.xi lryr... {I (5,791. 5; n+3 9:6 2: 54‘ 1i\~( ' E .'..l.l:.:¢ ,.}>.l).:l( (5.1;)...511 ({t..n}(¢oll €55... o drunk.» nan. . kufiflnmhniurfiunfi at!!! r . I '1'!!! ‘ 15:... r cliff: . 1...}; e 2:! l. I i .rrltlltuifv (FL I: .II. .) x! i I1!!!.¥$e-|rl '7 7:}... frir:b..5§$.1.i .x. I]! r :3) . 3 4.. Iflt. II... .. - if"! I II I .13.!!! I; r I153...) (fr, rn’llrvtl It!!! (5.}. . .;.. (1:51.. 9:} {2, l I: I .a‘. 1.. rr r»): .. )1)... r :4. (liféifiarrft'f. . a. ll.s;. r...(.lrv:| rlitf {9158's l”illlllflllljlillllfllllllll 319300 lll LIBRARY Michigan State University This is to certify that the thesis entitled HOUSEHOLD EXPENDITURE BEHAVIOR AND CONSUMPTION GROWTH LINKAGES IN RURAL NAMPULA PROVINCE, MOZAMBIQUE presented by Maria Nita Cau Dengo has been accepted towards fulfillment of the requirements for M.S. deg?9in Agricultural Economics Major professor Date December 23, 1992 0-7639 MS U is an Affirmative Action/Equal Opportunity Institution PLACE IN RETURN BOX to remove this checkout from your record. TO AVOID FINES return on or before date due. DATE DUE DATE DUE DATE DUE AUG 1 2 2001 :__T_ MSU Is An Affirmative Action/Equal Opportunity Institution c:\clrc\datedue.pm3-p. t HOUSEHOLD EXPENDITURE BEHAVIOR AND CONSUMPTION GROWTH LINKAGES IN RURAL NAMPULA PROVINCE, MOZAMBIQUE By Maria Nita Cau Dengo A THESIS Submitted to Michigan State University in partial fulfillment of the requirements for the degree of MASTER OF SCIENCE Department of Agricultural Economics 1992 ABSTRACT HOUSEHOLD EXPENDITURE BEHAVIOR AND CONSUMPTION GROWTH LINKAGES IN RURAL NAMPULA PROVINCE, MOZAMBIQUE By Maria Nita Cau Dengo This study reports the consumption expenditure behavior in rural Nampula Province, Mozambique, with special emphasis on the effects of cotton and cashew production and marketing activities on household consumption behavior and the potential derived consumption growth linkages. Both tabular descriptive analysis and econometric estimation of semi-log Engels curves, are used to address these issues. Data is from a survey of 343 farm households in rural Nampula province. Results of this analysis find that food expenditure is dominant, whatever the household income level and whichever the household production activity. Estimated average and marginal food budget shares are very high. Households pursue a strong subsistence strategy, with significant large own food produced and consumed budget shares. Food expenditure elasticities are above 0.80. Based on the model coefficient estimates and the estimated household average and marginal propensities to consume purchased goods, the study infers that support to smallholder cotton production has the potential to increase consumption growth linkages with local farmers and other sectors of the economy. To my dauther, Willie Ebenizério Chonguica iii ACKNOWLEDGEMENTS I would like to acknowledge a number of people and institutions who contributed to the completion of my master program. Thanks are due to Professor Michael T. Weber, my major professor for his interest in my work, for his guidance throughout my course work, and for encouraging and allowing me to use the Mozambique Food Security Project data set. His advice on policy oriented research contributed to shape this study. I am very grateful to Dr. David Tschirley for his friendly and productive guidance and supervision of the analysis and writing of this thesis. I also thank Professor Carl Liedholm from the Economics Department for his willingness to serve on my thesis committee and for his useful suggestions for extensions of this work. My special thanks go to the staff, graduate students and faculty members of the Agricultural Economics Department, who provideda warm environment which contributed in so many ways to my completion of a productive program. I would like to thank the staff in the Agricultural Economics computer center, who were always available for assistance. Special thanks go to Margaret Beaver and Wendy Halvorsen for assistance in the use of the SPSS/PC +. Elizabeth Bartilson and Jeff Wilson were also helpful during the editorial phase. Calea Coscarelli edited some of the tables of this work, I extend my thanks to her. iv My thanks go also to my fellow graduate students Cynthia Donovan and Paul Strasberg for their contribution. The Government of Mozambique and the African-American Institute (AFGRAD) deserve a special thanks for their financial support. Thanks to all members of my family, whose constant moral support and encouragement regarding my education have been invaluable. Finally, I thank my husband Eben. He has been a wonderful source of support and encouragement. I dedicate this work to my daughter Willie, my companion and source of joy. Any errors of omission and/or commission are solely mine. ABSTRACT TABLE OF CONTENTS ACKNOWLEDGEMENTS LIST OF TABLES LIST OF FIGURES CHAPTER F‘P‘F‘P‘H‘ UtJ-‘th-‘H N WNt—‘N no p~ CHAPTER 3. 3. 3. th-‘w INTRODUCTION AND STUDY CONTEXT INTRODUCTION . RESEARCH OBJECTIVES THE STUDY SETTING . LITERATURE REVIEW . THESIS ORGANIZATION . CONSUMPTION EXPENDITURE PATTERNS INTRODUCTION HOUSEHOLD DEMOGRAPHIC AND ECONOMIC CHARACTERISTICS RELATIONSHIPS BETWEEN HOUSEHOLD TYPE AND CONSUMPTION EXPENDITURE PATTERNS SUMMARIZED FINDINGS MODEL DEVELOPMENT: THE ESTIMATION OF ENGELS CURVES BRIEF BACKGROUND ON ENGELS CURVES THE CHOICE OF FUNCTIONAL FORM . MODEL RESULTS . . 3.3.1 INCOME EFFECTS . 3.3.2 EFFECTS OF COTTON AND CASHEW PRODUCTION 3 3 3 EFFECTS OF SELECTED OTHER HOUSEHOLD CHARACTERISTICS 3.3.2 EXPENDITURE BEHAVIOR OF THE AVERAGE HOUSEHOLD . 3.4 SUMMARY OF FINDINGS CHAPTER 4 POTENTIAL CONSUMPTION EXPENDITURE LINKAGES TO ECONOMIC GROWTH 99b COMP-t CHAPTER 5 5.1 INTRODUCTION . CONSUMPTION LINKAGES: THE IMPACT OF CASH CROPS CONSUMPTION GROWTH LINKAGES. RELAXING THE FOOD MARKET CONSTRAINTS . . SUMMARY OF FINDINGS CONCLUSIONS AND POLICY IMPLICATIONS SUMMARY OF RESEARCH FINDINGS vi ii iv viii 37 37 40 49 52 53 S7 58 62 63 63 64 67 71 73 73 5.2 RESEARCH IMPLICATIONS AND SUGGESTIONS FOR FURTHER RESEARCH . . . . . . . . . . . . . . . . . . . . . . . 75 APPENDICES . . . . . . . . . . . . . . . . . . . . . . . . . . . . 78 APPENDIX Al . . . . . . . . . . . . . . . . . . . . . . . . . 79 APPENDIX A2 . . . . . . . . . . . . . . . . . . . . . . . . . 80 APPENDIX A3 . . . . . . . . . . . . . . . . . . . . . . . . . 81 APPENDIX A4 . . . . . . . . . . . . . . . . . . . . . . . . . 82 APPENDIX A5 . . . . . . . . . . . . . . . . . . . . . . . . . 83 APPENDIX A6 . . . . . . . . . . . . . . . . . . . . . . . . . 84 APPENDIX A7 . . . . . . . . . . . . . . . . . . . . . . . . . 85 APPENDIX Bl. . . . . . . . . . . . . . . . . . . . . . . . . 87 APPENDIX BZ. . . . . . . . . . .’. . . . . . . . . . . . . . 93 BIBLIOGRAPHY . . . . . . . . . . . . . . . . . . . . . . . . . . . lOO vii TABLE TABLE TABLE TABLE TABLE TABLE TABLE TABLE TABLE TABLE TABLE. TABLE TABLE TABLE TABLE TABLE TABLE .10 .ll .12 .13 .14 .15 LIST OF TABLES HOUSEHOLD ECONOMIC AND DEMOGRAPHIC CHARACTERISTICS BY EXPENDITURE QUARTILE IN MONAPO DISTRICT . . . . . HOUSEHOLD ECONOMIC AND DEMOGRAPHIC CHARACTERISTICS BY EXPENDITURE QUARTILE IN RIBAUE DISTRICT . . . . . HOUSEHOLD ECONOMIC AND DEMOGRAPHIC CHARACTERISTICS BY EXPENDITURE QUARTILE IN ANGOCHE DISTRICT . MEAN PCCE AND TCE BY QUARTILES IN MONAPO FOR COTTON AND NON- COTTON GROWING HOUSEHOLDS HOUSEHOLD PCCE DISTRIBUTION AMONG COTTON AND NON- COTTON GROWING HOUSEHOLDS IN MONAPO AND RIBAUE HOUSEHOLD TCE DISTRIBUTION AMONG COTTON AND NON- COTTON GROWING HOUSEHOLDS IN MONAPO AND RIBAUE PCCE AND TCE DISTRIBUTION AMONG CASHEW AND NON- CASHEW SELLING HOUSEHOLDS IN MONAPO . MEAN PCCE AND TCE DISTRIBUTION AMONG CASHEW AND NON- CASHEW SELLING HOUSEHOLDS IN ANGOCHE MEAN HOUSEHOLD EXPENDITURE SHARE BY EXPENDITURE CATEGORY IN MONAPO, RIBAUE AND ANGOCHE MEAN SELECTED BUDGET SHARES BY PCCE QUARTILE IN MONAPO RIBAUE AND ANGOCHE . MEAN HOUSEHOLD EXPENDITURE SHARES BY EXPENDITURE CATEGORY AMONG COTTON AND NON- COTTON GROWING HOUSEHOLDS IN MONAPO MEAN HOUSEHOLD EXPENDITURE SHARES AMONG CASHEW SELLING AND NON- SELLING HOUSEHOLDS BY EXPENDITURE CATEGORY IN MONAPO AND ANGOCHE MEAN HOUSEHOLD FOOD EXPENDITURE SHARES BY SOURCE OF FOOD IN MONAPO, RIBAUE AND ANGOCHE . MEAN HOUSEHOLD FOOD EXPENDITURE SHARES BY SOURCE OF FOOD AMONG COTTON AND NON- COTTON GROWING HOUSEHOLDS IN MONAPO, RIBAUE AND ANGOCHE MEAN HOUSEHOLD FOOD EXPENDITURE SHARES BY SOURCE OF FOOD AMONG CASHEW SELLING AND NON-SELLING HOUSEHOLDS IN MONAPO, RIBAUE AND ANGOCHE REGRESSION RESULTS: PARAMETER ESTIMATES FOR SELECTED FOOD GROUPS IN MONAPO AND ANGOCHE REGRESSION RESULTS: PARAMETER ESTIMATES FOR SELECTED FOOD GROUPS IN MONAPO AND ANGOCHE