Nowcasting multidimensional poverty in the occupied Palestinian territory




Studying inequality and poverty in the Arab region is a challenging task. As pointed out in Atamanov and others (2020), the Arab region has insufficient statistical capacity. Few countries survey their population on income and expenditure. In Arab countries that run these types of surveys, statistical offices perform this exercise irregularly. In this context, analysts have to find alternative approaches to assess inequality and poverty in the region. One alternative approach is to look at multidimensional poverty in a multinational poverty index (MPI) framework (Alkire and Santos, 2014; Alkire and Foster, 2011). This approach relies on the type of data the analyst usually finds in the demographic health surveys. ESCWA and others (2020) has adopted the approach to the Arab region's specificity.
Following the standard approach of building such an index, ESCWA chose the dimensions and indicators to reflect the development goals of the Arab region. Unfortunately, because demographic health surveys do not include information on household income or expenditure, the Arab multidimensional poverty index does not include a monetary poverty dimension.

Since better data is available for the State of Palestine, Khawaja, Al-Saleh, Reece and Conconi (2020) propose a context-specific Palestinian multidimensional poverty index. The proposed index's dimensions and indicators include a monetary dimension and information on security and freedom, which are essential considerations in the Palestinian context. The choice of dimensions and indicators builds on the Palestinian basic law and Palestinian child protection and labour laws. Khawaja and others (2020) build the Palestinian multidimensional poverty index using the last version of the Palestinian Consumption and Expenditure survey (PECS) of 2016. The previous version of the same survey dates of 2011. Besides, the PECS 2011 does not include all indicators used in the Palestinian multidimensional poverty index. The presence of such data limitation points to the need to nowcast these poverty indices between two non-consecutive surveys and forecasting beyond the last survey.

Despite their strong assumptions, nowcasting and forecasting approaches to computing poverty indices remain an essential exercise from a policy perspective. Even if there is uncertainty on the exact magnitude of the index, capturing at least the direction of change is essential for the design of evidence-based policies. Some papers, such as Klassen and Lange (2012), proposed models to nowcast deprivation in specific dimensions such as health. However, until very recently, the literature was still silent on proposing models to nowcast multidimensional poverty except for a forthcoming paper (Ram, 2020), an OPHI working paper (Alkire, Nogales, Quinn and Suppa, 2020), and an ESCWA technical report (Makdissi, 2020).

The present report aims to contribute to this modelling exercise for nowcasting the Palestinian multidimensional poverty index. Our approach is related to Makdissi (2020) because it proposes a dynamic modelling approach of multidimensional poverty that separately models each indicator's change. This analytical choice allows for more flexibility on the modelling assumptions made to estimate the counterfactual change in each indicator's univariate distribution. In addition to Makdissi (2020), we propose solutions in a context where the indicators are not available in two consecutive surveys. The present paper's modelling approach considers political and security factors that may influence the dynamics of some of the indicators more than chronology and change in economic conditions. This modelling approach is handy in a context of a multidimensional poverty index that includes dimensions such as individual freedoms and security.

Another contribution is the Palestinian State model disaggregation into Gaza and West Bank regions and modelling them separately. This exercise is essential since the two regions are physically separated and have different experiences in terms of occupation type. The number of observations in Gaza being relatively low, the simulations at the regional level are less reliable, even within the theoretical model framework. Nevertheless, the simulation results' point estimates indicate that multidimensional poverty sometimes moves in opposite directions in both regions. In particular, the relative stability of multidimensional poverty in 2020 compared to 2019 at the state-wide level results from an increase in the West Bank's multidimensional poverty and a reduction of multidimensional poverty in Gaza during the pandemic. This difference is essentially due to a substantial decrease in Gaza injuries in 2020 and 2021. The reduction in the number of injuries in the West Bank was smaller than the one in Gaza. Besides, there was an increase in the number of demolitions in the West Bank in 2020 and early 2021.