myanmar-economic-monitor

Data Lab Strategic Brief: Myanmar Economic Monitor

A Strategic Brief is a high-level set of recommendations prepared by the Data Lab that address a given challenge. Recommendations may include internal and external data resources, specific colleagues and/or teams with relevant expertise, data management best practices, suggestions for exploration, and/or a list of relevant resources and similar projects.

Should you have any questions about the Brief, please contact: datalab@worldbank.org

Project Overview

The Myanmar Macroeconomics, Trade and Investment team is responsible for undertaking macroeconomic surveillance of Myanmar and conducting analytical studies to guide economic policy and development, including the bi-annual Myanmar Economic Monitor (P179106). However, published data can be delayed or unreliable, making traditional macroeconomic monitoring a challenge.

The Task Team seeks advisory on alternative data sources, especially remote sensing and satellite imagery, to better understand the current state of the Myanmar economy. Specifically, the team seeks to use these data to track economic activities in key sectors of the economy and assess how current developments – COVID, military coup, sanctions, power outages, and internal displacement due to conflict – are impacting economic outcomes.

The Data Lab advisory is presented in three sections:

Data Collection and Acquisition. Identified data resources that could support the Monitor update.
Data Management. Recommendations for managing derived project datasets.
Data Analytics. A menu of proposals for analytical work that could be coordinated through the Lab network.

Data Collection and Acquisition

This section includes a range of data collection and acquisition recommendations, such as open data resources, leveraging private data partnerships, current World Bank subscriptions and licenses, survey solutions, and remote sensing.###Official UN Earthquake and Refugee Impact Reporting

Population, Demographics, and Human Settlement Data

ID	Dataset	Description	License	Access
1	Meta High Resolution Population Density Maps	Meta, in collaboration with Center for International Earth Science Information Network (CIESIN), used artificial intelligence to identify buildings from satellite imagery and, with census and other data, derived population estimates at a 30-meter resolution. The 2020 dataset includes a spatial breakdown of population by gender and age.	Proprietary	Accessible by submitting a proposal to the Development Data Partnership.
2	WorldPop Population Density	WorldPop released the population dataset for Myanmar from 2020 at a 100m resolution. This dataset is available in a .tif format.	Open	WorldPop
3	UNHCR	UNHCR releases country focussed estimates of refugee movements that are publicly accessible	Open	UNHCR

Satellite Imagery for Detecting Land Surface Changes

ID	Dataset	Description	License	Access
4	United States Geological Survey Landsat and NASA MODIS	Satellite imagery data that can be used to track changes in land area and agricultural productivity. The Data Lab team used Normalized Difference Vegetation Index data in the past to measure change in agricultural productivity for Syria.	Open	Landsat; MODIS
5	Copernicus Sentinel Data	Copernicus Open Access Hub hosts Sentinel radar data that can be used to track land area changes, regardless of cloud cover.	Open	Sentinel

Geospatial Infrastructure Data (Roads, Buildings, Power Grids, Internet Connectivity)

ID	Dataset	Description	License	Access
6	Open Street Maps	Open Street Map hosts crowd-sourced points of interest, amenities, roads, and other physical features for Myanmar.	Open	OpenStreetMaps
7	Microsoft Building Footprints	Microsoft has used AI to generate recent baseline building footprints for Myanmar (dated November 2022) that are accessible through an open-source license.	Open	Building Footprints repository; Footprints for Myanmar:  WB SharePoint
8	Meta Electricity Grid Distribution Maps	Using gridfinder, an open-source tool for predicting the location of electricity network lines using night-time lights satellite imagery and OpenStreetMap data, Meta has released a map that shows the predicted electricity distribution grid and transmission lines.	Open	Electricity Distribution Maps
9	Ookla Speedtest Intelligence Data	Ookla provides internet connectivity data which allows to see quality of internet speeds across the country. Quarterly data at a subnational level is available publicly and the more temporally and spatially granular datasets can be accessed through The Partnership.	Open and Proprietary	Quarterly data: Ookla Open Data portal Daily data: accessible by submitting a proposal to the Development Data Partnership.
10	JBA Flood Risk Maps	JBA provides 30m resolution river and surface water flood hazard maps for 6 return periods. This company also worked to provide support to the farmers of Myanmar through their data.	Proprietary	Accessible by submitting a proposal to the Development Data Partnership.