viii 18 l9 19 20 21 22 23 23 26 29 32 33 34 35 35 50 51 TABLE TABLE TABLE TABLE TABLE TABLE TABLE CHANGE IN COMMODITY EXPENDITURE SHARE DUE A 1% CHANGE IN THE INCOME PROPORTION OF CASH CROP SALES DIRECTION CHANGES IN FOOD EXPENDITURE SHARES BY SOURCE AND NON-FOOD EXPENDITURE SHARES DUE TO MARGINAL CHANGES IN THE PROPORTION OF CASH CROP SALES . . . . . . . . . . . . . . . . . CONSUMER EXPENDITURE BEHAVIOR IN MONAPO AND ANGOCHE . CLASSIFICATION OF FOOD COMMODITIES ACCORDING TO THEIR EXPENDITURE ELASTICITIES FOR AVERAGE HOUSEHOLD . . . . . . . . . . . . . . . . . CONSUMPTION EXPENDITURE BEHAVIOR AMONG COTTON AND NON-COTTON GROWING HOUSEHOLDS IN MONAPO . MARGINAL PROPENSITIES TO CONSUME FOR CASHEW AND NON- CASHEW SELLING HOUSEHOLDS IN MONAPO AND ANGOCHE . . . . . . . . . . . . . . . . . . . MARGINAL BUDGET SHARES BY EXPENDITURE QUARTILE IN MONAPO AND ANGOCHE . . . . . . . . . . ix 56 57 59 61 65 67 69 FIGURE FIGURE FIGURE FIGURE FIGURE FIGURE FIGURE FIGURE FIGURE LA) NH NNNN bWNH LIST OF FIGURES LORENZ CURVE, ANGOCHE . LORENZ CURVE, RIBAUE LORENZ CURVE, MONAPO . . TOTAL FOOD BUDGET SHARE BY PCCE QUARTILES IN MONAPO, RIBAUE AND ANGOCHE MEAN SELECTED BUDGET SHARES BY PCCE QUARTILES IN MONAPO MEAN SELECTED BUDGET SHARES BY PCCE QUARTILES IN RIBAUE MEAN SELECTED BUDGET SHARES BY PCCE QUARTILE IN ANGOCHE . DEPENDENT VARIABLES INCLUDED IN THE REGRESSIONS INDEPENDENT VARIABLES INCLUDED IN THE REGRESSIONS 24 24 25 27 3O 3O 31 44 45 CHAPTER 1 INTRODUCTION AND STUDY CONTEXT 1.1 INTRODUCTION Despite Mozambique's well-endowed and diversified resource base, the country's economic performance has been declining sharply since the early nineteen-eighties. The main reasons for this decline relate to the simultaneous incidence of a series of adverse economic and non economic factors. The key economic factors have been the highly centralized economic policies of Frelimo, the ruling party since independence in 1975. Non economic factors include the recurrence of drought and the 15 years of civil war, both of which have further damaged the already weakened economy. These factors have had a severe adverse impact on the agricultural sector. This sector contributes about 40% to 50% of GDP, employs more than 80% of the labor force, and generates around 80% of the country's export earnings. More than 95% of all farmers are smallholders, of whom more than 60% are women (World Bank, 1992). The smallholder sector works approximately 80% of all cultivated land in the country. To reverse the economic decline and restore minimum levels of consumption and income, the Government adopted in 1987 the Economic Rehabilitation Program (ERP) with International Monetary Fund (IMF) and World Bank (WB) support. The ERP is a structural adjustment program similar to others implemented in Sub-Saharan Africa (SSA), with special emphasis on price and market liberalization. It's aim is to reestablish an economic environment with adequate incentives for private economic agents, and thereby to stimulate broad based and sustainable economic growth. 2 Although phase one (1987-1990) of the ERP has been successfully implemented, and significant market liberalization measures have been taken, the magnitude of Mozambique’s economic crisis is such that the attainment of self-sustaining economic growth still remains a long term objective to achieve (World Bank, 1992). The rural smallholder sector, due to its large size, is key to sustainable economic growth. Hence polices that enhance agriculture's contribution to the national economy and that encourage linkages between smallholders and local and regional economies are crucial. These policies are especially important now, given the rural devastation caused by the war and the subsistence strategy that smallholders have adopted in response to widespread food and labor market failures (MOA/MSU/UA Research Team, 1992a). 1.2. RESEARCH OBJECTIVES The principal goal of this study is to describe and statistically analyze the relationships between income and consumer demand in rural Nampula, in order to derive broad conclusions about potential rural growth linkages.' I 'The study has three main objectives: i) To understand the relationships between consumption and income in rural Nampula; ii) To identify types of households, according to their demographic characteristics and engagement in selected activities, whose demand pattern suggests strong growth linkages to the local economy; iii) To identify and draw broad conclusions regarding investment and policy priorities to promote local economic growth. 3 Given the current levels and sources of household income which determine the expenditure pattern, it is hypothesized that: H(l). Households growing cotton in Monapo have relatively high cash budget shares. Hence their marginal propensities to consume commodities locally produced and with potential growth linkages is higher than the non-cotton growing households. H(2). Households selling cashew in Monapo and Angoche have relatively high marginal propensities to consume non food commodities. H(3). Household access to land reduces household cash expenditure in food in all three districts. Empirical evidence reported particularly by the MOA/MSU/UA Research team and briefly presented below constitute the basis for formulating these hypotheses. The data for this study were collected during June, July, and August, 1991, as part of the MOA/MSU/UA Food Security Project. A sample of 343 households was drawn selected from 15 villages from three districts in Nampula: Monapo, Angoche and Ribaue. Because Nampula reveals marked ecological differences from east to west, which in turn are reflected in changing agricultural patterns, the three districts were selected purposively to reflect the ecological and economic differences within the region. Within each district, the set of secure villages was identified, and a two stage sampling procedure was used. In the first stage, villages were randomly selected, and in the second households were randomly selected from these villages (MOA/MSU/UA Research Team, 1992b). Each household was interviewed once during the data collection period. The survey instrument was designed to collect data on several aspects of the smallholder sector: household demographic structure, 4 patterns of use of family labor, access to and use of land, household participation in the market and the importance of cotton (MOA/MSU/UA Research Team, 1992b). The statistical results derived from this sample should not be extrapolated to all rural households, but to those in relatively secure areas, hence the use of terminology "rural Nampula" in this study should be understood in this context. 1.3 THE STUDY SETTING Monapo, Ribaue and Angoche are alike in general but have interesting contrasts. A brief description of the study setting follows, based heavily on MOA/MSU/UA Research Team (1992a). Perhaps the key characteristic of Monapo is the existence of cotton production and processing. Sixty percent of surveyed households cultivated cotton as a major cash crop in 1991. Ribaue produces basically food grains. Although cotton and cashew production contributed significantly to household income in the past, at the time the research was carried out, income from these crops constituted only a very small proportion of current household income. Angoche on the coast ' has fishing, cashew and food grains as important sources of household income. All three districts are very poor, with household average annual net income of about MT 380,000 approximately US$ 190 in Monapo and Angoche. Ribaue has the lowest income among the three. Within each district, income is highly correlated with land holdings. Approximately 85% of household average net income in all three districts is derived from on-farm activities, of which about half is cash income in Monapo and Angoche. In Ribaue, about one quarter of on-farm income is cash. 5 Households pursue a strong subsistence strategy due largely to market failures. The proportion of staple food retained with respect to gross household income is 40% in Monapo, 38% in Angoche and 64% in Ribaue. Monapo has the lowest share of food sales, 6% of household income. Cotton is an important cash crop in Monapo, contributing about 20% of gross household income and cashew sales are the second large source of on-farm cash income in Monapo. In Angoche food crops and cashew are important sources of cash income. Food sales account for almost one fourth of gross household income, most of this from peanuts and rice, followed by cashew sales with 14%. Although Ribaue has the highest proportion of food retained for consumption, still food sales are the most important source of cash farm income, accounting for 12% of gross household income. In Ribaue cashew and cotton sales contribution to household gross income is only 1,52. Household cash income as a proportion of total income is a crude indicator of household market integration. The level of households food market integration in rural Nampula is one of the lowest in sub-Saharan Africa. In rural Nampula households face high transaction costs associated with transport and risk costs aggravated by the war. The structure of household income determines the household cash expenditure pattern. In all three districts the share of own food produced and retained for consumption is high relative to other SSA countries, revealing an overall orientation to a strong subsistence strategy. This implies generally weak linkages with the outside economy for most households. This household strategy may be seen as a response to the very low level of services and the widespread food market 6 failures in terms of high cost and frequently unavailable supplies for purchase. 