Does the Poverty Level Minimum Basket Adequately Inform the Scope of Governments' Poverty Reduction Strategies?

Abstract

1. Introduction

Addressing poverty remains a key global development agenda as stipulated in the 2030 Agenda for Sustainable Development Goals (SDGs). Goal 1 of SDGs aims to ‘end poverty in all its forms everywhere’ while the African Union (AU) Agenda 2063 Aspiration One (1) emphasises on African people having a high standard of living, and quality of life, sound health and well-being. Historically, interventions aimed at addressing poverty have broadly emphasised on growing economic activity and redistribution of incomes and resources including through social protection programs. Nonetheless, poverty remains high in Sub-Saharan Africa (SSA) having reduced from about 51% in 2005 to 41% in 2015, relative to a decline by 11% from 21% in the same period at global level (Patel, 2018).

Kenya registered an average economic growth rate of above 5% for almost a decade. However, poverty remained persistent with a less than 1% annual decline. Results from the Kenya National Bureau of Statistics (2018) reveal that the overall poverty headcount rate reduced from 46.8% in 2005/06 to 36.1% in 2015/16. The overall headcount poverty in rural areas decreased from 52.5% in 2005/06 to 40.1% in 2015/16 while for urban areas it reduced by 2.7% to 29.4%. Food poverty headcount rate declined from 45.8% in 2005/06 to 32.0% in 2016, demonstrating the dominance of food poverty in the overall poverty level headcount. Reductions in food poverty incidence was registered in rural areas by 11.4% to 35.8% while the urban food poverty declined by 16.0% to 24.4% in 2015/16. While poverty rate has declined substantially, overall, the total number of the poor declined marginally from 16.6 million in 2005/06 to 16.4 million in 2015/16.

Government interventions should guide allocation of resources to priority areas that facilitate a decent life with adequate food consumption. Further, the basic needs basket used in the construction of poverty line need to adequately guide on intervention to eliminate the deprivations. For example, Bluhm et al. (2018) established that prior to 2000, economic growth largely influenced the poverty levels, however this changed necessitating the revision of poverty strategies. Westmore (2017) also noted that growth and social security system in China had served to reduce the number of people living in extreme poverty since 1970s. However, Anderson et al. (2018), in reviewing various studies found mixed results on the implications of government spending including pro-poor spending and health spending on poverty reduction. This implies that unclear identification of components of the basic-needs basket used to characterise the poverty line could result to unsuccessful government interventions sought to eradicate poverty.

Since 1990s, measurement of poverty has gained a multidimensional approach incorporating both monetary and non-monetary components. This means broadening the concept of poverty from its narrowest sense of lack of income, to encompassing multidimensional issues including basic needs and other avenues of accessing resources, social power relations (inequality) and human capabilities. Further, poverty is increasingly linked to deprivation which reflects on what poor people ‘have or do not have’ of the necessities of life (Alcock, 2006). In defining the poverty line, the cost of a minimum basket of necessities is considered. Such a basket can help to appropriately focus policies and inform the priority areas to be supported by government and development partners.

Goal 1 of SDGs considers the following components of poverty; income, hunger (food), malnutrition (food quality), access to education, social discrimination and exclusion and participation in decision-making. This emphasises that measurement of poverty can use varying components and approaches in investigating poverty (Jones and Tvedten, 2019). The multidimensional index includes ten indicators¹. A person is multidimensionally poor if she/he is deprived in three or more dimensions. The new index shows that 23.4 million (53%) Kenyans were multidimensionally poor in 2015/16 compared to 15.9 million (or 36.1%) considering income only.

It is possible that what a person living in poverty is deprived of among the necessities in one period, is different from what the same person is deprived in another period. This implies that the characterisation of poverty line should change across time. The composition of the minimum basket is therefore dynamic reflecting change in tastes and preferences across locality or gender. It could also be that certain components in the basket persist over time perhaps because the government interventions were not effective in addressing the issues. Thus, understanding the composition of the minimum basket would serve to understand how to ensure no one is left behind be it long-term poor, recently/poor or episodic poor (Alkire et al., 2017).

This study looks at the components of the poverty line minimum basket in informing the interventions undertaken by the government to understand the challenges and opportunities in appropriately focusing the mix of interventions to effectively address poverty in Kenya. This is also an attempt to finding out how identified indicators of measuring basic deprivation sufficiently inform policy makers in coming up with interventions that address deprivations adequately.

1.1. Objectives

The objective of the study is to generate evidence and knowledge on the effects of poverty reduction strategies. Specifically, the study aims to:

• a) Examine whether access to various interventions make a difference in the lives of the poor.

• b) Assess the impact of poverty reduction strategies on poverty and the cost of universalising the transfers.

