Raising revenue is half the job. Spending it well is the other half. African public budgets routinely show meaningful real growth in nominal terms while delivering deteriorating outcomes — the gap is the spending side. This module is about how to evaluate whether a public-spending programme is worth the revenue it consumes.
Targeting vs universal provision
Two paradigms for distributive spending:
- Targeted — identify the beneficiary group (poor households, the elderly, orphans-and-vulnerable-children, women heads of household) and direct resources only to them. Higher per-beneficiary impact, lower fiscal cost
- Universal — provide the benefit to everyone in an eligible status category (every primary-school child, every retiree). Higher cost, lower per-beneficiary impact, less targeting error
The targeting-error trilemma
Targeting errors come in two forms: • Type I (exclusion error) — a deserving beneficiary is left out. The poorest household isn't enrolled because they live in a remote area or don't have ID documents • Type II (inclusion error) — an undeserving beneficiary is included. Better-off households game proxy means tests; political clienteles capture the programme The trilemma: you can minimise exclusion error (cast the net wide), or minimise inclusion error (verify eligibility tightly), or minimise administrative cost — but not all three simultaneously. Universal provision essentially accepts maximum inclusion error to drive exclusion error to zero.
Hunger Safety Net Programme — Kenya's flagship cash transfer
HSNP is a regular unconditional cash transfer to chronically poor households in the four poorest counties (Mandera, Wajir, Marsabit, Turkana). Scale: ~100,000 beneficiary households, KES 5,400 paid bi-monthly via M-Pesa. Evidence base: multiple impact evaluations (Oxford Policy Management; World Bank; J-PAL affiliates) showing:
- Food-consumption impact +13-22% depending on cohort and season
- Asset accumulation effects (livestock, household goods) modest but positive
- Multiplier effects in local economies — every KES 1 of HSNP transfer generates KES 1.30-1.80 of additional economic activity in the receiving community (cash-multiplier studies, Egger et al. 2022 for analogue programmes)
- Targeting performance: Type I exclusion error ~15% (better than developing-country benchmark of 25-30%); Type II inclusion error ~22% (typical of proxy-means tests in low-administrative-capacity settings)
Cost-benefit analysis (CBA)
CBA is the workhorse evaluation tool for public spending. Net present value formula:
Net present value (NPV) of a public-spending programme
NPV = Σₜ (Bₜ − Cₜ) / (1 + r)ᵗ Where: • Bₜ = total social benefit in period t (in monetary equivalent, after shadow pricing) • Cₜ = total social cost in period t (including marginal cost of public funds for the revenue raised) • r = the social discount rate (the rate at which the social planner discounts future benefits to compare with present costs) • t = time period (years, typically) NPV > 0 → the programme passes the CBA test. NPV ≤ 0 → fails.
The social discount rate
The single most contested parameter in CBA. The Ramsey rule provides one formal answer:
The Ramsey rule for the social discount rate
r = δ + η × g Where: • δ = the pure rate of time preference (society's intrinsic discount rate; how impatient the social planner is intrinsically) • η = the elasticity of marginal utility of consumption (how diminishing the value of additional consumption is as we get richer) • g = the expected real per-capita growth rate of consumption Intuition: future generations will be richer (the g term) so a marginal shilling means less to them than to us, scaled by how curved the utility function is (η).
