PAY AS YOU DRIVE (PAYD) INSURANCE PILOT PROGRAM PHASE 2 MID-COURSE PROJECT REPORT. Prepared by Progressive County Mutual Insurance Company.

PAY AS YOU DRIVE (PAYD) INSURANCE PILOT PROGRAM PHASE 2 MID-COURSE PROJECT REPORT Prepared by Progressive County Mutual Insurance Company March 2007 This report was prepared in cooperation with the Texas Department of Transportation and the U.S. Department of Transportation, Federal Highway Administration, and Federal Transit Administration. The contents of this report reflect the views of the authors who are responsible for the opinions, findings, and conclusions presented herein. The contents do not necessarily reflect the views of policies of the Texas Department of Transportation, the U.S. Department of Transportation, Federal Highway Administration, or Federal Transit Administration. 1

Executive Summary The North Central Texas Council of Governments (NCTCOG) is testing the concept of a Pay As You Drive (PAYD) insurance program in which the cost of auto insurance would vary based on mileage driven rather than being a fixed annual or semi-annual amount. The NCTCOG is attempting to understand whether there is a causal relationship between financial incentives and miles driven, and what effect this may have on vehicle emissions and air quality. Progressive was selected as the auto insurance carrier partner for a PAYD insurance test in March 2006. Progressive has enrolled 3,014 of its customers who agreed to participate in the test. Participants will have the opportunity to earn an incentive by reducing the number of miles they drive during the test period. Mileage driven will be recorded via a TripSensor TM data-logging device that is plugged into the OBD port found beneath the steering column in each participating vehicle. Key Findings: Many customers feel powerless to influence when and where they drive. There are both structural and lifestyle-related factors that hinder the use of carpooling and public transportation. Customers associate reduced mileage with a reduction in auto emissions. Lower auto insurance rates are among the top incentives cited by customers; these lower rates would make them consider carpooling or using public transportation. There was a high level of interest in the pilot program among Progressive customers. Initial data indicate that the PAYD program may decrease miles driven. The effect of the PAYD program may be to reduce mileage during commute hours. As expected, there was no effect on time spent driving over 75 mph. Several factors were identified that could affect the credibility of the statistical results. Data collected during the second phase of this project may allow for more definitive conclusions. About Progressive Progressive has been in business since 1937 and ranks third in the nation for auto insurance based on premiums written and is the fourth largest auto insurer in Texas. Progressive is headquartered in Ohio, has a regional headquarters in Austin, and 35 claims offices located throughout Texas. In total, Progressive employees 1,981 people in Texas. Progressive seeks to be an innovative, growing and enduring business by providing our customers with great rates, superior online and in-person customer service, and innovative claims service. More information about Progressive can be found at www.progressive.com. Description of the TX PAYD Insurance program (Phase II) 2

Program Objectives The NCTCOG is testing the concept of PAYD insurance pilot that permits customers to purchase insurance on a per-mile basis rather than a fixed annual premium. The objectives of this study are: To determine if a PAYD option will induce drivers to reduce their mileage. To understand consumers attitudes and experiences involving the PAYD program, and the programs effect on their driving behaviors. This study assumes that driving less will result in fewer vehicle emissions and improved air quality. History of Progressive s Mileage-Based Insurance Programs Progressive is committed to using technology in innovative ways to reduce the cost of car insurance for consumers and providing consumers with choices in how to shop for, buy and own an auto insurance policy. This commitment has spurred our research and development of two usage-based automobile insurance programs. Progressive introduced the first-ever usage-based automobile insurance program available to consumers in the US in 1998. Called Autograph SM, the program was first test marketed in Houston and was made available to drivers throughout the state of Texas in 1999. Autograph was developed and patented by Progressive and involved the use of proprietary technology and global navigation positioning system (GPS) equipment retrofitted in customer vehicles. The Autograph test proved customers liked usage-based auto insurance because it saved them money and gave them more control over their auto insurance costs. Autograph customers in Houston saved an average of 25 percent on their auto insurance when compared to what they were paying. Progressive ended the test in 2001. In August 2004, Progressive introduced its second usage-based insurance program, called TripSense, in Minnesota. The states of Michigan and Oregon have been included in the pilot program subsequently. With TripSense, customers plug a data-logging device into a port in their car that collects information about the vehicle s use, including how many miles, the speed and time of day when the vehicle is driven. Customers can earn from 5 percent to 20 percent more in discounts based on how much and when the vehicle was driven. More information about this research program can be found at www.tripsense.progressive.com. Description of the Pilot Progressive recruited 3,000 volunteers from eligible existing Progressive customers in nine Texas counties (Collin, Dallas, Denton, Ellis, Johnson, Kaufman, Parker, Rockwall, Tarrant). These volunteers agreed to share vehicle usage data in exchange for a research participation award of $50 or more for each eligible vehicle they registered for the research study. Existing Progressive Direct Business customers with 1996 and newer model cars 3

