[MUSIC PLAYING]

EDWARD LAUMANN: Hello, my name is Edward O. Laumann,and I'm a professor at the University of Chicagoin the Department of Sociology.Now, I'm going to first introducethe key elements of social network analysis, whichhas become a very important methodological tool,a theoretical tool for that matter, in sociology.

EDWARD LAUMANN [continued]: So I'm going to be covering the following points--the analytic elements in social network analysis very briefly.I'm going to then turn to the National Social Healthand Social Life Survey, and we'regoing to be talking about numbersof sexual partners, social partnerchoice, and random mixing models versus social choice models,

EDWARD LAUMANN [continued]: the sociomatrix of choice in heterosexual population,estimating persons with AIDS from a population survey,and disparities in STIs by race, a network explanation.The last three items are really intendedto illustrate the power of this network approach.

EDWARD LAUMANN [continued]: Now, I'm going to, first of all, turnto the analytic elements in social network analysis,and we need to start with, of course,what is a social network?And then illustrate graphically howwe might represent these forms of relationshipsand the structural features of networks and their functions.

EDWARD LAUMANN [continued]: J. Clyde Mitchell was an anthropologistwriting a book Social Networks in Urban Situations in 1969.And the chapter he wrote on networks has become a classicand is a very good point of departure for our purposes.He offers the definition as, "A specific set of linkages

EDWARD LAUMANN [continued]: among a defined set of persons with the additional propertythat the characteristics of these linkages as a wholemay be used to interpret the social behaviorof the persons involved."Now, as you will note from the Mitchell definition,this is a very concrete anthropologist.He's studying real people in real situations and so on.

EDWARD LAUMANN [continued]: And this, unfortunately, puts a very major limitationin its utility as an approach for studying large populationsbecause, obviously, when we're talking about millionsof people living in a city or in a country,we are not going to be able to go very

EDWARD LAUMANN [continued]: far with that type of approach.So I proposed in 1973 a modificationof that definition as follows.It's a set of nodes, and nodes are essentiallypoints that could be persons, theycould be corporate actors like organizations,they could be socially defined categories such as status

EDWARD LAUMANN [continued]: groups like a race or ethnic groupor social classes, which again, involvelarge numbers of persons that arelinked by a set of social relationships.Social relationships are what Mitchell's linkages areand of a specified type.So social relationships could be neighbors,

EDWARD LAUMANN [continued]: they could be friendship, they could be confidants,they could be any number of operationswhich describe the nature of the relationship or the interactionthat characterizes the connection between twopersons or two nodes.Now, there are four functions of social networksthat can be readily identified.

EDWARD LAUMANN [continued]: You talk to people to exchange information,to find out something that the other person may notknow that might be of value to themin pursuing some purpose or goal that they have.It exercises social control.That is, you can have networks thatexpress disapproval or approval of a course of action

EDWARD LAUMANN [continued]: that ego is undertaking.Ego, by the way, is a convention referring to the referent actorin our discussion.You can provide socio-emotional support.You can talk to somebody, and he comes homefrom a bad day at the office, and youexpress support and concern about how terrible

EDWARD LAUMANN [continued]: it must have been and try to make him feel betterin those circumstances.Finally, you can provide for the exchange of goods or servicesor other physical entities, such as pathogens, as in sexuallytransmitted infections.Now, there are many features of social networksthat have attracted interest analytically.

EDWARD LAUMANN [continued]: Certainly, there is size.That is, how big is the network?Some people maintain contacts of a certain relationshipwith many, many people while othersare very parsimonious in the personsthat they would recognize as their friends.So the list of features here are different aspectsthat you could readily think about that

EDWARD LAUMANN [continued]: would affect intimacy, how tight do you feel with that person,how often do you see them?These are various features that aregoing to characterize the robustnessand the fluidity of these networks.Now, here is an example of a graphic representationof a network.

EDWARD LAUMANN [continued]: Ego is the focal actor, and he has named A, B, and Cas his best friends.And A, B, and C have, in turn, connections with others thatmight be their best friends.And notice that there is no connections among A, B, and Cthemselves.

