Tunisia - Poverty alleviation : preserving progress while preparing for the future (Vol. 2 of 2) : Annexes

Report No. 13993-TUN Republic of Tunisia Poverty Alleviation: Preserving Progress while Preparing for the Future (In Two Volumes) Volumrie II Annexes August 1995 Country Operations Division Country Department I Nliddle East and North Africa Regiotn Daicunent of the World Bank  Currency and Exchange Rates Currency Unit: Tunisian Dinar (TD) TD per US$ Period Averages 1980 = 0.4050 1981 =0.4938 1982=0.5907 1983=0.6788 1984=0.7768 1985=0.8345 1986=0.7940 1987=0.8287 1988=0.8578 1989=0.9493 1990=0.8783 1991 =0.9246 1992 =0.8844 1993= 1.0037 1994=1.0116 Fiscal Year January 1st - December 31st Weights and Measures Metric System  ABBREVIATIONS AND ACRONYMS AMG: Assistance Medicale Gratuite (Free or subsidized Health Care) BNA: Banque Nationale Agricole (National Agricultural Bank) CGC: Caisse Generale de Compensation (General Compensation Fund) CNRPS: Caisse Nationale de Retraite et de Prevoyance Sociale (National Pension and Social Insurance Fund) CNSS: Caisse Nationale de Securite Sociale (National Social Security Fund) CPI: Consumer Price Index (Indice des prix a la consommation) EU: European Union (Union Europeenne) FIAP: Fonds d'Insertion et d'Adaptation Professionnelle (Employment & Training Fund) FNAH: Fonds National d'Amelioration de l'Habitat (National Rehabilitation Housing Fund) FNRLR: Fonds National de Resorption des Logements Rudimentaires (National Fund for Slum Elimination) FNS: Fonds National de Solidarite (National Solidarity Fund) FODERI: Fonds Special du Developpement Rural Integre (Special Fund for Rural Development) FONAPRA: Fonds National pour la Promotion de l'Artisanat et des Petits Metiers (National Fund for Craftmanship Promotion) FOPROLOS: Fonds pour la Promotion des Logements Sociaux pour les Fonctionnaires a Bas Salaire (Housing Finance Fund for Low-Salary Civil Servants) INS: Institut National de la Statistique (National Institute of Statistics) LSMS: Living Standards Measurement Survey (Enquete sur le Niveau de Vie des Menages) MI: Middle-Income Countries (Pays a Revenus Moyens) MOA: Ministry of Agriculture (Ministere de l'Agriculture) PDR: Programme de Developpement Regional (Regional Development Program) PDRI: Programme de Developpement Regional Integre (Integrated Regional Development Program) PIC: Poverty Incidence Curve (Courbe d'Incidence de la Pauvrete) PNRPQ: Programme National de Rehabilitation des Quartiers Populaires (National Sanitation Program in Urban Areas) PDUI: Programme de Developpement Urbain Integre (Urban Development Program) SMAG: Salaire Minimum Agricole Garanti (Agricultural Minimum Wage) SMIG: Salaire Minimum Interprofessionnel Garanti (Non-Agricultural Minimum Wage) SNIT: Societe Nationale Immobiliere de Tunisie (Real Estate Company of Tunisia) SONEDE: Societe Nationale d'Exploitation et de Distribution des Eaux (National Water Supply Utility Company) SPROLS: Societe de Promotion des Logements Sociaux (Agency for the Promotion of Public Housing) STEG: Societe Tunisienne de l'Electricite et du Gaz (National Power and Gas Company) This report is based on the contributions of a team including Nawal Kamel, Stephen Mink, Martin Ravallion, and Dominique van de Walle of the Bank, and Fernando Clavijo, Juan Lopez, and Nicola Rossi, consultants. Valuable comments were received from, among others, Kathy Lindert and Miria Pigato. As task manager, Setareh Razmara was responsible for the overall preparation of the report. Lionel Demery and Jamil Salmi were peer reviewers for the report. At the time this report was prepared Mahmood Ayub was the Division Chief, John Underwood the Lead Economist, and Daniel Ritchie the Director of the department. When the Yellow Cover was discussed, Mr. Christian Delvoie was the Division Chief. Dominique Dietrich desk-topped the report. The cooperation of many Government departments in Tunisia is gratefully acknowledged. In particular, extensive assistance was received from the National Institute of Statistics, the Human Resource Division of the Ministry of Economic Development, Ministries of the Social Affairs, Public Health and Education. Without their help, preparation of this report would not have been possible.  