were eligible to participate. Not all vehicles in a household were required to participate in the program. Existing customers were sent invitations via direct mail solicitations beginning in April 2006. The year-long research study started in May 2006. It is limited to existing customers and the first 3,000 vehicles registered. Research study volunteers were required to plug a TripSensor data-logging device into the OBD port found beneath the steering column in each participating vehicle for one year. The TripSensor records mileage data during the test period. Participants receive a research participation reward of at least $50 per vehicle when they use the software provided to upload their driving data to Progressive after six months and again after 12 months. The amount each participant receives may exceed $50 if they actually drive fewer miles than expected during the study. Participants are eligible to receive an additional $25 for every 5 percent fewer miles driven compared to the expected mileage each six-month term. They can earn up to $175 per term, up to $350 in total. Expected mileage is determined by odometer readings taken from recent emission inspections. If no readings are available, participants are not eligible for a mileage reduction bonus in the first six-month test period. In the second six-month test period if the TripSensor shows they drove less compared to the previous six months they are then eligible for the reduced mileage bonus. Participation and receipt of research participation award is not contingent upon remaining a Progressive customer for the duration of the research study. The six- and 12-month research participation rewards to eligible research study volunteers are in the form of American Express Travelers Checks. Progressive will conduct statistical analysis on the data collected. We will also conduct three surveys, as well as focus groups, to gain feedback from participants. Study Costs Federal funding in the amount of $1.5 million was available for phase two of the PAYD pilot program. In addition there was an incremental local match in the amount of $375,000 for a total project funding level of $1,875,000. Progressive responded to the RFP with a Contract Pricing Proposal in the amount of $1,267,436. This included cost of the telematics device, technical, IT and marketing labor costs, customer recruiting expenses, consumer research expenses, and participation incentives for enrolled customers. Participant Interest The level of customer interest in the PAYD pilot exceeded initial expectations. Progressive had planned to have two direct mailings to solicit customers to enroll in the pilot. However, all 3,000 available slots were filled within 10 days of the first mailing. 4

