Councils are constantly seeking to deliver efficiencies and boost income, which makes revenue collection a logical area of focus. At Brent, Social Engine was appointed as our productivity expert, and their expertise was used to design and deliver a randomised control trial (RCT) to test application of behavioural insights to the rent arrears recovery process.
The intervention successfully resulted in increased payment rates those aged 31-40, men and BAME account holders, while those over 70 were less likely to pay their debts. We also found that those with the largest debts were significantly more likely to contact the council following receipt of the trial letter.
The challenge
Rent arrears cause the council to lose out on important income as well as incurring considerable costs in recovering the money – both in officer time and in sending high volumes of reminder letters. Early action to encourage payment or provide financial inclusion support can prevent the stress and hardship this places on tenants, as well as generating efficiency savings for the council. As Universal Credit is rolled out, ensuring that council tenants take accountability for managing their finances and paying their rent becomes increasingly important for both parties.
Our processes are regularly reviewed, and we use a variety of methods to contact our tenants, but at its core, the arrears recovery process is a series of letters and contacts to council tenants to notify them of their arrears and prompt them to either pay, or to contact the council for support. Behavioural insights wasn’t something that had been previously explored, and the council was keen to see how the learning from this might be applied to the full range of revenue collection activities.
The solution
Social Engine were tasked with developing and testing an intervention to encourage council tenants to pay their rent arrears earlier in the collection process. The RCT was applied to the first stage in the collection process, the first notification letter (RA1), on the basis that encouraging tenants to take action at the earliest possible opportunity (whether that was making payment or accessing support) was preferable. This would prevent escalation; reducing the cost of debt recovery and administration as well as helping tenants to avoid the stress and potential hardship arising from falling further into debt.
The intervention was to redesign the RA1 reminder letter to make it simpler and more salient – using accessible language and clearly indicating what the reader needed to do, as well as highlighting the consequences of inaction. In order to determine the effectiveness of the redesigned reminder letter, in comparison with the current reminder letter, we adopted the RCT methodology. RCTs can help to determine the extent to which impact is the result of the intervention, independent of changes that would have happened anyway without the intervention.
Each week all tenants with rent arrears due to be sent an RA1 letter were randomly assigned to either the treatment group or the control group. Those in the treatment group were sent the redesigned RA1 letter while those in the control group received the standard RA1.
Over 2,000 letters were sent out to the treatment and the control group each week, over a 14 week period. Resulting activity was measured at the individual household level, capturing data for each household two weeks after the reminder letter was sent (at the point when the second stage reminder letter would normally be issued). Specifically, we were seeking to identify whether or not each account holder took action to reduce their arrears in order to determine whether the intervention was effective. Taking action to reduce arrears included either paying off the rent owed or contacting the council to request support or establish a payment plan. Not taking action would mean no activity on the account, and progression to the next stage of collection action.
The Impact
The intervention successfully resulted in increased payment rates for specific subgroups of the test population. We found that those aged 31 to 40 were 11 per cent more likely to make a payment when they received the redesigned RA1 letter, while those over 70 years of age were less likely to pay their debts. Both of these results were statistically significant. Slightly below the academic threshold for being statistically significant (but only just below), were increases in payment rates for men and BAME account holders when they received the redesigned letter. We also found that those with the largest level of debt (those with rent arrears of £106 or more) were significantly more likely to contact the council when they received the redesigned letter than those in the control group.
This covariate analysis is particularly interesting, as it demonstrates that the change in letter prompted a different reaction in different groups, giving strong evidence of the need to create a segmentation model of our residents based on different characteristics and payment behaviours, and to tailor our recovery process accordingly. At present, we don’t have systems in place that would enable us to build a sufficiently sophisticated model, and our IT systems only make it possible for us to undertake fairly basic tailoring of our collection process.
Without undertaking significant analysis of payment behaviours, we can only fairly crudely estimate the potential financial gains that could be achieved if we were able to target different communications to different groups in this way. On average, 28 per cent of RA1s result in previously unscheduled payments from tenants (i.e. not standing orders, direct debits or directly paid benefits). These tenants pay, on average, £311 within two weeks of receiving a letter, which would be the point at which action would escalate if they had not paid. The trial achieved an 11 per cent increase in payment rates with the 31-40 age group against the control group. Extrapolated out to all 31-40-year olds within our tenant population, this would equate to roughly £20,000 of debt recovered over the course of the year. This group is roughly 20 per cent of our overall tenant population, so in theory if we were able to identify an intervention that had a similar impact for the remaining 80 per cent in would equate to roughly £100,000.
