Measuring Trust in Social Networks Tanya Rosenblat (Wesleyan University, IQSS and IAS) February 16,...
-
Upload
clarence-small -
Category
Documents
-
view
214 -
download
0
Transcript of Measuring Trust in Social Networks Tanya Rosenblat (Wesleyan University, IQSS and IAS) February 16,...
Measuring Trust in Social Networks
Tanya Rosenblat (Wesleyan University, IQSS and IAS)
February 16, 2006
What is Trust?
Dan lends money to Shachar. Having access to money allows Shachar to start a business and generate a profit which he can share with Dan (by paying a high interest rate for example). At the same time Shachar can harm Dan by refusing to repay the loan, for example (or repaying it late).
Muriel asks Tanya to look after her house or apartment or to take care of financial matters during a prolonged absence. Having Tanya take care of these errands is much less expensive than hiring a professional (a lawyer or an accountant, for example). At the same time, Tanya might turn out to be unreliable - pay utility bills late and rack up late charges etc.
What is Trust?
Dan lends money to Shachar. Having access to money allows Shachar to start a business and generate a profit which he can share with Dan (by paying a high interest rate for example). At the same time Shachar can harm Dan by refusing to repay the loan, for example (or repaying it late).
Dan trusts Shachar.
Shachar is trustworthy.
Why?
Trust Measures in Economics
Surveys (General Social Survey and World Values Survey) “Generally speaking, would you say that most people can be trusted or
that you can’t be too careful in dealing with people?” Highest trust countries are in Scandinavia; lowest trust – in South
America
Some problems with GSS type questions: What is the reference group? What is trust? (not defined in the question) Are participants truthful when answering this potentially sensitive
question?
Trust Measures in Economics
Trust (or Investment) GamePlayer 1 Sender
Player 2 Receiver
S sends x to R;
R receives 3xS R
R keeps y and sends 3x – y back to S
Trust Measures in Economics
Trust (or Investment) GamePlayer 1 Sender
Player 2 Receiver
S sends x to R;
R receives 3xS R
R sends y back to S and keeps 3x – y
Interpretation:
S is trusting if he sends x >0;
R is trustworthy if she reciprocates by sending y>0
Trust Measures in Economics
Trust game and GSS answers don’t coincide Trustworthy behavior predicts real life outcomes (e.g.,
repay loans) Trusting behavior possibly “gambling”
Some reasons to be cautious:
Trust Measures in Economics
Look at social networks and measure trust as an outcome of repeated interactions
Network gives1. Information about “types”
2. Mechanism for punishment of bad behavior
Research Strategy:1. Map the network
2. Measure how trust varies with social distance
Measuring Trust in Social Networks Two social networks:
1. Undergraduates at a large private university
2. Residents of shantytowns of Lima, Peru
Measuring Trust in Social Networks Two measurement techniques:
1. “Laboratory” web-based experiment
2. Field experiment using a new microfinance program
What is Trust? – some common definitions
“Firm reliance on the integrity, ability, or character of a person” (The American Heritage Dictionary)
“Assured resting of the mind on the integrity, veracity, justice, friendship, or other sound principle, of another person; confidence; reliance;” (Webster’s Dictionary)
“Confidence in or reliance on some quality or attribute of a person” (Oxford English Dictionary)
What is Trust?
“Confidence in or reliance on some quality or attribute of a person” (Oxford English Dictionary)
Define “trust” as willingness of agent to lend money to another agent. Define “trust” as willingness of agent to lend money to another agent.
What is Trust?
“Confidence in or reliance on some quality or attribute of a person” (Oxford English Dictionary)
Define “trust” as willingness of agent to lend money to another agent.Define “trust” as willingness of agent to lend money to another agent.
Trust will arise naturally in repeated interactions. Research Strategy – look at social networks.Trust will arise naturally in repeated interactions. Research Strategy – look at social networks.
Sources of Trust:2. Cooperative: Enforcement Trust2. Cooperative: Enforcement Trust
1. Information-Based:Type Trust1. Information-Based:Type Trust
Sources of Trust:
I know the other person’s type (responsible/ irresponsible with money).
2. Cooperative: Enforcement Trust2. Cooperative: Enforcement Trust
1. Information-Based:Type Trust1. Information-Based:Type Trust
Sources of Trust:
I know the other person’s type (responsible/ irresponsible with money).
