\ I" l' W I " I 0' [at C UNIVERSITY LIBRARIES I III IIIIII I4 IIIIIIIIII IIIIIIIIIIIII IL- 3 1293 0062 LIBRARY Michigan State L University This is to certify that the dissertation entitled Determinants of Rural Incomes in Communal Areas of Zimbabwe: Household Food Security Implications presented by Charles John Chopak has been accepted towards fulfillment of the requirements for Ph.D. Agricultural Economics degree in Raw II 59,me Major professor Date 16 Max 199]. MS U i: an Affirmative Action/Equal Opportunity Institution 0-12771 PLACE IN RETURN BOX to remove We checkout from your record. TO AVOID FINES return on or before date due. DATE DUE DATE DUE DATE DUE MSU Is An Affirmdive Action/Equal Opportunlry lnetitutlon em”! DETERMINANTS OF RURAL INCOHES IN COINUNAL AREAS OF ZIMBABWE: HOUSEHOLD FOOD SECURITY IMPLICATIONS BY Charles John Chopak A DISSERTATION Submitted to Michigan State University in partial fulfillment of the requirements for the degree of DOCTOR OF PHILOSOPHY Department of Agricultural Economics 1991 .0. tce‘u‘: .g. :; .‘fi “...-.~. ‘ .......,. I 13.33:: z: 2‘. :1 rxa‘ 523.1339! :: *4 1325;! l: «"134: .— 1133?- State a I n N . I ‘z 2.5““ 1' . a. . «$22.3 3.ch '33:. r! ,- we 5'50 4,. u “ ' ‘~ '3! a L. 3...; ‘C:ei ~ in: ' I522 . I ."up \3“ ‘ ‘ {u "u ‘ r.~ 1 II‘ ‘3» ‘ ‘ :‘ 3.13». ABSTRACT DETERNIIAITS OF RURAL INCODIES IN COIDIUNAL AREAS OF ZIMBABWE: SOUS'OLD FOOD SECURITY IIIPLICATIONS BY Charles John Chopak At least 100 million people live in absolute poverty in Sub-Saharan Africa. Although Zimbabwe is a grain surplus nation, a large portion of their population is food insecure. Furthermore, a lack of reliable data about the rural population has made it difficult for government to design policies to expand economic opportunities for the rural poor. This thesis analyzes the structure, level, and determinants of incomes in low rainfall areas of Zimbabwe to suggest alternative development strategies to expand income-earning opportunities for poor, rural households. The data were collected in twelve villages in Natural 39910118 IV (Hutoko and Mudzi Districts) and v (Buhera District) during the 1933/89 agricultural season, using a three-stage stratified-random Imple procedure. Household incomes were higher in Mutoko and Mudzi than in Buhera. nth‘mgh the distribution of incomes was highly unequal across districts, it was more unequal in Buhera. Households access to land, labor, and capital was greater in Buhera than in HutokO/Mudzi. Although the distribution of land and labor was mini-“1y equal across districts, oxen ownership was highly unequal. The environmental milieu was more favorable in MutokO/Huduv where rajJ‘flll was substantially higher and less variable than in 31111933- Althwgh government has made major investments to strengthen rural ml, at last? ‘..‘ A 'v 0"." ..~e ”L ,..R - 1.32.5 :1 :: fluff! m 24:59:32: .2 Lot: wists. ..... M‘s, gcl;:'.es 11. Lin; I services, the survey villages have benefitted minimally. Inter-household variability in total and agricultural income was largely due to differential access to physical resources. In contrast, labor characteristics determined whether households participated in local labor markets. Finally, policies that have driven Zimbabwe's agricultural revolution have had minimal (or negative) impact on resource-poor households. While government has limited ability to increase the agricultural productivity of the resource-poor households in the short run, government can help the rural poor by expanding food distribution schemes, public employment schemes, and human capital development. In the longer run, new technologies are required to reduce environment- related production risk, including soil and water conservation, crop improvement, and small-scale irrigation. Yet, to assist resource-poor households, future rural development programs must both increase agricultural productivity and expanding access to land (land resettlement), social services, and rural employment opportunities. Copyright by CIIARLES JOHN CHOPAK 1991 I;a:ef*.ily m1:- Emeritus: pcss It? treaties, Hm 1ft: '4: 5:212:91 “imbue an: a 3! :.-:s:a:.:';.-.; . ;.-".!:, mint! [:9 P “I Res...” .2: - -U\eb...". e 51‘1“- i...” nelea' ... ..‘ . . ’ to: ”'4; I gm: :5 tat" :‘ eh. h..‘- . '“e' 3;. Z ‘3: .A N‘“ r. to .r r. ‘s‘e 4.3.7, Ed R: ; s.“ ACKNOWLEDGEMENTS I gratefully acknowledge the contributions of all those who have made this dissertation possible. My sincere appreciation goes to the members of my comittee, Mike Weber, Carl Eicher, Eric Crawford, and James Shaffer, who provided numerous insights and invaluable research support both in Zimbabwe and at Michigan State University. The outstanding support and assistance of my advisor, Richard Bernsten, warrants special recognition and special thanks. His friendly advice and continual encouragement throughout my entire graduate program (especially during my field work in Zimbabwe), and his stamina in reviewing a great many drafts were instrumental to the successful completion of this dissertation. I wish to thank the Department of Agricultural Economics and Extension at the University of Zimbabwe, who provided office space, computer access, and logistical support. Specifically, I would like to thank Godfrey Mudimu, Jones Govereh, and Solomon Chigume, for their collaboration and team spirit during the field work. The enumerators--l'ungai Matikiti, Justin Chindunrey, 2vikomurero Mutize, Nancy Nyakuda, Peter Mutapaire, Jeremiah Katsande, Josamu ““Q‘dzuws, Evelyn Ziso, Henry Mandokota, Phineas Jori, Jekmore ”WWII: and Lazarus Mupinda--not only implemented the surveys, but ‘1“ Provided interesting insights about the study area and its people. smut. this work would not have been possible without the cooperation. ho‘pit‘ntY: and patience of the participating families and local leaders. I would also like to thank all those people at Michi9an State "“1““1111. Janet Munn, Christina DeFouw, Chris Wolf, Margaret 39““ ' 1:5 m; m f2: .2 :1 kgversm' 2:21;?! I;:es:e:: 211.2111 Dew-exam Lily, I ieixate emu: iii-d :1. ; :w ' . we" 13:3 {Lin 1: .- H‘- v p . ““92“ mt ua3, j .3" ' I ' / v _l_‘—'4 — Jeff Wilson, and Sherry Rich, who not only solved many problems, but did so, so pleasantly. Many thanks also to the Department of Agricultural Economics for its financial support during the course of my studies at )flSU; This study was funded by Michigan State University in cooperation with the University of Zimbabwe under a Food Security Research cooperative Agreement financed by the United States Agency for International Development (USAID) . Irinally, I dedicate this work to my wonderful family, especially my wife Danielle and my parents Marty and John, for all the years of love, support, and faith in me. Honey, Mom, Dad, this is for you! SI _ i .1. 1' ".- . cc... 9—} [gen - .q ‘0'... \ I U I d. u .0 a 53 3.12285 N) “I K h I (-0 “'N “.0081” a “I I, . . (A: K! 00 ’1 8) .‘V . A! K'Q . I. r) “I I) O‘O-‘ stew-«s- K. N N W, N N o a e c as K, K) N a ‘ K, K, r. ‘u, ~..‘F‘ 1.. . .1 'V 1....»wa 1.: ,‘x! Na K! ks () l “I D 3 A. _. c . TABLE OF CONTENTS PAGE LI ST OF TABLES O O O O O O O O O O O O O O O O O O O O O O O O O O O O O O O O O O O O O O O O O O O O O O O x 1 LI ST OF P Isms O O O O O O O O O O O O O O O O O O O O O O O O O O O O O O O O O O O O O O O O O O O O O O O x iv “BmIAT IONS O O O O O O O O O O O O O O O O O O O O O O O O O O O O O O O O O O O O O O O O O O O O O O O O xv I. Introduction ........................................... 1 1.1 Who are the world's poor? ....................... 1 1.2 Problem statement ..................... .......... 1 1.2.1 Zimbabwe's development objectives .......... 1 1.2.2 Constraints in achieving the development objectives ................................ 4 1.2.3 Zimbabwe's agricultural sector and food security ................................. 5 1.2.4 Food security equation .................... 6 1.3 Research objectives .............................. 7 1.4 Research hypotheses .............................. 7 1.5 Organization of dissertation ..................... 11 II. Literature review and theoretical considerations ........ 13 2.1 Rural poverty ................................... 13 2.1.1 Definition of poverty ...................... 13 2.1.1.1 Scope: general versus specific ..... 13 2.1.1.2 Measurement ........................ 15 2.1.1.3 Temporal dimension: chronic versus transitory ......................... 15 2.1.2 Incidence ................................. 16 2.1.3 Determinants of poverty .................... 18 2.2 Incomes and expenditures ......................... 21 2.2.1 Theoretical and methodological concepts .... 21 2.2.1.1 Income definitions ................. 21 2.2.1.2 Income distribution ................ 22 2.2.1.2.1 Alternative measures of central tendency ......... 23 2.2.1.2.2 Alternative measures of income equality .......... 24 2.2.1.3 Significance testing ............... 29 2.2.1.4 Monetization ....................... 29 2.2.2 Past income and expenditure studies ........ 30 2020201 world OOOOOOOOOOOOOOOOOOOOOOOOOOOOOO 31 202020221mbam OOOOOOOOOOOOOOOOOOOOOOOOOOO 32 2.3 Zimbabwe's communal sector ....................... 33 2.3.1 Characteristics of communal households ..... 33 2.3.1.1 Geographic dispersion .............. 33 2.3.1.2 Agricultural production ............ 35 2.3.1.3 Consumption preferences ............ 36 2.3.2 Rural household access to government services (since 1980) ............................... 37 vii I-‘ A 3 35:; mm...) " :z::ad::: 3.25;:wey If 34.2 .. 3“I3 U LI. I The :e‘ . 5:39.“: " 4.1.; r:_ L: “I “5 “-?|;| '1 (O In ..I u.:. 0‘. ‘ - (A! A KID. & O m “I e f, f) O CHAPTER . PACE III. Survey methodology ....................... ..... . ......... 41 3.1 Introduction ..................................... 41 3.2 Survey area ...................................... 41 3.2.1 Research area selection criteria ........... 41 3.2.2 Research location .......................... 44 3.3 Sampling procedures .............................. 44 3.3.1 Selection of households .............. ..... . 47 3.3.2 Sample size ................................ 47 3.4 Data collection .................................. 48 3.4.1 Enumeration ................................ 48 3.4.1.1 Selection of enumerators ........... 48 3.4.1.2 Responsibilities ................... 50 3.4.1.3 Training ........................... 50 3.4.2 Questionnaire design ....................... 50 3.4.3 Data entry ................................. 51 3.4.4 Survey instruments ......................... 51 3.5 Limitations of the data .......................... 52 IV. Income and expenditures: level, distribution, and composition .......................................... '54 4.1 The definition, structure, and distribution of household incomes and expenditures ............... 54 4.1.1 Household income and expenditure concepts .. 54 4.1.1.1 Description of net household receipts categories ......................... 54 4.1.1.2 Expenditures ....................... 60 4.1.2 Structure of household incomes and expend- itures ..................................... 64 4.1.3 Distribution of net household receipts ..... 64 4.2 Analysis of net household receipts ............... 64 4.2.1 Net household receipts levels .............. 64 4.2.2 Distribution of net household receipts ..... 68 4.2.2.1 Income quartile distribution ....... 68 4.2.2.2 Symmetry of net household receipts.. 70 4.2.2.3 Equality of net household receipts.. 70 4.2.3 Sources of net household receipts .......... 74 4.2.3.1 Structure of net household receipts, by villages and districts .......... 74 4.2.3.2 Structure of net household receipts, by income quartiles ................ 78 4.2.3.3 Structure of earned income, by income quartile ........................... 80 4.2.4 Monetization of households ................. 83 4.3 Analysis of household expenditures: empirical results ........................................... 85 4.3.1 Household expenditure levels ............... 85 4.3.2 Household expenditure pattern .............. 87 4.3.2.1 Expenditures by village and district 87 4.3.2.2 Expenditures by net household receipts quartiles ................. 89 4.4 Summary ........................................... 90 "~ Resource endowment and external environment ............. 94 5.1 Household resource definitions and measures of distribution ..................................... 94 5.1.1 Definitions ................................ 94 5.1.2 Measures of resource distribution .......... 96 5.2 Overview of household resource availability ...... 96 viii 5.3.2 Iqal'.‘ Slime-:11 9:: anti." .. S.I.‘. ' ‘cr 54.2 2.4:: . 54.3 Ca;.:a :='.A . en - an. cos-fl 0"!- ::'00 er . O-A' .- ....e..e.a..~ 533;?23273“ O‘. ’ e e 3. .: 79:1": ' ' :' 'w ill... . D e £22.35», .r .,. ' I seven" ' !. Oven O 5.1.. ‘Q .. oen‘ :r:e:‘ E s . 'O-s. p 6...;I "‘ ve:e__‘.:a-e: e‘n: kite; 4 I- 1 t. A 1. 09 I p R ' g 0! 0‘ h, w 1...: r' (‘1 w :9 e a e e 1.: On) '1 K! 5' f] ‘2§IE’°’-=.d . be.e:: c‘ . 1.1:. c II: 5" .." Ni: . 7: 9:43 Col 5...:r‘ p 7.3 2 :93? O £ng ‘ , I .3,‘ e ' I CBAPIBR V Continued PRO! 5.3 Distribution of household resources .............. 99 5.3.1 Symmetry of resource ownership ............. 99 5.3.2 Equality of resource ownership ............. 101 5.4 Resource endowment by net household receipts quartiles ........................................ 101 5.4.1 Labor ...................................... 101 5.4.2 Land ..................... ..... .. ..... ...... 105 5.4.3 Capital .................................... 110 5.5 Income level and sources by resource endowment ... 116 5.6 Interrelationships between resource endowment and socio-economic characteristics ................... 123 5.7 External environment ............................. 129 5.7.1 Physical environment ....................... 129 5.7.2 Institutional environment ....... ..... ...... 131 5.7.2.1 Access to services ................. 131 5.7.2.2 Changes in access to services since 1980 ............................... 135 5.7.3 Technology ................................. 135 5.8 Summary .......................................... 137 VI. Determinants of inter-household variation of incomes .... 143 6.1 Regression variables ............................. 143 6.1.1 Dependent variables ..... .......... ... ...... 144 6.1.2 Independent variables ...................... 144 6.1.2.1 Endogenous factors (to households).. 145 6.1.2.2 Exogenous factors (to households)... 148 6.2 Determinants of net household receipts ...... ..... 149 6.2.1 Model specification ........................ 149 6.2.2 Results of the model ....................... 150 6.2.2.1 Satisfaction of the assumptions .... 150 6.2.2.1 Results of the net household receipts regression model .................. 152 6.3 Determinants of net household receipt components . 156 6.3.1 Determinants of agricultural production (value) .................................... 156 6.3.1.1 Model specification ................ 156 6.3.1.2 Results of the regression model .... 158 6.3.2 Determinants of labor sales ................ 161 6.3.2.1 Model specification ................ 161 6.3.2.2 Results of the regression model .... 162 6.3.3 Determinants of transfers (received) ....... 165 6.3.3.1 Model specification ................ 165 6.3.3.2 Results of the regression model .... 166 6.4 Summary .......................................... 169 ‘EII- Rural development strategies ............................ 171 7.1 Household typology ............................... 172 7.2 Effect of current policies and services on rural income 00.00.000.000.....OOOOOOOOOOOOOOO0.0I..0... 174 7.3 Strategies to accelerate rural development ....... 179 7.3.1 Short and medium term strategies to help reflcurce‘poor hOUBBhOIdB 0.0000000000000000. 181 7.3.2 Long term rural development strategies ..... 184 7.3.2.1 Agricultural diversification ....... 184 7.3.2.1.1 Technology generation .... 184 7.3.2.1.2 Technology diffusion ..... 190 7.3.2.2 Rural development programs ......... 193. 704 smuy ......OOOOOOO0.0.........OOOOOOOOOOOO0.00. 195 ix I .1:- La: Ca; we Le- .4 I?! A-u' uto- ' . du‘u‘ ... .C'.C.':C.l‘ 53"“‘9 $13!." SJ. .’ e 2.4 in 13 V Castinnd . .. .. ,1» a. e l. .. e a“ a -. my a nu ca s .14 .4 I e no .6 c J o. w. o . n a“ 4"... a a “W a o ‘4 cu ... a. 6 II.- .. II n u. "I ~ I." I" a... ...—Itx... . 7’ L .‘ I ‘70 "...,H‘ . I. a .1. sud .. etc :3 O» \4& ~40 ‘3 6 Rd 6 6 RR 6 6 WI e O a. 6.1 .. 6.‘ M Zetenma: 5-2.1 .I.:«. i2 Fe 6. 6 C v CHAPTER V Continued PAGE 5.3 Distribution of household resources .............. 99 5.3.1 Symmetry of resource ownership ............. 99 5.3.2 Equality of resource ownership ............. 101 5.4 Resource endowment by net household receipts quartiles ........................................ 101 5.4.1 Labor ............................... ....... 101 5.4.2 Land .............................. ......... 105 5.4.3 Capital .................................... 110 5.5 Income level and sources by resource endowment ... 116 5.6 Interrelationships between resource endowment and socio-economic characteristics ...... ....... ...... 123 5.7 External environment ............................. 129 5.7.1 Physical environment ....................... 129 5.7.2 Institutional environment .................. 131 5.7.2.1 Access to services ................. 131 5.7.2.2 Changes in access to services since 1980 .OOOOOOOOOOOOOOOOOOO0.0.0.0.... 135 5.7.3 Technology ................ ...... . ...... .... 135 5.8 Summary ........................ ........ . ......... 137 VI. Determinants of inter-household variation of incomes .... 143 6.1 Regression variables .................. ........... 143 6.1.1 Dependent variables .................. ...... 144 6.1.2 Independent variables ...................... 144 6.1.2.1 Endogenous factors (to households).. 145 6.1.2.2 Exogenous factors (to households)... 148 6.2 Determinants of net household receipts ........... 149 6.2.1 Model specification ........................ 149 6.2.2 Results of the model ....................... 150 6.2.2.1 Satisfaction of the assumptions .... 150 6.2.2.1 Results of the net household receipts regression model .................. 152 6.3 Determinants of net household receipt components . 156 6.3.1 Determinants of agricultural production (value) .......................... ...... .... 156 6.3.1.1 Model specification ................ 156 6.3.1.2 Results of the regression model .... 158 6.3.2 Determinants of labor sales ................ 161 6.3.2.1 Model specification ................ 161 6.3.2.2 Results of the regression model .... 162 6.3.3 Determinants of transfers (received) ....... 165 6.3.3.1 Model specification ................ 165 6.3.3.2 Results of the regression model .... 166 6.4 Summary .......................................... 169 \EII, Rural development strategies ............................ 171 7.1 Household typology ............................... 172 7.2 Effect of current policies and services on rural income OOOOOOOOOOOOOOOOOOOOOO00.000000000000000.0. 174 7.3 Strategies to accelerate rural development ....... 179 7.3.1 Short and medium term strategies to help resource-poor households ................... 181 7.3.2 Long term rural development strategies ..... 184 7.3.2.1 Agricultural diversification ....... 184 7.3.2.1.1 Technology generation .... 184 7.3.2.1.2 Technology diffusion ..... 190 7.3.2.2 Rural development programs ......... 193. 7.4 Summary .......................................... 195 ix 8.1.111vel E. H. Or ... U. 0.! 5.1.3 Bauer 5?" ' 50.94 a»r¢a‘ 0 ill a u VIII. Summary and needed research ............................. 8.1 Summary of findings .............................. 8 1 1 Household resource endowment 1 1 S Determinants of incomes ..... 8.1.6 Effect of current policies and services .... 1.1 Level and sources of rural incomes ......... 2 Distribution of rural incomes .............. :3 .4 Distribution of resource ownership ......... 8.2 Policy adjustments ............................... 8.3 Limitations of the results ....................... APPENDIX 1 Schedule of research activities APPENDIX 2 survey mules ......OOOOOOOOOOO ...... .00.... APPENDIX 3 Analysis of inventory levels ........... BIBLImeY ......OOOOOOOOOOOOOO OOOOOOOOOO ......OOOOOOOOO x 211 212 214 216 .1 Prcpertms 5:14."; :f {save . 57- e. '9 vague... “;e’::‘.fio V‘s 5241:? cf ::e “’a‘. a... Q I..’ 5". Gov». ‘~‘_‘ v. ."‘e' “ 'k.‘ ‘sts . I I “we. .. ...v e‘.' I I ’9’- L'ZC; ser ‘1 '\ h "st-~31 in... w.“"~‘ A ..-“ $53.}?! 33:59:: .4 ed .. . lag 'Lr ‘ q . . _ " I. i~E‘. ‘2‘: My '1 Villa: 3-2 3-3 4‘9 4‘10 LIST OF TABLES Poverty in the developing world, 1985 and 2000 ..... Properties of alternative income equality measures . Summary of past income and expenditure studies in Zimbabwe 000.000.0000....0000000.000.000.00... ...... Distribution of research sites with respect to selection criteria and district, 1988-89, Zimbabwe . Summary of reasons why households were dropped from the final ‘ample 000.000.000.00000000000.00.0.0.000. Distribution of the sample of households for the University of Zimbabwe Food Security research prOject' Zimbabwe, 1988-89 0.00.00.00.00000000.00... Net household receipts by village, district, and total ample (ZS) 00.0.0.0..00..0.0.0.00.0000.000.0. Distribution of households (percent) among net household receipts quartiles by village and district Measures of central tendency and symmetry for net household receipts (per capita) by village, district, and ampl‘ 00000000000000.00000.0.00.0...000.000.... Distribution of net household receipts (per capita) by village, district, and sample ................... Components of net household receipts (per capita) by source, village, and district ...................... Percent contribution to earned income (per capita) by village, district, and sample ...................... Structure of net household receipts (per capita) by income quartile .................................... Structure of earned income (per capita) by income quartile 0000..000000.00000000000...0.000...00...... Disaggregation of potential farm sales (crops and livestock) by income quartile (including inventories) Cash versus noncash income by not household receipts quartiles 000.000.00...00..0000.0.000......00000000. Expenditures by village, district, and sample ...... xi PAGE 18 26 34 45 49 49 66 '69 71 72 75 76 79 81 82 84 86 ".I' - 5’. l A 5:23-30 .. “I 33121?” t‘“ Cancun-.3: < 313.1! A‘ I 07mm .. . 2:313:33: 5:.- ::'..‘. La: Ems h:'..‘. :e. .115 churn 212d use by : Li“: :se :1; : Cyan; one: 14;;11'. one: lame Love; zitfl lgve‘. :33“ 1931‘. 325933 9;: 33:11:: .. Mam en: 3830mm 9:: 3mm. pa: km: to “ Cm." an r I u..- ‘ .QC....C . 3;? . Reg:38".0r_ c 4-12 4-13 5-1 5-3 5-4 5-6 5-7 5-8 5-9 5-10 5-11 5-12 5-13 5-14 5-15 5-16 5-17 5-18 5-19 6-1 Structure of expenditures (per capita) by village and district, mdtOta1Iample 0.000.000.0000..00.000... Composition of expenditures (per capita) by income quartile 000.000.0000...00..0.00.000.00.00.000000000 Overview of household resource endowment by village (median), district (mean), and total sample (mean) . Distribution of household resources ................ Household labor characteristics by income quartile . Household head characteristics by income quartile .. Land characteristics by income quartile ............ Land use by district and total sample ......... ..... Land use by income quartile ........................ Capital ownership by district and total sample ..... Capital ownership by income quartile ............... Income level and sources by labor availability ..... Income level and sources by land ownership and use . Income level and sources by capital ownership ...... Resource endowment by household labor charac— teriatiC. 0.000.000.000.0000.0.0.000000.00.00.00.000 Resource endowment by land use ..................... Resource endowment by capital ownership ............ Rainfall pattern (millimeters) by district ......... Access to services by village ...................... Changes in rural access to services since 1980 ..... Technology awareness and adoption .................. Regression coefficients and test statistics for the econometric model examining inter-household variation of net household receipts .......................... Regression coefficients and test statistics for the econometric model examining inter-household variation of the value of agricultural production ............ Regression coefficients and test statistics for the econometric model examining inter-household variation Of 1‘”: 'ales 000000.0000000000.0000...00000.00000. xii PAGE 88 91 97 100 102 104 106 108 109 112 113 117 119 122 125 126 128 131 132 136 138 153 159 164 iegresm: c:é 9:3; in: 2" cf transfers : Scan of h;:. acts: hone? Smry effev museum :vr- 6-4 7-1 7-2 Regression coefficients and test statistics for the econometric model examining inter-household variation of transfers received by households ................ Summary of household assets and performance measures across household types ................. ..... . ...... Summary effects of current government policies across househOId tms 0.000.000...0.0000000000000000.00..0 xiii PAGE 168 173 175 F ‘ ’mvs‘u-enuwm" 2!! Loans: 3! :xpcrzents 2 Cayman c 0 ‘ ‘ .36 28.12.31 . 33.3th 4 .. wed es. A_M:W — LIST OF FIGURES Location of the survey areas ....................... Components of total annual net household receipts .. Components of total annual expenditures ............ The relationship between incomes, expenditures, and household food consumption ......................... xiv PAGE 46 56 61 65 5:2; 0" - .'V~'.. ~oa. eon-e... e I R'«~.‘. .va‘ Ileobiguie el 8 I k:‘.‘:.;.t:.'a. . sz-e-n'e...‘ ' teobflee'e e R 5:13: lane: n '4 .:.. 5:2: ;e p . .313. Stat. I i" Planet. a vex.” :: "‘ U. I! t H a 6!:- rammed—mm 1.-.,“ .. “"eev. H Rd. 131.1 A;:. Gm: burnt. ‘ '(lflhrflE-glflklkltflL LIL ..Oe'. .‘A“ One an.‘..~..a V! ‘ ‘ 9| P a... 'F ‘ £FW‘e5 C. tn.‘ L.‘ 5A....e'. " we... ... .. 3. e . . ~‘ . E“. " “e. “.e.': . WIMIONB Organisations AFC Agricultural Finance Corporation AGRITEX Agricultural, Technical and Extension Services AMA Agricultural Marketing Authority ARDA Agricultural and Rural Development Authority CMB Cotton Marketing Board CSC Cold Storage Commission 080 Central Statistics Office DMB Dairy Marketing Board DR&SS Department of Research and Specialist Services ENDA Environment and Development Agency FAO Food and Agricultural Organization (U.N.) GMB Grain Marketing Authority IFPRI International Food Policy Research Institute ROZ Republic of Zimbabwe SADCC Southern Africa Development Coordination Conference SEDCO Small Enterprise Development Corporation Technical terns W AEE Expenditures per adult equivalent CIGA Cash income generating activities cx Consumption expenditures BC Home consumption 168 Intermediate goods INV Investment INVNT Inventory (ending) NCR Net credit receipts NHI Net household income NHR Net household receipts NR Natural Region PCB Per capita expenditures 93C Production for home consumption TAB Total annual expenditures 1'31 Transfer received (in) T30 Transfers given (out) W (8:31. Coefficient of variation Standard deviation of the natural log of income XV ne .W" [cm-"- Hectares . . Leagues Q £4.22!!! 0". A’. v.05 V" !‘_E]:|u \ 32mm HA RCALS HM MT Hectares Kilocalories Millimeters Metric tons xvi .56: m the anti 22:21.1 a: lees 1":j' ' l ‘p- A " UsLL. .. . . ‘3 fie...- . .=....e§ .3 3; I '4‘ ea ...... “,5 399:9: . U‘ V ~ . . “‘9‘: .Ie’vcess. F“ -4 £53.: at.” \3e 4 nd‘ Ia ,‘.ess -a ‘H' 5“" .. “Q W P V; b *‘ti.:|‘: ‘ its 0 i. ‘1: - \ fl ‘ ‘ in \ - ‘)?é.‘ \‘Qr.;.‘e A ' 5 "- ‘Un MI INTRODUCTION 1.1 Who are the world's poor? There are at least 100 million people living in absolute poverty in Sub-Saharan Africa (FAO, 1986). They lack access to sufficient resources to acquire their basic food, clothing, and shelter requirements needed to lead a healthy and active life‘. Although the world's poor are a heterogeneous group, they have many similarities. The rural poor typically live in marginal agricultural areas, have poor access to institutions, and have limited voice in the policy process. First, the poor typically reside in deserts, coastal wetlands, mountainous areas, and other areas of the world with insufficient environmental stability (e.g., rainfall, soil quality, and landscape) to sustain the existing population (Leonard, 1989 and Chenery, 1974). Second, the poor have had limited access to education, and are employed on the fringe of the market economy as small farmers, shifting cultivators, artisanal fisherman, small livestock keepers, nomadic herdsman, landless laborers, or small artisans (FAO, 1986). Third, the poor have limited access to services such as credit, extension, and marketing outlets. Given their poor resource base and skill level, these limitations restrict their ability to break the P°V°rtY Cycle. rim-“11': the poor are often politically and socially disenfranchised, and live in rural areas (Al-Sudeary. 1933)- 1 . 0 itliils def-‘Lnition combines aspects of Sen's concept of ant em: (1985) and the World Bank's of food security (1986). he! :‘wmerut :51; to collect :L' :eqsztully, pcj semis; of ...e 14.2: :23: c: 3;: L3. 1:: cries : m' 3: 1| zzge: -..: . mime! :f :2 95:52.3», 7 2 These characteristics of the rural poor make it both difficult and costly to collect the necessary data to understand their situation. Consequentially, policies are often designed with insufficient understanding of the poor's aspirations, abilities, and constraints. Policies built on such misunderstandings are likely to produce uncertain results. In order to design policies that effectively increase rural incomes, it is imperative to study the economic, environmental, and cultural realities of the poor. 1.2 Problem statement 1.2.1 Zimbabwe's development objectives The Government of Zimbabwe's First Five-Year National Development Plan (1986-1990) clearly outlined the government's aspirations as: ”the establishment and development of a democratic, egalitarian and socialist society whose main aim is the develOpment and enhancement of the mental and cultural faculties, as well as efficient production and distribution of goods and services in order to raise the living standards of all Zimbabweans (Republic of Zimbabwe, 1986).” The plan highlighted the following six broad objectives for the overall economic development of Zimbabwe: (1) transformation and expansion of the economy, (2) land reform and increasing the efficiency of land usage, (3) higher living standards for the entire population, especially the rural population; (4) employment creation and manpower development; (5) develOpment of science and technology and (6) the need to incorporate environmental concerns into development programs. Of these six broad objectives, four impact directly on the well-being of the rural population: land reform, enlargement of employment opportunities ”“1 mafll><>wer development, higher rural living standards, and incorporating environmental concerns into development programs. First, land reform, a major objective in the countries struggle for Independence, continues to be an important issue. At independence, land ownership was highly skewed. Although communal farmer-households fig: 4: percent :1; P1532: I'v' a: I A 213. 5:: l;t‘....‘.'. 3‘13: 1;: d 2:: 21:12: ml: be a :42 11:5 has bee :23.) xly :3: ; 1:52: saLe. [.53 Send, the $1.“: 332.32: 1:: a; 2:35; 3f {853.1395 21'; a ”“00-4' . “flu-8.". I ~‘_ \.' u ‘ ‘en 833‘s 0 _'-: ”I 5*. 3 represent 40 percent of the population, 74 percent of their land is in Natural Regions IV and V (C80, 1986), agroclimatic regions considered marginal for agricultural production. Under the Lancaster Agreement, government agreed not to appropriate land from commercial farmers, but that land would be sold on a willing seller and willing buyer basis. Little land has been redistributed since Independencez. To date, primarily only commercial farmers in more marginal areas have offered land for sale. Also, budget constraints have prevented the government from purchasing all of the land offered for sale. Second, the government has had varied success in improving employment opportunities and manpower development. Although constrained by a shortage of resources, the government greatly improved rural access to primary and secondary education. Between 1980 and 1985, the enrollment in primary and secondary schools increased by 171 percent and 628 percent, respectively. Although government education expenditures increased by 130% during this period (C80, 1986), additional investment is needed to improve the quality of both primary and secondary education, especially in rural areas. On the other hand, the government has had less success at improving employment opportunities in both rural and urban areas of the country. For example, between 1980 and 1985, only one in ten school-leavers found work in the formal sector (C80, 1986) . Third, to increase the economic and social well-being of the rural Population, the government has sought to raise rural incomes by increasing agricultural productivity, and extending social and economic services to all rural areas (Republic of Zimbabwe, 1982). Between 1980 and 1935: government greatly increased expenditures to improve social services such as health (103 percent) and education (130 percent); and 2 A”hm-19h the government intended to resettle 162,000 families by 198 _ . (Palmer, 1990) .4. only 52,000 families had been resettled by 1989 '..Y “1.8 , '1'. '15": 'aett r 3?. tar-tress '- b m ' e “N @8tramt1 berserk. ha: I 13:.- .em.»., . r; O. I. w... ". 4.4!. 5e“.-::£ ‘04.: p... e . “. ”I... ‘0: 3'?- i ‘5 e ‘39.. de‘ ‘ 5“?“ W.‘. . H. "rcr 4 agricultural services such as extension (453 percent) and veterinary services (303 percent) (C80, 1986). Finally, the government has sought to repair the damage done to the environment as a result of deforestation, over-population, and overgrasing. The most extensive and severe soil degradation in Zimbabwe occurs in conmunal areas, representing approximately 3.8 million acres (Whitlow, 1988). Although the government currently promotes several environmental programs--including rural re forest at ion , land resettlement, and more emphasis on agricultural and conservation in schools--these programs have fallen short, given the enormity of the task. 1.2.2 Constraints in achieving the development objectives Government has experienced difficulty in achieving its objectives for two reasons. First, macroeconomic constraints have limited the number of interventions that the government has been able to initiate to improve rural living standards. Shortages of foreign exchange, budgetary shortfalls, inability to import foreign goods, foreign and domestic trade restrictions, and a large external debt (Zvinavashe, 1990) have been severe constraints since Independence. These problems are the consequence of both internal policies (interest rate, exchange rate, budget deficit, and trade policies) and external shocks (global recession, strong 0.8. dollar, and foreign trade policies). Second, a lack of reliable data about the rural population‘s characteristics and household objectives has made it difficult for government to design and target policies to increase access to economic OPPortunities for lower income households. Many researchers have highlighted the need to gain a better understanding of the structure, level, and distribution of rural incomes as a precondition for effective 9°11” design (Eicher and Baker, 1982 and World Bank, 1983). ,.z.3xmm1 W1“ xgced mt ether Leaf-:1 in screurr u. uLLets, and :7 mated agrirdtura 3.3. :creased pro: :5. :: caps: 34:: 15.: :f extensn: 311' 31:: use 2:: 1'” 3L1 (aggro; 11:92:11. For ext “U Cf :91; d: m“ stalled 1; :1. ”M395- 1.1:. 21' mince {‘3 3‘ § 5 1.2.3 Zimbabwe's agricultural sector and food security Compared with other African countries, Zimbabwe has been relatively successful in increasing food production; and creating large stocks of maize, millets, and sorghums. Policies adopted since 1980 that have stimulated agricultural sector include guaranteed prices for small grains, increased producer prices for grain and cash crops, improved access to output markets, increased availability of credit, and the expansion of extension services (Rukuni, forthcoming)3. Every year since Independence (except 1984) , Zimbabwe produced enough coarse grains (aggregate calories) needed to meet recommended energy requirements. For example, in 1982 and 1987 the energy equivalent value (kcals) of total domestic grain production (maize, millets, and sorghums)‘ equalled. 118 percent and 149 percent, respectively, of the total recommended annual domestic energy requirements (kcals)5. Further evidence that Zimbabwe produces enough coarse grains to meet aggregate energy needs is the fact that between 1985 and 1987, the closing stocks of maize, millets, and sorghums held by the Grain Marketing Board (GMB) rose from 462,000 to 1,806,000 metric tons, 4,360 to 89,000 metric tons, and 11,000 to 101,000 metric tons, respectively (GMB, various years). Yet, caloric equivalents and stock surpluses mask the prevalence of household food insecurity. National food availability does not guarantee individual household food security--defined "as a situation in 3For more details see Chapter 2 . " The energy equivalent of domestic grain production was calculated as the summation of the energy composition of the edible Portion of annual domestic grain production (Republic of Zimbabwe, 1987). 5 The total recommended annual energy intake for Zimbabwe (kcals) yas calculated as a summation of the recommended annual caloric Intake needs given its age-sex distribution (World Health Organization: 1985 and Republic of Zimbabwe, 1986) . ,5: 1;: in“ mat. the... 3;???“ '3 t e?! .3753: (the; : food sec-x: argjhfii (3 33:9: 9*: 13“.. sifted 533“ 3‘ as}: parent :5 he gets: .35? 3.191;; 3' :92: £11.21; Regs: . :2 9032:: cm in: mare h: ti: 152:3; rec ' . 3&1 {9337.33 § I. ‘~"."'a‘ 3 gov-ask. .“' a U. a same house; .M- ‘ "1 socuses c. as v. " . "a:e txl ‘41“: K N.%::‘ I. ~~ at” e.‘ ”<2? » "‘ \aflnaaN{ n 0! :f “(E “I P ‘ sesearflh 4; h . s 6 which all individuals in a population have access to a nutritionally adequate diet” (Eicher and Staatz, 1985). Recent micro-level studies suggest that even though Zimbabwe is a net grain surplus nation, a large portion of their population is food insecure. A World Bank task force on food security reported that 50 percent of Zimbabwe's population was malnourished (1983). Furthermore, Berg reported that although Zimbabwe exported grain, over 20 percent of children under the age of five suffered from second or third degree malnutrition; and that in as many as 30 percent of these children, growth was stunted (1987). The geographical incidence of food insecurity in rural Zimbabwe is largely determined by agro-ecological factors. Of the country's five Natural Regions, Natural Region I has the best, and Natural Region V has the poorest quality soil and lowest rainfall. The largest numbers of food insecure households live in Natural Regions Iv and V because these two natural regions have the highest population density (relative to their resource base) , lowest productivity, and highest incidence of agricultural-production risk (Waddington and Kunjeku, 1988). Because households in Natural Regions IV and V are most at risk, this study focuses on analyzing their food security status, and identifying alternative strategies for expanding access to food in these areas. 1.2.4 Food security equation There are two sides.to the food security equation: food availability and food access (Rukuni and Eicher, 1985). Food availability refers to an adequate amount of food being available to households--whether through domestic production, storage, or trade. Food access refers to a household' s ability to acquire food--whether through own production, market transactions (cash or in-kind), transfers. Since the objective of this research is to reduce poverty through increasing incomes, this study focuses on the food access side of the equation. ;.3 march oh :2 general ' J. A ‘. '? 'g.u’ U e - 0‘09 it 1 lav-rmfa'. try nerve: i... 12:93: :3. 1. JCSCI‘. mares a some: c manna 3795129: 1.3 Research objectives The general objective of this study is to provide a better understanding of the structure, level, and determinants of rural incomes in low-rainfall areas of Zimbabwe in order to identify alternative policy interventions to increase incomes of the rural poor. This study will address this general objective through five specific objectives. 1. Describe the level, distribution, and composition of household incomes and expenditures, including the contribution of the major sources of incomes (home-used production, cash income-generating activities, and transfers) and expenditures (consumption, investment, and transfers). 2. Describe the resource endowment of households in low rainfall areas and how they allocate these resources between alternative uses. 3. Identify the factors associated with the inter-household variability of incomes; especially for poor households. 