EDWARD LAUMANN [continued]: They only are connected through ego.This is called a radial network because egois the broker that connects these three disparate actorswith one another.In the second network, the ego-centered network,the interlocking network, we see that A, Band, C are connected to each other in friendship

EDWARD LAUMANN [continued]: as well as having some other people that they feel close to.This really looks like a clique or a setof persons who share a lot of direct ties with each other.And a clique is a very different kindof group formation or structure than thatof the radial network.One could easily imagine that in a radial network,

EDWARD LAUMANN [continued]: ego is really free to massage his relationshipswith his others as he likes.He can tell A things that are notconsistent with what he tells B, and A and Bare not going to know the differencebecause they don't communicate.Whereas, ego is under constraint about whathe can say to A, B, and C because they

EDWARD LAUMANN [continued]: are in communication with one another.And as a result, we would predict, among other things,that people in interlocking networksare going to have more attitude consistency, for example,than people in radial networks.A socio-centric network, and these arrowsrepresent the flows of information or whateverof interest, suggest, for example,

EDWARD LAUMANN [continued]: that F is a mediator that connects a lot of peopleto one another.And socio-centric network is essentiallylooking at all the information from the ego-centered networksexhaustively belonging to the set of persons thatbelong to this group, and we're nowlooking at the pattern of the whole structure of the system

EDWARD LAUMANN [continued]: rather than that of individual actors.In an ego-centered network which connects to other people,you can count the relationship as a tie or path from oneto another.So you can talk about a person being connected two steps.That is, he has to go to A and then

EDWARD LAUMANN [continued]: to some third party in order to get to that destinationposition.So path distances represent how faryou may be away from the source of informationor something of interest.The data that we're going to be spending some time with today

EDWARD LAUMANN [continued]: is the National Health and Social Life Survey.This was a survey that was done in 1992 in responseto the AIDS epidemic that had ravagedthe country during the 80s where the numbers of personswith AIDS was doubling every 10 months for a period of time.Here we're trying to describe a probability

EDWARD LAUMANN [continued]: sample of, in this case, about 3,200 persons between the agesof 18 and 59 who live in householdsthroughout the United States in 1992.Now, probability sample is a very important ideathat we need to distinguish from a convenience survey.

EDWARD LAUMANN [continued]: A probability sample is a sample in which we have identifieda population of interest, the adult population in the UnitedStates in this age range, 18 to 59,and we want to draw a sample of individualswho will represent or stand for those the population at large.

EDWARD LAUMANN [continued]: What we need to have is a situationin which every person in the United Stateshas a known probability of being included in our survey.And it's not a random sample because a random samplewould say that everybody had an equal probability of beingselected.Most of the time we're interested

EDWARD LAUMANN [continued]: in particular categories within that larger population.Let's say we're interested in African Americansor under-represented populations so we can have a largerprobability of inclusion for people of people of interestor groups of interest.So the probability sample is an unbiased selection,which is representative of the general population

EDWARD LAUMANN [continued]: and gives you a very accurate estimateof the percentages of the prevalencesof various kinds of things.The reason why we wanted this thingis because we were interested in preferencesof various kinds of behaviors.For example, partner turnover or the choice of partners or whatyou did with your partner.These are all matters of some concern,

EDWARD LAUMANN [continued]: and we needed to know the probabilityof these things occurring in the general population.So the comprehensive collection of informationon sexual practices and beliefs and attitudes is key here.We did not have any priors about whatis a good or bad idea or a good or bad attitude.The issue has to do with being clear

EDWARD LAUMANN [continued]: about what are the behaviors you want to measure,and are you adequately measuring them and representing themin this fashion?One of the outcomes or results of our survey, whichwas, as I said, about 3,400 people,was that we asked men and women a variety of questions

EDWARD LAUMANN [continued]: about who their sex partners were,what were their characteristics, how did they meet them,and a variety of issues of this kind.And we also certainly was able on the basis of interrogatingthem about all their sexual partners over their timeto be able characterize the partner turnover,sexual partner turnover of individuals.

EDWARD LAUMANN [continued]: Now, what is really interesting isthat in the sample of 18 to 59 year olds,the median number of partners that women reported was two,and the median number of partners that men reportedwas six.That means that half of the population had,

EDWARD LAUMANN [continued]: in the case of women, had two or fewer partners,and for men, six or fewer partners.There was then a long tail of otherswho'd had multiple partners that were larger than that number.But it's worth paying attention to the factthat this number is quite small, and it is not odd or aberrant.

EDWARD LAUMANN [continued]: It's, in fact, very typical because we had at the same timesurveys done on much larger samples in Englandand in France in 1992, to 18,000 people in both cases,and these numbers were essentiallythe same or very similar.There were very low differences in the number of partners.

