Caste and Punishment The Historical Legacy of Social Inequality for Norm Enforcement

1
Caste and PunishmentThe Historical Legacy of
Social Inequality for Norm Enforcement

• Karla Hoff (World Bank and Princeton University)
• Mayuresh Kshetramade (Affinnova Inc, Waltham,
  MA)
• Ernst Fehr (University of Zurich)
• October 2008

2
Motivation

• Well-functioning societies impose restraints on
  individuals opportunism in order to
• facilitate market exchange
• enforce basic social norms
• Legal rules and their enforcement provide such
  restraints
• Yet, legal enforcement is weak if not backed by
  informal enforcement mechanisms
• Informal enforcement mechanisms are based , e.g..
  on individuals willingness to punish
  transgressions of social norms

3
The Question

• How does an extreme social hierarchy the Indian
  caste system affect individual caste members
  willingness to punish violations of a cooperation
  norm?
• Punishment may be driven by opportunities or by
  preferences
• We conduct an experiment in which we keep
  opportunities constant so that we can infer
  preferences

4
Why is this question interesting?

• Dominance of one caste over the other could
  generally undermine the lower castes willingness
  to punish defection
• Reason
• Habituation to unfairness (fatalism)
• Lack of self esteem or esteem for the group
• Result
• Lower ability for collective action and public
  goods provision among the lower caste
• Perpetuation of social hierarchy even if legal
  obstacles are removed

5
Examples of shaming of low caste (dalits)

• the social experience of the low caste
  continues to be dictated by their caste status.
• More than 50 of villages deny low caste entry
  into non-low caste houses and into temples
• Almost 50 of villages deny low caste access to
  some water facilities, ban marriage processions
  of low caste, and prohibit their selling milk to
  a cooperative
• 25 of villages have separate drinking in water
  in schools for low caste
• Shah et al. Untouchability in Rural India (recent
  survey of over 500 villages)
• Their dependence on the better-off castes made it
  historically almost impossible for members of the
  low caste to engage in successful individual or
  collective actions against higher caste persons.
  Even today it is difficult.

6
Recent examples of retaliatory attacks against
dalits who tried to defy the traditional social
order

• When a dalit refused to sell bidis (hand
  rolled cigarettes) on credit to the nephew of an
  upper caste village chief, the upper caste family
  retaliated by forcibly piercing his nostril,
  drawing a string through his nose, parading him
  around the village, and tying him to a cattle
  post.
• When a dalit argued with an upper caste farmer,
  the upper caste villagers attacked 80 dalit
  families in retaliation. When the same dalit man
  then went to the police to report the incident, a
  social boycott was imposed on all of the dalits
  from his village they were thrown out of their
  village and denied every opportunity to earn
  their livelihood.
• Cited in NYU Center for Human Rights and
  Global Justice and Human Rights Watch (hereafter
  CHRJG), 2007, p. 60). The incident took place
  in June 2006 in Andhra Pradesh, India
• How might this environment shape preferencesi.e.
  choices in one-shot games-- in which there can
  be no retaliation?

7
Is the low caste in general conditioned to accept
injustice?

• Dominance of one caste over the other could
  undermine the low castes willingness to punish
  high caste members
• Fear associated with punishing high caste members
  because historically this was associated with
  fierce retaliation
• High caste members face fewer constraints on
  cheating low caste individuals
• (May inhibit trade and cooperation across castes)

8
Or, is the low caste conditioned to feel blind
hostility towards high castes?

• Dominance of one caste over the other may lead to
  hostility between the castes
• Could generate a generally high willingness to
  punish across classes regardless of whether a
  norm has been violated
• Dysfunctional punishment across castes (blind
  hostility)
• The caste system puts an extreme emphasis on the
  notion of superiority
• Could generate a desire to be superior to others,
  and one way of meeting this desire is to maximize
  income and status differences even at a cost to
  oneself (spite)
• See our paper on Spite and Development AER May
  2008

9
Caste system in Indiaan example of an extreme
social hierarchy

• General Castes (H), characterized by high
  ritual status and historically high power
• Scheduled Castes (L), formerly untouchables.
• Historically barred from courts, basic legal
  rights, and any but menial occupations
• Barred from temples, post offices, schools,
  village wells used by high caste people, eating
  houses, and most public gathering places.
• Untouchables were required to live outside the
  village proper and might enter its streets only
  for specific reasons, principally to work.

