Building and Measuring the Social Networks of Al Capone
Articles Blog

Building and Measuring the Social Networks of Al Capone

Okay well, we’ve had quite a, a number of great talks today
by well established speakers, professors, members of
the National Academy of Sciences. Associate professors who will
very soon be full professors. So we should round things out by
having an assistant professor, but we’ve gone further. Chris Smith is currently completing her
Ph.D in Sociology from UMS Amherst, and she will be,
coming to Davis permanently this fall. And following Chancellor Katehi’s
discussion this morning about the longevity of academics,
we look forward to her being here for, as an active researcher for
another 40 to 50 years. [LAUGH] So come on up.>>[APPLAUSE]
>>I’m not that young.>>[LAUGH]>>Great, thank you. Thank you for that introduction and
thank you for having me. I start officially on July 1st.>>[LAUGH]
>>I also wanna thank my collaborative, collaborator Andrew Papachristos. This project is very much
a product of serendipity and of course a social network. I have two goals in my talk today. The first goal is to explain the process
and methodology of moving from physical and digital archives, to the early
1900 Chicago organized crime network, or what I call the social
network of Al Capone. My second goal today is to show
some of the analytic potential of this kind of project. But first, let’s start with
the historical case of org, of Chicago organized crime and
of Al Capone. Organized crime in Chicago and across
much of, of the, across the United States in the early 1900s revolved
around three illicit economies. Specifically gambling and
prostitution, and then in 1920, the introduction of the prohibition
of the manufacturing, sale, and transportation of
intoxicating beverages. Organized crime was very much
the syndication of graft payments, protection, extortion. The city of Chicago was divided
up into distribution territories. There was price gouging,
theft, double crossing. Being a member of organized
crime required armed protection. And these groups were strong-arming
themselves into union, politics, and protection rackets. Near the center of Chicago organized
crime was America’s best known gangster, Al Capone. For those of you unfamiliar with his
biography, Al Capone was born in Brooklyn, around 1919 he moved to Chicago. He started off as a doorman to a brothel,
but by the mid-1920s was the boss
of the Chicago syndicate. He was very much the epitome of
the rags to riches American dream, the American gangster. In fact, sociologist Robert Merton
actually cites Al Capone by name in his 1938 paper, Social Structure and Anomie as the example of social
assent via corrupt means. Al Capone, was not convicted of
prohibition crimes or murder. He was convicted of tax evasion toward
the end of prohibition, and received the longest sentence ever handed down at
that time by a judge for tax evasion. So the question then, is how do we turn
this fascinating historical case, and this fascinating individual into
an object of sociological analysis? It first requires contextualizing the case
of organized crime as a hidden population. And by this I mean we don’t know
the population, the total population, we don’t have a list of
the total population. So it’s impossible to draw
a truly random sample or a sample that would be representative. The activities of organized crime
are hidden from public viewing and from records. Hidden populations in a social network
framework are often refered to as dark networks. And this not dark,
just as in covert or clandestine, but dark as in the lights aren’t on. We can’t see everything going on. By definition outsiders do not and cannot know the total
structure of a dark network. In fact, even insiders often only
see their local network, and are oblivious to the complete
overall structure. Some examples of dark network research
include Wayne Baker and Robert Faulkner’s work on price fixing conspiracies in
the electrical equipment industries. Carlo Morselli’s work on New York City’s
Gambino crime family, Padahzur and Perlier’s work on suicide bombers, and Marc Sageman’s work on
the global Salafi jihad. And I mention these for examples, because
all of them are research on dark networks, but they all relied on open source coding
of publicly available documents to create these dark networks. Specifically, using court records,
newspaper articles, and historical books, which is precedent for what I’ll
be talking about in just a minute. There is however another motivation for
using social network approaches to the study of hidden populations or
to study these as dark networks. In some cases, social network analysis may be the only
way to study particular crime groups. Interviews, ethnographies and survey
methods require a level of detail and personal information that is not always
possible with certain criminal cases. Especially the case of historical crime,
organized crime, let alone historical or crime. It is comparatively easier however,
to get more benign information and data on how people are connected. Cases with thin individual level data
should not discourage us from pursuing the thick relational level data that’s
available in some of these cases. So, what I have here in this early 1900
Chicago organized crime case is a dark network, but it’s also a historical
dark network of closed criminal cases, and this does require archival data, and
some creative approaches to the archives. This project required
two sampling strategies. A withdrawal of the first
basically requiring us the second. But we started off with a random seed
method of Chicago Crime Commission archives. The Chicago Crime Commission
began in 1919, it was a civilian organization
of of business elite and their goal was to protect business owners
in Chicago from the criminal class. It still exists today and
they house a very large archive. They had an important role in this
project in that they were interesting, interested in working with us. So we requested a random selection
of their consolidation files and their public enemy files. These were sort of the summary
files that named lots of names and described lots of activities. The logic here was that we wanted
this random seed of documents to locate individuals and
relationships in organized crime and they actually mailed us about ten boxes
of these archives to UMass Amherst. I’ll get there. [LAUGH] So, so my job at the beginning of
