A captive feasibility study is a multi-party analysis evaluating whether forming a captive insurer is financially viable and strategically sound for a specific parent organization. The study is the document the parent’s board reads before voting to fund a captive, and it is the document the domicile regulator reads before granting a license.
What the feasibility study actually is
Five professional parties typically contribute. The captive manager runs the study and integrates the pieces. Tax counsel handles the §832 analysis and the broader question of whether the captive will be respected as an insurance company for tax purposes. Domicile counsel handles the licensing application and the regulator’s specific filing requirements. The broker scopes the reinsurance program. The actuary produces the pro forma loss pick.
The actuary’s piece is one of five, but it is the piece that underlies every other piece. The tax case for forming the captive sits on the loss pick, because the §832 analysis turns on whether premiums are reasonable for the risk. The capital requirement sits on the loss pick. The reinsurance pricing the broker negotiates depends on the loss pick. The domicile regulator’s review of the feasibility study turns first on whether the loss pick is defensible.
This article walks through what the actuary specifically contributes at formation, where the loss pick numbers come from, what the analysis can and cannot tell the buyer, and what to require from the actuary’s report. If you are still deciding between a captive and other risk-financing structures, start there and come back. If you have read the tax case for forming a captive and want to understand the loss numbers that underlie it, this is the right entry point. The actuary’s report will refer to incurred-but-not-reported losses as a routine matter; that concept is the foundation of everything that follows.
Why the actuary’s piece is the hardest piece
The captive has no own data. By construction. It does not yet exist.
That is the first sentence of the actuarial challenge, and the rest of the challenge flows from it. Friedland frames the chain ladder method around the foundational assumption that claims recorded to date will continue to develop in a similar manner in the future (Friedland, p. 84). For a captive that has not yet written a policy, there are no claims recorded to date. Chain ladder is not available. Among the five core reserving methods, only the expected claims method runs at all, and Friedland is explicit that the expected claims method is the right answer when entering new lines of business or new territories (Friedland, p. 131). A feasibility study is that case in pure form.
The actuary therefore relies on three sources of evidence: the parent’s historical loss experience under prior risk-financing arrangements, industry benchmark data, and market data from reinsurance pricing. Friedland warns explicitly that benchmark data must be examined for comparability, because claims practices, policy coverages, deductibles, geographic mix, and coding differences can make published benchmarks noncomparable to the parent’s specific exposure (Friedland, p. 88). At formation, this caution applies in full force. The parent’s prior experience may have been written at different limits, under different state mixes, with different claims management. Benchmark data may be national when the parent operates in three states. Reinsurance market pricing reflects market conditions at the moment of the quote, not the captive’s eventual experience.
The practical implication is that the feasibility loss pick is the single most uncertain number in the entire feasibility analysis. A competent actuary acknowledges that uncertainty on the first page of the report and quantifies it with a range. A weak actuary reports a single point estimate and lets the reader assume it carries the precision of a fitted statistical model.
The exposure base decision
For a captive at formation, the expected claims method uses pure premium multiplied by exposure rather than earned premium multiplied by an expected claim ratio, because the captive does not yet collect premium in the insurer sense. The exposure base choice is therefore the first structural decision the actuary makes.
Friedland’s exposure base table by line gives the standard answers (Friedland, p. 132). Workers compensation uses payroll, usually per hundred dollars of payroll, sometimes per full-time equivalent for newer programs. Automobile liability uses vehicle count, with miles driven as an increasingly common refinement for fleets with telematics. General liability for corporates uses revenue or sales, with square footage substituting for retail and warehouse exposures. Property uses insured values. Crime uses employee count. Hospital professional liability uses occupied beds and outpatient visits.
The exposure base decision matters at formation more than at any later point in the captive’s life. The captive’s premium methodology, the parent’s premium allocation across operating subsidiaries, and the eventual reserving methodology all key off the exposure base. The annual reserve study five years from now will compare ultimate claims per exposure against the prior year’s number, and that comparison only works if the exposure measurement is consistent across years. A wrong exposure base at formation propagates through every downstream calculation for the life of the captive.
