Adrian Kavanagh, 29th June 2016 (Updated 6th July 2016 – see bottom of post)

As we head towards the 2016 US Presidential Election in November, there seems to be different messages coming from national-level and state-level opinion polls.

With well over four months left until the US presidential election on 8th November (the Tuesday following the first Monday in November), the national-level opinion polls are pointing to a clear win for Hilary Clinton. As data from RealClearPolitics shows, Clinton at present enjoys a lead over Donald Trump in national opinion polls that is currently averaging out at just under seven percent, although this lead narrows somewhat to around five and a half percent when Third Party candidates such as Libertarian Party candidate, Gary Johnson, and Green Party candidate, Jill Stein, are included.

Figure 1: RealClearPolitics overview of US Presidential Election national opinion polls (as of 29th June 2016)

If these figures were to be replicated in the November election, the expectation would be that Clinton would enjoy a commanding win in the Electoral College. However, even on these poll figures, in a uniform swing scenario (similar to the constituency-level analysis that I use to translate Irish opinion poll levels into potential Dail-seat levels) this level of Clinton/Trump support would not necessarily translate into a much larger Clinton win in the Electoral College than that which was won by Barack Obama over Mitt Romney in the 2012 election. In part, this reflects Obama’s success in 2012 in winning nearly all of the competitive swing states, or purple states, in that contest, with the exception of a narrow loss to Romney in North Carolina.

Table 1: Margin of Victory in US states at the 2012 presidential election and Estimated Margin of Victory in US states for the 2016 election, based on uniform-swing analysis of national-level poll figures. (Positive sign means a Democrat win, negative sign means a Republican win)

Table 1 shows the margin of victory for the successful candidate at the 2012 election in each of the US states and in the District of Columbia. (In this Table, a plus sign points to a Democrat/Obama win, while a minus sign points to a Republican/Romney win.) Table 1 also shows an estimated margin of victory for Clinton/Trump in these different states, applying a uniform swing assumption to translate the current average national poll levels in to state level estimates. With the lead currently enjoyed by Clinton over Trump in the national polls being larger than the margin of the Obama victory in 2012, this means that the analysis would expect her to be faring better in all of these states. Winning margins in blue states are estimated to be larger, while winning margins in red states are estimated to be smaller. Some of the red states that were won by Romney by relatively comfortable margins in 2012 are now expected, based on the figures produced in this analysis, to come “into play” in the 2016 contest, including states such as Georgia, Arizona, Missouri, Indiana (which Obama did win in 2008), South Carolina and Indiana. In the swing states/purple states that were won by Obama in 2012, the margin of victory changes from a narrow one to a rather more comfortable one in the 2016 support-level estimates. But…and this is an important point to note…the only state that would change hands in this analysis is the only purple/swing state won by Romney in 2012 – North Carolina. If this scenario panned out, Clinton would win the Electoral College vote by a 347-191 margin – a slightly larger margin of victory than Obama’s 332-206 margin of victory in the Electoral College at the 2012 election.

One thing that US politics does very well, however, involves the use of state-level opinion polls in the months leading up to US presidential election contests. State-level contests are crucial in the “winner takes all” Electoral College system and – especially in the context of a series of close elections from the 1992 election onwards – the contests in the most competitive states – the swing states or purple states – are even more crucial. With the Electoral College system, the winning candidates is not necessarily the candidate who wins the most votes, but is instead the candidate who wins the most Electoral College votes. Each state is allocated a number of Electoral College votes. (In the states of Maine and Nebraska some of the Electoral College votes are determined by the results in those states’ congressional districts.) The number of Electoral College votes per state, which is based on the number of Senators and Congress members elected by each state (and the District of Columbia), is roughly proportional to each state’s population, but not directly related, given that even the District of Columbia and the very smallest (population-wise) states (such as Wyoming and Vermont) all have three Electoral College votes. In this winner takes all electoral system, the candidate who wins the most votes in a state takes all that state’s Electoral College votes – no matter how large or how small their winning margin is. For instance, George W Bush won all the Electoral College votes in Florida in 2000 (and the US presidency, as a result) by a margin of just 537 votes. In the increasingly polarised political US landscape of the late 1990s and early 2000s, with a number of “red states” and “blue states” being won by consistently large margins by the Republican and Democrats candidates, respectively, the contest for the presidential elections has increasingly boiled down to the results of the contests in the handful of competitive swing, or purple, states, as is usually reflected in candidates’ campaigning and campaign spend geographies in the run-up to the November election.

