Assessing the Predictors of Lived Poverty in Kenya: A Secondary Analysis of the Afrobarometer Survey 2016

Abstract

Poverty remains pervasive in Kenya despite efforts to reduce it. This may be attributed to poor understanding of its predictors. Using Afrobarometer Round 7 data and employing multivariate logistic regression, the study found that age, religion and political affiliation were risk factors associated with poverty while education was a protective factor. The study recommends that investing more in all levels of education is key to reducing poverty in Kenya. Further, social protection policies and programmes for the elderly, as well as initiatives to enhance savings for those in the informal economy, are important for poverty reduction.

Keywords

Poverty lived poverty Afrobarometer education social protection Kenya

Background

Poverty reduction has been at the heart of global discussions for decades. Global efforts to eradicate poverty have achieved some significant gains as more than 1 billion people escaped extreme poverty between 1990 and 2015 despite a global population increase of 2 billion people (UNDP, 2016). By 2013, the global extreme poverty rate had declined to 11% from 35% in 1990. Recent evidence suggests that poverty is also dropping in most African countries. A study by Mattes et al. (2016) showed that between 2011 and 2014, lived poverty dropped in 22 of the 33 countries surveyed.

Even with this progress, there were more than 700 million people living on less than US$1.90 a day in 2013 Developing countries are still worst hit by extreme poverty. A Multidimensional Poverty Index (MPI) shows that almost 1.5 billion people in the developing countries live in poverty and a third of them are in Sub-Saharan Africa (UNDP, 2016). While the situation is bad in developing countries, developed countries are not left behind. Dulani et al. (2013) noted that lived poverty is still pervasive across Africa despite a decade of economic growth. The International Labour Organisation (ILO) estimated that in 2012, more than 300 million people lived in poverty in developed countries, suggesting that poverty is on the rise in developed countries (ILO, 2016).

Kenya has been at the forefront of eradicating poverty right from the time of independence when the founding fathers declared poverty as one of the three ills that had to be fought. Despite the concerted efforts, eradicating poverty has been elusive. In 2000, the world launched the Millennium Development Goals (MDGs) to drive global progress in, among other issues, poverty and hunger (Henao et al., 2017). These were superseded by the Sustainable Development Goals (SDGs) upon their expiry in 2015. At the heart of the SDGs is an ambitious target to end poverty by 2030 (UNDP, 2016).

Kenya’s Vision 2030 – the country’s economic blueprint to becoming a middle-income industrialized economy by 2030 – envisions poverty levels below 10% by 2030. Indeed, the goal was to reduce poverty rates by 2012 to a range of 30–35% (GoK, 2007). This was not achieved, as over 45% of Kenyans still live below the poverty line (UNDP, 2016). It is, however, not surprising that the poverty levels have not dropped over the last decade. While the Vision 2030 anchored poverty reduction under the social pillar, albeit as the last item, the Second Medium Term Plan (MTPII) that was to operationalize the Vision 2030 between 2013 and 2017 did not include poverty reduction as part of the social pillar, or as part of any other pillar within the MTPII framework.

While a number of studies have examined the determinants of poverty in Kenya, poverty has not been measured as an experiential phenomenon in prior studies; for example, works of Geda et al. (2001) and Mberu et al. (2014), while others like Kimani and Kombo (2010) have been designed as theoretical discussions. The experiential measure of poverty used in this paper is lived poverty. Lived poverty is a sum of responses on a set of questions about lack of basic necessities, namely food, water, medical care, cooking fuel and cash income (Dulani et al., 2013; Mattes et al., 2016). Given the importance of poverty reduction as a development agenda in Kenya, it is important to study why poverty is still pervasive despite the efforts to address it and to, especially, investigate the socio-economic features of the poor in Kenya.

The aim of this research is to assess socio-economic factors associated with high lived poverty in Kenya. As Kenya works towards achieving Vision 2030 goal of reducing poverty to single digits as well as achieving the SDG 1 (no poverty by 2030), it is important to evaluate some of the socio-economic factors that drive poverty in a bid to offer policy recommendations on where more effort should be targeted to eradicate poverty. This study cuts across five of the 17 SDGs (SDG 1, SDG 2, SDG4, SDG 8, and SDG 10). Using a unique dataset of Afrobarometer Survey – an African institution that collects opinion data from all over Africa on issues that affect Africa – this study provides a different perspective on how poverty can be addressed in Kenya by focusing on poverty as a multidimensional index of lack of basic necessities such as food, water, medical care, cooking fuel and cash income.

