|
|||||||||||||||||||
|
|||||||||||||||||||
|
|
||||||||||||||||||
|
|||||||||||||||||||
|
|||||||||||||||||||
|
|
||||||||||||||||||
Red-Blue Roundtable: Valdis Krebs
by Omnivoracious.com at 10:55 PM PDT, October 9, 2008
I started mapping book networks in the last century. It was 1998 when an on-line conversation raised my curiosity. Here is the original white paper I wrote about that initial investigation. After the late Tim Russert brought us the "red states – blue states" meme during the 2000 election I started to investigate patterns of political books. I tried various data collection techniques and found an interesting outcome –- no matter how I collected the data I ended up with highly similar patterns. I use snowball sampling -- start at a known point and follow the data out 1 or 2 steps. Once the snowball sample is complete, I start to eliminate the noise in the network -– I want to find the strong patterns that multiple overlapping networks provide. When the patterns emerge I usually see two strong clusters, with a minor cluster or scattering of books between the two large components. I only color the components after my network analysis software finds the emergent groups in the data –- then it is obvious which cluster is blue and which is red. Below is the first political book map I published on my web site. It showed the famous red-blue divide that had become common wisdom by 2003. It was ironic, and a commentary on our situation, that the center book -– holding both sides together -– was titled: What Went Wrong!
The sharp left-right divide remained in place for the 2004 US presidential election. Below are two graphs of the same data. The first graph is the emergent cluster view –- those similarly connected are closer together. This map was done about 1 year after the 2003 map above. They both contain many different books, yet reveal a very similar pattern and a strong divide.
The second view is of the very same data as above. This view was made after the emergent view showed us who was in which cluster. This view accentuates the divide by putting each side into facing arcs and then sorting the books alphabetically for easy reference. As in the diagram above most links are within the cluster with no direct links between clusters, only through 3 intermediary books that ended up spanning the boundary between red and blue.
During the 2008 US presidential election we expected a different pattern to appear. After all both candidates were initially talking of bridging the divide and the 2006 mid-term elections showed us several examples where the strong boundaries were becoming more fuzzy. The blues and the intermediates [often books outside of 2-party mainstream thinking] started to overlap. Ron Paul, Jesse Ventura, and Lou Dobbs were finding more blue readers than red readers. The map below was done just before the two major party conventions in 2008. We again see: different books, same pattern. We also now see [based on on current snowball sampling scheme] that the left reads a greater number of books than the right. This map does not indicate volume or quantity of sales. It is very possible that the right buys more books of a more focused set. As a general rule we do not compare quantities of books sold, we just use it a bar to include/exclude books in our starting sample -– a top % of Amazon’s bestsellers are chosen as our starting point. Our maps reflect patterns found in the bestsellers on Amazon –- we do not know what is happening amongst low volume books. Our maps capture the most common patterns.
As we have witnessed, after both national conventions this summer,
there is still a strong and vocal divide between red and blue. The war
of words and accusations grows louder as election day draws nearer.
Arms merchants do well in times of war –- no matter if the ammunition
are bullets or words! --Valdis Krebs Red-Blue Roundtable: Valdis Krebs
by Omnivoracious.com at 10:55 PM PDT, October 9, 2008
We vote in boxes, but most of us don't think in boxes. We think in networks -- those near to us [in social, not physical distance] influence what we know, how we think, and who we vote for. We are social animals, not logical animals, nor economically maximizing animals. Many vote against their economic self-interest. Many make illogical choices... or so it appears. Are we stupid? Are we sheep? Are we random? Our behavior often appears that way -- especially to outsiders who do not know the social ecosystem we are embedded in. Political pundits often get voting behavior wrong because they look at voters as independent, logical, demographically-driven, self-maximizing individuals. They miss the 800-pound gorilla in the room -- various social networks and their power to influence behavior and overwhelm demographics, economics, and geography. Birds of a feather flock together. This is a strong driver of human behavior and found throughout living systems in nature. Besides being a fascinating read, The Big Sort is a verifiable and happening dynamic. Yet, we will never live in totally homogenous tracts bereft of diversity. Even if we appear to do so on the surface -- "gee, they all look and dress the same" -- our largely invisible social ties may not be so homophilous. Simple organizing systems such as hierarchies and neighborhoods are never as they appear on the surface. Below is a picture of a corporate hierarchy -- simple reporting relationships, everyone in their own group. Blue nodes are managers, green nodes are directors, and the magenta colored node is the VP. It could be a picture of our state political structure. Precincts reporting into districts reporting into the state. Or broken down further, households into neighborhoods, into precincts. The simple organized hierarchy of boxes. The hierarchy below is viewed as a hub-and-spoke network, or tree, with the black lines showing reporting relationships and gold boxes being either departments or districts.
