Neighborhood Effects and Social: The case of Irrigated and Rainfed Rice Farners in Bohol

7/31/2019 Neighborhood Effects and Social: The case of Irrigated and Rainfed Rice Farners in Bohol

1/34

Neighborhood Effects and Social Behavior :

The Case of Irrigated and Rainfed Rice Farmers in Bohol

Takuji W. Tsusaka

~Emergence of Social Norms and Community Mechanisms~

Accepted by 2012AAEA (Agricultural and

Applied Economics Association of USA)

Valerien O. Pede

Keitaro Aoyagi

Thanks:

JICA

Lolit Garcia

Elmer Suaz

Edmund Mendez

Evangeline Austria

Ma. Indira JoseNeale Paguirigan


2/34

Experimental methods, spatial econometrics, and survey methods are nicely

blended in a single study.

The study provides innovative approaches to investigate the spillover effect ofsocial behaviors.

Excellent job!!

According to the reviewers of the conference,

1


3/34

Social Behavior

Social Interaction Mutual Trust

Social Behavior Social Relationship

Often used as synonyms and are considered to be forms ofsocial capital,

2

Empirical and theoretical supports:

Grootaert & van-Bastelaer (2008), Krishna (2007), Chou (2006), Isham et al. (2002),

Zack & Knack (2001), Knack & Keefer (1997), etc. etc.

y m y y n nn n nn ,

Plays a crucial role in rural development and poverty reduction.

Mechanism to induce effective collective actions and reduce

transaction costs in a variety of ways (Hayami 2009, Arrow 1999,

Solow 1999).


4/34

Neighborhood Effect and Social Behavior

(Objective)Our View: Since social behavior takes place between people, it is natural to suppose that such

behavior may affect other peoples behavior; i.e., there must be spillovers and externalities of

social behavior.

3

Blume, Bandiera, Brock, Durlauf, Moffitt,

Raudenbush, Sampson, Cook, Manski etc.

Bandiera, Rasul, Conley, Udry, Anselin,

Griffith, etc.

Neighborhood Effects

using spatial econometrics..

Hayami, Grootaert, van-Bastelaer,

Krishna, Chou, Isham, Zack,

Knack, Arrow, Solow, Keefer, etc.

Social Behavior/Social Capital

Ionnides,Topa

using behavioral game experiments

Onesa, Putterman, Dufwenberg, Muren,

Cook, Cooper, Eckel, Wilson, Bohnet,

Zeckhauser, Ben-Ner, etc.


5/34

Neighborhood Effect

Yi = Social Behavior

(e.g. Trusting)

Yi

Xi = Individual Profile

(e.g. Age)

For example,

XjXj

YjYj

i = Residual

Endogenous Social Effector Spatial Lag Effect

Xi

Xj Xj

Yj Yj

Exogenous Social Effect or

Cross Effect or ContextualEffect

Correlated Social Effect or

Perturbation Effect

j

i

j

j


6/34

Spatial Econometric Approach

(1) Neighborhood Definition Weight Matrix

n observations

n

0

0 0

1

1

0

0

Wafter normalization

5

Spatial Lag Operator

WX =

n

X =

x1x

2x3

xn

x averaged over neighbors for obs 1

x averaged over neighbors for obs 2x averaged over neighbors for obs 3

x averaged over neighbors for obs n


7/34

Spatial Econometric Approach

(2) Model Specification1) a-Spatial ModelY= 1X+ 1

2) Cross Regression Model

Y= 2X+2WX+ 2

Lagrange Multiplier Tests

Morans I

Lag Test

Error Test

Spatial Diagnostics

6

a) Lag Model (w/Cross) ( Spatial Durbin Model)

b) Error Model (w/Cross)

c) ARAR Model (w/Cross)

Y= 3WY+ 3X+3WX+ 3

Y= 4X+4WX+ 4 , 4 = 4W4 + 4

Y= 5WY+ 5X+5WX+ 5, 5 = 5W5 + 5

Robust Error Test SARMA Test


8/34

Behavioral Game Experiment

General Instructions

All participants receive a pre-paid show-up fee of P50 at the beginning. The games deal with actual money. Whatever money participants win in games will be theirs to take home.

