You are on page 1of 101

SocialandEconomic Networks: ModelsandAnalysis

Matthew O. Jackson
Stanford University, Santa Fe Institute, CIFAR,
www.stanford.edu\~jacksonm
Copyright 2013 The Board of Trustees of The Leland Stanford Junior University. All Rights Reserved. Figures reproduced with permission from Princeton University Press.

1.1:Introduction

WhyStudyNetworks?
Manyeconomic,political,andsocialinteractionsareshapedbythelocal structureofrelationships: tradeofgoodsandservices,mostmarketsarenotcentralized! sharingofinformation,favors,risk,... transmissionofviruses,opinions... accesstoinfoaboutjobs... choicesofbehavior,education,... politicalalliances,tradealliances Socialnetworksinfluencebehavior crime,employment,humancapital,voting,smoking, networksexhibitheterogeneity,butalsohaveenoughunderlyingstructureto model Pureinterestinsocialstructure understandsocialnetworkstructure

PrimaryQuestions:
Whatdoweknowaboutnetworkstructure? Howdonetworksform?Dothe`right networksform? Howdonetworksinfluencebehavior?(and viceversa...)

Synthesize
Manyliteraturesdealwithnetworks
Sociology Economics ComputerScience StatisticalPhysics Math(randomgraph)...

Whathavewelearned?Whatareimportant areasforfutureresearch?

ThreeAreasforResearch
Theory networkformation,dynamics,design... hownetworksinfluencebehavior coevolution? Empiricalandexperimentalwork observenetworks,patterns,influence testtheoryandidentifyregularities Methodology howtomeasureandanalyzenetworks

CentralFocus
Modelsforanalyzingandunderstanding networks:
Randomgraphmethods Strategic,gametheoretictechniques hybrids,statisticalmodels

Goals
Presumenopriorknowledge Introduceyoutoavarietyofapproachesto modelingnetworks(morebreadththandepth) Giveasenseofdifferentdisciplinestechniques andperspectives

Models
Provideinsightintowhyweseecertainphenomena: Whydosocialnetworkshaveshortaveragepathlengths? Allowforcomparativestatics: Howdoescomponentstructurechangewithdensity? Importantincontagion/diffusion/learning... Predictoutofsample: Whatwillhappenwithanewpolicy(vaccine,R&Dsubsidy, ...)? Allowforstatisticalestimation: Istheresignificantclusteringonalocallevelordiditappear atrandom?

Outline
PartI:BackgroundandFundamentals
DefinitionsandCharacteristicsofNetworks(1,2) EmpiricalBackground(3)

PartII:NetworkFormation
RandomNetworkModels(4,5) StrategicNetworkModels(6,11)

PartIII:NetworksandBehavior
DiffusionandLearning(7,8) GamesonNetworks(9)

SocialandEconomic Networks: ModelsandAnalysis


Matthew O. Jackson
Stanford University, Santa Fe Institute, CIFAR,
www.stanford.edu\~jacksonm
Copyright 2013 The Board of Trustees of The Leland Stanford Junior University. All Rights Reserved.

1.2:ExamplesandChallenges

Outline
PartI:BackgroundandFundamentals
DefinitionsandCharacteristicsofNetworks(1,2) EmpiricalBackground(3)

PartII:NetworkFormation
RandomNetworkModels(4,5) StrategicNetworkModels(6,11)

PartIII:NetworksandBehavior
DiffusionandLearning(7,8) GamesonNetworks(9)

TwoExamples
Ideaofdata Viewofapplications Previewsomequestions

PUCCI

BISCHERI PERUZZI LAMBERTES

STROZZI

Strozzi
RIDOLFI TORNABUON

GUADAGNI

Guadagni
GINORI

CASTELLAN

ALBIZZI

BARBADORI MEDICI

Medici
ACCIAIUOL

Padgett and Ansells (1993) Data (from Kent 1978) Florentine Marriages, 1430s

SALVIATI

PAZZI

PUCCI

BISCHERI PERUZZI LAMBERTES

STROZZI

Strozzi .104
RIDOLFI TORNABUON

GUADAGNI

Guadagni .255
GINORI

CASTELLAN

ALBIZZI

BARBADORI MEDICI

Medici .522
ACCIAIUOL

Padgett and Ansells (1993) Data (from Kent 1978) Florentine Marriages, 1430s

SALVIATI

PAZZI

Greece .18 France .13

Elliott, Golub, Jackson (2012)