4. Examine components of rural development strategies (short, medium, and long term) to increase incomes and expand opportunities of the rural poor. 1.4 Research hypotheses The hypotheses that guide the research are noted below. The 11m set of hypotheses examine the level, distribution, and composition of household incomes and expenditures. It is hypothesized that with respect to: (1) The level of incomes and expenditures: a) Households in Mutoko/Mudzi Districts (Natural Region IV) have higher per capita incomes than households in Buhera (Natural Region V). b) The levels of incomes and expenditures differ significantly between villages. 8 c) The differences in income and expenditure levels between villages in Buhera are greater than between villages in Mutoko/Mudzi. (2) The distribution of incomes and expenditures: a) Income is highly unequal within villages, between villages within districts, and for the entire sample. (3) The composition of incomes and expenditures: a) Lower income households earn a larger proportion of their incomes from home-production, than higher income households. b) Lower income households earn a smaller proportion of income from crop and livestock sales, than higher income households. c) Lower income households earn a smaller proportion of their income from non-agricultural product sales, than higher income households. d) Lower income households earn a larger proportion of their income from labor sales, than higher income households. e) Lower income households obtain a larger proportion of their income from transfers, primarily remittances, than higher income households. f) Lower income households are net grain buyers: while higher income households are net grain sellers. 9) Lower income households spend a larger proportion of their income on consumption, than higher income households. h) Lower income households spend a smaller proportion of their income on investments and purchases of intermediate goods, than higher income households. i) Lower income households spend a smaller proportion of their income on gifts, than higher income households. e Ah" v' 58 (I. 1 . .l as urged-hr... «KI-mur- 9 The second set of hypotheses examine household resource endowment and use in low rainfall areas. It is hypothesized that with respect to: (1) Labor endowment and use: a) In Buhera District, households have more resident members than Mutoko/Mudzi Districts. b) Households with more labor engage in more diverse agricultural and non-agricultural activities. (2) Land endowment and use: a) In Buhera District, households own more land per capita than in Mutoko/Mudzi Districts. b) In both districts, the distribution of land is unequal; with a higher degree of inequality in Buhera District. c) Farmers' cropping patterns are more diversified in Mutoko/Mudzi Districts. d) In Buhera District, farmers allocate a higher proportion of land to small grain production, while in Mutoko/Hudzi Districts farmers allocate more area to maize. (3) Capital endowment and use: a) In Buhera District, households own more traction animals per capita than in Mutoko/Mudzi Districts. b) In both Districts, the distribution of traction animals and agricultural equipment is unequal; with a higher degree of inequality in Buhera. The third set of hypotheses examine the intra-household variability of incomes and expenditures. It is hypothesized that with respect to: (1) Resource ownership and per capita household income are positively correlated. a) Households with more land per capita have higher per capita incomes. 10 b) Households with more traction animals and agricultural equipment have higher per capita income. c) Households with more resident household members have higher per capita income. (2) Household head characteristics and per capita household income are highly correlated. a) Households with more-educated household heads have higher per capita income. b) Female-headed households with the male working away from the household have the highest per capita income, followed by male-headed households, and finally female-headed households without a spouse. c) Households with older household heads have higher per capita incomes. The ‘fipgl set of hypotheses examine the effect of agricultural development policies and services on the income of rural households. It is hypothesized that with: ( 1) Agricultural development policies—-eg., price policies--have affected low and high income households differently. a) In absolute terms, agricultural development policies have raised the income of higher income households, but not affected those of lower income households. b) In relative terms, agricultural development policies have raised the share of total income going to higher income households and decreased the share going to lower income households. (2) Low and, high income households have different access to agricultural services. a) Lower income households participate in output markets less than higher income households. b) Lower income households borrow less money from the J [5.33 ke‘uu‘ n nv~l «‘0‘ I 53:54 1-5 Osmutlor. 2b.: :‘zsserta'. 1! L;:e:a"" 31:31:53 cf 1 L! :e“". § A. .‘“.bov“ , 'U‘ I :3 de"F-e., fi§¢e....‘ ”in In ..QL.’ :58;es. he. “i? \ - 0-. «5 .x.‘.c l'“ , " “fi-e: v ‘ ... q...“ s... ~ii;cn ‘ , “Sec ... I. ‘- 11 Agricultural Finance Corporation than higher income households. c) Lower income households send a smaller proportion of their children to both primary and secondary school than higher income households. 1.5 Organization of the dissertation This dissertation is divided into nine chapters. Chapter II reviews the literature necessary to evaluate the structure, level, and determinants of rural incomes in low rainfall areas of Zimbabwe. First, the definitions, incidence, and determinants of rural poverty are presented. Second, theoretical and methodological concepts--such as income definitions, measurement, distribution, monetization, and modelling issues--are presented. Finally, characteristics and rural development policies that impact communal farmers are presented. Chapter III presents the survey methods employed in the research; including ward/village selection, sampling procedures, questionnaire design, data collection procedures, and data limitations. Chapter IV examines the level, distribution, and composition of household incomes and expenditures. First, the definitions, structure, and approach used to evaluate incomes and expenditures are presented. Then, the level, distribution, and composition of incomes by sample, district, and village are described. Finally, the level and composition of expenditures by sample, district, and village are described. Chapter V analyzes the resource endowment and external environment of households in low rainfall areas. First, definitions and measures of distribution used to evaluate resource endowment are presented. Second, the level and distribution of resources are described. Then, resource endowment by per capita income (net household receipts) is examined. Next, the income level and source by resource endowment is evaluated. Finally, the external environment facing households is T [V 1 SI?" gesezzed. “a,” V v: EX rut," amass in per 0 . s A P f‘ '“. :; ":.' .eveorhs :51: 5:25 $21.1 t 1 . u ...- On, “'I o. l,..n§.-e.aa rev V" O 22;: : u.. use .1139; an: "~- I": . . C O .5197." to "' “ ' sue a hnoe' OVVO. s. . a ‘0‘ P: n L 1....95 future :1 12 presented. Chapter VI examines factors that explain the inter-household ‘variation in per capita incomes. First, the determinants of net household receipts (per capita) are evaluated. Then, the determinants of net household receipt components are evaluated, including the value of agricultural production, labor sales, and transfers (received). Chapter VII assesses the effects current policies on different income quartiles: and propose short, medium, and long term rural development strategies to increase incomes and expand opportunities of the rural poor. Chapter VIII presents a summary of the research results, and identifies future research needs. ‘\I ‘0‘" savour ' F 0-. I {on “‘9' 9“ '1‘-“ e. - . W I. ."e..‘ I "h: C. ‘4 \ . d‘“.__. ‘a‘.“. I “‘.. V“‘~ 1! ‘e .-. ‘ .~ \‘H“ ‘ . "3“‘I‘ e.‘§ i C “.59..- . t ' ‘g A ‘5 fl mu Literature Review and Theoretical Considerations This chapter reviews the literature that directly relates to the subsequent evaluation of the structure, level, and determinants of rural incomes in low rainfall areas of Zimbabwe. The first two sections focus on the general literature on rural poverty, incomes, and expenditures; and the third section develops these topics in the context of Zimbabwe's communal sector. 2.1 Rural poverty Poverty exists in all countries and in all geographic and agro- ecological settings. In developing countries--and particularly in Sub- Saharan Africa--the incidence of poverty is highest in rural areas (werld Bank, 1983), even after allowing for differences in consumption and living costs between rural and urban areas. 2.1.1 Definitions of poverty Definitions of rural poverty vary in terms of scope (general versus specific), how it is measured (relative versus absolute), and time (chronic versus transitory). 2.1.1.1 Scope: general versus specific Poverty definitions range from general to specific. General definitions of poverty emphasize deprivation , with respect to basic needs--primarily food, but also' clothing, and shelter (FAQ, 1986). Specific definitions emphasize deprivation with respect to indicators 13 "A - ".‘FI'JA tr? ‘ 1‘.'!-gqsx’r.-'. ' O A! fiz-I :a.'. 5"- kave 1:: ,...:5 cf 1 u" ;:e:a:d def ar' :7 IE e . - £ . - . O... ‘ Pv . ‘e :33; .0. e (.3 3::- ) Ow. Oge -.. a: ... “-.8 a ‘ -.‘u . -.v ~. ‘ u e J :‘ . ‘- _‘* 3rd \ “Ze: e ' ‘ 'v s " A~ :£:\,_ a ‘ K» . a & ,- 2‘ K ‘ El=‘_ 0 ‘\~ ‘ e. :N U‘ c \ 14 such as caloric intake or nutritional status. Glewwe and van der Gaag (1988) argue that poverty should be defined in terms of the actual measure used to draw (ie., calories) the poverty line; and define the poor as those households below this line. Poverty definitions range from general to specific: 1.) §g§i9_nggg§: This approach is the most general, and attempts to determine whether households' basic needs--food, clothing, education, health, and other needs--are being met. Households are defined as poor if these needs are not met. There are three criticisms of this measure. First, it is difficult to aggregate these needs into a single poverty measure. Second, the determining acceptable minimum levels is subjective. Finally, it is difficult to measure these needs (FAQ, 1986). 2.) 2§;_g§pi§g_igggmg: Per capita income is less general, and is the most comonly cited measure of poverty in the literature. This measure is constructed. by adjusting’ household income by family size. There are three criticisms of this measure. First, it fails to take into account inter-seasonal variation of incomes, and therefore fails to take nto account how households save/dissave depending on the year 2 Second, recall errors affect the accuracy of transactions data . Finally, households have other objectives besides maximizing income . 3.) £§;_ggpit§_ggn§gmp§ign: This poverty measure is more specific that per capita income, because it is constructed using what households actually spend on consumption, adjusted for household size. Critics argue that since this measure includes all consumption. goods, not just food, more households than those identified by this measure are actually living in poverty. 4-) MW: This measure is more specific than per capita consumption, and only requires information about food consumption. The advantages of this measure are that: 1.) it requires less data, 2.) recall is easier for food than for other consumption items, and 3.) food. price indices are easier to construct than non-food price indices. Although less data are required for this measure, one criticism is that because other non-food necessities are not included; it provides a less comprehensive understanding of poverty. The accuracy of this method depends on having an estimation of a households' propensities to consume (Anand and Harris, 1985). 5.) Iggg_;3§ig: This measure estimates the share of a household's budget that is spent on food. This measure stems from two observations made by anel: the share of the budget for food 1For example see Dione (1989). 2For example see Scott and ,Amenuvegbe (1990) and Lynch (1980). 3For example see Ellis (1988). decreases we sure ‘J I good the sine: 355837893 6.} Cal::; mama; «jam: r. c‘.m;f;e:' OO-IA ' . ...:vau & U. negate . IFOA. ': ....DV‘A (.- r I I “..I 5" nae-... 223.157.- :;::';:fe:e PC? rty. :eaLthy c teasures a masses 35636: 511331;: ‘ Let-tan, ; " 1 e nuan‘ h"nru e ‘16 1:5. ‘ “.... '§4.5~- 15 decreases as incomes increase and the share increases as family size increases. Therefore, the proportion of income spent on food is a good proxy of a household's welfare (Samuelson, 1980). On the other hand, Thomas (1986) questions whether Engel's first observation holds for the poorest households. 6.) W: Some researchers measure actual household and/or individual caloric intake, and compare the estimated levels against standard requirements. Households below some level are classified as in poverty. This measure is constructed from typical diets of the studied population, and tries to assess if an adequate amount calories are being consumed. Sen (1981) and Lipton (1980) challenge the objectivity of this measure. 6.) MW: Medical indicators of health and nutrit ion--such as height- f or-we ight , he ight- f or-age , arm circumference, and so on--are the most specific measures of poverty. These measures assume that poor households are not healthy or nutritionally well. For example, anthropometric measures are widely used for mass screening. Problems with these measures are unreliability--due to intra-observer and inter- observer imprecision--and population specificity--because the standard is not necessarily relevant to the studied population (Lohman, 1988; Lukaski, 1987; and Christakis, 1984). 2.1.1.2 Measurement The distinction whether poverty is defined as those households below some specific level, or in relation to other households is important. Relative poverty measures are concerned with the relative ranking of households with respect to income and consumption levels (FAO, 1986) . Conversely, absolute poverty measures attempt to determine if a household has sufficient income to meet it's basic consumption--mainly food--needs (FAO, 1986). A major problem with estimating absolute poverty levels is that one must first define a minimum level, against which households are evaluated. 2.1.1.3 Temporal dimension: chronic versus transitory The literature highlights the importance of the temporal dimension in analyzing the incidence of poverty (Glewwe and van der Gaag, 1988 and Poleman, 1984) . Households in chronic poverty are unable to produce or acquire enough food from year to year (Glewwe and van der Gaag, 1988). Normally, these households are resource poor, and live in unfavorable environments (is. low rainfall, poor quality soil). Households in :...3),' :21 to produce 3; 351;, 19::j9. 1.151;: or 1:. u:- 2.l.2 Imdcnce 1.23;;1‘. pcve: mt; cf dents. 52.5313 \ r“~nne he-«ve’.‘ SEE: -= 3i. researche. .5 . .l K‘- 1“.“ v: I ‘ '1 k e9'\ .1, tke .8 .3: k FEVER? ‘ I r: i \‘H s-‘:‘ ~‘ s e A ‘eenae I. *9 .\\'e I ‘4. 1“ t.‘A . ‘V‘q d D Q 3‘ e s l6 transitory poverty are generally able to meet their food needs, but are unable to produce or acquire enough food in a given year (Glewwe and van der Gaag, 1988). This is usually a result of seasonal fluctuations in rainfall or an unusual event (eg. a death of a family member). 2.1.2 Incidence Although poverty estimates are relatively inaccurate and employ a variety of definitions, the incidence of poverty is pervasive. We Although specific estimates vary considerably, available data indicates that poverty is a major worldwide problem. In the early 1970s, researchers estimated that, on the basis of available cross- section and cross-country observations, between 370 and 800 million people lived in absolute poverty (Fields, 1980). These studies estimated the numbers of absolute poor, but not their geographic dispersion. ‘In the late 1970s, FAO's Fourth World Food Survey (1977) reported that in 1972-74, based on estimates of food available for consumption, approximately 445 million people (25 percent of the total population of developing countries, excluding centrally planned Asian countries) were judged poor. In 1985, the World Bank estimated that more than 1 billion people lived in poverty throughout the world. Of this total, 520 million were in South Asia, 275 million in East Asia, 175 in Sub-Saharan Africa, 60 million in Europe, Middle East, and North Africa; and 70 million in Latin ,America and the Caribbean (World Bank, 1990). There is a disproportionate concentration of the world's poor in Sub-Saharan Africa. Although Sub-Saharan Africans accounted for only eleven percent of the world's population, sixteen percent of the world's poor lived there. Although the World Bank (1990) projects that, given current trends, 2:1591’93‘2 . e"! UP 9 eve” ... s e." r" angle. the O L‘:.:a LI es . '. at . ..-: ‘ ~ - J: ...“ ‘4 d _:_:..::‘.:.*.5 cf 1322:2538 :25 ' e :‘ . . . " grvezte . . ” "i O... ‘ fi " 7 an. . I ‘- 5 5‘ 5 a la.- ,'.‘:--. "I‘ . V .g: .. I “§‘\ ‘§ . 5" a3al‘;h ‘;‘ Q.‘.y o L ‘5 \‘a23 (‘8‘. ‘36: (I 17 by the year 2000 the numbers of the world's poor will decrease, this optimistic projection doesn't apply to all regions of the world. For example, the numbers of individuals living in poverty in Sub-Saharan Africa is estimated to increase to 260 million (Table 2.1). Poverty is largely concentrated in rural areas. FAO (1986) estimated that in 1975-1982 (based on data from 60 developing countries with populations of one million and over), for the countries considered, the percentage of the rural population in absolute poverty varied from 11 to 90 percent. In Sub-Saharan African countries, between 35 and 90 percent of the rural population lives in absolute poverty. 2mm The Zimbabwe Government's Transitional National Development Plan (1982) stated that poverty in Zimbabwe is concentrated in the rural communal areas. About half of these households had few or no cattle, and that about 20 percent had no land rights. Furthermore, in rural areas the average cash income was one-third of the agricultural workers' minimum wage, and one-sixth the cash income of mining and industrial workers. The widespread incidence of poverty in Zimbabwe is highlighted by the fact that over 70 percent of the population live in rural areas (C80, 1988). Yet, little is known about the characteristics and geographical dispersion of the rural poor in Zimbabwe because of the limited availability of rural income and expenditure data (Werld Bank, 1983). Rohrbach (1988) and Stanning (1985) have investigated related issues which are discussed in later sections. 5:1: 2.1. N are": ‘95 $1.? an its”? I" 58‘. “”3 if” I“: Q “‘3'? 1‘“: ~"-“ h‘u I inner lm:3‘; 3 3 £21.55 ‘v 3‘-‘ “a a, . 1‘“? 8;" . . \e‘e ‘5. ~ en‘s“ ‘- ' ‘. a‘: ‘9‘“ 3‘ K 18 Table 2.1. Poverty in the developing world, 1985 and 2000 , ‘ Geographic distribution of poor I Gunnumfl:rqfla1 a Huber X of X of lumber x of X of ! (millions) population world (millions) population world ! (1985) (poor (2000) poor I F South Asia 520 51 47 360 26 45 ! Sub-Saharan Africa 175 67 16 260 39 32 ( East Africa 275 20 25 65 4 8 ‘ Europe, Middle East, and North Africa 60 31 5 70 12 7 Latin America and the Caribbean 70 19 6 50 12 7 TOTAL 1 . 1110 ,-1. VH1 334“ __100_» "_ 8057_ 7 16 100 Source: Horld Develop-ant Report (1990). 2.1.3 Determinants of poverty Numerous studies provide insights on the determinants of poverty. General determinants Rural poor households are very heterogeneous, but they typically lack access to sufficient land, labor, physical capital, and human capital to acquire sufficient food--whether through own production, market transactions, or transfers (World Bank (1990) and FAQ (1986)). For most rural households, agriculture is ‘the single largest income source. Furthermore, the rural poor 1.) are vulnerable 'to inter-seasonal climatic changes, 2.) are ignored by agricultural policy makers, and 3.) have poor access to public services. The level and sources of household income depend on both internal ‘Excludes Eastern Europe. "Hm-n “"le In a A ‘ .no a- frat i- e 332.3193. r Oie. ' ‘QA ' ." -¢ O u Laden have "I as. a \_’; a $3.51“ at e.‘ r . K. 2‘: 23. I ‘ I 85. I I \ "i ‘ “-e t a “ .F' ‘ “\ C\ ‘ A " eV' e‘ e“ SC. ‘ -. ~.*H . as.~ . y“ ‘ ‘ '1» '4 “ \‘ “ a a ‘\ 19 (endogenous) and external (exogenous) factorss. A household's internal environment consists of the level and quality of resources--land, labor, physical capital, and human capital, and decision-making ability of household members to efficiently allocate them. am Own production is the major source of income in agricultural-based communities. Thus, access to land is critical to enable households to meet their food needs. Several studies show that small farmers and the landless have a higher incidence of poverty in South Asia, Southern Africa, and much of Latin America (World Bank, 1990; FAO, 1986; and de Janvry, 1981). As population increases, household access to land will decline even more. Household production is not only influenced by the quantity of land farmed, but also tenure arrangements. De Janvry (1981) argues that without clear user rights, farmers 1.) can not use land as collateral and 2.) inter-seasonal access is uncertain. When user rights are unclear, as is the case in many rural areas of Africa, farmers may lack the incentive to invest in land improvement because of the uncertainty of reaping the returns, resulting in eventual environmental degradation. um: Labor availability is also an important determinant of whether a household has the ability to produce enough food. In Africa, large households are desired because of the importance of children's contribution to household activities (ie. herding and weeding). Also, large families are needed to insure that the household has enough labor (World Bank, 1990). Household types that have a high incidence of poverty include the elderly (who have inadequate labor and capital) and younger households (which haven't accumulated enough resources). 5 The household's external environment consists of the agroclimatic, services, technological, and cultural environment; all of which influence household decisions, but over which the household has little control. 20 flgmaa_£asi£al Poor households generally have minimal access to education and health services. The quality of human capital is a shifter of the household's production function. Some studies have demonstrated a negative correlation between education and poverty. . For example, education improves technical efficiency in agriculture' by increasing farmer access to extension literature (Bernsten, 1978); and provides greater access to off-farm employment opportunities (Chuta and Liedholm, 1979). Furthermore, Schultz (1990) states that: "The decisive factor of production in improving the welfare of the poor are not space, energy, and cropland; the decisive factor are the improvement in population quality and advances in knowledge.” Ehzsisal_casital Finally, access to physical capital enables farm households to fully use their land and labor resources. In Sub-Saharan Africa, traction equipment and animals are key capital inputs required to increase labor productivity. Several studies (World Bank, 1990) show that poOr households lack access to these capital inputs which permit extensification (labor extending) when labor is scarce, and intensification (land extending) when land is scarce. Poverty in Zimbabwe To date, there is limited empirical analysis of the determinants of poverty in the communal areas of Zimbabwe (World Bank, 1990). Available evidence suggests that poverty is primarily associated with inadequate land availability, weak agricultural infrastructure development, and specific family characteristics. Access to land in Zimbabwe is highly skewed (ROZ, 1982). About 6,000 comercial farmers own 44 percent of the total agricultural land, located predominately in the better agro-ecological zones (1, II, and craved Cezverae'. «.4 ugly in ; meted that infirm-are. natal teas m1 pcpalat u: “4...-.. 2'1. :t is l rem-meted l: “1131 areas, 55:. Tina p04 3* 7‘“; PC2533 ix: £951.33: {tied a p: :3!!! fig 331-3th 1‘ - 1., 3.3." (£3333; F' I‘ e‘..;§l e‘ ‘ \‘;3‘ as 012;. 21 III). Conversely, 700,000 comunal farmers occupy 42 percent of total land, mainly in poorer agro-ecological zones (IV and V). The Government estimated that given the current levels of available technology, infrastructure, and management systems, the carrying capacity of the comunal areas is only 325,000 families, about half of the existing comunal population (R02, 1982). In addition, the distribution of agricultural infrastructure development is highly skewed. During the colonial period, Government underinvested in extension, marketing, education, and credit services in communal areas, which has affected agricultural productivity (ROZ, 1982). This poor access to services has contributed to impoverishing the rural population. With respect to family characteristics, a study in Gutu and Gwanda identified a positive correlation between poverty and family size, access to capital and draft power; and a negative correlation with land ownership (Economist Intelligence Unit, 1981). 2.2 Incomes and expenditures This section presents theoretical and methodological concepts relevant to the analysis of income and expenditure data; and reviews past studies . 2.2.1 Theoretical and methodological concepts This section 1.) presents various definitions, of incomes; 2.) evaluates alternative income distribution measures; 3.) presents techniques to test statistical significance, and 4.) discusses issues related to the monetization of households. 2.2.1.1 Income definitions Hany definitions of incomes are found in the literature. Most definitions only include some of the components of income, and therefore . "C “WE“ . *0 h .. ID 25:31:35 artaltxa :rtgzatlcr. :‘z'.at'.- ;: acre :2: a tze 8'2: l;.:.".t::a 2"e-e.-. ~h..-_.. 51.221333 5 ”flies tr 2.56.5: f :. . ' .-. “' e=:‘} z- -. \~He: a N 3". “=::~‘. ‘H . a L“\ o “‘5 ;Q \ .y ‘ \. egg V \b , S. 22 only partially describe the household's opportunity set. Partial income definitions are useful to evaluate returns to resources used to produce agricultural goods and to compare returns to alternative enterprise combinations, but they provide a misleading assessment of absolute and relative income levels. ' Hore inclusive definitions measure income as earned income--computed as the sum of cash or in-kind income--from both agricultural and non- agricultural sources (Hatlon, 1977; King and Byerlee (1977)). This definition of household income is more complete because it permits the estimation of the returns to available resources. Hayami (1978), Atkinson (1983), and Sen (1987) define total income in a more comprehensive manner, including transfers‘. This definition provides the most accurate measure of the total income available to the household for expenditures. For poor households in Zimbabwe, transfers (remittances) are an important income source (Stanning, 1985, World Bank, 1983). Thus, it is necessary to use this comprehensive definition to assess the adequacy of incomes to meet consumption needs. Analysis that uses expenditures as an income proxy tacitly uses this comprehensive income definition--including transfers and credit receipts (net)--because it is impossible to identify the income source used for a particular expenditure, so all income is included. This study uses both the earned income and total income concepts for the descriptive analysis; and uses the total income concept to identify determinants of income (Chapter 6) and in the policy analysis (Chapter 7). 2.2.1.2 Income distribution Eicher and Baker (1982), in their critical review of agricultural research in Sub-Saharan Africa, stated that research on income ‘This is the definition (net household receipts) that is used in the subsequent analysis. ., A --~vr 'be‘u" u"). g. —-_L .7 21.. 'rgltttion in :5 nation pr: strata: of in: 2.2.1.2.1 lltam 2: cr' : to stare ceztza‘. 1 31:71am: 1. Ensures of c 112, User. :59 3'4 attache: 23 distribution in rural areas is a high priority topic for the 1980s. This section presents the theoretical considerations concerning the estimation of income and its distribution. 2.2.1.2.1 Alternative measures of central tendency In order to select the most appropriate descriptive statistic to measure central tendency, it is necessary to determine whether the data are symetrical. Two measures of symmetry are skewness and kurtosis. Measures of central tendency are values that represent the average value, when the data are arranged according to magnitude, of a set of data (Bhattacharyya and Johnson, 1977) . The most commonly reported measures of central tendency are the mean and the median. If data are normally distributed, the mean is usually reported; the mean and the median are exactly the same when the distribution is perfectly normal. If the data are skewed, as is often the case with income data, the mean is a misleading indicator of central tendency because the mean is more sensitive than the median to extreme values (large degree of skewness) (Alreck and Settle, 1985; Steel and Torrie, 1980; and Ehattacharyya and Johnson, 1977). Skewness measures the degree of asymetry, or departure from symetry, of a distribution (Bhattacharyya and Johnson, 1977) . If a distribution is positively skewed, then there is a longer "tail" to the right of the central maximum; if it is negatively skewed, then there is a longer 'tail" to the left of the central maximum7. Kurtosis is the degree of peakedness of the distribution, usually with relation to a normal distribution (Ehattacharyya and Johnson, 1977). The higher the statistic, the more peaked the data distribution. 7Skewness is the mean subtracted from the mode, divided by the standard deviation. "I b ‘ ‘1“? ' 2.2.1.2.2 lit! :2: sectlc i 15355:: c 5 resent -, asses use: I . l 3“” a “re: '00... . '1 o .1: seal: ill! Lnéeper Refers I Se: «was ssa 1: n‘d As- ““ v“ a m”.A1e I! m“" e.‘.‘ . 1.. a _ ~ - ' a“? “as ‘ . ‘eo .. a. ”.9 .. e F’en- go . “a: “cs ,5! ‘ Ad 'I :P‘ 1 ‘z‘aezti ‘- 'e a" :C‘ ea ‘ .. F-..‘ Hg'.‘.'a‘ 9‘ e‘ :54 I “can ‘e 24 2.2.1.2.2 Alternative measures of income equality This section evaluates potential income inequality measures. First, a discussion of desirable properties of potential measures of inequality is presented, followed by a survey of potential measures. Finally, the measures used in this analysis are presented. ea e Three desirable properties for income inequality measures are: income scale independence, principle of population, and the principle of transfers (Sen, 1973). Income scale independence means that the income distribution should not depend on the level of total income (Cowell, 1977). In other words, as everyone's income changes (increase or decrease) proportionately, there shouldn't be a change in the inequality measure. The principle of population states that the measurement of inequality should not depend on the size of the population (Cowell, 1977). For example, if two identical economies (therefore, with identical measures of inequality) were added together, the inequality measure satisfies this principle if it is the same for the aggregated economy as for the individual ones. The principle of transfers examines the impact on inequality measures of a hypothetical transfer of income between two individuals. Their criteria can be satisfied in either weak or strong terms (Sen, 1973). The W is satisfied when the income transfer is from a richer individual to a poorer one, and is less than 1/2 the difference of the income between the two individuals; and when the transfer of income is made, inequality is decreased. The W is satisfied when the amount of the reduction in inequality depends only on the distance between incomes, not which individuals are chosen. The distance concept measures the difference in incomes between individuals. The stronger property is more desirable because it measures the ' ”IL". 21. YWJ M"erence betvee 251.3. :.="a." off-A.— rue-Den U. .eoku 2‘ .lter t3: 12?! 2.) re.at rattan, S.) s i In." A. “;A n via... gee...~ 45‘ L: gap.- ' l .32 ring. .1 :.::'e:e::e 59.39 -.‘ . " aw. anooe. 25 difference between income shares, and the inequality measure is derived directly. AuasmettBLJmQQMLinsmuuiELmsamnee The literature identifies six measures of income inequality: 1.) range, 2.) relative mean deviation, 3.) variance, 4.) coefficient of variation, 5.) standard deviation of the natural logarithm of incomes, 6.) Gini coefficient. Table 2.2 presents the formulas and properties of the six inequality measures. The range (R), the simplest measure of equality, measures the difference between the highest and lowest income observations as a ratio of mean income. The range8 is calculated as: The value of R falls between zero (income is divided equally between all individuals) and n (one individual receives all of the income). This measure ignores the distribution of incomes between the extreme values, and is sensitive to outliers. The relative mean deviation (1!) is a more complete measure than R because it looks at the entire distribution, not just the extremes. The relative mean deviation is calculated as: n 51:2 ly'yil ha? 1 The value of H falls between zero (perfectly equality) and 2(n-1)/n (all income to one individual). The main disadvantage is that similar ”For all income inequality measures, yi = income of observation 1 y mean income n number of observations “fair? 'Q" mm min was 131.2". 'WEIII EJ.T£ +0! IIII'IJ T} 26 Tulle 2.2. "met-ties of alterlmtive inn. insmmlity m [—1.717— INOEPENDENT OF OF PRINCIPLE DISTANCE PRWTIGIAL RANGE IN INEGJALITY OF CGICEPT INCREASES IN INTERVAL TRANSFERS INCGIES AND [0,1]? POPULATION 1. RANGE (R) FAILS ABSOLUTE NO NO DIFFERENCES ( INCREASES) (UNBGJNDED ABOVE) 2. RELATIVE W DEVIATIN FAILS ABSOLUTE NO NO (I) OI FFERENCES ( INCREASES) (UNBQINOED ABOVE) 3. VARIANCE STROIG ABSOLUTE NO NO (V) DIFFERENCES ( INCREASES) (UNBGJNDED ABOVE) ‘ DIFFERENCES IN j 6. CfiFFICIENT TNE LOG OF NO OF VARIATIGI WEAK INCGIE DIVIDED YES (UNBGJNDED (CV) BY THE INCOES ABOVE) THEMSELVES 5. STAIDARO DEPEIDS OI THE DEVIATIOI OF RANK NDER OF NO : TNE NATLRAL FAILS INDIVIDUALS IN YES (UNBGJNDED LOG OF INCGE A PWLATIOI ABOVE) (SOL) 6. CINI CGFFICIENT (G) ABSOLUTE DIFFERENCES ' ' 1 :lb)‘ malts can be re 3233- :; unanc m cacglete :2 seat, thus 27 results can be obtained with different distributions on the same side of the mean. The variance (V) is similar to the relative mean deviation, but is more complete because it squares the differences of observations from the mean, thus accentuating the differences. The variance is calculated V= 2 (Team / n 1 This measure has two advantages. First, it is sensitive to differences from the mean for all observations (called the Pigou-Dalton condition). Second, larger deviations from the mean are ”penalized" more, resulting in a higher value for V. The disadvantage of this measure is that a distribution could have a larger relative variation than another and still have a lower variance, if the variation around the mean income level is smaller than with the other distribution. The coefficient of variation (CV) is a more complete measure than V because it is both sensitive to differences from the mean like V and independent of the mean income level. The CV is the square root of the variance divided by the mean income level. The CV is calculated as: CV= (V/T'Y" The CV has the advantage that it: 1.) discriminates between distributions where weight is given to income differentials in the high income range, 2.) is independent of preportional changes in income or population, and 3.) it weakly satisfies the principle of transfers. This measure has two weaknesses: I.) the squaring procedure is arbitrary and 2.) it weighs differences equally. There is no a priori reason to use either of these procedures (Sen, 1973). .o‘.“ h’ «(O-’4 . :2 standard 3 mt useful 2:53;; greate 4 , V ' ..mvnufl‘ua ‘ 'hl .....laeu . c an ~ ‘ ‘ 1~i.;y I 28 The standard deviation of the natural logarithm of incomes (SDL) is the most useful income inequality measure if one is interested in attaching greater weight to differences in income between lower income individuals. The SDL is calculated as: SDL = [2 (10967) -log(y,-) ) /n]“2 1 The SDL: l.) eliminates the arbitrariness of the units used, 2.) gives greater weights to incomes in the lower range (more appropriate when interested in measuring extreme poverty), and 3.) is independent of proportional changes in income and population. The weaknesses of this measure are that: 1.) it uses an arbitrary squaring procedure (same as the CV), 2.) it fails the principle of transfers, and 3.) it is seldom reported, so it is difficult to compare this measure with results from other studies. The gini coefficient is the ratio of the area below the line of perfect equality and above the line representing the actual distribution of incomes, to the entire area below the line of perfect equality (if income equally distributed). The gini coefficient is calculated as: GINI=1+%-2/n3y[y1+2yz+-.- +nyn] where: y1 > y2 > . . . > y" The gini coefficient: 1.) is more sensitive to income differentials in the middle income range, 2.) is independent of proportional changes in income and population, 3.) has an appropriate distance concept, given the skewness usually found in income data, 4.) it avoids the arbitrary squaring procedure, 5.) it satisfies the weak condition of the principle of transfers, 6.) it is a direct measure of the income differences (ie., it looks at each pair of incomes), and 7.) the measure is frequently I 30 - Aflglh ~ e ...Uee-es-: 313.25 "9 .04 .--5 1;: ‘ ‘ ‘. -“e eoa - 5.?“ ‘ o~. . 'e a .Q...' . m 34;;23' : ~ tn. h 5 '2-3.“ .__ “ ‘5 . ~ha: . .“1 h.- . \~ EN“ "no! t‘. a K m . .‘ ‘\=Sa ls‘. \ $c'l \5“. ‘ ‘ ‘i “- R‘s- 29 reported so it is possible to compare it with results from other studies. These first three measures are not used to measure inequality in this study because they give misleading insights about the inequality of incomes. The range only looks at the highest and lowest incomes relationship to the mean. Both the relative mean deviation and the variance are dependent on the mean income level, and therefore don't examine the relationship of each pair of incomes in the sample. This study uses the coefficient of variation, the standard deviation of the natural log of income, and the Gini coefficient to measure inequality because they each give a slightly different view of the inequality of incomes. 2.2.1.3 Significance testing This section presents the statistical techniques used to test the null hypothesis that group means--between Districts and per capita income quartiles--are equal’. Student's t-test is used to compare District means (two independent groups). For multiple group comparisons, oneway analysis of variance is used to test the null hypothesis that means across income quartiles are equal. Duncan's multiple range test is used to obtain multiple comparisons between quartiles. This test identifies pairs of group means that are significantly different (five percent level). 2.2.1.4 Honetisation As rural economies develop, rural households increasingly rely on the cash economies to met their production and consumption requirements. Thus, monetization, measured by the degree households participate in the cash economy, is an indicator of rural economic development (Von Braun 9These tests are also used to compare household resource ownership across groups. g: lrzzerly, lass: ‘ xlzie tag: of t 3:23;: :za:‘.zat'.:r. rre feed 5 2:12:80 3;: . ”~-’ ‘ P 2..-. s b. “—4-... .. _ ...-A- _ ‘Br'e ;. _ ..-..~ ‘ee a‘ 'm R ..e -e . ‘Q‘I ._ _. a: 0. _ o~me.‘n; f .I "" ~11 9 .:~‘ . a _i . 'l" a‘.‘ ”ea-Lure Q. ' . “3’9 ‘ n ‘m. 'lm fi.“w ‘- e i ‘ ‘4 30 and Kennedy, 1986). Factors that contribute to the monetarization of rural economies include rapid urbanization, growth in the rural nonagricultural sector, and technological changes in agricultural production. First, rapid urbanization creates pressure to change food policy, either to import more food or to design marketing and production policies to extract marketed surplus from rural areas. Second, the growth of the rural sector is closely tied to the growth in food production, and households participation in markets (Mellor and Johnston, 1984). Third, technological change in agriculture usually requires farmers to apply purchased inputs and encourages enterprise specialization, both of which result in an increase in the monetization of households. The literature identifies both positive and negative impacts resulting from increased participation of semi-subsistence households into the cash economy. Studies by Pinstrup-Andersen (1988) , Dewey (1979), and Gudeman (1978) conclude that increased commercialization has a negative impact on nutrition and income; while studies in Kenya (Kennedy and Cogill, 1987 and Fleuret and Fleuret, 1983), Papua New Guinea (Harvey and Heywood, 1983), and Tanzania (Lev, 1981) suggest positive impacts; Alderman (1987) examined data from 15 countries and observed little impact; and studies in Kenya (Hitchings, 1982) show mixed results depending on crops studied. In Zimbabwe, Jackson and Collier (1988) found that as the percentage of income from cash sources increased, total per capita household-income increased. This study examines the impact of monetarization in Zimbabwe, the relationship between the percent of a households' income received from cash sources, and the household's level of per capita income. 