• c) Draw lessons on what to consider in designing the mix of policy interventions that are highly impactful in reducing poverty.

2. Review of poverty measurement approaches and surveys

2.1. Defining poverty level in Kenya

Determination of the poverty line is the most critical first step in poverty analysis. In the consumption-based, money-metric, approach to poverty, two methods are used to derive the poverty line: The Food Energy Intake (FEI)² method and the Cost of Basic Needs (CBN)³ method. Some studies have used both the FEI and CBN method in their analyses of poverty, arguing that these two approaches are not mutually exclusive and can be used in a complementary way—the former to set the food poverty line and the latter to arrive at a total poverty line (Piachaud, 1987).

It is important to recognise that poverty lines in Kenya have previously been estimated using a modified linear programming approach by Crawford and Thorbecke (1980), which assumed a daily adult equivalent calorie intake of 2,250 calories, using information available on current prices, food weight-to-calorie conversion factors, and the share of food in total expenditure. The Foster-Greer-Thorbecke (FGT) index of poverty measurement measures the head count, depth and severity of poverty in most studies⁴ (Foster et al., 1984; Ravallion, 1993; Ali and Thorbecke, 2000).

The food poverty line defines a certain minimum nutrition level an individual aim to achieve at the lowest cost; while considering or accounting for regional preferences and prices (Greer and Thorbecke, 1986). In a classical linear model, price of commodities is considered since other variables including the required dietary allowance, calories content of food and minimum quantities required for each commodity are exogenously determined. The classical linear programming solution may therefore imply consumption patterns that do not correspond to local tastes and preferences.

To ensure the consumption basket corresponds to the local consumption and production patterns. Wasay (1977) took the major food items in the budget of a low-income family and applied the percentage of the total calorie intake attributable to each item to calculate the desired consumption levels of the various food items. Further, Ali and Thorbecke (2000) provided estimates of the state and path of poverty in sub-Sahara Africa using income distribution between rural and urban areas. Finally, the minimum expenditure on each item was derived using the respective commodity prices. The total amount was then used to calculate the poverty level. This means that first, you determine the low-income families and define the cut-off point in terms of what to include in the basic need basket⁵.

Besides the food poverty, other nonfood expenditure items are crucial in determining the poverty levels, this including employment, schooling and other essential basic needs (Sen, 2004). Studies on non-monetary measures focuses on the deprivation and social exclusion approaches and examines the extent to which households have been excluded on social interventions. Most researchers identify a set of items that constitute the essentials of life—things that no family should have to go without—and these form the basis of the indicators of deprivation⁶ and exclusion⁷.

The deprivation indicators highlight the fact that many families are unable to afford basic needs. The exclusion indicators show how lack of access to basic services and social and economic activities affects poverty interventions. Addressing the different dimensions of social disadvantage aid in developing programs to combat the different forms of poverty levels in the society.

3. Does access to the various interventions make a difference in the lives of the poor?

This section focuses on methodological approaches for examining effects of various poverty interventions on the poor and the impacts of these interventions.

3.1. Methodology

The approach used focuses on explaining the role of public sector socio-economic interventions in reducing poverty. In addition, the methods employed explain the differing probabilities of benefiting from public sector interventions by poverty status of a household. To achieve these objectives, various estimation models were used. A binomial model was used to compute the likelihood of benefitting from a public sector intervention for the poor and non-poor. In this model, the dependent variable takes the value 1 if a household (or individual) is poor and zero otherwise. Poverty is defined based on per adult equivalent terms.

Besides the binary logit, a polychotomous logit model is used to assess or explain the differential effects of the interventions on the non-poor, poor and extremely poor. In this model, the population subsamples are ordered using poverty lines as cut-off points. The binary and polychotomous logit models may not explain the relationship between the interventions and the intensity of poverty. The study's approach therefore estimates a Tobit model—which is used to explain how receipt or non-receipt of an intervention may explain the intensity of poverty. In this model, the dependent variable is a poverty index restricted to the poor households while the non-poor households are assigned a value of zero.

Even so, the three sets of regressions may not answer the predicted impacts and costs associated with scaling up the public sector interventions. This aspect of the objective is achieved through ex-ante or ‘naïve’ simulations supplemented by propensity score matching (PSM).

3.1.1 Estimation equations

As for the other variables in the general form of the estimated models, $\delta$ is a constant term, $x_i^{\prime }$ is a vector of explanatory variables. Most of the explanatory variables⁸ were dummies taking a value of 1 if a household has a particular attribute of interest, e.g., received a particular cash transfer or consumed a particular good/service. A comprehensive list of control variables⁹ used were those found to be important in explaining poverty from past studies on poverty in Kenya. ${\beta }_i^{\prime }$ is a vector of parameter estimates each corresponding to the respective variables, while ${\mu }_i$ is the error term.

