.a. WORLD TOTAL 1,427,913 2,722,564 3,443,489 177,521 294,483 349,33/ Sources: Boyze, J. K. and R. E. Evenson, National and International Agricultura] kesearc;i and Extensicn Programs. (New York: IThe Agricultural Development Council, 1975); ana ni. Ann Judd, Jaines K. Boyce, anid Robert E. nvenson, "investing in Agri- cultural Supply" (Discussion Paper No. 442, Yale Universicy, Economic Growth Center, 1983). 4 Table 3 Research and Extension Expenditures as a Percent of the Value of Agricultural Product Public Sector Public Sector Agricultural Agricultural Research Extension Expenditures Expenditures Subregion ____1959 1970 1980 1959 1970 1980 Northern Europe 0.55 1.05 1.60 0.65 0.85 0.84 Central Europe 0.39 1.20 1.54 0.29 0.42 0.45 Southern Europe 0.24 o.61 0.74 0.11 0.35 0.28 Eastern Europe 0.50 0.81 0.78 0.32 0.36 0.40 USSR 0.43 0.73 0.70 0.28 0.32 0.35 Oceania 0.99 2.24 2.83 0.42 0.76 0.98 North America 0.84 1.27 1.09 0.42 0.53 0.56 Temperate South America 0.39 0.64 0.70 0.07 0.50 0.43 Tropical South America 0.25 0.67 0.98 0.34 0.71 1.19 Caribbean and Central America 0.15 0.22 o.63 0.09 0.18 0.33 North Africa 0.31 0.62 0.59 1.27 2.21 1.71 West Africa 0.37 0.61 1.19 0.58 1.24 1.28 East Africa 0.19 0.53 0.81 0.67 0.88 1.16 Southern Africa 1.13 1.10 1.23 1.64 0.67 0.46 West Asia 0.18 0.37 0.47 0.25 0.57 0.51 South Asia 0.12 0.19 0.43 0.20 0.23 0.20 Southeast Asia 0.10 0.28 0.52 0.24 0.37 0.36 East Asia 0.69 2.01 2.44 0.19 0.67 0.85 China 0.09 0.68 0.56 n.a. n.a. n.a. Country Group* Low-Income Developing 0.15 0.27 0.50 0.30 0.43 0.44 Middle-Income Developing 0.29 0.57 0.81 0.60 1.01 0.92 Semi-Industrialized 0.29 0.54 0.73 0.29 0.51 0.59 Industrialized 0.68 1.37 1.50 0.38 0.57 0.62 Planned 0.33 0.73 0.66 - - - Planned - excluding China 0.45 0.75 0.73 0.29 0.33 0.36 *For definition of Country Groups see footnote 4. Sources: Appendix Tables 1 and 2 and USDA, Indices of Agricultural Production, various issues. 5 Table 4 Research and Extension Workers Relative to the Value of Agricultural Product Extension Workers SYs per 10 Million per 10 Million (Constant 1980) (Constant 1980) Dollars Dollars Agricultural Agricultural Product Product Subregion 1959 1970 1980 1959 1970 1980 North Europe 1.05 2.01 3.14 2.76 2.56 2.61 Central Europe 0.80 1.21 1.56 2.19 2.77 2.73 Soutlhern Europe 0.93 1.17 0.96 2.00 2.76 2.69 Eastern Euroue 1.44 2.97 2.84 2.36 2.88 3.13 USSR 1.38 2.37 2.34 2.26 2.33 2.50 Oceania 1.91 2.64 2.43 2.26 2.17 2.11 NOrth Amerdca 0.84 0.89 0.84 1.44 1.31 1.08 Temperate South Amcrica 0.46 1.15 1.32 0.26 1.19 1.26 Tropical Soi,th America 0.41 1.41 1.77 1.71 3.95 6.46 Caribbean and Central America 0.53 0.86 1.20 0.82 1.53 3.12 North Africa 0.91 1.44 4.24 18.83 28.45 22.23 West Africa 0.33 0.61 1.42 7.61 14.01 18.08 East Africa 0.32 0.77 1.76 16.28 22.41 26.64 Southe_rn Africa 1.90 1.96 2.47 8.73 5.94 5.62 West Asia 0.33 0.84 0.88 4.39 7.25 6.54 South Asia 0.50 0.65 1.29 20.83 19.51 19.53 Southea.t Asia 0.47 1.28 2.07 9.81 13.07 19.72 East Asia 3.80 5.29 5.72 6.57 7.05 6.13 China 0.22 1.66 1.49 n.a. n.a. n.a. Countr-y Group Low-Incom7ne D)eveloping 0.43 0.67 1.40 18.14 18.61 20.43 Middle-lnco;ne Developing 0.69 1.31 2.40 8.89 14.68 15.98 Semi-Industrialized 0.70 1.21 1.36 2.80 4.95 5.21 Industrialized 1.24 1.71 1.85 2.37 2.31 2.12 Plar.ned 1.02 2.27 2.13 - - - Planned excluding China 1.40 2.54 2.50 2.29 2.49 2.63 Sources: Appendix Tables 1 and 2. 6 For extension, the picture is quite different. By 1959, low-income developing countries had attained very high extension intensities; 5 to 7 times greater than those attained in industrialized countries. By 1980, with a slight decline in these intensities for industrialized countries, the difference was even greater. Middle-income and semi-industrialized countries also increased their extension intensities. These worker intensities should not be interpreted as if there were no differences in the quality of workers among countries. There is little doubt that the general levels of training of both scientists and extension workers vary between countries and are lower in the developing countries. However, the differences are not as great as is generally supposed. There is also little indication that these differences have changed as research and extension spending has increased. These data do not include "extension type" spending associated with Rural Development Projects in developing countries. Were such data to be tabulated and included as extension spending, the magnitude of the differences in spending on extension relative to research in the developing countries would be even greater. Table 5 provides further insight into the motivation for the high extension worker intensities in developing countries. It shows expenditure worker ratios for research and extension. These ratios include salaries of scientists and extension workers and related costs, including laboratory costs and the costs of technicians. The ratio of research costs to extension costs is as much as 20 to 1 for the low-income developing countries and only 3 to 1 or so for the industrialized countries. Some of this difference is a quality difference (extension workers have quite advanced training in most industrialized countries and may have little training in low-income countries), and some is due to real cost differences. Many low-income countries do not have the capacity to train agricultural scientists and must incur high costs to train researchers and to purchase scientific equipment. 7 Table 5 Expenditures per SY/Extension Worker Extension Expenditures, Research Expenditures per Extension per SY Worker (000 Constant (000 Constant 1980 US$) 1980 US$) Region/Subregion 1959 1970 1980 1959 1970 1980 Western Europe 44 73 76 15 19 18 Northern Europe 52 52 51 24 33 32 Central Europe 49 98 99 13 15 16 Southern Europe 25 52 78 5 13 11 Eastern Europe '& USSR 32 29 29 13 13 14 Eastern Europe 34 27 27 14 12 13 USSR 31 31 30 12 14 14 North America and Oceania 90 127 127 28 40 51 Nort.h AVm.e:rica 100 142 130 29 41 52 Oceania 52 85 117 24 35 46 Latin America 56 44 54 18 19 18 Tem-,perate rourth America 85 56 53 28 42 34 Tropical Sout'h America 61 48 56 20 18 18 Caribbean end Central America 28 26 52 11 12 1. Africa 62 65 53 8 8 6 Nort'h Afri.ca 35 44 27 11 12 8 I-est Africa 108 97 83 6 8 7 Fast Africca 58 72 46 4 5 4 Southern Africa 59 56 50 19 12 8 Asia 23 38 39 2 3 3 West Asia 53 44 54 4 5 7 Southi Asia 22 28 34 1 1 1 Southeast Asia 20 22 25 2 2 2 East. Asia 18 38 43 3 9 14 ____la 4b.3 41 37 n.p. n,.a. -i - a. Coxuntry Group Low-Income Developing 34 40 47 2 2 2 1Middle-Inco.rne Deve1op4-ug 42 44 47 7 7 6 Semi.-Industrialized 41 45 46 10 10 11 1ndustrialized 55 80 93 16 25 29 P. anned 33 32 31 - - - Planned e%;ciuding China 31 25 30 1.3 13 14 Sources: See Tables 1 and 2. 8 Table 6 reports data on spending by commodity in the form of spending intensities. With few exceptions, developing countries cannot provide a commodity breakdown for their research spending. They do well to provide data on total spending. It is possible. however, to obtain publications data from the CAB Abstract system by commodity orientation. This was done for each of 25 countries for two periods 1972-75 and 1976-80. These data were then standardized into equal cost units utilizing Brazilian data. For Brazil real spending by commodity and CAB publications data were available. It was thus possible to standardize publications into cost equivalent units. Standardized publications were then used to allocate actual expenditures to commodities. The data show that spending intensities differ greatly by commodity in the 25 country sample (these 25 countries account for approximately 90 percent of total production in developing countries, excluding China). Spending intensities are low for coconuts, sweet potatoes and cassava and high for cocoa, coffee and livestock. The table also shows that the IARCs account for relatively low shares of the total research expenditures on the commodities they work on. Since expenditures per SY are very high in the IARCs (about 4-6 times the average for national spending), the IARCs are much less significant in terms of their share of scientific personnel devoted to these commodities. Table 7 reports the CAB publications data in the form of ratios of "basic" to "applied" research. Abstracting journals are classified as to whether they are oriented to relatively basic reseach fields or to relatively applied fields (see the notes to the table for the classification). While this procedure is very crude it does provide a basis for comparing the research programs of developing countries with the research programs of developed countries. The table shows that the 25 developing countries have slightly higher ratios of basic to applied research on crops and substantially higher ratios of basic to applied research on animals. 9 Table 6 Research as a Percent of the Value of Product. by Commodity. Average 1972-79 Period, 25 Countries REGION Spending by Ratio IARC Latin All International Spending -CONlIDITY Africa Asia America CountriesZ Centrs to Total Wheat 1.30 0.32 1.04 0.51 0.02 0.04 Rice 1.05 0.21 0.41 0.25 0.02 0.07 Maize 0.44 0.21 0.18 0.23 0.03 0.11 Cotton 0.23 0.17 0.23 0.21 - - Sugar 1.06 0.13 0.48 0.27 - Soybeans 23.59* 2.33 0.68 1.06 - - Cassava 0.09 0.06 0.19 0.11 0.02 0.15 Field Beans 1.65 0.08 0.60 0.32 0.04 0.11 Citrus 0.88 0.51 0.57 0.52 - _ Cocoa 2.75 14.17* 1.57 1.69 - - Potatoes 0.21 0.19 0.43 0.29 0.08 0.21 Sweet Potatoes 0.06 0.08 0.19 0.07 - - Vegetables 1.56 0.41 1.13 0.73 - - Bananas 0.27 0.20 0.64 0.27 - - Coffee 3.12 1.25 0.92 1.18 - - Groundnut 0.57 0.12 0.60 0.25 0.005 0.02 Coconut 0.07 0.03 0.10 0.04 - - Beef 1.82 0.65 0.67 1.36 0.02 0.02 Pork 2.56 0.39 0.60 1.25 0.02 0.02 Poultry 1.99 0.32 1.12 1.64 - - Other Livestock 1.81 0.89 0.42 0.71 - Sources: M. Ann Judd, James K. Boyce, and Robert E. Evenson, "Investing in Agricultural Supply" (Discussicin Paper No. 442, Yale University, Economic Growth Center, 1983); and USDA, Indices of Agricultural Production, various issues. (*) Ratios are high because production is very low. 10 Table 7 Ratios of Basic to Applied Research Crop Research Animal Research ______________________1972-75 1976-79 1980-83 1972-75 1976-79 1980-83 Argentina .13 .16 .08 .33 .59 .90 Brazil .18 .19 .17 .66 .97 .91 Chile .13 .13 .14 .38 .47 .59 Colombia .15 .17 .22 .34 .61 .90 Mexico .16 .10 .07 .32 .61 .90 Peru .25 .49 .26 .23 .15 .44 Venezuela .18 .14 .12 .51 .95 1.40 Ghana .12 .07 .12 .25 .48 .53 Kenya .15 .16 .18 .23 .71 .96 Nigeria .14 .22 .19 .32 .59 .64 Sudan .12 .04 .13 .58 .53 .60 Tanzania .04 .07 .13 .93 1.11 1.11 Tunisia .09 .05 .07 .57 1.18 2.10 Uganda .10 .06 .23 .29 .97 1.79 Egypt .14 .16 .16 .30 .41 .50 Sri Lanka .08 .09 .09 .33 .36 .26 India .21 .27 .26 .29 .43 .38 Indonesia .05 .10 .08 .64 .92 .43 South Korea .14 .15 .19 .58 .43 .61 Malaysia .22 .21 .17 1.07 .61 .51 Pakistan .10 .08 .09 .36 .43 .43 Philippines .19 .16 .15 .51 .37 .30 Taiwan .17 .29 .27 .76 .42 .30 Thailand .17 .16 .18 1.37 1.97 2.68 Turkey .41 .40 .28 .47 .73 .50 25 Developing Countries .18 .22 .21 .37 .52 .54 All Developed Countries .16 .15 .16 .23 .34 .30 Note: Ratios are based on counts of abstracted publications by class of journal defined as follows. Basic Crop Journal: Helminthological Abstracts (B); Rev. Plant Pathology Applied Crop Journals: Field Crops Abstracts, Herbage Abstracts, Horticultural Abstracts, Review of Applied Entomology, Soils and Fertilizers, Wood Abstracts. Basic Animal Journal: Helminthological Abstracts, Protozoologist Abstracts, Review of Med. & Vet. Mycology Applied Animal Journals: Animal Breeding Abstracts, Dairy Science Abstracts, Nutrition Abstracts (land and feeding), Rev. Applied Entomology (A), 1 1 2 Specifying the Determinants of Investment in Research and Extension If IARC impacts on national research and extension spending are to be measured, a specification relating national spending to "determinants," including IARC investment, is required. Such a specification should be consistent with economic logic and political reality. Since IARC investments are commodity based, it is natural to develop the specification for spending by commodity. The specification developed here is motivated by a project evaluation or planning perspective modified by political constraints. The specification includes variables that a rational planner would use to guide optimal investment. It also includes variables that reflect the political power of interest groups and political constraints. Before discussion of the specification it will be useful to discuss the data to be utilized and to list the variables in the data. Two data sets have been constructed. The first is a data set where the observations are for two periods, 1972-75 and 1976-80 for 24 countries.3 For this data set it was possible to obtain aid variables, thus allowing a test of the role of aid in influencing national spending. The second data set is for the same countries, for a reduced set of variables measured annually for the 1962-82 period. The observations in both data sets are on commodities (i.e., an observation is for a commodity, a country and a year) (or an average of 1972-75 or 1976-80 for the first data set). The field crop commodities are rice, wheat, maize, sorghum, millets, cassava, field beans, potatoes, sweet potatoes, groundnuts, sugar and soybeans, livestock and horticultural crops include bananas, coffee, coconut, beef, pork, poultry and other livestock. 