1.4. LITERATURE REVIEW Development strategies which strengthen the potential growth linkages of rural smallholders are not yet fully developed. However as reported by Haggblade, Hazell and Brown (1989), evidence from Asia suggests significant income and employment growth multipliers derived from production and consumption linkages. The issue is complex, and to fully address it would require the analysis of growth linkages derived from all factor and product markets. Under the current SSA socio- economic and institutional framework, consumption linkages account for about 80% of total agricultural growth multipliers. Research based on consumption expenditure data shows that growth linkages derived from rural consumption expenditure pressure demand for locally produced commodities, which in general are labor intensive and generate second round effects of growth to the local economy with strong employment links. In general it is believed that demand for labor intensive goods decreases as income rises, suggesting that investment directed to low income households whose marginal propensity to consume labor intensive goods is high, may improve income distribution while at the same time contributing most to local economic growth. This line of reasoning suggests that equity and maximized local economic growth may be complementary goals. But such a general statement must be evaluated with location specific analyses. In fact, the goals may be conflicting. Too, growth multipliers are only effective where there is an elastic supply of the commodities whose demand increases. Investment directed to low income households may 7 jeopardize long term development in part because low income households tend to have low saving rates. Hence there may be a trade off between equity and growth which should be taken into account. The importance of rural consumption linkages for economic development is centered on the following questions: which households exhibit expenditure patterns with potential to cause higher second rounds of growth and promote local and regional development? What household characteristics have strong impacts on local development? Household characteristics which are associated with strong linkages are open to policy influence. For example, if cotton growers tend to have stronger growth linkages, then strategic policies to expand cotton growing schemes so more households could participate should be appropriate. Households whose marginal propensity to consume goods with strong potential growth linkages would then be targeted for public investment and strategic policy. Also for households with weak linkages, policies to improve their income and consumption status are especially important. As noted by Hazell and Roell (1983) this partial approach has some drawbacks.which should be taken into account: (a) expenditure patterns per se do not determine the magnitude of income and emplOyment generated through growth linkages. The level of the multipliers depends on the elasticity of supply of the products demanded which in turn depends on the level of development in the area. (b) investing in low income households with high marginal propensities to consume locally produced goods or services may jeopardize long term development, because these households tend to have low savings rates. 8 (c) if expenditure patterns indicate that grain production is strongly associated with growth linkages (which is generally typical in poor economies), then to expect growth linkages requires that grain supply be elastic. This presupposes markets operating with neither barriers to entry nor price rigidities. Household consumption decisions are determined by household income, the price of goods and services and a set of other social and economic factors each household member as an individual faces. The complexity of the consumption and expenditure decisions that households undertake as both producing and consuming units in rural Nampula are explained and schematically analyzed in MOA/MSU/UA Research Team, 1992b (p21, 22). Consumption and expenditure patterns in rural Nampula are yet not well known. Given Mozambique's economic and social setting, characterized by fragile economic activity, deeply influenced by market instability and the majority of the population living in absolute povertyl, we expect household consumption expenditure patterns to be strongly influenced by the household engagement in cotton or cashew activities. Why may grOwing cotton or selling cashew determine different 1 consumption expenditure patterns?‘ The cash income obtained through these activities is obviously positively related to household good purchases. Furthermore in some cases the cotton and cashew structure of incentive schemes directly determines the household access to selected commodities. 1 Household is in "absolute poverty" if its members have an inadequate nutritional standard, even when the household is spending more than 60% of it's total income in food (Green R. , 1989). 9 A brief overview of the differences between cashew and cotton institutional support may help to understand the role of those crops in influencing household consumption expenditure patterns. Historically cotton and cashew have been important agricultural smallholder cash crops. Before independence (1975) cotton was produced with both forced labor and coerced production on smallholder farms. The state determined the areas to be planted, guaranteed the input supply, and supervised and controlled the entire production management process. In contrast to cotton, cashew production and management has been less controlled, i.e. households produced and marketed cashew based on their market price expectations and on the availability of consumer goods in the market. The cashew marketing policy has historically relied heavily on the provision of consumer goods to households selling cashew. Local traders played an important role in cashew marketing. They guaranteed the provision of consumer goods, mainly cloth and footwear, sugar and illumination oil to households selling cashew. After independence cotton production fell sharply as the institutional framework collapsed. Cashew marketing also declined as the rural trade system broke down. The current cotton policy is based on large private companies associated with the government. The company is responsible for cotton production, processing and marketing in selected areas and provides support for households growing cotton in their area of influence. This support includes the supply of production inputs and the provision of extension and market services. The private companies also provide hired labor opportunities. 