Transcript Diapositiva 1

The Effect of Education on Inprison Conflict: Evidence from
Argentina
Maria Laura Alzua
CEDLAS-Universidad Nacional de La Plata
Catherine Rodríguez
Universidad de los Andes
Edgar Villa
Universidad Javeriana
•
•
•
•
•
•
•
•
•
Motivation
Literature Review
Penal Legislation in Argentina
Educational Requirements of Inmates
Methodology
Data
Results
IV approach
Conclusions and Future steps
Motivation
• Two views of punishment in modern societies:
reductivism (i.e. reforming a prisoner lowers
future incentives towards violent and criminal
behavior) and retributivism (i.e. criminals should
be punished because they deserve it).
• Prison based education is thought to reform
prisoners so they have less incentives to
relapse: a) increasing opportunity costs through
potential future wages and/or b) affecting
behavior through preferences, risk aversion and
moral costs.
• Most empirical studies focus on channel a) for
ex-convicts and find evidence that prison based
education programs reduce recidivism rates.
• This study looks at individual behavior (violent)
within jails for prisoners that by law have to
participate in educational programs.
• Little on education in prisons.
• Little evidence for Latin America.
Literature Review
• Old criminological studies find in prison
education has a negative effect on crime
participation, but not controlling for selection into
programs.
• Steuer & Smith (2003) three state recidivism
study, educational programs lower the possibility
of re-arrest.
• Lochner & Moretti (2004), education reduces the
probability of incarceration.
• Tyler & Kling (2006) non white convicts increase
its after prison income after participating in
educational programs.
Policy Implications
• If violent behavior is acquired and has some
inertia, then lowering in-prison conflictivity may
lower crime in society once prisoners finish their
sentence. (Schnur 1949)
• If in-prison violent behavior is related to
recidivism, then education may lower crime.
(through different channels than that of
increasing legal income).
• Supporting reductivism view of the prison
system.
Penal Legislation in Argentina
•
•
•
Legislation changed in 1996. Progressive
system of 4 periods: Observation, Treatment,
Test and Parole.
Federal Penitentiary Service in charge of
federal prisons where convicted felons for
drug trafficking, money laundry, tax evasion
among others are sent.
Provincial Penitentiary Service in charge of all
other prisons at the province level.
Educational Requirements and
Provision
• Argentine educational system
–
–
–
–
Initial education (kindergarten): from 3 to 5 years
Elementary education EGB: from 6 to 15
Highschool Polimodal : 3 years
College and Graduate education.
• Law 24.660 (1996): prisons must ensure
education to (sentenced and remanded)
prisoners with less than 9 years of education.
• Prisoners must participate in educational
program unless they provide evidence of 9 years
of education. Good for exogeneity of treatment!
• Provision of primary and secondary schooling at
the province level is supplied by the province
and not the Federal Government which only
supervises.
• Province must guarantee a functioning school in
each prison. Severe shortage of teachers at
province level for adult-population makes the
mandate difficult to implement. Moreover, no
extra economic incentives for teachers to teach
at prisons.
• More restrictions for remanded prisoners to
attain education: a) high mobility of these
prisoners between prisons; b) valid ID card is
necessary to attain education (60% of inmates
do not have valid ID card).
Methodology
• Natural Experiment: law requires participation in
basic educational programs to achieve EGB but
not all prisoners that should comply with the law.
• Treatment: prisoners required by law to achieve
EGB and effectively do participate in a basic
educational program.
• Control: prisoners required by law to achieve
EGB but did not participate in basic educational
program.
• Identification assumption: Provinces
exogenously decide to provide basic education
program.
• If there is just partial exogenous variation calls
for IV for educ participation (see below).
• Linear Probability Model and Probit specifications
under prison fixed effects for pooled cross
sections:
Confit=a+beducit+dtimet+Xitc+gprisoni+Xitp+u
conf=1 if prisoner had a conflict in last 6 months
(sanctions and/or violent behavior),
educ=1 if prisoner belongs to treatment,
timet: time dummies,
prisoni: prison dummies.
