Fidelity Insurance: Pricing Defection Risk

Fidelity Insurance: Pricing Defection Risk

TL;DR

Fidelity insurance (crime insurance, employee dishonesty bonds) is a mature market that prices defection risk. Premiums are typically 0.3-2% of coverage, reduced significantly by control systems. Political fidelity insurance doesn’t exist due to adverse selection and moral hazard, but the theoretical design space is interesting. Dynamic, control-linked insurance products could create tighter feedback loops between delegation risk and pricing.

This document explores the insurance industry’s approach to pricing “bad actor” risk—directly relevant to the delegation accounting framework. If we can price defection risk actuarially, we can make delegation balance sheets more precise.

---

Part 1: The Existing Market

1.1 Primary Product Categories

Employee Dishonesty / Fidelity Bonds

The core product covering theft, fraud, and embezzlement by employees.

Attribute	Typical Range
Coverage limits	$50K -$5M+
Annual premium	0.3% - 2% of coverage
Deductible	$1K -$50K
Required controls	Background checks, dual controls, segregation of duties

Example pricing: A $1M policy for a small organization with minimal controls costs ~$2,000-3,000/year. With strong controls, this drops to ~$600-1,000/year.

Directors & Officers (D&O) Liability

Covers wrongful acts by leadership, including fraud, self-dealing, and breach of fiduciary duty.

Attribute	Typical Range
Coverage limits	$1M -$100M+
Annual premium	$2K -$15K for $1M (small org)
Key exclusions	Criminal acts (usually), prior known claims
Critical for	Nonprofits, startups (protects personal assets)

Cyber Crime / Social Engineering

Growing category covering wire transfer fraud, invoice manipulation, phishing attacks.

Attribute	Typical Range
Coverage limits	$100K -$10M
Annual premium	$1K -$20K+
Key exclusions	Voluntary transfers (tricky with social engineering)
Evolution	Insurers learned painful lessons 2015-2020

---

1.2 Why This Is Actually Insurable

Contrary to intuition, defection risk has favorable properties for insurance:

Actuarial data exists

• Decades of claims history from commercial insurers
• FBI Uniform Crime Reports provide baseline theft rates
• Association of Certified Fraud Examiners publishes biennial studies

Moral hazard is manageable

• Required controls as policy conditions
• Regular audits verify compliance
• Deductibles ensure skin in the game

Limited correlation

• Employee theft doesn’t cluster like natural disasters
• Economic downturns increase theft, but effect is modest (~20-30% increase)
• Not catastrophically correlated (unlike pandemic, earthquake)

Recovery infrastructure exists

• Forensic accounting recovers ~30-40% of fraud losses
• Criminal restitution adds additional recovery
• Subrogation allows insurers to pursue perpetrators

The Fundamental Equation

Premium = (Base Rate × Coverage) × Risk Multipliers × Control Discounts + Expense Loading

where:
  Base Rate ≈ 0.5-1% (reflects historical loss ratio)
  Risk Multipliers = f(industry, employee count, cash handling, prior claims)
  Control Discounts = 0.3-0.7 (strong controls → big discount)
  Expense Loading ≈ 20-30% (admin, investigation reserves)

---

1.3 Key Limitations and Exclusions

Standard exclusions (~90% of policies):

Exclusion	Rationale
Known bad actors	Can’t insure pre-existing fraud
Acts by owners/principals	Moral hazard too severe
Failure to prosecute	Requires criminal charges to claim
Collusion with insured	Prevents insurance fraud
Inventory shrinkage (general)	Too hard to prove employee theft vs. other loss

Underwriting friction:

• Extensive questionnaires about internal controls
• Background check requirements for covered employees
• May require audited financials
• Higher premiums (or denial) for prior claims
• Annual control attestations

---

Part 2: Control Systems and Pricing

2.1 The Control-Premium Relationship

This is directly relevant to delegation accounting: better controls = lower exposure = lower premiums.

