Most national statistical offices conduct a household expenditure survey or its equivalent. Data from these surveys are used in important programs such as the System of National Accounts and the Consumer Price Index and allow for the analysis of the determinants of household expenditures, including income. It is well known that measuring household expenditures using traditional survey methods has significant challenges, both on the input side as on the output side. Respondent burden is significant, and the conversion from input items to classified products is a challenge. 

While new methods such as Optical Character Recognition (OCR) and automated classification are more common, the gap between respondents’ input and data sets of good quality that meet the users’ needs has to be bridged. This invited session aims to bring together methodologists, data scientists and survey managers from national statistical agencies to present research and development work that has been recently completed, is underway, or is planned. Topics such as modern and innovative data collection tools like mobile applications, data processing involving machine learning algorithms and the use of alternative data sources will be discussed. At this time, an informal commitment has been obtained from four agencies (official presenters to be confirmed) and is subject to country or agency travel restrictions: 

- Office for National Statistics (UK), Joanna Bulman 

- Statistics Netherlands, Tim de Jong 

- Statistics Canada, Émilie Mayer 

- Office for Labor Statistics (US), Missy Pollock 

- Office for Labor Statistics (US), Nikki Graf 

In addition to sharing the experience of large statistical agencies, this session would offer an interesting geographical coverage. It would also provide an opportunity to learn more about the app used in the Netherlands where the receipt scanning pipeline from OCR to automated classification is presented. This pipeline is also adopted by other European countries. With inflation rates rising rapidly and at historically high levels, the pressure will be on these programs to provide quality expenditure estimates in order to derive reliable basket weights. Exchange of information among these agencies and with the statistical community would be relevant and timely.