Election Eve:

of support is within three percentage points of his observed level of 52% (i.e., between 49% and 55%). A national poll about the presidency that is accurate to within three percentage points would have a sample size of about 1500 .<br><br> That works out to one person selected for every 200,000 Americans. In other words, there would be roughly one person selected from a city the size of Boise Idaho or Rochester, New York. In national polling, lots of results that a highly inaccurate at the local level are combined to produce a result that is highly accurate nationally.<br><br> Moreover, the phrase ctoo close to call d is used rather loosely: If candidate A leads candidate Q by 52-48 in a poll with a 3-point margin of error, then, under standard assumptions, the probability is 91% that A is actually ahead in the full electorate. On 9/11/08, the website Real Clear Politics offered the following polling results about the popular vote standings of the candidates: McCain 46.2% Obama 43.8% But the president is chosen based not on popular votes but rather electoral votes . What can we infer from these numbers (Mc:46.2%, Ob: 43.8%) about the chance that McCain was ahead in electoral votes on September 11?<br><br> Care to guess? If you answered cvery little, d you get full credit! The website Electoral-Votes.com offered tallies for 9/11/08 that are in principle more illuminating: McCain 238 electoral votes Obama 273 Toss-Up 27 These tallies were based on the latest polling results in each state.<br><br> Whichever candidate was ahead was assigned all of the state 9s electoral votes, with no regard to the size of his lead . Why not conduct presidential preference polls with the explicit aim of estimating each candidate 9s electoral vote share? Two key points: " The ndings for individual states will be unreliable because they are based on small samples.<br><br> But that statement is already true for popular-vote polls. " Sampling theory provides estimates of the probability that each candidate is ahead in each state. That is good news.<br><br> For example, the poll might suggest that: The probability is 66% that candidate A is ahead in Connecticut, which has 7 electoral votes . Thus, candidate A will get 7 electoral votes from Connecticut with 66% probability and 0 electoral votes with probability 34%. With a little bit of Math and a poll of usual size, we can get: " An unbiased cpoint estimate d of how many electoral votes candidate A would receive in total if the election were held today.<br><br> " A strong estimate of the probability that candidate A is ahead (i.e. has at least 270 out of the 538 electoral votes). If we process the polling results in each state to estimate the probability that each candidate is ahead there, we reach these outcomes: " The probability is 76% that Obama was leading McCain in electoral votes as of 9/11 .<br><br> " The unbiased estimate of the distribution of electoral votes was Obama 287 , McCain 251 . But is there any alternative to the Electoral College that has a real chance of being implemented ? Hard to say, but let us get ready to consider one alternative by recalling what a weighted average is.<br><br> Possible cure: Candidate 9s national vote share is a weighted average of his vote shares in individual states, with the weight for each state proportional to its share of electoral votes (I.e. number of members of Congress) Weighted vote share (WVS) follows : WVS = e i p i (i=1,..51) Where e = electoral vote share and p = popular vote share For example, California 9s share of electoral votes is 55/538 = 10.2%, while Wyoming 9s share is 3/538= 0.6% Thus, in computing the national weighted average, California 9s popular vote result gets a weight of 10.6%, while Wyoming 9s gets a weight of 0.6 %. A candidate 9s weighted vote share would be expressed as a percentage .<br><br> The candidate with the highest WVS would be declared the winner of the Presidency. 2004 Election ActualWeighted Vote Share Vote Share George W Bush 50.7%51.0% John Kerry 48.3%47.9% So what? WVS has many advantages : (1) Unlike a shift to a popular vote election, WVS would increase the power of the small states rather than decrease it.<br><br> Percentage In uence of Small US States in 2004 Election, Under Three Criteria Electoral % of % ofWVS Votes Popular Vote Electoral Votes Impact Share 3 (8)2.2%4.5%10.2% 4 (5)2.4%3.7%4.3% 5 (5)3.3%4.7%6.8% 6 (3)2.8%3.4%4.7% Total10.7%16.3%26.0% (Numbers in parentheses show how many states have the corresponding number of electoral votes.) For elaboration of this chart, see Barnett and Kaplan (2007). (2) Yet the largest states would also gain under WVS 2004 Election California Irrelevant Texas Irrelevant New York Irrelevant Florida Hyper-swing Illinois Irrelevant (These four cirrelevant d states contain 31% of the US p o p ulation. ) (3) WVS would all but eliminate the worst consequences of the cwinner take all d rule.<br><br> (4) There would be no danger that the election would go to the House of Representatives, where the President would be chosen under bizarre rules . On average, a factor of 50 closer! WVS and popular vote shares differed by 2/5 of one percentage point in the elections since 1960.<br><br> (5) WVS would far more closely track the popular vote split than does the Electoral College . But aren 9t there simpler revisions to the present arrangements that would achieve the goals of WVS? Three recent proposals: " Move to popular vote without a constitutional amendment through a clever compact among larger states.<br><br> " Allow the winner in each congressional district to get the electoral vote that corresponds to that district. " Apportion a state 9s electoral votes in proportion to the state 9s popular vote outcome (rounding off to the nearest whole number). To put it brie y, these proposals have no future .<br><br> So, where are we?

less