Germany

.11

.11 .05 Portugal Italy

.12

Spain

WhatdoWeKnow?
Networksplayroleinmanysettings
Jobcontacts,crime,risksharing,trade,politics,...

Networkpositionandstructurematters
richsociologyliterature Medicis notthewealthiestnorthestrongestpolitically, butthemostcentral

``SocialNetworkshavespecialcharacteristics
smallworlds,degreedistributions...

Embeddednessof EconomicActivity
Fewmarketsarecentralized,anonymous Specificrelationshipsmatter...

NetworksinLaborMarkets
MyersandShultz(1951) textileworkers:
62%firstjobfromcontact 23%bydirectapplication 15%byagency,ads,etc.

ReesandShultz(1970) Chicagomarket:
Typist37.3% Accountant23.5% Materialhandler73.8% Janitor65.5%,Electrician57.4%

Granovetter(1974),Ioannides andLoury (2004)...

OtherSettings
Networksandsocialinteractionsincrime: Reiss(1980,1988) 2/3ofcriminalscommitcrimeswithothers Glaeser,Sacerdote andScheinkman (1996) socialinteractionimportantinpetty crime,amongyouths,andinareaswithlessintacthouseholds NetworksandMarkets Uzzi (1996) relationspecificknowledgecriticalingarmentindustry Weisbuch,Kirman,Herreiner (2000) repeatedinteractionsinMarseillefish markets SocialInsurance Fafchamps andLund(2000) risksharinginruralPhilippines DeWeerdt (2002) Tanzania,... Diffusion HybridcornadoptionRyanandGross(1943),Griliches (1957) DrugadoptionColeman,Katz,Menzel (1966) Sociologyliterature interlockingdirectorates,aidstransmission,language,successof immigrantgroups...

TheChallenge:
Howmanynetworksonjust30nodes? Person1couldhave29possiblelinks,person2couldhave 28notcounting1,....total=435 So435possiblelinks,eachcouldeitherbepresentornot, so2x2x2...435times=2435 networks Atoms in the universe: between 2158 and 2246

SocialandEconomic Networks: ModelsandAnalysis


Matthew O. Jackson
Stanford University, Santa Fe Institute, CIFAR,
www.stanford.edu\~jacksonm
Copyright 2013 The Board of Trustees of The Leland Stanford Junior University. All Rights Reserved.

1.3:DefinitionsandNotation

SimplifyingtheComplexity
Globalpatternsofnetworks degreedistributions,pathlengths... SegregationPatterns nodetypesandhomophily LocalPatterns Clustering,Transitivity,Support Positionsinnetworks Neighborhoods,Centrality,Influence...

RepresentingNetworks
N={1,,n}nodes,vertices,players gin {0,1}nn adjacencymatrix(unweighted,possiblydirected) gij =1indicatesalink,tie,oredgebetweeniandj Alternativenotation:ij in galinkbetweeniandj Network(N,g)

BasicDefinitions
Walkfromi1toiK:asequenceofnodes(i1,i2,...,iK) andsequenceoflinks(i1i2,i2i3,...,iK1iK)suchthat ik1ik in g foreachk

Convenienttorepresentitasthecorresponding sequenceofnodes(i1,i2,...,iK)suchthatik1ik in g foreachk

BasicDefinitions
Path:awalk(i1,i2,...iK)witheachnodeik distinct Cycle:awalkwherei1=iK Geodesic:ashortestpathbetweentwonodes

Paths,Walks,Cycles...
1 2 4 3 5 7 6 2 4 1 3 5 7 6
Path (and a walk) from 1 to 7: 1, 2, 3, 4, 5, 6, 7 Walk from 1 to 7 that is not a path: 1, 2, 3, 4, 5, 3, 7

