In the broadest sense, business statistics are figures relating to enterprises, including the agricultural sector and institutions. ‘Six years ago,’ recalls Ger Snijkers, CBS researcher and project leader, ‘the American Statistical Association asked me to compile a book on the occasion of the sixth International Conference on Establishment Statistics (ICES).’ The conference was scheduled for 2020, but as a result of the pandemic it was ultimately held in 2021. The book was to be the successor of the volume published following the first-ever ICES conference, almost 30 years ago. CBS researcher Arnout van Delden was also on the editorial team. ‘A lot has changed since then. Thirty years ago, we mainly based our statistics on sample surveys. Now we have all kinds of new sources. That requires new techniques and methodologies in compiling statistics.’
From data collection to output
The book consists of seven sections that cover the entire production process. Snijkers is clear: ‘We took the targeted statistical output as our starting point: which statistics are really needed? Currently, we are in a time of significant change in statistical output, for example with the new demand for figures relating to things like climate change and globalisation.’ Later sections deal with the production process of statistics, secondary sources such as government registers, survey data collection and the use of new sources like web scraping and big data. Finally, the actual generation of statistical output is treated in a section on sampling and estimationas well as a section on data integration.
Globally relevant
As Editor-in-Chief, Snijkers led an international team of eight editors including section editor Van Delden. The book includes contributions from more than 100 researchers from all over the world, from countries as diverse as the United States, Canada, Australia, Nepal and Indonesia. ‘We wanted to make sure the book truly was globally relevant,’ Snijkers explains, ‘so we invited contributors from non-Western countries that are generally under-represented. Sometimes those countries have totally different statistical production issues.’ Van Delden adds, ‘Even before the conference began we were off to a flying start, selecting topics and approaching writers. That made it possible for us to present a first draft of the chapters just a few months after the conference.’
In addition to all the new developments, the book also covers more conventional topics such as questionnaire design, probability sampling, sample coordination, and how to deal with missing data. As well as editing his section, Van Delden also co-wrote a chapter on administrative data. Almost every national statistical institute uses those data, as most of those offices have a policy that prioritises the usage of administrative sources. One problem with administrative data is that they involve dependence on external organisations. ‘Statisticians have a certain understanding of what is meant by an enterprise,’ Van Delden explains. ‘External sources’ definitions of what an enterprise is may not be the same, so it’s important to investigate how data from a variety of sources, such as Tax Office data and a business register, can be aligned.’ According to Snijkers, ‘In case of a survey, the statistical office decides whom to approach and exactly which questions to ask. Of course, that is not the case when a statistical office uses administrative data, so you always have to ask yourself: how can I turn this into a usable source?’
Real-time statistics
Given the rapid pace of developments in the field, it is impossible to predict whether this brand-new book will stand the test of the next 30 years. As Van Delden acknowledges, ‘The data revolution has only just begun. Artificial intelligence is another catalyst for change, for example if we can use it to search for business data on the internet.’ Snijkers shares his view: ‘I believe, in the end, we’ll move to (near) real-time statistics. As things stand, it can sometimes take a year to produce a statistic, for instance for an annual sample survey. IT is going to change that – but we need to remember that, although technology enables the collection and processing of all kinds of data in a timely manner, whether or not we can produce accurate statistics with use of these technologies depends on the development of suitable methodology.’
More data-minded enterprises
In the book, the authors also discuss the fact that enterprises are collecting more and more data about their own processes. ‘More and more enterprises are becoming data minded,’ Snijkers observes. ‘We have to start working out how best to make use of that fact right now, because in ten years’ time it will be too late.’ Statistical offices will have to develop new techniques to keep up with the data available within enterprises, as well as new methods to ensure that relevant outcomes from those data can be published. ‘Trust is essential: enterprises must have confidence that statistical offices handle their data properly.’
All knowledge bundled
‘Advances in Business Statistics, Methods and Data Collection’ is useful not only to statisticians around the world, but also to researchers at national banks, economists, students and their teachers. ‘You can select the parts that are of interest and relevance to you,’ says Snijkers. ‘Nowhere else can you find so much knowledge about business statistics, all in one place. It fills in a gap in our knowledge base. I was honoured that the American Statistical Association asked me to compile this book, and I’m very proud of the result the whole team has achieved. The content is in line with what we envisioned when we started, but it couldn’t have been published in this form without the contributions of all authors.’
