Behavioral Interventions to decrease dangerous driving in Armenia

There are two types of interventions envisaged.
First one utilizes the system of notifying the drivers about traffic speeding rules violations by providing additional information to them (both paper based and electronic). There are 5 active treatment arms which differ by the messages and behavorial biases they are targeted at. In addition, the subjects are separated into 3 different blocks depending on previous history of speeding violations.
Second intervention is a stand alone SMS based communication with the registered drivers which provides an update about the traffic violation points preserved/lost at this point of time in the year. The messages are meant to trigger social comparisons and within subject comparisons of performance over time. There are 4 active treatment arms envisaged.
The Final goal is to induce change in driving behavior which will be statistically measurable. .

2023-08-23

2023-10-15

For first intervention: Number of speed violations registered by speed cameras
For second intervention: Number of violation points preserved/lost

For first intervention: The days between two adjacent violations during the observation period

There are two types of interventions envisaged.
First one utilizes the system of notifying the drivers about traffic speeding rules violations by providing additional information to them (both paper based and electronic). There are 5 active treatment arms which differ by the messages and behavorial biases they are targeted at. In addition, the subjects are separated into 3 different blocks depending on previous history of speeding violations. In particular all drivers who have violations in 2021-2022, as well as first half of the 2023 are divided into three groups - with low number of violations (3-4 per year), medium (5-7 per year) and high (more than 7). The experiment is randomized for 5 active treatment arms for each of these blocks separately.


The second intervention is a stand alone SMS based communication with the registered drivers which provides an update about the traffic violation points preserved/lost at this point of time in the year. With this option we intend to use the point system introduced by the Road Police back in 2020, which provides the information about the driver’s performance via an electronic platform. Under this approach there is no need to wait for specific event to happen to be able to communicate with the experimental subjects. The communication can be chosen to happen at any time, using the contact details (mobile phone numbers) which are available in the system. The population of the drivers registered in the system as of the moment of the experiment (e.g. July 2023) will be block randomized into three active treatment arms and one control arm (which will receive no communication at all). The sample of the experiment participants and the allocation to treatment arms will be conducted in advance. The intervention will be delivered by the Road Policy using SMS and phone numbers identified in the database. Technical duration of intervention is very short (hours), but it would be preferable to monitor the delivery of the messages and conduct it over 2-3 days.

There are two types of interventions envisaged.
First one utilizes the system of notifying the drivers about traffic speeding rules violations by providing additional information to them (both paper based and electronic). There are 5 active treatment arms which differ by the messages and behavorial biases they are targeted at. In addition, the subjects are separated into 3 different blocks depending on previous history of speeding violations. In particular all drivers who have violations in 2021-2022, as well as first half of the 2023 are divided into three groups - with low number of violations (3-4 per year), medium (5-7 per year) and high (more than 7). The experiment is randomized for 5 active treatment arms for each of these blocks separately.
There are 5 types of messages envisaged for this intervention: 1. Informing about high number of casualties (appealing to social preferences), 2. Highlighting the aggragate of fines paid over the last few years (appealing to loss aversion), 3. Highlighting the savings on insurance due to better driving (appealing to possible gains), 4. Appeal to emotions by direct message from accident family member (appeal to empathy with the help of hand written mail from a daughter of deceases traffic accident participant). 5. Kitchen sink type intervention which joins messages from 1 to 4 into one letter.

The second intervention is a stand alone SMS based communication with the registered drivers which provides an update about the traffic violation points preserved/lost at this point of time in the year. With this option we intend to use the point system introduced by the Road Police back in 2020, which provides the information about the driver’s performance via an electronic platform. Under this approach there is no need to wait for specific event to happen to be able to communicate with the experimental subjects. The communication can be chosen to happen at any time, using the contact details (mobile phone numbers) which are available in the system. The population of the drivers registered in the system as of the moment of the experiment (e.g. July 2023) will be block randomized into three active treatment arms and one control arm (which will receive no communication at all). The sample of the experiment participants and the allocation to treatment arms will be conducted in advance. The intervention will be delivered by the Road Policy using SMS and phone numbers identified in the database. Technical duration of intervention is very short (hours), but it would be preferable to monitor the delivery of the messages and conduct it over 2-3 days.
There are three type of short messages envisaged for this intervention. 1. Information about the relative standing of the driver compared to other drivers (appeal to social preferences), 2. Information about the dynamics of behavior with respect to her/his own performance during the preceding period (appeal to self control). These two options have negative or positive frames depending on the drivers actual position. 3. A neutral sms to all drivers which highlights possible savings on insurance if driving performance is improved (similar to intervention 1 treatment 3).

Randomization is implemented using randtreat routine in Stata.

Individual driver

No

There are 3 blocks for each of the two interventions.

For each of the two experiments it is planned to send out around 10,000 notifications. Each of the experiments combined with control group will be composed of around 12,000 observations. This number might be not final due to specifics of the fieldwork.

Sample size is different depending on the blocks given that the power requirements can be different due to variations in means and standard deviations.
Intervention 1.
Block 1 - 450 per treatment - 2250 active + 450 control
Block 2 - 550 per treatment - 2750 active + 550 control
Block 3 - 700 per treatment - 3500 active + 750 control
Note: Sample sizes are large due to the assumption of low delivery rate of the letters (1 out of 3 sent) and the differences in number of observations per treatment are due to larger standard deviation in Block 2 and Block 3.

Intervention 2.
Block 1 - 900 per treatment - 2700 active + 900 control
Block 2 - 750 per treatment - 2250 active + 750 control
Block 3 - 750 per treatment - 2250 active + 750 control
Note: WE do not have reliable information to what extent the mobile numbers in the system are accurate. The sample size is relatively large due to assumption that 1 out of 2 numbers is correct and the message will be delivered.

Relatively low cost of intervention (sending additional page with the intervention 1 adds almost zero marginal cost and the price of one SMS in Armenia is low as well) provides the opportunity to have conservative estimations of sample sizes.

The power of the test is set at 80, and significance level is 5%. The MDES for all treatments is set at 5% level. These are the statistics for each of the interventions and blocks. Intervention 1. Speeding experiment (Letters and elecronic notifications) Block 1. average number of violations (normalized per month) - 0.27, st. dev - 0.03 Block 2. average number of violations (normalized per month) - 0.43, st. dev - 0.07 Block 3. average number of violations (normalized per month) - 0.89, st. dev - 0.32 Intervention 2. Violations points experiment (SMS) Block 1. average number of points preserved at the time of experiment (out of total 9) - 2.9, st. dev - 1 Block 2. average number of points preserved at the time of experiment (out of total 9) - 5.7, st. dev - 0.7 Block 3. average number of points preserved at the time of experiment (out of total 9) - 7.9, st. dev - 0.6