1 2 4 3 5

7 6 2

1 3 4 5

7 6

Simple Cycle (and a walk) from 1 to 1: 1, 2, 3, 1

Cycle (and a walk) from 1 to 1: 1, 2, 3, 4, 5, 3, 1

CountingWalks:
g=
0 1 0 1 2 1 1 1 1 0 0 1 1 2 1 1 0 0 0 1 1 1 1 0 1 1 1 0 1 1 0 3 2 4

g2 =

number of walks of length 2 from i to j

CountingWalks:
g=
0 1 0 1 2 3 1 4 1 0 0 1 3 2 1 4 0 0 0 1 1 1 0 3 1 1 1 0 4 4 3 2 2 4

g3 =

number of walks of length 3 from i to j

Components
(N,g)isconnectedifthereisapathbetween everytwonodes Component:maximalconnectedsubgraph
(N,g)isasubsetof(N,g) (N,g)isconnected i inNand ij ingimpliesj inN andij ing

Anetworkwithfour components:
2 3 4 5 1 6 10 7 8 9

SocialandEconomic Networks: ModelsandAnalysis


Matthew O. Jackson
Stanford University, Santa Fe Institute, CIFAR,
www.stanford.edu\~jacksonm
Copyright 2013 The Board of Trustees of The Leland Stanford Junior University. All Rights Reserved.

1.4:Diameter

Diameter,AveragePathLength
Howclosearenodestoeachother: Howlongdoesittaketo reachaveragenode? Howfastwillinformationspread?... Howdoesitdependonnetworkdensity?

Diameter
Diameter largestgeodesic(largestshortestpath)
ifunconnected,oflargestcomponent...

Averagepathlength (lesspronetooutliers)

Diameter:
K levels has n = 2K+1-1 nodes so, K = log2(n+1) -1 diameter is 2K

diameter is either n/2 or (n-1)/2

diameter is on order of 2 log2(n+1)

Smallaveragepathlengthanddiameter
Milgram (1967)letterexperiments median5forthe25%thatmadeit CoAuthorshipstudies Grossman(2002)Mathmean7.6,max27, Newman(2001)Physicsmean5.9,max20 Goyal etal(2004)Economicsmean9.5,max29 WWW Adamic,Pitkow (1999) mean3.1(85.4%possibleof50Mpages) Facebook Backstrom etal(2012) mean4.74(721millionusers)

Neighborhood andDegree
Neighborhood:Ni(g)={j|ij ing} (usualconventioniinoting) Degree:di =#Ni(g)

ErdosRenyi (1959,1960) RandomGraphs


startwithnnodes eachlinkisformedindependently withsomeprobabilityp Servesasabenchmark``G(n,p)

SequencesofNetworks
Linksaredenseenoughsothatnetworkis connectedalmostsurely: d(n)(1+)log(n)some>0 d(n)/n0: networkisnottoocomplete

TheoremonNetwork Structure
Ifd(n)(1+)log(n)some>0andd(n)/n 0 Thenforlargen,averagepathlengthanddiameter areapproximatelyproportionaltolog(n)/log(d) (Provenforincreasinglygeneralmodels: ErdosRenyi59 MoonandMoser1966,Bollobas 1981;ChungandLu01;Jackson08;...)

TheoremonNetwork Structure
Ifd(n)(1+)log(n)some>0andd(n)/n 0 AvgDist(n)P 1 log(n)/log(d(n)) samefordiameter

SocialandEconomic Networks: ModelsandAnalysis


Matthew O. Jackson
Stanford University, Santa Fe Institute, CIFAR,
www.stanford.edu\~jacksonm
Copyright 2013 The Board of Trustees of The Leland Stanford Junior University. All Rights Reserved.