10 Given the nature of the institutional support to cotton and cashew production and marketing, we hypothesize that these activities will have strong effects on consumption and expenditure patterns, increasing the linkages of these households with the local economy. The importance of consumption linkages as income rises in rural areas of LDCs has been widely studied as one key determinant of rural development. A rise in per capita income in low-income countries is associated with a substantial increase in the demand for food. Typically the income elasticity of demand for food is about 0.8 or even higher (Mellor, 1966). Household marginal propensities to consume are important in the analysis of consumption expenditure linkages, and are an important part of the story of growth and development. Questions to be considered include whose income is rising? For which goods and services does demand most increase with rises in income? And is the policy and institutional framework setting appropriate to promote growth? Knowledge of consumption patterns is recognized as one of the major contributors to economic planning and policy analysis in Africa. Understanding the nature and dynamics of consumer demand may be a quite valuable tool in po1icy and project design. As reported by King and Byerlee (1977), research on the relationships between income distribution and the pattern of consumer demand, and their implications for growth and employment in the total economy, has been growing and consumption based linkages are considered an important factor in the development process. The pattern of consumption expenditure is of particular interest because it varies with income. Hence, income distribution has an important influence on the pattern of expenditure, which in turn may ll affect employment in sectors adjusting to the demand changes, causing second round effects on the demand pattern (Mellor, 1977). Hazell and Roell (1983) find in Malaysia and Nigeria that households on the larger farms have the most desired expenditure patterns for stimulating secondary rounds of growth in the local economy, and conclude that those households may be appropriate targets for public investment to increase agricultural output. In the study area in Malaysia the average household spends 18% of its total budget on locally produced non-food goods and services, and allocates 37% of its marginal expenditure to services and non-food commodities. Similar indicators for the study area in Nigeria are relatively lower. This is expected, since this area has a less developed economy than the one in Malaysia, hence relatively low living standards and consequently households have fewer commodities to share among their members. This suggests that the development level of the region, as well as the institutional setting are important determinants of the size and incidence of consumption linkages. King and Byerlee in their study in Sierra Leone (1977), found that the marginal propensity to consume subsistence goods drops_dramatically as income increases. In the lowest income households, almost 70% of any increase in expenditure is allocated to subsistence food consumption while only 29% of incremental expenditures for the highest income households is allocated to subsistence food. Furthermore, King and Byerlee classify consumption patterns in Sierra Leone, in 1977, as quite labor intensive since 84% of all increases in consumer expenditures were on goods produced in small-scale agricultural, fishing, industrial and service sectors. It has been noted that most of the agricultural commodities for which demand expands rapidly with rising rural income, 12 are those which are labor intensive (Mellor, 1966), hence this provides a large market for farmers and absorbs more labor. Consumption expenditure patterns vary with household characteristics. Households are not homogeneous, and the differences among them affect consumption patterns (Cellis and Bliven, 1991). Deaton and Case (1985) note that households in Sri Lanka and Indonesia with more adults have expenditure patterns consistent with a higher standard of living than their per capita outlay would suggest. The following household characteristic effects were reported by Hazell and Roell (1983): In Nigeria family size has a significant negative effect on food budget shares, except for eggs and dairy products; the older the household head, the greater the share of the budget allocated to non locally produced food and the more educated the household head the more important livestock products, clothing and footwear, transport, education and health, and personal services and entertainment in the budget. Similar but less strong results were found in Malaysia. The poorest households tend to have large, young families, with limited potential to earn.additional income and significant child care demands. Similarly, female headed households are often poor and more vulnerable, hence the expected evidence of a negative relationship between those household characteristics and household per capita consumption. Consequently, these households are expected to fall under the category of low income households which tend to allocate a high proportion of incremental incomes to locally produced food. Cellis and Bliven (1991), found that in rural Zambia, the level of education of the head of household affects cash purchases of food. As the level of education increases, so does the share of cash purchased 13 food. This suggests different farm labor allocation, as off farm income opportunities rise. Furthermore, Cellis and Bliven found that the household revenue source matters in determining the household's consumption patterns. Households with high revenue proportions from off-farm income tend to have large cash expenditure shares. In Zambia, as the household's revenue share from animal husbandry increases, the maize expenditure share decreases and the meat and gathered food budget shares rises. Although the importance of consumption expenditure analysis and the related growth linkages effects is clear, it should be recognized that it is only part of the growth linkages story. A more complete picture of the role of linkages in economic growth requires the analysis of both urban and rural consumption, other input-output and factor market linkages and a deep comprehension of the institutional arrangements within and through which resources are allocated. 1.5 THESIS ORGANIZATION The paper has five chapters including this introduction. The following.chapter defines the most important concepts to be used thrOughout the study, and describes and conducts tabular analyses of consumption expenditure patterns and household characteristics. Budget shares are presented by commodity group and income class. Chapter 3 deals with the estimation of Engels curves, the function which captures the relationship between the demand for a good or service and total income. A brief discussion of the selected functional form is presented and model results are reported. Chapter 4 explores the potential consumption growth linkages derived from the model. The overall results 14 of the study and possible policy implications are summarized in the final chapter. CHAPTER 2 CONSUMPTION EXPENDITURE PATTERNS 2.1 INTRODUCTION In order to have a broad outline of the differences among households, and of the impact of household characteristics on consumption expenditure patterns, descriptive analysis on selected variables was performed. Special emphasis is given to highlighting differences between households with respect to their demographic characteristics, income levels, and engagement in particular activities like cotton or cashew production. The objective is to identify particular characteristics which may influence household consumption expenditure. Before addressing these issues, let's first broadly define the most important concepts used in this study: Consumption expenditure is used as a proxy for net household incomez, and is defined as the value of consumed production plus total household annual cash expenditure. Annual cash expenditure is the aggregate value of household cash expenditures in hungry and harvest seasons. The choice of consumption expenditure as a proxy for income is recommended as it is considered the appropriate indicator of permanent income (Friedman, 1957) and empirical research has often reported income data to be less consistent than household expenditure data (Hazell and Roell, 1983). 2Net household income is defined as the sum of: Value of own production consumed + Value of household production sold + Wages from off-farm employment - Value of production inputs. 15 16 Income classes are defined based on quartiles3 of annual per capita consumption expenditure (PCCE), rather than total household expenditure. Budget shares are the ratios between expenditure on a good/service or group of goods/services and total expenditure. Because they are dimensionless, they can be compared across households, across time, and across regions without the need to take into account prices and exchange rate conversions. The analysis of consumption expenditure behavior is done using budget shares. Budget shares are recommended variables for any demand analysis. Variables representing quantities consumed are more generally used in nutrition and detailed poverty studies. 