|Shares Out. (in M):||77||P/E||13.4x||0.0x|
|Market Cap (in $M):||3,130||P/FCF||0.0x||0.0x|
|Net Debt (in $M):||400||EBIT||310||0|
Link to PDF:
I believe AFSI is insolvent and incredibly aggressive in their accounting which makes it an attractive short.
One can see the that their earnings are low quality by looking at the cumulative change in tangible book from YE 2008 to now:
In other words, despite reporting $750m of earnings over 4.5 years, their tangible book value has only increased by $230m, $27m of which was the result of issuing a convert (tang book would be ~$100m lower if they recognized DTLs associated with earnings indefinitely). That's a lot of running relative to the distance traveled...
Amtrust Financial Services is a P&C insurance company operating in four segments: Small Commercial Business (workers comp, commercial package, other P&C for small businesses), Specialty Risk & Extended Warranty (historically warranty, but includes other specialty lines in Europe including Italian medical malpractice and UK legal expense), Specialty Program (workers comp, commercial package, general liability, etc), and Personal Lines Re (quota share with a private related party).
In addition, Amtrust earns "service and fee income" (revenue) from unaffiliated third parties for warranty administration, policy fees, management services for risk retention groups and affiliated third parties for brokerage, asset management, and IT services. This line has grown meaningfully in recent years as a result of Amtrust's frequent acquisitions (ex: Case New Holland Agency - $30m/yr, BTIS agency - $18m/yr, CPP NA (competitor to Lifelock)).
AFSI's mix has shifted around a bit in recent years, but generally stayed since 2010 (Italian Med Mal entry & launch of Personal Lines Re with related party NGHC).
AFSI booked rather low loss ratios in 2008/9, but they have been creeping up since then (largely as a result of adverse development).
AFSI's real advantage seems to be on the expense side, but as I will explain later we believe this is driven more by accounting tricks than the technology advantage that management claims.
Below are the resulting combined ratios. Unsurprisingly, most of the volatility comes from the loss ratio.
A life settlement is the sale of an existing life insurance policy to a third party for more than its cash surrender value, but less than its net death benefit. There are a number of reasons that a policy owner may choose to sell his or her life insurance policy. The policy owner may no longer need or want his or her policy, he or she may wish to purchase a different kind of life insurance policy, or premium payments may no longer be affordable. Policy owners often learn about settling their policies from a financial planner or advisor, insurance broker, attorney, friends or family, or estate planning presentations.
Life settlements are a legitimate asset class, but their reliance on assumptions along with the general opacity and illiquidity of the asset provide for a wide latitude of valuations. Unsurprisingly, this latitude tends to attract unsavory characters that have given LS a spotty track record as investments. Examples below:
Mutual Benefits Corp ($1b LS fraud that used internal LE estimates):
Life Partners (systemic underestimation of LE using internal estimates):
In determining the value of a policy, there are two primary factors: the life expectancy (LE) of the insured and the discount rate.
Investors in life settlements face different adverse selection issues than the carriers who issue the policies -- carriers want the insured to live longer so they receive more premiums and get to pay the death benefit later...LS investors want the opposite. As such, they use different estimates for LEs. Carriers generally use the VBT table (link below). Whereas LS investors use LE providers that specialize in adjusting for the different risks that LS investors face. Two providers have the lion's share of the market: 21st Services and AVS. Fasano is a distant third and EMSI is a further distant fourth.
As a result of the illiquidity, negative cash flow dynamics (paying premiums), adverse selection issues, risk of litigation, and risk of life extension (double-whammy as more premiums and longer wait for death benefit), the market generally uses a 15-25% discount rate for life settlements. I've confirmed this with private market investors and public company disclosures validate this (IFT - 20.6% MRQ , ALSO - 20-25%, GWG Holdings (public debt) - 13%).
Additionally, EEA Life Settlement Fund's auditors (E&Y) disagreed with their valuation when they applied a 10% discount rate to the "full credibility scenario" which incorporates a number of additional reserves beyond the move to a life expectancy based on 21st Services, AVS, etc.
"The discount rate of 10% in the “full credibility” value is, in our view, significantly lower than would be used by market participants in an arm’s-length transaction to purchase the assets of the fund"
Source: EEA Annual report - http://www.cisx.com/download_news.php?newsID=204669
Amtrust's Life Settlements:
Amtrust entered the life settlement space through a JV with ACAC (now NGHC) in mid-2010 with the acquisition of a portfolio of loans that were collateralized by life insurance policies. All of the policyholders defaulted on the loans, enabling Amtrust to foreclose on the policy. At this point, they marked to model the policy and recognized a non-cash, upfront gain.
Inception-to-date, Amtrust has recognized at least 40.7m of net income as gains (72.6m gains on I/S less 27.6m of NCI recognized). I say at least because AFSI's footnote disclosures show 162.5m of net income from life settlements (gross of NCI, net of profit commission). There is no disclosure or explanation in the filings for the difference and IR simply said they aren't related (hard to believe) and didn't want to talk about life settlements any more. Regardless, the company now carries them at $209m gross / ~99.2m net of profit commission and NCI (but also indirectly has another $10m of exposure through their stake in the JV partner). As such, they represent 17% of tangible book value.
