Working Paper · 2026

Accounting Under Pressure: Recognition Rules, Insurer Bond Sales, and Real Investment

John Manuel Barrios, Andreas Neuhierl & Linda Schilling

Read the Paper (PDF) View on SSRN ← Back to Research

We study whether corporate bonds held by insurers suffer smaller price drops in crises because of the accounting rules applied to these bonds or because of the intrinsic characteristics of the holding institutions. In a three-period model, historical-cost (HCA) insurers can continue to report an impaired bond at amortized cost. In contrast, mark-to-market (MTM) insurers must immediately book the loss. Since only realized losses reduce regulatory capital and can trigger asset sales, insurers' selling decisions, and the prices they help sustain, differ precisely where their statutory accounting treatments diverge.

Using bond-level holdings data, we find that a one-standard-deviation increase in insurer ownership reduces crisis-period drawdowns by 0.3 percentage points, about 3 percent of the average crisis decline. Even within the same insurance group, a P&C subsidiary is roughly 7 percent more likely than its Life affiliate to sell a speculative-grade bond they both hold. This gap vanishes when their accounting rules are aligned. Moreover, for the average firm in our sample, a 10-percentage point increase in insurer ownership corresponds to roughly $165 million more in annual post-crisis investment. These results indicate that during a crisis, the recognition rules applied to bondholders determine which investors are forced to sell into declining markets and which are able to hold, with significant implications for the firms whose bonds they hold.

The Recognition Rule, Not the Institution

Whether a bond cascades into a fire sale or dissipates benignly depends on the accounting rule the marginal holder follows. Historical-cost (HCA) insurers can carry an impaired bond at amortized cost and hold through the shock; mark-to-market (MTM) insurers must book the loss, and when it binds their regulatory capital, they are forced to sell.

How the recognition rule splits insurer behavior in a crisis The same shock hits a bond. A historical-cost insurer carries it at amortized cost, its reported capital is unaffected, and it holds through the shock, so the price is supported. A mark-to-market insurer must book the loss immediately, its regulatory capital binds, it is forced to sell, and the price falls further. Same shock hits the bond HCA insurer · historical cost Carry the bond at amortized cost Reported capital is unaffected Hold through the shock Price is supported MTM insurer · mark to market Book the loss immediately Regulatory capital binds Forced to sell into the decline Price falls further
One shock, two responses. The recognition rule the marginal holder follows — not its liability profile — decides whether a distressed bond is held or sold. A historical-cost insurer’s reported capital does not move with a temporary price drop, so it can hold; a mark-to-market insurer must book the loss, and once that binds its regulatory capital it is forced to sell, driving the price down further.

A Within-Group Natural Experiment

Life and P&C subsidiaries in the same insurance group hold the same bond in the same quarter but face different statutory rules once the bond is downgraded into speculative grade (NAIC 3–5). A P&C subsidiary is roughly 7 percent more likely than its Life affiliate to sell such a bond — a gap that vanishes when their accounting treatments coincide.

How much more likely a P&C subsidiary is to sell than its Life affiliate, by rating region Within the same insurance group and quarter, a P&C subsidiary is essentially no more likely than its Life affiliate to sell a jointly held bond at investment grade, where both use the same accounting, and about 7 percent more likely at speculative grade, past the NAIC 2/3 threshold, where only the P&C subsidiary must mark the bond down. 0 2% 4% 6% 8% P&C selling rate, % above Life NAIC 2/3 threshold ≈ 0 +7% Investment grade NAIC 1–2 · rules aligned Speculative grade NAIC 3–5 · rules diverge
The same bond, the same quarter, two rulebooks. How much more likely a P&C subsidiary is to sell a jointly held bond than its Life affiliate in the same insurance group, in relative terms. At investment grade both carry the bond the same way and the two are essentially equally likely to sell; once the bond crosses the NAIC 2/3 threshold into speculative grade — where only the P&C arm must mark it down — the P&C subsidiary is about 7 percent more likely to sell, roughly 1.4 percentage points in absolute terms.

Insurer Ownership Stabilizes Prices

Bonds with more insurer ownership fall less in crises. A one-standard-deviation increase in insurer ownership cuts crisis-period drawdowns by about 0.3 percentage points (≈3 percent of the average crisis decline), with a larger effect under a shift-share instrument. Mutual-fund ownership shows no such stabilizing effect.

Reduction in crisis drawdown per one-standard-deviation increase in ownership A one-standard-deviation increase in insurer ownership reduces a bond’s crisis-period drawdown by about 0.3 percentage points, roughly 3 percent of the average crisis decline. Mutual-fund ownership shows no comparable effect. Insurer ownership 0.3 pp ≈ 3% of the average crisis decline Mutual-fund ownership ≈ 0 0 0.1 0.2 0.3 Reduction in crisis drawdown per +1 SD of ownership (pp)
Stabilization is specific to insurers, not to institutional ownership as such. Reduction in a bond’s crisis-period drawdown for a one-standard-deviation increase in ownership. Insurer ownership cushions the drawdown by about 0.3 percentage points — roughly 3 percent of the average crisis decline, and larger under a shift-share instrument — while mutual-fund ownership, marked to market by rule, shows no such effect.

Real Investment Consequences

The pricing effect flows through to the issuing firm: for the average firm in the sample, a 10-percentage-point increase in insurer ownership corresponds to roughly $165 million more in annual post-crisis investment.

Ten percentage points more insurer ownership maps to about 165 million dollars more annual post-crisis investment For the average firm in the sample, a 10-percentage-point increase in insurer ownership corresponds to roughly 165 million dollars more in annual investment in the years after a crisis. +10 pp insurer ownership ≈ $165M more in annual post-crisis investment for the average firm in the sample
From bond prices to real investment. The pricing effect passes through to the issuer: for the average firm in the sample, a 10-percentage-point increase in insurer ownership corresponds to roughly $165 million more in annual investment in the years following a crisis.

Theoretical work has long argued that mark-to-market accounting can propagate shocks through commonly held assets while historical cost can dampen them. Our contribution is to make the composition of a bond's holders the operative variable and to characterize when and by how much: the price a bond sustains in a crisis is monotone in two sufficient statistics — the insurer share and the historical-cost fraction within it — so an identical shock dissipates or cascades according to the recognition rule of the marginal holder.

Empirically, the within-holding-company design isolates the recognition rule from the identity of the institution, holding fixed liability duration, redemption pressure, investment horizon, and other confounds that ordinarily separate insurers from other bondholders. The results reframe recognition rules — a policy choice, unlike an insurer's liability profile — as first-order determinants of who is forced to sell into declining markets, of the prices that remain, and of the real investment of the firms whose bonds they hold.

Barrios, John Manuel, Andreas Neuhierl, and Linda Schilling. “Accounting Under Pressure: Recognition Rules, Insurer Bond Sales, and Real Investment.” Working Paper, 2026. Previously circulated as “Accounting Under Pressure: Recognition Rules, Bond Prices, and Real Investment.”
Read the Paper (PDF) View on SSRN ← Back to Research