Science: Forecasting Political Results
Monday, August 25, 2008; 11:00 AM
Washington Post staff writer Shankar Vedantam answered questions about his Science Page story on Monday, Aug. 25 at 11 a.m. ET. Vedantam writes about how a geophysicist and a historian are using earthquake forecasting methods to predict the results of political elections.
He was joined by Allan Lichtman, a historian at American University.
Read the story here: In the Quake Model, Rumblings Favor Obama
Vedantam will also take questions on his weekly column exploring what the news tells us about human behavior -- this Monday's column explores why voters pay more attention to gossip and sex scandals than substantive policy issues.
Read the column here: Why Fluff-Over-Substance Makes Perfect Evolutionary Sense
A transcript follows.
Shankar Vedantam: Welcome to a chat discussing my story today about how the science of earthquake prediction can be used to forecast the outcome of the presidential election, and my Department of Human Behavior column exploring what the news reveals about human nature. Today's column argues that voters are evolutionarily predisposed to being fascinated by sensational and scandalous news -- such as John Edwards' affair -- and predisposed to be bored by news about policy, even when the policy issues directly affect their lives.
I am pleased to be joined by Allan Lichtman, a historian at American University, who helped develop the system that applies earthquake science to politics. Licthman and co-author Vladimir Keilis-Borok have come up with 13 "keys" or markers that predict the outcome of presidential elections -- or at least, the winner of the popular vote.
Washington, D.C.: Do you think that this model accounts for the element of Mr. Obama's race and background in the modelling? I think that for the past 8 years, most Americans put tribalism above the best interests of the nation or nationalism and in this respect have been little different than the ethnic groups in other nations who oppress minority groups - which is what seemed to have happened to Mr. Obama's father - as he was a Luo concerned that tribalism would trump Kenyan nationalism.
Allan Lichtman: The keys have accurately predicting election results since 1860 and thus have proven accurate despite enormous changes in our politics, economy, and society. Therefore I believe they will hold true this year as well. However, we will not know the influence of the potentially confounding issue of race until people enter the polling booth.
Shankar Vedantam: Most times when we want to understand something, we try to get closer to it. In an election, we try to interview individual Democrats, Republicans and independents about their views. Your approach steers clear of such micro-level analysis. Why should it work?
Allan Lichtman: My approach steers clears of micro-level analysis because it is impossible to model a complex system of 120 million voters at the individual level. It is possible, however, to distill the key components of the system and from those components model the behavior of the system as a whole. For example, it is not possible to model the behavior of the individual atoms of a saline solution, but it is quite feasible and useful to measure the salinity of the solution overall.
Shankar Vedantam: Allan, the science of earthquake prediction seems miles away from politics. What made you think there was a connection?
Allan Lichtman: Everything we know about politics we've already stolen from geophysics: tremors of political change, seismic movements of voters, political earthquakes, and volcanic elections. In truth, predicting earthquakes is very similar to predicting the outcomes of elections. Earthquake prediction seeks to distinguish between stability and upheaval (a major quake). Likewise an election prediction seeks to distinguish between stability (the electorate favors the party holding the White House) and upheaval (the electorate favors the challenging party). Earthquake predictions are based on patterns in the physical environment. Election predictions are based on patterns in the political environment.
Shankar Vedantam: Your keys suggest that the Democrats are likely to win the White House. Does the fact that recent polls show a closing of the race make you concerned about your prediction?
Allan Lichtman: My predictions are not based on polls and polls taken prior to the party conventions historically have no predictive value. For example, in 1988 I predicted that Republican George H. W. Bush would prevail even when he trailed Democrat Michael Dukakis by 17 points in the polls.
Shankar Vedantam: You suggest that campaigns don't play much of a role in the outcome of an election -- at least the skill in which a campaign is organized is not one of your 13 keys. Can you explain this counter-intuitive idea?
Allan Lichtman: My notion is that campaigns have little or no influence on elections for two reasons. First, voters are pragmatic. They base their decisions on whether or not the party holding the White House merits on third term based on performance as gauged by such big-picture considerations as the economy, foreign policy, social unrest, scandal, and policy change. Second, voters discount as political and self-interested what candidates say or do in a campaign.
Shankar Vedantam: According to your calculations, the Republicans lose 7 keys this year -- which means they should lose the White House in November. How much money would you be willing to bet on this forecast?
Allan Lichtman: I never bet on elections. I don't want to jinx my own forecasts. But anyone else should feel free to do so.
Middle America: Allan: In your model you predict an Obama victory. Current polls (take your pick) show Obama and McCain polling about even, give or take a few points. Given the reasons mentioned in the article predicting an Obama victory, why isn't Obama pulling ahead in polls by margins of 10, 20 or 30 percentage points?