2.2.2 Past income and expenditure studies This section first reviews income and expenditure studies conducted throughout the world; and then reviews studies conducted in Zimbabwe. 2.2.1.1 world I llterat; miles tend; tithe Carl Esra: Afr; awe: Hath ', ‘. 9,- FIC‘v'lse pe::er.t tze 7a; attlvlt Pat-.513. iii are :as': *- -04 sv.‘ ‘~ 5.‘ee 3. 5') 55;- ‘r \ ar- ~- '14.“ dfid 55, 32:5; l m.:an I y. 4': .‘~e - is . \fi . 5‘9. C 31 2.2.2.1 World A literature search identified 29 major income and/or expenditure studies conducted since the early 1970s--10 in Asia, 7 in Latin America and the Caribbean, 5 in North Africa and the Middle East, and 7 in Sub- Saharan Africa (Wahab, 1980 and Glewwe, 1990). These studies are compared with respect to”: l.) W: Only twenty-qne of the twenty-nine surveys provided a definition of income1 . The vast majority (86 percent) defined income as total household income, which included the value of home production, farm product sales, wages, off-farm activities, and transfers received. Two surveys--Botswana and Pakistan--reported total available household income, with transfer and credit outflows removed. The Reunion survey only collected cash income received by the household. 2.) W: In all cases, except in Sri Lanka, the implementing agency was the government statistics office. 3.) ove a e: In 74 percent of the studies, the coverage was national; while 17 percent included only rural areas, and 9 percent only urban areas. 4.) W: All studies used a one year reference period, with different starting times. The recall period for the studies was not reported. 5.) 53mp1§_§11g: The sample size of the studies ranged from 131 to 56,000 households. The national surveys interviewed between 1,000 and 56,000 respondents, between 131 and 1,700 respondents for the rural surveys, and 4,000 (only one reported sample size) for the urban study. 6.) ziglg_ptgfii: The skill level of the field staff employed to collect the data varied considerably between surveys. Of the sixteen studies that reported specifics about field staff, about 56$ hired temporary enumerators and 44 percent used permanent staff. For the eight country studies that reported both the sample size and the number of enumerators, the enumerator to respondent ratio varied between 1:4 and 1:150, the median being 1:55. 7.) W: All studies used a two or more stage stratified sample design. ° 10There isn't complete information for all surveys to compare all aspects of design and implementation. 11Six were household consumption and expenditure surveys, and two that did collect income data--Fiji and Sudan--didn't provide a definition of income. 5 “-1 9:3: .1... '33:" 20.35, . l- .1 1332'.” ..d- e ' 5 fi‘ .1 .e:;t.. .- "m-n. .- 5* bang!» . :AI-Z-Z Em! :°--e .-A- '.“' bur-I ...... . ma ::....ec __ m :tat‘.st;:e ( $3515 ‘ c .h 4.. m {13:5 2: ‘e - N'm \‘ ‘IA 32 This study used a total household income definition, covered rural households, used a one year reference period (with a one month recall period), had a sample size of 285 households, employed enumerators for the length of the study, and used a three stage stratified sample design (see Chapter 3). 2.2.2.2 Zimbabwe Since Independence (1980), researchers have conducted four studies designed to estimate household incomes: MLARR (1988-89), Central Statistics Office (1984-85)12, Stack (1985-87), Amln (1986-87), and Govaerts (1984-85)“. CSO conducted the most comprehensive survey, which estimated household incomes and expenditures in all provinces in Zimbabwe (Table 2.3). The estimates for Mashonaland East and Manicaland provinces serve as useful benchmarks to compare results from Mutoko/Mudzi and Buhera Districts--even though our study villages are in the poorest agro-ecological portions of these provinces. The CSO study used three income definitions developed by the United Nations: total household income (an earned cash-income concept), available household income (total household income plus net transfers and cash remittances), and income available for consumption (available household income plus in-kind income). Results from Stanning's income and expenditure study are most comparable to the Mutoko/Mudzi and Buhera District study because she collected data in similar agro-ecological zones; used the same length of recall period; and used the same income definition (including transfers and the valuation of home production). Stanning's study is particularly important because data were collected over two years, thus ‘ZJ. Jackson's research used the data collected by the C80. 13For specific details about these studies' results, and how they compare to this study's estimates, see Chapter 4. __. '; M. ' 3 .:.le, a‘ ‘ -' mne- ' s 'i- Una a new 5“ A -. uv.‘ 1 A. u A“ I hflC'boUee ‘ 1 . pique: : m V .."ms .h to :r . . :_‘ em...‘.:‘ 33 providing an estimate of inter-year variability in incomes. Jackson's analysis, based on data collected by the C80, used a total household income definition (including transfers and valued home production). Because this study used a long (annual) recall period and aggregated observations across Natural Regions, results may incorporate considerable measurement error, and are not comparable to the Mutoko/Mudzi and Buhera District study. The other two studies, Amin and Govaerts, only collected cash income transactions data-~farm product sales, non-agricultural sales, wages, and remittances. This limited. definition of income restricts the usefulness of these studies for comparative purposes. 2.3 Zimbabwe's communal sector Although communal households account for a majority of Zimbabwe's population (C80, 1987), little is known about the internal and external factors they face. This section describes the characteristics, institutions, and policies that influence communal household behavior in order to provide the necessary context to understand and interpret this study's results. 2.3.1 Characteristics of communal households This section presents the geographic dispersion, administrative structure, agricultural production, marketing, and consumption preferences of communal households. 2.3.1.1 Geographic dispersion Communal households accounted for 57 percent of the total population, they occupy only 42 percent of Zimbabwe's total land area (C80, 1987). This disproportionate allocation of land is exacerbated by its poor e':.'. '(‘s I 'ih" sum-7dr- F“ g: use Distr‘ct I . m I warms UN Mon: um "rm Stu Shawn M:- D"; h I”in: 1111 521's? , a: m use All 34 Tdale 2.3. m of pest inc. mid expellitIa-e stufies in litmus Nean j Researcher District Province Natural Samle Interview Income Income ‘ Region Size Freqmncy Definition Levels (25) eight including 414 MLARR Nutoko Nashonaland lV 48 Two Earned income 934/hh Nest Visits definition . Buhera Manicaland lv 58 713/hh liurmgwe Nashonaland lla, Ill 80 Monthly Total income 372/pc West definition; including Stack Shanna Nashonaland transfer and 260/pc Central lib, Ill 69 Nonthly the valuation of home , Binge Natabelelan production 108/pt: ‘, North V 20 Nonthly Chirau Nashonaland ll west Cash income Min 614 Single only; doesn't 428/hh Nagondi Nashonaland 111 Visit include home Nest prediction Total avail. Nanica. Central income 1237/hh 1 Statistics All All i - V 7000 Every 4 definition; Office months including transfers Nash East (net) and the 2992/hh valuation of home prochction Jackson All All 1 - V 600 Every 4 Same as TOO/hh months Stack Cash income Coveerts Nutoko Nashonaland iii, W 200 Single only; doesn't 442/hh East Visit include home - oduction Source: MLARR ( ‘7), Stack - . ope "1), Am n (1"); (:90 ( --;;), Jackson (1 ); . - Govaerts (1987). 3;".1', Cl ‘5 genre 53': Ne Se hi 5‘... ‘ . \e‘a‘ rain .‘Q N \ 5'31 b‘ R. Q‘ P); ‘ mi‘e‘.‘a1] “ a ‘K. “a 35 quality, classified as semi-intensive“ (17 percent of farms), semi- extensive15 (45 percent), and extensive16 (29 percent) farming (C80, 1988). 2 . 3 . 1 . 2 Agricultural production Crop and livestock production account for a large portion of both total household income17 (43 to 85 percent) and cash income18 (approximately 50 percent). 9:22! Despite government ' s restrictive recon'mendat ions about the appropriate crop choices for each Natural Region, conmunal farmers grow a diverse mix of crops. Rainfall and individual preferences determine the relative importance of different crops to household income and consumption. In the aggregate, maize is the most important crop in terms of area planted (55 percent); followed by bulrush millet (9 percent), sorghum (8 percent), groundnuts (7 percent), finger millet (6 1‘The semi-intensive farming area (Natural Region III) has an annual rainfall of 650-800 mm, is subject to fairly severe mid-season dry spells; and is recommended for livestock, fodder, and cash crops. Production of maize, tobacco, and cotton is considered marginal . ”The semi-extensive farming area (Natural Region IV) has an annual rainfall of 540-650 mm, is subject to periodic seasonal drought and severe dry spells during the rainy season; and is recommended for livestock and drought-resistance crop production. 1“The extensive farming area (Natural Region V) receives insufficient rainfall to even produce drought-resistant fodder and grain crops; and is recommended for extensive cattle and game farming . 1"Stack (Stack and Chopak, 1990) , based on data from the 1986/87 agricultural season. 18Govaerts (1987) , based on data from the 1984/85 agricultural season. ,Ve‘h ”Ill ..I’J'I ‘Il 1111' 1'1‘23'. I I CC...» C mars. rice. legs: 22., ea: - I ‘A' A‘ - .exmy .e. cvaes 7:23.: are OCH me 2128. treeteck 1;; 45-3343 (5.: ' ‘4 3~-~.‘.," ‘ f-e-ma.‘.a. “a r- e “p isi‘m “a l “Uni. x“ 9:1: -. Luca- 36 percent), cotton (6 percent), and sunflower (2 percent) (080, various years). Additional crops grown include bambara nuts, beans, cowpeas, soybeans, rice, as well as various fruits and vegetables. In Natural Region III, maize is most important, but small grains are also grown widely for consumption and cultural reasons. In Natural Region V, small grains are more important than maize, but most households still grow some maize. messes}: Livestock play an important role in insuring food security of rural households (Ndlovu, 1990). They provide draft power and manure for agricultural production; a source of cash-income (animal and product sales), a food source, and a store of wealth. Cattle dominates the number of ruminants owned by comunal households (64 percent); followed by goats (29 percent), sheep (5 percent), and pigs (2 percent). Households also raise chickens, ducks, and guinea fowl. 2.3.1.3 Consumption preferences There is a debate, with important policy implications, of whether households prefer maize or small grains. Rohrbach (1988) argues that communal households prefer maize since maize accounts for a large proportion of total farmed area, even in low rainfall regions. On the other hand, ENDA-Zimbabwe (1987) argues that households actually prefer small grains, but contend that households grow maize because processing small grains is labor-intensive. Furthermore, area allocated to small grains--which are better adapted to poorer agro-ecological regions-- would increase if appropriate hulling technology were available. 2.5.: ml 110 ll:::::': 1 :‘..'.".t::a‘. l ' .m- v, : . “assbil’ med :2: . n 1:“- m— — °- ‘21: . s- evdw) m " 1"“ F A; he“. I J a 'I 1 it: as ::e : I: 23937.: ‘ E." m... ‘ A, hie-......Ju M: 3:: 4v- Iiehfds ha 1'2 513w ; 32:22:: 5; :1: ‘9‘. 1 P h“! so. a. . “ 3‘ i Crztl 5952!! I: N. ‘ ~ $.~.' .v‘ ‘l-m :al An. ( .\ ‘1‘ 1’ ‘Us ”A"? :' u,A “Vs ' J57. d. 37 2.3.2 Rural household access to government services (since 1980) Although rural households had limited access to social and agricultural services before Independence, the government is committed to redressing these inequities (R02, 1982); and has significantly expanded rural access to services since 1980. lend In 1930, the Land Apportionment Act legalized the racial segregation of land. A desire to redress the historical inequitable distribution of land was one of the major reasons for fighting for liberation. At Independence, the government vowed to redress the inequitable land distribution by resettling 162,000 households on 10 million hectares of land by 1985 (Cusworth and Walker, 1988). By 1990, only 52,000 households have been resettled on 2.5 million hectares of land. The slow progress of resettlement has been due to both insufficient government funds to purchase land and few willing sellers, most of whom farm only in the lower potential zones. Thus, access to land continues to be a critical political issue”. W Before Independence, government extension served primarily the European farmers. In 1981, government established AGRITEX (Agricultural, Technical and Extension Services) by merging the Department of Conservation and Extension with the Department of Agricultural Development. Between 1980 and 1985, government expenditures for' extension. services increased, by 406 percent (C80, 1987), thereby lowering the extension to farmer ratio from 1:1000 to 1:850; and has set as a target a ratio of 1:400 (Eicher and Rukuni, 1990). 1"’For more information see Roth (1990), Blackie (1987), and Moyo (1987) . ' .7 ..V'Le - are ‘ C 1-.-.‘5N'a: re- he Depart—1% mantle for use: such a zzleu, seed figment-e, t menial far: :::eased by 3= D L ..11.": researc. ...; O ‘ S: H 7"“ e ...i I h... 1’. -4“; 6s..8 ass: Remy melt: " recanted. . 1&3; firmer: ”‘4'. ' Motscns; 2 . ) F? 3" "tabla it? Rude; ‘3 d ““1 in the de- § E‘s-t ‘3 ~¥I “.‘“ ‘ W K .4 39'9de 3“ a -. «.323 l O f '1 38 saw The Department of Research and Specialist Services (DR&SS) is responsible for agricultural research (crop, pastoral, livestock) and services such as regulation (plants and dairies), grading (meat and cattle), seed certification, and pesticide registration. Before Independence, these services almost exclusively served the needs of commercial farmers. Between 1980 and 1985, government expenditures increased by 35 percent (C80, 1987), and research was reoriented to address the problems of comunal farmers, including the establishment of on-farm research. Until 1980, agricultural research primarily addressed production constraints associated with the agro-ecological conditions of commercial farmers. Technology research focused on mechanization, hybrid seed (primarily maize), and management of chemical inputs. Since 1981, DR&SS has reoriented the agricultural research agenda to address the needs of comunal farmers by: l.) conducting varietal trials under communal area conditions; 2.) initiating a breeding program for sorghums and millets; and 3.) establishing a farming systems research unit to study, develop a FSR model, and provide information to assist policy makers. To date, new technologies (eg., hybrid sorghum and improved tillage methods) are still in the development phase (Shumba, 1990 and Mudimu, 1987). W The Department of Veterinary Services is responsible for prevention and control of animal diseases, including the cattle dipping. Before Independence, these were mainly available to commercial farmers. Since Independence, the importance of livestock in the poor agro- ecological areas has stimulated government to increase communal farmer access to veterinary services (Ndlovu, 1990). For example, between 1980 and 1985 government expenditures increased by 179 percent (C80, 1987) . Recently, AGRITEX began working with the Department of Veterinary 11:9! to 13°— gu (Eicher wary aerv‘ '~--.. -3 I -...'..'.':'a'. a.. 1’2 Agrlcalt mil: ted the 232113; 5ca::i 51': Iqu, a 75356! the ma: 3.2;;2: seed 532's, wheat, Li?- (1950), 5953? Ind: “time; new 213;); trees. 5333 6. 4w 0rde N ‘ _ Max:139 te 34:. 1957, .‘: .m ‘ ME. ‘39 1239: issued the 1 3;: fin). 39 Services to introduce animal health and management centers in communal areas (Eicher and Rukuni, 1990) to improve access and quality of veterinary services. 5911291§9£51_m3132£199 The Agricultural Marketing Authority (AMA) at the time of the survey coordinated the operations of the country's four marketing boards-~Grain Marketing Board (GMB), the Dairy Marketing Board (DMB), Cotton Marketing Board. (OMB), and 'the Cold. Storage Commission (CSC). These boards oversee the marketing of their individual commodities. Controlled crops include: seed cotton, cattle and sheep, milk and butter fat, maize, sorghum, wheat, groundnuts, soybeans, coffee, sunflower (1980), bulrush millet (1980), finger millet (1980), and edible beans. Before Independence, communal farmers had very limited access to marketing services; only three Grain Marketing Board depots served communal areas. Between 1980 and 1985, the government expanded market access in order to induce farmers to produce and market surpluses by constructing ten new depots and 55 collection points in communal areas (Muir, 1987). The government has attempted to increase rural incomes by offering incentive prices; for example in 1985/86, the government increased the sorghum (red and white) price by 120 percent (ZS80/mt to ZSlBO/mt). W Communal heuseholds had limited access to short term credit before 1980. In 1978, the Agricultural Finance Corporation (AFC) introduced a Small Farm Credit Scheme to promote agricultural by providing communal farmers credit for the purchase» of farms and agricultural inputs. Between 1980 and 1985, short-term credit extended to farmers increased by 142 percent (C80, 1987); primarily to communal farmers. :4. sense: 13:32:: amuse of t'.’ :esaeztl, cocr. 23:14.7 educatz mate: in a {3. .25 end, betweer mased by 133 ; 36m: 1%: i“- There has 31:; thus pen: 17:: 53 percent 32.12:; (350' 1? “0"" o ’vlerrum:: 40 £22i21_§grziseg Education Because of the liberation war and underinvestment by previous governments, comunal households have had limited access to primary and secondary education. The government, at Independence, declared that education is a fundamental right of every Zimbabwean (302, 1982). To this end, between 1980 and 1985 government expenditures on education increased by 130 percent. Between 1980 and 1985, primary education enrollment increased by 171%. There has been, though, an acute shortage of trained teachers. During this period, the number of teachers increased by 207 percent, but about 50 percent of the teachers in 1985 were temporary (untrained) teachers (CSC, 1987). The government is also committed to providing every primary school graduate at least four years of secondary school. Between 1980 and 1985, the shortage of teachers and school facilities was even greater for secondary school than with primary education. During this period enrollment increased by 628 percent. The government has attempted to address this problem by training more teachers; and in the meantime, they have resorted to double-sessioning the class rooms and have hired expatriots. Health Government has sought to increase health care services in the comunal areas. Between 1980 and 1985, government expenditures increased by 103 percent (C80, 1987) . Government efforts have focused on training village health workers 1981 and 1984, with 4,417 village health workers trained during this period; with a target number of 12,500 by 1993 (cso, 1987). I! :4 Introduction the survey 1119‘ mute the anal' mssed in Chap is: the :esea :LLectmn, and tr 3.2 km, ma T32 runey are: Iiémves and res 3'“ hunch m P: ‘I the gem hi. and deter: ‘5' ”Ht-"1. s 3»:le ”1d 30“} Therefi." a OVee, re! " m V. areas ( £315.! “Oils. t: 1!“ than 6' I“; rain: all '3‘ r99mm]. 13552: "":9 item 9% .. .Ca; m res«01 CHAPTER III SURVEY METHODOLOGY 3.1 Introduction The survey methods were designed to collect the data needed to estimate the analytical models required to address the research issues discussed in Chapter 1. This chapter describes the criteria used to select the research sites, the sampling frame, the mode of data collection, and the limitations of the data. 3.2 Survey area The survey areas were selected taking into account both the research objectives and resource constraints. 3.2.1 Research area selection criteria First, the general research objective was to assess the structure, level, and determinants of rural incomes in the more at-risk regions of the country; Secondary' data indicates that yields are lowest and production most variable in the poorest agroclimatic regions, defined by rainfall and soil characteristics (080, 1987; Rohrbach, 1988). Therefore, research sites (villages) were selected in Natural Regions IV and V, areas of the country with relatively less rainfall and less fertile soils. Rainfall averages 400-600 mm/year in Natural Region Iv and less than 600 nun/year in Natural Region v. Both regions have a unimodal rainfall pattern, distributed over only three to four months. The regional stratification provide a basis for assessing the differing income and expenditure pattern in the two contrasting agro- ecological regions. 41 42 Second, resource constraints required that the sites be located within a one-day drive of Harare, in order to enable supervision by University of Zimbabwe staff also engaged in teaching throughout the year. This constraint led to the identification of the Mutoko/Mudzi (NR IV) and Buhera (NR V) Districts. Subsequently, a rapid appraisal was conducted in these two districts with the assistance of local agricultural extension (AGRITEX) officials. Because almost no secondary data were available, key informant advice was relied upon to assess the variability in farm management practices, technology adoption, land allocation, marketing possibilities and non- agricultural activities across the wards in each district. This information was used to choose the wards and village for the more detailed survey work, using the criteria noted below. 1- W Crop production is a primary component of a household's ability to assure its own food security. Staple grains provide food for home consumption, with surpluses sold as cash crops to provide income to purchase food and meet other cash expenses. In addition, livestock sales are an important source of cash income, and store of wealth (Ndlovu, 1990). To assess differences in crop production across wards, all wards were evaluated in terms of the importance of maize; small grains (red and white sorghum, bullrush millet, and finger millet); and oilseeds (groundnuts and sunflower) .1 Consequently, villages were selected where households devote a high percent of their available land to small grains (millets and sorghums) and oilseeds. 1 Also, across these sites, livestock are of varying importance, but relatively more important in Natural Region V. 43 2. W. Previous research in comunal areas of Zimbabwe (Stanning, 1985 and Rohrbach, 1988) has shown that distance to market is an important determinant of a household's food security strategy. In particular, market access influences a households' production and consumption opportunities. Consequently, villages were chosen that range from 10 to 80 kilometers from the nearest Grain Marketing Board depot or collection point. 3. §ou£g§§ o; Off-Farm Income. Off-farm employment provides households with opportunities to improve their food security by supplementing agricultural income with wage earnings (Helmsing, 1987 and Chuta and Liedholm, 1979). Although not specifically selected with respect to this criteria, the sites have a diversity of opportunities for households to generate income through off-farm activities-- thereby providing an opportunity to analyze the role and contribution of off-farm employment on household food security. 4. P t w c . New agricultural technology can reduce food insecurity by both increasing and stabilizing crop yields (Waddington and Runjeku, 1988)--thereby enabling farmers to increase own production and their marketable surplus. The Department of Research and Specialist Services (DR&SS), Ministry of Agriculture, Lands, and Rural Resettlement (MLARR) conducts on-farm experiments, and the extension service (AGRITEX) managed on-farm demonstrations in rural areas. Therefore, four villages in NR Iv were selected near a set of Agritex trials/demonstrations with the expectation they' would. provide indicative ‘technical coefficients to assess the potential impact of new crop technology on improving household food security. 5- WW} Studies have reported that in recent years, households in low-rainfall areas have 44 increasingly substituted maize for small grains to meet food needs (BNDA-Zimbabwe, 1987). The primary reason cited was that the home-pro- cessing of small grains is more labor intensive, relative to maize. A site was selected near a small-grain dehuller to assess its impact on the role and uses of small grains in food security strategies of communal farmers. 6. o b r s. Public transfers, particularly food/cash provided through ‘food-for-work' programs in drought years, are an important means for improving household food security (Reut- linger, 1985). While not an explicit selection criteria, food-for-work programs have provided access to food in varying degrees across then villages selected. Table 3.1 shows the distribution of sites (villages) with respect to these six criteria. 3.2.2 Research Location The research was conducted in two survey areas: Mutoko and Mudzi Districts (140 kilometers northeast of Harare) and Buhera District (300 kilometers southeast of Harare). Figure 3.1 shows the location of the two survey areas. Although Mutoko District is in both Natural Regions III and IV, all research sites are in Natural Region IV. Mudzi District is entirely in Natural Region IV. Buhera District spans Natural Regions III, IV and V, but all the research sites are in Natural Region V. A total of 12 villages were selected, based on the criteria presented below. Six villages were chosen in each of the two survey areas. 3.3 Sampling procedures A multi-stage sampling procedure was used to select the household sample. The first stage involved the purposive selection of two natural regions (IV and V). The second stage involved the purposive selection of villages to insure diversity across selection criteria (see 3.2.1). 45 Table 3.1. Distribution of Research Sites with Respect to Selection Criteria and District, 1987-88, Zimbabwe. MUTOXO/MUDZI BUHERA CRITERIA 1 2 3 4 5 1 2 3 4 W919}! IV (400-600m) x x x x x v (Qcted to consume‘. Then, for households holding grain and oilseed ——\ 1 differhaximum grain and oilseed consumption levels were estimated report lTtly. Maximum grain consumption levels, based on FAO were ed consumption levels and discussion with key informants, “swedetimated at 300 kilograms per adult equivalent. It was given that households consumed grains sequentially: maize first-- craps its limited storeability and preference--then other grain estimgl The Zimbabwe dollar equivalent of maize was used as an lte of grain consumption except: In) If the amount of maize held by a household exceeded the oakimum consumption level, then HC was calculated as the value 23 300 kilograms of maize per adult equivalent, or 1:) If the amount of maize held by a household was less than V119. maximum consumption level, then HC was calculated as the ( alue of maize plus the Zimbabwe dollar value of other grains chillets, sorghums, and rice), not exceeding the maximum Hak9hsumption level. nuts, 111mm oilseed consumption levels (for groundnuts, bambara info Jtidney beans, and cowpeas) were established based on adult a‘tion provided by key informants. The maximum levels (per h equivalent) of these oilseeds are: groundnuts (Z$11.00) , 9118e§§ nuts (Z$2.50) , kidney beans (Z$1.50), COVpeas (Z$1.25) . Any 111C111 8 available to households below these maximum levels were agd as HC. HE Fim 4-1. w of total ml net household receipts. NET “a l Pnooucnou ,— son nous < coumnou CASN —- INCOIE < GENERAT ICNI ACTIVITIES —- TRANSFERS < -—— NET CREDIT RECEIPTS . GRAINS COISUIPTION OILSEEDS GRAINS INVENTORIES <——E OILSEEDS CRWS FARN SALES <——E LIVESTOCK NON -AGRI CULTURAL PRmUCTS AGRICULTURAL LAM SALES <-—C NON -AGR I CULTURAL BUSINESS INVENTUII ES OTHER FIX!) FOR MK GOVERNNENT < FM AID NUTRITIGI RENITTANCES (RELATIVES AUAY) OTHER PRIVATE (LWAL RELATIVES AND NEIGNBMS) LOANS RECEIVED (+) LOANS EXTENDED (') 57 stocks above this maximum, the surplus was transferred into the category, ending inventories. HC was estimated as the actual monetary value of the grain and oilseeds retained (net of outflows) for households that produced less than these maximum levels. . ' II: o s Receipts from cash income-generating activities (CIGA) is the income earned by household members (net of intermediate goods and services), regardless of whether the payment is made in cash or farm productsz. The five cash income-generating subcategories are: 1. ) Farm sales are cash receipts from sales of both crops and livestock sales. 2 . ) Non-agricultural product sales are cash receipts from the sale Of home-produced non-agricultural goods. 3 . ) Labor sales are cash or cash-equivalent (farm products) receipts from agricultural and non-agricultural labor sales by resident household members. 4.) Business inventories are grain and oilseed stocks that were Originally purchased for resale, but remained in inventory at the end of the year. 5.) Other income is cash and cash-equivalents (farm products) received by household members for other reasons (e.g., faith healing). In estimating CIGA, three types of intermediate goods and services (IGS)--expenditures made to purchase goods and services used in the pm> AVAILABLE < POP > OONSINIPIION l l I. l l NOUSENOLO LABOR CLOSING GIFTS LOANS rOOO OONSLBIPIION PAnIENIS INVENIOPY 66 1.91. 6.1. Net mu receipts by villus, district, and total sq». (3'), Make, mm, It! We Districts, Zim, 1 ...—— Distrifl/ SAMPLE PER HOJSEHOLD PER CAPITA PER ADULT EQUIVALENT village SIZE (11) MEAN SE MEDIAN MEAN SE MEDIAN MEAN SE MEDIAN Tm... 1 25 2395 862 1588 390 81 265 558 122 329 2 23 1056 167 882 282 127 110 356 151 168 3 20 899 150 786 139 23 116 189 28 167 6 26 799 119 626 116 15 97 172 22 156 5 23 1222 207 966 222 67 123 336 81 181 6 21 1328 312 769 175 32 117 275 56 175 District total 136 1306 178 872 225 29 125 320 39 176 MUtOkO/Nudzi 1 26 992 161 766 227 52 166 307 70 206 2 29 1657 136 1 179 296 26 263 379 36 335 3 15 689 62 506 161 36 161 227 55 190 6 27 720 178 696 115 20 93 153 26 116 5 29 706 65 797 126 - 12 100 181 19 165 6 23 1191 126 1016 229 29 169 339 66 263 District total 169 957 61 795 196 16 169 266 19 213 s“We total 285 1123 91 819 209 15 139 292 190 213 $0"We: Food Security surveys. ‘2:1.OO - US$0.60 I’Differences in District means were tested for statistical significance at the 1 and 5 percent level- NO differences in means were statistically significantly different. 67 Mean NHR‘ (per capita) averaged 25209 for the total sample, 25196 for ”utoko/Mudzi Districts (NR IV), and 25225 for Buhera District (NR V), although the district differences are not statistically significant. By comparison, Stack (Stack and Chopak, 1990) reported mean incomes (per capita) of 25260 and 25100 for Shamva and Binga Districts, respectively. The C50 (1988) reported mean incomes (per household) of 252,992 and 251,237 for Mashonaland East and Manicaland Provinces, respectively; which, based on an average household size, represents 25496 and 25196 per capita, respectively. The Mutoko/Mudzi and Buhera District estimates are similar to Stack's and the CSO's results since Stack's estimates are for a higher rainfall area (Bushu) and a poor rainfall year in a similar agro-ecological area (Binga) ; and the CSC estimates are provincial averages, which incorporate areas of higher rainfall than the survey area. Although mean NHR (per capita and per adult equivalent) were slightly larger in Buhera District households than Mutoko/Mudzi District, these differences were not statistically significant (5 percent level). These results were unexpected since it was hypothesized that Hutoko and Mudzi Districts would have the larger mean NI-IR because of its more favorable resource base and stable rainfall pattern. On the other hand, median NHR (Per capita and per adult equivalent) were larger for Mutoko/Mudzi mJitrict than Buhera. Since distributions of net household receipts are b°th highly skewed and peaked, the median is a more reliable measure of centr'al tendency. Within districts, there are large inter-village differences in median \ differe skewness other Stu ("The discussion of district and total sample incomes uses the the measure of central tendency. There are small IIces between the mean and median, which implies little of the data; also, the mean facilitates comparisons with dies. 68 EU. Median.per household NHR range from 25626 to 251,588 in Buhera; and 196 to 251,179 in Mutoko/Hudzi. Median per capita NHR range from 2597 25265 in Buhera; and 2593 to 25263 in Mutoko/Mudzi. Median per adult uivalent NHR range from 25167 to 25329 in Buhera; and 25114 to 25335 in toko/Mudzi. Except for small differences, the rank order of villages mains the same for all three income measures. 2.2 Distribution of net household receipts The distribution of net household receipts (NI-IR) are assessed by alyzing the distribution across quartiles, their symmetry and equality. 1.2.1 Income quartile distribution The distribution of households across NHR (per capita) quartiles is asented in Table 4.26. The differences between villages and districts 1 terms of the percent of households within income quartiles) is riking. First, villages range from having almost a majority of useholds in the lowest quartile (village 4 in Mutoko/Mudzi), to having at households in the upper quartile (village 1 in Buhera and village 2 Mutoko/Mudzi). Second, most villages in Buhera (excepts for village 1) VS a majority of households in the lower two quartiles; and most Llages in. Hutoko/Mudzi (except for villages 4 and 5) have most lBeholds in the upper two quartiles. These differences are discussed “ther in Chapter 6. ¥ Le discussion of village incomes uses the median as the of central tendency. The large difference between the mean median imply skewness in the data, and therefore the median ippropriate measure of central tendency. lt household receipts was chosen to determine quartiles it gives the most comprehensive income definition. table 6.2. Distrilution of Models (peront) m net MIC receipts mmrtiles by village lid 69 district (18), mtoko, Mi, mid liners Districts, litmus, 1m. F" PER CAPITA NNR WARTILES District] SWLE vi l lace SIZE LINER LMR-MIDDLE UPPER-MIDDLE UPPER (n) ( < 885 ) ( S85 - S139 ) ( $139 - 5263 ) ( > 5263) Ethan 1 25 6 20 26 52 2 23 30 22 30 17 3 20 35 35 5 25 6 26 29 S6 13 6 S 23 35 22 9 35 6 21 29 26 26 26 District total 136 26 29 18 26 Mutoko/Mudzi 1 26 19 27 27 26 2 29 7 3 28 62 3 15 33 13 60 13 6 27 66 33 15 7 5 29 36 31 31 3 6 23 6 17 52 26 District total 169 23 21 31 26 Same total 285 25 25 25 25 Source: Food Security surveys. 7O 4 .2.2.2 Sy-etry of per capita net household receipts The measures of central tendency and symmetry (skewness and kurtosis) :Lndicate per capita net household receipts (NI-IR) are asymmetrically distributed across villages, districts, and the total sample (Table 4.3) . The distribution of mm (per capita) is highly and positively skewed (tail to the right) in all villages, districts, and for the total sample, ranging from 0.6 to 4.4 across villages7. High positive skewness, characteristic of income data, indicates that for a majority of households, their NHR (per capita) are below the mean (a few households with large incomes are skewing up the mean). On average, positive skewness is higher in villages in Buhera District than in Mutoko/Mudzi Districts. In all but two villages (one in each district) the distribution of NHR (per capita) are highly peaked (kurtosis), ranging from 0.331 to 20.4618. High positive peakedness indicates that households are concentrated in a narrow income band at the lower end. The two villages with a negative kurtosis value, also have less skewness--which implies a more symmetric distribution of mm (per capita) in these villages. 4-2.2.3 Equality of net household receipts (per capita) All three measures of equality--coefficient of variation, the standard deviation of the natural logarithm of income, and the Gini coefficient-- indicate considerable differences in income distribution (Table 4.4). Mel sample results All inequality measures indicate a large inequality in mm (per caPita). For the total sample, the Gini coefficient is 0.4689; the °°°fficient of variation is 1.2488; and the standard deviation of the \ 7 O O Skewness is zero for a normal distribution. 8K1lit-tosis is zero for a normal distribution. 71 Isle 6.3. last. of ontral turkey mid sy-etry for net hmmahold receipts ( emits) by villus, district, all sqle, Mote, m1, I‘ll liters Districts, 21m, 1 . District] SANLE village SIZE NEAN NEOIAN SD SE SNENNESS KURTOSIS (n) were 1 25 390 265 1.01 81 2.525 7.563 2 23 282 110 609 127 6.631 20.661 3 20 139 116 96 22 .659 (.392) i 21. 116 97 75 15 2.000 5.261 5 23 222 123 221. 1.7 1.1.91. 2.169 6 21 175 117 11.7 32 1.370 1.1.31 District total 136 225 125 335 29 5.1.66 39.063 leutoko/hudzi 1 26 227 11.6 261. 52 2.767 7.621 2 29 296 263 138 26 .976 1.560 3 15 161 11.1 11.0 36 2.553 7.976 I. 27 115 93 101. 20 2.1.55 6.1.66 5 29 121 100 61. 12 .553 (.801) x 6 23 229 169 138 29 1.191 .311 -\ District total 169 196 11.9 166 11. 2.621 10.162 \ s-Plo total 265 209 139 261 15 5.959 52.860 \ smrc e ‘11. e. Food Security surveys. lMes in parentheses are negative users. 72 Isle 6.6. Distrilutim of net Mid receipts (par emits) by vi llqe, district, mud sqle, make, Mi, mu! liners Districts, 2i”, 1%. District] SANPLE CINI COEFFICIENT COEFFICIENT STANDARD OEVIAIION Village 6126 or VARIATION or THE NATURAL LOG (11) OF 1110016 Bthera 1 25 .6632 1.0332 .3676 2 23 .6376 2.1596 .5672 3 20 .3775 .7050 .3507 4 26 .3060 .6666 .2516 5 23 .5079 1.0090 .5506 6 21 .6300 .6600 .6623 District total 136 .5257 1.6889 .6689 Hutokolfludzi 1 26 .6772 1.1630 .3626 2 29 .2660 .6696 .2263 _3 15 .3269 .6696 .3296 4 27 .6056 .9063 .3300 5‘ 5 29 .2655 .5161 .2676 x 6 23 .3083 .6026 .2619 \ District total 169 .6060 .6557 .3661 ‘ sum: total 285 .6689 1.2666 .6073 \ so’~"‘¢=e: Food Security surveys. 73 natural log of income is 0.4073. District level differences All three measures of the distribution of NHR (per capita) are larger for Buhera District than for Mutoko/Mudzi Districts, indicating greater inequality in Buhera. Inter-village variability In terms of inter-village variability, there are three important results. First, NHR are unequally distributed. For example, the Gini coefficient demonstrates that two of the 12 villages have an high level of inequality, two are relatively high, one is moderate, and seven have a low level of inequality9. Second, the degree of inequality varies considerably across villages. For example, the Gini coefficients for Buhera District villages ranged frtm10.3080 to 0.6376, and from 0.2460 to 0.4772 for Mutoko/Mudzi; the CVs fOrBuheraDistrict villages ranged from 0.6466 to 2.1596, and from 0.4696 to 1.1630 for Hutoko/Mudzi; and the SDLs for Buhera District villages renged from 0.2514 to 0.5672, and from 0.2243 to 0.3824 for Mutoko/Hudzi. Finally, all three measures of NHR inequality provide generally the Same ranking of villages within each district, except for two villages in Buhera District. In one of these villages (#1), their relative ranking among villages for their CV is higher than their Gini, but smaller for their SDL. This ranking switch implies a relatively greater inequality in theimiddle income range for that village than other villages.‘ Conversely, in village 5 the SDL is relatively larger than both the Gini coefficient and CV, implying a relatively larger inequality in the lower income range in that village. h ’FAO (1986) defines a low amount of income inequality as .aVing a Gini coefficient less than or equal to 0.41; moderate inequality between 0.41 and 0.45; relatively high inequality tween 0.46 and 0.50; and a high inequality if greater than 0.50. 74 6.2.3 Sources of net household receipts This section discusses the magnitudes and contribution of different income sources to net household receipts (per capita). After presenting the sources of NHR across villages and districts, these sources are analyzed across mm (per capita) quartiles. 