Standard parameter ranges (UK Green Book; US OMB; World Bank guidelines):
- δ — usually 0% to 1% per annum (almost-no pure impatience for a benevolent social planner)
- η — typically 1.0 to 1.5 (consistent with constant relative risk aversion in the 1-2 range)
- g — country- and time-specific. Kenyan per-capita growth has averaged ~2.5% over 2010-2025
- Resulting r — for Kenya, roughly 3-5% real
Shadow pricing
Market prices may not reflect social opportunity costs. Three frequent corrections:
- Labour — in a high-unemployment context, the social opportunity cost of using a labour hour may be well below the market wage. Public-works programmes (e.g., Productive Safety Net in Ethiopia) deliberately exploit this: hiring workers off the unemployment margin has a near-zero social cost
- Foreign exchange — if the official exchange rate is overvalued relative to the equilibrium rate, the social value of a dollar of foreign currency exceeds the official rate. Shadow exchange rate adjustments shift CBA in favour of export-generating and import-substituting projects
- Marginal cost of public funds (MCPF) — every shilling of public revenue requires more than one shilling of distortion costs (the deadweight loss from the tax). World Bank estimates of MCPF in developing economies typically range from 1.2 to 1.5. CBA should value costs at the MCPF-adjusted level, not at the nominal currency cost
The Marginal Value of Public Funds (MVPF)
A newer and increasingly used framework (Hendren and Sprung-Keyser 2020). For a given policy, the MVPF is:
MVPF formula and interpretation
MVPF = (Willingness to pay of beneficiaries for the policy) / (Net government cost of the policy) Where: • Willingness to pay (numerator) is what beneficiaries would have paid in dollars to have the policy enacted — captures both direct cash benefits and any in-kind value • Net government cost (denominator) is the gross programme cost minus any fiscal externalities — e.g., increased income tax revenue from better-educated beneficiaries MVPF > 1 → beneficiaries value the policy at more than it costs the government → welfare-improving for society at any positive social welfare weight on the beneficiary group MVPF = ∞ → the policy pays for itself fiscally (e.g., investments in young children's health that more than pay back through future tax revenue) MVPF < 1 → beneficiaries value at less than the government cost → policy must be defended on distributional grounds (the policy targets a low-income group whose welfare weight is >1, justifying the inefficiency)
Empirical MVPF estimates for major US programmes (Hendren-Sprung-Keyser 2020):
- Health insurance for poor children (Medicaid expansion to under-5s): MVPF ≈ 10 (because of long-run fiscal returns through improved adult outcomes)
- Job training for low-income adults: MVPF ≈ 1.5
- Tax cuts targeted to upper-income earners: MVPF ≈ 0.7
- K-12 education spending: MVPF ≈ 1.5-4 depending on target and design
Why MVPF beats simple CBA for some questions
Traditional CBA requires shadow pricing of all benefits (some of which — kids' welfare, dignity, fairness — are genuinely hard to monetise). MVPF asks the easier question: what would beneficiaries pay? Combined with information on the welfare weight society places on the beneficiary group, MVPF gives a transparent decision rule. The empirical literature now contains 130+ MVPF estimates for OECD policies; the equivalent for African policies is in early stages but growing (the J-PAL Africa archive is a useful starting point).
Programmatic case studies
Free Primary Education (Kenya, 2003)
President Kibaki's 2003 abolition of primary-school fees produced a 22% jump in primary enrolment in one year (Lucas and Mbiti 2012, based on KCPE administrative records). Per-pupil unit costs fell because of crowding (60-70 students per teacher in some classrooms). Learning outcomes (KCPE composite scores) declined modestly in the first three years before recovering as teacher hiring caught up.
Pure CBA on the basis of human-capital returns: positive NPV, probably MVPF in the 2-5 range. The transitional decline in learning outcomes is the cost of universal access without proportional investment in capacity; not a permanent feature.
M-Pesa subsidy through digital-payment promotion
Not a direct subsidy, but the regulatory and infrastructure investment that enabled M-Pesa (mobile-money licensing, USSD-channel pricing, agent-network supervision) had massive positive externalities. Suri and Jack (2016, Science) estimated that M-Pesa access lifted 196,000 Kenyan households out of poverty by 2016, primarily through resilience to negative income shocks. The fiscal-and-regulatory cost was modest; the MVPF on this set of interventions is effectively unbounded.
Universal Health Coverage (rolling out 2024-)
The Social Health Insurance Fund (SHIF), launched October 2024, mandates 2.75% of gross income for all formal-sector earners and a per-household contribution for non-salaried. Replaces NHIF. Targets universal coverage by 2027.
MVPF analysis is preliminary — the empirical literature on universal health coverage in similar contexts (Mexico Seguro Popular, Thailand UHC, Indonesia BPJS Kesehatan) suggests MVPFs in the 1.5-4 range when accounting for productivity and out-of-pocket-cost savings. The Kenya-specific challenge is the funding mechanism's salience (a visible new deduction on payslips) and the design controversy around contribution-vs-benefit cohorts.
Exercise
A Member of Parliament proposes a Universal Basic Income (UBI) of KES 3,000/month to every adult Kenyan above age 18. The proposal cites the J-PAL-affiliated GiveDirectly randomised UBI study (Banerjee et al. 2024) that found durable consumption, asset, and well-being gains. (1) Using rough Kenyan numbers (adult population ~28 million, exchange rate ~KES 130/USD), compute the annual fiscal cost. (2) Express the cost as a share of GDP (Kenya 2024 nominal GDP ~$110 bn). (3) Compare with current revenue, expenditure, and budget deficit. (4) Apply MVPF reasoning to evaluate the proposal. (5) Recommend a position with the public-finance argument.