05 · Method
Methodology
We didn't model anything. We took a published curve, real temperatures, and official prices. Then we counted. Here's exactly how.
General principles
The analysis rests on two independent components, calculated separately, with no double counting. The first estimates income lost by heat-exposed workers. The second estimates the electricity surcharge for air-conditioned commercial spaces. These two populations are distinct by construction: a worker in an air-conditioned space is not heat-exposed.
Each component uses public data and declared assumptions. When a value is uncertain, we test three scenarios (low, central, high) to show the range. No parameter is tuned to achieve a predetermined result.
Analysis philosophy
This work is not an economic model. It's a counting exercise. We take an established physical relationship (the ILO/ISO curve), real weather data, and observable prices. The only "model" is the temperature-capacity curve, and we didn't invent it — the ILO and ISO published it after decades of research.
Component 1: Worker income loss
Temperature data
Hourly data for Nouakchott (lat 18.0858, lon −15.9785), period 2019–2025. Temperature at 2m, relative humidity, wind speed — the three components needed for WBGT calculation.
WBGT conversion
WBGT (Wet Bulb Globe Temperature) is the heat stress index used by ISO 7243. It combines dry temperature, humidity, and solar radiation. We use the simplified Liljegren (2008) formula to convert weather data to hourly WBGT.
Capacity curve
Work capacity curve at moderate intensity (200–300W metabolic). Below 26°C WBGT: 100% capacity. Above: measurable decline per degree.
Working hours
Covers typical working hours for Nouakchott's markets and shops. Friday is worked (half-day in some cases), Sunday excluded. This gives approximately 2,600 working hours per year.
Reference wage
Used as the base unit. The central scenario applies a 1.4× multiplier to reflect typical informal income.
For each hour h in [08:00, 18:00]:
capacity(h) = f_ILO(WBGT(h)) # between 0 and 1
loss(h) = 1 − capacity(h)
Hours lost/year = Σ loss(h) # ≈ 730 h
Cost/worker = hours_lost × (annual_wage / working_hours)
= 730 × (1.4 × 36,000 / 2,600)
= 14,742 MRU
City cost = 14,742 × 120,000 workers
= 1,769 M MRU
Exposed population
Informal and semi-formal economy workers exposed to heat: merchants, artisans, drivers, masons, street vendors. Excludes air-conditioned office workers and government administration. Conservative — Nouakchott's working population exceeds 300,000.
Component 2: AC premium
JICA (2018), Nouakchott Urban Master Plan, Table I-14. Includes shops, offices, workshops, services.
Declared local estimate. No official source available. Central (1 in 3) is consistent with observation of the commercial fabric.
Effective rate for professional consumers, non-subsidised.
Equipped spaces = 94,296 × penetration_rate
= 94,296 × 1/3 = 31,432 (central)
Monthly consumption per space ≈ f(month average temperature)
→ ranges from ~150 kWh/month (January) to ~450 kWh/month (October)
Monthly cost = spaces × monthly_consumption × 5.903 MRU/kWh
Annual cost = Σ monthly_cost (12 months)
= 869 M MRU (central)
Limitations and what we don't count
Any honest analysis must name what it doesn't capture. Here are ours.
Not counted — Health impact
Heat-related hospitalisations, mortality, chronic effects on exposed workers' health. These costs are real but require epidemiological data we don't have.
Not counted — Power outages
SOMELEC load shedding reduces the effective benefit of air conditioning. A shop experiencing 3 hours of outages per day in October doesn't fully benefit from its investment. We lack reliable data on outage frequency and duration.
Not counted — Degraded formal sector
Poorly air-conditioned offices — those whose AC can't handle the thermal load, or that suffer outages. These workers are neither "exposed" nor "protected" in our framework.
Not counted — Peri-urban agriculture
Crop yield losses in Nouakchott's market garden belt. Heat stress affects plants as much as workers, but that's a different calculation.
Each of these exclusions pulls in the same direction: the true cost of heat is likely higher than our estimates. We prefer a conservative, defensible number to an exhaustive, speculative one.
We didn't model anything. We observed the city, and counted.