Regarding simulations, specifically, the study measured the predicted change in poverty at national level using the Foster et al. (1984) poverty measures before and after the interventions. Three metrics are produced before-and-after the interventions—and these are measures for the poverty headcount (P0)—or the proportion of the population that is below the national poverty line. The poverty gap (P1) or the average amount by which the population is below the poverty line, and the poverty severity index (P2) the square of the poverty gap. The nominal costs of the transfers are easily derived from the transfer per household multiplied by the number of beneficiary households.

3.1.2 Construction of variables

The construction of variables and their descriptive statistics is summarised in Table 1. Poverty, the dependent variable, was defined based on per adult equivalent consumption and its construction is discussed in the previous subsection. The construction of each of the variables is described in the second column of the table. The composite index (compindex) is a variable that varies from 1 to 6. The index takes the value 1 if a household received only one of the public sector interventions (i.e., consumption of fortified maize (X), consumption of fortified wheat (Y), consumption of fortified fat (Z), consumption of free health services (Health1), Bursary, orphaned and vulnerable children cash transfer (OVC CT), or older persons cash transfer (OPCT)). Its value increases with the number of interventions received. Age squared is divided by 100 to scale up the estimated coefficients. The constructio\Noissue n of the rest of the variables are clear from Table 1.

Besides the monetary measure of poverty, this report also uses the multidimensional poverty index (MPI) measure. The MPI used as a dependent variable in some of the regressions follows the approach suggested by Alkire and Santos (2010) and Santos and Alkire (2011) and used in the Human Development Report. It has three dimensions, education¹⁰, health and living standard¹¹. Table 2 below summarises the deprivation definitions for each of the ten indicators.

Adults are considered malnourished if the BMI is below 18.5 while children are considered malnourished if their z-score of weight-for-age is below two standard deviations from the median of the reference population. A household has improved sanitation if it has a flush toilet, ventilated improved pit or composting toilet, provided that they are not shared. A household has access to safe drinking water if the water source is either piped water, public tap, borehole, protected well, protected spring or rainwater, and it is within a distance of 30 minutes' walk (roundtrip). Each of the three dimensions is given a weight of one third. We compute the indicators and in each case an indicator takes a value of 1 if the household is deprived and 0 otherwise. The results are summarised in Table 3 below.

The indicators are summed up into three dimensions each set of indicators weighted such that the weights add up to one third. The respective weights of each of the two education and health indicators are 1/6 while each of the five standard of living indicators has a weight of 1/15. Table 4 below indicates the descriptive statistics of the dimensions. A score of 0.33 in education or health would suggest that a household is deprived in each of the two indicators of these dimensions.

The MPI reflects both the incidence or headcount ratio (H) of poverty—the proportion of the population that is multidimensionally poor—and the average intensity (A) of their poverty—the average proportion of indicators in which poor people are deprived. The MPI is calculated by multiplying the incidence of poverty by the average intensity across the poor (H*A). The cutoff for poverty that we use is that a person is identified as poor if he or she is deprived in at least one third of the weighted indicators (i.e., a score of 0.33). The mean incidence of poverty, H, is 77.4%. The average share of weighted indicators in which the poor are deprived, A, is 61.9%. The MPI is 47.9%.

3.1.3 Data source and type

The data used for the analyses are the KIHBS 2005/06 and KIHBS 2015/16 from the Kenya National Bureau of Statistics. The data were representative of the whole country and covered about 13,000 and 21,773 households respectively. The 2015/16 KIHBS was conducted over a 12-month period (September 2015–August 2016) and was designed to provide estimates for various socio-economic indicators at the national and sub-national levels, and place of residence (rural and urban areas). The sample was divided into four quarters (a consecutive 3-month period) to capture seasonality. Each of the 2,400 clusters was randomly assigned into one of the quarters to generate nationally representative quarterly samples of approximately 600 clusters that can be analysed independently.

3.2. Regression results

The regression tables are presented at the end of the discussions. The poor have a higher probability to uptake the package of interventions being implemented by the government as compared to the non-poor as indicated by the positive and significant COMPINDEX variable. The probability is higher among the urban households as compared to the rural households (Table 5).

While the poor have a higher probability to uptake the interventions, it is much lower among the extreme poor as compared to the other poor group. While in both rural and urban areas, the probability for the extreme poor is less than that of the other poor households, it is much lower in urban areas than in the rural areas (Table 6).

The uptake of the interventions helps in reducing the poverty gap (Table 7). The rate at which the poverty gap is closed is higher in the rural areas as compared to the urban areas when the interventions are taken into account.

Although the uptake of the interventions and multidimensional poverty have the expected sign, the result are not significant in the overall, rural and urban samples (Table 8).