12 Table 8 provides a list of the variables for the two data sets with a short definition of the variable. Those variables marked with an asterisk are measured on a country rather than a commodity basis. That is, they are common to all commodities (accordingly their means are not comparable to the means of variables actually measured on a commodity basis). The variables are classified as endogenous, i.e., the choice variables being subject to analysis, partially endogenous, and exogenous. The exogenous variables are further classified as "economic" variables, "international transfer" variables, and "political-economic" variables. The dependent variables in the analysis are the variable measuring national research spending and national extension spending. RESEXP (measured in millions of 1980 dollars). EXTEXP (measured in millions of 1980 dollars. This variable is not measured on a commodity basis). The model by which this spending is determined is constructed in stages. The first stage is motivated by supposing that a planner is attempting to maximize the economic surplus, (i.e., both consumers' and producers' surplus) associated with the research or extension program. In the second stage the planner takes international transfer conditions into account. In the third, the planner takes political constraints into account. (This is the rationale for the classification of exogenous variables in Table 8). Before discussing these variables, it should be noted that several aid variables, AID, NDONORS, WBEXT, WBRES, NHSTAFF, and INTCR are also included in the model. These cannot be considered to be exogenous determinants of national spending, however, since actions by the recipient countries as well as choices by donors 13 Table 8 Variables Dictionary: Research and Extension Investment Analysis 1972-75 1976-80 Data 1962-82 Data Mean Std. Dev. Mean Std. Dev. I. Endogenous (Choice) Variables RESEXP: Annual Spending (millions of 1980 dollars) by Commodity on Research .9819 2.24 0.69 1.70 EXTEXP: Annual Spending (millions of 1980 dollars on Extension (all commodities) 30.68 41.95 26.50 39.60 II. Partially Endogenous Variables AID: Value of Aid from all Sources (millions of 1980 dollars) 25.00 17.67 n.a. NDONORS: The Number of Donors Providing Aid to Research 4.92 2.93 n.a. WBRES: World Bank Supported Research Programs (including national commodity) 10260 93445 WBEXT: World Bank Supported Extension Programs (including national components) 10383 67300 NHSTAFF: Number of IARC Scientists in Countries Other than IARC Host Countries 3.88 3.52 INTCR: Number of Joint IARC-Joint IARC-National Research Collaborative Research Agreements .27 1.44 BASIC: Ratio of Non-commodity Oriented Research to Commodity Research (See Table 7) 24.97 6.84 CONGRU: A Measure of Congruence Between Research Spending and Commodity Value .85 .13 CONGRU = 1 - E (V -C 2 where Vi is research share, Ci is Commodity share III. Exogenotus Variables A. Economic PROD: Value of Commodity Production (millions of 1980 dollars) 223.34 653.63 2113.62 8452.20 DIVER: Inverse of the Sum of Squared Shares of Production in Commodity Geo-Climate! Combinations 0.4118 .21 0.39 0.20 EXPRAT: Ratio of Expenditures per SMY to Expenditures per Extension Worker 10.14 9.99 9.44 9.10 ARABLE: Ratio of Arable Land in the Current Period to Arable Land Six Years Earlier 1.09 .11 1.05 0.10 CINTSP: Cumulated Research Expenditures on the Commodity in IARCs (millions of 1980 dollars) 6.17 13.78 4580.59 10148.80 (Table continued on the following page.) 114 Table 8 Variables Dictionary: Research and Extension Investment Analysis (continued) 1972-75, 1976-80 Data 1962-82 Data Mean Std. Dev. Mean Std. Dev. B. International Transfer RESNSR: Research Scientist Manyears on the Commodity by Neighboring Countries in Similar Geo-climate Regions (millions of 1980 dollars) 8.67 12.61 5.14 7.60 INTLOC: A Dummy Variable = 1 if the Country is Hosting the IARC Under- taking Research on the Commidity .019 .14 n.a. TOTALAREA: Total in Crops in the Country (000 ha) 10715.19 20902.44 10740.77 21558.60 C. Political - Economics IMPORTS: Value of Imports of the Commodity (millions of 1980 dollars) 16.39 71.68 n.a. EXPORTS: Value of Exports of the Commodity (millions of 1980 dollars) 24.46 100.75 n.a. UREARICE: Ratio of Prices Paid by Farmers for Urea Fertilizer to Prices Received for Rice 2.74 1.61 2.76 1.70 ECONAG: Percent of Economically Active Population Working in Agriculture 54.45 19.77 56.62 20.20 URBANPOP: Percent of the Total Population Living in Urban Areas of 100,000 Population or More 34.53 21.58 32.05 21.10 VIOLD: Percent of Population Killed in Domestic Political Violence in Past Decade .12(-10) .12(-9) 0.00 0.00 D. Other Ti: A Dummy Variable = 1 if time Period is 1972-75 0.05 0.5 n.a. RI: A Dummy Variable = 1 if Country is Located in Asia 0.4 0.49 n.a. R2: A Dummy Variable = 1 if Country is Located in Africa 0.32 0.47 n.a. 15 responding to characteristics of recipient countries determine this spending. Thus these aid variables must be regarded to be simultaneously determined along with national spending. (See the following section for a discussion of the econometric treatment.) Now consider the first stage of the planner's problem. A given research program can be expected to lower production costs per unit of production. The more units over which costs can be lowered, the higher the optimal level of research. Each commodity and each geo-climate region present different research problems to some degree. Hence units of production should be measured on a commodity-region basis. The two variables PROD (production) and DIVER (diversity) (and the interaction of these two variables) are designed to pick up these effects.4 National research spending is expected to rise as both production and diversity increase. For some (perhaps most) research programs a "minimum critical mass" of research effort may be required for an effective program. If so there will be a threshold level of production below which a research program cannot be justified. Small diverse countries are more likely than larger countries to face these problems. The variables EXPRAT and ARABLE are price variables reflecting prices of alternative sources of growth in supply. EXPRAT. the ratio of expenditures per SY to expenditures per extension worker, is designed to reflect the relative costs of pursuing growth through extension investment. (Expressing it in ratio terms avoids the need to specify an exchange rate.) It is expected that when the price of research resources falls relative to extension resources more spending in research will take place. The ARABLE variable (the ratio of arable land currently to arable land 6 years previously) is designed to reflect the price of supply -growth via land expansion. When the change in arable land 16 is small, reflecting land exhaustion, more spending on research is expected. Now turn to the second stage of the problem. The planner recognizes that technology may "spill-in" from other countries and from IARCs. He also recognizes, however, that the potential spill-in technology was designed for or "targeted" to geo-climate conditions in other countries. Other national programs will be targeting their research programs to their own geo-climate conditions. The IARCs may target to a broader range of conditions than are extant in their host countries, but in practice they lack the resources to provide technology targeted to more than a limited range of environments. Thus, the planner will find that some technology available on the international market is directly suited to use (i.e., it is targeted to domestic conditions but that much new technology (and related research findings) is "mismatched," i.e., it is targeted to geo- climate conditions differing from those of the country. It is hypothesized that the planner's response to closely matched technology from abroad will be to reduce domestic research investment since domestic research is a substitute for matched technology from abroad (extension spending may be inversed). Likewise, the planner's response to mismatched technology from abroad may be to increase domestic research investment since this mismatched technology offers domestic researchers an opportunity for modification and adaptation of the mismatched technology to domestic conditions. Of course, if the mismatch is too great it will not offer such opportunities. We would then expect planners to exhibit a mixed response to technology from abroad. On the one hand, they will "free ride" on the research of IARCs and neighboring countries to the extent that they see these research units as producing closely matched technology with little scope for adaptation. On the other hand, they will respond with increased adaptive research to the extent that they see these units producing mismatched technology 17 offering adaptation opportunities and to the extent that these units are producing "pre-technology" scientific discoveries that also enhance the productiveness of their own systems. The variables CINTSP (cumulated spending in IARCs on the commodity) and RESNSR. (SYs working on the commodity in geo- climate neighboring countries) are measures of the programs that a national planner will respond to. Whether the response will be a net negative free-riding response or a net positive adaptive opportunity response depends on the nature of the technology. The variable TOTALAREA is a measure of the size of the country and the interaction of this variable with CINTSP is designed to identify whether the response to IARC investment differs for large and small countries. Finally, the planner will respond to political constraints. The variables IMPORT and EXPORT measure the effects of international trade. Most countries implicitly place a higher value on international exchange than on domestic production. A unit of product that saves or earns foreign exchange is valued more highly than one that does not. A planner will respond to this by investing more in research on commodities that save or earn foreign exchange. Many countries intervene in agricultural markets. The UREARICE variable (the ratio of prices paid for urea fertilizer to prices raised for rice) is a measure of this intervention. A planner might attempt to "compensate" for some types of intervention by spending more or less on research. The variables, ECONAG, URBANPOP and VIOLD, are crude proxies for political organizations as well as for interest group power. A planner will respond to pressure from interest groups, for example to urban pressure groups by shifting resources from research to competing investments even though urban consumers are the major beneficiaries of agricultural research.5 High proportions of the labor force in agriculture are usually 18 associated with weak political power of rural people. If so, this could reduce spending on research and extension. These political variables. it should be noted, are proxies for many different combinations of interests and the ability to translate these interests into political action. In the absence of a political model little interpretation can be given to measured impacts. The justification for the inclusion of these variables in the model is simply that they may control for some difference in political conditions and reduce bias in the estimated parameters that can be given stronger interpretations. 19 3 Econometric Estimates: Investment Analysis Table 8 lists the variables discussed above. The actual specification requires a procedure for handling the partially endogerious variables. basically the aid variables. In addition the functional form has to be specified. The two-period data set (set 1) does not have sufficient observations to estimate investment relationships for each commodity. It does contain aid variables and is suited to a general analysis of research investment based on pooled commodity observations. The second data set for the 1962-82 period does contain sufficient observations to enable an analysis of determinants of spending for each commodity and for extension spending as well. It does not contain aid variables. 3.1 Aid Determinants - Two-Period Data The specification for the two period data set and for the aid analysis is considered first. This specification requires that national research spending and aid be treated as simultaneously determined. A Two-Stage Least Squares procedure is appropriate. The endogenous variables are: AID, NDONORS, NHSTAFF, WBRES. WBEXT, INTCR. CONGRU, BASIC, EXTEXP and RESEXP. The latter two variables are the most important from the perspective of this analysis. The model treats each of the first eight variables as dependent on both EXTEXP and RESEXP in addition to a number of exogenous variables. EXTEXP and RESEXP are treated as dependent only on aid (AID or WBRES and WBEXT) and a different set of exogenous variables. The econometric estimates based on this model are reported in Tables 9 and 10. Table 9 reports the results of the aid variables and for characteristics of national systems. Table 10 summarizes the main results showing determinants of investment in 20 field crop research. livestock and horticultural crop research and in extension. The functional form used is linear except that several multiplicative or interaction variables are used. These are: PROD2 = PROD x PROD PRDDIVER = PROD x DIVER PRDXPORT = PROD x EXPORTS PRDMPORT = PROD x IMPORTS INTSPLOC = INTLOC x CINTSP AREACINT = TOTALAREA x CINTSP AREADIV = TOTALAREA x DIVER The B001 notation identifies the endogenous variables in each equation. In Table 9, national research spending. RES and extension spending EXTEXP are the endogenous variables treated in determining aid flows and characteristics of national research systems (these variables are predicted in Table 10). As the table shows, aid agencies do appear to respond to national investment in extension but not to investment in research. Higher extension spending appears to reduce both the aid level to agricultural research and the number of donors providing that aid. A measure of general aid to extension is not available but the results do show that World Bank aid to extension responds positively to national spending levels. (Of course, as Table 10 shows, national spending responds positively to World Bank support as well. The two-stage least squares procedure is designed to identify the separate causal relationship.) Higher extension spending also appears to induce research programs with higher fractions of non-commodity oriented components.6 It also induces more IARC aid in the form of non-host staffing. The positive TOTALAREA and negative AREADIV coefficients in the AID, NDONORS, WBRES and WBEXT equations show that aid 21 Table 9 Estimated Coefficient and Statistics of Two-Stage Least Squares Equations for Determinants of Aid* Dependent Variables Independent Variables AID NN WBES BXT ITNA IN1gmR BASIC -_cam Intercept 21.541 6.93 -44.15 -39.45 4.70 2.22 13.38 .264 (2.02) (4.49) (2.55) (1.46) (2.36) (1.76) (3.38) (4.17) B00l.RES** .830 .022 1.31 .305 .112 -.010 .191 .0009 (1.27) (.23) (1.24) (.18) (.92) (.13) (.78) (.24) B001.E=IP** -.298 -.018 -.063 .316 .050 -.011 .087 -.0001 (5.49) (2.30) (.71) (2.29) (4.91) (1.71) (4.35) (.30) TOTLAREA .003 .0002 .003 .004 -.0003 .00005 -.0008 xo10,6 (10.46) (5.49) (6.72) (5.53) (6.10) (1.49) (6.84) (.54) AMADIV -.010 -.0008 -.010 -.014 .001 -.0002 .0025 1x106 (9.65) (5.63) (6.18) (5.44) (6.72) (1.43) (6.89) (.24) UREARICE -3.070 -.328 -4.82 1.96 .115 .057 2.48 .008 (4.88) (3.61) (4.73) (1.23) (.98) (.76) (10.67) (2.26) ARABLE -12.972 5.83 5.94 46.52 -.097 .618 10.35 .035 (1.88) (5.86) (.53) (2.66) (.08) (.76) (4.06) (.87) EC-NAC .595 -.048 .946 .099 -.024 -.029 -.180 .005 (4.50) (2.49) (4.41) (.028 (.96) (1.83) (3.66) (6.88) URBANPOP .119 -.131 .423 -.195 -.047 -.019 -.016 .007 (1.12) (8.49) (2.45) (.72) (2.39) (1.55) (.41) (11.40) VTCID 5547.1 2637.9 24723 68422 5399.7 -1200.7 22495 90.88 (.66) (2.22) (1.85) (3.27) (3.51) (1.23) (7.38) (1.86) INTLOC 5.766 .510 -2.94 15.24 1.54 3.09 -5.29 -.048 (1.40) (.86) (.44) (1.46) (2.01) (6.35) (3.47) (1.98) CflNSP -.005 .0015 -.035 .066 -.005 .003 .050 .0005 (.11) (.23) (.46) (.56) (.62) (.51) (2.44) (1.48) AREACINT -4x10-7 -4x10-8 -8x10-7 xO-6 -1x10-7 1x10-6 -3x104 -4xlO9 (.23) (.13) (.24) (1.55) (.33) (5.49) (.04) (.32) EXPRAT -.675 -.010 -.563 -.322 .003 -.031 .151 -.0008 (4.35) (.48) (2.24) (.82) (.10) (1.71) (2.62) (.88) RESNSR - - - - - - -.116 -.001 (3.16) (1.89) F 23.55 29.53 24.03 37.68 10.42 9.33 15.64 22.24 R2 .384 .438 .388 .4989 .216 .198 .308 .388 *Asolute values of asympotict-ratios in parentheses **he BCOI notaticnindicates that these variables ae treated as endogemus variales (See Table 10). Table 10 Estimated Determinants of Two Major Groups of Research and Extension Spending* Dependent Variable Horticultural Field Crop Crop and Livestock National Independent Research SRending Research Spending Extension SRending Variables (1) (2) (3) (4) (5) (6) Intercept 2.69 2.36 3.08 3.37 43.01 75.72 (2.48) (2.27) (1.94) (2.16) (1.56) (3.18) PROD .001 .001 .005 .005 - (2.98) (3.91 (5.92) (5.65) - PROD2 -1,2x10-7 -1.2x10-7 -lx106 -1x10-6 - (3.98) (4.16) (5.58) (5.58) - - TOTLAREA - _ _ .005 .003 - _ - - (4.79) (4.21) DIVER 1.09 .055 .287 1.17 9.15 8.89 (1.54) (.14) (.27) (2.08) (.28) (1.11) PRDIVER .001 .0005 -.005 -.004 - (1.17) (.63) (3.03) (2.61) - - AREADIV - - - - -.014 -.007 - - - - (3.88) (2.93) UREARICE .044 .033 -.125 -.088 -5.82 -5.16 (.81) (.06) (1.62) (1.29) (2.16) (5.51) ARABLE .574 .493 -1.20 -1.11 -2.51 -19.88 (1.03) (.91) (1.48) (1.38) (.17) (1.57) ECONAG -.060 -.039 -.015 -.034 .223 -.297 (3.35) (3.27) (.58) (1.91) (.23) (1.09) URBANPOP -.035 -.023 -.013 -.022 -.113 -.309 (2.91) (2.40) (.75) (1.50) (.24) (1.45) EXPRAT -.008 -.011 .054 .053 -1.87 -1.67 (1.05) (1.45) (5.17) (4.83) (7.31) (7.85) INTLOC -.285 -.095 1.27 .975 - - (.57) (.19) (.99) (.79) - - INTSPLOC .211 .185 .378 .342 - - (1.12) (1.00) (.70) (.63) - _ PRDXPORT 2x10-6 2x10-6 lx10-5 lx10-5 3x10-5 2.5x10-5 (5.71) (5.50) (10.85) (10.86 (3.59) (3.35) PRDMPORT 1.7x10-6 1.6x10-6 7x10-5 7x10-5 -3x1O-6 -3x10-6 (9.43) (9.53) (5.01) (5.04) (.90) (1.00) RESNSR .031 .024 .019 .023 -.179 -.277 (3.35) (2:95) (2.38) (3.17) (.58) (1.66) Ti .126 .048 .239 .283 -9.19 2.87 (1.03) (.42) (1.29) (1.57) (2.75) (.66) Rl -.451 -.156 -.786 -.951 -3.42 -4.23 (1.59) (.55) (1.88) (2.32) (.38) (.76) R2 .409 .204 -.111 .123 25.02 34.59 (1.26) (.70) (.24) (.29) (2.05) (5.65) CINTSP -.002 2x10-5 .026 .025 .018 .086 (.42) (.00) (3.00) (2.90) (.14) (.93) BOO1.AID .027 _ -.020 - .012 (2.04) - (1.02) - .01 BOO1.WBRES - .006 - -.001 - (1.38) - (.20) - - B001.WBEXT - - - - .367 w - ~~- (3.30) AREACINT lx10-6 lxl-6 1.6x10-6 1.6x10-6 3 7x10-6 2.5x10-7 (4.29) (4.07) (6.15) (6.05) {1.04) (.08) R2F 43.17 45.03 32.16 32.36 35.31 47.25 R2 .64 .65 .59 .60 .55 .62 *Absolute values of asymptotic t-ratios in parentheses. 24 agencies respond negatively to diversity. They provide more aid to large countries with little diversity. Countries with small areas and high levels of diversity are in some sense discriminated against by donors. This is in contrast to a result in Table 10 showing that national governments do not respond negatively to diversity in their own funding decisions. Interestingly the IARCs do respond positively to diversity in their non-host staffing decisions. It appears that when governments pursue high fertilizer/rice price policies (interpreted here as general policies discriminating against farmers and in favor of consumers) aid agencies respond by offering less aid to research (and possibly more to extension). They do not compensate for anti-supply policies by investing more in research. Their research programs are also more basic and more congruent. That is they are less commodity oriented and better matched to their commodity production patterns. Aid donors generally tend to respond to land exhaustion (i.e., low levels of the ARABLE variables) by offering more aid to research. The World Bank does not. Aid donors including the World Bank do appear to respond positively to the importance of the agricultural work force in the general labor force. This is in contrast to the tendency of national programs to spend less when the proportion of workers in agriculture is high. This is perhaps the one dimension where aid donors appear to be inducing more "qualitatively optimal" programs. Aid donors do not appear to respond to IARC locations in their programming. The IARCs, however, do favor IARC host countries in their placement of non-host staff and research contracts and collaborative agreements -- that is, centers tend to outpost staff and conclude agreements in countries where other centers are located. 25 The! qualitative dimensions of national programs appear to respond to political factors to some extent. A higher proportion of the labor force in agriculture appears to induce more commodity oriented and more congruent research programs. National programs also appear to respond to strong research programs by geo-climate neighbors by undertaking a lower proportion of non-commodity research. 3.2 Research and Extension Determinants -- Two-Period Data Table 10 reports the most important results of this analysis. It shows the determinants of national research spending on field crops research, on livestock and horticultural crops research and on extension spending. Two versions of each equatiorn are reported. In the first (e.g., 1, 3 and 5) general aid is treated as a determinant of spending. In the second (e.g., 2, 4 and 6) World Bank aid to research (or extension) is treated as the determining variable. Cumulated IARC spending (CINTSP) on the commodity is treated as an exogenous variable7 and tests whether IARC programs have stimulated or retarded national spending. This variable is also interacted with a variable measuring the size of the crop area in the country (AREACINT = TOTALAREA x CINTSP). This is designed to measure whether the IARC impact is related to the size of the country.8 Table 10 shows that IARC spending did not affect extension spending, but that it clearly did have a positive impact on both field crop research spending and on livestock and horticultural crop research spending. Further, the impact is positively related to the size of the country being affected. For field crop research the approximately zero coefficient on CINTSP shows that for small countries there is little or no IARC impact. For small countries the AREACINT variables has a low value. For large countries the positive impact is substantial. For livestock and horticultural crops it appears that a positive 26 impact holds even for small countries. These results are not affected by the choice of aid variables. The response of national research system spending to IARC spending is consistent with the estimated positive response to research undertaken by geo-climate neighbors. The RESNSR variable measures the scientist years devoted to the commodity by other countries in the same broad geo-climate zone. The positive response to this research and to IARC research shows that national systems see this research as opening up adaptive opportunities for their own research investment. The fact that countries do not respond to this research spending by spending more on extension is also consistent with a perception that the low technology being produced in these systems is not so well matched to their own production environments that they can simply facilitate its "spill-in" and adoption by investing in extension. Thus the pattern of response in both research and extension spending to both the IARC research and the research of geo- climate neighbors is consistent with the fact that agricultural technology has a high degree of location specificity. The typical developing country appears to have recognized that new technology does not easily spill-in from abroad and that low cost extension investment is not sufficient to facilitate its transfer. On the whole, technology produced abroad is mismatched to conditions at home. The degree of the mismatch is not so great, however, that it does not prevent new opportunities for adaptive research at home. In addition to mismatched technology. research institutions abroad are also producing pre-technology science of relevance. It too is of value at home only when a strong research capacity has been built. This interpretation of the IARC impact has important policy implications (as described below). The statistical measures reported in Table 10 support this interpretation. However, it is also important that the more general investment estimates be 27 judged against ja priori logic or expectations to determine whether the specific IARC impacts are part of a generally consistent investment relationship. To this end, consider the impacts of the economic variables on investment. For all research activities, the PROD and PROD2 impacts are significant and as expected. Holding geo-climate diversity constant, an increase in the units produced of a commodity offers a type of scale economy to a research system. Thus spending per unit of production will decline as shown by the negative production squared term. An increase in diversity itself does not have a strong impact on field crops research, (although it is positive), but does appear to stimulate more spending on livestock anQ horticultural research when production is low. High levels of diversity reduce the production inputs on this research spending. The same situation holds for extension spending. Higher levels of diversity lower the impact of total area on extension spending. This appears to be a kind of diseconomy or discouragement effect. The expected negative sign on the ARABLE variable is borne out only for the livestock and horticultural crops research (and possibly for extension). When the ratio of arable land currently to arable land 6 years previously is low it is indicating an exhaustion of arable land. The EXPRAT variable measures the ratio of a "price" of research services to a price of extension services. Since the dependent variable is expressed in expenditure terms if this variable has a zero coefficient, the actual price elasticity is -1.9 Since this ratio is probably measured with error its coefficient will be biased toward zero. It is important, therefore, that the standard error be considered in interpreting this variable. To facilitate this a range of price elasticities 28 (*1 standard deviation) is reported in the following section. This range shows that prices do matter. Those countries that have lowered this ratio by developing a capacity for training scientists at home and a reduced dependency on costly expatriate scientists have responded by buying more units of research and by spending more on research. The variables measuring political factors are important. They show very strong international trade effects. If a commodity is exported more research per dollar of product is expended for all commodities. Export orientation also stimulates extension spending. This impact is higher for the horticultural crops and livestock, perhaps reflecting post-colonial effects in which research or export commodities traditionally had strong "mother country" support. It is interesting, however, that the impact of imports of the commodities has a stimulus effect of roughly the same magnitude in field crops and of larger magnitude for the livestock and horticultural crops. Imports do not affect extension spending. This extra attention to traded commodities has several rational explanations. Most developing countries have pursued general economic policies that place a high value on foreign exchange. Demand elasticities for traded crops are high so supply can be increased without significant reduction in market prices. Increased imports of commodities may also provide political signals that something should be done about domestic supply. Of course, there still may be a colonial legacy reflected in the data but the import effects suggest that a more general set of factors are operating to favor traded over nontraded commodities. The variable proxy for agricultural price policies, UREARICE, does not have significant effects on research although countries pursuing price policies that discriminate against farmers (as measured by a high urea-rice price ratio) tend to 29 spend less on livestock and horticultural crop research. They also spend less on extension thus they do not attempt to compensate for negative price effects on supply by spending more on research and extension. The variables measuring the characteristics of the agricultural labor force and the urbanization of the population reflect very good political processes and cannot be given very clear interpretations. An increase in the percent of the population living in urban centers of 100,000 population and more tends to reduce spending on research and extensions particularly on field crops research. This presumably is measuring political power with an interest in directing government spending to nonagricultural interests. Countries with high proportions of their labor force in agriculture also spend less on research and extension, particularly field crop research. This variable is not measuring the same phenomena as the urbanization variable, but it is not inconsistent to suggest that farmer political power is actually weakest in the poorest economies with high proportion of workers in agriculture. Since this variable is also a proxy for the general wealth of a society it may be measuring a kind of wealth effect. If so it should be noted that there is a certain irrationality behind it since investment in research and extension is a production investment, not a form of public consumption. The results reported in Tables 9 and 10 are based on the two- period data set for which aid variables are available. The results with respect to the aid variables show that general aid for research (as measured by AID) does increase research spending for field crops research but not for livestock and horticultural crop research or for extension. The coefficients show displacement of aid effects on research spending of two sorts. First, research spending on field crops does not increase by the full amount of the aid. Second some reduction in livestock and horticultural crop research is induced by aid. 30 The results when World Bank aid is provided are similar for aid to research although the apparent displacement is more severe. World Bank aid to extension, on the other hand, provides a strong stimulus to national extension investment.10 The magnitude of the aid and other impacts on spending will be discussed further in the concluding policy section of the paper. Before turning to that discussion, results from the second data set are reported. 