How does Berkshire Hathaway's insurance float work?

It's actually incorrect to say the duration, lag, tail or time horizon of Berkshire Hathaway's float is short/small. It varies depending on the type of risk that is being assumed. Float is essentially calculated by the following:

[W]e have calculated our float — which we generate in large amounts relative to our premium volume — by:

[A]dding net loss reserves, loss adjustment reserves, funds held under reinsurance assumed and unearned premium reserves, and the
[S]ubtracting insurance-related receivables, prepaid acquisition costs, prepaid taxes and deferred charges applicable to assumed reinsurance.
— Berkshire Hathaway's 2002 Shareholder Letter

I would say the float at Berkshire Hathaway comes in generally 3 different durations:

1. Short duration revolver, credit card-like "float" that funds its claims with incoming premiums in almost a controlled Ponzi fashion (i.e. GEICO).

For example, the combined ratio year after year for GEICO is around the 90-95% range. This means for every $100 coming in the door in premiums from policyholders, I am paying back policyholders for their auto insurance claims in aggregate $95. And premium volumes are stable and actually growing. What this means is that I can keep funding my liabilities/claims with incoming premiums without ever touching/selling/buying or doing anything with my invested assets. Furthermore, the time lag, duration even for this GEICO and personal auto insurance isn't that quick either. It varies. The lag in claims payments for auto insurance claims is longer for BI or bodily injury coverage, meaning it takes T + 3, 4 years to pay off more than 80% of all the claims from T = 0, the date on which the coverage started. Whereas the lag for other collision and comp coverage claims payments is 0-2 years or so, with most of those claims (80% or so) being paid out more quickly in the same year of the policy coverage due to less litigation and time needed to adjust and settle the claims than for a bodily injury claim. Here is the lag for Progressive in the chart below. GEICO's lag is actually longer (on purpose) but this is a good baseline chart for what auto insurance claims payment lag (i.e. duration of float before "maturity"):

2. Short-duration insurance lines, with potentially fluctuating premium levels from year to year.

Why would this fluctuation happen? For personal auto insurance above, it's a mandatory coverage people must buy regardless of whether the economy is doing well or not, as long as you need to drive a car. Furthermore, GEICO is the lowest-cost provider by being able to sell directly (not using independent insurance agents who require 10-15% commission). Progressive is moving more and more to direct, as right now their distribution channel is half direct and half independent agencies and they've often been able to squeeze the commissions to below 10-15% from their independent agencies because they drive so much volume. So if you have a product line that's not lowest-cost, you will need to drive the top-line up and down depending on how hard or soft the insurance market is in the current premium cycle to maintain underwriting profitability. Here, you can't take as much volatility asset side. NICO, the first insurance carrier that Buffett bought in 1967 for $8.6 million, and the carrier to which he attributes as the baby that is responsible for starting it all and is responsible for at least HALF of Berkshire Hathaway's $373 billion market-cap today, falls into this second bucket of float.

3. Long-tailed primary insurance and long-tailed reinsurance float.

Long-tailed reinsurance float has been the biggest creator of float for Berkshire Hathaway since the arrival of Ajit Jain in 1985. Buffett started getting serious with reinsurance in the early 80's with loss portfolio transfers (LPTs) but really started in earnest when Ajit Jain arrived in 1985. Jain helped Berkshire become the reinsurer of last resort by reinsuring hairy stuff that many reinsurers would not touch. But given a long-enough tail, even an incremental, additional 4-5% that Buffett could return on the assets backing those liabilities compared to the base 2-4% that most reinsurers were getting from the same amount assets, compounded over 10-30 years, makes a massive difference. Compounding for so many years essentially made those liabilities much smaller on an NPV basis for Berkshire Hathaway than for any other reinsurer (even if Berkshire only got 200-300 bps of incremental return).  A great example of such "float" liabilities are runoff portfolios like the acquisition of Equitas, a long-dated, seasoned book of asbestos claims from Lloyd's that was in runoff. The February 2014 loss portfolio transfer (LPT) reinsurance transaction which essentially reinsured a runoff block of variable annuities (VA) policies belonging to Cigna, generated $2-$3 billion of float, with excess loss above a certain amount being backstopped by Cigna. No one talks about that - they pay attention to the Heinz deal that happened a few weeks later. Where do you think some of that money for the Heinz deal came from?