The actuary should document, in writing, what exposure base was chosen for each line, why it was chosen, what alternatives were considered, and how the exposure will be measured year over year. Group captives face an additional wrinkle: the exposure base must be measurable consistently across all member companies, which is harder than it sounds when members differ in size, payroll-reporting cadence, and segment mix.
Where the loss pick numbers come from
The actuary draws on three sources, in descending order of evidentiary strength.
The parent’s own loss experience. If the parent has historical losses from a prior fronted program, a large-deductible plan, or a self-insured retention, those losses are the strongest input. They reflect the parent’s actual operations, actual claims management, and actual geographic and segment mix. The actuary must adjust the historical data for changes in retention, limits, and exposure between the prior structure and the proposed captive structure. Friedland’s Berquist-Sherman policy-year substitution method handles the limit-change adjustment when retentions move (Friedland, p. 283). The actuary should disclose whether the adjustment was applied, what severity trend was used, and how sensitive the loss pick is to the trend assumption.
Industry benchmarks. NCCI for workers compensation. ISO and Verisk for general liability. RAA for excess and reinsurance. AIS and trade-association studies for medical professional liability. Friedland’s benchmark caution applies: claims practices, coverages, deductibles, geographic mix, and coding differences can make published benchmarks noncomparable (Friedland, p. 88). A competent actuary documents which benchmarks were used, why they are comparable to the parent’s exposure, where they diverge, and how the divergence was handled. The benchmark is a starting point, not an answer.
Market data from the broker. Reinsurance quotes embed market-level loss expectations. When a market quotes a per-occurrence excess layer at a given rate, the rate reflects the market’s view of expected losses above the captive’s retention. The actuary can triangulate against quoted reinsurance pricing to check the loss pick for reasonableness. If the actuary’s pick disagrees materially with what the reinsurance market is pricing, one of the two is wrong, and the actuary should be able to articulate which one and why.
A loss pick at this stage is not a single number. It is a range with a central estimate. ASOP 43 defines the actuarial central estimate as an expected value over the range of outcomes reasonably possible given the data and methods (Friedland, p. 10). The range communicates the uncertainty that the buyer’s downstream decisions, particularly capital and reinsurance attachment, must absorb. If the report distinguishes between pure IBNR and broader IBNR components, that is a positive signal: it means the actuary is thinking about the pieces of the eventual reserve number, not just the headline figure.
A weak loss pick uses a single source, reports a point estimate, and does not document its limitations. A strong loss pick triangulates across all three sources, reports a range with a central estimate, and lists what it cannot tell the buyer.
Retention analysis and reinsurance program design
The retention decision is two decisions. The per-occurrence retention determines how much loss the captive absorbs on any single claim. The aggregate retention determines how much total loss the captive absorbs across all claims in a year.
The actuary’s contribution to the per-occurrence retention decision is a frequency-and-severity model. Friedland’s frequency-severity framework (Friedland, p. 195) is the foundation: model how often losses occur, model how large each loss is, then combine. At each candidate retention level, the model produces an expected number of losses that pierce the retention, an expected severity above the retention, and an expected aggregate cost.
The actuary’s contribution to the aggregate stop-loss decision sits one step downstream. For a captive structure with per-occurrence excess of loss and aggregate stop-loss, Friedland specifies that the actuary typically estimates ultimate claims limited to the per-occurrence retention first, then applies the stop-loss as a final adjustment (Friedland, p. 332). Trying to model both simultaneously confuses the layering and produces aggregate numbers that double-count the per-occurrence reinsurance recovery.
The actuary’s analysis informs but does not replace the broker’s reinsurance market negotiation. The two work iteratively: the broker quotes, the actuary models the captive’s net exposure, the broker requotes if the structure changes, and the actuary remodels.