Before going into AFSI's assumptions, it's worth highlighting what FAS 157, which governs LS accounting says:
"The fair value measurement objective is to determine an exit price from the perspective of a market participant... unobservable inputs (ie: discount rate) shall reflect the reporting entity's own assumptions that market participants would use in pricing the asset or liability (including assumptions about risk)."
Level 3 inputs
30. Level 3 inputs are unobservable inputs for the asset or liability. Unobservable inputs shall be used to measure fair value to the extent that relevant observable inputs are not available, thereby allowing for situations in which there is little, if any, market activity for the asset or liability at the measurement date. However, the fair value measurement objective remains the same, that is, an exit price from the perspective of a market participant that holds the asset or liability. Therefore, unobservable inputs shall reflect the reporting entity’s own assumptions about the assumptions that market participants would use in pricing the asset or liability (including assumptions about risk). Unobservable inputs shall be developed based on the best information available in the circumstances, which might include the reporting entity’s own data. In developing unobservable inputs, the reporting entity need not undertake all possible efforts to obtain information about market participant assumptions. However, the reporting entity shall not ignore information about market participant assumptions that is reasonably available without undue cost and effort. Therefore, the reporting entity’s own data used to develop unobservable inputs shall be adjusted if information is reasonably available without undue cost and effort that indicates that market participants would use different assumptions.
AFSI's 2q13 10Q:
The fair value of life settlement contracts as well as life settlement profit commission is based on information available to the Company at the end of the reporting period. The Company considers the following factors in its fair value estimates: cost at date of purchase, recent purchases and sales of similar investments, financial standing of the issuer, and changes in economic conditions affecting the issuer, maintenance cost, premiums, benefits, standard actuarially developed mortality tables and industry life expectancy reports. The fair value of a life insurance policy is estimated by applying an investment discount rate based on the cost of funding the Company's life settlement contracts as compared to returns on investments in asset classes with comparable credit quality, which the Company has determined to be 7.5%, to the expected cash flow generated by the policies in the Company's life settlement portfolio (death benefits less premium payments), net of policy specific adjustments and reserves. The Company adjusts the standard mortality for each insured for the insured's life expectancy based on reviews of the insured's medical records. The Company establishes policy specific reserves for the following uncertainties: improvements in mortality, the possibility that the high net worth individuals represented in its portfolio may have access to better health care, the volatility inherent in determining the life expectancy of insureds with significant reported health impairments, the possibility that the issuer of the policy or a third party will contest the payment of the death benefit payable to the Company, and the future expenses related to the administration of the portfolio. The application of the investment discount rate to the expected cash flow generated by the portfolio, net of the policy specific reserves, yields the fair value of the portfolio. The effective discount rate reflects the relationship between the fair value and the expected cash flow.
These reserves sound an awfully lot like EEA, ALSO, and IFT...who all use much higher discount rates. Also, note that these reserves were not mentioned until after the OWS report.
1q13 CC - Ron Pipoly, CFO (emphasis added):
Additionally, there has been some confusion in the marketplace surrounding our life settlement portfolio. I want to take this opportunity to clarify and explain how we account for and how we value our life settlement portfolio. We utilize a very conservative approach in determining the fair value of the portfolio. There are many factors that go into determining the value of the portfolio not simply at discount rate.
To be clear, our effective discount rate on our life settlement portfolio for 2012 was 17.7%. If you were to reference our 2012 10-K you would note that we disclosed a discount rate of 7.5% and an internal rate of return of 17.7%. That internal rate of return is our effective discount rate on our life settlement portfolio.
To illustrate, the total future positive cash – gross positive flows in our life settlement portfolio are expected to be $795 million. Those cash flows are reduced by adding additional reserves that we apply. Among those reserves is a life expectancy reserve. Life expectancies are updated every year by two of the most widely recognized life expectancy providers and weighted to the more conservative of two life expectancies.
Additionally, we provided mortality adjustment reserve by reducing the standard mortality tables based on the assumption that higher net worth individuals will live longer based on access to better healthcare. We also provided additional reserve for future expenses, operational risk and the reserve for highly impaired lives.
For 2012, the total of those reserves were $356 million. That gives us future positive net cash flows of $439 million. That $439 million is then discounted by 7.5% resulting in our December 31, 2012 carrying value of $193 million and an effective discount rate of 17.7%. On both a quarterly and an annual basis, our evaluations are reviewed by a nationally-recognized life actuarial firm. If we would have simply used the discount rate of 7.5% applied to the expected positive future cash flows of $795 million, the carrying value of our portfolio would've been $383 million as opposed to the $193 million, which we actually carried in December 31, 2012.
To date, we have had four mortality events and have collected benefit on all four of those policies. The total benefit collected was $24.5 million. Those four policies were purchased for $1.3 million, and we paid a total of $400,000 of premium. We carried those policies at a value of $5.2 million. Those four policies generated a cash gain of $22.8 million.
We acquired these assets when the market for these assets was very distressed and that philosophy is consistent with being very contrary and buying assets cheaply that produce very high returns over the long term with very little capital at risk.
To summarize, what is referred to as our internal rate of return in our 10-K of 17.7% is our effective discount rate on our life settlement portfolio.
Here is how I interpret his comments:
If my interpretation of their "reserves" being the difference between VBT and specialty LE providers' estimates of LE, then their assumptions are vastly off market and clearly contradict the accounting rules for these assets.
Life Expectancies (LEs):
II. 1q13 CC (After OWS report):
In addition, AFSI's LE estimates have been steadily falling more than the passage of time (196 months at 12/31/11 to 155 months at 12/31/11 to 139 months at 12/31/12) despite
AFSI discount rates:
Despite attempting to twist the definition of discount rate to be the same as expected IRR (7.5% is not the "investment discount rate" rather than simply the discount rate and "Internal Rate of Return" is now "Effective Discount Rate"), the fact is they are discounting their investments at 7.5%.
Compare AFSI's assumptions to IFT (10K excerpts):
Valuation of Insurance Policies
..We currently use a probabilistic method of valuing life insurance policies, which we believe to be the preferred valuation method in the industry. The most significant assumptions which we estimate are the life expectancy of the insured and the discount rate.
In determining the life expectancy estimate, we analyze medical reviews from independent secondary market life expectancy providers (each a “LE provider”). An LE provider reviews the medical records and identifies all medical conditions it feels are relevant to the life expectancy of the insured. Debits and credits are then assigned by each LE provider to the individual’s health based on these medical conditions. The debit or credit that an LE provider assigns to a medical condition is derived from the experience of mortality attributed to this condition in the portfolio of lives that it monitors. The health of the insured is summarized by the LE provider into a life assessment of the individual’s life expectancy expressed both in terms of months and in mortality factor.
The resulting mortality factor represents an indication as to the degree to which the given life can be considered more or less impaired than a standard life having similar characteristics (e.g. gender, age, smoking, etc.). For example, a standard insured (the average life for the given mortality table) would carry a mortality rating of 100%. A similar but impaired life bearing a mortality rating of 200% would be considered to have twice the chance of dying earlier than the standard life. Historically, the Company has procured the majority of its life expectancy reports from two life expectancy report providers and only used AVS life expectancy reports for valuation purposes. Beginning in the quarter ended September 30, 2012, the Company began utilizing life expectancy reports from 21st Services for valuation purposes and began averaging or “blending,” the results of the two life expectancy reports to establish a composite mortality factor. However, when “blending” the AVS and 21st Services life expectancy reports, if the difference in the probability of mortality between the reports is greater than 150%, the Company will reduce the higher mortality factor until the difference is 150%.
Comment: Skewed to the more conservative estimate.
... However, beginning in the quarter ended September 30, 2012, the Company transitioned to a table developed by the U.S. Society of Actuaries known as the 2008 Valuation Basic Table, or the 2008 VBT. However, because the 2008 VBT table does not account for anticipated improvements in mortality in the insured population, the table was modified in conjunction with outside consultants to reflect these expected mortality improvements. The Company believes that the change in mortality table does not materially impact the valuation of its life insurance policies and that its adoption of a modified 2008 VBT table is consistent with modified tables used by market participants and third party medical underwriters.
Comment: making adjustments to the 2008 VBT table to reflect anticipated improvements in mortality is equivalent to "adding reserves".
...As is the case with most market participants at December 31, 2012, the Company used a blend of life expectancies that are provided by two third-party LE providers.
...The Company believes that investors in esoteric assets, such as life insurance policies, typically target yields averaging between 12%—17% for investments of more than a 5 year duration, and had historically used a 15%—17% range of discount rates to value its life insurance policies.
...Although the Company believes that its entry into the Non-Prosecution Agreement had a positive effect on the market generally and for premium financed life insurance policies specifically, the Company believes that, when given the choice to invest in a policy that was associated with the Company’s premium finance business and a similar policy without such an association, all else being equal, an investor would have generally opted to invest in the policy that was not associated with the Company’s premium finance business.
As of December 31, 2012, the Company owned 214 policies with an aggregate investment in life settlements of $113.4 million. Of these 214 policies, 171 were previously premium financed and are valued using discount rates that range from 16.80% to 33.80%. The remaining 43 policies are valued using discount rates that range from 14.80% to 21.30%.
Comment: AFSI's policies were almost entirely premium financed which means that the policyholder likely paid little to nothing on the policy...a "lender" advanced the premiums and foreclosed after the two year contestability period.
IFT's weighted average discount rate (mix of premium financed and non-premium financed): 24.01%.
