7 Liquidity Creation Through Information Insensitivity

The Gorton–Pennacchi Theory of Banks as Creators of Safe Securities

7.1 Overview

Chapter 6 developed the Kashyap–Rajan–Stein view of why deposit-taking and loan-commitment making belong under one roof: because both activities require costly liquidity buffers, and combining them allows sharing. The synergy in that chapter is fundamentally about costs — a bank that does both needs less liquidity in total than two specialized institutions.

This chapter develops a complementary but distinct answer. Gorton and Pennacchi (1990) ask a more foundational question that KRS leaves unanswered: why do agents want a safe, information-insensitive security in the first place?

A security is information-insensitive when its value does not depend on private information that some agents know and others do not. Bank deposits are the leading example: a depositor who holds a claim promising $\$1{,}000$ receives $\$1{,}000$ regardless of what the bank’s loan portfolio turns out to be worth. An insider who knows the loans are performing badly gains nothing by trading deposits against an uninformed depositor — the payoff is fixed. This is exactly what makes deposits safe to use as a medium of exchange: you can accept them without worrying that the counterparty knows something you don’t. Treasury bills share this property (backed by the government’s taxing power, not any specific risky project); stocks do not (their value depends entirely on information about firm prospects that insiders may know before you do).

GP’s answer to why agents demand such securities is rooted not in liquidity management but in the economics of asymmetric information in trading markets.

The central insight is that deposits protect uninformed traders from losses to better-informed insiders. When informed agents know more about asset values than ordinary investors, they can exploit the uninformed by trading against them in risky asset markets. A bank that pools risky assets and issues riskless debt against them gives uninformed agents a safe medium for transacting — one whose value cannot be manipulated by insiders because it is insensitive to the private information that insiders hold.

This logic explains several facts that the KRS framework takes as given:

Why bank liabilities are debt (safe, fixed-claim) rather than equity
Why the same institution holds risky loans and issues safe deposits
Why bank deposits, government bonds, and high-grade corporate debt all serve as “money-like” assets
Why the safety of a transactions medium depends critically on its information sensitivity

The chapter proceeds in four parts. Parts 1 and 2 develop the Gorton–Pennacchi model — the problem, the solution, and the alternatives. Part 3 examines the role of government when private contracting fails. Part 4 connects the GP framework to the modern literature on information insensitivity and safe assets, tracing its implications through the 2008 financial crisis and recent debates about the design of liquid financial markets.

7.1.1 Learning Objectives

By the end of this chapter, you should be able to:

Explain why uninformed investors face losses when trading against informed insiders in an asset market, and why this creates a demand for safe, information-insensitive securities.
Describe the Gorton–Pennacchi model setup and derive the insider trading equilibrium.
Explain how a bank intermediary — by splitting asset cash flows into riskless debt and risky equity — can break the insider coalition and restore efficient trade.
Identify the conditions under which private bank deposits can be made riskless, and explain the role of government deposit insurance when those conditions fail.
Explain what “information insensitivity” means, and connect the GP model to the broader literature on safe assets and the design of money-like securities.

7.2 Part 1: The Problem — Exploiting Uninformed Investors

7.2.1 A Different Question than Diamond–Dybvig

The Gorton–Pennacchi starting point is superficially similar to Diamond–Dybvig (Chapter 3). There are agents with uncertain consumption preferences — some will want to consume early, some late — and there are risky assets. But the key difference is this: Gorton and Pennacchi do not rule out a stock market. Agents are free to trade equity claims on the underlying capital at any time.

As Jacklin (1987) showed, when a stock market is available, equity trading weakly dominates bank deposits under full information: agents can achieve at least as good an allocation simply by trading equity claims at $t = 1$, with no strict need for a bank. This raises a challenge for any theory of banking that relies on liquidity provision alone — if markets work well, why would anyone use a bank?

GP’s answer: because a stock market, when populated by better-informed insiders, is not a safe place for ordinary investors to trade. The problem is not that markets are missing; it is that the equilibrium favors informed agents at uninformed agents’ expense. Informed agents can exploit uninformed agents who must trade for liquidity reasons. The bank — by issuing debt that does not expose the holder to the insiders’ information advantage — gives uninformed agents a safe alternative.

7.2.2 Setup and Assumptions

The economy spans three dates $t = 0, 1, 2$ with a single consumption good.

The Three Agent Types in the GP Model

1. Informed agents (Type I). These agents have known consumption preferences at $t = 0$: they consume at $t = 2$. Crucially, at $t = 1$ they observe both the return on capital ($R_H$ or $R_L$) and the proportion of early consumers ($\omega_h$ or $\omega_l$). They have aggregate capital endowment $M$.

2. Early consumers (a subtype of liquidity traders). These agents discover at $t = 1$ that they want to consume immediately (utility $U = C_1$). They must sell their capital for the consumption good.

3. Late consumers (the other subtype of liquidity traders). These agents discover at $t = 1$ that they prefer to consume at $t = 2$ (utility $U = C_2$). They may sell their $t = 1$ endowment of the consumption good for capital, or store it.