Below is a great analysis of the role of Berkshire's float from Goldman Sachs in its research report on Berkshire Hathaway in 2010, on a page titled "The Power of Float: collect-now, pay-later":

"Put simply, float is the amount of money held by insurers on behalf of other parties – the majority of which is typically funds held to pay future claims. With premiums often collected well before losses are paid, the insurer can invest these funds for its own account. Additionally, for longer-tail lines of business, the timing differential can be decades long. Thus, while any given year will see its share of claim payments go up or down, the amount of float held by an insurer will stay relatively steady to its premium in-take. Thus, for an insurer that is not shrinking, the float can take the form of permanent capital.

When valuing Berkshire, we believe it is important to ascribe a value to the float. We believe that the amount of investable capital held by an above-average investor has a tangible value. There are two important distinctions, however:

The cost of funds

Over time, there is only value to the float if the investment returns exceed the cost of funds – which for an insurer would be the underwriting profit or loss. As an industry, insurance companies have historically operated at an underwriting loss (i.e. the premiums were less than the combined expenses and claims). Thus, it is Berkshire’s proven ability and stated willingness to focus on profitability (as opposed to growth) in its insurance operations that has allowed the cost of its float to be essentially zero over its multi-decade history. This is also one of the reasons we do not ascribe a value to the float generated by the other insurance companies in our industry, where the track record to assess historical profitability is for most companies too short of a time frame.

The “callability” of funds

“borrowed funds” can only be truly invested if there is limited ability for the lender to call the funds. This is what distinguishes BRK’s model from that of a “levered” investment fund – i.e. the funds, for the most part, cannot be redeemed by the lenders (i.e. the policyholders). The one caveat to this however is a catastrophic insurance scenario in which some portion of the float would need to be returned to policyholders. However, as we noted in the section above, BRK’s billions of dollars of cash on hand helps to protect against this scenario.

The value of float

If you were to invest a certain sum of borrowed capital – where the cost of such funds was zero, there was no “callability” to the funds by the lender, and the entirety of the investment returns accrued to your benefit – you would want to maximize the amount of borrowed capital. This is essentially the value proposition for being a shareholder in Berkshire Hathaway – where the float is the borrowed capital."

This chart below is a nice graphical illustration from Goldman Sachs, as well, showing the consistent spread below the long-term gov't bond rate. This spread of at least -5% or 500 bps below the 10-year treasury RISK-FREE rate is evidence of an amazingly cheap source of funding. Berkshire has been borrowing at a cost of capital or interest rate 5% BELOW THE 10-YEAR TREASURY RISK-FREE RATE. Think about that.

This means Berkshire could have used its float that's being borrowed at -5% interest and invested it now in the 10-year U.S. gov't bond which is currently yielding a risk-free 2% to achieve A RETURN of 7% PER YEAR BY INVESTING IN RISK-FREE 10-YEAR BONDS!

Berkshire has leveraged its insurance underwriting profitability to reduce %22borrowing costs%22.png

Note the cost of float over time below. It's actually lower than shown b/c if the cost of float was negative (i.e. positive underwriting profits), then it was recorded in this table as 0% cost of float:

Float cost

These are another nice couple charts from Goldman showing the size of the float and its contribution to the intrinsic value of Berkshire Hathaway:

Estimated contribution to intrinsic value over time
$ value of intrinsic value

Note below the contribution of float from Berkshire's primary insurance-related entities. I comment below re: the big boost of float from the mid-1980's from reinsurance deals, which generated particularly long-tailed, long-duration float for Berkshire:

Components of float historically
To understand Berkshire, therefore, it is necessary that you understand how to evaluate an insurance company. The key determinants are: (1) the amount of float that the business generates; (2) its cost; and (3) most critical of all, the long-term outlook for both of these factors.
— Warren Buffett, Berkshire Hathaway Inc., Annual Letter 1999