Present bias is the single most-studied behavioural deviation. It explains why people simultaneously want to save and don't, want to exercise and don't, want to eat better and don't, want to start that business and never quite do. Understanding present bias and the commitment devices that counteract it is the most directly applicable behavioural insight for African financial services.
Standard exponential discounting
Exponential discount
U(consumption stream) = Σ_{t=0}^∞ δ^t × u(c_t) • δ ∈ (0, 1) = the discount factor per period • The discount RATE per period is r where δ = 1/(1+r) Key property: the discount factor between period t and period t+1 is δ — the same for all t. If you discount tomorrow at 95% of today, you also discount day 30 at 95% of day 29 and day 365 at 95% of day 364. Under exponential discounting, agents are TIME-CONSISTENT. If today you plan to save tomorrow's wages, tomorrow you will save them (no incentive to change the plan).
Quasi-hyperbolic discounting (β-δ)
Empirical observation: the gap between today and tomorrow feels larger than the gap between day 30 and day 31. People are more impatient about the immediate future than about the distant future.
β-δ model
U_today = u(c_0) + β × Σ_{t=1}^∞ δ^t × u(c_t) • δ ∈ (0, 1) = the long-run discount factor (typically 0.95-0.97) • β ∈ (0, 1) = the short-run impatience factor (typically 0.7-0.9) When β = 1, this collapses to exponential discounting (time-consistent). When β < 1, today is weighted DISPROPORTIONATELY relative to all future periods. The model is time-inconsistent — today's plan for tomorrow gets re-evaluated tomorrow because by then, tomorrow has become 'today.' Specifically: today you choose c_0 to maximise u(c_0) + β × δ × u(c_1) + β × δ² × u(c_2) + ... Tomorrow you choose c_1 to maximise u(c_1) + β × δ × u(c_2) + ... The weight on c_1 changed from β × δ today to β × 1 tomorrow. Specifically, the relative weight of c_1 vs c_2 changed. So tomorrow's choice may differ from today's plan.
Sophisticated vs naive agents
How much do present-biased agents know about their own present bias?
- Naive agents — present-biased but don't realise it. They plan tomorrow's savings, then find themselves unable to save tomorrow, and don't update their model. Persistent surprise at their own behaviour
- Sophisticated agents — know they're present-biased. They take steps in advance to counteract their future selves' impatience. They use commitment devices
- Partially sophisticated — somewhere in between. Most people, empirically
Why sophistication matters for product design
Sophisticated present-biased agents have DEMAND for commitment products. They KNOW they will overspend tomorrow; they want a device that prevents them. Naive present-biased agents DON'T demand commitment products. They believe (wrongly) that they will save tomorrow without help. They reject products that constrain them, missing the welfare gain. Product design implication: introducing commitment products to a naive population needs to FIRST raise awareness of the present-bias problem (educational marketing) and THEN offer the commitment solution. Marketing for commitment-savings products has heavy 'help yourself stick to your plan' messaging — this is the awareness-raising step.