1.5:DiameterandTrees

Diameter
Boundscanbedifficult theorems arenarrow,butintuitioniseasy Letsstartwithaneasycalculation Cayley Tree:eachnodebesides leaveshasdegreed

Intuition:
1 step: Reach d nodes,

Ideas:
1 step: Reach d nodes, then d(d-1),

Ideas:
1 step: Reach d nodes, then d(d-1), then d(d-1)2,

Ideas:
1 step: Reach d nodes, then d(d-1), then d(d-1)2, d(d-1)3, ... After l steps, totals roughly d l

Movingoutl linksfromrootineachdirection reachesd+d(d1)+....d(d1)l 1nodes Thisisd((d1)l 1)/(d2)nodes:roughly(d1)l Toreachn1,needroughly(d1)l =n orl ontheorderof log(n)/log(d)

Whatifnotatree,butERrandomgraph? allhavesamedegree reallyarerandom


showthatfractionofnodesthathave nearlyaveragedegreeisgoingto1

somelinksmaydoubleback
mostnodesuntilthelaststepare stillnotreached!

SocialandEconomic Networks: ModelsandAnalysis


Matthew O. Jackson
Stanford University, Santa Fe Institute, CIFAR,
www.stanford.edu\~jacksonm
Copyright 2013 The Board of Trustees of The Leland Stanford Junior University. All Rights Reserved.

1.6:DiametersofRandom Graphs

TheoremonNetwork Structure
Ifd(n)(1+)log(n)some>0andd(n)/n 0 AvgDist(n)P 1 log(n)/log(d(n)) samefordiameter

Movingoutl linksfromrootineachdirection reachesd+d(d1)+....d(d1)l 1nodes Thisisd((d1)l 1)/(d2)nodesorroughly(d1)l Toreachn1,needroughly(d1)l =nor l ontheorderof log(n)/log(d)

Whatifnotatree,butErdosRenyirandomgraph? allhavesamedegree reallyarerandom


showthatfractionofnodesthathavenearly averagedegreeisgoingto1

E[d]>(1+ )log(n)

Chernoff Bounds: Xisbinomialvariablethen Pr(E[X]/3 X 3E[X]) 1 eE[X]

Chernoff Bounds: Xisbinomialvariablethen Pr(E[X]/3 X 3E[X]) 1 eE[X]

http://en.wikipedia.org/wiki/Chernoff_bound

Chernoff Bounds:Linksbinomialimplies Probabilitythatnodehasdegreeclosetoaverage: d Pr(d/3 3d ) 1 e i Pr (d/3 3d) (1 ed)n (missingsteps:degreesnotquiteind.)

Chernoff Bounds: Pr (d/3

3d)

(1 ed)n

Ifd>(1+ )log(n)then Pr (d/3 3d) (1 1/n1+)n exp(n) 1

So: Ifd>(1+ )log(n)then Pr (d/3

3d)

Thus: Ifd>(1+ )log(n)thenwithprob 1:

log(n)/log(3d)<l < log(n)/log(d/3)

Avg distanceanddiameter: Larged:log(3d)&log(d/3)tendtolog(d) log(n)/log(3d)<l < log(n)/log(d/3) log(n)/log(d)l

somelinksmaydoubleback
mostnodesuntilthelaststeparestillnotreached,so mostlinksstillreachingnewnodes! Afterkstepsreachedarounddk nodesandn dk still unreached ifk<log(n)/log(d)thenn dk (much)biggerthandk somostnodesthatlinktoarestillunreached...

Ideas:
Most at maximum distance (100, 10000, 1000000, 100000000...) so average distance is actually same order as diameter

SocialandEconomic Networks: ModelsandAnalysis


Matthew O. Jackson
Stanford University, Santa Fe Institute, CIFAR,
www.stanford.edu\~jacksonm
Copyright 2013 The Board of Trustees of The Leland Stanford Junior University. All Rights Reserved.

1.7:DiametersintheWorld

TheoremonNetwork Structure
Ifd(n)(1+)log(n)some>0andd(n)/n 0 Thenforlargen,averagepathlengthanddiameter areapproximatelyproportionaltolog(n)/log(d) (Provenforincreasinglygeneralmodels: ErdosRenyi59 MoonandMoser1966,Bollobas 1981;ChungandLu01;Jackson08;...)

