Are your rate filings stuck in traffic? Don’t just sit there! Read what these experts are doing to manage their estimated times of arrival.

Imagine you just put the final touches on an insurance rate filing, dotting every proverbial “I” and crossing every figurative “T.” You proceed to complete your transmittal forms, remit your filing fee, sign your name and click submit in the System for Electronic Rates and Forms Filing (SERFF). Then you wait. Now imagine you are reviewing various companies’ insurance rate filings for compliance with your state’s laws and regulations and the filing we just mentioned lands atop your queue. You work your way through the rest of the queue and proceed to peruse the one you just received — but just when you are about to click approve, you observe a potential statutory issue you simply cannot overlook. You document your objection, return the filing to its sender and then you wait.

In recent years, the rate filing process in many states has at times felt like a waiting game — except that the “game” is not fun for any of the players and the stakes are potentially existential. “I think the number of filings overall has shot up because there was a huge spike in inflation,” says Scott Fischer, head of government relations and general counsel at Lemonade, who previously served as New York’s chief insurance regulator. “It is getting more expensive to do everything, and that quick rise in inflation meant lots more filings. You have less people at departments doing more complicated work and more of it. Things are going to slow down even in the best circumstances.” Slower approvals in high inflationary environments, in turn, challenge insurers’ rate adequacy and in extreme cases, destabilize markets.^[1]

“I make the analogy to traffic on the road,” says Gennady Stolyarov II, FSA, ACAS, lead property/casualty actuary at the Nevada Division of Insurance (DOI). “If there is a low-to-moderate amount of traffic on the road, every car can go at the speed limit, but there comes a point where the road is sufficiently congested that all cars move more slowly. You can try to redirect the traffic. You can try to encourage some vehicles to take exits. You can maybe even over time build more lanes. But if more and more vehicles keep coming in and continually congesting the roads, and we can channel the existing vehicles away from the main road only so fast, then this is a situation that’s going to last for some time.”

“I have heard people say, I really wish there was a fast lane for getting filings approved,” adds Dorothy Andrews, Ph.D., ASA, who is senior behavioral scientist at the National Association of Insurance Commissioners (NAIC). “One way to get closer to the fast lane is to make sure your filings are complete.”

Our goal with interviewing experts such as Andrews, Stolyarov and Fischer was to understand why some filings experience delays, how carriers and regulators can work together to navigate and minimize them, and what the future may hold for filers and reviewers. Unsurprisingly, our research did not discover any hidden shortcuts, but we did generally find that carriers that map their filing routes carefully in advance and keep compliance systems in good working order have the potential to navigate — if not to the fast lane — then at least to HOV lanes that can help minimize time spent in congestion.

Slow and steady

To understand how we arrived in the present state, it is useful to reflect on halcyon days when filings moved briskly. Stolyarov has held various responsibilities at the Nevada DOI since 2009. “I recall the 2010s as a period where we had the resources to delve in depth into many of the insurer filings and develop a kind of expertise that, in my view, placed us at the cutting edge of insurer rate reviews,” he says. “We were able to not just understand exactly what was happening but suggest refinements and improvements that not only made the outcomes more fair to consumers, but — as an additional benefit — helped improve the utility of these tools for insurers in some cases.” However, Stolyarov has observed several factors putting pressure on this public-private type partnership over time since then.

First, technology-driven efficiencies helped companies that may have once filed every two to three years move to annual cycles. “Previously enough time would have elapsed between filings that we would actually understand what happened with the previous filing, how it affected the rate level,” he says. Increased filing frequency has made unpacking feedback effects more challenging, time consuming and inferential.

The pandemic then brought about even more filings, for example, for auto — when insurers justifiably looked to incorporate impacts of drastic driving reductions into pricing. “I don’t think our society has quite returned to the way it was before the pandemic,” says Stolyarov. “Certainly, there have been objective indicators suggesting that the costs of losses to insurers have increased. As a result, the frequency and the magnitude of insurer rate filings skyrocketed. Some insurers might have been filing once or twice a year at that point, but it was not yet enough of a burden that it would strain capacity. Around early 2022, many insurers started filing two, three, four times per year, and the magnitude of each individual proposed rate increase was no longer that low or mid-single-digit percentage increase that it had been previously.”

Compounding matters, “During the period of the so-called Great Resignation or Great Reshuffle, we lost a lot of qualified, experienced staff members,” says Stolyarov.

To understand how we arrived in the present state, it is useful to reflect on halcyon days when filings moved briskly.

More broadly, “It is very challenging for state agencies to get people that are going to do this work when there is a lot of competition for this type of talent,” observes Fischer. “It is a staffing crisis, and I think it is probably across the board in a number of different states.” The challenge runs even deeper than backfilling vacancies or adding headcount within tight agency budgets.

“The learning curve can be fairly steep,” Stolyarov says. “There is an aspect of training that cannot be learned simply from studying actuarial texts. It is a matter of knowing the history and the precedents of an organization and having that institutional memory.” A newer reviewer may take longer to turn around approvals while acquiring said memory.

