Senegal - An assessment of living conditions (Vol. 2 of 2) : Annexes

Report No. 12517-SE Senegal An Assessment of Living Conditions (In Two Volumes) Volume II: Annexes May 5, 1995 Africa Region Western Africa Department Country Operations '4~~~~~~~~~4 4_ ~ - _ _  Currency Equivalent Currency Unit: CFA Franc Period Average: 1992 - CFAF 264.69/USS Period Average: 1994 - CFAF 555.20/US$ ACRONYIS AND ABBREVIATIONS AGETIP AAgence d'Execution des Travaux d'IntEret Public BCEAO Banque Centrale de Etats de l'Afrique de F'Ouest CCF Christian Children's Fund CNCAS Caisse National de Credit Agricole du Senegal CONGAD Conseil des ONG de Developpement CPSP Caisse de Perequation et de Stabilisation des Prix CSA Commissariat a la Securite Alimentaire DAARA Islamic School teaching the Koran DPS Division de la Prevision et de la Statistique DRC Domestic Resource Cost FDEA Femmes, Developpement, Entreprise En Afrique FED Fonds Europeen de Developpement FONGS Federation de Organisations Non Gouvemementales du Senegal GDP Gross Domestic Product GOS Government of Senegal IBRD Intemational Bank for Reconstruction and Development IDA Intemational Development Agency IFRPI Intemational Food Research Policy Institute ISRA Senegalese Institute for Agricultural Research NGO Non-Governmental Organization NPI Nouvelle Politique Industrielle ONCAD Office National de Cooperation et d'Assistance pour le Developpement PAGD Programme d'Appui a la Gestion et le Developpement SAED Societe Nationale d'Amenagement et d'Exploitation des Terres du Delta du Fleuve Senegal et de Vallees du Fleuve Senegal et de la Falemee SAL Structural Adjustment Loan SAR Societe Africaine de Raffinage SMIG Salaire Minimum Interprofessionel Garanti UMOA Union Monetaire Ouest Africaine UNDP United Nations Development Program UNICEF United Nations Children's Fund USAID United States Agency for International Development  ANNEX A - 1 TECHNICAL NOTE ON PROCEDURES AND DATA SOURCES FOR ESTIMATING REGIONAL POVERTY LINES FOR THE SENEGAL POVERTY ASSESSMENT 1. Methodology 1.1 Data sources. There are two primary sources of information on household income and expenditure in Senegal from which to estimate poverty lines: the first SDA Priority Survey in 1992 (PS), covering 10,000 households; and a much smaller sample (296 households) IFPRI/ISRA survey conducted over two agricultural years (1988-90 with some data from 1991) in the Groundnut Basin, Senegal Oriental (Tamabacounda) and Kolda. The PS represents the largest household survey ever conducted, and thus offers a high degree of confidence, even at the regional and sub-regional level of disaggregation, but collected information only on households' expenditures and not on quantities consumed or levels of auto-consumption. The ISRA/IFPRI survey did not cover the entire country and thus is not representative at the national level, but provides insights into consumption and production patterns not covered in the PS. The ISRA/IFPRI survey carefully calculated levels of auto-consumption, quantities consumed, and visited households more than once over a period of more than one year. 1.2 Approaches to defining poverty. One can estimate poverty lines based on relative terms - how one fares in relation to others in the same country/region - or on absolute terms - whether one attains a fixed "minimum" standard of living independent of the number of other people in the country who meet/do not meet this same standard. For example, using relative welfare one might look at households consuming less than one-third of average consumption levels in a country, and thus those falling into this category would not necessarily resemble each other from country to country (for example people spending less than one-third of average expenditure levels in the U.S. are certainly better off than a similar category of people in a lower-income country such as Nepal).' Therefore, as average consumption levels increased with growth, so would the poverty line thus making it difficult to tell if people actually were better off over time. In contrast, an absolute poverty line would stay relatively fixed, thus showing improvements in actual living conditions over time, and if definitions were similar, providing a point of comparison across countries. The drawback to this approach, however, is that determining what is a "minimum" implies numerous value judgements and possible error. 1.3 Overview of method adopted. The absolute approach to poverty lines adopted for the Senegal poverty profile is a variation of what Ravallion (1992) calls the "food energy method." The approach tries to minimize error and value judgements by focusing on one of the most universally accepted items of basic living standards - consumption of adequate food. We calculated how much it would cost (taking into account levels of auto-consumption) to attain a minimum caloric intake of 2,400 calories per adult equivalent per day, a level consistent with the level used in recent agricultural ' An example of the former is Boateng et al (1990) who fix a poverty line for Ghana as two thirds of the mean of the distribution of households by per capita expenditure, and a 'hard-core' poverty line as one-third of this mean.  