SmallWorlds/SixDegreesof Separation
n=6.7billion(worldpopulation) d=50(friends,relatives...) log(n)/log(d)isabout6!!

Examinedataanddiameter
AddHealthdataset Schoolsvaryinaveragedegree andhomophily Doesdiametermatchlog(n)/log(d)?

AverageShortestPathvs Log(n)/Log(d)84HighSchools AdHealth


7

GolubandJackson(2012)

4 AverageShortestPathvsLog(n)/Log(d) 3 Linear(AverageShortestPathvsLog(n)/Log(d))

0 0 0.5 1 1.5 2 2.5 3 3.5 4 4.5 5

ErdosNumbers
NumberoflinksincoauthorshipnetworktoErdos Had509coauthors,morethan1400papers 2004auctionofcoauthorshipwithWilliamTozier (Erdos #=4)onEBay,winnerpaid>1000$ KevinBaconsite....

Density:AverageDegree
HSFriendships(CJP09) 6.5 Romances(BMS03) 0.8 Borrowing(BCDJ12) 3.2

Coauthors(Newman01,GLM06) Bio 15.5 Econ 1.7 Math 3.9 Physics 9.3 Facebook(Marlow09) 120

SocialandEconomic Networks: ModelsandAnalysis


Matthew O. Jackson
Stanford University, Santa Fe Institute, CIFAR,
www.stanford.edu\~jacksonm
Copyright 2013 The Board of Trustees of The Leland Stanford Junior University. All Rights Reserved.

1.8:DegreeDistributions

DegreeDistributions
Averagedegreetellsonlypartofthestory:

DegreeDistribution,G(n,p):
probabilitythatnodehasdlinksisbinomial [(n1)!/(d!(nd1)!)]pd (1p)nd1

Largen,smallp,thisisapproximatelyaPoisson distribution: [(n1)d /d!]pd e(n1)p hencename``Poissonrandomgraphs

Randomnetwork p=.02,50nodes

Note
manyisolatednodes severalcomponents nocomponenthasmorethanasmallfraction ofthenodes,juststartingtoseeonelarge oneemerge

RandomNetwork p=.08,50nodes

Distributionoflinkspernode: Fattails(Price1965)
Morehighandlowdegreenodesthan predictedatrandom
CitationNetworks toomanywith0citations, toomanywithhighnumbersofcitationstohave citationsdrawnatrandom ``Fattailscomparedtorandomnetwork

Relatedtoothersettings(wealth,citysize, wordusage...):Pareto(1896),Yule(1925),Zipf (1949),Simon(1955),

Degree NDwwwAlbert,Jeong, Barabasi(1999)


log(freq)

Log(Degree)

ScaleFreeDistributions
P(d)=cda

log(P(d))=log(c)a log(d)

Bearman, Moody, and Stovels 04 High School Romance Data

RomanceNetwork
Bearman et al HS Network

fit: Uniform at Random .99 fit: Power .84


0.3 0.4 0.5 0.6

0 -0.4 -0.3 -0.2 -0.1 0 0.1 0.2

-0.5

-1 Log CCDF

-1.5

Series1 Series2

-2

-2.5

-3 Log Degree

SocialandEconomic Networks: ModelsandAnalysis


Matthew O. Jackson
Stanford University, Santa Fe Institute, CIFAR,
www.stanford.edu\~jacksonm
Copyright 2013 The Board of Trustees of The Leland Stanford Junior University. All Rights Reserved.

1.9:Clustering

Clustering
Whatfractionofmyfriendsarefriendsofeachother? Cli(g)=#{kj ing |k,jinNi(g)}/ #{kj |k,jinNi(g)} Averageclustering: Clavg(g)=i Cli(g)/n i k j Freq of this link?