EDWARD LAUMANN [continued]: In a sense then, the heterosexual populationin the United States is characterized-- in general,in the developed world is characterized by a relativelysmall number of partners.I later did a survey in China, a representative sampleof the survey in China, and in that case,the modal number of partners was one for both men and women.

EDWARD LAUMANN [continued]: Well over 85% of the women in Chinareported one lifetime partner.And for men, it was about 80%.So we're looking at very different characteristics,but they're small in frequency.

EDWARD LAUMANN [continued]: The issue of interest here is partner choice.Social networks lead you to pay attentionto how do people pick others to engage in whateverthe relationship is.And in general, we're going to say that equal statuscontact is the driver.If you have a choice, that is you can freely associate

EDWARD LAUMANN [continued]: with whoever you like and that either party has to, of course,agree to associate with you, so there is a mutual restrainton who picks whom, the general result-- and this has been donein all of the studies that I've done earlier on friendshipformation in the United States-- is that people pick others who

EDWARD LAUMANN [continued]: are similar to themselves on major social characteristics,such as age, gender, education, achievement, religion,political party preference, and so on.So these are-- what we say, the principleis that we pick others who are similar to us in social status,

EDWARD LAUMANN [continued]: and therefore, equal status contactis the basic driver that organizesthe way we choose others to associate with respectto age of the master statuses.To flesh out some of this argument about numberof partners, we can see here a tablethat lists in the first row a column, the number of partners

EDWARD LAUMANN [continued]: that range between 1 and 21 or more.And if you look in the last column,you will see the percentage of the general population thatare in that category, only 9% of the general populationhad as many as 21 partners or more partnersin their lifetime.And the internal two columns describe the likelihood

EDWARD LAUMANN [continued]: of reporting that you had ever had a sexuallytransmitted infection.And you will notice that the men generallyhave-- for every level of risk because, as you can see,the more partners you have, the more likelyyou're going to report a sexuallytransmitted infection-- you will see that men, in general,

EDWARD LAUMANN [continued]: on average have lower numbers for every given level of riskthan women.Now, from a network point of view, what we're going to sayis that one way to explain that differentialis that women pick partners who are riskierbecause men have higher numbers of partners in their lifetime.

EDWARD LAUMANN [continued]: And men pick partners who are less risky,that is women who don't have as many partnersin their lifetime.There are certainly other reasons why transmission mightbe facilitated in women for biological reasonscompared to men, but certainly one of them

EDWARD LAUMANN [continued]: is the network question that might account for this.Now, if we were talking about a random mixing model,we would be saying that there is no bias in the way youwould pick a sex partner.You would just simply pick somebodywho is the appropriate gender to have sex.

EDWARD LAUMANN [continued]: And this picture here graphically representsthat network.About 10% to 12% of those circles are singletons.That is, they did not report everhaving a partner in the past year.The dyads are connected where they are in couples,

EDWARD LAUMANN [continued]: and they are mutually exclusive for the last year.And then you have some people who report multiple partners,and you see chains occurring in that network.Now, it's easy to demonstrate that if the host human being isnot turning over partners very rapidly,

EDWARD LAUMANN [continued]: then the chances of infection, of the pathogen movingfrom one person to another is very limited.They're stuck there for the rest of their lives, if you will,of the pathogen's life.And it would simply result in sexually transmitted infectionsdisappearing from the population over several generations.But this is not the case, of course.

EDWARD LAUMANN [continued]: And the reason for that is because the choice of partnersis highly organized socially.We have four categories of relationships.We could talk about marriage.We could talk about being co-habers,that is, living together with a partner.You can be a long-term relationship, thatis you're having sex with a person,

EDWARD LAUMANN [continued]: but are not living with them.Or you have a one night stand or a short-term partner.What is incredible is how similarthe biases are with respect to partner choice,with respect to religion, to ethnicity, to religion,and age.So that all of them are really recruiting from people who

EDWARD LAUMANN [continued]: are very similar to themselves.So even a one night stand, which is supposedly with a stranger,is a stranger who, in fact, shares a lotof your social characteristics.So basic point we're trying to makeis that there's high social organization of their choice.Now, this table is called a sociomatrix.

