J' The World Bank i zs Democratic RepubkAh.,.-/t02 73 Ministry to the Presidency of the Republic, in Charge of Finance and Planning General Directorate of Planning Public Disclosure Authorized

/ Report No. 10273 Public Disclosure Authorized The Impact of the - Road: A Socio-Economic Study of the Andapa Basin

Public Disclosure Authorized Volume I

Prepared by Public Disclosure Authorized

/

Brigitta Mitchell Xavier Rakotonirina Transportation Department General Directorateof Planning

Washington, D.C., USA December,1977 Antananarivo, MADAGASCAR v

THE IMPACT IF THE ANDAPA-SAMBAVAROAD:

A SOCIO-ECONOMIC STUDY OF THE ANDAPA BASIN

MADAGASCAR

VOLUME I

-f TABLE OF CONTENTS

VOLUME I Page No.

SUMMARY AND CONCLUSIONS .. e...... i-x

I. INTRODUCTION ...... 0.0Oe0 ... 000000*o I

A. THE STUDY AREA ... oo...... oooooo ...... o 1 B. THE ROAD PROJECT 2 ...... o...... 2 C. THE ROAD IMPACT STUDY ...... ,....o..o.... oo 2

II. METHOD

A. GENERAL STUDY DESIGN ...... 6 B. THE SAMPLE SURVEYS: 1965-1975 ...... 8 1. The Baseline Survey: BCEOM, July 1964-June 1965 8 Household Sample ...... 9 Sample of Areas Under Different Crops ..... 10 Market/Price Survey *...... 11 Survey of Air Transport Movements to/from Andapa ..*.0.0... o...... 0....0o 0o 00.0.... 11 2. The Post-Road-Completion Survey: IBRD, July- December 1975 ...... 0...... 11 Household Sample (incl. area cultivated) .*. 12 Market/Price Survey ...... o ..... 16 Transporter Survey . .. 16 Origin/Destination Surveys ...... 4...... 16 Special Surveys .0..00.00. 000...... 0 o*o 17 C. PATA COLLECTION ..... 00 .....0 0 0 . .0.000 ...... 0 00 0 s The Base Survey - 1965 ...... 0...... 18 The Follow-up Survey - 1975 .... . 0.0...... 0 19 D. THE FRAMEWORK FOR THE ANALYSIS ...... 20

III. FINDINGS

A. REGIONAL STRUCTURES 1965-1975: COMPARATIVE PI.,TURE DRAWN FROM SECONDARY DATA SOURCES AND SPECIAL SURVEYS ...... 00...... 0 ... 28 1. Local Administrative Organization and Structure * 28 2. Social Services in the Study Area . 31 - Education .. 00 .00..0.00..0.000...0.0....0000. 31 - Health ...... o. 32 3. Infrastructure and Technical Services o...... *. 36 Banking .0*000000000.0...... 00...... o 36 Tax Collection ...... 36 Licensing/Registrations 00 0 ...... 0...... 36 Public Works Department .....o...... 37 Water and Forestry Department ...... o ...... 37 Agriculture Service ...... oo ... 38 OPACA/Rural Engineering ...... o ...... 38 -2-

Page No.

4. Agrict.lture: Area Potential and Area

Produiction:1965-1975 ...... e . 41 5. Prices of Major Agricultural Outputs and Inputs: 1935-1974...... 46

B. ECONOMIC EXCHANGESBETW'EEN THE REGION AND THE REST-OF TaE-WORLD ...... o. .. oo.oo. ... oo. 50 1. Air Exports/Imports 1965: Volume and Cost o....,50 2. Road Traffic: 1975 *e.oe...... ,e,, 52 The Local Road Transport Industry ...... 52 Transporters Operating Revenues and Costs 53

Freight and Passenger Transport Tariffs *.. 56 Area Demand for Transport 1975 ...... 56 Estimated Volume of Marketed Production ... 57 Results of O/D Surveys ...... o.... 59 Vehicle Traffic ...... 59 Passenger Traffic .o..... ,..... 64

C. THE ECONOMIC EVALUATION OF THE ROAD INVESTMENT o.... 67 1l Patterns of Crop Production...... 67 2. The Economic Benefits of the Road: The Producer Surplus Analysis ...... § .. 70 3. The Economic Benefits of the Road: The Road User Savings Analysis ...... ,,...... 77 4. Rate of Return Estimates ..... o,....., ..... 83

IV. INCOME DISTRIBUTION EFFECTS: CHANGES IN GENF'RAL WELFARE AND AREA RESIDENTS' LEVEL OF LIVING

A. INCOME LEVELS AND INCOME DISTRIBUTION: 1965-1975 .... 89 B. PATTERNS OF CONSUMPTION EXPENDITURES 111I...... C. SOME SOCIAL INDICATORS OF LEVEL OF LIVING ...... 124 1. Land Tenure Patterns and Agricultural Production Techniques 0 .....0 0 . .00 * , ...... 125 2. Shelter ...... o. o .. 0. 0000000..o. 135 3. Nutrition ...... ooooo*e* o...... 0...... 144 4. Some Demographic Characteristics 0...... 145

V. ATTITUDES AND OPINIONS ON THE IMPACT OF THE ROAD, AND ON INDIVIDUAL AND COMMLr'ITYDEVELOPMENT PRIORITIES * ...... 150 Table No. Table of Tables Page No.

I Universe and Sample of Andapa Area Households Stratified by Occupational Groups: 1965 *.e.....-o 9

2 Universe and Sample of Andapa Area Households Stratified by Occupational Groups: 1965-1975 ...... 16

3 Population of the Andapa Basin: 1965-1975 ...... 0 28

4 Attrition Rate of Primary/SecondarySchool Pupils - 1975 Enrolment ..o...,... ,,.ooo0oe voeoo@*, .00.000.0.. 32

5 Health Service Facilities: 1964/65 - 1975 ...... 33

6 Cultivable Land by Type of Soil, 1965-70-74 ...... 41

7 Major Crops Cultivated - Area Cultivated 1965-70-74 .. 43

8 Crop Production in Tonnes: Andapa Basin 1965-74 ..... 44

9 ComparativePrice Increase for Selected Crops: 1966-74 (ProducerPrice & Realized Export Price: FMG/Kg) *e.oGe.e.....o...... b...... oooo 46

10 ComparativeValue and Volume of Imported Agricultural

Inputs: 1970-74 ....***0 .*..***0 ...... 9 0...... 48

11 Volume and Value of Andapa Exports - 1964/1965 o*...... 51

12 Transporters: Andapa-SambavaRoad 1975 - Revenues & Expenditures by Type of Service & Type of VehiLle .... 54

13 Transporters: Andapa-SambavaRoad 1975 - Average Annual Operating Revenues and Costs per Operator and Vehicle by Gross-Revenue Class of Operator ...... 55

14 AgriculturalProduction and Volume Exported from Andapa: 1965-1975 ...... 58

-0 15 Origin Destination Survey: Types of Vehicles Counted 60

16 Lengths of Trip by Vehicle Type: 1975 o* ...... -...61

17 Traffic Flow by Weekday: 1975 o- ...... 61

18 Freight Transported Along the Road: 1975 ...... o 63

19 Purpose of Passenger Trips: 1975 ...... 64

20 Length of Trip by Purpose of Travel: 1975 ...... 65 -2-

Table No. Table of Tables Page No.

21 Passenger Flow by Weekday: 1975 ...... *§e..e...o 66

22 Area Cultivated, Yields and Producer Prices for Major Andapa Crops: Percent Increase, 1965, 1974 O.Oo..... 67

23 Cropping Patterns at Farm Level, 1965-1975: Percent of Households Growing Different Crops ...... 68

24 Yields, Distribution of Output and Producer Prices Reported by Sample Households - 1975 . .4oa 69

25 Sample Producer Surplus Dy Crop, 1965, 1975 e...... 72

26 Estimated Area Producer Surplus by Crop - 1975 Without vs. With Project . ... .o_ ...... o0.e.* 73

27 Estimated Transport Costs Savings - 1975 Without vs. With Project ...... 82

28 Economic Rates of Return and First-Year Benefits for Three Project Scenarios ...... ,.. 86

29 Proportion of Disposable Cash Income From Different Income Sources: 1965-1975 ...... 91

30 Average Household Cash Income and Production Expenditure - By Occupational Group: 1965-1975 ...... 93

31 Average Household Cash Income and Production Expenditure - By Distance Category: 1965-1975 ...... 94

32 Average Household Cash Income and Production Expenditure by Decile - 1965 ...... o*...... 96

33 Average Household Cash Income and Production Expenditure by Decile - 1975 . 97

34 Per Capita Income by Total Household Income Deciles - 1965 . 99

35 Per Capita Income By Total Household Income Deciles 1975 ...... 00,0 .... 0.e...000000000000. 100

36 Average Per Capita Cash Income and Production Expenditure by Decile: 1965 ...... o... 102

37 Average Per Capita Cash Income and Production Expenditure By Decile: 1975 ...... 103 -3-

Table No. Table of Tables Page No.

38 Comparative Tncome Decile Status of Panel

Households - 1965-1975 .,...... e...... o.. ao.eeoecOOO 104

39 Comparative Income Decile Status of Panel House- holds by Professional Group: 1965-1975 ...... 105

40 Average Change of Total Income of Panel Households, 1975 over 1965 - by 1975 Decile and Occupation . eoooo 108

41 Average Change of Disposable Cash Income of Panel Households, 1975 over 1965 - by 1975 Decile and Occupation o...... c. a ... ** .OO*..*OC. e@4*...O. 109

42 Per Capita Income of Panel Farm Households, 1965, 1975 - by Distance Category .o.....0...... e e 110

43 Proportion of Consumpt4in Expenditures for Different Types of Goods - 1965-1975 0*00.9-9 ....O .c.*e.... ec 112

44 Proportion of Major Expenditure Categories in Total Household Consumption Expenditure by Occupational Group - 1965-1975 .c...... ec...... e...... 114

45 Average per Household Lonsumption Expenditures by Occupational Group: 1965-1975 .... §...... *.. 115

46 Characteristics of Households by Deciles of Household Consumption Expenditures: 1965 .... eee.. ,cec.... 116

47 Characteristics of Households by Deciles of Household

Consumption Expenditures: 1975 ..... e....0.....c..... 117

48 Income Elasticities for Food and Non-Food Consumption by Occupational Group: 1975-1965 ...... ee.,eee 120

49 Income Elasticities for Food and Non-Food Consumption by Occupacional Group: 1975-1965 c..e...... e....o e... 121

50 Per Capita Consumption Expenditures by Categories of Expenditures and Deciles - 1965-1975 ... *...... 123

51 Average Size of Holding and Proportion Planted to Rice - 1975 ...... ce.ecc ...... 125

52 Proportion of Households Owning/Renting land - 1975 .. 126

53 Number of Households Selling and Buying Land - 1965- 1975 e oe e oee...... 127 -4-

Table No. Table of Tables Page No.

54 Acreage Bought and Sold - 1965-1975 . 128

55 Price Ranges Reported for Land Purchases and Sales: 1965-1975 .e.*.e ...... o* ...... o*o 128

56 Equipment Ownership by Area Households: 1965-1975 .O 130

57 Proportion of Sample Farmers Reporting Expenditures for Different Agricultural Inputs: 1965-1975 ...... 131

58 Proportion of Expenditure for Different Agricultural Inputs: 1965-1975 - By Occupational Group ...... 132

59 Animal Ownership by Area Households: 1965-1975 ...... 133

60 Wall Types of Houses Sampled - 1964-1975 ...... 136

61 Flooring Materials in Houses Sampled - 1964-1975 ..... 136

62 Roofing Materials on Houses Sampled - 1964-1975 ...... 137

63 Number of Rooms per House in Houses Sampled - 1964-1975 ..*...... eeeac,..eo.ooo ...... -.... 138

64 Percent of Households by Occupants per Dwelling Unit

1975 ...o D Oe.*.e . o o. * 0...... o * ... o ee.o .... 0..... 139

65 Distribution of Types of Lighting by Sub-Area, 1975 141

66 Distribution of Sources of Drinking Water - 1964-1975 ...... o...... e o 14

67 Age Distribution of Population - 1975 146

68 Percent of Heads of Household by Occupation - 1975 ... 147

69 Education Level of Andapa Population by Age and Sex: 1975 O.OOe..*[email protected]...... 148

70 Opinions on Impact of the Road - 1975 .00..00...... 150

71 Average Number of Trips Reported Before/After Road Opening * * * .... 00e ...... o v*o ....o -oo.oo .... 151

72 Individual Priorities: 1975 ...... ea..** 152

73 Individual Priorities by Location of Household and by Occupational Group -1975 ...... o...... 153 -5-

Table No. Table of Tables Page No.

74 Community Needs - 1975...... 154

75 CommunityPriorities by Location of Households and by OccupationalGroup - 1975.156

VOLUME II

TABLE OF APPENDICES

Appendices

A Sample Structure

B Miscellaneous Regional Statistics

C Area Exports and Imports - Air and Road Traffic

D Road Costs and Benefits

E Some Social Indicators

F Sample Household Incomes and Expenditures

G Questionnairesand Codes Acknowledgements

This paper, being a longitudinal study of a small regional economy, has drawn on the work and ideas of a great number of individuals and insti- tutions, many of them more intimately familiar with the Andapa area, its potential and its problems than the authors. The first acknow7ledgement is to the Malagasy ijthorities: without their active interest in tracing and evaluating the socio-economic impact of providing road access to a previously isolated region, the study would not have been undertaken at all. Particular - thanks are due to Mr. R. Andriamananjara, Director General of Planning, whose office provid.d counterpart support and facilitated efficient liaison with the many central and local administrative agencies, whose cooperation was essential to th -access of the 1975 follow-up survey work.

Much of the latter was based on the solid foundations that had been evolved in the design of the 1965 base survey by the two consulting firms AGRER and BCEOM. Both made earlier reports, drafts and supporting materials available to the present study. Messrs. Drachoussoff and Poirier of the AGRER team freely shared their first-hand knowledge of the study area and its socio-economic structure; Messrs. Odier and Brisson ensured the support of the BCEOM field office in Antananarivo that made it possible to reassemble the valuable baseline information completely.

The authors gratefully acknowledge the contributions of the 1975 fieldteam, which consisted of an interdisciplinary group of five post-graduates from the Swiss Federal Institute of Technology, Zurich, and of ten Malagasy field investigators recruited locally in the study area. Mr. A. Schlaepfer both supervised the coding of interview results in the field and prepared the basic descriptive tabulations of survey findings in Zurich. The Research Triangle Institute (RTI) of North Carolina produced the tables for the house- hold incomes and expenditures analysis. Ms. V. Kozel carried out the consid- erable computational work for the economic analyses.

In the preparation of this report, valuable comments were received from many colleagues. Mr. E. Vickery of RTI who cooperated over several years with the Transport Research Division on the problems of measuring the socio-economic impact of rural infrastructure investments, contributed much to the elaboration of the analytical approach used in this study. Messrs. C. Willoughby, C. Carnemark and particularly Mr. Martinez had con- structive critical comments on points of methodology and presentation. Ms. F. Johansen's thorough review of the draft clarified points of subs.ance and context. Mr. Rakotonirina contributed valuable insights on local institu- tional arrangements and practices, and considerably improved the presentation of the French versior.. Special thanks are due to Ms. Wendy Wright who cheer- fully persevered through the typing of successive drafts. And last not least a debt of gratitude is owed to the many people in the Andapa area, without whose willing participation and cooperation there would have been no studv project. The authors alone, of course, are responsible for the conclusions presented and for any errors that may remain. Abstract

In 1970, a first roadlink was opened between Andapa, a formerly isolated valley community in thlenortheast of Madagascar accessibleonly on foot or by airplane, and the port of Sambava, A socio-economicsurvey of the region had been undertakenin 1964/65, five years before the road was completed0 A follow-up survey was mad, in 1975 with the aim of measuring the Impact of the road on the regional ecr- and on residents' level of living. For the ex post economic analysis,b( ne traditional road user savings method and the producer surplus method were used to estimate project benefits. the road user savings measutre showed the road to have been a marginal invest- ment (economic rate of return - 6,3%) even if the area generates important quantities of rice for export along with the vanilla and coffee previously marketed by air. Significantincreases in rice production are contingent oniadditional investments(currently being undertaken) in irrigationand drainage to ake large-scaledouble cropping of rice feasible in the near fututre, The economic rate of return for the combined road and agriculture projects, based on a producer surplus approach which incluaes benefits on self-consumptionas well as surplus on marketed production,was estimated at 1 10 4Z,!

The likelihood of projected returns being realized is partly a funclJionof project planning and implementation,partly of area farmers' respornset.o new opportunities. The latter is explored through an analysis of changes in selected social indicatorswhich reflect farmer responses to- date, Examinationof area household ihcome and expenditurepatterns before/ without and after/with the project showed that, even with incipientdouble- cropping only, farm incomes have increased substantiallyin the follow-up survey period. Producer prices (centrallyadministered) of major Andapa cropG have been rising. Area farmers have responded to the opportunities of realizing better returns on their crops. They market greater amounts of coffee, a traditionalarea cash crop on which extension efforts have focussed most to date, They also export rice, traditionallygrown for subsistence only and for which there was only a negligible local market without the road. Labor force participationin the region is high, both for men and women. Prevailing patterns of landownership- mostly farming on own land - have promote. quitable sharing by area residents in overall income gains: in- come di-ribution in 1975 is less skewed than in 1965, Although food ex- penditures continue to dominate consumptionexpenditures for all but the wealthiest group, ownership of consumer durables has spread considerably from the time of the base survey. Individualpriorities of area residents for the future center on better housing and larger farm operations;main goals for community development are better local roads, improved health care and more schools. Based on past behavior and present aspirations,area farm- ers can be expected to respond well to future opporturitieswhich full imple- mentation of the agricultureproject will open up to them. CURRENCY EQUIVALENTS

1970 US$1.00 - FMG 277.71 iFG L - US$0.0036

1971 US$1.00 - FMG 277.03 IMG I - US$0. 0036

1972 US$1.00 - FMG 252.21 FMG 1 - US$0. 0040

1973 US$1.00 - FMG 222.70 FMG I - US$0.0045

1974 US$1.00 - FMG 240.50 FMG 1 - US$0. 0042

1975 US$1.00 - FMG 214.32 FMG 1 - US$0o0047 SUMMARYAND CONCLUSIONS SUMMARY AND CONCLUSIONS

Project Description

lo This study presents the results of two successive socio-economic surveys carried out in the Andapa Basin in the northeast of Madagascar. The area is small, but well defined: a fertile basin (500 m above sea level) of about 76,000 ha, cut off from neighbouring regions by mountain-ranges varying in altitude between 800 m above sea level in the south to over 2,000 m above sea level in the north and west. Spontaneously settled shortly before the turn of the century, by mostly Tsimihety farmers moving in from the regions to the southwest of the basin, the area currently has a population of 60,000. It has been producing a mixture of cash and subsistence crops: vanilla and coffee provided the cash income, while rice and manioc have been the tradi- tional staples. Despite its growing economic importance, the region remained physically isolated from the rest of the country (except for an air link) until 1970, when a first all-weather round connection of 107 kilometers from the Basin's administrative center, Andapa, to the coastal port of Sambava was opened to traffic. The two surveys in the area were carried out in 1964/65 and in 1975, about five years before and five years after the completion of the road project respectively. During this time, the first phase of an agriculture project was also implemented in the region.

Hypotheses

2. The study attempts to describe, analyse and measure the socio-economic changes that have occurred over the ten year study horizon, and to determine what portion of the change observed can be attributed to the road and the improved access to markets and other services that it has provided. Specifically, the data was used to test the following hypotheses:

(a) Production impact of the road:

- with reduced costs of transporting production inputs and outputs, producer surplus in the road area of influence should increase and the boundary of commercially feasible agriculture should extend. Share of marketed production in total production should be higher with than without the project.

(b) Incidence of road benefits (= income distribution effects of the road):

- The postulated production impact of the project is predicated on farmers/producers receiving a significant share of the road benefits. More of the transport cost savings from improved access will be passed on to agricultural producers with a higher degree of competition among the suppliers of transport and other - ii -

marketing services. Where transporters and/or traders are in a monopoly position, they can be expected to appropriate most of the road benefits and little change in agricultural production patterns would be observed. The relative share of road benefits between transporters/ traders and peasant-farmers should be reflected in

(i) the level of farmgate prices: if actually observed "with project" farmgate prices approach theoretical "with project" farmgate prices (i.e. if the major part of transport cost reductions is passed on by the transporters to the farmer/producer enabling more of them to profitably market a larger share of their crop), the road can be assumed to have con- tributed to a more equitable distribution of income.

(ii) observed income differentials between different sub- groups of the area population: if these decrease in the "with project" case as compared to the "without project" case, the road can be assumed to have contributed to a more equitable distribution of income.

(c) Road induced chano,es in area residents' patterns of consumption:

(i) as the level of income of area households rises, a lower proportion of income should be spent on food (Engel's Law).

(ii) consumption of high volume/low value goods (i.e. con- struction materials) may be expected to rise dis- proportionately due to high share of transport costs in total price, and therefore appreciable price reductions for this class of goods.

(d) Perception of project impact by target-population:

- if projects address the perceived needs of the target population, they are more likely to be implemented successfully. Area residents' development priorities are useful indicators of the impact of completed projects and provide inputs into future project planning and design.

Method

3. Two alternative approaches were used to estimate the contribution of the road investment to the regional economy (production impact). First agricultural producer surplus (PS) was employed to -.easure the project's benefits. 1965 and 1975 survey information on agri.cultural production techniques, cost of production, and share of production marketed and self- consumed was applied to estimate producer surplus on total area production without and with the project. - _lli -

4. Second, the traditional road usor qavin&s_(RUS) measure of project benefits was employed, using with project transport cost savings on freight and passenger traffic as the basis for calculating project benefits.

5. Economic rates of return were caJ.culited based on both measures. The costs of the road, as well as those of the agricultural investments were incorporated in the calculus, as the projected volume of agricultural pro- duction to be transported over the road assumed successful implementation of the agricultural project (OPACA) being carried out in the region; conversely, increased rice production attendant on the agriculture project could not be exported economically, nor indeed physically, without the road. Besides "lactualproject" returns, those for a number of different scenarios illustrating the effect of alternative timing and sequencing of "package" components were also established.

6. Producer prices of major area cash and subsistence crops could not be used as tmeasure of benefit incidence between sub-groups of the area popu- lation: producer prices of vanilla, coffee and rice are centrally adminis- tered and apply with little variation in all parts of the country. Farmers in the study area reported that they were paid the official 'weigh-station' price for their marketed crops. Similarly, transport tariffs for different standard roads are prescribed by provincial authorities and state marketing monopolies exist for major crops, leaving little room for price bargaining between producers and transporters.

7. Survey data collected on household incomes for the "without/before" and the "with/after" project case by contrast were detailed enough to permit analysis of comparative income differentials between occupational groups (farmers, craftsmen, employees, traders); between households living at various distances from the road; and between "income deciles". Observed distributions were examined in the light of social structure and prevailing institutional patterns (household composition, land-ownership, production technologies, producer price policy) as well as the change in transport services. A quan- tification of the road contribution to overall income changes observed was attempted, based on comparative income gains for households at varying distance from Andapa and the road.

8. With changes in level al.d distribution of income established, the study next examined how these are reflected in changing patterns of con- sumption expenditures, A test was made of the conformity of the household survey results to Engel's Law which holds that the proportion of total income spent on (sample) foods decreases as income increases. Area con- sumption patterns for the lowest-to-highest income groups were analyzed both statistically and based on the formula for estimating the income elasticity for food consumption and non-food items developed by Houthakker (19j7), Separately, comparative analysis of area housing standards at the time of the two surveys was undertaken to further illustrate changes in consumption of high volume/low value goods, and of modern consumer durables. iv-

9. Open-ended questions wete used io eliciL area re.,delLts' OpinionS of the road. In the same way, :espondents' individual aspiva3tions for themselves and their families, and their goals for developing different community services were establishied in an effort to gauge whethier their priorities and those of development planners were reasonably concordant.

Findings

10. The results of the economic analysis showed the Andapa-Sambava road and agricultural project package to have been a marginal investment, if measured in terms of road user cost savings. With a long construction period of seven years and abnormally high construction costs due to dif- ficult terrain and severe cyclone damages during construction, the project economic rate of return was a low 4.5%, if user transport cost savinigs and transporter surplus alone are considered. However, Madagascar has, over a number of years, been a net importer of rice which in 1974-5 was bought in the world market at prices exceeding local (subsidized) retail prices by FMG 28/kg. To reflect the true contribution of Andapa rice exports to the country's economy, the project was credited withi import subsidy savings of FMG 28,000/ton of rice marketed. This adjustment increased the rate of return to 6.3%.

II. Alternatively, measuring the project's 'impact at the level of the farm' using producer surplus in the without and with project situation to determine project impact considerably higher economic returns were derived amounting to 10.1% exclusive of rice subsidy savings, and to 11.1% including them. The difference reflects considerable gains in producei surplus on self-consumed crops, an important component of economic well-being in sub- sistence oriented economies and one that the road user savings analysis - focussed on the transport cost savings on traded goods fails to take into account. If self-consumption of rice is valued at retail, rather than producer price level, slightly higher returns of 11.4% and 12.3% result due to a steeper progression in retail than in producer price levels.

12. Two hypothetical project scenarios were also analysed to test (a) the impact of timely project implementation (i.e. "normal" construction period of 3 years) and (b) good phasing of project components (3 year construction period of the road; road opening coinciding with production of sizeable volume of rice (6,660 t) for export). Rates of return for the first scenario rise to 8.1% (RUS) and 15.9% (PS) respectively. Significant improvement, however, occurs only with optimal project phasing, when project returns increase to a healthy 10.1% (RUS) and 20.6% (PS). This is the situation that the ex-ante project appraisal estimate would perhaps be most likely to have reflected, and the difference between it and 'actual' project returns could be said to measure the not untypical margin between expectations and realization, between model assumptions and real life constraints. 13. Analysis of household income Ratterns with and without the project showed that almost all area households experienced significant income gains. Because of the dcifferentmethods used to establish consumption of own-produced goods in the base and follow-up surveys, average disposable cash income (- ADCI = total income net of value of self-consumption and production expenditures) was used to measure income gains. For the sample as a whole, these averaged 35% (in real terms) on a per household basis. Income gains differed for the different occupational groups: average 1975 disposable cash income as a proportion of 1965 average (constant 1975 values) was 249% for farmers, 148% for craftsmen, 127% for employees and 102% for traders. Differential between lowest and highest income groups narrowed substantially: in 1965, traders as a grcap disposed of an income 10 times higher than farmers; in 1975, it was only 4 times higher. Comparison of income by decile puts the finding in sharper focus: in 1965, the top (10th) decile received 52% of total disposa- ble cash income reported by sample households, while by 1975 this proportion had declined to 42%. ADCI per household in the top decile was 54 times that in the bottom decile in 1965; the ratio narrowed to 32 times in 1975. Part of the per household income differentials between deciles is explained by dif- ferent size of household -- this averages 2.6 persons in the lowest, 5.7 in the highest decile in 1965; 3.4 and 8.3 persons respectively in 1975. Per capita ADCI distribution therefore is markedly more equitable than that per household: in 1965 per capita income in the top decile is 25 times that of the lowest one; in 1975 the difference is halved to just under 13 times. Pet capita differentials between low income deciles are very narrow: in 1965, double the per capita income of the lowest decile is reached in the 6th decile only; quadruple in the eighth. In 1975, double the first decile income is achieved in the fourth decile (a steeper progression) but the quadruple still is reached in the eighth decile only. ln both samples, income from salaried employment begins to account for a significant portion of total per capita ADCI in this eighth decile.

14. The analysis of total sample household incomes by distance groups with respect to the new road showed no clear cut pattern of distance and differential income gains - the different representation of "high-income groups" (traders, employees) in these groups obscured the comparison. In both periods, however, per household ADCI was highest in Andapa (distance group 1), followed by distance group 4 (16-20 km from road); lowest were households - incomes in distance group 5 (21 and more kms from road). A more fine-tuned analysis of only those farm households included in both surveys (panel) served to clarify the impact of the road on income gains. While ADCI for all panel farmers more than doubled on average, those residing within 5 kms of Andapa and the road gained almost 50 percentage points more than less centrally located farmers.

15. Overall, the analysis thus demonstrates clear equity gains in the Andapa area over the 1965-1975 period, brought about mainly by a sharp rise in income from sale of agricultural products. Significantly higher producer prices for all of the products grown in the region, and better access to out- side markets have raised income earning potential. Prevailing patterns of land ownership and tenure have made it possible for the majority of area farmers -vi

to share ill exploiting the new opportunities. With few exceptions, families farm their oupi land. Commurity (fokonolona) land is still available, for free allocatlon to new households. Farm size is determined mainly by the number of economically active members in the household, and the degree of personal efforl and care tiey apply to the cultivation of their crops is the most i-ilportantdeterminant of the level of returns they can achieve. Wtiile agri- cultural production technology is still traditional, significant yield increases can be had from careful cultivation practices. Increasing scope for doutble-cropping with the agriculture project being implemented will meatn more clhangein teclhniology- attention must be paid to provide labor- enhancing small equipmaent appropriate to family farm operation.

16, Analysis of "without/before" and "with/after" project consumption exlj*nditores by area households in each of the two survey periods showed that thuy conform quite clearly to Engel's Law. Share of food expenditures in total expenditure for 1965 is 71% for farmers, 55% for craftsmen, 43% for employees anid 317/ for traders. The parallel values in 1975 are 79%, 71%, 54% and 41% vocpe.tively, In both periods, the share of food expenditure thus declines romnlowcst to highest income groups. For 1975, Houthakker's formula gives iood iucome elasticities that are all positive and less than unity, while uon-4ood elasticities (except for farmers) exceed unity. Similarly, food ciaskictiJes decline from farm to non-farm occupations while the non-food elasticities increase. 1965 data, while not as clear-cut, broadly follow Lhe same pattern. However, in the inter-temporal comparison share of food expend4ture for all occupational groups i higher in 1975 than in 1965, uhen average income levels were lower. This is probably explained by the fact thiat for many area residents imported foods are not an inferior good, ie. demanad for them does not fall with increasing income. Also, there have been differential price movements for different expenditure categories during the study period. The price of rice, the staple accounting for 50-57% of total food expenditures, more than doubled. Food prices generally have risen by about 90%, prices for other goods by 75% on average. For the latter, prices in Andapa were high in 1965 compared to the rest of the country due to high air transport costs, Savings from road transport may have counter- balanced inflationary price increases on imports, resulting in the observed higher share of food expenditures in total expenditures for 1975 compared to 1965.

11. Decile-by-decile comparison of per capita expenditure show a re- tiiarkablyclose range for the value of self-consumption in the first 8 deciles. No one is really witolout basic staples. People in the top decile have food expenditures of a different order of magnitude from the rest because they buy different types of food, Per capita cash consumption expenditures between 1965-1975 increase much more for the lower 7 deciles, where more than tlhree quarters of total expenditure is for food. Expenditure patterns for transport and travel change markedly. In 1965, only the top two deciles used transport services to any appreciable extent: air travel was a luxury available to few. In 1975, all deciles report expenditure for some travel: the road has extended mobilitv to all income groups. vii -

18. Changes in consumptionpatterns are further reflected in the type and quality of shelter enjoyed by area residents at the time of the two surveys. Constructionstandards for houses have changed axLddifferent materials for floors, walls and particularlyroofs are being used: three quarters and one quarter respectivelyof sample houses had thaLtchand metal roofs in 1965: in 1975 this proportion is reversed. The number of rooms per dwelling has increased (2.4 rooms/house in Andapa-town). Stone houses are being built by the wealthy and rank highest on the list of personal goals mentioned by area residents. Furniture and furnishingsare more modern and owned by a larger proportion of respondents. Lighting used is better and safer than ten years back. Piped water is available in the town. But sanitary facilities remain basic and a continued potential source of disease.

19, People in the study area are articulate about their personal goals and about developmentpriorities for their community. More than half want to purchase a better house, more than a third want to purchase land, just under a third want to buy agriculturalequipment or livestock. Concerning community priorities, the majority (79%) mention the need to improve roads in the Basin, two-thirdswant better medical services, 53% think schools are most needed. There are clear differences of priorities in Andapa-townwhere quite a few services already exist, in the Basin itself, and in the villages along the new road. Aspirations of area residents are for a share of the better life, which they have seen people enjoy elsewhere.

Some Observationson the Data Base

20. The present study is one of a very small number of longitudinal investigationsinto the socio-economicimpacts of a rural development project. The testing of alternativemethods of rural road evaluation and the tracing of peasant farmers' response to the 'stimuli-to-change'presented by the project have been the main objectives of the study. But a number of points that have emerged in the process of establishingthe data base itself may be worth retainingas well. The first observationto make is that the time "series" of up to five surveys of varying intensity planned when the baseline survey was first initiated, ended up as a series of two (before/afterproject implementation). Had there not been a fortuitous coincidenceof interest, on the part of the Malagasy authorities and the World Bank, to better under- stand peasant-farmerresponse when improved access to markets and to social services is provided, it may well have remained a series of one. This underlines the importance of active involvementof local institutions(with a "memory") and of local investigators(with a vital interest in the solution of local problems), in longitudinalstudies of the difficult phenomena of economic developmentand social change in rural arees of developing countries. The more a monitoring exercise for a particular project - or series of projects - can be made a part of ongoing local data gathering activities (which it may help to strengthen),the better are the chances that planned follow-up investigationsare actually carried out in a timely and com- petent manner. - viii -

21. Regarding data collection procedures, a few points stand out. The first concerns the vital importance,for any longitudinalstudy, of thorough documentationof data collection and data treatment procedures. When ten years after the base survey, the possibility of undertaking a follow-up study was first explored, none of the personnel involved in the data collection for the original survey were any longer available. But an unusually complete effort of project documentationnevertheless enabled the follow-up project to get underway with minimum delay.

22. Still, in this as in probably most longitudinalstudies in the real world, there have been differences in the data collected for the base and follow-up surveys. Time and money stricturesmade it impossible to replicate the original study design exactly, especially since the total study period in 1975 covered only five months, compared to a full twelve month cycle in 1965, During the base survey, detailed measures (weighings)of household consumption of food products were establishedby interviewersvisiting households several times daily and functioning as "participantobservers" for a sample week throughout the full year. In 1975, information on food and other consumptionquantities and expenditureshad to be elicited via respondent recall, in order to complete a requisite number of interviews in the time available for fieldwork. The informationthus gathered was suf- ficiently exact for the purposes of establishingcomparative consumption patterns but not for establishingdetailed nutrition profiles such as could have been developed from the base survey data. Even so, to minimize recall bias, a survey period covering a full year cycle would have been desirable, especially for rural household surveys.

23e Least satisfactory,for purposes of the analysis, was the compara- tive data base on volume and composition of freight transportedfrom the study area to the points of trans-shipmentand export in Sambava and in 1975. While the 1964/65 survey was able to establish a fairly complete picture of area exports and imports by monitoring the only channel of egress/ ingress: AIRMADAGASCAR,no comparablesource of informationwas available in 1975. In the complete absence of any traffic counts for the road since 1972, the best that could be done was a fourteen day origin - destination survey along the road during the period September- November of 1975, Such surveys should, ideally, have been carried out, at least two more times: once after the main rice harvest when peak traffic to evacuate paddy surpluses from the study area occurs (May-June);and once during the rainy season, when only essential traffic utilizes the road (December-March). In the absence of such counts, "informed estimates" (based on interviews with transport operators and on estimates of area production and exports) of "average daily traffic" had to suffice as a basis for the road user savings analysis when, ideally, traffic counts should have served as a cross-check on quantities reported as sold and exported by area farmers and traders.

24. Finally, in Andapa as with other rural populations,the survey teams have found respondentsalmost invariablymost cooperativein answering long series of questions. These often must have taxed their ability to recall actions and transactionsreaching back over periods - ix -

of a month, a year and longer. As a result, answers were considered carefully and given openly and freely. Despite this, it was found that quantitative data on some key parametersof the farm operation such as extent of area under cultivation,volume of crop production,or proportion of crops self- consumed, are indicativeof orders of magnitude rather than exact quantities or measures, established precisely to the last kilogram or square foot. In interpretinginterview results, it is important that the analysis develop appropriate aggregate categorieswhich allow meaningful distinction of relevant differentials.

25. Reliable objectivemeasures will rarely, if ever, be available for remote rural regions such as the study area. It is worth stressing,however, that not only do respondents'estimates often provide the only feasible source of information- they also provide valid informationbecause they describe a reality that is meaningful to them. The expert outsider may indeed correctly measure, say, a transport cost reduction of FMG 5/Kg. If area farmers report by contrast that transport costs heve not changed and are too high, this indicates not so much t1ia:farmers are not aware of the reduction but that, for them, it is not significantenough to change establishedpatterns of not transportinga product to market. The expert's calculationmay be 'exact' - but the farmer's response,because it reflects his threshold of risk aversion, may be the better predictor of changes in area patterns of production or exports.

Possible Extensions and Future Work

26. Considerablymore informationhas been collected than was presented in this report. Comparative analysis of new migrant and newly formed house- holds would be possible with available data; there is a wealth of demographic informationon the panel households included in both surveys; income and expenditure reports lend themselvesto more detailed further analysis, and 1965 data allow a thorough study of area nutrition patterns. On the other hand, the analyses carried out would benefit from more objectivelybased measures of area cultivatedto different crops, and of yields realized on land of different quality; on freight volumes exported from the region during different parts of the cropping cycle; on productionand trade in the communi- ties along the road as distinct from the Basin; and on production income and expenditure patterns in a control region such as , where transport patterns still are similar to those that existed in Andapa before the road opened. If additional investmentsin the region in connectionwith the agriculturalproject are contemplated,the opportunityto improve and extend the data base required for an impact analysis might present itself and should be exploitedwith a view to contributingto improved project design.

27. While focussed especially on the impact of improved access, the present study emphasizes a regional analysis framework that tries to combine measures of economic gains and indicatorsof the degree of equity with which they are shared among the local population. It is thereforehoped that it may be of interest not only to those engaged in the field of road construction and transport,but to a wider audience concernedwith various aspects of rural developmentplanning in traditionalpeasant economies.

V4~~~~~~~~~~~~~~~~~~ IBRD) 10331R MADAGASCAR POPULATIONDISTRIBUTION - 1971

Each dot represents2,000 inhabitants Hell -Vilnogj O Over60,000 Inhabitonts u.] 20.000 - 60,000 Inhabitants \ ri 10,000 - 20000 Inhobitants

Anlolaha

o M~~~~~~~ ~ ~ ~~~~~~~~~AJUNGA(- /

g e g o r o y ~ ~v. ~ ~./

I ; X z U,AmRnotn,(oznl

- 1 T1AMATAVE

\ t W~~~~~~.ANANA.91VE/ Q4 'Ts:roarwvim.d-y I<; IVEfl- Moromongo

- ) Ambato;ompy *.

' ,Antseraoe

N oMrondova - t qAmbos,tra

1 y ' e/~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~Mononlory yeV1-FI,NARANTSOA o

Manokora

W.Forafongono

_ ~~~~~~~ ~ ~~TULERe \

UPNII0 0 IO20k

bUAD GSCARf_ .~ ~ ~~~~~~~~~~~~~~~~~~~~~~~~~~5... .km 20 B,n II /44

MADAGASCAR TRANSPORTSYSTEM

FOURTHHIGHWAY PROJECT NO):SI + ,2

PREVIOUS IHIGAVY PROJECTS VII LAGI. I IVfLX 'AKPl') 11'O I V..

",,,14-.A MVI,, 00A( W' VII ,4 4. '4 ,4..' .. 411...... 1FASIBILITY B7Y If< IIDGC WlTAILESOPlANIFEDf4ING:E r ((11411', 5CNAIIIYVN.rR I, I

EXISTING ROADS fOURTH flIGHWAY PSI0JECT P'IL,L IV ,A,PI V'i.`11,'10,AI . ;

T - t^~~~~~IVID PAVI 1) R(MArlCC*4STfIUt.-i.0 W'_S ; * ---- Al IL Il44IY4lArR (w4- 544i44' R ,RAv', 4(40ADC(.03TR1C4.tlhr'l, 44014 ~'4 fl444,NtNT~~ ~ ~ ~ ~~~14

! J> AIR7t'.IROAItlC BOI7^a, +

IOVr RI PRE1FCT4LIFf UOUDA1' ;

4~ ~ ~ ~ 44~~~'~'7°() '43 '

I '8'V ''' ' ' ,,fw ,,*7?n7.,, .' W,'* '\ ; " ' 7ooy

4 V4 * ,' F-.~~~~~~~N- b

i -; ]7 / ) K - + o.

77'~~~~~~~~½ 444"0 'g' *44' 20- /3' ,,. Ch c a ; J ' _4 '''''''J]

F71.~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ i4 .V

120 lZ-r4.-u; ,; .;.A."', ...... r::-. l h; rWlt

VV

. , 4 -,A

Tuleor~~~~~~~~~~~~~~~~~~~Fe 1),;r, |~~~~~~~~~~Tlo '- - .- --. ; ! 4'~~~~~~~~~~~~~~~~~~~~~~~^tMtA, 44.

4?' J14,~~'i 4 6' ' ! I

[2.4.~ A I< ,120 4 12

LOCAION OF ANDAPA rg SITUATIOND'ANDAPA

y ~~~~~~~~tfhokh

1&~or~orive5 os

60

20"~~ -i 200a

'z.

CA-iuda. Pra;aatara~

Chgf.Iiou d. Soco-Pm.FsooLa

Limitia Province 2

i mi e oV Z

24" ECHELLE/10 ) 00000

440 480 I N t t $~~~~~AMBAVA

-o U

DMoany sX FOvhalana

MWaroambihy

Ambalamanasy

Ambodiangezoka ranomena - \G 0

Andrakataa C,

- Bea'npo A 5DAPA%

io -'Lake/Lac Ampthana

19 s

, _ l ~~~~~~~~~~Legende,Legnd

C-) Route Antp1-S8mbavat GO ~~~~~And~ep-SomnbEmaRoad Ro t dA ap A dpa od co RouQt Existante/ l c SguGon I Lo Map o ~~~~~~ExicqingRced Pa 3StainfLcto a - v/Piste pour lPoQns/ FootpaQh

ECHELLE XSCALE 11500 O00O CRAPTER I

I N T R O D U C T I O N

.~~ . 1 .. . I I . . I ...... , . I ...... I. INTRODUCTION

A. THE STUDY AREA

1.1 The Andapa Basin is situated in the north-east of the Democratic Republic of Madagascar (49028' - 49 39' east longitude, 14031' - 14°43 south latitude). The basin itself lies about 500 m above sea level. It is enclosed on all sides by high mountain ranges, varying in altitude between 800 m in the south to 2,000 m in the north and west. Administratively the study area is part of the district (prefecture) of ANTALAHA in DIEGO SUAREZ province. The town of ANDAPA is a sub-district cen:er that provides administrative services to seven cantons, five of which are situated in the Basin proper.

1.2 The Andapa Basin extends over some 76,000 ha 1/ and sustains a population of between 55,000 - 60,000. Spontaneous settlement of the region began late in the 19th century with immigrants arriving both from the eastern and the western coastal areas. The area population is unusually diverse in ethnical background, although a majority are Tsimihetys.

1.3 The study region is rich in fertile bottom lands for the production of Madagascar's main subsistence crop - rice. It also provides excellent con- ditions for growing coffee and vanilla, two of the country's most important export crops. Before the road to the port of Sambava was opened in 1970, the only way for products and people to enter or leave the area was by porterage (over 100-120 kms) or, more importantly since the nineteen-fifties, by air- transport (DC-3).

1.4 The economic importance of the Andapa region can be gauged by the fact that in 1960 its air passenger and freight movements were the second- highest in the country after Ivato airport serving the capital city of Antananarivo. Ivato handled a total of 8,600 tons, two-thirds of which was passenger traffic (10 passengers = I ton); while Andapa handled 4,300 tons, less than 20% of which was passenger traffic. 2/ Over 95% of the tonnage exported from Andapa went to Antalaha (56%) and Sambava (42%). 3/ Antalaha received almost all vanilla, the highest-value export crop; Sambava handled most of the coffee crop.

1.5 The air link however, was not only unreliable (with no flights pos- sible at all during the rainy season (Dec.-March) and frequent cancellations during the rest of the year due to dangerous cloud-cover in the approach zones), but also very expensive. An all-weather road connection of the Basin

1/ From map issued by "L'Institut Cartographique de Madagascar," 1:50,000.

2/ Bosch, J., Drachoussoff, V., Nemo, J, - Poirier, J: LA CUVETTE D'ANDAPA - Inventaire Agronomigue - Demographigue - Ethnologiciue et Socio- Economigue, Brussels, 1964, pp. 341, plus annexes, pp. 312.

3/ AGRER: MISE EN VALEUR DE LA CUVETTE D'ANDAPA - 3 Vol. (Rapports Pedologique, socio-economique et agricole), Andapa, May '67, p. 69. -2- to the coast was studied and found to promise great potential benefits to the local economy. Freight transport changes between Sambava and Andapa were exrected to be reduced by a factor of ten; passenger transport charges by a factor of four. This, in turn, could mean a potential increase in the price of traditionalcash crops to Andapa producers, as well as a first opportunity for the region to produce rice for export to neighbouringareas at a price competitivewith foreign imports.

B. THE ROAD PROJECT

1.6 In the early 1960s, the Malagasy Government approached the European Development Fund (FED) for financing of a road linking Andapa to the eastern coast road and the port of Sambava. The agreement to finance the road was signed in 1962, but due to the need for considerablerevisions of the road design, actual constructionstarted only in January 1964. The Italian firm MURRI Freres was in charge of the project. Extremely difficult terrain and the occurrence of three violent cyclones (1965, 1966, 1968) during the constructionperiod resulted in a total constructionperiod cf seven years: the road was opened to traffic in 1970.

1.7 The Andapa-Sambavaroad is 107 Km long. It is a typical mountain road which necessitatedconstruction of a large number of bridges and massive retainingwalls. Its bituminous pavement is 5.60 m wide, with shoulders (unpaved)of 1 m each. 1/ Including repair of the cyclone damage, the road cost a total of FMG 3,904 million to construct (approximatelyUS$15.6 million equivalent, 2/ or US$146,000 per kilometer). Sixty three percent of project costs were covered by the FED credit, the remaining 37 percent by the Malagasy Government.

1.8 An annual maintenance requirementof FMG 110,000/Kmwas estimated for the road in 1970. Actually, less than half this sum is currently avail- able for routine maintenance. A special allocation of FMG 12 mill4on was made in 1974 to repair major cyclone damage incurred in two locations along the road in 1973. The contractor in charge of the work went bankrupt after a third of the funds were spent; thus the repairs remained unfinished at the end of 1975, although a rough detour enabled cars to get through. A severe cyclone in February 1976 cut the road in about 15 different locations, besides laying waste about 95% of the houses in Sambava and 90% of those in Andapa.

C THE ROAD IMPACT STUDY

1.9 When the Andapa-Sambava road project was prepared for FED financing, it was realized that it presented a rare opportunity to monitor the impact of a completely new outlet road on an isolated economy of mixed subsistence and

1/ For detailed quantities of work and costs, see Table 2, Appendix D.

2/ Conversion factor used US$1 = FMG 250, as no year by year breakdown of constructioncost over the 1964-1970period available. - 3 -

a completely new outlet road on an isolared cconomv of mixed subsistence and cash crop farming. The area was compact and clearly defined - separated on all sides from the rest of the country by mountain ranges that were almost completely unpopulated. Current transport means and costs could be effectively monitored. The road was the first aind,at the tior, only infrastructure investment in the region.

1.10 In 1963, both FED and the French "Fonds d'Aide et de Cooperation" (FAC) agreed to finance the first two of a proposed series of socio-economic studies of the Andapa Basin. FED asked a Belgian consulting group, AGRER, to undertake a preliminary study of the cultural and socio-economic structure of the region. Based on secondary sources and selective interviews, this study assembled information on the region's physical characteristics; demo- graphy; infrastructure endowment; agricultural potential and current cropping patterns. It also described ethno-sociological patterns of family and village organization; land tenure patterns; technologies employed in agriculture, food preparation; and explored traditional value systems and receptivity to change. Finally, it established the basic structure of the local economy: imports into the region and cash crop production for export; local crafts and trade; financial exchanges between the region and the rest of the country; and fiscal arrangements and community budgets. A full census of all area households was also carried out as part of AGRER's surveys. Study results were presented in a detailed report 1/ in 1964, prepared jointly by the Association Inter- nationale de Science Economique Appliquee (ISEA) and AGRER (AGRER 1964).

1.11 With the Planning Commissariat of the Malagasy Republic providing coordination between the two studies, the Bureau Central d'Ecudes pour les Equipements d'Outre-Mer (BCEOM) was charged by FAC in 1964 to carry out a detailed statistical survey providing reliable quantified measures on the main aspects of the local economy. A careful survey design based on the unusually complete sampling frame provided by the AGRER census, was devel- oped 2/ to derive a representative sample of area households.

The main parameters treated in the BCEOM investigation were -

- household income (cash/kind) and their distribution

- household expenditures (cash/kind) and their dis- tribution

- income elasticity of consumption

- area under cultivation; crops/yields

- price of land

1/ Bosch, J., Drachoussoff, V., Nemo, J. - Poirier, J: LA CUVETTE D'ANDAPA - Inventaire Agronomigue - Demographigue - Ethnologigue et Socio- Economique,

2/ See detailed description, para. 2e7 ff. below. 4-

retail and producer prices of goods

physical and financia1 exchanges of the region witlh the rest of the country/wotld.

1.12 This survey, tabular results of whiclh were published in 7 volumes in 1966 1/ (BCEOM 1966) was conceived as a first phase (data collection) to be followed by a second (analytical) phase for establishing a regional ac- counts model. This would show the impact on the regional economy of changes in some key variables whichl would be measured in several re-surveys following the opening of the road. A f(nal analytical effort would then try to develop a methodology wlhichcould distinguish development impacts:

- that would have occurred without any project

- due to the road

- due to other planned infrastructure investments, especially irrigation/drainage works

Such a methodology would also take into account the influence of externali- ties suchI as weather (and its effect on crop yields), changes in world-market prices (and their effect on cropping patterns); and regional institutional parameters influencing marketability and profitability of different crops. None of the proposed follow-uD work under the auspices of FAC was carried out. FED however did finance a series of additional investigations in the region as part of feasibility studies for proposed agricultural and rural industries projects. The feasibility study for the agriculture project, carried out again by ACGER, summarized a considerable amounL of BCEOM find- ings, and collected a great deal of detailed information on the pedological, agro-economic, and ethno-social characteristics of three subareas for which irrigation/ drainage works and agricultural extension were being considered. Results were publislhed in 1967 (ACRER 1967). 2/ Additional reports on the potential for establislhing rural industries in the region were issued 1970/71.

1.13 In 1972, irrplementationof a FED financed agricultural project (OPACA) began. Total estimated project expenditure was US$5,401,000 equiva- lent; the major project components were irrigation/drainage improvements of 2,500 ha of rice-land, improvement to all-weather standards of 80-100 Km of roads in the Basin, and provision of extension services for coffee cultivation, vegetable productior., and animal husbandry. By 1975, the extension component

1/ de Buffevent, G., Francois P., Guy, M., Jullion, M., Morizot, M., Nemo, J., Parent, M.,: ETUDE ECOSOMIQUEDE LA CUVETTE D'ANDAPA, BCEOM, Paris, December 1966 - 7 volumes, pp. 1001.

2/ AGRER: MISE EN VALEUR DE LA CUVETTE D'ANDAPA - 3 Vol. (Rapports Pedologique Socin-Econom,iqueet Agricole), Andapa, May 1967. -5- of the nroject had been implemented effectively. But due to difficulties with contractors, only part of the irrigation/drainage works and none of the road improvements had been carried out.

1.14 In 1974, the availability of a unique data base on the Andapa Basin had come to the attention of the Transport Research Division of the World Bank who, at the time, had initiated a modest program of research and method- ology development aimed at improving the measurement and prediction of the likely socio-economic impacts of rural roads investments. When further investigation confirmed that only limited additional investments (the effect of which might have significantly confounded those of the road) had actually been carried out in the area, financing for a "follow-up survey" was sought and granted by the Bank's Research Committee. Authorization of the Malagasy Government was obtained to undertake the field work in Andapa in the second half of 1975.

1.15 Available funds precluded the engagement of full-time consultants; thus survey work was carried out by an interdisciplinary team of graduates from the Swiss Federal Institute of Technology, Zurich together with a counterpart of the Malagasy Planning Bureau in Antananarivo, who served as field team leader. Twelve local collaborators carried out the bulk of the interviewing, as well as the coding of the information gathered in the various surveys. Regional authorities in the study area cooperated by making their records available and facilitating the study team's operation in every way.

1.16 The present report summarizes the results of the base-and follow-up surveys and of the comparative analysis of the "before/after-without/with" road situation in the Andapa Basin. Chapter II deals with method and de- scribes survey design, data collection and analytical framework of the study. Chapter III presents the economic evaluation of the road project. Following a description of regional structures (social/economic) and of the economic exchanges between the region and the rest of the world as observed in 1965 and 1975, rate of return estimates for the road investment are derived applying both the road user savings and the producer surplus approach. Chapter IV analyses levels and distribution of income at the base-and follow-up survey time in an effort to gauge the incidence of project benefits and to assess whether they have contributed to a more or less equitable distribution of income in the region. In addition, consumption patterns are examined with a view to identifying and measuring changes in level of living to which the road mnay have contributed. Chapter V finally examines perceptions of the impact of the road project, by area residents and presents their opinion on future development priorities in the study region. CHAPTER II

M E T H O D SECTION II A

GENERAL STUDY DESIGN -6-

A. GENERAL STUDY DESIGN

2.1 In designing the follow-up survey the essential question was how to draw valid measures of road impact from two sets of observations on the impacted or target population:one collected five years before, and the other five years after the opening of a new all-wecther road connection. Evidently, not all of the ambitious goals of the initial research plan elaborated by BCEOM (see para. 1.11 above) in 1964 could be realized. Nor could one-follow-up survey be expected to yield the fine-tuned "regional accounts" model that might have been developed and validated with a time- series of 3-5 follow-up surveys of different intensity, as originally proposed.

2.2 Detailed documentation on survey instruments used, on survey design and on field procedures was available for both the AGRER and BCEOM base-studies. The latter also provided an elaborate coding framework for tabulating some household characteristics and especially household income/ expenditure information in great detail. Particular care was taken to account for types and quantities of food self-produced and self-consumed-- items that are often under-reported in the economic analyses of subsis- tence-oriented groups and societies.

2.3 With respect to the household survey, the meticulous design of the 1965 BCEOM survey (see paras. 2.7 ff below) suggested a one-on-one follow-up approach (panel) for at least 50% of the old sample households if that many could actually be relocated. The remaining sample should represent as adequately as possible the following types of new households in the area:

(1) newly formed (since 1966) households in the area;

(2) newly migrated households from other parts of the Andapa Basin to the sample villages;

(3) newly immigrated households from the rest of the country to the sample villages.

With such a sample, the most powerful statistical tests of "change" could be applied to the "old" households which could be individually matched for both surveys, while new household subgroups in the area population would be adequately represented for group matching. For the total sample, results of the 1975 National Census (counts of which were completed in Andapa in April 1975) would allow the establishment of exact sample expansion factors (parallel to those established by BCEOM from the 1965 AGRER census) for grossing up to regional totals. 2.4 Regarding the monitoring of exports from/imports to the regioti, the 1965 air traffic survey needed to be replaced by road transporter/ traffic surveys to establlslh (a) current transport volume, (b) the extent of transport cost reductions and (c) the main beneficiaries from these reductions: transporters? traders? peasant producers? peasant con- sumers?

2.5 Special surveys were planned to elucidate changes in local insti- tutions, price levels, etc. not due to improved road access, as these also would have independently influenced area residents' cropping and marketing patterns.

2.6 It was planned to establish 1975 measures of areas under different crops from aerial survey data. Aerial photography of the future road- alignment as of 1963/4 was available, as were detailed photo-maps and large-scale (1:5,000) crop-maps of most of the Basin for 1965. Combined with the results from interview surveys in the road-influenced area and in the study region, aerial photo-interpretation could provide objective evidence of expansion of cultivation in response to improved access. How- ever, because of budgetary constraints, the aerial survey could not be carried out. SECTION II B

THE SAMPLE SURVEYS: 1965-1975 -8-

B. THE SAMPLESURVEYS: 1964/5 - 1975

1. The Baseline Survey: BCEOM. Jul.:1964 - June 1965

2.7 In the baseline survey, samples were drawn from a universe of:

(a) Households (stratifiedby - (i) for estimation of household food occupation of head of house- consumption hold) (ii) for estimationof household income! expenditure patterns

(iii) for estimation of ownership of animals and agriculturalequipment

(iv) for estimation of volume of trade

(b) Households (non- stratified) - (i) for estimation of area under cultivation

(ii) for estimation of crop yields

(c) Markets/shops - for monitoring of prices of selected goods

(d) Air Madagascar records - complete record of one year's freight and passenger movements from/to Andapa. -9.

House1iold Sample

2.8 The household sample was a stratified clustered random sample of 723 households from a total of 9,900 1/ in the Andapa region survey r'{-t-,, Stratificationwas by occupationalgroups as shown in Table 1.

Table 1: UNIVERSE AND SAMPLE OF ANDAPA AREA HOUSEHOLDS STRATIFIED BY OCCUPATIONALGROUPS - 1965

Number of H in Number of H Coefficient of OccupationalGroup Universe in Sample Expansion

1. Farmers: Owners or Free Landholders 4,973 259 19.19 2. Farmers: Share- croppers/Renters, etc. 3,482 182 19.13 3.a) Traders and their Employees 128 64 2e00 b) Traders and their Employees* 290 21 13.81 4. Craftsmen and their Employees 468 56 8.35 5. Employees: Central/ Prov. Govt. 122 63 1.94 6. Emplovees: Local Admin. 109 28 3.88 7. Employees: Technical Services 110 20 5.52 8. Employees: Private + Coops. 92 11 8.32 9. Others 126 19 6.63

TOTAL 9,900 723 13.69

Source: BCEOM - Table 2.4.4, p. 2.63. *NOTE: A number of traders were found among farmers - for these, a different expansion coefficientapplies.

This breakdown by professionalgroups is adjusted from the original AGRER classificationby applying the occupationalbreakdown actually found in the sample to the universe total. In order to ensure correct grossing up to regional totals, different expansion coefficientshad to be applied to e.g. traders properly classified (samplingfraction 1/2) and those classified as farmers (samplingfraction 1/20). la HRH= Households (or Heads of Households).

1/ The total number of households establishedby AGRER was 10,156 - 256 of these were excluded from the BCEOM universe for various reasons (a- typical household such as foreign administrators: 17; and all house- holds whose 'head' had no occupation and that consistedof less than 4 persons: 239). - 10 -

2.9 The sampling frame was provided by the detailed census of area households undertaken by AGRER in early 1964. Within each of the 91 villages in the study region, households were listed by number within each occupationalgroup. The sample was drawn with random numbers in clusters of 4 contiguoushouseholds on the list for Groups 1 and 2 (farmers);clusters of 2 for Groups 3 and 4 (traders and craftsmen); unclusteredfor Groups 5-9 (employeesand others). Adjustments were made to keep each cluster within one village. A total of 723 households in 75 villages were interviewed.

2.10 Representativenessof the sample with respect to size of village and size of householdwas tested and reported adequate. Replacements of refusals and of households not locatable for various reasons were taken from the next consecutive cluster in the household list; they are assumed not to have introduced bias. Data collectionwas spread over a twelve month period between July 1964-June 1965. (see Table 4, App. 4 for detail.)

2.11 Starting in November 1964, a battery of questions concerning ownership of animals and agriculturalequipment was added to the house- hold survey questionnaire. Data on these were collected from 477 of the total 723 sample households. Sampling fractions and expansion coefficientsfor this part of the survey are detailed in Table 5, App. A.

2.12 All traders in the household sample (85/723) were also asked about their business. Total annual turnover during the past four years; volume of purchases and sales during the past 12 months; and a detailed analysis of purchase/salerecords for the three months prior to the interviewwere established.

Sample of Areas under Different Crops

2.13 Applying a sampling fraction of 1/50, a separate systematic sample of 202 householdswaa drawn from a list of 10,139 households 1/ in the 91 villages of the study area. Again, clusters of 4 contiguous householdswere selected. For each of the households,area under cul- vation for the main subsistence crop - rice, and the major cash crops - coffee and vanilla, was established;the former by means of simple topographicmeasurements, the latter by counting coffee and vanilla plants in the sample plantations. Area measurementswere carried out from mid-November 1964 to mid-January 1965; a subsampleof the plots thus measured were used for a later survey of yields.

2.14 The rice-plantationssampled were listed in order of size and added down the list cumulatively. Choosing every 9th hectare on the list, a 51 field subsample was established (the probability of inclusion in the sample thus being proportionateto the size of the

1/ This list excludes 17 'a-typical'households counted by AGRER. - ll1 plantations). From the 51 sample fields, a total of 318 yield samples were collected on randomly established emplacementsin the field. All risps were cut, transported to Andapa in plastic bags, hand-kernelled and dried for a few days - then weighed. In estimating total area yields, adjustmentswere made to account for losses due to less meticu- lous traditionalcrop handling methods.

Market/PriceSurvey

2015 Retail prices for selected products were checked monthly at ten different marketiDg outlets. While chosen to be representativeof small/ large villages and local/Chinese/Indiantraders, no satisfactoryway of extrapolatingresults beyond establishingarithmetic means were found. Price data derived from the household expenditure survey were found to be more reliable and subsequentlyused for area expenditureestimates.

Survey of Air Transport Movements to/from Andapa

2.16 All bills-of-ladingconcerning goods passing through Andapa airport were summarized monthly for the twelve month survey period to determine volume and value of exports from, and imports to, the study region. The major exporters/importerswere also identifiedand their importance for the local economy established. Porterage of goods for export had ceased to be of any importance at the time of the survey.

2. The Post-Road-CompletionSurvey: IBRD, July-December1975

2,17 Ihe AGRER/BCEOM survey design and execution,offered a sound 'before' set of data from a probability sample to examine change, includ- ing road impact, in the 1965-1975period. It was decided to match the 1965 sample design as carefully as possible (within available budget and time constraints)in order to facilitate testing of hypotheses in this near-experimentalfield survey. Samples were again drawn of:

(a) Households- (stratifiedby (i) for estimation of household occupation of head of household) food consumption

(ii) for estimation of household income/expenditurepatterns

(iii) for estimation of ownership of animals and agricultural equipment

(iv) for estimationof volume of trade

(Note: no separate sample) (v) for estimation of area under cultivationand yields - 12 -

(vi) for estimation of travel behavior/ expenditures

(vii) for estimating migra- tion patterns

(viii) for establishingpersonal/ communal attitudes about developmentpriorities

(b) Markets/Shops - for monitoring of (retail) prices of selected goods

(c) Transporters - to establish costs, tariffs and effectiveness,of road transport services

(d) Road Traffic - 14 day origin/destination surveys to establish patterns of passenger and freight movements along the road.

2.18 An aerial survey along the road (15 Kms each side) and of the Basin was to be undertaken to update measures of area under cultivation and establish current cropping patterns. Unfortunately,this survey had to be dropped because of lack of funds, an unexpectedlyhigh proportion of which had to be devoted to reconstitutethe base survey data tape from coded information.

Household Sample

2.19 The sampling plan for the 1975 household survey proposed the fol- lowing distribution:

Planned No. of interviews Andapa Basin 720

"Old" (1965 sample) households 360 (of 723)

"New" Households established in area 1966-70 180 established in area 1971-75 180

Along Andapa-SambavaRoad 140

Non-road-influenced"control"-households 140

TOTAL 1.000 - 13-

2.20 "Old" householdswere to be chosen suiccessivelyfrom among the following categories-

k.d) same l-eadof household,same place of residonc-

(b) same head of household,other residence in study area

(c) same head c' household, same residence- but head of household temporarilyabsent:spouse to be interviewed.

2.21 Interviewswith control groups in non-road-influencedareas outside the Basin were considered. But there had been no control group in 1965, and in 1975 the control sample had to be dropped because of time and money con- straints. Instead, it is attempted to approximate 'controls' in the analy- tical treatmentby examining differentialsfor sample households located at different distances (say less or more than 15 Km) from Andapa and the road.

2.22 Field work started with a search for the households surveyed in 1965. While instructionswere to find as many as possible, but to randomly drop any over 50% of the original 723 from the re-survey,all 472 households actually found were interviewed. This curtailed somewhat the size of the "new household" sample that could be accommodated.

2.23 For those heads of household not found again (251), 299 reasons were given why they were missing:

Reason % of Reasons

Died 60 20.1

Left permanently (mostly: assignment to new post, marriage, etc.) 64 21.4

Temporarily absent (travel) 23 7.7

Impossibleto interview (insane, prison, etc.) 13 4.4

Interviewererror 4 1.3

Refusal 1 0.3

Unknown in village, N.A., Don't Know 134 44.8

TOTAL 299 100.0

With respect to the total base survey sample, we thus know that in the ten-year interval between 1965 and 1975, some 65% remained,at least 9% of the heads of household permanently left the area, 8% died and 3% were temporarilyabsent. - 14 -

2024 To establish the sample of "new households", a random sample of 11 out of the 91 area villages was drawn. Andapa, the central place in the study area (with about 8.5% of all households) was also included. A complete mini-census of these twelve locationswas then conducted. All households were classified by occupation of head of household and as to whether they were:

old households, i.e. established in village before 1965

- newly formed households (from village) since 1966

- newly arrived households in the village (since 1966) from other parts of the Andapa Basin

- newly arrived households in the village (since 1966) from other parts of Madagascar.

2,25 Out of a total of 2,400 households thus classified, 1,480 (60%) had been in the locality before 1965, 466 (19%) were newly formed house- holds; 244 (9%) had come to the village from other parts of the study area, and 270 (11%) had migrated to the region from other parts of the country. (A third of the latter were government employees).

2.26 The "new household" sample was drawn randomly from the 12-village census lists. Samples of 52, 48 and 50 households respectivelywere inter- viewed from the three types of new households listed above.

2.27 The results of the 12-village census combined with the preliminary results of the complete National Census served to derive preliminaryexpan- sion factors for the 1975 survey. Details are given in Table 6 of Appen- dix A. (Final census results are not yet available to provide exact sampl- ing fractions - they should not vary much.)

2.28 A further additional sample of "new traders" was also selected to replace a sizeable loss (43 of 85) of traders from the 1965 survey who could not be re-located. Fifty-two "new traders" were randomly sampled from a list of heads of householdwho had acquired a business license from the licensing office between 1970 and 1975.

2.29 To complete coverage of the area of influence of the road, a final subsample of 100 households was selected from 3 villages directly adjacent to the new road from Andapa to Sambava. Following a rapid popu- lation count of all villages along the road (see Table 9, App. A), these were divided into three groups (near Sambava, mid-way, near Andapa) of somewhat differing agriculturalpotential due to varying topographical conditions. One village in each group was randomly chosen for survey, with every third household in the village being included starting with the first household on the road coming from Andapa and stopping when the requisite number of households had been interviewed. The final sample thus consisted of: - 15 -

1975 Household Sample Composition 1/

ANDAPA Basin 673

1965-Sample"Old" Households Re-surveyed 472

Newly-formed local households (1966-1975) 51

Households of immigrants from other parts of the Andapa Basin (after 1965) 48

Households of immigrants from other parts of Madagascar (after 1965) 50

"New" traders 52

Along ANDAPA-SAMBAVARoad

Belaoko-Lokoho - near Andapa 20 (= treated as "Basin" households in some analyses)

Manakana - mid-way 40

Ambodivohitra - near Sambava 40

Total 773

2.30 For the comparativeanalysis of base and follow-up survey results, the originally distinguishednine "strata" were collapsed into four broad categories: farmers, craftsmen, traders and salaried employees. Compara- tive sample distributionsin 1965 and 1975 and estimated coefficientsof expansion for each occupationalcategory and for the sample total are shown below.

1/ For comparative distributionof interviews by location and time of 1965 and 1975 surveys, see Tables 1-2 and 3-4, App. A. - 16 -

Table 2

Universe and SaDpR,e of Anda2a Area =louseholds Stratified by Occupational Groups, 1965-1975

Coefficients of N HOUSEHOLDS Universe Sample Expansion OCCUPAT'L GROUP 1965 /1 1975 12 1965 /1 1975 1965 1975

Farmers 8,581 10,800 460 460 18.65 23.48

Craftsmen 468 470 56 49 8.36 9.59

Traders 418 610 85 95 4.92 6.42

Employees 433 550 122 89 3.55 6.18

TOTAL 9,900 12,430 723 693 /3 13.69 17.94

/1 Source, 1965 data: BCEOM '966, p. 2.63. /2 For derivation, see Table 6, Appendix A. /3 This total excludes the 80 househ-olds interviewed in Manakana and Ambodivohitra but includes the 20 households in Belaoko-Lokoho which is part of the district of Andapa.

Market/Price Survey

2.31 Prices for some of the most commonly purchased staples (rice, sugar, salt) and 'luxury' items (soft drinks), as well as for specified local foods were established monthly, Augt'st to November 1975, in ten different marketing outlets (two in each canton). Interviews purchased small quantities of the products in question, which were subsequently weighed to establish 'actual' rather than 'posted' per Kg-price; This was particularly important for some of the local products customarily sold in ill-defined quantities such as bundles, (vegetables) or heaps (dried fish).

Transporter Survey

2.32 All transporters (29) providing regular passenger or freight ser- vices along the Andapa-Sambava road were interviewed to provide information on the structure of the local road transport industry and on volume, charges and costs,

Origin/Destination Surveys

2,33 Origin/destination surveys were carried out of all traffic on tile road between 6:00 am and 9:00 pm on 14 days. For seven days, counting was 17 carried out along the road near Andapa, on the other seven days, the teams were stationed near Sambava. Different vehicle types were distinguished; passengers were not only asked aboit thieir origins/destinations but also about the reason for undertaking the trip. The dates when counts were carried out are given in Table 6, App. C.

Special Surveys

2.34 ln addition to the sample surveys, 'special surveys' were carried OUL to apdate information on a number of regional activities and indicators:

(a) Administrative Structure of the Study Region:

Local administration: communal budgets

(b) Social Services:

- education: schools/teachers/enrollment

- health hlospitals/doctors/treatments

(c) Infrastructure:

- Enterprise MURRI Freres: data on Sambava-Andapa Road

- Public Works Dept: data on roads in the Andapa Basin

Air Madagascar, Port Authorities: Air/maritime traffic at Sambava

- Water and Forest Services: drainage; soil conservation SEM (Societe d'Electricite de Madagascar): drinking water, electricity.

(d) Agriculture: Crops/Yields/Marketing

- OrACA/AGRER:production/extension - general

- GNIV (Groupement National Interprofessional de la Vanille): Vanilla production and marketing

- Area Importers/Exporters: marketing of rice, coffee, vanilla ^~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~

SECTION II C

DATA COLLECTION - 18 - C. DATA COLLECTION

Some General Observations

2.35 The BCEOM baseline study provided unusually detailed and complete descriptionsof not only study design, but also procedures of data collection and data processing. These made it possible to modify follow-up survey procedures in such e way as to ensure minimum loss of informationdespite differences in approach dictated by differences in study finances and time horizon.

The Base Survey - 1964/5

2.36 The study horizon for the BCEOM surveys was a full 12 month period. In deciding on a preferred study design, two alternativeswere considered:

(1) a small sample of households to be surveyed several times per year;

(2) a larger sample of households to be surveyed once during the study period.

Based on earlier survey experience of one of the BCEOM consultants,the second solution was adopted, mainly because incomes and expenditureswere known to vary considerablymore between households than within households over time. Respondents'recall of non-food expendituresover the last 12 months was found to be quite detailed and reliable.

2.37 A total of ten interviewersand a field team leader resided in the study area for 12 months and carried out all data collection for the base survey. Each interviewerspent a full week with each cluster of village households (1-4): he visited each household at least once daily to weigh all food-items produced/purchased/consumed/thrownto waste and to record this and other survey information (see questionnaireguides, Appendix G), On average, interviewerssurveyed 7 households per month. In addition, they also carried out the surveys of areas under cultivation (see paras. 32-35 above); on average, 25 households were surveyed per interviewer/permonth in this survey.

2.38 As the instructions1/ accompanyingthe interview guides show, field procedures were thorough and detailed. Quality control of data gathered was provided by the field leader. Remaining inconsistencies were generally caught during coding, which was carried out by one or two persons in Tananarive between December '64 - October '66.

1/ See BCEOM 1966, Vol. 2.

I~~~... - 19 - The Follow-up Survey - 19/5

2.39 Due to the restricted time for which the field team was available, the follow-up survey field work had to be carried out in less than six months. Locating of households surveyed in 1965; mini-censuses in the study region Lo determine a sample frame for new households; traffic --ints;and household interviews in the immediate area of road-influence;compressed effective time available for household suiveys in the Andapa region to three months (for questionnaireguide and instructions,see Annex G),

2.40 Five survey teams, each consisting of two local interviewersand one Swiss counterpartwere organized to carry out household interviews in the five different cantons. Those groups working in the cantons closer to Andapa assisted groups dealing with more difficult cantons as necessary and/or took over a greater share of special survey work.

2.41 Because of time pressure, it was decided that informationab.ut food consumptionwould be elicited by questions rather than by detailed measurementsas effected by BCEOMo In this way, interviewerscould handle an average of 20 interviewsper month each.

2.42 Consistency checks were carried out by the graduate team on all completed questionnaires;discrepancies were cleared up with the interviewer if possible, by re-survey with the respondent if necessary.

2.43 Coding was carried out in December 1975 - January 1976 by the interviewersin Andapa under supervision of the Malagasy field teau leader and a Swiss junior consultant for data processing.

2.44 Whenever possible, identical codes were used for the 1965 and 1975 data. Appendix G specifies concordancesand divergences.

2.45 Both the BCEOM and the IBRD sponsored fieldteams found the col- laboration from the population in the region consistentlygood. Although respondentswere asked to spend considerabletime in answering long series of questions, they not only deliberated answers carefully,but even offered hospitality and small gifts to the survey teams after interviewswere com- pleted.

2,46 The study generated more than 85,000 IBM cards. Those prepared by BCEOM on the baseline survey (= 42,000) had been scrapped after the preliminary tabulationswere carried out and the data stored on magnetic tapes. Unfortunately,these data tapes were no longer locatable in 1975, Only by reconstitutinga new deck of cards (in Washington) from photos of the original code sheets (in Antananarivo)was it finally possible to es- tablish the detailed base survey informationagain.

2.47 The IBRD follow-up survey card deck was prepared in Zurich/ Switzerland,where also the bulk of the 1975 survey findings were tabulated by Messrs. Schlaepferand Rakotonirinawith the generous support of the Institute of Kulturtechnikof the Swiss Federal Institute of Technology (ETH). The comparative tables on 1965 and 1975 household income and expenditurepatterns were prepared by RTI (ResearchTriangle Institute), North Carolina. SECTION II D

THE FRAMEWORKFOR THE ANALYSIS - 20 -

D. THE FAMEWORK FOR THE ANALYSIS

2.48 Given - essentially- sample data from two time periods for examin- ing the social and economic impact of a first all-weather road connection to a previously isolated region and its communities,there are two basic types of generalizationsinvolved in analyzing the data: (1) generalization to the universe population, and (2) generalizationto theoretical relation- ships. Adding the time dimension, four levels or aspects of analysis can be distinguished:

Time Dimension Type of Generalization synchionic diachronic

Sample to structure: a) 1965r1 Changes in structure/2 Universe 1965-75: b) 1975 Constraints on hypotheses

Observation to Exploratory/3 a) Pattern changes: 1965-75 theoretical analysis of relationship patterns a) 1965 b) Confirmation/rejectionof b) *1975 with-without road hypotheses

2.49 Sample to population generaliz'tions(boxes I and 2) flow from the survey design adopted in the study and involve (overall and strata) sampling fractions,sampling errors, and the establishmentof confidence limits on universe estimates. Both the 1965 and 1975 surveys used samples that were stratifiedby occupationalgroups because it was hypothesizedthat -

(a) existing income/expendi.turepatterns for farmers, craftsmen, traders and salaried employeeswere different;and

(b) benefits from newly provided road access may accrue in different proportions to these subgroups of the regional population.

2.50 To ensure adequate subsample representativeness,non-farm groups had to be sampled at much higher rates. Their relative over-representationin the sample population therefore needs to be taken into account for those as- pects of the analysis where inspectionof preliminary survey results suggests differentialdistribution by occupationalgroup (e.g. income/expenditure patterns, travel behavior). On some aspects of social structure (e.g. age- structure of the populationwhich is a constrainton, say, number of persons entering the labor force each year), use of unweighted sample data will not introduce serious bias into the estimates. 21 -

2.51 While detailed sampling fractions for 1965 sample data are avail- able, 1975 sampling fraction for the different occupationalgroups are only approximations(see Table 6, App. A for derivation). Adjustment of these will only be possible once detailed area-returnsfrom tU 1975 National Cen- sus become available. Data presented in this report are therefore restricted to sample results (almost all unweighted).

2.52 The generalizationsto theoretical relationships(boxes 3 and 4) seek first to discern differencesin patterns of social and economic rela- tionships among different groups in the study region, as well as between the region and the rest of the country; second, to determine the changes in patterns (if any) that are reported from 1965 to 1975; and third, to estab- lish the extent to which such changes are attributableto improved access (the road investment). This third analysis includes tests of hypotheses about the impact of the road. The major focus of the analysis presented in this report will be on the latter aspects.

2e53 Information from such secondary sourcrs as are available will be examined together with the 1965 and 1975 data sets to establish broad pat- terns (box 3) and pattern changes (box 4) in basic dimensionsof social and economic activity in the region. Suxmmaryind cators of social activity include basic demographic data; information on local administrativeorgani- zation; infrastructureendowment; and level of social services provided. Important economic indicatorsare aggrezate agriculturalprodu':tion; pro- portion of crops marketed; and producer prices of agricultura'.inputs and outputs. This informationwill serve to indicate and circums2ribethe important changes besides the new road that have taken place during the study period. These have to be taken Tnto account when evaluating the impact of the road investment.

2.54 There exist a number of hypotheses about the relationshipsbetween improved accessibilityand change in agriculturalproduction that are well- developed theoreticallybut have not been demonstratedto hold practically. The major purpose of a study like the present one is to subject some of these hypotheses to as rigorous a test as feasible. Four major hypotheseswill be examined in the Andapa case.

Hypothesis1

2.55 Constructionof a new all-weather road from a major port to a pre- viously isolated community i:3 an event which is hypothesizedto have major impacts on productionactivities in the road-influencearea. Generally, the new road would be expected to increase producer surplus due to (a) reduced unit costs of transportingproduction inputs and outputs, (b) increasedpro- duction for marketing and probably (c) increased local consumption. Improved access should make it attractivefor farmers previouslyoutside the boundary of commerciallyfeasible agriculture to produce for the market, and induce those who already were marketing some part of their crop to increase the proportion of total output sold. - 22 -

2.56 This hypothesis will be tested using two different methods: the road user savings method (RU) and the producer surplus (PS) method. Ad- vantages and limitationsof each approach have been detailed extensively elsewhere. 1/ Essentially,the two approaches attempt to measure the same phenomena: they are thus alternativemeasures of net benefits attributable to the road, not additive. Given a freely competitive transport industry, pure cash crop production by farmers and zero home-consumptionof self- produced crops, both approacheswill give identical results in the benefit calculus. 2/

2.57 However, these conditions are not often satisfied in practice and the two measures will thus tend to show different project returns. If, for example, transport cost savings are not fully passed on to the farmer-producers,production increases (and thus transport demand) will be lower than theoreticallyexpected and the RUS approach will yield a downward biased measure of road benefits, which would need to be adjusted by the increased net income of transporters (or other middlemen) who appro- priate part of the benefits.

