3rd Webinar of the International Statistical Classifications Sprint

  • Virtual | New York
  • 31 January 2024


The webinar, organized by the United Nations Network of Economic Statisticians and the United Nations Committee of Experts on International Statistical Classifications, will explore the role of technology in making the process of classification more efficient. Practical examples from various NSOs and organizations using international classifications will illustrate this point. Additionally, the webinar will close with a panel of senior executives from organizations around the world who will address some of the biggest challenges faced by organizations when implementing international classifications, such as the frequency of revisions, building and maintaining capacity, modernizing the classification process and more.


Session 1 - Welcome and Opening Remarks

7:00 am - 7:05 am (New York time)

Opening remarks and objectives of the meeting

  • Chair: Vincent Russo, Supervisory Economist, Producer Price Index, Bureau of Labor Statistics; Co-chair, US Economic Classification Policy Committee
7:05 am - 7:15 am

Recap from meeting 2

  • Presenter: Claude Macchi, Eurostat

Session 2 - Role of Technology in the Statistical Classifications Process

7:15 am - 8:40 am

An expert system for semi-automated classification of product and industry descriptions

The matching between international standards and national descriptions of products and industries is a cumbersome but crucial task for data quality, analysis, and visualization. This presentation will show the holistic method developed within the AfCIOT project, which includes SUT-IOT-TiVA models and a dashboard. The method includes an algorithmic expert system in several steps, including automatic classification when possible, and expert manual allocation for the missing matches. The quality of the match is recorded during the process, allowing the stakeholders to trace how the classification was made, and providing an explanation to the model. The method needs two important modules: the preprocessing of all the standards and correspondences available across them, and a seamless user interface for carrying out all the processes.

Caliper - Collaboration to Improve the Interoperability of Statistical Classifications

The ongoing international effort towards the modernization of official statistics is increasingly involving statistical metadata and, in particular, statistical classifications. In fact, current expectations about statistical data include the possibility of easily looking up the exact meaning of any piece of data, and the possibility of smoothly reusing code lists in different information systems. At the same time, data collection is expected to keep up with the fast-evolving world in which we live. Since 2017, FAO has developed Caliper, a platform dedicated to improving the interoperability of statistical classifications. Currently, Caliper serves as the FAO dissemination platform for statistical classifications, and is being considered as the basis for a larger international alliance focusing on both the dissemination and the maintenance of international classifications. In this presentation we will describe the latest developments of Caliper, the progress in improving the dissemination and use of statistical classifications and possible future developments.

Statistics Canada's approach in managing statistical classifications

The presentation will explore the utilization of a robust Statistical Classification Management tool as a pivotal solution in optimizing data governance and improving decision-making processes. As Statistics Canada grapple with an ever-increasing volume and complexity of data, the need for a systematic and efficient approach to managing reference data becomes paramount. The presentation will delve into the key functionalities of a state-of-the-art Statistical Classification Management tool, emphasizing its role in ensuring data accuracy, consistency, and integrity across diverse business applications. By centralizing statistical classification management, the tool enables Statistics Canada to establish a single source of truth, minimizing the risk of errors and discrepancies in critical data products. Furthermore, the presentation will highlight the tool's ability to facilitate collaboration among other relevant statistical data standards.

The use of ML in NAPCS Classification for Retail Scanner Data and the Retail Commodity Survey

This presentation will illustrate how the Retail Commodity Survey at Statistics Canada has benefited from the use of Machine Learning in the commodity classification process. It will also cover the challenges of integrating and maintaining the Machine Learning technology into the survey processing systems, as well as lessons learned.

Classification of business activities by machine learning: The case of France

INSEE the French National Statistical Institute, manages the Sirene administrative business register. In November 2022, Insee introduced a new method to code the economic activity based on the description provided by the firm. This method is based on a machine-learning model using fastText, a versatile and efficient tool for learning word representations and sentence classification. Developed by Facebook's AI Research (FAIR) lab, this open-source library is acclaimed for its swift processing of large dataset. INSEE is updating its current activity classification, NAF 2008 (a more detailed version of the NACE Rev. 2), to the new NAF 2025 (a more detailed version of the NACE Rev. 2.1). The method relies on the firm's activity code as an output label of the model, requiring consistent alignment with both NACE and the current NAF. Retraining the model is essential to predict based on NAF 2025, ensuring the method's longevity. This is critical for efficiently classifying firms in flow formalities and updating the historical database of firms still containing textual declarations. Reclassifying firms inventory using FastText is particularly useful in cases where no direct correspondence exists between the old and new NAF.

Discussion (10 minutes)

Session 3 - Roles and Efforts of the ASEAN Community Statistical System (ACSS) on the Harmonisation of ASEAN Statistics

8:40 am - 9:00 am

This presentation will provide an overview on the efforts to enhance the quality (including comparability) of ASEAN statistics as well as enabling mechanisms, standards and classification currently implemented by ACSS, along with some key challenges.

Discussion (5 minutes)

Session 4 - High Level Panel on Key Issues

9:00 am - 9:50 am
  • Panel discussion
  • Moderator: Ivo Havinga, Consultant
  • Panellists:
  • •  Chettra Keo, Director of National Accounts National Institute of Statistics Cambodia
  • •  Jean Pierre Poncelet, Director Standards; Dissemination; Cooperation in the European Statistical System, Eurostat
  • •  Karin Orvis, Chief Statistician of the United States
  • •  Eric Rancourt, Assistant Chief Statistician, Canada
  • •  Stefan Schweinfest, Director of the United Nations Statistics Division

Concluding remarks

9:50 am - 10:00 am
  • Summary by Vincent Russo, Supervisory Economist, Producer Price Index, Bureau of Labor Statistics; Co-chair, US Economic Classification Policy Committee