EDWARD LAUMANN [continued]: And in the first set of entries on the rows,you have a list of the people characterizedby their age, gender, and race.And if you look across the row, theyare distributing their sexual partnershipsacross the 18 categories of choices that are available.

EDWARD LAUMANN [continued]: And as you will see, most of the actionoccurs on the main diagonal.I mean, the main diagonal is, in fact, a very small numberbecause that's a homosexual contact, a choiceof a same-gender person, whereas most of the actionoccurs just off the main diagonal whereheterosexual partner choice occurs.

EDWARD LAUMANN [continued]: So the only violation of the equal status contact ruleis the heterosexual preference informingpartners in this population.Now, a further point here then isthat the concentration of social choiceis so strong on the main diagonal

EDWARD LAUMANN [continued]: that there are many, many cells that are empty.That is, that there simply is no sexual transactions occurringbetween people in the row and the column.That means that there is no opportunityfor sexual transmission to occur.We were, on the basis of this and other considerations,able to conclude that the chances

EDWARD LAUMANN [continued]: of a heterosexual epidemic of AIDS in the general populationis zero.There is simply no mechanism thatwould allow for the spread of the disease in that fashionacross the general population.This table is just a stylized version of the other.A little easier to see the main diagonalization effect

EDWARD LAUMANN [continued]: to make the point.Now, thinking about networks again,we were faced with the issue of, howare we going to count people who are relativelyrare in the population, like persons with AIDS?The CDC does this by going to the boards of health

EDWARD LAUMANN [continued]: across the country and getting them to report--and this is a reportable disease-- all the peoplethat have been reported to them whohave AIDS-- infections of HIV.And they accumulate that in Atlanta and by regionand so on and so forth.

EDWARD LAUMANN [continued]: This is a case-by-case kind of approach,and we're interested in a person approach.So this is a bit of a problem, as we will see a little later.But in this case here, we decidedto try a different approach.We took a sample of 1,500 people from various other studies

EDWARD LAUMANN [continued]: that have been done.We estimate that the average person knows 3,000 people.So that about 4 and 1/2 million peoplewill be scanned if they're asked the question,do you know anybody who has AIDS,or do you know anybody who was murdered in the last year,or do you know anybody who committed suicide

EDWARD LAUMANN [continued]: in the last year?Now, the Uniform Crime Reports of the FBIand the vital statistics count suicides and homicidesand characterizes them by local, age, and gender.We are able to reproduce the distributions of homicide

EDWARD LAUMANN [continued]: and suicide victims from this process that we were describingof asking a cross-section of people,whereas we cannot do that for AIDS.There is major misrepresentationsof the distributions of AIDS in the general population.

EDWARD LAUMANN [continued]: For example, the East Coast and the West Coastare represented as having, in the CDC numbers,very high or elevated rates of the HIV.In fact, people living in the East Coastdidn't know people who had AIDS beyond their proportionsin the population.It was elevated in the West Coast.

EDWARD LAUMANN [continued]: But in Chicago, which had many of the factors that supposedlyencouraged the spread of the disease,they were, in fact, substantiallybelow the national average.So here was a major transport hubthat was presumably a major facilitatorof the spread of the disease, which

EDWARD LAUMANN [continued]: failed to register as having even average level of AIDSin the population.Whereas when we asked people living in this area,they knew people who had HIV, and itwas proportional to their numbers in the population.So this is a way of using networked ways of thinking

EDWARD LAUMANN [continued]: to supplement and complement some of the other waysthat we have for measuring these disease.So taking up another matter, the CDCcollects, as I said before, reportable sexuallytransmitted diseases.In this case, it's been well-known,as the data above on this slide shows you,

EDWARD LAUMANN [continued]: that African Americans are very much more likely to reporthaving gonorrhea than the white population, whichis at the bottom.And there's been various efforts to try to explain that.None of them are very compelling.Certainly, it doesn't have to do with minority or poverty status

EDWARD LAUMANN [continued]: since the Hispanic population is completely similar to thatof the white population with regardto the lack of gonorrhea in those populations.In Chicago, we had a cross-section surveyof the general population.And we asked people whether a health professional hadtold them that they had an STI.