List of Annexes Annex A.1:The Cost of Capital, 1983-1992 Annex A.2:Future Prospects for Poverty Reduction in Tunisia Annex B. 1:Poverty Lines in Tunisia: An Assessment of Old Methodology Annex B.2:Alternative Method to Construct New Poverty Lines for Tunisia Annex B.3:Different Poverty and Inequality Measures Annex B.4:Recommendations to Improve Social Data Annex C. 1:Employment Generation and Wage Policy Annex C.2:Informal Sector in Tunisia Annex C.3:Agricultural Value-Added and Distribution Policies Annex D. 1 :Education Annex D.2:Health Care System Annex D.3:Water Supply Annex D.4:Housing Annex D.5:Direct Transfers Statistical Annex  ANNEX A.1 Page 1 of 2 The Cost of Capital, 1983-1992 1. The cost of capital in Tunisia has been affected by a complex system of investment incentives and tax provisions aimed at redressing sectoral imbalances. In this Annex a simplified way to estimate the cost of capital has been computed. 2. Assuming that interest payments are fully deductible, as they are in Tunisia, the cost of capital is equal to: c = q (r (1 - t) + d) where: q= the price of investment goods; r= the real lending interest rate; t= the corporate tax rate; and d= the depreciation rate. 3. The deflator of total gross fixed investment was used as the price of investment goods, thus the detailed system of fiscal and financial incentives has not been considered. 4. The lending rate used is the money market rate plus 3 percentage points; the different preferential sectoral interest rates were not taken into consideration. 5. To simplify the calculation, a 50% tax rate is applied for 1983-88; and with the tax reform in 1989, the normal corporate tax of 35 % is applied for 1989-92. No differentiation between the wholly exporting and agricultural enterprises is considered, and various tax holidays have not been applied. 6. For d, a uniform value of 5 % was assumed. 7. One index of the cost of capital is calculated. Table A. 1 shows this series together with the ratio of the cost of capital to the average manufacturing wage (as cost of labor).  ANNEX A.1 Page 2 of 2 Table A.1: Capital Cost Index Year Capital Cost Index Capital Cost / (1983 =100) Avg. Wage Manuf. Index 1983 100 100 1984 109 105 1985 123 113 1986 149 137 1987 168 153 1988 167 145 1989 226 174 1990 247 173 1991 264 173 1992 267 166 Source: World Bank staff estimates.  ANNEX A.2 Page 1 of 2 Future Prospects for Poverty Reduction in Tunisia 1. As indicated in Chapter I, the reduction in poverty between 1985 and 1990 can be explained by both growth and the change in inequality.' With real consumption expenditure per capita increasing by 10% during this period,2 roughly two-thirds of the decrease in poverty is attributable to the growth in mean consumption, and the rest is attributable to improved distribution. For instance, if income distribution had not changed over the period, the same rate of growth would have reduced the poverty head-count index to 8.5% in 1990, compared with the rate of 7.4% actually achieved (and 16% for a 25% higher poverty line, compared with the rate of 14.1% actually achieved).3 Thus the pattern of growth over this period, combined with a more equal income distribution further reduced poverty. 2. The rate at which poverty is reduced in the future depends on many factors, including government policies, the rate of growth and its sectoral distribution, and the nature of any associated changes in relative inequalities. A consistent method of poverty measurement suggests that the pattern of economic growth seen in Tunisia in the 1980s has brought benefits to the whole population, as well as the poor. Based on a simple projection model, the prospects for poverty reduction through economic growth in Tunisia are quite promising: with a moderate increase in the rate of growth in mean consumption to around 2.5% per year, and a continuing fall in inequality consistent with recent experience, poverty in Tunisia could be eliminated by the year 2000, based on current poverty line standards. 3. Two scenarios are considered: (i) holding relative inequalities constant, and assuming growth of real consumption per person; and (ii) combining improvement in inequalities and growth of real consumption per capita. In both scenarios a low case of 1.6% per annum (roughly the rate experienced over 1987-93), and a high case of 2.5% (roughly the rate equivalent to 5-6%,annual GDP growth rate) for the growth rate of real consumption per person are considered (with the rate of population growth at around 2% p.a.). Table I.2 summarizes the main results. 1. Changes in poverty can be broken down into a growth component (a change that would have been observed if the Lorenz curve had not shifted), a redistribution component (a change that would have been observed if the average had not shifted), and a residual (the interaction between the growth and the redistribution effects). Thus: Change in poverty = Growth component + Redistribution component + Residual. The changes can also be decomposed into sectoral effects to estimate the relative importance of poverty changes within sectors and between them. (See Chen, Datt and Ravallion 1992). 2. Mean consumption in 1985 was TD 471 per person (TD 651 per person in 1990 prices), using the Specific CPI; mean consumption in 1990 was TD 716 per person. 3. This was estimated by calibrating a Beta Lorenz curve to the 1985 distribution and using this curve to simulate the distribution for 1990 if only the mean had changed; POVCAL was used for these calculations (Chen, Datt and Ravallion, 1992). In 1990 elasticities of the poverty measures, as well as low income measures, to growth in mean consumption are quite high. Based on new poverty lines estimated in Annex B for the head count index, the elasticity of the poverty line to the mean is -2.8; for the poverty gap index, it is -3.3; and for the poverty severity index it is -3.5. If the poverty line is increased by 25%, the elasticities are respectively -2.1,-2.9, and-3.2.  ANNEX A.2 Page 2 of 2 Table A.2: Simulated Effect of Different Economic Scenarios on Poverty Incidence of Incidence of Poverty in 2000 (based Annual Rate of Growth Poverty in on a 25% higher Scenarios in Consumption per Capita 2000 poverty line) Growth with constant income distribution Low case growth 1.6 3.0 8.0 High case growth 2.5 1.2 5.0 Growth with improved income distribution High case growth 2.5 0.0 2.0 Source: Staff calculations. 4. In the first scenario, assuming that there is no change in relative inequalities in the 1990s, all consumption levels will increase at the same rate. The low-growth case would reduce the proportion of Tunisia's population who are poor by 56% (and by 42% for a 25% higher poverty line), giving an incidence of poverty of 3% (or 8%). Based on the current poverty-line standards, the high-growth case would bring the poverty rate down to 1.2%(or 5%) by 2000.4 In the second scenario, approximately the same outcome could also be achieved with the low-growth case combined with the actual rate of improvement in relative inequalities observed during 1985-90. Moreover, the combination of a 2.5% growth rate of consumption per capita with a continuing reduction in inequalities could bring down the poverty rate to almost zero by 2000, based on current living standards (and to about 2 % based on a 25 % higher standard of living).5 m:\seareh\poverty\report\annexA 4. This would entail a 30% increase in real consumption per capita over the 1990s. This figure is actually quite consistent with the Bank's actual and projected rates of growth in the (per capita) private consumption component of the national accounts; although actual and projected growth rates vary considerably over time, the compounded effect from 1990 to 1999 is equivalent to a 27% increase over that period. 5. The rate of poverty reduction implied by the above scenarios would be even higher for measures of poverty and poverty lines that focus more on the poorest, because the elasticities to the mean of other poverty measures are actually higher than for the head-count index.  ANNEX B.1 Page 1 of 6 Poverty Lines in Tunisia: An Assessment of Old Methodology 1. Along with many other countries, Tunisia has set poverty lines using a version of the "food-share method".' These official poverty lines, used since 1980, were devised by National Institute of Statistics (Institut National de la Statistique - INS) and partly suggested by the World Bank.2 Although INS has aimed to assure consistency of their poverty measures within each sector over time, it has not attempted to assure consistency between sectors and to present the same level of living standards in all sectors and regions. With urban poverty lines set at higher standards of living than rural ones, urban poverty incidence appears to be higher than rural incidence. This discrepancy may confound policy choices aimed at reducing absolute poverty and policy choices concerning the priority given to urban versus rural development. Furthermore, this can also yield inconsistencies over time in the national poverty measures when there is mobility across sectors, as there certainly is