Agents of types 2 and 3 collectively form the set of liquidity traders (total mass $N$, large relative to informed agents). At $t = 0$, liquidity traders do not know which type they will be.

Endowments. Each liquidity trader receives 1 unit of the capital good at $t = 0$ and $e_1$ units of the consumption good at $t = 1$. Informed agents receive aggregate capital $M$ at $t = 0$ and consumption good $Me_2$ at $t = 2$.

Capital Good vs. Consumption Good: An Intuition

Think of the capital good as an illiquid productive investment — farmland, or a share in a private firm. You commit it at $t = 0$ and it pays off at $t = 2$, but it cannot be consumed directly in the meantime. The consumption good is cash: something spendable right now.

At $t = 1$ each liquidity trader also receives a paycheck ($e_1$ units of cash). Early consumers — those with utility only over $t = 1$ consumption — wish to exchange their farmland for cash, consuming $e_1$ plus the sale proceeds $p_{ij}$. Late consumers — those with utility only over $t = 2$ consumption — may use their paycheck to buy farmland (earning $R_H$ or $R_L$ at harvest) or simply store it. Informed agents are the insiders who already know at $t = 1$ whether the harvest will be good or bad. Liquidity traders do not.

Technology. Each unit of capital invested at $t = 0$ produces either $R_H$ or $R_L$ units of the consumption good at $t = 2$, with equal probability $\frac{1}{2}$, where $R_H > R_L > 0$. All capital units face the same aggregate shock — they all produce $R_H$ together or $R_L$ together. Let $\bar{R} = \frac{1}{2}(R_H + R_L)$ denote the expected return.

Consumption goods received at $t = 1$ can also be stored: one unit stored at $t = 1$ yields one unit at $t = 2$.

Preferences over early/late consumption. The proportion of liquidity traders who are early consumers is either $\omega_h$ (high, probability $q_h$) or $\omega_l$ (low, probability $q_l = 1 - q_h$), with $\omega_h > \omega_l$. This is the same taste-shock structure as Diamond–Dybvig, now augmented with a productive-asset return shock observed only by informed agents.

The information asymmetry. At $t = 1$:

Informed agents observe the full state $\{i, j\}$ where $i \in \{h, l\}$ (high/low proportion of early consumers) and $j \in \{H, L\}$ (high/low capital return).
Liquidity traders observe only their own type (early or late) — not the aggregate proportion of early consumers nor the return on capital.

This is the key asymmetry. Liquidity traders need to trade at $t = 1$ (early consumers must sell capital; late consumers must decide whether to sell endowment or store it), but they do so without knowing the state of the world.

Timeline.

7.2.3 The Full-Information Benchmark

Before introducing asymmetric information, let us establish the efficient benchmark where all agents observe the state $\{i, j\}$ at $t = 1$.

There are four states: $\{l, H\}$, $\{h, H\}$, $\{h, L\}$, and $\{l, L\}$. Let $p_{ij}$ denote the equilibrium price of one unit of capital in terms of consumption goods in state $\{i, j\}$.

At $t = 1$, early consumers wish to exchange their capital endowment for the consumption good; late consumers decide whether to use their $e_1$ paycheck to buy capital or store it. Market clearing requires:

\[N\omega_i \cdot p_{ij} \;\leq\; N(1 - \omega_i)e_1 \tag{1}\]

The left side is the total consumption good demanded by early consumers ($N\omega_i$ of them, each selling 1 unit of capital at price $p_{ij}$); the right side is the maximum supply from late consumers ($N(1-\omega_i)$ of them, each with cash $e_1$).

Two cases arise:

No storage (high capital return): If capital earns $R_H$, which exceeds the storage return of 1, late consumers sell all their endowment. Equation (1) holds with equality:

\[p_{ij} = \frac{e_1(1 - \omega_i)}{\omega_i} \tag{2}\]

Storage (low capital return): If $R_j = R_L$ falls to the storage return threshold, late consumers are indifferent between buying capital and storing, so:

\[p_{ij} = R_j \tag{3}\]

GP assume that in states where capital returns are high ($j = H$), no storage occurs, while in states with low returns ($j = L$), storage occurs. This requires:

\[R_H \;\geq\; \frac{e_1(1 - \omega_l)}{\omega_l} \;>\; \frac{e_1(1 - \omega_h)}{\omega_h} \;>\; R_L \tag{4}\]

The middle inequalities reflect the fact that when fewer people need cash ($\omega_l < \omega_h$), early consumers are scarce relative to late consumers and the no-storage price of capital is higher. The outer inequalities do the real work:

Left ($R_H \geq e_1(1-\omega_l)/\omega_l$): Even in the state with fewest early consumers — where capital’s no-storage price is highest — the return $R_H$ still exceeds that price. So late consumers always prefer to buy capital over storing when returns are high. No storage occurs in either high-return state.
Right ($e_1(1-\omega_h)/\omega_h > R_L$): Even in the state with most early consumers — where capital’s no-storage price is lowest — that price still exceeds $R_L$. So in low-return states, capital is a worse deal than storing endowment and late consumers store instead, pinning $p_{iL} = R_L$.