10
The Location

• District of Unnao in Uttar Pradesh

11
Subcaste (and thus caste) membership is exogenous

• Membership is fixed by descent, i.e., people are
  born into subcastes
• Low caste individuals (untouchables) cant rise
  in caste status
• High caste individuals may fall in status if they
  marry a lower caste individual
• However, marriage of non-low caste person to a
  low caste person is very rare
• And those rare occurrences may lead the upper
  caste to engage in public lynching of couples or
  their relatives, murder (of the bride, groom, or
  their relatives) ,rape, public beatings, and
  other sanctions NYU/Human Rights Watch,2007
• We concentrate on castes with the highest and the
  lowest status
• Causal impact of caste status on individual
  behavior can be examined

12
Subjects subcastes
Subjects are drawn from 2 low subcastes and 2
high subcastes
Note Divide in historical rights is at the
level of caste. Endogamous community occurs at
the level of subcaste
13
3rd Party Punishment Game with 3 stages Stage 1
2 are a Social Dilemma Game
100, 100
Sends to A
B
Keeps all the money
Sends to B
0, 200
A
A B each start with 50 rupees. If A sends his
50 to B, it triples in value. Then B has 200.
Note 100 1-2-days unskilled agricultural
wages
Doesnt send
50, 50
14
Stage 3 of game is third party punishment
C chooses how much to punish B
Sends to A
B
Keeps all the money
Sends to B
C chooses how much to punish B
A
Doesnt send
50, 50
We use strategy method to elicit Cs decision
when B cooperates and when he defects.
15
Why is this game interesting?

• The interaction between A and B represents an
  implicit exchange.
• Player A has a good that Player B values more
  thats the triplication of the money.
• But B can choose whether or not to pay A
• So this experiment mimics an exchange problem.
• A disinterested third party can sanction B for
  whatever he does. For each 2-rupee coin C spends
  on ghata imposition of a loss, B loses one
  10-rupee note. (We never use the world punish.)

16
Treatment conditions

• Focus is on the effect of players caste on norm
  enforcement
• 4 experimental treatments vary whether the
  players in a triple are H or L
• Problem Belonging to a subcaste means both
• belonging to a community
• being assigned a high or low social status
• How to avoid a confound in mixed triples between
• Effects of status differences across high and low
  castes
• Effects of ingroup/outgroup distinction across
  communities

HHH, HLH, LLL, LHL
17
Solution
Let subcaste of A subcaste of C ? subcaste of B
outgroup B (potential norm violator)
ingroup A (trustor)
ingroup C (punisher)
HHH, eg.
Thakur
Brahmin
Brahmin
HLH, eg.
Chamar
Brahmin
Brahmin
LLL, eg.
Pasi
Chamar
Chamar
LHL, eg.
Brahmin
Chamar
Chamar
18
How did we convey information about subcaste?

• The 3 players who interact in the game are not
  told the villages in which their partners live,
  but are told the names of their partners
• In the course of running the pilot, we learned
  the names of individuals in the villages, and in
  a pre-experiment, we established that nearly
  everyone could correctly identify the caste of
  individuals from their names
• So names are a natural way to indicate subcaste.

19
Subjects

• 205 triples drawn from low (L) and high (H)
  castes in 3 subdistricts of Uttar Pradesh, Unnao
  District
• 94 of L and 98 of H have always lived in their
  village
• Average age is 35
• C players (the punishers) were drawn from more
  than 40 villages. At most 6 subjects for a
  single treatment are from a single village.
• Group sessions (5 or 6 subjects) of 20 minutes,
  followed by individual sessions of 15 minutes
  each
• In the game, we never use the word caste or
  subcaste until the end of the individual
  session, when we ask the subject his subcaste in
  order to verify that we have not made a mistake
  about his subcaste membership

20
Measures of punishment

• Punishment for defection
• Punishment for cooperation
• We will focus on
• Results are similar if we use just Pun. for
  defection

Pun. for defection Pun. for cooperation
21
Hypothesis 1 Caste dominance/submission

• A high caste subject considers the same norm
  violation (defection) as more blameworthy if it
  is committed by a low caste subject
• L-violator upsets the established order
• HLH gt HHH
• Low caste subjects learned that being cheated
  by high caste subject is part of the natural
  order. H violator just asserts the traditional
  hierarchy of rights and privileges
• Ls dont demand fairness from high caste members
• They learned that punishing a high caste member
  can be very dangerous
• LHL lt LLL

22
Hypothesis 2 Blind hostility between high
and low castes

• Conflict between status groups may lead to
  hostility and resentment
• Many high castes are eager to preserve whatever
  they can of the distinctions and prejudices
  underlying the caste hierarchy, (Dreze and
  Sharma, 1995)
• whereas dalit political leaders are trying to
  overturn the hierarchy
• In a climate of hostility, a defection by a
  member of a different status group may be
  considered as particularly blameworthy

23
Blind hostility hypothesis, cont.

• Higher punishment in mixed treatments compared
  to single status group treatments.
• HLH gt HHH LHL gt LLL
• Note HLH gt HHH is the same as in caste
  dominance/submission hypothesis
• But LHL gt LLL differs from caste
  dominance/submission hypothesis

24
Hypothesis 3 Caste culture

• Dominance of one caste over the other could
  generally undermine the low castes willingness
  to punish defection
• Habituation to unfairness (fatalism)
• The world has always been and will always be
  unfair. One can do nothing about it.
• Low status groups may hold their own communities
  in low esteem, which psychological studies
  indicate undermines in-group affiliation
• Low status individuals may not be proud to be a
  member of a low status group. They may rather
  like to escape their low status.
• Why contribute to the benefit of the group if the
  group is not worth it?