between rural and urban areas of Tunisia (Box B. 1). The following sections describe the old INS method in detail, raising the methodological issues, and discusses why this method is not likely to produce a consistent poverty profile for informing policies aimed at reducing absolute consumption poverty. A. Description of the Old Tunisian Methodology 2. Setting poverty lines is largely a matter of normative judgement, and the judgements made will depend in part on the aims of measurement. In common with other countries, the old INS method of setting poverty lines entails first choosing a "reference group" whose consumption behavior is used as a benchmark for calibrating the behavioral parameters of the poverty line.3 Although the choice of the reference group is essentially a value judgement, the consumption behavior of the reference group is used for setting the comnosition of the diet (though not the total quantity of food, which is determined by the stipulated food-energy requirements, as discussed further below). The consumption behavior of the reference household also determines the allowance made for nonfood goods (as discussed further below). 3. Over time, INS has aimed to assure consistency of their poverty measures by adjusting the poverty lines for urban and rural areas separately so as to have constant value in terms of real consumption expenditure; Tunisia's Consumer Price Index (CPI) has been used for this purpose.4 Although this price index is not ideal, in principle the INS comparisons over time within each sector should be consistent in the above sense. 1. For a survey of the various methods found in practice see Ravallion (1993). 2. See Oueslati (1987); INS (1993), for a detailed description of the methodology. 3. The poverty line calculations are based on National Household Budget Consumption Surveys. Since 1975 the National Institute of Statistics (INS) has conducted household budget and consumption surveys every five years to collect detailed household-level information. These household consumption surveys do not constitute a comprehensive integrated portrait of living standards. Given the nature of the surveys, the information mainly focuses on aspects of consumption, and less is known about differences in nonconsumption dimensions of welfare, such as access to public services and sources of income. 4. To better reflect the consumption behavior of Tunisia's poor, it is preferable to use the "specific consumer price index". In the alternative methodology, described in Annex B.2, we will switch to the latter index.  ANNEX B.1 Page 2 of 6 Box B.1: Why Poverty Lines Should Yield to a Consistent Poverty Profile In all economic and social measurement, the choice of method to measure poverty will depend both on the data available and the specific purposes for which the measures should be used. The guiding principle in constructing a poverty profile for informing policy is that it should be consistent with respect to the policy objective. When that objective is to reduce absolute consumption or income poverty, consistency requires that any two individuals with the same real consumption or income are treated identically. When making comparisons over time, one should adjust solely for differences in the cost-of-living (using a suitable price index). The same principle also applies to comparisons across regions or sectors of the economy at the same date. A "poverty profile" is a decomposition of aggregate poverty, showing how living standards vary across regions or sectors of the economy. This information can have great bearing on the policy choices made in attempting to reduce aggregate poverty. A poverty profile is consistent if and only if the poverty lines used to compare different subgroups in that profile have the same value in terms of the underlying welfare indicator. The poverty line comprises two components: a food poverty line, giving the allowance for "basic foods"; and a nonfood poverty line, giving an allowance for "basic nonfood goods". The precise definitions of both components will vary from country to country, and they will also be subject to debate within any one country. However, there is no ideal method of setting consistent poverty lines, and the choices made will almost certainly be contentious. Poverty comparisons between urban and rural areas pose a number of problems. This is partly because "urban" can mean different things. Cost-of-living adjustments are also a