Data & Resources Front Page General Theories women in politics

Aristotle said that “The whole is greater than the sum of its parts.” The Battle for the Electoral College.

In his stunning 2016 Electoral College victory while losing the popular vote by three million votes, Donald Trump proved that individual parts can sometimes be greater than the whole. Obviously, Aristotle never studied the American election system.

The question is whether he can do it again. So I decided to take a peak at how Trump currently strands in each state he won in 2016. This is normally not an easy task, since key polling data is often not found in states like South Dakota. And in states where there have been surveys, the time differences often negates their usefulness.

But I recently discovered a survey site called Civiqs, which is an online opinion polling and data analytics company founded by Daily Kos founder Markos Moulitsas in March 2018. What is unique about this company is that it conducts daily tracking polls in all 50 states – in the same time frame.

Civiqs operates its own panel of Americans who have been recruited or volunteered to participate in political a surveys in all 50 states. These internet respondents complete a demographic profile that includes their residential location. Individuals are drawn randomly from voter based lists for political surveys and then randomly selected for an interview. Most importantly, they as a Trump Job Approval rating in every survey.

The Civiqs’ job approval surveys for this analysis began on January 20, 2017 and ended on March 3, 2020. During this period Civiqs interviewed 235,231 registered voters in all 50 states. The average job approval rating for this sample is 45.4%. The Real Clear Politics average for this same period was 45.3%.

As a political survey practitioner for over 30 years, I’m still skeptical of online surveys for political purposes. But for online surveys, their published methodology ( ) does address many of the concerns that researchers often have about internet surveys. Most importantly, their samples sizes are large and are conducted in each state within definitive time frames and they include the presidential job approval question. (For a review of the job approval importance, see the post “Will the Economy Save Donald Trump?”)

Why the Job Approval Rating?

The job approval rating is known as a global popularity variable, meaning it encompasses voters’ general impressions of Trump based on issues important to the voter. The correlation between vote choice and job approval is very high. Between 1972 and 2016, the correlation between the job approval rating and the presidential vote in my previous analysis was .840 (where 1 is a perfect match.) On a national basis, regressing the job approval rating on the percent of the two party vote from 1972 through 2016, the Rsq. was .699 (about 70% of the variance).

The importance of a positive job approval is highlighted by the fact that no incumbent president has won reelection since WW II with a job approval rating below 50%.

With this new survey data I can calculate Trump’s potential electoral college vote as if the election were held today and compare it to 2016. Just to be clear, this is not a prediction model that estimates Trump’s electoral college vote. It does, however, portends a possible outcome. In Figure 1 below are all the 2016 states that Trump won and the Electoral College voter for each.

Figure 1

In 2016, Trump carried thirty states for a total of 304 electoral votes. Using the Civiqs’ Trump job approval ratings for each state, I classified a state likely won by Trump if the approval rating was six or more percent above the disapproval rating. If Trump’s rating is upside down by more than five percent, I categorized it as a potential win for the Democrats. And if the state approval rating was within five or less percent above the disapproval rating, it was classified as a toss-up if the election were held today.

In Figure 2, I have listed each state and applied the new state by state electoral vote to each state that Trump won in 2016 based on Trump’s current job approval. For example, if you look at Arizona you will see that it lists a change of 11 electoral votes from Trump to the Democratic candidate, based on the fact that Arizona’s current approval rating of Trump is 46% vs. a disapproval of 52%, a difference of a minus 6%.

State 2016 Electoral Vote Change

Alabama Trump Win 9
Alaska Toss Up 3
* Arizona to Democrat Change 11 -11
Arkansas Trump Win 6
California Democrat Win
Colorado Democrat Win
Connecticut Democrat Win
Delaware Democrat Win
* Florida Trump Win Now Toss Up 29 29
Georgia Trump Win Now Toss Up 16 16
Hawaii     Democrat Win
Idaho Trump Win 4
Illinois Democrat Win  
Indiana Trump Win      11
Iowa Toss Up       6 6
Kansas Trump Win    6
Kentucky Trump Win 8
Louisiana Trump Win    8
Maine     1
Maryland Democrat Win
Massachusetts Democrat Win 
* Michigan to Democrat Change   16 -16
Minnesota Democrat Win
Mississippi Trump Win 6
Missouri Trump Win  10
Montana Toss Up    3 3
Nebraska Trump Win
Nevada Democrat Win 
New Hampshire Democrat Win
New Jersey Democrat Win
New Mexico Democrat Win
New York Democrat Win   
* North Carolina to Democrat Change 15 -15
North Dakota Trump Win   3
Ohio Toss Up    18 18
Oklahoma Trump Win  7
Oregon Democrat Win   
* Pennsylvania Democrat Win 20
Rhode Island Democrat Win
South Carolina Trump Win 9
South Dakota Trump Win   3
Tennessee Trump Win 11
Texas Toss Up  36
Utah Trump 6
Vermont Democrat Win
Virginia Democrat Win
Washington Democrat Win
West Virginia Trump Win 5
* Wisconsin to Democrat Change 10 -10
Wyoming Trump Win 3

Figure 2

I have also put an asterisk* next to each “battleground” state, where 80% of campaign funds are usually spent. Traditionally, these are the states where each campaign focuses their efforts to win the electoral college. As of this date, four of these battleground states that Trump won in 2016 are now classified as Democratic wins.

Using this arbitrary division, the state by state analysis showed that seven states would move from Trump’s win column to the Democrats favor. This would reduce his electoral college vote by 98. Which would put his total electoral vote at 206, sixty-four vote shy of victory. In addition, seven states moved from his win column to the toss-up category, totaling an additional 111 electoral votes that are now up for grabs. That’s if the election were held today!

The job approval as a predictor has a limited life span and can change as we near the election day. For example, two months before the 2012 election, Barrack Obama’s rating was 48%. But just before the election he crossed the 50% line and was reelected. As we near the 2020, I will revisit the Civiqs data and the state by state results. Stay tuned…

Front Page Money and Elections Theories

Money Can’t Buy Me Love or a Presidential Election

“Money is the Mother’s Milk of Politics.”

Big Daddy Unruh

It is well established that the key element to a successful campaign for public office is the amount of money it raises. For example, between 2000 and 2016, 90% of Congressional House seats were won by the campaign who raised and spent the most money. Here money does matter.

But does money buy success in a Presidential race? It would seem logical, since so much money is poured into a national campaign. Between 1972 and 2016, the Republican and Democratic candidates spent a combined 4.8 billion dollars. That’s B as in Billions! If that kind of money had no effect on the outcome, somebody should be embarrassed.

Studies certainly agree that more spending in lower level races correlates with winning, particularly at the Congressional level. The impact of presidential spending on winning is significantly sparse, even among academic journals. Some look at resource allocation and strategic use of campaign funds but I found no actual studies of how much candidates spend in a presidential election and their success.

So to cure my curiosity, I compiled presidential election data for every election since 1972. In particular, how much money each campaign spent in pursuit of victory, derived FEC records. These expenditures, however, do not include the amount of money spent by independent groups (independent expenditures), which are not required by the candidate to report.

Over the past 12 presidential cycles, the winning candidate out spent their opponent by $618 million dollars. Which on the surface would suggest that more money leads to victory. However, this advantage is not evenly spread throughout each election cycle, as graphically shown in Figure 1.

Winning Candidate Expenditures Less Losing Candidate Expenditures
Figure 1

The values above the zero horizontal line indicate the winner’s expenditures in millions. This chart shows that through most of these elections, the difference between winner and loser spending was relatively modest, until you reach 2008 when Barack Obama out spent John McCain by more than $520 million. This was followed by the $278 million spending advantage that Obama had over Mitt Romney in 2012. And in 2016 Hillary Clinton out spent Donald Trump by a whopping $211 million while losing the contest in the Electoral College.

Except for these three races, the spending deficit of losers does not appear to be overwhelming. For the years 1972 through 2004, the loser only gave up $30 million. But these aggregate figures don’t tell the whole story of whether out-spending your opponent usually leads to the White House.

First, lets see if there is a correlation between spending and winning or losing. We would expect, that if a candidate spends more money than his opponent, absent campaign effects, he would reign victorious. But in the last twelve presidential cycles, outspending the opponent is uncorrelated with winning!

Confirming this, is a scatter-dot graphic that compares the relationship between the incumbent’s (or his Party) percentage of the two-party vote and the difference between the winner and loser’s spending visually demonstrates that the vote is unrelated to winning or losing.

The scatter-dots in the graph are mostly horizontal along the zero spending difference level. We would expect that as the spending difference increases, the percentage of two-party vote would increase as well and that is not the case.

This is also confirmed by the R square of .064, which means there is almost zero relationship between the two variables. In other words, spending more or less than your opponent has no effect on the two-party share of the vote.

To corroborate that money has no effect on winning or losing, I regressed the winner-spending difference on whether the candidate won or loss (confirmed by logistic regression.) The significance level (.315) was far above the <.05 level needed to show a relationship, as shown in Table 1 regression coefficients. Again, it doesn’t matter that a candidate out spent his opponent or not, when it comes to winning.

  Standardized Coefficients             Beta                                    t SIGNIFICANCE LEVEL





So what’s going on here? The cardinal rule states that more money leads to winning in political campaigns, but this analysis says otherwise.

Let’s first differentiate a presidential campaign from every other political contest. In presidential campaigns, every minute is covered by national media and debated twenty-four seven by TV pundits. This is not a race for small town mayor. Unlike even U.S. Senate campaigns, every move by the presidential candidates is recorded, printed and debated. There is no escaping the daily deluge of media coverage unless you move to Mars. Every baby kissed, every misspoken word, and every speech given in a day are recorded and played over and over again.

And then there are the televised debates, where millions of people around the world watch. In 2016, some 84 million viewers watched Hilary and Donald duke it out.

As pointed out in a previous post (“Will the economy save Donald Trump?”), presidential campaigns are often shaped by non-campaign forces such as the economy or a national crisis. Some contests are over before the campaigns reach their stride. If you doubt that, just ask Jimmy Carter. By time election day arrives, even first graders have an opinion who they would vote for (if only the country changed the age requirement.)

Consequently, the effect of paid media diminishes as it is replaced by outside forces, sometimes even outside the control of both campaigns. Spending more on TV commercials at this point is wasting money.

Do I think presidential campaigns will not stop fundraising and buying TV time? Absolutely not! Some traditions are hard to break. Ask President Trump, his presidential campaign has set a fundraising goal of one billion dollars! If he doesn’t win, somebody is going to look pretty stupid.

Economy Front Page Polling Surveys Theories

Will the Economy Save Donald Trump?

After you brewed you morning coffee, many of you went on-line to check the latest updates on Washington’s political news. I know this because that’s what I do every morning. (My wife says I need to get a life. And she’s right, but my addiction is too far gone.) In this morning ritual, I undoubtably check Real Clear Politics’ latest polls, but the one I’m most interested in is the daily Trump Job Approval numbers.

In modern American elections, there is no more ubiquitous statistic than the President’s current job approval percent. Almost every day after a Presidential election, at least one polling company releases their latest results. In 1937, George Gallup created this simple question: “Do you approve or disapprove of the way (NAME) is handling his job as president?”

The original concept of the approval question rested on the theory that the president was the CEO of the country and that a simple and easily understood question would best gauge the county’s “impression” of how well the president was running the business of government. Today, most academics refer to the approval rating as a popularity measure and not a specific indication of how voters think the President is running the business of running the country.

Since Donald Trump was elected in 2016, there have been 1,014 surveys that asked the Presidential Job Approval question. His rating has ranged from a low of 32 percent approve, to a high of 57 percent, with an average approval rating of 45.2 percent during his two years and 360 days in office. As of this writing, that means pollsters have completed almost one survey a day since Trump said, “I swear.”

His average Approval Rating of 45.2%, qualifies him as having the lowest average rating of any first term president since Gallup started asking this question, with the exception of Jimmy Carter who had an average first term rating of 45%.

Since World War II, not a single incumbent presidential candidate has won re-election with a job-approval rating below 50 percent.

The Trump campaign understands that his Job Approval rating is a problem for the upcoming 2020 election, but they are hanging their strategy on the economy. The major talking point for Trump and all his supporters is the economy and not the polls.

Conventional wisdom supports the theory that the president is reelected when the economy is good and loses if it’s bad. And some academic studies seem to support this theory in part, that the economy predicts who wins or loses. The important question is what part of the economy matters most in helping Trump offset his lagging Job Approval ratings, if any?

Will the economy make up for President Trump’s lackluster Job Approval ratings come November 2020?

To determine that, we need to find economic measures that have a significant effect on presidential elections. Most political analysts have focused a three important economic measures that effect political outcomes: GDP growth, inflation, and unemployment.

Using linear regression (OLS) we can measure the impact of each economic variable on each Presidential election since 1972, such as the GDP growth rate, on the incumbent’s (or his party) two-party share of the vote .

For those statically inclined, I have included the models coefficients in the Table below. Of the five economic variables only three are statistically significant (denoted by *): Job Approval, nominal GDP growth and per carpita GDP. (GDP growth rate is significant at the <.08 level). When a variable is not significant it means that it has no correlation (no impact) on the dependent variable, which in this case is the two-party share of the vote. A variable is considered significant at the <.05 level. Some researchers include levels <.10, when the sample is small. Consequently, I’ve included GDP growth rate which is <.075.

Model Coefficients

  Unstandardized  Coefficients          B   Standard Error   Beta   Significance
Constant 55.78 12.047   .004*
Job Approval .327 .085 .783 .008*
GDP Growth Rate .-1.3 .623 -.465 .075*
Per Capita GDP -.003 .001 -.485 .025*
Unemployment -.176 -.047 -..039 .816
Inflation -.237 *.149 -.231 .499

In plain terms, unemployment and inflation have no effect (non-significant) on the President’s percent of the two-party vote, even as the he touts the “lowest unemployment figures ever.” No matter how good the unemployment numbers are (and they are), it won’t help him in November.

Consequentially, I have dropped both inflation and unemployment from the model. Without these two non-significant variables included, the Rsq. for the model is .887, or almost 90% of the model variance is explained by the remaining variables. The Beta coefficient shows that the Job Approval rating has far more impact on Trump’s percentage of the two-party vote than any other variable.

Below is a graphic representation the model’s final estimate for each election since 1972.

Each circle above the red line represents an incumbent (or Party’s) victory and, of course, the circles below a defeat. The model’s estimate of all 12 election’s percent of the two party-vote is off by less than three-tenths of a percent.

Using this simple equation, we can estimate what President Trump’s percentage of the two party vote using his current Job Approval rating, GDP growth rate and a per capita GDP, if the election were today.

Can Donald Trump Win?

The equation is simple: Vote Percent = Constant + Job Approval + GDP percent growth. When we insert Trump’s current Job Approval (45%), and latest GDP growth rate (1.6%) and GDP per capita we get the final equation:

(1) 50.2 + .372 (Job Approval) + (-1.4) x (GDP Growth rate)+ -.003 (Per Capita GDP)

(2) 50.1 + (16.74) +(-2.3) + -16.4 = 48.01%

Simply put, if the election were today, with the President’s current Job Approval of 45% and a GDP growth rate of 1.6% and a per capita GDP of $5463,Trump’s share of the two party vote would be 48%. In other words, today’s economy improves Trump’s vote by about 3%.

If President Trump can raise his Job Approval rating to 48%, his percentage of the two-party vote would 49.3%, increasing by only 1.3%. As his approval rating rises, the economic impact on the popular vote decreases as well, until he reaches 50%, when good economic numbers no longer effects his percentage of the two-party vote. In other words, a 50% approval rating is his Holy Grail.

How Accurate is this Election Model?

We can’t, obviously, measure the actual accuracy of the model’s estimates until after the 2020 election (which I will). But we can test the model’s accuracy against the past 12 presidential elections. When we do, the model estimates the actual percentage of all 12 elections cycles within an average difference of only .19%, which is graphically shown in Chart 9.

Chart 9

The two lines represent the actual vote percent (Blue) and the Model’s estimate (Red). The two images are nearly a mirror image of each other. As visually shown in the chart, the model is a very good fit for the data.

Professor Ray Fair’s January presidential model predicts Trump will win with 54.4% of the two-party vote. Fair’s model relies only on economic data and does not include Trump’s job approval rating in his equation. It also incorporates data back to 1918. His economic models have a good track record, but in 2016 his estimate was off by 5.4%.

Its important to note that the Job Approval rating has short shelf life, and predictions based on it more than two months prior to the election are less likely to accurately measure the two-party vote, even when controlling for the economy.

In addition, this model only applies to the popular vote and not the Electoral College. In modern times, the Electoral College usually mirrors the popular vote. But in the last two decades the popular vote deviated from the Electoral College vote in two key elections: George W. Bush in 2000 and Donald J. Trump in 2016.

I will address the Electoral College vote in a future post. This model, however, does not predict the Electoral College vote.

This Model also predicts the outcome without the the challenging Democratic candidate or the campaign each wages. (There is some evidence that candidate quality does effect the outcome somewhat.)

Political scientists have been predicting presidential outcomes for years with mixed success. After teaching graduate level political science courses for many years, I believe campaigns still matter, but not as much as pundits would have you believe. Each campaign takes place in an environment that is predicated on the political and economic environment.

But there are fundamental elements that shape every election that campaigns and candidates cannot change. Some candidates and campaigns are doomed from the start, as Jimmy Carter can attest to. But many presidential election campaigns can and do have some impact on the outcome. But an incumbent facing a recession headwind, knows the odds of winning are small.

So the question we started with: can the economy put Trump over the finish line even with a Job Approval rating below 50%? The answer is probably not. There is a reason why no incumbent President since World War 2 has never won re-election without an approval rating 50% or higher.

But even with a lagging approval rating this race is likely to be very close. Close enough for a replay of 2016, where Donald Trump carried the Electoral College while losing the popular vote. I will address this possibility in a future post.

When election day arrives, do what mother taught me: vote early and often!