Automobile Insurance, assigned risks are often determined by a national rat-

ing bureau (AIPSO, the automobile insurance plans

Private services office), not by the carriers themselves.
An alternative type of involuntary market (more
common in workers’ compensation and commercial
Because of the high cost of traffic accidents, auto automobile, but also used by a few states for per-
is the dominant line of insurance in most countries. sonal automobile) is a reinsurance facility or a joint
This article looks at its operation in the United States. underwriting association. In these plans, the costs of
There, for the first half of the twentieth century, assigned risks are allocated to all auto insurers in
auto insurance was priced by rating bureaus and sold the state as a percentage of direct written premium.
predominantly through independent agents; now most Rate suppression in the voluntary market and restric-
auto insurance is independently priced and sold by tions on risk classification in a few some states (such
direct writers with captive agents. The large premium as New Jersey with a mandatory joint underwriting
volume, independent pricing, aggressive competition, mechanism, and Massachusetts with a reinsurance
and the need for frequent rate changes make auto facility) have so greatly increased the assigned risk
insurance pricing the most common actuarial task. population and the costs of selling insurance that
Slightly more than half the US states use common many carriers have voluntarily left these markets.
law systems of tort liability; drivers are liable for Some drivers are uninsured, either because the
negligently caused bodily injuries and property dam- coverage is expensive or because the driver is not
ages, though most policies also provide limited first legally in the country. These drivers are generally the
party medical coverage. Dissatisfaction with adver- worst risks; the number of uninsured accidents is over
sarial proceedings, court delays, high legal costs, and twice the number of uninsured drivers. Uninsured
the lack of compensation for many accident victims motorists coverage provides injury compensation to
have led to the adoption of no-fault compensation the policyholder if the tortfeasor in an auto accident is
systems in nearly half the states. A few states have not insured. Underinsured motorist provides similar
choice no-fault systems, in which drivers may choose coverage if the tortfeasor’s policy limit is below a
between tort liability and no-fault when purchasing certain amount.
auto insurance. Choice no-fault requires an additional
mechanism for accidents involving drivers who have Ratemaking
chosen different compensation systems. Calendar/accident year data are typically used for
Strong opposition to no-fault by trial attorneys the liability coverages (bodily injury, property dam-
have led to compromise no-fault systems with low age, and personal injury protection); calendar year
tort thresholds that permit suits for many injuries. data are generally used for vehicle damage cover-
Inflation and liberal court interpretations of which ages (collision and other damages). Premiums are
accidents permit a suit have led to weak laws unable brought to current rate level using a ratio of adjusted
to contain rising loss costs. to unadjusted rate level. For example, with six-month
Tort liability covers all damages for which the policies, a rate change of +6% effective on Septem-
insured is legally responsible, whether economic ber 1 affects 2/3 × 2/3 × 1/2 = 22.2% of the calen-
damages (medical costs and lost wages) or general dar year’s earned premium, and the on-level factor
damages (pain and suffering). No-fault compensation to bring the year’s premium to the current rate level
systems cover economic damages only, not general is (1.06)/(1 + 6% × 22%) = 1.046. Alternatively, the
damages. policies in the experience period may be rerated using
All states have either financial responsibility laws the current rates (extending exposures), though the
or compulsory insurance laws. A postaccident, finan- additional complexity often outweighs the benefits of
cial responsibility law requires a driver either to show increased accuracy.
evidence of insurance coverage after an accident or to The exposure base of car-years is not inflation sen-
post a bail bond; compulsory insurance laws require sitive; no exposure trend is needed but rate changes
drivers to obtain insurance before driving. A driver are frequent. Mileage has been suggested as an alter-
unable to obtain insurance from a private carrier is native (usage sensitive) exposure base, but the poten-
assigned by the state to an insurer; the rates for these tial manipulation of speedometers by insureds makes
2 Automobile Insurance, Private

this impractical. During the high inflation years of the Some actuaries load all expenses as a percentage
late 1970s and early 1980s, many insurers switched of premium; others treat the expenses that vary
from twelve-month to six-month policies so that rates directly with premium as a percentage loading and
could be updated more frequently. the remaining expenses as fixed costs. For the lat-
Accident year losses are usually developed to ter method, the per-policy expenses are brought to
ultimate by chain ladder techniques using either the cost level of the future policy period and divided
reported losses or paid losses. Development is sig- by the corresponding written premium. The indicated
nificant only for the liability sublines; for vehicle rate change is the experience period loss ratio plus the
damage, calendar year data is generally used with fixed expense ratio divided by the complement of the
no adjustment for development. variable expense ratio: rate change = (loss ratio +
Allocated loss adjustment expenses (see ALAE), fixed expense ratio)/(1 − variable expense ratio).
or defense counsel fees and other claim specific Expense flattening procedures that reduced the
costs, are combined with losses for ratemaking and fixed expense loading for high premium policies were
sometimes for reserving in the liability lines of busi- introduced in the 1980s and are now required in some
ness (see Liability Insurance), including auto insur- states. Underwriting, policy issuance, and advertis-
ance. Unallocated loss adjustment expenses, or claims ing expenses are loaded as a flat rate, reducing the
department overhead, are added as a loading onto premium for high-risk insureds and increasing the
losses. premium for low-risk insureds. The lower persistency
Losses are brought to the cost level of the future of high rated classes along with their higher under-
policy period; in no-fault states, losses are also writing costs and not taken rates, lower limits on
brought to the current benefit levels. Trend factors liability coverages, and their frequent lack of physi-
cal damage coverages offsets the rationale for lower
are often based on countrywide industry paid claim
expense loadings. Insurers vary in their ratemaking
data, sometimes weighted with state specific data.
methods.
Econometric modeling of inflation rates has had with
The underwriting profit margin, which was set
limited success, and actuaries tend to rely on expo-
at 5% of premium by a 1920s regulatory mandate,
nential fits to historical loss cost data.
is now calculated to provide an adequate return on
The trend period runs from the average accident
invested capital; actuaries use discounted cash flow,
date in the experience period to the average accident
net present value, or internal rate-of-return pricing
date under the new rates; the latter is the effective date models. The short tail of personal auto insurance
of the rate filing plus one half the lag until the next makes sophisticated modeling less necessary than
rate filing plus one half the policy term. If the policy for the longer-tailed commercial liability lines of
term is six months, rate filings are made annually, business. Some insurers select a target operating ratio
and the anticipated effective date of the filing is but do not run financial pricing models for each state
January 1, 20XX, the average date of loss under review.
the new rates is January 1 + 6 months + 3 months = The basic limits data used for auto insurance
October 1, 20XX. ratemaking become a smaller portion of total premi-
Large direct writers can base rate indications for ums and losses as inflation reduces the value of the
major states on their own data. Smaller insurers and limit and as higher court awards leads drivers to pur-
even large insurers in small states may weight their chase higher limits of coverage. The price for higher
experience indications with the expected loss ratio, layers of coverage is based on increased limit factors,
adjusted by the same trend factors as the experience which are the ratio of losses up to a given limit to
loss ratio. The credibility (see Credibility Theory) basic limit losses. Increased limit factors were once
for a state’s experience has traditionally been based computed directly from empirical data; they are now
on a full credibility standard of 1084 claims with derived from size-of-loss distributions fitted to math-
partial credibility based on a square root rule of ematical curves.
