Property and Casualty Actuary: Pricing, Reserves and Catastrophes

Overview of the specialty

Under the LISF (Insurance and Surety Institutions Law), the property and casualty operation groups the lines of liability and professional risks, marine and transport, fire, agricultural and livestock, automobile, credit, bond, mortgage credit, financial guaranty, catastrophe risks and miscellaneous (article 25, section III). These are short-term contracts, almost always annual, that indemnify a loss of property or wealth: the insured receives what was lost, not a fixed sum.

That indemnity nature gives the risk two dimensions: frequency (how many claims per unit of exposure) and severity (how much each one costs). In addition, months or years may pass between occurrence, reporting and final payment: a car crash is settled in weeks, but a liability lawsuit or structural damage after an earthquake can take years. Estimating that development tail is the heart of reserving.

Mexico adds a layer few specialties have: severe catastrophe exposure. The 1985 and 2017 earthquakes and hurricanes such as Wilma (2005) and Otis (2023) showed that a single night can consume years of premiums. That is why the CNSF (National Insurance and Surety Commission) requires cumulative catastrophe reserves for earthquake, hurricane and hydrometeorological risks, along with a specific capital requirement based on the probable maximum loss (PML).

Auto is the largest line by number of policies and the most data intensive, with added pressure since article 63 Bis of the Ley de Caminos, Puentes y Autotransporte Federal (Federal Roads, Bridges and Motor Carrier Law) requires every vehicle using federal roads to carry liability insurance. Fire and commercial risks, by contrast, depend heavily on reinsurance and risk engineering.

Key functions

Technical pricing

Calculates the risk premium by segment with frequency and severity models (Poisson and gamma GLMs, or Tweedie) on exposure, and converts it into the gross rate with loadings for expenses, commissions, reinsurance cost and profit. In auto it works with dozens of variables: vehicle, use, postal code, driver, deductible and telematics.

Claim reserves

Estimates the reserve for outstanding obligations: reported claims with uncertain valuation, incurred but not reported claims (IBNR, known in Mexico as SONR) and adjustment expenses. It uses paid and incurred triangles by accident year and the methods the company has registered with the CNSF.

Unearned premium reserve and adequacy

Values the unexpired portion of the risk as a best estimate of future claims and expenses plus a risk margin, and checks that the premium charged is sufficient. If expected losses exceed the premium, the deficiency is reflected in the reserve.

Catastrophe risks

Calculates earthquake and hurricane risk premiums and PML with the CNSF systems (Sistema R and RH-Mex) or commercial models, controls accumulations by zone and sets up the catastrophe reserves. See catastrophe risk.

Reinsurance program design

Analyzes the loss distribution per risk and per event to propose retentions, quota share treaties and excess of loss layers. Quantifies how much capital each structure releases in the RCS (solvency capital requirement).

Portfolio monitoring and fraud

Builds dashboards of loss ratio, frequency and average cost by channel, agent and region. Develops models to detect suspicious claims and measures the effect of changes in deductibles or service providers.

Techniques, models and formulas

The property and casualty actuary lives between two questions: how much to charge for the risk ahead (pricing) and how much is owed for the risk already past (reserving). The techniques come from applied statistics, risk theory and, for catastrophes, earthquake engineering and meteorology.

Collective frequency and severity model

The aggregate loss S = X₁ + ... + X_N combines a random number of claims (Poisson, negative binomial) with random amounts (gamma, lognormal, Pareto). It is used for risk premiums, reinsurance and capital. See loss functions.

Generalized linear models (GLMs)

The industry standard for pricing: frequency with a Poisson log link using exposure as an offset, severity with a gamma, or pure premium directly with a Tweedie. They produce multiplicative relativities that are easy to audit. See pricing.

Chain-ladder

Projects claim development with age-to-age factors calculated on the historical triangle. It works well in stable lines with enough history, but it amplifies any anomaly in the latest accident year.

Bornhuetter-Ferguson

Combines an a priori expected loss ratio (from the rate or the plan) with the development pattern: the reserve depends on what remains to develop, not on what has already been paid. It is preferable for recent accident years or long-tail lines such as liability.

Stochastic reserving methods

The Mack model and the bootstrap on ODP models generate a full distribution of the reserve rather than a single point. They serve the risk margin and the RCS, and help explain to the board the range of possible outcomes.

Catastrophe models

They combine a stochastic event catalog, a hazard module (seismic intensity, wind speed), a vulnerability module by construction type and a financial module with deductibles and coinsurance. From them come the risk premium and the exceedance probability curve that defines the PML.

Frequency GLM with exposure

\[\log E[N_i] = \log e_i + \beta_0 + \sum_{k} \beta_k x_{ik}\]

N_i is the number of claims for risk i, e_i its exposure in policy years (offset) and x_ik its characteristics. Each exp(β_k) is a multiplicative rating relativity.

