Politics and the Weather: More in common than you might think

Posted by Matt Noyes October 23, 2008 at 4:26 pm

Image by Getty Images via DaylifeIt may seem a bit odd to find a political analysis piece authored by a professional meteorologist. After all, what could politics and the weather have in common that could give us insight on one or the other? Certainly, we’re all aware that both topics serve as excellent conversation pieces, but a closer inspection finds much more in common between a Presidential campaign and a blockbuster storm than one may imagine: guessed at by many, but mastered by few; behaving as predicted much of the time, but dishing out major surprises without warning; available statistical guidance and numerical models designed to guide, but not forecast; and a complex network of interactions that is often over-simplified to draw conclusions that were never meant to be reached on limited data.

That’s not to say that the weather is wholly unpredictable, and neither are politics - we’ve all seen excellent forecasts for both, and in fact, most of the time the guidance is correct in individual settings and scenarios. Complications arise, however, when multiple ingredients come together to form a much larger, much more intricate web of interactions. In such instances, meteorologists turn to what’s known as an ingredients-based methodology - that is, employing a method of forecasting in which you turn to each ingredient that’s playing a part of the whole, and determine how each of those parts should function, given what you know of the setup. It’s this meteorologist’s contention that a similar ingredients-based thinking can be effectively applied to the unpredictability of politics. In the end, you may or may not find the correct forecast, but you will have made much more sense of what can be a confounding collection of information.

To successfully employ this methodology, a careful analyst must approach the setup with a solid understanding of several facets - accurate and comprehensive observations and analysis, pattern recognition and historical context, sound forecasting techniques, and a healthy respect for inherent uncertainty. Whether dealing with weather or politics, if one can grasp all of these factors and incorporate them successfully into a forecast, chances are good we will yield the most accurate prediction possible, and adequately prepare the recipients of our forecasts for potential deviations in the final outcome, due to the uncertainty we’ve taken into account.

Current Observations - Political Polls

In the past days, weeks and months, I’ve seen a multitude of political polls, taken nationally and on a state-by-state basis, attempting to successfully gauge which presidential candidate is on top. We can consider these polls the observations of a storm - illustrating wind, temperature, and other information to show us what’s going on right now. It’s important to remember that current observations don’t determine the forecast, though if it’s snowing and there’s a north wind, chances are good the snow will continue. Similarly, if one candidate is significantly up in the polls and the wind is blowing their way, chances are good they will stay up. One big concern in the world of weather, however, is the quality of the reports we garner - is the data accurate? Do all of the observations make sense? Can you safely make an inference of what’s going on based only upon the limited data that is available? The same questions must be posed to analysis of polling data, and there are a few problems with these polls from a meteorological perspective.

The first problem comes from the method of sampling - were registered voters or likely voters surveyed? Was this a completely random survey? If so, where did the pool of individuals who were included in the “random” mix come from - from landline, listed telephone numbers? From cellular phone numbers? From some other source? How many individuals were polled - was it a sufficient number to ensure equal participation by registered Republicans, Democrats, Unenrolled and other voters? Were all demographics equally represented in the poll? Many of these questions are important in determining the quality of the data, but also, in fairness, are outside the control of pollsters. Nonetheless, these factors must be considered by the analyst.

The next problem comes with regard to the organization conducting the poll. Was there bias in the phrasing of the questions, or even the tone in which the questions were asked? Were sufficient checks in place to ensure data quality? Again, benefit of the doubt must be given to polling organizations, just like benefit of the doubt is given to the automated machines that report the weather at most locations nationwide. Back in the old days we had people at every observing site…now it’s mostly computerized observation devices. Known errors have emerged - under measuring of precipitation, an occasional report of “heavy snow” even though the sky is completely clear, and other nonsensical reports. Alas, such errors are an accepted inconsistency weather-minded folks must contend with, and it’s important to acknowledge that similar inconsistencies will arise in the closest thing we have to political observations - the polls - and be cautious of how much credence we give to each individual report, judging the whole rather than the parts.

