Another Metric that can Predict the Presidential Election Winner?

July 26, 2020

When I was doing research on how bookies make odds on elections, I found an article that noted how some bookies use favorability ratings of presidential candidates to calculate their odds. I personally have used these ratings in local and state surveys and found them to be quite useful at that level.

So I decided to see if these ratings had any impact on presidential races in the past. From Monmouth University’s monthly blog, I found data from their surveys going back to 1984, that included favorability ratings.

Their question gives voters three choices: favorable or unfavorable for Democratic candidate, unfavorable or unfavorable for Republican candidate, or neither candidate favorable. This question is strictly a popularity measure, but likely contains some perceptions of job performance.

The table below shows how voters ranked Democratic and Republican presidential candidates from 1984 to 2016 (this question wasn’t asked in 2008).


What jumps out at is the 2016 election, where 35% chose neither! Monmouth says they have not seen this kind of reaction since using this question. But this was an unusual pair. Both Trump and Clinton had the lowest favorable ratings of all presidential candidates since 1984.

But is there a relationship between favorable ratings and winning the election? Monmouth did not address this question but my curiosity forced me find out.

To test if candidate ratings effect the election, I have added a column for the winning popular vote candidate, where 1 represents the Republican candidate and 0 the Democratic candidate. It does not measure the Electoral College vote.

In all of these eight elections, the popular vote was won by the candidate with the more favorable ratings, even if it was as little as 1%.

The average of Clinton’s favorable ratings is 38.9% and Trump at 37.4%, a difference of only 1.5%. This difference isn’t great, and a test of the mean differences suggests that the means are are equal to or greater than each other (p=<.001).

So how do Trump and Biden favorable ratings compare at this point in the campaign? Using Real Clear Politics average of recent favorable ratings (July/June), shows Trump with an average ratings of 39.5% and Joe Biden with an average rating of 45.5%, a difference of 6%.

Like most polls, Job Approval ratings, betting odds, Job Approval ratings, and now this Favorability rating, all give Biden, at this point in time, a significant chance of winning at the national level. This does not include the swing states and the Electoral College.

Remember, we still have over three months until the finish line. Polls are most accurate when nearer election day. Go ahead place your bets, but don’t bet your house on it just yet.

In a later post, I’ll explore the favorability ratings in the swing states, which if Trump is going to pull out a victory it will likely occur there by winning the Electoral College vote. Be safe…


Do Likely Voter Polls Benefit Donald Trump?

July 22, 2020

The trend among most pollsters has always been to use a sample of all residents or registered voters early on in the election campaign and switch to likely voters as the election nears.

The reason for this protocol was explained to me by an early leader in modern campaign polling, Bill Hamilton, when we both taught at the University of Florida.

He explained that voters need some time to absorb the effects of the campaign and the competing candidates. Using a likely vote screen early on, leads many voters to fib about their likelihood of voting in an election still months away. And besides, he would say. “it’s cheaper.”

As we get closer to the November election, almost all polling firms will be utilizing likely voter screens.

But in the polling and campaign business, there are some who believe that the method of interviewing registered voters rather than likely voters early on, is an intentional method to under-report Trump’s real support in the polls.

This argument suggests that the more liberal public polling firms, by calling only registered voters is an intentional effort to increase Biden’s poll lead.

John McLaughlin, well known Republican pollster, commented how many pollsters are skewing the results by using registered voter samples.

The refusal to screen for actual likely voters is creating an under-polling of Republicans and therefore Trump voters. It seems intentional. It’s exactly what the media did in 2016. Let’s prove them wrong again.” ( ElectionsPolls & SurveysPress Releases, June 8, 2020)

For me, this argument begs the question, what are the polling differences between a registered voter sample and a so-called likely voter one.

The idea that Republican voters are more likely to vote in elections and Democrats less likely, has been asserted for many years. The argument goes like this, Republicans are better educated, wealthier and older. These are attributes that contribute to voting.

In 2010, the website FiveThityEight concluded that registered voters surveys favored Democrats by 5% over the actual election results and likely polls favored Republican candidates by 1%.

Oddly, the academic studies have not focused directly on this issue. It is true, that the opinion of pollsters and campaign consultants are all in agreement that more likely voters identify with the Republican Party and, of course, likely to choose a Republican candidate over a Democrat.

To address this issue on whether likely voter polls do favor Trump and registered voters Biden, I devised a way to test these assertions statistically by comparing two series of recent national surveys conducted by public polling firms.

