Basel endgame: whether mortgage insurance cuts LTV is where the capital actually moves

Risk-based capital on residential mortgages is driven off LTV ^[3], which is why a definitional question carries more weight than it sounds. Fold coverage into the ratio and an insured high-LTV loan is capitalized nearer to a loan with a large down payment, with nothing about the borrower, the property or the note changed. The case for doing so is made in an American Banker opinion column drawing on the mortgage insurance industry's comment letters ^[15].

Divide last year's insured origination volume by the borrowers it covered and the average lands near $389,000 ^[13]. At that size, one step in risk weight is worth real money per file, which is why the industry aimed at the LTV definition rather than at the size of a separate mitigant adjustment.

The net-versus-gross construction in the 25-year loan study deserves a close read ^[7]. Insured loans are measured after insurance recoveries; the comparison group is measured before any. That is the correct way to show what an insurance contract does, and a flattering way to argue how much capital relief it has earned. Severity is also one term in expected loss. Frequency is the other, and a severity result does not speak to it.

The sturdier argument is inconsistency inside the banks' own books. Under CECL, a bank already reflects mortgage insurance when it estimates credit losses ^[10], and Fannie Mae and Freddie Mac already receive MI benefit under their capital framework ^[2]. The same contract counts when a bank reserves and largely does not when it holds capital, and the ask is to close that gap and let pricing on retained loans follow ^[11].

On counterparty risk, the industry's answer is claims paid: more than $62 billion since the financial crisis, per its comment letters ^[8]. That is evidence about performance on the book insurers actually wrote, not about capacity in the stress a thinner capital requirement would be leaning on. Note also what the loss study is built from. It is GSE loan performance ^[7], which means loans underwritten to GSE standards, and the relief would attach to loans banks choose to keep on balance sheet ^[6]. Those two populations are not required to behave alike, and nothing in the published case establishes that they do.