For those who pay close attention to daily polls - often multiple polls per day - we’ve seen the grand swings from one poll to the next. Choosing one day as an example in national polls - we’ll look at a few polls that ended on or around October 20 - we saw an AP/GfK poll showing the presidential race tightening dramatically from the seven percentage point lead Obama had three weeks ago, with the most recent data in this poll putting Obama at 44% support among likely voters, and McCain at 43%. Meanwhile, most news reports indicate that Obama “continues to widen his lead,” which corresponds to polls such as the NBC News/Wall St. Journal poll, completed on the same day as the AP/GfK poll, that shows Obama with a solid 10 point lead. The day prior, the Pew Research poll was completed and put Obama at a whopping 14 point lead, nationally.

Taking a step back to gain “big picture” perspective - something we do regularly in making a weather forecast - these polls show Obama likely has a lead in national opinion, but the magnitude of that lead is highly uncertain. The Real Clear Politics average of pertinent polls, as of this writing, stands at just over a 7 point lead for Obama, but the difference from one poll result to another is astounding. In other words, the observations point to an Obama lead, but making an inference to the size of that lead would be taking a leap that is unsupported by the data.

Recognizing Historical Pattern - Votes in the Past

An important part of forecasting is to have the experience and insight to realize when a pattern resembles a previous point in history. Of course, no two storms are ever exactly alike, and the slightest difference can end up magnifying down the line, and be a huge player in the final outcome of how a storm develops. Similarly, taking the pulse of votes in the past, under similar circumstances, is absolutely essential to understanding the forces at work, but differences may or may not magnify down the line.

One historical pattern is to look at similar votes with similar approval/disapproval ratings of a sitting President and measure how the party performed in the following election. Of course, there is little to compare President Bush’s approval rating to, but all indications are that the party of the sitting President has the cards severely stacked against them when the approval ratings are low. What’s interesting here is the tremendous disapproval rating for the Democratically-controlled Congress, and what the meaning of this dual pronged disapproval is during the election.

Accurately, both major party candidates have assessed “change” is the key phrase, no matter what party one is a part of. Will the disapproval of Congress offset that of the President? One would think, logically, that this may go a part of the way, but certainly not completely offset the effects of a dismal approval rating in the President. Historical reference also will take into account issues such as who comprises actual voters. It’s one thing to be polled and give an answer of whom you’ll vote for, but it requires more effort to actually go out to the polling places on Election Day and cast the ballot. Has the demographic of voters changed in this election? This is one possible change to historical patterns, but remember the importance of the electoral college here - the question is not whether the demographic of the voter has changed, but rather, has it changed in states that hold key roles in the Electoral vote. In other words, getting many Democratic voters to come out in an already Democratically held state - or vice versa for Republicans - will not make any difference as to who is elected President of the United States, only in the populous vote, which doesn’t matter in the end, as we’ve seen in previous election cycles.

Another very important part of historical trends is to understand and accept the way certain states have voted in the past. For example, there’s been a lot of talk about traditionally-Republican states voting for Democrats in this election. That is entirely possible, but from an analytical standpoint, not highly likely. To relate this to weather, it’s the equivalent of seeing a storm approaching one of our country’s coasts, with a steady wind off of the ocean water. Even if there is cold air in place ahead of the storm, nine times out of ten a strong wind off the relatively warm ocean is going to warm the atmosphere enough for rain. It’s happen time and again in the past, and will happen time and again in the future. The only way I’d forecast snow in such a scenario is if there’s a sufficiently cold airmass in place ahead of the storm, such that even a strong wind off the water won’t scour out the cold that’s in place. Historical trends must be viewed in the same light - if you’re dealing with a traditionally strongly held Democratic or Republican state, it’s going to take a strong force to reverse the power of precedent.

Forecasting Techniques - An Art, or a Science?