The study is structured into six parts including this background, which provides the rationale for the study. The next section presents a literature review where the conceptual discussions, as well as the theoretical relationships, are made. The third section presents the methodology while the fourth section presents the results. The fifth section is the discussion of the results while the last section concludes and offers policy recommendations.

Literature review

Literature is replete with numerous socio-economic predictors of poverty. These range from demographic predictors (marital status, gender, age, place of residence, among others) to economic factors (such as employment, religion and education, among others). This section presents a review of the factors that are used in this study to explain lived poverty in Kenya.

Gender and poverty

A large set of current literature suggests that women experience poverty more than men and that gender plays a role in poverty (Alkire et al., 2014; UNDP, 2016). For instance, in 2012 there were 117 women in poor households for every 100 men in Latin America and the Caribbean (United Nations, 2015). In Portugal, studies have also found that women are more likely to be poor than men (Bastos et al., 2009). The severity of poverty for women increases with age as older women suffer more than older men (Shriver Center, 2016). A study carried out by UN Women (2015) found that women were 37% more likely than men to live in poverty in the European Union. Studies from developing countries have also shown that women are more vulnerable to poverty than men are (see Agbodji et al., 2013; Shimeles and Verdier-Chouchane, 2016; Vijaya et al., 2014).

For instance in West Africa, Agbodji et al. (2013) found that in as much as women were more or less likely to experience income poverty just as men, they were more likely than men to be multidimensionally poor. Gendered poverty is a problem in Uganda and this has seen the government coming up with programmes to address this problem (Hasaba, 2014). In South Sudan, Shimeles and Verdier-Chouchane (2016) found a negative correlation between female-headed households and poverty status. In Kenya, Geda et al. (2001) found that female-headed households were more likely to be poor than male-headed households. Mberu et al. (2014), while focusing on the slums in Nairobi, revealed that more female-headed households were chronically poor relative to the male-headed households. However, the studies on Kenya did not use lived poverty as the outcome variable. The findings of these studies show a consistent relationship between gender and poverty. Therefore, gender is included in this study as a predictor variable to explore its relationship with lived poverty in Kenya.

Age and poverty

There is a long-held view that poverty does not rise with age as it is always assumed that the older people had an opportunity to accumulate wealth over their lifetime and, therefore, are comparatively well off (Barrientos, 2002; Ogwumike and Aboderin, 2005). According to Barrientos (2002), this view needs to be re-examined as the standard measures underestimate old-age poverty, especially in developing countries. Ogwumike and Aboderin (2005) found that older people were more at risk of poverty than the younger adults in Nigeria and Ghana were. Evans et al. (2007)’s study in Viet nam provides mixed results on the relationship between age and poverty.

In Kenya, a study by Geda et al. (2001) revealed that age had a significant influence on poverty status. Mberu et al. (2014) also found that age had a U-shaped relationship with poverty as households headed by the youngest and the oldest heads are likely to be poorer than the rest of the groups. This was also the case in South Sudan (Shimeles and Verdier-Chouchane, 2016). The study on Kenya did not use lived poverty as the outcome variable and neither were they based on national surveys. This gap needs a further examination. This study attempts to assess the link between age and lived poverty in Kenya.

Education and poverty

Regarding the nexus between education and poverty, there is a rich literature that agrees that education is key to poverty alleviation initiatives. In Rwanda, Habyarimana et al. (2015) found that education level was a significant predictor of household poverty. As the study noted, education had a positive relationship with poverty. This shows that having a primary or secondary education was associated with decreased poverty levels as compared to no schooling at all. A study by Shimeles and Verdier-Chouchane (2016) in South Sudan revealed a negative relationship between education and poverty, with those with primary education earning 36.5% more than those without any form of education, while those with university degrees earning 188.6% more than those with no schooling .

In Kenya, Mberu et al. (2014) revealed an inverse linear relationship between education and poverty. They observed that as educational attainment increased, the proportion of households in poverty dropped monotonically. Geda et al. (2001) also found that education was an important determinant of poverty status, as primary education was particularly important as a poverty-reducing factor in rural areas, while having no education was associated with higher probability of being poor. Secondary and university education was important in alleviating poverty. The studies on Kenya were not, however, based on a national sample, so the results are not generalizable to Kenya. This study tests whether education plays a key role in lived poverty in Kenya.

