This is an unpublished paper which follows on from :
Victim Compensation Does Not Increase the Supply of Crime.
Journal of Economic Studies 16(4) 1989 53-60
ARE THERE ABSOLUTE CRIME INDUCEMENT EFFECTS OF VICTIM COMPENSATION?
I. INTRODUCTION
The provision of public assistance to crime victims has
come under strong criticism from public choice economists
[Meiners(1977,1978) Magaddino(1973)]. They identify two major
problems. Firstly, the supply of crime may increase due to 'moral
hazard' induced by the compensation payments [Becker &
Ehrlich(1972)]. Secondly, those responsible for administering
the system of payments will pursue their own self interest to the
extent of expanding the compensation program beyond its optimal
size. Despite the confidence with which these opinions are held
there has been little attempt to test them. A recent paper by
Cameron(1989) fails to find any support for the first argument.
The purpose of this paper is to take into account certain
limitations in Cameron's paper which looked only at marginal
inducement effects on the supply of crime. In it, crime rates
were regressed on the probability and size of compensation
awards. Such an approach limits the sample size so that the
absolute inducement effect can not be studied i.e there is no
comparison of crime rates between states with schemes and those
without them.
This paper studies the absolute inducement effect
by estimating crime supply functions which contain a dummy
variable for the presence of a victim compensation scheme. The
age of the scheme is also take into account. Some
support for the public choice school approach is found, for the
case of rape, when a restricted from of the supply function is
used. The addition of further socio-economic variables results
in a null effect of victim compensation scheme presence. No
support is found for any of the equations used for assaults.
II. THEORETICAL BACKGROUND.
In the absence of a compensation scheme individuals are
expected, in the traditional economic approach to crime
[Becker(1968),Cameron(1988)],to respond to the possibility of
crime in terms of maximizing subjective expected utility. The
presence of a compensation scheme increases individual welfare as
it adds to expected wealth. It should have substitution as well
as income effects. In the absence of compensation individuals
should optimize with crime avoidance strategies such as choosing
particular methods of coming home late at night or carrying
weapons or alarms. The might also reduce the ex post costs of
victimization by carrying less money and valuables or having some
form of insurance to cover recovery costs. The aforementioned
activities involve the loss of potential production or
consumption. The possibility of state provided compensation
should cause substitution away from private protection and
insurance towards more private production and consumption. Thus
compensation awards provide a subsidy to risk taking; this is
commonly known as the moral hazard proposition. The propensity
to apply for compensation will be some function of information.
It is well documented that not all those eligible for state
provided benefits actually take them up. In a world of full
information and optimizing behaviour individuals would be
expected to weigh up the expected marginal costs of applying
against the expected marginal gains of an application. In
Cameron(1989) a strong informational approach was taken as
claimants were expected to respond at the margin to small changes
in the expected value of awards and success of claims. Lack of
information can be an important constraint on victim compensation
programs; for example, Fujii & Mak(1979) report that many
victimized visitors to Hawaii leave without being aware of their
rights to apply for compensation. Lack of,or costly, information
may lead to satsificing which consequently involves behavioural
modification that does not take place at the margin. Individuals
may modify their behaviour if they are aware of a scheme. In
terms of the 'moral hazard' scenario potential victims become a
bit more lax in self-protecting behaviours because they think
they might qualify for some victim compensation. A satisficing
approach thus implies absolute rather than marginal crime
inducement effects. The presence of a scheme should decrease self
protection levels and thereby increase crime rates. Increasing
the generosity of a scheme may not influence behaviour unless
there is sufficient accompanying publicity which is also regarded
as credible.
If people satisfice they may exhibit a degree of rational
ignorance in that they may not be aware that schemes exist. A
reasonable hypothesis is that people are more likely to be aware
of a scheme the longer it has been in existence. We can allow
for this in the econometric work by including the age of the
scheme as an independent variable.
There is a counter to the 'moral hazard' argument. This is
that victim compensation schemes may actually deter crimes.
Catching perpetrators of crimes is well nigh impossible without
the cooperation of the victim. A condition of compensation
schemes is (see below) that victims collaborate with the police.
This may produce an exogenous upward shift in the police
production function. If we accept that this will deter crime,
which is by no means well established [see Cameron(1988)] then a
case could be made that the schemes will reduce crime rather than
increase it. 1
III. INSTITUTIONAL BACKGROUND.
Victim compensation schemes have been introduced at the
state level. By 1980, 30 states had schemes. available Although
the first scheme began in 1965, victim compensation is primarily
a mid to late 70's phenomenon as is clearly demonstrated by Table
I. All payments are subject to discretion and can be made only if
certain criteria are met.
These are:
(a) the crime must be unprovoked and carried out by a
non-relative of the victim
(b) victims must co-operate with the
police in the further pursuit of the offender
(c) there is usually an income requirement which excludes richer
victims of crime
(d) compensation is usually paid only to victims of violent
crime.
Given these conditions and the distress which would attend
complying with them it seems difficult to believe in the
anti-compensation position of Blackorby & Donaldson (1988) who
allege that "if rape victims are offered cash everyone has an
interest in qualifying "(p.691).
