Beyond Employment

What does being employed tell us about economic security in Kenya?

Lesley Maore · Portfolio case study · September 2026

Employment is a useful starting point. It tells us something about participation in economic activity. It does not, by itself, establish whether a household can meet its needs, whether a worker wants more paid hours, or whether a job provides protection.

I built Beyond Employment to examine those distinctions using real Kenyan public data. The project combines a historical household-survey analysis with a separate view of more recent lived experience. It does not merge different surveys into an invented population or treat different years as a single current picture of Kenya.

The strongest finding

In KNBS's 2015/16 Kenya Integrated Household Budget Survey, 29.7% of employed people aged 15–64 lived in households below the supplied overall consumption-poverty line. The approximate design-based 95% interval is 28.4%–30.9%.

This is household consumption poverty. It is not a statement that 29.7% earned below a particular wage, and it is not Kenya's current working-poverty rate. Employment follows the survey's native historical rules, including farming, unpaid family work, apprenticeship and volunteering. The finding nonetheless makes an important distinction: being counted as employed and living above a household poverty threshold are different outcomes.

Data and analytical choices

The KIHBS release contains 92,846 members, 21,773 households and a consumption aggregate covering those same households. I checked that household keys were unique, all members linked without row expansion, and household weights agreed across files. Household weights are applied once to each member; applying population weights to those same member rows would incorrectly multiply household size again.

I reconstructed employment from the questionnaire's work screens and return-to-work routing. It reproduced 26 published figures: employment rates and employed totals for ten age groups, the whole 15–64 population, and rural and urban residents. A simpler activity-response proxy misclassified one person. Another 29 working-age respondents had missing work screens and were kept unclassified in the poverty comparison.

The consumption calculation independently reproduced national overall poverty of 36.1% and food poverty of 32.0%. The weighted working-age population also matched the labour report. These benchmarks test specific calculations; they do not certify every variable in the release.

Uncertainty uses the released 92 strata and 2,387 observed clusters in a with-replacement Taylor approximation. All observed clusters remain in domain calculations. The intervals are approximate, omit finite-population corrections and do not quantify measurement or classification error.

Differences within employment

Household poverty among employed people was 34.7% in rural areas and 22.0% in urban areas. It was 40.6% among contributing family workers, 30.0% among own-account workers and 27.7% among paid employees. These descriptive groups differ in other ways; the analysis does not identify the causal effect of a job type or place of residence.

In a separate job-quality view, 59.7% of eligible employed respondents with valid answers wanted more paid hours in their main job. Desire alone is not an official underemployment measure: the hours, availability and reference-period requirements must also be satisfied.

The questionnaire routes casual workers past benefits and contract questions. Among non-casual paid employees routed to those questions, 37.0% of valid respondents reported a written contract. Employer NSSF and NHIF contributions were reported by 30.8% and 34.0%, respectively. The historical scheme names are retained. These conditional percentages must not be presented as coverage rates for every employed Kenyan. Valid and excluded counts accompany every result in the dashboard.

A separate lens on lived experience

Afrobarometer Kenya 2024 interviewed 2,400 adult citizens. I verified its household-adjusted weight against the published food-frequency and job-status distributions, then cross-tabulated all five deprivation items by cash-income job status.

Among respondents reporting a full-time cash-income job, 44.0% said they or a family member had gone without enough food at least once in the preceding year. The estimate becomes 44.8% with the older enumeration-area weight. Current job status and past-year family hardship have different time references. This result cannot establish that deprivation happened while the respondent held the current job. It is also not comparable to KNBS monetary poverty as if both measured the same thing.

What the project deliberately leaves unresolved

KCHSP 2020 has valuable job-quality questions and 84,859 individual records, but neither downloaded file contains an identifiable survey weight. National estimates are withheld; the design information required for national estimates remains unresolved. COVID-era conditions would require explicit treatment even after weights are supplied.

Four occupation/industry fields listed in the KIHBS catalogue are absent from its member file. Earnings are withheld from the headline analysis while basic, expected, in-kind and casual-payment treatment and extreme values remain unresolved. Direct ILOSTAT downloads were unavailable; five ILO-origin modelled series obtained through World Bank WDI are labelled as that fallback, with missing years preserved.

The data cannot rank people's usefulness, fully measure unpaid care, or estimate productivity per worker. Those are limits of the measures, not evidence that unmeasured activity has no value.

Deliverables and reproducibility

The interactive dashboard separates historical household poverty, conditional job quality, recent lived experience and modelled time context. Python scripts and executed notebooks preserve the calculations; SQL independently checks the poverty and Afrobarometer point estimates. A saved five-page Power BI Service report complements the public dashboard and requires report permission. It uses aggregate inputs; no Desktop PBIX or scheduled refresh is claimed.

The conclusion is specific: employment status describes one dimension of economic participation. A fuller account needs resources, hours and protection alongside it.

Sources

KNBS, 2015–2016 Kenya Integrated Household Budget Survey, catalogue 13, Labour Force Basic Report Table 3.3 and sampling chapter; Basic Report on Well-being; member questionnaire printed pp. 11–15. KNBS, KCHSP 2020 annual release. Afrobarometer Data, Kenya, Round 10, 2024, release, May 2025 codebook and October 2024 summary. ILO-origin series via World Bank WDI. Exact file hashes, source links, calculations and qualifications are retained in the project.