Meteorology is undoubtedly a science - the analysis of the world around us, and the formation of a hypothesis, with the world serving as the experiment. Then again, forecasting can certainly be considered an art - every meteorologist will come up with a slightly different forecast based upon their own interpretation of the data, and their own experience in given circumstances. Politics isn’t much different - the science is accurately interpreting the available data and putting forth a hypothesis to fit the data, and the art form is shaping the analysis and hypothesis in such a way that it will yield the most accurate result possible. Ideal forecasting techniques will utilize a careful analysis of all available data, and will combine the factors of observation, pattern recognition and historical context, and a respect for both uncertainty and something we in the world of weather call a “trend.” Political analysts use trends, too - in fact, they are the crux of most analysis, as a trend shows the direction something is heading.

For example, we may see a snowstorm slated for three days from now that’s forecasted to pass 300 miles to the east - safely away from us, and therefore, no snow is put in the forecast. As more data becomes available and we get closer to the formation of the storm, however, indications in the weather pattern and from our guidance suggest perhaps the storm will be 200 miles away, not 300. Light snow goes into the forecast. One day out, the storm is now forecasted to be 100 miles away - moderate snow. It doesn’t take a genius to see that each day, the forecasted storm is trending closer. It’s entirely possible, though not guaranteed, that the actual path of the storm may be directly over the region. So, to find a trend can certainly give an indication for the way a situation is unfolding, and the same can be said for politics - a trend toward a lead for one candidate or another can be a very telling sign.

Employing the Techniques - Forecasting the Presidency

Here’s the dangerous part. This is the part of the forecast where you make an assessment based upon all available data. The first necessary characteristic of a good forecaster is humility - the ability to realize that no projection is infallible, and while your prediction may verify, honesty is required in laying out where things may not go as expected. This ties back to the concept of uncertainty - in the weather world, that uncertainty comes inherent in nature, while in the political world, the uncertainty is tied directly to human nature; human actions and responses on the part of both the candidates and the voting public. This is an inherent uncertainty that can never be completely accounted for - no matter how carefully measured a response is from a candidate, no matter how certain situations are portrayed by the media, and no matter how significant a lead in the polls may be.

Now, that’s not to say that any of these factors should be ignored, but rather that they are all forecasting tools, and it is virtually impossible for any one factor to be determining. When all are figured together, however, increased certainty in a forecast may result. There is also, of course, the uncertainty of events unseen that have yet to transpire. With less than two weeks remaining until the election, the window is closing for defining events, but it remains a possibility until votes are cast on the first Tuesday of November.

Of course, neither politics nor weather are as simple as crunching numbers, otherwise both would be accurately forecasted without exception. There’s something to be said, though, for a scientific mind who dabbles in the unknown daily. If an individual who does battle with history, patterns, observations, analysis, trends and forecasts every day can broaden that focus to include other uncertainties, like politics, perhaps there is an insight that can be gained. Then again, perhaps not, but let’s investigate the election from the perspective of a meteorological mind.

First, as mentioned earlier, aside from illustrating a lead for Obama from a national perspective, the polls are relatively useless. If this were snowfall guidance, I’d ignore virtually all of it, except to know that Obama has some sort of a lead. How meaningful is that lead? Well, it’s meaningful because it exists, but remember that in a Presidential election, a national population opinion poll is nearly meaningless. There is so much emphasis put on the several polls like this that come out daily by the media, and therefore by the general population, but this will not win a Presidency: Electoral votes will. So, take the national polling data ranging from Obama +1 to Obama +14 and just file it away for now. It’s extra data that, aside from perhaps indicating a trend later in the forecast, won’t do anything but confuse you right now.

So, if the Electoral College is what matters, let’s take a serious look at the setup. In addition to national polls, which we’ve thrown out, we do still have state polls. Remarkably, these state polls aren’t run by nearly as many groups, or nearly as often, as the national polls, which is truly a shame because they are, in the end, all that matters. In the world of numerical analysis, there’s so much information to make sense of, often you must accept what appears to be a given. That is, if statistics are staring you in the face that seem to yield an obvious solution, the nature of the game tells us we should take the data at face value, unless there is a viable reason to question the inference we’re making. It’s important to realize, however, that the polling data in each state doesn’t imply anything - it’s only numbers - and perspective we gain is from our own inference, and if it’s erroneous, it’s our own fault for not properly acknowledging potential uncertainties.

