Trust and cooperation: Social capital and community resource management

1
Trust and cooperation Social capital and
community resource management

• Jetske Bouma, Erwin Bulte and Daan van Soest
• Department of Economics, Tilburg University

2
Introduction

• Social capital (SC)
• Thought to be an important determinant of
  economic performance
• Trust, norms, and networks that enable collective
  action (Fukuyama 1995, Putnam 1993, Bowles and
  Gintis EJ 2002)
• Lower transaction costs, better exchange of
  information, internalization of externalities
• Empirical evidence?

3
Introduction (2)

• Standard empirical approach
• Key element of SC is trust
• Trust the expectation that arises within a
  community of regular, honest and cooperative
  behavior based on commonly shared norms of the
  part of members of that society.
• Social capital measured via surveys (do you
  think other people can be trusted?).
• Results link between SC and economic growth not
  robust
• Positive relationship Putnam (1993), Knack and
  Keefer (QJE 1997)
• No relationship Miguel et al. (REStat 2005),
  Beugelsdijk and van Schaik (EJPE 2005)

4
Introduction (3)

• Potential issues/problems of the standard survey
  approach
• Methodological
• Actual versus stated preferences? (cf. CVM,
  Cummings et al. JPE 1997)
• Do we measure trust or trustworthiness? (Glaeser
  et al. QJE 2000)
• Practical
• Substitutability between SC and formal
  institutions?
• Spatial scale matters? Country vs village?
• Our objective
• measure SC using revealed preferences, and use it
  as an input in an analysis of actual behavior
  provision of a (semi-) public good Soil and
  Water Conservation in rural India.
• Literature
• Measure/explain trust Barr (EJ 2003), Carpenter
  et al. (2004, UrbStud 2004), and others.
• Comparison to real world behavior Cramb
  (AusJARE), Karlan (AER 2005)

5
Introduction (4)

• This paper
• Investments in soil and water conservation in
  rural India
• Try to explain key factors affecting social
  capital
• SC measured by trust game i.e., what factors are
  conducive to trust?
• Try to explain real contributions to resource
  conservation (soil and water conservation) using
  trust game results
• Individual play
• Play by community as a whole
• Policy relevance, scientific relevance

6
Introduction (5)

• Policy relevance
• Prioritization of poverty alleviation programs or
  natural resource conservation projects
• Urgency (state of depletion, extent of poverty)
• Expected effectiveness of the intervention
• Key considerations
• Public good/common pool characteristics
• Projects more likely to be successful if
  implemented in communities with more SC
• Can we predict success or failure using a simple
  game (i.c., a trust game)?
• World Banks budget for community based
  approaches is 7bln
• And how does it compare to just targeting
  homogenous communities (Kerr WD 2002)?

7
Introduction (6)

• Scientific relevance
• Literature on other-regarding preferences
• Evidence in laboratories and field suggests
  people care about distribution, intentions of
  others, etc.
• Other regarding preferences are shaped by social
  interaction and, possibly, the environment
• Henrich et al. (2004) experimental play mirrors
  patterns of interaction found in everyday life.
• But not that straightforward

8
Introduction (7)

• External validity
• simple games (e.g., gift exchange) predict
  behavior well in non-market situations (e.g.,
  contributions to charity see Falk 2004)
• but not in market situations (trade in sports
  memorabilia List JPE forthcoming)
• Extreme types of goods public versus private
  goods
• What if there are both private and public
  benefits? Soil and water conservation investments!

9
Introduction (8)

• External validity tests in this paper
• Direct external validity
• Is participants behavior in the Trust Game
  predictive of his/her behavior in the soil and
  water conservation project?
• Indirect does Trust game measure social capital?
• Are participants more likely to invest in soil
  and water conservation if his/her village acts
  more cooperatively in the Trust Game?
• And do we find similar results when using an
  alternative indicator of SC (caste homogeneity,
  directly observable)?