The first group includes 22 polls of registered voters only. The second group includes 22 likely voter polls only. Both survey groups were completed in June and July with similar end dates. The polls asked the identical question of choice for president.

When comparing Trump’s percent only in both registered and likely voter surveys, the results showed a small difference between the two groups. The chart shows how likely voters tended to support Trump more than the registered voter surveys.

Blue line = Likely Voter surveys /Red line = Registered Voter surveys

As you can see the per survey differences weren’t substantial, with a mean difference of only 1.73%, in Trump’s favor.

mean 41.5mean 39.7

But that’s only half of the story. This table and graph only shows the changes in Trump’s percentage without Biden. When you look at the percent difference between Biden and Trump in each poll, it reveals a completely different picture. After all, Trump isn’t running against himself (no comments, please).

The best way to measure the differences between two candidates is subtracting one percentage from another, creating a net difference variable. In this example, I’ve subtracted Trump’s results from Biden’s. The table below shows the net percentage difference between Biden and Trump by whether the sample was of registered (RV) or likely voters (LV). Positive numbers indicate a net percentage gain for Biden.

RV % DifferenceLV % Difference
RV mean = 9.2LV mean = 7.6
Biden / Trump Net % Difference in Registered Voters and Likely Voters

All the differences are positive except one likely voter survey with a minus 4%, which means that Trump had 4% lead over Biden in that one survey only. As the means for the two different modes show, Biden’s lead among registered voters is 9.2% and a 7.6% for the likely voter poll. In other words, Trump performs better in the likely voter sample by an average difference of 1.6%.

But is this modest difference real or just a random variation. To determine whether the differences are not random, I used a one sample T-test, which compares the mean of your sample data to a known value. In this case, is the mean of registered voters versus likely voters?

Without boring most of you with statistical data, the answer is yes, and the differences are statistically different at a P value of <.0001.

The individual differences between individual polls can sometimes be substantially larger than others. By comparing the results of two separate sets of polls, the differences are averaged out.

Notice that in this example, there are some polls showing Biden leading by 12% or even 14%. But when you combine those outliers with other polls, we average out those differences. (Remember the wisdom of the crowd.)

An example of that is an ABC News / Washington post poll released last Sunday. To their credit they used both registered and likely voters in the same survey. The registered voters had Biden at 55% and Trump at 40%, a 15 point differences. In the likely voter sample, Biden had 54% and Trump 44%, a 10% difference. The pattern is the same, Trump does better when the poll uses a likely voter sample, but the difference is based on only one survey.

This analysis confirms that Trump on average, does do better in likely voter polls than with traditional registered voter samples. It’s not at the level of the recent ABC / Washington Post Poll showing a 10% difference. This is not the case when we average a series of public polls in the same time period, using identical sampling methods.

With most recent public polls showing substantial leads for Biden, an average difference of 1.6% won’t help much. With almost four months and two Presidential debates to go, this campaign is just getting started. Stay tuned and stay safe…


Is There a Better Way to Predict Who Will Win the Presidential Election than Polls?

July 19, 2020

Probably most viewers of this blog keep an eye on the latest Biden vs. Trump poll numbers. That and Trump’s job approval ratings are the only metrics we can use to see who is winning and losing before the election itself.

But there is another measure of how well the candidates are doing and is possibly more accurate than polls (or what your neighbor believes).

But first I need to describe why this metric works. In 2004, James Surowiecki published a book entitled The Wisdom of Crowds: Why the Many Are Smarter Than the Few and How Collective Wisdom Shapes Business, Economies, Societies and Nations.

The book describes how the aggregation of information in groups increases prediction accuracy. As a real life example, he describes how a crowd at a county fair accurately guessed the weight of an ox when their individual guesses were averaged. The average of the guesses were significantly closer to the ox’s true butchered weight than the estimates of most crowd members.

His conclusion is that a diverse collection of independently deciding individuals is likely to make certain types of decisions and predictions better than individuals or even experts.

Although he doesn’t combine his insight with statistical sampling, the parallels are apparent. Surveys are, after all the aggregation of political opinions from a random selection of average voters.

The difference, of course, is that a traditional survey asks the voter who he or she will vote for and not who will win and by how much. Polls also become more accurate as the sample size increases (increasing the crowd). It is also the reason why combined polling averages can be more accurate than single poll as well.

In recent years, a number of prediction markets have appeared on the internet. These markets allow people to bet on the outcome of specific events such as elections. The concept of these markets is to allow people with or without expertise to make a bet on a specific event’s outcome, such as a presidential election.