3.3 Annual Data Analysis The annual data set, as noted earlier, does not have data on aid variables. It is, however, considerably richer in terms of observations by commodity. Accordingly the results reported in Table 11 are by commodity and for pooled commodity groups: cereals (maize, sorghum, millet, rice, wheat), staples (beans, cassava, groundnuts, potatoes, sweet potatoes) and commercial crops (soybeans and sugar). (Dummy variables for commodities are included in all pooled regressions.) The specification differs from that in Table 10 in three ways. First, since aid variables are not available, the variable VIOLD (a political unrest variable in the earlier analysis) is included in these regressions. Second, an effort is made to estimate both an area and production and hence yield impact on research spending. Third, international trade variables were not included in these regressions. These results are generally consistent with those reported in Table 10 and show a high degree of consistency across commodities. The IARC spending impact which is of central concern to this study has a statistically significant coefficient in regressions for maize, sorghum, rice, wheat, potatoes and sweet potatoes and in the pooled cereals and staples regression. Other studies have shown that the IARC contributions in terms of technology development and research contributions have been 3 1 Table 11 Estimated Determinants of Commodity-Specific National Agricultural Research and Extension Spending,Annual Data 1963-1980, 25 Countries Dependent Variable: Spending in 1980 Dollars Independent Variables Maize Sorghum Millet Rice Wheat Cereals Beans Cassava PRODUCTION 0.000024k 0.00013** 0.00074** 0.00041** 0.00047** 0.00023** -0.00168** 0.000024CC AREA 0.000045* -0.000013 -0.00033** -0.00056** -0.00031** _o.000o19** 0.00135** 0.00008** IARCSPENDING 0.000009** 0.000022* 0.000040 0.0000067* 0.000069** 0.000016** 0.0000065 -9.517E-07 UREARICEPRICE -0.0302 -0.0503** -0.0387* -0.0259 0.2594** -o.0971** -0.0188 _0.0452** RESNEIGHBORS 0.0217** 0.0307** 0.0355** 0.0121** -0.0556** 0.0129** 0.0434** 0.0672** PROPAGRWKIRS -0.0132** -0.0124** -0.0177** -0.0599** -0.0079 -0.0286** -0.0059* -0.0035** URBANIZATION -0.0049 -0.0064k -0.0078* -0.0539** 0.0241** -0.0266** -0.00049 -0.00092 RESEXTPRICE 0.0076* -0.0109** -0.0094* 0.0361** 0.0702** 0.0173** -0.0236** 0.0016 LAND EXHAUSTION -0.3481 -0.0993 -0.2721 0.0174 0.5191 -0.1600 -0.34S5 -8'.7u14** DIVERSITY 0.6024** 0.4825** 0.7590** 1.0257** -0.2170 0.3572 0.42421* -0.063. PROD X DIVERSITY 0.000015** 0.000019** 0.000022** 0.000038** 0.000054** 0.000017** 0.000016** -0.0000013* POLVIOLENCE -696.59** -555.88** -616.31** -2016.22* -3383.09** -649.20** -548.93** -113.82 R2 0.5554 0.5904 0.6905 0.7575 0.8179 0.6859 0.7526 0.2946 F 48.94 56.46 61.12 122.32 175.95 173.45 119.15 16.36 Independent Sweet Co_ercial Variables Groundnuts Potates Potatoes St D1e Soybeans Sujcar Crops PRODUCTION 0.00014** -0.00007** 0.0000012 -0.000028** 0.000082 0.000043** _0.000014** AREA -0.000031 0.0033** 0.00011 0.00023** 0.0011** -0.0018** 0.0014** IARCSPENDING 0.000026 0.0000042** 0.000019** 0.000006** n.a. n.a. n.a. UREARICEPRICE -0.0144 0.0238** -0.0426** _0.0240** -0.0206 0.0556** 0.0231* RESNEIGHBORS 0.0358** -0.0069 -0.0637** 0.0399** 0.0218** 0.0118* 0.0182** PROPAGRWKRS -0.0052** -0.0122** -0.0030* _0.0048** -0.0107** -0.0250** -0.0185** URBANIZATION -0.0021 -0.0054** -0.0052** -0.0015 -0.0081* -0.0047 _0.0060** RESEXTPRICE 0.0052* -0.0060** 0.0019 -0.0066** -0.0179** u.0023 _0.0088** LAND EXAUSTION 0.0086 0.2074 -0.1212 -0.0386 0.6474 0.0393 0.3016 DIVERSITY 0.4257** -0.3061** 0.0131 0.0855 0.4924** 1.0267** 0.7411** PROD X DIVERSITY 0.0000057** -0.000004** 2.967E-07 0.0000059** 0.000021** 0.000023** 0.000021** POLVIOLENCE -181.44* 72.86 -39.83 -250.59** -1378.33** -547.79** -969.57** R2 0.4169 0.6297 0.1432 0.3297 0.8886 0.7037 0.8203 28.00 66.61 6.55 98.45 341.67 101.67 396.02 *T ratio betweent 1.5 and 2.0 **T ratio greater than 2.0 32 higher in these commodities (other than sweet potatoes) than in beans, cassava and groundnuts. These latter commodities are generally regarded to present "tdifficult" challenges to researchers. To some extent this is due to the fact that they have received research attention for a shorter period of time than is the case for the cereal grains, where considerable research in developed countries has been undertaken over many years. The response of national expenditures to the research by geo-climate neighbors is positive in most commodities and in the pooled regression confirming the results reported in Table 10. An increase in production holding area constant, i.e., an increase in yields, stimulates research spending in the cereal grains and cassava, but yield is not generally highly correlated with research spending. An increase in general diversity does stimulate more research spending in almost all commodities and the production impact on research spending is higher for all commodities, the higher the level of diversity. These data show relatively weak land exhaustion effects. The relative price of research to extension services is a significant determinant of spending. It shows some bias in that a decline in the costs of doing research seems to stimulate research spending on wheat, rice and maize most. Land exhaustion effects are generally not significant. The political variables ECONAG and URBANPOP show effects similar to those reported for Table 10. Urbanization appears to be biased toward stimulating more wheat research and less research on other commodities. When the price policies of countries discriminate against farmers, they also discriminate against research spending except for wheat and potatoes. Political violence is associated with reduced spending for most types of research. 33 On the whole, the results for specific field crop commodities reinforce the conclusions of the earlier analysis. They show a high level of consistency across commodities. 35 4 Policy Implications of Investment Analysis The results of the econometric exercise reported in Tables 9, 10 and 11 have substantial policy relevance. While they do show a considerable degree of consistency with rational planning on the part of national governments it cannot be concluded that there is little reason for active policy interventions to change national government investments. Indeed another large body of evidence (see Evenson, Waggoner and Ruttan, 1981 and Ruttan, 1984) shows that research investments have produced extra- ordinarily high returns in terms of the increased agricultural output associated with research programs. The implication is that the!re is general underinvestment in research. Comparisons by region and by commodity show substantial variations implying underinvestment in at least some programs of research. With this in mind then it is useful to calculate the marginal impacts of alternative policy-related activities on national research and extension spending. Table 12 reports a number of such calculations based on the regression estimates reported in Tables 10 and 11. The table shows that as commodity production increases both research and extension spending rises but at a rate less than proportional to the production increases. The "elasticities" of spending with respect to production evaluated at the mean are in the 0.55 to 0.6 range. This means that at the mean of the sample a 10 percent increase in production induces a 5.5 to 6 percent increase in spending. This is probably due to fixed costs of undertaking research and extension programs and "real" scale economies to size. The implied scale parameter is essentially the inverse of this elasticity (i.e., 1/.6 = 1.66). However, it may also reflect an overestimate of real scale economies and a tendency on the part of governments to feel that once a substantial research program is in places it need not be expanded with the importance of the crop. Table 12 Calculated Impacts on National Research and Extension Investment (Millions of 1980 Dollars) Annual Research Spending Million Dollars (from Table 10) Livestock and Field Horticulture Extension Policy Variable Crops Crops Spending 1 million $ added to (elasticity) .551 .584 .592 Commodity production (dollars) .00164 .00396 .00624 1 million $ added to commodity exports .000634 .002277 .00695 1 million $ added to commodity imports .000472 .01253 -.000937 1 added SY by geo-climate neighbor .0305 .01901 -.1792 Ten percent decline in research costs per SY on ten percent spending + std. iviation .00005 -.00064 .00188 > rice is extension costs EW -.00017 -.00042 .00145 quantity elasticity + std. deviation -1.051 -.474 -1.456 -1.191 -.652 -1.591 1 million dollars added to IARC research stock a) first year .229 1.084 .105 b) after 10 years 2.290 10.840 1.050 1 million dollars general aid research 1.194 -.858 +.047 World Bank aid (to research or extension) .285 -.063 1.468 Research Spending by Commodity (from Table 11) Ground Sweet Maize Sorghum Millets Rice Wheat Beans Cassava nuts Potatoes Potatoes 1 added SY by geo-climate neighbor .0217 .0307 .0355 .0121 -.0506 .0434 .0672 .0358 -.0069 -.0637 Ten percent decline in urea-rice price .030 .050 .039 .026 -.259 .019 .045 .015 -.024 .043 1 million dollars added to IARC investment a) first year .225 .550 1.000 .168 1.725 .162 -.000 .650 1.050 .475 b) after 10 years 2.250 5.500 10.00 1.680 17.250 1.620 -.000 6.500 10.500 4.750 37 The table also shows that when the commodity being produced is exported research spending per unit of product is 1.39 times as high for field crops and 1.54 times as high as for livestock and horticultural crops as it is for non-traded commodities. (The number for an increase in traded commodity production is the sum of the commodity production numbers and the traded commodity lines.) When the commodity is imported, spending per unit of product is 1.29 times as high for field crops and over 4 times as high for livestock and horticultural crops (where imports are generally very low). Countries are apparently placing a high premium on foreign exchange. The positive response by countries to an added SY on the commodity by a geo-climate neighbor is quantitatively significant in field crops and appears to be biased toward all cereals except wheat and toward beans, cassava and groundnuts. The induced spending of $30,000 is large in view of the fact that the cost of the added SY may be only a little more than that. The computations for a 10 percent decline in the research costs per SY has policy relevance. Many countries have options to reduce these costs through improvement of their own capacity to train scientists and through better incentive structures to hold scientists in research positions. In Africa an expansion in the indigenous scientists component and a reduction in administrative costs can easily allow a reduction in costs per scientist. A decline in the research cost by 10 percent will result in a slight increase in spending on research. This means that the increase in quantity of SYs purchased will rise by a little less than 10 percent for field crop research and by approximately 6 percent for livestock and horticultural crop research. A 10 percent decline in extension costs, on the other hand, will increase the purchase of extension workers by 14.5 to 15.9 percent and will also increase total spending. 