2.58 Similarly, where self-consumptionof own-producedcrops is impor- tant -- a typical case for many developing countries -- transport demand will not reflect increase in home consumption and the RUS approach will thus fail to account for a potentially important part of total road bene- fits - those associated with the original and increased consumptionby farmers.

2.59 Using producer surplus to measure the economic impact of a road investment in rural areas remedies the deficiency of the RUS approach with respect to home consumption. By estimating the production effects of the road not at the level of the road (traffic volume and cost of transport services), but at the level of the farm (volume and value of agricultural productlon),it captures not only changes in output marketed, but also changes in the level and value of original and increased consumptionof own-producedgoods by area farmers. In its turn, however, ths producer s':rplusapproach fails to account for passenger benefits other than those realized on farmers' trips to market (which would be reflected in the decrease of production costs). Benefits to passengers commuting to work or travelling for pleasure -- which are fully accounted for by the RUS approach -- would need to be added to the benefits establi3hedby applying the PS method in order to reflect the full benefits of the road project.

2.60 The appropriatenessof using either approach also will depend on the completenessand reliabilityof the "without (before) project" data available for the analysis,and on the sturdiness of the assumptionsmade to forecast developments (traffic levels; crop yield levels; price levels)

1/ See: Carnemark et al: The Economic Analysis of Rural Road Projects World Bank Working Paper No. 241, Washington, D.C. August 1976, and The Socio-EconomicImpact of Rural Roads in Brazil, (unpublished manuscript) IBRD, Washington, 1976. Appendices 1,2, and 3, 1976. 2/ See Carnemark, op. cit., Figure 2, p. 6. - 23 - over the project's life. For remote rural areas, the analyst rarely disposes of sufficientlydetailed data to determine what combinationof improved subsistence/increasedexports/decreased imports, etc., are the most likely result of a road investment;only in rare cases will he have the time and means to generate, in the field, information that would allow him to apply the theoreticallymost appropriaceanalytical procedure.

2.61 The data collected for this study permit to test the hypotheses of the production impacts of improved accessibilityby both RUS and PS methods. First, the regional data gathered by special surveys in 1965 and 1975 (sup- plemented by systematic sample data which permit inferences to the entire population in the Andapa Basin) provide data on regional productfon and pro- duction costs, marketed share and self-consumptionthat allow a PS calculus to be carried out. Second, the data on comparative costs of air transport/ porterage vs. road access permit quite precise estimation of the potential impact of transportcost reductions on absolute and relative returns to production and sales of a range of locally produced crops (RUS). Finally, 470 of the households surveyed in 1965 were also surveyed again in 1975. This panel data will offer an unusually precise look at the impact of the road improvementby making possible an analysis of producer surplus using micro-data. 1/ Informationon the most importantexternal factors influenc- ing regional production such as relative price levels of major area crops, inflation level, etc., is used to isolate the impact of the road from other effects as far as feasible.

Hypothesis 2

2.62 A detailed comparisonwill be made of 1965 and 1975 income, con- sumption, and transportpatterns of sample households. A familiar hypothesis is that transportcost savings from road improvementswill be passed on from transportersto producers provided that active competitionexists among the suppliers of transportservices. (The presence or absence of such competi- tion will be establishedvia a special transportersurvey). The methodology suggested by Walters (1973) 2/ offers a test for the extent to which trans- port cost savings are shared between transportersand producers. This in- volves comparing actual 1975 farmgate prices with theoretical1975 farmgate prices (the latter being estimated as the level to which 1975 farmgate prices should have risen with inflation and by replacing air freight with surface transport).

1/ See The Socio-economicImpact of Rural Roads in Brazil, Appendix 4 (unpublishedmanuscript) IBRD, Washington, 1976.

2/ Walters, Alan: Yemen Arab Republic Feeder Road Study, (unpublished manuscript) IBRD, Washington, 1973. - 24 -

Hypothesis 3

2.63 Of particular interest for the road impact analysis are changes in level of income and income distribution. An early hypothesis of the "trick- ling down" approach to developmentwas that economic growth will be attended by a mo2e egalitarianincome distribution. Opposite results seem to emerge from several studies in Latin America. The data collected for this study should provide additional insights with respect to this hypothesis. First, the proportion of aggregate income which is received by different deciles of sample households arranged in ascending order of household income will be establishedfrom both the 1965 and 1975 survey. Changes in the proportion received by the different deciles will indicate whether income distribution has become more or less skewed in favor of lower income groups. Second, the same analysis undertaken separately for households in the different occupa- tional groups and for households living closer or further away from the new road will establishwhether any subgroup has disproportionatelybenefitted from improved access and/or other changes that may have taken place over this period.

Hypothesis 4

2.64 Regardless of whether an increase in level of income or improve- ments in income distributionoccur, the opening of the road can be expected to result in some changes of consumptionpatterns. These would mainly be in the consumptionof high volume/low value goods air transport cost of which made them prohibitivelyexpensive to area residentsbefore the road was built.

2.65 Assuming an upward trend in the level of real incomes in the region, there should be other indicationsof change in consumptionpatterns with a relatively smaller proportion of total expendituresbeing allocated to basic food consumption and increasing shares going to consumer durables, services and investments.

2.66 If initial statistical analysis of consumptiondata confirms the postulated trend, the analytical device used for examining consumption patterns will be Engel functions. These specify the relationshipbetween expenditureon consumptionof particular categories of goods and services and total income. 'Engel's Law' says that the proportion of income spent on food declines as income rises. The income elasticity of demand for food should thus be less than one, while that for clothing, consumer durables and other luxury-typegoods would exceed unity. If aggregate real incomes in 1975 are higher than those in 1965, the measured income elasticity of demand for food should have declined, that for "superior" goods should have increased. If the data are consistent with this hypothesis, it would provide a useful validity check for the follow-upand the base line surveys, as well as for the study results compared to data gathered elsewhere. - 25

2.67 Based on the works of Houthakker, 1/ Engel functions will be esti- mated in the following form:

log Y = ai + a1 log Xl + a2i log X2 + e, where

Yi = value of consumption for the i-th commodity group by household

Xi =value of total consumption (all commodity groups)

X2 =household size aoi, a,i, a2i = parameters

ei = error term

2.68 Engel's Law, therefore, requires that for food (particularlystaples) Ou-a11<1, where a f is the estimated regressioncoefficient (also, the income elasticity 2/ of emand) for food consumption. Further, a should have de- clined from 1965 to 1975, consistentwith the evidence tha freal incomes have risen over that ten-year period, and aif should decline as we move from lower to higher income groups within a given sample year. For non-food, the expecta- tion is that 1 < a2n< ', where a2n refers to the estimated income elas- ticity of demand for consumptionof non-foods items. 3/ Of course within the food and non-food aggregates, there will undoubtedlybe individual items which deviate from the expectationsfor the group average; imported food, for instance, is probably not an 'inferior'good for a sizeable part of area pop- ulation. Finally, differentialprice changes for food and non-food products also may result in a higher share for food expenditureseven when real income is increased. If we denote: £- the Engel elasticity of food, n - the price elasticityof food, p- cross elasticity of the demand for food with respect to other commodity prices, S - share of expenditureon food, Y - total income, Pf - price of food and P - price of non-food items, it then holds that

1/ Houthakker: An InternationalComparison of Household ExpenditurePatterns, Commemoratingthe Centenary of Engel's Law in: Econometrics,Vol. 25,4 (October 1957). 2/ Following Houthakker, the designation"income elasticity" is retained for brevity, even though total consumptionexpenditure is actually used as the principal explanatoryvariable. 3/ In fact as noted by Houthakker (1957), p. 264, the relationshipbetween the elasticitiesof food and non-food consumptioncan be summarized as EYiaI = XI, or that the sum of the elasticitiesof all consumption categories (each being weighted by the expenditureon the commodity concerned) must always equal unity. - 26 -

APf APn A S ="(e _ 1) . LY + 01+1) +1'

s y pf -n

By Engel's LawC -1 < 0; many d2mand studies have established -l0. Thus, the first component in the question above is negative, the second is positive, and the third may be positive or negative.

If the study shows that

AS Ay APf AP - > - >f n S 'Y ' Pf p the net effect of all the changes on the budget share of food could thus be positive without violating Engel's Law. CHAPTER III

FINDINGS - 27 -

SECTION III A

REGIONAL STRUCTIJRES- 1965 - 1975

(Non road-inducedchanges - constraints)

One of the least recognized reasons for misleading project return estimates is deficient definition of the without/beforeproject situation. Project components usually are clearly defined and described, but the regional system into which the project will be inserted and which it should "fit", often is not delineated in any depth and thus is little understood. When things "work", the project is often credited with effects to which other factors have substantiallycontributed; alternatively, project failure is sometimes assumed, where other componentsof the regional system should rightly be blamed. In this study, therefore, a detailed descriptionof regional struc- tures and t.eir changes from the base survey to the follow-up study period precedes the economic analysis proper so as to highlight potentialsystem constraintson project results. - 28 -

A. REGIONAL STRUCTURES 1965-1975: COMPARATIVE PICTURE DRAWN FROM SECONDARYDATA SOURCES AND SPECIAL SURVEYS

1. Local AdministrativeStructure

3.1 The sub-district(sous-prefecture) of ANDAPA is one of four sub- districts constitutingthe administrativedistrict (prefecture)of ANTALAHA. The latter in turn, is part of DIEGO SUAREZ province, the northern-mostof six provinces that make up the Democratic Republic of Madagascar.

3.2 The sub-districtof ANDAPA is further divided into 7 cantons, two of which are situated outside the Andapa Basin, which is the area of study. The five cantons (including the town of ANDAPA) have grown in popu- lation by 29% from 1965 to 1975, i.e. somewhat less than the country as a whole (33% - see IBRD Report No. 1099a-MAG,Annex V, Table 1.1).

Table 3

Population of the Andapa Basin: 1965-75

CANTON 1965 1/ 1975 2/

ANDAPA - urban 3,917 8,960 suburban 7,528 7,432 11,523 13,063 6,010 7,618 AMBODIMANGA 8,509 11,194 MAROVATO 8,824 11,459

TOTAL 46,311 56,726

3.3 The main administrativefunctions at the canton level are the collec- tion of taxes, the maintenance of communal property, and the assurance of law and order. Coordinationbetween the local administrationand the local depend- encies of various technicalministries such as Public Works or Agriculture,is provided at the sub-districtlevel by the sub-districtofficer (sous-prefet) who is a regular civil servant appointed to this position by the central authorities.

1/ Source: BCEOM 1966 p. 2.53: Households per canton multiplied by average size of household (4.56).

2/ Source: PreliminaryCensus Results, Province Diego Suarez, District Antalah&. February 1976. - 29 -

3.4 Alongside this formal administrativestructure there has also existed in Andapa as in other parts of the island, an 'informal' traditional form of community organization: the Fokonolona. These are associationsof people who occupy-cultivate-haveclaim to a common area of land and regulate its allocation to the individualmembers of the fokonolona. Land disputes have commonly been settled by these traditionalauthorities rather than by official court procedure.

3.5 The importanceof the fokonolona is that it extends communal bonds from an earlier exclusive focus on family and lineage (those related in time) to all who live in the same community (those related in place). Especially in regions where settlement has been relatively recent and by individuals/ couples of different ethnic affiliationand background as in Andapa, (rather than by homogeneous extended family groups), the fokonolona have played a valuable part in creating community cohesion in newly settled villages (see AGRER 1964 pp 163 ff). They have controlled allocation of communal land to new families, organized communal labor for such tasks as the construction and maintenance of roads or the installationand annual cleaning of drainage ditches and canals; and provided facilities for educating the children.

3.6 The Malagasy Government'spolicy of reorganizingand decentralizing local administrativestructure based on the fokonolona principles thus has not presented a drastic change in the Andapa region's administrativepattern but has rather served to formalize an 'informal' reality. As currently imple- mented, the new organizationstarts from the

"Fokontany"- as its basic cell: a village - or group of villages whose inhabitantschoose to group together and elect a president, secretary and councillorsto represent them. Next, a number of fokontany units in a

"Firaisampokotany"- roughly equivalent in size to the previous canton (the current reorganizationof the study area at this level is shown in Table I Annex B). The next stages planned for reorganizationare

"Fivondronam-pokotany"- an association of Firaisampokotany- roughly equivalent of sub-districtsand

"Farintany"- an associationof Fivondronam-pokotany- roughly equivalent of districts.

When the restructuringprocess is complete, local administrativeofficials up to the Farintany level will be elected to their office by the population they serve rather than be appointees of central government.

Community Budgets: 1963-1972

3.7 Unfortunately,no detailed communitybudget estimates are available for the post-1972 period, when the new administrativestructure began to be - 30 -

introduced. Overall, tax revenue is said to have declined after 1972, when the head-tax and the cattle-tax- two important revenue-generators until then - were abolished.

3.8 Comparison of 1963 with 1972 budgets of the rural communities 1/ in the Andapa Basin shows revenue estimates increasingby over 50%, over the decade. But with recovery rates dropping from 91% to 71% of assessed value, there is in fact very little change in actual community revenues. Expendi- ture estimates in both years allocate about 70% of total expenditure to operating expenses and the remaining 30% for investments;actual operating expendituresamounted to 74% of total revenue in 1963, and to 83% in 1972. Personnel costs accounted for half the operating expendituresin 1963, for almost two-thirds in 1972. The bulk of investment expendituresis for con- struction or purchase of administrativebuildings.

3.9 The correspondingbudgets for the township of Andapa, the admini- strative center of the region, actually show a slight decline in ordinary budget receipts and operating expenditurebetween 1963 and 1972. Forty and forty five percent respectivelyof total ordinary expenditureswere spent on operating expenses, while investmentsaccounted for sixty percent of ordinary expenditures in 1963 and fifty-five percent in 1972. The biggest single item (41%) of investment expenditurein 1963 was building construc- tion as in the rural communities;but in 1972 the bulk of investments (69% of total) were for the installationof an urban water supply system. (Extra- ordinary expenditure estimates consist mainly of provisions for repair of cyclone-damages).

1/ For details see App. B, Table 2. - 31 -

2. Social Services in the Study Area

Education

3.10 Educational facilities in the study area have increased very markedly from the time of the base-survey (1965) to the follow-up survey period (1975). A survey of all schools in the Basin showed that the number of schools and teachers more than doubled from 15 to 32, and from 40 to 87 respectively(for details see Table 5, Appendix B) at the primary school level. Secondary education was provided only by private (mission) schools in 1965; by 197r, a public secondary school (first cycle) has been estab- lished in Andapa. Pupils who want to continue beyond this level have to move to boarding schools in Antalaha or Diego Suarez.

3.11 In the absence of detailed population data from the 1975 census, it is not possible to establish precise enrolment rates for the population of school-age. Of the total population of 4,174 persons in the sample survey, 1,257 or 30.2% are in the 5-14 age group (see Table 4 Annex E). This would indicate a total population in the study area of over 18,000 in this age range, while total primary school enrolment is under 8,000. This compares with 4,090 primary school pupils in 1962. 1/

3.12 At the primary level, the number of boys and girls attending school is almost equal (53% and 47% respectively). But the attrition rate for girls at the higher levels is much greater than that for boys: the number of boys in the last (sixth) year of primary school is 20% of that in first grade, while the correspondingnumber for girls is only 13.5% (see Table 4 below). Compar- able data for the base survey period are not available.

3.13 There is a significantdifference in attrition rates between ANDAPA and the exclusively rural cantons in the area: in ANDAPA, I pupil in 3 stays the full course of primary school, while only 1 in 8 does so in the rest of the region. Partly, this reflects the fact that a number of the schools in the rural areas have not been establishedfor a sufficient number of years to build up to a full six-year program; but the figures also seem to indicate that the present curriculum- once it has provided pupils with the basic 'reading-writing-rithmetic'skills - offers little training in technical skills that would be relevant and useful for the bulk of the farming popu- lation.

1! See: AGRER 1964, page 71. - 32 -

Table 4

Attrition Rate of Primary/SecondarySchool Pupils - 1975 Enroluent

I II PRIMARY SCHOOLS N in First Year (CMI) N in Sixth Year (CM2) Ratio Male Female Total Male Female total II to I

ANDAPA 307 264 571 114 81 195 1:3

AMBODLANGEZOKA 397 343 740 57 28 85 1:9

ANDRANOMENA 152 134 286 24 15 39 1:7

AMBODIMANGA 347 324 671 61 26 87 1:8

MAROVATO 312 285 597 47 28 75 1:8

TOTAL 1,515 1,350 2,865 303 178 481 1:6

PERCENT 53 47 100 63 37 100

SECONDARY SCHOOLS First Year Fourth Year (6 eme) (3 eme)

N 102 42 144 60 21 81

PERCENT 71 29 100 74 26 100

3.14 Interviewswith local school officials revealed that the students going on to secondary school do so mainly with the ambition of qualifying for a post as salaried employee of the administration,which, as a rule, is able to absorb only a minority of them. Those who do not succeed in finding a position often remain unemployedrather than return to farming.

Health

3e15 In contrast to educationalfacilities, health services in the study region have not changed perceptibly from the base survey period, except for the number of paramedics,as Table 5 below shows* They are, however, expected to improve significantlywith the opening, in 1976/77, of a new hospital being constructed in Andapa by the Adventist Church. This will provide surgical as well as dental services in the study area for the first time. - 33 -

Table 5

Health Service Facilities: 1964/65- 1975

Facilities 1964/1965 1975

Hospitals I I Medical Centres - 1 Health Posts 3 3 Dispensaries 2 2 Pharmacies 8 9

Hospital Beds 42 38 Maternity Beds 42 32

Doctors 2 3 Dentists 1/6 - Midwives 3 8 Nurses 9 12 Nurses' Aides 6 12 N Patient Visits/mth - hospital Andapa 2,500* 3,830** N Patients/mth - hospital Andapa 1,500* 1,160** N Births-AndapaHosp/mth 47* 68** N Patients Hospitalized/p.a. n.a. 383** N Patients transferredto Antalaha Hosp/p.a. 10* 120**

Source 1964/5 data: AGRER 1964, pp. 65-70

* N for 1963 ** N for 1974

3.16 Hospital statisticswould seem to indicate that perhaps a smaller number of patients (with more serious conditions) see the doctors more frequently,while the improved availabilityof medicines through pharmacies has promoted treatmentof less serious cases by self-medication. A dramatic change, has however, occurred in the number of seriously ill patients be- ing transferredto the better-equippedfacilities in the district capital Antalaha: with the opening of the road, this has risen from an average of 10 patients per year to 10 patients per month. However, improved hospital services are still among the most important developmentneeds mentioned by area residents.

Diseases and their Prevalence

3.17 In 1964/65, the most serious illnesses reported 1/ in the study area were:

1/ Source: AGRER 1964, pp. 67-70. - 34

(i) Infectious and parasitic diseeses, especiallymalaria, and influenzaand measles among children;

(b) Tuberculosis;

(c) Venereal disease (syphilis found in 5% of pregnant women who came for obligatory examination;highest incidence among unmarried young adults);

(d) Diseases of the skin (especiallyd high concentration (1,052) of lepers many of whom initially sought refuge in the area because of its remoteness).

3.18 To judge from examinationof hospital records for 1974, the pattern of disease seems to have remained relatively unchanged, although efficient drug therapy for some of the conditions is available and drugs for such seri- ous conditionsas tuberculosisand bilharzia are provided free to patients seeking treatment. The majority of the lepra cases are non-contagious;of the roughly 1,000 cases on register in the region, 300 are currently reported to receive treatment. (Table 6 Appendix B details the most frequentlytreated conditions at the Andapa Hospital for 1974, the latest full year for which informationwas available). A total of 13,935 patients wvereseen for 14,321 different conditions. Among these, infectious and parasitic diseases (includ- ing malaria) were most prevalent with 5,719 cases (40%), Respiratorydiseases are the next most frequent with 3,253 cases (23%); followed by diseases of the digestive system (2,171 or 15%). Accidents, poisoning and shock account for 714 cases (5%).

3.19 Some pro2hylacticmedicine is provided in the area. Vaccination against smallpox is carried out free; also, antimalarials("nivaquine") are issued free of charge for children under five, and dispensed twice weekly to all children attending school. But treatment lapses during holidays and is only rarely pursued with regularity once children have left school.

3.20 People, especially in the rural cantons, tend to frequent tradi- tional healers rather than doctors and nurses for a range of problems both physical and mental. Employing traditionalmedicines - as well as consider- able skills in muscle-relaxing(massage) and bone-settingtechniques, these traditionalhealers register considerablesuccess and are highly esteemed in the community.

3.21 Although the problem of hunger is unknown in Andapa, nutritional patterns ir.the study area result in some deficiencies. The most serious of these is a calcium deficiency (too much rice, greene, sweet fruit; too little meat and milk) manifesting itself mainly in frequent dental problems. Iodine deficienciesand the number of goiter cases seem on the decline: since the opening of the road consumptionof fresh fish and other seafood imported from the coast has increased considerably. - 35 -

3.22 A major contribution to better health of the area population would be the more careful separation between sources of drinking water, water for cleaning, and waste disposal. Even in Andapa, where a piped public water supply exists, dishes and clothes are customarilywashed in the very same river or canal that carries away the debris. - 36 -

3. Infrastructureand Technical Services

3.23 As could be seen from the community budgets, the town of Andapa fulfills a number of 'central place' functions for the cantons and villages in the sub-district. Besides serving as the focus of trade and exchange, Andapa also provides the region with basic administrative,technical and social services.

3.24 In 1965, Andapa had no bank facilities. In 1975, there was one branch-office of the Banque Commerciale de Madagascar (BCM), but no informa- tion on number of accounts and volume of transactions is available. The annual volume of financial transactionshandled by the post-office has de- clined markedly over the ten-year period 1965-1974 (see Table 9 Appendix B): the level of deposits dropped from FMG 252 million in 1965 to FMG 155 million in 1974; withdrawals show a parallel decline from FMG 255 million (1965) to FMG 159 million (1974).

3.25 While collection of direct taxes is handled by an office in Sambava, indirect taxes and revenue from state monopolies (tobacco and matches) for the sub-districtare collected by an Andapa office. A special agency repre- sents the treasury at the local level: it handles the salary payments for all government employees in the area and collects taxes other than those mentioned above.

3.26 A registry office deals with some of the changes in land titles in collaborationwith the cadaster. But its main function is the registration and licensing of vehicles. The growth of vehicle ownership in Andapa from 1964-1975 is remarkable: in the five years since the opening of the road the number of vehicles registered has increased by a factor of 2.5 and com- pared to the time of the base survey, there are now thirteen times the number of vehicles registered in 1964 (229 vs. 17; for details sea Appendix C, Table 9).

3.27 Public supply of drinking water exist in Andapa-town (financed by a state credit of FMG 16 million and community funds of FMG 12 million), and in MIarovato(financed by the Fokonolona and the Catholic Mission). Construc- tion of the dam started in 1968, work was completed in 1973. Water is piped a total distance of 7 ims - this includes 1.2 Km between the dam3and the tank. Tank capacity is 150 m ; current da±ly consumption runs at 131 m . In all, there are 10 public standpipes, 18 administrative customers and 110 private connections. Connection fees are FMG 15,000 for pipes of greater than 15 cm diameter; FMG 11,000 for lesser diameter. Rental charges of FMG 102/month for the water meter as well as a fixed monthly fee of FMG 61 are levied. The charge for every cubic meter of water consumed is FNIG40.8. For 1976, extension of pipes by another 900 m is planned to satisfy 50 new connection requests. There is at present no plan to extend public drinking water fa- cilities to other communities in the Basin.

3.28 Electrificationof Andapa-town is foreseen for 1976, when a pumping station currently under constructionenters into service. Estimated elec- tricity consumption is 50 K'W. In addition, a central power station on the - 37 -

river Lokoho is planned as part of the irrigation/drainageprogram in the Basin. This would produce 700 KW - sufficient for the electricityneeds of the total study area. 1/

3.29 The Public Works department has a subsection in Andapa which, in 1975, employed a total of 16 people:

1 technician (section leader) 2 skilled workers 1 mason I housepainter 2 carpenters 8 casual laborers 1 apprentice

The unit is equipped with one 2.5 ton truck, one Landrover and one tractor- cum-mower. Additional equipment can be called in from the Public Works district center in Antalaha as necessary.

3.30 As in 1965, the unit handles both building constructionand civil works. Regarding the latter, it shares the responsibilityfor maintain- ing the ANDAPA-SAMBAVAroad with a unit in Sambava (annualmaintenance allocation = FMG 5 million), but its main function is the maintenance of two intercantonalroads (see plan):

(a) the "middle road": ANDAPA-ANDRANOMENA-AMBODIANGEZOKA(22 Km) bridges : 11; Maintenance Allocation: oll, FMG 1.5 million p.a.

(b) the "west road": ANDRANOMENA-AMBALAMANASYII (20 km) bridges : 6; Maintenance Allocation: FMG 14 million p.a.

3.31 All other roads in the Basin (some 50 Km) receive what little main- tenance they get from the local communes (see budgets App. B) and from the fokonolona. During the rainy season, most of them are passable by landrover only. Two road sections have been included in the FED project for upgrading: the "south road" from Ambodimanga to Ambodivaina (= 27 Km), and the north road from Marovato to AmbalamanasyII (12-15 Km). However, no contractors have been found to date to carry out the works.

3.32 The Water aud Forest department is reponsible for the controlled exploitationand preservation of forestry resources. With 85,000 ha of natural forest, these constitute an important regional asset. However, there is only one forestry agent in the whole Basin; and he has few if any legal means of stopping a dangerous and increasing trend by the region's growing population of burning and cutting down natural forest cover to extend their cropland for rice as well as coffee and vanilla. Between 1965 and 1975, 1,200 ha of forest in the Andapa - sub-districtwere destroyed. Only very little

1/ Both projects suffered considerablesetbacks from the cyclone Danae in February 1976. - 38 - reafforestationwith pine (pinus portula and p. cassia) and eucalyptus (e. rostrata and e. saligna) has taken place. Controlled exploitationof forest resources is for local use exclusively: in 1974, a total of 1,745 mature trees (measuring1 m in circumferenceat a height of 1.30 m above the ground) were cut. Most of the timber was used for house constructionand in local furniture production.

3.33 The Agriculture Service section for the Andapa Basin is comprised of - (the figures in parenthesisgive the 1964 equivalent staffing):

1 Sub-districtofficer (1) 1 Technical assistant (1) 1 Secretary - 3 Sector heads with (12) (technicians) 3 Extension officers to support them.

However, due to lack of qualified personnel,not all of these positions are currently filled.

3.34 Taking over part of the functions of the Agriculture Service as well as those of a previously existing office of Rural Engineering is OPACA (1'OPERATIOND'AMENAGEMENT DE LA CUVETE D'ANDAPA). This is an Agricultural Development Project financed by FED credits and implementedunder the direc- tion of the Ministry of Rural Development and the Rural Engineering Bureau. The organigrambelow details OPACA's current function and responsibilities.

ORGANIZATIONCHART OPERATIOND'AMENAGEMENT DE LA CUVETTED'ANDAPA (OPACA)

MINISTRY OF MlNlSTRY OF RUFRAL - RURAL DEVELOPMENT lENGINEERING

AGRER _ _ OPACA BRUSSELS CONTROL TECHNICAL ASSISTANCE

| 0 | W~VORIKiHOPS SUPERISIN EXTENSION AND O MISCELLANY PUBLIC WVORKS

OPERATION SU9 rlN OPERATION OPERATION L IVEFSTOCK TOPOGRA.'HA' AND RICF COFFEF -PIG F UPF WVISIUN

.... F..'Ra, 18640 CUVETTEDANDAPA km. 18.5A

ReseauRoutier/ km.21.5 10

RoadNet km. 22.5 Qkm. 17.0

m. km. 23. \ikm. 15.0 km. 25.5 9 9 km.26.5* kir. 13.5

km. 27.5 *km. 12.5

¢s |R~~~~~OUETE km. 460oQkm. tO.5

Amrbodiangezla% NoORTD

km. 370 M.3, Andanomena s Marovto

km. 39.o2 m 5 t;sW~~~~boiaina 00 4P~

akm. 440 30

Pm.4t.m.i 60.2*km. 64.3 km- 460

km. Sl.0 \ / 10

' ROADIN BADSTATE QD4BRIDGE (LENGTH) Ambodimanga i KILOMETERMARKER - 39 -

3.35 OPACA staff currently employed in Andapa are as follows:

I Director (expatriateexpert seconded from AGRER) 1 Assistant Director - Coffee expert 1 Agronomist - Rice and livestock I Engineer - Drainage works 1 Senior Extensionist- Coffee station 1 Workshop Director (supervising10 employees) I Drainage works supervisor 2 Topographers 2 Draftsmen 1 Accountant

In addition, varying numbers of auxiliaryworkers are employed throughout the year.

3.36 The initial phase of the project which encompasses improvement (irrigationand drainage) of 2,600 ha of rice-lands,as well as agricultural extension and some infrastructureimprovements in the Basin, is expected to cost 5,401,000 units of account (u.a.). Disbursementon the project to-date are as follows:

Disbursementson AgriculturalProject (OPACA) - 1972-1975 (FMG 000)

Extension service Constructionworks Year Technical Assistance Local costs Surveying Contractors

1972 45,476 6,554 13,688 79,266

1973 22,075 43,612 8,285 20,670

1974 31,928 3,021 7,828 75,761

1975 20,883 14,006 12,431 186,580

Total 119,562 67,193 42,232 362,277

Column A B C D

SOURCE: FED, 1976, Exchange rates used for u.a. to FMG: 1972/3 277.7 1974 = 285 1975 = 270 4 Oe

It is estimated that 10% of expenditures on items A, C, D, have flown into the local economy in payments to labor, materials and services; the proportion on item B was about 50% each for local and imported inputs (the latter mainly transportation, chemicals).

3,37 OPACA's first emphasis has been on improving coffee cultivation in the region, both by producing new plant material and by teaching better techniques of cultivation and plantation management.

3.38 As to the planned irrigation and drainage schemes, about 500 ha of rice land in the Basin have been improved to-date, making double-cropping and yield increases to 7 tons/ha (compared to current 2 tons/ha currently) technically feasible. Considerable extension inputs will be required, however, before such increases will be realized. Meanwhile, feasibility studies for improvement of another 2,500 ha in a planned second project phase (see map below) are under preparation. Andasibe -Kobahina

AmbodiangezokaMrvt'

Amodvaina\;k A ndranomenanS

i ZonesOPACA- lPhoases 1 et 2 Ambdimanga eso AREASecinAFetCTED BY AGRICLTURA

BealS Sonsa100000 W ~~~~Sections NotUnderosderto

. . ~~~~~~~~~~SCALE I/ lOOO0 - 41 -

4. Agriculture: Area Potential and Area Production: 1965-1975

Physical Characteristicsof the Study Region: A Brief Recapitulation

3.39 Previous reports 1/ have presented detailed analyses of the soils of the Basin and their morphology; of the nature of vegetative covers; of climate, rainfall and hydrologicalfeatures. These will be recapitulated here only in brief outline as necessary to indicate the region's agricultural potential.

Soils and Vegetation

3.40 The soils of the Basin itself are mostly clayey/sandy/loams: made up of alluvial deposits from the rivers that traverse the area. They con- stitute an area of about 17,000 ha, most o" which is currently under culti- vation. The soils on the mountain-sidesconsist of clayey loams on a basic of granite and gneiss - a mixture of alluvial and colluvial deposits and exposed base materials. Where clays predominate,they sustain dense vegetation;more xerophytic plant cover is found on quartz. The high altitude Marojezy moun- tain range (now a nature reserve) is formed of granite and sustains different types of forests ranging from tropical forest in the valleys to pines, lichens and moss on the peaks. Table 6 below summarizes the distributionof cultiv- able land in the Andapa Basin (34,000/76,000ha) by soil types.

Table 6

Cultivable Land By Type of Soil

Soil-types Mountain side Foothills Basin Total (tanety) (ha) (ha) (ha) (ha) clay on quartz 1,060 1,060 clay on gneiss 13,900 13,900 (not specified) 1,380 1,380 peat/bog 990 990 half-peat 1,540 1,540 mainly clay loams 4,500 4,500 terraces (sandy/ clay) 10,610 10,610

Total Cultivable Area 14,960 1,380 17,640 33,980

SOURCE: AGRER 1967, p.2

1/ In particular: AGRER 1967, Vol 1; AGRER 1964 - 42 -

Climate: Temperatures,Rainfall, Sunshine

3.41 The Andapa Basin has a climate somewhat different from that of the rest of the East Coast. It is characterizedby heavy rainfalls and fairly moderate temperatures(for detailed records of the Andapa meteorologicalsta- tion, see Annex B Tables 8-14). While there is some rain every month, the main rainy season extends from November-April,with a secondary peak in July-August. Average rainfall for the 1966-73 (latest available) period was 2,035 mm/p.a.; but variations can be considerable(Maximum: 1971:2,468 mm; minimum: 1972:1,494 mm). The highest monthly average temperature recorded in the past eight years was 250C (79 F) for Februarg; during the coolest months (July-August),the mercury did not drop below 18 C (64 F). The average daily hours of sunshine are lowest in July/August,highest in October. The Basin suffers from the devastating cyclones that batter the east-coast almost every year, especially during the three-monthperiod from December-February.

AgriculturalProduction Potential and Production

3.42 The physical characteristicsof the Andapa Basin outlined above show that it is well suited for an impressive range of crops. These include tra- ditional subsistence crops, such as rice, manioc, sweet-potatoes,bananas; as well as new garden crops like beans, potatoes, tomatoes, salads. Condi- tions are ideal for two major cash crops: coffee (robusta) and vanilla; also for ginger and cloves. They are less suitable for other east-coast crops: temperaturesare too low for oil-palms and coconuts; there is insufficientsun for tea; and the dry season is too short for 'arabica'coffee to flourish. Future agriculturaldevelopment of the Andapa Basin thus will need to concen- trate on more efficient production of currently grown crops rather than on further diversificationinto different products.

3.43 Between the time of the baseline survey (1964/1965)and the follow- up survey (1975), there were major shifts both in the extent of total area cultivated in the region, and in the relative importance of different crops. These are summarized in Table 7 below. - 43 -

Table 7

Major Crops - Area Cultivated 1965-70-74

A R E A I N H E C T A R E S CROP 1965 1/ 1970 2/ 1974 2/

Rice (Paddy) 4,830 11,512 17,340* Other Grains n.a. 27 650

Manioc n.a. 135 100 Saonjo n.a. 100 100 Potatoes n.a. - 20

Beans n.a. 155 150 ------__-- Bananas n.a. 400 870

Sugar Cane n.a. 375 500 ------__-- Ginger n.a. 6 Cloves n.a. 8 14

Vanilla 4,000 3,900 3,910 Coffee 2,800 3,920 3,950

Other (1,270) 15 380

TOTAL 12,900 20,412 27,990

PERCENT 100 158 217

1/ Source: BCEOM 1966, pp. 4.10 - 4.11.

2/ Source: Special Survey 1975: Service de l'Agriculture,Andapa.

* Includes tavy (= rainfed rice grown on mountainsides).

3.44 While some of the 'increase' in area under cultivationis no doubt attributableto better record keeping, there neverthelessseem to be some discerniblesignificant trends. Foremost is the increase in rice-cultivation: area under crop has grown by a factor of 3.5. By contrast, area planted to vanilla and coffee changes little, and there is incipientdiversification into new cash crops: cloves and ginger. Traditional root-cropcultivation also shows little change over time, and pulses (for home-consumptionand local markets) are not gaining in importance. - 44 -

3.45 Examination of quantities produced rather than area under cultiva- tion adds a measure of production efficiency and reveals a somewhat different picture. (Both 1965 and 1974 were average to good crop yearb).

Table 8

Crop Production in Tonnes: Andapa Basin, 1965-1974

CROP 1965 1/ 1970 2/ 1974 2/

Paddy 14,430 29,835 38,130 Manioc 1,100* 200 1,600 Saonjo 390* 30 150 Potatoes 34* 30 (200) ------Beans 50* 12 15 ------Bananas 427* 750 21,750

Sugar Cane 724* 600 (n.a.)

Cloves n.a. n.a, 8 Vanilla (green) 1,250 1,200 937 Coffee 1,330 1,500 2,285

1/ Source 1965: BCEOM pp. 4.41 - 4.43;* Table 4-2-10: Total Area Consumption.

2/ Source: Special Survey 1975 - Service de l'Agriculture, Andapa. Figures in parentheses are previous year estimates.

3.46 Area cultivatedto rice in 1975 was 360% compared to 1965-100, but production of paddy increased to no more than 265% of that in 1965, This would seem to indicate a drop in average yields due to the need to extend fields from the rich Basin soils up into the hillside, where irrigationis not feasible.

3.47 Area cultivated to vanilla declined only very slightly (just over 2%), but tonnage harvested went down by 25% over the 1965-1975 period. With a stagnating export market, Andapa farmers are either not replacing old vanilla plantationswhich have lower yields or neglect their vanilla planta- tions while turning to more easily marketed crops.

3e48 By contrast, area cultivated to coffee increased by 40% while yields reported in 1975 are 72% higher than those in 1965. This reflects the success of the OPACA extension efforts, which concentratedon coffee because of its better export prospects. Production of cloves and pepper also seems to be slowly gaining ground - a desirable trend to more diversifiedcash crops. - 45 -

3.49 Regarding subsistence crops other than rice, there seems to be welcome diversificationinto maize and various garden crops. But quantities produced are small and evidently destined for local markets at most.

3.50 To determine the relative importance for observed changes in local agriculturalproduction of road-inducedprice changes in the study area and of other changes in the prices of the main area agriculturaloutputs and inputs used, price movements at the national level are briefly examined nexto - 46 -

5. Prices of Major AgriculturalOutputs and Inputs - 1965-1974

3.51 For more than ten years, prices of all export crops in Madagascar have been subject to Government regulation. This was intzoduced to cushion producers from sometimes considerablefluctuations in world-marketprices and has worked through a number of stabilizationfunds, which establish annual 'producer prices' for each crop. These are binding throughout the country, and are weight station ("nu bascule") prices rather than farm gate prices. They serve as the basis for establishingguaranteed export prices at each of the country's ports, which include provision for transport costs, insurance and financing charges, packaging and shipping charges and, whole- saler margin. The difference between the guaranteed export price and actual world market price is paid into/out of the stabilizationfundse

3.52 Table 9 below (for details, see Tables 16-19 Appendix B) shows the 1966-74 developmentof producer prices and realized export prices for the major crops produced in the Andapa Basin. Producer prices between 1966 and 1974 increased appreciablybut, except for coffee, somewhat less rapidly than world market prices.

Table 9

ComparativePrice Increase for Selected Crops: 1966-74 (ProducerPrice and Realized Export Price: FMG/Kg)

1966 1974 % Increase: 66-74 Producer Realized Producer Realized Producer Realized Crop Price Exp. Price Price Exp. Price Price Exp. Price

Vanilla (185 x 4.6 ) (240 x 4,6) 851 2,503 1,104 3,308 + 29.7 + 32.2 Coffee 100 166 ]65 239 + 65.0 + 44,0 Cloves 110 172 320 791 + 190.9 + 359.9 Pepper 170 385 175 434 + 2.9 + 12.7

Paddy 12.6 65 /1 30 168 /1 + 138.0 + 158,5

LI The export price refers to 'de luxe' quality.

3.53 Traditionally,rice in Madagascar is mainly grown for subsistence: only 12% of total production in Madagascar was marketed in 1974, In 1973, all rice marketing was conferred to a central agency - SINPA (Societed'Interet National des Produits Agricoles). While previously private traders bandled the collection,processing and distributionof the crop, SINPA now has the monopoly to market rice and a range of other domestic crops, for which it establishesuniform producer prices and sets retail prices as well, - 47 -

3.54 In 1973, when decreasing local production for the market and soar- ing prices for imported rice were resulting in a rapid rise in food subsidies, the producer price for rice was doubled to 30 FMG/Kg. This made rice produc- tion for the market financiallymore attractive to peasant farmers than even before. But the difficultiesSINPA has been experiencingin building up an efficient marketing organizationhave been a serious handicap in moving increased production to urban centers of demand. Marketing inadequaciesthus threaten to jeopardize initially positive producer response to official price signals.

3.55 Even with the doubling of producer prices for rice between 1973 and 1974, Madagascar has remained a net importer of rice. Retail prices for this main staple were increased by 80% (from FMG 36/Kg to FMG 65/Kg) between 1973 and 1974, but still remained substantiallylower than the c.i.f. price for rice imports (FMG 93/Kg) which are subsidized by the government to protect the urban cost of living. Nevertheless, there has been a significant rise in the general (Antananarivo)cost of living index over the 1965-1975period which is estimated at 75%. 1/

Price of Inputs

3.56 All fertilizers,insecticides and herbicides,and agricultural machinery are imported into Madagascar. With the worsening terms of trade, especially following the oil-crisis, the volume of these imports declined drastically,even though expendituresstayed almost level. Table 10 below details price increases for major agriculturalinputs between 1970 and 1974: they averaged 60% for all fertilizers;38% for insecticides;86% for agri- culturAl machinery. (The increase in the general consumer price index for the same period was 44%).

1/ See IBRD Report No. 1099a - MAG Annex V, Table 9.1 The index refers to low-incomehouseholds in Antananarivo. - 48 -

Table 10

ComparativeValue and Volume of Imported AgriculturalInputs: 1970-74 (Currentprices c.i.fe)

Value (FMG mill.) Volume (tonnes) Price/t: FMG 000 Input 1970 1974 1970 1974 1970 1974

Fertilizers 598.3 404.8 30,346 12,760 19.7 31.7

Nitrogen 71.4 155.0 4,887 4,737 14.6 32.7 Phosphates 112.2 23.5 5,247 705 21.4 33.3 Potash 49.3 90.5 3,467 3,989 14.2 22.7 Compounds 353.1 112.9 15,849 2,856 22.3 39.5 Other (incl. guano) 12.3 22.9 896 474 13.7 48.3

Insecticides 355.9 425.1 1,399 1,210 254.4 351.3 Herbicides + fungicides 64.4 96.6 105 144 613.3 670.8 Rat poison etc. 4.1 13.5 10 19 410.0 710.5

Tractors 307.0 331.0 364 211 843.4 1,568.7 (wheeled)

Total 1,329.7 1,271.0

Derived from: Report No. 1099a - MAG, World Bank, December 1976, Annex V, Table 7.6.

3.57 The sharp reduction in volume of imported agriculturalinputs even more than their increasedprice suggests that there may be little chance of much of them being channelled into new local markets, when quantities im- ported are not even sufficient to satisfy previously existing demand. While the opening of the Andapa-Sambavaroad as well as the extensior.effort of the OPACA agriculturalproject would thus both be expected to contribute to increased use of modern inputs by area farmers in 1975, account must be taken of these 'external'determinants of modern input use in evaluatingtheir response.

3.58 In 1964/65 use of modern agriculturalinputs in .Andapawas certainly very low: the major share of expendituresfor agriculturalproduction were for wage, (36.4%)for animals (draught and breeding), (28.2%) and for tradi- tional farm implements,(8.1%). Only 2.8% were spent for chemical fertilizers - 49 - and modern agricultural equipment. 1/ (The remainder were payments for land and other miscellaneous expenditures). In 1964/65, wages for un- skilled labor in the Andapa region ranged between FMG 17-23/hr. in 1975, they attained FMG 36-49/hr (current prices) thus rising more steeply (112%) than the general cost of living index (75%). However, the increase in wages very closely parallels the increase in the producer price of paddy, the minimum wage per hour remaining slightly above the producer price of one Kilogram of paddy. (For comparative skilled labor wages and salary scales, see Table 15, Appendix B),

1/ Source: BCEOM 1966, Table 3.4.3. SECTION III B

ECONOMIC EXCHANGES BETWEEN THE REGION AND THE REST-OF-THE WORLD:

ANDAPA EXPORTS AND IMPORTS, 1965-1975

I 50 -

B. ECONOMIC EXCHANGESBETWEEN THE REGION AND THE REST-OF-THE-WORLD: EXPORTS/IMPORTS1965-1975

3.59 At the time of the base survey, practically all exports from, and imports into, the Andapa Basin were carried by airplane. The only important exception were cattle imports: these came into the area 'on the hoof' from the Ankaizina,a vast pasture area to the west of the Basin. Annual imports of beefcattle,both draught animals and cattle for slaughter, amounted to about 4,000 head, representinga value of about FMG 40 million. 1/

3.60 While short-distanceexchanges between the different cantons of the Basin (especiallythe rural communitiesand Andapa) were quite frequent (see BCEOM 1966, pp 5.69 - 5.86 for details) long-distanceporterage costs esti- mated at FMG 715/tonne kilometer 2/ made land transport of even high-value crops to the coastal ports uneconomical. Social/Administrativevisits (other than by plane) to Sambava or the district capital at Antalaha involved diffi- cult 2-4 day walks along 100-120kms of trails.

1. Air Exports/Imports: 1965

3.61 BCEOM's one-year complete record (August '64 - July '65) of all freight movements between Andapa and the rest of the country shows the follow- ing volumes of exports and imports:

1/ See BCEOM 1966, p. 5.88.

2/ See AGRER 1967, p. 77. - 51 -

Table 11

Volume and Value of Andapa Exports - 1964/65

I II III

Tonnage Exported to: Value of Exports Transport Charges AND - ANT/SAMB Antal. Samb. Other Total Mill FMG % FMG/tonne

Vanilla 301 - - 301 358.0 71.0 31,000

Coffee - 665 675 - 1,340 130.0 26.0 17,000

Other Agricul. prod. 20 58 20 98 2.9 0.5 19,000

Other 18 11 8 37 13.5 2,5 19,000

TOTAL (AVER.) 1,004 744 28 i,776 503.4 100.0

PERCENT 56 42 2 100

PASSENGER: One way ticket Return ticket

Source: I + II: BCEOM 1966, pp. 5.110 - 5.114 2,000

III: AGRER '64, p. 266 3,800

Note: Transport charges between Andapa and Antalaha, and Andapa and Sambava are the same as the two destinationsare practicallyequidistant from Andapa by air.

3.62 By far the highest export topnage (85% of total) is made up of coffee. But vanilla, with only 17% of tonnage, accounts for 71% of value exported. Almost all coffee exports from Andapa were made by private traders, while 70% of the vanilla crop was exported by cooperatives. Coffee exports - the highest in volume - were often delayed by 3-4 months due to the limited capacity of the DC3 planes serving Andapa. This added hioh storage fees to the already considerablehandling and air-transportcharges on the crop. The latter alone reduced farm gate prices for coffee in the Basin by FMG 18.50/Kg out of FMG 100/Kg guaranteed at port weighing station. 1/

1/ See: AGRER 1967, pp. 145/6. - 52 -

3.63 Eighty six percent of exports and imports were made by private traders. Large traders (importing over 50 t/p.a.) accounted for 57% of total imports, compared to 84% of total exports. This indicates a fairly high concentration of important marketing functions in a few firms.

3.64 Volume and value of imports recorded are considerably lower than those of exports, which exceed them by 36% and 67% respectively. As would be expected, imports are much more diverse in composition. (For itemized break- down see Appendix C Table 1). The highest single import in bulk is sugar - of which 122 tons were imported during the recording period. Carburants and lubricants are next with 102 tons, followed by cooking fuels (90 tons), salt (79 tons), and cement (76 tons). Textiles are the single highest value item imported. Antalaha and Sambava are the place of origin of 86X of the total tonnage imported to Andapa; the remaining 14%, coming mainly from Tananarive, do however account for over 40% of the total value of imported goods.

3.65 Reported air passenger movements 1/ are surprisingly high, averag- ing around 7,000 (arrivals and departures) annually for the 1955-1964 period. No indications on who these passengers are, or on purpose of trip is avail- able. But it seems surprising, that about ten people/day would make a trip costing FMG 3,800 (the average wage of an agricultural laborer in 1964/65 was FMG 120/day).

2. Road Traffic: 1975

3.66 At the time of the follow-up survey five years after the Andapa- Sambava road had been opened to traffic, there was no longer any regularly scheduled air connection from Andapa to the coast. All Andapa traffic moved over the new road-link to Sambava where the airport facilities had been improved to accommodate small jets.

3.67 A total of 34 transporters were registered with local authorities to provide freight and/or passenger transport for hire on the new road. Of these, 3 were no longer operating; 2 had merged their business, leaving a total of 30 transport(ers) firms active on the road. 29 of these were interviewed about their business (for questionnaire guide, see Annex G); one refused to cooperate in the survey.

The Local Road Transport Industry

3.68 Transporters regularly providing services along the Andapa-Sambava road are mostly located in the major towns: Sambava and Andapa. They handle either freight, or passengers. One of the freight transporters owned 3 vehicles and two operated 2 vehicles; the remainder were one vehicle opera- tions. Only 3 out of the 23 providing passe.agerservices owned 2 vehicles.

1/ See: AGRER 1964, pp. 311, 316. - 53 -

3.69 Of the 29 transporterssurveyed, twenty made their living exclu- sively from their transportbusiness. Of the remainder, 5 were also traders, 2 were tax-collectors,2 farmers.

3.70 Generally, buses and bush-taxis were owner-operatedexcept for second vehicles. By contrast, the majority of freight operators hired drivers to man their vehicles. In all, the 29 transportershad 21 employees - 4 of them part-timeworkers. Six of the 17 hired drivers were relatives of the vehicle owmers. Monthly salaries paid to drivers were between FMG 13,000 - 15,000.

3.71 The vehicle park is quite uniform: all trucks are Mercedes of 4.5 and 5t capacity,except for one Datsun pick-up (1.3t); only one truck (owned by a Sambava transporter)was purchased before 1970. The 19 buses are all 28 passenger vehicles: 11 Renault Saviem, 4 Renault SG2 and 4 Mercedes 508. 11 of these were purchased after the opening of the road. Taxis, with one exception (a Renault), were all Peugeots 404; all were bought second-hand at FMG 1,000,000or less: evidently these provide budding entrepreneursa first entry into the transport business.

Transporters'Operating Revenues and Costs

3.72 Revenues from, and expendituresfor, vehicle operation as reported by the transportersare summarized in Table 12 below (more detail is given in Tables 2-4 Appendix C). Adequately complete data are available for vehicles owned by 24 of the 29 transporterssurveyed. Vehicle operating costs include depreciationbut exclude returns to transporters. - 54 -

Table 12

TransDorters: Andapa-Sambava Road 1975 Revenues and Expenditures by Type of Service and Type of Vehicles

(FMG 000)

FREIGHT TRSPT PASSENGER TRSPT. ITEM TRUCK BUS TAXI-BE TOTAL

Number of Vehicles 10 15 6 31 Number of Trans- porters 6 13 5 24

Gross Revenue - 1974 141,732 82,173 8,222 232,127

Oper. Exp. - 1974 20,690 44,990 6,517 72,197

Net Revenue - 1974 121,042 37,183 1,705 159,930

Aver. Net Rev/Trsptr. 20,174 2,860 341 6,664

Aver. Gross Rev/Veh. 14,173 5,478 1,370 7,488

Aver. Oper. Exp./Veh. 2,069 2,999 1,086 2,329

Aver. Net Rev/Veh. 12,104 2,479 284 5,159

Aver*VOC/Km (FMG) 61 /1 32 19 39

/1 This average is based on 5 vehicles only.

3.73 Freight transporters report unusually large returns on their vehi- cles. Although the average age of the truck fleet is higher than that of the buses, average truck operating costs were well below those for buses, but reasonably in line with those reported elsewhere in Madagascar. 1/ Average

1/ See: IBRD Report No. 998a MAG, Annex p. 1, pp. 4-5. - 55 - net revenue per truck reported by respondents was 43 times that of the bush- taxis, and over 5 times that of the buses. 1/

3.74 Examination of vehicle operating costs for high-turnover vs. low- turnover transporters shows markedly higher costs for the latter group. This reflects both less efficient vehicle utilization and considerably higher maintenance costs for older vehicles (for derivation see: Table 2 Annex C).

Table 13

Transporters: Andapa-Sambava Road 1975

Average Annual Operating Revenues ard Costs per Operator and Vehicle, by Gross Revenue Class of Operator (FIG 000)

Rev. Class 1,999 2,000- 5,000 10,000- 20,000 Total Item 4,999 9,999 19,999 and over

N Operators 7 8 8 2 3 28

Of these operating:

Truck 1 4 - - 6 11 Buses - 5 9 4 - 18 Taxis 6 1 - - - 7

AVER,GROSS REV. - 1974 1,103 3,606 6,309 13,713 44,336 8,838 AVER.OPER. EXP. - 1974 900 2,467 3,249 8,220 4,579 2,936

AVER.O.Cost Per Veh. 1974 900 1,974 2,888 4,110 2,290 2,283

AVER.NET REVENUE - 1974 203 1,139 3,060 5,493 39,757 5,903

1/ It is unusual that truck operators should overstate their reveues, but if we assume their information to be correct, tariffs actually charged must be considerably higher than those quoted which are officially set by provincial authorities. For purposes of the economic analysis, transporter benefits were adjusted downward in line with volume and tariffs reported. - 56 -v

Freight and Passenger _Transport Tariffs

3.75 All survey respondonts said that they usually operate regular sclheduled rLns, Freight operators quoted tariffs by unit weight, and charges did not vary greatly between operators. There was, however marked differ- ence between tariffs on the Sambava-Andapa road and those charged on the poor roads inside the basin. The tariff for the 107 kilometer trip from Andapa to Sambava varied between FMG 2.5 - 3.0/Kg (FMG 23-28 per ton/Km); the c' rge for the 387 Km trip from Sambava to Diego Suarez was FMG 10/Kg (also FMC 25 tol/Km). On a 20 IKm ruin on earthroaas near Sambava, quoted rates were FIMC 1.0 - 1.5/Kg, equivalent to a cost per ton/Km of FMG 50-75. The truck tariff for moving one t-on()I freight from Ambalamanasy II to ANDAPA, also a 20 Km trip inside thie Basin is the same as that for the trip from Andapa to Sambava (107 Kin): FMC 2,500. This translates into ton/km rates of FMG 125 and FMC 25 respectively. These tariffs were confirmed by the 5 major area wholesalers who were al so asked about the transport charges they paid on goods exported/ impcrted. Costs per ton/Km on major (paved) roads were quoted at FMC 20-25; while Lhose on the poor earthroads in the Basin itself ranged from FMG 166-208. (See Table (, Appendix C for details).

3e76 Buses serve villages all along the main road between Andapa and Samnbava; sore even collt iniuc t:o and Antalaha. All operators charged the .aOTWfares, Thle L laxis tend to make shorter trips than buses. Fares reported for both buses and taxi ranged from FMG 150/person for a 40 Km trip, to FMG 10/person for a 6 Km local trip near Sambava. For tips inside the Basin where buses do not operate, taxis charged higher per kilometer prices to make up for higher operating cost and longer trip times. 1/ On average there are 6-8 passengers in a taxi-be per trip; they pay FMG 10/Km in the Basini coumpared to 1MG 3.5/Km on Lhe ANDAPA-SAMBAVA run. Sharper competition thus keeps the paEsenger fare-differential between paved and earthroads at 1:3, compared witlh J:5 for freight. Compared to the "without the road" sit- uation, the clharge for moving a ton of freight from Andapa to Sambava and Antalaha (expressed in 1975 prices) has on average declined by FMG 32,325 2/ and FMG 30,300 3/ respcctively; passenger fares (round trip) have been reduced by FMG 5,950 and FMC 5,250 for the same destinations.

Area Demand for Transport: 1975

3.77 None of the transporters interviewed expressed any concern about lack of business: demand for their services is high and allows especially truckers to cl_-irgevery remuinnerativerates. In passenger transport, charges are somewhat better a)igned witlh cost. The origin destination survey showedd (see paras. 3.90 ff below), that this has stimulated high 'people mobility' in the region.

1/ Sec Table 7, Appendix C mor details.

2/ Air tariff Andapa-Sambava FMC 20,000/ton x 175 35,000 - Road tariff 2,675.

3! Air tariff Andapa-Antala_ha FlMC20,000/ton x 1.75 35,000 - Road tariff 4,700. 577

Estimated Volume of Marketed Production

3.78 Andapa Basin demand for freight services can be roughly estimated from the reported vo3ume of marketed production. Oor crops sold outside the region, overall volume of products exported according to wholesalers' reports at least doubled (and perhaps quadrupled). Almost all of the volume increase is accounted for by rice exports. (See Table 14 below). There is, however, considerable uncertainty about the quantities of rice exported: no detailed figures could be obtained from SINPA, the monopoly marketing agency. If anything, rice exports seem underestimated significantly: assuming that per capita consumption of rice has not changed markedly from 1965 when it was 190 kg annually area exportable surplus in 1975 would have been of the order of 21,000 tonnes (paddy: 38,130 x 0.67 - 25,420 t rice nroduced; 1975 area population - 59,726 x 200 kg = 11,945 t rice consumed).

3.79 Reported transpcrt bottlenecks were mainly in moving paddy and other crops from the Basin to Andapa. Transport prices paid by SINPA were not sufficiently high to cover truckers operating costs on the poor local roads. By contrast, transport capacity on the Andapa-Sambava road was sufficient to have handled the volume.

3.80 No reliable information could be obtained about volume of imports which now flow into the region through a number of channels, rather than through one agent (AIR MADAGASCAR) only. A series of origin-destination surveys covering a full year period, supplemented by enquiries with major distributors in Antalaha/Sambava as well as in Andapa would be necessary to replicate the information collected for the base-survey. It can, however, be stated quite confidently that the volume of imports continues to be lower than the tonnage exported (see para. 3e89 below).

3.81 A survey of the four major wholesalers in Andapa tried to establish value of goods ex/imported by this group over the 1965/75 period. The earl- iest year for which complete information was available is 1971 - one year after the opening of the road. Imports valued at FMG 140 million and exports of FMG 146 million were handled by these firms then; the respective figures for 1974 were FMG 227 million of imports and FMG 230 million of exports. The Increase over the 4-year period is 62% and 58% respectively. Most of this probably reflects price inflation, rather than increased volume of goods moved (Cost-of-living-Index, Tananarive low-income households: 1970 = 100, 1974 = 144; import price-index, all products: 1970 = 100, 1974 = 193). 1/

1/ Source: IBRD, op. cit., Annex V, Table 9.5. - 58 -

Table 14

AgriculturalProduction and Volume Exported from Andapa: 1965-1975 (Tonnes)

1965 1975

Total Production

Paddy 14,430 38,130

Coffee 1,330 29285

Vanilla (green) 1,250 937

Cloves - 8

Reported Exports

Rice 80 2,000-4,000 (8,000)

Coffee 1,340 1,746 (2,290)

Vanilla (dry) 301 120 ( 240)

Cloves 8 18

Pepper - 2

Tonnage Exported 1,729 3,886-5,886 (10,250)

Source: 1965 - SERVICE DE L'AGRICULTUREand BCEOM 1966, p. 5.92. 1975 - Wholesale/ExporterSurvey (figures in parentheses Agriculture Service).

3082 From the earlier discussion on producer prices (paras. 3.51 ff above) and the above description of local marketing/transportingarrangements for different crops, it emerges clearly that the Andapa-Sambavaroad is not well suited to test Hypothesis 2, concerning the relative share of transport cost savings accruing to transportersand producers. Producer prices for all major crops are set nationally and apply, with only small variations, in all provinces. Provincial authoritiesalso set charges for transportingexport crops from various inland collection points to the nearest port. SINPA, the - 59 - monopoly agency marketing all rice and several other food crops, again sets producer and retail prices, as well as transport charges. Transporters thus do not directly face producers in offering their services; ra-her, they decide whether the tariffs set for a particular crop and a particular route will provide them an acceptable return or not. The producer who happens to be located along a route the transporter finds unpr-fitable to serve at the official rate therefore suffers not just a reduction in his farmgate price, but loses the opportunityto market his crops altogether (except by head- loading). High gross profits reported by freight transportersseem to in- dicate a high level of transport demand at remunerativerates in the region, and little necessity for truckers to compete for less attractivebusiness.

Results of Origin/DestinationSurveys - Andapa-SambavaRoad 1975

3.83 Hardly any informationhas been collected on the traffic carried by the Andapa-Sambavaroad since its official opening in August 1970. Traffic counts (near Sambava) were started in 1965 and repeated annually up to 1971 for the segments of the road open at the time. Over this seven-year period, average daily traffic on the road near Sambava grew from 102 to 178 vehicles (for details, see Table 8, Appendix C), much of it not doubt generated by the ongoing road constructionactivities. No figure for the years 1972 to-date are available.

3.84 To get some idea of the present volume and compositionof traffic on the road, an origin/destinationsurvey 1/ was undertaken during September/ November 1975, This is a dry season just after the harvest of the vanilla crop, when ireight traffic is average and passenger traffic perhaps higher than average. In total, the survey covered a period of 14 days, seven days at a counting point near Andapa, seven at a counting point near Sambava. Vehicles and passengers were counted between 6:00 g.m. and 9:00 p.m.; there is very little night travel except for emergencies. Counts were taken in three-and-four-dayclusters, covering the different days of the week evenly. Note was made of the types of vehicle passing the counting stations; their point of origin and destination;the freight and the number of passengers they carried; their origin and ultimate destination;and the purpose of their trip. Detailed results are presented in Annex C, Tables 10-13. These show passenger and vehicle informationfor each direction of flow at each counting point.

Vehicle Traffic

3.85 In all, some 2,178 vehicles were counted in the 14 day period. The following summary table gives a breakdown by vehicle type and counting station:

1/ Counting stations were located about 5 Km outside Andapa and 10 Km away from Sambava beyond the branchpoint of the Vohemar road. - 60 -

Table 15

Origin Destination Survey: Types of Vehicles Counted

COUNTING POINT SAMIBAVA ANDAPA TOTAL ADT VEHICLE TYPE N % N % N %

Private Auto 197 12 65 11 262 12 19

Taxi-Be 377 24 75 13 452 21 32

Minibus/Bus 113 7 113 19 226 10 16

Pickup Truck 414 26 228 39 642 30 46

Truck 179 11 67 12 246 11 18

Motorcycle/Bicycle 312 20 37 6 349 16 25

Oxcart - - 1 - i - -

Total 1,592 100 586 100 2,178 100 156

3.86 The vehicle count near Sambava was much higher than that near Andapa, accounting for some 73 percent of total vehicle traffic. The average daily traffic (adt) was calculated to be 227 vehicles at the Sambava counting station and 84 vehicles near Andapa. Observed adt for the entire road was 156 vehicles. Sambava clearly evidenced a higher proportion of taxi and motorcycle traffic to total traffic than Andapa, while Andapa had a high proportion of small buses and pickup trucks, vehicles capable of carrying more passengers.

3.87 Looking at origin and destination of vehicle trips by type of ve- hicle, traffic was broken down into local (short distance trips near the end points of the road) and through-traffic (trips which traverse the greater part of the road and may continue beyond it. Traffic to mid-road destinations accounts for remainder not included in Table 16 below). Taxis, pickup trucks, and motorcycles or bicycles were clearly most often used for local travel. Private autos were used for local travel around the endpoints of the route as well as for long-distance trips. Trucks tended to travel the whole distance of the route, and small buses were almost exclusively used for through travel.

I - 61 -

Table 16

Length of Trip by Vehicle Type: 1975

Length of trip Local Traffic Thru-traffic Vehicle Type N % of all traffic N % of all traffic

Private Auto 141 53% 103 39%

Taxi-Be 328 73% 95 21%

MinibusiBus 19 12% 178 79%

Pickup Truck 383 60% 201 31%

Truck 85 35% 133 54%

Motorcycle/ Bicycle 267 77% 27 8%

Total 1,223 56% 737 34%

3.88 The week days for which movements towards Sambava were highest are Mondays and Fridays; towards Andapa, movements peaked on Tuesdays and Fridays. Vehicle flow was clearly lowest on Sundays.

Table 17

Traffic Flow by Weekday: 1975 (N for two-day count)

Flow Day Towards Andapa Towards Sambava Total

Monday 157 180 337

Tuesday 176 155 331

Wednesday 142 134 276

Thursday 164 138 302

Friday 181 172 353

Saturday 163 155 318

Sunday 116 116 232

TOTAL 1,099 1,079 2,178 - 62 -

3.89 During the 14-day survey period, a total of 424 tonnes of freight moved along the road. 78% of this was long-distance(thru-traffic) between Andapa-Sambava;the remainder 'feeder' traffic around the endpoints of the road. Forty percent of total flow was towards Andapa; just under two-thirds of the products were transportedby truck, compared to 85% of the freight botnd for Sambava. Coffee, paddy and rice accounted for 83% of freight volume towards Sambava. 1/ 'Other goods' and 'other foods,' followed by wood, combustiblesand constructionmaterials are transportedtowards Andapa. The observed volume of 30.3 tonnes/day during this period can be considereda good approximationof "average" daily volume: this would be higher after the rice harvest (June-August),lower during the rainy season (December-March).

1/ Note that at the time of the survey the rice-mill that handles the paddy produced in the Basin was still located at Sambava. This necessitated some double transport- of paddy to Sambava and of rice back to Andapa. The transfer of the rice-mill to Andapa is foreseen for the near future - tonnage projectionsfor the economic analysis are therefore for rice, not paddy. - 63 -

Table 18

Freight Transported along the Road: 1975 (in tonnes, 14 days - SEP NOV)

0 N V C SN C A 0 T P 0 N 0 M 0 T A R F I T W B M T 0 D I F L H 0 U T H T D C E L E 0 S L E A Traffic Y E E A R D T S R L TOWARDS SAMBAVA

Truck 53.5 21.9 111.3 2.r 3.1 0.6 1.5 4.9 17.0 215.8

Pick-up 6.9 6.5 8.4 3.2 4.2 2.4 0.5 1.0 2.1 35.2

Other - 0.5 - 0.4 - - 0.9 0.1 - 1.9

SUB-TOTAL 60.4 28.9 119.7 5.6 7.3 3.0 2.9 6.0 19.1 252.9

TOWARDSANDAPA

Truck - - - - 32.3 14.1 12.7 15.9 33.7 108.7

Pick-up - 0.7 - 1.0 12.3 16.2 10.0 2.5 5.9 48.6

Other - - - 0.2 1.5 - 0.8 0.8 10.5 13.8

SUB-TOTAL - 0.7 - 1.2 46.1 130.3 23.5 19.2 50.1 171.1

COMBINED

Truck 53.5 21.9 111.3 2.0 35.4 14.7 14.1 20.8 50.7 324.5

Pick-up 6.9 7.2 8.4 4.2 16.5 18.6 10.6 3.5 8.0 83.8

Other - 0.5 _ 0.6 1.5 - 1.7 0.9 10.5 15.7

GRANDTOTAL 60.4 29.6 119.7 6.8 53.4 33.3 26.4 25.2 69.2 424.0

______11_ _ _ _ _1~ _ __ _ _ I______I______I______- 64 -

Passenger Traffic

3.90 Traffic during the survey period was composed of only 28% freight traffic, 56% passenger vehicle traffic, and 16% motorcycle (local) traffic. Passenger vehicles clearly satisfiedkeen demand: a total of 13,179 pas- sengers were counted on the road dur.-ngthe 14 day survey period. This is equivalent to 10.5 passengers per passenger vehiclc, assuming that half of the pick-ups carry passenger traffic. Of the total passengers, 65 percent were counted at the Sambava checkpoint, 35 percent at the Andapa one. The biggest group of travellersmentioned 'social visits' as the purpose of their trip (31%), followed by 'work' (13%), 'trade' (10%) and attendauce of traditional ceremonies (9%).

Table 19

Purpose of Passenger Trips: 1975

COUNTING POINT ANDAPA SAMBAVA TOTAL

REASON N % N % N %

Social Visits 1,338 29 2,731 32 4,069 3)

Work 561 12 1,165 13 1,726 13

Trade 376 8 929 11 1,305 19

Traditional Ceremony 421 9 750 9 1,171 9

Other 1,969 42 2,939 35 4,908 37

TOTAL 4,665 100 8,514 100 13,179 100

3M91 Many of those travelling for reasons of work were local travellets. Sambava had more local travellers and lower estimated passengervehicle load- ings: rough estimates show 9.5 persons per passengervehicle in Sambava and 12.7 persons per passengervehicle in Andapa than Sambava. Table 20 below gives a breakdown of length of trip by purpose. (The designations'local' and 'through'are the same as those used in the vehicular summary table; passengers not accounted for gave mid-road destinations). - 65 -

Table 20

Length of Trip by Purpose of Travel: 1975

PASSENGER LOCAL TRAFFIC THRU-TRAFFIC

REASON N h of otal N % of Total

Social Visits 1,771 46 1,652 43

Work 959 56 546 32

Trade 199 38 565 43

Traditional Ceremony 358 31 473 40

Other 2,259 46 1,830 37

TOTAL 5,843 44 5,066 38

3.92 Clearly, as with the vehicle counts there was more local passenger traffic than through traffic. While 56% of total vehicles travelled short distances in and around Andapa and Sambava, only 44 percent of the passengers fell into this category. About 34 percent of total vehicles counted travelled over most of the route, and 38 percent of total passengers. A rough calcula- tion shows estimated passenger vehicle loadings of 8.5 persons for local traffic and 10.6 persons for through traffic. Trips for purposes of work and those unspecified were mostly local, while trade trips and those taKen for the sake of official or traditional ceremonies tended to be long trips.

3.93 Table 21 below shows total passenger flow, distributed over the days of the weeks. Passenger patterns were much like the vehicle patterns summarized earlier. Travel towards Sambava was heavy on Monday, while travel in the opposite direction was heavier on Tuesday and Wednesday. Passenger flow in the later half of the week, Thursday through Saturday is heavier, with travel both towards Andapa and Sambava peaking on Friday: Saturday is market ddy in bot1. these tc:wns. - 66 -

Table 21

Passenger Flow by Weekday: 1975

DIRECTION Daily WEEKDAY Towards Andapa Toward Sambava Total Average

Monday 834 968 1,802 901

Tuesday 896 822 1,718 859

Wednesday 947 892 1,839 920

Thursday 1,003 928 1,931 965

Friday 1,247 1,243 2,490 1,245

Saturday 905 922 1,827 913

Sunday 617 955 1,572 786

Total 6,449 6,730 13,179 941

3.94 While there was practicallyno traffic of animal-drawnvehicles along the road, pedestrians used it quite heavily. An average of 265 people/ day were counted at the census point near Andapa. 112 of these were going to and from the wood to gather firewood. Between the villages along the road, a similar pattern prevailed; there, large groups of children walking to school were also regularlyobserved. Quite evidently, the road thus seems to have been "adopted" by the communitiesit was built to serve. Of all 773 respon- dents in the household survey, only one said the road was 'bad' - feeling it had increased the cost of living. 43 thought it had not made much difference but the great majority said it had been a positive development. SECTION III C

THE ECONOMIC EVALUATION OF THE ROAD INVESTMENT - 67 -

C. ECONOMIC EVALUATION OF THE ROAD INVESTMENT

3.95 As stated in Section II D (paras. 2.54 ff) above, the data collected for this study permit measuring the economic impact of the road "at the level of the farm" as well as "at the level of the road," i.e. project rates of re- turn will be established applying both the producer surplus (PS) and the road user savings (RUS) method. both approaches take as their point of departure changing agriculturalproduction in the area - observed changes for the 1965 - 1975 period; estimated or projected changes for the remaining project life.

1. Patterns of Crop Production

3.96 Patterns of crop production in the area were described in detail in Section III A above (paras. 3.41 ff). The 1965 and 1974 secondary statistical data on area cultivated,yields and producer prices of major crops are briefly resummarizedin Table 22 below.

Table 22

Area Cultivated,Yields, and Producer Prices for Major Andapa Crops: Percent Increase 1965, 1974

Producer Price Crop Area (ha) Yields (t/ha) % change 1965-1974 1965 1975 % change 1965 1975 % chan&e Current Constant '74

Rice 4,830 17,340 /1 + 259.0 2.98 2.19 - 26.5 + 138.0 + 50.0

Vanilla 4,000 3,910 /2 - 2.25 0.31 0.24 - 22.6 + 29.7 - 19.7

Coffee 2,800 3,950 /2 + 41.0 0.48 0.58 + 20.8 + 65.0 + 1.9

Other 1,270 2,790 + 119.7 n.a. n.a. - + 190.9 + 80.0 (- cloves) TOTAL 12,900 27,990 + 117.0 n.a. n.a.

L 1974 includes tavy = rainfed rice grown on mountainsides. ft Some of the differentialmay result from difference assumptions in converting trees to hectares.

3.97 The chapter on 'method' (paras. 2.7 ff above) presented the design of the 1965 and 1975 sample surveys, the results of which provide the basis for deriving the share of productionmarketed and self-consumed,used to measure the impact of the road investment. The surveys clearly indicate - 68 -

changes i.L. Lroppingpatterns. There is a noticeable shift to more "special- ization" i, the cultivationof crops. While a majority of farmers in both samples cultivated all three of the major crops, the majority declines from 822 in 1965 to 57% in 1975. The change is attributableexclusively to new patterns of production for the traditional cash-crops. The proportion of the sample growing rice remains an even 86% and 85% in 1965 and 1975 res- pectively, even though the chances of marketing rice at competitiveprices now exist. Cash crop production in the villages along the road is signifi- cantly lower than in the Basin because of different climatic conditions. For this reason, the 'road sample' households are shown separately in the comparison presented in Table 23 below. Quantities produced, self-consumed and sold, and (producer)prices reported by the sample are shown in Table 24. Figures for the "panel" households included in both surveys are shown sepa- rately: they served in several instances to estimate total sample "missing" values.

Table 23

Cropping Patterns at Farm Level - 1965-1975: Percent of Households Growing Different Crops ROAD /2 ANDAPA BASIN Type of Cultivation 196y5/ 1975 1975

RICE, COFFEE, VANILLA 82.2 57.3 41.3 RICE, COFFEE, no vanilla - 16.2 8.7 RICE, VANILLA, no coffee - 5.3 10.0 COFFEE, VANILLA, no rice 4.4 2.7 3.8 COFFEE only 0.5 1.7 1.2 VANILLA only 2.0 0.4 2.5 RICE only 3.9 6.4 25.0 Sub-Total 93.0 90.0 92.5

Other Crops only - - 2.5 Sub-Total 93.0 90.0 95.0

No Crops 7.0 10.0 5.0

TOTAL 100.0 100.0 100.0

OF THESE: PERCENT

CULTIVATINGRICE 86.1 85.2 85.0 CULTIVATINGCOFFEE 87.1 77.9 55.0 CULTIVATING VANILLA 88.6 65.7 57.4 N Households 202 693 80

/I SOURCE: BCEOM 1966, Table 4.1.1. L Note that the 20 households along the road near Andapa, which belong to the Andapa subdistrictare included in the ANDAPA BASIN sub-total. rable 24

YIELLS, DISTRXBUTION OF OUTPUT, PRODUCER PRICES REPORTED BY SAMOPLEAND PANEL hDUSEHfELDS: 1975

I.Area t*.1lt'.aaed \'iel.s 1total lutp'.t .utpur Sold 0o%tplt Cor, awed tutpu;t tored jUtpnt r$InPI To'?E kt r. arec) Not A,co-vnted Prices leceived by 1973 | in 175 I (1i7 Kg I V.' r.. :r For 11965 ' n Kg.' Farmers (Fl(KS. 197, 19tq75 1975 j 197i Y.'', 1965 1975 1965 J975 1975

.^nll11 t - ¢'r: z1 R.aRie a 32,487 5% 32,7?f6 79. 129 Pf .\1 :9.603 - :, 33 3.?, 0 341 21,60 ? 77 59792 5'.450 , .13' 322 ''1ee - Tobo1 San lI- "a 43,503 3.09 334,47. 1?9.610 27,887 30.191 992559 .432 165 ' s2 2.3 'tbor| Ca'hE- - lr:.41 Fa:nple a 1,437 | 203.03'i1 291.702 2>3,902 1 71?,3',', | 1 ar:- I 778 9738 219.0 '33940 1, 1; 21.,857 1I4,t52 ; ,CWC 9.749 4, 213 T-otsl -ample a 78,659 13.0 1,024,938 .14,790 591,974 1 208.271 Pan1.-i 1 43,973 50,02'3 9 '' I 34 14.2 712,938 130,336 421,7A4 I 101.082 9,726 ;4 Othor roods - - Toral Inmple a b 15,598 40 0 624,342 381,949 142,945 a'11 I 95,044 4,404 Oa10,249 11,326 41 3 467,455 3S8,136 47 107,304 '7,964 '5.949C 49

1'otal amSrple a 171,684 d 2,158,225 1,029,580 751,349 - Panel 102.490 114,127 320,224 57 072 d 1,552,067 795,386 539,681 '71,681 45.319 d

Notes: a. Not applirable since sample designs in 1964 and 1975 were dissimilar. 4 h. Estimated via 1 E C -where C = crop p category in 1965 rond P - the ratio 75 . 1,65, 1,65 75 4 5 7 (E7 I . P Z 0.9. All 7 5 this is for the data on original panel households only.

The area cultiv.ted is derived from total output. The yield of 203 is estimated as 0.93 (218.64), where 0.93 - the approximate average ratio of 1975 yields (total sample: original 218.64 HH), and = the 1975 yields for panel households only in 1975.

d. Not applicable.

e. Output sold but not yet paid due to difficulties with new cooperative organi:atlon.

Note: 1 are = 100 square meters 1 ton - 1000 kg

j' - 70 -

2. The Economic Benefits of the Road: Iw Yrodlicer St:rplusAnalysis

3.98 The economic benefits of thieroad will bo estimated in the first instance by measuring economic producer surplus in the "with" and "without" situation in 1975. I/ Only producer surplus £ >icultural!o. production is considered,as agriculture is by far the most important acivity in the region's economy. A majority of even those households that give crafts, trale or salaried employment as their main ocetupationalso cultivate crops. All estimates are based on market prices for crops and inputs, because available informationdid not allow shadow pricing. However, private and social producer surplus should not differ by a great margin: land is, as yet, no real constraint,so agriculturallabor in the area is employed almost fully producing rice for self-consumptionand vanilla and coffee for the market; and use of modern inputs is negligible.

3.99 The data base recuired for carrying out a producer surplus ana- lysis is not easily established. Secondary data on agriculturalproduction and yields are often of doubtful accuracy. Interview results on quanti- tative parameters are likely to reflect respondentbias in varying degree: informationon areas of land owned and cultivated;types and amounts of crops produced, sold, and self-consumed;levels and types of incomes earned, ani expendituresincurred for production (and household consump- tion) is likely to indicate orders of magnitude rather than exact measures, even when data collection is carried out meticulouslyand field control is exemplary. However, physical measures of area under crop, whether on the ground or via air-photo imagery, dS well as sample yield measures by trained personnel, while highly desirable are expensive and difficult to obtain, particularlyin areas of mixed tree cropping as in Andapa. Especially when the time available for survey work is less thatia full crop season, the farmers themselvesare the only source of information on which to rely.

3.100 This was true in the present studv, where summary sample data on proportion of crops sold and self-consumed(Table 24 above) were com- bined with available statistics on area under crop and production to serve as the main basis for the PS analysis. Given the uncertaintiesinvolved in translatingnumber of trees planted into hectare equivalents,average yields reported by the sample are reasonablyclose to those shown in Table 22 above. 2/ Average per kilogram price realized for different outputs reported

1/ Producer surplus is defined as the difference between income from production and cost of bought inputs, and thus represents the gross rtturn to family labor and owned land.

2/ Sample data allocated full acreage to each crop when interplanting was reported, this tends to downward-biassample per hectare yields. - 71 - by sample households again is very close to fixed producer prices. (The difference for vanilla probably results f:om a mixture, in reported sales, of previous years' stores of dry vanilla and current crop). Sample data were also used to estimate 1975 and 1965 area production expenditures. Table 25 below summarizes values of production,production expenditures and producer surplus reported by sample households, 1/, on which estimates are based.

3.101 Producer surplus for each major crop i -"with project" situa- tion was calculated by establishing 1975 sample f duction expendituresas a percentage of the value of production for each crop and applying this fraction to total area production reported. 1965 sample production expen- diture shares were calculated similarly.

3.102 To estimate producer surplus "without the project", it was assumed that 1965 area production wou±d have grown proportionateto popu- lation. Producer priceo, being government controlled,2/ would have been the same as in the with-projectsituation. With respect to yields, assump- tions vary by crop. Basically, little change has occurred in local pro- duction techniques employed in 1975 compared to those used in 1965. 3/ Even with the project, use of modern inputs remains negligible,although more people own plows and harrows than in the base survey period when hand-tools predominated. Reported lower yields for all products except coffee seem to indicate that poorer quality land is being brought into cultivationas popu- lation and marketing opportunitiesexpand. For rice, this would not be true in the without project situation:itis thus assumed that average without project yields would have remained at 1965 level. Improved coffee yields 'with the project' are assumed to be the result of extension efforts under the project; without project yields again are estimated to have remained at the 1965 level. For vanilla, facing a contractingmarket, with and without project production is assumed equal.

I/ In both the 1965 and in the 1975 sample surveys, cash incomes from different crops were reported;so were self-consumptionand total cost of agriculturalproduction. In addition, the 1975 survey provided figures on la'.. inputs by crop. Total production expenditureswere allocated to crops proportionate to repcrted labor inputs per crop for 1975. Assuming that production techniques remained basically unchanged between 1965 and 1975, estimates of 1965 production expenditureby crop were then derived by multiplying 1965 reported income per crop with the 1975 ratio of production expendituresto crop income (= estimated 1965 per crop expenditurebased on 1975 prices and techniques),and adjust- ing the results to 1965 price levels by multiplying this value with the ratio of the sum of estimated crop expendituresto actually reported total production expendituresin 1965. 2/ See paras. 3.51 ff above. 3/ See chapter on agriculturalproduction techniques,paras. 4.66 ff below. - 7i -

Table 25

SAMPLE PRODUCER SUFPLUS BY CROP - 1965,1975 (Current FMG 000)

;rops 1965 1975

1. Vanilla - Cash Income Rcd. 7,770) 27,045 t It" Due 2,571) - Self Consumption _ _ - Total Income 10,341 27,045 - Prod. Expenses 585 1,557 - Prod. Surplus 9,756 25,488

2. Coffee - Cash Income 2,884 21,437 - Self Consumption ___ - Total Income 2,884 21,437 - Prod. Expenses 143 885 - Prod. Surplus 2,741 20,552

3. Other Cash - Cash Income 61 2,452 - Self Consumption _ _ - Total Income 61 2,452 - Prod. Expenses 4 126 - Prod. Surplus 57 2,326

4. Rice - Cash Income 719 7,247 - Self Consumption 12,001 39,883 - Total Income 12,720 47,130 - Prod. Expenses 2,642 7,416 - Prod. Surplus 10,078 39,714

5. Other Foods - Cash lncome 960 18,089 - Self Consumption 10,125 3,710 - Total Income 11,085 21,799 - Prod. Expenses 177 273 - Prod. Surplus 10,908 21,526

6. All Crops - Cash Income 14,965 76,270 - Self Consumption 22,126 43,593 - Total Income 37,091 ].19,863 - Prod. Expenses 3,551 10,257 - Prod. Surplus 33,540 109,606

NOTE: Self-consumptionof own-producedcrops has been valued at retail price level. - 73 -

3.103 With respect to production expenditures, the project has had no impact on vanilla production: with and without project production expendi- tures are thus assumed equal. For coffee, yields have increased as a consequence of the OPACA extension efforts: production costs per ton thus decrease in the with project case, and are assumed to remain at 1965 level without the project. For rice, without project production expenditure are also estimated to stay at 1965 level; the beginning of double cropping with the project ;aa attsndant higher yields are assumed to have resulted in the lower per ton production expendituresin 1975.

3.104 Little information is available on the share of such crops as ginger, pepper, cloves and sugar cane in the total value of 'other cash crops' reported by the sample. Producer surplus on cloves only was cal- culated for this category, which is thus under-estimated.

3.105 Regarding the share of Production sold in the market, all vanilla, coffee and cloves are exported in both the with and without project case. The 1975 volume could have been handled by air as well as by road. As to rice, sample data indicate that the share of the crop marketed (locally) in 1965 was 9.5% of total production; that in 1975 grew to 26%, a consider- able part of it exported. Export of rice in appreciable volume would have been physically impossible as well as uneconomic with existing air transport facilities; producer surplus on exported rice is thus clearly imputable to the project. In the without project case, it is assumed that the share sold in local markets would have remained the same as in 1965.

3.106 The main assumptions underlying the estimation of producer surplus "with" and "without" the project and the rate of return estimates are detai.i,d 5- each crop in Annex D, Tables 12-15. Comparative "with" and "without" project estimates of producer surplus in 1975 are shown in Table 26 below.

Table 26: ESTIMATED AREA PRODUCER SURPLUS BY CROP - 1975 WITHOUT PROJECT VS. WITH PROJECT (Constant 1975 FMG '000)

1975 Value of Production Crop Without Project With Prolect

Vanilla 212,240 212,240 Coffee 275,683 362,400 Other Cash Crops (cloves) 2,990 5,985 Rice a /1 392,650 901,943 b /2 608,880 1,275,816 Total a)/1 883,563 1,482,568 b)/2 1,099,793 1,856,441

Incremental Producer Surplus /a 599,005 /b 756,648

/1 Self-consumptionvalued at producer price. /2 Self-consumptionvalued at retail price. - 74 -

Total agriculture producer surplus witn the project in 1975 thus exceeds without project surplus by FMG 599.0 - 756.6 million, depending on the valuation of self-consumed goods adopted. This translates into a per capita increase of producer surplus of between FMG 10,000-13,000.

3.107 The net increase in agriculturalproducer surplus does not measure the full contribution of the road and agriculture projects to the Malagasy economy. To establish this, three additional components should be added. The first is import subsidy savings on rice. As was pointed out in Section IIIA para. 3.54 above, Madagascar has been a net importer of rice for a number of years and was buying rice in the world markat at prices substantiallyabove those established locally: 1975 subsidy per ton of rice imported amounted to FMG 28,000. Insofar as the rice exported from Andapa substitutes for foreign rice imports, the net benefits of the project are the difference between the value of the production at internationalprices and the social cost of produc- tion. This difference can be broken down into the difference between the value of production at local prices and the social cost of production on the one hand and the difference between the value of production at international prices and that at local prices. The first component is the producer surplus which has been already estimated, while the second is the saving in the amount of the government subsidy which we are now including.

3.108 World market prices of rice were at unsually high levels in the 1974-;5 period when rice first began to be exported from Andapa, and are expected to fall in the longer run, although not by more than 50% of the increase since 1972. The analysis thus assumes subsidy savings of FMG 28,000/t for 1974-76, and of FMG 14,000/t for the remaining project life. Again, no adjustment for a shadow exchange rate has been made. If - as is likely - this is higher than the official rate of exchange, project benefits would be underestimatedsomewhat. Depending on the assumptions retained about tonnage of rice exported, a minimum and maximum value for 1975 benefit from this source would be:

FMG 000

Share exported, min. 3,000 t paddy x 0.67 - 2,000 t rice - 56,000 Share exported, max. 12,000 t of x 0.67 = 8,000 t rice - 224,000

Share exported retained in the rate of return analyses (see paras. 3.114 ff below) was 6,660 tons of rice, with associated subsidy savings of FMG 186.5 million.

3.109 Secondly, the road project has led to important savings on non- production related trips which, unlike production related trip savings, are not accounted for in the producer surplus measure. The results of the origin/destinationsurvey (Section IIIB, paras. 390 ff above) showed that non-productionrelated passenger trips constitute a majority of about 75% of all reported trips. For 1975, an average of 102,400 passenger trips from Andapa to Sambava and Antalaha were estimated compared to 7,000 air _ 75 - passenger trips in 1965. The volume of air passenger travel having remained very stable between 1960-70, the same trip number is assumed for the 1975 without situation. The following savings can thus be estimated (see para. 3.123 below for derivation):

Long-Distance Trips /2 LI Constant 1975 FMG '000 (a) (b) (a) /2 (b) /1 To Sambava: Bus 64,800 x 107 Km x 2.35 x 3.25 16,294 22,534 Taxi 7.200 x 107 Km x 2.05 x 3.25 1.579 2,504 Sub-total 72,000 17,873 25,038

Average cost per trip - road - Sambava 248 348

To Antalaha - Bus 27,360 x 188 Km x 2.35 x 3.25 12,088 16,717

- Taxi 3,040 x 188 Km x 2.05 x 3.25 1,172 1.857

Sub-total 30,400 13,260 18,574

Cost per trip - road - Antalaha 436 611

Cost per trip - air - both 3,395 /3 3,500

Saving per trip - Sambava 3,147 3,152

Saving per trip - Antalaha 2,959 2,889

Passenger Savings on normal traffic:

3,500 trips x 3,147 (Sambava) 11,015 3,500 trips x 2,959 (Antalaha) 10,357 21,372 Passenger Savings on generated traffic:

(68,500 trips x 3,152) x 0.5 (Sambava) 107,956 (26,900 trips x 2,889) x 0.5 (Antalaha) 38,857 68,500 trips x (348 - 248) (Sambava) 6,850 26,900 trips x (611 - 436) (Antalaha) 4,707

158,742

Total Passenger Savings (Long distance traffic) 179,742 Correction for non-production related trips (x 0.75) 134,807

/1 Road traffic charged in 1975: FMG/passenger/kilometer. /2 Road traffic charged in 1975 net of indirect tax portion of vehicle operating cost FMG/pass/km. /3 Air tariff charged in 1965, expressed in 1975 constant prices net of indirect tax portion of operating cost. - 76 -

3.110 For existing or normal traffic the full transport cost savings are claimed: trip makers realize an amount of consumer surplus equal to these savings. For generated traffic (the well-known triangle under the demand curve 1/), only 50% of transport cost savings are taken into account as usual. In addition to savinigson long distance trips there are short distance trip savings, calculated as the difference between the value of the trip hours x hourly wages (in constant '75 prices) and bus or taxi-travel. The total 1975 savings (see computer output, Appendix D) on short distance traffic corrected for non-production related trips are:

(FMG '000) 21,071 x 0.75 = 15.803

Passenger savings in non-production related travel thus add to a substantial (FMG'000) 150,610.

3.111 The third benefit component that remains outside the producer surplus calculus is the transporters' surplus. In the 'without project' case this vould be the difference between income and operating expenditures (net of taxes) of AIRMADAGASCAR for the Andapa-Sambava line. No data are available on which to base an estimate. However, state-o!ned airlines more often than not are heavily subsidized rather than surplus producing entities. It would therefore seem conservative to assume a break-even (or zero surplus) situation for the without project case.

3.112 In the "with project" case, transporter benefits were established from a survey of local transporters (Section IIIB, paras. 3.68 ff above). The reported difference between total revenue and vehicle operating costs was unrealistically high for freight transporters (FMG 12.1 Million per vehicle in 1974) and was adjusted downward in line with reported volume transported at reported tariffs (to FMG 2.9 millions per vehicle). For consistency, passenger transporters surplus was calculated in similar fashion. Adjusting from financial to economic values by netting out taxes, transporter benefits were calculated based on estimated freight and passen- ger volumes over the project life (for detailed volume estimates see tables 5-9 Appendix D) and added to other project benefits,

3.113 Total economic benefits of the road and agriculture projects in 1975 can now be summarized as follows:

1/ See for instance Carnemark, C. et. al: The Economic Analysis of Rural Road Projects, World Bank Staff Working Paper 241, p.6. - 77 -

1975 Source of ButiefitS Amount of Benefits (constant 1975 FMG '000)

Minimum Maximum

Producer surplus on agrIcultural production:

a) self-consumption ( producer price 599,000

b) self-consurnlptioui( retail price - 756,640

Import subsidy savings I'SS) 185,565 185,565

Non-production related passenger trips 150,610 150,610

Transporters' surplus - Freight 19,630 19,630

- Passenger 17,160 17,160

Total 971,965 1,129,605

Production related benefits (PS plus ISS) account for the bulk of the benefits (80% or 83% of the total depending on how self-consumption is valued). Of this, 61% (or 67%) are appropriated by local producers, while the rest accrue to the Malagasy economy as a whole. Non-production related passenger trips account for a sizeable 15% (or 13%) of project package benefits, and local transporters appropriate the remaining three to four percent.

3. The Economic Benefits of the Road: The Road User Savings Analysis

3.114 The more traditional approach to the economic evaluation of a road project is to measure benefits 'at the level of the road' by establish- ing user savings in the with vs. without project situation. This will be done next for the Andapa-Sambava road and results compared with those from the producer surplus analysis.

3.115 RUS benefits are measured in terms of cost savings on freight and passenger transport for normal and generated traffic (there is do diverted traffic as there are no alternative road or other transport connections). Transport cost savings were calculated as the difference between freight and passenger air tariffs (1964/5, expressed in constant 1975 prices) and road transport charges (1975) both adjusted from financial to economic values.

3.116 Projections of future traffic growth were based on the patterns observed since the road first opened (1970-1975), and are essentially a function of projected agricultural production and the expected share of exports. Passenger traffic (air) in the without project case, is assumed to remain stable as in the preceding 10 yc,r c-r-; so is "'midroad pedestrian" traffic vrom locations alaong tic traicrc of the new road towards its endroints in Andapa and Sambava, boL; distrlct administrative centers. Total pas. trtwith the prolect grow from 17,000 to 278,000 in the year the road first oDens to tra£fic 1/; iro1 ;975 on the estimated rate of growtb i. '!.C{% p.a. cmewIarahovf oh' population growth to reflect increasing ujobility. (For year-by-year ultiimates, see table 7, Appendix D).

3,117 llfgZht_Iraffic in the withoul ro_ct c5se was assumed to grow as detailed for the producer surplus aialyss. ;it.h the project, the overall implied growth rate of freight traffic is a high ]6% p.a. It has to be kept in mind, however, thiat traditional exports increase at a rate of 8% p.a. only, (mostly due to increased coffee exports). The jump in order of mag- nitude in volume exported is due exclusively to exports of rice made possible by the irrigation/drainage improvements of the agriculture project and the economic access to market provided by the road. (Detailed year-by-year esti- mates of tonnage imported and exported are presented in tables 5 and 6 of Appendix D).

3.118 Data and assumptions underlying the estimate of transport cost savings associated with the project are detailed in Appendix D, tables 9-11. Information on transport tariffs and operating cost for the before/without project situation is derived from a variety of sources, not all of them equally reliable. The corresponding values for the with project case were established from transporter, trader, and origin/destination surveys in the study area as detailed in Section IIIB above; they also took into account findings fr:ml other road projects being implemented in Madagascar. As pre- viously, the ^onsumer price index was used to adjust 1965 values to constant 1975 prices, the deflator used being 1.75.

3.119 To establish 1975 transport cost savings without and with the proj- ect for comparison with the producer surplus estimate, export volume without the project entimated in the producer surplus analysis is assumed to be flown out or Aidapa to Sambava (most of the coffee) and Antalaha (most of the vanilla) by D(C3. In the absence of any operating cost information from AIRMADAGASCI:.r break-even operation on the Andapa-Sambava line was assumed, i.e 0 operating costs equal average revenue of a full plane load. Experts estlmatoe -hat indirect taxes represent about 3% of financial operat- ing costt this was the factor used in adjusting financial to economic values.

3.120 Witih-project exports, again as estimated in the producer surplus analysis, lenv, the srea by road. Total freight traffic volume (inclulding imports) s, d>idd hetleen Samhava (75%) and Antalaha (25%). In addition,

1/ Estimate based on Ministry of Public Works traffic counts; this figure is higher Ghan the number of trips established from the 1975 origin/ destination survey. - 79 -

the volume of 'mid-road'freight was also calculated,based on origin destination survey results. 1975 freight tariffs and costs of vehicle operation (financial)are known from the transportersurvey (see Section IIIB above; also Appendix C, tables 3-4). Adjustment of financial to economic vehicle operating costs is based on a study undertakenby SETEC in Madagascar in 1973, which was updated by a Bank mission in 1975 (cf. Appraisal of a Fourth Highway Pruvect, IBRD Report 978a MAG, June 1976, Annex 1).

3.1z1 Freight transport cost savings are calculated as the sum of:

(a) the difference of without and with project economic charge per ton/Km, multiplied by without project ("normal")traffic volume, plus

(b) the difference of without project economic and with project financial charge per ton/Km, multiplied by the volume, of 'generated'traffic multiplied by 0.5, plus

(c) the difference of with project financialand with project economic charge, multiplied by the volume of generated traffic.

Based on estimated tonnage exported without and with the project in 1975 and associatedtransport costs, freight transport cost savings for long aistance traffice are derived below.

Constant 1975 FMG '000

SAVINGS ON NORMAL TRAFFIC 100,676

Vanilla - Sambava (53,090 - 2,311) x 48 2,437 - Antalaha(53,090 - 4,061) x 192 9,414

Coffee - Sambava (28,630 - 2,311) x 1,406 37,005 Antalaha (28,630 - 4,061) x 352 8,648

Other - Sambava (32,130 - 2,311) x 713 21,261 Antalaha (32,130 - 4,061) x 1,030 28,911

plus SAVINGS ON GENERATED TRAFFIC 126,175 - 80-

Constant 1975 FMG '000

SAVINGS ON NORMAL TRAFFIC

Coffee - Sambava (28,630 - 2,675) x (1,832-1,406) x 0.5 5,528 - Antalaha (28,630 - 4,700) x ( 458 - 352) x 0.5 = 1,268

Other - Sambava (32,130 - 2,675) x (7,661 - 713) x 0.5 = 102,327

- Antalaha (32,130 - 2,675) x (1,960-1,030) x 0.5 e 13,697

plus

Coffee - Sambava ( 2,675 - 2,311) x (1,832-1,406) 155 - Antalaha ( 4,700 - 4,061) x ( 458- 352) 67

Other - Sambava ( 2,675 - 2,311) x (7,661- 713) = 2,529 - Antalaha ( 4,700 - 4,061) x (1,960-1,030) 594

TOTAL FREIGHT TRANSPORT COST SAVINGS: Long Distance = 226,851

SUMMARY OF 1975 TRANSPORT COST AND VOLUME ESTIMATES

ITEM FMG/ton/trip - 1975 Constant Values 1975 WITHOUT PROJECT (AIR) WWITH PROJECT (ROAD) Tons Transported WITHOUT WITH FINANCIAL ECONOMIC FINANCIAL ECONOMIC PROJECT PROJECT

To Sambava:

Vanilla 54,150 53,090 2,675 2,311. 48 48 Coffee 26,690 28,630 2,675 2,311 1,406 1,832 Other - Rice 33,190 32,130 2,675 2,311 - 6,663 - Other exp. 136 158 - Imports 577 840

To Antalaha:

Vanilla 54,150 53,090 4,700 4,061 192 192 Coffee 29,690 28,630 4,700 4,061 352 458 Other - Imports 33,190 32,130 4,700 4,061 1,030 1,960 - 81 -

3.122 Cost savings on sthoit distance traffic (most generated), volume estimates of which were derived from the origin/destination survey add a further FMG 4,095,000 to this figure in 1975.

3.123 A parallel exercise is carried out for passenger traffic, results of which have already been presented in para. 3.109 above. Based on 1960-70 data, without project (air) passenger trips are assumed to stay level at 7,000 trips annually, which are evenly divided between Sambava and Antalaha. There also is some pedestrian traffic from surrounding "midroad" communities towards Sambava and Andapa estimated at 10,000 trips annually. With project nassenger trip estimates are derived from the origin/destination survey carried out ill 1975 (See Section IIIB above). Trips to Sambava and Antalaha in 1975 are estimated at 102,400; 'mid-road' passenger movements increase to 155,300. Long distance passengers take the bus in 90% of their trips, the bush-taxi on 10%. Local trips are split conversely, with 20% dsing taxis. Buses and taxis charge the same tariffs, but per passenger operating costs are considerably lower for the former (see Table 11, Appendix D for details).

3.124 Passenger transport cost savings for the observed mix of vehicles and destinations are then again calculated as the sum of:

(a) the difference of without and with proiect 'economic charge per passenger trip, multipliee by without project (normal) trip volume plus

plus

(b) the difference of without project economic, and with project financial charge per passenger trip, multiplied by the number of "generated" trips, multiplied by 0.5.

plus

(c) the difference of with project financial and with project economic charge per trip, multiplied by the number of generated trips.

3.125 Total estimated transport cost savings on 1975 freight and passenger volumes are summarized in Table 27 below. - 82 -*

Table 27

Estimated Transport Go5t Savin - 1975 (Constant 1975 FMG 000)

Item Freight Transport Passenger Transport

Long Distance

Normal Traffic 100,675 21,370 Generated Traffic 126,175 158,370

Sub-total 226,850 179,740

Short Distance

Normal Traffic 1,635 2,300 Generated Traffic 2.460 18,770

Sub-total 4,095 21,070

Total 230,945 200,810

3.126 User cost savings by themselves again do not measure the full con- tribution of the project to the Malagasy economy. To establish this, two additional comporents need to be added. The first is transDortersurplus. Paras. 3.110 - 3.112 above already described how this is derived as the difference between transportersrevcnue and operating expenditures (net of taxes) in the without and the with project situation.

3.127 The second component to be added is import subsidy savings on rice. Production of rice for the market in the study areas is clearly attributable to the project: when only an airlink existed to the rest of the country, paddy was grown exclusivelyas a subsistence crop. Para 3.107 above detailed how subsidy savings were established for 1974/75.

3.128 In the absence of sufficientlydetailed informationon 1975 area imports, it was not possible to estimate consumer surplus on imported goods. It can be expected, however, that transport cost reductions on such bulky items as cement, salt, sugar and even textiles would result in far from negligible consumer benefits. Their omission will lend a downward bias to the RUS benefit estimates.

3.129 Total economic benefits of the road and agricultureprojects in 1975 resultingfrom the RUS analysis can now be summarizedas follows: - 83 -

1975 Source of Benefits Amount of Benefits (constant 1975 FMG 000)

Freight transportcost savings 230,945

Passenger transport cost savings 200,810

Transporters'surplus 36,790

Import subsidy savings 186,565

TOTAL 655,110

Two-thirds of total benefits are cost savings on freight and passenger (pro- duction + non-productionrelated) transport; five percent are appropriatedby transport operators, and just under 30% accrue to the economy in the form of import subsidy savings.

3.130 A comparison of 1975 benefits estimated by the RUS approach with those establishedusing producer surplus shows that, excluding all adjust- ments, benefits measured at the level of the road are only 72% (or 57% depend- ing on the valuation of self-consumptionadopted) those measured at the level of the farm. This is because the latter include surplus on total agricultural production while the former only measure transport cost savings on marketed output. In Andapa, where the greater part of rice production in 1975 is still self-consumed,the difference between the two measures is considerable. Pro- ducer surplus in the with vs. without project situation increasesnot due to higher prices the producer receives in the former: producer prices are cen- trally administered;they apply uniformly all through the country and irrespec- tive of any regional developmentproject. The observed differentialreflects partly higher yields (due to the agricultureproject), partly expansion of area under crop in response to assured marketabilityof any crop excess over and above subsistenceneeds (due to the road). Increasedproduction with the project is accompaniedby lower production expenditure on bought inputs, so increased producer surplus indicates higher returns to more efficient use of family labor and owned land. While definitely project effects, these are not captured by the RUS approach which thus tends to give downward biased estimates particularlyin areas where subsistenceagriculture predominates.

4. Rate of Return Estimates

3.131 Rate of return estimates for the project were next derived using both a model currentlybeing developed by the TransportationDepartment of the World Bank (FAST) 1/ and a standard computerizedanalytical package (CBPACK) 2/. Input specificationsrequired are year-by-yearestimates

1/ FAST = Financial Analysis System.

2/ CBPACK = Cost/BenefitPackage (Release III) both: IBRD, Transportation Department. - 54 -.

of cost and benefit strearr,; toi the por:iod ol expecteid project 1ife; in this case, twenty years from couipletion ut c. nste:ictloii was assumed. Detailed cost and benefit streams utsed in t.h aoal.vs is airn giLven in Appendix D.

3.132 On the cost side, road cinint,ruitjiin cxpenliturcs, road maintenance expenditures and agriculture proje(t ex i'i(iIi i .,, ir the major items con- sidered. The first were. obt a JTied i I s Lc Iv f (Irm fromT FED wh o lhelped finance the project, and from represc.nt;;t iVti'; (i- Murri Jreres, the contracting firm who constructed the road; they inil itic a -;i.e;!oblc 1 'oponilit for repair of 'during construction' cyclone damages. (The figures obtained were assumed to be 'current' economic values an( hienice they were adjusted to constant 1975 price levels). With respect to maintenance cost, it could not be estab- lished firmly from the regional Public W4orks cenLer, whether regular main- tenance allocations were in fact speiit: actual expenditures for 1975 were used as indicative for "nornmal" majintenatice. To accouit for current and ex- pected cyclone damage repairs, a full resurfacing of tht'- road in years 8 and 16 after road completion was alsn included. A 'salvage valle' for the road at year 20 was estimated assuming a 'life' oi' 5O years for earthworks, 20 years for artworks and 8 years for the pavement.

3.133 As evidenced in the previous disecusx-i>lon . uroent and projected volume of production and exports (espU Ially hiiI VOllT;, I ow valne rice exports) are based on successful imilelicTitat i1 1.l10 (3IA(Aagricultural project. To avoid crediting the road project wick, ogilrcultural production effects that are attribu-able to agricultural ii.fr-asLruc ture (drainage works) and extension efforts, all OPACA project costs were inc luded in the rate of return calculus. Informationi on expendiit ,in i)currod to date and estimates for the remaining project impJc-meiitation pi r. :o -ine those p. ovided byt fED (see para. 3.36 above); they weere agai n a ssomf d be `' crrent' economic values and hence adjusted to constant 19-/5price levels.

3.134 The benefits considered in the PS and RUS approach respectively were discussed in detail above. The asFu1rp: 'Ins forl fprojecting agricultural production and freight and passc - olunies wvci t;ie mLl period of project life both in the without. and the Hiti p1 jt ',1 ta'tion are spelled out in Tables 12-15, Appendix D.

3.135 To gauge ttie iFcpiiicemr eYe; zil I' L L Ii et both with respect to the reliability of ths bhuc ;!t:1e h'he projections of future agricultural production and traffic levels, Lhe sensitivity of the rate- of-return results to varying assumptions (+20d- to -20% of "best estimates" on production costs and yields for thi PS u.i i , for volusle of freight transported in the RUS analysis) was als,o e7u ns9. FtirItheorore, in addiB- tion to the "real" case (= case A), -evelv-iil hiv. :,1cr4ai " s ceriarios" were modelled to examine the implications; of Inf2r:L nrojcct implernentation or sequencing schedules for project rcti"rnl. Te basik cost paramcters and the various sequences used in the annalyris of the different ca3cs .. re Sui- marized below (for details refer i:o Appendrid 1) - 85 -

CASE A reflects the actual construction period of seven years plus thie cost of cyclone damage repairs, as well as the actual/projected timing of the agricultural project.

CASE B models a "normal" construction period of three years reflected In lower construction costs, shorter mo- bilization/supervislion, etc.) - but retains Lhe actual timing of thie agricultural project.

CASE 0 assumes that the road and agriculture project are sequenced so that the road opens in the first years of significantly Increased rice production - the "optimal timing" case.

COST AND TIMIN( PARAMETERSFOR THREE PROJECT SCENARIOS (CONSTANT 1975 FMG)

Item CASE A CASE B CASE 0

Road Construction Period 7 yrs. 3 yrs. 3 yrs. Road Opening 1970 1970 1974 Agriculture Project Horizon 1971-1977 1971-1977 1971-1977 ------_------__------FMG MILLION

Road Construction Cost 6,519.2 5,240.6 5,240.6 Road Maintenance/yr: normal 5.3 5.3 5.3 Road Maintenance: exceptional 112.0 112.0 112.0 Repaving Cost 1,000.0 1,000.0 1,000.0 Agricultural Project Cost 1,569.5 1,569.5 1,569.5

3.136 The respective estimates of the ERR on the three project "scenarios" outlined above are summarized in Table 28 below. As would be expected from the 1975 benefit comparison above, the rates of return established using the PS approach are considerably higher than those obtained from the RUS analysis. Valuing self-consumption at retail price level, PS return estimates are in fact double those based on RUS; valuation of self-consumption at producer price level lowers PS return estimates by 1-2 percentage points.

3.137 Comparison of the actual case results with those of the two hypo- thetical scenarios elicits the following observations:

CASE A (actual) - High cost/long implementation period of the road, combined with the hiatus in phase with the Table 28

Economic Rates of Return (I:RR) , ;,xt:-Year Benefits for Three Proi_ct_ S tennrk

Ec:onoinc R.ite of Return M E T 11 0 D CASE A(ctual) CASE,B(etter) CASE 0(ptimal)

I PRODUCER SUPLUS

(a) Rice sold valued @ Producer Price, self- consumed @ R.etail Price

- including subsidy savings 12.3Z 13.9% 20.67 - excluding subsidy savings 11.4% 14.7, 18.5%

(b) All rice valued @ Producer Pricp

- including subsidy savings ll.1, 13.9% 18.3% - excluding subsidy savings 10.1% 12.7% 16.4%/

II ROAD USER SAVINGS

- including subsidy savings 6.3% 8.1% 10.1% - excluding subsidy savings 4.5% 5.8%. 6.9%

FIRST YEAR BENEFIT (10% Interest Rate) Year 7 Year /4 Year 4

I (a)- including subsidy saving 4.35 6.59 15.34 - exK1uding subsidy saving 4.35 6.59 12.36

(b)- including subsidy saving 3.73 5.72 13.19 - excluding subsidy saving 3.73 5.72 10.22

II - including subsidy saving 3.96 6.84 11.33 - excluding subsidy saving 3.96 6.84 8.35 - 87 -

agriculture project result in low project returns. Based on the RUS analysis, the project appears marginal; even when rice import subsidy savings are included, the ERR is only 6.3%. The comparable PS-based ERRs are 11.1% and 12.3% which, assuming an opportunitycost of capital of 10%, make the project economicallyacceptable.

CASE B - Compared to the actual case, this reflects the impact of the long drawn out constructionperiod on the economic return on the project. Given a 'normal' con- struction period, returns would improve by 1.3 - 1.7 percentage points in the RUS, by 2.6 - 3.6 percentage points in the PS analysis.

CASE 0 (optimal)- This highlights the importance of good project phasing for the economic "success" of the road and agricultureproject package. Assuming that the road is constructed on a normal schedule (see B above) and opened for traffic exactly at the time when rice is first produced in considerablevolume for export, RUS-based returns are more than half again those in the actual case and reach 10.1% if the adjustment for rice import subsidy savings is included. "S-based estimates reach a very satisfactory 18.3% and 20.6% ERR, respectively. Also, this is the only case when first year benefits are substantial (NOTE: The sensitivityof rates of return to "instant" (timely) re- sponse suggest the need for much more detailed information on peasant farmers' propensity to change their traditional patterns of production and marketing than is usually avail- able).

3.138 For the actual case (=A), the sensitivityof returns to different assumptions concerning agriculturalproduction costs, yields, and volume of freight transportedwere tested. Rates of return (includingrice import subsidy savings) are most effected by changing assumptions about yields: a 20% error in estimated volume of agriculturalproduction results in a change in producer surplus ERR of about two percentage points. Errors in production cost estimates of 20% are associatedwith ERR changes of about one-half of a percentage point. Variation of tonnage transported in the road user savings analysis by 20% changes rate of return estimates by one percentage point (see summary below). - 88 -

Sensitivity of Rate of Return Estimates to Changes in Production. Production Cost and Freight Transport Volume

I T E M CASE A(Actual) - 20% 0 + 20%

PRODUCER SURPLUS

a) Rice sold @ producer price, self-consumed @ retail price

Change yield by 10.2 12.3 14.1 Change production cost 12.7 12.3 11.9 ------___---

b) All rice @producer price

Change yield 9.2 11.1 12.9 Change production cost 11.6 11.1 10.7

ROAD USER SAVINGS

Change freight tonnage 5.3 6.3 7.2

3.139 In concluding,it should be emphasized that the ERR estimates de- rived above are conservativeestimates. The PS analysis is based exclusively on Andapa Basin production and excludes all producer surplus of farmers along the road (whose production patterns are quite different and from whom no before/withoutand only scanty after/with production estimates are avail- able). Secondly, savings to area residents in the form of consumer surplus on general consumptiongoods that previously had to be airfreightedin at high cost are ignored. Part of these benefits may be appropriatedby middle- men (traders,importers) but a significant share may be going to end-users. Finally - as always - no explicit valuation was made of improved access to social services and markets for the 95 percent of the populationwho pre- viously had to rely exclusivelyon their own limited productive and service resources. CHAPTER IV

INCOME DISTRIBUTION EFFECTS:

CHANGES IN GENERAL WELFAREAND AREA RESIDENTS' LEVEL OF LIVING

1965 - 1975 SECTION IV A

INCOME LEVELS AND INCOME DISTRIBUTION: 1965-1975 8')-

A. INCORE LEVELS AlND INCOME DISTRlBUTION: 1965-1975

4.1 Information oll110MUsl01d incomes and expenditures represents the core of the comparative data on the area economy collected by the 1965 and 1975 surveys. These sihowwihat the aggregate entities of the economic anal- ysis (an FRR of x% or an area producer surplus increase of y) mean at the level of the average finser Of crader family living in the region. They also reveal whether any one segment in the area population has improved, or worsenced, disproportionacLLy In oconomic position compared with other groups.

4.2 Hypothesis 3 in ')t'(:tion 11. above stated that area residents' in- comes are expected to rise -asa result of improved access, and that income distribution may impry;vc'givein certain 'structu.al' properties of economic and social organizati.n. Thie economic analyses carried out earlier (Section III C above) already iearly indicated a rise in level of returns to area farming activities between 1965 and 1975 above and beyond that which would have been expected withoui thc- road and agriculture investments. Analysis of incomes reported by sample households confirms these findings and contributes interesting additional Insight.

4.3 Some explanatory notes on the data presented in this chapter first. The analysis includes /22 of tlhe 723 households surveyed in 1965, and 770 of the 773 included in the 1975 survey. The latter thus includes households in the villages along the road as well as those in the Andapa Basin proper. If anything, this imparts a sliglit downward bias to 1975 comparative figures as cash crop production tends to be lower along the road.

4,4 Incomes (by sourcc) irot all economic activity by all members of the household are reported in 'current' FMG for 1965 and 1975 respectively. "Cash" income includes the cash equivalent of payments to labor or land that were actually paid in kind, but excIludes the value of self-consumption of own-produced goods whiicl enters in 'total income' valued at retail price level.

4.5 Sample totals l, ilucome received were calculated first and are shown in Tables 1-5 Appendix F, To fa'ilitate comparison (without resorting to the somewhat approximate 1975 expansion coefficients and inflation esti- mates), sample totals were next Lransformed into "per household" averages presented in thireeways:

a) average houselhold cash income by occupation of head of household: to test for differential income gains between occupational grouIps;

b) average houspheJd (ashi Income by location (distance from Andapa/Road) of households: to test for income differen- tials directlv ittii1utabic to varying accessibility; c) avo';ve' i isohol ds ch r( i. incorm-, (in:i 'Ivi I g .a1,ae of ;n1lcfc oc X 7A.Ir, the range of income differentials. Ii einnd highest incortie groups; show tendencies ilv t ht- i: t- .3opment; aind presenl hynothteses on tLhol aeIf, - i

4.6 Togc.ther with1i gross incomes i rum protdi:i( of agricul tural products, f7Won trrade, -nd from crafts, r.. a. penditures are shown in alI tables. investment in hoiinn. *ro)n. s Ov) i' also included as a 'productive' expenditure because of its con-i;1I-rabloimportance in the study area (see Section IV C, 2 below). The diffcrcnce between total cash incomes from all sources and production expenditures represents disposable cash income (D''I)for consumption and savings. Disposable cash income rather than total niet 1ncome (DCI plus value of self-consumption) was used for the comparative analysis because the daita on snif-consumption were collected in different ways in the two surveys (see parl. 2.41 above) and are likely to be downward biased for 1975. Generally, consrnumnion of self-produced goods in the study area would seem to conform to fairly stable patterns that have also been observea in similar subsistence economies elsewhere. 1/ Focus on the comparative disposable cash income received by households in the two study periods should therefore most clearly highl ight the changes that may have occurred.

Findings

4.7 Average disposable cash income per heuseholds received by area residents in 1975 was 235% of that received in 1965, rising from FMG 75,900 to FMG 17&,600 (current). Assuming a gonerai inl iOl' level in Andapa similar to that in the rest of the country of tHi+-order of 75% this repre- sents a real increase of FMG 45,775 per household (=35.).

4.8 The 1965 apd 1975 sample results indicate some pronounced shifts in the contribut ion of different income sour e. o total disposable cash incomes. While 'salaries and wages" were the cleo,rIvdominant source in 1965 contributing 44% of the all disposable cash income, foLlowed by income from trader sales (17%) and income from producPr sales of agricultural products 16%), 2/ the latter brŽcompetho- vo5t importaTit -v'c -al1 sources of disposable cash income (A-7%) In 1q75, outstripping s-d.-ris and trader sales which both contributo 22%. Adjustment of unwcoigh-ed snrpnle results for different occupatiounal -omnosition of t:he t7co samples (17/ of 1965 heads of household

1/ For cx,-lc a-- K±ug. Pr.X. and( '!eldon. P4,>... IncnmacDistribution and Levels of Living in Java, 1963-1970, in-, Erononic Development and Cul- t-ral_Change. University of Chicagn Press, vol. 25, number 4, July 1977, p. 707,.

2/ Note that this figure is low due to complications in the local vanilla market. - 91 - were 'eiployees"compared to 12% in the 1975 sample) would modify these re- sults somewhat but not reverse the picture.

Table 29

Proportion of Disposable Cash Income from Different Income Sources: 1965-1975

|Income Source | 1965 1975

| Producer Sales of Agricultural l I Products I 16.0* 47.1 l I Trader Sales of Crops and Goods I 16.7 21.7 l

| Income from Crafts | 6.3 11.1 l

I Salaries and Wages I 44.0 21.8 i II | Rent (Land/Buildings) I 1.6 1.1 l | Exceptional Sales I 13.4 3.2 l I Other Sources I 5.7 4.6 l

l l 103.7 110.5 I | (Housing ConstructionExp.) (3.7) (10.5) l l l 100.0 100.0 I

-l Average Disposable Cash Income/ l Household (FMG) X 75,900 178,600 l

* Note that this figure is low due to complicationsin the local vanilla market.