The market prices indicate what the crowd thinks the probability of a certain event occurring. Market bets range from 0 to 100%, so the amount of the bet represents what betters expected probability that the event will occur. For example, if a person bets 62 cents a share on a certain candidate winning, that tells us that person believes that the odds of that candidate winning is 62%.

To test the viability of a prediction market, I will be reporting the market odds of Biden and Trump winning the November election. I have chosen a Predictit, a prediction market that specializes in political events such as presidential elections. Don’t worry, it’s completely legal because it’s a not-for-profit associated with a university.

On this site you can bet on not only the national presidential election, but also individual states as well. As of July 16th, the best national election offer for a Biden share was 63 cents. In market terms, that indicates a probability for Bidden winning at 63%. Donald Trump, on the other hand, the best offer for Trump is 39 cents per share or a 39% chance of winning.

The change in price for Bidden share since July 10th was only three cents (3%). For Trump his per share order has declined two cents (2%) in the same period. In other words, the market currently believes the odds of Biden winning is increasing and for Trump decreasing.

The rational of predictive markets is based on the economic theory of the efficient-market hypothesis, which states that prices reflect all available information. In other words, the market price includes all information that is necessary to make a predictable outcome.

In the case of a presidential election, that would obviously include polls, but also other information such as news reports, pundits on talk shows, the Covid pandemic and even friends and neighbors opinions.

The table below displays the battleground states and the corresponding share bets (cents) for each state. It also includes the current two month average of the latest poll results over the past two months.

FLORIDA          633949.443.55.9
PENN            75264943.25.8
WISCONSIN       732847.842.45.4
OHIO            465647.744.7-3
MICHIGAN        752649.242.17.1
ARIZONA         643647.544.33.2
NORTH CAROLINA  584247.6452.6
GEORGIA         47544647.6-1.6
IOWA            445746.247.9-1.7
MINNESOTA       832052.341.311

Using Florida as an example, it shows that the current market bet on Biden winning is 63 cents a share. For Trump its 39 cents a share. In other words, the market believes the odds for Biden winning Florida is now 63%. At the same time, the latest polls show Biden with 49.4% and Trump at 43.5%, with Biden leading by 5.9%.

The data shows that market believes that Trump will only win three of the battleground states: Ohio, Georgia and Iowa. As you can see, their is also a consistency between the latest polling averages and the betters bets.

Why are people choosing on one candidate over another? To answer that, you have to remember that this is a for profit game, where winning is a financial gain and losing costs you money.

So the financial incentive should be paramount and not who you want to win. That should mean even Trump supporters could and are bidding on Biden, even if in their heart they would prefer the opposite outcome.

But that’s the very basis for the underlining wisdom of the crowd theory. It’s not who the better supports, but who he or she believes will win based on all available information.

The bidder calculus is based on the probabilities and the final payoff. Let me explain, if you bid 62 cents a share on Biden winning and he wins, your payoff is one dollar a share less the bid amount. Since you bid 62 cents, your net profit is 38 cents a share. If you bought a 1,000 shares, it cost you $620. Your profit is $380.

Now let’s say you think Trump can pull it out because the percent poll differences between the two candidates aren’t that great and he won that state in 2016. So instead of bidding on Biden, even though he is currently ahead, you buy 1,000 shares on Trump winning at a price of 32 cents a share for a total cost of $320. If he wins, your gross profit is a $1,000 less the $320 expense of purchase, for a net profit of $680, $300 more than the safer bet.

This scenario, however, only works if the bid price for a share is significantly lower than Biden’s, like the current bid price on Trump winning the national popular vote, which is currently at 38 cents a share. At the battleground state level, bid prices are usually higher due to their tendency to vote for both parties.

Do bidders use this scenario of bidding on the underdog to increase profits? Apparently they do when they think the polling differences aren’t that great or Trump is even leading.

Below is a regression graph showing how Trump bets per share in swing states increase as the poll differences decrease.

In the graph, the small circles are Trump bets on swing states. As the net percent polling difference increases in Biden’s favor (positive values), the number of Trump shares declines.

The sweet spot for Trump bets happen when the difference (% losing) is within the 3 and 7% difference in Biden’s favor. In that betting zone, is where the bulk of Trump bets occurred at prices they considered reasonable (25 cents to 42 cents a share).

Arizona is a good example. At this point, Biden leads Trump 3.2% and the best offer for Trump winning is 36 cents a share.

Although Arizona is suffering from a serious spike in new Covid-19 cases, the Republicans have won the last five presidential elections. The political website 270 to Win, has it as a toss up state. At 36 cents a share, for many betters this would be good buy, considering the potential payoff. (Now that I’m writing this, I’m considering buying a couple of Arizona Trump shares.)

