THE CASE OF NOUAKCHOTT Dr. Osman Nour, Arab Urban Development - - PDF document

1 / 18

New Approach to Measure Urban Poverty:

THE CASE OF NOUAKCHOTT

Paper presented at the XXV111 International Population Conference,. Cape Town, South Africa 29th of October- 4th of November 2017

• Dr. Osman Nour

Arab Urban Development Institute (AUDI) August, 2017

2 / 18

New Approach to Measure Urban Poverty:

THE CASE OF NOUAKCHOTT

• Dr. Osman Nour, Arab Urban Development Institute (AUDI)
• nour99@hotmail.com; onour@araburban.org

I-Introduction and Objectives of the Study: Poverty as a multidimensional phenomenon is perhaps most revealed in urban areas. Urban Poverty is not only characterized by inadequate income, but also by lack of adequate asset base, dwelling, provision of public infrastructure. In addition access to social services such as education and health can be scarce. Poor urban households often suffer limited social safety nets, powerlessness within political system (UN Habitat, 2003). The most common quantitative approach to measure urban poverty alone can underestimate urban poverty, because it does not take into account the extra cost of urban living, such as housing, education, health, electricity, water, transport, and other services. Therefore, income alone cannot capture the many dimensions of urban poverty. Several poverty measure have been introduced to estimate urban poverty, such as "Unsatisfied Basic Needs" (UBN), "Income Poverty Line" (IPL), and others, based on various characteristics of household pertaining to its dwelling conditions,

• wnership of assets, demographic characteristics, educational level,

and employment of household's members. Recently, the Arab Urban Development Institute (AUDI) and the UN Economic and Social Commission for West Asia (ESCWA) have constructed an index to measure Urban Poverty. The main purpose

• f the index was to measure urban poverty in a simplified and

straight forward manner through using short list of highly indicative

• variables. However, determining these variables require studying an

exhaustive list of potential indicators and performing an in-depth

3 / 18

analysis of their indicative power. The methodology for determining the indicators for the UPI is composed of three main steps: (1) running a series of correlation tests, (2) selecting the most significant variables and, (3) performing econometric analysis to determine the weight of each variable. This new urban poverty measure considers poverty as a multidimensional phenomenon, and emphasizes the concept of deprivation and looks at absolute and relative poverty at the same

• time. The index tries to avoid the traditional measure of poverty

through income poverty line. The new Urban Poverty Index is not

• nly used to compare it with other poverty measures, but it can also

be used to compare poverty level among the different districts of the city and among the various sectors, such as education, health, dwelling and economic situation. This advantage can help mayors, city officials, NGOs and related ministries in setting the priorities of interventions to reduce poverty, at the district, and at the sectoral levels. There are three main objectives for the present study. First, the study will determine the factors that are most correlated with urban poverty in order to construct an urban poverty index for Nouakchott. Secondly, the newly constructed index of urban poverty will be applied to Nouakchott Household Survey Data to measure urban poverty for the capital city and its nine districts. Finally, Nouakchott's UPI will be compared with income poverty index, to test of its validity. 2-Sources of Data: The data for the present study was collected and prepared by the National Statistical Department in Mauritania , using a well-designed long household questionnaire to permit the calculation of different poverty indices, for comparative purposes. The city of Nouakchott is subdivided into nine districts. The total sample size of 2700 households was distributed and weighed for the nine districts, proportional to the size of the population in each district(see Table1)

4 / 18

The percentages of weighted households range from 13% for Ksar district to 25% for Arafat district, which is the most densely populated area compared to other districts. The statistical unit is the household and it's members. The sample is well balanced from the gender perspective with 51% males to 49% females, and the same gender balance is maintained for most of the nine districts, with Sebkha registering a higher male to female ratio (55% males against 45% females).After the sample size in each neighborhood is determined, the households to be surveyed were selected randomly by the field supervisors.