3 8 The final calculations regarding aid and IARC spending are of most interest. The form of the model measuring IARC impacts was that the stock (i.e., cumulated expenditures in 1980 dollars) of IARC investment impacted on the annual flow of national research spending. Thus, a million dollar increment to IARC spending in 1978 would raise the value of the CINTSP variable in 1978, 1979, etc. If this IARC spending was in the field crops it would stimulate $229,000 added annual national research investment in the first year (1978). (This is calculated as the total of the spending impacts in the 24 countries in the sample. Presumably the scope of influence is wider than for these 24 countries, so this is an underestimate of the effect.) By 1988, a total of $2,290,000 added annual national research investment would have been stimulated by the 1 million dollar expenditure in 1978. With the data at hand it is not really possible to estimate the deterioration of this effect. It is conservative to suppose that it will last only 10 years (about the average time period for IARC investment in the data set). The results for individual field crops (based on Table 11 and the annual data) also show investment impacts that are generally large. Only cassava shows no impact. IARC investments of 1 million dollars in potatoes, sweet potatoes, wheat, sorghum and millets appear to stimulate an added million dollars in national spending within 1 or 2 years. Even for maize and rice the added national investment is significant. This may be compared with the estimates for direct aid. They show that 1 million dollars in general aid increases field crop research by more than 1 million dollars but at the cost of reducing spending on livestock and field crop research. Thus taking this displacement into account, only $336,000 net incremental research spending takes place for the 1 million dollar aid grant or loan. The same calculation made for World Bank aid shows an even more severe displacement effect. A million dollars in World Bank aid results in only a net increment 39 to spending of $222,000. In rather sharp contrast, it appears the World Bank extension aid has a large stimulus effect on extension spending.10 The aid inputs, it must be noted, are difficult to estimate and this will lead some policy makers to discount them. Most aid donors, however, are predisposed to believe that their aid has sufficient "strings" that it will not be displaced. Yet, most of it, in fact, is displaced and generally displacement is probably efficient. When accompanied by strong policy advice and pressure, as in the case of World Bank extension aid (the T and V system),, aid can have a large effect. It appears then that the IARC system has had significant and positive impact on national research (and extension) programs in the developing world. It has stimulated more spending in national systems and this impact is sufficiently large that an aid doncir interested in stimulating national research spending actually received more stimulus from a grant to the IARC system than from a direct grant to a national system. The IARC system has probably also had a significant impact on more qualitative aspects of national research systems as well. 41 5 Impact of Investment on Productivity A large number of studies showing relationships between agricultural productivity changes and investment in agricultural research programs in specific countries have not been undertaken (Norton and Davis, 1981 and Ruttan, 1984 provide reviews). However, in spite of the voluminous literature on the "green revolut:Lon." part of which was associated with International Agricultural Research Center (IARC) investments. little systematic study of IARC impact on productivity has been made. This is in part because the impact of an IARC is international in character. Some studies of productivity in a particular country (Evenson, 1983, for India) have inferred IARC impact on the basis of IARC-based high yielding variety (HYV) data. This, however, does not capture the full IARC impact because much of it is channeled through avenues other than HYVs and because it occurs in a number of countries. This section reports econometric estimate-s of impacts on crop productivity of national investment in crop-specific research, IARC research on the commodity, and national investment in extension. 5.1 Specification of the Productivity Relationship Since the focus of this section is on IARC effects, certain data limitations will have to be accepted. It will be necessary to pool data from several countries. Further, it will be necessary to deal with commodity-specific data since the interest is in particular IARC programs rather than in their general or average impact. This means that the only real crop-specific productivity variables which can be observed are measures of production and area harvested. In addition it is possible to measure i,rrigated area of all crops relative to all harvested area and fertilizer used. It is not really possible then to estimate a full production function or to compute a total factor productivity index by crop 42 for each country. The practical alternative options are to estimate one of the following specifications: (1) PROD/HA a+bHA=cI*+dF*+eR (2) LN(PROD) a'+b'LN(HA)+c'LN(I*)+d'LN(F*)+e'R where PROD is production in metric tons. HA is hectares harvested. I* is the ratio of irrigated area to planted area for crops that are normally irrigated. F* is the ratio of fertilizer used (valued at constant world prices) to acreage of crops normally fertilized. R is a vector of research-extension variables. These specifications are production function "proxies." The variables HA, actually has three roles in the specifications: (a) It measures productive services from land (b) It measures land expansion-contraction effects (i.e., where land quality for new planting may differ from the average land quality for the commodity) (c) It is correlated with other "left out" inputs such as labor and machine services and it may thus "pick-up" their effects. This study is not directly interested in the estimates of a', b', c', or d' (or a, b, c. and d) per .". Nor is the exact functional form of the production function an important issue since no attempt will be made to interpret coefficients as technical substitution parameters. The data available are not suited to addressing these relatively fine questions. The primary concern is with estimates of the e' vector of coefficients on the research-extension variables. Option (2) above is chosen as the more reasonable specification because left-out unmeasured inputs are likely to be proportional to cropped area (HA). The coefficient b' would, of course, not be an estimate of the marginal product of land in that case, but as noted, that is not of direct concern. The log- 43 linear relationship between the research-extension variables and production is also consistent with some evidence of research productivity. Griliches (1958) found that hybrid corn varieties tended to improve yields proportionately rather than additively. The I* and F* variables are included only for those crops that are either irrigated or fertilized. These variables are not measured on a crop-specific basis, but they are likely to be proportional to actual crop-specific variables and hence their inclusion can reduce bias. All specifications include country dummy variables. Thus "country effects" such as soil and climate factors, measurement errors, infrastructure, etc., that affect production or yield levels. but not their change over time, are picked up by these dummy variables. Specifications that pool commodities also include commodity dummy variables. Simultaneity problems may exist if national research and extension program investment responds to both production and area (i.e., to yield). A number of studies have dealt with this by simply arguing that the relationship is "recursive." That is, current research investment may respond to current yield performance, but current yields are responding to past research investments. In this study. the problem will be dealt with formally by utilizing the two stage least squares' estimates from Table 11 to construct the research variable. The actual variables specified to this study are defined as follows: 1959 (3) PRESIt = .2Rt1 +.4Rt2 +.6Rt-3 +.Ht-4 + I Rt-i i=5 where R*.t is predicted research spending in time t. The prediction is based on the investment analysis reported in Table 11.11 The weights used were indirectly estimated by constructing 44 an alternative stock using weights rising to one by year t + 9. This stock was slightly inferior to the specified stock. EXTDIV = (.tEXTt +.25EXTt-1 + .25EXTt-2)DIVER where EXTt is actual spending in 1980 dollars on all agricultural extension. DIVER = ES? where Si is the share of total production of a specific commodity in a specific geo-climate region. Livestock commodities are included in the construction of DIVER. Note that the weights for EXTDIV sum to one implying that no longterm impact from extension is realized. The full impact is realized by the end of year t+2. 1959 (4) INTRt = .2IARCt_,+.4IARCt-2+.6IARCt-3+ 8IARCt-4+ 7- IARCt_i i=5 where IARCt is spending by the IARC in 1980 dollars in time t. The following "interaction" variables were defined: EXTDIV = EXTDIV*PRESI INTRPRES = INTR*PRESI INTREXT INTR*EXTDIV One further modification was made to take into account the fact that IARC impacts are not likely to be the same in all countries in the data set. It would be, as a practical matter, nearly impossible to IARC programs to produce the same production impact in each of the 24 countries in the data set. The IARCs will in most cases be producing technology that is more closely matched to producing environments similar to its host country than to environments that are dissimilar. This should not only affect the productivity impact of the IARC program but its interaction with national research and extension programs as well. To attempt to take this into account, a variable, SR, is defined. This variable is equal to the proportion of the area 45 planted to the commodity in the country of observation that is located in the same geo-climate region as the IARCs central location. The geo-climate regions are defined by Papadakis (1965) and have been used in other studies of international productivity impact (Evenson et al., 1979, Evenson, 1983). The following variables were then defined: INTRSR = INTR*SR INTRESSR = INTRPRES*SR INTREXSR = INTREXT*SR The coefficients of these variables measure added impacts in similar geo-climate regions. The reasoning offered above would lead to the expectation that direct IARC impact via the provision of matched technology will be higher in similar regions, while the indirect impact via the provision of mismatched technology could be larger outside the similar region. It is possible, of course, that both effects will be larger in similar regions. 5.2 Productivity Impact Estimates The econometric analysis proceeded in three stages. In the first stage the predicting equations required for building the research stock variables were estimated (discussed in Part III above). In the second, crop productivity specifications were estimated for each of the 10 commodities in the study using data for all 24 countries. In the third stage, regional estimates for Asia, Africa and Latin America were obtained for maize, millets and sorghum pooled, all cereals pooled and all staple crops pooled. The results for stage two are summarized in Table 13. Table 13 reports the coefficients of the interaction terms in the model and the computed partial production elasticities for each commodity. The full regressions require over 20 tables to present. Copies are available from the author. All commodity regressions are pooled across maize, sorghum and millets, all Table 13 Estimated Crop Production Elasticities (Computed at the Mean) by Commodity,24 Countries, 1962-80 Interaction Effects Production Elasticities IARC Res X NRES IARC X NEXT National Res National Ext IARC Res NRES X Added Added Added Added Added COHMENTS NEXT_ GSR General -GSR GCeneral GCSR General CCSR General S nen &a Maize -.448(3)* -.222(5)* -.139(6) .596(5) -.743(7) -.0234* .0733* .432. .018 .136 .340** Millets .440(3)** -.197(3) -.154(4)* .349(5) -.139(2) -.065 -.019* -.067 .006 .728 .000 Sorghum -.251(2)** -.252(4)* .368(5)** -.167(3)* .212(5) 0.096* .068** -1.41 .188** 2.75** -.019* Maize, Sorghum, Millets -.109(2)** -.228(5) .416(6)** .428(5)** -.139(6) .079 .120** .197** .082** .240* .029** Rice -.336(3)** -.433(6)** .349(8) -.205(5)** .219(6)** -.102** .075** -.361** .091** .821 -.002 Wheat -.395(4) .379(6)* -.336(6)** -.986(5)** -.472(6)** .336 .271** -.622** .004 -.025 .044** Cereals -.322(3)** -.799(7) -.159(7) -.181(6)* .718(8) .050 .058** .036* .048** .189** .027** Beans .268(3) -.362(5)** .859(6)* -.170(5) .899(6) -.064** -.031* -.246 -.008* .030** .056** Cassava .195(2)** .548(5) -.776(5)** -.111(4)** -.911(6) .416 .419** -.236** -.059** .099** -.012 Groundnut -.758(3) -.823(5) .582(5)* .045* -.062 .001* Potatoes -.805(3) -.696(6) -.167(5)** .753(7) -.632(7) .141 .015** .091 .067** .054** .031** Sweet Potatoes .947(2)** -.123(3) -.385(4)** .774(5) -.525(6) -.001 .202 .232 .101* -.35** -.108** Staples .531(4) -.418(4)** .364(5)** -.598(5)** .111(5)** -.034 ** -.010 .008** .097* .073** .095** Notes: Number in parenthesis are E(-n) * "t" or comparable "F" indicates significance at the 5 to 10 percent level. ** "t" or comparable "F" indicates significance at the 5 percent or lower level. 47 cereals and all staples and show more stable and consistent elasticity estimates. It is important to bear in mind that most studies of research productivity impacts are in fact based on aggregated or pooled data. Consider first the interaction effects. The first column of Table 13 shows that national research and extension programs are substitutes in the cereals. IARC research is also a substitute for extension in rice and wheat in similar geo-climate regions. This means that spending more on extension lowers the marginal product of research and spending more on research lowers the marginal product of extension. For staples, it appears that national research complements extension in cassava and sweet potatoes where IARC research hasn't been effective. Where IARC research has been effective (as in cassava in similar regions) it tends to be a substitute for national extension. It appears that with the exception of the maize-sorghum- millets combination, IARC research has either no significant interaction with extension or it has a negative substitution interaction. The story that IARC research enhances the productivity of national extension programs is not generally told by these data. The interactions of IARC research with national research systems are also somewhat mixed. They are positive for sorghum, beans, ani staples generally and negative for wheat, cassava, potatoes and sweet potatoes. The IARC effect in similar regions is negative for maize, sorghum, rice, beans and staples generally. It is positive only for wheat. This result is consistent with the arguments regarding the matching of technology. Technology from the IARCs should be more highly matched to similar subregions and this should be manifested in lower IARC-NRES interactions in similar regions than in general. Wheat is the only case where the interaction is marginally significantly higher in similar regions. It has a strongly 48 negative extension interaction, however, where the same argument can be applied. Note that, for extension, the IARC-NEXT interaction is generally lower in similar regions. Of the 24 IARC interaction coefficients in Table 13 for similar regions, 17 are negative, and 12 are significantly negative. Only one has a marginally significant positive coefficient. These results provide general support for the underlying logic of specifications. The production elasticities are "partial" elasticities. The elasticity for national research shows the percent change in production