IV.EQUATION SPECIFICATION.
We investigate the inducement effects of victim
compensation by estimating crime supply equations for rape and
aggravated assault. These are the most suitable index crimes as
others do not generally by themselves bring eligibility for
benefits. The average award and award probability variables in
the Cameron(1989) equation are replaced by VCDUM which equals 1
for states which have a scheme and 0 otherwise. To control for
the possible influence of other variables HOUSE and the
respective clearance rates CRRAPE and CRAGASS are used.
HOUSE, a measure of population turnover is a portmanteau
variable. It captures three sets of factors:
(i) implicit surveillance; stable populations mean that criminals
are more easily found to be suspicious
(ii) risk shifts; newer residents are less well informed about
the safest way to conduct themselves in the neighbourhood
(iii) economic activity; more families will move into an area
which is expanding.
This variable was chosen in part in the earlier paper because of
the small sample size. Given the larger sample used in this
paper we added some of the variables commonly used in models of
crime.
The variables used are personal income (PINC), unemployment
rate (UR), poverty (POV), and the percentage of the population
black (PCBLK). PCBLK is expected to have a positive coefficient
because the inducements to crime for blacks are greater i.e. they
face worse labour market conditions etc. There is no suggestion
that blacks are more likely ceteris paribus to perform crime and
indeed there is no evidence of this [Gyimah-Brempong(1986)]. The
UR and POV variables are usually [see e.g.Ehrlich(1973)]
expected to have a positive sign on the grounds that they make
crime more attractive to a larger section of the population. The
income variable is often regarded as a proxy for 'loot' in that
higher incomes mean there is more to steal. There is of course
the opposite effect that higher income may mean that there is
less need to steal.
The effects of the above variables are rationalized in
terms of subjective expected utility [see e.g. Ehrlich(1973)] but
they might just as readily pick up environmental influences. This
seems particularly likely in the case of personal crime. To take
the Chicago model seriously we have to believe that a rapist
might sit down and calculate precisely that a 10% wage rise means
it is not worth his while to carry on raping as the opportunity
costs are too great.
The sources and definitions of all variables are given in
Table I.
V.RESULTS.
The basic equations replicating Cameron(1989) i.e.
regressions of crime rates on HOUSE, clearance rates and the
compensation dummy do not show a significant impact of schemes on
crime when the whole sample is used. This is still true for
assaults when we omit Hawaii and D.C. from the sample. However
the restricted sample produces the following result for rapes:
PCRAP = 0.19 -0.0045 CRRAPE + 0.014 HOUSE + 0.055 VCDUM
(1.41) (2.66) (4.72) (1.86)
n=49 R.SQ.ADJ = 0.49 {absolute 't' ratios in brackets}
This gives support to the public choice view as the
coefficient on VCDUM is significant at the 5% level on a
one-tailed test. Its point estimate is 17.24% of the sample mean
rape rate which seems a fairly large effect. The above
regression does not control for the age of the scheme. Doing
this by adding YRVCST gives us:
PCRAP = 0.14 -0.0035 CRRAPE + 0.014HOUSE + 0.67VCDUM -0.008YRVCST
(0.99) (2) (4.87) (1.89) (1.74)
n=49 R.SQ.ADJ.= 0.51 {absolute 't' ratios in brackets}
The result for VCDUM is not altered but the YRVCST variable
offsets the crime increasing impact of the scheme. Its negative
coefficient indicates that the rape rate is less in newer
schemes. It turns out that this offset effect is very small.
Adjusting VCDUM using the YRVCST coefficient gives us a
differential of -0.47 for the newest scheme (1980) and -0.51 for
the oldest (1965). These are massive point estimates given that
the mean of the rape rate is only 0.319. It is so large it seems
somewhat implausible suggesting that there may be severe omitted
variable bias.
The last equation is the most supportive of the public
choice school that we could generate. It is open to criticism on
the grounds that rape rate variations may well be due to factors
discussed above which have, thus far, been omitted from the
regressions. Including these variables removes the significant
impact of schemes found in the last equation. The results for
both equations are shown in Table II. These conclusions are not
altered if we move back to the full sample or drop either of the
VCDUM or YRVCST variables. Given that OLS estimation is used it
might be argued that the estimates are inefficient as the
covariance of the error terms in the rape and assault equations
is ignored. This is taken account of in the SUR {seemingly
unrelated regression} estimates shown in Table III. These do not
provide any support for the public choice view as the
coefficients on the compensation scheme variables are still
insignificant.
VI. CONCLUSION.
The body of public choice theory has, in the past, been
used to make propositions about potential crime behaviour in
response to victim compensation schemes. It has been argued that
the moral hazard problem results in an increase in crime rates
when compensation for victims is introduced. This paper estimates
crime supply functions for rape and aggravated assault to test
whether there are absolute inducement effects of victim
compensation schemes. Very little support is found for the moral
hazard proposition.
FOOTNOTES.
1. This is not taken into account in the empirical work reported
here. Experiments with a police production function including
victim compensation variables (not reported) failed to find a
significant positive impact.