Once a filing goes under the microscope, interactive dialogues can pressure timelines even further. “Sometimes you will see filings go six or seven rounds of questions,” says Andrews, who notes that the American Academy of Actuaries will offer a webinar on speed-to-market on June 10, 2024.^[2] “It may be a month or two between rounds. By the time you get to seven, you may be talking up to 12 months before approval. If both parties could settle the initial set of questions upfront, then it could be just one round — a month or two.” Andrews sees some of the back and forth as reducible on all fronts: “Regulators have to ask strong questions and companies have to answer those questions.”

Some questions are avoidable altogether. The NAIC Rate Model Review Team makes available a GLM filing checklist of items to include in a filing, yet Andrews observes that some model-based filings do not include all the diagnostics or narratives in the checklist.^[3] “If you want to improve the speed at which your filing is approved, you should consider making sure to include these types of items in the initial filing,” she says.

Stolyarov also authored recommendations carriers may utilize to help expedite filing review, which appear on the Nevada DOI’s webpage.^[4] These include providing underlying data and formulas where possible, defining acronyms and providing specific answers to specific questions that may be asked. “I actually wrote those back in 2010,” he reflects, “but they’re still relevant.”

Asked, answered, operationalized

The industry is innovating a number of technological and process solutions to address these basic but persistent issues that contribute to delays. The benefits of practices such as those Stolyarov and Andrews advocate become even more apparent when analyzed with emergent technology such as large language models (LLMs). Nickolas Alvarado, FCAS, CSPA, is a consulting actuary at Milliman and was a part of a multidisciplinary team that used LLMs to help thematically characterize recent objection letter dialog from SERFF.^[5] “We saw this as a way of unlocking value from documents that were just sitting there in the past,” says Alvarado. “There was rich data but how could you actually use it? It would take any person a very long time to do a fraction of what we could do with LLMs.”

The LLMs streamlined text from the objections, and machine learning (ML) clustered them into topics that legal and compliance professionals reviewed and analyzed. Many of the prevalent themes identified in the analysis related to matters such as following instructions, completing checklists and adhering to guidelines. “It was sort of confirming what we knew, which was that filers were not putting in as much detail as they perhaps should, or in some cases, were not including things that are rather obvious,” says Alvarado. “Following the directions correctly makes it easier for you and easier for the regulator.”

When the Milliman team sifted through the themes its algorithms identified, they also found that many topics required specific knowledge of state laws, requirements and customs to interpret. Stolyarov has one theory for why such objections arose. “There is a natural desire to save on work by submitting the same filing package to every jurisdiction,” he says. “We always encourage insurers to keep detailed records of what we have requested before, how those requests were responded to, and whether the resolution was satisfactory to us so that the approach can be carried forward in future filings.”

The industry is innovating a number of technological and process solutions to address these basic but persistent issues that contribute to delays.

Allstate has taken a comprehensive approach to operationalizing institutional knowledge such as this, which we learned more about during an interview with Alex DeWitt, FCAS, and Maggie Kong, FCAS, CSPA. DeWitt is a senior actuary and director who leads work connected to state filings while Kong is former director of pricing predictive modeling.

DeWitt notes that the filing process begins well before and ends well after a company submits its filing. Every filing travels a long and winding road that begins with analysis, continues with decisions of what to file and when, and (after the filing adjudicates) concludes with effecting the results in systems and storing appropriate documentation, he says.

“We started by mapping out the cycle of what we would call a flat rate change, which is a straightforward rate change,” says DeWitt. “How do we get that into market? We essentially mapped out a pipeline. We identified at least 72 discrete steps in that process as well as handoffs occurring across seven groups of people.” The individual steps, such as document preparation or electronic submission, were then automated and integrated with one another. Regarding state-specific practices, DeWitt adds, “The process is meant to encompass all of the markets we might be trying to put an analysis into, so it includes tailoring aspects that may be relevant for different jurisdictions. We are very thoughtful about preparing well-designed filings that think about the end user, which is our regulator.” Allstate’s forethought includes programming logic where, for example, a rate level analysis may indicate no rate change is needed, triggering a second human look to possibly determine not to proceed with a filing — sparing regulators’ queues from unnecessary congestion.

DeWitt notes there is rigorous ongoing user acceptance testing around Allstate’s process. “You have to do whatever you can to minimize the risk of error,” he says. “In the same way that you would when you set up a quality process for humans, you need the same guardrails for machines. Humans can make mistakes. Machines can make mistakes. That part is not different.”

Exhibiting model behavior

Filing automations may represent sound prevention against common objections on common filings, but novel filing situations or objections are trickier to streamline. “Anything that comes across a regulator’s desk that has the word model in it is more likely to draw a significant amount of questions,” says Kong. “There is also a lot more depth of the types of questions being asked.” She attributes this to various factors: “Generalized linear models (GLMs) have been in common use for so long that the industry has a much deeper proficiency and can ask deeper questions than 20 or even 10 years ago. On the other hand, with newer methods there is a curiosity where regulators may not know the techniques as well yet and would like to build greater working knowledge.” This fervor can create difficult cost-benefit decisions around how much to provide simply to satisfy curiosity. “I love when more people are interested in modeling and want to learn more. That eagerness excites me,” Kong says. “But it creates a balance of wanting to preempt questions versus not wanting to present an overwhelming amount of information that may not be material to review,” she says. “There are times when we will proactively provide information, and others when we will proactively have information ready to provide.” Over time and with experience, carriers can move closer and closer to that ideal balance.

