The Office for National Statistics (ONS) plays a vital role in British life. Without ONS statistics, government and local authorities would not be able to calculate or understand inflation, immigration, or employment reliably, nor could government design and implement effective policies to manage those issues.
Statistics determine how public funds are allocated and spent in communities, and inform countless decisions in our personal lives, from whether and where to go to university, to when and where to buy a house, and it helps us build a picture of how our population compares internationally.
Much of this information comes from Britain’s census, the once-every-ten-years survey of all UK households.
The delivery of the census is one of the largest public facing activities engaged in by the UK Government, and households that do not complete quickly can be a significant cost driver. Thousands of field workers are employed to follow up on each household that has not completed, which means that every field visit that can be avoided saves money.
Over the course of the last 9 months, BIT have undertaken an initial series of randomised controlled trials in collaboration with the ONS, aimed at helping to improve processes and the speed and quality of data collection.
The next census is in 2021, and an ONS priority is “designing with more cost effective methods of non-respondent follow-up”.
As part of its preparations, the ONS ran a 2017 test to trial new language, response modes, and methods of outreach. Unlike the official census, the test was voluntary, and it oversampled “hard-to-count” populations.
BIT worked with ONS to maximise response rates as part of the 2017 Census Test.
In one trial, we aimed to maximise Census Test response rates by individuals who had persistently not responded and therefore received a follow up visit by a member of the census field force.
BIT created two new versions of the calling card to be tested against the calling card designed by ONS. Both BIT versions made use of implementation intentions and prompts to help recipients plan when they would complete the census, these designs were informed by a previous trial run by BIT in Denver.
Treatment 1 – Implementation Intentions Calling Card
During the course of our exploratory work, field workers told us that they often personalised materials left at households, which they felt was very effective. To test this, in our second version we left a space for field workers to sign their name and added some personalised language.
We found that both versions outperformed the control group, doubling the rate of response within 48 hours. However, the response rates in personalised treatment were not significantly higher than for the other redesigned calling card.
Effects of the Interventions on Likelihood of Household Completing the Census Test within the Next 48 hours
These results suggest that increasing the use of communications informed by behavioural science could lead to cost savings for the ONS.
We are currently embarking on the second phase of our work with ONS, and look forward to reporting further on this collaboration.