Last but not least, AFSI has substantial exposure to Phoenix as a counterparty. This came to light when Tiger Capital sued PHL for increasing the Cost of Insurance on their policies. In the suit, Tiger disclosed 138 policies with Phoenix (out of 241 at the time / 261 now).
Phoenix policies trade at substantial discounts because of PNX's poor financial health and litigious approach to dealing with life settlement investors. For example, PNX's statutory surplus declined from $922.5m at 12/31/12 to $789.2m at 6/30/13 -- to support $13B of policy liabilities -- and PNX's AM Best rating has been downgraded below where most agents will take business.
I'll add more on the discrepancies here in the comment section.
In summary, the life settlement issues are:
I. Questionable LE estimates
II. Far off-market discount rates that violate the accounting rules
III. Substantial counterparty risk
AFSI has grown by leaps and bounds (33.6% CAGR of Net Premiums Written from 2007-2012) despite the soft market for P&C insurance which begs the question -- how were they able to grow so much and report results so much better than peers in a terrible insurance market? The short answer is that they weren't able to do so...
IBNR = Incurred but not reported. Case Reserves = reserves for claims already filed. Case reserves could also be a source of over/under-reserving, but IBNR is where management tends to have more discretion because IBNR is a more general reserve whereas case reserves are specific to claims.
Given AFSI's substantial underwriting leverage (NPW / Stat Cap; Loss Reserves & Unearned Prem / Stat Cap), tangible book is highly sensitive to changes in reserves. If AFSI were to boost reserves to where IBNR was 45% of total (holding case reserves constant) -- the 2010 YE level -- tangible book value would fall by more than half.
WorkersComp reserves are calculated using the "incurred development method" which:
Quarterly Incurred Development Method (Use of AmTrust Factors)
Quarterly incurred loss development factors are derived from our historical, cumulative incurred losses by accident month. These factors are then applied to the latest actual incurred losses and DCC by month to estimate ultimate losses and DCC, based on the assumption that each accident month will develop to estimated ultimate cost in a similar manner to prior years. There is a substantial amount of judgment involved in this method.
Yearly Incurred Development (Use of National Council on Compensation Insurance, Inc. (“NCCI”) Industry Factors by State)
Yearly incurred loss development factors are derived from either NCCI’s annual statistical bulletin or state bureaus. These factors are then applied to the latest actual incurred losses and DCC by year by state to estimate ultimate losses and DCC, based on the assumption that each year will develop to an estimated ultimate cost similar to the industry development by year by state.
...rely on historical development factors derived from changes in our incurred losses, which are estimates of paid claims and case reserves over time. As a result, if case reserving practices change over time, the two incurred methods may produce substantial variation in the estimate of ultimate losses. We have not used any “paid” development methods, which rely on actual claims payment patterns and, therefore, are not sensitive to changes in case reserving procedures. As our paid historical experience grows, we will consider using “paid” loss development methods.
Of the two methods above, the use of industry loss development factors has consistently produced higher estimates of workers’ compensation losses and DCC expenses.
In other words, AFSI uses the historical pattern of its own estimates (for case reserves) to estimate losses and acknowledges that using its own methodology is always lower than the industry.
The low end of the range is established by assigning a weight of 100% to our ultimate losses obtained by application of our own loss development factors. The high end is established by assigning a weight of 50% each to our ultimate losses as developed through application of Company and industry wide loss development factors.
From these disclosures, we can back into what the loss reserve would be if they only used industry factors -- $155m for just workers' comp.
Surprisingly, AFSI does not address Medical Malpractice reserving policies, particularly given the very long tail nature of this line and highly IBNR driven aspect (a pacemaker fouling up years later).
Specialty Risk and Extended Warranty claims are usually paid quickly, development on known claims is negligible, and generally, case reserves are not established. IBNR reserves for warranty claims are generally “pure” IBNR, which refers to amounts for claims that occurred prior to an accounting date but are reported after that date. The reporting lag for warranty IBNR claims is generally small, usually in the range of one to three months. Management determines warranty IBNR by examining the experience of individual coverage plans. Our consulting actuary, at the end of each calendar year, reviews our IBNR by looking at our overall coverage plan experience, with assumptions of claim reporting lag and average monthly claim payouts. Our net IBNR as of December 31, 2012 and 2011 for our Specialty Risk and Extended Warranty segment was $26.6 million and $52.9 million, respectively.