Clustering
Whatfractionofmyfriendsarefriends? Cli(g)=#{kj ing |k,jinNi(g)}/ #{kj |k,jinNi(g)} Averageclustering:Clavg(g)=i Cli(g)/n Overallclustering: Cl(g)=i #{kj ing |k,jinNi(g)}/i #{kj |k,jinNi(g)}

DifferencesinClustering
Average tends to 1

Overall tends to 0

AverageandOverallclusteringtendto0,ifmaxdegreeis boundedandnetworkbecomeslarge: Cl(g)=i #{kj ing |k,jinNi(g)}/i #{kj |k,jinNi(g)} issimplyp Ifdegreeisbounded,thenp(n1)isbounded Sopgoesto0asngrows

ClusteringinaPoisson RandomNetwork

High?ClusteringCoefficients
Prisonfriendships
.31 (MacRae 60)vs .0134
1 2 Freq of this link?

coauthorships
.15 math(Grossman02)vs .00002, .09 biology(Newman01)vs .00001, .19 econ(Goyal etal06)vs .00002,

www
.11forweblinks(Adamic 99)vs .0002

Clustering: .46 Random: .29

Padgett and Ansells data 1430s Florentine marriages and business dealings

Week1Wrap
Manyrelationshipsare``networkedandunderstanding networkstructurecanhelpunderstandbehaviorand outcomes Networksarecomplex,butcanbepartlydescribedbysome characteristics degreedistributions clustering diameter... Treelikestructuresaregeneratedbyrandomlinksleadto shortpaths Manyobservedsocialnetworksaremoreclusteredthan wouldariseatrandom

Week1:ReferencesInOrderMentioned
Jackson,M.O.(2008)SocialandEconomicNetworks,PrincetonUniversityPress Padgett,J.F.,andC.K.Ansell(1993)RobustActionandtheRiseoftheMedici,14001434,AmericanJournalofSociology98:12591319. Kent,D.(1978)TheRiseoftheMedici:FactioninFlorence14261434,Oxford:OxfordUniversityPress. Elliott,M.L.,B.GolubandM.O.Jackson(2012)``FinancialNetworksandContagion,SSRNworkingpaper2175056 MyersC.A.,andG.P.Shultz(1951)TheDynamicsofaLaborMarket,NewYork:PrenticeHall. ReesAJ,ShultzGP.1970.WorkersinanUrbanLaborMarket.Chicago:Univ.ChicagoPress GranovetterM.1973.Thestrengthofweakties.Am.J.Sociol.78:1360 80 Ioannides YM,DatcherLoury L.2004.Jobinformationnetworks,neighborhoodeffectsandinequality.J.Econ.Lit.424:1056 93 ReissAJ.1980.Understandingchangesincrimerates.InIndicatorsofCrimeandCriminalJustice:QuantitativeStudies.Washington,DC: Bur.JusticeStat. ReissAJ.1988.Cooffendingandcriminalcareers.InCrimeandJustice:AReviewofResearch,Vol.10,ed.MTonry.Chicago:Univ.Chicago Press Glaeser E,Sacerdote B,Scheinkman J.1996.Crimeandsocialinteractions.Q.JEcon.111:507 48 Uzzi B.1996.Thesourcesandconsequencesofembeddedness fortheeconomicperformanceoforganizations:thenetworkeffect.Am. Sociol.Rev.61:67498 Weisbuch,G.,A.Kirman,andD.Herreiner (2000)MarketOrganization,Economica 110:411436. Fafchamps M,LundS.2003.RisksharingnetworksinruralPhilippines.J.Dev.Econ.71:261 87 DeWeerdt J.2004.Risksharingandendogenousnetworkformation.InInsuranceAgainstPoverty,ed.SDercon.Oxford:OxfordUniv.Press Ryan,B.,andN.C.Gross(1943)TheDiffusionofHybridSeedCorninTwoIowaCommunities,RuralSociology8:1524. Griliches,Z.(1957)HybridCorn:AnExplorationintheEconomicsofTechnologicalChange,Econometrica25(4):501522. Coleman,J.S.,E.Katz,andH.Menzel(1966)MedicalInnovation:ADiffusionStudy,Indianapolis,Ind.:BobbsMerrill.