When Trump is leading in the polls, some betters are willing to pay a premium for a share since they consider it a safer bet, such as in Georgia and Iowa, where Trump leads in the polls by 1.6% and 1.7% respectively. The current price is 54 cents and 57 cents a share.

Many of the Trump betters are using the strategy outlined in the national poll scenario, making bids less than 27 cents a share in a couple of cases when the polling percent differences are within 7%. If you win, your profits increase significantly.

Betting markets can provide us with another metric on who is winning and losing based on free market forces, but it doesn’t replace polls in my judgement. When all the votes are cast, I’ll evaluate how well it did compared to other methods.

In 2016, Predictit shares also predicted Hillary Clinton would win. The day before the November 7 election, the going bet was 81 cents a share she would win. A lot of people lost money that day…be safe.


“Biden Opens Up Largest Polling Lead of the Year Over Trump …” Bloomberg News

July 16, 2020

I’m sure by now you have heard how Biden is extending his lead in the latest polls. The media has blasted that headline on every website and newspaper in the country. Although Bloomberg’s headline is correct, it’s not the whole story, which I think is even more interesting.

In the Chart below, we see how Biden’s net lead over Trump has increased over the past seven months.

Each bar represents the monthly average of all public polls with the trail ballot question between Biden and Trump. In January, Biden’s lead was 5%. By June it had increased to 12%. Statistically, that’s an 82% increase in six months.

Now let’s look at Biden’s and Trump’s share of the two party vote in the same time period in the chart below.

Biden Monthly Percent = Blue Line / Trump Red Line

Biden’s average percent in January was 49.8%% and his average in July was 49%, a 0.8% difference in seven months. Yet his average lead over Trump increased to 12%. In the chart above, notice the blue line (Biden) moves steadily along the 49% line, and takes only minor deviations from that line.

So how did Biden’s lead increase to 12%? The answer is that his numbers didn’t change but Trump’s declined, increasing Biden’s net gain even though his share of the vote hardly changed at all. In other words, there are not more for votes for Biden but less for Trump.

So where did all those votes go? They go where all voters go when they can’t make a decision: the undecided/don’t know category.

What does that mean? Simply put, these votes are now up for grabs by either candidate, but I would rather be Biden than Trump at this point. Some of these voters may not vote at all, which means they are contributing to Biden’s victory anyway.

If you are a Biden supporter, don’t go out and party just yet (bars are closed anyway). There are the Conventions and debates still ahead. Research suggests that many voters don’t start making up there mind until after labor day. The finish line is still a ways to go…Be safe.


Can Donald Trump Win the 2020 Election through the Swing States Again?

July 13, 2020

Donald Trump’s only path to another term in the White House is through the swing states and not the national popular vote. In other words, he has to repeat his wining strategy in 2016 where he lost the popular vote by 2.1% but won the Electoral College by 74 votes. (I eliminated the 7 faithless electors.)

Both his national polling numbers and job approval ratings suggest that without the help of divine intervention, his winning the national vote is highly unlikely with only four months and a pandemic.

But does this strategy have any chance of working at this point? To discover the answer to this question, we need to look to the past to see a possible future: 2016, 2012 and 2008 presidential elections.

In this analysis, I have included the three past presidential elections from ten swing states: Florida, Pennsylvania, Wisconsin, Ohio, Michigan, Arizona, Georgia,, North Carolina, Iowa and Minnesota. In the last three presidential elections, each of these states, except Arizona, voted for two different party candidates in separate elections.

In 2016, Donald Trump carried all but Minnesota, for a total of total of 131 electoral votes. Of all these swing states Clinton, the first woman to run for president from a major party, carried only one: Minnesota. The table below shows each of the swing states two party percent of the vote and the percent difference of votes cast and a graph of the data below it.

FLORDA          47.849.02-1.22
PENN            47.548.2-0.7
WISCONSIN       46.547.2-0.7
OHIO            43.651.7-8.1
MICHIGAN        47.347.5-0.2
ARIZONA         44.648.1-3.5
NORTH CAROLINA  46.248.5-2.3
GEORGIA         45.350.4-5.1
IOWA            41.751.2-9.5
MINNASOTA       46.444.91.5
AVERAGE %45.6948.672-2.982

The average percent difference between the two candidates for all ten states was nearly -3%, the percent Clinton lost by in these 10 swing states. Not much, but enough to put Hillary Clinton’s name in the history books with an asterisk. In Elections, unlike horseshoes, being close doesn’t matter.