In organizations we know that the interesting stuff -- learning, innovation, adaption -- usually does not happen within formal reporting structures. The good stuff happens in the "white space" on the organization chart -- between boxes, across groups, spanning boundaries. The invisible network that permeates every organization and every neighborhood is shown by the grey links below. Of course, in the age of the internet many of these grey lines cover large geographical distance.
Votes are counted along the black links in the hierarchy, but votes are created, influenced and reinforced along the grey links which represent overlapping social networks that we are all embedded in. Our friends, family, and colleagues, who influence our vote, are distributed through many neighborhoods, precincts, districts, and states near and far. This is why it is not only important to look at quantity of book sales by geography, but to look at networks of books and how they reveal the influence factors in each of our political sense-making processes. Networks of books? Do books have a social life? Not really, but Amazon provides us data to evaluate book purchases as a social system. Amazon's consistent feedback on every product page -- people that bought [this] also bought [those] –- allows us to create a social network of books. Of course what we are really evaluating are not the social dynamics of books but the social dynamics of buyers and readers of those books –- all without revealing the identity of the Amazon customers making their consumer choices. So, what do you think our network of books will reveal for this election? --Valdis Krebs Red-Blue Roundtable: Andrew Gelman
by Omnivoracious.com at 2:53 PM PDT, October 9, 2008
As I noted in my earlier entry to this discussion, I think Zogby's and Bishop's positions can be reconciled. Zogby's approach "has been to find broad areas of commonality, the overarching groupings that tens of millions of Americans find themselves in," while Bishop says that Americans are "sorting by economy, by ways of life, by education, by belief and, only every election day, by politics." How can both these statements be true? Most obviously, there can be unity across population groups amid geographic separation. It is easier than before to communicate with people who live in other cities; in my own world of academia, many have noted the transferring of allegience of faculty from their university where they work to the academic "field" they inhabit.
Almost half of Republicans do not describe themselves as being ideologically conservative. If we also consider issue preferences, the constraint of people's political preferences looks even weaker. Only 6% of respondents were Republicans who think of themselves as conservatives, oppose abortion, and have conservative views on affirmative action and health policy. Fully 85% of self-declared Republicans are nonconservative or take a nonconservative stand on at least one of these three traditional issues. A similar picture emerges if we look at Democrats. In this case, of the 49% self-declared Democrats in the sample, only 36% call themselves liberals. Overall, almost 90% of Democrats are nonliberal or have nonliberal views on abortion, affirmative action, or health policy. These numbers should not be surprising, given that in general, the correlation between party identification or ideology and opinion on political issues is low. Knowing somebody's political identification increases our chances to guess his or her issue preferences, but not by much. This supports Zogby's view of Americans as nonideological and Bishop's view of sorting based on lifestyle rather than politics. So, yes, most people are not consistently ideological in their attitudes. But people have strong views about the Democratic and Republican parties. Higher-income Americans in red states have distinctly different views than higher-income Americans in blue states. See this graph, which shows average ideological positions (as estimated from survey questions on economic and social issues) among poor, middle-income, and rich voters in red, purple, and blue states:
The higher the income level, the more distinct are the residents of red and blue states. This doesn't contradict Zogby's point that most Americans are broadly in the political center; it just shows that the distinctions that do exist manifest themselves geographically. It is a challenge of politicians, when campaigning, to make the most of these divisions and, when governing, to find the underlying unity. --Andrew Gelman Red-Blue Roundtable: John Zogby
by Omnivoracious.com at 2:53 PM PDT, October 9, 2008
There is a new investor class – The Investor Next Door – who reject the Michael Douglas character and the message of "greed is good" in the movie Wall Street. These new investors have modest expectations for a comfortable retirement and educating their kids and they increasingly prefer socially responsible investments. I even redefine age cohorts and try to fit them into a role they will play in a new America defined by living within limits.