this project was to scan these documents. Most of them were newspaper clippings,
but it also included some more primary sources such as investigator notes,
legal documents, letters. Copies of Al Capone’s arrest records,
original public enemy reports. The Chicago Crime Commission was
the first organization to use that label, public enemy, and create these reports of
who the public enemies of Chicago were. And entire files and sometimes even boxes
of documents on Al Capone such as this, inmate record from the Cook County jail. About nine months later they
moved to a new building and realized that these documents probably
shouldn’t be in Amherst, Massachusetts, so we returned them. I was able to visit the Chicago Crime
Commission in the summer of 2013, but only for a limited time. So we really needed to come up
with a second sampling strategy given our geographic restrictions. So the second strategy we
pursued was really focused on the Chicago Crime Commission’s
interest in Al Capone. So we treated Al Capone as our informant. He became the network ego for our study. The public fascination of Al Capone and
gangster era Chicago has preserved an eclectic collection
of archive and sources that open, and available to the public and
many of them have been digitized. So the strategy here was to search
the single informant across multiple archives to collect a corpus of
publicly available documents. To identify Capone his alters,
the relationships between them but also to rea, read beyond the Capone section of the
archive to identify additional alters and relationship between the alters. Examples of the documents in the second
informant based sampling strategy included lots of newspaper clippings. We had digital access to Chicago
Tribune Obituaries, bail bond cards, tax documents, court testimony,
a ship passenger roster to the Bahamas. Al Capone’s going to the Bahamas with
some Chicago politicians and bootleggers. And bloody fingerprinted letters written
in Italian that were death threats. [LAUGH]
This has nothing to do with Al Capone, but I always include this in my talks. [LAUGH] So here’s what I did. In total I visited four physical archives
all of them located in the city of Chicago. I accessed four digital archives,
and I collected some historical and contemporary accounts. In total coding over
4,800 pages of documents. And in the end we get this outsider
perspective of organized crime, but one that was recorded concurrently with
the activities of organized crime. We not so creatively called
the result the Capone database. We created the database in
Microsoft access because we wanted to take advantage of the relational
nature between the spreadsheets. We have spreadsheets on
individuals which I have a, a screenshot here of the dyads
between the individuals. Addresses so there’s spatial data here of
where the, these individuals hung out, the places that they owned. And also we link everything to
its original archival source, so a database on the sources. Coding for this project was manual and
whenever I talked to computer scientist they’re horrified that we
[LAUGH] did this manually. But we read every single document and every time a name appeared that person
entered the individual spreadsheet. We had to search to make sure
they hadn’t been entered already. And then we every time two or more names
were connected we created a dyad for those individuals. Individual level I, identify more
3,000 more than 3,300 people and whenever possible we included sex, year of birth, date of death, cause of
death, possible ethnicity and occupation. Though obviously there’s a lot of missing
data in some of those categories. Between these 3,300 individuals,
I identified 15,800 relationships or 15,800 dyads, but
what’s most interesting I think ab, you can see that I have some,
time stamps on some of the, especially when there’s events we
have a time, but more interesting is the different types of relationships
that I’ve identified in this database. I’ve over a 100 different
types of relationships. These include things such as families,
friendships, romantic associations, funeral attendance, criminal associations,
arrests, jail, and business associates. Some of these ties are directed,
such as paying someone else’s bail, giving someone else money or
shooting someone. Many of these are undirected
such as being brothers or traveling to the Bahamas together. Some of these ties are negative,
such as rivals, violence divorce. And then there are positive ties such as
friendships, political endorsements and mentorships. As a whole we get this massive
web of ginglang connections but the analysis does require subsetting
the data so we don’t have so much confusion in terms of direction,
positive, negative et cetera. So I subset the database to create theor,
networks of theore, theoretical interest to answer
various sets of questions. I briefly want to mention two projects
that have come out of this database so far. The first one is on network multiplexity, which you heard a little bit about
in the er, panel earlier today. Multiplexity is the network
property of two or more relationships existing
between a pair of nodes. And so in multiplexity research
the type of the tie or the content of the tie
is really important. So in this, this is just a fake example,
but if we imagine the black edges being criminal ties and the grey edges being
political ties, the only multiplex diet is between notes seven and eight because
they have both types of connections. In my organized crime research, I look
at these three discreet social worlds, the criminal world, the personal world,
and the legitimate world and then the overlap between those three. The criminal world includes things such
as co-arrests co, criminal associations, co-offending. The personal world is the informal world,
the friendships, the family members and
attending someone’s funeral. The legitimate world is the formal,
is formal but there is nothing criminal in that. So this is, this was actually
owning businesses together, being in unions together,
being politicians together. And then in the multiplex world,
we, so you, a multiplex tie show up here whenever you
had two or more of the other three worlds. So there were these two-plex ties and
these three-plex ties. And we were really interested in these
multiplex relationships because they’re in, they’re incredibly complex. In the corrupt and
violent context of organized crime, trusting your crooked neighbor