Tier 1 Controls (10-30% premium reduction)

Control	Mechanism	Detection Rate
Segregation of duties	Different people authorize, execute, record	High
Dual signatures	Required for transactions >$X	High
Mandatory vacation	Forces job rotation, catches ongoing fraud	Medium
Background checks	Criminal, credit, reference verification	Medium

Tier 2 Controls (additional 10-20%)

Control	Mechanism	Detection Rate
External audits	CPA reviews annually minimum	High
Reconciliation procedures	Daily cash counts, monthly bank recs	High
Access controls	Time-locked safes, multi-person vault	Medium
Surveillance	Cameras in cash-handling areas	Medium

Tier 3 Controls (additional 5-15%)

Control	Mechanism	Detection Rate
Universal bonding	All employees bonded, not just high-risk	Medium
Anonymous hotline	Operational fraud tipline	High (tips catch ~40% of fraud)
Regular training	Annual fraud awareness refreshers	Low-Medium
IT audit trails	Complete logging for financial systems	High

Tier 4 / Extreme Controls (additional 5-10%, diminishing returns)

Control	Mechanism	Detection Rate
Biometric access	Fingerprint/retina for financial systems	High
Real-time monitoring	AI/ML fraud detection	Medium-High
Third-party escrow	For large transactions	Very High
Blockchain audit trails	Immutable transaction records	Very High

2.2 The Optimization Problem

Total Cost = Premium + Control Implementation + Control Maintenance + Expected Uninsured Loss

Optimize where: d(Total Cost)/d(Control Investment) = 0

Practical breakpoints:

Organization Size	Optimal Control Level	Reasoning
<$500K budget	Tier 1 only	Control costs exceed premium savings
$500K-$2M budget	Tier 1 + partial Tier 2	External audit may be +EV
$2M-$10M budget	Tier 1-2 + partial Tier 3	Hotline, training worthwhile
>$10M budget	Full Tier 1-3	All controls +EV for insurance alone

The catch: Tier 4 controls are rarely +EV purely for premium reduction. They’re justified by:

• Regulatory requirements
• Reputational protection
• Deterrence beyond insured losses
• Grant/contract requirements

2.3 Dynamic Considerations

Substitution effects: Strong controls on cash → fraud shifts to procurement, travel expenses, phantom vendors. Insurers learned this and now require holistic control environments.

Control decay: Controls degrade over time as people find workarounds. Insurers assume ~10-20% annual decay without refresher training and audits.

Moral hazard from insurance: Better coverage → less vigilant monitoring. Optimal contracts include:

• Deductibles (skin in the game)
• Coverage caps (catastrophic only)
• Co-insurance provisions (insured bears percentage)
• Required control maintenance

---

Part 3: Political Fidelity Insurance

3.1 Why This Market Doesn’t Exist

Direct insurance against politician fraud is essentially non-existent. The theoretical barriers:

Adverse selection dominates

Politician Type	Willingness to Buy	Effect
Clean politicians	Low (overpaying for risk they won’t create)	Exit market
Corrupt politicians	High (cheap money laundering)	Dominate market
Result	Only bad risks remain	Market collapse

Moral hazard is catastrophic

• Insurance removes personal deterrent effect
• Politician knows they’re covered → more willing to steal
• Unlike employee theft, politician controls enforcement

Principal-agent problem

• Who’s the beneficiary? Taxpayers can’t contract with politician
• If politician is beneficiary, they profit from their own fraud
• Government as beneficiary creates circular incentives

Enforcement nightmare

• Proving “fraud” requires conviction
• Politicians influence prosecution
• Long delays between act and conviction
• Statutes of limitations

3.2 What Actually Exists

Campaign liability insurance

• Covers campaign staff errors and omissions
• Does NOT cover candidate fraud
• Protects against volunteer mistakes, event injuries

Government official surety bonds

• Some jurisdictions require for treasurers, tax collectors
• Surety (guarantor) pays, then recovers from official
• This is NOT insurance—official still owes the money
• Functions as credit enhancement, not risk transfer

D&O for appointed officials

• Covers negligence and honest mistakes
• Explicitly excludes fraud and criminal acts
• Protects against lawsuit defense costs

3.3 Theoretical Design Space

If you wanted to design political fidelity insurance, what would it look like?