FIGURE 1: DISTRICTS OF NOUAKCHOTT AND THEIR SHARES IN THE SAMPLE

DISTRICT NUMBER OF HH WEIGHTED NUMBER OF HH Teyarett 280 18,367 Ksar 200 13,554 Tefragh-Zeina 240 16,147 Toujounin 280 15,350 Sebkha 300 15,540 El Mina 480 22,113 Dar Naim 300 22,019 Arafat 380 25,068 Riad 240 14,268 TOTAL 2,700 162,424

3- Sample's Profile: a- Education: Nouachott's data show that 25% of the population had never been enrolled, a share that increases to 43% and 44% for El- Mina and Sabkha respectively. Inter-district discrepancies are more notable for lower levels of educational attainment. The illiteracy rate for Nouachott is 24%, with a high rate of 41% for El- Mina and Sebkha districts. On the other hand the percentage of those with university degree is as low as 4% for the city and this percentage reached 7% and 6% Teyarett and Tafrigh-Zeina. b- Work Status: Around 13% of Nouakchott's residents are salaried employees and 14% are self employed, with slight inter- district

5 / 18

• discrepancies. The share of unemployed averages 9% for the city

, reaching as low as 1% for Ksar district and as high as 14% for El- Mina district. The study shows that sales and service employees

• ccupy the highest share (35%), followed by unskilled jobs ( 25%).

Moreover,75% of the people work in private sector, versus 25% in the public sector. c- Health: Around 80% of Nouakchott's residents have no health insurance of any kind. This share increases to 90% for Sebkha and El-Mina and decreases to around 65% for Ksar and Trafragh-

• Zeina. The survey shows that 25% of the city's households don't

have vaccination records for their children, a share that reaches 60% for El-Mina district. About 80% of births take place in public

• hospitals. About 40% of the households reported visiting

physician for prenatal check-up, and another 40% go to certified

• nurse. Child deaths ( under five years) were reported for 8% of

the households. d- Social connections: When the respondents were asked about the extent of a person's social network, they reported an average of two close friends. When the respondents were asked how many persons outside their households they can resort to for money in case of emergency, the average was around 1.5 with minimum inter-district fluctuations. Finally, when the respondents were asked about how many people turned to them for help in the past year, the average was uniformly around one. In conclusion, there seems to be weak support for the hypothesis that the size

• n one's social network is related to monetary poverty.

e- Sources of information: Newspaper reading does not seem to be an established habit, as respondents reported reading the newspaper,show an average of only once per week. Listening to radio is a more common behavior, with 37% reporting that they did on a daily basis. Television is obviously the most established habit, with 67% of the households who reported watching it

6 / 18

every day. When the respondents were asked about their sources

• f information, around 60% mentioned their relatives, neighbors

and friends, 49% mentioned the local market, and 16% mentioned the workplace. In terms of Media as a sources of information, television came first , with 70%,followed by the radio 60%, and newspaper and magazine with 40%, and worldwide web with 10%. 4- Methodology: The questionnaire for the present study is designed in a way that permits the calculation of different poverty indices, such as Unmet Basic Needs (UBN), Income poverty (IP), and Urban Poverty Index (UPI). The questionnaire is divided into three main sections as follow: a- General information about the household: This section includes demographic and socio- economic characteristics of household members and household size. b- Questions to construct UPI: This is the main section which includes the questions that are going to be scored to measure urban poverty c- Additional descriptive Questions: This section includes questions related to UBN categories, namely education, health, dwelling conditions, dwelling amenities, and economic situation. These questions do not enter into the scoring system, but they can be used to generate additional descriptive information on the household living conditions, and will be used to set the intervention priorities. The methodology to determine the indicators for the UPI for Nouakchott is composed of the following main steps: running a series of correlation tests, to select the most significant variables, and performing regression analysis to determine the weights for

7 / 18

each variable. A basic criterion for a good poverty indicator is for it to be significantly correlated with poverty. The first step in determining the UPI indicators for Nouakchott was to determine which variables are indeed mostly correlated with

• poverty. The income poverty line was used to perform a

comprehensive set of correlation tests with the household related variables and those pertaining to the characteristics of the household head. These variables include the ones in the questionnaire, as well as others that were constructed later using the original questions of the survey. Cross-tabulations were performed along with Pearson's chi-square test in order to determine the significance of each variable. When looking at the cross-tabulation results, some variables showed to have higher degrees of variability between the poor and the non-poor, while

• thers did not vary much between the two groups. These latter

variables were not viewed as indicative since they did not show high variability and thus would be weak detectors of poverty. Therefore, they were removed from the list of variables to be selected for Nouakchott UPI. The final result was a list of 13 potentially indicative variables presented in Table (1) below.