Information about other agents decreases with social distance.
2. Cooperative: Enforcement Trust2. Cooperative: Enforcement Trust
1. Information-Based:Type Trust1. Information-Based:Type Trust
Sources of Trust:
I know the other person’s type (responsible/ irresponsible with money).
Information about other agents decreases with social distance.
The other person fears punishment in future interactions with me (or other players) if she does not repay me.
2. Cooperative: Enforcement Trust2. Cooperative: Enforcement Trust
1. Information-Based:Type Trust1. Information-Based:Type Trust
Sources of Trust:
I know the other person’s type (responsible/ irresponsible with money).
.
Information about other agents decreases with social distance.
The other person fears punishment in future interactions with me (or other players) if she does not repay me.
Fear of punishment can differ by social distance (differently afraid of punishment from friends, friends of friends, friends of friends of friends or strangers)
2. Cooperative: Enforcement Trust2. Cooperative: Enforcement Trust
1. Information-Based:Type Trust1. Information-Based:Type Trust
House Experiment - Social Network
Residential social network of (569) upper-class undergraduates (sophomores, juniors and seniors) at a large private university.
Students are randomly allocated to 12 residential houses after their freshman year (as a blocking group of 2-8 students).
Students make long-term friendships within the houses (since houses provide meals, entertainment and educational activities).
2 Houses used for the study
Network Measurement Methodology Need high participation rate in order to get meaningful
network data. In addition to participation fee and experimental earnings,
conduct a raffle with valuable prizes at the end of the study. A major publicity campaign that advertises experiment (letters
in the mail, posters, flyers, information table in the dining halls).
Direct emailing was not allowed until subjects signed up and agreed to receive emails.
Network Measurement Methodology Networks are usually measured through surveys Instead, use a coordination game with monetary payoffs to
induce subjects think more carefully about their answers Subjects name up to 10 friends and some dimensions of their
friendship (e.g., how much time they spend together during the week).
Network Elicitation Game:
Tanya Alain
Tanya names Alain
Network Elicitation Game:
Tanya Alain
Tanya Alain
Alain names Tanya
Tanya gets a prize of $1 if
Network Elicitation Game:
Tanya Alain
Tanya Alain
Alain names Tanya; Alain also gets a prize of $1
Tanya gets a prize of $1 if
Alain and Tanya get an additional prize if they agree on how much time they spend together each week.
Network Elicitation Game:
Tanya Alain
If T names A and A names T (coordinate) we call it a link; the link is stronger if there is agreement on the attributes of the relationship.
Network Elicitation Game:
Tanya Alain
In order to protect students’ feelings, each match is paid with 50% probability – so if they get 0, they don’t know whether this is because they were ‘rejected’, or because they were unlucky.
Network Data
In addition to the network game Know who the roommates are Geographical network (where rooms are located in the
house) Data from the Registrar’s office Survey on lifestyle (clubs, sports) and socio-economic
status
Network Data – Sample Description House1 - 46% (259); House2 - 54% (310) Sophomores - 31%(174); Juniors - 30% (168); Seniors -
40% (227) Female - 51% (290); Male - 49% (279)
5690 one-way relationships in the dataset; 4042 excluding people from other houses
2086 symmetric relationships (1043 coordinated friendships)
Symmetric Friendships
0 1 2 3 4 5 6 7 8 9 100
20
40
60
80
100
120
140
Symmetric Friendships
0 1 2 3 4 5 6 7 8 9 100
20
40
60
80
100
120
140
The agreement rate on time spent together (+/- 1 hour) is 80%
Network description
Cluster coefficient (probability that a friend of my friend is my friend) is .5841
The average path length is 6.5706 1 giant cluster and 34 singletons If ignore friends with less than 1 hr per
week, many disjoint clusters (175)
Experimental Design
Use Andreoni-Miller (Econometrica, 2002) GARP framework to measure altruistic types
A modified dictator game in which the allocator divides tokens between herself and the recipient. Tokens can have different values to the allocator and the recipient.
Subjects divide 50 tokens which are worth:1 token to the allocator and 3 to the recipient2 tokens to the allocator and 2 to the recipient3 tokens to the allocator and 1 to the recipient
Goals of the Experimental Design:
1) Measure Agent’s Altruistic Type and how their altruism varies with social distance (when allocators know the identity of the recipient).
