About This Research

What is Night Time Light (NTL) data?

Night Time Light data comes from the VIIRS Black Marble product (VNP46A4), a NASA satellite sensor that records the brightness of the Earth's surface at night at approximately 500-meter resolution. Each annual composite image averages cloud-free nighttime observations across an entire year, producing a consistent measure of artificial light emissions.

Why does it correlate with poverty?

Economic activity produces light. Municipalities with more commercial and industrial activity tend to be brighter at night, while rural or economically marginalized areas remain dark. Researchers use NTL intensity as a proxy for economic development, allowing comparisons across regions and over time without requiring detailed survey data.

Why Mexico?

Mexico's 2,469 municipalities span an exceptional range of urbanization — from Mexico City, one of the world's largest urban centers, to remote rural communities in Oaxaca and Chiapas. CONEVAL, the national poverty evaluation agency, provides standardized poverty measurements for all municipalities, making Mexico an ideal case for testing whether NTL data reliably tracks economic conditions across such diverse settings.

Data Corrections

The original analysis contained three errors in how night-light values were calculated. Each error inflated or distorted the results. Here is what changed and by how much.

BUG-001: Impossible Outlier Values

A normalization formula divided by the wrong number, making one municipality appear 1,252 standard deviations above average — a value that cannot exist in any real measurement.

Corrected range: −3 to +3 (standard distribution) vs. buggy range: −0.76 to +1,252

BUG-002: Undefined Values from Zero Light

Municipalities that recorded zero night light produced undefined results (−∞) when logarithms were applied. The corrected pipeline filters these before calculation.

Affected municipalities: 2 had null or −∞ log values in buggy data

BUG-003: Double-Averaged Values

The R script averaged pixel values twice — once to summarize each municipality, then again. This deflated values for most places and inflated others, reshuffling national rankings.

Mean NTL (log scale): 6.18 (buggy) → 1.33 (corrected)
Example rank changes after correction (out of 2,469 municipalities):
MunicipalityBuggy RankCorrected RankChange
Melchor Ocampo#73#2,400−2,327
Soto la Marina#29#2,350−2,321
Tijuana#2,378#117+2,261
Aguascalientes#2,419#159+2,260
San Francisco de los Romo#2,411#165+2,246