Crime and Punishment: An Economic Approach - · PDF file3 1. INTRODUCTION There is a large...
Transcript of Crime and Punishment: An Economic Approach - · PDF file3 1. INTRODUCTION There is a large...
Crime and Punishment: An Economic Approach
Gary Becker (JPE, 1968)
Notes by Team Grossman (Spring 2015)
Meagan Madden, Mavzuna Turaeva, Ke Xu
2
Contents 1. INTRODUCTION .................................................................................................................................... 3
2. THE MODEL ........................................................................................................................................... 3
The Cost of Crime ...................................................................................................................................... 3
Key Notation for Reference ...................................................................................................................... 4
Model Components .................................................................................................................................. 5
a. Damages ........................................................................................................................................ 5
b. Apprehension and Conviction ....................................................................................................... 6
c. Supply of Offenses ........................................................................................................................ 7
d. Punishment ................................................................................................................................... 9
3. OPTIMALITY CONDITIONS .................................................................................................................. 10
4. SHIFTS IN THE BEHAVIORAL RELATIONS ........................................................................................... 11
a. Effect of 𝐷′ ...................................................................................................................................... 12
b. Effect of 𝐶𝑝 ..................................................................................................................................... 13
c. Effect of 𝜀𝑓 and 𝜀𝑓 .......................................................................................................................... 15
5. FINES ................................................................................................................................................... 16
Welfare Theorems and Transferable Pricing .......................................................................................... 16
Optimality Conditions ............................................................................................................................. 16
The Case for Fines ................................................................................................................................... 17
Compensation and the Criminal Law ...................................................................................................... 17
6. SOME APPLICATIONS (BECKER) ......................................................................................................... 18
The Effectiveness of the Public Policy ..................................................................................................... 18
A Theory of Collusion .............................................................................................................................. 19
7. EXTENSIONS/CRITICISM ..................................................................................................................... 19
“Prison”- Community .............................................................................................................................. 19
“Expensive” Prison Time ......................................................................................................................... 20
“Cost” of Judicial Error ............................................................................................................................ 20
Why does murder have a higher probability of conviction? .................................................................. 20
3
1. INTRODUCTION There is a large economic literature related to crime which traces back to Gary Becker’s seminal
work “Crime and Punishment.” The article currently has 11,404 citations in Google Scholar.
Becker acknowledges the founding works of Bentham (An Introduction to the Principles of
Morals and Legislation, 1780) and Beccaria (On Crime and Punishments, 1764) who applied an
economic framework to decisions of crime and deterrence, and Becker suggests that his 1968
paper is “a resurrection, modernization, and […] improvement on these much earlier pioneering
studies.” Here, Becker acknowledges that zero crime is not optimal for society because
significant resources are devoted to fighting crime. Instead, he approaches crime from the
perspective of the losses that it inflicts on society. The concept of social loss was not novel in
Becker’s time (see The Social Loss of Dying Parents, Glaster and Strauss, 1964), but its
application to crime was a significant contribution.
The model is a function of social loss (borrowed from traditional welfare economics) which
incorporates the various costs associated with crime: (1) costs from criminal damages, (2) costs
to apprehend & convict criminals, and (3) social costs associated with punishment. The optimal
level of crime is that which minimizes the social loss function. The associated level expenditures
to fight crime and the associated penalties to deter crime can be derived from this optimum.
Becker then uses this theoretical framework to discuss public policy implications for crime and
punishment.
2. THE MODEL
The Cost of Crime Here, “crime” is considered any violation of the law, ranging from murder to tax-evasion to
traffic violations. As Becker states, crime is an economically important “industry” with
significant resources going to determent, enforcement, prosecution, and punishment of offenses,
as well as the damages lost (e.g. stolen property) from offenses. Table 1 gives a snapshot from
1965 of public expenditures on the crime “industry”.
4
In 1965 the total cost of crime equaled 21 billion, which was about 4% of national income.
Becker argues that the numbers are significantly understated and it would be much higher if
other unmeasured social losses were accounted for. For example, the cost of murder is measured
by the loss of earnings of victims and excludes the value placed by society on life itself, thus
represents an unmeasured loss/cost to the society; or the cost of gambling excludes both the
utility to the gambler and “external” disutility to some clergy. Also, private expenditures are
significantly understated, as many crime preventions activities cannot be easily measured, for
examples, choosing to live in the suburbs, taking taxi instead walking, and etc.
Key Notation for Reference
Notation Variable Description Functional Relationships
𝑶𝒊𝒋 Activity level of i
th crime committed by j
th individual;
measured as # offenses
Supply function for j:
𝑂𝑗 = 𝑂𝑗(𝑝𝑗 , 𝑓𝑗 , 𝑢𝑗)
𝒑 Probability of conviction (measured as ratio of
convictions cleared to total offenses) Market supply function:
𝑂 = 𝑂(𝑝, 𝑓, 𝑢) 𝒇 Cost of punishment
𝒖 Additional determinants of criminal activity
𝑯𝒊 Harm from i th
crime (to society) 𝐻𝑖(𝑂𝑖)
𝑮𝒊 Gain from i th
crime (to offenders) 𝐺𝑖(𝑂𝑖)
𝑫 Net damage to society 𝐷(𝑂) = 𝐻(𝑂) − 𝐺(𝑂)
𝑨 Police & judicial “activity” to apprehend & convict
criminals
𝐴 = 𝑓(𝑚, 𝑟, 𝑐);
𝐴 ≅ 𝑝𝑂
𝐴 = ℎ(𝑝, 𝑂, 𝑎)
𝒎 Manpower devoted to apprehend & convict offenders
𝐴 = 𝑓(𝑚, 𝑟, 𝑐); 𝒓 Resources devoted to apprehend & convict offenders
𝒄 Capital devoted to apprehend & convict offenders
𝑪 Cost to apprehend & convict criminals 𝐶 = 𝐶(𝐴)
𝒇′ Social cost of punishment
𝑓′ ≡ 𝑏𝑓 𝒃
Multiplier that transforms individual cost of
punishment to social cost of punishment
𝑬𝑼 Expected utility of criminal activity 𝐸𝑈 = 𝑝 ∙ 𝑈(𝑌 − 𝑓) +
(1 − 𝑝) ∙ 𝑈(𝑌)
𝑳 Social loss from crime 𝐿 = 𝐿(𝐷, 𝐶, 𝑏𝑓, 𝑂)
𝐿 = 𝐷(𝑂) + 𝐶(𝑝, 𝑂) + 𝑏𝑝𝑓𝑂
�̂� Optimal number of offenses 𝑉 = 𝐺′(�̂�)
When 𝑏 = 0, 𝑉 = 𝐻′(�̂�) 𝑽 Marginal monetary value of punishments
𝑮 Private gain coming from criminal offenses
5
Model Components
a. Damages The motivation behind outlawing or restricting an activity is a belief that other members of
society are harmed by the activity, resulting in diseconomies. Here, the level of harm from
criminal activities is a function of the number of criminal offenses. However, the individuals that
commit crimes receive some benefit, which is also a function of the number of offenses. It
follows that the net damage is the difference between the harm to society and gain to the
individual offender.
Variable Notation Description
𝑂𝑖 Criminal activity level of i
th crime; measured as #
offenses
𝐻𝑖 Harm from i th
crime (to society)
𝐺𝑖 Gain from i th
crime (to offenders)
𝐷 Net damage to society
Functional
Relationship Description
First- and Second-
Order Conditions Description
𝐻𝑖 = 𝐻𝑖(𝑂𝑖)
Harm from crime is
function of #
offenses
𝐻𝑖′ > 0 Harm increases with # offenses
𝐻𝑖" > 0 Increasing marginal harm
𝐺𝑖 = 𝐺𝑖(𝑂𝑖)
Gain from crime is
function of #
offenses
𝐺𝑖′ > 0
Gain to offenders increases with #
offenses
𝐺𝑖" < 0
Diminishing marginal gains to
offenders
Cost of Crime
a. Damages
b. Apprehension & Conviction
(Public Expenditures)
c. Supply of Offenses
d. Punishment
Protection & Apprehension
(Private Expenditures)
6
𝐷(𝑂)= 𝐻(𝑂) − 𝐺(𝑂)
Net damage is
difference between
harm and gain
𝐷′ = 𝐻′ − 𝐺′ ? 0
Both 𝐻𝑖′ and 𝐺𝑖
′ > 0, so sign of 𝐷′
depends on relative magnitudes of
𝐻𝑖′ and 𝐺𝑖
′; Becker assumes 𝐷′ > 0
except in section V
𝐷" = 𝐻" − 𝐺" > 0
Increasing marginal harm for
additional offenses, assuming
𝐻𝑖" > 0 and 𝐺𝑖
" < 0
b. Apprehension and Conviction Public resources for police forces and courts are necessary to enforce laws and prosecute
criminals. In 1965 public expenditure on police and court “activity” totaled to over $3 billion
dollars. According to the estimates prepared for seven major felonies, the average cost per
offense was $500 and per person arrested was $2000.
The cost of public expenditures depends on the “activity” produced by police, judges, juries, and
the sophistication level of technologies (e.g. fingerprinting, lie-detector tests). Becker models the
cost of apprehension and conviction (A&C) as a function of “activity” (𝐴). He uses the number
of convictions as a convenient approximation of 𝐴, and also uses a more general function for
activity. He derives the first and second order conditions for cost that are necessary for an
interior solution to the loss minimization problem.
Variable Notation Description
𝑝 Probability of conviction (measured as ratio of
convictions cleared to total offenses)
𝐴
Police & judicial “activity” to apprehend & convict
criminals; activity involves inputs of manpower (𝑚),
resources (𝑟), and capital (𝑐)
𝐶 Cost to apprehend & convict offenders
Functional
Relationship Description
First- and Second-Order
Conditions Description
𝐴 ≅ 𝑝𝑂
Becker suggests #
convictions (𝑝𝑂)
as approximate
measure of 𝐴
𝐴 = ℎ(𝑝, 𝑂, 𝑎)
More general
function
incorporates other
determinants of
“activity” (𝑎), e.g.
arrests;
ℎ𝑝 > 0
ℎ𝑂 > 0
ℎ𝑎 > 0
Reasonable assumptions that
activity level increases with 𝑝, 𝑂, 𝑎
7
𝐶 = 𝐶(𝐴)
Cost is a function
of the activity
level;
First, assumes
𝐶 = 𝐶(𝑝𝑂)
𝐶′ =𝑑𝐶
𝑑𝐴> 0
A&C cost increases with activity
level (e.g. adding more police force)
𝐶𝑝 =𝜕𝐶(𝑝𝑂)
𝜕𝑝= 𝐶′𝑂 > 0
A&C cost increases with respect to
increases in probability of
conviction (i.e. if more people
convicted via higher police &
judicial activity, then costs are
higher for higher activity level)
𝐶𝑂 =𝜕𝐶(𝑝𝑂)
𝜕𝑂= 𝐶′𝑝 > 0
A&C cost increase wrt increases in
# of offenses (i.e. if more crimes
committed then activity level must
increase to maintain same
probability of conviction, thus cost
is higher for higher activity level)
𝐶𝑝𝑝 = 𝐶"𝑂2 > 0 Assumes that 𝐶" > 0, which implies
that the Hessian matrix of second
derivatives for the cost function is
positive definite can find a
minimum
𝐶𝑂𝑂 = 𝐶"𝑝2 > 0
𝐶𝑂𝑝 = 𝐶𝑝𝑂 = 𝐶"𝑝𝑂 + 𝐶′
> 0
Then, applies more
general function
𝐶 = 𝐶(𝑝, 𝑂, 𝑎)
𝐶𝑝 > 0
𝐶𝑂 > 0
𝐶𝑎 > 0
Follows from assumptions that
ℎ𝑝, ℎ𝑂 , ℎ𝑎 > 0
𝐶𝑝𝑝 ≥ 0
𝐶𝑂𝑂 ≥ 0
𝐶𝑝𝑂 ≅ 0
Sufficient assumptions to get interior
solution; suggests that first two
conditions are reasonable but third
is less reasonable
c. Supply of Offenses Why do people commit crime? Diverse theories suggest that skull type, genetic predisposition,
family up-bringing, and disenchantment with society are determinants of crime. Among the
various explanations, there is consensus that an increase in the probability of conviction or
punishment would generally decrease the number of offenses committed. Thus, Becker models
the supply of offenses as a function of 𝑓 and 𝑝, as well as other determinants 𝑢.
Variable Notation Description
𝑂𝑗 Crime level committed by j
th offender; measured as
# offenses
𝑝𝑗 Probability of conviction for j th
offender
𝑓𝑗 Cost of punishment to the j th
offender
𝑢𝑗
Various other determinants of criminal activity for j th
offender (e.g. income available through legal or
illegal means, arrest history, willingness to break the
law)
𝑌𝑗 Income from a criminal offense for j th
offender
8
Functional
Relationship Description
First- and
Second-Order
Conditions
Description
𝑂𝑗 = 𝑂𝑗(𝑝𝑗, 𝑓𝑗 , 𝑢𝑗)
Supply of offenses
committed by
individual j is function
of 𝑝𝑗 , 𝑓𝑗 , 𝑢𝑗
𝑂𝑝𝑗=
𝜕𝑂𝑗
𝜕𝑝𝑗< 0
𝑂𝑓𝑗=
𝜕𝑂𝑗
𝜕𝑓𝑗< 0
If probability of conviction (𝑝) or
punishment (𝑓) increase, then the
supply of offenses with decrease
𝑂𝑢𝑗=
𝜕𝑂𝑗
𝜕𝑢𝑗 ? 0
Can reasonably anticipate the effects of
𝑢𝑗 components (e.g. an increase in
legal income should decrease #
criminal offenses)
𝑂 = 𝑂(𝑝, 𝑓, 𝑢)
Market supply of
offenses by all
individuals j,
simplified by using
average values 𝑝, 𝑓, 𝑢
𝑂𝑝 =𝜕𝑂
𝜕𝑝< 0
𝑂𝑓 =𝜕𝑂
𝜕𝑓< 0
𝑂𝑢 =𝜕𝑂
𝜕𝑢 ? 0
Same FOCs and SOCs as individual
supply functions
Becker uses theories of choice under uncertainty to explain the conditions for a person to engage
in an illegal activity: an offender’s expected utility from committing an offense exceeds the
utility from using his/her resources at other activities. Therefore, Becker argues, some people
become “criminals” not because they have different determinants for criminal behavior, but
because of different utilities and choice sets.
Functional Relationship First- and Second-Order
Conditions Description
𝐸𝑈 = 𝑝𝑈(𝑌 − 𝑓) + (1 − 𝑝)𝑈(𝑌)
𝐸𝑈𝑝 = 𝑈(𝑌 − 𝑓) − 𝑈(𝑌) < 0 Expected utility of an
offense decreases with
either an increase in 𝑝
or 𝑓 𝐸𝑈𝑓 = −𝑝𝑈′(𝑌 − 𝑓) < 0
Implication: An increase in the probability of being convicted “compensated” by an equal
percentage reduction in the “price” for convicted offense would not change expected income, but
would change the expected utility as it would affect the risk associated with the crime.
𝜀𝑝 = −𝐸𝑈𝑝
𝑝
𝑈= [𝑈(𝑌) − 𝑈(𝑌 − 𝑓)]
𝑝
𝑈
𝜀𝑓 = −𝐸𝑈𝑓
𝑓
𝑈= 𝑝𝑈′(𝑌 − 𝑓)
𝑓
𝑈
𝑈(𝑌) − 𝑈(𝑌 − 𝑓)
𝑓
𝑈′(𝑌 − 𝑓)
The widespread generalization that offenders are more deterred by the probability of conviction
than by the punishment itself turns out to imply that in the expected-utility approach that
offenders are risk referrers, at least in the relevant region of punishments.
9
Attitude towards risk Increase in 𝒇 Increase in 𝒑
Risk – averse (𝑈′′ < 0) Greater effect Lower effect
Risk – neutral (𝑈′′ = 0) Same effect Same effect
Risk – seekers (𝑈′′ > 0) Lower effect Greater effect
Implication: the level of crime depends not on the efficiency of police and the level of
expenditure on combatting crime, but on the offender’s attitude towards risk. Thus, if offenders
were risk takers at some level of probability and price, and risk avoiders at other, then policy
makers could minimize social cost by selecting 𝑝 and 𝑓 in the regions where risk is preferred, i.e.
where “crime does not pay”.
d. Punishment In the model, the cost of punishment to offenders (𝑓) is the monetary equivalent of the type and
level of punishment. Empirically, this can be done easily when fines are the method of
punishment, but it is much more complicated to measure the cost of imprisonment or other types
of punishment.
Except for fines, most punishments hurt not only the offenders but other members of society as
well. For example, money from taxpayers is spent on probation, parole, and imprisonment.
Therefore, Becker uses the total social cost of punishment:
Total social cost of punishment = Cost to offenders + Cost (−gains) to society
Variable Notation Description
𝑓𝑗 Cost of punishment to the offender j
𝑓′ Social cost of punishment
Non-Measureable:
- Value placed on freedom - Value placed on restricted consumption
Costs of Imprisonment
Measureable:
- Discounted sum of earnings forgone
- Expenditures on prison personnel, facilities, food, etc.
- Subjective - Increases with duration of imprisonment
- Varies with income potential - Increases with duration of imprisonment
10
𝑏 Multiplier that transforms 𝑓 to 𝑓′
For mathematical convenience, total social costs are expressed as the transformation of costs to
offenders by constant 𝑏:
𝒇′ ≡ 𝒃𝒇
3. OPTIMALITY CONDITIONS
Variable Notation Description
𝐿 Social loss from criminal offenses
𝐷 Damages
𝐶 Costs of apprehension & conviction
𝑏𝑓 Social cost of punishment
𝑂 # offenses
Functional
Relationship Description First- and Second-Order Conditions Description
𝐿 = 𝐿(𝐷, 𝐶, 𝑏𝑓, 𝑂) Social loss is function
of 𝐷, 𝐶, 𝑏𝑓, 𝑂
𝜕𝐿
𝜕𝐷> 0
𝜕𝐿
𝜕𝐶> 0
𝜕𝐿
𝜕𝑏𝑓> 0
Reasonable
assumptions that
social loss
increases with
increased damages,
A&C costs, or
social costs from
punishment
𝐿 = 𝐷(𝑂) + 𝐶(𝑝, 𝑂) + 𝑏𝑝𝑓𝑂
Less general, more
convenient function
for social loss equal
to loss in real income
from 𝐷(𝑂), 𝐶(𝑝, 𝑂),
and 𝑏𝑝𝑓𝑂 associated
with crime level (𝑂)
𝜕𝐿
𝜕𝑓= 𝐷′𝑂𝑓 + 𝐶′𝑂𝑓 + 𝑏𝑝𝑓𝑂𝑓 + 𝑏𝑝𝑂 = 0
𝜕𝐿
𝜕𝑝= 𝐷′𝑂𝑝 + 𝐶′𝑂𝑝 + 𝐶𝑝 + 𝑏𝑝𝑓𝑂𝑝 + 𝑏𝑓𝑂 = 0
First-order
optimality
conditions
𝜕2𝐿
𝜕𝑓2 = (𝐷+C)𝑂𝑓2 + 𝑏𝑝(1 − 𝐸𝑓)𝑂𝑓 > 0,
where 𝐸𝑓 =−𝜕𝑓
𝜕𝑂∙
𝑂
𝑓
Sufficient second-
order optimality
conditions to
minimize social
loss
For each of the first-order partials of the loss function, you can rearrange the terms, divide by 𝑂𝑓
or 𝑂𝑝, and substitute 𝜀𝑓 = −𝑓
𝑂𝑂𝑓 and 𝜀𝑝 = −
𝑝
𝑂𝑂𝑝 to get the following two equations:
𝐷′ + 𝐶′ = −𝑏𝑝𝑓 (1 −1
𝜀𝑓) (21)
𝐷′ + 𝐶′ + 𝐶𝑝1
𝑂𝑝= −𝑏𝑝𝑓(1 −
1
𝜀𝑝) (22)
Equations 21 and 22 represent the optimality condition that 𝑀𝐶 = 𝑀𝑅. The left-hand side is
interpreted as the marginal cost of increasing the number of offenses through a reduction in
11
either𝑓 (equation 21) or 𝑝 (equation 22). Similarly, the right-hand side gives the marginal
revenue with respect to 𝑓 or 𝑝. These optimality conditions are displayed in Figure 1 below.
With respect to either 𝑓 or 𝑝, the social loss is minimized when marginal costs are equal to
marginal revenues (the point of intersection). Marginal cost of increasing number of offenses, O,
is less when the probability, 𝑝, rather than “price” of punishment, 𝑓, is reduced. Same holds for
marginal revenues, but on the condition that Ɛ𝑝 > Ɛ𝑓 , where Ɛ𝑝 and Ɛ𝑓 are elasticities of 𝑝
and 𝑓. This is precisely the condition indicating that offenders are risk seekers and that “crime
does not pay”. The job of the policy makers is to choose certain level of 𝑝 and 𝑓 to satisfy the
optimal level �̂�.
4. SHIFTS IN THE BEHAVIORAL RELATIONS This section shows the effect of shifts in damages, costs and supply of offenses on the optimal
value of p and f. This section aims to explain why more damaging offenses, such as murder, are
punished more severely and more impulsive offenders, such as parking in a wrong spot, less
severely?
12
a. Effect of 𝐷′
Shift Effect Observation Optimal number
of offenses
D' (increase in
marginal
damages) (Fig.2)
Increases MC by
changing either p
or f.
p and f move in the same
direction - increase. The more
severe the offense, the more
resources we spend in both
prevention and punishment
decreases
C' (increase in
marginal costs)
Same as in D' Same as in D' decreases
13
b. Effect of 𝐶𝑝
Shift Effect Observation Optimal number of
offenses
Cp
(increase in other
components of cost)
(Fig. 3)
No direct effect on
MCf of changing
offenses, but reduces
MCp
p and f move in
opposite directions.
Reduces p and only
partially
“compensates” with
increase in f
increase
C' + Cp Some effect on MC.
Direction depends on
the relative
importance of the
changes.
increase OR
decrease
Note: 𝐶𝑝 and 𝐶′ differ significantly between different kinds of offenses. It is easier to solve
rape or armed robbery, than a burglary or auto theft because of the availability of the personal
identification, i.e. if you cannot identify your offender (auto theft), the chances of catching
him/her are slim (Table 2). For example, n increase in salary of policemen increase both C' and
Cp, while improvement in technology, say, fingerprinting will reduce both, but not by the same
extent. Implication: although improvement in technology may or may not reduce p and optimal
14
number of offenses, it does reduce f. It possibly explains why we are moving away from capital
punishment.
Table 2
*Adapted from the original paper. Only prison term in federal civil institutions and probability of
conviction of those found guilty in known offenses are included.
Murder Rape RobberyAggravated Assualt
Bugary LarcenyAuthoTheft
Probability of apprehension andconviction (percent)
57.9 37.7 25.1 27.3 13 10.7 13.7
Average prison time (months) 111 63.6 56.1 27.1 26.2 16.2 20.6
0
20
40
60
80
100
120
0
10
20
30
40
50
60
70
Probability of Conviction and Prison Time* (1960)
15
c. Effect of 𝜀𝑓 and 𝜀𝑓
Shift Effect Observation Optimal number of
offenses
Ɛf'
(increase in elasticity
of 𝑓)
(Fig. 4a.)
Given 𝑏 > 0,
Increases 𝑀𝑅𝑓 of
changing offenses
Decrease in optimal
𝑓 partially
“compensated” by
increase in 𝑝
increase
Ɛp'
(increase in elasticity
of 𝑝)
(Fig.4b.)
Given 𝑏 > 0,
increases MR of
changing offenses by
changing p
Decrease in optimal 𝑝
partially
“compensated” by
increase in 𝑓.
Increase
if 𝑏 = 0, i.e. in cases of zero social cost (fines), then both 𝑀𝑅 functions lie along horizontal axis
and changes in elasticities have NO effect
Note: The income of the firm would be larger if it could separate its total market into submarkets
by the level of elasticity of demand: higher prices would be charged in markets with lower
elasticities. In case of the “offense” market, total social loss could be reduced by “charging”
lower “prices” (p’s and f’s) in markets with lower elasticities.
Example: It is possible to separate persons who have committed the same type of crime into
different into groups that have different responses to punishment. For instance, groups of
unpremeditated murders versus serial killers (sane ones). Unpremeditated murders are supposed
to act impulsively and to be relatively irresponsive to the size of the punishment (i.e. low
elasticity of f). Or, young adults or insane people have lower elasticity of both 𝑝 and f.
16
5. FINES
Welfare Theorems and Transferable Pricing A well-known condition in welfare economics is 𝑀𝐶 = 𝑃, so welfare depends only on levels and
not on the slopes of marginal cost and revenue functions. For Becker’s theory of social loss from
crime, the slopes incorporated in the elasticity of supply do matter. Why the difference? The
reason is that the “price” paid by the offenders is not fully transferred to the society, so there is a
social loss incurred. With no social loss as in case of fine (𝑏 = 0), the elasticity of supply would
drop out of the optimality condition and 𝑀𝐶 and 𝑀𝑅 would lie on the x-axis.
When there is a social cost incurred (𝑏 > 0), as with imprisonment, it means some of the
“payment” by offender is not transferred to the rest of the society. In such a case the elasticity of
the supply (ex. Eq. 21) of the offenses becomes important in the optimality condition, because it
determines the change in the social cost caused by the change in punishment, 𝑓.
Optimality Conditions In the next section, Becker looks several cases where 𝑏 = 0 and varies the parameters 𝑓 and 𝑝.
He derives the optimality conditions for each case in order to show the optimal level and
mechanism for the penalty.
Variable Notation Description
�̂� Optimal number of offenses
𝑽 Marginal monetary value of penalties
𝑮 Private gain coming from criminal offenses
𝑯 Harm to society
𝑪 Cost of apprehension and conviction
Case Optimality Condition Description Implication
𝑏 = 0
𝐶′ = 0
𝐷′(�̂�) = 0 Marginal damage at the
optimal is 0 Because 𝐶 = 0, then can set
probability of conviction at
highest possible value 𝑝 = 1
without cost. Thus, monetary
value of punishment is simply
equal to the level of punishment
𝑓 = 𝐻′(�̂�).
𝑯′ completely compensated by
𝒇.
𝑉 = 𝐺′(�̂�)
Marginal $ value of
punishments should equal
marginal gain from
committing the offense
𝑉 = 𝐻′(�̂�)
Marginal $ value of
punishments should equal
marginal harm to society
𝑏 = 0
𝐶′ > 0
𝑝 = 1
𝐷′(�̂�) + 𝐶′(�̂�, 1) = 0
Optimality condition also
incorporates the marginal
cost of apprehension &
conviction
The optimal fine should equal
marginal harm plus marginal cost:
𝑓 = 𝐻′(�̂�) + 𝐶′(�̂�, 1).
Fines should also compensate
for costs to apprehend &
convict offenders.
17
𝑏 = 0
𝐶′ > 0 𝑓 = 𝑐𝑜𝑛𝑠𝑡𝑎𝑛𝑡
𝐷′(�̂�) + 𝐶′(�̂�, 𝑝)
+ 𝐶𝑝(�̂�, �̂�)1
𝑂𝑝= 0
𝑝 allowed to vary so
optimal condition
incorporates partials wrt
probability of conviction
Sign of 𝐶𝑝 depends on whether
penalty is changed through fine 𝑓
or probability of conviction 𝑝
The Case for Fines
Proposition: Social Welfare is increased if fines are used (whenever feasible)
Pros Cons
Fines do not use up social resources (no
expenditure on guards, supervisory personnel,
etc) (Table 1).
Need to know the 𝐺′, 𝐻′ and 𝑀𝐶 – not easy to
come by
Determination of 𝑂and 𝑓 is simplified Immoral*
Use of other 𝑓 in addition requires the
knowledge of elasticities.
No amount of money could compensate for
murder.
Other 𝑓 fail to compensate. Moreover, they
require “victims” to spend additional
resources to carry out punishment.
Whenever 𝐻 exceeds the resources of the
offender, they have to “pay” in other way:
through prison time (“debtor prisons”) **.
No anger and fear towards ex-convicts who
have not “paid their debt to society”.
Fines will discriminate against those who
cannot pay them. Thus, rich people will be
able to “buy” their freedom.
* Counterargument: fine is price in monetary unit, imprisonment is measured in time units.
** That’s why more serious crimes are punished by prison time, parole, or sometimes, death
sentence.
“Fairness” of the imprisonment in lieu of the full fine for those who cannot pay it depends on the
value placed on the prison time, which varies with offender’s financial standing.
“Value” of prison time
Poor Rich
Low: spend more time in planning their
offenses, like court appearances, to reduce
𝑝. The cost of conviction (fine) is relatively
larger to the value of their time
High: spend more money on planning
offenses, like hire good lawyers, even bribing
someone to reduce 𝑝. The cost of conviction
is smaller than the value of the prison time.
For example, in New York State, Class A Misdemeanors can be punished by a prison term as
long as one year or a fine no larger than $1000 (by effectively setting the exchange rate between
dollar and value of the prison time as $1000 365 = $2.74 ⁄ for a day). Often the poor are forced
to choose the prison.
Compensation and the Criminal Law If punishment by optimal fines became the norm:
The aim of legal proceedings would be not punishment and deterrence but assessment of
“harm”.
18
“Criminal” would be defined not by the nature of the activity, but by the offender’s
inability to compensate for the “harm”
Example: Monopolies are outlawed in the US because they pose constraints on trade that are
harmful to society. When accused in practices defendants often, become subject of damage suits
or are jailed. With optimal the gains to them would be less than “harm” they cause, thus, they
would cease their “harmful” activities.
However, it is not easy to correctly measure all the “harm” done and mistakes are inevitable. But
Becker argues that with experience the “margin of error would decline and rules of thumb would
develop”.
6. SOME APPLICATIONS (BECKER) Discussions of external economies are usually perfectly symmetrical to those of diseconomies,
however, there is no equivalent to police authorities to “apprehend” and “convict” benefactors.
There exist such awards as medals, titles, or Nobel Prize, to reward benefactors, but it’s a
relatively small number.
Similarities: the Model 𝐴 (𝐵) Net social advantage from
𝐵, benefits
𝐷(𝑂)
𝐾(𝐵, 𝑝1) Cost of apprehending and
rewarding, where 𝑝1 is the
probability of doing so
𝐶 = 𝐶(𝑝, 𝑂, 𝑎)
𝐵(𝑝1, 𝑎, 𝑣) Supply of benefits, where
𝑎 is reward per benefit and
𝑣 is other determinant
𝑂 = 𝑂(𝑝, 𝑓, 𝑢)
𝑏1 Faction of 𝑎 that is the net
loss to society
𝑏
Π = 𝐴(𝐵) − 𝐾(𝐵, 𝑝1) − 𝑏1𝑝1𝑎𝐵 Profit function showing
increase in income to
society
𝐿 = 𝐷(𝑂) + 𝐶(𝑝, 𝑂) + 𝑏𝑝𝑓𝑂
…………
Differences: Implications 𝜀𝑝 > 𝜀𝑓 – at the margin, benefactors are risk
avoiders
𝜀𝑝 > 𝜀𝑓 – at the margin, offenders are risk
seekers
Optimal 𝑝1 and 𝑎 are in the regions where
“benefits do pay”: the marginal income
available to benefactor is greater than the one
available to them in a less risky activity
Optimal 𝑝 and 𝑓 are in the regions where
“crime does not pay”: the marginal income
of criminals is less than that available to
them in a less risky legal activity
The Effectiveness of the Public Policy What determines the “effectiveness” of public efforts to discourage offenses? The model shown
earlier can be used to answer this question if social welfare is measured by income and if
19
E = 𝑀
𝐵
Where E means “effectiveness”, 𝑀 – maximum feasible increase in income and 𝐵 – increase in
income if all offenses causing net damage were abolished.
𝑀 is achieved by choosing optimal values of 𝑝 and 𝑓. E essentially depends on 𝐶, and elasticities
of 𝑝 and 𝑓. The smaller the cost and the greater the elasticities, the smaller the cost of reducing
offenses and thus the greater the E.
E differs among different offenses more because of differences in 𝐶, than in elasticities of
response. The cost 𝐴&𝐶 is higher for those crimes where there is a longer time period between
commission and detection of an offense. It means that effectiveness is greater for robbery than
for a related felony like burglary than for, say, antitrust regulation.
A Theory of Collusion “Collusion” – efforts by competing firms to collude in order to obtain monopoly profits.
𝐺(𝑂) - positively related to the elasticity 𝑀𝐶 curve and is inversely related to elasticity of their
collective demand curve.
𝑂- is setting price below or producing more than is agreed by colluding firms.
𝐻(𝑂)- harm to collusion depends on the number of violation and the elasticities of the 𝑀𝐶 and
Demand curves.
𝐶 = 𝐶(𝑝, 𝑂, 𝑎) − cost of discovering and “apprehending” violators. Similarly, it is increasing in
𝑝, 𝑂, 𝑎, where 𝑎 is the cost of punishing violators. Since fines are rules out, methods like
predatory price-cutting or violence has to be used, which are less costly than legal recourse. Thus
“syndicates” in US have advantage.
Elasticities are crucial: the prices and output levels would be set close to competitive ones, if
demand is elastic, there are many buyers and sellers, if punishments are less transferrable, and if
the government is hostile towards collusion.
7. EXTENSIONS/CRITICISM
“Prison”- Community This model primarily deals with the society as a whole: how to minimize the social loss to
society by reaching an optimum level of crimes. An interesting extension to the model would be
to consider communities that have federal prisons vis-à-vis those that don’t:
For “prison” - communities’ is 𝐺(𝑂) higher or lower?
Are such communities inclined to favor prison time over fines as a method of
punishment?
Are offenders in “prison”-communities more risk seeking?
20
“Expensive” Prison Time Supplementing fines with prison time is common practice, and in considering the “value” of the
prison time it seems that it is taken from the perspective of the poor person. An interesting study
would be to determine an “optimum exchange rate” - $/prison time, that would minimize social
cost to society. Consider the following:
“Harm” inflicted on offender who serves the jail time instead of paying fine: stigma,
being away from work, family… The length of prison time matters here
How significant is the length of time on recidivism
“Cost” of Judicial Error 𝑝 – the probability of conviction, as calculated as a ratio of the numbers of conviction over the
total number of offenses, is an important variable in Becker’s constrain minimization problem.
Nowhere in the model, however, is the probability of judicial error, 𝑝𝑒 , incorporated. In solving
the constrained minimization problem it would also be interesting to consider the following:
The social cost of futile legal proceedings
The Social loss of convicting innocent
𝜀𝑝 of convictions as , 𝑝𝑒 increases, i.e. are judges more inclined to leniency if they
know 𝑝𝑒 is high
Why does murder have a higher probability of conviction? Evidence suggests (Table 2) that there is a positive correlation between the severity of the crime
and the probability of conviction. Becker explains the correlation by the fact that “personal”
crimes are easier to solve because the traces of personal identification are present. While we
agree with the author, we also believe that there must be stronger reasons for that. Such heinous
crimes as murder and rape generate much more indignation, i.e. the emotional damage to society
is much higher. Thus, the society is much more driven and “unconsciously” allocates more
resources (monetary and physic) into solving them.