associated with a 1 percent change in the national research stock, holding national extension, IARC research and other variables in the equation constant. These elasticities are functions of the levels of other variables because of the interaction terms in the equations. They are evaluated at the mean of the data set. An 'IF" test is undertaken to test for the joint statistical significance of the coefficients entering the marginal production (and the computed elasticity). The elasticities are computed for countries outside similar regions and the incremental elasticity for similar regions is also shown.12 The IARC elasticities are computed on a presumption that IARC impacts will be realized in all 24 countries in the sample.13 The elasticities bear a relationship to rates of return on investment. Suppose that a country is presently spending one half of 1 percent of the value of product on cereals research. The elasticity estimate for cereals, 0.058, indicates that production will increase by 0.058 percent or 0.00058 times the value of production. Thus an investment in time t of 1 percent of the value of product will generate an income stream that will be zero in time t, 0.2* x 0.00058V in t+1, 0.4* x 0.00058V in t+2, 0.6* x 0.00058V in t43, 0.8* x 0.00058V in t+4, and 0.00058V 49 in all years thereafter.14 The discount rate which equates this earnings stream to the initial investment is approximately 35 percent. This is the internal rate of return to the research investment. Had the initial ratio of research spending been only 0.0025 instead of 0.005 the earnings stream associated with an elasticity of 0.058 would have yielded an internal rate of return slightly over 60 percent. The ratios of research spending to the value of product for the 1972-79 period by commodity were: wheat 0.0051, rice 0.0025, maize-sorghum-millets 0.0023, cassava 0.0011. beans 0.0032, potatoes 0.0029, sweet potatoes 0.0007 and groundnuts 0.0025. Table 14 shows the conversion of elasticities for both research and extension to internal rates of return for different ratios of spending to value of product. The low income countries in the sample had a ratio of extension spending to value of product 0.005. For the higher income countries it was 0.0075. With these conversions, the reader can see that national research investment has yielded generally high returns. National extension investment, as the table shows, must have above 0.059 to yield a return of 10 percent, under an assumption that its impact does not last beyond three periods.15 Extension impacts on cereal grain productivity and on potatoes and sweet potatoes productivity appear to be large enough to justify investment at the lower levels. Given the nature of the variable used, perhaps the most reasonable estimate is for the pooled cereal grains. This elasticity is sufficient to justify around one half of 1 percent on extension. Many countries# however, are currently spending roughly 1 percent of the value of product on extension. The estimate for cereal grains does not justify an investment of this magnitude.16 The estimates for both national research and extension should be interpreted with some caution. The productivity and effectiveness of both research and extension programs varies from Table 14 Internal Rates of Return Corresponding to Given Research and Extension Elasticities at Selected Ratios of Spending to Productivity Comparable Extension Comparable Research Elasticity Elasticity Internal Ratio of Spending Ratio of Spending Rate of to Productivity to Productivity Return .0003 .0025 .005 .01 .005 .0075 10% .0006 .005 .010 .0200 .059 .088 20% .0015 .0122 .0243 .0468 .068 .102 n 30% .0025 .0212 .0421 .0841 .077 .116 ° 40% .0043 .0353 .0766 .1412 .087 .131 50% .0051 .0416 .0851 .1702 .096 .145 60% .0066 .0547 .1094 .2188 .106 .159 70% .0081 .0675 .1350 .2700 .116 .174 80% .0113 .0808 .1616 .3230 .126 .189 100% .0131 .1088 .2175 .4350 .146 .219 51 country to country because of organization, leadership and general political and economic conditions. Studies in specific countries are required to investigate these issues further. The chief reason for resorting to international data in this study is that IARC impacts are international in character and cannot easily be measured in data for a single country. The production elasticities for IARC investment for the pooled maize-millets-sorghum data and for pooled cereals show that IARC investment has an elasticity of 0.027 for the developing world in general and a considerably higher elasticity for countries in similar regions. This impact is essentially the "green revolution" impact. It implies a very high rate of return because the ratio of IARC spending to the value of the product is low, ranging from 0.0003 for the cereals to 0.0008 for potatoes. Thus an elasticity of 0.017 implies an internal rate of return of 100 percent. These high rates of return are. of course, based on the fact that the IARC impact occurs not just in one country but in the entire region. Because the spending to product ratios are low, these high returns imply that substantial growth in productivity is produced by the IARCs. If IARC spending would have been 20 percent higher for cereal grains and had the same elasticities held (a questionable assumption), production of cereal grains would have been 0.027 x 0.2 = 0.0054 or one half percent higher per year (after the full impact is realized). This is a large growth increment from a relatively small investment. The results for IARC investment in rice are a little puzzling as they show very high returns in similar regions and none outside these regions. It also appears that IARC investment in rice has sharply reduced the marginal products of national research and extension in similar regions. The definitions of regions for rice may be a little too broad to capture the same effects as for other commodities. 52 For the staple crops. it appears that there is an IARC impact in all commodities except sweet potatoes. For cassava, the impact is confined to similar regions. For beans and potatoes, the impact extends beyond similar regions. The returns to this IARC research appear to be as high as for IARC research in cereal grains. Given the very high leverage factor with IARC research almost any measurable impact (in a statistical sense) will tend to have a high rate of return. The commodity-based results in Table 13 show that pooled commodity regressions tend to be more systematic than individual commodity regressions. Table 14 reports regional-based regressions for three pooled groups - maize-sorghum-millets. cereals and staples. All pooled regressions include commodity and country dummy variables. Table 15 does not include the similar region variables because the grouping of countries into the three broad regions achieves some of the same objectives. This table reveals patterns somewhat more clearly than did Table 13. The negative national research-extension interactions, for example, emerge for every region and every commodity group. The IARC-national research interaction is negative for cereal crops in Asia and Latin America, but is actually positive for staple crops in Latin America. The IARC-national extension interactions are generally positive, except in staple crops in Latin America. The estimated productivity elasticities are also somewhat more regular. National research investments are highly productive, except in Africa for cereal grains (presumably rice and wheat) and Latin America for staples. Implied rates of return are high. They range from 30 to 40 percent for maize in Latin America and maize and staple crops in Africa to 60 to 70 percent for maize and cereals in Latin America, cereals in Asia and staple crops in Asia. Table 15 Regional Impact Analysis Maize, Millets & Sorghum Cereal Crops Staple Crops Research- Extension Latin Latin Latin Coefficient America Africa Asia America Africa Asia America Africa Asia PRESSI .0121** .0393** .0314** .0146** .854(3) .0106** -.019** .0733** .0479** EXTDIV .0331** -.609(4) .0305** .0158** -.153(3) .0389** -.493(2) .939(2)** .0157* EXTDPRES -.117(2)** -.939(3)** -.172(2)** -.364(3)** -.228(3) -.597(3)** .318(3)** -.101(2)* -.457(2)** INTRIARC .286(5) .809(5) .213(6) .560(5)** .319(5) .171(5) .237(4)** .371(5) .514(5) INTRPRES -.179(6) .445(6) -.103(5)** -.193(6)** .157(7) -.644(7)** .685(6)* -.228(5) .105(5) Lnl INTRXEXT .129(5)** .178(6) .349(5)** .501(7) .222(6)** .755(6)* -.737(6)* .653(6) .188(5) ( PRODUCTIVITY ELASTICITIES National Research .0344 .0505** .1168** .1435** -.0060 .1135** -.0302** .0313** .1292** National Extension .1l08* -.0129 .165B** .0745*k .0128 .1921** -.0243** ** .1198** .0685 IARC Research .0317* .0355** .0416** .0298** .0543** .0428** .0412** .0187 .0312* Note: Numbers in parentheses are E(-n). "T" or comparable "F" indicate significance at 5 to 10 percent levels. "T" or comparable "F" indicate significance at 5 percent or lower level. 54 National investment in extension programs also generally appears to be productive, except in staples in Latin America and maize in Africa. The elasticities are high enough to justify a spending to value ratio of 1/2 to 1 percent but not much higher. IARC investment is productive across the board. The elasticities for cereal crops are highest in Africa and lowest in Latin America. The reverse is true for staples. The elasticities imply high internal rates of return to IARC investment, generally in excess of 100 percent everywhere. As a region, Asia does best with high productivity elasticities for all three forms of investment for all commodities. Latin America has benefited from all investments except in staples. Africa has mixed results. IARC investment has been least productive in staples. National investment has been most productive in the staple crops. 55 6 Policy Implications of Productivity Analysis This paper shows, as do many others, that research directed toward the discovery and development of new agricultural technology has a high payoff in terms of productivity growth. Not all research programs are successful, of course. In some cases, relatively new research programs may not be productive until a significant period of trial-and-error with scientific approaches and administrative and organizational change takes place. Most IARC programs are still quite young. Previous studies have documented high productivity of IARC research programs in wheat and rice, but relatively little systematic study of impact on other commodities has been undertaken. The chief objective of this study was to use international crop productivity data to measure IARC impacts in ten commodities. Certain data limitations had to be accepted in doing so and this study is not a substitute for more detailed country studies. Nonetheless* the study did identify and measure significant IARC impacts as well as national research and extension impacts on crop productivity. In addition it identified several interaction and regional impacts of interest. (The study also attempted to deal with the simultaneous relation- ship between productivity and research and extension investment). The major findings were: (1) Measurable positive IARC impacts on crop productivity were observed for all commodities except sweet potatoes. For pooled commodity groups, grains, cereals and staples, positive IARC impacts were measured for all groups in all regions. Computed rates of return to IARC investment are very high. (2) IARC impacts are higher in countries in the same geo-climate region as the IARC central location. In most commodities these IARC impacts lower the marginal product of both national research and national extension programs. The 56 IARCs produce technology that to some extent substitutes for the products of national research and extension. (3) Outside similar geo-climate regions, IARC impacts complement national research programs in some commodities, (maize, rice, beans) and substitute for others. (4) National research investment is highly productive in most commodities and in most regions. Internal rates of return to investment range from 30 to 70 percent for most commodities. (5) National research has a consistent negative interaction with national extension. Higher research spending reduces the impact of extension services. It appears that most extension services are not organized to directly channel or diffuse research products to farmers. (6) Extension services are also generally productive although their impacts are much more variable. Rates of return calculations show that few programs have been productive enough to justify extension spending-to-product ratios above 1 percent. The first part of this study examined the impact of IARC investment on national research investment. It concluded that IARC investment stimulated national research investment in most commodities, and concluded that the stimulus was probably because IARC research made national research more productive. The negative IARC-national research interaction terms for some commodities in this study raise some further questions on the issue. It should be noted, however, that the negative interaction term is estimated at the margin and may not hold for the average relationship between IARC and national spending. Further, it may be noted that IARC impact can stimulate national research productivity by making longer-term contributions that are not necessarily picked up in these data. The IARCs do produce technology that is a close substitute for some of the technology 57 produced in national programs. They also produce technology and pre-technology science that complements and thus stimulates national program work. The strongest IARC stimulation impacts occur in wheat, potatoes, millets and groundnuts. These commodities also have the weakest negative IARC-national research interaction terms. The policy questions to which these data speak are whether to expand the IARC system, whether to continue expansion and development of national research systems and whether to continue development of national extension programs. The maintenance and expansion of the IARC system itself is determined by international entrepreneurs and by donor country attitudes. This is in contrast to national spending on research and extension which is subject to national economic and political forces. The signals from this study are quite clear and quite strong. Investment in IARCs beyond the 1976-80 level is likely to be highly productive.17 A donor agency interested in getting the maximum increment of food supply in the developing world from a given aid grant will obtain it by investing more in an IARC. This study shows that IARC impacts on crop productivity are probably higher than are national research program impacts. Furthermore, investment in IARCs stimulates more national system investment than will a comparable amount of direct aid. These estimates of high productivity impact do not mean that all IARCs are optimally organized. What they do tell us is that the IARC concept is a good one. The IARCs have filled a vacuum, so to speak, and in their early years most have done so productively. The vacuum was the absence of strong science-based national research programs. It is now clear that national programs have made great progress, part of it due to IARCs. But a good deal more investment and institutional development is required before these systems will effectively substitute for the IARCs. 