Why informed agents do not trade in the full-information benchmark. In high-return states ($j = H$), the no-storage equilibrium price satisfies $p_{iH} = e_1(1-\omega_i)/\omega_i \leq R_H$. An informed agent who sold capital at $p_{iH}$ and stored the proceeds would earn $p_{iH} \leq R_H$ at $t=2$ — no better than simply holding capital. So informed agents have no incentive to sell. In low-return states ($j = L$), the storage equilibrium pins $p_{iL} = R_L$. An informed agent who sold at $R_L$ and stored would also earn $R_L$ — exactly the same as holding. They are indifferent, with no strict incentive to trade strategically. In neither case can insiders profit from their information.

Both informed agents and liquidity traders achieve expected utility equal to $\bar{R}$ per unit of capital endowment. This is the efficient benchmark.

7.2.4 The Insider Trading Equilibrium

Now suppose informed agents observe the state $\{i, j\}$ but liquidity traders do not. Informed agents form a small group (relative to the $N$ liquidity traders) and are assumed able to collude — to form what GP call an Insider Coalition.

The coalition’s strategy is to choose how much capital to supply to the market in each state. The key is that the coalition can make certain states indistinguishable to uninformed traders by controlling the price.

The Insider Strategy: Mimicking a Good State in a Bad One

Consider state $\{l, L\}$: few early consumers, low capital return. In the full-information equilibrium, capital would trade at $p_{lL} = R_L$ (storage occurs). But informed agents know capital is worth little and want to sell.

The coalition’s strategy: mimic state $\{h, H\}$ — many early consumers, high return — where the equilibrium price is $p^* = e_1(1-\omega_h)/\omega_h$.

To mimic $\{h,H\}$ in state $\{l, L\}$, the coalition supplies quantity $M_{lL}$ of capital alongside the $N\omega_l$ early consumers:

\[N\omega_l \cdot p^* + M_{lL} \cdot p^* \;=\; N(1-\omega_l)e_1\]

Solving for the coalition’s supply:

\[M_{lL} \;=\; \frac{N(\omega_h - \omega_l)}{1 - \omega_h} \tag{5}\]

Late consumers observe price $p^*$ and cannot distinguish state $\{l, L\}$ from state $\{h, H\}$. Believing (rationally, given available information) that they may be in the good state, they sell their endowment — receiving capital worth $R_L$ at $t = 2$ when they believed they were buying something worth $R_H$. The insiders receive good endowment in exchange for bad capital.

GP prove this coalition strategy is self-enforcing: no individual member benefits from deviating. If any informed agent tries to sell additional capital beyond their coalition share, the price would move and reveal the true state — at which point late consumers refuse to trade and the coalition’s profits evaporate. The threat of discovery disciplines coalition members.

Proposition 1 (Gorton–Pennacchi): Insider Trading Equilibrium

Under conditions on model parameters, there exists an Imperfectly Competitive Rational Expectations Equilibrium in which:

States $\{l, H\}$ and $\{h, L\}$ have fully revealing prices.
States $\{h, H\}$ and $\{l, L\}$ share the same price $p^* = e_1(1-\omega_h)/\omega_h$.
In state $\{l, L\}$, the insider coalition supplies $M_{lL}$ units, profiting at uninformed traders’ expense.

The expected utility gain to informed agents is:

\[\Delta U_{\text{informed}} \;=\; \frac{q_l}{2} \cdot w_m \cdot (p^* - R_L) \;>\; 0\]

where $w_m = M_{lL}/M$ is the fraction of informed capital sold in the bad state.

The corresponding expected utility loss to liquidity traders is:

\[\Delta U_{\text{uninformed}} \;=\; -\frac{q_l(\omega_h - \omega_l)}{2(1-\omega_h)}\left(p^* - R_L\right) \;<\; 0\]

The wedge $p^* - R_L > 0$ is the unit trading loss that uninformed late consumers incur by selling endowment for overpriced capital.

Economic interpretation. Informed agents are functioning as insiders: they know capital is bad (state $\{l,L\}$) but sell it at the good-state price. The uninformed buy capital they think is valuable. The mechanism is information concealment through strategic market supply — a formal model of the intuition behind insider-trading prohibitions.

Crucially, the uninformed agents’ loss does not reflect irrationality. They are optimizing given their information: observing price $p^*$ and believing (with positive probability) that the state is $\{h, H\}$, it is rational to buy capital. The problem is that this rational behavior is exploitable by agents who know with certainty that the state is $\{l, L\}$.

Connection to Akerlof’s Lemons Market

GP’s insider-trading result is structurally similar to Akerlof’s (1970) lemons problem, but in a secondary trading rather than a new-issue context. In Akerlof, sellers of used cars know quality while buyers do not, leading to adverse selection. In GP, informed agents know asset returns while liquidity traders do not, and exploit this by strategically supplying bad assets at good-asset prices.

