Chair: Krzysztof Jajuga
Room: S3B Sukiennice
Time: 11:30 - 13:00
Date: 28 June
Session 20 - papers & presentations
|Title: <<< Quality of official statistics in micro and small economies in globalized world >>>
The majority of economies and countries in the world of today (over 100) are micro and small economies (MSE). In globalized world the prerequisite of social and economic policy of governments of those economies and countries is the availability of complex, reliable, relevant and pertinent information adjusted to the specificity of each particular micro and small economy and relevant integrated data on their political, social, economic and ecological environment. The production and dissemination of such information for governments, social organizations and businesses is the duty of national systems of official statistics of the MSE. Micro and small economies are insisted by international organizations to submit statistical data following exactly the international standards that are adjusted to the specificity and the needs of large countries. However, those standards, methods, categories and data are often difficult for compiling, hardly interpretable or even useless for MSE. The objective of this paper is the analysis of social and economic attributes specific of micro and small economies that have the impact on information sources, methods and needs of national users of statistics. The typology of micro and small economies is proposed. The consequences of the specificities of different types of the MSE for the duties, functionalities and strategies of development of official statistics and problems of statistical capacity building for the MSE are discussed. In the conclusion it is formulated the need of creating the international, joint research institute providing scientific and educational assistance to national statistics and educating higher level statisticians (MOS - Masters in Official Statistics) for all micro and small economies.
|Cristina Pereira de Sá |
|Title: <<< The future role of international organisations: enhancement of relevance and quality of official statistics >>>
The authors analyse in the paper the future of official statistics, focusing in particular on how the international organisations through changes in governance or standard setting could contribute to increasing its relevance and quality. Official statistics are faced nowadays with a multitude of challenges (e.g. shrinking resources, improvement of timeliness, need to give responses of greater relevance to the specific requests of users, use of new data sources) and have to provide an effective answer through high quality official statistics in many domains at global level (e.g. SDGs, globalisation). Under these rapidly changing circumstances, it will however be essential to enable Official Statistics to play their important societal role through appropriate adaptation of the rules, the principles and resources, which frame their working conditions. Enhanced cooperation and communication between the main stakeholders at global level is essential for improving the credibility and relevance of official statistics. But in order to be able to face the current challenges the global statistical system needs to go a step further. The authors will discuss the feasibility of the application of the European Statistical System model to the global statistical system by considering the application of the ESS standard setting model and the transfer of aspects of the ESS governance and cooperation structure to the global statistical system. Is this possible without an underlying legal framework as it exists in Europe? Is there room for strengthening the agreements at global level and establish new cooperation mechanisms? In this context the authors will argue for a change in international statistical governance and address issues such as the new role of international organisations.
|Title: <<< Conceptualising quality for big data >>>
Whereas the importance of big data for official statistics is widely recognised, the quality of the big data is a concern and the statistical community has soon started to reflect upon a framework to assess their quality. For instance, the United Nations Economic Commission for Europe (UNECE) Big Data Quality Task Team (UNECE 2014) extended an administrative data quality framework; the AAPOR total error framework for big data (Kreuter, 2015) focused on an extension of the total survey error examining sources of errors specific for big data. Reis et al. (2016) have analysed a few alternative approaches on real case applications, concluding for the need to complement the different approaches and to structure the links between input and throughput quality and output quality. In the Essnet Quality of Multisource Statistics, a framework relating the output quality with the sources of errors is being proposed (Brancato, 2017). The proposal takes into account the main quality models for administrative data. This paper aims at investigating the applicability of the Brancato approach for big data and proposes the necessary extensions taking into account of the quality frameworks for big data. Finally, very preliminary thoughts on the relevant factors when producing experimental statistics are given.
- Brancato, G. (2017) Unpublished technical report WP1. Guidelines on the quality of multisource statistics, Essnet on Quality of Multisource
- Kreuter, F., Berg, M., Biemer, P., Decker, P., Lampe, C., Lane, J., ... & Usher, A. (2015). AAPOR Report on Big Data (No. 4eb9b798fd5b42a8b53a9249c7661dd8). Mathematica Policy Research.
- Reis F., Di Consiglio L., Kovachev B., Wirthmann A., Skaliotis M. (2016) Comparative assessment of three quality frameworks for statistics derived from big data: the cases of Wikipedia page views and Automatic Identification Systems, Q2016
- UNECE (2014) A Suggested Framework for the Quality of Big Data
|Title: <<< A knowledge-based approach to the statistical production process >>>
The debate on data quality in official statistics has recently shifted its focus from an output oriented to a production process perspective. This paper explores the link between the design and management of statistical production processes and data quality measurement and evaluation in the light of contributions from the knowledge economy and knowledge management literatures.The classification of statistical production either as a knowledge transfer or as a knowledge intensive process is crucial. In the first case, the knowledge embedded in data collected from respondent units is shifted to final users with no alteration in data processing, while in the second case the knowledge is transformed to generate more consistent, accurate and business relevant data. The knowledge economy literature tends to acknowledge that any production process is knowledge based, since the knowledge embedded in raw inputs (being goods or services) is either explicitly (codified knowledge) or implicitly (tacit knowledge) transformed into the final output by the human capital and technologies engaged in the process. The aim of this paper is twofold. Firstly, it highlights the key features of statistical production as a knowledge intensive process, also providing some concrete examples in the broad domain of globalisation, including recent development on profiling, establishment of large case Units and EU level early warning system. Secondly, it exploits the consequences of this approach on data quality with respect to different features of business survey management: questionnaire design, burden reduction, design of a statistical production process when knowledge is explicitly considered in the business model. This approach is important to enrich the output, to strengthen the consistency of official figures as well as to reduce the burden on respondents in terms of knowledge distance. It also calls for substantial changes in the way people are trained and work is organised in National Statistical Institutions.
|Title: <<< “Show me your code, and then I will trust your figures”: Towards software-agnostic open algorithms in statistical production >>>
This contribution aims at further promoting the development and deployment of open, reproducible, reusable, verifiable, and collaborative computational resources in statistical offices regardless of the platform/software in use. Motivated by the consensus that data-driven policies should be transparent, we argue that such approach is not only necessary for the practical implementation of statistical production systems, but also essential to reinforce the quality and trust of official statistics, especially in the context of a “post-truth” society. With the advent of open source in the scientific community, as well as other disrupting technologies and software emerging from data science and similar active communities, many statistical organisations are nowadays considering to introduce new software solutions in production environments so as to benefit from the many statistical libraries, advanced algorithms and innovative developments available. Owing to legacy issues, there is however a trade-off between the risks linked to business continuity and reliability against the need for efficiency, innovation and cost-effective solutions. We devise some practical requirements to gear the continuous and flexible development and deployment of (open and free, but proprietary as well) software production components before any solution being adopted. Together with some best practices derived from the open source community (version control, automated unit tests, generic documentation, continuous integration, and collaborative development) we propose to unleash the social power of open algorithms so as to create new participatory models of interaction between "produsers" (statisticians, scientists and citizens) that can contribute to a more holistic and extensive approach to production systems. Overall, a greater transparency in designing production processes is expected to result in a better grip on the quality of the statistical processes involved in data-driven policy-making. We illustrate this flexible and agile approach with various open, stand-alone software or source code used in the actual production of official statistics at Eurostat.