The amount they win depends on the decisions they make during the games.

The games must be taken seriously. They are not allowed to talk to other participants during the entire

event. If the fail to follow the rule the will not be iven the ost- aid show-u fee P100 at the end of all

Purpose: To elicit individuals social behavior quantitatively.

7

experiments.

Dictator Game

The game is played by pairs. Your anonymous partner is in a separate room. You never know who it is.

At the beginning, you (sender) get P100/Your partner (receiver) gets nothing.

Your partner will receive a transfer from you/You will not receive any money from your partner. The amount you keep is your payoff of this game/The amount your partner receives is his/her payoff of

this game.

How much do you transfer to your partner, if your partner is someone in your barangay?


9/34

B

?A (You)

?

Your Room (Senders) The other Room (Receivers)

Keep: P50Receive: P50

Keep: P100 Receive: P0

Send: P50

Send: P0

Dictator Game Example

C ?Keep: P0

Receive: P100

The amount you keep

is your payoff of this

game

The amount they

receive is their payoff

of this game

Send: P100

The transferred amount is recorded as

the game result, and is interpreted as a

measure of your altruistic behavior.


10/34

9


11/34

10


12/34


Public Goods Game (Two Rounds)

The game is played by groups of four people of which you are one. The three

anonymous partners are in a separate room. You never know who they are.

At the beginning, you (sender) and your partners (receivers) get P100 each.

1st Round

11

.

The total amount contributed will be doubled, and the doubled amount will be

shared equally among all members, regardless of your contribution.

The amount you keep plus the amount you receive back from the group is your

payoff of this game.

How much do you contribute to the group, if your partners are some ones in yourbarangay?


13/34

(50+50+50+50)

P 200

Contribute P 50 Contribute P 50 Contribute P 50 Contribute P 50

Keep P 50

You A B C

Keep P 50 Keep P 50 Keep P 50

The contributed amount is

recorded as the game result,

and is interpreted as aPublic Goods Game

Round 1, Example

P 400

Receive P 100 Receive P 100 Receive P 100Receive P 100

P 50

P 150

P 50

P 150

P 50

P 150

P 50

P 150Payoff Payoff Payoff Payoff

behavior to public goods.

12


14/34


Public Goods Game (Contd)

Therefore, the game has an aspect ofinvestment.

What if you increase/decrease your contribution, others contributions kept unchanged?

Case1 You Partner A Partner B Partner CContribution 50 50 50 50

Pa off 150 150 150 150

13


Payoff 165 135 135 135

Case3 You Partner A Partner B Partner C

Contribution 80 50 50 50Payoff 135 165 165 165


Payoff 180 180 180 180

Incentive for cheating (free-riding)

Importance of Collective Action


15/34


Public Goods Game (Contd)

You can secretly check the contribution from each partner by paying P1.

Then, you can send a message to particular partners to indicate that youare unhappy with that persons contribution, by paying P1 per message.

After the 1st Round

We will consider two variables in the analysis

14

The contributed amount is recorded as the result of the second round,

and is interpreted as a measure of your contributory behavior to public

goods in the presence ofmonitoring mechanism.

with the same partners as in the 1st round.Then, Play the 2nd Round

Free-riding Index (FRI) Group Members

Average Contribution

Your

Contribution

Check

Dummy

Indicator for degree of awareness of own free-riding


16/34

Our Study Site

San Miguel/Trinidad/Ubay, Bohol

Thanks to

15

Figure by Barkada Tours

Arnel & Nel !


17/34

Our Data

Primary Data

Agricultural and Socioeconomic Data (X)

4 crop seasons from 2009 to 2010 One half of the irrigated farmers (randomly selected): Volumetric pricing system

The other half: Area size based flat rate

Volumetric Pricing Dummy

16

Geographical Coordinates (W) Both farm plots and residences. Two types of neighborhood can be defined

Behavioral Game Results ( Y) 243 randomly selected farmers

Irrigated (N= 132) & Rain-fed (N= 111) The sample for this study


18/34

Data: Descriptive Statistics

Overall

(N=243)

Irrigated

Areas

(N=132)

Rainfed

Areas

(N=111)

t-test for mean

difference

[p-value]