Data for Understanding Critical Needs and Access to Services

ID	Dataset	Description	License	Access
11	GDELT Project (News data)	The GDELT Project “monitors the world’s broadcast, print, and web news from nearly every corner of every country in over 100 languages and identifies the people, locations, organizations, themes, sources, emotions, counts, quotes, images and events.”  While there are some limitations to the use of the GDELT APIs, if the team were interested in tracking news mentions of specific key words or sentiments by time and place, this is a useful starting point, since the data are free and open source.	Open	GDELT Project; For support running queries: datalab@worldbank.org
12	Premise Price Monitoring Surveys	Premise provides crowdsourced surveys that can be conducted across all countries. Currently, they have	Proprietary	Premise survey data availability list. To access, contact datapartnership@worldbank.org
13	Google Search Trends	Search trends data from the Google Trends API can be used to identify trends in search terms of specific words over time. For instance, a change in search terms related to ‘job’ or ‘prices’ can indicate a specific need in different parts of the country.	Proprietary	Submit a proposal to the Development Data Partnership.

Data for Measuring Economic Activity (Direct and Proxies)

ID	Dataset	Description	License	Access
14	Nighttime Lights	Nighttime lights have proven to be useful predictors of numerous dimensions of human activity: electrification, population, GDP, etc. Recent developments have made the entire nightly archive of nighttime lights available in the public domain, and novel tools have made generating a consistent timeline of nighttime lights across the DMSP and VIIRS sensors possible. These data are accessible through a WB collaboration with the US National Oceanic and Atmospheric Administration (NOAA).	Open	NightTime Lights Data. For support in utilizing Nighttime Lights, contact the Geospatial Operations Support Team (GOST): gost@worldbank.org or Rob Marty (DIME).
15	Meta Relative Wealth Index	The Relative Wealth Index from Meta identifies areas that are richer/poorer in comparison to other areas within the country.	Proprietary	Submit a proposal to the Development Data Partnership.
16	SpaceKnow Economic Activity Monitoring	SpaceKnow is a satellite data analytics company with expertise in using SAR and optical satellite imagery to track economic growth and activity across countries, sectors, climate risk, commodities, and companies. They provide insights into change in economic activity by using Satellite Imagery data	Proprietary	Submit a proposal to the Development Data Partnership.
17	AIS Data	Ship location data aggregated by Spire is made available through a UN Collaboration. This data is available from 1 Deb 2018 – 31 Aug 2022. The Data Lab team has used this data in the past for seaborn trade activity estimation.	Proprietary	To request this data, please contact the Geospatial Operations Support Team (GOST): gost@worldbank.org or Andres Chamorro

Data for Monitoring the Movement of People

All the datasets within this category can be accessed by submitting a proposal to the Development Data Partnership.

ID	Dataset	Description	License	Access
18	Outlogic Observation Panel (Mobile Device GPS Data)	Outlogic collects a mobile location data panel that includes mobility metrics (speed, bearing, altitude, vertical accuracy) and other detection capabilities (IoT, Wi-Fi, and Beacon). The Data Lab team has used these data to track movement across borders between Syria and Lebanon.	Proprietary	Submit a proposal to the Development Data Partnership
19	Veraset Movement Data	Veraset provides a similar service to Outlogic– mobility data comprises population location and movement data derived primarily from mobile device GPS, Wi-Fi, and IoT signals. A similar analysis to Outlogic can be performed using Veraset data.	Proprietary	Submit a proposal to the Development Data Partnership
20	Mapbox Movement	Mapbox Movement is a global data set derived from 20B+ location updates daily, which may be used to understand aggregate activity, density, and movement over time at the city, regional, or country scale.	Proprietary	Submit a proposal to the Development Data Partnership
21	Mapbox Traffic Matrix API	The Mapbox Matrix API is built on anonymized mobile device telemetry data and supports large scale traffic and road network analyses. The API may be used to identify areas poorly served by critical services. This dataset has been used in the past by WB colleagues to analyze spatial accessibility of health facilities. The usefulness of this dataset depends on the availability of Mapbox data for Maynmar.	Proprietary	Submit a proposal to the Development Data Partnership

Data Management

This section includes data privacy policy, data storage and access policy and infrastructure, compute infrastructure, data license compliance, data security classifications, data sovereignty policy, etc.

