This page was created for Carlos Petricioli's thesis presentation. You can get a copy of this work at the Download tab.
All images presented in this page were created for this work.
As a result of the work made and with the aim of promoting research and statistical/econometric development of techniques that involve using innovative alternatives to create better estimations of the national accounts, please do not hesitate to contact me for the dataset produced a result of this work. If you plan to use them, please cite. BibTeX citation.
Any comment regarding this work is welcome.
This work shows how to use the nighttime lights satellite images time series provided by NOAA. It develops the necessary methodology to make the images consistent inter-temporarily and inter-satellite (for some years the images are based in different satellites) so that it makes sense to use the nighttime light intensity from space data, measured with a Digital Number (DN) with values from 0 to 63, to estimate the economic growth of the Mexican Federal Entities' GDP. Each pixel on each image is an annual average of the nightlight visible from space for an area that corresponds to around one square kilometer. This means that the complete series is made from around 290 millions of pixels just for Mexico. This is why this work shows how to deal with the data even from the download. To start, this work presents an extensive literature of how to use spatial time series and a historic review of how this nightlight time series has been used. Moreover, a calibrated nightlight time series was developed and its available for download. Additionally, this work developed a subnational time series of spatial nightlight inequality Gini indexes that can be used to measure inequality in each Municipality among each Federal Entity in Mexico, a completely new metric in the national accounts. Finally, this work presents the theoretical framework necessary to develop an economic growth model for the States' GDP that covers from 1992 to 2013 and uses the nightlight DN as a proxy. The results were desirable thanks to the calibration process. It was found that the nightlight visible from space represents a good proxy that can be used to correct the economic growth that the government reports for the States. It was found that in some cases, the government can actually over or under estimate the long term (10 years) States' GDP growth as much as 15%. This work represents a viable option to correct this bias.
Topics: Geospatial analysis, DMSP-OLS, Nightlights, GIS, Satellite intercalibration, Subnational GDP estimation, NLDI, Gini, Proxy, Economic growth, R, Rgeos.
Here you can see a video of my executive presentation at Google's Data Point, an event for the Data Science community that took place at Google New York, NY.