Mid-Term Statistical Results for the Texas PAYD Pilot The PAYD pilot covers two six-month test periods. Participants are instructed to upload data to us at the end of each period. The first period of the test has been completed, and we have collected first term data on the driving behavior of the total participant pool. In addition, the participants were divided into two groups based on whether baseline mileage data was available from the NCTCOG. Progressive received the baseline mileage data on only 93 of the 3,014 participants (3%). For the remainder of the participants, the first phase of the test represented the first opportunity to measure mileage driven. Both groups could earn a participation incentive by reporting their data to us, and the former group had an ability to earn an incentive during the first period by reducing their mileage driven. The first statistical study looked at whether incentives encouraged participants to reduce mileage driven by comparing their baseline mileage to their first term mileage. Only data from the 93 participants with baseline mileage readings from the NCTCOG were included in this study. Progressive found that the average annual mile decrease for this group was 1,237 miles per year. This reduction is statistically significant at the 95 percent confidence level, meaning that there was a credible decrease in mileage driven due to the incentive. (Refer to Exhibit 1). Since the group for whom we had a baseline mileage reading could earn an incentive by reducing their driving during the first term, while the second group could not, the two groups were compared to see if they behaved differently in terms of mileage driven and percentage of time spent driving above 75 mph. Customers who were incented to drive less, drove 961 fewer miles, on average, than customers in the other group (Refer to exhibit 2A). This difference was statistically significant at the 90% confidence level. Mileage was also broken out into three driving time of day groupings-- low risk time, medium risk time, and high risk time. It was found that customers who were incented to drive less, drove an average of 504 fewer medium risk miles (mainly commute-hour miles) than the customers in the other group (Refer to exhibit 2). Results were credible at the 95 percent confidence level. There was no significant mileage reduction during other time periods (Refer to exhibit 3). Although mileage declined, when comparing the two customer groups, there was no significant reduction in the percentage of non-idle driving time spent above 75 mph (Refer to exhibit 4). This was not unexpected, since earning an incentive wasn t contingent upon changing driving behaviors other than mileage driven. Factors That Could Affect the Credibility of the Results Reporting Bias. Some calculations were done by comparing data reported from the NCTCOG to data from the Tripsensor device. Any differences in the accuracy of the two reporting methods could have a negative impact on the validity of the analysis. Data collected during the second six month test period should mitigate this potential bias and allow for a comparison of two measurement periods using the Tripsensor device. 5

Sampling Bias. A sampling bias could be introduced if the participants for whom mileage data was available from the NCTCOG are in some way different than those for whom mileage data wasn t available, and that difference affects mileage driven. This bias would affect the second statistical test, which compared these two groups. All vehicles in household not participating in the test. This test was constructed without the condition that all vehicles in a household participate. Because of this, it would be possible for participants in multi-vehicle households, who only signed up one vehicle, to reduce mileage driven on that vehicle simply by using the other vehicle more. In this way, participants could earn an incentive without actually reducing mileage driven. To see what effect this may have on the results, the 93 participants for whom we had mileage data from the NCTCOG were evaluated to see how many signed up all vehicles in their household, and how their mileage driven during the first test period compared to the mileage reading from the NCTCOG. We found 38 of the 93 participants had signed up all vehicles, and that the average annual mile decrease for this group was 555 miles per year (Refer to Exhibit 5). However, because of the small sample size and sample variance, this wasn t a statistically significant difference in mileage driven. Data collected during the second phase of this project may allow for a more credible conclusion. Small sample size. Even though the results of the statistical tests appear to be credible, readers should keep in mind that these conclusions were reached by evaluating only 93 participant mileage readings. Once again, the second reporting period will allow for more credible estimates to be calculated using a larger group of participants. Pay As You Drive Insurance Consumer Interest Survey PRE-PILOT SURVEY METHODOLOGY Prior to the launching of the Pay As You Drive (PAYD) pilot, a pre-pilot survey of customers 1 was conducted in order to gain knowledge of their receptivity to insurance pricing based on mileage driven, and their awareness of the link between mileage driven, auto emissions and clean air. An online survey of Progressive customers 1 residing in the nine county area surrounding Dallas/Fort Worth, Texas was conducted from April 6, 2006 until April 30, 2006. A total sample of 1,183 Progressive customers was achieved. The objectives of this study were: To determine the interest in insurance pricing based on mileage driven To determine the interest in and ability to change one s driving behaviors in order to get a lower insurance premium To determine if insurance premium reduction will induce drivers to reduce mileage 1 This survey included Progressive customers only, not the general market. Therefore, results are only generalizable in so far as Progressive s book of business is representative of the overall customer auto insurance market. Other data sources suggest that our current customer base may not be reflective of the general US consumer auto insurance market; therefore, results may not be generalizable to the larger population of auto insurance customers. 6

