When examining the outlook for our local political parties, there is too often a tendency to “nationalize” our perspective. For example, many Colorado Republicans will savor the moment as Donald Trump gives his Inaugural address, because whatever their personal feelings toward him before and even after his election, they are for the most part glad their team won. This is an appropriate response – it’s hard to object to that impulse to celebrate what is in a lot of ways a new lease on life for the national Republican Party, one that was largely unexpected by most observers.
Colorado Democrats, on the other hand, will be tempted toward excessive handwringing, consuming dire reports of the demise of their party from the local to the presidential level and reluctantly resigned to their status in the wilderness as the opposition party for the next two years at the very least. I do take issue with this response, because there is a simple cure for this depression: Look at our state.
Colorado Goes for Clinton
Consider Hillary Clinton, who in the aftermath of the election has been pilloried from all sides as a historically flawed candidate who lost the election more than Trump won it, and then consider this: She won Colorado.
And if that doesn’t cheer you up, if you see her 48% of the vote and think, well, if it weren’t for third-party candidates, Trump surely would’ve won, consider this: Even if you gave Donald Trump every single of one of the votes cast for Evan McMullin and Gary Johnson – an outlandish thing to do considering that Johnson’s support here in Colorado skewed younger and that the Johnson campaign appealed directly to younger voters – as long as you in turn give all of Jill Stein’s votes to Hillary, she still wins. That’s right. Clinton + Stein voters still outnumber Trump + Johnson + McMullin voters by 1,785.
While you digest that statistic, let me be clear about the case I’m making here, or perhaps more precisely the case I’m not making. I’m not saying that I believe that 100% of Jill Stein’s voters would’ve voted for Hillary Clinton were Stein not on the ballot, nor that 100% of Evan McMullin’s voters would’ve voted for Trump were McMullin not on the ballot, nor that all of Gary Johnson’s voters would’ve chosen either Trump or Clinton were Johnson not on the ballot. Surely, these are all voters who voted the way they did for a reason, and a large number of them probably would’ve simply left their choice blank or chosen a different or write-in candidate in this hypothetical scenario. I’m simply trying to establish that the political atmosphere here in Colorado is not an atmosphere that reflects the results of the Electoral College (though Hillary Clinton’s 48.16% of the vote is remarkably similar to her 48.20% in the national popular vote).
Donald Trump’s Performance Compared to Mitt Romney’s in 2012
So that being said, it’s important to ask ourselves the question: Why? Why didn’t Colorado’s voters follow the lead of voters in states like Michigan and Wisconsin, flipping our state red and contributing to Donald Trump’s Electoral College victory? We can begin to see a pretty clear answer in the following scatter plot, as we take a look at the differences between support for Trump and support for his predecessor as the Republican nominee, Mitt Romney. The points are sized to reflect each county’s relative size compared its state, so that, for example, a county that comprises 3% of registered voters in Colorado is the same size as a county that comprises 3% of registered voters in Michigan. I’ve chosen to include Colorado, Michigan and Wisconsin together so that we can see the three states as a group and play a little game of “One of these things is not like the others”:
A couple things stand out here. For one, it’s clear that Colorado’s population is concentrated within a smaller group of counties – there are more relatively “large” points for Colorado than there are for Michigan or Wisconsin. Denver, El Paso, Arapahoe and Jefferson make up 47% of registered voters in the state, and Adams, Larimer, Boulder, Douglas and Weld are all right at or above 5% of registered voters. In Wisconsin, to get to 47%, we have to drop down to the top seven counties (Milwaukee, Dane, Waukesha, Brown, Racine, Outagamie and Winnebago), and only the top three are above 5% of registered voters. And in Michigan, only four counties (Wayne, Oakland, Macomb, Kent) are above 5%.
So let’s focus on these larger counties within each state. In those top nine counties in Colorado, Donald Trump received 881,280 votes. In 2012 in those same counties, Mitt Romney received 885,349 votes. Even with increased turnout across the board, Trump failed to match Romney’s vote total. We can see this in the plot above as only two of these counties gave Trump a higher percentage of the vote than they gave Romney, and even then just barely: Adams (which Trump still lost) and Weld (which Trump won but which is a historically Republican county anyway).
In the remaining seven counties, Trump performed worse than Romney. And even if you point to the increase in votes for Gary Johnson, and to a lesser extent Evan McMullin’s inclusion on the ballot, that still speaks to an overall rejection of Trump’s candidacy.
The stories in Michigan and Wisconsin are very different, albeit for different reasons. In those four largest counties in Michigan, Trump received 890,583 votes, while Romney only received 858,166. Considering that Trump only won Michigan by 10,704 votes, that is a significant difference. He outperformed Romney in Wayne County and flipped Macomb County, and so even before we begin accounting for the outlying counties where Trump smashed Romney’s vote totals, we can see part of the reason why Trump was able to succeed in Michigan.