Commitment devices in African financial markets
Behavioral-economics research over 15+ years has documented strong demand for commitment products in African contexts:
- Chamas and ROSCAs — rotating savings groups where contributions are visibly committed at predetermined intervals; the contribution cycle is a commitment device. Empirically robust evidence of widespread use across Kenya, Tanzania, Uganda, Rwanda
- SACCOs — mandatory monthly contributions tied to loan eligibility. The mandatory feature explicitly limits the member's ability to skip savings
- Mobile-money lock products — M-Shwari Lock Savings (Safaricom + NCBA Bank); Stawi accounts; KCB-M-PESA Save. Customer agrees that funds are inaccessible for a defined period
- Christmas-club savings — banks offer; customer cannot withdraw before December 1; popular for school-fees savings tied to the school-year cycle
- Standing-order arrangements — automatic deduction from current account into savings/SACCO/pension. The automation prevents the customer from 'choosing' to save each month
- Layby and credit purchases — buying goods on installment with payments fixed; harder to spend the cash elsewhere because it's been pre-committed
Field-experimental evidence
Multiple RCTs have measured the effectiveness of commitment-savings products:
- Ashraf-Karlan-Yin (2006) Philippines — RCT of commitment savings account. Customers who took up the product saved 81% more 12 months later than the control group. Strong demand for commitment among the subset of savers who were aware of their own present bias
- Brune-Giné-Goldberg-Yang (2016) Malawi — RCT of commitment savings tied to harvest income. Treatment-group farmers saved 60% more, used the savings for input purchases, and increased subsequent harvest by 22%
- Dupas-Robinson (2013) Kenya — RCT of providing simple money lockboxes to small business owners in Western Kenya. Treatment group saved 66% more for health and business investment over six months
- Karlan-Mullainathan-Roth (2019) — meta-analysis of 22 commitment-savings RCTs across 9 countries. Take-up rates 20-40%; average savings increase 50-90% among those who took up
Commitment products and welfare
An important nuance: not all 'commitment savings' is welfare-improving for the customer. • Some customers face genuine emergencies and need flexibility. A lock product becomes harmful if a child needs medical care and the funds are inaccessible • Some commitment products are designed for the provider's benefit (high fees on early withdrawal, hidden penalties). These extract rent from the customer rather than helping them • The welfare gain from commitment depends on the customer's underlying time inconsistency — agents without present bias don't gain from commitment; they lose flexibility • Behavioural-paternalism critics argue: who decides whether the customer NEEDS commitment? If we agree the customer should choose, we agree the commitment device should be voluntary The robust design: voluntary commitment products with reasonable but not punitive early-access provisions. Customer is free to choose the commitment level but not punished for genuine emergency access. M-Shwari Lock Savings broadly follows this model — funds are locked but the lock period is customer-chosen (1, 3, 6, or 12 months) and early-access fees are non-prohibitive.
Wage-payment frequency and self-control
Field experiments suggest payment frequency matters for saving behaviour:
- Weekly vs monthly pay — workers paid more frequently (weekly) tend to save less overall, because each pay-period decision is small and present bias dominates each one. Monthly pay forces a bigger one-shot decision
- Lump-sum vs streaming benefits — recipients of one-time lump-sum payments (cash transfers, tax refunds) often have higher savings rates than recipients of equivalent monthly streams. Counterintuitive but consistent with mental-accounting and present-bias models
- Bonus vs continuous pay — year-end bonuses are saved at higher rates than equivalent salary distributed across the year
Policy implications
- Pension and retirement design — auto-enrolment with default contribution rate captures most workers. Increasing the default rate increases saving without forcing it. The Kenyan NSSF design and many SACCO arrangements take this approach implicitly
- Cash-transfer timing — for the recipients, the timing of cash transfers (single lump sum, periodic stream) affects accumulation outcomes. HSNP's bi-monthly model is intermediate — large enough payments to support periodic asset purchases but not so frequent as to dissipate via routine spending
- Borrower protection — interest-rate caps don't address present bias (which is on the borrower side). Better protection: clear disclosure (engages System 2), cooling-off periods for high-rate loans (creates a delay that reduces present-bias-driven decisions), repayment-history-based credit scoring (rewards demonstrated discipline)
- Default-option design — making the welfare-improving choice (save more, contribute to pension, enrol in health insurance) the default captures status-quo bias and present bias simultaneously
Exercise
A Kenyan SACCO has 50,000 members. Mandatory savings: KES 1,000/month. Members can voluntarily save more but few do (average voluntary savings: KES 400/month, equivalent to 40% of mandatory). The SACCO board is considering a new product to address present bias: (A) Voluntary commit-to-save where members commit to extra savings at specific milestones (year-end, school-fees season), with no penalty for missing. (B) Automatic save-the-change: every transaction rounds up to next KES 100; the change is auto-deposited. Optional opt-out. (C) Linked savings goal: member sets a goal (school fees, dowry, asset purchase); automatic monthly contribution toward it from current account. (1) Which is most likely to increase savings rates given what we know about commitment-product demand? (2) Estimate the savings-gain magnitudes (rough order-of-magnitude). (3) What are the implementation challenges for each? (4) Should the SACCO offer all three or focus on one? Justify.