25
Caste culture hypothesis

• HHH ? HLH gt LLL ? LHL

26
Does the caste submission/dominance or blind
hostility hypothesis hold?
27
Does the evidence support the caste culture
hypothesis ?
Mean relative
Mean absolute
punishment
punishment
p0.03
8
7.63
5.76
6
p0.01
rupee coins
4.40
4
-
Two
2.33
2
0
(N123)
(N82)
(N123)
(N82)
Punisher is high caste (HHH HLH)
Punisher is low caste (LLL LHL)
28
Does the same cooperation norm hold across all
treatments?

• Differences in punishment could arise either...
• Because in HLH HHH, the obligation to
  reciprocate cooperation is stronger than in LHL
  LLL, or
• Because low castes have a genuinely lower
  preference to punish defection
• Player Bs expectation of punishment is an
  indicator of the perceived obligation to
  cooperate
• In all treatments,
• between 80 and 90 of Bs expect punishment for
  defection
• between 5 and 10 of Bs expect punishment for
  cooperation
• So the perceived obligation to punish is
  identical across treatments

29
Expected punishment across treatments
Cooperation is perceived as the normatively
correct action in all treatments Caste
differences in punishment reflect a lower
preference to punish norm violations
30
Do income or wealth differences affect the
preference to punish?

• We have three variables that are strong
  predictors of real per capita (adult equivalent)
  consumption
• Lives in a mud/brick house or a pure brick house
• As soon as people have enough resources for food
  they want to live in a brick house
• Completion of high school (completion is
  necessary to be able to read)
• Land ownership

31
Predictors of real per capita consumption

• World bank living standard survey on southern and
  eastern Uttar Pradesh
• Living in a brick house is associated with a 35
  increase in real per capita consumption (relative
  to living in a mud house)
• Completion of high school raises consumption by
  24
• Elasticity of consumption with regard to land is
  between 0.12 and 0.15 (2 instead of 1 bigha of
  land implies 12-15 higher consumption)

32
Considerable overlap across castes in predictors
of income
33
Scatter plot illustrates the overlap
34
Do richer individuals punish more?(house
ownership)
Same results for absolute punishment of defection
35
Do richer individuals punish more?(land
ownership)
Same results for absolute punishment of defection
36
Do richer individuals punish more?(education)
37
Do income differences explain the punishment
pattern?(predicted consumption based on World
bank coefficients)
If anything, punishment levels are lower by
richer than poorer individuals . (Consumption
is predicted from LSMS by land owned, house
quality, caste, education)
38
Tobit estimates of caste status effect on
punishment
39
Caste or subcaste?

• Recall There are many subcastes with high status
  and many subcastes with very low status
  (untouchables)
• If caste status drives punishment, then we should
  observe
• little variation between the high subcastes (at
  apex of caste hierarchy) and between the low
  subcastes, but
• much variation between those groups at the apex
  and the outcaste groups.

40
Is caste culture hypothesis supported at the
level of the individual subcastes?
No difference in pun for defection pun for
cooperation in HHH between Brahmin and Thakur
punishers.
..
41
or between Chamar and Pasi punishers
..
42
Thus the key divide across subcastes with respect
to punishment is the H-L divide
43
What is the psychological mechanism behind this
divide?

• Earlier studies have shown that in-group
  affiliation affects punishment in two ways
  (Bernhard et al. and Goette et al. 2006)
• If the punisher and the victim belong to the same
  in-group, punishment tends to be higher
• If the punisher and the norm violator belong to
  the same in-group, punishment tends to be lower
• Do H- and L-punishers show the same concern for
  their in-group members?

44
Recall, AC condition held up to now.
C
B
A
HHH, eg.
Thakur
Brahmin
Brahmin
LLL, eg.
Pasi
Chamar
Chamar
45
Next consider BC condition
A (trustor)
B (potential norm violator)
C (punisher)
HHH, eg.
Thakur
Brahmin
Brahmin
LLL, eg.
Pasi
Chamar
Chamar
46
Prediction

• If H and L punishers show similar concern for
  in-group members, then
• Pun in AC gt Pun in BC,
• This gap will be the same for HHH and LLL

47
Only H shows in-group affiliation

48
A more detailed look H shows much more
concern for in-group members
49
But L makes no distinction
50
Conclusion (1)

• How does an extreme social hierarchy the Indian
  caste system affect individual caste members
  willingness to punish violations of a cooperation
  norm?
• We examined the impact of low and high caste
  status on punishment while controlling for
  ingroup/outgroup, wealth, income and education
  effects
• Caste dominance/submission hypothesis refuted
• Blind hostility hypothesis refuted

51
Conclusions (2)

• Caste culture hypothesis is supported
• High caste has a generally higher willingness to
  punish regardless of whether the norm violator is
  high or low caste
• High caste punishment pattern is consistent with
  the view that Hs care more for in-group members
• Low caste members seem to care less for their
  ingroup members
• The lower willingness of L to punish norm
  violations that hurt members of their own
  community is consistent with the view that a
  history of social exclusion from legal rights and
  limited access to market exchange undermines
  preferences for norm enforcement

52
Conclusion (3)

• Lower preference for norm enforcement among low
  caste members could contribute to
• a lower ability for collective action and public
  goods provision among the lower caste
• a perpetuation of social hierarchy even if legal
  obstacles are removed