4.9 The second salient change is in the relative level of reported pro- duction expenditures. For the sample as a whole, these averaged 66% of total cash income in 1965, leaving 34% disposable for consumption. In 1975, by con- trast, only 22% of total cash income went for production expendituresand 78% remained "disposable"for other purposes. The major drop is in production expenditure in trade (from 91% to 22% of total cash income), but there is a similar downward trend for crafts (41% to 19%) and agriculture (29% to 15%). A significantpart of this expenditure reduction can certainlybe attributed to reduced transport costs due to the newly opened road. (seoc TThbL ';i _are sig- nifi cmLt diiI .;aking A average dispoz-il) , 1965 had . lope, t i mes their amnul v - . oii bctter off in 1975, bLut t01C0 .1 v•&ii I .b bV t 30% more ll iljnf.'. ' .,. ' folur t iu'cwb th_eAldO .

4.1 1 Tic L jW...l. e nmainly tC propOrt O-nat]\ _.: I Leg. L oups. Averag.i 1975 d isF, O a. "' ;49% was fo9 fa-rmers, 148%. ;., 7d t ders. in bothl survey perL u . ;. LI .- :,. 'c vo;ced sales of agri.u I c . ;''I611 C i njcome camns aclmost exclus.2vi . for 43Z and 'J')3 -- C..'iy, Lraders draw a greater s. ,'i -oduc ed agricuiltural pr.,'civ i to0"' :;reflect a change in th WC.col;l-gie 'd fewer "spc-cialis s.] . :> ' .fl sional traders' ', .-- t u sized traders w7! - , :r- i l diversified inv ia. fi,.) '0I Il cash income in 1'

4.12 fn ha 1 -. sharin,g of regi. .x. 0- horizon. Rt- u. . . ,.cpU±a7iu! stabilizes, more ii spor. 3 alizcc skill to earn their livihe ) ..rL t, a trend no doilht:or . , I S yea rs.

4.113 Ta-lb:- . to thc nn- r(ab , '- ! differfontiaJ 5 potter,ns bceuewin ' if trader an-ir d . ( level and 7I1<'. 4) vere eXcA.lic .

4 .*A HowIq sourres ir And income there iS `t 2

1/ In con-ta7 '' -. .- V andf. 19, ?.. Table .IU AVERAGE IIOGSEHOLDCASHi INCOMIE AND PROD1UCTIONEXPEN(DITU!RE - BY'(X.C:PATIONAL CRO": 19635 1975 (Fp1C0000'

I 9 ~ 51 9 7 5 I' Per Household Avera e 3000 ______Per FAR*NERSCI(AFTSW.N TRAMP~S Household Averape - 9M47000 EMPFLOYEES TM71AL FARJIFRS VCRAfCTS1E.NI RADE.RS EMPLCYFE'1 TOTAL Producer Sales Cash Crops 20 8 2 7 3.9 6 14 -Subsistence Crops 5 64 1 46 4 I S 0 7 0.4 0.5 ~ 4 4 I bq.6 66 3 -Animal 1 3 23.9 ProdUCtS 1 4 0 9 0.4 434 24 265 0 3 0 21 I.4 ~53 4 64 `RTA,Producer Sales of Ag. Prjd. 24.0 43 4 7 6 4 27 2 92 2 59. A~~r,cultura3 5 264 98 4 990 Production(4 93Exyendi~~~~~~urel ~~~5.03 (4 2' 5 23 I (4 93 (9 33 (18 0' 44 23 (14(149~ 9' ~~424 8' Trader sales of Cro_p,_&Goods 1.1 1.9 2.279.0 2.2 150.0 7 0 21.7 (Trade-Expenditures3 346 9 22 49 7 . .7) (5 43 21 8 3 63 (137 13 (2.53 (3.73 (63 73 23 I(ncomefrom Crafts (9) (10 9, 0 8 89 6 1 9 24L Q I 7 4 68.9 50 2 (Craft 73 0 24 3 Expenditures) (-3 e38 83 (0.33 12 03 (3.33 CUb 10 7' (24 93 Salaries and t4.2' (4 73 Wages 3 2 3.3 1).9 271 2 33 3 7 7 16 3 455 f, 2300 Rent Qland/Buildings~ 38.9 0.8 0.8 2.2 2 2 1 2 1 4 2.8 3,q Exceptional 2 9 2.0 Saels .0.4 0.8 82.6 1 9 10.2 2.7 10 2 23 2 Other Incorne 2 2 5 7 I 5 1 9 17 1) 7.0 4.3 3 5 3 1 29 7 16 Average Total 2 82 Cash Income/Household 31 8 102.6 1,407 3 193.2 224.4 220 8 172,5 664.4 434.4 'Ilomqe ConstrcictionFxpendituires) 227 8 (2 31 7) (I (4 4' (5 1) (3 Cl( (4,2' (6 0' '125 83 'Total Production (20 6' (28 8' Expenditures) (7 93 (50 93 (1,174 73 (15 93 (148.3i (16 63 (38 43 t248 63 38 Average Disposable ~ 93 (49 23 Cash Income Per 23 9 51.7 232 6 177 3 76 1 10. 134 1 45f.g oHi us eh o l d s_ 395 5 2786 in Group ______460 56 L 84 122 ______j __ _ _ 5236 49 95 90 770 I -T T F K. I-- - -cig

-'- J -- - I "'''-1".' I II I K -IcC- - Ic-{

c.r.c - C:, C,,

I I C) , 0.i cc. J ' I

C- I -I

C-! 'C-

I __ 0' I _-2 *I .. i o p-IC' War- I .-.. - ,- C - I I -, - -. II _ -- . 1 i K - ri v--i >2K19 c--C-, 17 1- 0' IC - -, o li1L :-- K' Ii I --- I. -- L C, - --.-- 4- -4-- ____- - p. - I

Ii

I 'II I -, IC- -I-. -'-I . I* I j C --- 1

'I-I Ii, Ira., 0' C-I C I IIZ C-. Ip*

I --. 2- __ --- 4- - ----4 C- P-II* I C, I'I ph I 'I - - -. -, '1 ac-a z 2 I 4 I !' - IN - t------h------------------2 '-I '' , I' I i--P -I,-: - IC

-I .1 . -I . C -. C,

I-- -

- -- -a- - - - - I -- P-' <: .5- 1 I-lI - 1 ,-I I ____ 2 V24 ______-- ;. --- I,.

T1 Z

I I I I cK'f I p,.i -r I I I 1:,,.' >1

I , C - ' I-' '-I I -hI I-* ', (' P.. II

I, '--I I,

CII - I'P.,

IC .il I

1,1 C

- I - 95 -

from salaries and wages. Crafts and other sources account for the small remainder. In the distance group 4 (16-20 km), just under half (47%) of income is from sale of agriculturalproducts. In groups, 2, 3, and 5 (6-10 km, 11-15 km, 21+ km), agriculturalproducts sales dominate with 60%, 62% and 71% respectively. The share of income from agriculturalproduct sales (62%) for the road sample is also high, although the level of average income is low: many farmers sampled grow rice only. Andapa (distancegroup 1) is the only sizeable market for sales of animal products.

4.15 In the 1965 data, the relative contributionof various income sources is quite different. Over half of disposable cash income in Andapa (distance zone 1) originatedfrom salaries and 30% from (exceptional)sales. Agriculturalproduct sales contributedonly a very small portion of disposable income (under 4%). For distance zones 2, 3, and 5, agriculture generated 32%, 26% and 60% respectively. Salaries contributeda proportionatelyhigher share of disposable incomes in zones 2, 3, and 5 than in 1975 (34% vs. 17%, 20% vs. 6%, and 21% vs. 11%); many previous employeesnow farm their own account. Also, the disturbed vanilla market probably explains some of the variation observed.

4.16 Changes in income distributionwere next traced by systematically dividing the sample population into income deciles and comparing level, compositionand dispersion of income for the various sub-groups. Tables 32 and 33 below show average sample household cash income and production expen- ditures for "total income (i.e., cash-income+ value of self-consumption) deciles" in 1965 and 1975. The "equal share coefficient",1/ a useful in- dicator of income concentration,was establishedat 97.5% for 1965 sample households, and at 88.5% for 1975 sample households. Thus 50 percent of gross cash income available to all sample households was held by 2.5% of the households in 1965, and shared by 11.5% households in 1975.

4.17 In 1965, the upper (10th) decile of income earners among the house- holds sampled received 52% of total disposable cash income reported. In 1975, this proportionhad declined to 42%. Average disposable cash income (ADCI) per household in the top decile was 54 times that in the bottom decile in 1965; the same ratio narrowed to 32 times in 1975.

4.18 The major source of income in the 6 lower deciles in 1965 is sales of agriculturalproducts; 2/ in the top 4 deciles salaries and wages are increasinglymore important. In 1975, ADCI from agriculturalproduct sales dominates contributionsfrom other income sources up into the 9th income decile, being outweighedby trade in the top decile only. Market-orienta- tion in agriculture thus seems to become more pronouncedwith improving

1/ See Alker, H.R., Jr. "Measuring Inequality"in The QuantitativeAnal- ysis of Social Problems, Addison-WesleyPublishing Co., Reading, Mass., 1970.

2/ Remember again that income from this source is understatedas one-third of the vanilla crop sold remained unpaid during 1965. Table 32

AVERAGE HOUSEHOLD CASH INCOME AND PRODIICTIONEXPENDTTURE BY DECILE - 1965 (FMG 000)

I T E M ____ D E C T L E T O T A L 1 2 3 4 5 6 7 8 9 10 _ Producer Sales - Cash Crops 4.2 5.8 8.5 12.0 12 7 17.8 19.4 24.5 32.6 - SubsistenceCrops 11.1 14.8 0.6 1.4 1.4 1 3 2.1 1.1 2.0 - Animal 1 3 1.5 0.4 1.3 Products 0.1 0.3 0.7 1.0 1 3 1 6 1 9 1.8 1.1 0.5 1.0 I.SUlB-TOTAL: Producer Sales of Agric Prod. 4.9 7.5 10.6 14.3 16.1 20.5 23.3 27 6 35 2 12.0 17.1 2.(Agricultural ProductionExp.) (1.7) (2.2) (2.4) 3.7) ( (4.4) (4.2) (8.4) (8.4) (8.1) (5.7) (4.9) 3*Trader Sales of Crops & Goods 0.1 0.3 0.4 1.3 0.8 1 8 2.3 4.6 33 1 1,420.1 150.0 4.(Trade Expenditures) (O 04) (O 1) (0.3) (0.6) (1.3) (1.1) (I 2) (1 0) (32.0) (1,336 9) (137 O0 5.Income from Crafts 1.2 1.5 1 6 1.9 2.1 5.6 4.1 9.5 10.3 42 4 8.1 6.'Craft Expenditures) (0.07' (0.01) (0.02' (0.05' (O 07) (0.2) (0.1) (O 71 (4.5' (27.5) (3.4' Z Salaries and WaRes 2 4 4.7 2.8 6 2 6.6 8.6 18 1 38.0 116 5 228.6 33.5 8.Rent (Land/Building) 0 1 0.2 0.3 0.6 0.3 0.5 0.8 2 2 2.0 4.7 1.2 9.Exceptional Sales - 0.3 0.4 0.2 0.5 0.1 1.0 0.5 1.5 95.5 10.2 QOther Income 0.8 0.8 0 8 1.2 1.4 2.3 1.9 1 1 4.3 27 6 4.3 .Average Total Cash Income per Household 9 5 15.3 16.9 25.7 27 8 39.4 51.5 83.5 202.9 1,730.9 224.4 2 l .(HouseConstruction Exp (0.51 (0.7) (1.0) (1.61 (1.41 (2.11 (3.6' (5.2) (6.51 (7 7) (3.0) L3(AverageTotal Production Expenditure) (2,31 (3.0) (3.7) (5.8) (6.61 (= 2 + 4 + 6 + 12' (7.8) (13.3) (15.31 (51.1) (1,342.91 (148 5)

Average Disposable Cash Income per Household 7 2 12.3 13.2 19.9 21.2 31.6 (= 11 - 131 38.2 68.2 151.8 388.0 75 9 ; N Households in Group 72 72 72 72 72 72 72 72 72 74 722 Table 33

AVERAGE HOUSEHOLD CASH INCOME AND PRODUCTIONEXPENDITURE BY DECILE - 1975 (FMG 000O

I T E M D E C I L E T O T A L 1 2 3 4 5 6 7 8 9 10 Producer Sales - Cash Crops 15 3 27 2 34 9 40.4 45 5 51 3 72.1 81.8 110.9 182 2 66 1 - SubsistenceCrops 3.3 6.0 11.2 7 4 1.9 4.5 25 0 37 0 47.3 93 6 26 - Animal Products 0.1 5 1 5 0.1 0.7 3.8 3.2 1 8 4.7 1.9 46.2 6 4 I.SUB-TOTAL:Producer Sales of Agric.Prod. 18.7 34 7 46 2 48.5 62 2 76 0 98.9 93 5 160.1 322 0 99 0 2.(AgriculturalProduction Exp ) (2.7) (4.3) (6.21 (8_0) (9.21 (9.41 (11.3) (15.4) (27.31 (54.91 _------…------_------_------(14 9) 3.TraderSales of Crops & Goods 0.9 1.6 1.3 7.6 8.9 11.7 14 2 18.2 78.7 354.4 49 7 4.(TradeExpenditures) (0.1) (0.2) (0.2' (0 91 (4.4) (4.4) (4 7) (1.0) '17 91 (76 5) _--- _------'11.01 …------_------_------_------_------_ ------___ -______-- ___ -_-__ -__ -__ ___ -______-_ 5.1ncomefrom Crafts 1 7 4.8 3 8 5.6 15.5 13.7 26.6 20.2 21.5 129 3 24.3 6.(CraftExpenditures) (0 06) (0 071 (0.11 (0.02) (_ (0.6) (0.51 (8 0) (16 4) (21 7) (4.81 _------_----_ ------_------_------_------_-_--_-_-__-_-_--_-_------7.Salariesand Wages 1.9 3.5 8.0 12.3 8.1 13 8 17.5 58 0 79 0 187 3 38 9 B.Rent (Land/Building) 0 1 0.6 0.8 2 1 2.0 1 0 2.5 3.5 1.3 5,6 2.0 9.ExceptionalSales 0.9 1.3 1.6 2.1 3 2 4.7 3.9 7.7 10.3 20.9 5.7 DQOther Income 3.0 3.5 19 6.7 3 2 2.1 4.6 5.6 14 7 36 8 8 2

IL-AverageTotal Cash Income per Household 27 2 50.0 63.6 84.9 103.1 123.0 168.2 236 7 365.5 1,056.3 227 8 12 tHouse ConstructionExp.) (1.5) (2.0) (2.1) (2.7' (3.1) (5 9) (4.01 (4.9) (5 2) (157.0) '18.9) 13.Average Total ProductionExpenditure) (4.41 (6.6) (8.6) (11.5) (16.81 (20.31 (20.51 (29.31 (66.8) = 2 + 4 + 6 + 12) (309.81 (49.5)

4 1 .AverageDisposable Cash Income per Household 22.8 43.4 55.0 73.4 86.3 102.7 147.7 207.4 298.7 t=11 - 13)(ADCI)I 746.5 178.3

N HouseholdsG in roup 77 77 77 77 77 77 77 77 770 1975 ADCI as % of 1965 317 353 417 369 407 325 387 304 |199 192 235 - 98 -

returns to agriculture in the region especiallywith regard to the marketed share of previously mainly "subsistence"crops.

4.19 The difference in the ratio of agriculturalproduction expenditures to producer sales values (around 302 in 1965, around 152 in 1975) is again partially explained by outstanding payments for part of the vanilla sales. The ratio is fairly stable throughoutall ten deciles in 1975, with slightly higher expendituresin the top three deciles (more use of hired labor).

4.20 Reported income from salaries and wages increases from the lowest to the highest decile with a significantjump in the eighth decile in 1965. Per household salary average in the top decile is 53 times that in the bot- tom decile, and 3 times that in the eighth decile. Correspondingvalues are 98 times and 3 times respectivelyin 1975, indicating a relative gain in high level salaries over the study horizon. Net incomes from trade and crafts reported in 1975 exceed those in 1965 by a considerablemargin: underreportingof production expendituresin 1975 could be a possible ex- planation for part of the divergence,however.

4.21 Decile by decile overall gains (1975 as percent of 1965 ADCI per household)-areshown on Table 34. The third and fifth deciles register the highest gains, with 1975 levels more than 2.3 times those in 1965 (constant 1975 prices adjusted for inflation). In all other deciles except the top two, 1975 levels are over 175X those of 1965. Decile nine and ten improve only slightly by about ten percent, essentiallyholding their absolute in- come level, but losing some of their relative (disproportionate)advantage.

4.22 Tables 34 and 35 below combine summary income data with a number of basic household characteristicsfor sample households in each income decile. Both in 1965 and 1975, households in lower and upper income brackets exhibit a number of characteristicfeatures. First, households in the two lowest deciles average just half the number of persons per household of those in the two upper deciles. There seems to be a clear-cut trend for income to rise with the number of economicallyactive members in the family. Second, there is a predominanceof "traditional"occupations (farmer, craftsman) in the lower deciles; a predominanceof the more "modern" occupations (trader, employee) in the upper income deciles. Third, 50% of female heads of house- hold in the 1965 survey, and 35% of female heads of household in the 1975 survey were comprised in the lowest two income deciles, although a minority was representedup through the top income bracket in both surveys. Fourth, it would seem that the very young and the very old are overrepresentedin the lowest income decile, although more clearly so in the 1975 than in the 1965 sample. For the 1975 survey, data on literacy levels of heads of households show a concentrationof illiteratesin the lowest income deciles, while the majority of household heads with a secondary education (41/50) are found in the top three deciles. Area cultivated per household - again available for the 1975 survey only - increases fairly steadily by income decile, most likely a function again of family size and number of economicallyactive.

4.23 In the next set, sample household incomes were calculatedon a per capita rather than a per household basis to correct for household income Table 34

PER CAPITA INCOME BY TOTAL HOBSEHOLD INCOME DECILES - 1965 (FMG 000O

D E C I L E S TOTAL

I T E M 2 3 4 5 6 7 8 9 10 AVEA-GE

HOUSEHOLD INCOME RANGE IN DECILE - UPPER 28.6 38.7 49 3 59.0 70.2 84.5 105.7 161.8 343.9 10,908.1 - LOWER 5.4 28 6 38 8 49.4 59.2 70.3 84.5 105 7 163 0 354 1

PER CAPITA TOTAL INCOME 7 3 11.7 11.5 12 2 13 9 15.0 17 0 24.2 40.3 325.9 56.0 PER CAPITA VALUE OF SELF-CONSUMPTION 3 7 6 3 7 1 6.5 7.9 7 2 7.6 8 6 5.6 6 4 6.9 PER CAPITA CASH INCOME 3.6 5.4 4 4 6.7 6.0 7.8 9 4 15.6 34.7 319.5 49 1 PER CAPITA PRODUCTION EXPENDITURE 0.8 1 0 0.9 1 2 1 4 1 5 2 4 2 8 8.7 247 8 32.4 PER CAPITA DISPOSABLE CASH INCOME 2.8 4.4 3.5 4 5 4 6 6.2 7.0 12.8 26..0 71.7 16 7 PER CAPITA CASH EXPENDTTURE 9.9 8.2 6 5 8.0 6.9 7.9 9.0 12 7 23 6 56 9 16.5

HOUSEHOLD CHARACTERISTICS COMPOSITION

N Households 72 72 72 72 72 72 72 72 72 74 722 N Persons 192 209 273 321 331 368 396 386 422 401 3,299 N Adult Equivalents 146.3 170.2 210.8 245.4 260.2 281 3 307 9 298 6 331 3 314.8 2,566.8 0

OCCUTPATTONOF' HEAD OF HOUSEHOLD

N Heads of Household: Farmer 65 59 59 57 59 53 48 37 21 2 460 Craftsmen 3 4 7 5 7 7 8 8 3 4 56 Traders - 3 3 3 3 3 2 5 14 48 84 Employees 4 6 3 7 3 9 14 22 34 20 122 722 AREA CULTIVATED (ha) NlT AVAI ABLE

DEMOGRAPHIC CHARACTERISTICS OF HEAD OF HH

N Heads of Household: Male 46 53 64 61 67 66 68 70 69 70 634 Female 26 19 8 11 5 6 4 2 3 4 88

N Heads of Household: Age 29 16 21 16 15 12 16 11 19 18 11 155 Age 30-49 31 26 33 30 32 34 40 37 36 30 329 Age 50 + 25 25 23 27 27 22 20 16 18 32 235 N.HA. - - - - Ih1 l3 - 1 - -

N Heads of Household: Literacv I~T AVA LABLE 722 I ---A

...... ~~~~~~~~~~~~~~~~*.*--t

I I'

;; ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~.. Ic.

K . - i' -

;£I . I ; .

II = _ ,,- ,-' : -:._

-I

I , I

I I . - 101 - differentials due to different household composition. As might be expected from the data presented above, per capita ADCI distribution (see Tables 36 and 37 below) is more 'equitable' than per household distribution. In 1965, per capita income in the top decile is 25 times that of the lowest one; in 1975, the difference is halved to just under 13 times. Differentials between low income deciles are very narrow: in 1965, double the per capita income of the luest decile is reached in the sixth decile only; quadruple in the eighth. In 1975, double the first decile income is achieved in the fourth decile (a steeper progression), but the quadruple still is reached in the eighth decile only. (In both samples, it is in the eighth decile that in- come from salaried employment begins to account for a significant portion of total per capita ADCI).

4.24 In summary, the decile analysis of overall household income levels and distribution shows a clear gain in area income levels for the lower eight deciles on the income scale. The top decile is still separated from the rest by an order of magnitude. But per capita differentials, even more than per household differentials, demonstrate that there have been clear equity gains in the Andapa area economy over the 1965-1975 period. In the total sample comparison, the specific contribution which the road has made to this im- provement could not be separated out and quantified, mainly because price regulation of producer prices for major crops as well as for selected trans port tariffs obscure and, in fact, supplant the expected impact of transport cost reductions.

4.25 In a final analysis, "panel" data on these households that were included in both the 1965 and 1975 surveys are examined in an effort to -

a) trace area household "mobility" with respect to absolute and relative income level; and

b) examine the relative contribution of accessibility (dis- tance to Andapa and the road) to a).

4.26 It has been shown previously (see Tables 34 and 35 above) that total income levels (household income range) by decile increased in real terms from 1965 to 1975 for all but the top 2 deciles - between 40-45% for the lower 7 deciles, by 22% for decile 8. Thus, except for the top quintile, households that either advance to a higher or remain in the same decile in 1975 compared to 1965, experience income gains; those who drop by one decile maintain a more or less stable income, while the remainder suffer a definite deterioration in income status. Table 38 below shows comparative income decile categories for all 468 "panel" households included in this analysis (separate breakdowns by professional group are given in Tables 11-12, Ap- pendix F).

4.27 Clearly there is considerable "income mobility" in the study area: of the total 468 panel households, 200 (=43%) were in a higher income decile in 1975 than in 1965; 81 (=17%) were in the same decile in both years; 83 (=18%) had declined by one decile; and for 104 (=22%), income level had -.ib le 36

F~.A)ER CAP!V 11 ,~ A%'.,N' Ii C,£X(NY' C. C.LE- '95

c, ro Sa I e- C ,,i - . K

,I tic D. O~~~~2 I .1

'd(-Tr S.1 Io ~ ~ ~ ~ ~ 01 ¼04 0 00 6

Ex,zqd I ~0 ?)Il '712' ------*------. ~4------

nri ILx.pnn1 icrc ---- '1

------

f-r.~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~1 t ,ad3' io, .1I3

ona p-,~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ I 31)0 -,

'q'er. Income . . 03 ;~ 0 I

- .-.--- TI!..,n ir~re1(o1 7 .2 .3 ~ oi 0 1i______..

''P.1 (11!iTlCtiT1f,xo'dien,t,res ))r 00 4) '~ 'I

"'-~ePor .ipiti3 isplbec:jsb nco-c' 2' 4 4 41>7 Af 7.0~ 13.0 Th 2

Itr~ons( ill Group 192 2096 336t223 Table 37

AVERAGE PER CAPITA CASH INCOME AND PRODUCTIONEXPENDITURE BY DECILE. 1975 (FMG 000O

D E C I L E s | TOAL I T E m ~~~~~~~~~~~~~~~~~~~~~~~~~-SAMPLE 1 2 3 4 5 6 7 8 9 10 AVERAGE

Producer Sales - Cash Crops 4.5 6.7 8 4 8.0 8.7 8.9 12.1 13 1 16.1 21.9 12.0 - Subsistence Crops 1.0 1 5 2.7 1.5 2 5 3 3 4.2 5.9 6.9 11 2 4.8 - Animal Products - 0.4 - 0.1 0.7 0.6 0.3 0 8 0.3 5.5 1.2 SUB-TOTAL: Producer Sales of Agric.Prod. 5.5 8.6 11.1 9 6 11.9 12 8 16.6 19.8 23.3 38 6 18.0 (AgriculturalProduction Expenditure) (0 7) (1.0) (1.4) (1 5) (1.7) (1.6) (1.8) (2 4) (3.9) (6.5) (2.6) Trade Sales of Crops & Goods 0.3 0.4 0.3 1.5 1.7 2.0 2.4 2.9 11.4 42.6 9.0 (Trade Expenditures) (-) (-) (0.1) (0.8) (0.7) e0.7) (0 1) (2.5) (9.1) (1.9) Income from Crafts 0.5 1.2 0.9 1.1 3.0 2.4 4.5 3.2 3.1 15.5 4.4 (Craft Expenditures) (-) (-) (-) (-) () (0.1) (-) (1.2) (2.3) (2.7) (0.8) Salaries & WaRes 0.6 0.9 1 9 2 5 1.6 2.4 2.9 9 3 11.5 22 5 7.1 Rent (Land/Building) - 0.1 0.2 0 4 0.4 0.2 0.4 0.6 0.2 0.7 0.4 Exceptional Sales 0 3 0 3 0.4 0.4 0.6 0.8 0.7 1.2 1.5 2.5 1.0 Other Income 0.9 0.9 0 5 1.3 0.6 0.4 0.8 0.9 2.1 4.4 1.5 Average Total Cash Income 8 1 12 4 15.3 16 8 19.8 21 0 28.3 37.9 53.1 126 8 41 4 (House ConstructionExpenditure) (0 4) (0.5) (0 5) (0 5' (0.5) (1.0) (0.6) (0.7) (0.7) (18.8) (3.4) tTotal ProductionExpenditures, (1 1) (1.5) (1 9) (2.1 (3.0) (3.4) (3 1) (4.4) (9.4O '37.1) (8.7) Average Per Capita DisposableCash Income 7 0 10.9 13.4 14.7 16 8 17.6 25 2 33.5 43.7 89.7 32.7

N Persons in Group 261 313 319 387 402 444 460 480 531 641 4,238 - 104 -

Table 38

Comparative Income Decile Status of Panel Households 1965-1975

DECILE 1975

DECILE 1965 1 2 3 4 5 6 7 8 9 10 Total Z

1 9 9 4 2 4 5 2 5 3 - 43 9

2 3 46 3 6 2 3 2 3 1 44 9

3 4 1 71 8 5 3 5 3 1 1 49 10

4 6 6 7 \8 3 9 3 4 2 50 11

S e~6~ ~ ~~ 4 ~ 5~ ~7 ~ 7_6~ ~ ~9 ~~~~~7 6 1 358 127

6 5 1 4 9 8 \7 3 3 9 41 10

7 - 6 7wd 3 5 1 5 5 50 11

8 1 2 4 2 1 6 11_9\5 48 10

9 - 1 1 3 1 1 t fr their3 44 9

10 - - - I - - 1 5 133 7

4ercent 8 10 10 8 10 10 11 12 11 10 100 deteriorated decisively. Some e hh more pronounced changes reflect dif- ferent stations in family life cycles: some relatively impecunious single people in 1965 are part of a wealthy extended household in 1975; heads of a large household in 1965 whose children have left to form their own families decline in income status.

4.28 Extent and direction of change vary considerably between occupa- tional groups: farmers have the highest percentage gaining higher decile status (46Z); traders have the highest proportion staying in the same decile (30%) and craftsmen most frequently lose in decile status (27%). Employees have the smallest proportion with worsening decile status (15%). - 105 -

Table 39

Comparative Income Decile Status of Panel Households by Professional Group: 1965-1975

Income Position Percent Farmers Craftsmen Traders Employees Total

Higher decile 46 40 22 40 43

Same decile 15 11 30 25 17

One decile lower 16 22 28 20 18

Two+ deciles lower 23 27 20 15 22 100 100 100 100 100

N Panel Households 336 37 40 55 468

Z HH in Category 72 8 8 12 100

N All Sample Households 522 49 97 105 773

Z HH in Category 68 6 12 14 100

4.29 Income gains and losses were next established in money terms. Ad- justing 1965 income to constant 1975 level (using the increase in consumer price index (+75%)), the amount of total income and or disposable cash income gained or lost was established decile by decile for the total panel, as well as for the four occupational groups. The 'total income' comparison highlights how much the trader and, to a lesser degree, the craftsmen subsamples obscure total panel figures. The significant drop in gross income for the majority (22/40) of panel traders and one third of panel craftsmen in fact reflects mainly a considerable drop in production expenditures, i.e., a change in the value of inventory now that the connection to wholesalers in Antalaha and Sambava is assured year round. Reported changes in disposable cash income (i.e., total income net of production expenditures and the value of self-con- sumption) more meaningfully express gains and losses: 80% of panel farmers gained an average of FMG 100,000 per household, 75% of craftsmen gained an average of FMG 115,000 per household; 64% of employees gained an average of FMG 245,000 per household and 60% of traders gained an average of FMG 635,000 per household. Conversely, 20% of farmer-households were worse off by FMG 35,000; 25% of craftsmen-households lost FMG 200,000 on average, one-third of employee-households experienced average income declines of FMG 135,000, and - 106 -

40% of panel traders reported a loss in disposable cash income of FMG 680,000 on average, compared to 1965. The analysis of panel households thus confirms the findings of the total sample comparison: tne majority of area farmers have experienced significant income gains, as have employees who advanced in the hierarchy; traders and craftsmen register sometimes substantial income decline; overall, income distributionis more equitable in 1975 "with the project" than it was in 1965 "without the project".

4.30 The effort to isolate the specific contributionof market proximity and good access to overall income gain concentratedon panel farmers as the most homogeneous group with respect to all factors other than distance from Andapa and the road. Per household incomes for this subsample in 1965 and 1975 were first calculated for the same distance categories as were used in Table 32 above for the total sample. (See Table 13, Annex F.) To facilitate comparison,and correct for different household composition in the two sur- vey periods, 1965 values were again adjusted to constant 1975 level and all figures expressed on a per capita basis. Distance group I (less than 5 km from Andapa and the road) being the only one to differ significantlyfrom the rest, the remaining four groups were combined for the comparison the results of which are shown below.

4.31 As a group, panel farmers have improved their income status con- siderably from the time the base survey was carried out. Compared to total sample incomes, panel farmers occupy the middle to upper middle ground (between the 6th and 8th deciles, cf. Tables 36 and 37). On average, per capita total income increasedby 38%, the relative share of the value of own produced goods for self-consumptiondropped by 15% (but see para. 4.6 above). Per capita cash income reported by the panel doubles and, as pro- duction expendituresincrease only slightly by 9%, per capita disposable cash income in 1975 is 125% above that in 1965.

4.32 There are considerabledifferences, however, between farmers in the close vicinity of Andapa and the road and those who are located more than 6 kilometers (an hour's walk or more) away from it. The former experiencean increase in total income of 58% (vs. 35% for the Basin households)with cash income increasingby over 120% and value of own-producedgoods self-consumed falling to three quarters that in 1965. Productionexpenditures of farms centered around Andapa drop by one third compared to 1965, while those for the farmers in the remainder of the study area increased by 17%. Taken together, these differentialchanges result in a tripling almost (+193%) of per capita disposable income for Andapa and road centered farm-households compared to an increase by 113% for householdsnot so advantageouslylocated.

4.33 Comparing actual amounts of per capita income in the two distance categories, it can be seen that those living close to Andapa started with some advantage in 1965 already, their total per capita income was 118% that of the remainder of the area: per capita cash income 146%, and per capita disposable cash income 131% (- production expendituresin Andapa were more than double those in the rest of the Basin at that time). While all panel farmers increased their income substantiallyby 1975, the Andapa centered - 107 -

farms increased their relative advantage over the rest - their total income is now 139% that of the second distance group, per capita cash income 171%, and per capita disposable cash income 180% (- per capita production expendi- ture in Andapa now exceed those in the basin by only 20%). -^tcould thus be argued that, as a rough measure, proximity to the road has added between 21 (total income per capita) and 49 (disposable cash income per capita) percen- tage points to income gains over and above those of farms in the less central parts of the study area. The high priority ranking Andapa Basin residents give to road improvement (see Chapter V below) indicates that they are well aware of its economic importance to them. Table 40

Average Change of Total Income of Panel Households,1975over 1965 - by 1975 Decile and Occupation (Constant 1975 FMG 000)

Occupation Farmers Craftsmen Traders Employees Total Net Average 1975 Increase Decrease Increase Decrease Increase Decrease Increase Decrease Increase Decrease Change Decile

1 11.8 45.4 _ 42.0 _ - _ 40.0 11.8 45.1 - 34.6 N HH 7 29 - 1 _ _ _ 1 7 31 38 2 32.4 40.8 8.0 84.0 29.0 239.5 67.0 81.0 32.5 69.2 - 11.7 N HH 22 14 1 3 2 2 1 1 26 20 47 3 39.5 43.1 33.7 - 23.0 124.0 - 61.5 38.2 57.8 - 2.3 N HH 22 14 3 1 3 2 26 19 45

4 47.6 106.8 35.5 4.0 - - _ 265.0 46.6 110.5 - 13.8 N HH 22 13 2 1 1 24 15 39

5 62.0 116.3 61.7 9.0 76.5 _ 70.0 - 63.4 89.5 + 50.6 N HH 35 3 3 1 2 4 44 4 48

6 71.8 110.3 105.4 12.0 102.0 _ - 22.0 76.5 73.0 + 60.o N HH 36 3 5 1 1 1 42 5 47

7 122.3 176.3 60.5 - 62.0 - 54.3 25.0 111.9 125.8 + 85.9 N HH 41 4 4 1 3 2 49 6 55 8 184.2 106.5 38.0 937.3 _ 133.1 108.8 145.0 157.9 609.5 - 51.3 N HH 27 4 1 3 4 12 4 40 55

9 283.2 - 264.8 332.0 313.0 5.702 144.0 255,7 260.2 3,893.2 - 717.0 N HH 27 5 1 1 8 6 3 39 12 51 10 524.2 - 599.0 703.0 1,549.2 3,459.4 747.4 362.6 889.3 2,120.9 + 273.6 N MW 31 1 InI 5 11 3 35 9 44

TOTAL 126.6 67.8 123.8 347.2 900.2 3,140.4 294.6 167.6 187.1 602.7 - 42.2 N HH 252 84 25 12 18 22 37 18 332 136 468

NET AVERAGE +:78.0 - 28.9 - 1320 + 143.4 - 42.4 CHANGE 336 37 40 55 468 Table 41

Average Change of Disposable Cash Income of Oane1 xouseholAs,1075 over 106R - hy 1075 Deci'e an, "ccupatlon (Constant 1975 FMr Ono)

Occupation Farmers Craftsmen Traders Employees Total Net Average 1975 Increase Decrease Increase Decrease Increase Decrease Increase Decrease Increase Decrease Change Decile

1 12.8 20.2 - 0.5 - - - 56.0 12.8 20.9 - 4.9 N HH 18 18 1 1 18 20 38

2 25.7 27.3 20.0 72.3 35.0 49.5 45.0 24.0 26.6 40.8 7.4 N HH 29 7 1 3 2 2 7 7 3.3 13 4.7

3 35.0 14.8 41.0 - - 38.0 40.0 149.0 35.9 27.4 10.6 N HH 23 13 41 1 27 18 45

4 42.9 42.5 44.0 16.0 _ - - 116.0 43.0 46.4 15.5 N HE 25 10 2 1 1 27 12 39

5 59.8 47.0 82.3 93.0 67.0 - 73.0 - 63.0 56.2 50.5 N HH 34 4 3 1 2 4 43 5 48

6 63.9 38.0 69.7 13.0 33.0 - - 16.0 63.7 27.7 50.1 N HH 35 4 4 2 7 1 40 7 47

7 110.0 84.4 89.5 - 129.0 - 63.0 42.3 106.6 71.8 74.1 N HH 38 7 4 1 2 3 45 10 55

8 150.2 87.5 77.0 1,446.0 84.0 267.6 79.6 116.6 126.0 235.6 33.9 N HH 27 4 3 1 1 3 10 7 41 14 55

9 233.3 16.0 226.1 - 335.6 981.6 122.4 175.7 226.2 600.8 1 47.8 N HH 26 1 6 3 6 5 4 40 11 51

10 475.9 47.0 263.5 - s 3 920.0 559.7 429.5 668.1 1,206.5 N HH 12 ______497.7 NH 12 1 _2 _ 12 _ 2 40440_4__ 494

OTAL 101.6 36.4 115.6 199.8 634.5 6. .5 137.5 153.2 157.2 zo.7 b9 26 24 16 35 20 354 114

73.3 38.9 109.2 106.9 77.6 336 37 40 55 468

% of HH 80% 20% 75% 25% 60% 40% 64% 36% 75% 25% - 110 -

Table 42

Per Capita income of Panel Farm Households, 1963,1975 - by Distance Category (Constant 1975 FMG 000)

Distance Up to 5 kms More than 5 kms Total Panel Item ategory from Andapa + Road from Andapa + Road

TOTAL INCOME - 1975 48.0 34.5 35.8 - 1965 30.3 25.6 26.0 / Change + 58% + 35% + 38X

OF THIS : SELF-CONSUMPTION VALUE - 1975 9.4 11.9 11.6 - 1965 12.9 13.7 13.7 % Change - 27% - 13% - 15%

CASH INCOME - 1975 38.6 22.6 24.2 - 1965 17.4 11.9 12.3 % Change + 122% + 90% + 97%

PRODUCTION EXPENDITURE - 1975 4.1 3.4 3.5 - 1965 6.2 2.9 3.2 % Change - 34% + 17% + 9%

DISPOSABLE CASH INCOME - 1975 34.5 19.2 20.7 - 1965 11.8 9.0 9.2 % Change + 193% + 113% +125%

N Households 1975 35 301 336 1965 33 303 336

N Persons 1975 191 1,714 1,905 1965 139 1,478 1,617

Average Household Size 1975 5.5 5.7 5.7 1965 4.2 4.9 4.8 % Change + 31% + 16% + 19% SECTION IV B

PATTERNS OF CONSUMPTIONEXPENDITURES - III -

B. PATTERNS OF CONSUMPTION EXPENDITURES

4.34 Under hypothesis 4 in Section II D above it was proposed to use the detailed Andapa survey results on household expenditures to test "Engel's Law" that relative expenditure foi (staple) food will decline as levels of income rise. The previous chapter documented a substantial rise in household incomes during the 1965-/5 period. In this section, therefore, comparative patterns of consumption expenditures for the two survey periods will be examined for conformity with Engel's Iaw.

4.35 First, the consumption expenditures reported (item by item) by the 1965 and 1975 sample households were summarized into 19 expenditure categories.l/ Among food items, self constumption of rice and "other local foods" were dis- tinguished, along with purchases of rice, other local foods, and imported food. Among other consumer goods, products purchased were grouped into six categories: other household products (such as soap, cooking fuel, etc.) textiles and clothing; adornment (e.g., jewelry, watches, etc.); entertainment (e.g., alcoholic beverages, cigarettes and tobacco, etc.); furnishings and furniture; and consumer durables (e.g., stoves, ice-boxes, sewing machines, etc.). Among services, expenditures for (rented) housing, education, health, and travel/transport were distinguished. Transfers comprise repayments of debts; gifts and other; taxes are shown separately because imrportantchanges in taxation have occurred during the study period.