This is why a focus on state-level support trends – with specific reference to these swing/purple states – is important in the context of a US presidential election. Hence, state-level opinion polls warrant a great degree of focus in the run up to the November election. In 2012, opinion polls in the swing states were more likely to suggest that Obama would win the Electoral College than the messages that were being gleaned from national-level polls, which tended to put Romney ahead in the weeks leading up to the actual contest. Even though there was a swing against Obama nationally relative to his stronger performance in 2008, he comfortably won the Electoral College vote in 2012 because he managed to fare better where it mattered – the swing against him was not as notable, by and large, in the swing/purple states and in the end he only lost North Carolina and Indiana out of the states that he had won in the 2008 contest. Effectively, Obama performed better in the swing/purple states than national opinion poll support trends would have suggested. So, votes obviously matter when it comes to determining who wins a presidential election contest. But another important aspect relates to where you win those votes. Candidates with a more efficient geography of support – where votes are not “wasted” in large majorities in your strong states/narrow losses in the swing/purple states – will be at an advantage in this electoral system – as with Obama in 2012, but also Bush in 2000.

Figure 2: RealClearPolitics overview of US Presidential Election state-level opinion polls (as of 29th June 2016)

Given that, how do current state-level opinion polls support trends compare with the state-level support estimates (as listed in Table 1) that would be expected based on current national opinion poll levels? Some interesting trends emerge here and these suggest that the contest between Clinton and Trump might be a lot closer than the national opinion poll trends would suggest.

Table 2: Support trends in state-level US presidential election polls (as detailed on the RealClearPolitics site) compared with state support estimates based on uniform-swing analysis of national opinion polls, as of 29th June 2016 (Where a number of polls have been carried out over the past few weeks in a state, the state-level poll figure quoted here is the RCP-average for that state – for instance, see here for the Pennsylvania figure.) 

Table 2 suggests that the Trump support levels are holding up better in most of the swing/purple states than would be expected arising out of the analysis of trends in the national opinion polls. For instance, the state-level polls points towards much closer contests in states such as Colorado, Pennsylvania, New Hampshire and Ohio than would be estimated based on the analysis of the national opinion poll trends. Furthermore, the national analysis suggests that the Democrats would gain North Carolina, but the state level polls suggest that Trump is ahead there – albeit by a very narrow margin.  This does not all amount to good news for Trump however. The Arizona polls show that he is trailing Clinton in a state that Romney won by a margin of nine percent in 2012 and a state where the national-level poll analysis would be still expecting him to win in by a somewhat comfortable margin. The Latino-vote factor could well be playing a role here, as indeed could also be the case in the much reduced size of the Republican lead in the state of Texas, which is suggested by state-level polls there. One other interesting trend to be evidenced here is the suggestion that the size of the Democrat lead in the blue states might, in reality, be much smaller than would be expected based on national opinion poll trends, as evidenced in the recent polls in New Jersey, Connecticut and California. On the other hand, the size of the Republican lead in the red states also appears to be much smaller than would be expected based on national opinion poll trends, as evidenced in the recent polls in Utah, Arkansas and Texas. From the perspective of Trump, much-reduced margins of victory in the red states are not necessarily a bad thing – unless the reduction in the Republican lead ultimately pushes these states “into play”, as seems to be the case in Arizona, at least – as in the Electoral College system a “win is a win”, no matter how large or how small the margin of victory is. These trends – and it is worth noting that we are still very early in the contest here, with 131 days still left until the actual vote on November 8th – could suggest that the unusual nature of this particular election contest is changing the rules of the game – at least to some extent – in terms of the Republican/Democrat geographies of support. One potential knock-on effect could be to leave Trump with a more “efficient” geography of support than that the 2012 Romney support pattern. In this scenario, a recovery in Trump national poll figures could see contests in the swing states become even more competitive than the current state-level poll figures are suggesting.

Ultimately, while national opinion poll figures may be suggesting that the US presidential election contest is already “game over”, the state level polls suggest instead that it’s very much a case of “game on”…

Update (6th July 2016):  National opinion polls in the final week of June and first few days in July have also consistently pointed to a Clinton lead of a few percentage points, but the gap between Clinton and Trump has narrowed somewhat. The RealClearPolitics Poll Average on 6th July suggested that the gap had narrowed to just under five percent; with Hilary Clinton estimated to be at 44.9% and Donald Trump to be at 40.3% (4.6% gap).

Table 3: Support trends in state-level US presidential election polls (as detailed on the RealClearPolitics site) compared with state support estimates based on uniform-swing analysis of national opinion polls, as of 6th July 2016 

Ironically, over the very same time period, Clinton’s performances in opinion polls in the swing, or purple, or toss up, states has, on average, have improved somewhat, although it is worth noting that these polls were carried out before the publication of the FBI findings and recommendation on the Clinton email server controversy. At this point, research from fivethirtyeight.com suggests that Clinton is, on average, currently polling much better in the red states than Obama did in 2012 and also polling slightly better in the purple states (although the trend across these is by no means a uniform one), while Trump is, on average, seen to be performing slightly better in the blue states. This fivethirtyeight.com research shows that states in which Clinton is performing (at least in the opinion polls) notably better than Obama did in 2012 includes red states such as Utah, Kansas, Idaho, Arkansas, Oklahoma, Tennessee and Arizona and in purple states such as Florida, North Carolina, Wisconsin and Minnesota, as well as a few blue states, including Maryland and Massachusetts. By contrast, the same research shows that states in which Trump is performing (at least in the opinion polls) notably better than Romney did in 2012 includes blue states such as Connecticut, Maine, New Jersey and New York and in purple states such as Nevada and Colorado, as well as a few red states, including Montana and West Virginia.