A-2 surveys. In reality, minimum requirements may vary substantially within gender/age cohorts according to level of physical activity (i.e. pregnant and lactating women, farmers during planting and harvest season, and manual laborers burn up more calories than urban residents working in an office). Sensitivity analysis (below) provides an indication of how changes in minimum caloric intake shift poverty line estimation and poverty incidence. Then, rather than estimating what households spend on non-food items, we took observed data on non-food expenditures from those households just spending enough to attain the minimum caloric intake and added this amount to the food poverty line to obtain the general poverty line. However, to get to this point, there were several adjustments that needed to be made. 1.4 Specific procedures followed. First, using the PS database we calculated total monthly adult equivalent expenditures for the food goods which account for the bulk of caloric intake - millet/sorghum, rice, peanuts, bread, sugar, and vegetable oil - dividing all people into twelve household expenditure groups (from those spending a total of 1,000 CFAF or less per capita per month to those spending over 25,000 CFAF per capita), and for both urban and rural areas in each of the ten administrative regions. Then, in order to understand how much food was purchased with these expenditures, the expenditures were divided by prices in each region obtained from other sources (see section 3 below on price data sources). The resulting quantities were then transformed into caloric equivalents, using coefficients from the Office de Recherches Sur I'Alimentation et la Nutriion Afticaine (ORANA) which is based in Dakar. This gave average total purchased calories and average expenditures on the six commodities for each expenditure group. Then, since the six commodities for which purchased calories were calculated represent roughly 85 percent of total calories consumed in Senegal, an additional 15 percent of these base calories was added to adjust for calories from other foods (a percentage supported by other studies on consumption in Senegal) for which prices were not available. The composition of the food basket required to satisfy a minimal level of caloric intake was varied by region to reflect the unique consumption patterns of each region. 1.5 Adjusting the Poverty Lines for Auto-Consumption. Purchased calories were then adjusted upwards to account for the fact that the PS did not estimate levels of auto-consumption (which would normally make households which spent less money yet consumed more home produced foods look poorer than they actually are). This was done by calculating coefficients for auto- consumption for urban and rural areas in each region using data from the IFPRI rural households surveys and other agricultural surveys detailed below. Thus the resulting poverty lines reflect the fact that in regions where residents rely more on auto-consumption, and where prices for basic food goods tend to be lower, the amount of expenditure required to achieve a minimum caloric intake is lower than in urban areas or regions (St. Louis for example) with lower levels of auto-consumption. 1.6 Non-Food Expenditures. Because estimates of the 'appropriate" expenditures on non-food items to achieve a minimum standard of living are likely to vary from region to region, and are often prone to mistakes, no judgement was made on a minimum level of non-food expenditure; rather, under the assumption that individuals first satisfy their basic food needs, actual observations on non-food expenditures of those just achieving the minimum caloric intake were taken directly from PS data and added to the required food expenditure to equal the region-specific poverty line. All of the above was calculated first on an adult equivalent basis, and then reconverted into per capita terms, so that the population distribution of households with per capita expenditures below the poverty line could be calculated for each region, urban and rural. Sensitivity analysis and regional poverty analysis were based on household level data. This slightly underestimates poverty in areas with larger  A-3 than average household size (such analysis can be performed although this would require substantial data manipulation by the DPS while the regional differences if recalculated are probably not large). 1.7 Results and Comparisons with Relative Poverty Lines. Tables AL.1 and A1.2 show the results of regional and national poverty analysis. Relative poverty lines have also been calculated for comparison: at two-thirds of average expenditure in the same area of the country, 53.26% of the overall population could be considered poor (with 53% of the Dakar population falling below two-thirds of the average consumption in Dakar, 44% of the rest of the urban population, and 43% of the rural population). Alternatively, an even lower-bound poverty line of one-third mean expenditure would place approximately 25% of the population in poverty (13.8% of Dakar, less than 10.8% of other urban areas, and 16% of rural areas).2 The high level of relative poverty using an upper poverty line in Dakar, points to the substantially higher average consumption level in Dakar compared to the rest of the country. 