Goals of the Experimental Design:
1) Measure Agent’s Altruistic Type and how their altruism varies with social distance (when allocators know the identity of the recipient).
2) Distinguish between information-based and cooperative trustworthiness by varying the degree to which the recipient finds out about allocator’s actions.
Goals of the Experimental Design:
1) Measure Agent’s Altruistic Type and how their altruism varies with social distance (when allocators know the identity of the recipient).
3) Measure Recipients’ expectations about actions of allocators to understand to what extent recipients know about trustworthiness of allocators and how accurately it is alligned with the decisions of allocators (use this to study trusting behavior)
2) Distinguish between information-based and cooperative trustworthiness by varying the degree to which the recipient finds out about allocator’s actions
Experimental Design
Each allocator participates in 4 treatments in random order: Baseline: anonymous allocator and anonymous
recipient (AA). Anonymous allocator and known recipient (AK) Known allocator and anonymous recipient (KA) Known allocator and known recipient (KK)
With some uncertainty (always 15% chance that allocations are made by computer)
Sources of Trust:
The other player is altruistic and takes my utility into account.
Anonymous Allocator/Anonymous Recipient (AA), Anonymous Allocator/Known Recipient (AK)
2. Cooperative (Enforcement) Trust:2. Cooperative (Enforcement) Trust:
The other player fears punishment in future interactions with me (or other players) if she does not take my utility into account.
Known Allocator/Anonymous Recipient (KA), Known Allocator/Known Recipient (KK)
1. Information-based (Type) Trust:1. Information-based (Type) Trust:
DirectFriend
DirectFriend
Direct Friend
DirectFriend
Allocator
For Allocator choose 5 Recipients (in random order): 1 direct friend; 1 indirect friend of social distance 2; 1 indirect friend of social distance 3; 1 person from the same staircase; 1 person from the same house.
IndirectFriend2 links
IndirectFriend3 links
Sharestaircase
Samehouse
Who is the Recipient when known? (AK and KK)
Experimental Design – What Do Recipients Do?
Recipients make predictions about how much they will get from an allocator in a given situation and how much an allocator will give to another recipient that they know in a given situation.
One decision is payoff-relevant:
=> The closer the estimate is to the actual number of tokens passed the higher are the earnings.
Incentive Compatible Mechanism to make good predictions
Get $15 if predict exactly the number of tokens that player 1 passed to player 2
For each mispredicted token $0.30 subtracted from $15. For example, if predict that player 1 passes 10 tokens and he actually passes 15 tokens then receive $15-5 x $0.30=$13.50.
DirectFriend
DirectFriend
Direct Friend
DirectFriend
Recipient
Recipients are asked to make predictions in 7 situations (in random order): 1 direct friend; 1 indirect friend of social distance 2; 1 indirect friend of social distance 3; 1 person from the same staircase; 1 person from the same house; 2 pairs chosen among direct and indirect friends
IndirectFriend2 links
IndirectFriend3 links
Sharestaircase
Samehouse
Recipients’ Expectations
DirectFriend
DirectFriend
Direct Friend
DirectFriend
Recipient
Recipients are asked to make predictions in 7 situations (in random order): 1 direct friend; 1 indirect friend of social distance 2; 1 indirect friend of social distance 3; 1 person from the same staircase; 1 person from the same house; 2 pairs chosen among direct and indirect friends
IndirectFriend2 links
IndirectFriend3 links
Sharestaircase
Samehouse
Recipients’ Expectations
A possible pair
Experimental Design
Within-subject design with randomized order of presentation: either all choices with “will find out” on one screen followed by “will not find out” screen; or “will find out/will not find out” on one screen for each choice.
Timing - Allocators:
AA and AK
or
AA and AA
Session 1; 1 decision from 1 pair chosen for monetary payoff (max $15)
Timing - Allocators:
AA and AK
or
AA and AA
OR
KK and KA
or
KA and KK
Session 1; 1 decision from 1 pair chosen for monetary payoff (max $15)
Session 2 (1 week later); 1 decision from 1 pair chosen for monetary payoff (max $15)
Analysis
Identify Types
Analysis (AM)
Selfish types take all tokens under all payrates.
