View
0
Download
0
Category
Preview:
Citation preview
Social Network Analysisin Historical Research
Claire Lemercier
National Center for Scientific ResearchCenter for the Sociology of Organizations
(CNRS-Sciences Po, Paris)
KULeuven, 21 September 2012
Social Network Analysisin Historical Research
• Doesn't exist? (in network terms)(but Redes, 2011,
Österreichische Zeitschrift für Geschichtswissenschaften, 2012, The Connected Past, etc.)
Social Network Analysisin Historical Research
• Doesn't exist? (in network terms)• What is historical research?
– That thing that historians do?– Research based on traces?– Longitudinal research?
Outline
1. Network? Analysis? What is it all about?2. Sources – What can be/has been done3. Time – What should be done4. Data and Software – How do I do it?
1. What is Social Network Analysis?
• A strong research tradition (with handbooks)• Not a school/theory
but not only a set of techniques/softwarea way to look at reality (a lens?)
• Not the only way to talk about networks/social ties
– cf. physicists, etc.– cf. "normal" historians
1. What sort of network? What sort of analysis?
• "Networks": a way to view things, not a separate type of things
• A network is not a group(but a set of ties and non-ties between entities)
1. What sort of network? What sort of analysis?
• "Networks": a way to view things• A network is not a group• There is no such thing as
"the network of an individual"(nothing we can "map", anyway)
1. What sort of network? What sort of analysis?
• "Networks": a way to view things• A network is not a group• There is no such thing as
"the network of an individual"• Software does not find networks in sources
you have to find them, and thus first to define ties that will be systematically recorded, i.e. abstracted
1. What sort of network? What sort of analysis?
• "Networks": a way to view things• A network is not a group• There is no such thing as
"the network of an individual"• Software does not find networks in sources• "To analyze" a network
=to find a structure/patternsat different scales
11
Some networks do not need much analysis...
… Others do
Lemercier, 2005
1. Why bother?• A fresh look at things• Find structure in a mess
borders! hierarchies!structure is created by
absent ties
1. Why bother?
• A fresh look at things• Find structure in a mess• Discuss ties
in a more symmetrical/systematic waydo the successful have many ties?do ties explain success?
1. Why bother?
• A fresh look at things• Find structure in a mess• Discuss ties
in a more symmetrical/systematic way• Test hypotheses
what do ties do? cf. Padgett and the Medici(but sometimes, ties do not do much – e.g. Fertig)
1. Why bother?
• A fresh look at things• Find structure in a mess• Discuss ties
in a more symmetrical/systematic way• Test hypotheses
what do ties do?where do ties come from?cf. Padgett and the Medici, part 2
1. Why bother?
• A fresh look at things• Find structure in a mess• Discuss ties
in a more symmetrical/systematic way• Test hypotheses• Graphs for exploration/communication,
indicators/statistics for tests
1. Thinking about ties
• Not "Is there a network here?" (there always is)
1. Thinking about ties
• Not "Is there a network here?" (there always is)• But: when looking for ties in sources:
– What is this tie? Who says so?– Is it a river or a riverbed (interaction vs. potential)?– Is it strong? In which ways?– When did it appear? When did it disappear?– Did actors consciously build/maintain/break it?
1. Thinking about ties
• Not "Is there a network here?" (there always is)• But: when looking for ties in sources (...)• And when interpreting results of network analysis:
– Is this tie/structure enabling and/or constraining?– What might local patterns (e.g. structural holes)
mean for this tie?
1. Thinking about ties
• Not "Is there a network here?" (there always is)• But: when looking for ties in sources (...)• And when interpreting results of network analysis:
– Is this tie/structure enabling and/or constraining?– What might local patterns (e.g. structural holes)
mean for this sort of tie?– How do different sorts of ties overlap/interact?– How do they interact with attributes of actors?
2. Networks from historical sources
• No scarcity of sources!
2. Networks from historical sources
• No scarcity of sources!• … well, it depends...• Not only ties between people
– Ties between places, words, artefacts, organizations...
– "2-mode networks", e.g. affiliationsperson-organization→ person-person→ organization-organization
Traces of interaction
Pierre Gervais, transcript of Gradis accounts
Armement du Navire La Renommée Dt a Divers £ 14188.3.6 Pour Le montant du dit armement non Compris 745 debitées cy devant lesquelles Jointes a cette Some de £ 14188.3.6 font £ 14933.3.6 debités cy Contre p la totalité du dt Armement Sçavoir à Caisse pour Journées & autres y Compris 2/m a l'Equipage £ 7858.15 à la Ve Coureau Forgeron 380.1 à Marchand pour quincaille 65.7(...)