It is our institutions and leaders that are seriously lagging. The people get it just fine. There is a new political consensus – for change, for problem-solving, for consensus-building – instead of partisanship and there is a new American consumer. --John Zogby Red-Blue Roundtable: Bill Bishop
by Omnivoracious.com at 2:53 PM PDT, October 9, 2008
America really is different from place to place. To show that, Bob Cushing developed a simple scheme. He divided the nation's 3100 counties into four groups based on the 2004 election results. There were counties that voted in a landslide (more than 20 percentage points) for George Bush. They are bright red in the charts below. Those counties that voted for Bush but at margins under 20 points are in a lighter red (well, kinda purple). The counties that voted for John Kerry in a landslide are bright blue and the counties that voted for the Democrat at margins under 20 percentage points are light blue.
Over time, these county groups were collecting different kinds of people. For instance, here we can see that the blue counties from 2004 were pulling in more people with college degrees. The greater the vote for the Democrat, the higher the proportion of the population that has a college degree. (Get used to this stairstep pattern.)
Meanwhile, the dark red counties from '04 were collecting more people who went to church. Not surprisingly, we also found that 45% of the people in dark red counties attended Bible study. In strong Kerry counties, 28.7% of the people attended a prayer or Bible group.
And red counties collected white Americans.
Opinions differed by geography. Here are our four county groups showing the percentage of people who feel "strongly" that homosexuality should be accepted.
And finally, here are the percentages of self-described liberals in 2004 who said invading Iraq was the right decision.
To me, this is the most interesting chart of the bunch. The meaning of "liberal" changes from place to place. A liberal in a heavy Bush county is a lot more conservative than a liberal in a Kerry dark blue community. This tells me that we are all powerfully affected by our neighbors — and that we have so segregated ourselves that the basic language to describe politics has different meanings in different places. No wonder we're buying different books in different places! --Bill Bishop See the whole Red-Blue Roundtable. Red-Blue Roundtable: Andrew Gelman
by Omnivoracious.com at 2:53 PM PDT, October 9, 2008
I attribute part of the difference in focus between Bishop and Zogby to their different goals: Bishop is interested in America's political divisions and thus writes about the increasing number of local areas that are becoming politically more monochromatic. In contrast, Zogby, as a pollster, is particularly interested in groups of people who can be persuaded--swing voters--and their relation to the economy, which remains the most important issue in deciding people's votes. As a result of these different focuses, Bishop sees Americans as divided whereas Zogby sees the country as fundamentally centrist. What do the data say? There's truth in both Bishop's and Zogby's perspectives; it all depends on how you slice the population. First, in defense of Bishop's view of polarization, there's lots of evidence that partisanship is much more ideological than it used to be. For example, here are the average positions of self-declared Democrats, Republicans, and independents on the issue of abortion:
As late as the Reagan years, the parties were indistinguishable on abortion. As we discuss in chapter 8 of our Red State, Blue State book, voters have become polarized in their attitudes to their parties and in economic, social, and foreign policy issues, in a way that they weren't, 30 or 50 years ago. Yes, the voters are divided by party. But where do they stand on the issues? Here, Zogby's hypothesis of moderation is supported by detailed modeling of survey data. When we put voters and congressmembers on a single left-right scale, we found that most voters are in the middle with their elected representatives sitting to their left and right:
These ideologies are not self-declared "liberalism" or "conservatism"--we don't necessarily trust responses on these politically-loaded terms--but are estimated from positions on a bunch of issues. In the book, we also break these down by Democratic, Republican, and battleground states.
See the whole Red-Blue Roundtable. Red-Blue Roundtable: Bill Bishop
by Omnivoracious.com at 1:21 PM PDT, October 6, 2008
The sign outside my local elementary school last spring wished students a "great summer" in a typical Travis Heights way: "Whatever Higher Power(s) You May Or May Not Believe In...YEE HA!"
Oh, and one more thing. Travis Heights Elementary is exactly one block from the house where liberal writer Molly Ivins lived. (If Molly were still alive and writing, our zip code would be bright blue on the Amazon book map.) To the cable television boys who finger paint on the electoral map, Texas is always solid red. (The CNN guy never taps his digit down in our direction.) But here, in the neighborhood where I live, we're bluer than Vermont. And that's the point about all those red and blue state maps. They are good for showing what has been a remarkably static division in the Electoral College. But they miss how people are living and, I suspect, buying books. The divisions state to state are real, but they are nothing compared to how Americans are sorting themselves from community to community. Statistician Bob Cushing and I traced the votes at the county level from 1948 to 2004 in our book The Big Sort. We could see that majorities, Republican or Democratic, were piling up in communities. The last five presidential elections have been as close as any in the last 100 years. But an increasing number of people live in counties where elections aren't close at all, where either one party or another wins in a landslide. So, in 1976 — the nearly dead-even contest between Jimmy Carter and Gerald Ford — about a quarter of all voters lived in a county where one side or the other won by 20 points or more. By 2004 — the nearly dead-even contest between John Kerry and George W. Bush — almost a half of all voters lived in places, like Travis Heights, where 20 points or more decided the election.