Leontieff (fair) types divide the surplus equally under all payrates.
Social Maximizers keep everything if and only if a token is worth more to them.
Analysis (AM)
About 50% of agents have pure types, the rest have weak types.
Force weak types into selfish/fair/SM categories by looking at minimum Euclidean distance of actual decision vector from type’s decision.
Comparison of Anonymous/Non-Anonymous Dictator Games
0.2
.4.6
mean
of share
ANONYMOUS FRIEND
Recipient does not find out
1 - Selfish 2 - Fair
3 - Social Maximizer
0.1
.2.3
.4.5
mean
of share
ANONYMOUS FRIEND
Recipient does find out
1 - Selfish 2 - Fair
3 - Social Maximizer
Comparison of Anonymous/Non-Anonymous Dictator Games
0.2
.4.6
mean
of share
ANONYMOUS FRIEND
Recipient does not find out
1 - Selfish 2 - Fair
3 - Social Maximizer
0.1
.2.3
.4.5
mean
of share
ANONYMOUS FRIEND
Recipient does find out
1 - Selfish 2 - Fair
3 - Social Maximizer
0.54
0.180.28
0.48
0.220.30
0.45
0.30 0.25 0.25
0.48
0.27
Comparison of Anonymous/Non-Anonymous Dictator Games
0.2
.4.6
mean
of share
ANONYMOUS FRIEND
Recipient does not find out
1 - Selfish 2 - Fair
3 - Social Maximizer
0.1
.2.3
.4.5
mean
of share
ANONYMOUS FRIEND
Recipient does find out
1 - Selfish 2 - Fair
3 - Social Maximizer
0.54
0.180.28
0.48
0.220.30
0.45
0.30 0.25 0.25
0.48
0.27
If the recipient does not find out the identity of the allocator then allocators are only slightly less selfish towards friends than anonymous recipients.
Comparison of Anonymous/Non-Anonymous Dictator Games
0.2
.4.6
mean
of share
ANONYMOUS FRIEND
Recipient does not find out
1 - Selfish 2 - Fair
3 - Social Maximizer
0.1
.2.3
.4.5
mean
of share
ANONYMOUS FRIEND
Recipient does find out
1 - Selfish 2 - Fair
3 - Social Maximizer
0.54
0.180.28
0.48
0.220.30
0.45
0.30 0.25 0.25
0.48
0.27
If the recipient does find out the identity of the allocator then allocators are much less selfish towards friends than anonymous recipients.
Analysis
There is only weak evidence for directed altruism. Almost half the sample are altruists (fair or SM) – but they do not appear to discriminate.
Friends are treated better if the allocator fears that they might find out that she acted selfishly.
Analysis
We find similar results for beliefs about other allocators’ types.
Repeated game concerns make friends more trustworthy and more trusting.
Field Experiment
Location – Urban shantytowns of Lima, Peru
Trust Measurement Tool - a new microfinance program where borrowers can obtain loans at low interest by finding a “sponsor” from a predetermined group of people in the community who are willing to cosign the loan.
Types of Networks
Which types of networks matter for trust? Survey work to identify
SocialBusinessReligiousKinship
Survey Work in Lima’s North Cone
Who is a “sponsor”?
From surveys, select people who either have income or assets to serve as guarantors on other people’s loans.
25-30 for each community If join the program, allowed to take out
personal loans (up to 30% of sponsor “capacity”).
Presenting Credit Program to Communities in Lima’s North Cone
Experimental Design
3 random variations:Sponsor-specific interest rate
Helps identify how trust varies with social distance
Sponsor’s liability for co-signed loan Helps separate type trust from enforcement trust
Interest rate at community level Helps identify whether social networks are efficient
at allocating resources
DirectFriend
DirectFriend
Direct Friend
DirectFriend
Sponsor 1r1
Sponsor-specific interest rate is randomized
IndirectFriend2 links
IndirectFriend3 links
Random Variation 1
DirectFriend
DirectFriend
Direct Friend
DirectFriend
Sponsor 1r1
Sponsor-specific interest rate is randomized
IndirectFriend2 links
IndirectFriend3 links
Sponsor 2r2 < r1
Random Variation 1
DirectFriend
DirectFriend
Direct Friend
DirectFriend
Sponsor 1r1
Sponsor-specific interest rate is randomized
IndirectFriend2 links
IndirectFriend3 links
Random Variation 1
Sponsor 2r2 < r1
The easier it is to substitute sponsors, the higher is trust in the community.