4.2.3.1 Structure of m by village and district The three major components of net household receipts (per capita) are: earned income (production for home consumption and cash income—generating activities (net of intermediate goods and services), transfers received, and net credit receipts (Table 4.5). Disaggregated net household receipts In terms of the relative importance of these major sources, three POints stand out. First, for all villages in both districts, earned income accounted for the major share of NHR--ranging from 88 to 99 percent in Buhera; and 69 to 93 percent in Mutoko/Mudzi‘o. Second, transfer income (transfers received) was large for the total sample (15 t of mm (Per capita)) , but more important in Mutoko/Mudzi District. For example, transfers accounted for over 256 of mm in only one village in Buhera In~81:ri.c:t, but for over 25% in 5 of 6 villages in Mutoko/Mudzi villages. Finally, although net credit receipts were less than 10 percent of NHR (except in one village in Buhera District), they were generally negative-- indicating a credit burden. Disaggregated earned income The following sections explores the subcomponents of earned income (Table 4.6). rec . 10Note: The cumulative percent of earned income and transfers ne e1Ved can exceed 100 percent because many households had a gative outflow (therefore negative percent) of credit. 75 this 6. 5. We of mm) M hOIdeold receipts (per mite) by mos, village, as! district, Make, Mi, me! there Districts, lime, 1W. TRANSFERS NET CREOIT PER CAPITA District] SANPLE EARNED INCONE RECEIVED RECEIPTS NNR village SIZE (1)) ANOONT PERCENT ANOONT PERCENT AMOUNT PERCENT ANOONT PERCENT <26) 12:) (2:) 12:) Buhera 1 25 367 99 36 9 (30) 1 6) 390 100 2 23 269 95 11 6 2 1 262 100 3 20 122 66 35 25 (18) 113) 139 100 6 26 115 99 10 9 1 9) ( 6) 116 100 5 23 216 96 11 5 ( 2) ( 1) 222 100 6 21 169 97 16 6 1 6) 1 5) 175 100 Dist rict total 136 3g 96 111% 6 (2;) ( 5) 225 100 Moholludzi 1 26 156 69 71 31 1 (<1) 227 100 2 29 239 61 56 19 2 1 296 100 3 15 136 66 36 26 (15) ( 9) 161 100 6 27 107 93 9 6 1 1) 1 1) 115 100 5 29 96 77 26 21 1 1 126 100 6 23 176 77 66 29 (12) I 5) 229 100 '\ Dim-m total 6 153 79 23 3 1 2) 196 100 __ LL 1.6 in? Sept.- total 285 163 88 32 15 ( 7) 1 3) 209 100 g sz‘ce: Food Security surveys. 'Differences in District mans were tested for statistical significance at the 1 C") and 5 9') percent level. this 6.6. 76 Make, “1, “MDistricts, If”, 1 Parent cmtrihutia) to eemed inc. (per mite) by village, district, PRODUCTION FOR CASH INCONE GENERATING ACTIVITIES District] WLE H015 001151”th village SIZE NONE INVE!T- EARN NON-AG LABOR BUSINESS OTNER k (n) OONSLNE l SALES I SALES | SALES lINVENTORv| CASH 1 Mere 1 25 9 66 10 6 21 2 1 2 23 11 63 16 2 1 3 1 3 20 22 32 19 6 6 16 2 6 26 25 60 6 9 19 3 1 5 23 12 36 5 <1 62 1 <1 6 21 7 6 9 16 53 3 6 District total 136 12 62 11 6 .241. 3 2 Mutoko/“2i 1 26 17 31 32 7 6 1 3 2 29 16 56 16 3 6 <1 <1 J 15 22 51 15 9 2 <1 0 6 27 26 26 25 9 16 1 o 5 29 30 67 <1 5 16 1 <1 \6 23 17 36 16 16 16 <1 1 fltrict total 169 20 63 16 7 11}, 1 1 Staple total 285 15 63 13 7 17 2 2 i_ s“Mme: Food Security surveys. mu! sqle, ‘Differences in District values were tested for statistical significance at the 1 (‘1') and 5 0") perc t level. For a discussion of the inventory category, see footnote 11. _-( 77 Min First, production for home consumption was the most important component of earned income, accounting for 58 percent of earned income; of which 43 percent was held as inventories11 and 15 percent was assumed to be home consumed. Second, cash income-generating activities accounted for 42 percent of earned income; of which 17 percent was from labor sales, 13 percent from farm sales, 7 percent from non-agricultural product sales, and 4 percent from business inventories and other cash sources. t c v co so In both Districts, three similarities stand out. First, production for home consumption accounted for over one-half of earned income; 54 percent in Buhera District and 63 percent in Mutoko/Mudzi. Second, farm and labor sales accounted for a two-thirds of earned income from cash income- 9enerating activities (CIGA). Finally, non-agricultural product sales, business inventories, and other cash income sources contributed little to earned income; ranging from 2 to 6 percent in Buhera District and from 1 t9 7 percent in Mutoko/Mudzi. Two major inter-district differences between production for home °°nsumption stand out. First, production for home consumption accounted for a larger share of earned income in Mutoko/Hudzi than Buhera (63 Percent versus 54 percent), due to greater estimated consumption from home Production (20 percent versus 12 percent). The percent contribution of \ 11The large mean inventory holdings--as a proportion of per pre a earned income--is misleading. Means values were used to capgent the percent contribution of different subcomponents to per de lta earned income. Analysis of the data indicates a large ingree of skewness and peakedness in the level of total and grain me d?htories (per capita) held by households, suggesting that the tenlan' not the mean, is the most appropriate measure of central 1 dency. An examination of median grain inventories reveals that n 9111}? two villages did households hold more than 1.5 bags of graln per capita. Furthermore, 73 percent of the total sample held 398 Of grain or less. For a more thorough discussion see APpendix 5. Cdp it: 78 inventories was similar between the two districts (42 percent in Buhera and 43 percent in Hutoko/Mudzi) . Second, although farm sales and labor sales were the major source of earned income in both districts, farm sales were more important in Mutoko/Mudzi (18%) and labor sales were more important in Buhera (21%). In contrast, non-agricultural product sales accounted for a similar percentage (7 percent) of earned income in both districts, and both business inventories and other cash sources were minor sources of earned income . Was Across villages, the production for home consumption share of earned income varied greatly; ranging from 11 to 77 percent. Among subcategories, home consumption ranged from 7 to 30 percent and inventories ranged from 4 to 63 percent (Table 4.6). The three most important cash income-generating activities in all villages, were farm sales and labor sales; and in two villages non- figricultural product sales. Business inventories and income from other cfiflh sources were small. Conversely, the contribution of CIGA was e“til-emely variable across villages, ranging from 23 to 89 percent. Among a‘lhcategories, the contribution. of labor sales ranged from 1 to 53 Percent, farm sales ranged from <1 to 32 percent, non-agricultural product 3also ranged from <1 to 18 percent, business inventories ranged from <1 to 14 percent, and other cash sources ranged from <1 to 6 percent. 41.2.3.2 Structure of net household receipts, by income quartiles Across net household receipt (per capita) quartiles, four similarities Etand out (Table 4.7). First, across all income quartiles, earned income contributed the largest share of mm (87 to 92 percent); which is composed Of production for home consumption (50 to 62 percent) and cash income- generating activities (32 to 37 percent). 79 tile 6.7. Stu-mm of net hand-old receipts (per ceita) by item qmrti le (25), Make, lhallzi, ml! Mere Districts, line, I”. LGER LINER MIDDLE lN’PER NIDDLE UPPER NET NGJSENOLD ( < 2385 ) ( 2385 - 23139 ) ( 23139 - 23263) ( > 23263 ) RECEIPTS MEAN PERCENT MEAN PERCENT MEAN PERCENT MEAN PERCENT EARNED INCOIE PNC 33 a 62 56 a 52 95 a 51 261 b 50 CICA 17 a 32 60 ac 37 69 be 37 180 bed 37 TOTAL 69 a 92 % d) W 166 b 88 622 c 87 EARNED INCGE TRANSFERS 8 a 15 16 at 15 32 he 17 n bf! 15 NET CREDITS (6) ( 8) ( 5) ( 5) ( 8) ( 5) (10) ( 2) TOTAL NNR (pt) 53 I 100 107 at 100 187 be 100 686 Ind 100 Saree: Food Security surveys. ‘ Dmcan's Multiple Range test was used to assess the statistical significance of the difference of scam, men there are three or more grows (means). where that are statistically different (5 percent level) across qmrtiles have different letter“) assigned to them. No letter after a nulber signifies that there was no statistically significant difference across martiles. 80 Second, as expected, production for home consumption accounted for a larger share of NHR (per capita) for the lowest quartile (62%) than for the higher quartiles (50 percent for the highest quartile) households. Third, the share contribution of transfers was similar across income quartiles, ranging from 15 to 17 percent. Finally, net credit receipts were, on average, negative and small across quartiles (8 percent). This shows a higher repayment burden for the poorer households. 1 Earned income (by income quartile) is further disaggregated for additional analysis (Table 4.8). 4.2.3.3 Structure of earned income, by income quartiles Across NHR quartiles, three important points stand out (Table 4.8). First, for all income quartiles production for home consumption (home consumption plus inventories) accounted for a similar share of earned income (57 to 67 percent). On the other hand, as NHR increase, households required a smaller share of income to meet recommended consumption levels. This is in part because an upper bound was placed on the maximum level of per capita consumption; but more importantly, as income rose the level of inventories significantly increased as a percent of earned income. Second, for all income quartiles cash income-generating activities (CIGA) accounted for a similar share of earned income (43 to 35 percent); and the contribution of individual components of CIGA.were similar across quartiles. In contrast, farm sales constituted a much larger share of earned income for the highest three quartiles. Surprisingly, labor sales constituted a larger share of CIGA in the highest income quartile (19 percent) than for the lowest quartile (14 percent). Third, a review of potential farm sales (Table 4.9)12 shows that for “Potential farm sales equals total crop and livestock sales, plus inventories. Inventories are included as sales because they represent a reserve grain and oilseed stock that has not been consumed and is available for sale. 81 Tdale 6.8. Strata-e of earned inc. (per smite) by incm q-rtile (23), Make, Mi, dd Buhera Districts, line, 1%. LGER ( < 2385 ) as!» PERCENT LINER MlDDLE (2385 - 23139) MEAN PERCENT UPPER MIDDLE (23139 - 23263) MEAN PERCENT UPPER ( > 23263) MEAN PERCENT mun HI I: WT!“ m COISllIED 23 67 29 30 30 18 32 8 INVENTMY 9 a 18 28 ab 29 65 b 60 210 c 50 TOTAL PK 33 67 56 58 95 58 261 57 CASM Im- EERATIH ACT. FARM SALES 3 a 6 13 ac 13 23 be 16 63 bed 15 NOIAG WTS 6 a 8 7 ab 7 17 b 10 23 c 5 LA“ SALES 7 a 16 15 a 16 26 a ’ 16 81 b 19 BUSINESS NV 2 6 3 2 1 1 8 2 OTHER CASH 2 6 2 2 2 1 5 1 TOTAL CTGA 17 35 60 62 69 62 m 63 TOTAL em I“ 69 a 1m % d: 1” 166 b 100 622 c 100 :ource: Food Security surveys. ‘ Dmcan's Multiple Range test was used to assess the statistical significance of the umbers that are statistically difference of means, men there are three or more craps (means). different (5 percent level) across quartiles have different letter(s) assigned to them. No letter after a usher signifies that there was no statistically significant difference across martiles. 82 Tine 6.9. ”Win of pot-Rial fare aales' arm/livestock) by inn-e quartile (inclldim 1min), Intake, sum, and mu Districts, zine-hue, was/a9”. LOaER LINER MIDDLE UPPER MIDDLE UPPER POTENTIAL FARM ( < 2385 ) (Z385 - 23139) (23139 - 23263) ( > 23263 ) SALES MEAN PERCENT MEAN PERCENT MEAN PERCENT MEAN PERCENT CI? TOTAL GRAINS 12 a 50 29 a 55 61 b 55 215 c 66 OILSEEDS 6 a 25 13 a 25 31 a 28 75 b 23 FRUITSSVEG 1a 6 1a 2 6a 5 17b 5 COTTGI 1 6 <1 <1 1 1 1 <1 CRO SALES 20 a 83 63 a 81 99 b 89 303 c 96 LINESTGK SALES 6 17 1D 19 12 11 19 6 1 POTENTIAL FAQ 26 a 1m 53 a 100 111 a TN 327 c 100 , SALES Source: Food Security surveys. ' For a definition of potential fare sales see footnote 13. Dmcan's Multiple Range test was used to assess the statistical significance of the difference of means, then there are three or more grows (seam). Nunbers that are statistically different (5 percent level) across quartiles have different letter(s) assigned to them. No letter after a number signifies that there was no statistically significant difference across quartiles. 83 all income quartiles, grain sales accounted for the largest component of farm sales (50 to 66 percent), followed by oilseed sales (23 to 28 percent), and livestock sales (6 to 19 percent). While important for individual households, fruit and vegetable and cotton sales contribute relatively little to farm sales across all income quartiles. 4.2.4 Monetisation of households As household income increases, economists observe that households earn a greater share of income from cash sources, making it possible for households to invest in agricultural and non-agricultural capital which stimulates further household income growth (von Braun and Kennedy, 1986 and Matlon, 1977). To test this hypothesis, the contribution of 1.) cash (and non-cash income) to total per capita income and 2.) the composition of cash income (non-farm and farm)” were estimated. First, as per capita incomes increase, the cash share of income increased substantially, and the non-cash share decreased (Table 4.10). Second, for all income quartiles farm sales represented the largest share of cash income per capita (44 to 62 percent), although the contribution of different cash sources varied between income quartiles. For the lowest NHR.quartile, farm sales were the largest component of cash income (44 percent), followed by transfers (26 percent), labor sales (19 percent), and net credit receipts (11 percent); and net credit receipts represented a large negative outflow of cash income. In contrast, for the highest NHR quartile, farm sales were also the largest component of cash income (62 percent); but labor sales (18 percent) were the second most important source, followed. by transfers (14 percent). Net credit receipts, business inventories, and other cash income contributed little to per capita cash income. 13Cash income is the summation of all receipts received by household members from all cash sources. Non-cash income is the summation of all income received by household members in-kind. 84 lists 6.10. Cash veni- noncash incue' by net Mid receipts (per chita) q-rtiles, Make, Mi, Id Illnera Districts, 21m, 19mm. LowER LowER MIDDLE UPPER MIDDLE UPPER CASN VERSUS NCNCASN mm 1 < 2385 ) (2385 - 2:139) (2:139 - 2:243) ( > 2:243 ) MEAN PERCENT MEAN PERCENT MEAN PERCENT MEAN PERCENT ToTAL Tacos moans NON-CASN 31 53 I.I. 36 59 28 123 22 CASII 27 47 77 6!. 154 72 61.0 78 moans or CASII Tacos FARM SALES 12 u. I.1 53 89 58 273 62 NON-ASRT SALES I. 15 7 9 17 11 23 s LABOR SALES 5 19 13 17 26 17 78 18 BUSINESS INVENT. 2 7 3 I. 1 1 8 .2 OTHER CASII 1 I. 2 3 1 1 6 1 TRANSFERS 7 26 TI. 18 29 19 63 14 NET CREDITS (+/-) I 3) (11) I I.) ( 5) ( 8) ( 5) (10) ( 2) TOTAL CASN TNcoTE 27 100 154 100 440 100 urT ty surveys. ' For a definition of cash and noncash income, see footnote 13. 85 Finally, although transfers constituted a large share of cash income for the lowest quartile (26 percent), they were also quite large for the other three quartiles (18, 19, and 14 percent, respectively). 4.3 Analysis of household expenditures: empirical results This section analyzes both annual household expenditure levels and patterns in Mutoko/Mudzi and Buhera Districts. The mean is used as the measure of central tendency for expenditures at the sample and district level; while the median is presented for the village level analysis. 4.3.1 Household expenditure levels Mean and median expenditure levels were estimated per household, per capita, and per adult equivalent--by village, district, and the total sample (Table 4.11). t v tur For all three measures (per household, per capita, and per adult equivalent), mean expenditures were larger in Buhera than in Mutoko/Mudzi District. This result is consistent.with.earlier analysis which indicated that estimated incomes--for similar measure of income--were larger in Buhera than Mutoko/Mudzi District. For the total sample, expenditures per household averaged 25839; compared to 25963 for Buhera District and 25725 for.Mutoko/Mudzi District. Expenditures per capita averaged 25149; compared to 25156 for Buhera District and 25143 for Mutoko/Mudzi District. Expenditures per adult equivalent averaged 25210; compared to 25222 for Buhera District and 25199 for Mutoko/Mudzi District. 8GB Tdale 4.11. Ewes-ea by villne, district, III sqle (23), Make, Mi, lid “Tera Districts, 21m, Toss/89'. DIStI‘ICt/ SAMPLE PER WSEIIOLD PER RESIDENT PER ADULT EWIVALENT village SIZE (n) MEAN SE MED IAN MEAN SE MED I AN MEAN SE MED IAN Buhera 1 25 1221 168 855 226 27 173 329 47 250 2 23 995 87 1130 174 22 141 229 30 175 3 20 946 134 975 165 34 100 219 41 160 4 24 886 231 559 127 29 88 187 41 130 5 23 613 67 579 99 13 77 144 21 108 6 21 1111 198 764 141 19 117 218 35 184 District total 136 1E5: 66 760 156 11 114 222 16 160 MutokoflMudzi 1 26 1224 156 872 262 56 195 355 75 254 2 29 434 45 371 86 10 65 111 14 84 3 15 564 69 492 174 30 129 237 41 197 4 27 378 40 334 61 5 60 81 7 77 5 29 560 53 593 99 7 87 145 13 121 6 23 1201 208 969 216 33 149 316 47 224 District total 149 E 53 584 143 13 102 199 18 145 Sample total 285 839 43 658 149 8 107 210 12 151 Source: Food Security surveys. 'Differences in District means were tested for statistical significance at the 1 (**) and 5 (*1 percent level. 87 e -v n ur s The expenditure data analysis, as with incomes discussed earlier, confirms the hypothesis that expenditure levels vary considerably between villages, although the rank order of villages was generally the same for all three measures. Expenditures per household ranged from 25613 to 251221 in Buhera District and 25378 to 251224 in Mutoko/Mudzi; expenditures per capita ranged from 2599 to 25226 in Buhera District and 2561 to 2526241n Mutoko/Mudzi; expenditures per adult equivalent ranged from 25144 to 25329 in Buhera District and 2581 to 25355 in Mutoko/Mudzi. 4.3.2 Household expenditure pattern This section examines the levels and composition of expenditures (per capita) to identify policy interventions to raise the level of available income--through increasing incomes, or reduce specific expenditures to increase available income. 4.3.2.1 Household expenditures by village and district Expenditures (per capita) are grouped into three categories-- consumption, investment, and transfers granted-~with mean values estimated by village, district, and for the total sample (Table 4.12). s t on s For both districts, the composition of expenditures per capita were similar. For the total sample, consumption represented 71 percent of expenditures (70 percent in both Buhera and Mutoko/Mudzi Districts); investments represented 26 percent (26 percent in Buhera District and 24 percent in Mutoko/Mudzi); and transfers represented 4 percent (4 percent in Buhera District and 5 percent in Mutoko/Mudzi). 138 Tble 4.12. Statute of Mitts-es (per cqaite) by village, district, as! total sqle (ZS), mete, Mi, lid Idlera Districts, 21m, 1”. D i striCtI SAMPLE CGISINIPT IGI INVESTMENT TRANSFERS vi l loge SIZE (n) WT PERCENT AWNT PERCENT ”MINT PERCENT Buhera 1 25 154 68 62 27 11 S 2 23 119 69 58 30 1 1 3 20 99 60 62 38 3 2 4 24 87 69 29 23 11 9 5 23 81 81 14 14 4 4 6 21 114 80 23 16 4 3 District total 136 110 7D 41 26 6 4 Mutoko/Mudzi 1 26 170 65 58 22 34 13 2 29 64 74 22 26 <1 <1 3 15 125 71 49 28 <1 <1 4 27 52 85 9 15 1 2 5 29 77 77 19 19 2 2 6 23 145 68 70 32 1 <1 District total 149 101 70 35 24 7 5 Swle total 285 105 71 38 26 6 4 Source: Food Security surveys. 'Differences in District means were tested for statistical significance at the 1 (**) and 5 (*) percent levels. No means were statistically significantly different. 89 As expected for low rainfall agricultural areas, consumption dominated household expenditures. Yet, households also invested a significant share of their income“, and subsequent analysis will show that these investments were mostly made to pay school fees expenses. Although transfers given to others represented up to 13 percent of expenditures across villages, they were not an important expenditure in most villages. - t d tur As was the case at the district level, the analysis confirmed that consumption expenditures dominated total expenditures in all villages; but the composition of these household expenditures varied greatly between villages. Consumption expenditures (per capita) in Buhera District ranged from 60 to 81 percent and in Mutoko/Mudzi from 65 to 85 percent. Investment expenditures in Buhera District ranged from 14 to 38 percent and in Mutoko/Mudzi from 15 to 32 percent. Transfers in Buhera District ranged from 1 to 9 percent and in Hutoko/Mudsi from less than one to six percent. 4.3.2.1 Expenditures by not household receipts quartiles The analysis of expenditures by net household receipts (per capita) quartiles provided three interesting insights into the relationship between income levels and the composition of household expenditures (Table 4.13) . As incomes rose: 1.) consumption expenditures as a percentage of total per capita expenditures fell15 (82 to 63 percent); 2.) investment 1"School expenses are often considered a consumption expenditures. This analysis classifies them as investments since households perceive them as a way to increase future income. ‘EAlthough home-consumed.production has an assumed upper bound (see footnote 1, p. 55), its influence is small. For the lower quartile, home-consumed production represented 37% of consumption expenditures, which is an accurate estimate since these households had small inventory levels. For the upper quartile, home-consumed production represented only 20% of consumption expenditures. 9O expenditures rose (16 to 30 percent); and 3.) transfers given rose (1 to 7 percent). Analysis of the disaggregated expenditure components points out three important patterns. First, for all income quartiles, food and clothing purchases dominated both total (54-69 percent) and consumption expenditures (84-87 percent). Second, as income rose, absolute expenditures on education rose; the absolute level is relevant because households had similar numbers of school-age children. Third, as incomes increased, agricultural production investments rose--both absolutely and relatively; which gave wealthier households a larger capacity to produce. Finally, as incomes increased, transfers given rose, but were a small share of expenditures for all quartiles. 4.4 Summary This section presents an overview of the empirical analysis of the level, distribution, and composition of incomes and expenditures. This profile of household incomes and expenditures will be used to generate hypotheses that will guide the bivariate and multivariate analysis presented in Chapters v and IV, respectively. Incomes Estimated mean not household receipts (per capita)--25209 for the total sample, 25194 for Mutoko/Mudzi Districts, and 25225 for Buhera District-- are consistent with results reported in studies conducted in Zimbabwe by MLARR (1990), the Central Statistics Office (1988), Stack (Stack and Chopak, 1990), Amin (1990), and Govaerts (1987) under similar agro- ecological conditions. Although mean NHR (per capita) were larger in Buhera District, median mm (the more appropriate“ measure of central tendency) were larger in Mutoko/Mudzi District. The distribution of households across net household receipt (per capita) quartiles varied greatly across villages. For example, in each 91 Title 4.13. Smith» of esp-flutes (par emits) by inc. gmrtile, mtoko, Mi, mid mt. Districts, 21m, 1W. —-=—=—— LMR LGER MIDDLE UPPER MIDDLE UPPER EXPENDITIRE COIPOSITIGI ( < 2385 ) I 2385 - 23139) (25139 - 23243) ( > 25243 ) MEAN PERCENT MEAN PERCENT MEAN PERCENT MEAN PERCENT WIN Fill) AND CLOTHING 52 a 69 70 a 64 100 b 63 136 c 54 TRAVEL 2 a 3 3 ac 3 6 be 4 11 d 4 PM PROCESSING 3 a 4 5 a 5 5 a 3 8 b 3 WSING 2 a 3 2 a 2 16 b 10 6 a 2 OTHER 5 6 5 5 4 2 7 3 TOTAL WT!!! 62 a 82 83 a 76 114 b 71 162 c 63 INVESTENT EDUCATIOI 6 a 8 12 a 11 11 a 7 33 b 13 AGRI PRCDUCTIQI 4 a 5 9 a 8 15 a 9 36 b 14 TOTAL INVESTENTS 12 e 16 23 a 21 41 b 26 75 c 30 TRANSFERS 1 a 1 3 a 3 5 a 3 17 b 7 TOTAL EXPEDITIIES 75 e 100 109 e 100 160 1: 1m 54 c 100 (per capita) __ — Source: F Secur ty surveys. ‘Dmcan's Multiple Range test was used to assess the statistical significance of the difference of meats, men there are three or more grows (semis). liners that are statistically different (5 percent level) across qmrtiles have different letteris) assimed to them. No letter after a WP signifies that there was no statistically significance difference across mat-tiles. 92 district there was one village with greater than 50 percent of the households in the upper quartile; and one village with at least 75 percent of the households in the lower two quartiles. More households in Buhera District (55 percent) were in the lower two quartiles. All measures of inequality indicated considerable income inequality across the total sample. District level analysis indicated that incomes were more unequally distributed in Buhera District, compared to Hutoko/Hudzi District. Three major sources of NHR (per capita) were earned income, transfers received, and net credit receipts. Earned income was the largest source of NHR in all villages and both districts. Transfers were»more important in Mutoko/Mudzi District, compared with Buhera District (although they were similar across NHR quartiles). Net credit receipts were small and negative across the total sample. Subcomponents of earned income were production for home consumption (PHC) and cash income-generating activities (CIGA). There was little variability in PRC between districts, but there was considerable variability between villages. Conversely, as incomes increased, assumed home-consumed production fell and inventories increased. The overall level of CIGA varied little across districts. In Buhera District, labor sales were the largest source of CIGA; followed by farm sales; while in Hutoko/Hudsi District, farm sales were the largest source of CIGA, followed by labor sales. Expenditures For all three mean expenditure measures (per household, per capita, and per adult equivalent), expenditures were larger in Buhera District, compared to Mutoko/Mudzi District. The analysis confirmed that expenditure levels vary considerably across villages. Expenditures were grouped into three categories--consumption, investment, and transfers granted--with mean levels estimated by village, 93 district, and for the total sample. For the total sample, consumption was the largest expenditure category (70 percent); followed by investments (25 percent); and transfers granted (4 percent). The composition of expenditures varied across income quartiles. First, as NHR (per capita) increased, the level of consumption expenditures fell. Second, as NHR increased, investments increased (mostly for education and agricultural production). Across all quartiles, transfers granted represented a small share of the budget. CHAPTER V RESOURCE ENDOWMENT AID EXTERNAL ENVIRONMENT Both internal (endogenous) and external (exogenous) factors influence household income. Internal factors, partly under the household's control, include both the level of resources available to the household, and their ability to allocate them efficiently. External factors include ‘the .agroclimatic, institutional, technological, and cultural environment; all of which influence household decisions, but over which the household has little control. This chapter describes the sample households' resource endowment, the external environment facing these households, and how they allocate resources to generate income. First, definitions and statistical measures to analyze resource distribution are presented. Second, household resource endowment levels are estimated. Third, the distribution of resource ownership is examined. Fourth, household access to key resources--labor, land, animals, and equipment--are evaluated across income quartiles. Finally, the external environment-- physical, institutions, and technology--facing households is described. 5.1 Household resource definitions and measures of distribution Key household resources, and the methods used to evaluate their distribution, are defined below. 5.1.1 Definitions The figs; set of definitions relate to labor resources: 1.) flggggng1_. A household is composed of family members who are related to the household head, live together, and collectively 94 95 make arrangements for feeding, budgeting, and other essentials of living. 2.) M. Resident household members are family members who live at the homestead the entire agricultural season (land preparation through harvesting). 3.) ggggghglg_h§;g. The household head is the resident household member who makes the major agricultural investment decisions. 4.) 8 use d ead. This classification incorporates the gender and residency of both the household head and spouse. The three categories are: male-headed, female-headed with male non-resident, and female-headed with no male (divorce or death). The ggggng set of definitions relate to land resources: 1.) ngure. In communal areas, there are four types of land tenure: a.) Household use rights: This is the predominant type of tenure, and means that households have a secure right to cultivate or graze the land. b.) Rent: This means that a household pays (or receives), in cash or kind, to temporarily use the land. c.) Share: This means that a household pays (or receives) a percent of the harvest to use the land. d.) Borrow: This means that a household temporarily gains (or gives) access to a piece of land, without an explicit payment (cash or kind). 2.) §911_g351;§y. The household head's assessment (poor, average, or excellent) of the soil quality (fertility and drainage) of each parcel. The EDIE set of definitions relate to physical and human capital resources: 1.) Animg1_§;agtign_gla§§gg. The three classes of animal traction ownership are: a.) Non-equipped: The household owns 99 oxen 9; traction equipment. b.) Semi-equipped: The household owns either a plow g; oxen, but not both. c.) Totally equipped: The household owns DEED a plow and two oxen. 2.) W. The animal traction index measures household ownership of animal traction equipment. A household is assigned a value of is zero if it has no animals or equipment (non-equipped), one-half if it has some animals or equipment (semi-equipped), and one if it has a full complement of animals and equipment (totally equipped). 3.) figgtgg_figgggg. A master farmer is someone who has completed 96 AGRITEX's two year Master Farmer course‘, adheres to husbandry practices recomended by AGRITEX for that area, and attain crops yields as good as the upper 25% of area farmers over a five year period (AGRITEX, 1984). 5.1.2 Measures of resource distribution The analysis uses skewness and kurtosis to assess symetry; the median (village-level analysis) and the mean (district, sample, and per capita income quartile) to assess central tendency; and the Gini coefficient to assess equality of resource availability. See Chapter 2 for a more detailed explanation of these measures, and why they were chosen. 5.2 Overview of household resource availability The three most important household resources are land, labor, and capital. Household access to these resources varied greatly across villages, districts, and the entire sample (Table 5.1). June: In the study sites, household members were the primary source of labor (6.6 members per household). Generally, household labor was more abundant in Buhera District, where mean residents per household2 averagd 7 .4 compared to 5.9 in Mutoko/Hudsi Districts3. Furthermore, in Buhera District, households were less variable in size. For example, median residents per household ranged from 6 to 7 in Buhera villages, and 4 to 7 in Mutoko/Mudzi villages. ‘Participants must attend 24 one-day classes annually, plus a four day veterinary course, and a four day farm machinery course. 2 See footnote 4 from Chapter 4. 3These means are similar to Stack (Stack and Chopak, 1990) estimates for Shamva and Binga Districts (6.3 and 12.9 residents per household, respectively); and Govaerts' (1987) estimate (7.8) for Mutoko District. 97 Tbie 5.1 Overview of Mid reams-cs W by village (adieu, district (me-1), as! total sqle (an), 21m, 1“. 1.6598 0 LEVEL OF SAMPLE TOTAL LAND LAND 1m ADCREDATION SIZE RESIDENT NON-RESID AREA Pc PAE m m (1!) (ha) (ha) (ha) HUMERA DISTRICT 1 25 6 5 4.1 .7 .9 .78 2 23 7 5 6.2 .8 1.1 .83 3 20 7 5 5.1 .8 1.2 .91 I. 24 7 5 5.4 .9 1.3 .89 5 23 7 s 3.7 .5 .7 .81 l 6 21 6 5 3.7 .6 .9 .75 ‘ DISTRICT TOTAL 136 7.4 ... 5.3 5.6 ** 1.0 * 1.4* .83 ** mom/TIMI 1 26 7 4 2.4 .4 .6 .75 2 29 4 3 3.7 .6 .8 .76 3 15 4 3 3.1 .8 1.1 .70 ) 4 27 7 5 4.0 .6 .9 .56 5 29 5 4 3.0 .5 .7 .69 6 23 6 4 1.8 .3 .5 .70 ‘ DISTRICT TOTAL 149 5.9 *- 4.3 3.3 *1- .7 * .9* .70 *1- i I SAMPLE TOTAL 285 6.6 4.5 4.4 .9 1 2 .76 J} Source: Secur ty surveys ‘The median is used to assess central tendency for village level estimates; mile the district and total semis level estimates are means. ifferences in district means were tested for statistical significance at the 1 U") and 5 0') percent level. 98 Land All measures of land availability (per household, per capita, and per adult equivalent; by village, district, and the total sample) indicated large differences in household access to land. Mean cultivated area averaged 4.4 hectares for the total sample, 5.6 hectares for Buhera District, and 3.3 hectares in Mutoko/Mudsi Districts. These means are somewhat smaller than Stack's (Stack and Chopak, 1990) estimate for Binga District (7.9 hectares per household); and slightly larger than Govaerts' (1987) estimate (2.6 hectares per household) for Mutoko District‘. .As expected, Buhera District households had greater access to land than Nutoko/Nudsi Districts households because the population density is lower in Natural Region v than in Natural Region IV. Yet, district level averages tend to obscure the large inter-village differences in land availability. For example, land per household (median) ranged from 3.7 to 6.2 hectares in Buhera District; and 1.8 to 4.0 hectares in Nutoko/Nudzi Districts. Similarly, land per capita (median) ranged from 0.5 to 0.9 hectares in Buhera District; and 0.3 to 0.8 hectares in Mutoko/Nudzi Districts. These differences are even greater when converted to a per adult equivalent, ranging from 0.7 to 1.3 hectares in Buhera District; and from 0.5 to 1.1 hectares in Mutoko/Nudsi Districts. W‘ Several studies5 have identified access to traction animals and equipment as important factors that enable households to produce enough food. Overall, Buhera District households had greater access to traction. ‘Binga District is agro-ecologically similar to Buhera ZDistrict. Our study's estimate for the three villages in nutoko District was 2.7 hectares per household. 5For example, see Dione (1989). 99 The mean animal traction index was 0.76 for the total sample, 0.70 for Nutoko/Mudzi Districts, and 0.83 for Buhera District. As expected, Buhera District households had a higher animal traction index. Because Buhera District is less favorable for crop production, households place greater emphasis on livestock production, compared to Mutoko/Nudzi farmers. As was the case for land, within both districts there are large inter-village differences in access to traction. Median animal traction index ranged from 0.75 to 0.91 in Buhera District and 0.56 to 0.76 in Nutoko/Mudsi Districts. 5.3 Distribution of household resources The distribution of land, labor, and physical capital are assessed by analysing their spread--symmetry and equality--across the total sample and districts. Land was measured as hectares per capita; labor as residents per household; and physical capital as the number of oxen owned per household. 5.3.1 Symmetry of resource ownership The measures of symmetry (skewness and kurtosis) indicate an asymmetric distribution of resources across districts and the total sample (Table 5.2). Household residents, land per capita, and oxen owned had positively skewed distributions (tail to the right); and showed a high amount of kurtosis. This indicates that ownership was clustered at the low end and was spread over a very narrow range. Although all three resources were assymetrically distributed, the distribution of land was the most skewed; followed by oxen, and finally residents. These same total sample trends hold for the districts, although for Buhera District the distribution of all three resources was more clustered at the lower end (skewness) over a narrower range (kurtosis) 100 This 5.2 Distrihitim of Mid resets-ces, Make/Mi mid Idlers Districts, little, 1”. MGJSENOLD SAMPLE SYMETRY GINI RESORCE SIZE COEFFICIENT (f) MEAN MED IAN SD SE SKEUNESS KURTOS i S RESIDENTS DUNERA DISTRICT 136 7.4 7 4.6 .39 2.212 8.508 .3073 mom/men 149 5.9 5 2.8 .29 0.347 (0.570) .2657 DISTRICTS 285 6.6 6 3.8 .23 2.135 10.125 .2926 SAMPLE TOTAL GEN DUNERA DISTRICT 136 2.2 2.0 3.1 .26 4.695 26.511 .5764 iIJTOKO/IuJZI 149 1.6 1.0 1.9 .16 2.509 8.032 .6047 DISTRICTS 285 1.9 2.0 2.6 15 4.959 34.078 .5947 SAMPLE TOTAL LAD DUHERA DISTRICT 136 1.0 0.7 1.9 .17 9.648 104.745 .4442 MJTNOIIRDZI 149 0.7 0.5 0.6 .05 5.496 43.131 .3646 DISTRICTS 285 0.9 0.6 1.4 .08 11.625 166.319 .4175 SAMPLE TOTAL Source: Food Security surveys. ‘Values in parentheses are negative Tamers. 101 than for Nutoko/Nudsi Districts. 5.3.2 Equality of resource ownership The Gini coefficient is used to assess equality of resource ownership across the total sample, and for each district. For the total sample, the Gini coefficient indicated a low level of inequality for residents (0.29), a moderate amount of inequality for land (0.40), and a high degree of inequality for oxen (0.66)‘. For all three resources, Buhera District households had larger Gini coefficients than those in Mutoko/Mudzi, indicating that all three resources are less equally distributed in Buhera District than Mutoko/Nudzi. 5.4 Resource endowment by not household receipts quartiles This section explores several hypotheses about the relationship between household resources and income by analysing the distribution of land, labor, and capital across per capita net household receipts (income) quartiles. 5.4.1 Labor Household labor resources varied in terms of their composition, and age and gender of the head (Table 5.3). Household composition Household composition varied greatly across per capita income quartile. The poorest (lowest quartile) households had, the largest families (11.7) and the most resident household members (8.6). Although households had similar numbers of non-residents, for the upper income ‘FAO (1986) classifies degrees of equality, based on the following ranges in the Gini coefficient: less than 0.41 as low; 0.41 to 0.45 as moderate; 0.46 to 0.50 as relatively high; and greater than 0.50 as high. 102 Tile 5.3 Mid Idler characteristics by incm qmrtile (an), MWI mui Idlers Districts, ZidldsIa, 1W. NOISEIIOLD LMST LWER MIDDLE UPPER MIDDLE UPPER CGIPDSITIGI ( < 2585 ) ( 2585 - 25139 ) ( 25139 - 25234 ) i > 25234 ) NOISENOLD (8) RESIDENTS 8.6 a 7.0 b 5.8 c 5.0 c MOI-RESIDENTS 3.1 2.5 2.8 3.4 TOTAL HEIRS 11.7 a 9.5 b 8.6 c 8.4 c AGE DISTRIIITIN (2) (Residents) < 6 22 29 28 33 6 - 18 40 38 38 35 > 18 38 34 34 32 GENDER DISTRIBUTIGI (X) (Residents) MALE 49 47 47 46 FEMALE 51 53 53 54 iource: Food Securfiy surveys. a/ Dmcan's Multiple Range test was used to assess the statistical significance of the difference of means, Then there are three or more grows (leans). percent level) across martiles have different letter(s) assigned to them. signifies that there was no statistically significant difference across quartiles. Nubers that are statistically different (5 No letter after a Turber 103 group, 40 percent of the household members were non-resident, compared to only 26 percent for the lowest income quartile, a potential source of remittances. Contrary to expectation, higher income households tended to have a larger percent of resident children (under six years) and fewer adults (older than 18 years) than lower income households. Finally, higher income households tended to have a slightly higher proportion (54 versus 50 percent) of female residents (including children) than the lowest quartile households, which is consistent with the results indicating that higher income households were more likely to have non-residents employed outside the community. Household head Household heads in all income quartiles were similar in age (48 to 50 years) and gender (82 to 87 percent male) (Table 5.4). On the other hand, the cross classification of households by gender and residency status of the head provided unexpected results. First, as expected male-headed households were most common (82 to 87 percent) in all income quartiles; followed by female-headed households with the male away (8 to 15 percent) and female-headed households with no male (1 to 10 percent). Second, it was hypothesized that these female-headed households would earn the lowest incomes--unless they receive a significant amount of transfers-~because they would have less access to labor resources. Although this relationship held for the lower three quartiles, it does not for the highest income quartile where there were more female-headed/no male households in the upper income quartile (10 percent) than any other quartile. 