To determine if drivers associate reduced mileage with reduced auto emissions resulting in cleaner air Pre-Pilot Survey Results Based on the survey results, it was found that lower auto insurance premiums are ranked high as an incentive to carpool or use public transportation. However, the necessity of driving (and therefore not changing one s driving behavior) often supersedes the insurance discounts that come from reducing miles driven. Drivers have little control over when and where they drive. Less than 20% say they have a lot or quite a bit of control to reduce the number of miles they drive in a week and less than 15 percent say they have a lot or quite a bit of control to change the time of day they drive. However, customers do feel very in control of how they drive; 93% of customers say they have a lot or quite a bit of control over how fast they drive (Refer to exhibit 6). Although in control of how they drive, childcare issues, work schedule, work location, and lack of available public transportation may prohibit drivers from participating in mileage reduction programs. 54% of respondents selected my schedule prohibits them from carpooling; the most frequent answer for public transportation is no service available in my work/home area. I need my car to take my children to and from daycare/school was ranked as the top reason for not carpooling and the second ranked reason for not using public transportation. Lifestyle factors may also play a part in whether people are willing to carpool or use public transportation. I prefer to have the ability to come and go as I please ranked highly as a reason for not carpooling and using public transportation. In addition, I don t want to rely on other people,, and I need my car to run errands before or after work, were cited by 30% and 28% percent of participants, respectively, as reasons for not carpooling (Refer to exhibits 7 and 7b). Despite this, drivers do associate reduced mileage with a reduction in auto emissions. Car emissions are listed as the most common contributor to air pollution and rank clean air as the third most important factor for a city to have (Refer to exhibit 8). When asked what would lead them to reduce miles driven, lower auto insurance rates are sited among the top incentives. Participants were asked to select from a list of incentives (which included lower auto insurance rates) that would make them consider carpooling and using public transportation, and then to rank those incentives in order of importance. Lower auto insurance rates were the most frequently mentioned incentive to carpool, and ranked second in importance among all incentives to carpool. Saving money on gasoline is ranked first among all incentives to carpool and is mentioned as an incentive to carpool by 39.9% of respondents. The most frequently mentioned incentive to use public transportation is if there was service available close to my home (49%). Service also ranked first in importance among all incentives to use public transportation. Lower auto insurance rates ranked fifth in 7

importance, among all incentives to use public transportation and were mentioned as an incentive to use public transportation by 30% of respondents. The biggest obstacle to carpooling reported by customers is my schedule prohibits (54%). The biggest obstacle to using public transportation is no service available in my work/home area (48%) (Refer to exhibits 9 and 10). Mileage reduction programs need to account for the barriers drivers face in reducing mileage, which involve daily responsibilities and access to public transportation, both of which are out of the customers control. As one customer said during recruiting for the pilot, Our society, and our cities and towns were developed in such a way that does not promote less driving but actually more commutes. I live in Dallas, Texas. I may have to drive anywhere from 30-45 minutes to get to work. But that is not of my own design, that is how society and city planners decided to design this world. Drivers recognize the connection between car emissions and air pollution and how they contribute to the problem. However, customers generally feel powerless to reduce their mileage. Next Steps Participant Survey The participants in the pilot completed the first downloads of their driving data at the end of December 2006. We are currently in the process of developing a survey to be administered to all participants who were recruited into the study. The survey is scheduled for fielding in April 2007. The purpose of the survey is to provide feedback from participants on their experience with the TripSense device, including specifics related to why a participant may have defected from the pilot program, participants expectations of the pilot program, participants attitudes toward a Pay As You Drive system and what effects the program has had on participants driving behaviors. Focus groups Focus groups will be conducted at the end of the program anticipated administration is July 2007 to gain further feedback from Pay As You Drive participants. Goals are to learn: How consumers reduced their mileage Did they increase their use of public transportation Did they otherwise change their behavior and attitudes toward driving Better understand the opinions of participants regarding the process Gauge participants satisfaction with their participation in the pilot program Participant recommendations about how they would reduce their mileage Collect More Data and Analyze Second period results will include results from the larger pool of participants for whom we didn t have sufficient starting odometer reading data. This will lend more statistical credibility to our conclusions. Prepare Final Project Summary 8