TABLE2: LIST OF SIGNIFICANT VARIABLES

No. Field Variable 1 Dwelling Area of dwelling 2 Amenities Type of service water 3 Amenities Sanitary sewerage 4 Economic Bank account 5 Economic Not going to the doctor for financial reasons 6 Nutrition Weekly consumption of chicken 7 Nutrition Weekly consumption of sweets 8 Equipment Possession of cable subscription 9 Equipment Possession of air condition 10 Equipment Possession of fixed telephone line 11 Equipment Car ownership 12 Information Reading newspapers and magazines 13 Information Browsing the internet outside work

8 / 18

9 / 18

The second step is to look for the most indicative variables to be included in the regression equation to determine Nouakchott UPI. This is performed by regression analysis to the 13 selected variables in order to test for their significance in terms of their causal relationship with poverty. Some of these variables coincided with the variables from the Tripoli-UPI. However, a group of variables was part of the Tripoli-UPI, but did not make it to the list of indicative variables for Nouakchott. We added these variables to the regression equation to check how they will perform1. Ordinary least squares (OLS) was used for this purpose, where the econometric equation takes the following form: 𝐽𝑜𝑑𝑗 = 𝛽 + 𝛾𝐸𝑥𝑓𝑚𝑗 + 𝛿𝐵𝑛𝑜𝑗 + 𝜀𝐹𝑑𝑝𝑜𝑗 + 𝜈𝐹𝑟𝑣𝑗𝑞 𝑗 + 𝜖𝐽𝑜𝑔𝑝 + 𝜄𝐼𝑓𝑏𝑒𝑗 + 𝜁 Where𝐽𝑜𝑑𝑗 is the value of income per capita for each household “i”. Dwel is the variable related to dwelling, namely area per capita. Amn is the matrix of variables related to dwelling amenities. Econ are the set of variables pertaining to the economic situation. Equip are the variables of equipment ownership. Info are the set of variables related to information and communication. Head are the variables related to the head of household characteristics; a set of dummy variables related to his/her education level in this case. (ɛ) is the white-noise error term.

1 The UBI methodology was first applied to the Household data for Tripoli (Lebanon) using the same questionnaire

10 / 18

TABLE3: REGRESSION RESULTS, DEPENDENT VARIABLE: INCOME PER CAPITA

Variable Beta (Constant) area per capita 0.155*** Water tanks as main source of drinking water

• 0.061***

Sewerage pit or no sewerage 0.041*** Not going to the doctor for financial reasons 0.047*** Weekly consumption of chicken 0.088*** Weekly consumption of sweets 0.032*** Possession of cable subscription 0.026*** Possession of air condition 0.158*** Possession of fixed telephone line 0.108*** Reading newspapers and magazines 0.094*** Browsing the internet outside work 0.18*** Car ownership 0.002 Possession of internet subscription at home 0.091*** Having a bank account 0.076*** head of household having university education

• 0.044***

mineral water as main source of drinking water

• 0.001

Visiting cafes and restaurants 0.149*** R 0.564 R-squared 0.318 Adjusted R-squared 0.318 F-statistic 4305.57*** Note: * for p ≤ 0.05, ** for p ≤ 0.01, ***for p ≤0.001

All variables showed to be highly significant and thus good explanatory variables for poverty, with the exception of car

• wnership and mineral water as a main source of drinking water.

The majority of variables measure wealth rather than poverty and therefore has positive signs when regressed against income per

• capita. Only water tanks used as a source of drinking water showed a

negative sign and indeed this is so because this variable reflects habits of poor rather than non-poor households. By the same logic,

11 / 18

the lack of sanitary sewerage or the sole dependence on a sanitary pit should have resulted in a negative coefficient since one would expect that better-off households live in better neighborhoods with better infrastructure; however, the results did not confirm this logic

• f analysis. This could indicate that an unsatisfactory state of

infrastructure could exist in some regions regardless of the wealth status of their residents. In addition, and surprisingly, the household head university education showed to be negatively related to

• income. This could be due to errors in the data or should be further

studied as a phenomenon specific to the Nouakchott context. 5- THE NOUAKCHOTT UPI: SCORES AND RESULTS A- THE SCORING SYSTEM The results of the regression analysis were used to build a scoring system for the Nouakchott UPI and apply it to the data. The two insignificant variables (mineral water and car ownership) as well as the household-head university education variable were removed from the analysis. The remaining 14 variables were used to compose the scoring system for the Nouakchott UPI. The values of the coefficients from the regression were converted to scores by being multiplied by 1,000 and reconverted to base 1,000. The final scores are presented in the Table4 below.