The primary actuarial methodology used to project future losses for the unexpired terms of contracts is to project the future number of claims, then multiply them by an average claim cost. The future number of claims is derived by applying to unexpired months a selected ratio of the number of claims to expired months. The selected ratio is determined from a combination of:
We began writing general liability, commercial auto and commercial property (jointly known as CPP) business in 2006. In order to establish IBNR for CPP lines of business, we rely on three methods that utilize industry development patterns by line of business:
Yearly Incurred Development (Use of Industry Factors by Line). For each line, the development factors are taken directly from Insurance Services Office, Inc. (“ISO”) loss development publications for a specific line of business. These factors are then applied to the latest actual incurred losses and DCC by accident year, by line of business to estimate ultimate losses and DCC
Expected Loss Ratio. For each line, an expected loss ratio is taken from our original account level pricing analysis. These loss ratios are then applied to the earned premiums by line by year to estimate ultimate losses and DCC
Bornhuetter-Ferguson Method. For each line, IBNR factors are developed from the applicable industry loss development factors and expected losses are taken from the original account level pricing analysis. IBNR factors are then applied to the expected losses to estimate IBNR and DCC
...Because we determine our reserves based on industry incurred development patterns, our ultimate losses may differ substantially from our estimates produced by the above methods.
Because of the numerous third party administrators we use, we have utilized only limited incurred development methods based on historical loss development patterns, or methods that rely on paid development factors. Paid loss development methods rely on actual claim payment patterns to develop ultimate loss and DCC estimates.
Comment: Implies they use "Expected Loss Ratio" and B-F Method. No disclosure of gross/net/case/IBNR for these lines of business.
In each of the last three years, AFSI has seen adverse development, particularly for years 2009 and 2010 when the company was booking aggressive initial loss ratios of 54.3%/57.1%. However, because they have grown their underwriting so aggressively in recent years, the adverse development has been swamped by new business that is also subject to aggressive initial loss picks.
The questions are:
AFSI discloses the premiums, losses and expenses ceded to MHLD in related party footnotes and MHLD breaks out the AFSI quota share as a separate segment disclosure. AFSI's CEO is the Chairman of MHLD and they share BDO as an auditor...so why do they report such different results for MHLD's share of the AFSI quota share and, more importantly, why does AFSI have a 710 bps lower loss ratio than what they cede to MHLD on a quota share when it covered 84.6% of their underwriting (cumulative since mid-2007)??
Note that the disclosures match up on the NPE line (8m difference - qtrly timing of recognition).
Neither AFSI nor MHLD had an explanation for this, but MHLD confirmed that those disclosures are mirrors of one another in what they cover.
Two competitors have recently had adverse development that was across similar business lines (WC, Commercial Multi-Peril, and Commercial Auto) and account sizes (small biz) - Meadowbrook (MIG) and Tower Group (TWGP). MIG was subsequently downgraded by AM Best and forced to find a reinsurance partner and TWGP is now under review.
Note that all three were relatively strong performers in 2008-2010 (they were also each growing more rapidly than the industry.
Some of AFSI's customers require a rating of "A-" or higher (and agents would re-direct new policies). Given the high degree of leverage and relative illiquidity of some of AFSI's large assets, a severe liquidity crunch is very possible at AFSI (particularly given the significant debt funding at the holdco level subject to rating and net worth covenants).
Below is an estimation of AFSI's consolidated statutory surplus and the residual stat surplus after allocating capital to specifically disclosed subsidiaries.
Below is the resulting consolidated underwriting leverage at AFSI (using YE surplus to be generous with the exception of 2013 which uses 2012 YE).
In comparison, MIG (recently downgraded by AM Best) was writing at 1.5x NPW / Stat Surplus.
So why is AFSI able to use so much more leverage? Bermuda!
They cede over 70% of GWP to Amtrust International Insurance (AII) which then cedes 40% of the original GWP to MHLD (another related party). AII then collateralizes the statutory subsidiaries' reinsurance recoverables with a combination of letters of credit (off balance sheet debt of 152.8m at YE 2012) and a collateral trust (MHLD also has a collateral trust for the benefit of AII/US Stat subs). In total, the US subs have 1.76B of reinsurance recoverables that are not subject to a haircut since they're collateralized .
However, AII lacks the liquid assets it requires to fund the entirety of their share of the recoverable, so they use (1) the previously mentioned letters of credit (2) a collateral funding agreement with MHLD for ~168m and (3) repo financing of $292m.
AII 2010 financial statements that show all of AFSI's consolidated repo financing is attributable to this entity:
AII's incremental source of cash flow is premium payments (via US subs). So if AFSI were forced to slow down underwriting growth (or heaven forbid shrink/stop), AII would not have the liquid assets to fund the repo repayment.
What assets does AII have to sell beyond those tied up? Equity investments in:
Given this lack of liquid assets and callable debt (repo), AII is the weak link in the chain. So what, you say, the regulators don't care...it's been going on forever!
Ben Lawsky, the Superintendent of Financial Services & NYDFS are working on it and getting some scalps (including Metlife). I believe AFSI's structure is exactly the sort of hidden risk they are seeking to expose and correct.
Background on DFS Investigation into Shadow Insurance
In July 2012, DFS initiated an investigation into shadow insurance at New York-based insurance companies and their affiliates.
Insurance companies use shadow insurance to shift blocks of insurance policy claims to special entities — often in states outside where the companies are based, or else offshore (e.g., the Cayman Islands) — in order to take advantage of looser reserve and regulatory requirements. Reserves are funds that insurers set aside to pay policyholder claims.