A final project summary for phase two of the PAYD Insurance Pilot Program will be completed. This will include a summary of the statistical analysis performed on the driving data, as well as a summary of key findings from the participant surveys and focus groups. Exhibits Summary of Mileage Driven by Group 12500 12360 Annualized Mileage 12000 11500 11000 10500 10000 11321 10766 11123 11830 Odometer Reading Tripsense Reading 9500 Odometer Reading & All vehicles Participating All Participants with Odometer Readings Participants without Odometer Readings 9

Exhibit 1: Baselne Mileage Compared to First Term Mileage (Annualized) t-test: Paired Two Sample for Means Uploaded Mileage (TripSense Device) Baseline Mileage (State Odometer Readings) Annualized Average Mileage 11123 12360 Variance 34220012 32616101 # Observations 93 93 Pearson Correlation 0.71 Hypothesized Mean Difference 0 Degrees of Freedom 92 t Stat -2.73 P(T< = t) one-tail 0.00 t Critical one-tail 1.66 Exhibit 2 Annualized Commute Hour Mieage Comparison t-test: Two-Sample Assuming Unequal Variances Participants who could not earn an incentive during the first period Participants Who Could Earn an Incentive by Driving less Annualized Average Mileage 4562 4058 Variance 9193066 6840385 Observations 1081 96 Hypothesized Mean Difference 0 Degrees of Freedom 119 t Stat 1.78 P(T<=t) one-tail 0.04 t Critical one-tail 1.66 10

Exhibit 2A Annualized Total Mileage Comparison t-test: Two-Sample Assuming Equal Variances (Outliers eliminated) Participants who could not earn an incentive during the first period Participants Who Could Earn an Incentive by Driving less Average Annual Mileage 11830 10869 Variance 31528621 30829152 Observations 1039 94 Pooled Variance 31471105 Hypothesized Mean Difference 0 Degrees of Freedom 1131 t Stat 1.59 P(T<=t) one-tail 0.06 t Critical one-tail 1.28 11

Exhibit 3 Annualized Mileage Driven During Low Risk Time Periods t-test: Two-Sample Assuming Equal Variances Participants who could not earn an incentive during the first period Participants Who Could Earn an Incentive by Driving less Average Mileage 6455 6178 Variance 12755881 12489636 Observations 1087 98 Pooled Variance 12734050 Hypothesized Mean Difference 0 Degrees of Freedom 1183 t Stat 0.73 P(T<=t) one-tail 0.23 t Critical one-tail 1.65 Annualized Mileage Driven During High Risk Time Periods t-test: Two-Sample Assuming Unequal Variances Participants who could not earn an incentive during the first period Participants Who Could Earn an Incentive by Driving less Average Mileage 169 183 Variance 84453 107097 Observations 1081 95 Hypothesized Mean Difference 0 Degrees of Freedom 107 t Stat -0.40 P(T<=t) one-tail 0.34 t Critical one-tail 1.66 12

Exhibit 4 Percent of Time Spent Driving over 75 MPH t-test: Two-Sample Assuming Equal Variances Participants who could not earn an incentive during the first period Participants Who Could Earn an Incentive by Driving less Average % Time 0.96% 1.00% Variance 2.50 2.15 Observations 1072 96 Pooled Variance 2.47 Hypothesized Mean Difference 0 Degrees of Freedom 1166 t Stat -0.20 P(T<=t) one-tail 0.42 t Critical one-tail 1.65 Exhibit 5 Annualized Total Mileage Comparison Participants Who Included All Vehicles in the Test t-test: Two-Sample Assuming Equal Variances Uploaded Mileage (TripSense Device) Baseline Mileage (State Odometer Readings) Annualized Average Mileage 10766 11321 Variance 27986996 25306033 Observations 37 37 Pooled Variance 26646514.59 Hypothesized Mean Difference 0 Degrees of Freedom 72 t Stat -0.46 P(T<=t) one-tail 0.32 t Critical one-tail 1.67 13