Chain-ladder development factor

\[\hat f_j = \frac{\sum_{i} C_{i,j+1}}{\sum_{i} C_{i,j}}, \qquad \hat U_i = C_{i,J_i}\prod_{j\ge J_i}\hat f_j\]

C_i,j is the cumulative amount for accident year i at development age j; the sums include only the years with both ages observed. U_i is the ultimate estimate from the latest diagonal J_i.

Bornhuetter-Ferguson reserve

\[R_i^{BF} = ELR \cdot P_i \left(1 - \frac{1}{CDF_i}\right)\]

ELR is the a priori expected loss ratio, P_i the earned premium for year i and CDF_i the cumulative development factor to ultimate; 1 - 1/CDF is the proportion not yet developed.

Regulatory framework in Mexico

LISF arts. 25 and 27 · Property and casualty lines

They define each line and its coverage. Article 27, section XV, defines catastrophe risks as events of unpredictable frequency and severity that accumulate liabilities, and includes earthquake, volcanic eruption, hurricane and other hydrometeorological risks.

LISF arts. 216 to 219 · CUSF Chapters 5.1 and 5.2

The unearned premium reserve and the reserve for outstanding obligations (which includes IBNR and allocated adjustment expenses, article 217) are valued as a best estimate plus a risk margin. The actuarial methods are registered with the CNSF, signed by a certified actuary and backed by an independent actuary's opinion.

CUSF Chapter 5.6 · Catastrophe reserves

Earthquake and hurricane build cumulative reserves from the earned portion of the retained risk premiums calculated with the CNSF technical models, plus their investment income. In agricultural and livestock, the monthly increase is 35% of the retained earned gross premium, with a maximum limit tied to the PML. They may be drawn down only in the cases the CUSF (Single Insurance and Surety Circular) provides for.

CUSF Title 6 · RCS for PML

In addition to the requirement for technical and financial risks, earthquake, hurricane, agricultural and mortgage credit portfolios generate a requirement for risks based on the probable maximum loss, net of catastrophe reserves and excess of loss covers. See solvency.

CUSF Title 9 · Retention limits

The board of directors sets the maximum retention limits by line on a technical basis, and the reinsurance committee monitors compliance. The actuary supports those limits with volatility and capital analysis.

LCPAF art. 63 Bis · Mandatory auto liability

Every vehicle using federal roads must carry third-party liability insurance; the SHCP (Ministry of Finance), after consulting the CNSF, issues the rules and the minimum coverage amount. It is a mass-market product with simple pricing and high volume.

Case study: auto claim reserve with chain-ladder and Bornhuetter-Ferguson

An insurer has the following triangle of cumulative paid auto claims (millions of pesos) by accident year and development age in years. 2022: 40.0, 62.0, 70.0, 72.0. 2023: 44.0, 68.2, 77.0. 2024: 48.0, 75.0. 2025: 52.0. Assume 2022 is fully developed. Earned premium for 2025 was 128 million and the expected loss ratio in the rate is 70%.

Age-to-age factors: f₁ = (62.0 + 68.2 + 75.0) / (40 + 44 + 48) = 205.2 / 132 = 1.5545; f₂ = (70 + 77) / (62 + 68.2) = 147 / 130.2 = 1.1290; f₃ = 72 / 70 = 1.0286.
Cumulative factors to ultimate: 2023: 1.0286; 2024: 1.1290 × 1.0286 = 1.1613; 2025: 1.5545 × 1.1613 = 1.8053.
Chain-ladder ultimates and reserves: 2023: 77.0 × 1.0286 = 79.20, reserve 2.20; 2024: 75.0 × 1.1613 = 87.10, reserve 12.10; 2025: 52.0 × 1.8053 = 93.87, reserve 41.87. Total reserve: 56.17 million.
Bornhuetter-Ferguson for 2025: 0.70 × 128 × (1 - 1/1.8053) = 89.6 × 0.44607 = 39.97 million, with an ultimate of 91.97. Using BF for 2025 and chain-ladder for the other years, the total reserve is 54.26 million.
Sensitivity: if a hailstorm had pushed 2025 payments to 60.0 instead of 52.0, chain-ladder would project 60.0 × 1.8053 = 108.32 (reserve 48.32), while BF would keep the reserve at 39.97 (ultimate 99.97). Chain-ladder multiplies the atypical event by 1.81; BF recognizes it only once.
For the regulatory reserve, future cash flows are discounted with the risk-free curve, adjustment expenses are added and the cost of capital risk margin is included. The chosen method must be registered with the CNSF.

Lesson: No method is right on its own: chain-ladder trusts the data and BF trusts the rate. The actuary chooses by accident year, documents the judgment and explains the difference, especially when an atypical event contaminates the latest diagonal.

Where they work and how they grow

Analyst (0 to 3 years)

Builds exposure and claims databases, updates triangles and loss ratio dashboards, runs GLMs under supervision and prepares the regulatory reserve reports.