Banchereau Dt a Chabbert & Banquet n/C £ 1416.13 p Sa traite Sur les dits pble a la fin d'Avril Suivant Sa Lettre du 19 courant
Mad d'Interville Dt; a Caisse £ 9.2 p fraix & droits de notre envoy d'une Caisse Liqueur & &
Pierre Gervais (in progress) from Gradis accounts
Networks from texts
Tilly, 1997
Düring 2012
Networks from lists
Düring 2012
Verbruggen 2007
Verbruggen 2007
Stuber & Krempel, forthcoming
Annuaire Desfossés, édition 1938
Thinking about sources...
3. The (historical) evolution/dynamics of networks
• Work in progress!• An atypical example: Isabelle Rosé's study of
Odon de Cluny
879
880
881
[...]
899
900
901
[...]
940
941
942
3. The (historical) evolution/dynamics of networks
• Work in progress!• An atypical example: Isabelle Rosé's study of
Odon de Cluny– An ego-network in the strong sense– Coding ties– Dating ties!
Do the oldest have more social capital?When do ties die?Do dead people have ties?
– Representing change: what do we move, what do we keep in place?
3. The (historical) evolution/dynamics of networks
• Work in progress!• An atypical example: Isabelle Rosé's study of
Odon de Cluny• Alternative strategies
– A time-scale
Sigrist & Widmer2011
3. The (historical) evolution/dynamics of networks
• Work in progress!• An atypical example: Isabelle Rosé's study of
Odon de Cluny• Alternative strategies
– A time scale(by the way, you can also project on a map...)
– A time scale with an innovative view of organizations
François, 2009
3. The (historical) evolution/dynamics of networks
• Work in progress!• An atypical example: Isabelle Rosé's study of
Odon de Cluny• Alternative strategies• Modeling (micro)dynamics,
not only describing change?"Siena" modeling (based on simulation)
Lemercier & Rosental, 2010
Changes from migration ca. 1860 to migration ca. 1880
Rate parameters: 0. Rate parameter 17.0456 ( 1.0511)Other parameters: 1. eval: outdegree (density) -1.8468 ( 0.1625) 2. eval: reciprocity 0.4143 ( 0.1180) 3. eval: transitive triplets 0.1041 ( 0.0143) 4. eval: 3-cycles -0.0895 ( 0.0426) 5. eval: betweenness -0.0951 ( 0.0183) 6. eval: same language 0.9669 ( 0.1304)
4. Data formats and software• Good news: generic, manageable data formats• Not-so-good news: a complex, evolving
software landscape(but you can do it!)
4. Data formats and software• Good news: generic, manageable data formats
Edgelists (lists of ties)
4. Data formats and software• Good news: generic, manageable data formats
Edgelists (lists of ties)Generally accompanied by an "attribute file"
4. Data formats and software• Good news: generic, manageable data formats
Edgelists (lists of ties)Generally accompanied by an "attribute file"Also works for "2-mode networks"
(software converts to 1-mode)
4. Data formats and software• Good news: generic, manageable data formats
Edgelists (lists of ties)Nodelists
one line per person, all his/her contacts on the lineor: list of members of the organization!
4. Data formats and software• Good news: generic, manageable data formats• Not-so-good news: a complex, evolving
software landscape– Zillions of possibilities (esp. visualization), often free– Communities to guide you
– Excel plugin, basic functionalities: NodeXL– Nice visualizations (but not-so-easy data entry?):
Visone– Very easy to use for beginners, great for visual
exploration of (not-too-big) data: Netdraw
4. Data formats and software• Not-so-good news: a complex, evolving
software landscape– Excel add-on, basic functions: NodeXL– Nice visualizations: Visone– Very easy to use for beginners: Netdraw– Classical SNA calculations:
Ucinet (non-free software - but cheap+community)– If you have larger networks (> 5000 "nodes"): Pajek– For modeling (and/or geeks):
assorted R packages (sna, igraph, statnet, RSiena...) / pnet
claire.lemercier@sciences-po.org
Historical Network Research website
The Networks Network
Recommended