2.2 HOUSEHOLD DEMOGRAPHIC AND ECONOMIC CHARACTERISTICS Despite significant inter-household heterogeneity the demographic structure of the "typical" sample household in the three districts is quite similar. Most families (66%) in Ribaue have two or more children. In Monapo and Angoche the proportion of households with two or more children are respectively about 48% and 39%. Comparative analysis among different types of households shows that households growing cotton are more likely than other households to have two or more children. In Monapo among cotton growing households, 65% have two or more children, while this proportion is only 23% among households not growing cotton. Relatively large families tend to have large numbers of children. Households without children are about 31% in Monapo, 26% Angoche, and 20% in Ribaue. Tables A1 - A7 in Appendix A 3 By definition, households are equally spread among the four quartiles, in ascending order of consumption expenditure. 17 show the frequency distribution of households according to their size and number of children. There are no more than two elderly“ people per household. In Monapo 10% of surveyed households have one or two elderly people. In Ribaue and Angoche this proportion is respectively 5.7% and 7.2%. The dependency ratios is largely determined by the number of children in the household. The average dependency ratio is 0.99, 1.00 and 0.85 in Monapo, Ribaue and Angoche respectively. Among cotton growing households relatively higher dependency ratio is found: 1.1 in Monapo and 1.02 in Ribaue. The annual average per capita consumption expenditure (PCCE) is about 75,000 Meticais, approximately US$ 506 in Monapo and Angoche, and 25% lower in Ribaue. Thus it is estimated that the income per capita of about three fourths of the population surveyed in all three districts falls below US$ 100. This result reveals a picture of general acute poverty in rural areas, even though incomes may be underestimated, since some households reported zero expenditure in some commodity expenses and the recorded products may not be exhaustive. Zero expenditures may appear for more than one reason: the household may never consume that good or service, or the household did not consume that good or service during the recall period. The infrequency of certain expenses may cause some recall problems too. “ Elderly people are defined as adults over 65 years old. 5 Dependency ratio is defined as the number of children up to 10 years old plus elderly people divided by number of adults aged 18 to 65. This differs from some measures of dependency ratio, indicating the number of dependents per adult in the family. 6 USD 1.00 = Meticais 1450,00 (Average for 1991. World Bank,l992 ) 18 In all three districts both household size and the dependency ratio decline as income per capita rises. One probable explanation of this result may be that relatively large households are associated with large numbers of children, which have relatively lower productivity, and hence the inverse relationship between income per capita and household size in these households. 0n the other hand, the low variability of these indicators reflect general similarities among rural households. Economic and demographic characteristics of the households in each income class are shown in Tables 2.1, 2.2 and 2.3. Table 2.1 HOUSEHOLD ECONOMIC AND DEMOGRAPHIC CHARACTERISTICS BY EXPENDITURE QUARTILE IN MONAPO DISTRICT Expendi- Mean Upper Lower Average Dependency ture Annual PCCE* PCCE* household ratio Quartile PCCE* bound bound size -------- Meticais -------- 1 22,341 31,770 5,439 4.4 1.497 2 47,826 63,419 35,721 4.7 1.014 3 76,130 96,826 63,791 4.2 1.106 4 149,612 289,078 99,250 2.5 0.419 District Mean 76,150 3.9 0.980 *PCCE: Per capita consumption expenditure Source: Nampula Smallholder Survey, 1991 The data suggest some income differences among districts, particularly for Ribaue, were the mean consumption eXpenditure per capita is at least one fourth lower than in Monapo and Angoche. Average PCCE is relatively higher in the districts with cashew production (Angoche and Monapo), and cotton production (Monapo). Table 2.4 shows the PCCE and the TCE distribution among cotton and non cotton growing households. Comparative analysis of the average 19 Table 2.2 HOUSEHOLD ECONOMIC AND DEMOGRAPHIC CHARACTERISTICS BY EXPENDITURE QUARTILE IN RIBAUE DISTRICT Expenditure Mean Upper Lower Average Dependency Quartile annual PCCE*' PCCE* household ratio PCCE* bound bound size ------- Meticais ------- 1 23,808 32,568 6,940 6.2 1.300 2 40,991 49,099 32,983 5.0 1.043 3 58,746 72,650 49,369 5.4 1.043 4 108,263 181,142 75,265 3.4 0.620 District Mean 56,865 5.0 1.010 *PCCE: Per capita consumption expenditure Source: Nampula Smallholder Survey,1991 Table 2.3 HOUSEHOLD ECONOMIC AND DEMOGRAPHIC CHARACTERISTICS BY EXPENDITURE QUARTILE IN ANGOCHE DISTRICT Expenditure Mean Upper Lower Average Dependency Quartile annual PCCE* PCCE* household ratio PCCE* bound bound size ------ Meticais ------- 1 29,451 39,156 17,715 5.6 1.468 2 51,123 62,739 40,994 3.9 0.685 3 76,735 95,250 65,028 3.5 0.694- 4 157,645 369,993 96,603 3.0 0.553 District Mean 78,446 4.0 0.853 *PCCE: Per capita consumption expenditure Source: Nampula Smallholder Survey,199l income among selected groups of households shows that households growing cotton in Monapo district surprisingly do not have higher mean PCCE than the households who do not grow cotton. 20 Table 2.4 MEAN PCCE AND TCE BY QUARTILES IN MONAPO FOR COTTON AND NON-COTTON GROWING HOUSEHOLDS Expenditure PCCE“) TCE‘b) Quartiles Cotton No Cotton No n - 63 Cotton n - 63 Cotton n - 41 n =41 ------------ Meticais ------------ 1 24,928 20,122 124,195 89,470 2 47,760 44,267 242,443 142,449 3 71,246 90,108 340,581 236,319 4 130,039‘ 172,721 384,226 350,592 Overall Mean 67,791 89,009 273,548 217,108 (a) PCCE: Per capita consumption expenditure (b) TCE: Total consumption expenditure Source: Nampula Smallholder Survey, 1991 However, a positive income effect associated with growing cotton, is reflected for these households in the two lowest income quartiles in Monapo. The mean PCCE of the households growing cotton in the first and second quartiles in Monapo district is 24% and 8% respectively higher than the PCCE for non cotton growing households in the same quartiles.- 'Furthermore, total household consumption expenditure (TCE) is much higher for cotton growers than for non cotton growers, especially in the. lowest quartiles.A possible explanation for the relatively lower mean PCCE among cotton growing households in the third and fourth quartiles, is associated with the fact that households growing cotton tend to have larger families, with larger numbers of children. The dependency ratio shows that for every adult member in cotton growing households there 21 are 1.1 dependents, while in non-cotton growing households this relation is 1.0 to 0.78. The dependency ratio difference is larger in the third quartile. Inter distrital PCCE distribution among growing and non-growing cotton households is shown in tables 2.5 and 2.6. TABLE 2.5 HOUSEHOLD PCCE* DISTRIBUTION AMONG COTTON AND NON-COTTON GROWING HOUSEHOLDS IN MONAPO AND RIBAUE Expenditure Cotton Non-Cotton Quartile Monapo Ribaue Monapo Ribaue n-63 n-l9 n - 41 n - 74 --------- Meticais --------- 1 24,928 21,829 20,122 24,291 2 47,760 41,984 44,267 40,933 3 71,246 63,429 90,108 57,858 4 130,039 121,251 172,722 106,696 “era” mean 67,791 58,846 89,009 56,352 *PCCE: Per capita consumption expenditure Source: Nampula Household Survey, 1991 _The above results support the hypothesis that, cotton has a positive income effect in Monapo7. Tables 2.7 and 2.8 show the income distribution by quartiles in Monapo and Angoche among households selling and those not selling 7 Due to very poor cotton performance associated with institutional collapse of the Secretariat of State for Cotton (SEA) in Ribaue, the cotton issue is not examined there. 22 TABLE 2.6 HOUSEHOLD TCE* DISTRIBUTION AMONG COTTON AND NON-COTTON GROWING HOUSEHOLDS IN MONAPO AND RIBAUE Expenditure Ribaue Quartile Cotton Non- Cotton Non- Cotton Cotton ------- Meticais ------- 1 124,195 89,470 154,005 146,941 2 242,444 142,449 259,815 186,362 3 340,581 236,319 358,441 304,226 4 384,226 350,592 379,004 349,869 Mean 273,548 217,108 288,185 244,878 *TCE: Total Consumption Expenditure Source: Nampula Smallholder Survey, 1991 cashew. We hypothesized that cashew production increases growth linkages regardless of its impact on income in Monapo and Angochea. Comparative analysis of income distribution among households selling cashew and those who do not, show that TCE and PCCE are higher for the households who sold cashew in Monapo and Angoche. Mean PCCE is about one fourth and one third higher in Angoche and Monapo respectively for ' cashew selling households. The above statistical results do support consistently the hypothesis that cashew production has a positive effect on household income. Hence we expect these households to exhibit consumption expenditure patterns consistent with relatively high cash budget shares. To complement the analysis of income distribution, Gini coefficients on TCE have been calculated for the three districts and have the following values: 0.361, 0.318 and 0.376 for Angoche, Ribaue and Monapo 8 Ribaue is not considered because of the pest problem which destroyed the cashew harvest. 