104 Table» 5.4 Moussholdl hssdl characteristics by' incoasr quartile (mean), MutokoMMudsi and lluhere Districts, Zimbabwe, 1 . HOUSEHOLD LOHER UPPER HEAD LONEST MIDDLE MIDDLE UPPER CHARACTERISTICS ( < 2585 ) ( Z585 - 25139 ) ( 25139 - 25234 ( > 25234 ) ) AE (years) 50 50 48 48 GENDER DISTRIIUTIDN (2) MALE 85 82 87 82 FEMALE 16 18 12 18 CLASSIFICATION OF HOUSEHOLD NEAD(%) MALE HEADED 85 82 87 82 FEMALE HEADED/ MALE AUAY 10 15 11 8 FEMALE HEADED/ NO MALE 6 ab 3 a 1 a 10 b W@ Source: Security surveys. ‘ Dmcan's Multiple Range test was used to assess the statistical significance of the difference of means, then there are three or more groups (means). percent level) across cpsrtiles have different ietter(s) assigned to them. signifies that there was no statistically significant difference across quartiles. Nubers that are statistically different (5 No letter after a hater 105 5.4.2 Land In an agricultural based economy, access to land is an important determinant of a family's income-earning potential. Land availability, tenure, and quality Although household access to land varied across income quartiles, there was no statistically significant relationship between income quartiles and tenure, soil quality, or distance to fields (Table 5.5). Land availability (both per capita and per adult equivalent) increased from the lowest to highest income quartile. Strikingly, the highest income quartile households had twice as much land as the lowest income quartile households. For all income quartiles, land was predominantly family owned7 (93 to 96 percent). Furthermore, there was little difference in soil quality (fertility and drainage) across income quartiles. Heuseholds reported that over half (52-59 percent) of their land was of average soil fertility; and over 74 percent assessed their soil fertility as average or excellent. Most households rated their soil drainage as excellent (49 to 62 percent) or average (25 to 40 percent). Finally, the mean distance from the homestead to their fields was similar across income quartiles (10 to 13 minutes). Land use rt was hypothesised that given the differences in rainfall between sites, land use patterns would vary across districts and across income quartiles. 7Ownership means households had long term use rights, but couldn't sell the land. Tfile 5.5 Lid characteristics by inc. qmrtile (an), MtokoMi 106 and More Districts, liddama, 1 r 1 LAND LGIEST LINER MIDDLE UPPER MIDDLE UPPER CHARACTERISTICS ( < 2585 ) ( 2585 - Z5139 ) ( 25139 - 25234 ) I > 25234 ) AREA AVAILABLE Per HH 4.8 4.6 3.7 4.5 Per capita 0.6 a 0.7 a 0.7 a 1.4 b Per adult 0.8 a 1.0 a 1.0 a 1.8 b TEIIIE (2) Own 95 93 96 96 Rent/share 0 0 1 1 Sorrow in 4 a 7 b 3 a 3 a Sorrow out < 1 < 1 < 1 < 1 Total 100 100 100 100 ”IL GIALITT Fertility (2) Poor 26 25 26 20 Average 58 52 55 59 Excellent 16 23 18 20 Total 100 100 100 100 Drainage (2) Poor 12 13 6 12 Average 40 25 34 36 Excellent 49 62 60 53 Total 100 100 100 100 DIST TO 11 13 11 10 FIE iource: Food Security surveys. ' Dmcan's Multiple Range test was used to assess the statistical significance of the difference of men, men there are three or more grOLps (means). percent level) across quartiles have different letter(s) assigned to them. signifies that there was no statistically significant difference across quartiles. The distance to field is a weighted average of the distance (minutes) of all fields from the homestead. FIELD DISTANCE where: i = field Tuner 3 i: DISTANCE 1 .0 TOTAL *AREA AREA NLwers that are statistically different (5 No letter after a nuber 107 W Two similarities across districts stand out (Table 5.6). First, households in both districts allocated similar proportions of land to crops (88 to 89 percent) and fallow (11 to 12 percent). Second, the relative amounts of land allocated to crop types was similar; grain occupied the majority of available land (78 to 82 percent), followed by oilseeds (14 to 15 percent), and other crops such as cotton, fruits and vegetables, and intercropping (3 to 6 percent). In contrast, there were three major differences in cropping patterns across districts. First, although maize and small grains were the major crops in both districts, households in Buhera District allocated a significantly larger share of their land to small grains (61 percent) and less to maize (18 percent) than households in Hutoko/Hudzi Districts (36 and 37 percent, respectively). This result was expected since available maize technology is more appropriate for Mutoko/Hudzi Districts, which have relatively higher rainfall. Second, although oilseeds occupied similar portions of available land in both districts, households allocated a significantly larger share of their land to groundnuts in Buhera District (7 percent) than in Mutoko/Mudzi Districts (3 percent); but they allocated a significantly larger share of their land to sunflower in Mutoko/Mudzi Districts (10 percent) than in Buhera District (2 percent). Cropping patterns also varied greatly across per capita income quartiles8 (Table 5.7). Crops were grouped as grains (maize, small grains9, and rice), oilseeds (primarily groundnuts and sunflower), fruits and vegetables, cotton, and mixed (intercrops). I’The importance of different crops varied for individual households, even within quartiles. 9Small grains include millet (bulrush and finger) and sorghum (white and red). 108 Tble 5.6 Land me by District ml! total sqle (an), Mata/Mi mid fliers Districts, 21m, 1 . — LAND TOTAL DUHERA MUTOKD/MUDZI use SAMPLE DISTRICT DISTRICTS LAIm ass (10 CULTIVATED 88 89 88 mm: 12 11 12 cam ALLOCATIIII'kx) GRAIN MAIZE 27 18 ** 37 ** SMALL GRAINS 48 61 ... 36 “1 MAIZE/SMALL GRAINS 2 2 3 RICE 2 1 2 TOTAL DRAIN 79 82 78 DILSEED CROINIDNDTS 5 7 .. 3 *1- SDNELONER 7 2 n 10 ** 0TNER° 4 6 2 TOTAL 15 1!. 15 COTTON 1 0 2 FRUITS I. VECETASLES 1 1 1 INTER- CROPPED 3 2 3 Source: F SecurI ty surveys. 'Diffsrences in district means were tested for statistical significance at the 1 (1") and 5 (1") rcent level. ercent of crops allocated to cultivated land. cOther oilseeds include Mara nuts, coupeas, and Richeys beans. 109 Tau: 5.7 La: Tee by In. gmrtile (nun), Make/Ishi and mu Districts, zine-Ina, Toes/89'. UPPER LAND LDUEST LDHER MIDDLE UPPER MIDDLE USE ( < 2585 ) ( 2585 - 25139 ) ( 25139 - 25234 ) ( > 25234 ) LA.) USE (2) CULTIVATED 85 a 90 ab 87 ab 91 b FALLIMI 15 a 10 ab 13 ab 9 b CHIP ALLOCATIoI'Ix) GRAIN MAIZE 22 a 25 a 33 b 28 a SMALL GRAINS 59 a 43 b 41 b 40 b MAIZE/SMALL GRAINS 1 I. 1 4 RICE 1 1 3 1 TOTAL GRAIN 84 a 83 a 78 b 72 c OILSEED GRGINDNUTS 4 a 4 a 4 a 8 b SUNFLOHER 4 a 5 a 7 ab 10 b OTHER DILSEEDS'= 3 3 5 4 TOTAL 12 a 12 a 16 b 22 c conm 1 1 1 1 FRUITS 8 VEGETABLES O a 1 ab 1 ab 2 b INTER- CROPPED 3 3 3 3 Source: Food Security surveys. ‘ Duican's Multiple Range test was used to assess the statistical significance of the difference of means, Then there are three or more grows (means). percent level) across quartiles have different letter(s) assigned to them. signifies that there was no statistically significant difference across cpsrtiles. "Percent of crops allocated to cultivated land. cOther oilseeds include were nuts, coweas, and kidneys beans. Nuwers that are statistically different (5 No letter after a Timber 110 Three similarities stand out across per capita income quartiles. First, in all income quartiles grain crops dominated area planted (72-84 percent). Second, the rank order of area (percent) by crops was similar across income quartiles. Grains were planted to the largest share of available land; followed by oilseeds (12-22 percent), inter-cropped plantings (3 percent), fruits and vegetables (1-2 percent), and cotton (1 percent). In contrast, there were four important differences in land use that occur as per capita incomes increased. First, as incomes increased farmers tended to allocate a smaller share of their cropped land to grains (84, 83, 78, and 72 percent); and a larger share to oilseeds (12, 12, 16, and 22 percent). Second, as incomes increased the relative importance of individual grain crops changed. For example, as incomes increased, the maize share tended to increased (22, 25, 33, and 28 percent), and the small grains share decreased (59, 43, 41, and 40 percent). Third, as incomes increased, the proportion of land planted to individual oilseed crops increased. For example, higher income households planted a larger percent of their area to groundnuts (4, 4, 4, 8) and sunflower (4, 5, 7, and 10 percent). Similarly, higher income households planted a larger share of their land to fruits and vegetables (0, 1, 1, and 2 percent). 5.4.3 Capital ownership The amount and quality of capital--physical, human, and financial-- available to households is an indicator of both a household's wealth and its ability to' cultivate available land in a timely manner. Physical capital In the low-rainfall areas of 2imbabwe, animals and equipment were the most important physical assets owned by households. Cattle, sheep, and goats were the most comonly held livestock, but some households also 111 had pigs and donkeys. Flows were the most commonly owned agricultural equipment; but some households had cultivators, ridgers, barrows, sprayers, and ox carts. Win It was hypothesized that livestock ownership would vary across districts and per capita income quartiles. Differences across districts Two important inter-district differences in livestock ownership were identified (Table 5.8). First, households in Buhera District owned significantly more cattle (7.5 per household), oxen (2.2 per household) and small ruminants (10.6) than households in Mutoko/Mudzi District (4.7, 1.6, and 4.0, respectively). As expected, livestock were more important in Buhera District because: 1.) historically laws limiting herd size were enforced more rigorously in higher density areas (more like Mutoko/Nudzi than Buhera); and 2.) Buhera District has more grazing land, which permits households to manage larger herds. The second inter-district difference was that Mutoko/Hudzi Districts households owned more pigs (0.7 per household) than Buhera District households (0.1)--possibly because pigs are fed maize in communal areas, which is more plentiful in Mutoko/Mudzi Districts. Differences across income quartiles The only statistically significant relationship between income and livestock ownership was for large and small ruminants (Table 5.9). Higher income households generally owned more total cattle (4.9, 4.8, 6.9, and 7.4) and oxen (1.5, 1.5, 2.0, and 2.5). In contrast, the relationship between income and small ruminants and non-ruminants animals is less clear. For example, higher income households generally owned more sheep (0.6, 1.2, 0.6, and 2.3), except for the upper middle income households; but there was no consistent Twle 5.8 Cwitel emisrship (ms-I) by district mu! total sqle, “WI ms! Idlers Districts, 21m, TM. 112 CAPITAL TOTAL DUHERA IIJTWDIIIDZI MERSHIP SAIN’LE DISTRICT DISTRICTS PHYSICAL CAPITAL l r i (l) Cattle OMEN 1.9 2.2 * 1.6 * P BULLS .3 .3 .2 STEERS .3 .6 ** < .1 ** HEIFERS .9 1.1 .7 DAIRY 1.7 2.1 * 1.3 * CALVES 1.1 1.2 .9 TOTAL 6.2 7.5 ** 4.7 ** Small ruminants SHEEP 1.2 2.4 ** .1 ** COATS 5.9 8.2 ** 3.9 ** Other DGIKEYS .2 .3 .1 PIGS .4 .1 ** .7 ** was! In Plows 1.3 1.6 *' 1.0 ** Cultivators/Ridger .1 < .1 ** .2 ** Narrows < .1 < .1 < .1 Ox carts .3 .4 ** .1 ** IMAM CAPITAL W “909 24 17 *1! 31 so > 1 years 76 83 ** 69 ** r r No 83 88 * 77 * Trainee 12 9 * 16 * Master farmer 5 3 7 FINAEIAL CAPITAL AF No 96 95 98 Yes 4 5 2 Source: F53: Security surveys. 'Differences in district means were tested for statistical significance at the 1 I”) and 5 (*) percent level. 113 This 5.9 Cwitel Nip (an) by per cwite in. qmrtile, "WI and More Districts, 21m, 1%. CAPITAL LINER LINER MIDDLE UPPER MIDDLE UPPER MERSHIP ( < 2585 ) ( 2585 - 25139 ) ( 25139 - Z5234 ) ( > 25234 ) PHYSICAL CAPITAL W Cattle OMEN 1.5 a 1.5 a 2.0 ab 2.5 b BULLS .3 .2 .3 .3 STEERS .2 a .3 ab .3 ab .5 b HEIFERS .7 .7 1.1 1.0 DAIRY 1.3 1.4 2.0 1.9 CALVES .9 .9 1.2 1.2 TOTAL 4.9 a 4.8 a 6.9 ab 7.4 b Small rusinants SHEEP .6 a 1.2 a .6 a 2.3 b GOATS 5.0 ab 7.2 a 4.5 b 6.8 ab Other DONKEYS .2 < .1 .3 .2 PIGS .4 .4 .3 .4 m m Plows 1.2 1.3 1.2 1.4 Cultivators] Ridgers .1 .1 .1 .1 Harrows < .1 O .1 < .1 0x carts .2 a .2 a .3 ab .4 b IIAM CAPITAL Matias (1‘) None 27 26 22 21 > 1 years 73 74 78 79 was No 85 80 85 81 Trainee 10 15 10 14 Master farmer 4 4 5 5 FIMIAL CAPITAL 4 1 1 7 urity surveys. ‘ Dmcan's Multiple Range test was used to assess the statistical significance of the difference of means, men there are three or more grows (means). percent level) across (partiles have different letter(s) assigned to them. simifiss that there was no statistically significant difference across qmrtiles. NIflaers that are statistically different (5 No letter after a meter 114 relationship between donkey, goat, or pig ownership. W This section presents the reported levels of household equipment ownership; for the total sample, district, and across per capita income quartile. Differences across district In both districts, plows were the only agricultural implement that were commonly owned, with 85 percent of the households reporting owning at least one plow (Table 5.8). Although few households owned ox carts, they were more common in Buhera District (0.4 per household) than Hutoko/Mudzi (0.1). On the other hand, cultivators were more available in Hutoko/Mudzi Districts (0.2 per household) than in Buhera (< 0.1). Finally, few farmers in either district owned ridgers, harrows, or sprayers. Differences across income quartiles Ownership of agricultural equipment (mean number owned) was surprisingly similar across per capita income quartiles (Table 5.9). Plow ownership was relatively constant across income quartiles (1.2 to 1.4 per household); and few households owned cultivators, ridgers, harrows, or sprayers, regardless of income quartile. On the other hand, although few households owned ox carts, ownership appeared associated with rising income levels. Human capital Formal and non-formal education contribute to strengthening human capital stock, which serves to increase an individual's ability to exploit income-earning opportunities. Two measures of human capital are education and participation in extension training (eg, master farmer 115 program). First, the proportion of household heads who had attended school differed across districts, but not across income quartiles. For example, significantly more household heads in Mutoko/Mudzi District had no formal education (31 percent), than in Buhera District (17 percent). In contrast, slightly more household heads in the upper two quartiles had attended school (78 and 79 percent, respectively), compared to the lower two quartiles (73 and 74 percent, respectively). The proportion of household heads who had participated in the Master Farmer Program differed across districts, but not across per capita income quartiles. In Mutoko/Mudzi District, about twice as many household heads were master farmers (7 percent) or master farmer trainees (17 percent) than in Buhera District (3 and 9 percent, respectively). In contrast, the rate of household head participation in the Master Farmer Program was similar across income quartiles (4 to 5 percent). Financial capital Farmers made minimal use of formal--government and commercial-- credit. For the total sample, only 4 percent of the households borrowed from the Agricultural Finance Corporation (AFC). These households were concentrated in three villages in Buhera and one in Mutoko/Mudzi District; and received only small loans (the median loan was 259.45). Rohrbach (1988) reported similar results. Households reported they didn't use AFC credit because they don't produce enough to repay the AFC (45 percent), don't want to sell crops (23 percent), were dissatisfied with AFC's lending policies (13 percent), and other reasons (20 percent). There was no systematic relationship between credit use and income quartiles, although upper income quartile households used AFC credit most frequently (7 percent), followed by the lowest quartile (4 116 jpercent), and the two middle quartiles (1 percent each). 5.5 Income level and sources by resource endowment To assess the relationship between median net household receipts-- level and components--and resource endowment, households were classified by resource endowment (labor, land, and capital). Labor The level and sources of income varied by household size; and gender and age of the household head (Table 5.10). Hsnashels_aiss Smaller households (fewer than 5 residents) earned more NHR (per capita) than larger households--small households reported incomes of 25229; compared to 25124 for households with five to seven members, and 2597 for households with more than seven members. While the sources of incomes were similar across household size, larger households earned slightly more income from labor sales. WM Income level and sources were quite similar for male-headed and female-headed households, although male-headed households reported slightly larger incomes (25141) than female-headed households (25133). On the other hand, there were major gender differences with respect to transfers received, and credit obligations. Female-headed households received larger transfers (2518) and also incurred larger credit obligations (255) than male-headed households (258 and 251, respectively). Furthermore, male-headed households earned slightly larger receipts from farm. sales (2515) and labor sales (256) than female-headed households (2511 and 254, respectively). 117 5.10 Inc. level I'd sauces by labor eveiidsility (medium), Mountain Districts, ZiwdmIe, 1W. T LADOR SAMPLE NET NET owNERSMIP SIzE RECEIPTS INCOME TPNc" CIGA° TES" TLs’ NAP' TRI' NCRh (Pc) (PC) Mid size <5 95 229 175 120 88 26 3 1 24 (1) 5-7 88 124 114 61 64 12 8 3 5 (3) >7 102 97 87 44 37 12 6 1 6 (1) MOI-dioidhead gmmhr Female 39 133 128 73 64 11 4 4 18 (5) Male 243 141 118 61 61 15 6 2 8 (1) 1 MOI-dioldhead ' use <35 68 149 136 57 67 10 6 <1 5 (2) 35-55 117 117 98 53 57 15 8 3 12 (2) >55 97 152 131 87 53 17 3 2 9 <1) - Securi ty urveys. ' Values in parentheses are negative meters Total production for home consumption ‘ Cash income-generating activities ‘ Total farm sales ' Total labor sales ' Hon-agricultural product sales 9 Transfers received " Net credit receipts 118 MEMOS Although younger household heads (less than 35 years old) earned similar incomes as the oldest household heads (greater than 55 years old), they earned less from production for home consumption, farm sales, nonagricultural product sales, and transfers. In contrast, younger households earned more from market transactions and labor sales. Land ownership and use Access to land, and how households allocate it to individual crops, are important determinants of income. This section first examines the relationship between land availability, and income levels and sources. Then, it examines the relationship between area allocated to crops and income levels and sources. 1.289418118211131 Households with the least land (< 0.61 hectares per capita) earned less income (25109) than households with 0.61 to 0.96 hectare (25141) and much less than households with more than 1 hectare (25231) (Table 5.11). A similar relationship existed between land availability and production for home consumption, market transactions, and farm sales. In contrast, households with less than 0.61 hectares earned more income from labor sales (> 257) than those with more land (254), implying that households without sufficient land available seek off-farm employment as a strategy to earn income to meet household needs. Finally, land poor households received more transfers (259) than households in the middle two quartiles (253-8), although households with the most land received the greatest amount of transfers (2518). M Analysis of the relationship between land use and income level and sources provided several insights. 119 5.11 Inc. level III moss by ill! mnership III! as (media), Make/Mi mid Idlers Districts, 21m, 1988789'. _ LAND SMLE NET PRmUCTION CASH INCOIE FARM LABOR NON-AC TRANS- NET MERSHIP SIZE HWSEHOLD FIR HOE GENERATING SALES SALES PRwUCTS FERS CREDIT AND USE RECEIPTS COiSlHPTION ACTIVITIES RECEIPTS Lad per cwite (ha) < .41 63 109 46 56 11 8 2 9 (2) .41 - .61 75 100 52 46 9 7 2 8 (1) .61 - .96 75 141 67 50 15 4 1 3 (2) > .96 72 231 105 90 40 4 2 18 (2) Grain area (2) < 69 69 197 79 91 43 4 4 12 (1) 69 - 82 73 142 76 61 24 6 1 7 (2) 82 - 96 70 101 56 59 12 6 3 6 (1) > 96 72 113 45 46 6 7 1 12 (1) Maize area (2) < 12 71 100 45 46 9 6 1 5 (1) 12 - 24 72 150 61 61 13 8 2 11 (0) 24 - 40 70 144 87 67 22 5 1 6 (1) > 40 72 159 73 66 18 5 6 20 (4) hll grain area (3) 0 23 265 114 98 41 9 3 46 (1) 1 - 43 89 168 69 81 41 3 5 18 (3) 43 - 64 86 119 68 46 11 6 2 5 (1) > 64 86 100 38 45 8 7 1 6 (1) Oilseed area (2) 0 84 118 54 45 6 8 1 12 (1) 1 - 14 71 106 54 48 8 4 2 4 (1) 14 - 27 64 136 71 61 25 9 1 8 (3) > 27 65 217 84 96 52 3 5 12 (3) Cotton area (3) 0 274 139 65 61 13 6 7 (7) > 0 10 128 65 68 51 3 1 4 (3) Fruit/Vegetwl area (2) 0 269 139 64 60 14 6 2 8 (2) > 0 15 121 65 70 29 4 0 14 0 ourc: ...' Secur t surveys. ' Values in parentheses are negative WPS. 120 Grain area As the percent of the cropped area allocated to grain production increased, median incomes declined, households sold more labor, and generally received more transfers. For example, households that allocated more land to grain (> 96 percent) earned lower income (25113) than households that allocated less (< 69 percent) area (25197), and also earned less from production for home consumption, CIGA, farm sales, and nonagricultural product sales. Maize area On the other hand, maize area (is) was positively related to median incomes. For example, as the proportion of cropped area allocated to maize increased, incomes increased from 25100 (< 12 percent maize) to 25159 (> 40 percent maize). Also, households that devoted a larger share of their land to maize tended to earn more income from production for home consumption, CIGA, farm sales, nonagricultural product sales, and transfers; but less from labor sales. Small grain area In contrast, small grain area was inversely related to median income. For example, households that did not grow small grains earned almost twice the income (25265) as small grain producers (25168, 25119, and 35100). Furthermore, households with more land in small grains earned much less income from production for home consumption, market transactions, farm sales, and transfers. Oilseed area Area in oilseeds was positively related to income. For example, households with a larger share (>27 percent) of their land in oilseeds tended to have larger median incomes (25217) than households that did not plant oilseeds (25118). Also, households that allocated more land 121 to oilseeds earned more from production for home consumption (2584), market transactions (2596), farm sales (2552), and had larger credit obligations (253), than households with less oilseed land. Cotton area Only 4 percent of the sample households grew cotton. Compared to non-growers, those households earned slightly smaller incomes (25128 versus 25139), and sold less labor (256 versus 253); but earned more from market transactions (2568 versus 2561), farm sales (2551 versus 2513), and transfers (257 versus 254) than non-growers. Fruit and vegetable area Only 6 percent of the households grew fruits and vegetables. These households earned slightly lower incomes (25121 versus 25139) than non- growers. On the other hand, fruit and vegetable growers reported higher market transactions (2570 versus 2560), farm sales (2529 versus 2514), and transfers (2514 versus 258) than non-growers. Capital ownership This section analyzes the relationship between capital (physical, financial, and human), and income levels and sources (Table 5.12). W Analysis of the data found a positive relationship between income and the level of physical capital owned by households. Oxen ownership Households owning an oxen team (two or more oxen) earned more income (25184) than households with less than two oxen (< 25122). Also, households with an oxen team reported greater production for home consumption (2568 versus 2560), market transactions (2571 versus 2546), 5.12 [mm level mid sou-css by cwitai omership (wdien), ”Mi 122 .11! ”Tera Districts, 21m, 1 . fi CAPITAL SAWLE NET PRwUCTIGi CASH INCOIE FARM LABOR NON-AG TRANS- NET MERSHIP SIZE HWSEHOLD Fm HOIE GENERATING SALES SALES PRODUCTS FERS CREDIT RECEIPTS COISlMPTICMi ACTIVITIES RECEIPTS Physical cwitel 95m omierghjp < 2 142 122 60 46 9 6 4 10 (1) > 2 143 184 68 71 23 6 1 8 (2) P ow r hi 0 44 118 56 41 6 3 2 12 (<1) > 1 241 141 65 65 17 6 2 8 (2) Fin-Iciei cwitel SEE—lean No 275 138 62 62 15 6 2 8 (1) Yes 10 252 138 69 14 21 0 19 (1) mm cwitel Education 9f head 0 68 122 69 37 8 3 2 10 (1) > 1 214 143 61 65 18 7 2 8 (2) source: Food Secur ty surveys. ' Values in parentheses are negative nuIbers. L.) '- e 123 farm sales (2523 versus 259), but less non-agricultural product sales (251 versus 254), than households without an oxen team. Plow ownership As with oxen, there was a strong relationship between income and plows owned. Plow-owning households had larger incomes (25141) than non-owners (25118). Also, plow-owning households reported greater production for home consumption (2565 versus 2556), market transactions (2565 versus 2541), farnl sales (2517 versus 256), labor sales (256 versus 253), and credit obligations (252 versus < 251) than non-owners. 1W Although only 4 percent of sample households borrowed from the Agricultural Finance Corporation (AFC), borrowers earned larger incomes (25252) than non-borrowers (25138). Credit-using households also reported much greater production for home consumption (25138 versus 2562), labor sales (2521 versus 256), and transfers (2519 versus 258) than non-borrowers. Waite}. Education was positively related to income. Households heads who had some schooling reported higher incomes (25143 versus 25122) than heads without schooling. In addition, household heads who had attended school participated more in the market (2565 versus 2537), earned more from farm sales (2518 versus 258) and labor sales (257 versus 253). 5.6 Interrelationships between resource endowment and socio-economic characteristics This section explores interrelationships between the major household resources. Crosstabulations are presented to identify socio-economic factors associated with differences in household land, labor, and 124 capital endowments. Labor Three measures of labor availability are household size, and the gender and age of the household head. Although larger households cultivated less land (per capita) than smaller households, 0.47 versus 0.98 hectares, they owned more oxen (2 versus 1) (Table 5.13). Surprisingly, male-headed and female-headed households were quite similar. Female-headed households had slightly fewer resident household members (5 versus 6), which partially accounted for their having slightly more land (0.75 versus 0.62 hectares per capita). In contrast, female-headed households owned more oxen (2 versus 1), possibly because their husbands had non-farm employment which enabled the household to invest in oxen. As expected, households with older heads tended to have more residents than younger households (7 versus 5) , cultivated more land (0.78 versus 0.56 hectares), and owned more oxen (2 versus 1). Land use Analysis of the data highlights important differences in land use patterns (crop priorities), associated with household resource endowment (Table 5.14). Grain area The relative importance a household placed on grain production was inversely related to farm size. Households that allocated the largest proportion (> 96 percent) of their land to grains cultivated less land per capita (0.51 versus 0.79 hectares), than households that allocated a smaller proportion of land to grains (< 69 percent). 125 5.13 Resumes cub-mt by hOImehold iwor charecteristics (ndimi), Intake/Mi ms! ”IO" Districts, little, I”. - LAM SQIC Nudaer of Area Oxen CHARACTERISTIC Size Residents Avai labls Ownership MOI-shad size 1 < 5 95 4 .98 1 , 5-7 88 6 .55 1 ; > 7 102 9 .47 2 Numdooid head , gmsier ! 1' Female 39 5 .75 2 5 Male 243 6 .62 1 [ Nonmdioid head ' Q! 1 < 35 68 5 .56 1 5 35-55 117 7 .59 1 1 > 55 97 7 .78 2 Source: Sec surveys. Haize area Similarly, except for households that grew almost no maize, the share of land planted to maize was inversely related to land availability. For example, households that planted a smaller shares (< 24 percent) to maize cultivated more land (0.62 versus 0.50 hectares), than households that allocated more land (> 40 percent) to maize. Small grain area In contrast, the relationship between small grain area and land availability is less clear. Farmers who planted the largest share of their land to small grains (>40 percent) tended to have more land per capita (> 0.60 versus 0.55 hectares) than farmers that did not grow small grains. Also, households that grew more small grains tended to have larger families and more oxen. This may be explained by the fact that poorer (larger) households tended to depend more on small grains, as did older households (taste preferences) who had accumulated more traction capital over their lifetime. 126 5.16 Russia-cs m by 1“ we (.dim), Mote/Idai all than Districts, Zim, 1”. T—————_) Land Suple Nunber of Area Oxen Use Size Residents Available Omership (8) (fl) (HA) (f) Grain area a) < 69 S .79 2 69 - 82 73 7 .66 2 82 - 96 70 6 .72 2 > 96 72 6 .51 1 Isize sras G) < 12 71 8 .62 2 12 - 26 72 6 .73 1 26 - 60 72 5 .65 2 > 60 70 6 .50 1 -ll grain area m 0 23 S .55 1 1 - 63 89 6 .63 2 63 - 66 86 6 .66 2 > 66 86 7 .60 1 Source: Food Security surveys. 127 Cash crops Oilseed, cotton, and fruits and vegetables were generally grown as cash crops. The analysis of grain crops suggested that households first attempt to meet food needs through grain production. As expected, once these needs were met, farmers tended to grow cash crops. This hypothesis is supported by the data that shows that households allocating the largest share (> 27 percent) of their land to oilseeds, had the largest cropped area (0.77 hectares). Snmilarly, cotton producers cultivated slightly' more land (0.66 versus 0.62 hectares) than non-growers. The fact that cotton producers had more household members (8 versus 6) and more oxen may be explained by the greater labor intensity of these crops and its high profitability, which facilitates investment in draft power. In contrast, fruit and vegetable production is both land and labor intensive. The data suggests that land poor households (0.50 versus 0.64 hectares per capita) tended to grow fruits and vegetables which may enable them to more fully employ their larger (7 versus 6) household labor supply. Capital ownership Analysis of the data highlights the complementarity between resources available to households (Table 5.15). W Oxen were the most important capital asset in the communal areas. Households owning a full team (2 or more oxen) also had more family labor (5 versus 7) and access to more land (0.66 versus 0.55 hectares per capita). 128 Vials 5.15 W ash-it by mitsl ounarship (ndim), Mob/Mi ltd Idlers Districts, little, 1m. Capital Suple Huber of Area Oxen Omership Size Residents Available Ownership (8) U) ("M M) Physical ”ital r < 2 162 5 .57 0 > 2 163 7 .70 2 W 0 66 5 .57 0 > 1 261 6 .66 2 Fit-mist witsl 8111219 lo 275 6 .62 2 Yes 10 6 .85 0 129 Limsialmital Although few households (4 percent) reported borrowing from the AFC, borrowers had smaller families (4 versus 6) and less oxen (0 versus 2) than non-borrowers. On the other hand, borrowers from the AFC had more land (0.85 versus 0.62 hectares per capita). W There were only small differences in resource endowment between household heads who had attended school and those that had no education. This may be partially explained by the fact that older farmers had less opportunity to attend school, but younger farmers had less land, which shows conflicting influences. 5.7 External environment Parmers' income levels and structure is influenced by several factors exogenous to the household. This section presents the components of the external environment that define the household's opportunity set-- physical environment (rainfall), services, and technology. 5.7.1 Physical environment In semi-arid areas like Natural Regions Iv and v, rainfall patterns play a dominant role in guiding household resource allocation decisions. To evaluate rainfall patterns between the two sites, historical rainfall data were analyzed. Since long-term rainfall data were not available for the survey villages, the Mutoko and Middle Save rainfall stations were selected to represent the Mutoko/Mudzi and Buhera Districts, respectively. The Mutoko station was chosen because it was the closest rainfall station to the Mutoko/Mudzi survey area, although it was located in Natural Region III and our survey sites are in Natural 130 Region IV)”. Thus, the survey areas probably received both less rainfall and had greater year-to-year variability. The Middle Save station was selected to represent the rainfall pattern in Buhera since it is the station closest to the Buhera sites and is Natural Region v”. Comparison of the rainfall data for 1980/81 to 1983/84 shows that in Mutoko/Hudzi District, rainfall was substantially higher (706 versus 477 mm, four year average) than in Buhera District (Table 5.16). Both areas have similar long term intra-seasonal rainfall distribution, with November to March being the peak rainfall period. Furthermore, the coefficient of variation of rainfall is larger for Buhera (34%) than Nutoko/Mudzi Districts (26$) based on 32 and 36 year averages, respectively. These data indicate that Mutoko/Mudzi District has a greater agricultural potential, and implies that household resource allocation patterns will differ across sites, in order to cope with the differential risk associated with rainfall. 10Rainfall data from the Mutoko station is a valid proxy for Mutoko/Mudzi rainfall pattern, since Natural Regions are not distinct, but only general indicators of rainfall. 11The Buhera station is located in the Natural Region III portion of the district, so it was not selected. 131 rials 5.16 Rainfall pattern (silli-tsrs) by District, Zidadue, wan/a6. District JUL AUG SEP WT NOV DEC JAN Jill TOTAL 1 Make District 1980/81 1981/82 1982/83 1983]“ LR average'I ilIera District 1980/81 1981/82 1982/83 1983/86 : Average over 36 years (1952/53 to 1987/88). Average over 32 years (1952/53 to 1983/86). 5.7.2 Institutional environment 5.7.2.1 Access to services Rural services are important catalysts to agricultural development. Since Independence (1980), household access to output and input markets, grain processing, and public transport, education, health, veterinary services, and extension has improved. Yet, access still varied considerably across villages (Table 5.17). QEERHS.E§£B§E§ Households primarily marketed their crops through Grain Marketing Board (6118) depots and collection points, approved buyers‘z, non- approved buyers, and marketing (output) cooperatives. Only the non- approved buyers and cooperatives were located in villages. ”The. distance of households to approved buyers is similar to GMB depots in Mutoko/Mudzi Districts and collection points in Buhera District. 132 twle 5.17 Access to services by villna, ”Mi Districts, line, 1%.. Bmera District Hutoko/Hudzi Districts Service village villages 1 2 3 6 5 6 1 2 3 6 5 6 m s 6118 depots (has) 60 82 82 100 90 160 37 66 30 26 37 65 6118 collect points (kills) 50 52 12 30 20 20 na na na na na na Non-approved buyers (8) 0 3 5 2 6 O 0 0 0 0 O 2 Output cooperatives (fl) 0 0 0 1 O 1 0 0 O 0 0 0 [nuts Shops: dry goods (fl) 1 6 1 3 6 1 0 6 1 0 3 5 Shops: seed (8) 1 1 1 3 6 1 0 1 0 0 1 2 Shops: fertilized!) 0 1 O 0 0 O 0 1 0 0 1 0 "put cooperative (I) 0 0 0 0 0 0 0 1 O 0 0 1 Processim Maize sill (Inns) 5 0 2 36 5 2 2 0 na 0 na 1 Sorghun dehuller (Inns) 51 70 0 67 50 20 29 0 na na na 23 tm Buses per week (I) 3 2 16 1 0 1 0 2 2 2 3 6 Months without bus service 0 0 0 0 0 0 0 0 0 6 O 2 Edmation Primry schools (6) 1 2 1 0 1 1 0 1 0 1 0 0 Secondary schools (I) 0 1 0 0 1 0 0 0 0 0 0 0 health Clinics (fl) 0 1 0 O O 0 0 0 0 0 0 0 Veterimry services Cattle dips (I) 1 1 0 0 1 1 0 1 0 0 0 1 Extension Extension workers (8) 1 1 0 1 1 1 - Secur ty sueysrv. ‘na seam these services were not available. 133 Given the existence of alternative marketing channels, it was difficult to clearly assess the degree of access to markets across districts and villages. Although none of the villages were close enough to a GMB depot for households to transport crops by foot or ex cart, households in Buhera District were much farther (60-140 kms) from 6143 depots than households in Mutoko/Mudzi Districts (26-46 kms). 0n the other hand, only Buhera households had access to GMB collection points (located between 12-52 kilometers from their villages)“. Also, non- approved buyer were more prevalent in Buhera District (4 of 6 villages, compared to one village in Mutoko/Mudzi Districts). Thus, although Mutoko/Mudzi households have greater access to GMB depots, the presence of collection points and non-approved buyers in Buhera District provided households the opportunity to market surpluses W Households in both districts had limited local access to purchased inputs. Although all but two villages had dry good stores, only 74 percent of these stores sold seed and only 11 percent sold fertilizer. More specialized inputs, such as herbicides and insecticides, were even less available locally. In Mutoko/Mudzi Districts, poor access to privately sold inputs was partially mitigated by the existence of input- purchasing cooperatives in two villages, although no village in Buhera had an input cooperative. W Maize mills were generally more available to households than sorghum mills. Households in Mutoko/Mudzi District lived closer to maize mills (0-2 kms) than households in Buhera District (0-34 kms). Pew households lived close to a sorghum dehuller (one in each survey area), with the 130143 collection points had not been established in Mutoko/Mudzi Districts during the survey period. 134 median distance in Mutoko/Mudzi District being less (29 kms) than in Buhera District (49 kms). W Bus service was infrequent (median of 2 buses per week) in all but one village in each district, which both had daily service. Transport was available year around, except in two villages in Mutoko/Mudzi District, where roads were inaccessible during’ the higher rainfall periods. 811924134211 Village-based primary education was more accessible than secondary education. A majority of villages (67%) had primary schools, with better coverage in Buhera District (83%) than in Mutoko/Mudzi Districts (50%). In contrast, only two villages had secondary schools, both located in the Buhera District. 59.11311 In both areas, households had poor access to modern health care facilities-~only one village (Buhera District) had a health clinic. E§§§£135£¥_£§£2i2§§ The distribution of veterinary services reflected the greater importance of livestock in Buhera than in Mutoko/Mudzi. For example, 67 percent of the Buhera villages had cattle dips, compared to only 33 percent in Mutoko/Mudzi Districts. mm All villages were served by AGRITEX extension agents, but each agent was assigned to cover several villages. Buhera District households had better access to extension because more villages had resident extension 135 workers (83%) than Mutoko/Mudzi District (33%). 5.7.2.2 Changes in access to services since 1980 Since 1980, the Roz has invested heavily to strengthen rural services. Yet, in the survey villages, there has been little improvement in household access to services (Table 5.18). For example, since 1980 the number of non-approved buyers has increased in only two villages (one in each survey areas). Similarly, even though the number of dry good stores increased from 15 to 23, only three additional stores sell seed and one additional store sells fertilizer. Furthermore, access to primary and secondary education has changed little since independence. For example, government has constructed only two additional primary schools (one in each area) and two secondary schools (both in Buhera District). Similarly, local access to modern health care services has not improved significantly. For example, only one survey village (Buhera District) had a clinic, compared to none at independence. Veterinary services have improved only marginally since 1980, with household access to cattle dips increasing for only two villages, one in each survey area. Finally, access to extension services has changed little, as indicated by only one additional resident extension worker living in each survey area. 5.7.3 Technology Household adoption of improved technologies depends mainly on its appropriateness (technical suitability) and availability. This section uses household awareness and adoption of technology as proxy indicators of household access to technology. 136 Ifile 5.18 Chane in ru-al access to services since wan, Motollhflzi all liters Districts, 21m, was I'd was. Change in rural services sum. District Hutoko/Hudzi Districts "7 F 1 2 3 6 5 6 1 2 3 6 5 6 ‘1 1 l 1 Market access ‘ Ion-approved buyers (I) I 1980 0 1 2 2 6 0 0 0 0 0 0 0 ) 1988 0 3 5 2 6 0 0 0 0 O 0 2 ‘ . Shops selling dry goods (6) l 1980 1 1 1 3 6 0 0 2 1 0 2 3 ; 1988 1 6 1 3 6 1 0 6 0 3 5 ' . Shops selling seed (f) ‘ l 1980 1 1 1 3 6 O 0 0 O 0 0 1 1 a 1988 1 1 1 3 6 1 0 1 0 0 1 0 y Shops selling fertilizer-(6) { . 