Volumetric Pricing Dummy

0.561(0.498)

Age 51.062 49.689 52.694 3.004 *

(12.019) (12.248) (11.585) [0.052]

Gender Dummy 0.708 0.758 0.649 0.109 *

Agricultural & Socioeconomic Variables (X)

Sample Mean

(Standard Deviation)

17

(0.456) (0.430) (0.480) [0.063]

Years of Schooling 6.395 6.144 6.694 0.550

(3.0384) (2.922) (3.159) [0.160]

Ln Asset 10.578 10.444 10.738 0.295

(1.132) (1.193) (1.038) [0.718]

Field Size (ha) 1.585 1.167 1.754 0.586 ***

(1.058) (0.682) (1.228) [0.000]

Household Size (head count) 5.936 6.144 5.689 0.455

(2.302) (2.321) (2.265) [0.125]

Household Female Ratio 0.500 0.484 0.519 0.035 *

(0.162) (0.148) (0.176) [0.092]

Value for the last season

Averaged over the 4 seasonsStatistical Significance: *** 1 %, * 10%


19/34

Data: Descriptive Statistics

Overall

(N=243)

Irrigated

Areas

(N=132)

Rainfed

Areas

(N=111)

t-test for mean

difference

[p-value]

Dependent Variables (Y)

Dictator Game 30.041 32.197 27.477 4.719 *

(20.236) (21.555) (18.314) [0.070]

PGG Round 1 54.403 53.182 55.856 2.674

Behavioral Game Results (Y)& Control Variables

Sample Mean

(Standard Deviation)

18

(23.033) (22.080) (24.139) [0.368]

PGG Round 2 52.140 51.818 52.523 0.704

(24.350) (23.633) (25.279) [0.823]

Controls for PGG

Risk Preference 53.786 54.470 52.973 1.497

(25.898) (24.380) (27.686) [0.655]

PGG R1 Message Receipt Dummy 0.280 0.273 0.288 0.016(0.450) (0.447) (0.455) [0.789]

PGG R1 Free-riding Index -0.110 0.455 -0.781 1.235

(15.335) (14.746) (16.049) [0.533]

Statistical Significance: * 10%


20/34

Defining the Neighbors

Weight Matrices (W)

2 types of neighborhood

Residential Neighborhood

Plot Neighborhood

2 Areas

Irrigated

Rainfed = 4 Weight Matrices

How did we choose neighbors?? Threshold Distance

19

Too short Some people will have no neighbor

Those observations will be dropped.

Too long Boundary Bias Problem

(e.g. Griffith, 1983)

You

d

You

d

Impose the shortest possible threshold distance.(GeoDa does it)


21/34


Individually Calculated 4 Threshold Distances

0.959

1.302

0.956

1.376

(a) Plot Neighborhood/Irrigated Farmers

(b) Plot Neighborhood/Rainfed Farmers

(c) Residential Neighborhood/Irrigated Farmers

(d) Residential Neighborhood/Rainfed Farmers.

km

Use this as a uniform

threshold distance

Field Plot Neighborst-test for

Residential Neighborst-test forNeighborhood Structures;

20

mean

difference[p-value]

mean

difference[p-value]

IrrigatedAreas

RainfedAreas

IrrigatedAreas

RainfedAreas

Weight Code (a) (b) (c) (d)

Number of Observations 131 109 132 110

Total Number of Neighbor Relations 860 1166 866 1292

Nonzero Weights (%) 5.01 9.81 4.97 10.68Avg. Num. of Neighbors 6.565 10.697 4.132 6.561 11.746 5.185

(2.649) (4.309) [0.000] (3.119) (5.409) [0.000]

Avg. Distance bet. Neighbors (km) 0.603 0.587 0.016 0.583 0.574 0.009(0.236) (0.239) [0.293] (0.243) (0.252) [0.564]

Statistics@Threshold Distance = 0.956km


22/34


0

5

10

15

20

1 2 3 4 5 6 7 8 9 10111213141516171819202122

0

5

10

15

20

(a) Plot/Irrigated

Distribution of the Number of Neighbors(b) Residential/Irrigated

21

0

5

10

15

20

1 2 3 4 5 6 7 8 9 10111213141516171819202122

0

5

10

15

20

1 2 3 4 5 6 7 8 9 10111213141516171819202122

(c) Plot/Rainfed (d) Residential/Rainfed


23/34


Graphical Representations

(b) Plot/Rainfed (d) Residential/Rainfed

22

(a) Plot/Irrigated (c) Residential/Irrigated


24/34

Summarized Hypotheses

H1) Social behaviors of individual farmers are influenced by their neighbors social

behaviors and personal attributes.

H2) Neighborhood effects on social behaviors, particularly contribution to public

goods, are higher in the irrigated areas vis--vis in the rain-fed areas.

23

H3) In the irrigated areas, farmers are more contributory to public goods when they are

engaged in volumetric water pricing system than in area-based flat rate system.

H4) The endogenous social effects on public goods contribution are larger among farm

plot neighbors than among residential neighbors.


25/34

Results: Spatial Model Diagnostics

Game Experiment Dictator Game Public Goods Game, Round 1 Public Goods Game, Round 2

Neighborhood Plot Residential Plot Residential Plot Residential

Ecosystem Irrigated Rainfed Irrigated Rainfed Irrigated Rainfed Irrigated Rainfed Irrigated Rainfed Irrigated Rainfed

Moran's I 0.042 * -0.031 0.131 *** -0.126 -0.087 0.060 *** 0.004 -0.010 0.119 *** -0.016 0.162 *** 0.014 *

(0.050) (0.385) (0.001) (0.990) (0.849) (0.004) (0.246) (0.219) (0.000) (0.241) (0.000) (0.081)

Error Correlation 0.616 5.332 ** 1.511 5.001 ** 8.135 *** 0.083

(0.433) (0.021) (0.219) (0.025) (0.004) (0.773)

La Correlation 3.034 * 7.854 *** 1.831 10.961 *** 9.849 *** 0.974

Lagrange Multiplier Tests on Cross Regression Residuals for Spatial Dependence Identification

. . . . . .

(0.082) (0.005) (0.176) (0.001) (0.002) (0.324)

Error Correlation 12.977 *** 2.540 0.623 0.375 0.214 1.165

Robust (0.000) (0.111) (0.430) (0.540) (0.644) (0.281)

Lag Correlation 15.395 *** 5.062 ** 0.943 6.335 ** 1.928 2.0555

Robust (0.000) (0.024) (0.332) (0.012) (0.165) (0.152)

SARMA 16.011 *** 10.394 *** 2.453 11.336 *** 10.062 *** 2.138

(0.000) (0.006) (0.293) (0.003) (0.007) (0.343)

Spatial Model of

Our Choice

Lag

&

Cross

Cross

Lag

&

Cross

Cross Cross Cross Cross Cross

Lag

&

Cross

Cross Cross Cross

For RobustnessCheck

ARAR&

Cross

Lag&

Cross(Statistical Significance: *** 1 %, ** 5%, * 10%, 15% )


26/34

Results: Spatial Estimation

Dictator Game Neighborhood Plot Residential

Ecosystem Irrigated Rainfed Irrigated Rainfed

Spatial ModelLag &

Cross Cross

Lag &

Cross

ARAR &

Cross Cross

Endogenous Social Effect () 0.239*

0.352***

0.331

Correlated Social Effect () 0.034

Neighbors' Characteristics ()

Volumetric Pricing Dummy -13.630 * -11.492 -11.601

Age 0.123 0.060 -0.105 -0.128 0.445

Gender Dumm 4.382 -24.624 * 11.062 11.584 8.701

WYW

Years of Schooling -0.750 -4.222 * -1.288 -1.322 1.294

Ln Asset -0.887 10.944 * 5.075 5.230 -1.344

Field Area (ha) 16.206 *** 3.016 8.419 8.361 4.183

Household Size -2.513 -2.327 -1.876 -1.895 -2.070

Household Female Ratio -2.364 -2.837 21.705 23.196 -5.491

Own Characteristics ()

Volumetric Pricing Dummy -2.131 -0.327 -0.371

Age -0.201 -0.091 -0.263 * -0.266 * -0.121

Gender Dummy 2.914 3.098 3.526 3.605 5.631 Years of Schooling 0.610 0.282 0.221 0.213 0.334

Ln Asset -0.374 0.854 -0.308 -0.290 -0.347

Field Area (ha) -0.118 2.322 -0.956 -0.920 2.324

Household Size -0.323 0.261 -0.377 -0.387 0.462

Household Female Ratio 29.147 ** 3.188 30.608 ** 30.845 ** -4.975

WX

X

Statistical Significance:

*** 1 %, ** 5%, * 10%, 15%.


27/34


Public GoodsGame, Round 1

WY

W

Neighborhood Field Plot Residential


Spatial Model Cross Cross Cross Cross

Endogenous Social Effect ()

Correlated Social Effect ()


Volumetric Pricing Dummy -9.637 -11.411

Age -0.623 0.548 -0.682 * -0.520

Gender Dummy -9.676 19.125 -10.821 35.160 **

Years of Schooling -0.152 -0.088 0.533 -1.508

X

Statistical Significance:*** 1 %, ** 5%, * 10%, 15%.

Ln Asset 0.013 -3.671 3.425 -3.891

Field Area (ha) 8.575 -7.042 2.966 5.672Household Size -1.822 -0.502 1.453 0.633

Household Female Ratio 11.186 11.712 -8.118 31.177


Volumetric Pricing Dummy 0.221 0.217

Age -0.450 *** -0.713 *** -0.441 *** -0.743 ***

Gender Dummy -4.709 -3.032 -5.778 -2.514

Years of Schooling 0.112 -0.194 0.236 -0.044

Ln Asset 2.323 2.589 1.307 0.981

Field Area (ha) 5.586 * -2.125 5.754 * -1.965

Household Size -0.742 -0.717 -0.575 -0.889

Household Female Ratio 20.000 -18.463 18.812 -14.913

Control

Risk-Taking Behavior 0.227 *** 0.132 0.215 *** 0.135


28/34


WY

W

WX

Neighborhood Field Plot Residential


Spatial Model

Lag &

Cross Cross Cross

Lag &

Cross Cross

Endogenous Social Effect () 0.332 *** 0.284 ***


Volumetric Pricing Dummy 1.337 -11.436 * -6.470

Age -0.317 -0.584 -0.276 -0.221 -0.579

Gender Dummy 0.439 14.381 -3.750 -1.077 25.875

Years of Schooling -2.103 * 3.266 1.360 0.428 1.365

Ln Asset 2.964 -0.095 -1.520 -1.265 -3.677

Public GoodsGame, Round 2

WY

W

WX

X

Statistical

Significance:

*** 1 %, ** 5%,

* 10%, 15%.

. . . . .

Household Size -0.577 3.104 0.798 0.224 4.020

Household Female Ratio -8.270 19.684 -35.690 -28.779 -19.813


Volumetric Pricing Dummy 3.199 3.314 4.083

Age 0.167 -0.169 0.179 0.187 -0.089

Gender Dummy 2.492 0.145 0.690 2.059 -0.156

Years of Schooling 0.371 0.214 0.385 0.366 0.322

Ln Asset 1.746 0.954 1.110 1.330 0.387

Field Area (ha) -0.716 -2.529 1.028 0.302 -2.671

Household Size 0.613 0.219 0.567 0.517 0.144

Household Female Ratio 7.872 -5.448 2.114 5.034 -6.961

Controls

Risk-Taking Behavior 0.126 ** 0.252 *** 0.193 *** 0.161 *** 0.240 ***

Round 1 Message Receipt Dummy 7.312 ** 11.574 ** 7.139 * 7.416 ** 11.731 **

Round 1 Free Riding Index 0.212 * 0.014 0.232 * 0.266 ** 0.093

Round 1 MRD x FRI 0.440 ** -0.358 0.471 * 0.419 ** -0.289

Round 1 Own Contribution 0.847 *** 0.542 *** 0.821 *** 0.840 *** 0.584 ***

X


29/34

Summary of the Results

H1) Social behaviors of individual farmers are influenced by their neighbors social behaviors and

personal attributes. The endogenous social effects among irrigated farmers are found in dictator game and monitored public goods game.

The exogenous social effects are minor on the whole. No correlated social effects are found.