Week1:ReferencesContd
Milgram,S.(1967)TheSmallWorldProblem,PsychologyToday2:6067. Grossman,J.W.(2002)TheEvolutionoftheMathematicalResearchCollaborationGraph,inProceedingsofthe33rdSoutheastern ConferenceonCombinatorics (Congressus Numerantium,Vol.158). Newman,M.E.J.(2001)Scientificcollaborationnetworks. I.NetworkconstructionandfundamentalresultsPhys.Rev.E64,016131 Goyal,S.,M.vanderLeij,andJ.L.MoragaGonzalez (2006)Economics:AnEmergingSmallWorld,JournalofPoliticalEconomy 114(2):403412. Adamic,L.A.(1999)TheSmallWorldWeb,inProceedingsoftheECDL.LectureNotesinComputerScience1696,Berlin:SpringerVerlag. Backstrom,L.,P.Boldiy,M.Rosay,.J.Ugander S.Vignay (2012)``FourDegreesofSeparationarXiv 1111.4570v3 Erdos,P.,andA.Renyi (1959)OnRandomGraphs,Publicationes Mathematicae Debrecen6:290297. (1960)OntheEvolutionofRandomGraphs,PublicationoftheMathematicalInstituteoftheHungarianAcademyofSciences5:1761. (1961)OntheStrengthofConnectednessofaRandomGraph,ActaMathamatica AcademyofSciencesofHungarica 12:261267. JW Moon,L Moser ( 1966)``Almostall(0,1)matricesareprimitive,Studia Sci.Math.Hungar, BBollobs (1981)``DiameterofRandomGraphs, Trans.Am.Math.Soc.,1981 ams.org Chung,F.,andL.Lu(2002)TheAverageDistancesinRandomGraphswithGivenExpectedDegrees,ProceedingsoftheNationalAcademyof Sciences99:1587915882. Jackson,M.O.(2008b)AverageDistance,Diameter,andClusteringinSocialNetworkswithHomophily,arXiv:0810.2603v1,inthe ProceedingsoftheWorkshopinInternetandNetworkEconomics(WINE2008),LectureNotesinComputerScience,EditedbyC. PapadimitriouandS.Zhang,SpringerVerlag,BerlinHeidelberg. Golub,B.andM.O.Jackson (2012)``NetworkStructureandtheSpeedofLearning:MeasuringHomophilyBasedonitsConsequences, AnnalsofEconomicsandStatistics 107/108JulyDec.,2012.

Week1:ReferencesContd
Currarini,S.,M.O.Jackson,andP.Pin(2010)Identifyingtherolesofracebasedchoiceandchanceinhighschoolfriendshipnetwork formation,intheProceedingsoftheNationalAcademyofSciences,107(11):4857 4861 Bearman,P.,J.Moody,andK.Stovel (2004)ChainsofAffection:TheStructureofAdolescentRomanticandSexualNetworks,Chicago, UniversityofChicago,manuscript. Banerjee,A.,A.G.Chandrasekhar, E.Duflo, M.O.Jackson(2012)``DiffusionofMicrofinance,NBERWorkingpaperw17743 Marlow,C.(2009).MaintainedRelationshipsonFacebook. mimeo PriceDJS.1965.Networksofscientificpapers.Science149:51015 Pareto,V.(1896)CoursdEconomiePolitique,Geneva:Droz Yule,G.(1925)AMathematicalTheoryofEvolutionBasedontheConclusionsofDr.J.C.Willis,F.R.S.PhilosophicalTransactionsofthe RoyalSocietyofLondonB213:2187. Zipf,G.(1949)HumanBehaviorandthePrincipleofLeastEffort,Cambridge,Mass.:AddisonWesley. Simon,H.(1955)OnaClassofSkewDistributionFunctions,Biometrika 42(34):425440. Albert,R.,H.Jeong,andA.L.Barabasi (1999)DiameteroftheWorldWideWeb,Nature401:130131. MacRae,J.(1960)DirectFactorAnalysisofSociometric Data,Sociometry 23:360371.

You might also like