In Wisconsin, in those seven largest counties, Trump received 551,718 votes compared to Romney’s 605,043. That would seem to suggest a result similar to Colorado, but look at the difference in those vote totals even though the states are similar in size. In Colorado, Trump is relying upon the nine largest counties for a very significant 73% of his overall vote, while in Wisconsin he only needed the top seven counties to account for 39% of his overall vote in order to win. What this means is that Trump’s unpopularity in Wisconsin’s larger counties, particularly in the big three of Milwaukee, Dane and Waukesha (a county which he won handily but performed worse than Romney), is more than canceled out by his performance in the rest of the state.
And that brings us to something else that stands out in the scatter plot. Look at the top right corner of the plot, where we have counties where Trump not only received a majority of the vote, but also exceeded Romney’s performance in doing so. There are some good-sized Wisconsin counties there, including the remainder of our “big seven”: Brown, Racine, Outagamie and Winnebago. Those are all competitive counties, and Trump swept them. We also see Macomb County there, as a large Michigan county that Trump was able to swing significantly into the Republican column. And we see a large number of smaller counties in both Michigan and Wisconsin, several of which saw significant double-digit shifts to Donald Trump. What we don’t see there are hardly any large Colorado counties, Weld being the only one of Colorado’s top fifteen counties where Trump both won a majority of the vote and outperformed Mitt Romney. To find the next largest county that fits that criteria, we have to go reliably Republican Fremont County, which accounts for less than 1% of all registered voters in Colorado (it’s the dot around 4 point Trump-Romney increase and 70% Trump).
Median Household Income and its Impact on Trump’s Performance
While Fremont County does not account for a significant portion of the vote in Colorado, with a median household income of $40,423 it is an ideal segue to a second scatter plot that tells the rest of the story as we account for the difference between these three states. Because simply showing Donald Trump’s performance relative to Mitt Romney isn’t enough. Now, we have to answer why that difference exists. Well, here’s one possible answer:
What we’re looking at here is a plot showing median household income (latest 5-year estimates from the American Community Survey) and how income relates to the Trump-Romney difference, again with a county’s relative size reflected in its size on the plot. What we can see is a clear sign of Trump’s unpopularity in counties with higher median household incomes, led by Douglas County, Colorado with a median household income of over $100,000. While Donald Trump still won Douglas County, he only received 54.7% of the vote compared to Romney’s 62.1% in 2012. Most of Colorado’s larger, more well off counties follow this pattern, with, again, the only outliers being Adams and Weld.
The three largest Wisconsin counties fit the mold as well, as Waukesha County and its $76,545 median household income saw a larger Trump drop-off than Milwaukee County and its $43,873 median household income. Is there something to be said for the fact that Waukesha County is a heavily Republican county, and so there is more opportunity for drop off than there is in Milwaukee, where Romney only received 31% of the vote? Sure. But the fact is that in a statewide election it doesn’t matter where the votes come from. We see another example of this in Michigan, where Trump received 15,119 more votes than Mitt Romney in heavily Democratic Wayne County ($41,210 median household income), while neighboring Oakland County ($67,465 median household income) did not contribute to the Trump wave as Romney received 7,311 more votes than Trump despite higher 2016 turnout overall. Macomb County ($54,582 median household income) was the site of perhaps Trump’s most impressive performance in Michigan as he won 53.6% of the vote, and Macomb was widely covered as an indicator of how Trump was able to win statewide and in the Electoral College. And though the median household income of $54,582 in Macomb County is slightly higher than the national average, there is evidence that the economic recovery from the crash of 2009 has not been felt among the middle class there.
Looking back at Colorado, it is clear that a significant portion of our state’s registered voters reside in counties with relatively high median household incomes. Nearly half (47%) of registered voters in Colorado reside in a county where the median household income is above $60,000, including two of the four largest counties in the state in Arapahoe and Jefferson. When we drop down to counties where the median household income is above $55,000, because there are three larger counties (Adams, El Paso and Larimer) within the $55,000-$60,000 range, there is a significant increase to 74% of registered voters. That is an astonishing figure when compared to Michigan and Wisconsin. In Michigan, only 19% of registered voters reside in counties with a median household income above $60,000, and that figure only increases to 24% when we drop down to $55,000. And in Wisconsin, 25% reside in counties with a median household income above $60,000, and 35% reside in counties with a median household income above $55,000.
If you’re looking for the reason why Donald Trump did not win Colorado, consider his performance relative to Mitt Romney in 2012, and then consider the important statistic of median household income by county and how that affected voters’ opinion of Trump. For Donald Trump to have won in Colorado, he would have had to significantly improve upon Mitt Romney’s performance in 2012. He was unable to do so. And a large reason why he was unable to do so is because three out of four Colorado voters reside in counties with a median household income above $55,000. They do not share in the economic pain and frustration at a slow recovery experienced by fellow voters in Rust Belt areas in Michigan and Wisconsin. To the contrary, many Colorado voters enjoy the benefits of an economy that by most measures outperforms the rest of the country. Those who don’t enjoy those benefits (see: Pueblo County) swung their support to Donald Trump. There just simply weren’t enough of them to make a difference statewide.