Place of residence and poverty

The influence of rural residence on poverty has been studied by a number of scholars over the years. A study by Habyarimana et al. (2015) in Rwanda showed that place of residence was a significant predictor of poverty. More specifically, the study found that an urban household had a positive association with the household asset index, which suggests that being an urban household decreases poverty as opposed to being a rural household. Shimeles and Verdier-Chouchane (2016) found a positive relationship between living in rural areas and poverty levels in South Sudan. In Kenya, Geda et al. (2001) found that those living in urban areas were less likely to be poor as compared to those living in rural areas. The study on Kenya used a different outcome variable to measure poverty and, therefore, provide a gap that requires further exploration. Given these prior findings, it is important to examine whether residence explains lived poverty in Kenya.

Employment and poverty

A number of studies have found a strong relationship between employment and poverty. Studies generally examine the form of employment and how it affects poverty levels. For instance, Shimeles and Verdier-Chouchane (2016) also revealed that being employed in an industry or service sector had negative relationships with poverty status in South Sudan. In Kenya, Mberu et al. (2014) found that households whose heads were formally employed were less poor than those whose households were either self-employed or were employed in the informal sector. Further, Geda et al. (2001) revealed those engaged in the agricultural sector are more likely to be poor. These studies confirm that employment influences poverty status. This is dependent on the sectors of the economy in which one is employed. However, the studies use a different measure of employment, which offers the need to examine this link further using a different set of measurements. This study examines whether employment and unemployment influence lived poverty in Kenya.

Religion and poverty

It is said that religion and poverty are strange bedfellows. Thus, the subject of religion as a predictor of poverty has been studied by a number of scholars. Keister (2008) examined whether religious affiliation affected wealth ownership for conservative Protestants and revealed that indeed the affiliated affected wealth both directly and indirectly. Another study by Keister (2007) showed that individuals raised in Roman Catholic families accumulated more wealth than other groups. The study also revealed that Catholic values contribute more to a culture of wealth accumulation. In Kenya, Mberu et al. (2014) showed that more Muslims than Christians were in chronic poverty over the four-year study period, and that fewer remained out of poverty than Christians did. However, this study did not measure poverty using the experiential measure of poverty. Therefore, this study examines the relationship between religion and lived poverty in Kenya.

Political affiliation and poverty

The literature on the influence of political party affiliation is not well developed. The available literature focuses on the attitudes towards poverty as observed from various political ideologies and party affiliations (Cozzarelli et al., 2001; Wagstaff, 1983). For instance, Wagstaff (1983) investigated the attitudes to poverty by examining political affiliations in Britain. In the study, those who supported the Conservative Party had more negative attitudes towards the poor than supporters of the Labour Party did. In a study by Cozzarelli et al. (2001), political affiliation strongly predicted the effect towards the poor, stereotypes about the poor and attributes towards the poor. This might suggest that poverty had a political affiliation. However, a 2013 report by The Brookings Institution concludes that poverty does not have a political affiliation (Berube et al., 2016). One study, however, stands out as having examined the relationship between political party and poverty. In an attempt to study cultural and structural causes of poverty, Jordan (2004) found that the Republican Control Index (RCI) had a positive effect on poverty. This suggests that poverty rose with RCI. No study is available in Kenya that has explored this link. The present study addresses this gap by examining whether affiliation with government or opposition affects lived poverty in Kenya.

Methodology

This study used data from Round 7 of the Afrobarometer Survey. The survey was conducted in 2016 as part of an opinion survey on a number of issues that affect selected African countries. Data for this study – on Kenya – were provided by the Institute of Development Studies of the University of Nairobi in an SPSS format. This data was then converted into Stata file for this analysis. The sample size for the survey was 1599 and covered all the 47 counties of Kenya. The sampling procedure and the data collection process for Afrobarometer Surveys are available on their website (Afrobarometer, 2017).

The primary outcome for this study is lived poverty – an experiential measure of poverty through a set of five questions on lack of basic necessities (Dulani et al., 2013; Mattes et al., 2016) namely food, water, medical care, cooking fuel and cash income. This measure mixes both subjective and objective approaches to measuring poverty (Bratton, 2006) and has strong construct validity and reliability (Mattes, 2008). As much as this measure is one-dimensional (Meyer and Keyser, 2016), it is preferred here as a measure of poverty since it is the best index for measuring poverty from the Afrobarometer data than using the asset-based wealth index.

The responses on lived poverty were on a five-point Likert scale ranging from ‘never’ to ‘always’. The responses from each of the items were summed up to come up with an index for each respondent hence forming a continuous variable with a score of zero to five. However, this study used a revised binary lived poverty measure where the index was grouped into ‘no/low lived poverty’ and ‘moderate/high lived poverty’ in order to meet the logistic regression requirements. The new binary measure was the outcome variable in this study.