1980 0 1 0 0 0 0 0 0 0 0 0 0 . ; 1988 0 1 0 0 0 0 0 1 0 0 1 0 ‘ 1 Elastic: Primary schools (t) | 1980 1 2 O 0 1 1 0 0 0 1 0 0 ' 1988 1 2 1 0 1 1 0 1 0 1 0 0 Secondary schools (l) 1 1980 0 0 0 0 0 0 0 0 0 0 0 0 ; 1988 0 1 0 0 1 0 0 O 0 0 0 0 - health -' Clinics (6) ! 1980 0 0 0 0 0 0 0 0 0 0 0 0 J 1988 0 1 0 0 0 0 0 0 0 0 0 0 i Vetsrimry services jI Cattle dips (I) 1 1980 1 1 0 0 1 0 0 1 0 0 0 0 1988 1 1 0 0 1 1 0 1 0 0 0 1 music: ‘ ‘ Extension workers (8) I . 1980 1 1 0 1 1 0 0 1 0 1 0 0 i ‘ 1988 1 1 0 1 1 1 0 2 0 1 0 0 ' Souerc: Food Secur ty Surveys. 137 A majority of households reported they were aware of most technologies recommended by AGRITEX (Table 5.19). Across districts household awareness of recommended technologies was similar, except in Buhera District for technologies that were either not available (soil analysis) or more risky given the low level, and erratic distribution, of rainfall (soil liming and herbicide use). Mutoko/Mudzi District households reported higher adoption rates for a majority of recommended technologies, than in Buhera District. In Mutoko/Mudzi District, households had high adoption rates for most field preparation and variable input technologies (ie., contour construction, field pegging,and fertilizer and insecticide application) primarily because these technologies were better suited to the agroecological conditions found in this area. In addition, the greater adoption of fertilizer and insecticide by Mutoko/Mudzi District households was consistent with household land use (more area allocated to crops such as maize, cotton, fruits, and vegetables). In contrast, Buhera District households had similar or higher adoption rates for recomended technologies that required the use of animals traction equipment and animals than Mutoko/Mudzi District households. This result supports the earlier finding that Buhera District households owned more oxen and plows than Mutoko/Mudzi District households. 5.8 Su-ary Analysis of the survey data indicated that households differed considerably, in terms of access to owned resources--land, labor, and capital. Although these differences in resource endowment contributed to explaining inter-household variability, inter-village differences in the external environment were also important explanatory factors. 1J38 Idale 5.19 homology m as! edptim, Mokomdzi mi liters Districts, Zim, 1%. Awareness Adoption Recon-ended (X) (X) Technology Mutoko/Mudzi Buhera Hutoko/Mudzi Buhera Fielg Prgpgrggjgg ‘ Post harvest plowing 98 100 63 ‘ Secondary plowing 85 90 63 Field pegging 88 98 12 Tine ridging 96 56 . tied ridging 25 7 5 Contour construction 100 50 ' Soil analysis 55 9 16 . Variable igpgg . Kraal ssnure 98 60 hybrid seed 80 Basal fertilizer 95 58 2 Top dressing fart. 86 62 Soil lining 5 0 Insecticide 87 25 Herbicide 16 0 m hitch assembly 16 Yoke size 39 - mm Dry planting 50 Sole cr...’ _' _._ .. ._ _“ _ 52 139 Resource endowment Households in Buhera District had more land (hectares per household), labor (residents), and capital (oxen) than households in Mutoko/Mudzi Districts. In contrast, the distribution of labor (residents) and land was relatively equal across districts and the total sample, while oxen ownership was highly unequal. am Households averaged 6.6 resident and 4.5 nonresident members. Buhera District households had both more resident (7.4) and nonresident (5.3) family members than households in Mutoko/Mudzi Districts (5.9 and 4.3, respectively). Households in both districts had similar age and gender structures. Lang Households in Buhera District had greater access to land (1.0 hectares per capita) than Mutoko/Mudzi households (0.7 hectares per capita). Across districts, land use was similar in two ways. First, in both districts households cultivated a similar proportion of their total land (88-89 percent). Second, grains dominated household land use (78- 82 percent of cultivated area), followed by oilseeds (14-15 percent), intercropped (2-3 percent), fruits and ‘vegetables (1 percent), and cotton (0-1 percent). . Across districts, land use differed in three major ways. First, although households in both districts grew mostly maize and small grains, Mutoko/Mudzi District households allocated a larger share of their land to maize (37 percent) and less to small grains (36 percent) than did households in Buhera District ( 18 and 61 percent, respectively). Second, although households in both districts allocated the same proportion of their land to oilseeds, Buhera District households allocated a larger share to groundnuts (7 percent) and less 140 to sunflower (2 percent) than households in Mutoko/Mudzi District (3 and 10 percent, respectively). Finally, only survey households in Mutoko/Mudzi District grew cotton. These differences in cropping patterns reflect the preference of households in Buhera District to rely more on drought tolerant crops, since the rainfall is lower and more variable than in Mutoko/Mudzi Districts. 9.12mi Household capital assets included primarily livestock and draft-drawn euipment. Capital ownership had pronounced differences across districts. First, in Buhera District households owned more cattle (7.5 per household) and small ruminants (10.6) than households in Mutoko/Mudzi Districts (4.7 and 4.0, respectively). In addition, Buhera District households owned more plows (1.6 per household) and ex carts (0.4) than households in Mutoko/Mudzi District (1.0 and 0.1, respectively). These results reflect the fact that since Buhera District has a lower population density and is more arid, cattle play a more important role in the farming system than in Mutoko/Mudzi Districts. Income level and sources by resource ownership Labor availability contributed to explainning differences in income level and sources. For example, larger households earned less total income (NHR) and income from production for home consumption (PHC), but earned more from labor sales. Unexpectedly, there was no relationship between household head gender (except female heads received more' transfers) and income. With respect to land availability, households that owned more land employed more of their available labor on their own farm (is. as total cultivated area increased, PHC and farm sales increased, but labor sales declined). Second, households primarily produced small grains for home 141 consumption (ie. as small grain area increased, farm sales declined). Third, households grew oilseeds as a cash crop (ie. as oilseed area increased, farm sales increased). Finally, households pursued a food- first strategy (ie. as grain area decreased and oilseed area increased, net household receipts, production for home consumption, and farm sales increased). With respect to capital ownership, households who owned a full complement of animal traction equipment and animals earned higher total income and had larger farm sales than non-owning households. In addition, households that borrowed from the AFC (4 percent of the total sample) had much larger NHR and PHC than non-borrowers. External environment Several factors exogenous to the household influenced the level and structure of net household receipts. W In Mutoko/Mudzi District, rainfall was substantially higher and less variable (lower cv) than in Buhera District (Table 5.16). In contrast, both areas had similar intra-seasonal rainfall pattern, with peak rainfall occurring from November to March. 53:11.92: Although the R02 has made major investments to strengthen rural services, the survey villages had not benefitted greatly. Since 1980, there had been little improvement of household access to output markets (GMB depots, approved buyers, and non-approved buyers), input markets, primary and secondary education, modern health care, and extension services (Table 5.18). 142 13921112129! _ For most technologies recommended by AGRITEX, farmers in both districts reported similar levels of awareness. Yet, Mutoko/Mudzi District households have adopted more of the recommended technologies (especially variable inputs) than Buhera District households, except for technologies that required oxen or plows. This implies that these technologies were better suited to Mutoko/Mudzi districts more favorable agroecological conditions. CHAPTERVI DITERNINANTS OF INTER-HOUBIOLD VARIATION OF INCONBS Both endogenous and exogenous factors explain inter-household income variability. Endogenous factors are resources available to the household for allocating to competing economic opportunities ; including land, labor, capital, and purchased inputs. IExogenous factors are elements external to the household that influence how households allocate their endogenous resources; including institutions, technology, and the physical environment. In poorer agroecological areas, exogenous factors are particularly important determinants of household income. This chapter is divided into four sections“ The first section presents variables hypothesized to explain the inter-household variation in incomes. 'The second section examines regression results that identify the determinants of income (NHR). The third section presents and examines regression results that identify determinants of several subcomponents of income (NHR)--the value of total agricultural production, labor sales, and transfers (received). The final section summarizes the results of the chapter. 6.1 Regression variables This analysis attempts to explain inter-household variability in net household receipts and its major subcomponents--the value of agriculture production, labor sales, and ‘transfers (received)--using independent variables identified in Chapter 5. 143 1“ 6.1.1 Dependent variables First, net household receipts (NHR) was chosen as the dependent variable in the regression model of aggregate income. Because NHR is the most inclusive income measure, it best indicates a household's ability to meet its consumption requirements. Second, the three major subcomponents of NHR, (ie., the value of agricultural production‘, labor sales, and transfers (received)) were chosen as the dependent variable in the regression models of income subcomponents because they accounted for at least ten percent of net household receipts (NHR)2. The specifications of individual regression models are presented in sections 6.2 and 6.3. 6 . 1 . 2 Independent variables Both endogenous and exogenous independent variables were hypothesized to explain inter-household variation in total income, and its three major subcomponents. Due to the highly aggregated nature of net household receipts, it is not possible to infer a causal relationship between the NHR and the independent variables included in the model. On the other hand, it is possible to infer greater causality between the dependent variables in the components models (value of agricultural production, labor sales, and transfers received) and their associated independent variables, since there exists a clearer theoretical relationship between these independent and dependent variables. This section describes all of the independent variables used in all of the subsequent econometric models . 1Production for home consumption and farm sales are included in the agricultural production (value) model. 2Subcomponents not included. are nonagricultural sales (6 percent of NHR), business inventories (2 percent), and net credit receipts (3 percent). 145 6.1.2.1 Endogenous factors (to households) Households allocate both owned resources and purchased inputs to income-generating activities. 1. Owned resources. These include labor, land, and capital (physical and human). a. m: These variables measure the size and composition of the household labor assets. 1. Household size: Measures of household size are the number of resident members (RESIDENT), non-resident members (HHNONRES), and male non-resident members (M_NONRES). 2. Household composition: Measures of household composition are the number of adult equivalents3 (ADULTEQV) and the dependency ratio” (03111101) . 3. Gender of the household head: The household head's gender was incorporated in two ways: a dummy variable for male/female (HHSEX); and two dummy variables that distinguish between households that were male-headed, female-headed/male away (DHDSTATl) , and female-headed/no male (DHDSTATZ). 4. Household head's age: The household head's age was incorporated as both a continuous variable (HHAGB) and a dummy for age cohort group (D_HDAGEl and D_HDAGE2). b. Land: Land variables measure the amount, quality, access, or productivity of land available to the household. 1. Available area: Available area was measured by the amount of land (hectares per capita) available to the 3Adult equivalents are a measure of household size, adjusted for the age-sex composition of the household. ‘The dependency ratio is the number of resident household members per worker. 146 household (POLAND). 2. Land quality: Land quality was represented by the proportion of land which had average or excellent soil fertility (SFRTADEQ). 3. Land access: Land access was estimated as the mean distance of all fields from the household (DISTAREA). 4. Productivity: Land productivity was represented by the ratio of gross harvest value to the total cultivated area (P_TIVITY). c. Capital: Capital variables measure the availability of physical and human capital. 1. Physical capital: Three variables measure physical assets held by households. a. Plows: The number of plows owned per capita (PLOWPC) . b. Oxen: The number of oxen owned per capita (ostPC) . c. Animal traction: Two sets of dummy variables were constructed to measure animal traction availability. One of these variables differentiated households as to whether or not they had a full complement of equipment and oxen . (D_AT)5. Another set of variables differentiated whether or not a household was non-equipped, semi-equipped (D_AT1), or totally equipped (D_AT2). 2. Human capital: Two sets of variables represent the quality of human capital. a. Formal. education: Formal education is 5A full complement is defined as having a minimum of 2 oxen and one plow. 147 represented by a continuous variable that measures the number of years of school attended by the household head (EDUC_Q), a dummy variable to whether or not the household head attended school (D_EDUC), and a dummy variable to indicate whether or not the household head was literate in English (D_LIT). b. Nonformal education: Exposure to nonformal education is represented by the household head's attendance at extension.meetings or participation in AGRITEX's master farmer program. One dummy variable (D_EW) represents whether or not the head attended extension meetings; and two dummy variables represented whether or not the head attended meetings (never, sometimes (D_BWl), or always (D_EW2)). Similarly, one set of dummy variable differentiates whether or not the household head was a master farmer trainee (D_MFl) or already was a master farmer (D_MF2). 2. Variable inputs. 'These include household use of purchased inputs and borrowed credit. a. M: A continuous variable measured the amount (dollars per capita) of agricultural inputs used by the household (INPUT_PC). b. Q;g§1§__ggg: Use of government-provided credit is represented by both a continuous variable (AFC__PC) which indicates the dollars per capita borrowed, and a dummy variable which represents whether or not households borrowed (D_AFC) from the AFC. the 148 6.1.2.2 Exogenous factors (to households) Exogenous factors include the agroclimatic conditions that characterize the farming environment; including household access to marketing outlets, input markets, extension, and technology. 1. Agroclimatic conditions. Agroclimatic conditions are represented by a set of dumy variables, which were constructed to reflect household perceptions about the previous year's rainfall level, indicating ‘whether they ‘thought it had been a poor, average (D_RAINl), or good (D_RAINZ) season. 2. Services. These variables were constructed to measure household access to marketing outlets, input markets, and extension‘. a. Mgkgtipg outlets: Access to marketing outlets is a composite variable which includes distance to the. Grain Marketing Board's (GMB) depots and collection points, and the existence (in the village) of either non-approved buyers or marketing cooperatives. Households were evaluated as having either poor, average (D_MRTI), or good (D_MRTZ) access to marketing outlets. b. Inpgt_mgrkg§§: Access to input markets is represented by two dummy variables: which represent whether either improved seed (primarily hybrid maize) was sold in their village (D_INPUTl), or both improved seed.gpg fertilizer were sold in their village (D_INPUTZ). c. fixggpgigp: Household access to extension is represented by a dumy variable to indicate whether or not an AGRITEX extension worker lived in the village (D_EXT). ‘The Agricultural Finance Corporation (AFC) scheduled meeting daYs in all villages when households could apply for credit. Information was unavailable to distinguish inter-village access to credit services, so only household use of AFC credit was included 1n the analysis. 149 3. Technology. Technology was represented by variables that reflect household use of improved technology; including household access to input markets, use of variable inputs, and equipment ownership. 6.2 Determinants of net household receipts The independent variables in the net household receipts model provide insights about factors associated with inter-household variation of net household receipts. This section first presents the specification of a multiple regression model. Then, the results of the regression model are presented. 6.2.1 Model specification To explain inter-household variation in net household receipts, a multiple regression model-~using both endogenous and exogenous factors-- was fitted to the survey data using ordinary least—squares (OLS). Ammuuuusne The assumptions of the regression model, including linearity, are those comon. to a classical multiple regression model7, and when satisfied provide» OLS estimators which are both unbiased and consistent. An evaluation of whether or not the assumptions are satisfied is discussed for each regression model. 7The assumptions of a multiple regression model are (Pindyck and Rubinfeld (1981, pp. 75-76) : 1. The model is specified by the following equation: Y = B -F13X2 -+ + ... + B”)! + e 2. The independent variables are nonstochastic; and there is no exact linear relationship between one or more of the independent variables. 3. The error term has a zero expected value and constant variance for all observations. Errors corresponding to different observations are uncorrelated. The error variable is normally distributed. 150 e o 9 Five sets of independent variables are included in the econometric model; including variables that measure exogenous factors (services, agroclimatic factors, and technology), labor characteristics, land availability, capital ownership, and the use of purchased inputs. The final model specification is represented as: NHR t C + 3101+ BZDZ + ... + 86136 + 3711+ BBDT + B9X2 + smxs + B11x6 + ‘31.2% + 31309 + 311.910 + 815011 + 3149‘s + 31736 + 9 Where: NHR = Net household receipts C = Constant term D1 8 Output market access: medium (dummy) D2 = Output market access: high (dummy) 03 = Input market access: seed availability (dummy) 0, = Input market access: seed and fertilizer available (dummy) 05 = Rainfall rating: average (dummy) D6 = Rainfall rating: good (dummy) x1 = Household head age (years) D7 = Household head gender: male (dummy) x2 a Land per capita (hectares) x3 = Number of oxen (per capita) x, = Education of household head (years) DB a Household head is a master farmer trainee (dumy) D9 = Household head is a master farmer (dummy) D10 = Household head sometimes attends extension meetings (dummy) D11 8 Household head always attends extension meetings (dummy) x5 = Input use per capita (25) x6 = Amount borrowed from AFC per capita (25) e a Error term 6.2.2 Results of the model This section presents the degree to which the assumptions of multiple regression were met, and the results when the data were fitted to the model. 6.2.2.1 Satisfaction of the assumptions The model was examined for specification error (functional form), the relationship between independent variables, and the characteristics of the error term a 151 Pu C O A linear regression model was specified for two reasons. First, a linear relationship between the dependent and independent variables was chosen because no a priori information existed that suggested otherwise. Second, a review'of scatterplots between each independent variable and net household receipts also suggested a linear relationship. e o w en nde ndent variables The presence of multicollinearity was tested both by examining the correlation matrix of the independent variables, and observing if the T statistics are low and the goodness of fit (R2) is high. An examination of the correlation matrix identified that the highest correlation existed between the amount of land available and oxen ownership (a zero-order correlation coefficient of 0.60). To assess the impact of this correlation, the variable representing oxen ownership was then dropped, but the model was weakened, specifically it had less predictive power (F statistic) and a poorer goodness of fit (R2). Therefore, both variables were included in the regression model. 8 e e t The regression residuals were examined for heteroskedasticity, serial correlation, and the normality of their distribution to test if the regression model's results are BLUE (best linear unbiased estimators). Heteroskedasticity, or unequal variance of the error terms, was tested by reviewing the standardized scatterplot of the residuals of the actual and predicted values of net household receipts. This procedure demonstrated that except for a few outliers, the error term displayed a relatively equal variance. The presence of serial correlation was evaluated in two ways. First, a visual examination of the residual plot indicated that serial correlation did not exist. Second, the Durbin-Watson statistic was 152 calculated (DW - 1.91) for first-order serial correlation. These result imply that the null hypothesis-~that no serial correlation exists--should be accepted. Finally, the error term was evaluated for the normality of its distribution. An examination of a histogram of the standardized residuals showed that the distribution was generally normal, but slightly peaked. 6.2.2.2 Results of the net household receipts regression model The results of the regression model indicated that there was a strong statistical relationship between net household receipts and the included independent variables (Table 6.1) . The results of the regression model is divided into its performance and interpretation. e r ssi n de The performance of the regression model is evaluated using the F statistic, R2, and adjusted R?. The F statistic (24.79), which tests for linearity between the dependent and independent variables, was statistically significant at the one percent level. The included independent variables explained nearly 62 percent (R2 - 0.615), and 59 percent when adjusted for the degrees of freedom (adjusted R2 - 0.590), of the variation of net household receipts for the sample households. W The estimated regression coefficients provides several insights. First, household reliance on agriculture tolearn income is highlighted by the large magnitude and statistical significance of production related variables--land availability (1 percent level), oxen ownership (5 percent 153 this 6.1 Reressiai coafficimlts ms! test statistics for the scam-tric mdel aa-inim inter- halmdiold vsriatim of net hmmdiold receipts, ”mi ms! Idlers Districts, lime, 1%. Independent Regression Standard Mean Standard Variables Coefficient Error Deviation f or MH 1. Services a. Output markets Medium 69.36 * 51.68 0.07 0.26 High 73.27 1“ 60.26 0.32 0.67 b. lrput markets Seed 81.66 *1" 33.19 0.58 0.50 Fertilizer 27.71 38.61 0.18 0.39 2. Agroclintic Average 196.26 1"" 62.38 0.28 0.65 Good - 2.97 65.69 0.56 0.50 8 MH 1. Labor a. Mead's age (years) 0.08 0.80 68.86 15.73 b. head's gender 6.61 32.62 0.86 0.35 2. Land a. Per capita availability (ha) 121.98 1"“ 9.21 0.86 1.63 3. Capital a. Physical Oxen (I) 69.66 ”1' 21.72 0.37 0.62 b. lumen (head) Education (years) 8.37 ** 6.50 3.68 2.90 Master farmer progrul Trainee 110.56 ”*1” 39.19 0.10 0.30 Master farmer 6.56 55.69 0.05 0.21 Extension meeting attendance Sometimes -26.33 27.06 0.69 0.50 Always -61.09 65.02 0.12 0.33 6. Variable imut use a. [nut use per capita (ZS) 7.67 *1" 3.21 1.10 3.32 b. Amomt borrowed from AFC (28) 9.10 12.68 0.13 0.81 mm -89-88 74-75 Sl-aary statistics Smla size 285 Multiple R 0.786 R «pare 0.615 Adjusted R aware 0.590 F statistic 26.790 Sign. of F statistic .00005 level Significance level: * 20 percent 5 percent 1" 10 percent 1 percent 154 level), and input usage (5 percent level)8. This result complements earlier results which showed that households earn a majority of their net household receipts (62 percent) directly from agricultural production. The regression coefficients of land availability implies that, ceteris paribus, a one hectare increase in land (per capita) was associated with a 25121.17 increase in per capita net household receipts. 0f the resources available to households, land appeared to have the largest impact. Similarly, oxen ownership has a large regression coefficient, implying that an additional ox (per capita), ceteris paribus, was associated with an additional increase of 2549.64 to net household receipts. Finally, an additional dollar of inputs used, ceteris paribus, would result in an additional 257.47 in net household receipts. Second, the importance of factors exogenous to household was demonstrated by the statistically significant relationship between net household receipts and household access to output markets, input markets, and rainfall. Household access to agricultural marketing outlets .was statistically significant for both medium (20 percent level) and high access (10 percent level). Both variables impky a positive, but weak relationship between output market access and the level of net household receipts; though caution should be used when interpreting this result since other factors9 strongly influence whether households produced enough to participate in these markets. Household access to input markets also provided interesting results. First, the variable used to assess the impact of household access to a1:<:>res that sold improved seed (within village) was statistically s.i-gnificant (5 percent level) and large (2581.44), thus implying a strong .8Plow ownership was also highly correlated with net household receipts (0.59), but due to multicollinearity with oxen ownership and land availability, it was not included in the regression model. 911's cluding household resources (land, labor, and capital) and Other exogenous factors (access to input markets and rainfall). 155 influence on income. In contrast, household access to stores that sold both seed and fertilizer (within village) was not statistically significant (at the 20 percent level), possibly due to the generally low level of fertilizer use. The village rainfall dummy variables provided mixed results. The average rainfall (previous season) variable was statistically significant (1 percent level) and large (25194.24), illustrating ‘the impact. of rainfall on income. But good rainfall in the previous year was not statistically significant even, at the 20 percent level. Third, the proxies for human capital gave mixed results. Although education of the household head was statistically significant (10 percent level), an additional year of education, ceteris paribus, was associated with only in an additional 258.37 to net household receipts. The influence of whether the household head was involved in the master farmer trainee program was statistically significant (1 percent level), implying that household heads which participated in the program, ceteris paribus, earned an additional 25110.54 per capita. Yet, caution is needed in interpreting these results because it is impossible to discern the direction of causality”. On the other hand, the influence of the household head being a master farmer graduate and attendance at extension meetings was not statistically significant, even at the 20 percent level. Finally, household head characteristics were not significantly (20 percent level) related to the level of net household receipts. The nonsignificance of both the age and gender of the household head suggests that resources available to the household were more important than the household head's individual characteristics. 1°It is not possible to determine whether households with larger net household receipts wanted to be master farmers, or if the participation in the program had resulted in higher levels of net household receipts. 156 6.3 Determinants of household receipt components Net household receipts is composed of several components. For each major component, a regression model was estimated to explain inter- household variability. All regression models were fitted using ordinary least squares (0L8), and use the classical multiple regression model assumptions. This section presents the model specification and regression results for four specific household receipts components-~the value of agricultural production, labor sales, and transfers (received). 6.3.1 Determinants of agricultural production Agricultural production was the largest income source, comprising 62 percent of net household receipts. The value of agricultural production (25) was chosen as unit to aggregate all agricultural goods produced by the household. 6.3.1.1 Model specification Twenty independent variables--representing factors both exogenous and endogenous to the household--were used to explain inter-household variability of agricultural production (value). 157 The final model specification for the value of agricultural production is represented as: AGPROPC a c + 3,131 + 3202 + + 3,137 + sax1 + egos + swxz + Where: AGPRODPC ans + 31sz + 3139‘s + 31136 + 815% + 816010 * 3171311 + 318912 + 819"? + 320x13 * 9 Agricultural production (per capita) Constant term Output market access: medium (dummy) Output market access: high (dummy) Input market access: seed availability (dummy) Input market access: seed and fertilizer available ( dumy ) Rainfall rating: average (dummy) Rainfall rating: good (dummy) Extension worker lives in village (dummy) Household head age (years) Household head gender: male (dummy) Dependency ratio Land per capita (hectares) Mean distance to fields Number of oxen (per capita) Education of household head (years) Household head is a master farmer trainee (dummy) Household head is a master farmer (dummy) Household head sgmetimgs attends extension meetings ( dam? ) Household head always attends extension meetings ( dumy) Input use per capita (25) Amount borrowed from AFC per capita (25) Error term 158 6.3.1.2 Results of the regression model This section presents evidence that the classical multiple regression model assumptions were satisfied, and the results when the data were fitted to the model. a o tions Following the procedures outlined in section 6.2.2.1 (p. 189), the model was examined for specification error, the relationship between independent variables, and the characteristics of the error term. The estimated model satisfied all of the assumptions of a classical multiple regression model. t e model The independent variables explained nearly 38 percent, and 34 percent when adjusted for the degrees of freedom, of the inter-household variability of agricultural production (Table 6.2). The F statistic (8.10) was statistically significant at the one percent level. The estimated regression coefficients provide several insights. First, among endogenous variables, capital (physical and human) and land resources contribute towards explaining the level of agricultural production; highlighted by the large magnitude and statistical significance of the independent variables--oxen ownership (1 percent level), land availability (5 percent level), mean distance to fields (10 percent level), and whether the household head was a master farmer (5 percent level) or trainee (1 jpercent level). These relationships supported the results from the earlier regression model of net household receipts, which suggested a strong household reliance on agriculture. Oxen ownership had a large positive regression coefficient, implying that an additional ox (per capita) was associated with, ceteris paribus, an additional 2588.54 in the value of agricultural production, demonstrating the importance of timely land preparation. The impact of oxen ownership 159 Tdile 6.2 Regression coefficients and test statistics for the eccentric model ex-inim inter- hoimdlold variatim of the value of nricultis‘al predation, Mokollhmlzi and filters Districts, Zimbabwe, 1988II9. independent Regression Standard Mean Standard Variables Coefficient Error Deviation Wm 1. Services a. Output markets Medias -63.53 66.16 0.07 0.26 high 0.21 31.56 0.32 0.67 b. input markets Seed 76.32 **** 25.72 0.58 0.50 Fertilizer 56.66 ** 31.77 0.18 0.39 2. Agroclimatic Average 128.29 **** 36.50 0.28 0.65 Good 66.67 36.83 0.56 0.50 3. Extension 27.39 21.88 0.60 0.69 Eggggggggg factors (NH) 1. Labor a. Mesd's age (years) - 0.27 0.66 68.85 15.76 b. head's gender 6.67 25.06 0.86 0.35 c. Dependency ratio ~28.59 * 18.18 1.75 0.56 2. Land a. Per capita availability(ha)16.21 *** 7.17 0.86 1.63 b. Mean dist to fields (min) -1.19 ** 0.69 11.28 12.39 3. Capital a. Physical Oxen (i) 88.56 **** 16.80 0.37 0.62 b. human (head) Education (years) 2.92 3.53 3.67 2.89 Master farmer program Trainee 92.61 **** 30.27 0.10 0.31 Master farmer 82.39 ** 62.83 0.05 0.21 Extension meetings Sometimes -15.26 20.86 0.68 0.50 Always -33.03 36.59 0.12 0.33 6. Variable input use a. Input use per capita (25) 6.13 *** 2.55 1.11 3.32 b. Borrowed from AFC (25) 13.53 * 9.81 0.13 0.81 mm 15-68 72-95 Summary statistics Sample size 285 Multiple R 0.619 R square 0.386 Adjusted R square 0.336 F statistic 8.096 Sign. of F statistic .00005 level Significance level: *1 20 percent '** 5 percent ** 10 percent **** 1 percent 160 was larger, as expected, on agricultural production than on net household receipts. Although highly significant, the coefficient for land availability was smaller (2514.21) in agricultural production income model, compared to the net household receipts model. The»mean distance of all fields to the homestead was negatively related to agricultural production, possibly suggesting that fields farther away from the homestead received less crop management inputs. Households whose head was either a master farmer or trainee had large positive regression coefficients (2582.39 and 2592.41, respectively), implying that participating households benefitted from the advice provided. In contrast, household head's education (years) and attendance at extension meetings were not statistically significant, even at the 20 percent level. Second, the importance of factors exogenous to the household was illustrated by the statistically significant relationship between the value of agricultural production and household access to input.markets and rainfall. Household access to inputs markets was significant for both villages with stores that sold only seed (1 percent level) and stores that sold fertilizer (10 percent level). Both variables imply a positive relationship between agricultural production and access to input markets, but the relative size of the coefficients imply that access to seed stores was more important. The village rainfall dummy variables gave mixed results. Average rainfall in the previous season was significant (1 percent level) and large (25128.29), suggesting a strong influence of rainfall on agricultural production. In contrast, the dummy variable representing good rainfall in the previous season was not statistically significant, even at the 20 percent level. (explain) Third, the influence of household labor characteristics on agricultural production was weak. Only the variable ”household's dependency ratio" (eg., the ratio of household size to the number of workers) was weakly significant (20 percent level). Both the household head's age and gender 161 were not statistically significant, even at the 20 percent level--further reinforcing the argument that physical resources (land and capital) were more important in explaining agricultural production than household head characteristics. Finally, household use of purchased inputs and credit were both significant (5 and 20 percent levels, respectively) and large (256.13 and 2513.53, respectively), implying that their use has a positive impact on agricultural production. However, caution should be used in evaluating the impact of purchased inputs and credit because the mean value used by households was low (251.11 and 250.13, respectively). 6.3.2 Determinants of labor sales Labor sales comprised an average of 15 percent of net household receipts for the entire sample, ranging from 1 to 51 percent across villages. Payment for these services was primarily received as cash (95 percent). Although labor was predominantly sold by the household head and spouse (85 percent) for agricultural services (greater than 80 percent), this dependent variable also incorporates labor sold for other purposes. 6.3.2.1 Model specification Fourteen independent variables--representing factors both exogenous and endogenous to the household-~were used to explain the inter-household variability of labor sales. 162 The final model specification for labor sales is represented as: LS(PC) - c + 31131 + 32132 + 3303 + 3,0,. + 135x1 + 3605 + 37% + 38x2 + 39x3 + 8101!, + BHXS + 81207 + 313138 + B1‘X6 + e Where: LS(PC) = Labor sales (per capita) C = Constant term D1 = Output market access: medium (dummy) D2 = Output market access: high (dummy) D3 = Input market access: seed availability (dumy) D, = Input market access: seed and fertilizer available (dummy) x1 = Household head age (years) D5 8 Household head gender: female head/ male away (dummy) D6 2 Household head gender: female head/ no male (dummy) x2 = Dependency ratio x3 = Land per capita (hectares) x4 = Land productivity x5 = Education of household head (years) D7 a Household head is a master farmer trainee (dummy) D8 = Household head is a master farmer (dummy) x6 = Input use per capita (25) e = Error term 6.3.2.2 Results of the regression model This section presents whether the classical multiple regression model assumptions were satisfied, and the results when the data were fitted to the model. Was Following the procedures outlined in section 6.2.2.1 (p. 189), the model was examined for specification error, the relationship between independent variables, and the characteristics of the error term. The estimated model satisfied all of the assumptions of a classical multiple regression model. 