Principal investigator: Adriano Valladão Pires Ribeiro
Article title: COLLECTION OF DELINQUENT FINES: AN ADAPTIVE RANDOMIZED TRIAL TO ASSESS THE EFFECTIVENESS OF ALTERNATIVE TEXT MESSAGES
Article authors: Laura Haynes, Donald Green, Rory Gallagher, Peter John and David Torgerson
Location of the intervention: United Kingdom
Sample size: 5450 individuals
Main theme: Economic Policy and Governance
Type of Intervention: Sending a message to a mobile device
Primary variable of interest: Amount paid for outstanding fines
Evaluation method: Adaptive experimental evaluation (RCT)
Policy Problem
The collection of overdue fines resulting from, for example, non-payment of taxes, traffic fines, and criminal proceedings is a major challenge for the public sector. There is a significant cost involved in recovering these amounts. Therefore, in search of a strategy to increase the percentage of outstanding fines collected, the United Kingdom experimented with sending notifications via mobile phone to debtors.
Evaluation Context
In the United Kingdom, one million new fines are imposed each year, but only 50% are collected in the first six months. Recovering the amounts owed is costly in terms of administrative expenses and time. A team is responsible for trying to contact debtors by phone, and if this method fails, a bailiff is assigned to go to the debtor's residence, and may even seize property in some cases. Text messages are emerging as a more cost-effective alternative. Among their advantages are the fact that they can be sent by automated systems and their content can be easily altered at almost no cost, even allowing for personalization by recipient.
Intervention Details
The experiment took place in three regions of southeast England between January and April 2012 and had two phases. The experiment population consisted of people whose mobile phone numbers were held by HMCTS (the agency responsible for collecting outstanding fines) and who had failed to pay a debt to the point of authorizing the seizure of assets by bailiffs. At the beginning of each week, a list with the names of new debtors from the previous week was collected, and only those cases where the agency had the phone number were kept; the remaining names were discarded. In the first phase, 1817 people were selected, while in the second phase this number was 3633 debtors, totaling 5450 individuals.
With the list in hand, HMCTS sent an individual text message to the participants' phones. The content of all messages reminded them of the debt, warned of the consequences of not paying it, and instructed the person to call a helpline with a reference number. The types and messages sent are in Table 1. If the message did not reach the recipient, the sender was notified, and these messages were classified as undelivered; 54.5% of the 5084 messages sent were in fact delivered. Finally, if the amount was not paid within 7 days of notification, a bailiff was assigned to the case.
The motivation for this type of experiment stems from three factors. First, there is a clear indication that something has been breached, that is, the outstanding amount has not been paid by the agreed date and would thus have a greater chance of punishment. Second, the message text required immediate action with a phone to make a call and warned that a bailiff would be assigned if payment was not made. Third, in some cases the message was personalized with the name and amount owed, signaling the government's ability to use information that could lead to a penalty.
Table 1: Message associated with each condition
| Condition | Message Sent |
| No | No text message was sent. |
| Sensors | You haven't paid your fine. Pay immediately or a summons will be served to a bailiff. Call [phone number] ref [number] div [number]. |
| Personalized Name | [Name], you have not paid your fine. Pay immediately or a summons will be issued to a bailiff. Call [phone number] ref [number] div [number]. |
| Customized Value | You have not paid your fine of [amount]. Pay immediately or a summons will be issued to a bailiff. Call [phone number] ref [number] div [number]. |
| Personalized Nominal Name and Value | [Name], you have not paid your fine of [amount]. Pay immediately or a summons will be issued to a bailiff. Call [phone number] ref [number] div [number]. |
Exclusive
The experiment consisted of two phases. In Phase 1, between January and the beginning of February, individuals were divided into 5 groups with equal probability: a control group, which received no message, and groups with different treatments, where the message sent could be the standard one, a personalized one with the name, a personalized one with the amount due, or a personalized one with both name and amount due. This phase primarily allowed us to verify whether text messages, given any treatment, would be effective compared to not sending a message. Since individuals were randomly allocated to each group, the difference in the experiment's results could be attributed to the sending of notifications. Phase 2, between February and April, excluded the control group from the sample and aimed to obtain a better evaluation of the effect of each type of message.
The treatment effect could be measured by comparing the difference in the average amount paid by people who received each type of message. However, there is a complication in the experiment. Even though individuals were randomly allocated to each group, the message might not arrive and be classified as undelivered. Therefore, the observed treatment effect is not directly captured by the averages. The average measures the effect among those intended for treatment, that is, among all those assigned to treatment regardless of whether the message was delivered or not. Another effect is on those who underwent treatment, that is, only among those to whom the message was delivered. This impact is obtained by dividing the average of each group of those intended for treatment by the fraction of people assigned to each group who actually received the messages.
Finally, it should be noted that the variable analyzed was the amount paid by each person, so the vast majority of individuals in each group paid nothing, while a small portion paid amounts ranging from small to higher.
Results
1 phase: The group that received no message paid an average of £4,46 (in pounds sterling). The average for the group with a personalized message was £12.87, a 189% increase compared to the control group. Personalization with the amount due generated an average payment of £10,53. Finally, the personalized message with a value had an average of £11,74. When only the people to whom the messages were delivered are considered, the average amount paid by each group varies to £7,43, £15,14, £11,57, and £12,68, respectively for the standard group, personalized message, personalized value message, and personalized message and value message groups. These results show that sending a text message has a positive effect on the payment of an overdue amount; all groups showed a higher average compared to the average of those who did not receive a message.
2 phase: Having established that sending a text message with the payment request is effective, the next step is to explore the best type of message. To do this, the group that would not receive a message was excluded, and all individuals were then classified into one of the other four groups. The averages for the standard, nominal, value, and nominal and value groups were, in order, £8,34, £11,21, £8,82, and £9,68. After adjusting for the people who actually received the message, the averages rose to £15,40, £20,58, £15,83, and £18,34, respectively. Again, the personalized message with only the name had the highest average amount paid.
Lessons in Politics
The experiment presented demonstrates that it is possible to improve the efficiency of government agencies using less costly means. Sending a text message to debtors with outstanding fines proved effective in collecting the amount owed, especially when the text is personalized with the person's name. Following this procedure, the amount collected from overdue fines would be almost triple compared to not sending any text message.
Reference
Haynes, Laura C; Green, Donald P.; Gallagher, Rory; John, Peter; Torgerson, David J.. “Collection of Delinquent Fines: An Adaptive Randomized Trial to Assess the Effectiveness of Alternative Text Messages,” Journal of Policy Analysis and Management, vol. 32(4), pages 718-730, September. 2013.