Seven predictors were examined. These were gender, age, education, employment, religion, residence and political affiliation. Gender was measured as a binary variable of male and female. It measures the sex of the respondents. Prior studies examining socio-economic predictors of various outcomes have used gender. Age was a continuous variable in the original dataset ranging from 18 to 93 years. This variable was re-categorized based on six age groups. Education measured the level of education of respondents. The variable was re-categorized into five distinct groups from ‘no formal schooling’ to ‘university’.

Employment is an economic predictor variable measuring whether the respondent is employed or not. This was re-calculated from the original variable by re-categorizing it into a binary variable of ‘employed’ and ‘unemployed’. Religion measured whether the respondent was a Christian, a Muslim or practised other religion (or did not practice any religion at all). This measure was also a re-calculated one from the original measure in the dataset. Finally, place of residence is a binary variable which measured whether the respondent lived in a rural or urban setting.

Political affiliation measured whether the respondent was affiliated to a political party or a coalition that currently forms the government or not. If the affiliated was for the party or affiliated that was in support of the current government, then the response was coded as government, otherwise it was coded as opposition.

The analysis was carried out using Stata v12.1. The svyset commands were first applied to ‘account for oversampling of urban primary sampling units (PSUs), and to adjust for clustering of observations within PSUs and stratification’ (Manzi et al., 2014) by county. The predictors were tested for collinearity during model building. The study found a low correlation among the predictors. A multivariate logistic regression model was used to explore the predictors, which were identified through a manual backward stepwise regression technique. This paper reports the odds ratios (ORs) and 95% confidence intervals (CI).

Results

Descriptive results on the outcome variable are presented in Tables 1 and 2. The results in Table 1 reveal that 47% of the respondents had gone at least once without food, 48% had gone at least once without water, 53% had gone at least once without medical care, 33% had gone at least once without cooking fuel, and 83% had gone at least once without cash income. Thus, in order of severity, cash income is the item most households had gone without followed by medical care, water, food and cooking fuel.

Table 1.

Percentage of respondents who went without basic necessities.

Gone without …	Never	Just once	Several	Many times	Always	Total
Food	53	20	19	7	1	100
Water	52	13	17	13	5	100
Medical care	47	18	22	9	3	100
Cooking fuel	67	16	11	5	0	100
Cash income	17	19	34	24	6	100

Table 2.

Descriptive statistics on lived poverty index in Kenya.

Lived poverty	Frequency	Percent
No lived poverty	183	11.5
Low lived poverty	704	44.2
Moderate lived poverty	531	33.3
High lived poverty	175	11.0
Total	1593	100

The responses on lived poverty were further re-categorized into the severity of lived poverty from no lived poverty to high lived poverty. Table 2 shows that 11.5% of the respondents experienced no lived poverty, 44.2% experienced low levels of lived poverty, 33.3% experienced moderate levels of lived poverty, and 11% experienced high levels of lived poverty. Thus, most respondents experienced low to moderate lived poverty.

In order to examine what socio-economic factors explain lived poverty levels in Kenya, it is important to examine whether lived poverty differs across the predictors. In order to carry out this, a chi-square analysis was carried out for all the predictors. The results are presented in Table 3.

Table 3.

Bivariate relationships between predictors and lived poverty in Kenya.

Characteristics	N	%	[95% confidence interval]	p-value
Gender				0.5393
Male	364	45.12	[39.68, 50.68]
Female	360	43.55	[38.69, 48.55]
Age group				0.0010
18–24	101	33.24	[27.17, 39.93]
25–34	240	42.64	[37.06, 48.41]
35–44	176	46.75	[39.71, 53.91]
45–54	99	51.45	[42.93, 59.88]
55–64	59	53.92	[41.92, 65.49]
65+	44	53.43	[40.05, 66.34]
Religion				0.0000
Christian	549	40.73	[36.49, 45.11]
Muslim	94	65.04	[55.42, 73.57]
Other	72	54.53	[43.88, 64.79]
Employment status				0.0001
Unemployed	428	51.64	[46.15, 57.08]
Employed	289	36.88	[30.92, 43.26]
Education				0.0000
No formal education	87	77.24	[68.94, 83.85]
Primary	324	52.68	[47.3, 58]
Secondary	229	38.42	[33.15, 43.98]
Post-secondary	58	28.27	[21.73, 35.88]
University	18	20.48	[14.26, 28.51]
Residence				0.0044
Urban	226	36.97	[32.44, 41.74]
Rural	498	48.45	[42.24, 54.71]
Political affiliation				0.0013
Government	221	41.26	[34.86, 47.96]
Opposition	202	56.11	[49.21, 62.78]