The post-formation version of this same analysis sits in the gross-to-net bridge, where annual reserves are split between gross, ceded, and net columns. The feasibility analysis determines the shape of that bridge before it exists. Frequency-severity dominates over the development-pattern methods at formation because there is no development data yet.
Capital adequacy and confidence level selection
How much capital does the captive need to hold to support the retained losses at a chosen confidence level?
That is the capital question, and the answer is one of the most consequential numbers in the feasibility study. Typical confidence levels in captive feasibility work range from 75% to 95%, with 80% to 90% the most common selections. The confidence level expresses the probability that the captive’s actual loss outcomes will fall at or below the funded amount. At 90% confidence, the funded amount covers the captive in nine years out of ten; in the tenth year, additional capital is needed.
The mechanics involve a distribution. The actuary fits an analytical severity distribution (lognormal, gamma, or Pareto for heavy-tail lines), a frequency distribution (Poisson or negative binomial), and combines them into an aggregate distribution through convolution or simulation. The chosen percentile of the aggregate distribution is the capital requirement.
The variance assumption is the input that matters most. A more variable loss distribution produces a fatter right tail and a materially higher 90th-percentile capital number. Two actuaries running the same expected loss pick but different variance assumptions can produce capital requirements that differ by a factor of two or more. The actuary should document where the variance assumption came from, whether it was fit to the parent’s own data or pulled from industry data, and how sensitive the capital number is to it.
The confidence level selection itself is a buyer decision, informed by but not made by the actuary. High-confidence funding ties up more capital but reduces the probability of a mid-year capital call. Low-confidence funding frees capital but raises the probability that the captive runs short. ASOP 43 defines the central estimate (Friedland, p. 10) but does not prescribe a confidence level for capital. The buyer must choose, with the actuary’s modeling as input.
A dedicated piece goes deeper into the percentile selection mechanics and the buyer trade-offs. For the feasibility stage, what matters is that the buyer understands the decision is theirs, that the variance assumption drives the answer almost as much as the confidence level does, and that the actuary’s report should make both transparent. If the capital floor is the binding constraint on going forward as a standalone captive, a cell captive or protected cell company is the natural alternative to evaluate alongside the standalone form.
Sensitivity testing and stress scenarios
A feasibility study that reports only a central case is incomplete. The analysis must test how the loss pick and the capital requirement move under stress.
Standard stress scenarios for a captive feasibility study include frequency increases of 20%, 50%, and 100%; severity increases of 20%, 50%, and 100%; the addition of a new line of business in year three; an adverse legal or regulatory development such as a jurisdictional venue shift or a tort reform reversal; and reinsurance market hardening at first renewal, which raises the captive’s net retained exposure if excess coverage becomes unaffordable.
Friedland’s worked example of a paid bodily-injury triangle on the most recent accident year, where the cumulative development factor to ultimate is 90.00 (Friedland, p. 134), illustrates how leveraged a single year’s projection becomes when data is thin. At captive formation there is no data at all, so every projection inherits the maximum possible leverage. Sensitivity testing is the only tool the buyer has to size how much that leverage matters.
The actuary should produce a tornado chart or equivalent display showing which inputs move the central loss pick most. Variables that move the central estimate by more than 10% under plausible stress deserve explicit discussion in the report body, not just a footnote in an appendix. The buyer should know which two or three assumptions are doing most of the work, because those are the assumptions the board will want to revisit when actual data starts to roll in during years one through three.
A feasibility study that assumes the captive can buy excess coverage at year-one pricing for the next ten years is making a heroic assumption. Modeling at least one renewal cycle with hardened pricing tests the captive’s ability to absorb the cost shock.
What the feasibility study cannot tell you
The feasibility study has hard limits, and a competent report names them.
It does not predict actual losses. It projects expected losses under a set of assumptions, each of which is uncertain. The actual losses the captive experiences in year one may fall anywhere within the range the actuary identifies, and occasionally outside it.
It does not bind the domicile regulator’s review. Regulators apply their own reserving and capital standards, often more conservative than the feasibility study assumes, and may require a higher confidence level for licensing capital than the buyer chose.