Exhibit 6: Customers indicate they have high control over how fast they drive but little control over when and where they drive To reduce the number of trips you drive in a year Over how fast you drive To use public transportation To be a part of a carpool To change the time of day you drive To reduce the number of miles you drive in a w eek 0.00% 10.00% 20.00% 30.00% 40.00% 50.00% 60.00% 70.00% 80.00% 90.00% 100.00 % A lot Quite a bit Some A little No control Exhibit 7: Lack of available public transportation prohibits use 5 most frequent answers mentioned for not using public transportation (multiple response) No service available in my work/home area 48.1% Not available for my travel needs 34.8% My schedule prohibits/irregular work schedule 25.3% Closest stop too far from my home 24.9% I prefer to have the ability to come and go as I please 20.3% Top 5 ranked reasons for not using public transportation (mean rankings) No service available in my work/home area I need a car to take my children to and from daycare/school I need my car for work/business reasons My schedule prohibits/irregular work schedule Not available for my travel needs Base = 1183 14

Exhibit 7b: Customers schedules and lifestyles prohibit carpooling 5 most frequent answers mentioned for not carpooling (multiple response) My schedule prohibits 53.59% I prefer to have the ability to come and go as I please 34.65% I don t know anyone to carpool with 30.43% I don t want to rely on other people 30.34% I need my car to run errands before or after work 28.01% Top 5 ranked reasons for not carpooling (mean rankings) I need a car to take my children to and from daycare/school My schedule prohibits carpooling/irregular work schedule I need my car for business reasons I don t know anyone to carpool with I need my car to run errands during the day Base = 1183 Exhibit 8: Auto emissions rank as most significant contributor to air pollution 5 most frequent answers mentioned as common air pollutants (multiple response) Auto emissions 77.0% Factory/industrial emissions 68.6% Diesel fuel emissions 46.0% Burning fossil fuels to make utilities such as electricity or natural gas 42.2% Gasoline vapors 23.8% Top 5 ranked common air pollutants (mean rankings) Auto emissions Factory/industrial emissions Diesel fuel emissions Burning fossil fuels to make utilities such as electricity or natural gas Gasoline vapors Base = 1183 15

Exhibit 9: Lower insurance rate premiums and reduce gasoline costs rank as top incentives to carpool 5 most frequent answers mentioned as incentives to carpool (multiple response) If it lowered by auto insurance rates 41.3% If I am saving money on gasoline 39.9% If there was a ride home program in case of emergency or changed work schedule 31.1% If my employer offered incentives to carpoolers 31.1% If my work stop and start timers were flexible 27.4% Top 5 ranked incentives to carpool (mean rankings) If I am saving money on gasoline If it lowered my auto insurance rates Nothing, I will never carpool If my company provides me with a car for work If my work stop and start times were flexible Base = 1183 Exhibit 10: Top incentives to use public transportation include availability of service and lower auto insurance rates 5 most frequent answers mentioned as incentives to use public transportation (multiple response) If there was service available close to my home 49.5% If it lowered my auto insurance rate 29.7% If I am saving money on gasoline 28.3% Nothing. I will never use public transportation 21.8% If my employer offered incentives for using public transportation 20.3% Top 5 ranked incentives to carpool (mean rankings) If there was service available close to my home If my company provides me with a car for work If I am saving money on gasoline Nothing. I will never use public transportation If it lowered my auto insurance rates Base = 1183 16