becomes really important. So who do you have those most
multiplex relationships with? These trusting and reciprocal ties. So one example of a multiplex
dyad that shows up in this data is Al Capone’s friend,
Daniel Serritella, who was a politician. So they have this friendship tie that
shows up in the personal network. Al Capone donated a lot of money
to Daniel Serritella’s campaign. That showed up in the legitimate network
cuz it was a legitimate campaign contribution. However, they were involved
in many scandals together, including diverting an entire truckload
of duck feed to Al Capone’s soup kitchen. So they show up in
the criminal network as well. So between this pair of men, they have all
three of these relationships represented. So in the analysis we wanted to figure out how exactly multiplexity was working
in this world of organized crime. We, I found out that multiplexity
was actually quite rare. Only 9% of the relationship, of the total
dyads were multiplex in this sample, but even though multiplexity was rare, it was incredibly relevant to the
undergirding structure of organized crime. And what we did to test this was look
at bi-variant exponential random graph models. That tested the dependencies
between these three networks. And we found that,
we found strong dependencies between them even when controlling for Al Capone
and other internal network properties. So though not pervasive multiplexing glues
these three roles of organized crime together above and beyond
the personalities of famous gangsters and other internal network properties. The second project coming
out of this database is my research on gender inequality and
organizational change. Here I’m looking at the shifting
structure of organized crime. And specifically what did, what happened to organized crime during
the exogenous shock of prohibition. So what we see in the first
network on the left. That is the pre-prohibition
organized crime network of the 19, I just picked 1900 as my starting point,
but the 19 years before prohibition, and then the 13 years of prohibition. In terms of the structure, what we see is that before prohibition
organized crime was obviously smaller, but it was also more decentralized, and
it had more of this clustered formation. During prohibition however
there was massive growth. The network came to include three
times more people, four torn, four times more ties. But it still became sparser. More and more centralized. So it became this more
monopolistic organization. What I find is that the shifting structure
of organized crime left women behind. When I look at the decentralized
clustered structure, women made up almost 20% of the nodes and
resided on about 20% of the ties. It was about the same as womens’
participation in the labor market during the early 1900s. However, during prohibition women
occupied only 4% of the nodes and resided on only about 4% of the ties. In addition to this descriptive statistic, I calculated gender gaps based on
structural positions within the networks. And obviously, there was gender inequality
in both, but what I found is that in terms of structural position, gender
and equality got worse over every single measure moving from the pre-prohibition
network to the prohibition era network. And here I’m speaking to an organizational
theory that suggests that networks are opportunity structures
containing resources. Minority groups have limited access
to resources in decentralized clustered structures. But resources are hoarded by, hoarded by the groups in power
in monopolistic structures. So as organizations change and
adapt to exogenous shocks, so to do categorical inequalities
within organizations. I wanted to just mention a few
limitations of the Capone database and this type of project. The first limitation is
obviously of missing data. Dark networks have to assume there
are missing data in our projects at the individual level and
at the relational level. And this happens when undocumented and unremarkable relationships
are under sampled. So the small time bootlegger, who maybe
showed up in the archives once, but never appeared again. But also it occurs when remarkable
relationships that escaped documentation are under sampled, such as
the really smart and undetected wife. Or the sneaky police officer,
those people who never got caught for their crimes never show
up in the archives. The second limitation of this da,
of this type of project is the seed bias. Al Capone certainly introduces
an informant bias to the sample or to to the database, but it is a bias that
captures the reality that family and friends friends and bootlegging
partners were not random events. And instead what we could do with this
data base is model Al Capone’s influence on models on the networks and identify
locations where he might matter less. These limitations do have implications for
analysis. The findings are likely underestimated and conservative, there
are potential spotlight effects. The effects that we find might be
an effect of the archival spotlight rather than an effect of the population. When you use exponential random graph
models, you are assuming complete data and incomplete data is at risk of
producing unstable predictions. However, I will argue that the benefits
of this type of project greatly outweigh its limitations. I have this fascinating case
of a unique historical moment that’s broad in scope and
broad in its content of the relationships. As researchers, we benefit from
using social network analysis as a logic of discovery for a partic,
particular set of events, group of people, and historical moment just as much
as a moment of a logic of proof. The future directions for this project,
this actually, this database is quite new. It took me eight years to build but
it is still new in terms of what we, we’re doing with it. The future directions
include a paper on the su, on the structural
insignificance of Al Capone. We do find that he doesn’t
appear to be a robust actor. He fails to reside on structural holes,
and he has many redundant ties. And we think that’s kind of cool, given
the, our sampling strategy and everything. We’re also doing a project on the violence
diffusion throughout the network. Over, in the next few years,
after we publish everything that we want, we are making the database
publicly available and we look forward to future collaborations. [LAUGH] Thank you. [APPLAUSE]

Leave a Reply

Your email address will not be published. Required fields are marked *

Back To Top