Nowcasting emerging-market growth with three monthly pulses

Quantamentry — how electricity, ports, and factories update the growth score between GDP prints.

Why does annual GDP data fail for EM monitoring?

Annual GDP is the wrong instrument for watching an emerging market in real time because it arrives far too late and far too smoothed to catch a turn while it still matters. An emerging-market annual GDP print typically lands one to three quarters after the period it describes, and it then gets revised for years afterward. By the time the number confirms that growth rolled over, the currency has already moved, the central bank has already reacted, and the position you would have taken on the signal is long gone.

The fix is not a better GDP model. It is different data — high-frequency physical activity series that move now and stand in for growth until the accounts catch up. We built three of them into the growth dimension. Each is monthly, each is free public data, and together they form what we call the growth pulse.

The intuition behind the pulse is old and well-tested: when you cannot measure output directly and on time, measure the physical inputs and throughput that output requires. Central banks and statistical agencies have nowcast GDP from electricity, freight, and factory data for decades. What is new is that the raw series are now free, public, and clean enough to run daily against 100+ countries — so a nowcasting approach that used to require a subscription desk can be built on open data and applied consistently across the whole panel, gold tier and frontier alike.

The three pulses

Each pulse is a year-on-year change in a physical measure of activity — chosen because physical throughput is hard to fake, hard to revise heavily, and available months before GDP.

1. Electricity generation year-on-year (Ember). Electricity is the closest thing macro has to a real-time metabolic rate. Factories, offices, and data centres all run on it, so monthly generation tracks the pace of activity with very little lag. We take the year-on-year change in generation to strip out the strong seasonal cycle.

2. Port-trade momentum (IMF PortWatch / Kpler AIS). We take the trailing-30-day mean of port-call tonnage, anchored at the latest observation, and compare it year-on-year. This is a direct read on the volume of goods physically moving through a country's ports — an external-demand and trade-cycle pulse that updates daily with only a few days' lag.

3. Industrial production year-on-year (Eurostat). For the European geographies Eurostat covers — including Türkiye, Ukraine, and the CEE bloc — monthly industrial-production volume is the classic activity series, and its year-on-year change is the third pulse. Of the three it is the one most directly designed to proxy output, which is why we lean on it wherever Eurostat publishes it.

The three are chosen to be complementary rather than redundant. Electricity captures broad economic metabolism including services and households; port tonnage captures the traded-goods cycle and external demand; industrial production captures the factory sector specifically. A country can have soft factories but resilient trade, or falling electricity demand with steady ports — and because the three pulses view activity from different angles, a divergence among them is informative rather than a defect. We do not force them to agree; we average them and let the spread tell its own story.

How does the blend work?

Each raw pulse is a percentage year-on-year change, and we map it to a 0–100 sub-score with a single, deliberately simple rule:

pulse_score = 50 + clamp(yoy, -20, +20) × 2.5

So a flat pulse (0% YoY) scores 50. A pulse growing at its +20% cap scores 100; one contracting at the −20% floor scores 0. The clamp keeps a single wild print — a redenomination artefact, a base-effect spike — from dominating. The mapping is intentionally linear and legible: no hidden weighting inside a pulse.

The three pulse scores are then averaged and blended into the growth dimension:

final = 0.7 × base + 0.3 × mean(pulse scores)

where base is the existing structural growth score (GDP percentile, WEO deviation, unemployment). The pulses get 30% — enough to move the reading when activity turns, not enough to let one noisy month swamp the fundamentals. Crucially, the blend renormalises when a pulse is missing: a country with electricity and ports but no Eurostat IP is scored on the mean of the two it has, not penalised for the one it lacks. Coverage differs by country, and the formula respects that instead of pretending to uniform data.

A worked example: Türkiye

Take Türkiye in the growth score computed for July 23, 2026. Its three pulses read:

Run the mapping and the numbers fall out directly: electricity at −5.9% gives 50 + (−5.9 × 2.5) = 35.3; industrial production at −0.27% gives ≈ 49.3; trade momentum at +1.6% gives ≈ 54.1. The pulse mean is 46.2 — a hair below the neutral 50, telling you Turkish activity was running slightly soft in the window, dragged mostly by the contraction in electricity generation while trade held up.

That 46.2 then blends into the structural base at the 0.3 weight to produce the growth sub-score for the July 23, 2026 reading. The virtue of the worked example is that you can check every step — the raw pulses, the mapping, the mean, the blend — against the published number. No black box.

It also shows why the multi-pulse design matters. If we had only the electricity pulse, Türkiye's activity read would have been 35.3 — a fairly gloomy number — and the growth score would have leaned hard on a single soft series. Bringing in industrial production (49.3, roughly flat) and trade momentum (54.1, slightly expanding) pulls the picture back toward neutral, because the three series disagree and the average is the honest summary of that disagreement. No single physical proxy is a reliable stand-in for growth on its own; the point of blending three is that their idiosyncratic noise partly cancels while their common signal survives. When electricity, ports, and factories all turn together, the pulse mean moves decisively — and that agreement is exactly the moment worth paying attention to.

Honest limits

The pulses are a genuine improvement, but they are not GDP, and a few boundaries keep the claim honest.

• Nightlights are deliberately excluded from the pulses. We ingest monthly satellite nightlights (World Bank via NASA VIIRS) for ~170 countries, and they are a lovely activity map — but the product publishes with roughly a seven-month lag. That disqualifies it from anything called a "pulse." We use nightlights as a structural cross-country read only, never as a nowcast, because a seven-month-old signal dressed up as timely would be exactly the overclaim this platform avoids.
• Retail trade is ingested but not scored. Eurostat retail-trade volume is in the warehouse, but it is not yet wired into the pulse blend. When it is, it will be a fourth pulse; today it is reference data, and we would rather say so than quietly imply it is in the model.
• Electricity demand is not GDP. Generation tracks activity, but the relationship drifts with the energy-intensity of an economy, with weather, and with structural shifts toward services. A pulse is a proxy, weighted at 30% precisely because it is a proxy. Treat a moving pulse as "activity is turning," not "GDP is now X."
• Coverage is uneven. Not every country has all three pulses; some bronze-tier economies have none, and their growth score stays structural-only. The renormalisation handles this gracefully, but it means the sharpness of the nowcast varies by country.

The full mechanism — every pulse, the mapping, the weights, and how the growth dimension folds into the seven-dimension composite — lives on the methodology page.

To watch the growth pulse move on the countries you care about, Create a free account — watch 3 countries.

Sources: Electricity generation — Source: Ember (CC BY 4.0). Industrial production and retail — © European Union, Eurostat, CC BY 4.0. Port-trade momentum — Sources: Kpler; UN Global Platform; IMF PortWatch (portwatch.imf.org).

— Quantamentry, July 24, 2026