The connection matters for the solution: just as Akerlof suggested that warranties — credible commitments about quality — could cure the lemons problem, GP show that a credible commitment to a safe payoff (bank deposits) can cure the insider trading problem. The bank is a warranty on the value of the claim.

7.3 Part 2: Private Liquidity Creation

7.3.1 The Intermediary Contract: Splitting Cash Flows

Now suppose that liquidity traders, anticipating their losses in the stock market, are allowed to contract. At $t = 0$, they can form a bank intermediary that pools capital and issues two types of claims: deposits (safe debt) and equity.

The bank collects capital from $N_I$ liquidity traders (who receive deposits) and $M_I$ units of informed-agent capital (who receive equity). Total bank capital is $A = N_I + M_I$. The bank’s total return at $t = 2$ is $A \cdot R_j$.

The critical constraint on the contract is that deposits must be riskless:

\[D \cdot R_D \;\leq\; A \cdot R_L \;=\; (D + E) \cdot R_L \tag{6}\]

where $D$ is the total deposit principal and $R_D$ is the promised gross return. Constraint (6) says that even in the worst state ($R_L$), the bank can pay depositors in full. Rearranging:

\[\frac{E}{D} \;\geq\; \frac{R_D - R_L}{R_L} \tag{7}\]

The equity cushion $E/D$ must be large enough that, in the low state, equity is wiped out before any loss falls on depositors. This is the classic equity-as-buffer function: equity absorbs downside risk so that deposits are safe.

The key economic insight concerns what a deposit holder needs to know. An agent who will receive $D \cdot R_D$ regardless of whether the capital return is $R_H$ or $R_L$ has no need to monitor or observe the capital return. The deposit is information-insensitive: its value does not depend on the private information known only to informed agents.

Equity, by contrast, is information-sensitive: it pays $A \cdot R_H - D \cdot R_D$ in the good state and zero in the bad state. Equity holders care deeply about the realized return. This is why informed agents are the natural equity holders: they have information relevant to equity value, and they earn a return from contributing it to the bank.

7.3.2 Breaking the Insider Coalition

The key question is whether informed agents will voluntarily become equity holders in the bank, rather than continuing to operate the Insider Coalition.

GP prove that the answer is yes — under conditions on $M/N$ — through two propositions:

Proposition 2 (Gorton–Pennacchi): Nonexistence of Stock Market Insider Equilibrium

Consider a small number of liquidity traders forming a bank (while most remain in the stock market initially). If the ratio of informed-to-uninformed capital $M/N$ is sufficiently large, there exists a deposit rate $R_D$ such that:

Deposits are riskless (condition (6) holds).
Liquidity traders prefer to hold bank deposits rather than stock market equity.
Individual informed agents prefer to hold bank equity rather than join the Insider Coalition.

In other words: if informed agents have enough capital to provide the required equity cushion, the insider trading equilibrium unravels.

Why do informed agents defect from the coalition? The coalition’s profits depend on having uninformed agents to trade against. If liquidity traders leave the stock market and hold deposits instead, the stock market becomes a lemons market: the only sellers of capital in the stock market are insiders selling in bad states. Late consumers know this and will only pay a price reflecting $R_L$ for any stock market capital. With no one to exploit, the insider coalition earns nothing.

Given that the coalition earns nothing once uninformed agents hold deposits, informed agents prefer bank equity — which earns a competitive return on their capital — to stock market equity with a worthless insider coalition.

Proposition 3 (Gorton–Pennacchi): Existence of an Intermediary Equilibrium

If $M/N$ is sufficiently large, there exists an equilibrium where:

All liquidity traders hold riskless bank deposits.
All informed agents contribute equity capital to the bank.
The stock market Insider Coalition has no profitable strategy.
All agents achieve expected utility arbitrarily close to the full-information benchmark.

The required deposit rate satisfies $R_D \leq \bar{R}$: depositors earn close to the expected asset return and are fully protected from downside risk.

Why is $M/N$ large enough a necessary condition? The bank’s deposits can only be riskless if informed-agent equity is sufficient to cover depositors in the worst state. From constraint (7):

\[M_I \;\geq\; D \cdot \frac{R_D - R_L}{R_L}\]

The more informed capital relative to the mass of liquidity traders, the more easily the bank can capitalize equity to backstop deposits. If informed agents are scarce, private deposits may not be achievable without government support (addressed in Part 3).

7.3.3 The Role of the Bank: A Summary

The bank’s role in the GP model is fundamentally different from the roles identified in earlier chapters:

Earlier chapter	Bank’s role
Diamond–Dybvig (Ch. 3)	Risk-sharing: insurance against uncertain preferences
Diamond (Ch. 5)	Delegated monitoring: efficient agent for observing borrowers
KRS (Ch. 6)	Liquidity buffer sharing: shared cost of reserves
Gorton–Pennacchi (this chapter)	Information intermediation: creating safe securities that protect uninformed traders

In each case, the bank holds risky assets and issues safe liabilities — but the reason this arrangement is efficient differs. The GP reason is that the bank absorbs information risk on behalf of its depositors: by pooling capital into a bank and holding riskless deposits, liquidity traders no longer need to worry about whether someone smarter than them is on the other side of the trade.