Should I try to get
sponsored by Sponsor1 or Sponsor2?
DirectFriend
DirectFriend
Direct Friend
DirectFriend
Sponsor 1r1
Sponsor-specific interest rate is randomized
IndirectFriend2 links
IndirectFriend3 links
Random Variation 1
Sponsor 2r2 < r1
Measure the extent to which agents substitute socially close but expensive sponsors for more socially distant but cheaper sponsors.
Should I try to get
sponsored by Sponsor1 or Sponsor2?
Randomization of interest rates Decrease in interest rate based on slope:
SD1 SD2 SD3 SD4
Slope 1 0.125 0.250 0.375 0.500
Slope 2 0.250 0.500 0.750 1.000
Slope 3 0.500 1.000 1.500 2.000
Slope 4 0.750 1.500 2.250 3.000
Each client is randomly assigned a slope (1,2,3,4):
Close friends generally provide the highest interest rate and distant acquaintances the lowest, but the decrease depends on SLOPE
Demand Effects
The interest rate on the previous slide for 75% of the sample and 0.5 percent higher for 25% of the sample to check for demand effects (people borrow more and for a different reason when interest rates are lower?).
DirectFriend
DirectFriend
Direct Friend
DirectFriend
Sponsor 1r1
Sponsor’s liability for the cosigned loan is randomized (after borrower-sponsor pair is formed)
IndirectFriend2 links
IndirectFriend3 links
Random Variation 2
Measure the extent to which sponsors can control ex-ante moral hazard.(can separate type trust from enforcement trust by looking at repayment rates).
Sponsor’s liability might fall below 100%
Community 1
Low r
Community 2
High r
Random Variation 3 Average interest rate at community level (to measure cronyism)
Under cronyism, the share of sponsored loans to direct friends (insiders) increases as interest rate is reduced.
The setting: Urban Shantytowns in Lima’s North Cone Some MFIs operate there, offering both individual
and group lending, with varying levels of penetration but never very high.
Work has been conducted in 2 communities in Lima’s North Cone.
Experimental Process
Household census Establish basic information on household assets and
composition. Provides us with household roster for Social Mapping Provides us with starting point to identify potential
sponsors Identify and sign-up sponsors through series of community
meetings Offer lending product to community as a whole
Microlending Partner
Alternativa, a Peruvian NGO Lending operation (both group and individual
lending) Also engaged in plethora of “community building”,
“empowerment”, “information”, education, etc.
The Lending Product
Community ~300 households We identify 25-30 “sponsors” who have assets and/or
stable income, sufficient to act as a guarantor on other people’s loans.
A sponsor is given a “capacity”, the maximum amount of credit they can guarantee.
A sponsor can borrow 30% of their capacity for themselves.
Individuals in the community are each given a “sponsor card” which lists the sponsors in their community and their interest rate if they borrow from each sponsor.
Results so far…
So far work has been conducted in 2 communities in Lima’s North Cone.
The first community has 240 households and the second community has 371 households.
Characteristics of Sponsored Loans
The average size of a sponsored loan is $317 or 1040 soles.
The average interest rate for sponsored loans is 4.08%
Social Distance of Actual Client-Sponsor by Slope
0.5
11.5
2m
ea
n o
f sd
1 2 3 4
All Communities
Social Distance of Actual Client-Sponsor by Slope
0.5
11.5
2m
ea
n o
f sd
1 2 3 4
All Communities
Greater slope makes distant neighbors more attractive due tolower interest. We see substitution away from expensive closeneighbors.
Social Distance of Actual Client-Sponsor by Slope
0.5
11.5
2m
ea
n o
f sd
1 2 3 4
All Communities
Effect is mainly driven by clients substituting SD=1 for SD=2 sponsors.There is less substitution of SD=2 sponsors for SD=3,4 sponsors.Therefore, slope 2,3,4 look different from slope 1 (where all interestrates are essentially equal) – but not so different from each other.