As of now, based on current state and national opinion polls, the figures suggest that Hilary Clinton is on course for a comfortable win in the Electoral College, along the same lines as Obama’s 332-206 margin over Romney in 2012. Her scope for a larger win, at present, seems to be limited by the fact that Obama won 10 of the 11 closest state contests in 2012 and hence only North Carolina appears to be an immediate potential gain for the Democrats.

Figure 2: The closest state contests for the 2016 Presidential Election, according to fivethirtyeight.com polls analysis as of 5th July 2016.

However, state poll figures suggest that some red states, including Arizona and Kansas – but also Utah – the state that offered Romney his biggest win in 2012 – may be coming “into play” for the Democrats. Figures are suggesting that some of the swing/purple states in 2012 (and indeed in most of the other recent presidential election contests) including Wisconsin, Minnesota and Michigan, may not be as competitive in 2016, while currently in the polls the contests in other states, such as Florida, are not as close as they proved to be in the 2012 election, as shown by the fivethirtyeight.com analysis of recent state polls.

This analysis suggests that some red states may figure among the closest state contests in 2016 – with the states most likely to turn from “red” to “purple” including Arizona, Kansas, Missouri and South Carolina. If the Clinton campaign was able to bring these states “into play” and go on to win these, as well as North Carolina, then Clinton would be in line to win the Electoral College by a considerable 383-155 margin – which would be the largest winning margin by a successful presidential election candidate since 1988 (426-111) but not as large a winning margin as that enjoyed by Ronald Reagan in the landslide 1984 525-13 win in the Electoral College over Walter Mondale.

Against that, although some of the purple states in 2012 may seem to be edging more solidly towards the Democrats in 2016 based on the opinion polls carried out to date, as noted above, there are still a number of purple states, which were won by Obama in 2012, that are still in play for the Republicans. The fivethirtyeight.com polls analysis suggests that that purple states, won by Obama in 2012, that might be most likely to be won by Trump in November 2016 include Nevada, Colorado, Ohio, New Hampshire, Florida, Pennsylvania, Virginia and Iowa. (He also seems to be competitive in the second congressional district in Maine, according to a recent poll carried out there.)

Assuming that he can hold the states won by Romney in 2012 (and at this point in time that is by no means guaranteed, especially given the close contest in the polls in North Carolina), Trump would not need to win all of this group of states to secure the presidency. But he needs to win a sufficient number of states to ensure he wins 63/64 more electoral college votes than then 206 won by Romney in 2012. (If he (and Hilary Clinton) were to both win 269 electoral college votes, then the result would be decided by the Congress/House of Representatives, which would “presumably” be more likely to vote for Trump.) Given the numbers involved, Trump effectively needs to at least gain/win two of the three largest Democrat-leaning swing/purple states (Ohio, Florida and Pennsylvania) and two or three of the less populous Democrat-leaning swing/purple states (Virginia, Nevada, Colorado, Iowa and New Hampshire). If Trump fails to win Florida, then he will probably need to win most, if not all, of the swing states in the “Rust Belt”, as well as Virginia.

Assuming that he can hold the states won by Romney in 2012, Trump could win the presidency by winning/gaining the following groups of “swing states” (and this is by no means an exhaustive list!!!):

• Florida (29), Ohio (18) and Pennsylvania (20) – 67 electoral college votes
• Florida (29), Ohio (18), Colorado (9), Nevada (6) and New Hampshire (4) – 66 electoral college votes
• Florida (29), Ohio (18), Virginia (13) and New Hampshire (4) – 64 electoral college votes
• Florida (29), Pennsylvania (20), Colorado (9) and Nevada (6) – 64 electoral college votes
• Florida (29), Pennsylvania (20), Colorado (9), New Hampshire (4) and Maine Congressional District 2 (1) – 63 electoral college votes
• Ohio (18), Pennsylvania (20), Virginia (13), Colorado (9) and Nevada (6) – 66 electoral college votes
• Ohio (18), Pennsylvania (20), Virginia (13), Colorado (9) and New Hampshire (4) – 64 electoral college votes
• Ohio (18), Pennsylvania (20), Colorado (9), Nevada (6), Iowa (6) and New Hampshire (4) – 63 electoral college votes

Overall, the poll trends in early July 2016 suggest that Hilary Clinton is well placed to win the presidency, but – given that a week is a long time in politics and four months are much longer – nothing is guaranteed at this point in time.