News 29 Aug 2019
All pieces have to fall in place in quality assurance of statistics
The quality of statistics has to be honed to perfection in order for the data to be reliable and usable. It takes more than one trick to ensure the quality of published data. Every step of the process from planning and selecting the best data sources to revision of data, publication and continuous development work has to be in order, Statistics Finland’s experts say. Thus, what is needed are skilled information and statistics professionals, high-quality data, the right tools and methods.
The starting point for several definitions of quality is that the product must meet the user’s expectations – in the case of statistics, this means, for example, that the end result truly describes the social phenomena it is intended to describe. In addition, it has to be up-to-date and easily accessible. Changes in society have to be taken into account in the content, methods and publication modes of statistics.
“In the future, we will try to produce new data even more rapidly and flexibly”, says Quality Manager Outi Ahti-Miettinen.
Independence and best methods
Quality assurance starts with impartiality, openness and independence: Statistics Finland works independently.
“Statistics Finland operates under the Ministry of Finance and is managed by results, but the Ministry does not intervene in the compilation or publication of statistics. Statistical legislation alone ensures this”, Senior Adviser Leena Storgårds emphasises.
The next step in quality assurance is using the best scientific methods. This is related to, for example, forming the sample, processing and analysing data and correcting possible errors.
“When it comes to surveys, the interviewers are continuously trained, and we make sure they understand how, for example, to prevent the interviewer’s own actions from affecting the results”, says Ahti-Miettinen.
“We also have long traditions of how sample surveys can be made as representative as possible”, she adds.
Many methods to hone data
All data are checked before publishing. Material provided by enterprises and authorities can also sometimes contain errors or have missing data, when the data masses are big. For example, data concerning enterprises must be combined with Statistics Finland’s data frame using the Business ID. If some of the IDs’ are missing from the received material or have been entered incorrectly, the right enterprise must be combined with the data concerning it by other means, Project Manager Antti Santaharju describes. On the other hand, sometimes the data on, for example, turnover has by mistake been entered in thousands of euros, when the correct form is euros, and this mistake must be corrected.
Exceptionally large or small figures that should be revised can be found in the material with statistical methods, for example, based on probability. Missing data, in turn, can be searched for in other sources.
According to Santaharju, several of these corrections can increasingly be done automatically in the future.
Support from criteria and legislation
Quality work is also based on legislation and different quality criteria and ethical principles.
“The Quality Criteria of Official Statistics of Finland used by Statistics Finland are based on regulations and we are committed to fulfil them”, Ahti-Miettinen emphasises.
Statistics Finland is also bound by the Fundamental Principles of Official Statistics defined by the United Nations Statistics Division and the directive on European statistics. Statistics Finland’s principles of professional ethics comply with the Declaration on Professional Ethics of the International Statistical Institute (ISI). It contains the ethical principles obliging the statistical field with respect of society, customers and partners.
Evaluation and openness support confidence
In order for quality not remaining only a proclamation, quality is evaluated and assessed continuously in various ways. Statistics Finland internally audits around eight sets of statistics per year – when Statistics Finland’s own experts evaluate the work done by others and the work processes, obtaining information about how the processes function.
The Statistical Office of the European Communities, Eurostat, on the other hand, carries out audits of individual statistics, to supervise that the statistics in question is compiled according to regulations.
“Every five to six years, a peer review of the entire European statistical system is carried out with the help of evaluation teams. At that time, also the media, research institutes and ministries are interviewed, and it is a huge effort. The next round is scheduled for 2021 to 2022”, Ahti-Miettinen adds.
When the quality is in order, there is still one step to take so the public can rely on the quality of statistics: openness. Therefore, the data sources, methods and procedures of statistics are published in the quality description attached to each set of statistics, often also in the even more detailed methodological description. Possible changes to the statistics made over the years are justified. The operating principles and, naturally, also the legislation pertaining to Statistics Finland are also public.
The other side of the coin is data protection. Data on individual persons or enterprises are confidential, the protection of them is carefully ensured and they are to be used only for statistical purposes.
And if, despite all this, an error is at some point found in the published statistics or in a chart presenting the results, it is corrected swiftly and openly, Ahti-Miettinen and Storgårds state.