KRS vs. GP: Two Different Rationales for the Same Balance Sheet

KRS and GP both explain why a bank holds risky loans and issues deposits — but they identify completely different mechanisms.

In KRS, the rationale is about cost: combining deposits and loan commitments under one roof reduces the total liquidity buffer required, because the two demands for cash are imperfectly correlated. You can prove the efficiency gain with a formula and measure it in dollar terms.

In GP, the rationale is about information: the bank’s risky asset portfolio, when partly financed by informed-agent equity, makes it possible to issue deposits whose value does not depend on private information. This is a necessary complementarity — risky loans, equity, and riskless deposits must all coexist for the information problem to be solved — but it is not a cost saving in the KRS sense. There is no formula showing that the combined institution needs less of some input. Instead, the combined structure is what enables a riskless claim to exist at all.

Together, the two papers answer different questions about the same balance sheet. KRS: why hold loans and issue demandable deposits rather than specializing? GP: why issue debt (safe, fixed-claim) rather than equity against those loans? The answers are independent and mutually reinforcing.

7.3.4 Corporate Debt and Money Market Mutual Funds as Alternatives

The GP framework does not make bank intermediation the unique solution. The same logic applies to any institution that can split risky cash flows into safe debt and risky equity.

A firm with sufficient assets can issue riskless corporate debt — commercial paper or bonds backed by a large enough equity cushion — that serves the same transactions purpose as bank deposits. Liquidity traders can hold corporate debt instead of bank deposits and be equally protected from insider trading.

GP explicitly make this point (Section IV.B of the paper): if firms can issue sufficiently safe debt, uninformed agents need never interact with the stock market. A money market mutual fund holding a diversified portfolio of such debt extends the idea further — diversification across issuers reduces the risk that any single firm’s private information contaminates the fund’s share value.

Money Market Mutual Funds as GP-Style Intermediaries

A money market mutual fund is structured exactly as the GP model predicts:

The fund holds a diversified portfolio of short-maturity, high-grade debt (commercial paper, Treasuries, agency debt).
The fund issues shares designed to maintain a stable $1 net asset value (NAV) — a riskless claim on the portfolio.
The stable NAV is informationally insensitive: a holder of MMMF shares does not need to know the credit quality of each underlying issuer, because diversification and credit quality screening have already been done.

GP note in their conclusion that MMMFs “provide a potentially important transactions medium” and that their analysis gives “no reason to prefer bank debt over money market mutual funds.” This is a striking claim in 1990 and anticipates the debates about narrow banking that intensified after 2008.

7.4 Part 3: When Private Contracting Fails — Government’s Role

7.4.1 The Limits of Private Liquidity Creation

The GP intermediary equilibrium requires $M/N$ to be large enough: informed agents must hold sufficient equity to make deposits riskless. There are natural circumstances under which the private solution is insufficient:

Insufficient informed-agent capital. If $M$ is small relative to $N$, informed agents cannot provide enough equity to backstop all deposits.
Sufficiently risky assets. Making deposits riskless requires the equity cushion to cover depositors in the worst case — the bad state $R_L$. From condition (7), $E/D \geq (R_D - R_L)/R_L$: the smaller $R_L$ is, the larger the required cushion. In this sense “risky” means a severe downside, not high variance per se — $R_H$ does not appear in the constraint at all. If $R_L \to 0$, the equity requirement grows without bound and private riskless deposits become infeasible.
Asset opacity or fraud. If uninformed depositors cannot verify that the bank holds the assets it claims to hold, they cannot be confident deposits are riskless even when formally they should be.

In these cases, private bank intermediation either fails to create fully riskless deposits or is too costly to operate at scale. Government intervention can fill the gap.

7.4.2 Deposit Insurance: Government as Risk-Absorber

Suppose deposits are risky — $D \cdot R_D > A \cdot R_L$ — because the equity cushion is insufficient. A government deposit insurance system can replicate the allocation achieved under riskless private deposits through a tax-and-subsidy scheme:

When the bank fails (state $R_L$): The government taxes late-consuming agents — those with income at $t = 2$, the only agents with resources available when payoffs are realized — to cover the shortfall $T = D \cdot R_D - A \cdot R_L$, paying depositors in full.
When the bank survives (state $R_H$): The government collects a fair insurance premium $\Theta$ from the bank, redistributed to agents.

Proposition 4 (Gorton–Pennacchi): Deposit Insurance Replicates Riskless Deposits

With a fair insurance premium, the government’s tax-subsidy scheme gives all agents the same expected utility as in the riskless-deposit equilibrium.