12 / 18

Table 4

Variable Score Area per capita 131points if area is at least 20m2 / 0 if no Water tanks as main source of drinking water

• 52 points if yes / 0 if no

Sewerage pit or no sewerage 35 points if yes / 0 if no Not going to the doctor for financial reasons 40points if yes / 0 if no Weekly consumption of chicken 74points if at least twice a week / 0 if no Weekly consumption of sweets 27points if at least once a week / 0 if no Possession of cable subscription 22points if yes / 0 if no Possession of air condition 133points if yes / 0 if no Possession of fixed telephone line 91points if yes / 0 if no Reading newspapers and magazines 79points if yes / 0 if no Browsing the internet outside work 152points if yes / 0 if no Possession of internet subscription at home 77points if yes / 0 if no Having a bank account 64points if yes / 0 if no Visiting cafes and restaurants 126points if yes / 0 if no Total 1, 000 points

The scores are automatically generated with the exception of:

• 1. Area per capita: The area needs to be divided by the household

size and then compared to the threshold of 20 m2 per capita.

• 2. Weekly consumption of chicken: needs to be compared to the

threshold of 2 times per week per household

• 3. Weekly consumption of sweets: needs to be compared to the

threshold of 1 time per week per household. The thresholds for chicken and sweets consumption were based on the median values of the data in the questionnaire. The only negative score is the one pertaining to drinking water from

• tanks. If the answer to this question is yes, then 52 points need to

be deducted from the overall score of the household. This would potentially create negative overall scores if the households have not scored much in the other variables.

13 / 18

B-THE RESULTS:

In order to test the validity of the Nouakchott UPI potential results, the questions and the scores were tested on the current data and then compared to the income poverty results. Table5 below shows the distribution of households according to the Nouakchott UPI. A cumulative percentage of around 60% of the households had a score

• f 300 or less. The highest concentration of households is within the

score bracket of 100 to 200. Very few households fall within the higher score brackets. This distribution reflects high level of poverty in the city that have been depicted in all the indicators that have been calculated. TABLE5: DISTRIBUTION OF NOUAKCHOTT UPI RESULTS

Score Percent Cumulative Percent 0 to 100 12.4% 12.4% 100 to 200 24.7% 37.1% 200 to 300 22.6% 59.7% 300 to 400 18.3% 78% 400 to 500 10.8% 88.8% 500 to 600 7.5% 96.3% 600 to 700 2.4% 98.7% 700 to 800 0.6% 99.3% 800 to 900 0.4% 99.6% 900 to 1,000 0.4% 100% Total 100%

Figure1 below shows the distribution of the Nouakchott UPI by

• quintiles. The histogram is skewed to the left in comparison to the

normal distribution, indicating a higher concentration of households within the lower score brackets.

14 / 18

FIGURE 1: HISTROGRAM OF NOUKACHOTT UPI SCORE DISTRIBUTION

When looking at the distribution of Nouakchott-UPI across the districts of the city (Figure2), we notice that Sebkha and El Mina have the highest concentrated of households within the score brackets of 0 to 200. This confirms the previous observations (in section3) that these two districts are indeed the poorest districts of the city, and the Nouakchott UPI was able to capture this. Ksar appears to be the most well off district on average since it has almost no households with scores lower than 200, and more that 50% of the households in Ksar district have a score of 400 to 600. Tefragh- Zeina seems to witness that highest diversification and inequality since it has very poor (17% of the households have scores less than 200) as well as very well off households (Also 17% of the households have scores of 600 to 800).

