Saturday, October 6, 2012

Update on the Polls


From my last post, the polls on the Presidential race slightly fluctuated during the last two weeks, but more importantly, this is the last post where I'm going to refer to Silver's graph showing when the Convention bounces will evaporate:


2012: 09/20-09/24

What's important to note from the graph is that Convention bounces were predicted to have faded by the 28th of September. In line with the graph, the day following my last post brought a set of polls that prompted Silver to make an interesting post wherein he said, “The. Polls. Have. Stopped. Making. Any. Sense.” The reason that NYTimes blogger wrote that particular line is because at the time pollsters released data that contradicts their fellow pollster's data. For example, a Marquette University Poll and Real Clear Politics each put out polls showing President Obama with a 14 point lead and the other put Governor Romney ahead by 3 points. What is amusing about how Silver rationalized disagreement between the polls is how he used Statistical Analysis to say that one pollster sometimes releases data that lies outside the consensus of several pollsters. In fact, Silver pointed out that other polls released in the same day showed the President with a lead. What that means is that Silver is correct in asserting that some polls do fall outside the consensus of other pollsters, but at the end of the day his graph correctly predicted the President leading the polls on that day.


Again looking at the graph, the President was expected to have declining numbers; yet, on the 20th of last month just the opposite happened. Silver put out a post that argued that Obama continues to get favorable numbers from the polls. Without reading too much into Silver's article, perhaps Romney's comment about 47% of the country did boost the President's polling. Another possibility as to why the polls are showing the President with a lead is because the Conventions ended about two weeks earlier, but that does not explain improvements in Obama's poll numbers. Finally, the incident in Benghazi could be providing a "rally behind the flag effect" in favor of the president---which does not favor the challenger. What's clear from the graph is that the President is expected to be leading, but his improving lead in the polls postpones when the President's polling will decline.

Even during the weekend that started on the 21st of last month, Obama's numbers kept improving, even if just marginally. Though the polls did not change radically, the fact that the President's numbers actually improved during that weekend means that he's slowing down how fast his Convention bounce is dissipating. Again, the graph from the FiveThirtyEight forecast projects that the President's bounce should decline, but the President's poll numbers show otherwise. What's more is that on that same day Silver added about 100 more polls to his forecast model, thus, making his predictions more robust. Silver also made a point on the following day, however, that makes sense from impartial standpoint and that is that days where polls hold even count as a lost day for Romney. The blogger's statement should be taken with caution, lest readers claim that the NYTimes blogger has a bias for the President and a bias against Romney. On the contrary, Silver made it explicit that his articles will have a good tenor to whoever the polls are favoring at the time.

Most telling off all the articles posted during the last two weeks was one published on the 24th of September because in the article Silver wrote that historically during this time in the Presidential Election polling was robust enough to predict who will win in November. Here's a chart that Silver put together that showcases his point: 


There's been two instances where polling data from 45 days before the General Election did not predict who would win in November.  The anomalous case happened when each Dewey and Gore led 45 days before the Election, but Truman and Bush, respectively, prevailed at the end of the season. The table shows the incumbent party in purple and the challenging party in orange. The "Leader" column reads what candidate was leading the polls and by what amount. Discarding the outlying Elections, the Presidential candidates who led 45 days from the Election won in November. In regards to the data, the NYTimes blogger made a humorous statement when he wrote, "Of course, if Mr. Obama led in the race by 30 percentage points — as Lyndon B. Johnson did in 1964 — there wouldn’t be much need for such detailed analysis, and FiveThirtyEight might be free to blog about the baseball playoffs." Silver is well known for his addiction to baseball, but the point he is trying to convey is that if President Obama were leading in the polls with 30 points, then the President would have a lot room for error to win in November. More ominous is the fact that the NYTimes blogger wrote that, historically, at this point in the race there has not been a tendency for the Election to go the challenger who is trailing in the polls.

Again, in the same post Silver pointed out that incumbents have had a larger lead during this time in the General Election, however, the NYTimes blogger does point out that the President is running the country under a recession, thus, explaining the President's relatively-mediocre polling. In addition, the post alluded to another article that Silver wrote in regards to poll volatility because he mentioned that late changes in the polls are not frequent because voters are tuning in earlier to the General Election and because voters are more polarized. Another point that the blogger makes is that the number of voters declaring themselves 'undecided' is misleading because people choose third party candidates (in polls that include a third party candidate's name) in protest, but fear throwing away their votes and opt for a major party in November. Overall, the 24th of September brought good numbers for the President which falls in line with historical data, that is, historically 80% of candidates who led the polls 45 days before the General Election won the vote in November (the 80% probability is in line with the FiveThirtyEight forecast that has the president winning reelection with a 77% chance).

2012: 09/25-09/27

The week on which Obama's lead was predicted to decline brought a few poll numbers putting him ahead in Ohio, but more importantly the week brought many posts by Silver including one that put forth a strategy for Romney to win the Election without Ohio. I won't go into the details of that particular post because the good NYTimes blogger does a better job at explaining how Romney can win the General Election without winning Ohio. What's important to take from the post is that if Romney loses Ohio, then Romney cannot lose a combination of Iowa and Wisconsin. In fact, it is through Iowa that Silver says that Romney can hope to win the Election given Romney loses Ohio. The polls released during the day favored the President, meaning, that Silver is actually predicting how the Romney campaign will react to changes in the polls, particularly, in Ohio.