Collecting Data. If data acquisition involves contracting a vendor for data collection, consider utilizing a ToR Template prepared with checklists of what you can include to ensure data quality, proper data handling and transfer of data throughout the data collection process. Contact Cathrine Machingauta (DECDG) for assistance.
Sourcing Data from External Sources. When obtaining data from external sources (e.g., vendors, contractors, Government) the team should ensure that any licensing or terms of use are documented and stored along with the data. Metadata documentation should also be prepared for any data received to describe the structure and content of the data. This greatly assists in data transparency and reuse. Please contact Cathrine Machingauta (DECDG) if you have questions about data and licensing.
Personal Data Privacy. If your project involves processing personal data i.e., collection, storage, use, transmission, disclosure or deletion any information relating to an identified or identifiable living individual, contact your business unit’s Privacy Focal point for further guidance and compliance with the Bank’s Privacy policy. You can find your Privacy Focal point on this Data Privacy SharePoint site.
Storage and Discoverability. To effectively manage source data and project outputs, the Task Team should consider depositing data in the Development Data Hub (DDH) in accordance with the Bank Procedure for Development Data Acquisition, Archiving and Dissemination and Guidance on Development Dataset Acquisition and Archiving. The Data Catalog provides long-term storage for data, with support for versioning and updates as required. The Catalog makes data easily discoverable, downloadable and accessible via API. This would ensure data is organized during, and remains available beyond, the project for follow-up study or reuse by other Bank staff (as appropriate), in effect creating another useful deliverable from the project. Contact the Development Data Hub Lab Lead Rochelle O’Hagan (DECDG) for assistance.

Data Analytics and Insight Dissemination

With acquired data and sufficient data management procedures and infrastructure in place, how do we responsibly generate and share insights from these data? This section includes economics and statistical analysis, data science (AI/ML), app development, geospatial analytics, code collaboration best practices, reproducible code best practices, data product licensing, etc.

After a brief check-in with the Myanmar Economic Monitor team, the following ideas for analytical contributions were discussed.

I. Understanding Cross-Border Movement and Internal Displacement Patterns

The Data Lab can undertake the following activities to better understand population movements.

