`Measuring Economic Growth from Outer Space` Weil
Aim explore the usefulness of NTL as a different proxy for economic activity
Motivation Gross Domestic Product (GDP) = key measure of economic growth
Badly measured in developing countries:
- Large informal sector + Weak government statistical infrastructure
- periods or regions for which GDP data are not available at all
- creates problems for policy making: (IMF) and World Bank both rank countries regarding their national
statistics + ratings focus solely on the quality of national accounts data
Context: Use NTL in SSA: If coastal area, primate cities, malaria-prone areas grow faster
NTL Better than GDP:
- Error prone: lights observable from space
- Coverage on fine spatial scales = NTL data on 1km square + observe at city level
- high frequency = resolves countries not having city-level GDP data
- Proxies economic activity at geographical and temporal scale which previously unavailable
How it reflects economic activity: Consumption and investment require lights, rises with income
- Fig 2: south korea light intensity reflect long-term growth:increase of 119% in GDP
Data NTL; focus on human-made light & exclude clouds, auroral activity etc .
- digital number (0- 63) represents light intensity year fixed effects bc varying sensor settings over
time
Data problems (overcome by adding country fixed effects)
- light for the same GDP vary with production among different activities and population density.
- lighting technology varies which affects relationship between light and GDP.
Data Quality – World Bank grading scheme = info on sets of countries with better or worse national statistics,
Identification 3 Regression of income growth on lights growth where estimate (ψ) inverse of elasticity of lights wrt income)
- Country panel for 1992–2008 with country and year fixed effects , relies on
within country variation in income and lights overtime
- Add country-specific time trend to see how lights predict deviations from
growth trend (considers rachet effects) – account for annual fluctations
- focus on long-run growth - period 1992 -2006 bc missing data after
Assume national income accounts is uncorrelated with measurement error in using lights to measure growth.
Ratchet issue: New light installations may result in non-decreasing light levels, masking economic downturns.
Results o Table 2: highly sig ψ: 0.277 NTL strongly correlates with economic activity + predicts GDP
o Table 4:estimated elasticity is 0.3, indicating lights' predictive power for income fluctuations.
o Table 5: low- to middle-income countries elasticity of growth in night lights wrt true GDP growth
(β)=1.15 = close to 1: long-term rate of lights growth approximately = rate of true income growth
Fig 7: issues with GDP measurement in the WDI.
- Points near the 45-degree line indicate similar results between (WDI data and lights data)
- adjustments lead to higher growth rates for Republic of Congo and Haiti, while certain countries with
extreme recorded growth rates, like Burundi, undergo significant revisions
Application to Subsaharan Africa NTL is provides measure in places where data is unreliable:
Coast vs Interior: Argue coastal has more growth BUT NTL data = inland regions experienced greater
growth in lights & coastal areas grew more slowly.
Primate Cities vs Hinterland: in ssa, hinterland areas grew slightly faster than primate cities
(concentration of growth in dominant cities) challenging that primate cities are sole drivers of
development
Effect of Malaria on Growth: regions with historically lower malaria experienced faster growth rates,
even after the antimalarial campaigns = reductions in malaria doesnt translate into increased GDP
Mechanism rate of increase in lights may diminish as income rises bc w urbanization, increased population density could
block some lights from reaching space= elasticity < one.
Policy lights data key role in analyzing growth at supranational levels, where income data at spatial level are
Aim explore the usefulness of NTL as a different proxy for economic activity
Motivation Gross Domestic Product (GDP) = key measure of economic growth
Badly measured in developing countries:
- Large informal sector + Weak government statistical infrastructure
- periods or regions for which GDP data are not available at all
- creates problems for policy making: (IMF) and World Bank both rank countries regarding their national
statistics + ratings focus solely on the quality of national accounts data
Context: Use NTL in SSA: If coastal area, primate cities, malaria-prone areas grow faster
NTL Better than GDP:
- Error prone: lights observable from space
- Coverage on fine spatial scales = NTL data on 1km square + observe at city level
- high frequency = resolves countries not having city-level GDP data
- Proxies economic activity at geographical and temporal scale which previously unavailable
How it reflects economic activity: Consumption and investment require lights, rises with income
- Fig 2: south korea light intensity reflect long-term growth:increase of 119% in GDP
Data NTL; focus on human-made light & exclude clouds, auroral activity etc .
- digital number (0- 63) represents light intensity year fixed effects bc varying sensor settings over
time
Data problems (overcome by adding country fixed effects)
- light for the same GDP vary with production among different activities and population density.
- lighting technology varies which affects relationship between light and GDP.
Data Quality – World Bank grading scheme = info on sets of countries with better or worse national statistics,
Identification 3 Regression of income growth on lights growth where estimate (ψ) inverse of elasticity of lights wrt income)
- Country panel for 1992–2008 with country and year fixed effects , relies on
within country variation in income and lights overtime
- Add country-specific time trend to see how lights predict deviations from
growth trend (considers rachet effects) – account for annual fluctations
- focus on long-run growth - period 1992 -2006 bc missing data after
Assume national income accounts is uncorrelated with measurement error in using lights to measure growth.
Ratchet issue: New light installations may result in non-decreasing light levels, masking economic downturns.
Results o Table 2: highly sig ψ: 0.277 NTL strongly correlates with economic activity + predicts GDP
o Table 4:estimated elasticity is 0.3, indicating lights' predictive power for income fluctuations.
o Table 5: low- to middle-income countries elasticity of growth in night lights wrt true GDP growth
(β)=1.15 = close to 1: long-term rate of lights growth approximately = rate of true income growth
Fig 7: issues with GDP measurement in the WDI.
- Points near the 45-degree line indicate similar results between (WDI data and lights data)
- adjustments lead to higher growth rates for Republic of Congo and Haiti, while certain countries with
extreme recorded growth rates, like Burundi, undergo significant revisions
Application to Subsaharan Africa NTL is provides measure in places where data is unreliable:
Coast vs Interior: Argue coastal has more growth BUT NTL data = inland regions experienced greater
growth in lights & coastal areas grew more slowly.
Primate Cities vs Hinterland: in ssa, hinterland areas grew slightly faster than primate cities
(concentration of growth in dominant cities) challenging that primate cities are sole drivers of
development
Effect of Malaria on Growth: regions with historically lower malaria experienced faster growth rates,
even after the antimalarial campaigns = reductions in malaria doesnt translate into increased GDP
Mechanism rate of increase in lights may diminish as income rises bc w urbanization, increased population density could
block some lights from reaching space= elasticity < one.
Policy lights data key role in analyzing growth at supranational levels, where income data at spatial level are