Keeping in spirit with giving a candidate a good tenor given the day's polls, Silver gave Governor Romney a grim outlook for November, especially, if the poll numbers continue to come out as they did on the 26th of September. The article for this day actually goes more into pollster's methodology and requires some patience to read because Silver specifically addresses how the President can over perform his poll numbers compared to his poll numbers in 2008. Indeed, Silver makes a concession and says that writing about the Electoral math is academic and will not appeal to the average reader. Yet, that did not stop NYTimes blogger from stating that if the President continues to get solid poll numbers from two more swing states, then the President will win in November. Some positive news for the incumbent President is that the President will over perform his numbers on Republican leaning states like Florida, Ohio, Alaska and Arizona, but what is disturbing from reading the post is that Silver wrote, in a clinical manner, that states such as Kentucky and Tennessee will have better numbers for Obama in this Election because the race of a candidate is important in those states (people in these two states tend to vote for Whites according to Silver), thus, voters in Kentucky and Tennessee will be more comfortable voting for an African-American President.

Oddly enough, on the day that the President's Convention bounce is supposed to evaporate Silver published a chart that superimposes political happenings on top of the Popular Vote curve. Below is a graph from Silver's Now-Cast:


The reason why Silver published this chart in his article is because he wanted to make a point about how a gaffe's effect dissipates as November approaches. The blogger's exact words were, “Ninety percent of ‘game-changing’ gaffes are less important in retrospect than they seem in the moment.” The graph above shows the contrary to Silver's statement---that is, in this particular gaffe the polls shifted towards President Obama---because Romney's poll numbers were declining after his comments on the Benghazi incident. Furthermore, what is evident is that the 47% comment did not help Romney's numbers, especially, at a time when he was recovering from his Libya comments. Part of the reason that Romney may have had declining numbers during the Benghazi incident is because the News media criticized Governor Romney harshly. Luckily enough, however, Silver has two forecasts: the Now-cast and the Nov. 6 Forecast, where one predicts what would happen if the election were held today and the latter predicts the results in November. Romney's numbers look much better for the November forecast than they do in the Now-cast.

2012: 09/28-09/30

If the Now-Cast brought bad news for Romney, then November Forecast should be favorable because the 28th of September is the long awaited weekend projected by the FiveThirtyEight forecast where the President's Convention bounce dissipates. Coincidentally, the 28th brought economic data and poll numbers that made the President's chances of winning the Electoral College fall from 83.9% to 82.7%. Without a doubt Silver had his comments on the matter and his thoughts are that President Obama will continue to see erosion in his poll numbers. One of the more striking statements in Silver's post is that he mentioned that voters, specifically Democrats, are optimistic about the direction of the economy and this is thanks to Bill Clinton and President Obama because both expressed optimism in their 2012 Convention speeches. The blogger made the point that optimism in the economy is an impetus to get people to start spending, thus, jump starting the economy. From a psychological standpoint, Obama and Clinton's speeches will have a self-fulfilling effect in the sense that optimism will improve confidence in the economy. In regards to the FiveThirtyEight forecast, the November forecast takes into account economics factors, which is why Silver mentions the economy in his post.


Not only did the weekend bring favorable numbers to the President, but on Saturday there was more bad news for the Romney campaign because a poll came out showing Obama with a lead in Iowa. On the post relevant to the day, Silver again reiterated how Romney cannot lose Iowa because from 25,001 simulations; Romney only won 2% of the time given he loses Iowa. From the three articles pertaining to the weekend on the 28th of September, one post particularly simplified the General Election because the post is written with a sports perspective. For example, Silver likened the General Election to a football game divided into quarters wherein the first quarter consisted of the Republican primaries, the second quarter consisted of announcing the Republican Vice-Presidential Candidate, the third quarter being the Party Conventions, and the final quarter is the one we are currently experiencing at this stage of the Election where Convention bounces have dissipated. Comparing the election to a football game sounds unorthodox, but the manner in which Silver compares the Election gives some insight as to how to think about the win probabilities, for instance, Romney has a 15% chance of winning which is approximately equal to the chances of a team down by a touchdown (with 10 minutes remaining) winning just 16 percent of the time.

2012: 10/01-10/03

If any good news came for Romney it was from the Monday leading into the first 2012 Presidential Debate because the day brought a post by NYTimes blogger wherein he discussed the possibility of an Electoral tie---a situation that would lead a Republican-majority House of Representatives to select the next President of the United States (POTUS). Silver pointed out that letting the House of Representatives to select the President is problematic because the party controlling the House will elect someone from their party, in this case Romney. Moreover, the NYTimes blogger pointed out that the decision to elect Romney would be controversial, especially, if President Obama wins the Popular vote by two or three points. From 25,001 simulation a tie in the Electoral College happened 0.6% of the time or about 152 times. It's safe to say that there will be definitive leader in the Electoral College because the number of occurrences for a tie in the Electoral College is small.


Finally, the Tuesday leading into debate also brought a post by Silver wherein he wrote that Romney's numbers are starting to improve. The fact that Romney's numbers are improving implies that the President's lead is going to deteriorate within the next few weeks. In fact, the days following the first Presidential Debate will bring poll numbers that will reinforce that trend, especially, if voters perceive the President as losing the Debate. What's more is that on the day of the debate Silver published an article wherein he states that the first Presdential debate in the General Election usually favors the challenger---which is true in this election because the first debate in the 2012 Presidential Debate will be about Domestic Policy, meaning, Romney will hold the advantage due to the fact that the most important domestic issue is the economy. The most ominous statement Silver made in his article is that no candidate trailing as much as Romney heading into the first debate went on to win the General Election. If it's any consolation to Romney, candidates who do well in the first debate improve their number by 3 points.

No comments: