Contracts versus Trust in Water Allocation:

1
Contracts versus Trust in Water
Allocation growing sharing the pie in
Northeast Brazil Alexander Pfaff
(alex.pfaff_at_duke.edu) Duke University -- Public
Policy, Economics, and Environment Maria
Alejandra Vélez (mav_at_adm.uniandes.edu.co) Universi
dad de los Andes -- Facultad de Administración
Funded by CRED, Columbia University (NSF-funded
DMUU center)
2

• We explore the efficiency and equity resulting
  from allocation of a fixed resource within a
  bargaining institution.
• We wish to compare enforceable contracts with
  settings where responders must trust.

3

• We use generically framed field experiments
  (artefactual) based on water allocation in NE
  Brazil.
• 570 participants in Ceará, in Fortaleza or the
  Jaguaribe (largest agricultural) Valley.

4
State of Ceará in Northeast (Nordeste)
(popul gt 7m) (rural gt 2m) 75 below poverty line
Jaguaribe Valley
cover box
Brazil
5
City Tourism and Industry
These areas are soon to be further connected by a
large canal to bring water towards Fortaleza.
Its completion signals the arrival of new
allocation decisions for which no decision
process or body has been announced.
Rural area Mainly rice production
6
The Experimental Design

• Games are neutrally framed. Thus, we do not
  mention water but just tokens and describe
  actors not as locations, valley and city, but as
  players A (proposers) and B (responders).
• Each set of paired participants had to allocate
  10 tokens in a one shot experiment.
• Proposers are more productive than responders
• 21 productivity ratio case each chip was worth
  R2 for proposers but 1R for responders (low
  productivity case).
• 41 productivity ratio case R4 for proposers
  but R1 for responders (high productivity case)

7
The Experimental Design

• Our experiments have three steps
• 1 proposers set the agenda by requesting an
  amount of the resource
• 2 responders accept or reject, the latter
  giving both a small default payment of R 5
• 3 if responders accept, then proposers must
  decide whether to send back some of their
  earnings.
• For 21, the maximum possible total earnings
  is R20 for 41, it is R40.

8
The Experimental Design

• No Communication provides no signal to
  responders of a proposers intention of sharing
  in the 3rd step, so the responder decides based
  on the initially requested split of the resources
• Message provides with the initial proposed
  split a non-binding written message stating what
  transfer of surplus will occur in the 3rd step
  if, in the 2nd step, the responder accepts the
  request
• Contract provides with the initial proposal a
  binding written message about that transfer.

9
The Experimental Design

• No Communication and Message allows for Trust.
• Proposers, essentially invite responders to trust
  by asking for a larger share of the resource.
• If accept, responders give up a sure gain, the
  R5, for the chance of gaining more yet also a
  risk of gaining less
• Our design is different from the classical trust
  or investment game (Berg et al. 1995 or Guth et
  al. 1997), where the first mover must trust the
  second mover for trusts gains to arise.
• Trust is productive, as the largest total pie is
  created by initially sending everything to
  proposers.
• For instance, in Ceará, output rises as water is
  shifted from flooded rice to tourism and industry.

10
RESULTS
11
RESULTS

• In our one-shot game, as previously reported in
  the literature, we find evidence of trusting
  behavior, by responders and expectations of trust
  by the proposers.
• Trust exists and it pays off in the sense of
  raising not only efficiency, i.e. total earnings,
  but also the earnings of the responder.
• However, we also find that contracts do even
  better than trust in terms of efficiency and
  equity.

12
RESULTS No Communication

• We see proposer asking for 66 across
  productivity ratios in No Communication.
• Thus, proposers to some extent expect trust
  (leaving initial a allocation for responders less
  than the default).
• They are right, acceptance is high for the high
  productivity (86) despite higher requests for
  resources, though lower for the low productivity
  case (70).
• In accepted cases, the Average Ask For in No
  Communication is 59 for low productivity case
  and 61 for high productivity which yields less
  than R5 for responders

13
RESULTS No Communication

• Trust is present on average because the
  responders accept on average an initial split of
  resources that gives them less than the default
  payment they would receive if they reject the
  initial request.
• The reason to do so is trusting that the expected
  3rd-step transfer will be sufficient for a gain
  greater than the default theyll get with
  rejection.

14
RESULTS No Communication

• Trust pays off
• Transfers in the 3rd, sharing step are not high
  (R4.8 in high productivity , R 2.5 in low
  productivity ) but added to the initial resource
  split they beat the default payment (R5) a
  rejection would earn

15
RESULTS Message

• With a non-binding Message (coordination
  device) raises further the initial allocation to
  the proposer and thus efficiency.
• The average proposer request rises to 70 for low
  productivity and 69 for high productivity with
  accompanying messages about 3rd-step sharing
  averaging R4 for low productivity and R8.3 for
  high productivity.
• Proposers expect trust and they are right.
• Acceptance increased to 88 for low productivity
  and while lower for high productivity , at 73,
  the efficiency is still greater on average with
  the Message.

16

• Again trust pays off

• But is not fully justified
• Proposers are lying!

17
RESULTS Message

• 3rd-step actual transfers are lower than the
  non-binding messages.

18
RESULTS Contract

• Both efficiency and equity rise with
  enforceability..
• Proposed initial resource splits rise to 83 in
  low productivity and 78 in high productivity and
  the modal proposed split in both 21 and 41 is a
  striking 100.
• Acceptance for low productivity is down to 68,
  likely due to binding transfers being low,
  while acceptance for high productivity is 80.
• Transfers are higher too, with an average of R6
  for accepted offers for low productivity ,and
  R10 for high productivity.

19
Efficiency Results

• For 21, the maximum possible total earnings is
  R20 for 41, it is R40.
• Contract earnings ranked first in efficiency
  (Earnings Sum), closer to the possible maximum

20
Equity Results

• Contract does better for equity as measured in
  three key ways
• 1. The ratio of proposers to responders
  earnings.
• 2. Aiding the lowest earner
• 3. The frequency of earnings being allocated as
  if by a 50-50 distribution rule.

21
Equity Results
22

• Contract had also the greater percentage of
  accepted observations where the total earnings
  were distributed evenly
• - In Contact 37 in low productivity and 27
  for high productivity .
• - In No Communication the percentage of
  accepted observations where earnings were
  distributed evenly was 26 for low productivity
  and 13 for high productivity
• - In Message the percentage of even sharing
  was 11 for both productivities

23
To Conclude

• Trust exists and is socially productive
• However, both efficiency and equity rise with
  enforceability.
• This could provide insights into the design of
  any new institution to allocate water in Ceara
  but also other relevant settings.
• Our reports of trust obtained in a one shot game
  could be considered as the minimal trust.
• We expect trust to increase with a repeated game
  since punishment, reputation and learning could
  help to develop more trust.
• Next experiments, will do repeated to allow for
  learning and reputation.

24
To Conclude

• We presented the results of a neutrally framed
  game.
• Thus, we do not mention water but just tokens
  and describe actors not as locations, valley and
  city, but as players A (proposers) and B
  (responders).
• We plan do water framed experiments where valley
  and city actors actually interact

25
To Conclude

• For this preliminary draft, we focus on only the
  experimental results.
• However participants answered a survey with
  socio-demographic questions, a open ended
  question asking to explain their behavior in the
  game, the General Social Survey (GSS) with
  trust-related questions and, to learn their
  attitudes to risk, a risky-choice task
• The risky choice task was the following A the
  end of the game, each participant had to choose
  between option A which gave 10R for sure or
  option B which implied a lottery with 10 chance
  of R0, 20 chance of R5, 40 chance of R10,
  20 chance of R15 and 10 chance of R20 (same
  expected value than option A).

26
(No Transcript)
27
(No Transcript)