EDWARD LAUMANN [continued]: And as you can see in the second column,there was a substantial number of peoplewho reported this, but were diagnosed by doctors.And the little blue at the bottomrepresents public health clinics.In Atlanta, the number of cases having a diagnosis of gonorrhea

EDWARD LAUMANN [continued]: was reported 80% of the time by public clinics rather.Now, this means that our image of the sourcesor the distribution of this diseaseis very biased because these are public health clinics, whichare free, which tend to serve minority populations,

EDWARD LAUMANN [continued]: and which are very likely to report STDsto the Centers for Disease Control.So we're looking at a situation whereour image of the distribution of STDsis very much biased toward minorities havinga lot of it when, in fact, if we were looking

EDWARD LAUMANN [continued]: at it from the standpoint of diagnosisby the health care system, we wouldsee the vast majority of it are done by doctors,and these are often privately paid for,and therefore, the image of where the disease is foundin the general population is very misrepresented

EDWARD LAUMANN [continued]: in this strategy of measuring STDs.Now, we've introduced before the ideaof a random mixing population, and weargued that, in fact, there is a very strong tendencyfor some folks to be very similar to each other

EDWARD LAUMANN [continued]: with regard to the behavior of, in this case, sexual practicesthemselves.So we can describe sort of three categories of the populationbehaviorally.There's the peripheral population,which runs about 80% of the population wherethey have no sex partner in the past year or only one.The adjacents are people who have two or three partners

EDWARD LAUMANN [continued]: in the last year.And the core group, which is about 5% of the population-- 4%have four or more partners in the past year.Now, in order for the model for representing the distributionSTIs to work, there has to be a very strong biasin favor of going to people like yourself behaviorally.

EDWARD LAUMANN [continued]: You will find that as you move from peripheralto adjacent to core, you become more racially exclusive,so that the white core group has very little sexual transactionswith black core group and vice versa.So the explanation that we would offerfor the concentration of gonorrhea

EDWARD LAUMANN [continued]: in the African American populationhas to do with the fact that the core groupin the black community happens to be seeded with gonorrheaand passes it among themselves.But there is no broker or network mechanism

EDWARD LAUMANN [continued]: that would allow them to pass it to the general populationbecause they are so much confinedto their particular racial group.Turning to another example of a network story,I mentioned earlier that in Chinathe vast majority of people only have one lifetime partner.

EDWARD LAUMANN [continued]: Now, the question has been, how has sexually transmitteddiseases moved in this population given that there'ssuch a low turnover rate?It happens that about 30% of men in Chinahave sex with commercial sex workers.That number is very low in the United States.

EDWARD LAUMANN [continued]: It's less than 1%.So that's a difference right away.The Chinese public authorities wereconvinced that the major transformation of China that'sgone on for the last 25 years meantvast movements of rural population to the cities

EDWARD LAUMANN [continued]: to help build the cities and infrastructure and so on.They were not usually allowed to live within the city itself,so they often lived in the periphery and commuted in,and that they were the culprits in the rapid rise of STDsin this population.We have a different image.Chlamydia, which we measured by taking urine samples

EDWARD LAUMANN [continued]: from the population, both men and women--and we had to do that because people could notreport that they had chlamydia since it's 90% asymptomatic,and they didn't know they had it.And what we see here is arrayed the likelihood or the riskof reporting having chlamydia by what the male income was

EDWARD LAUMANN [continued]: of your partner ranging from the lowestdecile to the top decile.And as you can see, for women, people in the very top decilehave a close to 40% likelihood of having chlamydia.So that the greatest risk for committeeor are for high-status women, or at least women associated

EDWARD LAUMANN [continued]: with high-status men to contract the disease.So a public health intervention that we would suggestis that you go to airports and not to bus stationsto do intervention work of this kind.This again, shows you the power of a network explanationfor these things.

EDWARD LAUMANN [continued]: So what I've tried to do is to demonstratehow social network connectivity differs across populationsand how these differences help to explain the differentialspread of STDs or STIs and that more work on thinking about howthe sexual networks are organized

EDWARD LAUMANN [continued]: will pay very high dividends in helpingus to explain and possibly to help intervene in controllingthe spread of these infections.We have a set of questions here thatwould be helpful in maybe pulling togethersome of the ideas that have been presented in this talk.One is, what are the essential elements

EDWARD LAUMANN [continued]: of social networks as a tool for explaining the spread of STIs?How would one contrast random mixing modelswith social ordered models of partner choice?What are the social factors likely to influencesexual partner choice among heterosexual couples?What is a sociomatrix?

EDWARD LAUMANN [continued]: And why should you prefer a probability-based samplingdesign over a convenience sample?