23 TABLE 2.7 PCCEa AND TCEb DISTRIBUTION AMONG CASHEW AND NON-CASHEW SELLING HOUSEHOLDS IN MONAPO Expenditure PCCE“) TCE“) Quartiles Cashew Non- Cashew Non- Cashew Cashew -------- Meticais -------- 1 23,426 21,081 109,534' 86,941' 2 51,663' 44,878* 232,986 235,130 3 80,033: 72,772: 302,290 291,241 4 153,045 125,503 390,828 319,167 Overall Mean 83,525 67,550 268,840 236,704 (a) PCCE : Per capita Consumption Expenditure (b) TCE : Total Consumption Expenditure * : Significant at a - 0.001 Source : Nampula Smallholder Survey, 1991 TABLE 2.8 MEAN PCCE‘.AND TCEb DISTRIBUTION AMONG CASHEW AND NON- CASHEW SELLING HOUSEHOLDS IN ANGOCHE Expenditure PCCE“ TCE” Quartiles Cashew Non- Cashew Non- cashew Cashew -------- Meticais --------- 1 34,520' 24,706* 189,996* 151,221* 2 53,010*. 44,434' 208,706' .139,944* 3 83,814' 71,347: 284,772 261,579 4 168,386* 132.461 470,842 394,597 Overall Mean 85,747 65,236 290,361 228,530 (a)PCCE : Per Capita Consumption Expenditure (b) TCE : Total Consumption Expenditure * : Significant at a - 0.001 Source : Nampula Smallholder Survey, 1991 respectively. Figures 2.1, 2.2 and 2.3 show the associated Lorenz curves . 24 These coefficients reflect typical levels of inequality in income distribution in rural areas of SSA. This pattern is expected in Nampula rural areas, where farmers households have similar production practices. $ CUM OF TOTEXP 400 80" SD ' 40“ 20 .. all“ confluent. 0.31 0 I'IIT11ll1liill1]lIII]:UI+II#IJITIT:IIILIIITIIIIIII:IIITT111:IT 2 7 12 17 22 27 32374247 5257 8257 72 77 B2 87 92 97 :5 CW 0F WULATIN *NW $01M 0F WTION FIGURE 2.1 LORENZ CURVE, ANGOCHE Source: Nampula Smallholder Survey, 1991 X CW G"- TOTEXP 100 80‘ 60" Glml eta-"mum . 0.31- 20 " O HIIIIIIIIiIIIIiIIITiIIliiritlilnlitlnilIlliiillinIIJIHII 2 1D 18 27 35 43 52 80 BB 77 85 93 ”CW 0F ”ULATICN ' SCUEX SClMOFmATION FIGURE 2.2 LORENZ CURVE, RIBAUE Source: Nampula Smallholder Survey,1991 25 X CW CF TOTEXP 100 80’ 40" mam mutton: .- 0.370 20 t l L l I l l l I l l l l l l l l 1 O III]lllllIIIIIrTITIlIlIIIUTIIIIIIIIIIIIIIIIIIIfiITIIIllIllll 2 7 ‘12 17 22 27 32 37 42 47 52 57 82 67 72 77 82 B7 92 97 X CW OF FUULATIW ' SW SCIMOF WTION FIGURE 2.3 LORENZ CURVE,MONAPO Source: Nampula Smallholder Survey, 1991 Income inequalities in Monapo and Angoche are slightly more pronounced then in Ribaue. King and Byerlee estimate a gini ratio of 0.32 for rural Sierra Leone in 1977, and in a recent study Abdoulaye Fall (1992) found gini coefficients ranging from 0.31 to 0.51 in rural Senegal. 2.3 RELATIONSHIPS BETWEEN HOUSEHOLD TYPE AND CONSUMPTION EXPENDITURE PATTERNS Sets of eight mutually exclusive and exhaustive commodity groups have been created, including all food and non food expenditures. Table 2.9 shows the average budget share9 for each commodity group by district. As expected, expenditures on food are very high, above 70% of total expenditure in all three districts. This result reveals the level of absolute poverty throughout rural areas. Cereals, beans and cassava 9 Shares are means calculated from individual household shares. 26 TABLE 2.9 MEAN HOUSEHOLD EXPENDITURE SHARE BY EXPENDITURE CATEGORY IN MONAPO, RIBAUE AND ANGOCHE Expenditure Category Monapo Ribaue Angoche ------- Budget shares -------- Food 0.801 0.832 0.747 (0.13) (0.13) (0.14) Cereals 0.176 0.338 0.141 (0.16) (0.19) (0.11) Beans 0.069 0.128 0.037 (0.08) (0.11) (0.05) Cassava 0.330 0.229 0.249 (0.20) (0.16) (0.13) Fish 0.106 0.033 0.206 (0.09) (0.05) (0.17) Meat 0.028 0.043 0.020 (0.05) (0.08) (0.05) Other foods 0.092 0.060 0.090 (0.11) (0.08) (0.08) Non Food 0.199 0.168 0.253 (0.14) (0.13) (0.11) Cloth & Footwear 0.099 0.068 0.103 (0.08) (0.08) (0.08) Education & Health 0.002 0.015 0.017 (0.01) (0.02) (0.04) Other non food 0.098 0.084 0.134 (0.10) (0.09) (0.11) Note : Values in parentheses are the standard deviation. Source: Nampula Household Smallholder survey, 1991 constitute 58% of total household food expenditures in Angoche, 72% in Monapo, and 84% in Ribaue. Expenditure on cassava is relatively higher in Monapo (33%), and about one fourth of total expenditure in Ribaue and Angoche. As expected, Angoche has the highest mean budget share for fish, constituting the third major component of household consumption expenditure in Monapo and the second in Angoche. Non food expenditure shares range from about 17% in Ribaue to 25% in Angoche. Education and health budget shares are very low, around 2% in each district. This 27 reveals the large deficiency in the provision of these services in rural areas. The analysis of budget shares by income class can provide insights on the pattern of expenditures on luxuries and necessities. Luxuries are goods or services whose budget share rises with income, while budget shares for necessities fall with increases in income. Figure 2.4 illustrates the average food budget shares by income class (PCCE quartiles) in Monapo, Ribaue and Angoche. MEMORIES ll “-55 : 1/ O 8- N M 0.5 _ 0.4- 0.2 " O ' '- "l 2 3 4 “MILE! —*— MONAPO —*— n I BAUE + ANGOCHE FIGURE 2.4 TOTAL FOOD BUDGET SHARE BY PCCE QUARTILES IN MONAPO, RIBAUE AND ANGOCHE Source: Nampula Smallholder Survey, 1991 The food budget share is dominant in all three districts, and there is no significant change on total food budget share as PCCE rises in Monapo and Ribaue. In Angoche the average food budget share rises 28 from 66% in the lowest quartile to 82% in the highest. This increase is not expected despite the great poverty. There is quite large food budget shares variation among households in the same income class (quartiles) as shown by the standard deviation in relation to its mean. Table 2.10 and Figures 2.5, 2.6 and 2.7 show the demand patterns for selected commodities by income class for each district. Given the general situation of acute absolute poverty, and widespread market shortages, demand patterns for some basic commodities like cereals in Monapo and Ribaue, and cereals and fish in Angoche, are positively related to income. However, as expected, cassava budgets shares decline as income rises in all three districts, from 48% to 27% in Monapo, from 26% to 19% in Ribaue, and from 27% to 20% in Angoche. Fish budget shares rise with income in all districts, though they do so most strongly in Angoche. Overall, the data indicate that households tend to move towards better diet patterns as they become relatively better off. The budget share pattern for other non-food items, which includes ' among others soap, tobacco, and some durable items (radio), as’a complement to total food expenditure shares, tend to decline as income rises in Angoche. Probable economic explanation for this behavior is as mentioned above, the level of poverty and the deficient commodity supply. The mean cloth and footwear budget share ranges from 7% to 11%, and is almost equally distributed among households in different income classes. 29 HoNH .%o>u5m umpHosHHmEm mHsmamz “ condom mcoHumH>op pumpcmum mum monogamoumm SH mosHm> H ouoz chaquaomxo GoHuQESmaoo muHmmo Mom . moom R AOHV ANHV AHHV HmHv Adv ANHV AmHv AmHv AoHv ANHV AmHv Ade ”H GN mN Hm NH mH wH mH NH mN ON NH poom-¢oz ANHV Amav AMHV Amav AmHV AeHv ANHV AeHV Haas AeHv Aeav Aowv HN RN mN NN mH HN mN mN 5N mN mm we m>Mmmmo AmHv HmHv ANHV AmHv va ANV ANV Amv Amy va va Ame mN mN wH «H m m.N H.N N.¢ HH HH NH w.w LmHm Amv ANV Hoe Amv AHHV HdHV AHHV ARV Ame Aev Ame Adv e m o m mH mH MH HH m.m N.w m.m H.