H1 is accepted to the extent that it depends on the irrigation availability and the type of social

behavior.

H2) Neighborhood effects on social behaviors, particularly contribution to public goods, augment in

the irrigated areas vis--vis in the rain-fed areas. There exists a clear contrast in the result between the two ecosystems.

Hypothesis Verification

H1

H2

28

The endogenous social effects and the impact of FRI are found only in the irrigated areas.

H2 is definitely supported.

H3) In the irrigated areas, farmers are more contributory to public goods when they are engaged in

volumetric water pricing system than in area-based flat rate system Volumetric water pricing makes no difference in the outcome of dictator game and pre-monitoring public goods game

It has a minimal positive effect in monitored public goods game.

H3 is only weakly supported.

H4) The endogenous social effects on public goods contribution are more salient among farm plot

neighbors than among residential neighbors. The spillover of public goods contribution under monitoring is stronger among plot neighbors than among residential

neighbors.

H4 is clearly accepted.

H3H3

H4H4


30/34

Concluding Remarks

I. Emergence ofsocial norms and community mechanisms in irrigated society.

i. Farmers altruistic behavior and contributory behavior spill over to their neighbors,

indicating that collective actions required in irrigation water management induce the

emergence ofsocial norm: farmers decide on their social behavior more or less by

following the way their neighbors behave socially.

ii. Cooperative resource management also promotes a community mechanism: free

riding acts are corrected.

29

iii. This irrigation system was introduced in 2008. It is implied that, by intervention

schemes such as the construction of gravity irrigation, changes in social norm and

community mechanism occur rather shortly than slowly.

Essential Assumption: There was no intrinsic difference in behavioral spillover among

irrigated farmers and among rainfed farmers prior to the construction of irrigation,

which is partially supported by the descriptive tables.


31/34

Concluding Remarks

II. Limitations & Possible Extensions:

i. The contrasting result between the two ecosystems may be partially attributed to the

relative concentration of neighbors in the rainfed areas, (in addition to the relatively

independent agricultural practices). The result would be more convincing if we could

separate out the former factor.

ii. Number of neighbors (k-nearst neighbors) and Distance to neighbors (threshold

distance) are 2 conflicting criteria commonly used in constructing weight matrices.

30

e ave assume t e a rness o mpos ng t e same t res o stance across t e

types of neighborhood. It may be interesting to use a k-nearest neighbor criterion,though the choice of k has to be justified some way.

iii. The next step of this research may be to estimate social behaviors using social

distance instead of geographical distance. Such a study will explore how personal

relationship affects the spillover of social behavior (and will also be applicable tosituations in which geographical distance is not a hard constraint).


32/34

Appendix

31


33/34

Order of Approximation bias (%)

7th 0.5000 0.5217 4.35

5th 0.5217 4.34

3rd 0.5208 4.17

Marginal Effect vs. Coefficient

Consider the endogenous social effects.

Let be the marginal effect.

+ 3 + 5 + 7 + = = 1

k1

k2

1

k3

1

j=1

2k-1

kj-1

7th 0.4000 0.4110 2.745th 0.4110 2.74 marginal effect () coefficient () bias (%)

3rd 0.4107 2.67 0.5000 0.5217 4.35

0.4000 0.4110 2.74

7th 0.3000 0.3046 1.52 0.3000 0.3046 1.52

5th 0.3046 1.52 0.2000 0.2013 0.67

3rd 0.3045 1.50 0.1000 0.1002 0.17

7th 0.2000 0.2013 0.67

5th 0.2013 0.67

3rd 0.2013 0.67

7th 0.1000 0.1002 0.17

5th 0.1002 0.17

3rd 0.1002 0.17

32


34/34

Marginal Effect vs. Coefficient

Consider the exogenous social effects.

Let be the marginal effect.

=

1-

(1+ + 2 + 3 + 4 + .........) = j = (if 0 < < 1)j=0

When = 0.3 , then 1.4

33

Neighborhood Effects and Social: The case of Irrigated and Rainfed Rice Farners in Bohol

Documents

Transcript of Neighborhood Effects and Social: The case of Irrigated and Rainfed Rice Farners in Bohol