word count: 1977 words
Steele, J. (2017, February 19). Mobile phones can create high-resolution poverty map. Retrieved February 23, 2021, from https://www.indiatoday.in/technology/news/story/mobile-phones-can-create-high-reolution-poverty-map-959791-2017-02-09
The article I read was one that pertained to how mobile phones can create high-resolution poverty maps in India. Researchers found that if geospatial data from satellites were combined with mobile data, it would allow them to produce poverty predictions. The article points out that censuses, although useful, are inefficient. “Census and household surveys are normally used as data sources to estimate rates of poverty. However, they are not regularly updated and in low-income countries, surveys can be patchy,”, says Jessica Steele, the lead author of this study. Phones are constantly updated second by second, leaving it the prime source of information as a whole. When this is paired with satellite data, a dynamic view of poverty and spread is laid out. It was very interesting to see how much information is packed into the phone: the apps, data usage, number of texts sent/received, etc. All this information is used to paint a picture not yet conceived, and it is absolutely genius on how data scientists do this. This micro-level information can reveal if a person is migrating to a different area due to better economic opportunities or a household’s access.
This article relates to Amartya Sen’s definition of human development, which is simply expanding the freedoms that people enjoy. By laying out a dynamic image of poverty of various nations, financial and economic aid can be administered effectively and efficiently. The obvious dimension of human development that is being explored here is poverty assessment and analysis, and the goal of this is to pull citizens out of the crutches of poverty. The way this is currently being done is slow and ineffective, but with the use of data analysis, true human development in the realm of poverty can be achieved. The geospatial datasets that were used were satellite imagery/RS (remote sensing) data, while the geospatial methods that were used to help create these poverty maps are the assessment of: levels of data usage, number of texts sent, times calls were made and their duration, monthly credit consumption on mobiles, and various movements of mobiles and their use of networks. All this information coupled together can still paint a significant map, regardless of the fact that some of the poorest in society do not own a cellphone. Finally, I think that the scientific question in this article is: Can something as simple as a cellphone truly map out poverty effectively? To which both I and the author of the article believe is true.
Steele JE et al. 2017 Mapping poverty using mobile phone and satellite data. J. R. Soc. Interface 14: 20160690. http://dx.doi.org/10.1098/rsif.2016.0690
This article also relates to the mapping of poverty using mobile phone and satellite data. They start with pointing out that traditional approaches to targeting poverty rely heavily on census data. However, in low-middle income countries (LMICs), this information is either out-of-date or fully unavailable. The article suggests that if public and private data sources were to be combined, a novel insight into the distribution of poverty could be mapped out. As stated in the article “Mobile phones can create high-resolution poverty map”, there are various elements in a cellular phone and the information it contains. However, in this article we take a deeper look in how this data is collected and paired with satellite information. RS data or remote sensing data collects physical properties of land, while CDR (call data record) data collects data collected at the level of physical cell towers. When RS and CDR data are combined, this complementary pair can capture distinct human living conditions and behavior. The poverty data that was being represented in this study was asset, consumption, and income-based measures of wellbeing for citizens in Bangladesh. The dataset that was being studied showed that when RS and CDR data was coupled, it produced a higher R-squared value when compared to the CDR data or RS data by itself. The R-Squared value is a value that shows its statistical significance on what it is trying to predict or model.
This article relates to Amartya Sen’s definition of human development, which is simply expanding the freedoms that people enjoy. By laying out a dynamic map of poverty, financial and economic aid can be administered effectively and efficiently. This article shows that the data being collected and modeled surpasses anything that the census could. Poverty assessment and analysis has reached new heights with the administration of CDR and RS data. The geospatial data sets and methods that were specifically used for this study the CDR and RS data. The study really opened my eyes on the progress data has made to eliminate or help problems, such as wide-spread poverty. It is clear that data collection and analysis can only get better from here, and human development in the realm of poverty is apparent.
“Pape, Utz; Parisotto, Luca. 2019. Estimating Poverty in a Fragile Context : The High Frequency Survey in South Sudan. Policy Research Working Paper;No. 8722. World Bank, Washington, DC. © World Bank. https://openknowledge.worldbank.org/handle/10986/31190 License: CC BY 3.0 IGO.”