Actuary (3 to 7 years)

Owns a rate or the reserve for a line: selects development factors, proposes relativities, drafts technical notes and presents results to underwriting and finance.

Senior actuary or manager (7 to 12 years)

Signs methods and valuations as a certified actuary, coordinates reinsurance program design and catastrophe modeling, and defends the figures before the CNSF and the independent reviewer.

Technical director or chief actuary (12 years or more)

Sets pricing strategy and risk appetite, decides retentions and reinsurance structure with the board, and is accountable for reserve adequacy and the RCS of the property and casualty operation.

Typical employers in Mexico

Property and casualty and multiline insurers, with strong demand in auto
Agricultural insurers and agricultural insurance funds
Reinsurers and representative offices of foreign reinsurers
Reinsurance intermediaries and corporate insurance brokers
Actuarial consulting practices and audit firms
Catastrophe modeling and risk engineering firms
Comisión Nacional de Seguros y Fianzas
Auto and usage-based insurance insurtechs

Common positions

Auto pricing actuaryReserving actuaryCatastrophe modelerCommercial property and fire actuaryAgricultural actuaryProperty and casualty reinsurance actuaryClaims and fraud data scientistProperty and casualty technical managerHead of the actuarial function

Skills and tools

Technical

Collective risk model and loss distributionsGLMs and gradient boosting modelsDeterministic and stochastic reserving methodsCredibility theoryCatastrophe modeling and exceedance curvesValuation of reinsurance structuresBasics of risk engineering

Tools

Python and R (GLM packages, ChainLadder, chainladder-python)SQL and claims databasesExcel for triangles and validationCNSF Sistema R and RH-MexCommercial catastrophe modeling platformsPricing software (Emblem, Akur8 or similar)Geospatial tools (QGIS)

Personal

Judgment in selecting factors and assumptionsAbility to explain uncertainty rangesCollaboration with underwriting, claims and reinsuranceSkepticism toward atypical dataExecutive communication of extreme risks

Certification and training

CONAC certification in property and casualty

A core requirement, together with the cédula profesional, for signing technical notes and reserve valuation methods for the property and casualty operation before the CNSF.

ACAS and FCAS from the Casualty Actuarial Society

The CAS, founded in 1914, is the actuarial society devoted entirely to property and casualty. Its exams cover ratemaking, reserving, reinsurance and predictive analytics, and are the international benchmark in the field.

Catastrophe modeling certification

Model vendors and some universities offer specific training in seismic hazard, vulnerability and model use; it is highly valued in reinsurance.

Data science training

Machine learning and MLOps courses complement the classic GLM. See certification and actuarial AI.

Trends reshaping the field

Telematics and usage-based insurance

Driving data (mileage, braking, time of day) make it possible to price on behavior as well as profile. The actuary must integrate high-frequency variables without losing interpretability or credibility.

Climate change and hydrometeorological risks

Rapidly intensifying hurricanes such as Otis call into question catalogs calibrated on past climate. Climate trend adjustments are growing in hurricane, flood and agricultural drought models.

Parametric insurance

It pays based on an observable index (magnitude, central pressure, accumulated rainfall) rather than the adjusted loss. In Mexico it is used by governments and in agriculture; the actuary designs the trigger and measures basis risk.

Reserving automation

Reserving processes are moving to version-controlled code in Python or R with bootstrapping and automatic validation. This frees time for root cause analysis and for reconciling regulatory and IFRS 17 metrics.

Frequently asked questions

What exactly is the IBNR reserve?

It is the estimate of incurred but not reported claims (SONR in Mexico): events that occurred before the valuation date but that the insurer does not yet know about. The LISF includes it in the reserve for outstanding obligations together with allocated adjustment expenses, and it is estimated with development triangles.

When do I use chain-ladder and when Bornhuetter-Ferguson?

Chain-ladder suits mature years and stable lines with a long history. Bornhuetter-Ferguson is better for recent accident years, long-tail lines or new portfolios, because it relies less on the few payments observed. It is common to use BF for the latest one or two years and chain-ladder for the rest.

Is the CAS recognized in Mexico?

The ACAS and FCAS credentials are highly valued at reinsurers and subsidiaries of international groups. To sign before the CNSF, however, what counts is the cédula profesional and certification by the professional college or accreditation by the Commission.

What distinguishes the catastrophe reserve from the unearned premium reserve?

The unearned premium reserve covers the expected value of claims for the unexpired coverage period. The catastrophe reserve is cumulative, built from the retained earthquake or hurricane risk premium and its investment income, and exists to pay the retained portion of an extreme event, with a limit tied to the PML.

Go deeper

Pricing and GLMs Technical reserves and triangles Catastrophe risk in Mexico Distributions and loss functions Reinsurance Credibility theory All specialties In-depth actuarial topics

Sources consulted