problem, as spatial cost-of- living indices are far less common than inter-temporal indices, such as the CPI. However, some common methods of setting poverty lines for each sector almost certainly do not assure that the same standard of living is treated the same way in each sector, with urban poverty lines set at a higher standard of living than rural ones. Indeed, urban poverty incidence can appear to be higher than rural incidence, but only because one has used an appreciably higher real poverty line in urban areas (see, for example, Ravallion and Bidani 1994). The same point can apply to comparisons of poverty made across regions, by employment sector, or over time. Clearly, this may confound policy implications aimed at reducing absolute poverty. 4. The old INS poverty lines have been anchored to the consumption behavior of the twentieth percentile in 1980. Mainly because of this choice of reference group, the old INS methodology could not assure consistency between sectors,5 and does not yield poverty lines that are constant in terms of the standard of living that they imply. For instance, the reference group is well above INS's current estimates of poverty incidence in Tunisia (about 7% of the population in 1990 were deemed poor). Moreover, the twentieth percentile in urban areas are likely to be better off than the twentieth percentile in rural areas, and the better-off households will naturally tend to spend more on each calorie consumed (reflecting a more expensive diet). 5. The choice of reference group alone does not determine the poverty line; it determines the type of consumption behavior that the poverty line is consistent with, that is, the shares of consumption going to various goods (food and nonfood). The overall level of the food component of the poverty line is usually determined by the food-energv requirements that are assumed, typically based on the food-energy 5. In theory, one must make a first guess of who the poor are, use this guess to set the reference group and calculate the poverty line, and then revise the definition of the reference group accordingly. Proceed in this way until there is (hopefully) convergence.  ANNEX B.1 Page 3 of 6 needs of the population set by nutritionists. This practice is followed in Tunisia. The average caloric unit value (CUV) is calculated by dividing food expenditure by the estimated number of calories consumed by the reference group (the twentieth percentile in Tunisia). The CUV is then multiplied by the predetermined food-energy requirement for the average Tunisian to obtain the food component of the poverty line.6 The CWU sets the composition of the diet, and the poorer the reference group, the "cheaper" the diet will tend to be (more starchy food staples, less protein and complex carbohydrates). To incorporate an allowance for nonfood goods, the INS food poverty line is then divided by the average budget share devoted to food of the reference group. In Tunisia (as elsewhere) these steps are carried out separately within urban and rural areas. 6. Combining these three elements, the old INS poverty line for sector i is given by: = v120k{r/s (1) where v,1 denotes the caloric unit value for the twentieth percentile in sector i, kIr is the caloric requirement for sector i, and sil is the food share for the twentieth percentile in sector i. 7. Based on this old method, the urban poverty line for 1990 is set at TD 278 per person per year, and the rural poverty line is set at TD 139 (Table B. 1). B. Properties of a Poverty Profile Based on the Old INS Poverty Lines 8. Are the value judgements built into these poverty lines appropriate for the purpose of a poverty profile which is to help inform policy decisions aimed at reducing poverty? To understand the properties of poverty measures based on the old INS poverty lines, it should be first noted that the CUV is not a price; it is chosen by each consumer and is therefore a function of the consumer's budget constraint (prices and total expenditures) and tastes. The same is true of the food share. Thus it can be assumed that both the CUV and the food share will vary with prices (denoted by the vector pi for sector i) and the total consumption expenditure per person (in the twentieth percentile) in each sector (xi'). Furthermore, tastes may differ between sectors. The poverty line for sector i will then vary with prices and the expenditure level of the twentieth percentile according to: Zl.(pi,Xi2O) =~ Vi2