This article relates is about how poverty is estimated in South Sudan. The article starts out by revealing why poverty is so apparent in South Sudan today, and why not much is known about the welfare of many of the citizens there. Civil war broke out after the republic of South Sudan gained their independence, and this conflict resulted in the displacement of about a third of the population. Due to the census being the slow entity that it is, the last recording of wellbeing and poverty dates back all the way to 2009. To combat that gap in data, researchers from the National Bureau of Statistics (NBS) implemented a High frequency South Sudan Survey (HFS). The HFS measured various aspects: consumption, poverty, welfare, currency devaluation, and inflation. All these components together resulted in extremely high poverty levels, in fact the HFS was said to have estimated that 4 out of 5 people across 7 states of South Sudan were living under the international poverty line of $1.90 PPP. Along with the HFS survey, supplemental data from was collected thanks to satellite imagery. All this data combined helped statistical models fill in gaps of poverty for inaccessible areas that were plagued with conflict. It was very interesting to see how impactful HFS and satellite data was for representing the crippling poverty in those areas of South Sudan. This shows the implications of data science and how useful big data is to solve big problems. The census, in my opinion, does not represent poverty levels properly and does not care to access small areas such as the ones represented in the study.
This article relates to Amartya Sen’s definition of human development perfectly, in that this data can be used to expand freedoms of impoverished people and give them a better life. The dimension that is being addressed in this research is poverty assessment and analysis. The obvious development goals for this study are to remove people from poverty and give them access to the help and resources they need. The geospatial datasets and methods that were used in this article were satellite data and the HFS . The development pattern or process that is being explored here is the estimation of poverty within South Sudan, and the spread of this poverty. Unlike the other articles whose purpose was to map out poverty, this article delves more into the estimation of poverty in areas that would not normally be surveyed and depict how severe this problem is and the reasons behind for it. Lastly, the scientific question that the author is trying to seek the answer for is: How can we estimate poverty in inaccessible areas effectively and efficiently?
“Pape, Utz; Wollburg, Philip. 2019. Estimation of Poverty in Somalia Using Innovative Methodologies. Policy Research Working Paper;No. 8735. World Bank, Washington, DC. © World Bank. https://openknowledge.worldbank.org/handle/10986/31267 License: CC BY 3.0 IGO.”
Somalia is a country that has experienced a lot of trauma which has led to immense poverty within the country. Everything ranging from terrorist attacks to drought has resulted in the severe internal displacement of Somalia’s population. As with a lot of low socio-economic countries, Somalia is highly data-deprived leaving policy makers not enough data to work with and improve the poverty levels within the nation. The last time the government conducted a full population census dated back to 1975. To fill this vast gap in data, a Somalia High Frequency Survey (SHFS) was implemented in 2016 by the World Bank. This is very similar to HFS (High frequency South Sudan Survey) discussed in the previous study that tried to estimate poverty in South Sudan. However, getting this data would be far more of a challenge when compared to South Sudan. Somalia lacked the statistical infrastructure which posed a number of problems for effectively implementing a household survey and estimating poverty. Due to no recent census, inaccessible areas (risk to safety of field staff), and the nature of the nomadic population, obtaining significant and accurate data would be a challenge. This study outlines how the SHFS overcomes these challenges through methodological and technological adaptations. Geospatial techniques along with high-resolution imagery were used in the SHFS to model population distribution and build a probability-based sampling frame. This was administered in order to overcome the lack of a recent population census, and this method relies heavily on satellite imagery. The geo-spatial data that was collected was used to create a “security assessment access map”, which allowed researchers to easily see the inaccessible/unsafe areas within Somalia. The way that the SHFS and HFS were implemented in their respective countries now seems to be virtually identical. Both methodologies show that even with huge challenges in obtaining data can still be overcome through geospatial data and satellite imagery. It was very interesting to see how all of the articles that I summarized and assessed are so similar in their approach to fixing the problem of poverty in various nations. I love how this study specifically was aimed towards mapping out inaccessible areas through satellite imagery in order to alleviate poverty and inequality within Somalia in a safe and effective fashion. This study was also very good at outlining the hardships that come with data collection in these dangerous and unfamiliar lands. The author was able to overcome these hardships through: building a probability-based population sampling frame, estimating poverty derived from satellite imagery and geo-spatial data, and employing special sampling strategies for the outlier of nomadic populations. This goes to show that data collection is a fragile yet necessary work. It was also interesting to see how the author noted that his work could be further refined with applications incorporating predictors with higher spatial frequencies which would improve estimates. Overall, this was my favorite study to read and really made me ponder more about the applications and precision that comes with data analytics.