Z = N/1084. The developed, trended, and credibility Trend is lower for basic limits losses than for
weighted experience period losses are divided by total limits losses, and a basic limits rate review
premiums at current rate levels to give the experience understates the needed change in total premium
loss ratio. unless accompanied by a change in the increased
 Automobile Insurance, Private 3

limit factors. During years of high inflation, insurers in the complexity of its class structure and in the
updated increased limit factors routinely, either by relation of its classes to the loss hazards. Life insur-
reexamining empirical data or by calculating the ance classification uses age, property insurance (see
implied change in the size-of-loss distributions. Property Insurance – Personal) uses construction
and protection classes, and workers’ compensation
Illustration. A January 1, 20X7, rate filing for six uses industry, which are intuitively correlated with
month policies is based on calendar/accident year the benefits or losses; auto uses sex, marital sta-
20X5 data developed to 15 months since inception tus, garaging territory, and credit rating, all of which
of the year (March 31, 20X6). The 20X5 basic have been alleged to be discriminatory or socially
limits earned premium is $10 million, the basic limits improper. Some states prohibit certain classification
reported losses are $5 million, the variable expense dimensions and curtail the use of others, though the
ratio is 20%, and the fixed expense ratio to written empirical support for the risk classes leads most actu-
premium is 15% in 20X5. Rate changes of +4% aries to support their use.
and +5% were effective on January 1, 20X5 and Merit rating (see Bonus–Malus Systems) for auto
20X6. The accident year loss development factor insurance uses past accidents and (moving) traffic
from 15 months to ultimate is 1.2; loss adjustment violations. Because the class system is so refined
expenses are 10% of losses; the loss trend rate is and claims are infrequent, the credibility of merit
6% per annum and the fixed expense trend is 3% rating for a single car is low (about 5 to 10%).
per annum. A target underwriting profit margin of Some states mandate a uniform merit-rating plan with
2% is based on a financial pricing model, though the higher credibility to provide financial incentive to
insurer would write business even at a −1% margin. obey traffic laws.
We determine the indicated rate change. Many insurers incorporate the merit-rating dis-
Using the pre-January 1, 20X5, rates as the count or surcharge as an additional class dimension
base, the average rate level in 20X5 is 3/4 × (1 + in a multiplicative model; a surcharge for two traffic
1.04) = 1.03, since the 4% rate change on January violations might be +4% of the premium. States that
1, 20X5, affects half of the first half of calendar year mandate uniform merit-rating systems generally often
20X5 earned premium and all the second half. The prefer dollar discounts and surcharges; a surcharge for
current rate level is 1.04 × 1.05, and the premium two traffic violations might be $75.
on-level factor is (1.05 × 1.04)/1.03 = 1.060. Losses Ratemaking for collision coverage is based on
and loss adjustment expenses developed to ultimate the value and damageability of the vehicle, often
are $5 million × 1.2 × 1.1 = $6.6 million. The trend represented by its model and age.
period runs from July 1, 20X5 to October 1, 20X7, Underwriting cycles in auto insurance cause dis-
or 2.25 years; the trended losses are $6.6 million × cernible, though irregular, patterns of high and low
1.062.25 = $7.52 million; and the experience loss ratio profitability. The causes of the cycles are disputed,
is $7.52 million/$10.60 million = 70.97%. The fixed ranging from the cobweb theory of agricultural cycles
expense ratio trend period runs from July 1, 20X5, (see Underwriting Cycle) to game theory (see Coop-
to April 1, 20X7; the ratio at future cost levels erative Game Theory; Noncooperative Game The-
is 15% × 1.031.75 = 15.80%. The target rate change ory) explanations of oligopolistic pricing. The phase
indication is (70.97% + 15.80%)/(1 − 20% − 2%) = of the underwriting cycle often has as much influ-
1.112, or +11.2%; the minimum is (70.97% + ence on an insurer’s pricing decisions as actuarial
15.80%)/(1 − 20% + 1%) = 1.071, or +7.1%. The cost indications have.
filed rate request would lie between these bounds,
depending on the intensity of competition.
Classification ratemaking stems largely from auto (See also Automobile Insurance, Commercial)
insurance. Other lines of business may also have a
large number of classes, but, auto insurance is unique SHOLOM FELDBLUM