4.36 Total sample consumption expenditures by category are detailed -.n Tables 6-10 in App. F. The proportion of total expenditure going to purchases of different types of goods in the "before/without road" base period and at the "after/with road" follow-up time gives a first summary indication of changes in overall expenditure patterns.

4.37 The summary sample data do not seem to bear out the hypothesis of decreasing expenditures on food with increasing income. Quite contrarily, total expenditures seem not to have increased in real terms, and the relative shares of food expenditures has risen by 11% - from 52.4% in 1965 to 63.5% in 1975. Expenditures for services also have increased slightly (+1.4%), while the shares spent on consumer goods (-7.1%), transfers (-2.5%) and taxes (-2.9%) have declined considerably.

4.38 It must be borne in mind, however, that these are 'total sample' comparisons not adjusted for sample composition. Transforming sample values into per household average expenditures for 1965 and 1975 (in a similar fashion as was done for incomes above), it can be shown that reiative share of food expenditures in total expenditures is indeed different for the four occupational groups (which also broadly represent four income classes) dis- tinguished in both surveys.

4.39 The differential shares of the five summary expenditure categories in total expenditures (derived from Table 45 below) for each of the two surveys

1/ For classification of products into product groups, see Appendix G. 1 1 -

-. 11

Prq)nt-4t c.n f Cons pi,n 4, <-f It.

D if 'n r an. L lp 4r , fSf- _

Expend-ituiro l.t-sms 965r) 1 91975

FOOD ^2.74 | __

Rice bc,9ht 7,1 1 1. .; - self-consused 15 . 27.1 Other I)rcql f.,I - I-ugnt 14., self-consumod 19.2 Imported food bought 1 .c 3 2

CONSUIMERGOODS - OTHTER 2: 16. 7

Household P'rolicms -. Ai.re Entertainnen t3 2.:

Adornment 0. 9 Consumer r 1 T r ninhinysFd i

ISERV-jG&i J ,3

Rent (Housing-) 2.4 Education 3G

Travel/ ,,'^ . ,o -- /.

TRANSFS - | 0

Do , *, oy, o i De3bt }p.iyrrlnxl,t I r). Gix:ts .. Othaer ).5

TOT-AL 00. i.0,. .

N EM; 000) 76 , 177327.7

1J lousaliolds nfep rLring 722 770

IN t) JrQ ;s 1 | / . ,i ffi^1l., J , b a, ,.,> ia,, 8

]F. ) Cin . 2,C'. - _. ... .

?c,) ;:--- - r *%< ,(. " 10) - 113 -

separately clearly seem to follow Engel's Law; the share of food expenditures in 1965 ranges from 71% for farmers (lowest income group) down to 31% for traders (highest income group). Similarly in 1975 farmers devote 79% of total expendituresto food, while traders' share is only 41%. For other consumer goods and services, the inverse trend prevails, with farmers noticeably low in expenditure for services both in 1965 and 1975. Traders by contrast spend almost as large a share on services as on food. The share of tax expenditures in totel expendituresreported d,ops for all groups: in 1975, farmers pay one quarter, craftsmen one fifth, employees one third and traders two-thirds of the share they were assessed in 1965.

4.40 Examination of the more detailed expenditurebreakdown in terms of average per household expendituresfor various types of goods reveals a basi- cally very stable pattern of expenditure from 1965 to 1975, apart from the increased share of food expenditurescompared to expenditureson other goods. Payments for services remain stable as a share of total expendituresexcept for employees, who increase consumptionof services and pay relativelyless for other consumer goods than in 1965.

4.41 The seeming 'suspension'of Engel's Law in the inter-temporalcom- parison can probably be explained by different price movements for different expenditure categories over the study period. The price of rice, the staple food accountingfor 50-75% of total food expenditures,more than doubled betwieen1965 and 1975. Food prices generally rose 90% above those paid in 1965. Prices for other goods, by contrast, are reported as having increased by 75% in the country as a whole. In Andapa, where high transport costs in 1965 contributedto untypically high cost of other goods at the time, the upward movement in price for these goods was counter-balancedby appreciable decreases in transport cost, especially for high volume staples (sugar,salt, canned food) and bulky materials. It is plausible that with the completion of the road, a more varied supply of non-local foods became available in Andapa for the first time, and that previouslymostly self-sufficientfarmers increased food expendituresfor a better diet in the first instance. While the declining share of self-producedversus bought local foods in 1975 compared to 1965 may partly result from differences in the way informationon food cQnsumptionwas collected, the value of imported goods purchased definitely increased in real terms: by 56% for farmers, by 34% for craftsmen,by 66% for employees and by 315% for traders. Regarding non-food expenditures,Andapa prices should have increased relativelyless than those in Tananarive,because they included sizeable transport components in 1965. As no quantity informa- tion on items purchased by households is available for 1965, it has not been possible to compare prices for selected items and derive a local inflation measure; it would seem, however, that purchases of other goods did in fact increase (see data on equipmentsowned by households, Section IVC below) even though expenditureshave not gone up accordingly.

4.42 Parallel to the analysis carried out on the income side, households were also classified into 'deciles of total consumptionexpenditure' to high- liglz differences in spending patterns with different levels of expenditure. Tab!-s 46 and 47 below show the characteristicsof households in the various PROPORTIONOF T,;JO.REXP,,NDLITTRE CATEGORIES IN TI)TALHOUSEHOLD CONSUMPTION_EYPENBJTmPJREBY CCCUPATIONAL GROUP 19~ 75

Paeraxut of Total Consurption Expenditure Expenditure Types F'AMT?MANi CP!.: 'TSMEN EMPLOYEES TRADERS

FOOD 1975 /9 71 54 41 1965 71 55 43 31

CONSUMERGOODS 1975 1 l 17 21 19 1965 118 23 32 25

SERVIc;ES 1975 8 22 31 1965 9 1:3 31

TRANSFERS 197$ , 2 2 3 1965 3 5 14

TAXES 1975 1 2 1 6 '165 l j 3 9

TOTAL '))100 100

FMG OOO/Household 1975 13721 158.8 278.6 1425.0 1.65 67. ! 92.9 167 0 238.5

1975 Exp. as ,% 6 of 1965 expendi-uures 169 178

i ~~~~~~~~~~~~~~~~~~~~~I , , . . , , . , ______N Households 1979 i i gn 97 R%orting . 19C 122 814 Table 45

AVERAGE PER HOUSEHOLD CONSUMPTION EXPENDITURES BY OCCUPATION&L GROUP: 1965-1975 (FMG 000)

Expenditure 1 9 6 5 Items 1 975

FARMER CRAFTSMEN EMPLOYEES TRADERS TOTAL FARMER CRAFTSHEN EMPLOYEES TRADERS TOTAL FOOD 48.0 50.7 72.7 75.7 55.5 107.9 113.1 , 149.7 173.8 Rice - 121.4 bought 3.6 8.5 1f.5 15.5 7.5 24.7 18.9 - self-consumed 20.9 13.0 J.4 35.0 27.6 25.8 Other local food 6.2 16.6 53.1 56.1 45.2 - bought 8.1 14.3 32.0 48.5 51.8 32.5 15.4 22.3 30.4 51.0 - self- 14.4 12.8 74.8 32.8 10.4 18.1 14.0 5.0 consumed 2.6 5.7 4.0 4.8 Imported food - bought 1.0 2.1 4.4 3.4 2.0 2.8 5.1 12.8 18.9 6.2 CONSUMER GOODS 11.8 21.8 54.1 59.2 25.3 19.3 26.3 59.6 79.0 32.0 Household Products - other 2.5 7.0 11.7 Entertainment 20.5 6.5 5.7 9.3 1.6 3.3 27.7 34.2 12.1 6.8 6.0 3.2 3.4 4.8 Textiles and Clothing 6.4 8.3 10.9 5.0 8.8 23.6 25.3 11.7 8.3 Adornment 10.6 20.9 26.6 12.3 0.3 0.4 3.4 1.5 Consumer Durables 1.0 0.4 0.2 0.8 0.3 1.2 3.8 1.8 0.6 Furnishings 3.1 1.3 0.8 0.6 & Furniture 0.7 1.1 0.7 2.2 0.9 4.8 2.8 1.6 0.7 0.8 1.2 3.3 1.1 SERVICES 2.7 8.2 22.5 73.7 14.9 6.4 13.1 61.0 133.2 29.2 Rent (Housing) 0.4 2.4 4.2 12.1 Education 2.6 0.3 2.1 9.4 0.7 1.0 3.0 9.5 2.7 Health 26.0 4.1 1.6 4.1 10.2 0.7 1.6 2.9 34.8 6.9 Travel/Transport 6.8 1.9 3.0 3.7 0.9 3.2 9.1 23.1 6.3 12.4 28.8 6.3 1.5 3.2 32.3 65 8 13.3 TRANSFERS 2.2 3.2 12.7 8.9 4.8 2.4 3.6 4.5 11.4 3.8 Debt Repayment 0.8 - 2.8 Gifts 0.9 1.1 0.2 0.1 0.8 1.8 7.5 - 0.1 0.1 Other 3.5 2.3 1.8 1.3 0.6 1.4 3.8 7.7 2.7 2.4 4.5 1.4 0.4 2.2 0.7 3.6 1.0 TAXES 2.7 9.0 5.0 20.9 5.6 1.1 2.5 3.8 27.6 4.9

TOTAL CONSUMPTION VALUE 67.4 92.9 167.0 238.5 106.1 137.1 158.8 278.6 425.0 191.3 of thise Cash consumptionexp. 32.1 67.1 147.0 214.2 75.5 79.0 100.1 227.7 372.5 13437 rable 46

CHARACTFRTSTTCS Or HoUSEHOLDS BY DECILES OF HOUSEHOLDCONSUMPTIONEXPENDITURES: 1965 I FMG UOUV

I T E M t E C I L E S 1 2 TOTAL SAMPLE 3 4 5 6 7 8 9 10 AVERAGE HOUSEHOLD CONSUMPTION EXP RANGE - UPPER 34.8 44 9 54 3 63 1 71.6 83.1 100 0 129.8 143 7 880 9 _ IN DECILE - LOWER 5 7 35 3 45 0 54.5 63 1 71.8 83.2 100.4 130.4 197.1 _ PER HOUSEHOLD CONSUMPTION EXPENDITURE 25 2 40.4 49 7 57.9 67 0 77.1 91.4 113.9 159.6 371 9 106.2 PER CAPITA CONSUMPTION EXPENDITURE 12.7 11 5 14 7 14 7 13.7 14.1 16.9 20.3 27.2 64 7 23.2 OF THIS: PER CAPITA VALUE OF SELF 6 8 5 8 7.4 7.6 7.3 6.6 CONSUMPTION 7.2 7.9 5.6 4.9 6.8

PER CAPITA CASH EXPENDITURES 5.9 5.7 7.3 7.1 6 4 7.5 9.7 12.4 21 6 59.8 16 4 PER CAPITA PRODUCTION EXPENDITURES 4 1 1 0 1.5 2.2 1.8 4.0 12.1 3.2 20.9 208 4 32.5 PER CAPITA TOTAL INCOME 12.9 10.8 14 7 14.9 14 7 16.9 28.2 22 5 47.2 284 0 56 0 HOUSEHOLD CHARACTERISTICS

HOUSEHOLD COMPnSITION

N Households 72 72 72 72 72 72 72 72 72 74 722 N Persons 143 251 242 283 350 392 389 402 422 425 3,296 N Adult Equivalents 115 3 197 0 191.0 220.2 270 0 294 0 307.0 315.4 323.7 333 2 2,566.8 OCCUPATION OF HEAD OF HOUSEHOLD

N Heads of Household : Farmers 61 66 55 55 60 52 46 42 20 3 460 Craftsmen 1 3 10 9 2 10 9 5 5 2 56 Traders 4 - 3 5 4 2 11 5 16 34 84 Employees 6 3 4 3 6 8 6 20 31 35 122 722 AVERAGE AREA CULTIVATED PER HOUSEHOLD (ha) NOT A AILABLE

DEMOGRAPHTC CHARACTERISTICS OF HEAD OF HH

N Heads of Household: Male 40 61 62 60 66 64 71 70 68 72 634 Female 32 11 10 12 6 8 1 2 4 2 88 722 N Heads of Household: Age 29 21 17 18 11 14 15 7 21 13 18 155 Age 30-49 23 28 28 37 36 35 41 32 40 29 329 Age 50 + 28 27 26 24 22 21 24 18 19 26 235 N A. _ _ _ 1I 1I I 7233 -722 N Heads of Household: Illiterate Primary Educ. Second.Educ NOT AVAILABLE Bac + N.A. Table 47

CHARACTERISTICS OF HOUSEHOLDSBY DECILES OF HOUSEHOLDCONSUMPTION EXPENDITURES: 1975

D I T E M E C I L E s TOTAL SAMPLE 1 2 3 4 5 6 7 8 9 10 VERAGE HOUSEHOLD CONSUMPTION EXP RANGE - UPPER 64.0 83.1 101.0 116 2 135.7 159.0 BY DECILE 190.7 229 5 326.4 2,553.5 - LOWER 0 64.1 83.2 101 5 116.3 136 1 159.4 191.0 229.9 332 7 - PER HOUSEHOLD CONSUMPTION EXPENDITURE 47 4 73 9 91 5 108.4 125.9 146.7 174.6 207.3 273 4 663.7 191.3 PER CAPITA CONSUMPTION EXPENDITURE 21 8 21 1 24 6 23.8 24.7 25.3 25.7 27.2 34.9 131.6 34.7 OF THIS : PER CAPITA VALUE OF SELF 8 9 9.8 10.5 10.3 10.0 10.5 CONSUMPTION 10.9 9.9 10.2 59.1 10.2

PER CAPITA CASH EXPENDITURES 12 9 11.3 14 1 13 5 14.7 14.8 14 8 17 3 24.7 72 5 24.5 PER CAPITA PRODUCTION EXPENDITURES 2 3 2.4 2.6 4.5 3.5 5.3 3 4 2 8 12 6 32.2 8.9 PER CAPITA TCTAL INCOME 34.1 32.7 35.4 38.4 33.1 35.9 36.1 37.4 62.6 130.2 54.0 HOUSEHOLD CHARACTERISTICS

HOUSEHOLDCOMPOSITION

N Households 77 77 77 77 N Persons 77 77 77 77 77 77 770 168 269 286 350 392 445 523 58S N 603 616 4,238 _ Adult Equivalents 138.1 223 7 2254 292 4 320.6 360.3 422 0 481.6 495.5 472.9 3,432.5 OCCUPATION OF READ OF HOUSEHOLD

N Heads of Household w Farmer 67 65 69 67 57 55 55 53 30 17 535 Craftsmen 3 5 2 5 8 10 4 2 8 2 49 Trader 2 2 3 2 6 7 8 9 22 35 96 Employees 5 5 3 3 6 5 10 13 17 23 90 77-0 AVERAG. AREA CULTIVATED PER HOUSEHOLD- ha) 1.58 1.92 2.29 2.16 2.38 2.57 2.55 2.65 2.66 3.39 2.42

DEMOGRAPHIC CHARACTERISTICS OF HEAD OF HH

N Heads of Household: Male 53 59 65 63 67 70 70 72 69 71 659 Female 24 18 12 14 10 7 7 5 8 6 111 N Heads of Household: Age 29 11 14 10 17 10 5 4 4 6 3 84 Age 30-49 26 36 33 24 37 40 41 36 39 40 352 Age 50 + 40 27 34 36 30 32 32 37 32 34 334 N A. ------770 N Heads of Household: Illiterate 42 39 33 37 27 27 24 21 8 5 Primary (read & 263 write) 27 35 36 35 43 39 46 41 46 45 293 Secondary + I - 2 1 1 2 2 6 12 23 Bacc + - - 50 ------I I N A 7 3 6 4 6 9 5 9 11 3 63 770 - 118 -

deciles. (Note that the consumptionexpenditure deciles differ somewhat in compositionfrom the 'total income' deciles: in Andapa as in other parts of the world, spending habits do not necessarily always concord with disposable income.)

4.43 The first thing to note concerning 1965 expendituredata is an excess of total expenditures (consumptionand production expenditures)over total income reported by households in all but the top two deciles. This results from the income shortfall on vanilla about a third of which, while sold, remained unpaid by the end of the survey period. However, on the strength of the promissory notes received, people bought on credit to satisfy basic consumptionneeds. Variations in consumptionexpenditures between 1965 and 1975 are thus much lower than reported variations in incomes.

4.44 'Consumptionexpenditure deciles' even more clearly than income deciles differ with household composition: average number of persons per household in the top decile in 1965 is 5.7, that in the bottom decile 2. Correspondingfigures in 1975 are 8 and 2.2. As most consumption expenditure is for the necessitiesof life, the level of expenditures is closely related to the number of mouths to feed and bodies to clothe and shelter.

4.45 Farm households again predominate in the lower deciles, while traders and employees are the majority in the top two deciles. Half of all female household heads are found in the two lowest consumption deciles in 1965; 38% in 1975. Average area cultivatedper household increases slightly but steadily with each decile from 1.58 ha in the lowest to 3.39 ha in the highest decile.

4.46 Although per household consumption expendituresincreased percep- tibly for each decile, the only real "jump" is between that of the ninth and tenth deciles. The top decile expendituresare 2.3 times the ninth and 15 times the lowest decile expenditurein 1965; 2.4 times the ninth, and 14 times the lowest decile expenditurein 1975.

4.47 Preliminary tests of Engel's Law are provided by the regression analyses reported in Tabl_s 48 and 49. Food consumptionand non-food con- sumption expendituresare the dependent variables, total consumptionexpen- diture (as proxy for income) and household size are the independentvari- ables. Estimates of income elasticitiesfor food and non-food consumption and household size elasticitiesare first establishedfor each of the four occupationalgroups and for both the 1975 and 1965 samples (see Table 48). The 1975 estimates (i.e., the a for variable X1, total consumption)reveal classic consumptionpatterns: all of the food income elasticitiesare positive but less than unity; dal etdspt one (fot farmers) of the non-food income elasticitiesexceed unity, and Id eVery case the regressioncoeffi- cients attain high levels of statiatidalsighificance. Also consistentwith theoreticalexpectation is the revealdd bthator of the elasticitiesacross occupationalgroups: the food eldatiditi§sdetline from farm to non-farm occupations,while the non-food elastititiesincresge, These results are in line with Houthakker'sreseerch on cohsUmptionpatterns. - I I f, -

i .1S Thcl I')/,,t uwlSC, IOI !( /ie eI.sti(itiei (i e . tthc: a Tur Vol l X,, household si ze) aI.l-oC(T'f1 tt1 thL find I ings of previous t esi-i;ttI I ;.

').1tlihkkv L (1957) poiilts ,u:, I nncc th w'i aror pa t I lnl i itiitleIs, I l L-f c t of a change in fam:;t . L,n'- oii "i11 Cotisumpt io n musxt be zero, nsl ,i ii '0_; -co'llsequellt_ I It i 1,01 possil)( I or Iilit "I f ui vaii J le .. alIil lsit iv' or aIl 1 Iva. Fol 1975, Table 48 shows all hownel i- 1i¾tLIl I -or f ood (confsolmpt. ior, to he posi t ivy It.1 css than nii LY , 1i, ,1 Ni(pt 011o1 (which in- i.t ;l. h;i(ll.Lyinsignificaint) aire negat ive lor i o'o0 &o, llsunipti ori.

'fi.'u InL 1( ' (,, l i.l1e less persuasive thari those for 1I>. ,', V, h(-ltlss broadly -wu ; i., withi cxpectation. Income elastitfl ie. it,(i tClid to be ½:;r . t, wn..lhilu thieyvare typically greater I tlhoioifl I,rIIon-food c(-lisumRp iII !)1w hotisehold size elasticities are also ge(,.I ' ,t eot witli thlc result That-l. the results for tthe base surve ni. pp f are more err. t i I 'l.n tbhose of 1975 cani probably be attributed t- it,;Lrt ions rest.iijt iI , roir,-o0ll-pa nellt,of area farnaers, for part of t - . i.llY(p.

*',1 'able 49 sUti'VO ci?.C the resul t s of ar, at t empt to test whet hr .r

`wr,%,c-dstatist Ical 'i t" nigjt be ach ieved from expandding tlie Ii si aanaturvI variab. es. i'SeIc. Laa of missing data in thte 1965 survvv, nl,'utho i d.sd ulrVeyed iii b.f ii nriods are included in the analysis. V. \ t' t.e-:Idi r e:1,\t i,(I fr-) 1ronilAndapa-) , X4 (-,i(lj, c! I I L ; r-s , u - n), X coiipi cied primary education: I = ves, C u-A;) ('e(1 5'cciolniry ('etu'.t ion: I = yes, 0 - no), and X (com', etd -. - I i rll1- 1 = yf S, = ne) t.reidded (tlhe responldent ror variabi; 1 II I'' X i s th-e honI:I; InIo Id I ). There is only slight overal] ir-,-i- r COrniiarud to . hlt- 41;, hlie patterns reported for variables X . i.- . vi lv rhoi u-.i, :, i lahie 1+8, wltit nio improvemient lii Ii. --.. . on.L-,!ly 1p `ab],(ulu fniasticities for food consutnption oi cral ,r'i eCplop ees in 19t6. Nonl :'. tho newly included variables X. - X, a . ".t ittical]l si .ni II I I,ei'csed distance from Andapa tends .i1 r 'irid ansi iicoeFr -oo-'icd consumlption0, perhaps attesting to inc>- "r,ull, s;" iS vil laices. lacaCt- d lar from Aildapa. Age of household 1w *.. ii,1-11A} S l-ar pa ttL_ I o . ff luence. Illiteracy (X seems to I)(-otf - ' ' ' i!.-. 1ni_ Vuci-dit'Vk'|S C)ransuinrpt ion in butl1 surveys, and wii t !I I ., d uom,wmt ce in V /'on] J. No recognizable directon it.1 .li O i-(s liii ' il . i ls '. X 8 thie eslimat-ed e]as.icitits L - -a l'l atz signJ niId I .,;L Atiral sigriifieanCe.

,.:. iI l 1 ,1lit 'rI I r,rI. . ibt,ve)I4 cst of relau t' V

i ?Cuea-*.i li< . i s,- nti,i uty va ijables might improve ovlers-] '-ii

,*, i , ' klI-Y; ,,ii. i iVat. C.:C'P du, to data constraints. Hewovi'' I i, i i1 -,; t-'4v hto t oncludcd t:'at cn.in"'1t,i i-i

1 .+;.i l'^ :,r nA-n. -i -n-s (.-.o:h a s l-anges ir, producor surp 1, - 120 -

Table 48 INCON ILASTICITIES FRFO(I) AND NON-FOODCONSUNPTION,

(Total 5aqle)

2 Y - Food Con!WtionX / I - Non-Food ConsuwptionV

Const. X1 31/ X2 RCont. 1 129' R^

Farmers s a 1.47 0.83 0.12 -0.45 0.88 0.02 (n-532) t (6.50) (40.65) (t.92) 0.85 (-0.82) (17.67) (0.39) o.b8 Craftsmen: J4.51 0.55 0.24 -3.80 1.22 -0.27 (n- 49) t (5.41) (7.51) (3.56) 0.76 (-2.17) (7.85) (-1.94) 0.61 2ployees: 4.64 0.53 0.30 -5.17 1.36 -0.29 (n - 90) t (8.84) (11.62) (6.18) 0.82 (-5.11) (15.40) (-3.11) 0.77 Traders a 6.26 0.41 0.25 -2.79 1.13 -0.10 /+ t^- oV' (Jr R^5 !" !, 066 (-2.81) (14.44) (-0.95) 0.70 - 121 -

Table 49 INCOMEELASTICITIES FOR FOODAND NON-FOOD CONSUWTION. BY OCCUPATIONALGROUP: 1975 AND1965 (Panel Households Only)

Y - Food Consumption -/

Const. X1 X2 X3 X4 X5 X6 X7 X8 R2

1975

Farmers : -0.01 0.93 0.07 0.02 0.08 0.07 0.08 -0.07 - (n = 334) t (-0.02) (36.57) (3.11) (1.11) (2.07) (2.32) (1.50) (-0.33) - 0.88

Craftsmen: a 3.22 0.71 0.15 0.05 -0.11 0.02 -0.02 0.42 - (n - 37) t (2.99) (7.41) (2.53) (1.29) (-0.65) (0.25) (-0.14) (1.88) - 0.88

Employee : * 3.08 0.61 0.31 0.06 0.04 0.44 0.41 0.41 - (n - 55) t (3.62) (10.43) (4.86) (1.60) (0.25) (1.71) (1.61) (1.58) - 0.88

Traders -a 5.61 0.42 0.21 0.11 0.10 -0.01 -0.06 0.10 -0.17 - (n = 42) t (4.89) (6.26) (2.26) (1.42) (0.50) (-0.05) (-0.22) (0.34)(-0.41)0.70

1965

Farmers : a -5.79 1.34 0.31 0.08 0.26 -0.15 -0.53 0.20 - (n * 321) t (-3.63) (9.70) (2.31) (0.66) (1.40) (-0.82)(-1.75) (0.17) - 0.38

Craftsmen: i 28.91 -0.55 -0.82 -0.10 -2.88 -0.61 -0.33 0.97 - (n - 34) t (4.53)(-1.17) (-1.14)(-0.24) (-2.15) (-0.51)(-0.21) (0.41) - 0.29 Employees : a 7.32 0.25 0.70 -0.35 0.08 -0.04 -0.38 0.04 (n - 48) t (2.50) (0.99) (2.69) (-2.46) (0.20) (-0.05)(-0.45) (0.04) -- 0.46

Traders : 5 5.53 -0.18 1.00 0.74 1.42 -0.94 0.32 1.19 1.01 - (n - 39) t (1.77) (-0.90) (2.49) (2.13) (1.92) (-1.00) (0.29) (0.88) (0.52)0.39

Y = Non-Food Consumption"

Const. X1 X2 X3 X4 X5 X6 X7 X8 R

1Q75

Farmers e -0.42 1.01 -0.01 -0.07 -0.26 -0.20 -0.22 0.32 - (n - 334) t (-0.45)(14.13) (-0.24) (-1.26) (-2.36) (-2.29) (-1.50) (0.57) - 0.52

Craftsmen : a -11.00 1.86 -0.44 -0.24 0.08 0.21 0.18 -0.82 - (n - 37) t (-3.89) (7.39) (-2.76) (-2.33) (0.18) (0.80) (0.47) (-1.40) - 0.74 Employees : a -4.06 1.41 -0.31 -0.18 -0.19 -0.68 -0.76 -0.48 - (a - 55) t (-2.22) (11.18) (-2.29) (-2.20) (-0.62) (-1.23) (-1.39)(-0.87) - 0.81 Traders : a -6.81 1.50 -0.20 -0.08 -0.06 -0.41 0.01 -0.20 0.15 _ (n - 42) t (-6.56) (24.61) (-2.44) (-1.16) (0.31) (-2.35) (0.03) (-0.80) 0.39 0.96

1965

Farmers : a -1.06 1.01 -0.06 0.04 -0.14 -0.01 0.04 -0.43 - (n - 321) t (-1.49)(16.30) (-1.08) (0.76) (-1.67) (-0.07) (0.28) (-0.82) - 0.54

Craftsmen : a -4.56 0.93 0.27 -0.11 1.01 0.08 0.49 0.10 - (n - 34) t (-2.80) (7.81) (1.46) (-1.09) (2.96) (0.26) (1.19) (0.17) - 0.82

Employees : * -3.47 1.23 -0.10 0.20 -0.05 -0.12 -0.06 - - - (n - 48) t (-2.44)(10.02) (-0.80) (2.98) (-0.26) (-0.31) (-0.14) - - 0.84

Traders : i 2.55 0.77 -0.13 -0.06 -0.29 -0.21 0.50 0.32 -0.44 - (n - 39) t (2.30) (11.00) (-0.88) (-0.51) (-1.10) (-0.61) (1.26) (0.67) (-0.62) 0.85

NOTE: 1965 Householdsfor which any item of informationwas missing were excludedfrom the regressionsresulting in a smallersample N for 1965.

1/ Food conusumptionincludes self-consumption of own-producedfood. 2/ lion-food consumption is based on cash expenditures only and excludes Ravings. - 122 -

4.52 Calculation of consumption expenditureson a per capita rather than a per household basis further underlines the basic similaritiesin the consumptionpatterns of eight tenths of area households. Per capita value of self-consumptionespecially is remarkably steady, fluctuating between FMG 5,800 - 7,900 in the lower 8 deciles in 1965, between FMG 8,900 - 10,900 in 1975. Everyone in the study area is adequately provided with the basic food staples; people in the top decile have food expendituresof a different order of magnitude from the rest because they purchase different kinds of goods besides the locally grown staples.

4.53 Per capita cash consumption spenditures increase decile by decile, but cash purchases in the eighth decile are only slightly more than double those in the bottom decile in 1965; in 1975, the eighth decile spends only one-third more than the first. On a per capita basis, consumptionexpenditure increases from 1965 to 1975 are much more marked for the 7 lower deciles where more than threequartersof total expenditures is for food, than for the upper three, who spend a proportionatelylarger part on other consumptiongoods and services.

4.54 Expendituresfor transport and travel show an interestingdevelop- ment from 1965 to 1975 as might be expected. In 1965, households in the four lowest deciles report no transport expendituresat all and those of the next four deciles are miniscule. Only in the top two deciles are transport services used to any appreciable extent, accountingfor 13.5% and 4.4% of total per capita consumptionexpenditure respectively. Air travel is a luxury available to few. In 1975, all deciles report expendituresfor some travel: the road has made travel possible for all income groups. But share of transport in total per capita consumption expenditurefor the top decile increases as well to 16.9% of total consumption expenditure: this is the group where some families own a vehicle, while those in the other deciles rely on bus and taxi services for their occasional trips.

4.55 To conclude: patterns of consumptionexpenditures in the Andapa region do conform to Engel's Law as between income groups for each survey separately but differentialprice movements of food and other goods between 1965 and 1975 distort the intertemporalcomparison. It may also be instruc- tive to revise the test for subsistence-orientedeconomies by including "imported foods" in the category of other consumptiongoods, as they will hardly be considered 'inferior'by local farm households.

i Table 50 PER CAPITACONSUMPTION EXPENDITURES BY CATEGORIESOF EXPENDITURESAND DECILES: 1965-1975 (FMG000)

E-xpenditure Category D E C I L E __ _ Total 1 2 3 4 5 6 7 8 9 10 Sample

FOOD 1975 17.3 17.3 19.6 18.8 19.5 19.4 19.7 20.0 22.1 35.9 22.1 1965 9.1 8. ? 10.8 10.9 9.9 10.1 11.3 12.9 14.5 18.7 12.2

OTHERCONSUMER GOODS 1975 3.4 2.8 3.5 3.6 3.4 3.6 3.9 4.4 7.3 14.8 5.8 1965 2.6 2.0 2.6 2.5 2.6 2.7 3.6 4.4 7.5 18.2 5.5

SERVICES 1975 o.6 0.7 0.9 1.0 1.3 1.5 1.6 2.0 4.0 26.0 5.3 1965 0.1 0.2 0.3 0.4 0.4 o.6 0.7 1.4 2.8 19.0 3.2

(OF THESE: TRANSPORT) 1975 0.2 0.2 0.3 0.3 0.2 0,5 0.3 0.4 0.9 14.1 2.4 1965 - - - - 0.1 0.1 0.1 0.3 1.2 8.8 1.4

MISCELLANEOUSTRANSFERS1975 0.? 0.1 0.6 0.3 0.3 o.6 0.3 0.5 0.9 2.6 0.8 1965 0.4 0.4 0.3 0.3 0.4 0.3 0.5 1.0 1.6 3.7 1.1

TAXES/LICENSES 1975 0.3 0.2 0.2 0.2 0.2 0.3 0.2 0.4 0.6 4.3 0.9 1965 0.5 0.3 0.8 O.6 0.5 0.5 0.8 o.6 0.9 5.2 1.2

TOTALPER CAPITA 1975 21.8 21.1 24.8 23.9 24.7 25.11 25.7 27.3 34.9 83.6 34.9 CONSUMPTIONEXP. 1965 12.7 11.2 14.8 14.7 13.8 14.2 16.9 20.3 27.3 64.8 23.2

N Persons in Decile 1975 168 269 286 39 523 586 603 616 4,238 1965 143 251 242 283 35° 389 402 422 425 3,296 SECTION IV C

SOME SOCIAL INDICATORS OF LEVEL OF LIVING - 124 -

C. SOCIAL INDICATORS

Introduction

4.56 While determinationof area household income and expenditurepat- terns were the main focus of the sample survey efforts, a range of physical and institutionalparameters were monitored also, to provide both separate measures of comparative level of living and additional insights into some of the determiiiantsof the economic changes observed. Two areas were selected for detailed analysis: land tenure patterns and agriculturalproduction techniques- because of their importantbearing on income distributionpat- terns in rural economies; and the quality of 'shelter' (housing,furniture and furnishing,utilities and facilities)- because it satisfies (or not) one of the basic human needs and tends to reflect changes in consumptionexpenditure patterns quite effectively. In addition, some demographicvariables were briefly examined as potential indicatorsof change in area population charac- teristics. - 125 -

1. Land Tenure Patterns

4.57 Given the importanceof export (cash) crops in Andapa region agriculture,it might be expected that some form of 'plantationagriculture' using wage labor on large estates would be found in the area. Inspection of data on landownershipand tenure proves otherwise. Both export crops and foodstuffs are grown by peasant farmers, mostly on their own land.

4.58 The 1964/65 surveys give no informationon size of holdings owned and cultivatedby sample households. There seem to have existed at that time two or three large estates of several hundred hectares each, but these have been required to sell most of their holdings to local cultivatorswho, in any case, produced the majority of area exports even in 1965.

4.59 The 1975 survey data show that over 98% of farmers, 90% of crafts- men, and 70% of traders and employees cultivate some land. Average size of holding reported is 2.7 ha. Only 57/701 (-8%) of farmers cultivatemore than 5 ha; 10/701 (=1.4%) cultivate 10 ha or more. The largest reported holding is 34 ha; the smallest 0.04 ha. On average, 43% of area farmed is planted to rice, the main subsistencecrop.

Table 51

Average Size of Holding and ProportionPlanted to Rice - 1975

N TOTAL AREA AREA PLANTED SIZE OF HOLDING OCCUPATION CULTIVATING CULTIVATED TO RICE MEAN MAX MIN (ha) (ha) ha ha ha

FARMERS 545 1,347.7 558.7 2.6 11.1 0.1

CRAFTSMEN 44 102.8 47.6 2.3 12.0 0.1

TRADERS 68 286.5 132.8 4.2 34.0 0.5

EME-OYEES 64 138.4 65.5 2.2 9.0 0.04

TOTAL 701 1,875.4 804.6 2.7 34.0 0.04

4.60 Two-thirds of total sample households own all the land they culti- vate, and over 80% own more chan half. The proportion of those renting land is lowest among farmers, highest among employees,many of whom have been in the area only for a shor time. (Establishedemployees and traders own the - 126 -

land they farm). The "landless` in t. .e Andapa region are clearly the advan- taged, not the disadvantaged: they are the traders and officialswho can af- ford not to cultivate their own crops.

Table 52

Proportion of Households Owning/RentingLand - 1975

LANDOWNERSHIP LANDOWNERSHIP FARMERS CRAFTSMEN~~~PE R TRADERSC E N T EMPLOYEES TOTAL

Own - do not cultivate 0.2 - - 1.1 0.3

Farm Own Land 72.3 55.1 54.6 43.3 65.6

Farm over 50% Own Land 18.4 20.4 6.2 8.9 15.9

Farm less than 50% Own Land 2.8 10.2 4.1 3.3 3.5

Farm All Rented Land 4.1 4.1 5.2 14.5 5.4

Own/Farm No Land 2.2 10.2 29.9 28.9 9.3

TOTAL 100.0 100.0 100.0 100.0 100.0

N Households 537 49 97 90 773

Of these: N Landless 12 5 29 26 72

4.61 Expressed as a proportion of total area cultivated by sample farmers, rented land constitutesno more than 11% (207/1,875ha). More than four-fifths of it is rice-land, only rarely are cash-crops grown on rented land. (This seems to be a persistent pattern: in 1965, 29% of land planted to rice was rented, but only 7% of coffee-land and 2.5% of vanilla-land 1/).

4.62 Data on land purchases and sales made by sample households over the last decade were recorded with a view to establishingwhether the opening of the road would result in greater numbers of land transactionsand, possibly,

1/ See: BCEOM 1966k pp. 4.67, 4.70, 4.73. - 127 - increased land values. Table 53 below shows clearly that selling and purchas- ing activity increased considerablyfrom the 1965-70 period to the 1971-75 "post-road"years, but other factors besides the road (e.g., dissolution of large 'plantation-holding')may have contributed to this change.

Table 53

Number of Households Selling and Buying Land - 1965-1975

ITEM FARMERS CRAFTSMEN TRADERS EMPLOYEES TOTAL

N Selling 1965-1975 40 6 3 3 52 of these: % Selling 1965-1970 22 - 33 33 21 % Selling 1971-1975 78 100 67 67 79

N Buying 1965-1975 112 18 18 20 168 of these: Z Buying 1965-1970 36* 17 55* 40 37* 4 % Buying 1971-1975 76 83 50 60 65 3=______N Households Sampled 537 49 97 90 773

* NOTE: Some famers buy in both periods - percentage adds to more than 100.

4.63 In all, 53 ha (-2.8% of total currently cultivated),were sold, and 198 ha ("10.6% of total currently cultivated)were bought in the transactions reported above. Employees and craftsmen purchased over half the land they currently cultivate during the past ten years; farmers and traders acquired about a fifth of their present holdings during the same period. On average, people who bought land in the past decade cultivate 4.25 ha; sellers by con- trast currently farm no more than 2.4 ha. Purchases and sales seem to move with family-lifecycles, with older cultivators selling off part of their holdings after their children have left the home.

4.64 The trend in the prices paid for land is more difficult to evaluate properly because there is insufficientinformation on the quality of land purchased. For the main crop lands, the range of prices paid over the past ten years shows an upward trend in the highest prices paid, though lowest reported prices (presumablyfor unimproved land) remain steady. Prices paid for building can exceed FMG 1 million/ha.

4.65 Overall, it can be concluded that land in the study area is not as yet a limiting factor. Family holdings mostly are kept to the size which the household can effectivelycultivate using family labor. Only a few of the - 128 -

Table 54

Acreage Bought and Sold - 1965-1975

ITEM FARMERS CRAFTSMEN TRADERS EMPLOYEES TOTAL

Hectares Sold 1965-75 41.7 3.8 4.5 2.5 52.5 of these: % Sold 1965-1970 30 - 67 20 31 % Sold 1971-1975 70 100 33 80 69

Hectares Farmed by Sellers 103.5 9.5 8.5 4.7 126.2

Hectares Bought 1965-75 115.7 28.8 17.6 36.9 198.0 of these: % Bought 1965-1970 42 24 36 53 41 % Bought 1971-1975 58 76 64 47 59 ------__---- Hectares Farmed by Buyers 504.3 59.6 89.1 61.1 714.1

Table 55

Price Ranges Reported for Land Purchases and Sales: 1965-1975 (FMG 000/ha)

/I CROP Before 1965- 1965-1970 1971-1975 Low - High Low - High Low - High

Rice n.a. 6 50 5 110

Vanilla 4 55 4 75 12 100

Coffee 5 20 3 200 4 100

Mixed n.a. 7 36 3 150

Other n.a. 5 200 2 1,850

/1 SOURCE: BCEOM 1966, p. 4.84. - 129 -

traders and employees rely mostly on hired labor for their farming. Young families "rent" land (often free of charge) from older relatives or from the "fokonolona"- community;as they work and improve it, they acquire title to the land or they purchase land from the proceeds of their work. Higher producer prices benefit area farmers directly and evenly: differentialsare due to more or less careful farming practices than to tenure differences.