Beneficiary structure options:

Structure	Mechanism	Problems
Taxpayer fund	Insurance pays government treasury	Politician doesn’t care (not their money)
Bond market	Municipal bond insurers pay bondholders	Incentivizes bondholder monitoring, but limited scope
Escrow/bond posting	Politician posts bond, forfeits on conviction	Functions as delayed compensation, not insurance
Victim restitution	Insurance pays fraud victims directly	Verification nightmare, coverage scope unclear

Pricing impossibility:

Parameter	Estimate	Uncertainty
Baseline corruption charge rate	2-5% of politicians	High variance by jurisdiction
Conviction rate if charged	60-70%	Varies by offense type
Average theft amount	Bimodal: $10K vs$10M+	Extreme variance
Detection rate	20-40% of fraud detected	Very uncertain

Implied pricing:

• Actuarially fair premium: ~1-3% of coverage annually
• With adverse selection adjustment: 10-20%+
• At 20% premium, corrupt politician stealing $1M pays$200K for $800K net—still profitable
• No equilibrium exists where clean politicians participate

3.4 Better Mechanisms Than Insurance

Performance bonds (infrastructure model)

• Contractor posts bond for project completion
• Third-party surety monitors performance
• Works because completion is observable
• Applicable: Politician posts bond, forfeits if convicted

Clawback provisions (executive comp model)

• Deferred compensation recovered if fraud discovered
• Pension forfeiture for convicted officials
• Many jurisdictions have these laws
• Problem: Enforcement is discretionary

Prediction markets

• Bet on whether official will be convicted within N years
• Information aggregation, not insurance
• Legal in some jurisdictions for research
• Example: Polymarket has had markets on politician investigations

Escrow mechanisms

• Politician’s salary held in escrow for N years post-service
• Released only if no conviction
• Creates substantial contingent liability without insurance

---

Part 4: Research Literature

4.1 Actuarial Foundations

Association of Certified Fraud Examiners (ACFE)

The ACFE’s biennial Report to the Nations is the primary data source on occupational fraud:

Finding	Value	Implication
Median loss per case	~$117,000	Significant but insurable
Median duration before detection	12 months	Controls that accelerate detection are valuable
Tip-based detection	~40% of cases	Hotlines are highly effective
Internal audit detection	~15% of cases	Audits catch less than expected
Owner/executive fraud	5× higher losses than employee fraud	But harder to insure

Key papers:

• Holtfreter, K. (2005). “Is occupational fraud ‘typical’ white-collar crime?” Journal of Criminal Justice. — Demographic analysis of fraud perpetrators
• Hollow, M. (2014). “Money, Morals and Motives: An Exploratory Study into Why Bank Employees Commit Fraud.” Journal of Financial Crime. — Qualitative analysis of rationalization
• Murphy, P. (2012). “Attitude, Machiavellianism and the rationalization of misreporting.” Accounting, Organizations and Society. — Why people convince themselves fraud is acceptable

4.2 Insurance Economics

Moral hazard and adverse selection:

• Rothschild, M. & Stiglitz, J. (1976). “Equilibrium in Competitive Insurance Markets.” QJE. — Foundation of adverse selection theory
• Shavell, S. (1979). “On Moral Hazard and Insurance.” QJE. — Optimal insurance under moral hazard
• Winter, R. (2000). “Optimal Insurance Under Moral Hazard.” Handbook of Insurance. — Comprehensive treatment

Crime insurance specifically:

• Boyer, M. (2007). “Resistance (to fraud) is futile.” Journal of Risk and Insurance. — Models optimal enforcement vs. insurance
• Dionne, G. & Wang, K. (2013). “Does insurance fraud in automobile insurance increase claims?” Journal of Risk and Uncertainty. — Empirical evidence on moral hazard in fraud contexts