Topic	Description
Data Sources	Mobile location data, Movement Data from Mapbox, Population data layers, UNHCR migration data (where available), Open Street Map Points of Interest, Border crossing locations (provided by Task Team), High resolution satellite imagery (to be purchased from Orbital Insight), Rader-based vehicle detection and counts (to be purchased from SpaceKnow)
General Approach	With the popularization of smartphone usage and connectivity across the world, the last decade has witnessed the emergence of massive mobility datasets, such call detail records, geo-tagged posts from social media platforms and generated from GPS devices. These datasets have propelled a rich scientific production on various applications of mobility analysis, ranging from epidemiology to disaster resilience, urban planning and transportation engineering. Through the Development Data Partnership,staff now have access to high-frequency location-based mobility data, constituted of high temporal and spatial resolution timestamped geographical points generated by GPS-enabled devices. By overlaying population and socioeconomic data with mobility data, the project team will be able to produce an analysis based on the estimation of home locations and population displacement to be delivered on a weekly basis. The results could shed light on the movement patterns in different areas of the country. In addition, a point of interest rate visit analysis may show to what degree the decrease in movement and activity in business and commercial centers may correlate to the economic outlook, comparing to historical data.The project team may leverage mobile device generated data, road traffic data, as well as other data sources as they become available to support cross-validation. All analytics will be supported by at least two data sources, where possible – for example, combining Mapbox’s movement statistics with mobile-device location data and comparing with official UNHCR data. For cross-border movement related to trucks, the team will combine the above analytics wtih high resolution satellite imagery for snap-shot counts by mode, as well as radar-based vehicle detection for larger count samples (but without modal split). See also: Mobility for resilience: displacement analysis — mobilkit documentation , Observed Cross-Border Mobility Traces in Lebanon and Syria, and Traffic Trends at Border Crossings: Syria
Outputs	Data Pipeline and Pre-Processing. The Development Data Partnership handles data delivery, data management and data engineering on behalf of all staff and, in doing so, creates economies of scale. Transfer and storage costs are absorbed by the Lab. Datasets. All datasets used in the analysis will be documented and hosted on the World Bank Development Data Hub (for ease of sharing) and a project SharePoint (for the task team’s ease of use). Code and Documentation. All methods will be fully documented and reproducible, so that the analysis can be quickly updated by the task team as new data are available. Indicators. Observed displacement by home locality and time period, and comparative statistics on border crossings abserved through the data Maps. Dynamic maps showing movement / displacement over time, aggregated by a common geospatial index for cross-indicator comparison, and/or aggregated by area of interest. Training and Support. The Lab will provide the task team with training so they may tweak the methodology and continue updating the insights on their own. The Lab would be available for technical assistance, as needed.
Limitations	The methodology relies on private intent data in the form of mobile location data. In other words, the input data was not produced or collected to analyze the population of interest or address the research question as its primary goal but repurposed for the public good. The< benefits and caveats when using private intent data have been extensively discussed in the World Development Report 2021. On the one hand, the mobility data panel is spatially and temporally granular and readily available, on the other hand it is created as a convenience sampling which constitutes an important source of bias. The panel composition is not entirely known and susceptible to change, the data collection and the composition of the mobility data panel cannot be controlled. Due to the nature of the problem and the cardinality, the results can be complex and better analyzed and interpreted when looking at the interactive maps, instead of annual indicators.
Estimated Resources	Compute and Storage. AWS S3 and AWS EC2 r6i.16xlarge (under Data Lab’s custody). ITS charges for mobility data compute and storage services: $5,000 (est.) per month (charged directly by ITS to project code). Staff Resources. <80 hours GF-level staff time. Data Purchase. High resolution satellite imagery and radar-based vehicle count data for select time periods and border crossings (estimate to be obtained upon request by the Task Team).

II. Understanding Economic Impacts

The Lab proposes focusing on the following key business activity and business supporting infrastructure indicators:

II.A. Observed Electricity Usage at Night

Changes in nigttimelight activity can indicate changes in availability of electricity, mass movements of people, relative economic activity intensity, and changes in gas flaring activity. Some of our sample work using this methodology can be seen as part of the Syria Economic Monitor.

Topic	Description
Data Sources	Nightly VIIRS nighttime lights Global Gas Flaring Reduction Partnership for locations of gas flaring facilities. Locations of industrial areas (provided by Task Team) Covid incidence statistics (provided by Task Team)
General Approach	1. Download, clean, pre-process data for areas of interest (including industrial areas). 2. Generate separate map layers for lights observed in gas flaring locations and lights in other locations. 3. Generate maps and statistics showing changes in nighttime light intensity over time and areas of interest. Here is an example. 4. Overlay maps with Covid incidence data to identify potential correlations.
Outputs	Datasets. All datasets< used in the analysis will be documented and hosted on the World Bank Development Data Hub (for ease of sharing) and a project SharePoint (for the Task Team’s ease of use). Code and Documentation. All methods will be fully documented and reproducible, so that the analysis can be quickly updated by the Task Team as new data are available. Indicators. Bi-weekly percent change in nighttime light intensity by area of interest. Maps. Static maps showing percent change in observable nighttime lights (with and without flaring) over time, aggregated by a common geospatial index for cross-indicator comparison, and/or aggregated by area of interest. Training and Support. The Lab will provide the Task Team team with training so they may tweak the methodology and continue updating the insights on their own. The Lab would be available for technical assistance, as needed.
Limitations	Nighttime lights are a common data source for measuring local economic activity. However, it is a proxy that is strongly—although imperfectly—correlated with measures of interest, such as population, local GDP, and wealth. Consequently, care must be taken in interpreting reasons for changes in lights.
Estimated Resources	Compute. Nighttime light analysis compute costs scale with the area of interest. Staff Hours. 24 hours GG-level staff time