¢ mcmom AMHV on AHHV AHHV ANNV Ava HHNV Ava Ava AmHv AmHv AHHV eH mH mH HH oe On On Hm mN mH. wH H.m mHmouOo AoHV AMHV HHHV AmHv _ Adv ANHV AmHv AmHV AoHV ANHV AmHV Ade Nw mm mm mm ww Hm Nw Nm mm mu ow mm poom ---- eHHuupso xmuum Ne ensuwpdedxm Hence we N ..... e m N H d m N H H m. N H huomoumo onoowa< odmnHm oamcoz endqucOmxm mmoouz< oz< MD0, the budget share increases as income rises, while for those with Y1 <0 the budget share decreases as income rises. Expenditure shares independent from income have 71 - 0. If'lqj < 0, then the share of good 1 decreases with the increase of the jth household characteristic, holding constant the remaining factors. The identity (3.6) reflects household budget rearrangements, in response to different household characteristics, meaning that whatever the household budget shares reallocation, the effects cancel out to conform to the income unity constraint. The model is estimated using ordinary least squares (OLS) for each single equation, hence it is assumed that the classical linear regression assumptions hold. The classical regression method has been applied in similar studies by Hazell and Roell (1983), King and Byerlee (1977), Deaton and Case (1985) and others. The parameter estimates allow the calculation of marginal budget shares (MBSi), average budget shares (ABSi) and expenditure elasticities (E1) as shown by equations (3.8), (3.9) and (3.10): 43 M381 = aEi/aE 2 p1 + Yi(1 4' M) + Elijzj' (3.8) 31 11331 = 31 =91 + yiLnE + 221.323 (3.9) d 5, = {<9E1/E315}w:E/15i = mam/ABS, (3.10) As usually done these are short run (point) indicators calculated for the average household, evaluated at the sample mean expenditure (E) and the sample mean value of household characteristics (23 ) for each district. The marginal budget shares are important predictors of consumption linkages as income changes (Hazell and Brown, 1989), although the growth linkage response from good i depends on good i's supply elasticity. The marginal budget shares tell how a one unit change of household income would affect the demand of a specific good or service holding constant all other factors. The expenditure elasticity allows the categorization of goods according to their income elasticity. If the £1 is less than or equal to one and greater than zero, goods 1 are often called normal necessities. Goods with elasticities greater then one are considered normal luxuries, and goods with elasticities less then zero are inferior necessities (Layard and Walters, 1978). As income rises one would expect food expenditures to increase less rapidly than income, hence food in general is a normal necessity. Expenditure on luxury goods would behave in the opposite manner. Endogenous variables 44 Variables included in the analysis The dependent variables in any household consumption expenditure analysis are a group of commodity expenditure shares which are regressed against income and other selected independent variables. Figure 3.1 shows the 13 commodity budget shares used in this study. Name Commodity Subgroup FDBSHARE Total food budget share FCBSHARE Cash purchased food budget share FOBSHARE Own produced consumed budget share CEBSHAREa Cereals budget share BEBSHARE Beans budget share CSBSHARE Cassava budget share FSBSHARE Fish budget share CRBSHARE Meat budget share OPESHAREb Other food budget share NFBSHARE Non-food budget share CFBSHARE Cloth & footwear budget share EHBSHARE Education & Health budget share OTBSHARE° Other non-food budget share 0‘93 petroleum, other household utilities. soap, Maize, Maize flower, Sorghum, and Rice; Sugar, Cooking oil, groundnut, salt, vegetables, and coconut; radios, transport, tobacco, taxes, and FIGURE 3.1 DEPENDENT VARIABLES INCLUDED IN THE REGRESSIONS concern . The product aggregation is consistent with the study's main We are particularly concerned with household cash purchased food shares, fish budget shares, and non-food shares as principal indicators of linkages. Although expenditures on fish are already included in the variable cash purchased food, the regression coefficients associated with fish budget share are important per se as they indicate the importance of household linkages with the non- agricultural economy. Furthermore any RHS variable that has a 45 significantly positive impact on these endogenous variables represents a household characteristic that is associated with linkages to the outside economy. Conversely, any RHS that has a significantly negative impact indicates a household characteristic that is strongly associated with household subsistence orientation. Exogenous variables Figure 3.2 illustrates the set of exogenous variables included in each district. The independent variables were selected taking into account the household demographic, economic and institutional framework. Name Description Unit INTERCEPT Intercept. LNTOTEXP Ln of per capita total expenditure. Meticais FAREA_PC Per capita access to land. Ha/people HMEM Total members in the household. People HHAGE Age of head of household. Years HHED Head of household level of formal Years education. PWMHH Proportion of adult women in the % COTSHARE (or AZAMSHAR) CAJ SHARE DUMMY_GR DUMMY_NT DUMMY_FC household. Cotton (or rice and Groundnut) sales as % of total income. Cashew sales as % of total income. Dummy for head of household genderz‘ Male headed hOusehOld -1; otherwise -0. Dummy for native household: Native household -1; otherwise -0. Dummy for household with dry and wetland: Household with both dry and wetland -1; otherwise -0. % Z , FIGURE 3.2 INDEPENDENT VARIABLES INCLUDED IN THE REGRESSIONS Although income is usually the dominant variable in explaining expenditure patterns in cross-sectional studies (assuming households 46 face the same price), it is widely accepted that household characteristics have to be taken into account in the explanation of household consumption expenditure behavior. Household size, simply defined as the total number of members in the houéehold, is generally considered to be an important explanatory variable in household consumption expenditure patterns. It is expected that large families will have larger total food budget shares than small families ceteris paribus. The proportion of women in the household, the level of formal education and the age of the household head were included in the model to trace their possible association with household consumption expenditure patterns. We expect that households with a larger proportion of women are likely to have a larger share of own food produced and consumed, and the higher the level of education and age of the household head, the larger the cash food expenditure share. In addition to household characteristics, other key explanatory variables are included: total expenditure per capita, proportion of selected cash crops12 sales in total expenditure, and the household's per capita access to land. Where all factor and product markets eXist and operate 4 competitively, one would expect that the proportion of cash income would not affect relative food and non-food budget shares allocated to each type of food. Household choice of crop mix and allocation of household resources would be only based on price expectations. But the current market situation in rural Nampula constitutes an additional constraint on household consumption decisions. 12 Cotton sales for Monapo and Ribaue; Rice and groundnuts for Angoche and Cashew for all three districts. 47 Most households are very poor and face widespread market failure for both factor (there is no formal capital market in rural Nampula, and the labor market is very weak) and product markets. The predominance and association of these facts (poverty and a large degree of market failure) lead to great difficulties in purchasing food and other commodities, and constrains the choices open to the household. As income rises, households would obviously tend to move towards improved diets. But households in rural Nampula have limited means to improve their consumption. Hence we hypothesize that as household income rises, holding constant all other factors, households would improve their diets through two adjustments. First the households would adjust in their production mix, which would affect household consumption patterns. Households would move towards the production of more preferred items. Particularly in rural Nampula we expect that as income rises households would substitute out of cassava into cereals and beans. The second household adjustment would be the response to the strong limitation on access to cash purchased food and non-food commodities due to market failures and the relatively high cost of purchased food. "Purchased food, driven largely by fish is between 29 and 70 times more expensive than the value of retained own production" (MOA/MSU/UA Research Team, 1992c). Therefore it is hypothesized that households with larger proportions of cash income would tend to have larger expenditure shares on purchased food and non-food. This consumption pattern would lead to a positive relationship between income and total food budget share, given the larger share of cereals and fish in household total food budget. Land