As a follow-up, if McCain wins, will you do another story showing that either your modeling was incorrect, or that you were wrong?
Allan Lichtman: Early polls are an extremely unreliable indicator of the outcome of a presidential election. For example, at the time that I picked George H. W. Bush to win in 1988 he was trailing his Democratic challenger Michael Dukakis by 17 points in the polls. There is also this year the historically unprecedented factor of an African-American nominee. The unmeasurable issue of race may be influencing the polls.
Shankar Vedantam: What I especially like about the Lichtman-earthquake model is the extent to which it offers us a falsifiable hypothesis. By publishing his forecasts months ahead of an election, the researchers are certainly opening themselves and their theory up to test -- which is what science must do. I should add, however, that this theory is not a crystal ball approach. As with earthquakes, being right 70, 80 or 90 percent of the time is a good deal better than being right 50, 40 or 30 percent of the time!
Munich, Germany: I'd always believed that sexual scandals made a bigger social splash than tales of incompetence or failure, for example, because of religious mores. Dr. Davis researched many newspapers from many countries, but did he look at cultures unrelated to the monotheistic Christian-Muslim-Jewish religions?
There might not have been many newspapers for audiences of tribal religions that worship multiple deities, and India was probably too keen on abiding by colonial British values.
Shankar Vedantam: Hi Munich: Thanks for the question on my Department of Human Behavior column today. Dr Davis looked at eight countries that were quite diverse -- and that also have individually changed a great deal over the past 300 years. He found the content of scandalous or sensational news did not change either over time or between locales. So religion might be a red herring here.
Harrisburg, Pa.: Several decades ago I recall someone claimed to have developed a model that predicted Presidential elections on economic data. Do such models continue to show they are strong economic predictors in who may be elected President?
Allan Lichtman: Economic models are not sufficient because they do not take into the account the full range of factors that influence elections. That said, most economic models are predicting a Democratic win this year.
Seattle, Wash.: I did a similar (much less official) study on when voters pick upheaval over stability. The thing that I stumbled across was third party vote. You mentioned that one of the criteria was the absence of a third party candidate.
I came to the opposite conclusion. In the last 17 elections, there have been 5 elections where the incumbent's popularity was less than 45% and 5 elections where a third party got more than 5%. Four elections showed up on both lists. My hypothesis was that this was not a coincidence.
People can't quite picture themselves voting for the other party even where their own party has left them severely disenchanted. A third party candidate gives them a middle option.
I've discussed this with some people and they seem to think that it was racism and not disenchantment with Democrats that created the 1948 and 1968 third party challenges. However, one has to question why no racist candidacies gained serious vote totals in the other elections of the era. Racism was certainly present throughout the era. My theory is that it was only when southern democrats were disenchanted with the party in general and not quite ready to defect to the Republicans (still associated with Hoover) that they gave into their racist third party urges.
The non-racist 1980 Anderson and 1992 Perot elections also fall into this category.
If my theory is correct, Bob Barr will get far more votes than any one anticipates. Republicans won't quite be able to defect the whole way to Obama, but they'll be more than happy to express their dissatisfaction with the Bush administration.
Of course this all relies upon whether or not McCain is able to effectively distance himself from Bush. 1952 was the only time a pres or VP wasn't on the ticket and also the only time a less than 45 percent incumbent didn't lead to a significant third party challenge.
Allan Lichtman: I agree that third party campaigns are usually detrimental to the party holding the White House. As you indicate, they are a sign of discontent with the party in power and the thrid party candidate usually goes after the candidate of the incumbent party -- e.g., Perot in 1992. I do not believe Barr will get enough votes to reach the 5 percent criterion. However, like Nader in 2000 he might influence the outcome in individual states.
Arlington, Va.: Instead of all the modeling, which is interesting, couldn't you just ask the president of the company that makes most of the voting machines who is going to win?
Allan Lichtman: The issue of whether all our votes are fully and fairly counted is a serious one as illustrated by the situation in Florida and the current controversy over electronic voting machines. We need a much stronger regulatory presence to make sure that in the world's oldest running democracy we have voting mechanisms that are accurate and tamper-proof. The problem is not yet solved.
Arlington, Va.: The keys seem to predict national elections only and don't appear to relate to regional clusters because, otherwise, the South would react to the same keys as the rest of the country instead of voting something called conservative principles, even if the principles work against personal interest.
Allan Lichtman: You are absolutely correct. The keys predict national presidential elections only. They do not predict senatorial, gubernatorial, or congressional elections and they do not predict regional patterns in voting. The Keys have held in recent decades when the south was solidly Republican, they also held in earlier times when the south was solidly Democratic. They also do not predict individual voter decisions. The point is that they are broad national trends that hold regardless of patterns among regions or voter groups. To cite another example, the keys accurately predicted elections in the early twentieth century when African-Americans, the most reliable Democrats today, were the most reliable Republicans.
