Project

General

Profile

Nighttime Lights Study Ideas

We have two very high quality datasets available to us for original analysis of the economic effects of SEZ: The UNCTAD database of SEZ and a harmonized, temporally calibrated fusion of the DMSP-OLS night-time light and VIIRS night time lights datasets. We can combine these datasets to assess the economic effects of the creation of SEZ.

A dataset of worldwide political boundaries may also be required.

The basic procedure for a difference-in-differences analysis would involve:

  • Select a set of SEZ that have creation dates at least two years after the start, and two years before the end of the range of nighttime light (NTL) data. The NTL dataset runs from 1992 to 2018, so the SEZ creations should be from 1994 to 2016.
  • Select a geographic region around each SEZ, as an area of influence for the SEZ. ( treatment zone)
  • Select the remainder of the administrative region that contains the SEZ — a province or county — as a comparison area. ( control zone )
  • For each SEZ, compare the NTL intensity change of the treatment zone to the control zone.

We will need to do some preparatory analysis to determine if the use of the balance of the administrative region is a sensible control, particularly for SEZ that are close to a boundary

I haven't started a literature review, which may generate other Ideas for analysis.

Notes

  • Harvard's China map may have additional geo data that would be useful
  • It would be valuable to be able to exclude the influence of other projects, such as rail lines, which may have been built near the same time as the SEZ.
  • Consider the proximity to: airports, universities, ports, train stations.
  • We can control for GDP in top level provinces, possibly sub-provinces. The prefecture data comes from a set of statistical years books, so it may be difficult to get original data ( rather than from Wikipedia. )
  • Using Synthetic Control Method to build a control model for the administrative regions before the start of the SEZ. A good example of the technique here https://www.cato.org/sites/cato.org/files/pubs/pdf/working-paper-41.pdf
    • Advantage as a control that it doesn't require specific knowledge about the prefectures of a country. You can just regress past economic data on an equation with the economic data of every other prefecture, get a tight fit for the pre-SEZ years, then extrapolate and compare after the SEZ.
  • Use Open Street Map to get information about roads, buildings and estimates of urbanization.

Finding Control Zones

The crux of the difference-in-differences technique will be identifying appropriate control areas — areas that do not get SEZ but are similar to the treatment areas that did get SEZ.

  • Compare pre-treatment growth. Locate areas that have the same size as the treatment zones and have the same pre-treatment growth trajectory in NTL.
  • Match areas by pre-treatment business intensity, road density, population density.