Of the 1593 people who responded to the questions on lived poverty, 724 of them (44.3%, 95% CI: 39.8%, 49.0%) had high lived poverty. Table 3 shows that the likelihood of experiencing high lived poverty increased with age as one moves from the 18–24 age bracket (33.2%, 95% CI: 27.2%, 39.9%) to 55–64 age bracket (55.9%, 95% CI: 41.9%; 65.5%). The results on education show that the poverty rates decline as you go up the education ladder from no formal education (77.2%, 95% CI: 68.9%, 83.9%) to a university degree (20.5%, 95% CI: 14.3%, 28.5%).

In the bivariate analysis (Table 4), the following factors were significantly associated with higher lived poverty: age (p = 0.0010), religion (p < 0.001), employment status (p < 0.001), education level (p < 0.001), place of residence (p = 0.0044), and party affiliation (p = 0.0013). Gender did not have a significant association with higher lived poverty (p = 0.5393). These predictors were included in the analysis with two separate models – full and reduced model (the model without gender, as it was insignificant in the bivariate analysis).

Table 4.

Multivariate logistic regression model for socio-economic factors influencing lived poverty in Kenya.

Characteristic	Full model				Reduced model
	OR	p	[95% confidence interval]		OR	p	[95% confidence interval]
Age group
18–24	1.00
25–34	1.33	0.21	0.85	2.07	1.33	0.21	0.85	2.08
35–44	1.67	0.06	0.97	2.86	1.71	0.06	0.99	2.96
45–54	1.86	0.05	0.99	3.50	1.93	0.04	1.05	3.56
55–64	1.93	0.10	0.89	4.21	2.02	0.07	0.94	4.32
65+	1.46	0.37	0.63	3.37	1.55	0.29	0.68	3.54
Employment
Unemployed	1.00
Employed	0.82	0.29	0.56	1.19	0.83	0.35	0.57	1.22
Education
No formal schooling	1.00
Primary	0.65	0.24	0.31	1.34	0.67	0.27	0.33	1.37
Secondary	0.37	0.01	0.18	0.77	0.39	0.01	0.19	0.79
Post-secondary	0.26	0.00	0.11	0.61	0.27	0.00	0.12	0.62
University	0.18	0.00	0.07	0.49	0.19	0.00	0.07	0.51
Religion
Christian	1.00
Muslim	2.11	0.02	1.16	3.84	2.17	0.01	1.19	3.94
Other	1.00	1.00	0.50	2.02	1.02	0.95	0.51	2.06
Residence
Urban	1.00
Rural	1.20	0.41	0.78	1.84	1.21	0.38	0.79	1.85
Political affiliation
Government	1.00
Opposition	2.02	0.00	1.41	2.90	2.03	0.00	1.42	2.91
Gender
Male	1.00
Female	0.86	0.40	0.61	1.22

In the reduced model (Table 4), several factors were associated with higher lived poverty: belonging to the 34–44 age bracket (OR = 1.71, 95% CI: 0.99, 2.96), or to the 45–54 age group (OR = 1.93, 95% CI: 1.05, 3.56) or to the 55–64 age group (OR = 2.02, 95% CI: 0.94, 4.32) versus belong to the 18–24 age group; being a Muslim (OR = 2.17, 95% CI: 1.19, 3.94) versus being a Christian; and being affiliated to an opposition political party or coalition (OR = 2.03, 95% CI: 1.42, 2.91) versus being affiliated to a government political party or coalition.

Different factors were also associated with lower lived poverty scores: having secondary education (OR = 0.39, 95% CI: 0.19, 0.79) or a post-secondary education (OR = 0.27, 95% CI: 0.12, 0.62) or a university education (OR = 0.19, 95% CI: 0.07, 0.51) versus having no formal education. Surprisingly, the results showed that employment and place of residence did not have any significant relationships with lived poverty.

Discussion

This was a secondary analysis of the 2016 Afrobarometer Round 7 survey data. The study found that more than two-fifths of the respondents experienced higher lived poverty. This number compares fairly with the current national poverty rate of over 40% (UNDP, 2016). Belonging to certain age groups, being a Muslim and being affiliated to an opposition party was associated with the probability of higher lived poverty. Those aged between 34 and 44 years were 1.71 times more likely to experience higher lived poverty than those aged 18–24 years. This increased to 1.93 for those aged 45–54 years and to 2.02 for those aged 55–64 years. Indeed, more than half of the elderly population experienced higher lived poverty. Thus, the severity of poverty increased with age.

