Portfolio choice and insurance represent two fundamental markets for risk-sharing — one where individuals actively acquire risk exposure for potential gains, the other where they pay to transfer risk away.

In this post, I explore the mathematical equivalence between these choices, how both markets thrive on imperfection and share vulnerability to correlated risks, but also retain important differences that make each market unique. I have benefited from discussions with Xuelin Li.

Introduction

At their core, portfolio choice and insurance decisions both represent risk-sharing mechanisms, though they operate through different market structures. This equivalence can be expressed mathematically, as described in Armantier, Foncel, and Treich (2023).

The Insurance Problem

In the canonical insurance model (Mossin, 1968), an individual with wealth \(w\) faces a potential random loss \(\tilde{L}\). The individual can purchase insurance coverage at level \(\alpha\), where \(\alpha\) represents the proportion of the loss covered by insurance. The insurance premium is \(\alpha\pi\), where \(\pi\) is the premium rate for full coverage.

The individual’s objective is to maximize expected utility:

\[\max_{\alpha} \mathbb{E}u[w - \alpha\pi - (1-\alpha)\tilde{L}]\]

This expression has a clear economic interpretation: the individual’s final wealth is their initial wealth (\(w\)) minus the insurance premium paid (\(\alpha\pi\)) minus the portion of the loss they bear themselves (\((1-\alpha)\tilde{L}\)). When \(\alpha=0\), the individual has no insurance and bears the full loss; when \(\alpha=1\), the individual has full insurance and pays no losses out of pocket (but pays the full premium \(\pi\)).

The Portfolio Choice Problem

In the standard portfolio choice problem (Pratt, 1964), an investor with initial wealth \(w_0\) decides how much to invest in a risky asset offering a random excess return \(\tilde{X}\).

The investor’s objective is to maximize expected utility:

\[\max_{a} \mathbb{E}u[w_0 + a\tilde{X}]\]

Here, \(a\) represents the amount invested in the risky asset. The investor’s final wealth is their initial wealth (\(w_0\)) plus the return on their risky investment (\(a\tilde{X}\)). When \(a=0\), the investor takes no risk; as \(a\) increases, the investor takes on more risk.

The Equivalence

These two problems are mathematically equivalent under the following change of variables:

• \(w_0 = w-\pi\) (initial wealth in the portfolio problem equals wealth minus full insurance premium)
• \(a=1-\alpha\) (amount invested in risky asset equals proportion of loss not insured)
• \(\tilde{X}=\pi-\tilde{L}\) (excess return equals insurance premium minus potential loss)

This transformation has a fascinating economic interpretation. Not buying insurance is equivalent to making a risky investment: When you decline insurance (\(\alpha = 0\)), it’s mathematically equivalent to fully investing (\(a = 1\)) in a risky asset whose payoff is the saved premium (\(\pi\)) minus the potential loss (\(\tilde{L}\)).

On the other hand, full insurance is equivalent to risk-free investing. In insurance, you start with full exposure to risk and pay to reduce it; in portfolio choice, you start with no risk exposure and pay (by accepting volatility) to increase it. But the underlying mathematical structure—and therefore the optimal decision logic—is identical.

But does this mathematical equivalence extend to how these markets function at the broader level?

Parallel #1: The Need for Imperfection

Fundamentally, both insurance markets and the investment industry share a fascinating property: they require some level of imperfection to function properly.

• In asset markets, we have the Grossman-Stiglitz paradox (Grossman and Stiglitz, 1980). If markets were perfectly efficient (all information reflected in prices), there would be no incentive for anyone to gather information or trade actively. The paradox is that market efficiency requires active participants, but perfect efficiency would eliminate the profit motive for such participation. Asset markets therefore exist in an equilibrium of partial inefficiency—just enough to reward informed traders while preventing outsized arbitrage opportunities.

• Insurance markets face a similar paradox. If insurers could perfectly predict individual outcomes, true risk-pooling would collapse. If an insurer knew exactly when each person would die, they would charge each individual precisely the present value of their future payout. This would transform insurance from risk transfer to mere saving, eliminating its core function. Similarly, if insurers could predict with certainty which specific houses would burn down this year, they would charge those homeowners the full replacement cost upfront, while charging nothing to those whose houses would remain safe. This would make insurance unaffordable precisely for those who need it, effectively eliminating the market. The industry thrives because outcomes remain probabilistic rather than deterministic, allowing risks to be pooled across many individuals.

In essence, both markets depend on enough uncertainty to sustain active participation. Complete information would destroy both.

Parallel #2: Correlated Risks Create Systemic Stress

Both industries also face similar challenges when risks become correlated.

• Insurance markets struggle when many policyholders face the same adverse outcome simultaneously. Climate change exemplifies this problem: as weather-related disasters become more frequent and severe, insurers face potentially catastrophic, correlated losses. Traditional insurance works best when risks are independent and can be pooled effectively. When risks become systemic, the pooling mechanism breaks down.

• Similarly, financial markets face their greatest stress when investor behaviors correlate. When many market participants rush for the exits simultaneously (as in a financial crisis), liquidity evaporates, and prices can disconnect dramatically from fundamental values. Crowded trades and fire sales amplify market stress precisely because diversification benefits disappear when behaviors become correlated.

In both domains, correlation undermines the ability to diversify or share risk efficiently, threatening the stability of the entire system.

Difference #1: Who Knows More?

Despite these parallels, there’s a crucial asymmetry in the information landscape of these markets. And this difference creates opposing regulatory concerns: asset markets worry that sellers know too much; insurance markets worry that buyers know too much.

• In asset markets, the supply side (issuers of securities) typically has superior information. Corporate insiders understand their business prospects better than outside investors. This creates a classic adverse selection problem (Akerlof, 1970) — investors must worry that they’re buying assets the issuer knows are overvalued.

• In insurance markets, the information advantage usually lies with the demand side (Rothschild and Stiglitz, 1976). Individuals know more about their own health, driving habits, or property risks than insurers can observe. Insurers must worry about attracting high-risk customers who know they’re bad risks. This drives practices like medical underwriting, deductibles, and exclusions for pre-existing conditions.

This informational asymmetry leads to fundamentally different regulatory approaches:

• Securities regulation focuses on disclosure requirements and insider trading restrictions to protect less-informed investors from exploitation by better-informed issuers and insiders. The regulatory philosophy centers on leveling the information playing field through mandatory disclosures, waiting periods, and penalties for trading on non-public information.

• Insurance regulation instead emphasizes consumer protection and solvency requirements. Regulators allow practices like risk classification and medical underwriting that might seem discriminatory in other contexts, precisely because they help insurers manage adverse selection. Simultaneously, regulations ensure insurers maintain sufficient capital reserves to fulfill their promises to policyholders.

In the modern era, one could plausibly argue that insurance markets could be evolving toward the patterns we observe in asset markets. As insurers develop increasingly sophisticated predictive models and access larger datasets (including wearable technology data, telematics in vehicles, and smart home devices), the information asymmetry may shift. This is the situation that Brunnermeier, Lamba, and Segura-Rodriguez (2023) call “inverse selection.”

Difference #2: Primary vs. Secondary Markets

Another point of difference lies in the existence of robust secondary markets.

• In asset markets, vibrant secondary trading allows continuous risk reallocation. Investors can adjust positions at virtually any time, responding to new information or changing risk preferences. This creates liquidity, price discovery, and flexibility for market participants.

• Insurance markets, however, predominantly function as primary markets only. Once an insurance contract is written, policyholders generally cannot sell their protection to third parties. If circumstances change, the typical options are limited: continue the coverage, cancel (often with penalties), or in some cases, renegotiate directly with the original insurer.

This difference reflects a deeper structural distinction in how risk is transferred:

• Asset markets operate through decentralized trading mechanisms where many buyers and sellers interact directly through exchanges or electronic platforms. Risk transfer occurs continuously through price discovery without requiring a central intermediary to pool risks.

• Insurance markets function through centralized pooling mechanisms where an institutional intermediary (the insurer) aggregates many similar risks. The insurer serves as both the counterparty and risk manager, rather than simply facilitating peer-to-peer risk transfers.

It turns out that this asymmetry leads to some important implications.

For example, secondary markets for securities allow continuous incorporation of new information into prices. Insurance premiums, by contrast, typically adjust only at renewal periods, potentially creating lags in risk pricing. In addition, investors can dynamically hedge and adjust exposure as conditions change. Insurance policyholders face more rigid structures, limiting their ability to optimize protection as their risk profile evolves.

Of course, some limited exceptions exist — life settlements allow some life insurance policies to be sold to third parties, and catastrophe bonds create tradable securities linked to insurance risks. However, these represent narrow segments rather than comprehensive secondary markets for insurance protection.

These differences suggest that even mathematically equivalent decisions may lead to divergent outcomes in practice due to the institutional frameworks within which these choices occur.

Difference #3: Clean vs. Noisy Leverage

Another key difference lies in how each market allows participants to scale their exposure to risk.

• In asset markets, leverage provides a clean mechanism for scaling systematic exposure. An investor holding the market portfolio can borrow to increase exposure to aggregate risk without adding idiosyncratic noise. This ability to separate systematic bets from idiosyncratic risk makes leverage a powerful and well-defined tool.

• In insurance markets, the closest analog to leverage is to under-insure — to reduce coverage and thereby increase exposure to risk. But this scaling is “noisy”: it exposes the individual to both systematic risk (economy-wide or correlated events) and idiosyncratic risk (personal losses). There is no mechanism to selectively scale only the systematic component.

This structural difference matters. In asset markets, leverage amplifies clean, priced risk; in insurance, “leverage” entangles the individual with unpredictable, non-compensated idiosyncratic variation. As a result, while portfolio choice naturally incorporates leverage decisions, insurance decisions do not.

Conclusion

The symmetry between portfolio choice and insurance extends far beyond their mathematical equivalence. Both markets thrive on imperfection, struggle with correlation, and deal with adverse selection—albeit from opposite directions.

The obvious takeaway from this is that understanding these parallels offers insights for financial planning and regulation. But more interestingly, technological advances are challenging these traditional paradigms. Big data and predictive analytics are reducing information asymmetries in insurance markets, while algorithmic trading and information dissemination are changing the dynamics of market efficiency.

It would be interesting to see how these technological transformations reshape the symmetry between these two domains.