Vahid Fouladgar; Mehdi Feizi
Abstract
1- INTRODUCTION
The existence of accurate and sequenced economic statistics and information can be seen as a prerequisite for any economic conditions assessment for countries. However, for several reasons, economic data in many developing countries, including Iran, are produced with low accuracy or ...
Read More
1- INTRODUCTION
The existence of accurate and sequenced economic statistics and information can be seen as a prerequisite for any economic conditions assessment for countries. However, for several reasons, economic data in many developing countries, including Iran, are produced with low accuracy or low sequence. The weakness in the variety and quality of economic data is an important obstacle to socio-economic research and understanding the country's conditions correctly for policymakers.
Sending satellites and imaging the earth's surface at different hours and publishing the information of these satellites in recent years, in addition to providing a platform for geographical and military investigations, has also given researchers in the fields of human sciences the opportunity to use this free, non-manipulated and universal data in different countries and times, for their research (Chen & Nordhaus, 2019)
This research uses the new generation of monthly night light data (VIIRS) to estimate Iran's seasonal GDP for the first time. What distinguishes the current research from previous studies is evaluating the explanatory power of nighttime light images for shorter periods (seasons) in Iran (including night light, GDP and population) and using new data set (VIIRS vs. DMSP).
2- THEORETICAL FRAMEWORK
The use of seasonal periods, especially before the development of information technology and the development of real-time data-gathering methods, facilitates researchers and policymakers to monitor changes in economic indicators and reduces data collection and publication costs (Bell and Hilmer, 2012).
The review of internal research using nighttime light images indicates that few studies have been conducted. Most of these studies were focused on geographical and remote sensing applications, only Akhbari et al. (2017) used the old version of NTL (DMSP) for economic studies in Iran.
In foreign studies, researchers at first focused on the feasibility of using the mentioned data to represent the number of economic activities at the country level (Chen and Nordhaus (2011), Henderson et al. (2012), Shi et al. (2014)) and after those studies focused on using these data in different countries and for new purposes such as inequality analysis, integration with other data, etc. (Chen and Nordhaus (2015), Beyer et al. (2018). Then and after the release of VIIRS as a new generation for satellite images, many studies were conducted to compare the efficiency and accuracy of these data with the previous generation in different countries. After that, this information was used for more detailed analysis: in short periods, at more minor geographic scales, and for analyzing social and political shocks. Some of these studies can be found in Chen and Nordhaus (2019), Wang et al. (2019), Sun et al. (2020), Farzangan and Hayo (2018), Farzangan and Fischer (2021).
3- METHODOLOGY
In this research, the night light satellite images were extracted from the University of Mines database. After refining the GeoTiff files and cutting them according to Iran's map shapefiles (using QGIS software), the amount of emission during 37 seasons (from spring 2012 to spring 2020) has been calculated. In the next step, this variable is used as an independent variable in 4 linear regression models. Its ability to estimate seasonal GDP has been investigated by combining population data and virtual variables for seasons. In the last step, to evaluate the predictive power of the selected model, GDP data for three seasons after the training data have been predicted and the error rate of this prediction has been calculated.
4- RESULTS & DISCUSSION
According to the results of this paper, R2 for a Univariate model with night light data for estimating seasonal GDP in Iran (without oil) is 0.828; if this model improves with the seasonal population variable and seasons dummy variable, its explanatory power can increase up to 0.945.
Also, it was observed that the selected model, which includes night light, seasonal population information and virtual variables of the seasons, can estimate the dependent variable (seasonal GDP) with less than 2% error in three seasons after the mentioned period.
5- CONCLUSIONS & SUGGESTIONS
Considering the delay in national economic indicators publishment (such as seasonal GDP), besides the complications and costs of calculation for this index, this paper shows that nighttime light (NTL) can be used as a good proxy for calculating the GDP in Iran. Also, the combination of this variable with two more variables (population and the dummy variables for seasons) can increase the accuracy of the estimates (R2=0.945) and estimate the dependent variable (seasonal GDP) with less than 2 percent error for the next three seasons.
Finally, it is suggested that future studies focus on estimating the amount of economic activity at more minor geographical levels (provincial and city), using these data and evaluating the efficiency of these data for estimating other economic indicators.