This is consistent with Ogwumike and Aboderin (2005), who observed that the elderly were more at risk of poverty than the younger household heads in West Africa. Therefore, social protection programmes that target the elderly should be expanded. In the 2017/18 and the 2018/19 budget statements, the Government of Kenya announced that it would include every Kenyan aged 70 years and above in the social protection programme. However, given that the retirement age in Kenya is 60 years, the life expectancy at birth is 60 years for men and 64 years for women (UNDP, 2016), and poverty incidence in this study was worse for those aged above 64 years, the social protection programme should include all non-working adults aged 64 years and above.

Consistent with prior studies on the nexus between religion and poverty (Keister, 2007, 2008), this study found that religion played a significant role in lived poverty status. Being a Muslim was associated with a higher probability of having higher lived poverty scores. From the results, Muslims were 2.17 times more likely to suffer higher lived poverty than Christians were. As the bivariate results showed, more than two-thirds of the Muslims experienced high lived poverty as opposed to just two-fifth of the Christians. This was also observed by Mberu et al. (2014), who noted that Muslims were more likely than Christians to experience chronic poverty. This calls for further investigation into why Muslims are more affected by poverty than Christians in Kenya. Such an investigation would document whether Muslims suffer just because they are a minority group or because of other factors that can be addressed through policy decisions by various state and non-state actors.

The analysis also showed that being affiliated to an opposition party or coalition as opposed to the government party or coalition was associated with the probability of higher lived poverty. Those affiliated to the opposition were twice as likely to experience higher lived poverty as those affiliated to the government. The bivariate results showed that more than half of those affiliated to the opposition suffered higher lived poverty. While not much literature is available to compare the results on the influence of party affiliation on poverty, this view is consistent with the findings of Jordan (2004) on how party control affects poverty. As much as suggesting that this is the case for Kenya may not be politically correct, the finding cements the urge for those not affiliated to the government to demand equitable distribution of national wealth to all Kenyans regardless of how they voted in a general election.

On the other hand, only education was associated with lower lived poverty. This is consistent with prior literature (Habyarimana et al., 2015; Shimeles and Verdier-Chouchane, 2016). The study showed that having secondary education was associated with a lower lived poverty as compared to those with no education. From the analysis, while over three-quarters of those without formal education experienced a high lived poverty, only just about two-fifths of those with a secondary education experienced higher lived poverty. This dropped slightly below a third of those with post-secondary education and to just 20% of those with an university education. This is consistent with the monotonic drop in poverty with increased educational attained as observed by Mberu et al. (2014) as well as the finding by Geda et al. (2001) that those without education were more likely to experience poverty. Education, therefore, is a sure tool for escaping poverty. Thus, more investment in education is required to improve enrolment rates in primary schools, completion rates in both primary and secondary schools, and transition rates from primary to secondary and from secondary to colleges and universities. Since those without formal education suffer the most, it may be important for the government to invest more in adult education as a method to provide the aged out-of-school population with a chance to have formal schooling. This should also include vocational and technical training to equip them with the necessary skills to be economically productive by starting their own businesses.

Surprisingly, gender, employment and rural residence were not significant predictors of lived poverty. It was expected that the unemployed in the sample would report higher lived poverty. Indeed, the bivariate results had shown that over half of the unemployed experienced high lived poverty while just over a third of the employed reported high lived poverty. From the regression analysis, however, these differences in reported lived poverty did not affect poverty levels significantly. This could be attributed to the measurement of the predictor in this study. Prior studies that have examined employment have modelled it as a combination of another predictor such as age (Evans et al., 2007) or focused on employment in a specific sector of the economy or employment type (Geda et al., 2001; Mberu et al., 2014; Shimeles and Verdier-Chouchane, 2016). Thus, future studies may need to measure employment differently to test its effect on lived poverty. Another reason could be that the sample contained just as many unemployed (51%) individuals as the employed ones (49%).

The study also expected lived poverty to be influenced by place of residence. Given that nearly half of those living in rural residences reported high lived poverty as compared to just over a third of those living in urban areas, the study expected lived poverty to be significantly predicted by place of residence. This is contrary to prior research that has established this link (Geda et al., 2001; Habyarimana et al., 2015; Shimeles and Verdier-Chouchane, 2016). However, this might be explained by the under-representation of the rural (or over-representation of urban) population in the sample used in this study. The 2016 HDI report shows that 74.4% of the population in Kenya is rural (UNDP, 2016), while this study had 64% of the the rural population in the sample. The poverty incidences among the rural population should not be ignored, and investments in projects to reduce poverty, such as more funding for agriculture, roads, electricity and other investments that improve the income generating activity of the rural poor, should be continued.

