Meeting

WP5 meeting, Brussels, 20 July 2016

Attendance

• Present: David Salgado, Albrecht Wirthmann, Marc Debusschere, Françoise Dupont (WebEx), Pasi Piela (WebEx), Fernando Reis (WebEx), Monica Scannapieco (WebEx)

Agenda

1. Preparation of the Workshop on Public-Private Partnerships for Mobile Phone Data for use in Official Statistics, Luxembourg, 22-23 September 2016

David Salgado, as WP coordinator, informs about the current state of the organization of the workshop:

• All invitations have been dispatched. The list of invited MNOs will be distributed among the WP co-partners.
• 8 confirmations have been already received.
• STATEC has accepted to take charge of the logistics (venue, coffee breaks of both days and walking lunch for the first day). Practical info will be requested to disseminate it among the participants.
• It is agreed to use the wiki as the dissemination platform.
• It is agreed that both the preliminary analysis of the diverse aspects of the access to mobile phone data and the responses to the questionnaire submitted to the ESS will be relevant as inputs for the workshop.

The preliminary draft agenda was approved and each item was revised to debate and agree on the contents.

• Welcome address and initial presentation.- It is agreed that Eurostat, Statistics Spain (as WP coordinator on behalf of the ESSnet), and STATEC (upon their approval after consultation) jointly address the participants under a coordinated message offering the appropriate context for the event.
• Digitalisation of human activity and Official Statistics.- It is agreed that Eurostat makes an initial 30-min presentation to convey two first important ideas, namely (i) the importance of Big Data for Official Statistics as the example of digitalisation of human activities and (ii) the statistical problems lurking in their use for the production of official statistics as the example of the Google Flu Trends shows.
• Examples of on-going initiatives.- It is agreed that the collaborations Orange France/INSEE, Positium/Statistics Estonia, and Proximus/Statistics Belgium will offer a wide and deep overview of the potentialities of the partnership between MNOs and NSIs. It is urged to have the titles of the talks as soon as possible.
• Round table on lessons from these experiences.- Depending on the final number of participants, since we want to receive feedback from as many MNOs as possible but making the round table as agile as possible, we consider two options: either to select some MNOs and 6-7 aspects of the collaborations to debate about or to select just 3-4 very relevant aspects of these collaborations and make it open targeting to collect as many opinions as possible from everybody. The details will be carefully prepared by the members of the WP and Eurostat. In any case, we will need a moderator/facilitator and we consider STATEC's director of methodology as a potential actor in this role (a consultation will be made). It is remarked that the round table needs to be opened with an adequate summary to put the debate into context.
• Brief summary of sessions I and II.- It is agreed that Spain starts the second day with a brief summary of the main conclusions so far.
• Aspects in accessing mobile phone data.- It is agreed to include two twin talks on the diverse aspects of the access to mobile phone data, namely, one for the perspective of Official Statistics and other for the perspective of the MNOs. Spain, as WP coordinator, is considered the best choice for the first one while Telefonica is considered upon their acceptance as an excellent speaker for the second. These talks will be jointly prepared (still waiting for concrete preparation meetings due to summer holidays period). It is agreed that the exhaustive list of aspects prepared for the first deliverable of the WP will provide a good skeleton for these talks.
• Tour-de-table on these aspects.- It is agreed that we must target to listening to as many MNOs as possible regarding all these aspects (at least most of them). A tour-de-table arises as most appropriate. We agree on the proposal of consulting DG Connect to find a moderator/facilitator for this tour-de-table.
• Joint conclusions for the future.- This session must collect the main conclusions and above all prepare the next steps to establish long-term collaborations. It is agreed to propose the creation of a working group involving both NSIs, Eurostat, and MNOs to jointly find solutions and agreements upon the preceding issues found during the workshop. Apart from formalities about this group (taks force, working group, etc.), it is recognized that the creation of this collaboration group would be an important step forward.

Complementarily, we agree on:

• sending a reminder for the registration together with the preliminary agenda and an explanatory note to reinforce the views expressed above;
• sharing the list of invited MNOs and the list of confirmed attendees;
• organizing an informal pay-your-own dinner for the 22nd September for all participants (at their will);
• sharing with participants the lists of issues to tackle in the round table and tour-de-table so that they will be ready for providing feedback.

2. Questions/remarks on the draft proposal for the work plan of WP5 for the SGA-2

David Salgado, as WP coordinator, makes an overview of the proposed work plan for the SGA-2 for this meeting. The proposal revolves around three main poles:

• Concrete issues regarding official statistical production:
  • Statistical methodology, especially, (i) the novel use of machine learning techniques upon mobile phone data, which ultimately are proxy variables, (ii) the inference problem from mobile phone data sets to the population of analysis (the general population, the population of tourists, the population of commuters, ...), (iii) the computation of accuracy as the (almost) most important aspect of the quality of the results.
  • Technology requirements, especially, the potential need of distributed computing, the need of programming languages (Python, R, Julia, ...) and any other aspect regarding computer technologies arising from the use of this information source.
  • Quality assesment, especially, (i) the assessment of the product obtained with the analysis with the current quality framework, in particular regarding accuracy, and (ii) the complementary assessment of the current quality framework for this new statistical product.
• Concrete statistical output:
  • Although the approach must be bottom-up, the focus is on the process and not on the final product. The statistical output is a tool to assess the former aspects of the process. As statistical output we will focus on the distribution of the population across the terrritorial and time grid agreed upon by the WP in a preliminary task of SGA-2.
• Concrete working approach:
  • The goal is not to reproduce what research centers and universities have already done, but to reproduce as much as possible actual production conditions in NSIs to investigate how the statistical production process is affected by this information source. In this sense, we propose to simulate the continuous production and update of the distribution of the population upon the periodical collection of data (say, every day, every week, ...).

The methods to do that in practice seems to be a bit unclear at this point and David Salgado clarifies the following:

• The proposal is detailed enough from the production methodological point of view to satisfy the SGA-2 requirements explicitly demanded by the ESSnet coordination during the Tallinn meeting. At the same time, it offers enough room to each country member to develop the different tasks in the proposal to pursue the different national interests (tourism/mobility/etc.) and diverse access restrictions to the data.
• To exchange findings, ideas, doubts and the work in general it is suggested that we follow the generic approach in scientific work, i.e. to compose draft or working papers including the main findings and ideas so that we all can debate about them, to physically meet (two physical meetings are planned for the SGA-2), to use Webex facilties as much as possible, and standard tools (as the email).

List of actions