REFERENCES
Becker,Gary S. & Ehrlich,Isaac H. 1972. "Market
Insurance,Self-Insurance and Self-Protection." 80 Journal of
Political Economy
Blackorby,C. & Donaldson,D.1988. "Cash versus Kind, Self Selection
and Efficient Transfers." 78 American Economic Review 691-700
Cameron,Samuel. 1988. "The Economics of Crime Deterrence: A
Survey of Theory and Evidence". 41 Kyklos 301-323
Cameron,Samuel. 1989. " Victim Compensation Does Not Increase
the Supply of Crime". 16 Journal of Economic Studies 53-60
Ehrlich,I.1973. "Participation in Illegitimate Activities: A
Theoretical and Empirical Investigation". 81 Journal of Political
Economy. 521-564
Fujii,J. & Mak,J. 1979. "The Impact of Alternative Development
Strategies on Crime Rates: Tourism Versus Agriculture in
Hawaii". 13 Annals of Regional Science 42-56
Gyimah-Brempong,K. 1986. "Empirical Models of Criminal Behavior:
How Significant a Factor is Race?". 15 Review of Black Political
Economy 27-44
Magaddino,Joseph.P.1973."Crime,Victim Compensation and the Supply
of Offenses".21 Public Policy 437-440
Meiners,Roger.1977. "Public Compensation of the Victims of
Crime". Ch.14. in Barnett,R.E. & Hagel,J. 1977. "Assessing the
Criminal: Restitution,Retribution and the Legal Process".
Ballinger. Cambridge,Mass.
Meiners,Roger.1978. Victim Compensation: Economic,Legal and
Political Aspects. Lexington: D.C.Heath.
Ramker,Gerald F. & Meagher,Martin S. 1982. "Crime Victim
Compensation: A Survey of State Programs" 46 Federal Probation
Reynolds,Helen. 1981. Cops and Dollars. Springfield:C.C.Thomas
Stigler,George J. 1970. "The Optimum Enforcement of Laws" 78
Journal of Political Economy 526-536
TABLE I: DEFINITIONS AND SOURCES OF DATA
PCRAP = Per Capita Rapes
PCASS = Per Capita Aggravated Assaults
HOUSE = Percentage of Households moving into residential unit
March 1979- March 1980
CRAGAS = aggravated assaults dleared up divided by the number of
aggravated assualts
CRRAPE = rapes cleared up divided by the number of rapes
PCBLK = percentage of population black
PINC = Per Capita personal income
UR = unemployment rate
POV = percentage of families below the officially defined poverty
level
VCDUM = Dummy variable =1 if state has a victim compensation
scheme ; zero otherwise
YRVCST = last two digits of the year in which a victim
compensation scheme was introduced.
All data, except for YRVCST, refer to 1980.
TABLE II: OLS ESTIMATES OF SUPPLY OF RAPE AND ASSAULT FUNCTIONS
Dependent Variable: PCRAP PCASS
Independent Estimated Estimated
Variable Coefficient Coefficient
INT -0.31 -0.42
(-1.16) (-0.17)
PINC 1.67e-005 2.12e-004
( 0.9) ( 1.09)
UR 1.79e-002 7.23e-002
( 2.35) ( 0.86)
POV -1.93e-003 2.19e-002
(-0.26) ( 0.27)
PCBLK 0.59 5.59
( 3.53) ( 2.98)
HOUSE 1.74e-002 7.45e-002
( 6.66) ( 2.6)
CRRAPE -2.24e-003
(-1.36)
CRAGAS -2.94e-002
( 2.01)
YRVCST -2.98e-003 -1.81e-002
(-0.7) ( 0.38)
VCDUM 0.26 1.03
( 0.79) ( 0.28)
Number of Observations 49 49
R-squared 0.72 0.49
Corrected R-squared 0.67 0.38
NOTES
1. Figures in brackets are 't' ratios.
TABLE III:SUR ESTIMATES OF SUPPLY FUNCTIONS OF RAPE AND ASSAULT
Dependent variable: PCRAP PCASS
Independent Estimated Estimated
Variable Coefficient Coefficient
INT -0.36 -0.11
(-1.60) (-4.93e-002)
PINC 1.88e-005 2.03e-004
( 1.14) ( 1.16)
UR 1.84e-002 6.89e-002
( 2.68) ( 0.9)
POV -1.55e-003 2.22e-002
(-0.23) ( 0.3)
PCBLK 0.59 5.55
( 3.89533) ( 3.28)
HOUSE 1.75e-002 7.4e-002
( 7.45) ( 2.86)
CRRAPE -1.79081e-003
(-1.38)
CRAGAS -3.2e-002
(-2.83)
YRVCST -3.12e-003 -1.75e-002
(-0.81) (-0.41)
VCDUM 0.27 0.98
( 0.91) ( 0.3)
Number of Observations 49 49
R-squared 0.72 0.49
Corrected R-squared 0.67 0.38
System R-squared 0.49
NOTES
1. Figures in brackets are 't' ratios.
2. Equations are estimated by the Seemingly Unrelated Regression
(SUR) method.