Gender also failed to predict lived poverty. Therefore, it was left out as a predictor in the final model. The 2016 report on human development showed that in Kenya, women’s Gross National Income (GNI) per capita was US$2357 as compared to US$3405 for men (UNDP, 2016). This income inequality shows that women in Kenya are generally poorer than men are. Therefore, the study expected to find a significant link between gender and lived poverty. Thus, this failure is surprising since prior studies have found a significant link between gender and poverty (Bastos et al., 2009; Geda et al., 2001; Mberu et al., 2014; UN Women, 2015). However, the bivariate results showed that women (43.6%) experience poverty just as much as men (45.1%), just as Agbodji et al. (2013) had observed in West Africa. This may explain the failure of gender to predict lived poverty as women have been found to suffer when multidimensional poverty perspective is examined – one that was not the subject of the current study.

This study is not without its limitations. First, as a secondary analysis of the 2016 Afrobarometer Survey, not all the socio-economic predictors were available for examination. For instance, important predictors such as marriage and the size of the household are not available in the survey. Thus, their association with lived poverty could not be explored despite being important predictors of poverty. Moreover, the secondary analysis may have been influenced by recall bias. This is particularly true regarding the question on the number of times the respondents had experienced a shortage of the basic necessities that formed lived poverty – the outcome variable in this study. Further studies should explore whether other factors not included in this study affect lived poverty.

Secondly, recent studies have shown that these primary predictors of poverty do so through an interaction effect (Evans et al., 2007). This has seen studies explore the influence of women with no education v. women with education, old women versus old men, and female head of households in rural areas versus those in the urban areas, among others on poverty. Therefore, further studies should examine how these predictors influence lived poverty. Finally, since several of these surveys have been conducted over the years, it would be worthwhile for further studies to examine whether predictors of lived poverty have changed over time. This perspective was not the subject of the current study.

Conclusion

The study explored predictors of lived poverty (the likelihood to report higher lived poverty scores) in Kenya. The study found that several socio-economic factors were associated with lived poverty. Therefore, the study proposes specific interventions to address lived poverty in Kenya. Education remains the most significant policy tool to eradicate lived poverty. While education remains mostly a function of the central government, policy interventions at the national level to improve the educational attainment of individuals are preferred. More investment in basic education to improve enrolment rates, transition rates and completion rates is therefore key. Further investment in technical and vocational educational training institutions is also preferred to equip the youths with more marketable skills for employment and entrepreneurship. For adults without formal education, more investment in adult education is preferred.

Old age is a risk factor for lived poverty. The population of Kenyans working in pensionable jobs is very small. Therefore, as they age and stop working, a majority of them do not have a means of survival other than expecting remittances from their children or families. With most of them lacking retirement benefits, the severity of lived poverty rises. As such, there is a need for more social protection programmes targeting the old to be a priority for both the central government and county governments. These include more focus on improving pension schemes, cash transfers and health insurance covers to all the non-working elderly population above the national retirement age. However, as a long-term measure, policymakers should improve the national social benefit plans and find innovative ways to entice all young and working Kenyans in the informal economy to save for their future. This will help cushion them against poverty as they age.

Footnotes

Funding

The author received no financial support for the research, authorship, and/or publication of this article.

ORCID iD

Fredrick Onyango Odhiambo

Author biography

Fredrick Onyango Odhiambo is an independent researcher and a monitoring & evaluation consultant with ResearchPro Solutions, a private research consulting firm based in Nairobi, Kenya. He is also an IDS Scholar at the University of Nairobi.

References

Afrobarometer (2017) Surveys and Methods: Sampling principles and weighting. Available at: http://www.afrobarometer.org/surveys-and-methods/sampling-principles (accessed 14 April 2017).

Agbodji

Batana

Ouedraogo

(2013) Gender Inequality in Multidimensional Welfare Deprivation in West Africa: The case of Burkina Faso and Togo. Washington: The World Bank.

Alkire

Apablaza

Jung

(2014) Multidimensional Poverty Measurement for EU-SILC Countries. London: University of Oxford.

Barrientos

(2002) Old age, poverty, and social investment. Journal of International Development 14(8): 1133–1141.

Bastos

Casaca

Nunes

et al . (2009) Women and poverty: A gender-sensitive approach. The Journal of Socio-Economics 38(5): 764–778.

Berube

Kneebone

Williams

(2016) Suburban Poverty Traverses the Red/Blue Divide. Washington: Brookings Institution.