10
Introduction (9)

• Setup of the rest of the presentation
• The experiment
• Results of the trust game
• 3) Determinants of trust
• 4) Social capital and natural resource management
  in reality
• 5) Conclusions

11
1. The experiment

• Trust game
• Two anonymous participants investor i and
  trustee j.
• Investor i receives endowment E,
• She invests an amount 0 ? Si ? E
• Trustee j receives Rj 3Si (experimenter triples
  Si),
• And returns amount 0 ? Sj ? Rj (i.e., share Rj/
  Sj)
• Maximize aggregate payoffs invest Si E (and
  split?)
• But investor only invests if she expects to
  receive back more than she invested...
• Game theory versus behavioral economics!

12
The experiment (2)

• What is it that Trust game measures?
• Trust?
• Altruism (Cox GEB 2004)?
• Risk preferences (e.g., Karlan AER 2005,
  Schechter JEBO 2006)?
• For measuring SC should not be contaminated by
  risk preferences
• Evidence inconclusive Karlan ( AER 2005) and
  Schechter (JEBO 2006) vs Ashraf et al. (2004) and
  Eckel and Wilson (JEBO 2004)
• Our study
• No reason for systematic variation
• Possibility to control for risk attitudes (via
  income etc.).

13
The experiment (3)

• Instructions similar to those of Barr (EJ 2003),
  but added
• (i) double-blind play, and
• (ii) double roles...
• Implications
• More observations
• Additional control variable (share returned by
  participant)
• Bias investor behavior unaffected, but trustee
  behavior is (lower share returned, Burks et al.
  JEBO 2000)
• Focus on investor behavior (trust)!

14
The experiment (4)

• Research area
• Five villages, three different watersheds

15
(No Transcript)
16
The experiment (6)

• Subject pool
• 2003 random sample of 20 households per village
  
• 2004 invited to participate in the Trust game
• Main source of income agriculture
• 33 illiterate
• Variation in caste heterogeneity, average income,
  income inequality, plot size, aridity, soil
  conditions

17
The experiment (7)

• Implementation
• Instructions in local language written but also
  read aloud by the instructor
• Test questions
• Random selection of anonymous partners (draw
  number from box)
• Transactions with coded envelopes (double blind)
• Showup fee, bonus for no communication during the
  experiment
• E Rs 50 (US1.15 one days wage)
• Main problem 6 hours, instructions and game
  separated loss of control
• Cheap talk
• Controlled for by means of dummy (2 villages).

18
Instructions
19
Participants in Kadampally
20
Meanwhile, outside
21
2. Trust game results

Double role

Single role



This study





Subjects




Indian rural






households
Number of playing pairs






92
Initial endowment





US 1.15
Relative value initial
Full days
endowment






wage
Proportion of investors





0.13
investing zero

Mean investment by





0.49
investor
(
S
)

i
Mean return on investment





0.87
for investor
i

(
R
/S
)

j
i

22
2. Trust game results

Double role

Single role



This study
Burks et al.





(2000)
Subjects




Indian rural
US students






households
Number of playing pairs






92
46
Initial endowment





US 1.15
US 10.00
Relative value initial
Full days
Two hours
endowment






wage
wage
Proportion of investors





0.13
0.06
investing zero

Mean investment by





0.49
0.62
investor
(
S
)

i
Mean return on investment





0.87
0.78
for investor
i

(
R
/S
)

j
i

23
2. Trust game results

Double role

Single role



This study
Burks et al.
Berg et al.
Burks et al.
Barr





(2000)
(1995)
(2000)
(2003)
Subjects




Indian rural
US students
US students
US students
Zimbabwean






households
farmers
Number of playing pairs






92
46
32
22
141
Initial endowment





US 1.15
US 10.00
US 10.00
US 10.00
US 0.80
Relative value initial
Full days
Two hours
Two hours
Two hours
Half days
endowment






wage
wage
wage
wage
wage
Proportion of investors





0.13
0.06
0.06
0.14
0.09
investing zero

Mean investment by





0.49
0.62
0.52
0.65
0.43
investor
(
S
)

i
Mean return on investment





0.87
0.78
0.89
1.31
1.28
for investor
i

(
R
/S
)

j
i

24
Investment behavior across villages
Group discussions Kadampally and Bicchawara
(not significantly different from Karji)
25
Trustee behavior across villages Share returned lt
1/3 self-interested SR 1/3 reciprocal SR
gt1/3 altruistic
26
3. Determinants of SC

