Ethnic and political dynamics of poverty, employment, and wages in Kenya

Magero, Joshua

Ethnic and political dynamics of poverty, employment, and wages in Kenya

Thesis / Dissertation

2024

Publisher

University of Cape Town

Department

School of Economics

Faculty

Faculty of Commerce

Abstract

The association of ethno-political bias and economic development outcomes remains central to policy considerations in most African countries, especially in the sub-Saharan region. Perceptions and realities of ethnic and political favoritism in wellbeing have been a major cause of ethnic conflicts, erosion of civic participation, and the undermining of an inclusive national identity and other aspects of social cohesion in Kenya. In turn, this has more broadly affected her socio-political and economic stability adversely. Although it is widely accepted that observed economic wellbeing differentials can be attributed to a myriad of factors spanning geographical climate, economic factors, socio-cultural, and colonial historical impacts, it has been argued equally that the observed disparities in Kenya might have been shaped by the continued entrenchment of discriminatory policies by the successive post-independence regimes. The general perception is that political leaders, particularly those in the executive arm of the government, have tended to manipulate resource distribution in favor of their ethnic groups and/or counties, and this has been enabled by the spatial concentration of the ethnic groups across spatial units in a nonrandom manner. These claims of ethnic and political favoritism, however, largely only abound as anecdotes as prior work in this field remains inadequate. A few pieces of evidence contend that there exists a consistent systematic bias in favor of ethnic groups and Presidents' regions. This thesis corroborates this anecdotal evidence by means of empirical investigation. The thesis seeks to identify the underpinnings of the relationship between ethnopolitics and resource distribution in Kenya. We do this by mapping ethnically based resource allocation over time in Kenya, from the 1990s through to the present, in three interrelated but independent papers: (1) Characterize the trends and patterns of multidimensional poverty levels; (2) Examine the relationship between coethnicity with the President and poverty levels and establish whether governance reforms in the post-2010 period has reduced the ethnic bias in resource distribution, and (3) Investigate the role played by ethnicity and the political regime in power at the time of entry into the labour market on present-day employment and earnings. Essentially, the thesis seeks to identify empirically the political underpinnings of poverty levels and the labor market over time. We introduce the thesis in chapter one, proving the background and motivation for the thesis. In chapter two, we derive reasonably comparable datasets from interrelated independent large surveys; three rounds of 10% samples of the Kenya Population and Housing Censuses (1999, 2009 and 2019) and the Kenya Integrated Household Budget Surveys (KIHBS) 2016, and construct some of the key variables. This chapter pays special attention to the imputation of ethnic identity when explicit individual-level measures are absent. The ethnicity variable is not explicitly in some of the datasets. This discussion lays out the primary independent variable, ethnicity, in the later analyses. Chapter three discusses the elements and construction of the primary outcomes of interest – a multidimensional poverty measure. The chapter provides some descriptive and visual evidence of a positive correlation between Kikuyu ethnic identity and poverty measures as well as simple correlations between poverty measures and county-type. We construct the multidimensional poverty index (MPI) (from three domains: education, dwelling conditions, and access to key basic amenities-tap water, electricity, clean cooking fuel) and characterize its distributional patterns between 1999-2019. From the pseudo panel dataset of counties constructed, we can observe poverty experience over time and examine the extent to which poverty is sustained in specific regions. The following findings emerge from this chapter: poverty experience in Kenya is persistent rather than static, for most counties. We find a high degree of persistence of poverty status of counties: 82% of counties that were the least poor in 1999 have remained the least poor in 2019; 56% of those that were moderately poor in 1999 are still moderately poor in 2019, while 66% have remained trapped in poverty, being the poorest between 1999 and 2019. While counties in the Mount Kenya region have simultaneously exhibited the lowest poverty levels and best improvement in poverty reduction, counties in the Western have remained moderately poor or moved to the bottom of the poverty distribution. Notably, counties in Northern Kenya have remained “trapped” in poverty at the tail end of the distribution as the poorest over the 25-year period. Although poverty has declined gradually over time (the poverty rate dropped from 48.5% in 1999 to 42.5% in 2019), the cross-country disparities in poverty outcomes have persisted, exacerbating the inequality situation that already exists. Our results also reveal contemporaneous clustering of poor counties with specific ethnic groups. Counties that have the highest share of President's coethnics are the least poor and have the greatest reduction in poverty levels over time. Counties, where the Kikuyu tribe are the majority (Central Kenya and lower Eastern Kenya), have consistently simultaneously exhibited the lowest poverty levels and the most significant reduction in poverty over time. On the contrary, the inhabitants of Western Kenya (the Luo and the Luhya tribes) have remained moderately poor and the Northern Frontier Districts have stayed “trapped” in poverty (mostly the Kenyan Somali). The key contribution of this chapter is its consideration of the dynamic poverty phenomenon, as opposed to the transient nature where poverty is experienced only at a specific point in time. Chapter four attempts to deepen the institutional/governance mechanism discussed in chapter one, locating the source of these inequalities in politics using regression-based analysis. The main objective of this