Do higher‐quality nighttime lights and net primary productivity predict subnational GDP in developing countries? Evidence from the Philippines

• Published date: 01 September 2022
• Author: Jesson A. Pagaduan
• Date: 01 September 2022
• DOI: http://doi.org/10.1111/asej.12278

ORIGINAL ARTICLE

Do higher-quality nighttime lights and net primary

productivity predict subnational GDP in developing

countries? Evidence from the Philippines

Jesson A. Pagaduan

Macroeconomics Division of the Economic Research and Regional Cooperation Department (ERCD), Asian

Development Bank (ADB), Metro Manila, Philippines

Correspondence

Jesson A. Pagaduan, Macroeconomics Division of the Economic Research and Regional Cooperation

Department (ERCD), Asian Development Bank (ADB), Metro Manila, Philippines.

Email: jpagaduan.consultant@adb.org

Abstract

Nighttime lights (NTL) data from satellites are a useful proxy for local economic

activity in developing countries where economic data are sparse. Yet most analyses

use the flawed DMSP NTL data, a poor proxy for GDP in less densely populated

and highly agricultural rural areas. In this article, we augment a novel NTL dataset

of the newer and better VIIRS NTL data with more ubiquitous remotely sensed

data, namely, net primary productivity (NPP) and land cover, and we test whether

these satellite data predict subnational GDP in both urban and rural sectors of the

Philippines. The results confirm that the higher-quality VIIRS NTL data predict

urban economic activity sufficiently well for both light-intense and dimly lit regions

but still do not explain rural economic activity very well. The use of croplands

NPP as an intensive measure of agricultural productivity, however, dramatically

improves the performance of land cover as a proxy. We demonstrate that remotely

sensed data can be useful in various applications, including evaluating the long-run

dynamics of province-level GDP growth, the local impact of natural disasters, and

the effects of infrastructure projects at the city and municipal levels. Such applica-

tions point toward the need for empirical analysis of growth at finer scales of

aggregation.

KEYWORDS

GDP, net primary productivity, nighttime lights, Philippines, VIIRS

JEL CLASSIFICATION

O10, B22, E01, R12

DOI: 10.1111/asej.12278

© 2022 East Asian Economic Association and John Wiley & Sons Australia, Ltd.

Received: 25 July 2021; Accepted: 25 January 2022

288 Asian Economic Journal. 2022;36:288–317.

wileyonlinelibrary.com/journal/asej

1|INTRODUCTION

Nighttime lights (NTL) data from satellites are a useful proxy for local eco-

nomic activity in developing countries where economic data are sparse.

Most nighttime economic activity requires light, hence higher intensity of light in a

given area implies a greater level of income, which is increasing in both income per

capita and number of people (Henderson et al., 2012). As a proxy, NTL predict

subnational gross domestic product (GDP) data, which are typically unavailable in

poor countries where there is weak government statistical infrastructure. These data

encompass a much wider geographic coverage and higher frequency—weekly or

even daily—compared to national income accounts, which often are released only

after some delay. In addition, NTL data are collected in a consistent and objective

manner by sensors aboard satellites and thus are not subject to government manip-

ulation (Donaldson & Storeygard, 2016).

Yet most analyses use the flawed Defense Meteorological Satellite Program

Operational Linescan System (DMSP) NTL data product (Gibson et al., 2020), a

poor proxy for GDP in less densely populated and highly agricultural rural areas

(Keola et al., 2015;Gibsonetal.,2021). The DMSP sensors suffer from blurring

and top-coding issues, which give rise to spatial inaccuracies, as well as from lack

of on-board calibration, not to mention inter-satellite differences that cause tem-

poral inconsistencies in the NTL series. Moreover, agricultural activities, mainly

concentrated in the rural sector, emit less observable lighting and hence are not

easily detectable by the sensors. In contrast, economic activities in urban clusters,

such as manufacturing and construction, services, retail, and transportation, are

more likely to be captured by NTL data (Chen & Nordhaus, 2019). Thus, using

DMSP data to proxy for local economic activity may introduce sizable measure-

ment errors, especially in the case of developing economies that often consist of,

on the one hand, relatively less densely populated and dimly lit areas at night,

and, on the other, a sizable share of agriculture in total GDP.

This article explores the potential of the higher-quality Suomi National Polar-

Orbiting Partnership Visible Infrared Imaging Radiometer Suite (VIIRS) NTL

data product, augmented with net primary productivity (NPP) and land cover, to

predict subnational GDP in both urban andrural sectors in a developing economy,

the Philippines, inthe period 2000–2018. The Philippines,an archipelagic country,

features great subnational heterogeneity and hence provides valuable cross-

sectional variation.

1

Although subnational GDP is available for the country’s

17 administrative regions, income accounts at the province level, within

which much of the interesting variation in economic growth takes place, are

unavailable.

2

VIIRS NTL, NPP, and land cover are of interest for two reasons.

1

Many studies find that the lights are more useful in predicting economic variables in the cross-section than in

predicting time-series changes in these variables (Chen & Nordhaus, 2019; Gibson & Boe-Gibson, 2020; Goldbla tt

et al., 2020; Nordhaus & Chen, 2015).

ASIAN ECONOMIC JOURNAL 289