• Social capital measured by amount invested by
  investor (trust). Interval regression.
• Sik constant ?1Xik ?2Gik ?3Ck ?ik
• X vector of investor characteristics
• G game information (average trustworthiness,
  or own behavior as a trustee)
• C village / context variables
• Potential endogeneity problems
• Income, agricultural dependency (non-farm income)
• IV regressions (instruments literacy, age,
  watershed dummy, soil).

27
OLS
IV
OLS
Determinants of trust/SC
Constant
Watershed dummy

Average income village
Average share ret.
Share returned by participant
Pre-game discussion

Majority caste

Per capita income (/pred)
Per capita landholding
Access to irrigation
Agricultural dependency (/pred)
Female
Literate
Land fertility
Age
No. of observations
Pseudo-R2
28
Determinants of SC (3)

• Determinants of investments
• Village characteristics
• Village income negative effect
• Average share returned does not matter
• Group discussion does not matter cheap talk
• Individual characteristics
• Introspection matters
• Dependency on cooperation of others matters
  (irrigation)
• Belonging to a majority group positive effect
• Income and agric dependency do not really matter
• More land less willing to contribute

29
4. Social capital and resource management

• Two proxies of SC
• Trust average amount sent
• Social homogeneity (directly observable of
  households belonging to majority caste)
• Two dependent variables community development
  activities, data obtained in large household
  survey (2003)
• Soil and water conservation (subsidized SC may
  not matter)
• Operations and maintenance ( not subsidized SC
  is expected to matter)

30
Social Capital and Resource Management (2)

• The equation that is estimated (Probit)
• Zik ?0 ?1SCk ?2Xik ?3NGOk ?4Ck ?ik
• Zik SW or OM of individual i in village k.
• SCk social capital indicator (trust, SH),
  village k
• Xik vector of individual/household
  characteristics
• (caste, income, education, land holding,
  irrigation, etc.)
• NGOk SWC implemented by NGO? (dummy)
• Ck vector of context variables
• Again IV estimates to deal with endogeneity of
  variables (income, land, amount sent by
  participant in Trust Game
• First Consider the results for the subsidized
  activity (SW)

31
(No Transcript)
32
Social Capital and Resource Management (4)

• Results for investments in SWC
• Social capital, however measured, is not a
  significant determinant of household investment
  in SW when such activities are subsidized.
  Private decision (CBA).
• Income matters () compete for attractive jobs?
• Number of household members matters.
• Next turn to the unsubsidized activity
  (operations and maintenance)

33
Probit-IV
Probit-IV
Probit
Investment in OM
Constant
Watershed dummy
Average income village
0
Amount sent by prtcp in TG (/pred)
Average amount sent in TG
Village homogeneity
Majority caste

Maintenance fund
Per capita income
Per capita landholding
Access to irrigation
Agricultural dependency
Land fertility
Number of household members
No. of observations
Pseudo-R2
34
Social Capital and Resource Management (6)

• Results for OM
• Social capital based on trust has a significant
  and positive impact on the provision of the
  public good (significant at 1 level).
• Individual play not generalizable perhaps CPR
  resembles a market more closely than a charity
  context (List / Falk)
• Village homogeneity is a decent proxy for trust
• Average and own income (),
• Member of majority caste ()
• NGO (-) because of maintenance fund crowding
  out?
• Influence decision-making matters ()

35
5. Conclusions

• Focus on how the determinants of social capital
  and how SC affects provision of public good with
  private good characteristics
• Determinants of social capital
• Trust difficult to explain (low R2), but
  dependency on others (irrigation) and majority
  membership are important determinants
• External validity
• Trust game behavior is poor predictor of
  individual contributions to SWC
• But higher contributions at the village level
  seem to reflect SC (as evidenced by the results
  on village homogeneity)