58 The signals from this study regarding national research system investment are also quite clear. In spite of variation in organization, skill levels and other characteristics, most national system programs are productive. Returns to investment are high. Most estimated elasticities are sufficiently high that they imply high returns to investment even if they are over- estimated by a factor of 2 or 3. A blanket recommendation that all national systems should be expanded without regard to their existing organization and structure is not justified by these data. However, an expansion of well-organized systems is called for and the data clearly show the potential for high payoff national system investments in all countries in the developing world. Finally, the signals regarding extension investment, while generally positive, do call for caution. While it was assumed that extension does not produce a long-term income stream, it is of course, possible that some permanent gains are due to extension. This possibility was not investigated in this study. There is a minimum productivity impact below which large investments in extension cannot justify extension spending to produce value ratios of much more than 1 percent of the value of agricultural product. Perhaps the more serious issue regarding extension, however, is the lack of evidence that extension complements research. The strong negative interaction terms between research and extension suggest that extension productivity is based, not so much on extending research results but on more general productivity improving effects through improving farm management. There is nothing wrong with this, but this finding suggests that more systematic study of the research-extension link is called for. 59 Notes 1 Judd, Boyce and Evenson, 1983 provide details. Appendix Tables 1 and 2 to this paper summarize changes in national system development. 2 The definition of country groups is that used by The World Bank in its World Development Report 1984. 3 See Table 7 for a list of the countries; for the analysis, Taiwan is excluded. 4 Diversity is measured at the country level. It is defined as n DIVER SR i=1 where Si is the share in total agricultural product of the ith crop geo-climate combination. 5 Many studies show that while consumers are the major gainers from agricultural research, they are not strong supporters of research (See Binswanger, in press, and Rose-Ackerman and Evenson, 1985). 6 The variable BASIC does not necessarily measure "basic" research. Non-commodity oriented research can include farming systems and economic research. 7 The CINTSP variable is a naturally exogenous variable since IARC spending is undertaken in a specific location and thus cannot respond to country specific conditions. It can, of course, respond to commodity conditions. 8 Note that this is not the area of the crop on which the research observation is made, but the area of all crops. 9 Note that d(PQ) = dP(0) + dQ(P) = Q + dQ (P) dP dPQ - ;Q= PD + Q1 + n 10 The World Bank is a relative late-comer to the research and extension support field. It provided very little support prior to 1974. Its lending since then for research and extension has been: Research Extension 1974-76 $227.5 million $ 314.4 million 1977-80 $271.9 million $1,033.0 million 1981-84 $890.0 million $ 740.5 million 60 As can be seen, the Bank became a major factor in extension support after 1977 and a major factor in research after 1980. 11 The weights in (3) were "estimated" by comparing the residual squared error of the equation with an alternative to (3) where the weights rose to one at tg instead of t5. Specification (3) was slightly superior. 12 The elasticity for similar regions is the sum of the two elasticities. 13 This is actually an underestimate of the elasticity since the coefficient estimates may apply to all developing countries, not just to the 24 countries in the sample. However, excluding the Peoples Republic of China, the 24 countries in the sample account for more than 85 percent of crop production in the developing world. 14 Note that this presumes that spending occurs at the beginning of year t and productivity doesn't appear until the end of the year. Thus one full year is added to the implicit time lags built into the specification. A 6-month lag could have been used. This calculation is thus conservative. 15 No attempt to test whether the impact lasts beyond three periods was made. However, had a different time configuration been built into the extension specification, its coefficient and its elasticity would have changed. The rate of return would probably not have changed very much. 16 Caution in interpreting extension results from international data is warranted. Even if these estimates are unbiased, they represent an average impact from programs varying greatly in quality. Well-managed extension programs with skilled extension workers will have an impact higher than this average estimate indicates. 17 This is the case even though the IARCs are relatively high cost institutions. Expenditures per scientist-year are 2 to 3 times those of national systems because of international salary levels and more elaborate technical support (See Judd, Boyce and Evenson, 1983). Appendix Table 1: Agricultural Research Expenditures and Worker Years,by Region (A Constructed Time Series, 1959-1980) WESTERN EUROPE Expenditures (000 Constant 1980 US $) Worker Years (Number) Country 1959 1962 1963 1968 1971 1974 1977 1980 1959 1962 1965 1968 1971 1974 1977 1980 Denmark 4,797 9,310 15,504 26,741 24,889 24,835 28,308 32,267 170 200 300 500 530 560 638 727 Finland 3,949 5,360 6,976 8,089 8,664 11,080 14,935 17,803 136 152 165 180 215 242 326 389 Iceland 493 559 960 754 1,064 1,298 1,583 1,422 22 20 19 19 27 35 44 47 Ireland 3,949 11,284 16,612 19,047 24,654 26,171 25,956 44,824 130 250 310 350 422 490 486 300 Norway 12,696 11,989 17,262 19,829 22,776 26,744 32,122 37,511 260 280 300 308 393 480 577 674 Sweden 6,769 14,104 20,763 26,091 29,350 28,655 34,180 40,205 100 120 170 205 250 300 359 422 U.K. 62,065 70,527 78,902 106,973 141,350 152,827 166,005 235.495 1,000 1,300 1,850 2,578 2,840 3,310 5,551 5,468 Northern Europe 94,718 123,134 156,980 207,523 252,747 271,610 303,089 409,527 1,818 2,322 3,114 4,140 4,677 5,417 7,981 8,027 Austria 3,949 3,949 5,814 8,349 10,331 8,979 10,978 13,415 80 90 100 105 110 110 134 164 Belgium 12,696 14,104 14,866 18,552 19,488 29,228 30,599 35,709 260 550 650 650 650 800 838 978 France 22,569 49,369 96,897 203,511 187,840 201,541 179,770 221,590 440 720 850 1,086 1,130 1,240 1,868 2,191 H Germany 59,242 141,056 193,797 234,819 234,800 229,240 242,763 252,044 1,300 1,700 2,100 2,500 2,750 3,000 3,177 3,298 Netherlands 36,659 56,422 76,688 70,445 79,832 106,980 220,106 277,762 638 720 820 900 981 1,100 1,538 1,724 Switzerland 5,924 8,180 10,796 23,482 35,220 48,714 55,892 70,713 170 210 250 285 295 325 373 472 Central Europe 141,054 273,082 398,859 559,157 567,511 624,682 740,108 871,233 2,888 3,990 4,770 5,526 5,916 6,575 7,928 8,827 Greece 7,899 7,927 9,413 8,871 9,392 9,362 11,809 12,683 195 212 280 280 325 390 492 528 Italy 22,569 28,211 33,222 46,965 76,310 84,054 59,668 106,988 600 900 1,091 1,025 1,099 1,200 1,218 636 Portugal 4,231 7,053 8,305 11,740 18,784 19,103 19,427 19,757 300 300 350 400 450 500 372 378 S?ain 4,513 9,310 13,841 31,308 46,960 53,490 60,928 69,400 450 550 580 615 640 670 1,004 1,144 Southern Europe 39,212 52,501 64,781 98,884 151,446 166,009 151,832 208,828 1,545 1,962 2,301 2,320 2,514 2,760 3,086 2,686 Regional Total 274,984 448,717 620,621 865,564 971,704 1,062,301 1,195,029 1,489,58 _6,251 8,274 10,185 11.986 13.107 14.752 18.995 19,540 Appendix Table 1: continued LATIN AMERICA Expenditures (000 constant 1980 US$)2/ Worker Years (Number) Coantry 1959 1962 1965 1968 1971 1974 1977 1980 1959 1962 1965 1968 1971 1974 1977 1980 Argentina 28,211 32,442 48,991 41,968 42,978 70,441 53,490 59,750 320 420 670 7650 880 880 890 1,065 Cnile 1,693 2,963 6,229 6,915 16,436 10,315 11,960 11,319 32 58 113 162 171 192 171 177 Paraguay 423 564 554 730 775 1,146 2,529 5,357 5 10 15 20 26 31 48 63 Uruguay 761 1,411 1,385 2,087 2,348 3,437 3,399 3.821 7 35 35 60 75 100 180 222 iamperate South America 31,088 37,380 57,159 51,700 62,537 85,339 71,378 80,247 364 523 833 892 1,152 1,203 1,289 1,527 nolivia 507 669 693 653 587 427 6,459 11,374 20 29 40 50 60 86 66 125 Brazil 11,284 22,569 41,527 60,008 70,440 114,620 130,735 174,012 200 400 800 1,350 1,650 2,000 3,121 2,935 Coluobia 14,104 13,428 17,746 25,464 30,806 31,329 29,668 32,231 200 338 300 550 809 870 824 881 Ecuador 704 1,411 2,768 4,226 5,260 8,901 8,132 6,100 12 20 34 64 94 200 183 208 buyana 348 519 814 1,198 1,380 1,851 1,601 2,678 6 10 15 23 29 36 27 41 Peru 1,073 2,104 4,154 8,479 11,740 12,895 6,871 8,163 32 65 131 155 180 220 295 290 Venezuela 6,772 11,193 13,677 19,829 17,845 15,283 34,509 34,885 100 176 184 155 226 354 329 360 Tropical South America 34,792 51,893 81,379 119,857 138,058 185,306 217,975 269,443 570 1,038 1,504 2,347 3,048 3,766 4,865 4,840 Barbados 172 103 244 295 449 593 514 652 3 4 5 7 8 10 13 23 Costa Rica 775 930 1,108 1,043 2,747 2,374 1,935 2,168 40 48 59 55 61 71 60 75 El Salvador 1,186 1,186 1,108 1,174 1,409 1,815 2,507 2,391 50 50 48 60 83 85 78 78 Guatemala 862 1,025 1,218 1,474 2,247 2,963 4,083 5,332 19 22 27 43 47 58 71 123 NI Baiti 86 103 122 147 225 296 356 452 8 9 11 18 20 24 33 37 nrnduras 1,129 1,411 1,660 1,827 1,878 1,719 831 1,047 35 43 51 60 67 72 66 60 Jamaica 172 205 244 295 1,132 1,810 1,639 935 15 17 20 32 55 88 85 40 Aexico 5,079 5,924 6,922 8,871 14,558 22,924 20,393 70,929 190 220 280 520 540 711 1,074 1,079 Sicaragua 451 803 1,385 1,827 1,878 1,719 1,711 2,211 8 10 17 22 29 34 44 57 Panama 345 410 487 589 899 1,185 1,515 2,482 11 13 16 25 44 49 29 51 irinidad & Tobago 172 205 244 295 449 593 832 709 10 11 14 22 23 29 39 40 ,ominican Rep. 690 820 975 1,179 1,798 2,370 3,486 2,514 10 11 14 22 23 29 40 40 Caribbean a Central1I Aerica- 13,676 16,144 19,332 23,390 36,493 49,644 '48,956 112,941 491 564 691 1,090 1,230 1,552 2,015 2,167 aegional Total 79 .556 _05417_ 157.860 194.947 237.088 320,289 338,309 462.631 1.425 2,125 3.028 4,329 5.430 6,521 8.169 8,534 1/ Includes adjustment for missiag countries based on estimates (1% of subtotals): CATIE (IICA) 4 Cuba 19 T3 2! Our method for converting currencies to constant U.S. dollars tends to yield lower expenditure levels for the 1970's and 1980 for couutries wnicn nave experienzed high rates of inflation (e.g., Mexico, Brazil, Argentina) than the expenditure levels reparted by others (see Oram and 3indlish, 1981). Appendix Table 1: continued NORTH AMERICA AND OCEANIA Expenditures (000 Constant 1980 US$) Worker Years (Number) 1959 1962 1965 1968 1971 1974 1977 1980 1959 1962 1965 1968 1971 1974 1977 1980 Aub :rO.iia 76,169 95,918 156,421 169,591 281,760 267,447 286,823 306,199 1,500 1,700 1,900 2,130 3,000 3,200 2,425 2,589 X,e. Zland 14,952 16,927 27,131 29,484 44,612 68,773 73,713 78,683 250 250 450 475 590 700 707 713 bc a/ 91.577 .. 113,408 186.553 200,071 328.004 337,901 362,339 386,806 1.759 1,960 2,362 2,618 3,608 3,919 3,132 3,302 0u4,664 108,614 129,013 211,336 234,800 229,240 277,925 241,246 950 1,050 1,150 1,300 1,450 1,520 1,820 1,836 Unia. StaEes 564,224 648.858 808,409 939,275 1,056,600 1,050,683 1,072,880 1,094,338 5,740 6,150 6,570 7,000 7,400 7,500 8,303 8,469 Nor:T. .,erica 6o8,889 757,472 937,423 1.150,612 1,291,400 1,279.923 1,350,805 1.335,584 6,690 7,200 7,720 8,300 8,850 9,020 10,123 10,305 iab._.al Io ral 703.460 870880 I i2o 9 1.350 652 13619404 1.617 i 13 390 87449 99,160 10,082 10,918 12.458 12.939 13,255 13,607 1i _-_ludes adjustment for miissing countries based on estimates: 0.5% of subtotals Appendix Table 1: continued EASTERN EUROPE AND USSR Expenditures (000 Constant 1980 US$) Worker Years (Number) Country 1959 1962 1965 1968 1971 1974 1977-80 1959 1962 1965 1968 1971 1974 1977-80 Bulgaria 11,284 13,823 15,781 27,657 37,568 38,019 38,264 250 300 350 650 981 960 966 Czechoslovakia 101,560 115,242 122,645 130,194 129,140 143,115 162,458 1,470 1,770 2,070 4,015 3,150 4,100 4,654 Hungary 5,642 7,335 33,222 67,836 61,048 69,050 67,737 400 500 1,500 1,560 1,420 1,500 1,471 Poland 22,569 39,496 57,308 69,923 77,484 93,263 95,233 1,240 2,170 3,210 4,100 4,700 5,150 5,259 Romania 19,747 33,853 49,834 61,053 68,092 82,560 95,398 650 850 1,285 1,900 2,500 3,200 3,698 Yugoslavia 14,104 14,248 14,396 20.611 28,176 34,386 35,017 1,080 1,100 1,140 1,720 1,890 1,970 2,006 Eastern Europe- 195,896 250,877 328,368 422,546 449,642 508,032 553,400 5,701 7,493 10,702 15,618 16,400 18,906 20,220 USSR 372.388 688,354 744.595 781,682 gin.554 997.900 939,383 12,000 20,400 24,450 25,600 29.800 33.350 31.,394 0. Regional Total 568.284 939.231 1.072.963 1.204.228 1.360.196 1.505.932 1.492.783 J7.701 27.893 35.152 41.218 46.2Qo 52.256 51 6 1/ Inzludes adjustment for missing countries based on estimates (% oc subtotals): D.D.R. 11% Albania 1% 12% Appendix Table 1: continued AFRICA Expenditures (000 Constant 1980 US$) Worker Years (Numher) Country 1959 1962 1965 1968 1971 1974 1977 1980 1959 1962 1965 1968 1971 1974 1977 1980 Morocco 2,116 3,386 4,154 5,217 6,574 