Xc represent individual characteristics of prisoners:
age, working in prison dummy, marital status,
time in prison, if unemployed when entering,
sport activities dummy, prison break attempt,
medical assistance dummy, personal vists
dummy,
Xp represent prison characteristics that vary in time
and within prisons: number of prisoners,
average age of inmates, percentage of
murderers, rapists and thieves, average
education levels and percentage of failed prison
breaks.
Data
• Annual Census data from 2002 to 2005 of
prisoner population in Argentina.
• Data collection did not allow us to construct
panels at individual or prison level.
• Detailed information:
– Characteristics of prisoners when arriving at prison
– Type of crime committed at entry and stage of the
process in which the prisoner is at
– Participation in education program and the level of
education at entry
– Participation in conflicts due to bad behavior that was
sanctioned and if prisoner was involved in violent
behavior in the last year.
•
Measures of conflict within prison:
i) extreme violence (extvio): if prisoner
participated in violent behavior that ended up
in injuries and/or deaths
ii) violence (vio): if prisoner participated in violent
behavior that ended up in material damages
iii) Sanction (sanc): if prisoner had a sanction for
his behavior in last year
iv) Severe Sanction (sevsanc): if prisoner had a
sever sanction (individual confinement for
more than 15 days, transferred to another
facility)
Sample consists of Argentine males in the
prison state system, sentenced in the
“treatment" period, with incomplete legal
educational requirements.
Table 1
Descriptive Statistics of Prisons in Argentina 2002-2005
Gender
%
Schooling Level when
entering
Male
94.9
None
7.1
Single
69.4
Female
5.1
Primary Incomplete
23.1
Married
12.9
Primary Complete
47.9
Widow
1.2
Secondary Incomplete
13.2
Divorced
1.5
Secondary Complete
2.9
Living with Partner
12.8
%
Marital Status
%
Age
%
Legal status
%
Felony
%
Less than 18
0.1
Sentenced
40,4
Robbery/burglary
40,2
18 a 24
30.2
Remanded
58,0
Homicide
11,3
25 a 34
39,7
Minor robbery
7,3
35 a 44
17,7
Drug trafficking
7,0
45 a 54
7,9
Rape
4,8
55 a 64
2,8
other
28,0
65 or more
0,7
Period of Process
%
Observation
8,8
Treatment
54,3
Test
11,5
Parole
0,6
Table 2
Descriptive Statistics of Census 2002-2005
Number
%
Total Prisoners
34.349
100
Prisoners that had violent behavior (vio)
7.968
23,2
Prisoners with some type of sanction (sanc)
9.621
28,01
7.828
22,79
454
1,32
5.110
39
Prisoners with sever sanction (sevsanc)
Prisoners that had extreme violent behavior (extvio)
Participation in EGB
Table 3
Descriptive Statistics of Sample: Participation in Basic Educational Program at
Province Level during 2002-2005
Province
Total Prisoners
Participation in EGB
%
Buenos Aires
1.376
567
41,21
Catamarca
180
2
0,01
Córdoba
3.576
1.529
42,76
Corrientes
678
395
58,26
Chaco
635
447
70,39
Chubut
1.125
868
77,16
Entre Ríos
449
110
24,50
Formosa
306
138
45,10
Jujuy
171
62
36,26
La Pampa
199
70
35,18
La Rioja
22
0
0,00
Misiones
808
142
17,57
Neuquén
432
161
37,27
Río Negro
414
164
39,61
San Juan
338
43
12,72
San Luis
289
9
3,11
Santa Cruz
147
129
87,76
Santa Fe
1.773
225
12,69
Tierra del Fuego
24
4
16,67
Ciudad de BsAs
161
45
27,95
Results
Table 4: Pooled Probit Regressions
Extvio
Extvio
Extvio
Vio
Vio
Vio
-0,09
-0.105
-0.105
-0.175
-0.133
-0.146
(0.061)
(0.065)
(0.077)
(0.048)**
(0.051)**
(0.072)*
Individual controls
No
Yes
Yes
No
Yes
Yes
Prison controls
No
No
Yes
No
No
Yes
9165
9165
9165
9411
9411
9411
Educ
Observations
Robust standard errors in parentheses
One-tail test: * significant at 5%; ** significant at 1%
Table 4: (continued) Pooled Probit Regressions
Sancs
Sancs
Sancs
Sevsancs
Sevsancs
Sevsancs
-0.011
-0.040
-0.148
0.033
0.002
-0.118
(0.027)
(0.028)
(0.030)**
(0.028)
(0.029)
(0.031)**
Individual controls
No
Yes
Yes
No
Yes
Yes
Prison controls
No
No
Yes
No
No
Yes
10858
10858
10858
10299
10299
10299
Educ
Observations
Robust standard errors in parentheses
One-tail test: * significant at 5%; ** significant at 1%
• Sign of b is negative when all controls are
included for every variable used to measure
conflict.