A fair insurance premium equates expected premium collected with expected payout. Since the model assumes $R_H$ and $R_L$ each occur with probability $\frac{1}{2}$ (the equal-probability technology assumption from the setup), the bank fails with probability $\frac{1}{2}$ and survives with probability $\frac{1}{2}$:

\[\underbrace{\frac{1}{2} \cdot \Theta D}_{\text{expected premium in good state}} \;=\; \underbrace{\frac{1}{2} \cdot T}_{\text{expected payout in bad state}}\]

The $\frac{1}{2}$s cancel, giving:

\[\Theta D \;=\; T \;=\; D \cdot R_D - A \cdot R_L\]

This implies a fair premium rate per dollar of deposits:

\[\Theta \;=\; R_D - \left(1 + \frac{E}{D}\right) R_L \tag{8}\]

The premium equals the promised deposit rate minus the risk-adjusted recovery rate. When deposits are just barely riskless ($D \cdot R_D = A \cdot R_L$), the fair premium is zero — no insurance is needed.

The government’s advantage over private contracting is its ability to tax. Private parties cannot credibly commit to covering depositors from future income when the bank fails; the government can, because it has coercive taxing power. This is a genuine comparative advantage — not merely a regulatory artifact.

The GP Rationale for Deposit Insurance

GP provide a clean positive rationale for deposit insurance distinct from the standard Diamond–Dybvig motivation (preventing runs) or the consumer-protection motivation (protecting small depositors from bank mismanagement).

In the GP framework, deposit insurance serves one function: it creates a riskless transactions medium when private intermediation cannot. The government is issuing its own guarantee on top of the bank’s risky assets, converting a potentially risky deposit into a riskless one.

This is a supply-side story: it’s not (only) that depositors are uninformed and need protection; it’s that the economy needs a riskless transactions asset and private parties may not be able to create enough of it.

7.4.3 The Treasury Market: An Alternative Safe-Asset System

Government bonds can play the same role as insured bank deposits — as an alternative rather than a supplement to the banking system.

If the private economy cannot supply enough riskless debt (because firms’ assets are too risky or insufficiently diversified), the government can issue claims backed by its taxing authority: Treasury securities. Unlike bank deposits, Treasuries are directly riskless without requiring any private equity cushion.

Treasury Securities as the Ultimate Information-Insensitive Asset

In the GP framework, a Treasury bond is the ultimate information-insensitive security:

Its payoff is backed by the government’s taxing power, not by any specific risky asset.
No private agent has superior information about the government’s tax base that would allow them to exploit uninformed holders.
Treasuries are therefore the ideal transactions medium for uninformed agents — even safer than private bank deposits.

GP note (Section V) that a government debt market “can improve uninformed agents’ welfare by providing additional riskless securities” and that this “parallels that of the provision of deposit insurance since, in both cases, the government’s role is to create a risk-free asset.”

Deposit insurance vs. Treasury securities: two roads to the same place. GP show that both instruments achieve the same allocation. The choice between them is a policy decision:

Deposit insurance keeps the banking system central: private banks originate loans (preserving their comparative advantages in monitoring and relationship lending), while government guarantees solve the information-insensitivity problem on the liability side.
A Treasury-based payments system could potentially replace banks as the transactions-medium provider. Agents would hold Treasuries (or MMMFs investing in Treasuries) for payment purposes and turn to private intermediaries only for loans. Deposit insurance would be unnecessary.

This is the core tension in the narrow banking debate — whether commercial banks need deposit insurance to operate safely, or whether separating the payments system (safe government-backed liabilities) from credit intermediation (risky private loans) would be more efficient. GP’s framework clarifies exactly what would need to be true for such a substitution to work: the supply of information-insensitive assets must be sufficient to meet the transactions demand.

7.5 Part 4: Information Insensitivity as a General Principle

The Gorton–Pennacchi paper was published in 1990, but its central concept — information insensitivity — became the organizing framework for a major strand of post-2008 financial crisis research. This section traces the idea forward.

7.5.1 Holmstrom (2015): Debt Is Designed to Be Opaque

Gorton and Pennacchi show that banks create information-insensitive securities. Bengt Holmstrom (2015) generalizes this to argue that debt itself — as a security design — is fundamentally an information-insensitivity device.

The key insight: when a security pays a fixed return over a wide range of outcomes, it becomes very difficult for an informed agent to profit from trading against an uninformed holder. There is nothing to know that would help — the security pays the same amount regardless of what the insider knows.

Holmstrom (2015): The Logic of Debt as a Safe Security

Consider a firm with assets worth $V \in [V_{\min}, V_{\max}]$. A debt claim promises face value $F$ to the holder.

If $F \leq V_{\min}$: debt is riskless. An informed agent who knows $V$ has no advantage over an uninformed holder — debt always pays $F$ regardless. Trading in this debt is immune to the insider-trading problem.
If $F > V_{\min}$: debt can default. An informed agent who knows $V$ is low can trade against uninformed holders who believe $V$ is high.