In a typical shadow insurance transaction, an insurance company creates a “captive” insurance subsidiary, which is essentially a shell company owned by the insurer’s parent. The company then “reinsures” a block of existing policy claims through the shell company – and diverts the reserves that it had previously set aside to pay policyholders to other purposes, since the reserve and collateral requirements for the captive shell company are typically lower. Sometimes the parent company even effectively pays a commission to itself from the shell company when the transaction is complete.
This financial alchemy, however, does not actually transfer the risk for those insurance policies off the parent company’s books because, in many instances, the parent company is ultimately still on the hook for paying claims if the shell company’s weaker reserves are exhausted through a “parental guarantee.” That means that when the time finally comes for a policyholder to collect their promised benefits after years of paying premiums — such as when there is a death in their family – there is a smaller reserve buffer available at the insurance company to ensure that the policyholders receive the benefits to which they are legally entitled.
Lawsky praised MetLife’s announcement in a Twitter message today, citing “progress on making the shadowy world of insurance ‘captives’ more transparent.”
MetLife Chief Executive Officer Steve Kandarian said the offshore reinsurance program began in 2001 and “makes less sense” now.
“The New York Department of Financial Services’ industry inquiry regarding captives was an important factor in our taking a closer look at our offshore reinsurance subsidiary,” Kandarian, 61, said at the investor presentation. “We’re going to take steps to bring these businesses back on shore and to a more highly capitalized, U.S.-based and U.S.-regulated entity.”
AFSI acquires renewal rights and/or managing general agents, sometimes even paying them contingent consideration as a % of future premiums written -- these look a lot like commission expenses, but because they are bought instead of built, they do not hit the income statement (any intangibles that get amortized are over long periods even those are added back to "operating earnings").
Even though they're following GAAP (except for the add-backs), it is a misrepresentation of the economics of the business. One way to see through the M&A accounting fog is to look at tangible book growth (as we did at the beginning) which is lacking.
Back of the envelope, I think this "trick" brings AFSI's expense ratio down by 500-600 bps based on ~$1.2b of GWP (~50% of 2012 GWP) from acquired distribution channels at 10-12% commission rates.
For the sake of argument, let's say that if the distribution had been third parties, AFSI would have had a ~30% expense ratio (about in-line with MIG/TWGP) but since they acquired half of their distribution, they don't have to pay those commissions (renewal rights roll and MGAs are done with contingent consideration to keep them incentivized). That brings the expense ratio on a gross basis down to ~24-25%.
Then AFSI cuts MHLD in on the quota share, but wants reimbursed for the cost of generating those premiums -- at the (fair based on my assumptions) rate of 30% ceding commission for premiums ceded (about what they get). AFSI gets to run this through the income statement and nets it against expenses for the expense ratio.
If $100 of premiums are written and 40% are ceded pro-rata for a 30% commission, AFSI will get $12 of ceding commissions to count against their $25 of gross loss expense, making the net expense ratio $13. Since they retain $60, the net expense ratio becomes 21.67% (13/60).
Voila! Better expense ratio via acquisition & ceding...assuming you don't mind that you probably overpaid for the distribution on the front end which hurt the cumulative change in tangible book.
To keep this "trick" going, they have to continue generating at least the same amount of their distribution as they did before. That's fine if they don't need to grow, but AFSI needs to keep growing to cover up adverse development from under-reserving and keep cash flowing to AII.
However, they're starting to run out of tangible book and consolidated statutory surplus to spend with tangible book sitting at ~$600m and needle moving deals likely to cost $75m+ and be entirely comprised of intangibles if they are purely distribution.
Given the length of this report, i'm going to leave this for the comments section as a follow-up.
Given the length of this report, i'm going to leave this for the comments section as a follow-up.
AFSI has math problems. That's troubling for an insurance company.
Below is a comparison of press releases and SEC filing disclosures. They never explained any of the changes, and the income statement, shareholders equity, and goodwill/intangibles were never changed.
How does premium receivable go up $47m between the PR and 10K without impacting revenues or unearned premium?? I'll refrain from accusing them of something more nefarious than clever accounting gimmicks and aggressive assumptions, but I'm certainly suspicious...
There are a myriad of other disclosures that we have not been able to reconcile for AFSI against its own disclosures as well as related parites NGHC and MHLD. I'll touch on some of them in the comments section.
With all these red flags, you might think that AFSI would trade at a discount to the space. Hardly!
For the most part, compas trade between 60-160% of stated book and 90-150% of tangible book. AFSI is off the charts on these metrics. At 5x tangible book (with inflated assets and deflated loss reserves), AFSI is anything but cheap, unless you think I'm dead wrong and/or that the game will go on forever.
The table below doesn't match in some instances because it pulls from CapIQ without making any of the relevant adjustments (ie: AFSI 10% stock dividend).