is one of the most important household assets in rural Nampula. Poor households are associated with small land holdings and income is highly correlated with land holdings (MOA/MSU/UA Research 48 team, 1992f). Therefore it is expected that access to land will exhibit the similar sign pattern as income, except obviously for purchased food budget share. Households with more access to land would have a consumption expenditure pattern corresponding to a relatively high income households, with larger budget shares for cereals and fish, and relatively lower budget shares on cassava. Cash crops are in general perceived as means to increase smallholder income and stimulate growth linkages with other sectors in the local economy. The general hypothesis is that revenues from cash crop would be positively related to cereals, beans and fish budget shares and negatively related to cassava budget shares following the similar pattern as income. It is also hypothesized that cash crops have positive effects on non-food budget shares, particularly for cashew and cotton, whose institutional support may increase the availability of some non-food commodities to households selling cashew or growing cotton. Due to agroecological differences among the three districts cotton and cashew crops were selected for Monapo, and rice and groundnut and cashew for Angoche. To take into accountqualitative household characteristics, like the effect of household access to both types of land: dry and wetland; native household and the gender of head of household, three dummy13 variables were included in the model RHS. It is hypothesized that households with two types of land have relatively higher expenditure shares on own food produced consumed. Male headed households are hypothesized to have higher cash purchased food and non-food budget ‘3 A dummy variable is a binary variable constructed such that it takes the value of unity whenever the qualitative phenomenon it represents occurs, and zero otherwise.(Kennedy, 1943) 49 shares. Native households are expected to be associated with higher own food produced and consumed budget shares. 3.3. MODEL RESULTS Tables 3.1 and 3.2 show the parameter estimates for selected household expenditure shares grouped by commodity and origin. The complete regression results are shown in Appendix Bl and B2. The overall significance of the regression can be assessed by the F-tests against the null hypothesis that the expenditure shares are not influenced by any of the exogenous regressed variables. As expected the null hypotheses is rejected for almost all the expenditure shares regression at less than 10% level of error, except for fish and meat budget shares in Monapo and beans, meat and own produced food budget shares in Angoche. Empirical data shows that the average fish expenditure share in Monapo (29.5%) is relatively lower than in Angoche (36%) (MOA/MSU/UA Research team, 1992a Table 11 pp 18). In Monapo fish is less available and has a higher cost then in Angoche. The cost of dried fish in Monapo is almost twice (1.78) that in Angoche (MOA/MSU/UA Research Team, 1992c). It is therefore not surprising that the fish budget share regression is nOt significant at 10% in Monapo while in Angoche is strongly significant (0.0065). The discussion of the model resultsl“ will concentrate on variables directly related to the hypotheses to be tested. The coefficient estimates predict the extent and direction of 1"In interpreting the results it should be noted that the derivation of Engels curves using commodity share equations results in lower R2 statistics than if actual expenditures were used as dependent variables. 50 HmmH .>m>eam emvponppmsm p—zasmz “mor30m mmpnm_tm> ucmucmamo we no»: mew: m:o_um:uo modern me=u_u:maxm Pmauom cw can» mowum_umuw em LOHHmEm HHHmo_axu o» wvmmp wcozmzvm 0.»me we coszK—mm mp: .mLoLLo Upmvcmuw Hymn—manta 50.; Human—.3030 .mpm>m_. 00:00 .chmwm 0L0 mmwmsucmgmn E. mmgam w... ”OHOZ Hmoo.o Novm.o mooo.o o~oo.o eon.o HmHo.o ooo.o HooN.o NVNo.o ooo.o vmvo.o NvHo.o Humumtmvwcmww. momH.o mmoo.ou omHH.o moHH.o omHo.o Homo.o mmHm.o Nomo.o- HoN.ou NmHm.o vomo.o ovNH.o em .uumamn< HmHmm.ov HNemm.ov ANNoo.ov HHNNo.ov HvHoo.ov HNomw.ov HmomH.ov Homon.ov HHooo.ov Hooo.ov Avoo.ov HoNvo.ov moNo.o wao.o- mmNo.H- NmNm.o ONHN.o Hemo.ou co-.o- Homo.o m~mo.o- NHNo.H .omov.o mome.ou Hcmumcoo HoooH.ov anmm.ow HowNH.ov vaNN.ov Hmmmm.ov HNon.ov AoNoH.ov HmNNm.ov HmHmH.ov HHmoH.ov Hemmm.ov Hmmmv.ov oomo.o oNHo.ou NHmo.o- mNHo.o moHo.o- mcoo.o oHoo.o- moHo.o Hoqo.on HNoH.o mmHo.o meo.ou mwl>zzso ANNmm.ov HoNHm.oV HVNNN.ov Hmmmo.ov HooNN.ov Avmmov Hono.ov ANNNo.oV HONon.ov HommN.ov vaNH.ov Hmsmo.ov NHoo.o- wooo.ou oHHo.o wooo.o ovoo.o- oooo.ou mvHo.o- wooo.o mNoo.o mmHo.o- m~oo.o- HmHo.o zmz: HmHoo.ov HNmom.oV HHooo.ov Ammmm.ov HNmmN.ov HmHNm.ov Amoco.ov Homem.ov HomON.ov Hooo.ov HmvHo.ov HNmHo.oV mooo.o HoHo.o mmHH.o mvNo.ou NHHo.ou NNHo.o Ammo.o meoo.ou mmHo.o NmmH.ou vaoJo- mmoo.o deHOHZE HmooH.ov HNva.ov Amoco.ov Homoo.ov Hommm.ov Hovoo.ov HomNN.ov AmomN.ov HomHo.ou AeomH.ov HoNHo.ov Homom.ov meo.o mmoo.o oNNH.o- Hmoo.o- Hvoo.o mmNH.o NNoo.o ovvo.ou muoo.ou owo~.o mmvH.ou mmHo.o ma<=mwzzso Hoooo.ov vawm.ov ANNNN.ov AmHoo.ov Homeo.ov vaoo.ov HQHNm.oV Amomv.ov HmNNo.ov Hoomv.ov AmmoH.ov HNHwN.oV evNH.o oeoo.o oomo.os NmmH.o- NNmo.o- vomH.o oHHo.o- omHo.ou mvoH.o vaH.o: Nmmo.o- HNNo.o Nzzao wmn=m LmnHOLHHmEm mpaqspz ”moraom .mm_np_em> “contempt no new: mew: mu:u_ncmnxm Hpauop »_ coca mu_um_umun em LOHHpEm appmo_axu ou nnpm_ mco_um:cm oepzm mo co_upewumo one .n_m>wp mocpo_w_:m_n urn mononucmtma c_ mor=m_u "muoz mooo.o Hwoo.o mNeN.o mooo.o mvHo.o mooo.o mooo.o mvHo.o Aupunnavmcm_m mooH.o oomH.o oNNo.c mooH.o ¢¢~H.o AmmH.o HmHN.o eVNH.o «a .numznn< Hooo.ov Amoco.ov Amueo.ov HNoNH.oV Hosvm.ov Hommfi.ov Hvooo.ov Hoooo.ov veov.H mmem.H- vo~m.o meo¢.o- «moH.o- mmev.o- mmHo.H vooH.H unpumcou Asovm.ov HHomv.ov Hmnmm.ov HNovo.ov AHHoN.oV “Non.ov HoNoN.ov HHHoH.ov mHNo.o mmeo.o- mHNo.o mHNo.o- omHo.o memo.o- HHNo.o omHo.o- mo|>zzso Hch¢.ov Hmme.ov HHHNv.ou HvHov.ov AsmoH.ov HVNNm.oV memv.ov HwaH.ov mmoo.o- mmHo.o oooo.o- omoo.o mmHo.o HHoo.o- «NHo.o- mmHo.o- zuzz Hoooo.ov Hoooo.ov Huvsm.ov Hoooo.ov HoNN~.oV Ammoo.ov HHooo.ov HoNNN.oV mmoH.o- vaH.o ono.o- omoH.o m-o.o ammo.o mooH.o- mNNo.o- axNHOHzE HchN.ov HNQNH.oV HvNNH.oV HumoN.ov Ammmm.cv HHomm.ov amsmm.ov Hmwmm.ov moNo.o- HovH.o- oNoH.o moNo.o mHoH.o- m~oo.o vmmo.o mHoH.o wzzzzo AHNHN.oV Homeo.ov HmoH~.ov HHNNN.ov HmmmH.ov HHmoo.ov Ammoo.ov HmmmH.ov HmHo.o oeHo.o ommo.ou HoHo.o- AmoH.o mmNN.o NmNm.o- HmoH.o- mzz=o ua ezuozudwo 2.00% ounce: naps-73> acoucoaoucH mmoouz< 92¢ omu3m HopHonHHmEm mHsmamz ”condom 59 mn.o o.mH H. wN «H.H N.NN N.NH . poom-¢oz em.H N.o¢ m. mN mm.H ¢.wN N.HN poom Sumo mN.o H.¢¢ e. Ne mw.o N.w< N.wm poom poospoum :30 nH.H o.mm N. mN NN.o m.NN H. on boom Hmuoe Ho.H H.N e. N mo.H N.mH d. N mpoow nonuo oo.H N.N N. H Nw.o m. N m. N new: ow.H w.Hm N. NH NH.H m. NH m. OH anm 0N.o m.MN w. eN Nm.o N. NH N. Nm m>mmmmo wo.o N.N d. m mm.o N. m N. o mammm No.H N.mH N. «H mm.H N. mN w. NH mHmouOo ............ OHSOHUGOQxO kuou mo OGOOHOQ ------------- moumsm mouwzm mOHOSm moumcm NuHOHummHm uowpdm powpsm NOHOHummHm powpsm uowpsm oudquGdem HmcHwHflz meum>< OHSquCOme HmanHmz meuo>< Nuowoumo Osoowc< oamcoz ousqucomxm mmoouz< 924 omtam coupospmeN «HanEmz “ monsom .mcpozmmaog NcpxoeN couuooscoc new couuoo to» mmpnmNLp> mzm pr do Ho>wH cums “m umuppsopmu ” « NH.H c.NH w.eH NH.H o.o~ N.m~ coou-coz ov.H ¢.mN H.wH Nm.H N.om o.mN noon comproead zmmu em.o o.Nm H.No Ho.o w.mv o.mm coed umozuoea :zo NN.o ¢.Nm N.mm mN.o o.vN m.oN wood HmuoN Nv.H N.NH N.HH Nm.H N.mH m.N poo» emcee mo.o m.N m.~ Nw.o N.N N.N ppm: NN.H N.oH N.m NH.H o.mH N.HH ;m_u vo.o m.mN o.om Nm.o m.NH H.om m>pmmmu Nm.o m.v o.w Nv.o H.m v.0 mcmmm mm.H o.mN N.oH mm.H o.mN N.mH mHMOemo ........ ------ mt:u_vcmaxm Hpuou mo acoutma -uu---u--- mmtmcm mmtmcm wormsm mmepcm 36:85 883 page D 3:3: 885 “page mrzuwucmexm Hpcwmrpz mazpomn< wuzuwccqum —m:_thz manponn< NEONmumu couuou-coz capaoo oe=u_pcmnxm oN<:mm mazNHoszxm onNstmzou H.v m4mtzm cmvpocHHmEm ppaasmz " moczom N o m.mN N.oH «.HN N.HN H.vH N.mN o.oN voou-coz m we N.Nm N.mm N.vm N.om N.Hv N.NN o.HN . poem nmnpcouzm smog N we .ov N.Nv N.Nv o.mv H.vv m.wv o.Nm . pood noozpotd :30 N mN N H.mm o.NN . N.NN N.mN m.oN o.om noon kuop N oH H H.N N.N N.NH N.w H.NH N.oH voom tmzuo e.m N o.m o.N v.N N.m m.N H.N . “no: N mm m mH N.NN N.NN N.NH «.mm N.mH m.oH ;m_u H NH N.oH N.mN o.NN m.vH m.NN N.NH N.vm m>pmmmu N N N.N N.N m.N N.N N.N o.m m.o mcwmm N NH N eN H.mH m.mH N.NN m.mH «.mN N.mH mHmmtmo v N H v m N H zeommumu usoomc< cameo: . mtzanconxm wzuowz< oz< omm mmm