Bratton

(2006) Poor People and Democratic Citizenship in Africa. Cape Town: Afrobarometer.

Cozzarelli

Wilkinson

Tagler

(2001) Attitudes toward the poor and attributions for poverty. Journal of Social Issues 57(2): 207–227.

Dulani

Mattes

Logan

(2013) After a Decade of Growth in Africa, Little Change in Poverty at the Grassroots. Cape Town: Afrobarometer.

10.

Evans

Harkness

McKay

et al . (2007) The Relationship Between Old Age and Poverty in Viet Nam. New York, NY: United Nations Development Programme.

11.

Geda

de Jong

Mwabu

et al . (2001) Determinants of poverty in Kenya: A household level analysis. ISS Working Paper Series/General Series 347: 1–20.

12.

GoK (2007) Kenya Vision 2030. Nairobi: Government of the Republic of Kenya.

13.

Habyarimana

Zewotir

Ramroop

(2015) Determinants of poverty of households in Rwanda: An application of quantile regression. Journal of Human Ecology 50(1): 19–30.

14.

Hasaba

(2014) Women and poverty eradication efforts in Uganda: Why is ending gendered poverty still far-fetched? In Falola

Abidogun

(eds) Education, Creativity, and Economic Empowerment in Africa. New York, NY: Palgrave Macmillan, pp.43–59.

15.

Henao

Moyer

Namakula

(2017) Africa 2030: How Africa Can Achieve the Sustainable Development Goals. Kigali: Sustainable Development Goal Centre for Africa (SDGC/A).

16.

ILO (2016) World Employment and Social Outlook 2016: Transforming jobs to end poverty. Geneva: International Labour Organisation.

17.

Jordan

(2004) The causes of poverty: Cultural vs. structural: Can there be a synthesis? Perspectives in Public Affairs(Spring) 1: 18–34.

18.

Keister

(2007) Upward wealth mobility: Exploring the roman catholic advantage. Social Forces 85(3): 1195–1225.

19.

Keister

(2008) Conservative protestants and wealth: How religion perpetuates asset poverty. American Journal of Sociology 113(5): 1237–1271.

20.

Kimani

Kombo

(2010) Gender and poverty reduction: A Kenyan context. Educational Research and Reviews 5(1): 24–30.

21.

Manzi

Munyaneza

Mujawase

et al . (2014) Assessing predictors of delayed antenatal care visits in Rwanda: A secondary analysis of Rwanda demographic and health survey 2010. BMC Pregnancy and Childbirth 14: 290–298.

22.

Mattes

(2008) The Material and Political Bases of Lived Poverty in Africa: Insights from the Afrobarometer. Cape Town: Afrobarometer.

23.

Mattes

Dulani

Gyimah-Boadi

(2016) Africa’s Growth Dividend? Lived poverty drops across much of the continent. Cape Town: Afrobarometer.

24.

Mberu

Ciera

Elungata

et al . (2014) Patterns and determinants of poverty transitions among poor urban households in Nairobi, Kenya. African Development Review 26(1): 172–185.

25.

Meyer

Keyser

(2016) Validation and testing of the Lived Poverty Index Scale (LPI) in a poor South African community. Social Indicators Research 129(1): 147–159.

26.

Ogwumike

Aboderin

(2005) Exploring the links between old age and poverty in anglophone West Africa: Evidence from Nigeria and Ghana. Generations Review 15(2): 7–15.

27.

Shimeles

Verdier-Chouchane

(2016) The key role of education in reducing poverty in South Sudan. African Development Review 28(S2): 162–176.

28.

Shriver Center (2016) Older women and poverty. WomanView 19(9). Avaiable from http://www.ncdsv.org/SSNCPL_Woman-View-Older-Women-and-Poverty_3-30-2016.pdf.

29.

UN Women (2015) Progress of the World’s Women 2015–2016: Transforming economies, realizing rights. New York, NY: United Nations Entity for Gender Equality and the Empowerment of Women.

30.

UNDP (2016) Human Development Report 2016: Human development for everyone. New York, NY: United Nations Development Programme.

31.

United Nations (2015) The Millennium Development Goals Report 2015. New York, NY: United Nations.

32.

Vijaya

Lahoti

Swaminathan

(2014) Moving from the household to the individual: Multidimensional poverty analysis. World Development 59: 70–81.

33.

Wagstaff

(1983) Attitudes to poverty, the protestant ethic, and political affiliation: A preliminary investigation. Social Behavior and Personality 11(1): 45–47.