One half of U.S. voters live in counties that have remained unchanged in their presidential preference since 1980; 60 percent live in counties that have not changed since 1988; and nearly 73 percent live in counties that have not changed since 1992. Orange and Los Angeles counties in California are side by side, but local political majorities have been growing in opposite directions since 1976, a phenomenon found in two-thirds of U.S. communities. Our sense is that people aren't moving to be around others who feel as they did about the Iraq War or single-payer health plans. People are clustering around others who live as they do — people who have similar lifestyles, who read similar kinds of books. And every four years those ways of life align with political party. Marketing folks have known for some time that demographic factors have little meaning these days. People don't define themselves as "single, male, college-educated, 25 to 35 years of age." They think of themselves as environmentalists, car-racing enthusiasts, or, as one woman told me, "I'm an ocean-oriented person." They know that to learn about another's politics you consider the way they live, not their age, race, or level of education. In a radio talk show in Minneapolis, three callers told me they realized they had moved into a community with political opposites when they saw their neighbors using lawn chemicals. (It was a public radio show, in case you couldn't tell.) Politics these days aren't about issues. People don't line up with a party because they agree with a set of policy position papers. One political scientist described the choice of a political party this way: You have a choice of attending one of two parties being held in two rooms off the same hall. You look into each room and you look at the people — how they appear, their gestures, what they're wearing. You get a vibe and then you join with the group you think is most like you. You might even look at the books they are reading. That's how people pick neighborhoods these days. It's also how they pick churches and civic clubs. It's the reason people with college degrees are clustering in particular cities and why some places are succeeding economically while others are slipping further behind. That's The Big Sort. And it plays out more in culture, in books, than in politics. --Bill Bishop
|
For a social network scientist, Amazon is a great sandbox for experimenting and searching for interesting patterns! 
Voting is simple... precinct, district, state, add 'em up, send 'em in. Outside of Florida 2000 and Ohio 2004 this hierarchy of geographical boxes works well. Boxes are simple, you are either in this one or that one, or not eligible to play. Simple clear rules. Simple clear math.
A conversation between a pollster, a journalist, and a professor seems like a great idea--we all do research, but using different methods and with different goals. But in this particular forum we seem to have been talking past each other, with Zogby emphasizing how most Americans have a fundamental non-ideological view of policy, Bishop talking about partisan divisions, and me going on about red and blue states. Perhaps our areas of expertise are just too far apart. Nonetheless, I will try one more time to make some linkages. I am interested to hear my co-discussants' views on these issues.
We have pretty well established the many ways in which Americans have figured out to divide ourselves into warring political cultures. There are a number of ways to move beyond this. One of my colleagues has written very eloquently about "microtrends," small little clusters of new demographic groups that provide guideposts for strategists and marketers. My approach has been to find broad areas of commonality, the overarching groupings that tens of millions of Americans find themselves in. That is how we build bridges, reach consensus, and move a nation forward, as opposed to locking it into gridlock and political entropy. And the exciting thing is that these large clusters and meta-movements are being driven by the people themselves. Thus, instead of a quadrillion little pieces of religious fragments, tens of millions of Americans are adopting a Secular Spiritualism – a myriad of ways to reject and avoid materialism and crass consumerism, replacing it with a search for a broader purpose to their lives. 
Gender neutral, faith neutral, God neutral...that's my neighborhood here in Austin, Texas. In 2000, George Bush came in behind both Al Gore AND Ralph Nader in Travis Heights. When 70 percent of Texas voted in favor of a ban on gay marriage, my precinct voted 90% against. In my zip code, 90% of the federal election contributions this year went to Democrats.
When you look at local voting results over time, you see that counties tip Republican or Democratic, and then they keep tipping as more Republicans move into Republican counties and Democrats cluster in Democratic communities. (Or, at the same time, counties grow increasingly lopsided as members of the minority party decamp for more politically hospitable environs.)