Paper short abstract:

This paper evaluates multidimensional poverty vulnerability in Algeria and Tunisia using M-gamma measures by Alkire and Foster (2019), following Gallardo's approach (2022). It employs Bayesian networks classifiers to model the joint probability of poverty and deprivation in each dimension. Vulnerability is measured through the mean risk approach (VMR).

Paper long abstract:

1. Research Context

MENA countries, like others globally, have embraced the UN's agenda for 2030 and its Sustainable Development Goals (SDGs), emphasizing the imperative of leaving no one behind. Despite progress in poverty reduction, the region faces challenges due to its fragile context and ongoing crises, which threaten advancements made. This highlights the importance of understanding vulnerability to poverty to develop effective, sustainable policies.

2. Methodology

Inclusion of vulnerability in poverty analysis dates back to the 2000’s following the pioneering study by the World Bank on social risk and management (2001). A number of approaches have been proposed to assess and estimate vulnerability to poverty but they are not yet been widely adopted. Indeed, since vulnerability is by definition forward looking, most measures require long panel data. However, for many countries, only cross-sectional data are available. Although poverty is now-well recognized as a multidimensional phenomenon, empirical studies on vulnerability assessment are dominated by the monetary approach to poverty. Yet, vulnerability should also reflect the fact that it can occur in different dimensions of well-being. Few studies address multidimensional vulnerability, notably Calvo (2008), Abraham and Kavi (2008), Feeny and McDonald (2016), and recent works by Gallardo (2020, 2022). However, there's a dearth of research on vulnerability in the MENA region. This study aims to bridge this gap, building on previous work by Bérenger (2021) and Gallardo (2022), to examine vulnerability to multidimensional poverty in Algeria and Tunisia.

The objective of the present study is to fill this gap. Drawing on the study by Bérenger (2021) which assesses levels and trends in multidimensional poverty in Algeria, Iraq and Tunisia the present paper proposes to examine vulnerability to multidimensional poverty following the approach developed by Gallardo (2022) and to investigate the complex relationship between multidimensional poverty and vulnerability in Algeria and Tunisia. According to the study by Bérenger (2021), although these two countries have very similar levels of multidimensional poverty, it is interesting to examine whether their population face the same risk of poverty in the future. Vulnerability to multidimensional poverty is estimated using the downside mean semideviation approach proposed by Gallardo (2013). To estimate the risk of being multidimensional poor in the future we draw on Gallardo (2022) that implements multidimensional Bayesian network classifiers. This study is currently one of the two rare applications of Bayesian networks to the analysis of welfare and poverty.

3. Analysis

Utilizing data from UNICEF's Multiple Indicators Cluster Survey (MICS) for Algeria (2012/13, 2018/19) and Tunisia (2012, 2018), we adapted the MPI to the Arab context with insights from ESCWA (2017, 2021).

Analogous to the famous FGT-alpha indices in unidimensional case, our study provided measures of vulnerability to multidimensional poverty when γ=0,1 and 2 and decomposition of vulnerability by distinguishing severe and moderate vulnerability. Four key findings are noticeable.

First, both in Algeria and Tunisia, vulnerability headcount ratios are significantly higher than poverty headcount ratios suggesting that current poverty estimates tell us only part of the story. Despite similar levels of multidimensional poverty, vulnerability measures are higher in Tunisia than in Algeria. In addition, the achievements in poverty reduction are more fragile in Tunisia than in Algeria. Similar trends were also observed with the measures that account for the intensity and inequality among the poor.

The second finding is that moderate vulnerability prevails over severe vulnerability in both countries. Trends over time indicate that in Algeria, vulnerability seems to be shifting more towards moderate vulnerability as the contribution of severe vulnerability to overall vulnerability decreased between 2013 and 2019. Opposite trends were observed in Tunisia where the most significant decreases in vulnerability concerned moderate vulnerability particularly in rural areas. However, severe vulnerability increased in urban areas suggesting that some moderately vulnerable people have slipped into severe vulnerability. As a result, in Tunisia, vulnerability seems to be shifting more towards severe vulnerability between 2012 to and 2018.

Third, the dimensional decomposition of the two main components of vulnerability enabled us to identify the indicators where the differences between severe and moderate vulnerability were the largest are early pregnancy, mortality, nutrition, access to water and school attendance in Algeria; mortality, early pregnancy, sanitation, nutrition followed by school attendance in Tunisia. They correspond to dimensions of structural poverty. In contrast, deprivations in the remaining indicators are much more similar between the two groups of vulnerable. This makes it possible to differenciate the indicators or dimensions that would require specific attention for the design and implementation of social policies. Especially, it is worrisome to note that the risks of vulnerability seem to have increased. particularly in the nutrition indicator in both countries, in early pregnancy in Algeria and in sanitation in Tunisia.

Fourth, the analysis of the overlap between different forms of vulnerability and poverty showed that chronic poverty among the vulnerable is more important in Tunisia than in Algeria. Among the currently poor in Algeria and Tunisia, 38.5% and 48.3% (resp.) remain chronically poor with a high probability of experiencing multidimensional poverty in the future while 26.4% in Algeria and 21% in Tunisia are infrequently poor suggesting that they are likely to escape from poverty. Our results revealed different trajectories in the evolution of the vulnerability components in these two countries. Severe vulnerability among the poor decreased in Algeria between 2013 and 2019 while moderate vulnerability slightly increased particularly in rural areas. On the other hand in Tunisia, the decrease in moderate vulnerability among the poor came at the expense of an increase in severe vulnerability between 2012 and 2019.