Investment in life settlements — The Company has elected to account for the life settlement policies it acquires using the fair value method. Due to the inactive market for life settlements, the Company uses a present value technique to estimate the fair value of our investments in life settlements, which is a Level 3 fair value measurement as the significant inputs are unobservable and require significant management judgment or estimation. The Company currently uses a probabilistic method of valuing life insurance policies, which we believe to be the preferred valuation method in the industry. The most significant assumptions are the estimates of life expectancy of the insured and the discount rate.
In determining the life expectancy estimate, we analyze medical reviews from independent secondary market life expectancy providers (each a “LE provider”). An LE provider reviews the medical records and identifies all medical conditions it feels are relevant to the life expectancy of the insured. Debits and credits are then assigned by each LE provider to the individual’s health based on identified medical conditions. The debit or credit that an LE provider assigns to a medical condition is derived from the experience of mortality attributed to this condition in the portfolio of lives that the LE provider monitors. The health of the insured is summarized by the LE provider into a life assessment of the individual’s life expectancy expressed both in terms of months and in mortality factor.
The resulting mortality factor represents an indication as to the degree to which the given life can be considered more or less impaired than a standard life having similar characteristics (e.g. gender, age, smoking, etc.). For example, a standard insured (the average life for the given mortality table) would carry a mortality rating of 100%. A similar but impaired life bearing a mortality rating of 200% would be considered to have twice the chance of dying earlier than the standard life. The probability of mortality for an insured is then calculated by applying the blended life expectancy estimate to a mortality table. The mortality table is created based on the rates of death among groups categorized by gender, age, and smoking status. By measuring how many deaths occur during each year, the table allows for a calculation of the probability of death in a given year for each category of insured people. The probability of mortality for an insured is found by applying the mortality rating from the life expectancy assessment to the probability found in the actuarial table for the insured’s age, sex and smoking status. The Company has historically applied an actuarial table developed by a third party. However, beginning in the quarter ended September 30, 2012, the Company transitioned to a table developed by the U.S. Society of Actuaries known as the 2008 Valuation Basic Table, or the 2008 VBT. However, because the 2008 VBT table does not account for anticipated improvements in mortality in the insured population, the table was modified by outside consultants to reflect these expected mortality improvements. The Company believes that the change in mortality table does not materially impact the valuation of its life insurance policies and that its adoption of a modified 2008 VBT table is consistent with modified tables used by market participants and third party medical underwriters.
The mortality rating is used to create a series of best estimate probabilistic cash flows. This probability represents a mathematical curve known as a mortality curve. This curve is then used to generate a series of expected cash flows over the remaining expected lifespan of the insured and the corresponding policy. A discounted present value calculation is then used to determine the value of the policy. If the insured dies earlier than expected, the return will be higher than if the insured dies when expected or later than expected.
The calculation allows for the possibility that if the insured dies earlier than expected, the premiums needed to keep the policy in force will not have to be paid. Conversely, the calculation also considers the possibility that if the insured lives longer than expected, more premium payments will be necessary. Based on these considerations, each possible outcome is assigned a probability and the range of possible outcomes is then used to create a value for the policy.
Historically, the Company has procured the majority of its life expectancy reports from two life expectancy report providers and only used AVS Underwriting LLC (“AVS”) life expectancy reports for valuation purposes. Beginning in the quarter ended September 30, 2012, the Company began utilizing life expectancy reports from 21st Services, LLC (“21st Services”) for valuation purposes and began averaging or “blending,” the results of the two life expectancy reports to establish a composite mortality factor.
On January 22, 2013, 21st Services announced revisions to its underwriting methodology and on February 4, 2013, announced that it was correcting errors discovered in its previously announced revised methodology. According to the 21st Services, these revisions have generally been understood to lengthen the average reported life expectancy furnished by this life expectancy provider by 19%. At March 31, 2013, the Company had not received any life expectancy reports from 21st Services utilizing its revised methodology and, to account for the impact of the revisions and based off of market responses to the methodology change, the Company lengthened the life expectancies furnished by 21st Services by 13% prior to blending them with the life expectancy reports furnished by AVS.
As of June 30, 2013, the Company received 98 updated life expectancy reports from 21st Services that utilize its revised methodology. These life expectancies reported an average lengthening of life expectancies of 15.8% and, based on this sample, for the six months ended June 30, 2013, the Company increased the life expectancies furnished by 21st Services by 15.8% on the rest of its portfolio of life settlements prior to blending them with the life expectancy reports furnished by AVS. The Company expects to continue to lengthen life expectancies furnished by 21st Services that have not been re-underwritten using their updated methodology. Since the Revolving Credit Facility necessitates that the Company procure updated life expectancies on a periodic basis, the amount of policies that are lengthened by the Company in this manner will decrease over time and the fair value calculations in future periods will, accordingly, reflect the actual impact of the revised 21st Services methodology on a policy by policy basis as updated life expectancy reports are procured. At June 30, 2013, had the Company not applied a 15.8% increase to the 21st Services life expectancy reports, the portfolio would have increased from the reported amount of $265.8 million by $5.3 million.