1.8 The Gini index has been calculated based on expenditure data, and, as noted in chapter 1, this presents some problems as expenditures do not take into account levels of auto- consumption or price differences. Nonetheless, other available evidence and studies point to a gini coefficient of at least .4. The gini coefficient for Dakar is likely to be the most accurate (as home consumption and prices are uniform within Dakar), while the coefficient is probably overestimated for rural areas. 2. Caveats in Interpretation of Poverty Lines & Limitations of the Approach 1.9 Two difficulties present themselves in determining the relative welfare among rural regions based on income and/or expenditures from a period of a few months, as was done with the Priority Survey. The first difficulty encountered in generalizing about poverty from the "snapshot" measurement of the priority survey stems from the fact that rural consumption patterns change dramatically over the course of the agricultural cycle, from "hungry" season to harvest. The PS measured expenditures and revenues at the harvest season, when incomes and expenditures were likely to be at their highest, and prices at their lowest. Therefore, the rural poverty line presented above represents a bottom-line estimate with increases in the incidence of poverty likely at other times of the year. Sensitivity analysis presented below demonstrates that if all other factors are held constant, if auto-consumption levels decreased by 30%, as could be the case during the hungry season, poverty in rural areas could jump to 60% of households because of the large number of people around the poverty line. This highlights the dramatic variability of rural consumption levels, particularly in regions without access to remittance income. Sensitivity analysis also demonstrated the robustness of assumptions used for the model. Second, large interannual variations in rainfall patterns can affect the overall level of poverty or reverse the relative welfare ranking among rural regions from one year to the next, at least in regions with high dependence on rainfall and low remittance income. In the case of Senegal, the PS probably provides a fairly accurate representation of average poverty levels during harvest season over a period of years if one uses groundnuts, the 2 In Ghana and elsewhere, analysts have defined an 'upper' and 'lower' bound poverty line as 2/3 and 1/3 mean per capita expenditure. Here, the numbers are approximate as they represent interpolations of grouped data.  A4 largest income source for the majority of the rural population, as a proxy for the 1991/1992 crop year; groundnut production during this crop year was fairly typical.3 1.10 Any methodology chosen for estimating poverty lines has its drawbacks.' One drawback of the "food energy intake" approach is that it is static in nature. While all poverty line estimation procedures are static (i.e. a snapshot at one point in time), this approach is hampered by the fact that food consumption patterns are determined in a complex and fluid way that combines demand responses to relative price movements, with changes in income levels and tastes and preferences. Another drawback is more philosophical in nature: this approach implicitly makes a normative assumption that people perceive meeting caloric requirements as their highest priority. With specific regard to Senegal, there is widespread anecdotal evidence that some ethnic groups consider other types of expenditure -- housing, ceremonies, religious tithes - as very high priorities and are willing to go deeply into debt to pay for them. There are also technical problems from a nutritional standpoint regarding choice of caloric intake levels, which may be different for a more sedentary urban dweller than for a physically active farmer. 1.11 This procedure has several other limitations that have more to do with data gaps encountered in Senegal than with the methodology itself. Because home consumption was not recorded under the PS, the analysis assumes constant proportions of home consumption versus purchases across expenditure groups. There are many reasons to believe that this varies across expenditure groups in rural areas. Recent survey work by ISRA/IFPRI5 clearly demonstrate great variability in household rates of cereals self-sufficiency within zones by income quartiles. There is a strong positive correlation between income level and cereals self-sufficiency rates. If this is true, taking a simple average for home consumption by region has the effect of underestimating rural poverty as poor people should have a lower actual home consumption coefficient than the regional average and should therefore be consuming fewer calories than the calculations indicate. A second issue is that taking coefficients for home consumption from other data sources may also introduce bias because none of these other surveys were designed to be representative on an administrative region basis. Third, with regard to prices, there are obvious problems related to taking regional averages as prices faced by all consumers in a given region. In addition, different expenditure groups may face different prices for the same goods (as poorer groups often purchase goods in smaller quantities at higher prices). 