This article relates to Amartya Sen’s definition of human development, because the manufacturing of poverty maps or security assessment access maps would allow for citizens of Somalia to expand their freedoms and gain a better life. The dimension of human development that is being addressed by the authors’ research is poverty assessment and analysis. The development goals that the author stated was to refine poverty estimations by incorporating predictors with higher spatial frequencies and building footprints. The geospatial data sets used in this study were satellite imager and the SHFS itself, while the methods could range from anything between simple sampling strategy to survey design. The human development pattern that the author is investigating pertains to poverty assessment of all of Somalia’s people, including the nomadic ones. Finally, I believe that the scientific question being addressed here is: How can we effectively map out poverty in inaccessible/dangerous areas?
In conclusion, all of these articles really opened my eyes on what data science can achieve. The implications are endless and can branch out even further than the realm of poverty analysis and assessment. I feel that with all my analysis of these articles, the next direction would be to study the effects of all this research on a broader scale. All of these studies are confined to a certain region or area, but I want to see this data utilized by governments to cause real change. The data is there, all we need to do is utilize it. Overall, I really enjoyed reading and assessing these articles and I feel that I laid out a perfect blueprint for Assignment 2.
Castelan, C. R. (2019, July 9). Making a better poverty map. Retrieved March 22, 2021, from https://blogs.worldbank.org/opendata/making-better-poverty-map
This article has to do with enhancing the detail and data inclusion of existing poverty maps. Basically, they are just using satellite data, CDR data, and RS data in order to fill in the gaps of poverty maps and make them much more comprehensive. In order to enhance machine learning of these predictive models and maps: Validation must be accelerated, local capacity should be strengthened, improvement of data infrastructure, and the investment of data access arrangements. Much of the rest of the article is very similar to Dr. Jessica Steel’s reports; however, this article pertained to more the advancement of those maps and predictive models.
Horton, M. (n.d.). Stanford scientists COMBINE satellite data, machine learning to map poverty. Retrieved March 22, 2021, from https://pangea.stanford.edu/news/stanford-scientists-combine-satellite-data-machine-learning-map-poverty
This article is by Stanford scientists, and how they point out that machine learning and satellite data can produce poverty maps and predictive models. During the day, satellite imagery can pick up conditions of roads and homes and predict the poverty levels in the region. During the day, satellite imagery can see the amount of light and cellular activity, and this information can indirectly depict poverty levels (typically, the more light and cellular activity in a region usually means the more developed it is). It was concluded that the power and capability of machine is second to none, in that it does an amazing job of predicting poverty distributions and actually outperforms existing approaches.
Kumar, A. (2020, July 06). How to understand global poverty from outer space. Retrieved March 22, 2021, from https://towardsdatascience.com/how-to-understand-global-poverty-from-outer-space-442e2a5c3666
This article provided me with further knowledge on how global poverty can be assessed via satellite imagery from outer space. In this specific study, satellite imagery and neural networks were used to predict asset wealth. Similar to the study done by the Stanford scientists, economic livelihood was assessed through nightlights. In order to complete the task of utilizing computer vision to map and predict poverty in Rwanda, five steps needed to take place. 1. Download Demographic and Health Surveys (DHS), nightlight satellite imagery, and daytime satellite imagery. 2. Test whether nightlights can predict wealth accurately. 3.Test whether basic features of daytime imagery can also predict wealth accurately, and extract image features. 4.Construct a convolutional neural network (CNN) leveraging a combined dataset of daytime and nightlight images and apply transfer learning. 5.Construct maps showing the predicted distributions of wealth. In conclusion heatmaps of wealth predictions were able to be plotted.
Martinez, A., Jr. (2021, March 11). Using machine learning on satellite images to map poverty. Retrieved March 22, 2021, from https://development.asia/insight/using-machine-learning-satellite-images-map-poverty
This article is virtually the exact same as the Kumar paper, in that poverty models and maps were created in order to assess economic prosperity or economic hardship. In this article, instead of assessing poverty in Rwanda, it was about assessing poverty in the Philippines and Thailand. Though, my focus was on Eastern Africa; The same principles still apply, and further improved my understanding on how satellite imagery is used to predict poverty.