• Educ participation is statistically significant for all
measures of conflict except for extreme
violence.
• Marginal effects: participating in education
program decreases probability of conflict by
– 0.5 percentage points for Violence,
– 3.5 percentage points for Sanctions,
– 5 percentage points for Sevsanc.
• Similar results are obtained with a linear
probability specification.
IV approach
– With our current knowledge, we cannot say
explicitly how education is allocated between
prisoners, even in the presence of a
protocole.
– IV estimation in order to overcome the
problem of selection into educational
programs.
– Instruments: a) Number of adult teachers per
prisoner (Teachers per 100 inhabitants); b)
Expenditure in education at province level
Table 5: First Stage of IV
Dependent variable: Educ
Number of teachers per prisoner
Expenditure on education at the province level
Prisoner works in prison
Years in prison
Age
Married
Unemployed entering prison
Number of felonies
Prison Controls
Observations
R-squared
Robust standard errors in parentheses
Two-tail test: * significant at 5%; ** significant at 1%
(1)
(2)
Probit
OLS
0.028
0.009
(0.004)**
(0.001)**
0.0011
0.0004
(0.00015)**
(0.00005)**
0.110
0.037
(0.027)**
(0.009)**
-0.029
-0.009
(0.005)**
(0.002)**
-0.007
-0.002
(0.001)**
(0.000)**
-0.111
-0.035
(0.045)*
(0.014)*
-0.106
-0.033
(0.030)**
(0.010)**
0.127
0.043
(0.018)**
(0.006)**
Yes
Yes
11346
11346
------
0.14
Table 5: IV Estimation
Probit
Extvio
Vio
Sancs
Sevsancs
0.504
-1.159
-0.603
-0.433
(0,44)
(0,64)*
(0,25)**
(0,25)*
Yes
Yes
Yes
Yes
Observations
9094
9339
10853
10294
R-squared
-------
-------
-------
-------
Educ
Individual and prison controls
Bootstrap standard errors (1000 replications) in parentheses
One-tail test: * significant at 5%; ** significant at 1%
• First stage shows that instruments are relevant
at 1%. Overidentification test for LPM shows that
joint exogeneity cannot be rejected at 1%.
• Sign of b is negative for every measure of
conflict except extvio which is positive but is the
only one in which educ is not statistically
significant.
• Educ participation is statistically significant for all
measures of conflict except for extreme
violence.
• Marginal effects increase: participating in
education program decreases probability of
conflict by
– 4 percentage points for Violence,
– 23 percentage points for Sanctions,
– 15 percentage points for Sevsanc.
• Similar results are obtained with a linear
probability specification (higher actually for
violence around 28 percentage points).
• Critique: participating in education program is
just capturing the effect of “not having free time”
for prisoners to eventually get into conflicts
Hence no “reform of prisoners” is taking place by
educating them.
• Effect found is simply a way of decreasing
probability of conflict by decreasing time
available for leisure (busy effect).
– We do control in all specifications “time spent on
working” among inmates
– Robustness check: same effect on all specification.
Conclusions and future steps
• Education Participation seems to reduce bad
behavior and/or conflict within prisons.
• Information of 2005-2006 census allows to form
a panel at individual and prison level.