1.12 Despite these problems and "data headaches," it was felt that an absolute poverty line estimation approach had greater practical value and was potentially more credible to decision-makers than a relative poverty approach. This was because the poverty line was based on more meaningfiul criteria than setting the line at some proportion of mean national income or as some percentage of the population - essentially arbitrary cut-off points in terms of meeting basic human needs. I Groundnut yields and production were respectively 831 kg/ha and 701,000 MT, compared to 1980/81 to 1992/93 averages of 830 kg/ha and 730,000 MT (see Annex C, Table C. 1 for the data). ' See Ravallion (1992) for a summary discussion of the strengths and weaknesses of the various common approaches taken. 5 Kelly, Valerie A. Aspects Economiques de la Production el de la Commercialisation des Produits Agricols au Niveau des Mdnages du Bassin Arachidier et du Centre du SUneigal Oriental. March, 1993.  A-5 Alternative Absolute Poverty Lines. Since this poverty line represents the minimum needed to meet daily caloric needs, one could also test the effect of a lower bound (ultra-poor) and an upper-bound poverty line of 90% and 1 10% of the base poverty line. Those who one might consider among the ultra poor, who spend 90% or less of the absolute minimum, constitute 29.7% of the population. Using a more generous poverty line of 110% of the base raises the percentage of the population that is poor to 37 %. 3. Data Sources 3.1 Price Data 1.13 Price data used in the poverty profile were collected from three sources: the Commissariat a la Securite Alimentaire (CSA) market price information system; the GOS Direction de Commerce; and the 1988-90 ISRA/IFPRI survey work. 1.14 Data from the CSA were used for millet/sorghum and broken rice. The CSA collects cereals price data6 in 24 urban and 31 rural markets around Senegal on a weekly basis. The regional distribution of these markets is presented in Table A 1.5.7 Prices are collected at the retail, wholesale, semi-wholesale and producer price level. 1.15 For this analysis, a simple average of retail prices were taken for urban and rural markets by region for the four month period that data were collected under the PS (October 1991 to January 1992). For regions where no rural price data were collected (Ziguinchor and Dakar), the urban average was taken. This results in a downward bias in prices for imported cereals (either from overseas or other regions of the country), while the prices of commodities produced and traded locally are biased upwards. 1.16 For bread, sugar, and vegetable oil, Direction de Commerce prices were used. These are official prices (adjusted slightly for distance from Dakar), not prices actually observed in the market. However, with the exception of sugar, official prices are more or less respected in most parts of the country. The official sugar price was adjusted downwards (from the official price of 340 CFAF/kg to 250 CFAF/kg) for regions bordering The Gambia (Tambacounda, Ziguinchor, Kolda, Kaolack) to reflect widespread availability of smuggled sugar. 1.17 Finally, average groundnut prices (shelled) were used from the ISRA/IFPRI data for those regions in which the survey operated (Groundnut basin, Kolda, Senegal Oriental). For all other regions, the average of the ISRA/IFPRI prices were used. While this method leaves a lot to be desired, the PS data show that expenditures on groundnuts is very small in the purchased food budget (both in value and caloric terms), and altering it would have little effect on poverty line estimation. 6 Millet, sorghum (local and imported), rice (paddy, local milled, imported whole rice and broken), and maize. The PS did not collect data on maize expenditures, so these CSA data were not used. 7 In recent years, coverage has been very limited in Ziguinchor due to the political situation.  A-6 3.2. General Caloric Intake and Home Consumption Patterns 1.18 The most important sources for calculating home consumption coefficients are Kelly et al, Volume II, Part II (1992) for the Groundnut Basin and Senegal Oriental, various surveys cited in Kite et al (1992) for Oriental and the Casamance, Horowitz et al. (1992) for St. Louis. This is supplemented with Martin (1988) who calculated cereals self-sufficiency coefficients for all regions. 1.19 Although most of these data emanate from detailed and well-executed surveys, their reliability is limited by the fact that sample sizes are small. In addition, most samples were determined based on stratification by agro-climatic zone. For purposes of estimating poverty lines, the PS data are organized using administrative boundaries as criteria, so this creates adds an additional complication to the issue of comparability between data sets. That said, home consumption coefficients are considered to be reasonable "first guesses" that hopefully can be refined and improved over time. 1.20 Table A1.6 presents aggregate caloric intake patterns in Senegal from Food and Agricultural Organization (FAO) food balance sheets for 1984-86. Although these estimates are somewhat dated and national aggregates are of limited usefulness, they serve as a rough reference point that allows us to get a sense of the importance of different food sources for caloric intake. The six commodities included in estimating caloric intake with PS data represent roughly 85 percent of total calories consumed according to FAO data. This figure has been used to adjust caloric intake for the six commodities (purchased and home production) upwards by 15 percent to derive total caloric intake. 