It is not a Statement of Actuarial Opinion. Statements of Actuarial Opinion are annual, post-formation, signed under ASOP 36, and addressed to the captive’s board and domicile regulator on the adequacy of carried reserves. The feasibility study is none of those things. It is a planning document.
It assumes the proposed business is captured by available industry data. New or unusual exposures, such as industry-specific cyber risk, emerging regulatory liability, or novel product warranties, are particularly weak in feasibility analyses because benchmark data either does not exist or does not match the captive’s specific exposure. The report should name the gap and propose how the captive will close it over the first few years of operation.
A good feasibility report names these limits clearly. A weak report does not, or buries them in an appendix the buyer never reads. The buyer’s job is to read the limits section as carefully as the loss pick.
What the buyer should require from the actuary
A concrete checklist follows. None of these items is unreasonable for a feasibility study at the price point captive owners typically pay. If a report lacks them, the buyer should ask why before signing off.
The actuary should provide a documented methodology section identifying which sources contributed to the loss pick and how the three sources were weighted. Named benchmark sources with publication dates. A range with a central estimate, not a point estimate. A sensitivity analysis showing how the central estimate moves under stressed frequency, severity, retention, and exposure assumptions. A documented limitations section that names what the analysis cannot tell the buyer. A named signing actuary with credentials, typically FCAS or ACAS with MAAA membership for U.S. captive work.
Domicile-specific reserve requirements should be addressed for the proposed domicile. Major captive jurisdictions have published expectations for actuarial input at licensing, and the actuary’s report should map cleanly to those expectations.
A transition plan for years one through five should describe how the loss pick will evolve from a pure a priori estimate (year one) to a Bornhuetter-Ferguson blend with industry development patterns (years two through three) to a development-method blend on older accident years (years four through five). The buyer needs to understand that the feasibility study is the first input to a process that will refine itself over time, not the last word.
A statement of independence from the broker and the captive manager belongs in the report. The actuary is a fee-for-service professional, and the buyer needs to know that the analysis is not contingent on the broker placing the program or the captive manager winning the management contract.
The diligence questions in the buyer’s guides on what to ask in an RFP and on interviewing the actuary apply at the feasibility stage as well as at the annual review stage. The actuary the buyer hires for the feasibility study is the actuary the captive will likely retain for the first three to five years of annual reserve work.
What comes after the captive is formed
The feasibility study is the first chapter of a multi-year story.
In year one, the feasibility loss pick becomes the a priori expected claim ratio for the captive’s first annual reserve study. With one or two diagonals of own data, the expected claims method (Friedland, p. 131) dominates. The reserve analyst is essentially restating the feasibility loss pick with a year of own loss experience as a credibility-weighted check on the original assumptions.
In years two and three, the Bornhuetter-Ferguson method dominates because the captive has more own data but still not enough to support chain ladder on recent years. Friedland reports that actuaries rely on Bornhuetter-Ferguson almost as often as they rely on chain ladder (Friedland, p. 152). The development pattern is borrowed from industry benchmarks or from the parent’s prior experience, weighted by the BF credibility logic against the captive’s emerging own data.
In years four and five, chain ladder enters the method blend on the older accident years where the captive’s own data has matured enough to project. The newer accident years still rely on BF. By year five and beyond, the captive looks like a standard insurer reserving exercise: chain ladder on mature years, BF on recent years, expected claims as a check.
The actuary’s relationship with the captive becomes ongoing. Annual reserve studies are the floor. Interim monitoring catches drift between full studies by comparing actual quarterly development against what the prior full analysis projected (Friedland, p. 345). For a captive that started from a feasibility loss pick with no own data, interim monitoring matters more than it does for a mature carrier, because the original loss pick has more uncertainty to absorb.
The reserving cadence for single-parent captive reserving, group captive and RRG reserving, and hospital captive reserving all build on the foundation the feasibility study sets. The feasibility loss pick is the first number; everything that follows is a refinement of it.