4.3 Political Economy

Corruption pricing:

• Mauro, P. (1995). “Corruption and Growth.” QJE. — Cross-country evidence on corruption costs
• Fisman, R. (2001). “Estimating the Value of Political Connections.” AER. — Uses firm values to price political relationships
• Khwaja, A. & Mian, A. (2005). “Do Lenders Favor Politically Connected Firms?” QJE. — Banks price political connections into loans

The “selectorate theory” framing:

Bueno de Mesquita et al.’s The Logic of Political Survival (2003) provides a framework where politician behavior is predictable based on:

• Size of selectorate (who could potentially support leader)
• Size of winning coalition (who actually keeps leader in power)
• Private vs. public goods provision

This maps to insurance: smaller winning coalitions → more extraction → higher “premiums” would be needed.

4.4 Mechanism Design

Relevant theory:

• Holmström, B. (1979). “Moral Hazard and Observability.” Bell Journal of Economics. — Optimal contracts under partial observability
• Tirole, J. (1986). “Hierarchies and Bureaucracies.” Journal of Law, Economics, & Organization. — Delegation chains and information
• Aghion, P. & Tirole, J. (1997). “Formal and Real Authority in Organizations.” JPE. — When delegation is optimal despite agency costs

Corruption-specific mechanism design:

• Becker, G. & Stigler, G. (1974). “Law Enforcement, Malfeasance, and Compensation of Enforcers.” JLE. — Efficiency wages as corruption prevention
• Mookherjee, D. & Png, I. (1995). “Corruptible Law Enforcers.” RAND Journal of Economics. — Optimal monitoring under corruption risk

---

Part 5: Novel Insurance Structures

5.1 Parametric Triggers

Traditional insurance requires proving a specific fraud occurred. Parametric insurance pays based on observable indices:

Potential design:

Trigger	Threshold	Payout
”Financial irregularity index”	>2 standard deviations	Automatic
Audit findings count	>N material findings	Scaled
Whistleblower reports	Verified reports >X	Per-report
Forensic accounting score	Below threshold	Automatic

Advantages:

• Eliminates investigation costs
• Faster payout
• Less litigation
• Objective triggers

Disadvantages:

• Gaming the index
• False positives
• Index construction is hard
• May not cover actual losses

5.2 Prediction Market Hybrid

Premium adjusts based on internal prediction market on fraud risk:

Premium_t = Base_Premium × f(Market_Probability_of_Fraud_t)

where Market_Probability comes from internal betting market

Mechanism:

• Employees bet on whether fraud will be discovered in next N months
• Betting reveals private information
• Premium adjusts in real-time
• Information aggregation + incentive alignment

Challenges:

• Manipulation (bet against, then report)
• Thin markets in small organizations
• Legal/regulatory issues
• Cultural acceptance

5.3 Mutual Insurance Cooperatives

Industry peers cross-insure each other:

Structure:

• Pool of similar organizations contributes to fund
• Claims paid from pool
• Surplus returned as dividends
• Members have information advantages over commercial insurers

Existing examples:

• Credit union leagues (mutual fidelity coverage)
• Church denominations (clergy misconduct pools)
• Trade associations (industry-specific risks)

Advantages:

• Better information sharing
• Aligned incentives (peers monitor each other)
• Lower overhead than commercial carriers
• Specialized underwriting expertise

Disadvantages:

• Limited capital for large losses
• Correlation risk within industry
• Governance challenges
• Adverse selection within pool

5.4 Dynamic/Real-Time Pricing

Insurance that adjusts continuously based on observed risk indicators:

Data inputs:

• Access pattern anomalies (from IT systems)
• Transaction velocity changes
• Segregation of duties violations
• Employee sentiment indicators
• Financial ratio changes

Mechanism:

Premium_daily = Base × Σ(Risk_Factor_i × Weight_i)

where Risk_Factors update daily based on telemetry

Enabling technologies:

• API integration with financial systems
• ML anomaly detection
• Real-time audit trail analysis
• Behavioral analytics

Current status:

• Cyber insurance moving this direction (Coalition, Corvus)
• Fidelity insurance lags (data integration harder)
• Regulatory barriers in some jurisdictions

---

Part 6: Implications for Delegation Accounting

6.1 Pricing Delegation Risk

The insurance market provides market prices for defection risk. This makes delegation balance sheets more concrete:

Balance Sheet Item	Insurance Analog	Market Price
Exposure (theft)	Fidelity bond premium	0.3-2% of coverage
Exposure (executive fraud)	D&O premium	0.2-1.5% of coverage
Contingent liability (if caught)	Policy limits + exclusions	Defines maximum recovery

Example: Alice delegating $1M to Bob

If Alice can buy fidelity coverage at 0.5% ($5,000 premium):

• Market is pricing Bob’s defection risk at ~$5,000 expected
• This is Alice’s insurable exposure
• Uninsurable exposure (coverage exclusions) adds to this

6.2 Control Investment as Exposure Reduction

The control-premium relationship maps directly to delegation accounting:

Control Investment	Premium Reduction	Implied Exposure Reduction
$0 (baseline)	0%	0%
$5K (Tier 1 controls)	25%	25%
$15K (Tier 1-2)	45%	45%
$30K (Tier 1-3)	60%	60%

Marginal analysis: If $10K in controls reduces exposure by$15K (premium savings + uninsured loss reduction), invest.

6.3 Why Political Insurance Fails Inform Oversight Design

The impossibility of political fidelity insurance reveals what makes delegation risky:

Insurance Failure Mode	Oversight Implication
Adverse selection	Need universal coverage (like bonding requirements)
Moral hazard	Need monitoring independent of the insured
Enforcement capture	Need external enforcement (not self-policing)
Detection difficulty	Need information systems politician doesn’t control

Design principle: If you can’t insure it, you need structural controls instead.

6.4 Research Directions

Open questions:

1. Optimal control portfolios: Which combinations of controls maximize exposure reduction per dollar?

2. Dynamic pricing feasibility: Can real-time risk indicators predict fraud well enough to price dynamically?

3. Political mechanism design: What non-insurance mechanisms best approximate insurance for political delegation?

4. Cross-organizational mutual insurance: Could EA/rationalist organizations create a mutual fidelity pool with information advantages?

5. Parametric trigger construction: What observable indices best predict fraud without being gameable?

6. Moral hazard quantification: How much does insurance actually increase fraud? (Empirical estimates vary widely.)

---

Summary

Concept	Key Finding
Fidelity insurance exists	Mature market, 0.3-2% of coverage
Controls reduce premiums	30-70% reduction possible
Political insurance doesn’t exist	Adverse selection + moral hazard + enforcement capture
Novel structures possible	Parametric, prediction market hybrid, mutual, dynamic
Implication for delegation accounting	Insurance premiums provide market prices for defection risk

The insurance industry has spent decades pricing defection risk. Their methods—actuarial analysis, control requirements, exclusion design—are directly applicable to delegation accounting. Where insurance fails (political contexts), the failure modes tell us what structural controls are needed instead.

---

Further Reading

Academic

• ACFE (2024). Report to the Nations: Occupational Fraud. — Primary data source
• Rothschild & Stiglitz (1976). “Equilibrium in Competitive Insurance Markets.” — Adverse selection foundation
• Holmström (1979). “Moral Hazard and Observability.” — Contract design under partial observability
• Bueno de Mesquita et al. (2003). The Logic of Political Survival. — Selectorate theory

Industry

• International Risk Management Institute (IRMI). Crime Coverage Guide. — Practitioner reference
• Nonprofit Risk Management Center. Coverage Guides. — Sector-specific
• Surety & Fidelity Association of America. Loss Statistics. — Claims data

Related Pages

• Delegation Accounting Overview — Balance sheet framework
• Power Struggles — Application to political contexts
• Risk Decomposition — Accidents vs. defection