1.21 As part of his work on representative farm models for Senegal, Martin (Table Al.7) presents figures for cereals self-sufficiency in major agro-climatic zones for average and poor rainfall years. These are used as a rough check on other data sources, and as will be seen below are used to replace some of the counter-intuitive home consumption rates from other surveys. 3.2.1. The Groundnut Basin and Senegal Oriental 1.22 The main source of home consumption for these regions is the ISRAIIFPRI rural household survey. Kelly et al (1992) calculated caloric intake on an adult equivalent basis by source (purchases, home production, gifts) for the following commodities: millet, sorghum, maize, local rice, imported rice, cowpeas (niebe), peanuts, vegetable oil, wheat bread, milk, and tubers. Tables A1.8 and A 1.9 take the ISRA/IFPRI tabular data to derive home consumption coefficients by survey zone, and then apply them to administrative regions. For zones in which data were collected over a two year period, a weighted average of home and purchased consumption was taken. Taking an average was considered reasonable as in most survey zones, 1988/89 was a somewhat poor rainfall year, while 1989/90 was better than average. 1.23 For the most part, the ISRA/IFPRI figures correspond to what one would expect: higher home consumption appears fairly well-correlated with higher rates of rainfall. Geographically, as one proceeds from north to south, home consumption rates rise. There are a few exceptions, such as the very high rate of home consumption recorded by the ISRA survey in the city of Niakhar which is in the Fatick region but also near the border of Thies which is a well monetized region. However, in some cases, cultural traditions (i.e. the Serrer ethnic group are often reported to place a high value  A-7 on food security gained through growing one's own crops) can cause variations from village to village in the same region. Since Thies is well monetized part of the country with somewhat irregular rainfall (at least relative to areas such as Tambacounda, Kolda, and parts of the Casamance), one would assume a home consumption coefficient more along the lines of Colobane (around 50 percent) if one were searching for a figure for the entire Thies region. Martin uses a figure of 70 percent for cereals self-sufficiency in the Central Groundnut Basin. That figure is used for Thies cereals home consumption in the poverty line analysis.8 3.2.2. Ziguinchor 1.24 Data in Table Al.10 are from Jolly et al, with citation on page 11-15 of Kite (vol 11). Data are old (1983), but presumably, any trends in deficits would only have gotten worse, due to declining rainfall, increased salt intrusion (especially in lowland villages south of the Casamance River), population pressure, and political unrest. Sample size is also small (9 villages and 196 households). Nearly all farms were cereals deficit (5 of nine were 40 percent or more deficit, 40 percent had stocks sufficient to cover more than 6 months of need, and only 5 percent held supplies in excess of annual needs) with farms south of the Casamance River having a greater structural deficit than those north of the River. The worse situation south of the river appears to be strongly influenced by land availability and productivity. Southern farmers cultivated only 0.378 ha each while the figure was more than double for Northern farms (0.863 ha/worker). Southern yields were also substantially lower than those in the North (maize, 221 kg/ha versus 838 kg/ha, rice, 1511 kg/ha versus 888 kg/ha, and groundnuts, 621 kg/ha versus 954 kg/ha). As a result, farms south of the River have turned increasingly to non-farm income. For their sample villages, Posner et al estimated that southern farms received the majority of their income (59 percent) from outside of agriculture. North of the River, this figure was only 20 percent. 1.25 Because the situation has probably gotten worse in the last ten years for the reasons cited above, and one surplus village skews results somewhat, the unweighted percent (rather than the weighted figure of 18 percent), rounded to 30 percent is chosen as the cereals deficit for Ziguinchor, giving a home consumption coefficient of 70 percent (this is 5 percent higher than the figure cited in Martin). 3.2.3. Kolda 1.26 Kite (Volume II) discusses a Canadian-financed survey implemented in 1989 as part of a forestry project in the Department of Kolda. 269 interviews were conducted in 52 randomly selected villages. In addition, all 7 quartiers of the city of Kolda were visited (however, the number of interviews is not known). Survey results are only indicative, due to the small sample size. I Incomes (and associated food consumption patterns) may be highly variable within the Thies region relative to other regions of the country. There is anecdotal evidence of severe pockets of poverty in rural Thies, while there are other areas where incomes are quite high due to participation in fruit tree cultivation, truck gardening, and maritime fishing. Judging by the low levels of caloric intake registered by Kelly et al, and their judgment that diets were not well-diversified due to low levels of cash on hand (especially in 1988189), the Niakhar zone is probably one of the poorer zones of the Fatick/Thies regions.  