Who will fill in the internet census?
By Chris A Garrington, on 19 March 2021
This weekend (Sunday March 21, 2021) we will all be asked to fill in our census forms. There’s a key difference this time: the Office for National Statistics, which runs the operation, is aiming for at least 75 per cent of our returns to be submitted online, and the early signs are that millions of people have already responded.
What difference will this make? We do have some information, because the option to complete the Census online was also offered in 2011 – around one in six people opted to do so.
And why does it matter that as many people as possible fill in their returns? I help to run the Office for National Statistics Longitudinal Study, a huge resource based on a one per cent of Census returns since 1971: more than a million individual records. Access to the Census data allows us to build a rich picture of people and places in England, to help us look at what’s happened over time and to anticipate what might happen in the future. So it matters a great deal.
It matters that we get as many people completing online as possible, because that will help the data to be gathered and processed more quickly than has been the case in previous censuses.
And it also matters because we know that without targeting, some groups could be under-represented in a Census which takes place mainly online. Prior to the pandemic, there had been plans to hold events – in schools and local libraries for example – to help people complete the census online if they couldn’t do it at home. Obviously now that can’t happen. The census agencies have provided paper forms to all people in areas that think have poor internet provision, or who are less able or willing to complete online. But there’s obviously a risk that some people might be confused, and may not realise how they can request a paper form.
Previous ONS analysis has shown us that younger people and those living in larger households are more likely to submit their response online, along with those living in large urban areas compared to rural ones.
The ONS analysis was cross-sectional, but the ONS Longitudinal Study (LS) allows us to explore these questions further, by taking a set of sample members who previously completed a census in 1991 and examining their response mode in 2011. Why should we do this when there obviously wasn’t an online option before 2011? Well, one of the things that we were interested in was the question of whether people who have missed a census might be more likely to return to it when there is an online option.
By starting with people who completed a census in 1991, and also in 2011, we know that they were definitely present at both points. Some of them did not complete a census in 2001. There are various reasons why that might have been the case, and we can’t know, but we suspect that for some it was simply a lack of interest, a feeling that the census wasn’t going to be of any benefit for them.
We were able to analyse a sample of just over 360,000 study members who were living in England or Wales in both 1991 and in 2011 and had submitted census returns. We found that online completion of the 2011 census was lowest amongst the over-55s, with 46 per cent of paper returns coming from this age-group compared to 24 per cent of the online returns.
Online returns were used less by those who did not have post-18 educational qualifications. With 13 per cent coming from this group compared to 27 per cent on paper. The online mode was also used less by those in semi-skilled and unskilled social classes, along with those who lived in socially rented accommodation.
There were also differences related to ethnicity, with higher proportions of online returns coming from the Indian, Pakistani, Bangladeshi, Chinese and Black African ethnic groups compared to paper returns, and also a higher proportion of online returns for people for whom English (or Welsh, if living in Wales) was not their main language.
Finally, individuals living in smaller households were less likely to use the online submission route; the online platform is more suitable for larger household sizes as the paper form only allows full information on six individuals to be provided.
We also looked at the number of questions that were imputed (that is, because the original question was left blank by the person completing the census) as a measure of the quality of the census response. Imputations were lower amongst those who completed the form online compared to those who completed on paper, suggesting that the overall response was high quality. This is a promising sign given that the majority of people will complete online in 2021/22.
Finally we looked at whether the online submission route encouraged individuals to submit a census return who might not otherwise have done so, by looking at sample members who were present in 1991 but missed the 2001 Census. This analysis indicated that those who submitted a return in 2001 were more likely to do so again, even if other factors such as age, region, employment status or education qualifications might have mitigated against that.
For those ‘returning’ in 2011 having missed the 2001 census, whilst adults aged over 50 were likely overall to complete using a paper form, this association was not as strong as for their peers who had been present in 2001: the returners had a slightly enhanced likelihood of completing online.
People living in the North West, East Midlands, South East or South West were less likely to submit their return using the online platform in 2011 than those in other regions, with odds even lower if they also did not submit a census return in 2001.
We found that having GCSE or equivalent or a degree were associated with greater likelihood of submitting an online return in 2011.
The ONS will need to visit those who don’t fill in the Census online, in order to ensure that it captures a fully representative sample of the population. We hope that our research will help in targeting the groups most at risk of being missed.