6,112 8,633 8,026 17 25 34 43 60 65 543 686 Sudan 2,820 4,513 6,922 7,828 9,580 8,213 11,388 13,600 50 50 50 55 128 140 144 150 Egypr 11,284 16,363 21,040 24,785 24,067 21,015 22,325 23,717 400 500 600 700 750 800 850 903 Tunisia 1,204 1,800 2,383 2,807 2,428 3,874 5,320 6,764 35 44 53 62 72 140 212 285 Libya 973 1,456 1,927 2,270 2,413 2,125 2,541 2,793 39 58 77 91 97 80 112 123 North Africa- 20.789 31.095 41,161 48,485 50,920 46,713 56,734 62,037 590 738 887 1,037 1,?07 1,335 2.028 2,340 Cameroon 564 1,129 1,684 2,374 3,052 3,437 3,364 3,788 10 15 30 48 72 96 94 106 Chad 282 564 831 1,043 1,174 1,146 1,369 1,602 7 10 16 23 26 30 36 42 Danomey 564 1,129 1,799 1,956 2,043 1,719 2,053 2,403 7 10 13 15 16 16 16 19 Gambia 23 28 42 47 52 47 56 66 5 5 6 6 5 5 6 7 Gabon 68 80 89 124 141 239 285 334 4 4 6 5 5 5 6 6 Ghana 3,386 4,513 5,537 6,262 6,574 5,731 12,443 12,655 60 80 100 120 140 304 301 352 ON L-I Ivory Coast 5,642 8,462 11,073 13,045 14,088 12,036 12,399 12,771 40 60 80 100 110 110 113 116 Liberia 94 94 141 211 282 282 360 394 14 16 25 18 14 16 18 20 Mali 845 1,411 1,738 2,869 3,992 4,393 5,246 6,141 12 15 21 16 25 35 47 68 Mauritania 115 151 207 223 259 258 402 284 3 4 6 6 7 7 11 8 Nigeria 14,104 22,569 33,222 31,308 37,568 38,207 147,429 121,840 110 170 170 195 300 300 843 1,084 Senegal 3,668 4,513 4,982 6,001 6,574 7,640 8,369 9,726 45 55 55 85 130 160 148 172 Sierra Leone 282 394 444 470 564 573 687 698 16 22 28 23 30 36 34 35 Upper Volta 451 507 636 730 740 669 1,087 1,105 5 6 7 10 10 11 12 12 Zaire 8,462 4,797 6,922 7,828 8,230 8,608 5.949 5.095 20 25 35 30 66 85 113 97 West Africa-2 44,333 57,892 79,750 85,664 98,133 97.733 231,723 205,737 412 572 676 805 1,099 1,398 2.068 2,466 Appendix Table 1: continued AFRICA (continued) Expenditures (000 Constant 1980 US$) Worker Years (Number) Country 1959 1962 1965 1968 1971 1974 1977 1980 1959 1962 1965 1968 1971 1974 1977 1980 Burundi 282 423 721 1,043 1,017 958 3,332 3,608 10 13 16 20 24 27 22 41 Ethiopia 845 1,411 2,214 3,372 3,412 3,437 3,370 3,400 8 12 25 30 52 65 110 155 Kenya 1,411 1,975 3,322 5,480 7,748 13,492 19,844 22,712 25 40 70 140 210 280 299 400 Madagascar 2,256 5,079 6,091 6,915 6,409 6,125 5,309 4,878 25 40 50 65 70 80 76 68 Malawi 704 1,129 1,660 2,087 2,818 3,437 4,641 5,660 15 22 35 44 57 208 242 276 Mauritius 1,411 2,116 2,768 4,567 5,870 6,208 7,450 7,879 25 30 35 41 51 61 46 50 Rwanda 564 648 664 653 859 763 894 945 9 10 8 8 16 18 23 24 Tanzania 1,552 2,116 2,768 5,480 7,748 9,933 7,436 7,214 45 60 65 60 100 145 194 212 Uganda 1,411 2,116 3,322 5,480 7,748 6,687 5,804 7,452 20 32 40 52 80 80 135 175 Zambia 1,252 2>042 2,824 4,209 7,394 7,176 5,575 5,202 21 31 41 55 81 79 104 96 East Africa-/ 12,740 20,770 28,726 42,822 55,615 63,455 69,384 75,156 221 316 420 561 808 1,137 1,364 1,632 Botswana 42 141 444 521 775 629 2,803 4,977 1 2 2 10 33 30 46 61 LesoLho 28 70 110 209 303 324 429 465 1 2 3 3 7 10 13 14 Zimbabwe 1,411 1,411 2,076 3,783 5,119 7,640 7,467 10,560 140 100 134 135 172 180 155 201 a, South Africa 39,496 46,422 77,519 62,619 46,960 47,758 63,441 64,519 550 720 900 900 900 1,000 1,328 1,351 Swaziland 310 437 512 521 695 669 1,357 1,306 4 6 9 11 12 12 24 23 South Africa 41,287 58,482 80,661 67,653 53,852 57,020 75,497 81,827 696 830 1,049 1,059 1,124 1,232 1,566 1,650 Regional Total 119.149 168,239 230,298 244.624 258.520 264.921 433.338 424.757 1,919 2,456 3.032 3.462 4.236 5.102 7.026 8.088 Notes: 1/ North Africa totals adjusted for missing countries (% of subtotals): Expenditures Manpower Algeria 13 9 2/ West Africa totals adjusted for missing countries (E- of subtotals): Angola 4 CAR 2 Congo 3 Guinea 2 Niger 2 Benin 1 Guinea- Bissau 1 15 31 East Africa totals adjusted for missing countries (X of subtotals): Mozambique 7 Somalia 2 9 Appendix Table 1: continued ASIA Expenditures (000 constant 1980 US$) Worker Years (Number) Coun_ ry 1959 1962 1965 1968 1971 1974 1977 1980 1959 1962 1965 1968 1971 1974 1977 1980 Cyprus 423 704 831 1,005 915 944 1,599 2,410 15 i8 20 24 37 54 56 58 Iran 4,231 7,448 12,458 16,699 23,480 34,386 39,840 45,163 55 110 250 360 550 580 457 518 Israel 11,566 14,104 16,335 19,568 18,314 22,578 25,558 30,209 170 220 270 327 440 500 566 630 Jordan 128 175 243 339 427 852 869 849 6 8 14 17 23 40 30 35 lurkey 4,797 6,206 9,690 16,960 21,367 22,924 24,640 26,463 150 200 397 440 485 540 580 623 Syria 282 704 1,219 2,219 2,700 3,057 4,045 4,963 5 10 15 40 75 110 145 179 ,estAsil1/ 24,427 33,449 46,485 64,741 76,611 96,605 110,068 125,465 457 645 1,101 1,377 1,835 2,079 2_OQ_ 2a29 dangladesh - - - - 2,348 2,677 15,735 27,613 - - - 150 190 1,234 1,320 Sri Lanka 3,104 3,940 4,982 6,286 6,340 5,731 4,244 5,057 50 65 80 95 105 130 287 422 N~epal 906 1,109 1,337 1,519 2,163 2,229 2,556 2,634 71 87 104 119 169 184 224 226 India 24,825 29,622 41,020 45,717 66,108 66,868 103,855 120,167 1,150 1,160 1,450 1,800 1,950 2,150 2,244 2,345 PaListan 2,256 3,386 4,982 5,741 4,696 4,776 38,528 29,899 120 180 270 350 250 280 1%60 1 212 South Asia- 32.024 39.199 53,891 61,041 84,105 84,749 169,866 190,931 1,433 1,537 1,961 2,435 2,703 3,022 j,711 5,691 indonesia 564 2,256 4,705 6,783 8,688 8,023 42,229 33,200 15 70 140 240 340 592 914 1,473 ON Malaysia 3,386 5,924 9,136 9,653 11,740 11,463 19,564 30,391 40 30 150 156 195 149 284 386 - ?nilippines 2,781 3,633 4,255 4,877 5,499 6,844 8,637 9,533 200 300 400 500 600 620 630 640 Thailand 1,552 4,231 7,476 9,652 11,740 11,463 23,547 21,600 150 250 350 475 600 725 ,134 1,264 Souchla"t Asia- 9,028 17,488 27,873 33,752 41,057 41,194 102,435 103,249 441 774 1,135 1,494 1,891 2,274 5,229 4,102 Caina 54,166 169,265 332,223 469,638 535,344 623,434 633,420 643,555 1,250 4,000 8,000 11,000 13,500 16,000 1' 000 17.272 nong Kong 141 183 195 195 200 190 118 132 9 8 8 8 10 12 8 8 Japan 135,414 197,479 334,992 420,064 575,260 611,306 645,543 684,276 7,200 8,500 10,000 11,500 13,700 14,000 14,784 15,671 Soutn Korea 2,538 2,820 3,322 4,567 23,381 24,400 26,607 29,012 300 320 340 450 744 807 880 960 laiwan 1,975 3,245 3,877 4,539 5,400 5,539 12,520 14,000 250 273 310 350 375 400 404 452 Last Asia-/ 141,469 205,765 345,809 433,659 610,283 647,849 691,636 734,694 7,837 9,194 10,765 12,431 15,008 15,371 16,237 17,262 kegional Total NJlL. 46'i Lhh 8O6 981 1 062 831 1 347.400 1493 831 1 707425 1 797 894 11.418 16.150 22.962 28.737 34.937 38.746 44 277 4h6 Notes: 1/ West Asia totals adjusted for missing countries based on estimates (% of subtotals): Iraq 2, Lebanon 6, Others 6. 2/ South Asia totals adjusted for missing countries based on estimates (% of subtotals): Afghanistan 2, Others 1. 3/ Southeast Asia totals adjusted for missing countries basee on estimate (5% of subtotals). Missing countries: Burma, Cambodia, Laos, Portuguese Timor, Singapore, Vietnam. 4/ East Asia Totals adjusted for missing countries based on estimates (1% of subtotals), Missing countries: Mongolia, North Korea. Appendix Table 2: Agricultural Extension Expenditures and Worker Years, by Region (A Constructed Time Series, 1959-1980) LATIN AMERICA Expenditures (000 Constant 1980 US $) WXki,r Years (-nmh-a) Average Country 1959 1962 1965 1968 1971 1974 1977-80 1959 1962 1965 1968 1971 1974 1977 1980 Argentina 4,513 5,642 22,149 31,308 39,212 23,879 37,412 100 165 260 286 350 360 359 359 Cnile 564 618 2,768 5,217 8,922 6,495 10,176 80 91 500 500 800 649 748 847 Paraguay 101 183 387 547 564 498 780 5 10 20 30 36 42 83 124 Eruguay 564 1,129 1,108 1,305 1,409 1,338 2.096 20 30 40 50 60 70 102 133 Temperate S.A. 5,741 7,572 26,413 38,378 50,106 32,210 50,464 205 296 820 866 1,246 1,121 1,292 1,463 Bolivia 282 615 387 653 383 249 1,370 48 60 73 84 81 70 87 120 Drazil 22,851 33,008 42,358 81,403 129,610 179,570 285,039 1,688 1,916 2,196 4,275 6,972 12,600 11,641 14,428 Colombia 5,924 6,318 6,424 4,696 7,514 7,640 12,593 140 161 224 287 350 425 515 609 icuador 564 1,723 1,583 2,087 2,348 2,292 3,778 50 115 130 145 160 270 327 387 Peru 845 3,104 9,303 5,011 5,870 5,922 9,491 80 252 420 600 780 960 1,152 1,344 Venezuela 16X363 16,589 15,863 15.889 15,027 12,799 21,097 340 355 450 622 675 735 901 1.067 Tropical S.A.-/ 47,296 61,971 76,676 110,835 163,051 210,557 336.702 2,369 2.888 3.528 6,073 9,108 15.211 14,769 18,135 Costa Rica 902 902 1,007 1,305 3,005 2,254 3,531 40 40 38 59 104 105 155 205 El Salvador 479 564 674 679 704 1,146 1,795 36 55 81 91 106 140 212 283 donduras 394 423 369 664 751 859 1,346 35 40 40 50 63 75 164 253 ',exico 2,538 3,668 6,368 8,871 10,589 19,103 29,929 200 250 220 460 800 1,300 1,843 2,115 :;icaragua 507 704 1,038 888 939 763 1,195 16 24 32 28 30 30 43 49 Jamaica 72 94 142 186 240 362 464 126 158 159 266 426 723 949 957 Caribbean- 8.414 10.931 16.509 __21.660 27,912 42.118 65.807 779 975 980 1,641 2,630 4,082 5,790 6,643 kegional Total 61,451 80,474 119,598 170,873 241,069 284,885 452,97_3 h 1 4,159 5,328 8,580 12,984 20,414 21,851 26,241 1/ Includes adjustment for missing countries (plus 1%) I cludes adjustment for missing countries based on estimates (% of subtotals): Barbados 2 Cuba 25 Guatemala 10 Haiti 19 Panama 9 Trinidad & Tobago 5 Other 2 72 Appendix Table 2: continued WESTERN EUROPE Expenditures (000 Constant 1980 US$) Worker Years (Number) Average _ountrv 1959 1962 1965 1968 1971 1974 1977-80 1959 1962 1965 1968 1971. 1974 1977 1980 ianmark 15,516 16,927 17,995 20,873 21,132 19,103 22,340 742 788 790 945 947 949 951 954 Finland 12,414 16,081 20,460 18,786 18,784 24,835 26,720 670 750 861 825 750 743 685 634 Iceland 704 845 970 939 939 956 1,192 42 41 42 42 43 44 47 51 Ireland 5,079 6,488 8,305 9,392 13,384 14,137 17,309 345 385 436 465 504 540 551 578 Norway 11,848 12,977 13,841 14,351 12,679 10,125 15,047 666 678 650 645 640 640 815 989 Sweden 13,823 14,952 15,781 15,915 15,262 13,182 16,584 740 800 844 852 817 705 760 815 U.K. 53.600 67,707 80.288 99,147 112,704 93,603 114,886 1,588 1,693 1,650 1,700 2,100 2,300 2,419 2,554 jorthern Europe 112,983 135,977 157,640 179.403 194,884 175,943 213,078 4,793 5.130 5,273 5,474 5,801 5,921 6,228 6,575 Austria 11,284 14,104 16,612 18,260 18,784 17,192 22,619 726 700 700 680 650 620 699 777 delgium 1,242 1,552 1,827 2,010 2,066 1,911 2,773 345 398 340 284 280 275 342 409 France 23,132 28,702 83,056 75,664 65,744 70,874 139,796 2,460 3,668 4,400 5,200 5,700 6,300 6,530 6,790 Germany 49,369 57,834 63,675 62,098 61,048 53,490 57,698 2,936 4,400 4,400 4,500 4,812 5,100 4,714 4,874 O Netherlands 15,234 23,980 31,839 37,821 41,090 39,352 27,800 1,228 1,598 1,500 1,500 1,500 1,250 1,446 1,471 Switzerland 2,820 3.808 4,705 6,391 7,396 7,640 9,336 170 270 370 480 505 530 555 582 Central Europe 103,082 146,417 201,714 202,254 196,128 190,460 260,022 7,865 11,034 11,710 12,644 13,447 14,075 14,286 14,903 Greece 3,668 4,034 4,318 4,226 3,569 3,344 3,933 330 440 400 480 839 900 907 913 Italy 11,284 19,747 29,071 37,831 37,568 33,431 42,046 2,000 2,500 2,500 3,050 3,250 3,500 3,772 4,042 Portugal 845 5,642 10,244 10,697 10,566 9,552 12,009 500 650 692 850 970 1,100 1,185 1,270 Spain 2,153 8,462 13,841 18,263 19,958 18,148 23,932 500 500 700 920 1,050 1,200 1,356 1,512 Southern Europe 17,950 37,885 57,474 71,018 71,661 64,474 81,920 3,330 3,590 4,292 5,300 6,109 6,700 7,220 7,737 Regional Total .016 320.279 416.829 452.676 462673 430.877 556.020 15.988 19.759 21.275 23.418 25.357 26.69 27.734 29.215 Appendix: Table 2: continued NORTH AMERICA AND OCEANIA Expenditures (000 Constant 1980 US$) Worker Years (Number) Average 1959 1962 1965 1968 1971 1974 1977-80 1999 1962 1965 1968 197A 1974 1977-80 Country Australia 30,576 50,780 55,371 62,619 93,920 95,517 113,478 1,700 1,750 1,800 2,000 2,250 2,300 2,400 New Zealand 7,899 8,462 9,136 10,958 11,740 16,239 19,296 370 375 375 400 450 450 300 Oceania- 50,466 59,538 64.828 73,946 106,188 112,314 132,774 2,080 2,136 2,186 2,412 2,713 2,g64 2,714 Canada 50,780 56,422 69,212 78,273 84,528 85,965 102,140 1,500 1,500 1,750 2,000 2,100 2,200 2,200 U.S. 282,112 310,323 332,223 391,365 469,600 477,583 567,388 10,000 10,000 10,200 10,400 10,600 10,800 9,653 Norch America 332,892 366,746 401,435 469,638 554,128 563,548 669,528 11,500 11,500 11,950 12,400 12,700 13,000 11,853 Regional Total 383,358 426,284 466,263 543.583 660.316 675,862 802,302 13,580 13,636 14,136 14.812 15,413 15,764 14,567 1/ totals adjusted for missing countries based on estimates of 0.5% of subtotals. 73 References Boyce, James K., and Robert E. Evenson (1975), "National and International Agricultural Research and Extension Programs," Agricultural Development Council. Commonwealth Agricultural Bureau, Abstracts-computerized data base. Evenson, Robert E. (1982), "Observations on Brazilian Agricultural Research and Productivity," Revista De Economa Rural, Vol. 20, No. 3, July/September. (1983), "Economics of Agricultural Growth: The Case of Northern India," 1§es la Third World Developmejnt, K. C. Nobe and R. K. Sampath (eds.). Westview Press, Boulder. Evenson, Robert E., Waggoner, P. and Ruttan, V. W. (1979), "Economic Benefits from Research: An Example from Agriculture." Science, 205. September. Food and Agriculture Organization of the U.N., Yearbook fQ Production, annual. Griliches, Z. (1958), "Research Costs and Social Returns: Hybrid Corn and Related Innovation," Journal Qf Political Ecofnomy 66:419-31. Jamison, D. T., and Lau, L. J. (1982), Farm Edo and Frjn Efficisncy, Baltimore, Johns Hopkins University Press. Judd, M. Ann, Boyce, James K., and Evenson, Robert E. (1983), "Investing in Agricultural Supply," E.G.C. Discussion Paper 442, Yale University, Economic Growth Center. Kislev, Yoav, and Evenson, Robert E. (1975), "Investment in Agricultural Research and Extension, An International Survey," Economic Development Cultural Chanae, April 23. Norton, G. R., and Davis, Jeffrey S. (1981), "Evaluating Returns to Research," American Journal .2 Agricultural Economics, Vol. 63, No. 4, November. Oram, Peter and Bindlish, Vishva (1981), "Resource Allocation to Agricultural Research: Trends in the 1970s (A Review of Third World Systems)," Washington, D.C., International Food Policy Research Institute. Papadakis, J. (1965), Agricultural Climates gf the Q Buenos Aires. Ruttan, Vernon W. (1984), Aaricultural Research Poligy, Minneapolis, University of Minnesota Press. World Bank (1983), World Development Report, New York, oxford University Press. The International Agricultural Research Centers zz a, o0t-