Nighttime lights as a GDP proxy: how reliable is it for project appraisal?

Night-light data shows up in a lot of project appraisal memos these days, usually as a chart with a dotted trend line and a caveat buried in a footnote. The pitch is simple: satellites pick up visible light emitted at night, and more light generally means more economic activity, more grid connections, more commerce running past dark. For a portfolio economist waiting on census data that's three years stale, that's an appealing shortcut. The question worth asking before you put it in a results framework is how far the shortcut holds.

Where the GDP-proxy claim comes from

The idea has real academic roots. Economists have been testing nighttime lights against national accounts data since the early 2000s. The finding that holds up across most of that literature is that lights track growth rates reasonably well, while saying much less about the absolute level of GDP, in places where official statistics are thin or unreliable. Lights don't tell you a country's GDP is $40 billion. They tell you a region grew faster or slower than its neighbor over the last few years, which is a narrower and more modest claim.

That distinction sets the boundary on what you can write in an appraisal memo. Estimating the absolute size of an informal economy in a programme area needs more than a lights read; lights alone won't get you there. Testing whether a district that received a grid-extension loan three years ago is growing its commercial footprint faster than a comparable district that didn't is the kind of claim lights can support.

The reliability problems nobody puts in the footnote

A few things break the proxy in ways that matter for how you use it.

Urban cores saturate. Bright city centers hit the sensor's upper limit, so a district that's already well-lit can show flat or even declining light values while its economic activity keeps climbing. The sensor simply can't register more brightness past a certain point. This is a known issue with the older DMSP-OLS record in particular, and it means the proxy is weakest exactly where GDP is largest.

Blooming distorts footprints. Light from a dense settlement spreads into surrounding pixels, so a bright town can appear to electrify a wider radius than it covers. For appraisal work tracking grid extension into specific unserved settlements, this matters: you can mistake glow from the regional capital for progress in the village fifteen kilometers out.

Rural and low-density activity undersells. A newly electrified smallholder area running a few irrigation pumps and a maize mill at night produces far less light than the same dollar value of economic activity in a dense urban strip. Lights are better at catching commercial and industrial intensification than at catching the first wave of rural electrification, which is often exactly the stage a development finance project is trying to document.

Sensor changes break continuity. Comparing a baseline survey done with one satellite generation against a follow-up read from a newer sensor without adjustment will show a jump that has nothing to do with the project.

None of this means the proxy is useless. It means it's a trend indicator, best used the way the original research used it: comparing change over time within a consistent dataset, rather than comparing one snapshot's absolute brightness against another's.

What this means for project appraisal between surveys

For a results framework on a financed electrification project, the honest use case is narrower and more useful than "GDP proxy." A monthly read of light intensity over the settlements in your programme area tells you where the grid extension is showing up as visible activity and where it isn't, in the gap between the baseline survey and the three-year follow-up. It won't give you a GDP number for your implementation completion report. It will flag, this quarter, which clusters in the loan area still look dark months after the utility reported connections completed, which is exactly the kind of discrepancy worth a field visit before the formal survey catches it. Electrification Map runs that monthly read as a single panchromatic layer over your programme area, built for tracking progress on financed projects between the surveys that measure outcomes.

If you're weighing whether to add a luminosity layer to your monitoring plan, the honest starting point to discuss is what it's for.