Meanwhile, Andrews and her colleagues on the NAIC Rate Model Review Team are attempting to drive greater consistency in this type of questioning. She estimates that more than half of states presently utilize her team’s services to provide assistance reviewing filings that carriers may file in any given state. One resource the team maintains is a database of previously reviewed model filings. “If a company, for example, wants to file the same model in five different states, we may have already written a report for one state that the other four states can go in and look at,” she says. Reading the reports also has knock-on benefits to the quality and consistency of questions carriers receive. “Knowledge gets transferred,” Andrews says. “Our goal is to help state reviewers become more independent, and I think we are seeing that because regulators are asking more technical questions without our assistance.”

Filing automations may represent sound prevention against common objections on common filings, but novel filing situations or objections are trickier to streamline.

Besides innovating on conduct around the filing itself, filers are also testing ways to model and manage regulator workloads. In addition to the LLM-driven analysis discussed previously, Alvarado and teammates also empirically analyzed the average number of days from filing to approval over time in several different prior approval states.^[6] They found that time to approval is roughly consistent over time in any given state but can trend positively or negatively depending on factors such as staffing and procedural changes. Companies can potentially perform analyses of their own times to approval or that of their peers to avoid unnecessary corrective filings. “If you are doing a rate filing, for example, then you have to select a trend that you expect over, say, the next year,” says Alvarado. “The rates are for that prospective period for which you priced. If your filing isn’t approved when you expect it to be, then those rates technically need to be revised and refiled.” With the richness of diagnostic data available in SERFF, optimization analysis can also potentially be used to marshal resources towards filings more likely to lag or to monitor for lower traffic periods during which to submit filings.

Paving the road forward

As the pandemic fades into the rearview mirror, the road ahead for filings may contain fewer potholes. “One would hope that there would be some easing of the inflationary pressures,” says Stolyarov. He has cautions, however. “The pandemic has taught us that the future is radically uncertain,” he says. Some changes could potentially future proof the filing process further against unforeseen stresses that may arise.

One area to do this could be through statute and regulation. “There might be things that could be accomplished in terms of having more inflationary aspects embedded into rates,” says Fischer. “You could build in a little bit more flexibility. Nobody likes rates going up, but it might be more palatable for consumers to experience a 2.5% increase year over year, than to get virtually nothing for five years and then be surprised when you’re getting a 15% increase.” Such approaches are not unprecedented: Several states already permit some level of “flex rating” where prior approval is not required when rate changes are within a certain range.^[7] Additionally, almost all states permit model year rating for auto physical damage whereby far out emergent model years receive higher rates than current ones before any data even comes in. “Flexibility does not give up any of the control of the prior approval, but it says that at the beginning of the process, we are going to think about the unknowns,” Fischer adds.

Fischer also points to the Interstate Insurance Compact as having improved speed-to-market for life insurance, where a number of states have agreed to delegate away limited regulatory functions traditionally conducted within their own insurance departments. While there are a number of state-specific considerations in property/casualty that are not ripe for delegation, the NAIC Model Rate Review Team is an example of how limited delegation can distill efficiencies into bogged processes.

For now and in the near future, regulators and filers are likely to continue finding themselves waiting longer than they prefer from time to time. However, the participants in the ecosystem are not sitting idly by waiting on the world to change. Innovations such as those described in this article are permeating almost every dimension of the filing experience, allowing innovators to figure out which parts work best and run with them. If these improvements happen to also coincide with an easing in inflationary pressures, it is not hard to envision a scenario where better conditions arise sooner than we expect.

Jim Weiss, FCAS, CSPA, is a vice president for Crum & Forster and is editor in chief for Actuarial Review.

^[1]The California property market is a topic on its own but is an example of where some view filing delays as a destabilizing factor. For more information, read Dale Porfilio’s coverage in the March-April AR.

^[2]https://www.actuary.org/Speed-to-Market-NAIC-Presentation.

^[3]https://content.naic.org/sites/default/files/call_materials/NAIC%20Reviews%20GLM%20document%20list.pdf.

^[4]https://doi.nv.gov/Insurers/Property-Casualty/Filing-Information/Personal-Automobile-Insurance/.

^[5]https://www.milliman.com/en/insight/analyzing-insurance-product-filings-artificial-intelligence-llm.

^[6]https://www.milliman.com/en/insight/regulatory-insurance-intelligence-rate-filing-days-approval-february-2024.

^[7]https://www.leg.state.nv.us/session/76th2011/exhibits/assembly/cmc/acmc279j.pdf.