Call Detail Records (CDRs) are metadata (data about data) that capture subscribers’ use of their cell-phones — including a timestamp, subscriber identification code and, at a minimum, the location of the phone tower that routed the call for both caller and receiver. Large operators standardly collect over six billion CDRs per day. Note that the scope of the term CDR has expanded extensively beyond the meaning implied by its original acronym ("Call Detail Records"). In current usage CDR data refers not only to calls, but to all network events (made by cell phones, tablets, etc.) which the operator records. This includes, but is not limited to, calls, sms, data usage sessions and mobile money transactions. The exact form of a single CDR is dependent on the type of network event and data retention policies within the operator.

For this phases we did not have any CDR data to perform analysis on the 9 zones. In lieu of, an overview on using CDRs and indicators that can be dervived from them is provided below. A more detailed treatment and methodology is provided in the accompanying report "Initial proposals for the development of Economic Micro-indicators from Call Data Records" by N/LAB.

Using CDR data

The use of CDR data for demographic and economic analysis is of increasing interest to the scientific community, and we expect for this interest to have extended into broader sectors by mid 2019. Traditionally, economic indicators are obtained from a number of logistically challenging methods. In particular surveys (which are costly and produce relatively small sample sizes) and direct statistical contributions from private sector (which are extremely logistically challenging to undertake in LDC and LMIC contexts). For example The PMI index to assess the manufacturing sectors is based on a highly intensive monthly surveying task, but acts only over a limited amount of “carefully selected companies”. Household Final Consumption Expenditure, the main component of assessing GDP, is extremely challenging to estimate even in western economies, a challenge that is exacerbated when moving to developing contexts (where much of the market is hidden, and much consumption not reported or embedded in an exchange economy). Indeed, in under-developed economies the data on national income, per capita income and per capita consumption are often simply not available. Due to lack of statistical data, it is difficult to assess economic development accurately in such countries. In such situations we should actively seek new data streams to complement current estimation techniques.

CDR Micro-indicators

We now briefly consider below potential micro-indicators that are extractable from CDRs (vitally in strict anonymized and privacy preserving fashion) and that deserve exploration as to suitability for use as a form of “proxy” or augmentation for traditional macroeconomic indicators.

New Business Startups

Identification of Sellers/Businesses. The identification of both formal and informal business startups is detectable from mobile money transfer patterns in countries where appropriate systems are available (as is the case in numerous LDCs) - this is particular relevant in regions which have seen high uptake of mobile money (such as m-pesa in East Africa), and may offer unparalleled insight into economic grown and traditionally hidden economies. This is likely more powerful than EO analysis alone, as it reflects change in building footprint usage in areas that are already established infrastructurally.

Employment

Workplace Identification. As shown in the World Bank “Reports for the Development of Origin-Destination Matrices Using Mobile Phone Data” it is possible to identify work locations through anonymized CDR. This in turn provides the basis for a powerful economic indicator in aggregating figures for the size of the current workforce. While holding must potential, this must be “ground-truthed” via a traditional survey, in order to calibrate models

Manufacturing Activity

Identification of Industrial Activity. Simple analysis of network tower activity gives a high confidence estimate of whether SEZ building footprints are actually being used, and the extent of that activity.

Manufacturing Activity

Land use classification: Network activity profiles (i.e. daily profiles) are distinctly different between industrial and residential areas of cities. While EO imagery can be analysed to identify land use classifications CDR also has the ability to do so (or used in unison to validate/improve results), and hence provide estimates of contraction/expansion of industrial footprint

Income & Wages

Mobile Money Receipt Magnitudes. Many workers in developing countries are increasingly receiving wages via mobile money means. This is particularly the case in the informal economy, and as such CDR receipts hold much potential to improve wage indicators.

Socio-Demographic

Population Tower entropy/mobility. The range of movement and mobility of a population can be derived from CDR data, and this is now an established correlate (amongst other with poverty). See http://www.unhcr.org/innovation/wp-content/uploads/2016/11/blumenstock-science-2015.pdf

Socio-Demographic

Geographical Call Entropy. the extent to which the population’s members are “socially mobile”, reflected in the range of their calls is another indicator of socio-demographic progress. If all calls for example tend to be confined with an individual cluster of towers, with little range it is indicative of poverty. If calls tend to traverse districts, regions and even countries, it is likely indicative of increased affluence. See http://www.pnas.org/content/106/36/15274.full

Manufacturing

Commercial vehicles movement. Industrial and Manufacturing transportation has distinct CDR patterns, as goods are driven across country (to external regions) or via ports and other infrastructural hubs. Identification and measurement of the magnitude of these patterns promises to offer a strong indication of manufacturing output.

Employment

Mobile Money Receipt Regularity. consistent patterns are synonymous with regular employment, and as such highly illuminating in the construction of unemployment estimates. Inconsistent patterns may also provide a measurement of the split between formal and informal employment.

Retail consumption / Household Final Consumption Expenditure

Mobile Money Spend . Retail sales indicators are particularly hard to amass, due to the logistics and coordination they require between public and private sector. In areas where mobile money is prevalent (e.g. Kenya, where mobile money transactions reflect 25% of GDP), retail consumption will be a strong correlate.

Hours Worked

Identification of Working Patterns: Increase of employment accompanied with a decrease in hours worked is also said to be an indicator of economic development, with increased wages likely and improved work conditions likely giving a strong sign of economic and social development.

Socio-Demographic

Data Usage. As “Data” usage increases as a proportion of total network usage, this is also as an indication of increased affluence.

Level of Building Permits

When activity in a cell tower area jumps it is a strong indicator of new building - this can be used in combination with computer vision analysis to provide a proxy for the traditional “new building permits” indicator. Balance of Trade / Export Activity International Call levels. At the heart of much SEZ investment is the goal of improving direct foreign investment and in particular export activity. Level of international calls from a region offers a simple insight to the level of such activity.

GDP / Household Final Consumption Expenditure

Micro-loans Frequency/Magnitude. - many mobile companies in LDCs and LMICs are beginning to provide through third party support (e.g. Jumo, a South African company working in Uganda and Kenya for financial services such as micro-loans. Their uptake may contain insight into expenditure, and ultimately HFCE.

Balance of Trade / Export Activity

Airport travel/ Port Activity . Again, a myriad of movement patterns are discernable from anonymized CDR data, providing aggregate values for levels of freight transport activity, especially at major ports. This is likely to give a strong correlate for export activity, and hence insight into balance of trade.

Tourism

Short Term/International Sim Uptake. a clear indication of both tourism and foreign investment in a region, the purchase of short terms sims will also likely be a productive area for investigation.