Washington, D.C.: Professor Lichtman: How do you deal with the question of subjectivity for the two charisma keys? While I support Obama, I don't find him particularly charismatic, though I can see where some would. McCain seems to me not charismatic at all, but surely others see him as an exciting "maverick" and "war hero" and so on.
Also: White Protestant Nation is a TERRIFIC book. Are you disappointed in the attention it's received? Seems to me there should have been more.
Allan Lichtman: The charisma keys are the most subjective or the most judgmental of all 13 keys. No election has ever turned on the charisma keys and this one does not. The point, however, is that there are always individual voters and some voters groups that will find particular candidates charismatic or inspiring. To reach the threshold for charisma under the Keys system, however, the candidate must be broadly recognized as inspirational. Such candidates are very rare. In recent decades only John F. Kennedy and Ronald Reagan have reached the threshold.
Thanks for your very kind comments on White Protestant Nation. I do wish the book would gain more attention. It seems that these days to get attention you have to write a polemical not a serious book. However, for a nearly 600 page work of history it has gotten considerable attention, including reviews in the NYT, the St. Louis Post, and the Chicago Tribune, as well as places such as Kirkus, Publisher's Weekly, Book Forum, and Library Journal. Unfortunately, the NYT review was written by a former Bush speechwriter and was anything but serious and objective. All other reviews have been excellent.
Vero Beach, Fla.: Earthquakes are predictable in the sense that relative hazard can be predicted (Oakland is far more likely to be hit than Sacramento), but predicting timing is difficult. The four-year American cycle (didn't Walt Whitman call it the Presidentiad?) seems a big help to the prognosticator. Is anyone using keys for British elections?
Allan Lichtman: You are correct. The regular cycle of presidential elections make such contests far more predictable than earthquakes. We don't have the same timing problem or geographic reach. I have been asked to apply the Keys system to other nations, but I lack the expertise in the politics, history, social context, and economics of those societies. Certainly, however, an authority on British or French politics could apply the Keys methodology and approach to elections in those nations.
Shankar Vedantam: Allan, you mention in your paper that four keys alone could predict every election between 1860 and 1980. Can you talk about those four keys and why they are so powerful in predicting the outcome of elections?
Allan Lichtman: The most powerfully predictive keys are first the contest key. It may seem counterintuitive that contests for the nomination of the the incumbent party are so important. But a major contests within the party holding the White House is a nearly sure sign of discontent with governing. However, an internal battle for the nomination of the challenging party is not predictive of failure at the polls.
The second most powerful key is the short term economy. Invariably, incumbent parties lose when the short-term economic key is turned against them. The reverse, of course, is not always true.
A third key that is not predictive overall, but has mad the difference in close calls -- as in 1992 -- is the third-party key.
The policy-change key is also important, although it confounds the conventional wisdom that a leader should always move to the center.
McLean, Va.: Is there a way to see the "accompanying graphic" mentioned in the story online? I didn't see a link on the story page. Thanks!
washingtonpost.com: Here's Today's Graphic - 13 Keys to Keeping the White House
Allan Lichtman: The graphic is here. It is an important and convenient short-hand to the Keys. But to understand the Keys system fully and make correct calls on individual Keys, please consult my book, The Keys to the White House, 2008 Edition.
Vienna, VA: Given that macro variables predict elections but micro variables (voters) cast votes, doesn't your model imply that individual decisions are shaped by broad social forces whether people perceive them or not?
Allan Lichtman: That is an excellent but perhaps unanswerable questions. The Keys deal with the electorate overall rather than individual voters. It does suggest, however, that the collective decisions mad by the American people are indeed shaped by the broad forces outline in the keys.
Allan Lichtman: Thanks for the excellent questions. Beyond helping to predict and understand elections, the Keys have implications for the conduct of politics in the United States. they indicate that we need not continue to conduct empty, consultant-driven, sound-bite style general election campaigns that fail to elevate the level of political debate, and inspire activism at the grass roots. Rather, the Keys suggest that candidates should fire the consultants and handlers and use campaigns to build a foundation for governing the country.
Shankar Vedantam: Thanks so much to Allan Lichtman for joining us for this chat. The November election will tell us whether his research is correct! Have a good day, everyone.
Editor's Note: washingtonpost.com moderators retain editorial control over Discussions and choose the most relevant questions for guests and hosts; guests and hosts can decline to answer questions. washingtonpost.com is not responsible for any content posted by third parties.