This is why debt is the natural design for a liquid, transactions-purpose security: by promising a fixed payoff at or below the worst-case asset value, debt becomes information-insensitive. Equity is the opposite: its value depends entirely on realized firm value, giving insiders maximum scope to exploit uninformed holders.

Corollary: Equity is suited for markets where sophisticated, informed investors participate (stock exchanges with mandatory disclosure, institutional ownership). Debt is suited for transactions use by uninformed agents who trade frequently and cannot monitor continuously.

Holmstrom’s framework explains several empirical regularities:

Short-maturity debt is more information-insensitive than long-maturity debt (less time for private information to matter before repayment).
Collateralized debt is more information-insensitive than unsecured debt (the collateral narrows the range of outcomes the holder needs to assess).
Repo and other collateralized short-term lending instruments are the most liquid instruments in modern finance — exactly what information insensitivity predicts.

7.5.2 Dang, Gorton, Holmstrom, and Ordonez (2017): Banks as Secret Keepers

Building on GP and Holmstrom, Dang, Gorton, Holmstrom, and Ordonez (2017) take the argument one step further: banks are not just issuers of information-insensitive securities; they are actively suppressors of information production.

DGHO (2017): The Optimality of Opacity

The DGHO result is counterintuitive: the optimal bank design makes its assets as opaque as possible, consistent with keeping deposits riskless.

In the strict GP model, once deposits are made riskless through the equity structure, information production by outsiders cannot undermine them — depositors’ payoffs are fixed regardless of what anyone discovers. DGHO extend this to a richer setting where deposits are not guaranteed to be fully riskless, and where outsiders can spend resources to investigate whether the equity cushion is actually sufficient.

The DGHO mechanism. If an outside analyst produces private information revealing that a bank’s loan portfolio is worse than believed — say, that the effective $R_L$ is lower than assumed — this can make deposits appear risky in secondary markets, even if the face value is nominally fixed. Agents who learn this trade against uninformed holders of deposits, recreating the kind of information asymmetry GP identified. By keeping loan-level detail opaque, the bank removes the informational raw material that would make such investigation profitable in the first place.

This is the precise opposite of the transparency ideal in standard disclosure economics. For the specific purpose of creating a safe transactions medium, opacity preserves the conditions under which deposits remain information-insensitive.

This has a striking empirical prediction: banks that disclose more about their loan portfolios should be less able to issue truly safe deposits, because disclosure invites information production that reintroduces the insider-trading risk. This provides a novel justification for the traditional bank practice of not disclosing individual loan details.

7.5.3 Safe Assets in the Modern Economy

The GP framework connects to the broader macroeconomic literature on safe assets — high-quality claims whose values are insensitive to changes in economic conditions and are therefore highly liquid and widely accepted as collateral.

Safe Asset	GP Category	Information Sensitivity
Government bonds (Treasuries, Bunds)	Government-issued, tax-backed	Lowest: backed by taxing power
Insured bank deposits	Government-backstopped bank debt	Low: FDIC insurance removes default risk
High-grade short-term corporate debt (CP)	Private riskless debt	Low but contingent: depends on issuer quality
MMMF shares (government-only)	Diversified government debt	Very low: only Treasury-backed assets
MMMF shares (prime)	Diversified private debt	Moderate: depends on portfolio quality
Senior ABS tranches	Structured private debt	Higher: depends on model and collateral

A major finding of post-2008 research (Gorton and Metrick 2012, Krishnamurthy and Vissing-Jorgensen 2012) is that the global demand for safe assets regularly exceeds the private supply. This “safe-asset shortage” is a macroeconomic manifestation of the GP insight: the economy’s need for information-insensitive transactions media is not automatically satisfied by private markets. The shortfall must be covered by government-issued riskless claims.

7.5.4 When Information Insensitivity Breaks Down: The 2008 Crisis

The most important application of the GP framework is the understanding it provides of the 2008 financial crisis. GP’s core prediction is that when safe securities become information-sensitive — when their value depends on private information that some agents hold and others don’t — their role as transactions media collapses.

The Crisis as a Breakdown of Information Insensitivity (Gorton 2010)

Before 2007, the senior tranches of mortgage-backed securities (MBS) and collateralized debt obligations (CDOs) were treated as informationally insensitive:

They were rated AAA by rating agencies.
They were widely used as collateral in repo markets — the modern equivalent of the $t = 1$ trading market in the GP model.
Uninformed institutional investors (pension funds, money market funds, foreign banks) held them without detailed analysis of the underlying mortgages.

This treatment was based on the belief that house prices would not fall simultaneously across the country — that geographic diversification made senior tranches riskless in the same way that bank diversification makes deposits riskless in the GP model.

When it became clear in 2007 that this belief was wrong, the securities became information-sensitive. Holders realized they didn’t know what the securities were worth. Worse, they worried that dealers and originators did know, and were trying to offload the bad ones — precisely the insider-trading dynamic in GP’s model.