163 Wane]. The independent variables explained nearly 27 percent (23 percent when adjusted for the degrees of freedom) of the inter-household variability of labor sales (Table 6.3). The F statistic (6.86) was statistically significant at the one percent level. The estimated regression coefficients provide several insights. First, female-headed households--whether there was no male or he was away--had a strong negative relationship with labor sales by resident household members. In both cases, the coefficient was large and negative (-78.76 and -77.46 for female-head with male away and no male, respectively), and highly statistically significant (1 percent level), suggesting that both spouses are needed for households to take advantage of available employment opportunities. Second, variables associated with labor composition were also important in explaining the variation in labor sales. The houseshold's ”dependency ratio" was significant (5 percent level), implying that households with more children (is. , a smaller number of workers per household) sought wage employment, possibly due to their greater need for additional income to supplement their agricultural income. Third, the relationship between agricultural related variables and labor sales was mixed. Although land productivity and the level of purchased inputs were statistically significant (1 and 20 percent levels, respectively), land productivity had a small positive influence (0.13) and input usage had small negative influence (-2.22) on labor sales. The amount of available land was significant (1 percent level), but small (251.54), implying that additional land would have had little impact on household labor sales. Fourth, among variables measuring human capital resources, only household participation in the master farmer program, whether a master farmer or trainee, were significant (1 and 5 percent level). Households heads that participated in the master farmers had a negative statistical 166 This 6.3 legreasim coefficients mu test statistics for the mtric ml minim inter- hoimduold varieties: of liar sales, Moko/Idzi ml! Mere Districts, litmus, 1”. Independent Regression Standard Standard Variables Coefficient Error Mean Deviation x r h 1. Services a. Output markets Medit- 75.82 '1'" 16.81 0.08 0.26 high -0.33 9.89 0.32 0.67 b. "put markets Seed 12.28 11.33 0.58 0.50 Fertilizer -6.56 13.83 0.18 0.39 5% factors (NH) 1. Labor a. head's age (years) 0.03 0.36 68.86 15.76 b. head's gender Female headed/male away -61.21 “1” 19.17 0.85 0.36 Female headed/no male -57.63 "'1' 22.56 0.10 0.31 c. Dependency ratio 23.66 1'“ 9.25 1.75 0.56 2. Land a. Per capita availability (ha) 1.56 m‘ 9.21 0.86 1.63 b. Productivity 0.13 '1'" 0.02 89.77 179.25 3. himsn capital Education (years) 1.68 6.50 3.67 2.89 Master farmer program Trainee 26.75 m 13.66 0.10 0.31 Master farmer -67.63 '1'” 19.57 0.05 0.21 6. Variable irput use a. "put use per capita (25) -2.21 ** 1.26 1.11 3.32 92mm 3-24 33-68 Sunary statistics Samle size 285 Multiple R 0.515 R scansre 0.265 Adjusted R aware 0.227 F statistic 6.86 Sign. of F statistic .00005 level Significance level: * 20 percent ”1' 5 percent *1' 10 percent "*1" 1 percent 165 relationship with labor sales (they sell 2544.39 less labor), implying that households with master farmer heads focused more on their agricultural activities. Conversely, household heads that were training to become master farmers had a positive statistical relationship with labor sales. Both the level of formal education and attendance at extension meetings were not statistically significant (20 percent level) in explaining labor sales. This is not surprising, given that most local employment opportunities were for agricultural labor. Variables representing physical capita1--oxen and plow ownership--were initially included in the model, but were dropped both because they were not statistically significant and were correlated with other independent variables. Finally, the importance of factors exogenous to the household provided mixed results. Only medium household access to output markets was significant (1 percent level), suggesting that access to marketing outlets (primarily non-approved buyers) encouraged households in these villages to intensify agricultural production. Household access to input markets and high access to output markets were not statistically significant, even at the 20 percent level. 6.3.3 Determinants of transfers (received) Transfers, on average, comprised 15 percent of household receipts for the entire sample, ranging from 4 to 31 percent across villages. Although households received transfers from both government and private sources, almost all (99 percent) were from private sources (primarily remittances). 6.3.3.1 Model specification Fifteen independent variables--representing factors both exogenous and endogenous to the household--were used to explain the inter-household variability of transfers received. 166 The final model specification for transfers received is represented as: TRI = c + 31131 + 32132 + + 3909 + smx1 + snx2 + s12):3 + aux, + Bulls + Ihsx6 + e Where: TRI = Transfers (received) C = Constant term 01 = Output market access: medium (dummy) D2 = Output market access: high (dummy) D3 = Input market access: seed availability (dummy) D4 = Input market access: seed and fertilizer available (dummy) D5 = Rainfall rating: average (dummy) D6 = Rainfall rating: good (dummy) D7 8 Household head age cohort group: 35 to 55 years old (dummy) D8 = Household head age cohort group: more than 55 years old (dummy) 09 = Household head gender: male (dummy) x1 = Dependency ratio x2 = Number of male nonresident household members x3 = Land per capita (hectares) x‘ = Land productivity x5 = Education of household head (years) x6 - Input use per capita (25) e 8 Error term 6.3.1.2 Results of the regression model The model was examined for specification error (functional form), the relationship between independent variables, and the characteristics of the error term e 532W Following the procedures outlined in section 6.2.2.1 (p. 189), the model was examined for specification error, the relationship between independent variables, and the characteristics of the error term. The estimated model satisfied all of the assumptions of a classical multiple regression model. 167 We]. The independent variables explained nearly 25 percent (20 percent when adjusted for the degrees of freedom) of the inter-household variability'of transfers (Table 6.4). The F statistic (5.74) was statistically significant at the one percent level. The estimated regression coefficients provide several insights. First, the age of the household head was positively related to the level of transfers received. Both middle-aged and older household heads were statistically significant (10 and 1 percent, respectively), with older household heads receiving larger amounts of transfers (2530.36) than middle-aged households (2525.18). This result suggests that older household heads had offspring that were old enough to help supplement their family's income. Second, the number of male nonresident household members was also statistically significant (1 percent level), suggesting that male nonresident members had greater income earning opportunities than females. This result is consistent with cultural practices that associate obligations for males with their household, and females with their husband's household. Third, there was a relationship between agriculture and transfers received. Both agricultural productivity and purchased input use were statistically significant (10 and 1 percent level, respectively). Agricultural productivity and transfers received had an inverse relationship, implying that nonresident family members played an important role in assuring their household had an adequate income level. The positive relationship between transfers received and use of purchased inputs demonstrates the importance of transfers as a source of cash to invest in agriculture. 6.6 {09' {3914 mm! M- / lndeoe'dent variables #— {1&5 factor 1. Services :. Mput ear hedlur High b. Imut mark Seed Fertil Z. Agroclllstic Averag Good £1096on facto 1. Labor :. Head‘s age Middle Older b. Head's gem C- ‘5me :1. Inner of l 2. lard '« Per capita h. Proactivit 3- M capital Educati. l. Variable input " “Wt use p m \ 5"? statistics le si: Mlllple ) R 5951‘: Adimmed g F statisti Sim. of r 1J58 Tble 6.6 Ragrassiai coefficients ms! test statistics for the mtric ndel ex-inim inter- hotmdiold veriatim of trmmfers received by hoimdlold, ”Mi ml! Idlers Districts, Zifidsae, was/a9. Independent Regression Standard Mean Standard variables Coefficient Error Deviation Wm 1. Services a. Output markets Medium 1.16 18.20 0.08 0.26 high -6.66 12.95 0.32 0.67 b. input markets Seed 20.16 ** 12.36 0.58 0.50 Fertilizer 10.27 16.10 0.18 0.39 2. Agroclimatic Average 69.98 **** 16.02 0.28 0.65 Good 16.13 16.83 0.56 0.50 f c r MM 1. Labor a. head's age cohort group Middle aged 26.08 ** 12.96 0.26 0.63 Older 29.62 **** 10.25 0.61 0.69 b. head's gender -7.86 11.61 0.86 0.35 c. Dependency ratio 8.50 8.52 1.75 0.56 d. Mixer of male nonresidents (R) 13.66 m‘ 3.67 0.87 1.13 2. Land a. Per capita availability (ha) 3.06 2.69 0.86 1.63 b. Productivity -0.06 ** 0.02 189.77 179.25 3. human capital Education (years) -0.38 1.65 3.67 2.89 6. Variable input use a. Input use per capita (25) 7.95 **** 1.21 1.11 3.32 Qggg;ggt_§ggg 12.51 25.57 Summary statistics Sample size 285 Multiple R 0.695 R square 0.265 Adjusted R square 0.203 F statistic 5.76 Sign. of F statistic .00005 level Significance level: * 20 percent *** 5 percent ** 10 percent ***‘ 1 percent Final were Bi! variable average reepecti and less to outp rainfall 6.4 Sui Intel income) labor (Contro Control increas Firs inter-h agrocli maIketa libor 2 Sec: 169 Finally, although the importance of factors exogenous to the household were significant, their interpretation is problematic. Although the variables, household access to stores that sold seed and villages with average rainfall, were statistically significant (10 and 1 percent levels, respectively), one would expect households in poor rainfall environments-- and less access to inputs--to depend more on transfers. Household access to output markets, access to stores that sold fertilizer, and good rainfall were not statistically significant, even at the 20 percent level. 6.4 Summary Inter-household variability of per capita net household receipts (total income), and three subcomponents (ie., value of agricultural production, labor sales, and transfers received) is influenced by endogenous (controlled by the household) and exogenous (beyound the household's control) factors. 'This analysis identifies constraints households face to increase their income. First, factors exogenous to the household were important in explaining inter-household variation in total income (output and input markets and agroclimatic conditions) and value of agricultural production (input markets and agroclimatic conditions) models, but explained little in the labor sales and transfers (received) models. Second, endogenous factors--primarily access to land and oxen--were important in explaining inter-household variation in total income and the value of agricultural production, but explained little in the labor sales and transfers (received) models. This result demonstrates household dependence on agriculture, and suggests that poorer households could have increased their total income and agricultural productivity by acquiring more land and oxen. Third, household labor characteristics were important in explaining inter-household variation in labor sales (household head's gender and the dependency ratio) and transfers (household head's cohort group and the number of value of hypothes: agricultt househol' advantag- with tra 170 number of male nonresidents), but explained little in the total income and value of agricultural production models. This result further supports the hypothesis that physical resources are more important in explaining agricultural production. Furthermore, it suggests that individual household characteristics determine whether households are able to take advantage of local employment opportunities or supplement their income with transfers. Policies must take i effect of interest 1'; agriculture agricultural or saving to creating an . Second, p01 i economic cha inpact of po] household ac: Sadoulet, 195 If Policy PIObable Cone on househOIdg theY would b CHAPTER VII RURAL 13mm STRATEGIES Policies that maximize the impact of government on rural development must take into account two factors. First, they must anticipate the effect of macroeconomic policies--exchange rate, foreign exchange, interest rates, and inflation policies--on the performance of agriculture and national development objectives. For example, the agricultural sector can promote national development by l.) generating or saving foreign exchange, 2.) reducing inflationary pressure, and 3.) creating an effective demand for commodities produced by other sectors. Second, policies must be designed with an understanding of the socio- economic characteristics of the poor (and the subsequent short-run impact of policy changes on the poor) since their impact will depend on household access to resources and their enterprioe mix (de Janvry and Sadoulet, 1989). If policy makers had sufficient information to evaluate, ex ante, the probable consequences-—intended and unintended--of alternative policies on households with different income structures and productive assets, they would be able to identify policies having the most desirable effects on food insecure households (Strauss, 1986; Bigman, 1985; and Behrman and Murty, 1982). This chapter draws on the insights from previous analysis to examine the effect of current policies on rural households, and suggests a rural development strategy to raise incomes of the rural poor. First, a typology of households is presented, based on access to resources. Second, the effect of current policies and services on different 171 household 1 gmpoeed 1 technologY change . 1.1 House Analysi strategy d access to Zimbabwe ( p «C ant as 172 household types is evaluated. Finally, rural development strategies are proposed under two scenarios--first, for the short run (existing technology), and then for the long run, assuming proposed technological change. 7.1 Household typology Analysis in Chapter 6 found that a households' income-earning strategy depends on its access to land, labor, and capital. Based on access to resources, three types of rural households are found in rural Zimbabwe (Table 7.1). , l. Beggggg§;pggg_hggggnglg§. These households were in the bottom two per capita income quartiles. Although they had relatively more labor, they owned relatively little land and capital assets (oxen), and used minimal purchased inputs. 2. flaggigglgfigrm_hgg§§nglg§. These households were in the upper- middle per capita income* quartile. .Although they were also relatively land poor, these households had less labor but owned more capital assets (oxen), and used more purchased inputs compared to resource-poor households. 3. §m§11;£§zm_ngg§ghglg§. These households were in the upper per capita income quartile. In contrast, although these households had the least amount oflabor, they owned twice as much land, more capital assets (oxen), and used considerably more purchased inputs than resource-poor and marginal-farm households. Across the villages studied, mean incomes (per capita) ranged from 25115 to 25390. Observed differences in access to resources also serves to explain these inter-village income differences. Villages with a higher mean incomes, had a larger proportion of their households in the upper income quartile (Table 4.2), and these households had greater access to assets. For example, in village #1 (Buhera District), the 52 percent of the households in the upper quartile owned more than twice the land (per capita), and almost twice as many oxen, as households in the other three quartiles. Alternatively, in village #4 (Mutoko/Mudzi District), the 44 percent of the households in the lower income quartile, owned only 65 percent as much land (per capita), and 55 percent as many oxen, as households in the other three quartiles. ml: 7.1 Variat 1'0 ‘9 \ Ptffor. Labor : Land p Prod, ‘§ SWFCQ: F 173 File 7.1 a-ry of bin-dield assets as! perfor-Ice nests-es across hundiold types, "Wi and mu Districts, um, 1W. assumes-econ MARSHAL-FARM SHALL-FARM I : "WSEHOLDS HWSEHOLDS WSEHOLDS 1 % loudneld assets ‘ 3' Land (per capita) 0.7 e 0.7 a 1.1. b 1’ Residents (I) 7.8 a 5.8 b 5.0 b i | Capital assets } Oxen (R) 1.5 a 2.0 ab 2.5 b : Plows m 1.3 1.2 1.1. 1 Variable capital 1 input (23/c)” 4.3 5.7 7.7 I I ‘9 imtmt 6.6 a 15.0 a 35.7 b 1 Perfor-Ice measures I Labor prodactivity" 102 a 231 b 585 c 1 Land ”mtfvityd 116 a 225 b 302 c I Pr“. W 80 a >100 b >100 c I i Source: Food Security surveys. ‘ Dmcan's Multiple Range test was used to assess the statistical significance of the difference of means, when there are three or more grows (meam). Riders that are statistically different (5 percent level) across household types have different letter(s) assigned to them. No letter after a tuner signifies that there was no b statistically significant difference across household types. includes only those households that purchased imuts. ‘ Ratio of net household income to the amber of mu equivalents. Ratio of the gross harvest value to total cultivated area (hectares). ' Percent of food and clothing met throidi promotion for home consuption (including inventories). Given policies: needed” ‘ (Eicher: I 1.1 Bffeci Since incomes t access (i education, Althoug revolutior households households policies, impacts, based rura Pricing pc In 2m] increase s dClaimant c AlthOug the new marginal- to “debt other ent: (residezit) “Tied lass were 1988 E uterpriae 174 Given this socio-economic differentiation of households, "diverse policies, technological packages, and institutional innovations are needed” to address the varying needs of different household types (Eicher, 1990). 7.2 Effect of current policies and services on rural income Since 1980, the Government of Zimbabwe has sought to increase rural incomes through policies designed to influence output prices, market access (inputs and outputs), credit, extension, agricultural research, education, and small scale enterprises. Although these policies are credited for Zimbabwe's agricultural revolution, they have largely benefitted marginal- and small-farm households, and had minimal (or negative) impact on the resource-poor households (Table 7.2). The following analysis reviews the respective policies, and identifies factors responsible for their differential impacts, in order to identify opportunities to stimulate more broad- based rural development. Pricing policies In Zimbabwe, government has relied heavily on incentive prices to increase small grain, maize, and oilseed production since these are the dominant crops in low rainfall areas (Chapter 2) . Although government anticipated that these policies would increase the incomes of all communal farmers, they have primarily benefitted marginal- and small-farm households because these households had access to under-utilized resources, or were able to reallocate resources from other enterprises. Although resource-poor households had more labor (resident) than either the marginal- or small-farm households, they owned less land, less oxen, and invested less in agriculture. Thus, were less able to take advantage of price incentives to alter their enterprise mix and thereby generate an agricultural surplus--even in the \\ Existi Polici and se ......— Price Susi Iaiz Oils Iarket Inpu am: I Credit I Extens I ASricu Resear. ~ Educat \ Slallfl enterp // fierce: l 175 Ible 7.2 81.7 effects' of cts-rot W policies across handheld types, Mtokonhflzi lid Illera Districts, Zim, 1m. . Existim RESWRCE-m MARGIMAL-FARM SMALL-FARM .‘ Policies < < 2:139 ) (2:139 - 23243) c > 23243 ) l and services sun” c‘ a“ use” cc In" use” cc Md " : Prise ] Small grain 0 - 0 0 0 0 + + + g } Maize o - o + + + . | ; Oilseeds 0 0 0 + + + + + + { J Market ggggsg ‘ lrputs 0 0 0 + 0 + + 0 + I Outputs 0 0 0 + 0 + + 0 + ’ Credit 0 0 0 0 0 0 + + + ’ Extension 0 0 0 0 0 0 + 0 + 1 Agricultural ‘ Research 0 0 0 + + + + + + \ Education 0 0 0 0 0 O 0 O 0 Small-scale enterprises 0 0 0 O 0 0 0 0 0 Souce: tisur—veys" . ___-_____ #v‘i ‘ 0 8 no effect; + = positive effect; and - a negative effect. b Met household receipts ‘ Cotisuiption ‘ Marketing rel! resc incr COIlS Markt H: house marks Re (fert produ a prc house} To agriCt farm 1 oilsee market 176 relatively good rainfall season such as 1987/88. In addition, since resource-poor households purchased more maize (and mealie meal), increasing the producer price actually has had a negative effect on consumption (increases cost of food budget). Market access Household participation in input and output markets1 depends on a household's ability to bear environment-related production risk (input markets) and generate an agricultural surplus (output markets). Resource-poor households purchased few agricultural inputs (fertilizer, insecticide) other than hybrid maize seed, due to the risky production environment (Table 7.1). The high rainfall variability makes a profitable return uncertain; and if the crop fails, resource-poor households had insufficient capital to bear the risk. To participate in output markets, households must generate an agricultural surplus. But in the study areas, only marginal- and small- farm households owned sufficient land and/or oxen to generate maize and oilseed surpluses; and only the small-farm households produced a marketed surplus of small grains (sorghum and millet). Credit Household access to credit appears unbiased, since the Agricultural Finance Corporation (AFC) held meetings in all villages and accepted loan applications from all applicants. Although few households borrowed from the AFC, small-farm households borrowed most frequently and took larger loans. A likely explanation is that small-farm households were more likely to borrow because they were better able to bear the consequences of crop failure by being able to repay a loan from savings. 1Household access to input and output markets was a village phenomenon, not varying across household types. cont: their Hutoi they atten these or th AgriC‘ Un1 const: rainf; Commui on DE chemi¢ Six Viriei 177 Extension Although AGRITEX extension workers are mandated to work with all farmers, time and resource constraints make this impossible. First, the low extension agent to farmer ratio (1:817, 1:946, and 1:900 for Mutoko, Hudzi, and Buhera Districts, respectively), limited potential farmer contact. Second, some extension agents had no transport, thus limiting their geographical coverage. Finally, in many cases (primarily in Mutoko/Hudzi Districts) extension agents didn't live in the villages they served, further limiting their contact with farmers. Analysis of the survey data showed that small-farm households attended AGRITEX extension meetings more frequently, which suggests that these farmers either perceived extension advice to be of greater value, or that extension was biased towards recruiting wealthier farmers. Agricultural research Until 1980, agricultural research primarily addressed production constraints associated with the agro-ecological (mainly soil quality and rainfall) conditions of comercial farmers (and to a lesser extent communal farmers in Natural Regions I, II, and III). Research focused on mechanization, hybrid seed (primarily maize), and management of chemical inputs (fertilizer, insecticide, and herbicide). Since 1981, DRESS has reoriented the agricultural research agenda to address the needs of communal farmers in three ways: 1.) they conducted varietal trials under conniunal area conditions: 2.) they initiated a breeding program for sorghums and millets; and 3.) they established a farming systems research unit to study communal farmers' problems, develop a FSR model, and provide information to assist policy makers. To date, new technologies for low rainfall communal areas (eg., hybrid sorghum and improved tillage methods) are still in the development phase (Shumba, 1990; Mushonga, 1986; and Nyati and Nyamudzera, 1984). Currently, hybrid maize and fertilizer are the most "appropriate" of the ai rnnfi adopti Becau: resou: const: Throug is a coeffi Distri resouz given Educat Alt more benefi Alt buildi Slippli BChOOl To their inCOme 178 the available technologies for increasing agricultural production in low rainfall areas since both are land-extending technologies. Hybrid maize adoption is high because it is the only readily available maize seed. Because fertilizer is land-extending, one would normally expect resource-poor and marginal-farm to apply more, since they face a land constraint. Yet, fertilizer use was highest for small-farm households. Throughout the study area, very few farmers used fertilizer because it is a risky technology, given the high variability of rainfall (a coefficient of variance of 26 and 34 for Hutoko/Hudzi and Buhera Districts, respectively). Furthermore, adoption was lowest for resource-poor households because they were least able to bear the risk, given their lack of financial capital. Education Although the government has extended rural access to education to more communal households since Independence, many children have not benefittedz--especially children of resource-poor households. Although primary education was "free”, parents were required to pay building fees (maintenance, repair) and purchase uniforms and school supplies: which resource-poor households were less able to afford (ie., 92 percent of resource-poor households' children attended primary school, compared to 98 percent for small-farm households, respectively). To attend secondary school, parents must pay school fees to enroll their children. Because these fees were high, relative to the household income, many children weren't able to attend school. For example, only 38 percent of the children of resource-poor households attended secondary school, compared to 68 and 90 percent of the marginal- and 2Primary and secondary school enrollment was similar across districts, with Mutoko/Mudzi District having slightly higher enrollment than Buhera District. small‘ re sour which financ Snail- Gov focuse and en effect partic and Bi resour invoiv produc market 7.3 s. The develo: of the EXtent PhYBic. rural resouri key p1 househc °°mpron ”ample 179 small-farm households, respectively. Thus, a smaller percent of resource-poor households' children earned secondary education diplomas which enable them to migrate and find urban jobs to help their family financially (remittances). Small-scale enterprises Government interventions to stimulate small-scale enterprises have focused on creating the Small Enterprise Development Corporation (SEDCO) and encouraging cooperatives in rural areas. These programs have had no effect in our survey area since none of our sample reported participating in SEDCO-sponsored activities. While few marginal-farm and small-farm' households sold nonagricultural products, almost no resource-poor households reported sales. Furthermore, most sales involved traditional, natural resource-based (eg., clay and reeds) products that were generally of low quality. Therefore, production and marketing of these products were limited geographically. 7.3 Strategies to accelerate rural development The previous analysis suggests that to date, agricultural-based rural development policies have had minimal impact on increasing the incomes of the majority of rural households. The resource-poor--and to a lesser extent the marginal-farm--households did not own sufficient land or physical capital to benefit from these policies. Therefore, a broader rural development strategy is required that addresses the needs of resource-poor households. A successful rural development strategy must take into account four key practical realities. First, policies must take into account- household resource endowments and local economic opportunities, but not compromise national food security and macroeconomic goals. For example, policies intended to help resource-poor households should increase the demand for the factors of production (quality and quantity) to 1' gover Se const produ quick resou: benef. adjust housel small help t Thi consun prefer 900:, incent farmer or exp Pin sPecif Policy m°8t i: belief i1 leakag, nudersi This 11101-3“ lcng n ChRCQEs 180 to which poor households have access, but not interfere with government's broader objectives (eg., efficient resource allocation). Second, unless policies are targeted to relieve the resource constraints facing the resource-poor, households with access to key productive resources (ie., land and oxen) will be able to adjust more quickly and more vigorously to general policy interventions than resource-poor households, and thus capture the policies'intended benefits. For example, in the mid 1980s when relative prices are adjusted to encourage the production of small grains, only wealthier households were able to reallocate available resources to generate a small grain marketable surplus, thereby undermining the objective to help the rural poor. Third, policies must avoid introducing long-run production or consumption distortions by taking into account both consumer tastes and preferences and production possibilities of all households (wealthy and poor, rural and urban). For example, Zimbabwe's introduction of an incentive producer price for small grains, designed to help communal farmers, induced large grain stocks for which there was limited domestic or export demand. Finally, policy makers should decide whether they want to help a specific group (targeted policy) or all rural households (nontargeted policy). Targeted interventions reduce the cost of helping individuals most in need of assistance (resource-poor households), but often provide benefits to nontargeted individuals (leakages). Therefore, to minimize leakages, targeted interventions must be designed with a clear understanding of the socio-economic characteristics of the poor. This section first examines a short-run rural development strategy to increase incomes of the rural poor, given current technology. Then, long run rural development strategies, which incorporate technological changes, are explored. 7.3.1 She: In the agricultux short-run the agric because t1 extension) poor house land. Pi survey ar specifical viii requi which ads; thid-ridge: environmem from currei Second, are likely resOurce-pc ‘h‘d limits Purchase 1, loins in Ci Poor not 0 househoma’ reSource.pc ““88 to 1 181 1.3.1 Short and medium term strategies to help resource-poor households In the short and medium term, it will be difficult to raise the agricultural productivity of resource-poor households. Traditional short-run agricultural policies will have a limited effect on raising the agricultural productivity and incomes of resource-poor households because these policies (ie., agricultural product prices, credit, and extension) do not address the two major constraints facing resource- poor households-~farming in a risky environment and limited access to land. First, for all households, the agricultural potential of our survey areas was limited by environment-related production risk, specifically low, erratic rainfall3. Therefore, the research system will require several years to develop more "appropriate" technologies which adapt crops (varietal improvement) or cropping practices (eg. , tied-ridges) to better use the available rainfall; or modify the environment (eg. , irrigation) to achieve higher and more stable yields from current crop varieties. Second, traditional policies designed to intensify crop production are likely to have little impact on the resource-poor. Not only did resource-poor households have limited access to land and oxen, they also ' had limited financial capital, which restricted their ability to purchase inputs (ie., fertilizer), and bear the risk of repaying input loans in case of crop failure. Recall that marginal-farm and resource- poor not only owned half the land (per capita) compared to small-farm households, but had less than half the financial capital. While resource-poor households had relatively more labor, because they lacked access to land, it is likely that their labor was not fully employed. However, in the short-run government can help resource poor- households though transfer programs designed to immediately insure food 3In both survey areas, rainfall was low (long-run averages 0f 47 7mm and 706mm for Buhera and Mutoko/Mudzi Districts, respectively) and the coefficients of variance were high (34% and 26% for Buhera and Mutoko/Mudzi Districts, respectively). SECU pros impr' publ 1990 Targi dist: cppoz exper incon offer 182 security, while longer-run interventions are developed. Three such programs that are consistent with the government's stated objectives of improving rural services and raising incomes are: food-based policies, public employment schemes, and human capital development (van der Wells, 1990) . Targeted food policies The government should examine the effect of food pricing and distribution policies on poor households in order to identify opportunities for expansion. Since food was the single largest expenditure item--and poor households spent a larger proportion of their income on food than other household types--expanding these programs offers great potential for supplementing incomes of the poor. For example, a current government program links regular clinic visits of children under five years, with a supplemental feeding program (Tagwireyi, 1989). This program is important because it links access to food and health services and targets these benefits to an especially vulnerable group (young children). A non-targeted intervention to reduce the food cost of not only the poor, but for all households is proposed by Chigume and Jayne (1991). Chigume and Jayne observe that since there are already maize grinders in rural areas, permitting local sales of mealie meal would lower the cost of meal. Currently, the G143 assembles grain in the rural areas, transports it to Harare for milling, and then sends it back to rural areas for sale as meal. This redundant movement of grain increases the price of purchased grain, which places a disproportionate burden on poor households who spend a larger share of their income on food purchases. and l rural 183 Public employment schemes Government should continue, and possibly expand, the cash-for-work program in low rainfall areas--especially after poor rainfall years". First, it provides an important welfare function by creating a market for the surplus labor of resource-poor households (since they are relatively labor abundant), especially after a poor rainfall season when own-production is lowest. Second, although this program currently focuses on constructing rural roads and buildings, it could be expanded to include environment projects, as is done in Botswana (Asefa et al, 1989). For example, cash-for-work could be targeted to construct soil conservation gullies and terrace fields, in addition to repairing roads and building schools--all of which would contribute to agricultural and rural development in the long run. Human capital development Subsidizing secondary education is a potential short-run intervention that could help raise (long run) rural incomes, and provide equal opportunity to education for resource-poor households. As discussed earlier, although primary and secondary enrollment has increased substantially since independence, not all household types have benefitted. Only 38 percent of the secondary school age children of resource-poor households' attended secondary school, compared to 68 percent and 90 percent for marginal-farm and small-farm households, respectively. Government could improve resource-poor households' access to secondary education through a subsidy targeted to this group. The subsidy would transfer the risk inherent in the traditionally low pass rate of rural students to the government. Survey results suggest that an additional 16 percent of resource-poor households would enroll their l'The data collection period coincided with a good season for most households, so these programs were not wide—spread in our survey areas. childr elimin 1.3.2 A r take i. produc1 the ‘pc househc opportt activit 7030201 This diffuse facing 1 7~3 - 2 . 1. It it GBpecial Pflfiucti ainsult efficien ‘\-“‘~‘ ‘ine fseholds flseh‘flds I a 184 children if school fees were halved, and 34 percent if school fees were eliminated. 7.3.2 Long term rural development strategies A rural development strategy to assist resource-poor households must take into account the resource availability, relative factor abundance, productivity potentials, and local alternative economic opportunities of the poor. This strategy should seek to both assist resource poor households to increase their agricultural income, but also expand opportunities for marginal producers to diversify into non-agricultural activities, or migrates. 7.3.2.1 Agricultural diversification This section presents suggestions for technological generation and diffussion needed to address the environment-related production risk facing households in low rainfall areas. 7.3.2.1.1 Technology generation It is clear that a major constraint to raising agricultural incomes-- especially those of rural poor households-~is environment-related production rick. A shortage of water is the main constraint limiting agricultural productivity, so technologies that improve water-use efficiency are needed (Waddington and Kunjeku, 1987). To reduce the risk associated with the high inter-seasonal variation and low levels of 5The discussion focuses on assisting resource-poor households, because it is assumed that marginal- and small-farm households are better able to adjust to policy changes than resource-poor households, and they will also be able to benefit from benefit from policies designed to assist resource-poor households. Harginal- and small-farm households have more financial capital and larger agricultural inventories which allow them to bear more risk. Conversely, it is less likely that policies designed to assist small-farm households will help the resource poor. rain CORS' In degrac growt} and 01 to 19 introd includ Compul with inten). accoun. 0f ll‘ “plexus 185 rainfall, government should expand efforts in soil and water conservation, varietal improvement, and irrigation develoment. 1, 591]. and 333;; cgnggrvation: In both their ”Growth with Equity” (1982) and ”First Five-Year National Development Plan" (1986) policy papers, the Government of Zimbabwe recognizes that natural resources in many parts of the country have been poorly managed. Although the government currently promotes several environmental programs--including rural reforestation, land resettlement, transferring more administrative decision-making power to local authorities, more emphasis on agricultural and conservation in schools, and increased expenditures on infrastructure and extension--Whitlow (1988) argues these programs have fallen short in the face of the enormity of the task‘. 7 are severely In communal areas, approximately 3.8 million acres degraded (Whitlow, 1988). Whitlow cites land tenure and high population growth as the main causes, resulting in deforestation, over-grazing, and over-population. Conservation efforts in communal areas date back to 1936 when contour ridges and stormwater drainage techniques were introduced to conserve soil and water. Subsequent interventions included retiring degredated land, reduction of livestock herds, compulsory conservation methods, and replacing traditional land tenure with one based on individual rights. Yet, these conservation interventions had little impact because they: 1.) failed to take into account the socio-cultural (status) and economic (store of wealth) role of livestock; 2.) lacked the necessary manpower and finances to implement them: 3.) were compulsory, and therefore resisted; and 4.) . ‘It was estimated in the late 19305 that, given the available resources, it would take 250 years to complete the needed anti-erosion work in communal areas (Whitlow, 1988) . “Ibis is a conservative estimate since it is based on aerial Photographs, which only detect advanced levels of soil erosion. 186 were promoted during periods of political, economic, and security problems (pro-independence) . Therefore, available evidence suggests that techniques to control soil erosion and conserve soil moisture are a high priority, and should be more vigorously examined, tested, and promoted to ensure the long-run sustainability of communal agriculture. Two conservation techniques, improved tillage methods and mulching, can relax some of the soil and climatic constraints. First, potential conservation tillage methods include winter plowing, deep tillage and soil inversion, and tied-ridging (Sanders, 1989). Winter plowing allows farmers to plow at the end of the year so fields are ready for sowing when ‘the first rains come (Shumba, 1990). Deep 'tillage and soil inversion methods increase water penetration to the plant's roots. In West Africa, farmer adoption of these methods has increased yields of millets, sorghums, maize, upland rice, and cotton (Charreau and Nicou, 1971). Tied—ridges, a practice that holds surplus water to allow more infiltration into the soil, also has the potential to increase yields, reducing risk, and slowing soil erosion in communal areas. Research conducted by DR&SS at the Chiredzi Research Station (Natural Region V) since 1982 suggests that tied-ridges increased yields for sorghum (25 percent), maize (15 percent), and cotton (34 percent)8 (DR&SS and AGRITBX, 1987). Second, mulching with crop residues, leaves, and stems can improve the soil's physical properties, add nutrients and organic matter, and reduce soil temperatures (Lal, 1987). Mulching also conserves soil moisture by decreasing the amount of runoff and evaporation. Yet, mulching has less potential impact in communal areas (especially Natural Region V) than conservation tillage because crop residues are an important source of livestock feed (Ndlovu, 1989). 8Four year averages for 1983/84 to 1986/87. 