Welcome to the tenth week of the undergraduate social networks at the University of Maine at Augusta, in which we consider the role of social networks in politics. Politics is a collective act in which people deliberate and answer the question, “What is to be done?” The answer to that question is some form of policy. In order to influence the answer to the question “What is to be done?” movers and shakers attempt to shape the answer to a prior question, “Who is to decide what is to be done?” through various appointments and elections. In the wake of a monumental presidential election last year and its aftermath in the Congress and state legislatures of 2017, these questions loom not only in hearing rooms and conference rooms and courtrooms across the country but in campaign events, in living rooms and ultimately in voting booths. If for better or worse politics determines what is to be done, with lives and livelihoods hanging in the balance, then to the extent social networks shape politics social networks matter a great deal.
This lecture touches on a single, but dense, piece of assigned academic reading: James Fowler’s “Legislative Cosponsorship Networks in the U.S. House and Senate.” This reading is written at a professional level and are thick with meaning; as a result getting through it can be a challenge this week. You may find it useful to consult lecture materials before you dive into them to provide some context. Be sure to take the time needed to understand what the authors are doing and what the authors find; their ideas may be of good use to you as you start thinking about upcoming homework assignments and research projects.
Our lecture subjects this week are:
- A Shift in Homework Expectations
- A Variety of Political Relations and Affiliations (from a variety of sources)
- Unpacking Cosponsorship Networks
- Assignment: Finding 2-Mode and 1-Mode Relations in the Maine State Legislature
A Shift in Reading and Homework Expectations
As you may have noticed, the readings and homework assignments of September and October were occasionally rote and technical in nature, designed to take you through a series of tasks necessary for you to use the tools of social network analysis. But in the last period of this course gets underway, you’ll notice that the nature of the readings and homework assignments has changed. You’re going to start reading professional social network research articles now, and that kind of reading is more difficult. In your homework, I’ve been starting to ask you to make a series of choices about the objects of your study, and with some skills under your belt I’ll leave it up to you to identify meaningful network patterns. The majority students in this class are doing well in making this adjustment, but some of you haven’t yet made the leap. If you’re struggling to adjust, please get in touch with me and let me know. It’s normal to struggle a little bit with this kind of material — persistence through struggle is how we grow. I’ll do my best to lend a hand if you give me a bit of advance notice.
Guidance: On Reading a Research Article
This week, I’ve asked you to read just one academic research articles. By page length, it doesn’t look too bad, but there’s a lot of information packed in that article. Here’s how not to read a dificult research article:
- Pick up the article and read it from beginning to end.
- Be sure you understand every word written in a sentence before moving on to the next sentence.
- On the second page, throw down the article in despair.
Research articles aren’t meant to be read like other pieces of writing. Instead, they’re meant to be read in pieces by different audiences. I’ll tell you a secret: almost no professional reads a research article straight through from the front to the back! Instead, we drop in to selected parts of the paper to get oriented and obtain a quick sense of the paper’s mission and findings. Only then (and only if we find material of value to our inquiry) do we read the rest of the paper to obtain the necessary context and evaluate its quality. Experts will certainly read the entire article, but newcomers to a subject may gain knowledge by only reading a few parts of the article. Remember that the vast majority of readers will be unable to understand some aspects of a article. Your job is to gain as much understanding from an article as possible. Try following these easiest-to-hardest steps in order:
- Read the title to get a rough idea of the subject of the article.
- Read the abstract to understand what the article is trying to accomplish.
- Skip to the discussion or conclusion section of the article and see if you can figure out what the authors found in their research.
- Head back to the introduction and literature review to identify the dependent and independent variables in the author’s research.
- Advance to the methods section (at least in a research paper like James Fowler’s) and note how each of the variables is specifically measured. In the first round of reading, ignore technical details.
- Take a quick look at research results. What are the broad patterns in results?
- Start asking critical questions. A few examples:
- do the authors honestly and completely describe the current state of knowledge regarding their research question?
- are there hypotheses in this article predicting the ways in which variables affect one another? If so, what are they?
- do variables actually measure what the authors say they measure?
- do the results described in the results section of the article support the authors’ conclusions?
- To answer those critical questions, reread the paper more closely.
If you proceed in this order, then even if you don’t make it to the final step you’ll have gained some essential understanding regarding the research described in the article. Understanding and successfully reading a research article isn’t a simple task; it’s a skill that has to be practiced for you to become proficient at it. But with practice, you’ll be surprised at how quickly your comprehension of research articles improves. You’ll have taken a significant step toward becoming a scientifically literate consumer of knowledge, independent of major media spin.
A Variety of Political Relations and Affiliations (from a variety of sources)
Traditionally, politics has been thought of as a realm of striving individuals who think very hard and make rational calculations based on interests, but recently social network analysis has revolutionized our understanding of political behavior. For example, what if the Congress is not a unique set of exceptional leaders, but just another group of people building and maintaining ties to one another like any other? What if businesses and political action groups trying to influence the Congress can be thought of as networks too? What if these networks influence our political thought and action in ways that aren’t strictly rational?
Political network analysts inside academia spend their careers trying to find out the answers to these questions, and political activists quickly move to apply academics’ insights in order to explain people’s behavior in the past and predict what people will do in the future. Read this article from the October 2012 New York Times to find out how the Obama and Romney presidential campaigns applied social network insights in the last U.S. presidential election to great and perhaps unsettling effect:
In interviews, however, consultants to both campaigns said they had… authorized tests to see if, say, a phone call from a distant cousin or a new friend would be more likely to prompt the urge to cast a ballot…. “You don’t want your analytical efforts to be obvious because voters get creeped out,” said a Romney campaign official who was not authorized to speak to a reporter. “A lot of what we’re doing is behind the scenes.”
In the weeks before Election Day, millions of voters hear from callers with surprisingly detailed knowledge of their lives. These callers — friends of friends or long-lost work colleagues — typically identify themselves as volunteers for the campaigns or independent political groups.
The callers are guided by scripts and call lists compiled by people — or computers — with access to details like whether voters may have visited pornography Web sites, have homes in foreclosure, are more prone to drink Michelob Ultra than Corona or have gay friends or enjoy expensive vacations.
The callers are likely to ask detailed questions about how the voters plan to spend Election Day, according to professionals with both presidential campaigns. What time will they vote? What route will they drive to the polls? Simply asking such questions, experiments show, is likely to increase turnout.
After these conversations, when those targeted voters open their mailboxes or check their Facebook profiles, they may find that someone has divulged specifics about how frequently they and their neighbors have voted in the past. Calling out people for not voting, what experts term “public shaming,” can prod someone to cast a ballot…. “I’ve had half-a-dozen conversations with third parties who are wondering if this is the year to start shaming,” said one consultant who works closely with Democratic organizations. “Obama can’t do it. But the ‘super PACs’ are anonymous. They don’t have to put anything on the flier to let the voter know who to blame.”
The Obama and Romney campaigns, as well as affiliated groups, have asked their supporters to provide access to their profiles on Facebook and other social networks to chart connections to low-propensity voters in battleground states like Colorado, North Carolina and Ohio.
When one union volunteer in Ohio recently visited the A.F.L.-C.I.O.’s election Web site, for instance, she was asked to log on with her Facebook profile. Computers quickly crawled through her list of friends, compared it to voter data files and suggested a work colleague to contact in Columbus.
In politics, networks matter, networks are being tracked and network data are being used. If you have an inclination to build a career in politics, an understanding of social networks is essential.
You should recall from earlier lectures this semester that social networks can be characterized and utilized in very complicated ways, but arguably boil down to only two very simple sorts of information:
- One-mode information: Ties between nodes
- Two-mode information: Membership of nodes in events, organizations, categories or actions
In short, networks are about relations and affiliations, and politics is chock-full of those.
The political sphere provides some of the richest social network data to be found because the law mandates that information about politicians’ ties, events, organizations, activities and categorical identifications be made public. This information is made public to promote civic participation and for the sake of history, but as a side benefit it is available to social network researchers who either are interested in the substance of politics or are interested in finding network information on which they can test their general theories.
At the national level of United States politics alone, the following are just a few of the many available sources of information about political ties and affiliations:
- White House Visitor Records (2.6 million events at last count)
- Federal Contractor Data
- White House Advisory Board and White House Fellow memberships
- Congress.gov, a searchable database of bill sponsors and cosponsors (see James Fowler’s article in this week’s readings)
- The Congressional Record of official legislative remarks
- House of Representatives: members, leaders, committees and votes
- Senate: members, leaders, committees and votes
- Lobbying Disclosure: Lobbyists, Lobbying Firms, Lobbying Clients
- Candidate and Committee Receipts and Spending
- Individual Campaign Contributor Search
- Independent Expenditures
- Section 527 advocacy group financial disclosure
- Securities and Exchange Commission EDGAR database: Executive Compensation and Directorate shares held via proxy statements (see “DEF 14A” reports)
- Executive List and Board of Directors List for Interlocking Directorate research (see Mark Mizruchi’s article in this week’s readings) via Lexis-Nexis database (log on with University of Maine system library card, search under “Get Company Info,” then download tables in “Top Executives” and “Board of Directors” sections)
Unpacking Cosponsorship Networks
As you shift in your undergraduate work to reading more professional pieces of writing, the investment required to read an 11-page paper shifts dramatically upward. James Fowler’s 11-page “Legislative Cosponsorship Networks in the U.S. House and Senate” contains deep meaning in nearly every sentence, so to understand it you need to read slowly, think carefully and interpret deeply. To help you in your work, let’s look at a few tricky elements.
Sponsorship and Cosponsorship
This distinction is actually pretty easy, but also it’s pretty important. When a bill is proposed by a member of Congress who would like it to be passed into law, the member who proposes the bill is called the sponsor. After a member of Congress proposes a bill, other members can sign their names onto the bill in formal support of that bill. Those supporting members of Congress are called bill cosponsors.
At times, in his effort to speak to a mathematically-skilled audience, Fowler can make himself a bit inaccessible to the mathematically-unskilled. Consider this passage:
“p(k) = k−γ“? “−γ is the linear slope of the distribution when a histogram of papers per author is displayed on a log–log plot”? Does this make your head spin? Let’s simplify.
Simplification #1: Negative Exponents Turn into Fractions. Just as x-2 is the same as 1/(x2), p(k) =k−γ is the same as p(k) = 1/kγ.
Simplification #2: Get Rid of Greek letters. γ is just the Greek letter gamma, and gamma doesn’t mean anything more here than “some number.” Imagine that γ=2. Or 3. Or 1.5. Then see what happens to the equation to get the general gist of what the equation does. To do this, let’s simplify in a third way…
Simplification #3: Graph it. If so, then p(k) = 1/k2, which would appear on a graph looking like this:
What if γ=3? If so, then p(k) = 1/k3, which looks like this on a graph:
What if γ=1.5? If so, then p(k) = 1/k1.5, which looks like this on a graph:
(I recommend you use the handy Wolfram Alpha as a graphing tool, which is what I just did.)
What’s the same in all of these graphs? The value of p(k) (in the y axis) is high when k (in the x axis) is low — but it quickly drops down to very low levels as k gets bigger.
So what’s that mean?
Simplification #4: Say the idea in words. Remember that k = “number of academic papers” and p(k) is “the number of scientists who have written that number of papers.” The plain-spoken pattern Fowler’s talking about is that most scientists have written very few academic papers, and very few scientists have written many academic papers.
Fowler’s next step is to say that legislators cosponsoring legislation is not like academics writing papers, and to show you how the curve looks different:
More legislators write a medium-to-large number of bills than a very small number of bills, a pattern different than the pattern in the curves above. Considerable legislative productivity in the Congress seems to be the norm. That’s the gist of what Fowler’s trying to express.
Histograms, Histograms, Histograms
Fowler’s paper is fond of histograms. A histogram is nothing more than a bar graph in which the x-axis features some characteristic of legislators that varies and the y-axis tells you how many legislators have how many of that characteristic. Let’s look at Fowler’s Figure 3:
How many circular dots are there in the histogram? 10. How many square dots are there in the histogram? 10. Although Fowler doesn’t say this outright, this means that each dot covers a 10% range in the x-axis. Knowing this, the circle in the upper-left of Figure 3 tells you that about 23% of legislators in the House have 0-10% of their colleagues in the House cosponsoring their bills. The circle at the lower-right of Figure 3 tells you that fewer than 1% of of legislators in the House get cosponsorships for their bills from 90-100% of their colleagues in the House. The pattern of circular dots for the House tells you that in the House, most legislators get their cosponsorships from a minority of their colleagues.
Moving to the Senate, the square in the lower-left of Figure 3 tells you that only 3% of Senators get their cosponsorships for their bills from 0-10% of their colleagues in the Senate; the square in the upper-right tells you that about 21% of Senators obtain cosponsorships for their bills from 90-100% of their colleagues. This pattern leads Fowler to conclude that the Senate is much more collaborative across political divisions than the House of Representatives.
A final hurdle to a solid understanding of Fowler’s paper is his notion of “connectedness.” The paper tells us that it is related to closeness centrality (how close a node is to other nodes in terms of network distance). So why not simply use closeness centrality? Why create a new measure? One important reason is that Fowler has a social network in which ties aren’t just binary, either existing or not existing. Fowler’s ties between members of Congress have strength, having a stronger weight if one member supports a large share of another member’s bills and a weaker weight if one member supports a smaller share of another member’s bills. Fowler (2006: 461) has a formula for this weight…
wij = Σl aijl/cl
… which may look like gibberish to you, but I promise you that it means something. We just have to read Fowler’s page 461 to decode that line.
wij: the “weight” of the tie between two members of Congress (as we’d understand it in this class, the “tie strength“). w refers to the weight, and ij means that the weight is measured between two members of Congress: member i and member j.
l: a letter used to refer to bills. Why the letter “l”? No reason; we just need to roll with it.
cl: the total number of cosponsorships of bill l
Σl: The Greek letter Sigma and the letter “l” means “make the measurement to the right of Sigma for all bills, and add all up the measurements to get a grand total.”
aijl: this (see the text) equals 1 bill l, sponsored by member j, is cosponsored by member i. It’s 0 otherwise.
Σl aijl: therefore is the total number, all added up, of bills sponsored by member j that are cosponsored by member i.
Σl aijl / cl is a number between 0 and 1 (if you multiply it by 100, it’s a percentage), the share of all cosponsorships of j’s bills that are made by i.
The bigger that number is, the more connected member j is to member i because member j gets a greater share of support from member i. Fowler’s “connectedness” measure (which I don’t expect you to replicate) counts ties of strong connection to other people more strongly when figuring out how well-connected members of Congress are to one another.
I hope that these hints help you to read Fowler’s paper with a bit more clarity, and to understand what an impressive feat he has accomplished to study patterns in bill cosponsorship over 31 years to find the most connected members of Congress.
Why does this “connectedness” matter? Why care about network connections in politics? Because politics is not just about votes, votes, votes by disconnected individuals. Politics is about making connections, building coalitions of influence, coalitions that can usher a particularly well-connected politician into higher and higher office. It is no mistake that, as Fowler puts it:
“Several of these individuals would eventually be candidates for President, party leaders, cabinet officials, mayors of large cities, UN ambassadors and other household names in American politics.”
Connections put those politicians over the top.
We’re in perhaps the hardest time of the semester, when all the pressures to complete work seem like they’ll never cease. I remember what it felt like as a student at this time. Before we rush to the end of the semester, I’d like to provide a bit of relief. Your homework assignment for this week involves three steps:
- Breathe in.
- Breathe out.
Really — that’s it.
What about your class participation exercise? Here it is:
- Breathe in.
- Breathe out.
During this time of the semester, the pressure often seems infernal. Take a few minutes to pause, rest, relax and reflect on all the work you’ve completed so far. That’s your work for this week.