Even “margin of error” can be thrown out the window, for the most part, because there have been numerous instances where the actual result of an election is far outside the margin of error. We see this in snowfall forecasts - 6″-12″ of snow may be predicted, but when 4″ or 14″ falls, we’re not shocked…we just blew our interpretation of the data, and more importantly, didn’t adequately consider potential sources of uncertainty.

So, as we make the following inferences, understand that we may be making a mistake in doing so, but in the interests of expediency, we will have to consider them to be safe inferences. The first step here is to find states that were strongly Democrat or strongly Republican in the last two elections, then compare to the polling data. If we see strong Democrat or strong Republican for two election cycles, and the same party leading in the polls, we have to assume the result is probably going to be quite similar.

Solidly Democrat in 2000 and 2004 (greater than 10 point margin each time), and polling well for Obama now:

California - 55 electoral votes
Massachusetts - 12 electoral votes
Connecticut - 7 electoral votes
New York - 31 electoral votes
Illinois - 21 electoral votes
Vermont - 3 EV
Rhode Island - 4 EV
Delaware - 3 EV
Maryland - 10 EV
Hawaii - 4 EV

Running Obama Total Electoral Votes: 150 (270 needed to win)

Solidly Republican in 2000 and 2004 (greater than 10 point margin each time), and polling well for McCain now:

Texas - 34 electoral votes
Mississippi - 6 electoral votes
Kentucky - 8 electoral votes
Kansas - 6 electoral votes
Alaska - 3 electoral votes
Alabama - 9 electoral votes
South Carolina - 8 EV
Oklahoma - 7 EV
Nebraska - 5 EV
Wyoming - 3 EV
Utah - 5 EV
Idaho - 4 EV

Running McCain Total Electoral Votes: 98 (270 needed to win)

Now that we’ve grouped what we feel most confident about, it’s time to take a step down in confidence and continue to build our forecast. The next level down will be states that showed moderate or weak wins, or even losses, for the candidate’s party in ‘00 and ‘04, but are polling strongly now for the candidate. As we add these states in, keep in mind that this is with increased uncertainty, and we will address that with analysis beside each state of October poll data - anything older has been discounted.

Showing Moderate/Weak/No Support in ‘00/’04, now polling well for Obama:

Minnesota - 10 electoral votes - has voted slightly in favor of Democratic candidate in past two cycles, and state polls range from Obama +6 to Obama +19. We’ll call it safely Obama.
Wisconsin - 10 electoral votes - also went slightly in favor of Democrats last two cycles, Obama up 8 to 14 points. Safely Obama.
Michigan - 17 electoral votes - went Democrat both cycles. Obama up 16 to 22 points. Safely Obama.
Iowa - 7 electoral votes - barely Bush in ‘04, barely Gore in ‘00. Obama up 13 points in two polls. Forecasting sense says beware of such a closely divided state, but the polls suggest this should be safely Obama.
New Jersey - 15 electoral votes - Moderate Democratic ‘04, strong ‘00. Obama up a wide ranging 8 to 23 points. Appears to be safely Obama.
Oregon - 7 electoral votes - weakly Kerry and Gore, but Obama up 13 to 17 points. Appears to be safely Obama.
Washington - 11 electoral votes - Moderate wins for Democrats last two cycles, Obama up 10 to 19 points. Safely Obama.
Pennsylvania - 21 electoral votes - weak to moderate support for Democrats last two cycles, Obama up 8 to 13 points. Appears to be safely Obama, though it’s a state to watch with McCain making a push.
Maine - 4 electoral votes (electoral votes may be split by region in the state) - Solidly Democrat last two cycles. Obama up 12 to 17 points. Safely Obama, but one electoral vote still up for grabs in Eastern ME.