chapter is to investigate the drivers of poverty at the household level and county level, and the effect of a change of governance structure on the distribution of poverty across counties, and whether this transition has effectively redressed the ethnic-identity-based resource allocation problem that has bedeviled the country for a long time. We profile trends and patterns of material well-being in Kenya and decompose the county poverty gaps. We find suggestive evidence that coethnicity with the president matters for household poverty outcomes. Our estimation results show that the key drivers of household poverty are the ethnicity status of the household head, the level of education, engagement in gainful employment, location of the household, gender of the household head and the number of household members. The estimations reveal that being coethnic with the incumbent President is contemporaneously associated with a lower likelihood of being poor, and much more, staying coethnic major counties confers additional benefits to the coethnic households. More importantly, the result points to the fact that even though ethnic inequalities may exist at the individual level, they could be overridden by regional redistribution. The spatial dimension of poverty imbalances seems stronger than the ethnic dimension. A household is able to gain more welfare if they moved to other regions where presumably more resources are being channelled. The implication of this finding could be that removing barriers to movement of people, or rather, strengthening social cohesion amongst communities could help reduce the ethnic inequalities, thereby improving the overall welfare. Decomposition of the poverty gap reveals that the significant differences in poverty levels between coethnic majority counties and non-coethnic majority counties that cannot be solely attributed to differences in county characteristics which hints at potential “discrimination” or “favoritism” that benefits president's coethnics and regions where they are a majority. A significant portion of the poverty differences therefore remains unexplained. Further analyses suggest that despite the implementation of devolution, the importance of patronage networks continues to be witnessed, implying that character of ethnic patronage politics persists. We find no doubt that president's regions disproportionately benefit from public goods and services. The Kikuyu dominant counties more especially on average have substantially lower poverty levels, higher access to employment opportunities, better road networks, high school quality, adequate health infrastructure, proper water distribution and connection to electricity (resources which are largely publicly provided) although the reasons for such disproportionate allocations remain unclear. Although we are unable to estimate whether local politicians also influence the distribution of these resources, nor whether the outcomes would be different in politically contested areas, our evidence largely suggests that institutional reforms, through devolution in this case, has not effectively reduced practice of ethnic favoritism in the distribution of public goods. Instead, the patterns are more likely to be consistent with clientelist public investment behaviors motivated by national presidential election outcomes. The final empirical chapter of the thesis, chapter five, extends these bias-related concerns to a new arena – the general labor market – examining differences in labor market outcomes as a function of coethnicity before and after the new dispensation in 2010. We estimate the coethnic effects, and the regime of entry effects in the labour market outcomes-wages and employment; whether sharing the same ethnicity as the President today, or indeed, sharing ethnicity with the incumbent President when one was first eligible to enter the labor force, matters for contemporaneous employment and earnings. Due to the richness of data, most of the analyses in this chapter are primarily based on KIHBS 2015 data. However, evidence from the other datasets is also presented. We utilize the quasi-experimental technique; the regression discontinuity and probability score matching in these analyses. Our results reveal evidence of coethnic biases in different sectors and types of employment across both earnings and employment in the general labour market. Overall, we find that when Kalenjins are in power in 1999, they have higher chances of getting wage employment relative to non-Kalenjins. The same pattern is observed for the Kikuyu in 2009, and 2019, when they control the presidency. This finding suggests that coethnic biases continue to matter in the labour market, even in the post devolution, even though the magnitude has dampened. The coethnicity status at the regime of first entry into the labour market plays a significant role in today's employment and earnings. Sharing the same ethnicity with the ruling elite at the time one enters the labour market is associated with significantly higher employment probabilities today. While the coethnic effects especially for access to paid job seem manifest at the time of entry into the labour market, these effects dampen for the coethnics who have stayed longer in the job market. The recent coethnic cohorts seem to have higher employment probabilities than the earlier regime entrants. Hence, the biggest coethnic differentials are therefore pronounced at the time of entry but do not persist for long. This chapter also finds that the coethnic effects of employment are prevalent both in the private sector and the public sector. Different from earlier studies that do not find pronounced ethnic biases in employment in the Kenyan public sector, our findings reveal the existence of such biases across all regimes of entry. Our findings are consistent with the evidence from the administrative datasets derived from the audit of parastatals in Kenya, that find oversubscription of certain ethnic groups. We find results indicative of this evidence using the household survey dataset. Chapter Six of the thesis provides a summary of the thesis findings, policy implications and document areas for further research, highlighting the main findings and suggested policy implications and recommendations from the thesis. In conclusion this thesis has elucidated the pronounced