Results using logistic regressions: Direct social neighbor has the same effect as a 3-4
percent decrease in interest rate
Even acquaintance at social distance 3 is worth about as much as one percent decrease in interest rate
Independent effect of geographic distance: one standard deviation decrease in geographic distance is worth about as much as a one percent drop in interest rate
Repayment rates of clients and sponsors
020
40
60
80
10
0m
ean
of share
left
0 1 2 3 4 5 6 7 8 9 10 11 12
48 sponsor loans and 49 non-sponsor loans
6dN
Non-sponsor loan Sponsor loan
020
40
60
80
10
0m
ean
of share
left
0 1 2 3 4 5 6 7 8 9 10 11 12
55 sponsor loans and 89 non-sponsor loans
Los Olivos
Non-sponsor loan Sponsor loan
Repayment rates of clients and sponsors
020
40
60
80
10
0m
ean
of share
left
0 1 2 3 4 5 6 7 8 9 10 11 12
48 sponsor loans and 49 non-sponsor loans
6dN
Non-sponsor loan Sponsor loan
020
40
60
80
10
0m
ean
of share
left
0 1 2 3 4 5 6 7 8 9 10 11 12
55 sponsor loans and 89 non-sponsor loans
Los Olivos
Non-sponsor loan Sponsor loan
Repayment rates after n months (n=1,2,..,12) are similar for sponsorsand non-sponsors in both communities.
Effect of Second Randomization0
20
40
60
801
00
mean
of share
left
0 1 2 3 4 5 6 7 8 9
18 loans with 100 percent sponsors and 5 loans with 50 percent sponsors
Low quality clients
100 percent sponsor resp. 50 percent sponsor resp.
020
40
60
801
00
mean
of share
left
0 1 2 3 4 5 6 7 8 9
19 loans with 100 percent sponsors and 7 loans with 50 percent sponsors
High quality clients
100 percent sponsor resp. 50 percent sponsor resp.
Note: This graph only includes loans which are 6 months and older.
Effect of Second Randomization0
20
40
60
801
00
mean
of share
left
0 1 2 3 4 5 6 7 8 9
18 loans with 100 percent sponsors and 5 loans with 50 percent sponsors
Low quality clients
100 percent sponsor resp. 50 percent sponsor resp.
020
40
60
801
00
mean
of share
left
0 1 2 3 4 5 6 7 8 9
19 loans with 100 percent sponsors and 7 loans with 50 percent sponsors
High quality clients
100 percent sponsor resp. 50 percent sponsor resp.
Note: This graph only includes loans which are 6 months and older.
Higher sponsor responsibility increases repayments rates of BAD clients(defined as having paid back less than 50 percent after 6 months).No effect of repayment of high-quality clients.
Effect of Second Randomization0
20
40
60
801
00
mean
of share
left
0 1 2 3 4 5 6 7 8 9
18 loans with 100 percent sponsors and 5 loans with 50 percent sponsors
Low quality clients
100 percent sponsor resp. 50 percent sponsor resp.
020
40
60
801
00
mean
of share
left
0 1 2 3 4 5 6 7 8 9
19 loans with 100 percent sponsors and 7 loans with 50 percent sponsors
High quality clients
100 percent sponsor resp. 50 percent sponsor resp.
Note: This graph only includes loans which are 6 months and older.Evidence for enforcement trust!
Conclusion: We develop a new microfinance program to measure
trust within a social network. Preliminary evidence suggests that social networks
can greatly reduce borrowing costs (measured in terms of interest rate on loan).
Evidence that sponsors pick clients who are as likely to repay as they are (micro-finance organization is no better) (type trust)
Evidence that sponsors can enforce repayment for a chosen client (enforcement trust).
Sources of Trust:
I know the other person’s type (responsible/ irresponsible with money).
.
A direct friend is worth about 3% (monthly) interest rateA friend of a friend is worth about 1% (monthly) interest rate
The other person fears punishment in future interactions with me (or other players) if she does not repay me.
Fear of punishment enforces loan repayment for bad clients
2. Cooperative: Enforcement Trust2. Cooperative: Enforcement Trust
1. Information-Based:Type Trust1. Information-Based:Type Trust
Future work:
More communities Decompose trust by link type Distinguish type and enforcement trust
AND: Cronyism
Logistic regressions confirm earlier graphs and quantify the size of thesocial distance/interest rate tradeoff: a direct link to a sponsor is worthabout 4 interest rate points. A link to a neighbor at distance 2 is worthabout half that much.