Looking at the population as a whole, the only statistically significant impact was on participants contacting the council to discuss their arrears; increasing the rate of contact by 27.6 per cent in the treatment group. More detailed analysis of these calls was carried out, but didn’t identify any clear themes such as an increase in requests for support or arrangements to pay. As a result, it is difficult to draw conclusions from this, except that any call made can reasonably be expected to be a better outcome than ignoring the issue.
Given that the criteria for sending the letters didn’t change, this prompted a hypothesis that residents may have become desensitised to contact from the council. This has resulted in the desire to undertake further analysis of recovery activity and payment patterns over an extended period, with the aim of understanding the relative effectiveness of the various activities in the collection process, so that we can build an evidence base for more wholesale changes to our early stage collection process.
The council is part-way through an ambitious digital transformation programme; meaning not only that our digital landscape is undergoing significant change, but also that the skills set within the council is evolving and capacity to undertake “big data” projects such as segmentation and predictive analytics is growing.
While we do not yet have all the building blocks in place to do this, we have begun to embrace more complex analytics. Since undertaking the trial, we have begun developing a prototype rent dashboard that combines some of the characteristics that we expect would be early indicators of those tenants most at risk of falling into significant arrears, so that we are in a better position to focus our resources on the highest risk cases. This prototype should be operational by Spring 2020.
How is the new approach being sustained?
The redesigned letter didn’t increase payment rates across the board, but we felt it demonstrated sufficient impact on contact as to warrant adoption as the standard first stage letter – particularly given the statistically significant impact on those with highest balances contacting the council to discuss their arrears. In addition, the same principles have also been applied to the subsequent communications in the arrears pathway. As described above, the groundwork has also been laid to use more sophisticated analytics, which will help us to identify different groups of behaviours and eventually enable us to tailor our processes to those groups, once the tools are in place to do so.
Social Engine also delivered initial training to a group of transformation and service improvement staff to build knowledge of behavioural insights throughout the organisation. This in turn was used to develop a shorter “lunch and learn” session to introduce the principles of the MINDSPACE framework and RCT methodology, and to give officers a practical opportunity to apply these in a workshop setting to build their understanding of the concepts. This has now been delivered to over 100 staff from a variety of departments, and has anecdotally been applied to range of activities; from increasing responses to consultations, and revamping the way we promote registration and nationality services, to boosting attendance at events for childcare providers in the borough.
Guidance is now in place to help officers apply the framework and design future RCTs. While we have not yet run any further trials, we are giving consideration to potential areas for doing so, notably in relation to council tax recovery, a function which the council has recently brought back in house.
Lessons learned
- Within a varied population, make sure you look at the different groups within it to assess impact – at a global level our data told one story, but looking at the different groups within that population told a very different one. This difference in itself gives powerful evidence for the need to be more sophisticated and tailor our process to maximise it’s effectiveness.
- Take the opportunity to look at the whole process, rather than focusing only on the specific intervention you’re testing. Our review of the letter prompted fresh review of associated processes, such as looking at the customer journey for online payments, which unearthed a number of opportunities for streamlining and improvement, in addition to the application of behavioural insights to the letter itself.
- Initial investigations of the various areas proposed for the RCT (council tax, housing benefit overpayments and rent arrears) highlighted that many of our IT systems made it difficult to twin-track two different recovery processes and follow the impact of these at the individual account level. Working through this took time and resources, but has given us valuable insight into the basic requirements needed to run an RCT without extensive manual input. Not all of our IT systems are supported in house, and where that is the case we can now be clearer at the outset about what is needed, in order to assess viability of future RCTs much more effectively and minimise abortive costs. A range of improvements are being undertaken as part of the council’s digital strategy, which will improve our capacity and ability to undertake analytical work, and will increase the level of sophistication we can apply to revenue collection, making these types of projects a simpler undertaking.
- Similarly, developing tracking reports surfaced a number of issues around data quality and disconnects between how our data is mapped in reporting systems and how it is used operationally. This has led to necessary improvements in our data documentation, which will support our data quality strategy and are a necessary precursor to “big data” projects that we wish to undertake in future.
- Particularly in times of austerity, any proposal to move away from tried and tested recovery methods could be met with hesitance. If the hypothesis that residents have become desensitised to our recovery activity is correct, one potential solution would be to raise the threshold for commencement of recovery activities and seek other means of notifying residents to their account balance before this point. RCT’s and / or data analysis can help to provide the evidence base for decisions such as these, but will require some resource investment up-front; it’s vital that this is understood, and that you have senior leadership commitment to this if these types of projects are to be a success.