Prior to the quarter ended June 30, 2013, when blending AVS and 21st Services’ life expectancy reports to derive a composite mortality factor, the Company would cap the higher mortality factor at an amount that was 150% above the lower mortality factor. This was done so as to reduce variances between life expectancies furnished by these two life expectancy providers. For the period ending June 30, 2013, the Company terminated its use of such a cap. The Company believes that the elimination of this cap is consistent with valuation methodologies employed by other market participants in connection with 21st Services’ revised methodology.
The discount rate used to calculate the present value of the life settlement contracts was determined based on the following at August 31, 2012:
The Company utilizes actuarial pricing methodologies and software tools that are built and supported by leading independent actuarial service firms such as Milliman USA and Modeling Actuarial Pricing Systems, Inc. (“MAPS”) for calculating expected returns. The MAPS (formerly Milliman) system is the most widely used platform for generating mortality reports. The Company uses life expectancy data from major non-related third-party life expectancy data providers. MAPS uses the life expectancy data and calculates a multiplier to the VBT data to reflect the current state of health of the insured. MAPS generates a report encompassing three individual valuations, two valuations based on two separate life expectancy reports and one valuation based on the weighted average of the two life expectancy values. The Company utilizes the valuation generated from the larger life expectancy value. The next step in the valuation process is determining the appropriate discount rate for the asset type. The discount rate incorporates current information about market interest rates, the credit exposure to the insurance company that issued the life settlement contracts and the Company’s estimate of the risk premium an investor in the policy would require. The Company believes that investors in alternative investments typically target yields averaging between 13 - 16% for investments of more than a 5 year duration. A 16% rate of return over a five year duration is a reasonable target for investments that are highly illiquid, relatively new and more volatile (i.e. life settlement contracts) than standard investments. In recent months, there have been a number of government investigations of several life settlement companies. These investigations, and subsequent Securities and Exchange Commision Inforcement Actions, in one instance, have caused a temporary dislocation in the life settlement market. Liquidity has tightened even further. The current market is still in a state of flux brought on by economic circumstances in Europe (traditionally the largest investors in the life settlement markets). These factors have caused investors in life settlements to demand a higher yield (averaging between 20 – 25%) because of the perceived risk in life settlements. Additionally, in the past 6 months, the Company, in the normal course of business, sold two fair value policies (each with face amounts below $15 million) to another investor, for a discount rate of between 19% and 22%. Additionally, in discussions with investors for potential debt offerings, and in discussions with other investors for potential sales of policies, the discount rate, for policies with face amounts below $15 million, indicated a rate of between 18% and 22%. For policies with face amounts above $15 million, of which there are fewer purchasers present, investors indicated a rate of between 25% and 28%. In light of all of these factors, the Company believes the perceived risk or uncertainty over these assets has changed and therefore the risk premium an investor would require has changed, therefore, in the third quarter ending May 31, 2012, management made a change in accounting estimate and adjusted its discount rate from 16% to 20%, for policies with face amounts below $15 million, and to 25%, for policies with face amounts above $15 million. Additionally, management has elected to report the more conservative of the values generated by the MAPS system, by utilizing the values linked to the longer life expectancy report. Management believes that a more conservative discount rate is appropriate due to the recent sales of policies and in discussions with other investors in the life settlements market.
The Company utilizes the MAPS system for valuations. MAPS is a generally accepted third party pricing system utilized by funds and investors engaged in the purchase and sale of life settlements.
Life expectancy reports are generated by third party medical underwriting firms, such as AVS, 21 st Century, EMSI and Fasano. These firms review medical data for an individual and grade using a series of debits and credits. The resulting values are used to generate a life expectancy value. The MAPS system utilizes this life expectancy data to calibrate underlying mortality curves generated for each individual case.
MAPS utilizes the appendixes to the VBT2008 report as a base for mortality projections. This chart established 25 unique values corresponding to 25 statistical values indicative of the next 25 years of a persons life. The value is the statistical probability that the individual will meet an untimely end in that given year. When a life expectancy provider produces a report, it indicates a value at which point a certain individual will achieve a 50% chance, statistically, of dying. This is calculated by randomly making 1000 simulations and calculating an exact point of death for each simulation. The point at which the cumulative deaths equal 500 is the median point or the life expectancy.
In order to calibrate this curve, a multiplier is formulated to cause the median mark to equal the life expectancy value provided by the life expectancy provider. For every case, 1000 simulations are run, each simulation resulting in a unique death month. For example, if a life expectancy equals 50 months, the underlying data would show 500 deaths prior to the 50th month and 500 deaths after the 50th month.
The MAPS system takes the results of the modified life expectancy curve and applies it to the premium and death benefit projections stream. For the 500 deaths prior to the 50th month, utilizing our previous example, it would apply the percentage of the simulated runs that died in each 12 month period to the death benefit and premium schedules to form estimated cash flows. The net present value, using a discount rate defined by the user, of the streams of values created by this method form the indicated value of the policy.