Running Obama Total Electoral Votes: 252 (270 needed to win)
Flexible Obama Total Electoral Votes - High End (adjusted for “toss up” electoral votes): 252
Flexible Obama Total Electoral Votes - Low End: 251

Showing Moderate/Weak/No Support in ‘00/’04, now polling well for McCain:

Louisiana - 9 electoral votes - strong, then moderate Republican support last two cycles, NO October poll taken in this state - September poll had McCain up 15 points. We’ll assume safely McCain.
Arkansas - 6 electoral votes - weak to moderate Republican last two cycles, NO October poll, McCain +9 in September poll, down from +16 in March. Will assume safely McCain.
Tennessee - 11 electoral votes - solid Republican in ‘04, weak Republican in ‘00, McCain up 12 points, safely McCain.
Arizona - 10 electoral votes - Moderate Republican both cycles, NO October poll, late September polls had McCain up 7 to 21 points - a huge range, but we’ll call it safely McCain in his home state.

Running McCain Total Electoral Votes: 134 (270 needed to win)
Flexible McCain Total Electoral Votes - High End (adjusted for “toss up” electoral votes): 134
Flexible McCain Total Electoral Votes - Low End: 134

This leaves a number of states unaccounted for. It’s interesting to note that Obama is running strongly in all states that were strongly Democratic in the previous two election cycles, but John McCain cannot say the same on the Republican side. For instance:

North Carolina - 15 EV - Favored Bush by about 12 and a half points both election cycles, now state polls range from a tie to Obama up by 4 points. There is no definable trend in recent polling. Though the polling data favors Obama slightly, the strong history of voting for Bush is great reason for pause. McCain held a lead here until mid-September. With such history of voting Republican, I actually would expect McCain to pull this one out. Though I’ll give him the electoral votes, the electoral votes will also be considered in both candidates’ high and low range.
Obama Running Total, then (high/low) range: 252 (267/251)
McCain Running Total, (high/low) range: 149 (149/134)

Indiana - 11 EV - There are wild swings in the state polls, ranging from McCain up 5 to Obama up 10. There is no clear trend, only wild and unpredictable swinging. With the polls rather useless at this point, one has to look to voting history in the state. Bush by 16 in ‘00, Bush by 21 in ‘04. Ingredients-based methodology says this goes to McCain.
Obama Running Total, then (high/low) range: 252 (267/251)
McCain Running Total, (high/low) range: 160 (160/134)

Georgia - 15 EV - The trend has been rather steady here, but it’s been steady with McCain leading between 3 and 8 points. The history was Bush by 12 and 17 in the last two elections. This should go to McCain.
Obama Running Total, then (high/low) range: 252 (267/251)
McCain Running Total, (high/low) range: 175 (175/134)

Montana - 3 EV - McCain has been leading here, and still leads by 4 to 8 points, but the trend has been for that lead to shrink. The margin in ‘04 and ‘00 was Bush by 20.5 and 25 points, respectively. Though the shrinking lead for McCain implies Obama could take this state, the history is strongly against it. Call it McCain.
Obama Running Total, then (high/low) range: 252 (267/251)
McCain Running Total, (high/low) range: 178 (178/134)

District of Columbia - 3 EV - No polls have been run here. Is expected to go strongly Obama, however.
Obama Running Total, then (high/low) range: 255 (270/251)
McCain Running Total, (high/low) range: 178 (178/134)

West Virginia - 5 EV - McCain’s lead varies from +2 to +9 in the polls, but the trend is for this lead to increase. The history here is Bush by 6, and Bush by 13 in the last two election cycles. That’s a ringing Republican endorsement, and with McCain’s lead growing, it makes sense he’d take the state.
Obama Running Total, then (high/low) range: 255 (270/251)
McCain Running Total, (high/low) range: 183 (183/134)

Now things get tricky, as we’ve entered the battleground states. All of the methodology in the world really shouldn’t be able to offset the inherent uncertainty. Therefore, I will continue expanding ranges, in addition to the actual running total electoral vote forecast.