In 2008, we had two non-incumbents facing off for the ultimate political prize. John McCain, the U.S. Senator from Arizona. McCain started the Republican primary season as an underdog, but made a comeback when he won the New Hampshire primary by defeating future nominee Mitt Romney.

Through out the general election campaign, McCain was considered an underdog against the first African-American candidate to run for president, Barack Obama.

And it showed in our ten swing states, where John McCain almost matched Hillary Clinton’s 2016 loss. The Table below, shows both Obama and McCain’s percentage of the vote in each state.

FLORDA          50.948.222.68
PENN            54.544.210.3
WISCONSIN       56.242.313.9
OHIO            51.546.94.6
MICHIGAN        57.340.916.4
ARIZONA         44.953.4-8.5
NORTH CAROLINA  49.749.40.3
GEORGIA         46.946.90
IOWA            53.944.49.5
MINNASOTA       54.143.810.3

This image has an empty alt attribute; its file name is image-7.png

The only state McCain won was Arizona, his home state. Nationally, he lost to the country’s first Black American president by 5.9%. Another asterisk in probably the most monumental election in history.

The 2012 presidential election pitted Obama against Mitt Romney, the first Mormon presidential candidate to run for the highest office in the land. (A lot of firsts in these three elections!) In the Table below, you can see that Romney won three of these swing states.

FLORDA         50.0149.130.88
WISCONSIN      52.845.96.9
OHIO           50.747.73
ARIZONA        44.553.5-9
NORTH CAROLINA 48.3550.4-2.05
GEORGIA        45.553.3-7.8
IOWA           51.946.25.7
MINNASOTA      52.644.97.7
AVERAGE %50.16648.2331.933
Table 3 2012 Obama vs Romney Swing State Results

This image has an empty alt attribute; its file name is image-8.png

Of the three election cycles, Romney did better than either Clinton or McCain. He only lost by 1.9% in all of these 10 swing and carried Arizona, Georgia and North Carolina.

The question is whether Trump can repeat his 2016 swing state victories in 2020? This question will likely be determined by two variables: voters overall appraisal of his job approval and their comparison of Joe Biden as an alternative.

At this point in the campaign, our best gauge of how either candidate is doing in each swing state are the latest polls conducted there. Using the average of state level surveys conducted since June 1, 2020 from Real Clear Politics, I have added it to each swing state’s data, as seen in Table 4 below.

North Carolina47.544.23.3
OHIO 44.744.70
AVERAGE %48.2443.364.88

Biden is behind Trump in only one state, Iowa by a -1.3% and tied with him Ohio. That doesn’t mean Trump can’t win any of the swing states, but it suggests the road to victory through the swing states is going to be a rocky one.

The other problem for Donald Trump is that Joe Biden is not Hillary Clinton. By that I mean his unfavorable ratings are significantly less than her’s were in 2016 prior the general election.

Real Clear Politics unfavorable average for Clinton right before the 2016 election was 54.4%, compared to Biden’s current average of 46.1%, over 8 points less than her negative ratings in 2016. Donald Trump’s current unfavorable rating is 56%, some 12 points higher than Biden’s unfavorable rating.

So to sum up, if Trump is unable to win the popular vote he will have to again rely on the swing states, many of which he won in 2016. But the times have changed and I will repeat again, he doesn’t have Hillary to help him this time. In recent swing states polls, he is upside down in all but Iowa.

He has four months to make up some ground on Biden. He has the convention bounce and of course, the Presidential Debates and maybe some good news on the pandemic. I will be returning to swing states again as we get closer to the election. I love politics, don’t you….Be safe.


Trump Supporters Don’t Despair (Yet)

July 11, 2020

In previous posts, I have stated that no incumbent President has won re-election with an approval rating less than 48% on the last poll before the election since Truman.

The reasoning for this statistical fact is based on the theory that the approval rating gauges both the president’s popularity as well as his job performance. In essence, the job approval of a incumbent president is a referendum on his performance. Most scholars consider this rating the best single measure for reelection, but only within a couple of months of the election.

It’s no secret that Trump’s approval ratings have been low for sometime. His rating has averaged 42.9% over the last six months, 5% less than it should be on election day. But how does this compare to other president’s ratings at the same period in the past? And does he still have time to recover?

I decided that the 2012 election would make a good comparison since Barack Obama also suffered from a job approval deficiency during much of his final year before his reelection.

In a comparison of Trump and Obama’s job approval rating from January through June of their final year in office. In Table 1 below, are the approval ratings of both candidates’ from polls conducted on the same dates — but, of course, from different years: 2011 /2012 for Obama and 2019 / 2020 for Trump.


The survey dates are identical for both series to keep the comparison in the same time frame prior to the next election. The only difference is the year of the poll.

The last column shows the two candidate’s approval differences. Positive numbers indicate that Obama’s ratings were higher and the negative values shows that Trump’s ratings were higher (Obama rating ,- Trump rating).

During these two campaign time frames, Trump’s average approval rating is 43% and Obama has an average rating of 47%. The final two approval ratings for Trump in this series were his lowest at 37% each and trashed his average. In comparison, that gave Obama a 17% rating lead over Trump. The figure below shows how the two candidate’s approval ratings varied in the respective time periods.

Red line Trump Approval ratings / Blue Obama Ratings

Obama (blue line) started out higher during this time period, but then took a dive around August and didn’t recover until October of 2011. Trump was doing better see-sawing closer to the 50% mark, his highest ratings of his presidency.

But in last two July polls, his ratings collapsed by 20 points. His falling approval rating is likely caused by the rising cases of Covid-19 and his response to the crisis. If that is the case, his approval ratings may not bounce back anytime soon, if at all.

In the November of 2012, Barack Obama defeated Mitt Romney with 51.1% of the popular vote. His last approval rating before the election was 47.9%. (OK, its not quite 48%).

Donald Trump could still increase his ratings in the next four months, but time is running out and the virus is again raising its ugly head.

I know most folks are watching the trial ballot polls between Trump and Biden, but we should keep an eye on the job approval ratings as well. Be safe…


How did Obama Win Florida and How did Clinton Lose it?

July 9, 2020

Reading a recent article about how Barack Obama won the two terms for president, it made me start to wonder how he managed to win Florida twice and Hillary Clinton lost it in 2016. His victories were close, of course, and his victory over Mitt Romney in 2012 was less than one percent. Hillary’s loss in 2016 was 1.2%.

My curiosity led me to compare Obama’s and Clinton’s Florida county by county results to see why he succeeded and she failed. In Table 1 below, I have listed Obama’s 2012 and Clinton’s 2016 percent of each Florida county vote.

Alachua 5758.97
Baker   20.2916.72
Bay     27.5324.92
Brevard 42.9438.02
Broward 67.1266.51
Calhoun 26.9120.41
Citrus  38.3728.59
Clay    26.6126.13
Collier 34.6235.77
DeSoto  42.2234.95
Dixie   25.8417.63
Duval   47.6747.54
Flagler 45.7738.3
Gadsden 70.0167.92
Glades  40.0329.2
Gulf    28.2623.58
Hardee  3428.34
Hendry  46.5141.52
Holmes  15.2510.02
Indian River38.4136.3
Jackson 35.0230.4
Lake    40.8936.86
Lee     41.3238.27
Leon    61.1360.52
Levy    33.1526.3
Liberty 28.6419.75
Madison 47.8541.46
Manatee 43.2239.8
Marion  41.3235.51
Martin  38.0935.18
Monroe  49.5644.66
Nassau  25.2323.32
Orange  58.5660.39
Osceola 61.7360.95
Palm Beach58.1456.57
Pasco   45.8637.38
Polk    45.9541.3
Putnam  37.1330.48
Saint Joe30.5921.04
Saint Lucie53.4242.73
Santa Rosa23.0647.09
Sumter  32.2529.53
Suwanee 26.8521.2
Taylor  30.2423.16
Union   24.7617.81
Volusia 48.7841.82
Wakulla 35.2128.32
Walton  23.3420.44
Table 1 Obama and Clinton County Percent

Overall, Obama outperformed Clinton by 4.3% in all 67 counties. The chart below displays the counties where Obama either out performed or under performed her (red bars) in 2012.

I subtracted Obama’s percent of the vote from Clinton’s to calculate a net performance rating. In all but six counties (red bars), he outperformed her in the same counties in his 2012 election. Santa Rosa County showed a particular dislike for Obama in 2012, and I’m hesitant to speculate the reason for that anomaly.

So would have Hillary beat Donald Trump in 2016, if she had performed at the same level as Barack Obama in 2012? Realizing that Mitt Romney and Donald Trump are two very different candidates and that Hillary carried her own political baggage, the answer is yes. She only lost by 1.2% and if her campaign had carried the same counties at the same level as Obama, she would have defeated Donald Trump by 3%.

I know what you are thinking, would that have denied Trump’s Electoral College victory? NO… Donald Trump’s Electoral College vote was 304. If you subtract Florida’s 29 electoral votes, that would still leave Trump with 275 votes, five more than needed to capture the White House!

This analysis shows that the small Republican leaning counties still matter when combined together. In other words, if you are a Democrat, you don’t have to win a county as long as you eat away at the Republican vote.

Florida Democrats have long depended on the larger urban counties that traditionally vote Democratic and spent less time and money in the smaller Republican leaning ones.

Obama changed that paradigm, as shown in Table 1. You can still win the big Democratic counties, and at the same time eat away at Republican’s advantage in the smaller, less urban ones.

Hillary Clinton outperformed Obama in most large urban counties such as Broward and Miami-Dade but he outperformed her in almost all of the smaller Republican leaning counties. Yes, both elections were close but that only counts in horseshoes. In the end, the candidate with the most votes wins and not the number of counties you carried. Be safe…


Online vs. Live Caller Polls: Which is Better?

July 6, 2020

I was just beginning my survey career in the 1980’s, just as pollsters were transitioning from in-person interviews to telephone surveys. Gallup finally gave up the in-person survey in 1988, marking the end of public polls using in-person surveys.

Today the new kid on the block is the online poll where interviews are conducted through the internet via computers and cell phones. This change has blossomed for a number of reasons, the most important is cost. A telephone survey is far more expensive to conduct.

The other problem is the response rate, which is the percentage of people who complete a telephone poll. Today that rate has dropped to 6% of voters. This has led some national firms, like Pew Research, to shift most of its surveys to an online format.

But there are downsides to this change. The most important is the lack of randomness that telephone surveys enjoy. The key issue for any poll is that every respondent has an equal probability of being selected and interviewed. Probability Theory mandates that randomness is necessary to make mathematical assumptions as to the reliability of the poll. In other words, whether we have confidence in the results based on expected error.

Opt-in polls are self-selecting and are not random, consequently the results are not testable. That doesn’t mean the results are wrong, but it does mean you cannot use probability theory to estimate the potential error. Some major survey firms that use online polls, compensate for this by using national panels of voters that are recruited via probability methods like the American Trends Panel.

Even with these issues, most national surveys have switched to online polls. The ultimate question is how different are they from live-caller surveys? To analyse this, I have adopted a method called an observational study which does not require the randomization the subjects involved. This is often used in medical, economic and political studies where the randomization of participants is not possible.

In June, most polls started to show Joe Biden expanding his lead over Donald Trump in national and state polls. Taking the results of all surveys conducted in June, I selected national polls using two different survey methods: online and live caller interviews that occurred within two days of each other. The object was to measure if the results of the two methods resulted in significantly different results.

In June there were 10 live-caller surveys and 10 online only surveys. In both mode types, Biden led in every poll ranging from seven percent to fourteen percent. As shown in Table 1 below, the number in each column is the percentage lead Biden had in each survey mode.

Table 1: Biden poll lead over Trump in July

The average percent lead for Biden in the online polls is 8.3%. In the live-caller polls, it is 10.5% or a difference of 2.2%. In the chart below, we see a graphic representation of the both poll methods.


The red line shows the live-caller poll’s percent lead for Biden and the blue line the online-poll percent lead for Biden. Although some results showed both polls with similar outcomes, overall it is apparent that Biden in general has a small lead among live caller surveys. But is there a statistical difference in the two modes?

Using a simple student T-Test, we can confirm if the differences are significant or just a random occurrence.

One-Sample Test
 Test Value = 0                                       
MODE TYPESig. (2-tailed) 95% Confidence Interval of the Difference
TABLE 2 significance level <.0001

As shown in Table 2, the differences are significant (Sig. level =.000) is less than .05, so we have confidence that the two modes are significantly different from each other.

So we have two types of surveys, one group using live callers from a randomized list and another group from online surveys with opt-in panels taken in the same time frame. Both sets show Biden in the lead but the online surveys show the lead significantly less than the live-interviews surveys.

In 2016, there was a similar pattern that showed non-live interviewer polls narrowing the gap after June between Trump and Clinton, but still showing Clinton leading. In the end, the online surveys in many battleground states had the race closer but with Clinton still losing.

Non-live interview firms have always asserted that some voters are more honest about their voting intentions when responding to either an online or IVR (robocall) poll.

This would usually be attributable to what political scientists call social-desirability effects, that is when a voter gives an answer that is considered more acceptable even though they don’t believe in it, when they are talking to a real person. And the online survey mode may give some Trump voters more courage to respond truthfully, since there is no live interviewer.

So what is a poll watcher to do? My advice is to average all the polls for that period and ignore the poll mode. It won’t guarantee the actual results are correct, but it increases the number of surveys and, of course, the total respondents. The “law of large numbers” can increase the accuracy. More on this in a future post. Be safe…


The Changing American Voter

July 1, 2020

It’s one of those laws of the universe, isn’t it? Nothing stays the same. Things change.” – Author: Sam Gayton

Sometimes you become immune to demographic changes. They sneak up on you until one day you wake up and find that the world you thought you knew has moved on and left you with new people and ideas that no one ever mentioned before. This is a good thing and you must grow with it.

People grow old and younger folks take their place, and with them comes a new political environment. The U.S. is now facing such an demographic revolution as the decline of the Baby Boomer generation accelerates. America is becoming more diverse, with new ideas and beliefs.

With the recent exposure of police brutality regarding Black Americans has bared this nation’s history of unequal treatment for many of its citizens.

Some historic and long dead figures in American history, now face new public scrutiny about what they both said and acted on during their life on earth. And Mississippi has finally removed the Confederate symbol from its own flag. As Bob Dylan once sang, “The Times They are A-Changin’.”

Politics are not immune from these changes. White voters who have dominated American politics since the Revolution are starting to show their age. In the following chart, we can see how white voters are declining as a percent of the electorate since 1990, by some 14% in just 28 years.

Source: United State Election Project

Some of this decline is due to deaths, but most of it is caused by the increasing number of other ethnic groups, particularly Black voters. As an example of the increasing voting power of Black voters demonstrated in the 2012 presidential race where turnout for White non-Hispanic voters was 64.1% and for Blacks 66.2%. This is the first time since the Census Bureau began recording turnout numbers that the Black voter turnout was greater than White’s.

This high turnout was due largely to Barack Obama on the ballot, but illustrates the increasing influence on Black voters in election outcomes. This statistic alone should convince the Biden campaign to add a Black woman to his presidential ticket.

The next chart gives some overall perspective of how this change is effecting turnout in recent elections.

Source: United States Elections Project

The blue line in this chart shows the yearly turnout rate for non-Hispanic Whites since 1986 and the red line Black voters. You can see how Black voters have slowly gained on whites turnout rates. In the 2016 presidential election, Black voters fell 5% from the previous presidential election. With the loss of these voters, Hillary Clinton lost the election.

In 2020, the composition of the turnout will likely determine the outcome. Trump voters are still likely to show up and Biden is not Hillary Clinton. Biden’s success will depend on Black voters. If White voters continue to decline, the Republican Party will over time will suffer. “Demographics are destiny” said Auguste Comte. Be safe…


The Polarization of the American Electorate: Part 3

Race and Guns

June 28, 2020

Our continuing series on the differences between Trump and Biden supporters touches on the how the two groups differ fundamentally on race relations and guns, both issues that consume the nation today.

On the issue of whether blacks are have gotten less than they deserve, most Trump devotees strongly disagree, as shown below.

Pro-Trump folks also believe that Blacks need to try harder and Biden supporters disagree with that statement with equal intensity as displayed below.

To reinforce how Trump supporters see Blacks, I have recoded the black thermometer ratings into four categories: 0-25 degrees, 26-49 degrees, 50-75 degrees and 76-100 degrees and compared these categories to Trump’s thermometer ratings, as shown below.

The first bar represents voters who ranked Blacks between zero and 25 degrees, the most negative ratings. These same folks gave Donald Trump a thermometer rating of 85.2 degrees (very warm).

Do Trump supporters like their guns? Boy do they! The Chart below tracks the number of gun owners by how they rate Donald Trump. Gun ownership starts at zero and maxes out at 999 and his ratings rise pretty much with the number of guns they own.

The person with the lowest rating of Trump owns no guns. What is remarkable (at least for me) is that the greater the Trump support, the more guns they own. The strongest Trump rating (99.3 degrees) has almost 1000 guns. I don’t know about you, but I’m not going to mess with a strong Trump supporter!

After seeing how many guns pro-Trump supporters have, I’m not surprised by the following chart, which asks if they are worried about being a victim of a mass shooting.

The Trump devotees are packing and no mass shooter better mess with them! Trump supporters with an average arsenal of nearly 12 guns are ready for that shooter. For the poor Trump supporter guy with only kitchen knife, he should be extremely afraid.

Although much of this is humorous, these charts demonstrate the serious political divide this country faces. Can America still call itself the “United States of America” when a Grand Canyon separates us on key issues? I’m not sure anymore…Be safe.