Where the Signals Point Now

In the same fortnight, Kenya can publish an inflation reading, a central-bank rate decision and an Ebola case count. The numbers do not answer the same question. One describes the price level households have already faced; another describes the monetary setting; the third tracks a disease outbreak through confirmed cases and deaths. Read together, they offer context. They do not become one forecast simply because they appear on the same page.

As of the latest available Central Bank of Kenya table on 24 October, annual inflation was 6.76 percent in September, up from 6.59 percent in August. The CBR remained at 8.75 percent in the most recent listed policy decisions. Separately, the World Health Organization’s 23 October update reported 9,704 confirmed Bundibugyo Ebola cases and 4,687 deaths in the Democratic Republic of the Congo through 20 October; it also reported nine cases in Kenya, including one death. These are dated measurements with different denominators and reporting systems. CBK’s inflation series, its CBR history and WHO’s outbreak update provide the underlying records.

The task is to decide what each signal can support, what it cannot, and how much it should change a prior view. That discipline matters more than the number of indicators on a dashboard.

Oil is a price, not a diagnosis

An oil price can move before fuel arrives at a Kenyan depot. The EIA’s Brent Spot FOB series records one benchmark for physical crude, while futures contracts are different instruments tied to specified delivery months. The latest daily observations in the EIA series describe the market through 20 October 2026. A reader should state which series, date and instrument is being used before claiming that “Brent rose” or “settled.”

Even a verified rise in Brent is not a direct forecast of the retail price in Nairobi. EPRA’s formula for regulated pump prices includes landed product cost, handling, storage, transport, financing, margins and taxes. It uses an import-cost window that runs from the 10th of the previous month to the 9th of the pricing month; regulated prices are published monthly. EPRA’s pricing formula shows why a crude-price move can arrive late, partly offset or magnified by other inputs.

The exchange rate is part of that chain. A dollar-priced cargo must be converted into Kenyan shillings, and shipping and insurance are priced separately from the crude benchmark. The CBK notes that exchange-rate depreciation can contribute to inflation in a net-importing economy, while energy costs can affect transport and other consumer prices. That is a mechanism, not a promise that every rise in Brent appears at a pump in the same proportion or on the same date. CBK’s monetary-policy explainer describes the currency and energy channels.

Inflation remembers more than one shock

The September 2026 inflation reading is a signal about prices across the consumer basket, not a tally of costs caused by oil. In its June Monetary Policy Statement, CBK reported that non-core inflation had risen to 15.1 percent in June, while headline inflation was 6.4 percent; its later statistical series records headline inflation at 6.76 percent in September. Those categories move differently. Food, fuel, exchange rates, transport and services can all contribute to the measured CPI. A headline number cannot identify the cause without the category detail. The CBK June statement and the monthly series show the distinction.

Rates belong to a different time scale. The CBK rate history shows 8.75 percent at its April, June, August and October decisions. A rate hold does not say inflation has vanished; it says policymakers chose not to change the instrument at those meetings. Nor can one infer from the hold that a hike is imminent or impossible. A monetary committee weighs the inflation outlook, exchange rate, credit and output conditions, among other inputs.

Forecasts should preserve this separation. A reader may form a view that high energy prices could keep inflation elevated, but that is a conditional claim. It depends on the size and duration of the supply shock, currency movements, domestic prices and policy responses. The measured September rate is an observation; “oil will keep inflation high” is a forecast and should be labelled as such.

An outbreak count is not a transmission map

The health signal has its own denominator problems. WHO’s 23 October report says DRC had recorded 9,704 confirmed cases and 4,687 deaths through 20 October. It separately reports nine Kenyan cases, including one death; eight cases after the first imported case were known contacts, and WHO said there was no evidence of wider community transmission at that cutoff. These are confirmed-case counts at a stated date. They do not count every infection that may have gone undetected, and the Kenyan finding should not be generalized to the DRC or to a later date.

WHO’s report describes the disease as caused by Bundibugyo virus, a species of Ebola virus. It says symptoms can resemble other febrile illnesses and that confirmation requires laboratory testing. The response therefore depends on timely diagnosis, isolation and care, contact tracing, safe burials and community engagement. WHO’s 23 October Disease Outbreak News explains both the case totals and the response requirements.

This distinction matters when several indicators are placed side by side. A rising count can reflect new transmission, an expansion in testing, delayed sample results, or some combination. A falling count can mean improvement, but it can also reflect missing reports. WHO’s July situation report explicitly warned that newly reported cases might include backlogged samples and therefore might not be newly acquired infections. That is why a case series is stronger when read with testing, location, onset and contact-follow-up data, rather than by its headline total alone. The July WHO report records that caveat.

The danger of counting the same news twice

When signals seem to reinforce each other, ask whether they are independent. A geopolitical event may affect crude, shipping insurance, exchange rates and pump prices. Those are several observations, but they may all be downstream of one underlying shock. Treating them as four independent pieces of evidence can exaggerate confidence. The same is true when headline inflation, fuel inflation and transport inflation partly reflect the same increase in imported energy.

An analyst should draw the causal chain before multiplying probabilities. First comes the event or disruption. It changes expectations about supply and freight. Those changes may reach the landed cost of petroleum products, then pump prices, then transport and some consumer prices. The lags and pass-through rates vary. A second indicator is useful when it offers genuinely new information—for example, verified cargo delays alongside prices—not simply because it is another number affected by the same news.

The countercase also belongs on the page. A higher global oil price can be offset locally by a stronger shilling, inventory contracts or a lower price in the import window used for the next EPRA review. Inflation can ease if food or other categories fall even while energy costs remain high. An outbreak may expand in one province while the number of active zones declines elsewhere. These are not contradictions; they are evidence that the underlying systems have more than one moving part.

What calibration can and cannot say

Forecast probabilities become useful when outcomes and deadlines are defined in advance. A binary forecast can be scored with the Brier score, the squared difference between a probability and the eventual outcome. If an event was assigned 70 percent and occurs, the score is (0.70 − 1)² = 0.09. If it does not occur, the score is (0.70 − 0)² = 0.49. Lower average scores are better, but one forecast says little about skill. Over many comparable forecasts, calibration asks whether events assigned a probability near 70 percent happen about seven times in ten.

A constant 50 percent forecast scores 0.25 on every binary outcome. That is a handy arithmetic reference, not always the right benchmark. If an event usually happens 90 percent of the time, a useful comparison is the historical base rate, not a coin toss. Forecast quality also depends on a well-defined event, consistent deadlines and enough cases to evaluate. A tidy score cannot repair a badly chosen question.

The same principle applies to present-day monitoring. State the measure and its date, keep the underlying source available, identify which other indicators share the same cause, and say what evidence would change the view. Separate confirmed cases from suspected cases, price observations from retail prices, and current measurements from future scenarios. That makes it possible to update a forecast without pretending the earlier one was certain.

The signals do not point in one direction just because they are simultaneous. Some reveal current strain; others show how a shock may travel; still others are incomplete records of events that began earlier. The work is to keep those meanings intact. Numbers are useful when they narrow uncertainty. They become theatre when the analyst makes them say more than they can.

Flooding near Garissa, Kenya, seen by Copernicus Sentinel-2 on 29 April 2024. Historical satellite image illustrating climate exposure in Kenya, not an image of October 2026 conditions or the health and price indicators discussed in the article. Credit: Contains modified Copernicus Sentinel-2 data 2024 / European Union; image published 3 May 2024. Image source · Copernicus Sentinel data terms; free reuse with source attribution and a statement that the EU does not endorse the user’s activity