New Hampshire - 4 EV - Has long been a battleground state and went Republican in ‘00 and Democrat in ‘04. Obama is up by 7 to 10 points in the polls but the trend is downward. Speaking solely from an analysis standpoint, this is a state that’s becoming more and more up for grabs, but as of this writing, would have to guess Obama by a hair.
Obama Running Total, then (high/low) range: 259 (274/251)
McCain Running Total, (high/low) range: 183 (187/134)

Virginia - 13 EV - Another battleground. Bush by 8 points in the last two election cycles. Obama up in the polls between 2 and 10 points, but the trend is down. In this case, given the strong Republican history and a falling lead for Obama, I would think McCain pulls the state.
Obama Running Total, then (high/low) range: 259 (287/251)
McCain Running Total, (high/low) range: 196 (200/134)

South Dakota - 3 EV - McCain is leading by 7 in the latest poll, though the trend is for a shrinking lead. Still, this traditionally Republican state should be carried by McCain.
Obama Running Total, then (high/low) range: 259 (290/251)
McCain Running Total, (high/low) range: 199 (203/134)

North Dakota - 3 EV - Polls show somewhere between a tie and Obama leading by 2. The trend is for a decreasing lead, however, and the history is Republican. Like its neighbor to the south, probably going to McCain.
Obama Running Total, then (high/low) range: 259 (293/251)
McCain Running Total, (high/low) range: 202 (206/134)

New Mexico - 5 EV - With polls showing Obama up between 7 and 13 points, and an increasing trend, it looks like this previously Republican state in the next election will go Democrat. Give it to Obama.
Obama Running Total, then (high/low) range: 264 (298/251)
McCain Running Total, (high/low) range: 202 (206/134)

Florida - 27 EV - The state that was barely Bush in ‘00 and more strongly Republican in ‘04 has come back with varying poll numbers, but McCain held the lead until the end of September, lost it through a couple of weeks in October, and now the trend is in his direction. A Republican history coupled with a trend back toward what was a low-level steady-state McCain supporting Florida favors McCain taking this one.
Obama Running Total, then (high/low) range: 264 (325/251)
McCain Running Total, (high/low) range: 229 (233/134)

So, with a forecast electoral vote of Obama at 264 and McCain at 229, and McCain’s high end estimate lower than Obama’s low end estimate, the following battleground states remain in play:

Ohio, Missouri, Colorado and Nevada.

Six electoral votes gives Obama the Presidency. That means McCain would absolutely have to win Ohio, Missouri AND Colorado. Obama winning any of these would secure the Presidency, under my forecast, while only winning Nevada would leave him one electoral vote shy. Keeping in mind the perspective of a meteorologist, this equates to the following: Winter. A strong Canadian disturbance and a strong Pacific disturbance both come racing into the Eastern United States. Cold air follows the northern storm, moisture follows the southern disturbance - together, they’d create a blockbuster storm (in this case, let’s call it a McCain comeback win). But everything would have to come together perfectly - the warm, the cold, the moisture, and the timing of the disturbances. Right now, the statistical and numerical guidance strongly favors the storm developing 300 miles east. For those who say “don’t explain the weather, just give me the forecast in 30 seconds,” this equates to: “Partly sunny, storm stays east.” For those who savor the juicy details, all the components are there - evident on satellite and radar, and evident in the observations. The question is whether they’ll come together.

Nevada - 5 electoral votes - Recent polls range from a tie to Obama ahead by 5. Though Nevada is historically Republican, and the Republican gubernatorial candidate leads in the polls, there’s been a fairly persistent signal of Obama ahead in this state. I would assume he’ll take it, but it will contribute to both candidates’ ranges.
Obama Running Total, then (high/low) range: 269 (330/251)
McCain Running Total, (high/low) range: 229 (238/134)

Missouri - 11 EV - McCain lost a long-standing lead in this state at the end of September. There are signs he’s coming back to life in the latest polls - some of which put McCain in the lead - but those polls showing Obama with a lead are indicating a greater lead than he’s ever enjoyed in Missouri. This has the makings of a duel to the death. The nature of the state would tilt it back toward Republican - there’s not much time to do that and Obama’s numbers appear to be growing a bit in some polls - but further analysis shows these numbers as outliers in the last few weeks. Believe it or not, when outliers are removed, the state is a 1 point race and tightening. Believe it or not, a forecasting perspective encourages a McCain win in Missouri.
Obama Running Total, then (high/low) range: 269 (330/251)
McCain Running Total, (high/low) range: 240 (260/134)

