To make sure that the calculated/aggregated mobile phone data provided by the MNO are consistent and can be used to calculate the indicators, quality assurance is required. There is no specific checklist for assessing the quality of aggregated home location information; a logic check is required to answer the following questions inter alia:
- Is the geographical distribution of home locations consistent with official statistics?
- Is the distribution of home locations consistent with the reference data (WorldPop population statistics or NSO population statistics)?
- In areas in which the usage of slow or fast Internet was expected, do the data support the assumptions?
If household survey data exists for the same geographic areas, it is valuable to compare the results from the MPD with official estimates of Internet use collected through household surveys. In Brazil, the results from using MPD was compared with official household survey data from the PNAD Contínua sample Q4 2018. While PNAD Contínua was not designed to produce estimates at such low geographical levels, the results need to be interpreted carefully and only as an additional element to contribute to the discussion on the validity of the study. There are also other differences to consider, including that the MPD study only includes access via the owner's mobile phone, not access by other devices (computers, borrowed phones, etc.). Another difference is that the MPD used was only sourced from one operator which may have a customer base that does not represent the entire population.
Nevertheless, the comparison showed similar results between the MPD study and the household survey data at an aggregated level, with a difference of up to 1.04 per cent in the Rio de Janeiro Metropolitan Area (Table 6). One hypothesis is that the structure of the MPD obtained was similar to that of the mobile phone users, meaning that the MNO have a broad client base which is representative of the population. Further analysis could include examination at a more granular geographical level, with the focus on matching of tower/geographical disaggregation, coverage, and the statistical precision of survey data.
Table 6. Comparison of Internet use estimates from mobile and survey data, March-April 2019
Source: mobile phone data and Q4 2018 PNAD Contínua Survey/IBGE.
In Indonesia, a comparison between the Socio-Economic Survey (Susenas) and MPD revealed greater differences. The estimates from MPD showed higher proportion of Internet users than in the household survey. However, considering that not all 2G phones have Internet access capability, it is possible that the MPD is slightly over-estimated. Figure 11 shows the proportion of the population using the Internet in municipalities on the island of Bali.
Figure 11. Comparison of the proportion of Internet users between survey and mobile data in Bali, January-December 2019
Source: