I thought that it would be useful to put the data that we are here to discuss and the points that you have listed into a bit of context, in terms of both the statistics and the processes that are used to collect, assess and publish those data.
A range of data collection exercises are undertaken on education throughout the year. A number of those are census-type collections, which are conducted in parallel over the autumn. We are at the end of the 2018 collections across a range of things. Although all those collections are related and happen in parallel, they are distinct exercises and have important differences between them. No doubt we will come to some of those throughout our conversation.
Our primary sources of data are the pupil census and the teacher census, which are probably the most well-recognised collections across the system. We also conduct the non-teaching school staff census from which the support staff data come, the primary school class data collection and the school establishment collection. Every two years, we collect data on attendance, absence and exclusions in schools.
A range of other collections are carried out throughout the year, including collections on school meal provision and physical education provision, and a range of data exercises relating to pupil performance and outcomes.
The autumn administrative collections are supplied to the Scottish Government directly by local authorities and, in some cases, by schools, particularly grant-aided schools. The quality assurance process starts with their initial collection of the data, which they use for the day-to-day running of the education system. They hold those data on their management information systems for regular use.
Quality assurance is built into each stage of the process, so schools, local authorities and the Scottish Government all have a role to play in that process. The precise arrangements vary a little bit between the collections, depending on the nature of the data and the amount of detail that we collect. Ultimately, the data are signed off by directors in local authorities prior to being published by the Scottish Government.
During those administrative data collections, we collect and process what amounts to tens of millions of pieces of data relating to 2,500 schools, around 700,000 pupils, 70,000 teachers and more than 20,000 support staff across the system. That enables us to publish more than 100,000 statistics. In addition, we release bespoke data sets and analyses throughout the year for a wide range of users, including researchers and academics, the media, the general public, politicians and parliamentary committees.
The code of practice for official statistics sets a framework for our approach to handling data and producing statistics. It is there to ensure the quality, value and trustworthiness of the statistics that are produced by organisations such as the Scottish Government, and it provides us with guidance on specific aspects of producing statistics, such as ensuring that the burden on data providers is proportionate and that appropriately qualified professionals are used throughout the process. The ultimate responsible for ensuring that the code of practice is adhered to in the Scottish Government rests with the chief statistician.
The purpose of the administrative data collections in education, as with all other sectors, is to describe—it is to paint a picture of the subject to which the data relate. They can only ever play back practice in the system or paint a picture of the system. They do not and should not define or constrain practice in a system. However, they do and should facilitate debate and discussion about practice in the system. Any statistical collection or publication does not exist in isolation or for its own sake; it needs to remain relevant, reflecting current or likely future situations, while ensuring that as robust and accurate a picture as possible can be presented.
What we publish and how we publish it needs to be a balance between a number of factors. Those include known limits or restrictions to the data, previous practice on publication, public interest, current issues and the context in which we are presenting the data, and the availability of other sources of information on the same or similar topics. Our approach is generally to publish as much information as possible in as accessible a format as possible, without releasing data that we know to be misleading, incomplete or erroneous.
Education data are probably more in the spotlight now than at any other time in recent memory. That brings a requirement to continually improve their fitness for purpose; it also more often brings us closer to the practical limits of the power of the data that we collect. Therefore, simply replicating what has gone before will not always provide the most useful or accurate data, and it is right that we take action to investigate and address any anomalies in the data or issues raised by data providers throughout the process.
No decision to make changes to the presentation of official statistics is taken lightly. In this case, the requirement to reflect on-going changes in the relative importance of support staff data, as against other elements of education data—teachers and class sizes, for example—meant that a deeper examination was warranted when a potential issue was highlighted to us. The resulting changes provide a more reliable data set than would have been the case if we had left those unaddressed.