A-8 1.27 The one interesting result for purposes of deriving home consumption coefficients is that urban residents grow much of their own food. Based on an annual standard of 2,000 kg/household (10 family members at 200 kg/yr each), and reported cereals purchases of 1,025 kg/yr, Kite gives a ballpark figure of 50 percent as cereals home consumption. According to the PS, average household size in urban Kolda is 8.7. Therefore, we can revise this down to 40 percent home consumption (8.7x200= 1740, 1025/1740=59 percent purchased).9 1.28 Concerning rural home consumption of cereals, Kite derives a rough figure of 75 percent. This is approximately in line with Kelly et al's home consumption figure (incorporating a wider array of commodities) of 77 percent. The Kelly et al figure of 77 percent is retained, as it is the more reliable of the two data sources. 1.29 Eighty-two percent of the families interviewed were not cereals self-sufficient, with the bulk of purchases occurring in the soudure (lean season). Cereals produced are largely for home consumption, as only 2-3 percent of cereals were marketed. Proportions marketed were higher for other crops (vegetables 33 percent and cassava/sweet potato 20 percent marketed). The main on-farm cash crops are cotton (97 percent marketed) and groundnuts (67 percent marketed), while total cash revenues were highest from groundnuts (30 percent of family revenues). 1.30 Womens' incomes are especially low, reflecting limited access to land and inputs, and their traditional concentration on rice production (little of which is marketed). Fifty percent of the women had cash incomes of less than 5,000 CFA/yr. 1.31 Forest products (construction wood, fuelwood and charcoal, are also an important secondary source of cash revenue for some households, as is also hunting and gathering. However, these sources (as well as agriculture) are increasingly threatened due to over-exploitation of the natural resource base, population pressure, declining soil fertility and erosion. Fallow periods have shortened as 58 percent of households fallow only 4 years and only 15 percent fallow more than 5 years (while 8 years is appropriate). 3.2.4. Saint Louis 1.32 IDA data collected by Horowitz et al (1992) show that households in Saint Louis derive a relatively low proportion of food from home production, around 20-25 percent of the value of food consumed according to preliminary results from 3 villages. More extensive data in the table below (from IDA Phase II which covered 9 villages) appear to indicate a slightly higher home consumption coefficient on the order of 35 percent. These are not necessarily inconsistent as figures in the table are on a calorie equivalent basis. Because purchased calories are generally more expensive, (major expenditure items include oil, sugar, fish, etc. while home consumption consists of low value cereals), this makes sense. As Table A 1.7 shows, Martin uses home consumption figures for cereals in the range of 50 to 80 percent in average years and 20 to 50 percent in bad years. Using the figures in Table Al. 11 from Horowtiz et al. (1992), if cereals consumption represents approximately 60 percent of Fleuve caloric intake, and all non-cereals calorie sources are purchased, I It is possible that urban home consumption in a number of other regions is substantially higher than the one percent figure used. This is probably true for regions where 'urban' centers (other than perhaps the regional administrative seat) are really towns bordering agricultural areas.  A-9 this works out to 30 percent to 48 percent of food consumed is produced at home (50% x 60% and 50% x 80%) in average years. On this basis, a home consumption coefficient of 35 percent is chosen for the St. Louis region, roughly similar to the IDA figure for 9 villages. 1.33 Using IDA household consumption data for the Fleuve, the six commodities covered in the estimation of regional poverty lines (using PS data) account for 86 percent of total caloric intake. This coincides closely with the 15 percent figure chosen to adjust total caloric intake to account for those commodities not explicitly considered in calculating poverty lines. 4. Sensitivity Analyses 1.34 Sensitivity analysis can provide an indication of the robustness of the analysis performed to changes or revisions in assumptions. It can also shed light on how higher prices occuring at other times of the agricultural cycle (other than harvest when the PS was conducted) might alter the poverty profile. The following tests were performed: prices of all six food commodities were increased by 10 percent; the minimum daily caloric requirement was lowered by 10 percent (from 2,400 to 2,160 calories per adult-equivalent); the "other foods" coefficient was raised and lowered by 10 percent (from 15 percent to 16.5 and 13.5 percent); and the rural home consumption coefficients were raised and lowered by 10 percent for each region. 1.35 Table A1.12 presents the results of the sensitivity analyses in terms of levels of expenditure needed to maintain a minimum level of caloric intake.'