Predicting Multidimensional Poverty with Machine Learning Algorithms: An Open Data Source Approach Using Spatial Data

Round 1

Reviewer 1 Report

The scientific papers require an impersonal writing while your proposed paper includes "we" as far as 21 times and "our" another 10 times. For the elegance of the writing these elements should be replaced by impersonal formulations.

Author Response

Please see the attachment. Your review is the first in the document (Reviewer 1). 

Author Response File: Author Response.pdf

Reviewer 2 Report

Congratulations on the paper. The text is well-organized and easy to follow. Using alternative data to estimate multidimensional social phenomena is evolving and contributing to public management. I agree that the census's cost and frequency of publication are solid motivations for applying the approach proposed in the research. However, some points must be corrected, and others must be improved before publication.

Major comments:

1.) Lines 41-43 ("The impacts of Covid-19 on multidimensional poverty are estimated, showing that 460 million more poor people would enter multidimensional poverty"). Some attention is needed in using satellite images for the longitudinal analysis of poverty. The physical aspects of cities are relatively stable. Therefore, it is important to highlight the limitation of the approach proposed in this specific case. Furthermore, the limitations of using satellite images to measure longitudinal poverty directly affect the proposal in Line 391. Please, rephrase the sentence: "Another highlight is that the estimation can be updated automatically, avoiding waiting for a new survey." Then, highlight the limitations of the research approach and its implications for the longitudinal analysis of multidimensional social phenomena.

2.) Sentences starting on lines 58 and 118 conflicts with the literature review presented between lines 82 and 93 and with the literature itself [1,2,3]. Please adjust and strengthen your arguments.

3.) The terms "auxiliary data" or "supplemental data" are used interchangeably in Lines 67 and 68. However, the literature defines the data type used in the research as "alternative data." That is unconventional data extracted from other data (images, texts, social media, geolocation) using big data techniques [3,4].

4.) Please enter the year of all data. I could not find year information from the sub-indicator data for calculating the composite poverty indicator and Open Street Map data.

5.) Please justify the criteria for defining the distance limits (Lines 303 and 304)

6.) Line 413: For future studies, it is also possible to implement algorithms that minimize the confusion between exposed soil and buildings (roofs of houses) to improve the overall classification [4].

Minor mistakes:

·       Line 41: For example, Through simulations.

·       Line 122: Reference [?]

·       Lines 140 and 150: What is IPM? Do you mean MPI?

·       I could not access the database at the link: https://https://github.com/sandboxDANE/IPM-Pobrezamultidimensional

·       Please maintain homogeneity (e.g., Line 213: multidimensional poverty indicator / Line 228: multidimensional poverty index) throughout the text!

·       Finally, proofread the text to correct other typos and grammatical errors.

The recommended literature to enrich the work:

[1] di Bella, E., Leporatti, L., & Maggino, F. (2018). Big data and social indicators: Actual trends and new perspectives. Social Indicators Research, 135, 869-878.

[2] Suel, E., Bhatt, S., Brauer, M., Flaxman, S., & Ezzati, M. (2021). Multimodal deep learning from satellite and street-level imagery for measuring income, overcrowding, and environmental deprivation in urban areas. Remote Sensing of Environment, 257, 112339.

[3] Niu, T., Chen, Y., & Yuan, Y. (2020). Measuring urban poverty using multi-source data and a random forest algorithm: A case study in Guangzhou. Sustainable Cities and Society, 54, 102014.

[4] Libório, M.P., de Souza, J.B., Guimarães, S.J.F., & Ekel, P.I. (2022). Estimating municipal economic activity: An alternative data-based approach. Remote Sensing Applications: Society and Environment, 28, 100877.

Author Response

Please see the attachment. Your review is the second in the document (Reviewer 2). 

Author Response File: Author Response.pdf

Reviewer 3 Report

I reviewed the manuscript and have the following comments for the authors:

1.       The abstract is not explaining the conducting study completely. To make the abstract more attractive, it is advised to add the data collection techniques, Sample size, the finding of the study, etc. This will help readers to better apprehend the true meaning and significance of your study. Also, this statement is not clear “The methodology can be used to predict multidimensional poverty at granular spatial 5 levels”. The methodology can be used, or methodology has been used??

2.       The author used the night-light images taken from space to predict socioeconomic conditions and wealth index indicators as a proxy of the multidimensional poverty index. How this can be justified? Night-light images taken from space can consider the proxy for energy poverty, but how it can explain someone’s socioeconomic conditions or poverty level.? It is suggested to review and edit this concept accordingly and also modify the study title to only energy poverty instead of multidimensional poverty index. See the following articles: 10.1111/aswp.12152: 10.1016/j.rser.2022.112157

3.       Using OpenStreetMaps (OSM), the features gathered were related to distance to police stations, hospitals, schools, universities, churches, airports, banks, train stations, and bus stops. But how this data will be related to the multi-poverty index? Have you used this as an explanatory factor that can affect the poverty level? For which households? Does the date you have taken from different sources are the date for the same households? How this will be ensured?

4.       Linear regression is the used model for prediction; however, I have not found any findings related to Linear regression.

5.       How the authors measured the reliability, validity, and robustness of the data? Which techniques have been used for this?

6.       How the findings of this study can be generalized?

Author Response

Please see the attachment. Your review is the third in the document (Reviewer 3). 

Author Response File: Author Response.pdf

Round 2

Reviewer 3 Report

N/A