II.B. Internet Connectivity Availability and Quality

Topic	Description
Data Sources	Ookla Speedtest Intelligence Data
General Approach	1. Obtain Ookla baseline and most recent subnational data for internet quality through the Development Data Partnership. 2. Identify areas where there is higher quality internet speed.
Outputs	Datasets. All datasets used in the analysis will be documented and hosted on the World Bank Development Data Hub (for ease of sharing) and a project SharePoint (for the Task Team’s ease of use). Code and Documentation. All methods will be fully documented and reproducible, so that the analysis can be quickly updated by the Task Team as new data are available. Indicators. Change in internet speed/network connectivity at a province/district level \| Identified districts where there might be people that are difficult to reach via the internet Maps. Network connectivity map using Ookla dataset. Training and Support. The Lab will provide the Task Team with training so they may tweak the methodology and continue updating the insights on their own. The Lab would be available for technical assistance, as needed.
Limitations	Ookla speedtest data is a collection of speed tests conducted on the Ookla platform. This dataset also does not cover the entire population. Additionally, their dataset is crowdsourced, and their sampling methodology is untested.
Estimated Resources	Staff Hours. 20 hours GF-level staff time

II.C. Shipping Trends

To monitor seaborn trade activity in ports, the Lab proposes use of the Automatic Identification System (hereafter, AIS). AIS was originally developed by the International Maritime Organization in 2004 to prevent collisions between large vessels. This system requires all commercial ships (gross tonnage greater than 300) and passenger ships to broadcast their position via ground stations and satellites. A nascent literature has been dedicated to convert raw AIS messages into economic values of trade, utilizing dynamic information on ship movements, static characteristics of each ship, and reported draft (depth of submergence), to estimate the amount of goods offloaded or loaded at a certain port. More details about this methodology may be found here: Seaborn Trade Activity Estimation

Topic	Description
Data Sources	Ship location data aggregated by Spire and made available through a UN collaboration (available from 1 Dec 2018 - 31 Aug 2022).
General Approach	1. Extract, clean, and pre-process AIS data for areas of interest. 2. Make estimates for any missing data. 3. Estimate trade flows.
Outputs	Datasets. All datasets used in the analysis will be documented and hosted on a project SharePoint (for the Task Team’s ease of use). Since these data are proprietary, they cannot be made openly available. Code and Documentation. All methods will be fully documented and reproducible, so that the analysis can be quickly updated by the Task Team as new data are available. Indicators. Estimated monthly trade volumes based on AIS signals, by port. Training and Support. The Lab will provide the Task Team with training so they may tweak the methodology and continue updating the insights on their own. The Lab would be available for technical assistance, as needed.
Limitations	The accuracy of the trade estimation method depends heavily on the level of draft reporting in ports, which varies globally. The methodology doesn’t currently account for transshipment (goods that are offloaded at a port and then loaded onto another vessel without going through customs). We will have to explore whether data on transshipment rates are available for the ports of interest. This methodology does not capture trade conducted by vessels that turn off their AIS signals. The team intends to explore this topic with further research.
Estimated Resources	Staff Hours. 72 hours GF-level staff time

II.D. Agricultural Productivity

Satellite-derived climate datasets that can be monitored with different temporal resolutions (monthly, daily, and forecast). The variables covered include precipitation, temperature, evapotranspiration, and drought indices. Monitoring these variables in a real-time fashion enables the early detection of water stress in vegetation areas, and helps various actors prepare for climate impacts.

Growing season can be easily observed through phenology metrics measurements using remote sensing (earth observation – EO) data. State of planting and harvesting estimates are determined by importing Vegetation Indices (VI) data into TIMESAT – an open-source program to analyze time-series satellite sensor data. TIMESAT conducts pixel-by-pixel classification of satellite images to determine whether planting has started or not.

Phenological events are sensitive to climate variation. Therefore, phenology data provide important baseline information to assess ecological trends and identify climate change impacts. Comparing current VI values with the long-term averages, and with minimum and maximum values, helps to better understand the performance of the vegetative season and its expected productivity.

Topic	Description
Data Sources	Climate related data (precipitation, temperature, and the derivative products), vegetation indices from optical and radar based data.
General Approach	1. Exploratory Data Analysis 2. Zonal Statistics based on Admin boundaries 3. Generate a Standard csv or Stata-style csv output for each variables
Outputs	Datasets. All datasets used in the analysis will be documented and hosted on a project SharePoint (for the Task Team’s ease of use). Code and Documentation. All methods will be fully documented and reproducible, so that the analysis can be quickly updated by the Task Team as new data are available. Indicators. Various temporal data related to climate based products (precipitation, standardized precipitation evapotranspiration index, consecutive dry days, extreme rainfall exceeding 90th percentile), and vegetation indices based products (anomaly, phenological metrics: start of season, end of season). Maps. Static maps showing indicators over time, aggregated by a common geospatial index for cross-indicator comparison, and/or aggregated by area of interest. Here are examples of similar outputs for work completed for Ukraine: Map and Chart. Training and Support. The Lab will provide the Task Team with training so they may tweak the methodology and continue updating the insights on their own. The Lab would be available for technical assistance, as needed.
Limitations	All the data and analysis is merely based upon the most current publicly available remote sensing data. As the climate phenomena is a dynamic situation, the current realities may differ from what is depicted in this document. Ground check is necessary to ensure if satellite and field situation data are corresponding.
Estimated Resources	Staff Hours. 80 hours GF-level staff time

To get a sense of “need intensity,” the team can explore use of the Google Health Trends API (available through the Development Data Partnership) to determine temporally and by region changes in queries related to key terms provided by the Task Team (e.g., “water,” “money transfer,” “pediatrician,” “hospital,” “Covid”). The method can be adjusted to incorporate other datasets, such as news (GDELT) or social media data.

Topic	Description
Data Sources	Google Search Trends API\| Network Connectivity map obtained through Ookla internet quality \| ACLED data \| Covid incidence rate (provided by Task Team) \| GDELT news data \| Poverty HH survey data
General Approach	1. Identify search terms that can be queried on Google Trends API 2. Map the changes in search terms as they correspond to ACLED-reported conflict data and Covid spikes. 3. Verify if this change can be attributed to a change in connectivity network coverage 4. Identify areas where there is an unnatural spike in number of times a search term appears and compare to ACLED , Covid-incidence data, and news data 5. Compare results with HH survey data to identify corrolation.
Outputs	Datasets. All datasets used in the analysis will be documented and hosted on the World Bank Development Data Hub (for ease of sharing) and a project SharePoint (for the Task Team’s ease of use). Code and Documentation. All methods will be fully documented and reproducible, so that the analysis can be quickly updated by the Task Team as new data are available. Indicators. Changes in identified needs by type and the lowest-level of aggregation possible Maps. Dynamic map of needchanges in surveyed goods by location over time, aggregated to H3s for comparison Training and Support. The Lab will provide the Task Team with training so they may tweak the methodology and continue updating the insights on their own. The Lab would be available for technical assistance, as needed.
Limitations	Google Trends relies on people with an active internet connection to search, which excludes people without internet access or who do not use Google search.
Estimated Resources	Staff Hours. ~ 40 GF-level staff hours.

IV. Dissemination and Capacity Building

Since analytical results from this work could support additional teams and counterparts, the team create a centralized repository of all datasets used in analytical work an ensure all data are suitably documented and made accessible (where licensing permits), and that all methodologies are similarly made available through GitHub, so that others can reproduce the results. The Support for the Syria Economic Monitor is an example.

Additionally, the Lab can produce a web-based map for layering indicators for ready comparative analysis, as well as an Excel workbook for ease of indicator dissemination across the widest possible audience.

Additional Resources

Development Data Partnership’s Community and Documentation

The Development Data Partnership fosters a community of data practitioners and maintains a robust data documentation and code collaboration platform based on GitHub, which is recognized as a good practice in the World Bank and in partner international organizations. See more at https://docs.datapartnership.org.