Ohio - 20 EV - The battlegrounds continue. These polls range wildly in recent days, from McCain by 2 points to Obama by 14 points! There is no reliable trend. The state’s history is weak support for Bush in both elections. Ohio will continue to be a pivotal state and is nearly impossible to call from a numerical standpoint. Another duel to the death, with the ground support of the Obama organization favoring his campaign. Missouri and Ohio equate to the energetic disturbances aloft - vigorous, and though small by themselves, if they combine forces in the right way, the makings of a blockbuster storm for McCain.
Obama Running Total, then (high/low) range: 289 (330/251)
McCain Running Total, (high/low) range: 240 (280/134)

Colorado - 9 EV - This state has shown consistent Obama support for several weeks, and though McCain erased the deficit in some polls for a brief time, the trend has since stagnated at an Obama lead of 4 to 9 points. Pending a game changer or some serious Colorado stumping from McCain - especially considering Biden was in Pueblo today - this is probably going to Obama, even though the Rocky Mountain state went 8 points in favor of Bush in ‘00 and 5 points in ‘04.
Obama Running Total, then (high/low) range: 298 (339/251)
McCain Running Total, (high/low) range: 240 (285/134)

So at the end of the day, this forecaster has what appears to be a decisive victory for the Democratic candidate, Barack Obama, and John McCain’s comeback blockbuster storm misses coordinating by small amounts in some spots, which compound over time. There are, however, some VERY important caveats that keep uncertainty in the forecast, and it really does go back to the battleground states, which I view very much like the storm ingredients. Consider the following: just like in meteorology, we know that our observations, let alone the forecast guidance, is not completely accurate. In fact, from the observations (the polls) consider the following Real Clear Politics poll averages with just under two weeks until the election:

Colorado: Obama +5.4 (lead stable)
Ohio: Obama +6.0 (lead stable or slightly growing)
Florida: Obama +1.0 (lead shrinking)
Nevada: Obama +3.3 (lead shrinking)
Missouri: Obama +2.7 (lead stable)
North Carolina: Obama +2.0 (lead shrinking)
Virginia: Obama +7.0 (lead shrinking)

Next, consider the adjusted numbers if we remove “outliers” - or polls that are substantially removed from the pack:

Colorado: Obama +4.5 (lead stable)
Ohio: Obama +2.5 (lead stable or slightly growing)
Florida: Obama +1.0 (lead shrinking)
Nevada: Obama +3.3 (lead shrinking)
Missouri: Obama +1.0 (lead shrinking)
North Carolina: Obama +2.0 (lead shrinking)
Virginia: Obama +7.0 (lead shrinking)

This implies Ohio should be the extremely close race that’s expected, and may go to McCain, rather than to Obama, as I’m expecting. The keys for Obama are to keep everything exactly the way they are for two weeks - that would ensure a win and open the door to the landslide (note that Obama’s high range forecast for electoral votes is a whopping 339 if he pulls out all of the states he’s contending in!). So, that’s what you Obama supporters are looking for - stability or padding the lead. McCain supporters should be watching all of these numbers, though from a forecasting perspective I’ve given you my best estimate, which means you’ll want to watch Colorado and Virginia, because McCain would either need both, or just Virginia if everything else goes as I’m expecting.

Of course, there could be surprises in any one of the more uncertain races, including New Hampshire. Many pundits still put Pennsylvania in play, with good reasoning, but unless there’s some sort of skewing in the polling data I’m unaware of, it seems like the temperature is about 45 and the winds have shifted to the north - meaning colder air is on the way, but I’m not sure the critical freezing mark will be reached before the storm shuts down.

Keep in mind that I’m not a political analyst by trade, nor do I know the delicate workings of precincts and the like. Rather, I’m an analyst of uncertain and seemingly unpredictable circumstances on a daily basis, and see a strong connection between these two seemingly unrelated topics.

Of course, the most fun part of weather…and politics…is waiting to see what really transpires. And I know we have plenty of weather watchers in the field for this one.

Matt