15 / 18

FIGURE 2: DISTRIBUTION OF NOUAKCHOTT UPI SCORES ACROSS REGIONS

C- COMPARATIVE ANALYSIS BETWEEN THE POVERTY INDICATORS NOUAKCHOTT UPI VERSUS INCOME POVERTY When comparing the Nouakchott UPI with the income poverty calculated for the city, the results show that there is a high degree of compatibility between the two poverty measures. Table 6 shows that the average score for non-poor is 403.7, while that for the poor is much lower (206.3).

TABLE 6: COMPARING NOUAKCHOTT UPI WITH INCOME POVERTY

extreme income poverty Mean Non-poor 403.6904 Poor 206.3062 Total 275.9072

Further analysis was performed on the Nouakchott UPI results by comparing the UPI score deciles to the income poverty shares within each decile. The results show that the poor are indeed concentrated

16% 17% 39% 89% 82% 22% 29% 38% 67% 45% 22% 40% 10% 17% 63% 46% 51% 15% 52% 44% 20% 15% 19% 9%

0% 10% 20% 30% 40% 50% 60% 70% 80% 90% 100%

0 to 200 200 to 400 400 to 600 600 to 800 800 to 1000

16 / 18

in the lower brackets (34.8% in the 100 to 200 score bracket), while the non-poor are concentrated in the upper ones. For example, 16.3% of the non-poor are within the 500 to 600 score bracket, while

• nly 2.6% of the poor are within this same bracket. This gives us

further confidence in the adequacy of the Nouakchott UPI to capture the poor in the city. TABLE 7: UPI DECILES AND INCOME POVERTY

Score Non-poor poor 0 to 100 2.0% 18.1% 100 to 200 6.2% 34.8% 200 to 300 17.2% 25.7% 300 to 400 27.4% 13.3% 400 to 500 22.1% 4.7% 500 to 600 16.3% 2.6% 600 to 700 5.5% 0.6% 700 to 800 1.5% 0.1% 800 to 900 0.9% 0.1% 900 to 1,000 1.0% Total 100.0% 100.0%

6- Concluding Remarks: The purpose of this report is to measure urban poverty in the capital city of Nouakchott, Mauritania. Three main poverty indicators were discussed: (1) income poverty index, (2) Tripoli-UPI, (3) Nouakchott-

• UPI. Income poverty was calculated based on the national poverty

line of 2008. The Tripoli-UPI composed of seven variables was applied to the Nouakchott data and gave quite satisfactory results. The Nouakchott-UPI was constructed using the given dataset and its results were compared to the previous two indicators in order to test its validity. Several conclusions come out of this research endeavor:

17 / 18

• 1. By all standards and measures used, extreme poverty in

Nouakchott is very high, where more than 50% of households face harsh livelihood conditions.

• 2. The previously designed Tripoli-UPI was able to capture the

extremely poor despite the fact that its constitution variables had emerged from a different urban context. This indicates that most of the Tripoli-UPI variables are rather transversal variables that can be used to measure urban poverty in different contexts.

• 3. The Nouakchott-UPI, which was calculated following the same

methodology as the ‘Tripoli-UPI’ gave more accurate and stronger results than the Tripoli-UPI. This can be explained by the fact that the Nouakchott-UPI has emerged from the Mauritanian data, and has thus taken into consideration the specificities of the city and its social, economic, and demographic characteristics.

• 4. Four main variables showed to be significant for both

indicators: area per capita, having a bank account, having an internet subscription at home, and visiting cafes and

• restaurants. These variables actually reflect, important

aspects of urban living conditions that affect the well-being of urban households. They can hence be considered as proxy- variables for “universal” characteristics in relation to poverty.

• 5. The present study shows that the districts with low levels of

socio-economic characteristics, such as El-Mina and sabkha districts, have low scores on UPI, while the districts with relatively high level of socio-economic levels, such as Ksar and Tafregh-Zeinah, have high scores on the UPI.

• 6. Looking at the distribution of Nauakchott's UPI scores ( divided

into quintiles), across the various districts of Nauakchott, we noticed that the districts with the highest income poverty (El- Mina and Sabkha districts) are the ones having the lowest Naoukchott UPI scores. Similarly, Ksar and Tafregh-Zeinah, with the lowest levels of income poverty have the highest Naoukchott UPI scores.

18 / 18

References 1-