The result: the repo market froze. Counterparties refused to accept MBS as collateral — not because the securities were certain to be bad, but because they had become information-sensitive and no one on the receiving end of a trade could rule out being on the wrong side of an insider. The loss of information insensitivity is the financial crisis in the GP framework.

This reading of the crisis has direct implications for regulatory design:

1. The crisis was not primarily about runs on retail deposits (Diamond–Dybvig) but about the collapse of the wholesale safe-asset market. The relevant “deposits” were repo agreements, asset-backed commercial paper (ABCP), and prime MMMF shares — not retail bank accounts. The GP framework is better suited to understanding this episode than the Diamond–Dybvig framework, even though both involve a collapse of a transactions medium.

2. Disclosure requirements can be counterproductive for safe assets. Requiring banks and ABS issuers to disclose detailed asset information — a natural regulatory response — can paradoxically destroy information insensitivity by giving sophisticated analysts the raw material to acquire information advantages over uninformed holders. GP and DGHO argue for caution here.

3. Government guarantees are the only reliable backstop. Just as in the original GP model, private information-insensitive securities can fail; government guarantees are the only reliable source of truly riskless assets. The 2008 crisis confirmed this: the assets that maintained their safe-asset status throughout were Treasuries and insured bank deposits — the two government-guaranteed instruments in GP’s analysis.

The Tri-Party Repo Market and GP

The tri-party repo market — where financial institutions borrow overnight by pledging securities as collateral — is the modern equivalent of the $t = 1$ trading market in the GP model. Repo borrowers are the analog of early consumers (who need liquidity); repo lenders (money market funds, foreign banks) are the analog of late consumers (who have liquidity to invest).

When MBS became information-sensitive in 2007–2008, repo lenders stopped accepting them as collateral — exactly as late consumers in the GP model refuse to trade if they believe insiders are on the other side.

Gorton and Metrick (2012) document repo haircuts rising from near zero to 45% on structured products during the crisis — a direct measure of the increasing information-sensitivity of the collateral. A haircut is the discount demanded by the lender before accepting collateral: a 45% haircut on MBS means lenders were only willing to lend $0.55 per $1.00 of face value, because they feared the remaining $0.45 reflected insider information they couldn’t observe.

7.6 Summary and Connections

This chapter has developed the Gorton–Pennacchi framework and traced its implications through the modern literature.

Paper	Question	Answer	Method
Gorton–Pennacchi (1990)	Why do banks issue safe debt against risky assets?	To create information-insensitive securities that protect uninformed traders	Theory: informed/uninformed trading model
Holmstrom (2015)	Why is debt the natural design for liquid assets?	Debt is information-insensitive by construction	Mechanism design / security design
DGHO (2017)	Why are banks opaque?	Opacity destroys incentives for costly information production, preserving insensitivity	Extended GP model with endogenous information
Gorton–Metrick (2012)	How did the 2008 crisis manifest in wholesale markets?	Repo haircuts ≈ loss of information insensitivity in collateral	Repo market data during crisis
Krishnamurthy–Vissing-Jorgensen (2012)	Is there a “convenience yield” on safe assets?	Yes: investors pay a substantial premium for information-insensitive assets	Treasury vs. corporate yields

Connections to previous chapters.

Diamond–Dybvig (Chapter 3) shows that a bank’s demand-deposit contract can insure against taste shocks — but this requires ruling out equity trading. The GP model shows why equity trading is problematic: it exposes liquidity traders to insiders. The two models are complementary: DD explains the insurance value of deposits; GP explains why deposits rather than equity are the right instrument for providing that insurance.
Diamond (Chapter 5) explains that banks monitor borrowers on behalf of depositors, justifying the fixed-rate deposit contract as a disciplining device. The GP result adds a second function for the same fixed-rate deposit: it is also informationally inert, protecting depositors from having to evaluate asset quality. Monitoring (Diamond) and information insensitivity (GP) are distinct but mutually reinforcing reasons for the same institutional arrangement.
KRS (Chapter 6) explains why deposit-taking and loan-commitment making coexist: they share a liquidity buffer. The GP result explains why the liability side takes the form of fixed-rate deposits rather than equity: deposits are information-insensitive, equity is not. Together, KRS and GP explain both the combination (deposits + loans) and the form (deposits = safe debt) of the standard bank balance sheet.

The GP Insight in One Sentence

A bank creates a safe transactions medium by absorbing the information risk of its asset portfolio onto its equity holders — allowing uninformed depositors to transact without fear of exploitation by better-informed insiders.

This single insight — that the bank’s balance sheet is fundamentally an information-allocation device, placing information-sensitive claims with informed equity holders and information-insensitive claims with uninformed depositors — unifies the rationale for bank deposits, deposit insurance, government bonds, and the design of liquid financial markets.

7.7 References

7.7.1 Primary Source

Gorton, G., and Pennacchi, G. (1990). Financial Intermediaries and Liquidity Creation. Journal of Finance, 45(1), 49–71.