tendencies of ethnic imbalances and their relationships with the political processes, to support policy formulation in redressing the longstanding challenges of poverty, equity, and inclusivity in Kenya. The findings from this thesis vitally contribute to the ongoing attempts to undo the legacy of ethnopolitical favoritism, not just in Kenya, but in sub-Saharan Africa more broadly, where these issues are widespread. We find that coethnicity with the President comes out as an important element in the conduct of social and political affairs in Kenya, with direct implications on poverty and the labour market. We also find persistence of poverty experience within counties. Most counties have remained in the same state over the study period. Particularly, counties dominated by the President's coethnics have remained relatively richer than other counties, besides having the greatest poverty reduction, while others have remained trapped in poverty. Our findings also reveal that transition to devolved system of governance from the central government structure has not significantly altered the resource distribution based on ethnopolitics, but the coethnic wage gap has reduced significantly in the post-2010 constitutional dispensation. Although it might be too early to comment on the effectiveness of devolution since Kenya is only at its initial stages, our evidence suggests that devolution has not reduced the ethnic favoritism in regional development. The persistence of ethnic favoritism in the post devolution era shows that the new constitutional dispensation and other legislations in pursuit of equitable access to resources and opportunities have not been effective in addressing the long-term challenges of county inequalities. Political exclusion and inequitable and unequal distribution of resources and opportunities have been longstanding issues in Kenya. This study, however, has been subject to limitations including data and the scope of coverage. A lot of potential further research in this area remain to be desired. With availability of data, over a long period of time, more accurate empirical investigations can be conducted to guide policy formulation on poverty reduction and narrowing the inequality. While we have attempted to control for all the possible empirical factors given our data-time trend, extent of urbanization, being Presidents coethnics, the proportion of the population with post-secondary schooling, and the proportion of the population with paid work, the empirical data may not convincingly rule out many other potential explanations. Some of these unobserved factors in our data may include better equipped bureaucracies, stronger representation in the national assembly, more effective county executives, among other factors, which we have not controlled for. Although we have convincingly provided evidence of persistence of ethnic imbalances beyond the governance reform period, future studies may consider adding additional empirical material that could support the assertions in this thesis. To exhaustively attribute the differences to political favoritism would require additional analysis using granular data with information on the conduct of bureaucrats-economic planners, resource allocation and electoral outcomes for instance. Another potential area of consideration could be whether political and ethnic favoritism in resource distribution play out differently in politically contested areas from the areas regarded as strongholds. Due to data limitation, we have estimated only the overall extent of favoritism, but not investigated the potential sources of these biases. However, assessing accountability channels by comparing different contexts (resources or index of resources in general) can confound other differences, posing an empirical challenge-especially when these projects are implemented in different phases. There is need to go beyond the canonical existence of political favoritism, and investigate further, the mechanisms through which the favoritism is implemented. Moreover, our study has only relied on the on- and off-election cycles (between political regimes). It might be useful to evaluate political favoritism using data that allows us to study how public goods provision tracks political developments at a high frequency. Investments in public projects for instance may accelerate in the weeks immediately before voting, and slow down significantly in the post-election period. Rather than assuming that coethnics continue to benefit from public goods provision throughout the Presidency of the incumbent leaders, mapping out the temporal process of ethnic and political favoritism gives more granular information which does not conceal within-period biases. This would contribute to understanding the timing of fiscal spending around the election cycle. Finally, one of the key limitations in this thesis is data limitation, especially with regards the ethnicity variable, which we inferred by from proxies. This might have an impact on the robustness of our result. Key policy implications arise from this thesis. First, there is need for significant tracking and scrutiny of government spending behaviour to curb the persistent ethnic and political biases. There is need for increased media coverage of distribution of public investments. This has an effect of increased accountability. Evidence shows that public investments which were hard to track and hidden from the public eye, especially within the implementation phases—exhibit significant favoritism, while stages which are disclosed publicly and receive widespread media coverage. Therefore, increasingly active independent media scrutiny as well as increasingly robust democratic institutions, expanding constraints on executive power, and donor oversight may partly curb favoritism. Rigorous democratic institutions and a free and transparent press can empower citizens to hold their elected officials accountable. International donor agencies should increasingly place strict conditions on the use of their funds to restrain corruption as strict donor conditionality on donor-funded projects. Evidence shows that oversight by international aid donors may help restrain favoritism in stages when the project is under the donor's scrutiny.

Keywords

economics

Reference:

Collections

PhD / Doctoral

Full item page