187 mm: Since agriculturalists believed that Natural Regions IV and V are not suitable for intensive rainfed cropping, prior to independence DRESS conducted little research to assess which crops and cropping techniques were most appropriate under these rainfall and soil conditions (Mudimu, 1986; Nyati and Nyamudeza, 1984). This research orientation ignored the fact that 74 percent of the communal land area is located in these natural regions, and that these households depend on rainfed agriculture. Analysis of the survey data suggests two reasons why DR&SS should give high priority to increase drought tolerance in staple grain crops. First, resource-poor households met only 80 percent of their food and clothing needs through own production, compared to marginal- and small- farm households who met more than 100 percent of their needs. Second, all households tended to allocate grains to a smaller proportion of their land area, once their food needs were secure. Thus, increasing staple crop yields enabled resource-poor households to meet their food and fiber needs and enabled them to expand the proportion of land allocated to cash crops-~thereby increasing their cash earnings potential. Because breeding for drought tolerance is a lengthy and costly undertaking, the government must consider: 1.) consumer preferences, 2.) yield potential, and 3.) available researcher expertise. First, DRESS should give high priority to continuing its efforts to develop maize varieties that are better suited to low rainfall areas because households throughout communal areas have a strong preference for maize. For example, about 97 percent and 83 percent of the households in Hutcko/Mudzi (NR 4) and Buhera (NR 5) Districts, respectively, grew maize. Even in low rainfall villages where maize fails in more years than it succeeds, farmers continue to plant it annually. Furthermore, recent research results suggest that maize has considerable potential in low rainfall areas. For example, in mass 188 cereal comparison trials (1983-1988) at Makoholi Experimentation Station (NR IV), hybrid maize outyielded sorghum (Shumba, 1990). Since these trials were conducted at an experimentation station, additional research is needed to fully exploit the yield potential of maize under local farmer conditions. Finally, emphasizing' maize improvement is also consistent with Zimbabwe's long and successful history of maize breeding. Over the past 50 years, Zimbabwe has developed many breeding lines and has developed a cadre of experienced breeders (Rattray, 1989). Second, sorghum research is more developed than millet research in Zimbabwe, with DR&Ss currently developing and testing higher—yielding sorghum hybrids (Mushonga, 1986). To support these efforts, the ICRISAT/SADCC initiated a regional sorghum improvement program in 1983/84 at the Hatopos Research Station (near Bulawayo). This program's objectives are to both collect traditional varieties to conserve genetic diversity, and to assist national programs strengthen their varietal improvement efforts (House, 1987). Although sorghum's yield potential is high and this crop has lower water requirements than maize, at present there is little urban consumer demand for sorghum. Recent research suggests that the demand for sorghum could be increased through changes in food technology. First, Gomez et al. (1987) reports that with available composite flour technology (ie., the blending of sorghum and wheat to make bread). It is possible to partially replace wheat with sorghum, which would significantly reduce wheat imports, and thereby save foreign exchange. Second, making sorghum processing technology available in rural areas has the potential to increase rural consumption by reducing the processing constraint (ENDA-Zimbabwe, 1987). Therefore government should continue to support research to increase sorghum yields and select for drought tolerance to help poorer households assure their food supply (especially in low rainfall years), but also explore new commercial uses for sorghum and assist in improving 189 rural household access to village-level processing technology. 1L_1;;1gg§;gn: The potential of small-scale, village-based irrigation to reduce environment-related production risk is recognized at the regional, national, provincial, and district level (Rukuni, 1989). Not only does SADCC's Food Security Programme support a project (#12) to develop irrigation and improve management techniques throughout the region, but government is also committed to promote irrigated cultivation in communal areas (R08, 1986). Furthermore, in our survey areas, the Provincial and District develOpment plans cited small-scale irrigation schemes as an important development strategy (Development Plans for Hashonaland East and Manicaland Provinces, and the Mutoko, Mudzi, and Buhera Districts, 1988). In many parts of Sub-Saharan Africa, small-scale irrigation projects were generally' more jprofitable ‘than larger-scale projects, although there is a need for more research on the technical, economic, social, and environmental impacts of these smaller schemes (World Bank, 1987). Rukuni (1989) suggests that small-scale irrigation. has considerable potential to reduce food insecurity where it is introduced as a supplemental enterprise in a rainfed-based cropping system. Despite this potential, small-scale irrigation development faces several constraints. First, schemes--such as the ones in Buhera District, administered by local farmer committeea--have suffered from mismanagement, conflicts between extension workers and farmers, and poor administration (Buhera Development Plan, 1988). Yet, in spite of these difficulties, the average income of participants was higher than the district average for communal households. Second, irrigation development has proven to be very expensive. Consequently, government has been unable to invest heavily in expanding these schemes (Rukuni, 1989). Thus, given the high development costs and the need to improve the 190 food security of a large number of small farmers, future research and development efforts should focus on development schemes that integrate rainfed and irrigated agriculture, as recommended by Rukuni. Food crops--such as maize, small grains, and some oilseeds--could be grown as rainfed crops; and high value cash crops--such as cotton, tobacco, oilseeds, and vegetables--could be grown as irrigated crops. Such a strategy would maximize the private and social profitability of the development costs of irrigation. 7.3.2.1.2 Technology diffusion Successful farmer adoption of these risk-reducing technologies require government to both improve household access to limiting resources and put in place complementary agricultural policies. For example, without improved, access to ‘traction animals, resource-poor households can not adopt improved tillage methods. Similarly, households will not adopt new crops and associated management practices unless research is undertaken to relax prevailing technical constraints, and provide incentive price, marketing, credit, and extension policies, as discussed below. Agricultural research Continued applied agricultural research is needed to field test risk- reducing technologies to see which are technically feasible and economically viable with resources owned by resource-poor households, under the varying agroclimatic (specifically, soil texture and water availability) and socio-economic conditions that characterize Natural Regions IV and V. For example, small-scale irrigation is only feasible in areas with sufficient water reserves, and clay soils that can hold water. Second, DRESS should evaluate the level of resources (including human capital) that households require to adopt these technologies. For example, farmers will require training to successfully manage an 191 irrigation scheme; and fall plowing requires traction animals. Third, DRESS in conjunction with AGRITEX should develop farmer recommendations that will facilitate household adoption of these technologies, including where these technologies are agroclimatically feasible, what management techniques are required, and what cropping system are suggested. Fourth, DRESS should coordinate with AGRITEX to develop a supporting extension program to help farmers adopt these technologies, including written materials (for farmers and extension workers), field days, and demonstrat ions . Pricing policy In the absence of other policy changes, higher prices alone will likely raise only the incomes of marginal- and small-farm households. If the government wants to stimulate agricultural production with pricing policies, they must set prices at levels that correspond with those prevailing in the market place to not burden the budget. Furthermore, only crops for which there is market (either domestic or foreign) should be promoted. larket access As new technologies are developed farmers will require additional complementary input and expanded output marketing services. For example, a more stable production environment will increase the demand for both inputs that are currently used at low levels of intensity such as fertilizer, and new inputs required to grow the promoted crops. Currently, inputs are distributed in comunal areas through private sector retailers who purchase inputs in anticipation of future sales. Due to capital constraints, initially these small retailers may be unwilling or unable to carry larger and more varied inventories required to support the nacent increased demand. To address this initial constraint, the RFC could assess the feasibility of providing input 192 inventory loans to retailers to induce them to stock these inputs locally. The major output marketing problem currently facing communal farmers is an inadequate number of trucks and the unreliability of transporters (Chigume and Shaffer, 1989). Farmers reported difficulties in securing transport, even after contracting transporters to collect their crops. During informal interviews, transporters reported they were unwilling to service more remote villages for two reasons. First, transport was insufficient to meet demand because they lacked spare parts to repair disabled vehicles, due to a foreign exchange constraints. Second, traveling to villages served by poor quality (dirt) feeder roads caused excessive wear and tear on trucks, which shorten their useful life. Government should consider two strategies for alleviating the transport constraint. First, to expand the supply of private transport, government should increase the foreign exchange allocation for spare parts for heavy duty trucks. This would permit existing transporters to increase their utilization capacity, thereby allowing them to better serve communal areas. In addition, in good rainfall years when marketed surplus is large, government could make available District Development Fund (DDF) trucks to collect crops, and charge farmers by deducting a transport charge from the farmer's GMB payment check. Making available DDF trucks to supplement private sector transport services would enable households to market their crops immediately after harvest and receive early payment. Furthermore, in locations targeted for the promotion of perishable vegetable crops, it will be necessary to develop special marketing arrangements (ie., contracts) to insure timely delivery to market. Credit Survey results showed that households, especially resource-poor households, had limited working capital. Unless additional working 193 capital is made available, only the small-farm households will be able to adopt the new technologies, and expand input use. To meet the expanding demand for working capital, AFC will need to expand credit services for all farmers. In addition, to meet the special needs of the resource-poor households, the AFC should examine the feasibility of establishing a targeted lending scheme to the households to acquire a maximum of two oxen per household (to take advantage of the new ‘technologies), with credit terms designed to minimize leakage to marginal- and small- farm households. Extension Close collaboration between the farming systems research activities of DRESS and RGRITEX's extension program is needed to insure that new technologies effectively address farmer's constraints. .As new technologies become available, agricultural extension should be redirected to support their diffusion. First, extension staff should be trained how to adapt these new technologies to farmers' environment and resources. Second, demonstrations should be held to make farmers aware of the newly developed technology. Third, extension workers should work with farmers to teach them how to modify the new technology to fit their circumstances. Finally, extension workers should be encouraged to work more with resource-poor households, and incorporate them into the Master Farmer program. 7.3.2.2 Rural develop-ant programs Although the strategies outlined above are intended to increase the agricultural productivity of resource-poor households, a more comprehensive rural development strategy is required to substantially raise their incomes. To assist these resource-poor households, government will have to initiate a rural development program which incorporates land reform/resettlement, greater access to social 194 services, and expanded rural employment opportunuities. Land reform/resettlement Although this study did not explicitly study land reform or resettlement, two findings suggest that access to land is highly associated with income levels. First, small-farm households owned twice as much land (1.4 hectares) as resource-poor households (0.6 hectares) (Table 7.1). Second, the multivariate analysis (Chapter 6) of both net household receipts and agricultural production showed that the land availability coefficient was both large and statistically significant. These results suggest that government could substantially alleviate the plight of the resource-poor households by expanding opportunities for them to resettle in new project areas. In addition, rural outmigration and land redistribution (and associated land degradation) would alleviate land pressure for the remaining households. Access to social services As part of its long run strategy for reducing poverty, the World Bank places high priority on targeting investment in human capital towards the poor (van de Walle, 1990). Earlier analysis demonstrated that resource-poor households (in particular) continue to have limited access to education and health. For example, many resource-poor households could not afford to send their children to secondary school. Furthermore, only 58 percent of the survey villages had primary schools, 17 percent had secondary schools, and 8 percent had a clinic. Consequently, government should continue to expand social services for rural households, and explore the feasibility of expanding access to education for resource-poor households through targeted subsidies. 195 Employment creation As population pressure grows, it is unrealistic to expect on-farm agricultural activities to absorb the increase in the labor force, and have agriculture continue to provide the primary source of income for rural households. Although in our sample labor sales averaged only 17 percent of per capita net household receipts (Table 4.6), they represented a particularly critical source of income for poor households. Not only did labor sales finance food purchases, but they also provided an opportunity for resource-poor households to monetarize their relatively abundant factor of production, family labor. Currently, households reported that a majority of their labor sales were as agricultural labor on other households' fields. Although the introduction of more labor intensive crops will increase the demand for labor--which will particularly benefit poor households, government should expand its efforts to expand nonagricultural employment in rural areas. Eilby and Liedholm (1988) argue that small scale enterprises have the potential to absorb surplus labor if investment are made to develops infrastructure. 7.4 Summary Zimbabwe's agricultural policies and services were designed to raise the incomes of rural households. An implicit assumption underlying these policies was that all households have sufficient land and capital to respond to these policies; and thereby increase agricultural production and generate an agricultural surplus. Empirical evidence shows that many households actually had limited access to land and omen--thereby limiting the resource-poor households from taking advantage of these new opportunities. Thus, government needs to formulate a more comprehensive rural development strategy. Potential short and long run rural development strategies were examined to assess their likely effect on the 196 agricultural production and incomes of the poor. In the short run (current technologies), it is apparent that government has limited ability to increase the agricultural productivity of the resource poor households. On the other hand, government can implement policies that will improve the food security of the poor in the short run, and invest in longer run. by expanding food-based policies to increase poorer households' access to food, using public employment schemes to promote conservation interventions and infrastructural development; and investing in human capital by increasing access to secondary schools for children of resource poor households through a targeted subsidy. In the long run, technology development should focus on reducing environment-related risk since this is a major constraint to increasing agricultural productivity in low rainfall areas. Three strategies that would address this constraint include interventions to promote soil and water conservation, crop improvement, and small-scale irrigation. Soil and water conservation techniques--primarily improved tillage, terracing, and crop management methods--have successfully relaxed soil and‘ climatic constraints in several semi-arid areas of Sub-Saharan Africa. Crop improvement, targeted at increasing drought-tolerance, also has potential to reduce the risk of crop failure due to water stress. Government efforts to stabilize yields through incorporating greater tolerance to environmental stress (primarily drought) should concentrate on maize and sorghum. Finally, although small-scale irrigation schemes represent a potential strategy, further research is needed to evaluate its role in the communal areas because historically, small-scale irrigation schemes have experienced administrative problems and historically high development costs. Small-scale irrigation should focus on the promising opportunity to integrate rainfed (food crops) and irrigated (high value cash crops) enterprises into a crop-livestock based farming system. Furthermore, a rural development program designed to assist resource- 197 poor households must incorporate a broader set of interventions than traditionally included in an agricultural development strategy. Resource-poor households typically have insufficient land to take advantage of strategies to increase income through incentive prices and improved marketing infrastructure. Since agriculture alone can not absorb the increase in the labor force, the government should explore assiting resource-poor households by expanding the current land resettlement program, increasing rural access to social services, and stimulating rural employment creation. cum VIII MY m LIMITATIONS 0" m RESEARCH .At Independence, Zimbabwe's stated development objectives included: (1) transformation and expansion of the economy, (2) land reform and increasing the efficiency of land usage, (3) higher living standards for the entire population, especially the rural population; (4) employment creation and manpower development; (5) development of science and technology and (6) incorporation of environmental concerns into development programs. Of these six broad objectives, four impact directly on the well-being of the rural population: land reform, expanding employment opportunities and manpower development, raising rural living standards, and incorporating environmental concerns into development programs. But, government has experienced difficulty in achieving its objectives for two reasons. First, macroeconomic constraints have limited government's ability to implement interventions to improve rural living standards. Since Independence, Zimbabwe's development has been constrained by shortages of foreign exchange (which have reduced the country's ability to import capital goods, and resulted in budgetary shortfalls) foreign and domestic trade restrictions, and a large external debt (Zvinavashe, 1990). These problems are the consequence of both internal policies (interest rate, exchange rate, and trade policies) and external shocks (global recession, strong U.S. dollar, and foreign trade policies). Second, many researchers have highlighted the need for a comprehensive understanding of the structure, level, and distribution of 198 1*!“ 199 rural incomes as a precondition for effective policy design (Eicher and Baker, 1982 and World Bank, 1983). To date, few studies have addressed this need. Thus, Zimbabwe's lack of reliable data about the target rural population's characteristics and household objectives has made it difficult for government to design and target policies to increase access to economic opportunities to lower income, rural households. Therefore, the general objective of this study is to provide a better understanding of the structure, level, and determinants of rural incomes in low-rainfall areas of Zimbabwe in order to 1.) assess the impact of current strategies on increasing the rural poors' incomes and access to services, and 2.) prOpose new policy options to better serve the needs of the rural poor. This study will address these general objectives through four specific objectives. 1. Describe the level, distribution, and composition of household incomes and expenditures, including the contribution of the major sources of incomes (home-used production, cash income-generating activities, and transfers) and expenditures (consumption, investment, and transfers). 2. Describe the resource endowment of households in low rainfall areas and how they allocate these resources between alternative uses. 3. Identify the factors associated with the inter-household variability of incomes, especially for poor households. 4. Assess the impact of recent policies on rural incomes, and propose alternative rural development strategies (short, medium, and long term) to increase incomes and expand income opportunities for the rural poor. This chapter sumarizes the research findings, and then presents policy prescriptions and needed research. 200 8.1 Su-ary of findings The most important findings relate to the distribution of incomes, the level and sources of income, household resource endowment, the distribution of resource ownership, determinants of incomes and specific subcomponents, and the effect of current policies and services. 8.1.1 Level and sources of rural incomes All measures of incomes (net household receipts)-—per household, per capita, and per adult equivalent--indicated large income differences across villages and districts (Table 4.1). For example, median per capita incomes ranged from 2597 to 25265 in Buhera; and from 2593 to z5263 in Mutoko/Hudzi. These results are similar to other studies conducted under similar agroclimatic conditions. Ilajor income sources across villages and districts The analysis specified three major income sources: earned income (production for home consumption and cash income generating activities, net of intermediate goods and services), transfers received, and net credit receipts (Table 4.5). Across the twelve villages, the relative importance of these major sources varied greatly. First, earned income accounted for the major share of the total income (per capita)---ranging from 88 to 99 percent in Buhera; and 69 to 93 percent in Hutoko/Mudzi. Second, transfer income (transfers received) was large for the total sample (15 t of total per capita income), but was more important in Mutoko/Mudzi District. Transfers received exceeded 25 percent of total income (per capita) in only one village in Buhera District, but accounted for greater than 25 percent in five of six villages in Mutoko/Mudzi District. Finally, although net credit receipts equalled less than ten percent of total per capita income (except in one village in Buhera District), they were generally negative. sal COD ind in Hut! Furi inve Diff hous hous qua: rang: grea1 T1 accou Peres Perce PErCe; Conan; 10west 201 Across districts, there were three major similarities with respect to earned income components. First, production for home consumption accounted for over one-half of earned income; ranging from 54 percent in Buhera District and 63 percent in Mutoko/Mudzi. Second, farm and labor sales accounted for three-fourths of income from cash income-generating activities (CIGA) in both districts. Finally, non-agricultural product sales, business inventories, and other cash income were minor contributors to earned income. Furthewr analysis of income from cash income-generating activities indicated that farm sales and labor sales are the major source of CICA in both districts, although farm sales were more important in Mutoko/Mudzi (18%) and labor sales were more important in Buhera (21%). Furthermore, non-agricultural. product sales accounted for' a similar percentage (7 percent) of income in both districts, and both business inventories and other cash sources were minor sources of income. Differences in income sources and levels across income quartiles To further evaluate income sources and levels, the sample of households was divided into four income quartiles. Mean income (net household receipts per capita) was less than 2585 for the lowest quartile; ranged from 2585 to 25139 for the lower-middle quartile; ranged from 25139 to 25243 for the upper-middle quartile; and was greater than 23243 for the upper quartile. This analysis illustrates several points. :First, earned income accounted for the largest share of income for all quartiles (87 to 92 percent); while production for home consumption ranged from 50 to 62 percent and cash income-generating activities ranged from 32 to 37 percent (Table 4.7). Second, as expected, production for home consumption accounted for a larger share of household income for the lowest quartile (62 percent) than for higher quartiles (50 percent for the highest quartile). Third, across quartiles there was little 8.1 capi villi I: liboz Buher Dist: 7 in 1 All adult large Distric Distric Mfiion ‘ to Obac “mam. 202 difference in the share contribution of transfers, ranging from 15 to 17 percent. Finally, although not credit receipts were, on average, negative and small for all quartiles, the lowest quartile reported the largest net outflows (8 percent). 8.1.2 Distribution of rural incomes The three measures of equality--coefficient of variation, the standard deviation of the natural logarithm of income, and the Gini coefficient-~showed considerable differences in income equality (Table 4.4). All three measures of per capita NHR distribution indicated greater income inequality in Buhera District than in Mutoko/Mudzi Districts. 8.1.3 Household resource endowment The three most important household resources were land, labor, and capital. Household access to these resources varied greatly across villages, districts, and the entire sample (Table 5.1). In the study sites, household members were the primary source of labor. Generally, household labor was more abundant in Buhera District. Buhera District households had both more resident and nonresident family members, and were less variable in size, than households in Mutoko/Mudzi Districts. For example, median residents per household ranged from 6 to 7 in Buhera villages and 4 to 7 in Mutoko/Mudzi villages. All measures of land availability (per household, per capita, and per adult equivalent; by village, district, and the total sample) indicated large differences in household access to land. As expected, Buhera District households had greater access to land than did Mutoko/Mudzi Districts households because population density is lower in Natural Region V than in Natural Region IV. Yet, district level averages tend to obscure the large inter-village differences in median land availability, which ranged from 0.3 to 0.9 hectares per capita. 8.] EXP rec the 38v: inc: COef avg; {257 203 As expected, Buhera District households had a higher animal traction index. Since Buhera is less favorable for crop production and has a lower population density, more pasture land was available than in Mutoko/Mudzi Districts. As was the case for land, within both districts there were large inter-village differences in access to traction, which ranged from 0.70 in Mutoko/Mudzi Districts to 0.83 in Buhera District. 8.1.4 Distribution of resource ownership For the total sample, the Gini coefficient indicated a low level of inequality for labor (0.29), a moderate amount of inequality for land (0.40), and a high degree of inequality for oxen (0.66). Across Districts, the Gini coefficients were larger for Buhera District than for Mutoko/Mudzi, indicating that all three resources were less equally distributed in Buhera District. 8.1.5 Determinants of incomes This section presents the dependent and independent variables used to explain the inter-household variation in per capita net household receipts (income) and its most important subcomponents—-specifically, the value of agricultural production, labor sales, and transfers (received). W The included independent variables explained 59 percent (adjusted R2) of the variation in income for the sample households. The estimated regression coefficients in the income model provide several insights. First, agriculture's important contribution to earn income is highlighted by the large (and statistical significant) coefficients for production-related endogenous variables--land availability (25121.98), oxen ownership (2549.64), and input usage (257.47). These results complement earlier analysis which showed that hou: agri land fact incl rain was and inccm resul acces stati acces not 3 0f fe: Fir househ resour househ. ’ 1 11531; The 204 households earned a majority of their income (62 percent) directly from agricultural production. Second, among the household-owned resources, land appears to have had the largest impact on income. Third, several factors exogenous to the household were statistically significant, including household access to output markets, input markets, and rainfall conditions. Household access to agricultural marketing outlets was marginally significant for both medium (2569.34, 20 percent level) and high access (2573.27, 10 percent level). Both variables imply a positive, but weak relationship between output market access and income‘. Household access to input markets also provided interesting results. First, the variable used to assess the impact of household access to stores that sold improved seed (within village) was statistically significant and large (2581.44). In contrast, household access to stores that sold both seed and fertilizer (within village) was not statistically significant, possibly due to the generally low level of fertilizer use. Finally, household head characteristics (age and gender of the household head) were not statistically significant, suggesting ‘that resources available to ‘the household ‘were more important than the household head's individual characteristics. W The included independent variables explained 34 percent (adjusted R2) of the inter-household variability of agricultural production income (Table 6.2). The estimated regression coefficients provide several insights. First, several endogenous independent variables were statistically 1Caution should be used when interpreting this result since other factors--including household resources (land, labor, and capital) and other exogenous factors (access to input markets and rainfall level)--strongly influence whether households produce enough to participate in these markets. 205 significant and made a major contribution to explaining agricultural production income--oxen ownership (2588.54), land availability (2514.21), mean distance to fields (- 251.19), and whether the household head was a master farmer (2582.39, 10 percent level) or trainee (2592.41). These relationships support the results from the earlier regression model of net household receipts, which indicated a strong household reliance on agriculture. As expected, oxen ownership had a larger impact on agricultural production income than on total income (NHR) . Second, the importance of factors exogenous to the household is illustrated by the statistically significant coefficients for two input access proxy variables--villages with stores that sold only seed (2576.32) and stores that sold fertilizer (2554.44, 10 percent level). Both variables imply a positive relationship between agricultural production income and access to input markets, but the relative size of the coefficients imply that access to seed stores was more important. Third, unexpectedly, variables selected to measure the impact of household labor characteristics, and the household head's age and gender were not statistically significant (even at the 20 percent 1evel)-- further reinforcing 'the argument that physical resources (land and capital) were more important in explaining agricultural production than household head characteristics. Finally, the purchased inputs variable (256.13) was significant, but the credit variable (2513.53) was only significant at the 20 percent level. While these results imply that purchased inputs had a positive impact on agricultural production, the mean sample values for these variables were low (251.11 for purchased inputs and 250.13 for credit). LADQE_§§1§§ The included independent variables explained nearly 23 percent (adjusted R2) of the inter-household variability in labor sales (Table 206 6.3). The estimated regression coefficients provide several insights. First, labor sales were significantly lower for households with no male (- 2557.63) or where he was away (- 2561.21). This suggests that both spouses are needed for households to take advantage of available local employment opportunities. Second, the labor composition variable was also important in explaining the variation in labor sales. The significant household's ”dependency ratio" variable implied that households with more children (ie., a smaller number of workers per household) sought wage employment due to their greater need for additional income to supplement their agricultural income. Third, among agricultural-related variables, both the land productivity 'variable (250.13) and land (251.54) were statistically significant. Since these coefficients are relatively small, they imply that both agricultural productivity and land had only a minimal effect on labor sales. I Fourth, among variables measuring human capital resources, only household participation in the master farmer program was statistically significant. Households heads that completed the master farmer program sold less labor (- 2547.43), implying that households with master farmer heads focused more on their agricultural activities. WM). The included independent variables explained nearly 20 percent (adjusted R2) of the inter-household variability in transfers (Table 6.4). The estimated regression coefficients provide several insights. First, both middle-aged (2524.08) and older (2529.62) household heads received significantly more transfer income than younger household heads, suggesting that older household heads had children that were old 207 enough to work away from home and help supplement their family's income. Second, households with more male nonresident members (2513.46) earned significantly more transfer income, suggesting that male nonresident members had greater income-earning opportunities than females. This result is consistent with cultural practices that obligates males to assist their parents, and females to help their husband's household. Finally, two agriculture-related variables were statistically significant. Households with greater agricultural productivity (- 250.04, 10 percent) received more transfers income, implying that households with lower agricultural productivity relied more on nonresident family members “to supplement their income. The positive relationship between transfers received and purchased inputs (257.95) suggests the potential importance of transfers as a source of cash to invest in agriculture. 8.1.6 Effect of current policies and services Overall, agricultural policies have had minimal (or negative) impact on resource poor households, but have benefitted marginal-farm and small-farm households. Resource-poor households do not have access to sufficient physical (especially land) or financial resources to generate an agricultural surplus, or bear the risk of intensifying their production (crop failure). Although these resource-poor households had relatively more labor than other households, they were unable to exploit this advantage due to insufficient land or capital. Consequently, short-run agricultural-focused policies (ie. , agricultural product prices, credit, and extension) will have a limited impact on raising the incomes of resource-poor households because they fail to address the two major constraints to increasing their agricultural productivity: access to land and the quality of their environment (soil and rainfall). 208 Yet, in the short run, government can help resource—poor households through transfer programs directed at raising their incomes. For example, targeted food programs would directly increase poorer households' access to food; public employment (cash for work) would directly increase their labor income; and education subsidies would both reduce the cost of educating their secondary school age children, and contribute to human capital development. 8.2. Policy adjustments and needed research To improve the incomes of households in low-rainfall areas of Zimbabwe, government must initiate a broad-based rural development strategy. Although improved technologies are needed, the strategy must include a broader set of initiatives if it is to benefit resource-poor households. Because the major constraint to increasing agricultural productivity in low rainfall areas is environment-related production risk, technology development should focus on reducing this risk. To reduce the risk associated with the high inter-seasonal and low levels of rainfall, government could promote three technologies: soil and water conservation, crop improvement, and small-scale irrigation. In various semi-arid regions of Africa, research has demonstrated that soil and water conservation techniques--primarily improved tillage and crop management methods--can effectively reduce soil degradation and climatic constraints. Similarly, additional crop improvement research (especially breeding designed to stabilize yields by incorporating greater resistance to environmental stress factors--(primarily drought tolerance)--into maize and sorghum is needed to develop improved varieties and management practices appropriate for low rainfall areas. Finally, research evidence from Zimbabwe suggests there is considerable potential for small-scale irrigation--particularly systems that integrate rainfed (food crops) and irrigated (high-value cash crops) 209 agriculture. Yet, further research is needed to exploit this potential because throughout Africa, irrigation schemes have been plagued by administrative problems and high development costs. In addition to increasing agricultural productivity, the suggested rural development program should incorporate a broader set of initiatives to specifically assist resource-poor households, since they do not have sufficient resources to benefit from agricultural development programs alone, and agriculture alone can not absorb the projected increase in the rural population. In order to address this need, government should allocate greater resources to expanding income- generating opportunities for resource-poor households by, for example, accelerating the on-going land resettlement program, providing greater access to social services, and investing in rural employment creation. 8.3 Limitation of the results The results of this study are limited by the survey methods employed. First, the sample included only households in the NR IV portion of Mutoko/Mudzi Districts and the NR V portion of Buhera District. These households are not representative of all households in NR's IV and V because there exist considerable differences across these agro- ecological regions in terms of rainfall, production systems, access to markets, and ethnic background. Second, income and expenditure data are subject to main potential sources of non-sampling error: households' difficulty to recall information, which may have led to their under-reporting of sales, inventory levels, and illegal activities. While it is possible to increase recall accuracy by shortening the recall period, this may increase non-sampling error associated with respondent fatigue. Finally, this study is based on observations from a single year. Thus, it was not possible to observe how households actually adjust their 1 patter! 210 their income and expenditure strategies in response to varying rainfall patterns and changes in personal circumstances. 23L]. APPEIDIX 1: SCHEDULE OF RESEARCH ACTIVITIES 1937 1988 mm < ------------------- >< --------------------------------------------------- > bgnd Plant Heed < ----- flarvestino ----- > and Production Activities 8 <---SE---><---->< ------- >< ------------------------ > < -------- > 8 Marketing Activities