AgriculturalProduction Techniques

4.66 The kinds of agriculturalequipments used by area cultivators as well as expenditureson different types of agriculturalinputs can serve as summary indicatorsof agriculturalproduction techniquesprevalent before and after the road was built.

4.67 Ownership of traditionalagricultural implementshas changed somewhat during the 1965-1975period. Basic hand tools continue to be most commonly used, although the number of sickles and machetes reported per household in 1975 is less than half that in 1965. Use of ploughs and (wooden) harrows is on the increase: one household in ten owned a plough in 1965, two householdsin ten do now; one household in ten had a harrow in 1965, three in ten do now.

4.68 A range of other equipmentswere reported in 1975 that did not appear in the i965 survey. Practicallyevery household owns a mortar for home food preparation,but there are now a few machines for husking rice and coffee commercially.

4.69 Transport equipment also has made its appearance. Over 10% of sample householdsown a bicycle, 4% have a motorcycle, and 1% own a car. (Details Table 10, Appendix E).

Expenditureson Agricultural Inputs

4.70 Analysis of the 1965 and 1975 expenditure patterns for different agriculturalinputs confirms the relative stability of farming techniques. The proportion of area farmers purchasing different types of inputs changes very little. Almost all expendituresare for traditionalinputs - labor, land, animals, basic tools. Modern inputs, even five years after the road ha- been opened, are quite insignificant: crops will grow well without them anu they are not used.

4.71 The greatest change that has occurred between 1965 and 1975 is a shift away from payments to labor and land "in kind" to payments "in cash". Along with this shift, a smaller proportion of farmers hire labor or cultivate rented land. The percentage of households purchasing animals and land during 1965 and 1975 is steady at 23% and 4% respectively. Fewer households purchase tools and seeds in 1975 than in 1965. - 130 -

Table 56

Equipment Ownership by Area Households: 1965-1975

Number Owned per Household

Item 1965 1975

Production Equipment

SHOVEL 1.3 1.4

ANGADY 0.7 0.8

SlCKLE 2.7 0.9

MACHETE (Coupe-Coupe) 2.0 0.8

HOE 1.2 0.5

HARROW (Wood) 0.1 0.3

FIHAVANA 0.5 0.5

PLOUGH 0.1 0.2

HARROW (Iron) 0.03 0.01

Food Processing Equipment

MORTARS n.a 0.9

RICE HUSKERS n.a 0.1

COFFEE HULLER n.a 0.01

Transport Equipment

BICYCLE n.a 0.1

MOTORCYCLE n.a 0.04

CART (Hand/Ox) n.a 0.02

BOAT n.a 0.01

CAR n.a 0.01

Other

RIFLE n.a 0.01

1I Source: BCEOM 1966, p. 4.111 - 131 -

Table 57

Proportion of Sample Farmers Reporting Expenditures For Different AgriculturalInputs: 1965-1975

ITEK ITEM ~~~~1965P E R C E N T 1975

Labor Wage - Cash 40.4 41.4 - Kind 7.0 -

Animal Purchase 22.5 23.8

Land Rent - Cash 7.3 21.5 - Kind 28.8 10.2

Land Purchase 4.4 4.0

Tools 59.0 52.3

Other (Anim/Eq.Hire) 3.3 1.4

Seeds/Plants 15.6 5.3

Insecticides/Pestic. 0.1 1.0

Fertilizer - 0.1

N Reporting 723 773

4.72 Examining the share of different types of expendituresin 1965 and 1975, a very stable pattern is shown for farmers, while there are appreciable shifts in allocation for the cultivatorswhose main occupation is outside farming. All of the latter incur much higher expendituresfor labor in 1975 than in 1965, while expenditureson purchase of animals decline considerably. Traders spend much less for purchase of land in 1975 than in 1965, while employees spend much more. Table 58 below details proportion of expenditure for different inputs, while total expendituresreported by sample households are summarized in Table 11, Appendix E. - 132 -

Table 58

Proportion of Expenditure for Different Agricultural Inputs: 1965-1975- By OccupationalGroup

Percent of Total Input Expenditures Expenditure Item Farmers Craftsmen Traders Employees Total

Labor Wage - 1975 28.7 35.0 71.0 54.3 47.7 - 1965 28.5 18.6 40.2 44.4 31.8

Animal Purchase - 1975 23.7 20.7 15.5 11.7 19.0 - 1965 21.7 29.8 17.9 25.6 22.7

Land Rent - 1975 25.1 25.3 4.7 9.6 15.8 - 1965 24.4 26.5 11.7 8.4 20.5

Land Purchase - 1975 12.1 14.6 3.0 16.8 9.5 - 1965 15.1 16.6 13.0 9.7 14.0

Tools - 1975 8.1 4.1 3.7 5.1 5.8 - 1965 6.6 4.1 7.5 4.5 6.1

Other (Animal - 1975 1.2 - 1.8 1.8 1.3 Eq. Hire) - 1965 1.3 2.6 2.1 2.4 1.7

Seeds/Plants - 1975 1.0 0.6 0.3 0.5 0.6 - 1965 2.8 1.8 5.6 5.0 3.0

Insecticide/ - 1975 0.1 0.1 - 0.1 0.4 Pesticide - 1965 - - 1.9 - 0.1

Fertilizer - 1975 - - - 0.2 - - 1965 -

TOTAL EXPENDITURE (FMG 000) - 1975 5,010.2 886.7 4,201.3 1,343.4 11,441.6 - 1965 2,270.3 285.8 360.2 605.0 3,521.3

N Households in Group - 1975 537 49 97 90 773 - 1965 460 56 85 122 722

N Households CultivatingLand - 1975 525 44 68 64 701 - 1965 Not Available - 133 -

4.73 There has been very little change in animal ownership over the study period. In 1975 as in 1965, most households owned chicken (though fewer than previously due to fowl pest in recent years) or water fowl. The number of bovine cattle reported has increased very slightly from 0.6 to 0.7 per household. There is an appreciable increase in pigs raised for sale, up from 0.1 to 0.7 per household: a consequence of the OPACA extension work in the region. But on balance, area farming remains thoroughlycrop oriented - besides the draft cattle, only animals that 'look after themselves'are raised (for details, see Table 17, Appendix E).

Table 59

Animal Ownership by Area Households: 1965. 1975

Number Owned per Household ITEM 1965 /L 1975

Bovine Cattle 0.6 0.7

of these: Draft Anim. 0.4 0.5 Breeding 0.08 0.08

Porcines 0.1 0.7

Fowl

Roosters, hens, chicken 9.4 4.8

Ducks 1.6 1.8

Geese 0.9 1.3

Turkey 0.04 0.06

/1 Source: BCEOM 1966, pp. 4.109-4.110.

4.74 The extent to which double cropping of rice is practiced also might provide an indication of changing production technology. Actually, the pro- portion of sample households reporting some double cropping of rice is still small: 33/701 or 4.72 of those cultivatingland. They include 29 farmhouse- holds, 3 craftsmen and 1 employee household. The proportion of double-crop- pers is the same in the Basin sample and in the sample road villages. As a group, they differ from the total sample in that they own less land (1 ha on average, compared,yith2.7 ha for the total sample) and have larger - 134 -

families (7.4 persons per household compared with 5.5 for the total sample). Being labor-rich and land-poor, this group adopted the new production tech- nology, but to assure self-sufficiencyrather than to generate a surplus for the market (less than 6% of their total production is reportedly sold).

4.75 In summary, it is thus neither the change in production tech- niques nor in input use that can explain increased returns to agricultural production. These seem to be due almost exclusively to the combined effects of higher producer prices and access to new markets (rice). Extent of area cultivated is currently determined by the amount of (family) hand-labor available to farm it. If double cropping of rice in the area to be irrigated under the OPACA project is to expand and production of rice for export to increase significantly,introduction of small-scalemodern equipment ("labor- extenders", "not labor-replacers")such as those developed by the IRRI (In- ternationalRice Research Institute) in Manila would probably be necessary. Yield reponses to improved planting, weeding, and harvesting techniques (the "efficiency differential"1/ between good and poor farming) are high, even without significantmodern input use. Prospects for the intensificationof cultivation by provision of improved technology appropriate to family farming seem promising, because present returns to area farmers are high enough to permit most of them to invest in small equipments.

/1 See: Shapiro, K. H.: Efficiency Differentials in Peasant Agriculture and Their Implicationsfor Development Policies;Discussion Paper No. 52, Center for Research on Economic Development,The University of Michigan, June 1976. - 135 -

2. Shelter

Area Housing

ConstructionStandards

4.76 Traditionally,houses in the Andapa region have been constructed of locally available materials. Rectangular in shape, they were built of wood. Floors were raised two-three feet above the ground (- there is frequent flooding during the rainy season - ), and usually a porch surrounded the house. The most frequentlyused material for wall-constructionwere ribs of the raphia palm (baobao) which can be fitted tightly (LO I) to provide quite effective shelter against wind, rain and cold. Traditional roofing was straw of about one foot thickness. In most houses, the kitchen was a separate building set apart from the living quarters. Grain stores, equally rectangular,were set on stilts more than a meter high, while the coops for chicken and other fowl were not raised above the ground.

4.77 No informationon the housing standards of the 1965 sample house- holds was collected. However, the AGRER 1964 report includes data on 690 houses in 9 villages 1/ in the Basin, which were used in the comparisonwith 1975 sample households. Changes in the standards of house constructionare indicated in the following three tables showing distributionsof flooring, wall, and roofing materials found in the houses surveyed. Data from the 1975 survey are shown separately,for the town of Andapa, where changes can be expected to be most pronouncedand for villages along the road where available materials are different.

4.78 There is a clear change of materials used in wall construction, with boards and other woods replacing traditionallyused raphia-ribs even in the Basin outside Andapa town. In the villages along the road, walls are often constructedfrom barkstrips woven in lattice patterns.

1/ Ibid, p. 122 ff. The nine villages were: Andapa-Brousse,Analambe, Ambodiangezoka,Andranomena, Tsaratanana, Amboahangihazo, Anjaribe, Antsahanongo,. - 136 -

Table 60

Wall Types of Houses Sampled - 1964-1975

| Percent of Houses Wall Material I 1964 1975 !_Wall_Material___iBASIN ANDAPA-TOWN BASIN ROAD TOTAL | | Baobao | 94.2 79.3 79.0 31.0 72.9 |

| Boards I 5.7 16.0 10.4 4.0 10.6 I

| Wood - other | - - 9.0 61.0 14.0 |

Earth - other | 0.1 3.4 1.2 1.0 1.6 |

| Brick/Stone - 1.3 0.4 3.0 0.9 |

| TOTAL L 100.0 100.0 100.0 100.0 100.0 I

I SAMPLE N | 689 150 521 100 771 |

|No Answer 1 - 2 - 2 |

Similar shifts can be observed in the floor materials used:

Table 61

Flooring Materials in Houses Sampled - 1964 - 1975 i 1 Percent of Houses IFloor Materials' 1964 1975I | Floor Materials BASIN ANDAPA-TOWN BASIN ROAD TOTAL |

| Board | 98.6 82.7 62.2 26.0 61.5 | I i I | Wood - other | 1.4 8.0 36.7 71.0 25.5 |

| Concrete i - 9.3 0.8 3.0 2.7 |

|Earth j - - 0.4 - 0.3 |

| TOTAL | 100.0 100.0 100.0 100.0 100.0 |

I SAMPLE N j 690 150 521 100 771 I

|No Answer j - - 2 - 2 | - i37-

One third of the 'Basin' households now use 'other wood' for floors instead of boards. In Andapa town and in the villages along the road 9% and 3% re- spectively now report concrete floors.

4.79 The most striking change is observed in the roofing material used. If the proportion of houses with corrugated metal roofs is truly the "quick" indicator of development it has often been claimed to be, things have improved amazingly during the 1965-75 decade:

Table 62

Roofing Materials on Houses Sampled - 1964-1975

Percent of Houses Roof Material 1964 1975 BASIN ANDAPA BASIN A-S ROAD TOTAL

Corrugated Metal 28.7 91.3 74.3 68.0 76.8

Other Metal - 0.2 - 0.1

Tiles - - - 3.0 0.4

Straw + Leaf Mat. 71.3 8.7 25.5 29.0 22.7

TOTAL 100.0% 100.0 100.0 100.0 100.0

Sample N 690 150 521 100 771

No Answer - - 2 - 2

The roportion of houses with corrugated metal roofs and with traditional thatch in the Basin has practicallybeen reversed:from 29% and 71% respec- tively in 1964, to 74% and 26% in 1975. In Andapa town, less than 10% of houses sampled had thatched roofs; along the road the correspondingshare is 29%.

4.80 Along with the change in house constructionmaterials used, there also seems to be a trend towards larger houses. While the traditionalhouse had one room only, the majority of houses in the Andapa Basin now have two rooms. The average for Andapa town is two-and-a half rooms per house. (For detailed breakdown of 1975 data, see Table 8, App. E). - 138 -

Table 63

Number of Rooms per House in Houses Sampled - 1964 - 1975

1964 1 9 7 5 Item BASIN ANDAPA BASIN A-S ROAD TOTAL

Number of rooms 891 370 941 138 1,449

it " houses 690 150 519 * 100 769 *

Average rooms/house 1.3 2.4 1.8 1.4 1.9

* NOTE: There were 4 no-answers.

4.81 All of the above analysis indicates that area residents have been in- vesting quite consistentlyin improving their living quarters. It should be pointed out that there is an element of "forced savings" involved here: fre- quent and heavy loirricanesperiodically destroy a significantproportion of buildings, making reconstructiona necessity. Over 80% of the houses in the 1964 sample and 67% of those in the 1975 sample were less than 15 years old; most had been "reconstructed"one or more times. Reconstructionusually in- corporates improvements,if the household can afford them at all. Purchase of a (preferablystone) house was the single item (402/742mentioning) most frequently mentioned as a priority personal developmentgoal (see paras. 5.5 ff below) by area residents.

House Ownership and Occupancy Rates

4.82 The 1975 survey examined current patterns of house ownership,but there is no comparable informationfrom the 1965 surveys. Over-whelmingly, people in the study area own the houses in which they live: only 9% rent accommodation. Renters are heavily concentratedin Andapa town and are mostly employees and traders - only nne percent of farmers live in rented accommodation (see Table 7, App. F).

4.83 Occupancy rates per dwelling unit vary by subareas. Andapa has a high proportion of single person households; 3-5 person households dominate in the villages along the road; Andapa again has the highest proportion of 6-8 person households,while the majority of extended family units (9 and over) sharing a common dwelling unit are found in the Basin. - 139 -

Table 64

Percent of Households by Occupants per Dwelling Unit - 1975

N of Occupants Percent per Dwelling Unit ANDAPA BASIN ROAD TOTAL

1 12.8 5.6 10.0 7.5

2 10.7 12.1 12.0 11.9

3-5 32.2 39.7 43.0 38.7

6-3 27.5 22.8 25.0 24.0

9 and over 16.8 19.8 10.0 17.9

TOTAL 100.0 100.0 100.0 100.0 - 770

N A 1 2 - 3

N Households 150 523 100 773

Facilities and Utilities

4.&4 A few summary statisticswill serve to indicate the types of facilities and utilities available to, and commonly used by, Andapa area residents.As was pointed out earlier (para. 4.76 above), kitchens in the Andapa region are mostly located away from the living quarters. The sample households show this to be true for about 60% of area households: - 140 -

Distributionof Types of Kitchen - 1975

Separate House 60

Separate Room in House 10

Cooking facility in living quarters 29

N.A. 1

TOTAL 100

N Households 773

Almost all area householdsuse wood for fuel to cook meals:

Distribution of Types of Fuel Used - 1975

Wood 96

Gas - butane 2

Kerosene 1

Alcohol 1

Charcoal

TOTAL 100

N = 773

4.85 LightinR is provided mostly by simple Kerosene lamps locally produced. In Andapa especiallyfancier (and safer) Kerosene-basedlamps are found more frequently. A few households have several hours of elec- tric light a night provided by (privatelyowned) diesel generators. - 141 -

Table 65

Distributionof Types of Lighting by Sub-area - 1975

Type of P E R C E N T Lighting ANDAPA BASIN ROAD TOTAL

Kerosene Lamp- local 51 77 71 72

Kerosene Storm lamp 31 16 23 19

Petromax lamp 15 7 6 8

Electricity 3 - - 1

TOTAL 100 100 100 100

N Households 150 523 100 773

4.86 In an area as abundant in rainfall as the Andapa Basin, water supply as such is not a problem. But the quality of the water used for human con- sumption very well is. The main source of water for most area households are the rivers meandering through the basin. Significantminorities have access to a public water supply (in the Andapa township). Also, spring water is used whenever available. It is fair to say, however, that people in the study area are not sufficientlyaware of the damaging effects of variously cont,aminatedwater supplies on their health and well-being. There is a pro- mising field for a major educationaleffort in basic sanitationprinciples. - 142 -

Table 66

Distributionof Sources of Drinking Water - 1964, 1975

1964 1 9 7 5 SOURCE BASIN ANDAPA BASIN ROAD TOTAL

River 66.2 22.9 69.6 98.0 54.4

Public, pipe 4.0 72.2 5.6 - 17.6 water, Supply, tap - 0.9 0.8 - 0.7

Spring 28.6 - 17.1 2.0 11.9

Pond 1.2 2.0 5.1 - 3.8

Rain-tank - 2.0 1.8 - 1.6

TOTAL 100.0 100.0 100.0 100.0 100.0

N.A. 30 6 9 1 16

N Households 690 150 523 100 773

The above table shows that not much has changed over the past decade in the Basin proper, but that the town of Andapa now predominantlydraws its water from a monitored supply system. Villages along the road almost exclusively use river water, which is of reasonable quality being much faster flowing than in the basin area, but often not easily accessible.

4.87 An amazing 111/773 sample households report shower facilities in the home. Two thirds of these are located in Andapa, the remainder in the basin. Toilet facilities are of a very basic nature. Most households re- port using a "hole" or "ditch" - but given the prevailing ground water levels, this is a rather optimistic description, - 143 -

Distributionof Types of Toilet - 1975

Hole 73

Ditch 7

Nothing, N.A. 19

W.C. I

100

N = 773

In a region, where waterborne disease is as prevalent as in the Andapa Basin, effective drinking water and sewerage control is likely to have a significant and immediate impact on residents' health and welfare.

Furniture and Furnishings

4.88 Respondents in the 1975 survey were also asked about the numbers and types of furniture and furnishings ("consumer durables") they possess. Results can again be compared with the 1964 data presented by AGRER. 1/

4.89 On all items but one reported in both 1964 and 1975, the average number of pieces of furniture owned per household has increased. The excep- tions are storage trunks or chests, which appear to be replaced by more sub- stantial cupboards. The average number of chairs and tables per household has almost doubled in the Basin; in Andapa it is three times the 1964 aver- age.

4.90 Ownership of lamps was already high (1.4/householdin 1964), but the increase of 0.3 lamps/householdhas been all in the more elaborate 'petro- max' type. None of the traditionallamps burning oils or animal fats were reported in use any more.

4.91 One household in five owned a sewing machiner in 1964 - and the same is true in 1975. But radio-ownershiphas increased tenfold: 6 out of 10 households in the Basin and along the road have a radio; in the town of Andapa 9 out of 10 households do. A whole range of products not mentioned in 1964 appear in the 1975 list, but most of these are found in the town only. Four percent of area households own a camera, one percent have a tape recorder and one percent use a (butane) gas stove. (For further details, see Table 9, Appendix E).

1/ AGRER 1964, pp. 125-6. - 11+4-

4.92 Overall, there is not only a larger spread in ownership of pre- viously available consumer durables, but also a wider range of new and dif- ferent goods imported into the region. The latter are mostly purchased by the urban (employee- trader) households in Andapa, although some products are beginnng to appear in the rest of the study area as well.

4.93 In summary, the quality of shelter has been improving over the study period, most notably so in Andapa town. Area residents are investing a sizeable portion of their increased income in housing. The road has par- ticularly induced wider ownership and use of previouslytoo expensive bulky items like metal roofing and cement foundations. With the possibility of transportingwood from the high mountain areas along the road, local produc- tion of furniture has become possible and has evidently found a ready local market: consumer durables and goods that help to make the house a home are owned by an increasingproportion of the Basin's population.

3. Nutrition

4.94 It was briefly pointed out in the chapter on data collection (para. 2.37 above) that detailed data on food consumptionof sample houLeholdswere gathered in the base survey. These would permit a very thorough analysis of the nutritional adequacy of the region's diet which, as consumptionpatterns seem quite stable, would still be potentiallyrelevant at this time. - 145 -

4. Demographic Characteristicsof the Study Area

4.95 As full results of the 1975 population census should be available in the near future, this report will outline just a few of the more salient demographic features of the study area.

Population Growth in the Study Area vs. The Country as a Whole

4.96 As mentioned earlier (para. 3 above) in the report, the Andapa Basin has only been settled since the end of the 19th century. Being an area of high agriculturalpotential, it attracted a steady flow of migrants and counted a total of 46,300 inhabitants at the time of the AGRER/BCEOM sur- veys in 1964/5. With the opening of the road, a renewed influx of people into the region could have been expected. It is, therefore, remarkable to find that population growth in the study area over the 1965-1975period is lower than that reported for the country as a whole: the total population of Madagascar increased from 6,650 million in 1965 to 8,833 million in 1975 - an increase of 32.8%; the population of the survey area went from 46,300 in 1965 to 59,700 in 1975 - a 29% increase. 1/

4.97 Urbanization in the Andapa Basin, however, was more pronounced than in the country as a whole. In 1965, 11.8% of Madagascar'spopulation was classed as urban (i.e. living in places with 5,000 + population). In 1975, the nroportion so classified had grown tQ 14.5%. Comparable figures in the stuay area were 8.5% in 1965, and 15% in 1975.

4.98 Sex and age distributionin Andapa again is quite similar to that in the rest of the country. The proportion of males and females is around 50% each. Age groups compare as follows:

1/ This discrepancymay be more apparent than real though: a detailed population count was made in Andapa in 1965, while it is said that the previous census underestimatedpopulation in the country as a whole. - 146 -

Table 67

Age Distributionof Population- 1975

Age Percent of Population (years) ANDAPA MADAGASCAR 1/

0-4 16.1 17.6 5-9 16.0 10-14 14.2 15-19 12.6 58.9 11.0 57.3

20-29 10.5 15.3 30-39 9.5 9.8 40-49 8.1 7.2 50-59 7.1 5.0 60 + 5.9 5.4 100.0 100.0

Sample N 4,174

Universe Population 59,726 8,883,000

I/ See: Report No. 1099a-MAG:MADAGASCAR - Economic Memorandum on Current Economic Position and Prospects and Selected Development Issues, World Bank, December 20, 1976, Table 1.2.

These are young populations: close to 60% are under 20 years of age. The slight discrepancy in the 20-39, and the 40-59 age groups is probably ex- plained by the structure of the Andapa sample, which included a high share of re-interviewswith b'eadsof households included in the base line survey in 1965 and thus is slightly biased towards older age groups.

Demographic Characteristicsof Area Population: 1965-1975

4.99 Examination of the study area's population profile shows only small changes in ethnic compositionbetween 1965 and 1975. The largest group are the Tsimihety, whose share of population (61%) is higher than that of heads- of-household (57%), pointing to larger families in the longest established and numerically dominant group. Change from 1965 when 50.9% of heads of households were Tsimihety, indicates a significantgain in their majority.

4.100 The two next most importantgroups are the Betsimisarakaand the Merina/Hova. T-e former remained stable with 11% of heads of households - 147 -

in both 1965 and 1975. The proportion of Merina/Hova declined from 11% to 8% of all heads of household over the ten-year period. There also was a significantdrop in the share of non-Malagasyheads-of-household from 6.2% in 1965 to 2.2% in 1975. (See Tables 1-2, Appendix E, for details on distributionof heads of households in the sample by age, sex and ethnic groups.)

4.101 Occupationaldistribution of heads of households in the Andapa Basin remained quite stable from 1965 to 1975:

Table 68

Percent of Heads of Household By Occupation

OCCUPATION 1965 1975

Farmers 86.7 86.9

Craftsmen 4.7 3.8

Traders 4.2 4.9

nIployees 4.4 4.4

100.0% 100.0%

Universe household N 9,900 12,430

Sample household N 723 693

There is a gain of almost 200 traders over the ten years, while the number of craftsmen remained almost exactly the same (468 to 470). The propor- tion of employees remains stable. The majority of farmers (close to 87%) is, if anything, understated:most craftsmen and two-thirds of the traders also farm.

4.102 No data on educationalstatus were collected in the 1965 survey. 1975 survey results show that adult men are more educated than women, while in the population of school-age,literacy levels of boys and girls are very close. This difference between age groups confirms improvements in schooling also noted in Section III A (paras. 5 ff) above. - 148 -

Table 69

Education Level of Andapa Population

by Age and Sex: 1975 (Percent)

Literacy 5-14 years 15 + years Total - All ages Level M F M F M F

Illiterate 19.2 19.8 25.2 37.8 29.0 43.3

Primary School 78.7 79.6 62.6 59.0 62.9 53.8

Secondary School 2.1 0.6 12.1 3.1 8.0 2.8

Baccalaureat - - 0.1 0.1 0.1 0.1

TOTAL 100.0 100.0 100.0 100.0 100.0 100.0

SAMPLE N 530 494 1,037 1,463 1,695 1,603

No Answer 112 120 110 249 421 455

4.103 Table 69 would seem to indicate that at least 20% of the school-age population ("illiterate")and probably up to 33% ("illiterate"+ "no answers") are not currently enrolled in schools. There is a significantdifference in education level between the population of the Andapa Basin and that in the villages along the Andapa-Sambavaroad, where 43% of males aLud55% of females were illiterate.

4.104 Economic activit reported for the total sample population in 1975 shows some interestingfeatures. Two-thirds of males aged 15 and over, and 90% of those over 20 years old report being economicallyactive. Of these, more than a quarter (220/762) pursue a second activity (e.g. work as farmer and craftsman). The comparable figures for females over 15, are 42% economi- cally active; for those over 20 years of age that rate is 52%. 69/460 or 15% report a secondary activity as well, and 'housewives'also pursue secondary activities,mostly crafts and trade. Female labor-forceparticipation is thus amazingly high. There were close to two females for every three males giving their primary occupation as farming. Craftsmenwere equally divided (34 male/34 female) in the primary activity, but more females than males claimed handicraftsas a second job (if housewives with a second job are included). Employees are mostly male. - 149 -

4.105 In each occupation,modal age of males was higher than that of females. Distributionsshowed the highest concentrationof male farmers in the age groups 40 and over, while females were spread fairly evenly from the twenties to the fifties. Craftspeoplewere mostly young and normally (Gaussian)distributed through the age ranges. Traders were more often older men or younger women. Employees were young and mostly male. (For details, see Table 6, Appendix E).

4.106 Compared to other areas of Madagascar, the populationof the study area is thus average with respect to age and sex, and urban/rural dwellers. Ethnic composition is much more varied than in most regions. Edu- cation level is comparativelyhigh; so is labor force participation,es- pecially that of women. CHAPTER V

ATTITUDES AND OPINIONS ON THE IMPACT OF THE ROAD,

AND ON INDIVIDUAL AND COMMUNITYDEVELOPMENT PRIORITIES - 150 -

V. ATTITUDESAND OPINIONS ON THE IMPACT OF THE ROAD, AND ON INDIVIDUALAND COMMUNITY DEVELOPMENT PRIORITIES

Road Impact

5.1 Survey respondents in 1975 were explicitly asked their opinion about the impact of the new road. Table 70 below shows that, overall, the econom_c of consequencesof _ne road outweighed other social impacts in the perception the target population. Only the households in the road-side villages mentioned social impacts more often than economic.

5.2 Among occupationalgroups, three-quartersof the traders, three- fifths of craftsmen and over half the farmers credit the road with positive economic impacts. For employees, social and economic impacts are about equally important. About 5% of respondentsfeel that the road has 'made no difference' in their lives and one half of one percent think it has brought in its wake higher prices and lower levels of living.

5.3 The opinion survey thus underlines what the origin/destinationsur- veys indicated as well: Andapa area residents use the road ext ,isivelyand appreciate the positive contributionit has made to the regiorrf economy and to closer integrationwith the rest of the province and the co'r Wry.

Table 70

OPINIoNlS ON IMPACT OF THE ROAD - 1975

Percent Respondents Mentioning

Type of Impact OCCUPATION IBASIN ROAD FRESCRAFTSMEN TRAllRS IEMPLODYEES ANDAPA 54. 6 .6 O Economic Impacts 51.2 61

Improved opportunity for sales 16.? 22.1 21.9 23.6 15.0 of products 20.3 28.6 3h.0 Improved.access to markets 14.9 16., 20.7 13.3 15.5 18.5 15.7 10.0 (for goods/labor) 9.6 13-9 9.2 5.0 Higher level of living 8.4 12.3 12.4 12.2 6.7 7. 9.1 Cheaper goods 7.6 4.1 8 46.7 LhO.5 _3.8 39.1 56.0 Social Impacts 439 32.6 19,5

26.7 26.5 19.9 25.5 42.0 Easier Travel 29.8 14.3 1h.th 12.2 7.5 7.3 8.0 5.0 More visits/cultural exchanges 7.L 6.1 L.1 7.8 6.5 6.6 5.9 9.0 Improved access to health serr. 6.7 12.2 1.0

1.1 4.6 8 1.0 No ifference 47 6.1 4.1 - Lower Level of Living 0.2 - 1.0 30.6

100.0 100.0 100.0 100.0 100.0 Total 100.0 100.0 100.0 90 773 151 522 100 N Responding 537 4&9 97 - 151 -

5.4 Reports by the 1975 sample on trips made and visitors received be- fore and after the opening of the road confirm this picture: For the 1965-1970 period, 482 of all householdsreport making no trips, and 66% receive no visitors. For the 1971-1975period, these proportionsdecline to 302 and 412 respectively. Those who travel averaged 3.4 trips in the 1965-'970per- iod, 10.8 during 1971-1975. Average number of visitors received doubles from 3.2 to 6.6. The gain in mobility is most marked, however, for traders:

Table 71

Average Number of Trips Reported by 1975 Sample Household Before/AfterRoad Opening

N of Trips Reported Occupation 1965-1970 1971-1975

Farmers 3.4 14.9

Craftsmen 2.1 4.8

Employees 4.0 12.6

Traders 4.0 29.4

N Reporting Trips 404 539

TOTAL N 773 773

Poals for Development: Individual and Community Priorities

5.5 This final section summarizesthe answers by 1975 sample respondents to questions about their aspirations for t.iemselvesand their families, and about the most desirable improvementsfor the communitiesin which they live.

IndividualPriorities

5.6 People's goals for themselvesand their families were established by asking an (open-ended)question about what they would most want to achieve in their lives. The total sample ranked the following ten concerns most im- portant: - 152 -

Table 72

Individual Priorities: 1975

1. trchase of a house 54

2. Purchase land 37

. :P-..rchaseagricultural implements 30

4. Purchase livestock 27

9 !tart a business 26

Farm more land 16

7. Enlarge business 13

8. Purchase draft animals 11

9. Purchase motorized agric. equipment 10

10. Purchase high qualIty food 10

N Reporting 742

N in Sample 773

5.7 Priorities differ somewhat by location of households. In Andapa, purchase of agriculturalimplements is mentioned most frequently,with buy- ing a house second and buying land third. Householdersin the Basin rank purchase of a house (52%) and purchase of land (38%) highest in priority, but about one third want to purchase livestock or start a business. Better bousing is a concern for 80% of householdersalong the road, followed by the desire to farm (rent) more land (36%), or purchase land (30%). 9% of residents along the road would like to buy a car.

5.8 The variation in personal goals is greatest among different occu- paticnal groups. For farmers, acquiring a better house ranks highest (57%); purchase of land second (39%) and purchase of livestock third (31%). Crafts- men want tools most, then to start a business, and third to buy a house. Next to purchasing a house, employees seem most interested in expanding their farming activities (purchase implements,especially motorized; pur- chase land). Traders, as would be expected, have as first priority enlarg- ing their business; they rank buying a house second and buying land third. Table 73 below presents details on 'minoritypriorities'. - 153 -

Table 73 Individual Priorities ly Location of Household and by Occupational OrouP - 1975

PRIORITIES Percent Respondents Mentioning

PRIORITIES _ OCCU ATION TOTAL LOC -TION

FARMERS CRAFTSMEN TRADERS EMPLOYEES ANDAPA BASIN ROAD FOOD 10.2 8.7 14.4 2.3 9.7 2.8 12.0 8 Housinga Purchasehouse 57.1 37.0 44.4 55.7 54.2 44.0 51.9 80 Improve house 2.3 2.2 3.3 5.7 2.8 5.7 2.4 1 Purchasefurniture 6.2 4.3 3.3 1.1 5.1 - 4.2 17 PurchaseHH goods 9.8 8.7 6.7 3.4 8.6 6.4 5.8 26 AgricultureI Purchase Land 39.4 34.8 23.3 37.5 36.9 36.9 38.3 30 PurchaseAg.eq.Tools 28.2 41.3 11.1 51.1 29.6 61.0 25.5 6 n " Motor 6.2 23.9 6.7 27*3 9.8 30.0 5.2 5 Ox/hand cart 1.4 - 1.1 1.1 1.2 .7 1.6 - Farzmmore land 18.4 10.9 4.4 13.6 15.6 7.8 13.8 36 Livestock:Purchase/axt.live- stock 30.7 26.1 17.8 15.9 27.1 10.6 33.5 18 Purchasedraft animals 12.9 17.4 6.7 3.4 11.3 8.5 13.2 6 Purchase other animals 1.9 - - 2.3 1.6 2.1 1.6 1

Commerce:Start business 28.2 37.0 17.8 18.2 26.3 12.1 31.3 21 Enarge business 6.6 10.9 56.7 5.7 12.8 12.1 14.8 4.

Handicrafts:Eblarge craft .8 13.0 - 1.1 1.5 4.3 1.0 - Purchasecraft materials .6 2.2 - 1.1 .7 1.4 .2 2 Credit 1.2 2.2 1.1 3,4 1.5 2.8 .8 3

Education: Continue studies 1.0 - - 1.1 .8 3.5 .2 -

Bettereduc. for children 4.1 - 3.3 8.0 4.2 5.7 4.2 2

Travel:Purchase car 4.4 10.9 16.7 9.1 6.9 7.1 6.4 9 rurchase bicycle .4 - - 1.1 .4 _ .4 1 Purchase easier travel 2.3 - 3.3 2.3 2.3 1.4 2.0 5 Tripsand visits 1.9 - 2.2 2.3 1.9 2.8 1.8 1

Other 2.3 - 4.4 2.3 2.4 2.8 2.4 2 ------_ ------_ ------_ ------…-- _ _ --_____ -_ ------GrandtotAl mentions 1,442 134 224 243 2,043 384 1,375 284

N Mentioning= % base 518 46 90 88 742 141 501 100

TotalN 537 149 97 90 773 151 522 100 - 154 -

Community Priorities

5.9 The list of priorities for community "development" goals differs compietely from individual preoccupations. Reflecting the present availa- bility - or absence - of social, and technical services, community priorities differ much more by location of households than by the occupation of the household heads.

5.10 The most important needs (expressed by more than five percent of respondents) mentianed by the total sample were:

Table 74

Community Needs: 1975

1. Roads 79

2. Medical Services 63

3. Schools 53

4. Electricity 37

5. Drinking Water 17

6. Motorized Agricultural Eq. 10

7. Credit 10

8. Water Management & Flood Control 9

9. Veterinary Services 6

N Reporting

N in Sample

5.11 Looking at priorities by location of households, the need for im- proved roads is highest for residents in the Basin (89%) and in Andapa (76%), who have long been promised better communications within the region itself, but a low 31% for households in roadside villages. The latter put highest priority (85% of mentions) on schools, which rank third in the Basin and fourth in Andapa where facilities are much more adequate. Health services are second highest priority in all locations. - 155 -

5.12 Electricityranked third in priority in Andapa, fourth in the rest of the Basin, and fifth in the villages along the road. For the latter, drinking water is of more concern (30% of mentions), which takes 5th priority (17%) in the Basin and is ranked 8th by Andapa residents,most of whom have access to public water supply.

5.13 A sizeable 10% of mentions from households in the Basin and along the road calls for water management and flood control. 10% and 12Z in Andapa and the Basin respectivelyexpress a need for motorized agriculturalequip- ment (felt evidently beyond the reach of individual households). The need for credit agair is voiced by an even 10%, 10% and 11% in Andapa, the Basin, and along the road. Veterinary services are of greatest concern in roadside villages.

5.14 Among occupationalgroups, there were few divergences from the over- all pattern of commuaal priorities. Employees (who mostly live in Andapa) ranked electricitysecond just above health facilitiesand schools. Traders ranked access to credit fifth. Craftsmen put higher priority (5th ranking) on acquisition of motorized agriculturalequipment than farmers. Employees and traders are more aware of the need for soil protection than craftsmen and farmers.

5.15 Table 75 below summarizes communal development goals and shows rank- ings in detail.

5.16 In sum, area residents' personal concerns are with improvementsin their material well-being. Most have no real worry about food, but they want better housing and the greater security and prestige that goes with ownership of land and cattle. A significantminority see trade as the best avenue to improve their level of living. Aspirations for community improvementsare high. Area residents have travelled on the paved road to Sambava; seen public drinking water installed in Andapa town; and watched a few wealthy households generate their own electricity. They know what health and education facili- ties are available to residents in the district capital. They aspire to a similar level of services for their own communitiesas well.

5.17 There also is a healthy awareness of the need for improved agri- cultural production techniquesas a condition of continued improvementsin output of the region's major crops. As population begins to press on avail- able land resources,awareness of the need for better land management is be- ginning to emerge as a concern.

5.18 Overall, the list of prioritiesfor personal and community develop- ments reflects the aspirations of a peasant society well along the road to a 'modern' community. Sensitive to (world) markets and their functioning through long years of cash crop production, the area population is increas- ingly turning earlier subsistence sectors into new parts of the cash economy. This process may become further accentuated,if the region can be more closely integrated into the national economy. But probably a link to the western road net and Tananarivewould be needed to really change its present 'enclave' status and transform the Basin into a dependable granary for some of the deficit regions of the country. Table 75

CommunityPriorities by Location of Householdsand by Occupationof Head of Household

Percent Respondents Mentioning

PRIORITIES . OCCUPATION TOTAL -CATION FARMER CRAFTSMENTRADERS EMPLOYEES ANDIAPA BASIN ROAD

INFRASTRUCTURE& SOCIAL

Roads 78.7 87.5 78.9 76.7 79.1 75.9 89.1 31.3 MedicalServices 66.1 62.5 51.6 59.3 63.3 74.5 59.3 67.7 Schools 54.4 47.9 46.3 52.3 52.8 26.2 53.9 85.9 Electricity 30.5 45.8 44.&2 60 5 36.6 65.5 31.2 22.2 DrinkingWater 20.0 8.3 11.6 8.1 16.8 6.9 17.1 30.3 AGRICULTURE Water Management and FloodControl 10.9 4.2 6<3 4.7 9.2 1.4 11.2 10.1 Soil Protection 1.7 2.1 3.2 4.7 2.2 h.8 1.6 2.0 MotorizedAgric.Equip. 10.2 16.7 11.6 7.0 10.4 10.3 11.6 4.0 VeterinaryServices 6.4 6.2 3.2 4.7 5.8 3.4 6.0 8.1 Fertilizer 1.5 - - - 1.1 - 1.2 2.0 INSTITUTIONAL Credit 8.3 6.2 21.1 9.3 9.9 9.7 9.7 11.1 Self-Management 2.6 2.1 12.6 2.3 3.8 9.0 2.3 4.0 MISCELLANEOUb 7.0 6.2 3.2 10.5 6.8 6.9 4.8 17.2

TOTAL MENTIONS 1,584 142 279 258 2,263 427 1 ,543 29i

N Mentioning % base 531 48 95 86 760 145 516 99 N in Sample 537 49 97 90 773 151 522 100