Legal Marijuana Is Incredibly Popular With Marginal Voters

Sean McElwee (@SeanMcElwee)

Get the data here.

Marijuana legalization is gaining momentum within the Democratic Party, with Senate Minority Leader Chuck Schumer endorsing decriminalization, Dianne Feinstein primary challenger Kevin de Leon announcing his support for Cory Booker’s legalization bill and an increasing number on red state candidates seeing the issue as a way to partially ameliorate the opioid crisis. In our report, Future of the Party, we made the case that politicians were behind the curve, and even among young Republicans support for legalization is greater than opposition.

Here, I took a look at the possibility that marijuana legalization could energize marginal voters and that marijuana advocates might benefit from focusing on bringing marginal voters into the electorate. I used the American National Election Studies 2016 survey, which includes the following question: “Do you favor, oppose, or neither favor nor oppose the use of marijuana being legal?”

ANES includes a validated vote measure for each national federal election from 2012 to 2016. Individuals who voted in none of those elections I categorize as “non-voters,” individuals who voted in one or two I categorize as “marginal voters” and individuals who voted in all three I call “consistent voters.” For all analysis below, individuals who were ineligible to vote in 2012 are excluded.

As the chart below shows, non-voters and marginal voters overwhelmingly support legalizing marijuana, while consistent voters on net oppose it. While there are often policy divides between consistent voters and marginal voters, the divides on the issue of marijuana legalization are quite dramatic. This suggests that marijuana legalization advocates would benefit from focusing on turnout and mobilization.


I also examined support for marijuana legalization among Democrats by voter status. Here again, I find that inconsistent Democratic voters are more likely to support marijuana legalization. As Democrats work to mobilize the missing Obama millions, marijuana legalization presents one issue that would appeal to these marginal voters.


Anti-Immigrant Sentiment Is Most Extreme In States Without Immigrants

Sean McElwee (@SeanMcElwee)

Yesterday, we released new analysis showing where people believe that undocumented immigrants commit more crimes, based on a question Pew asked in August of 2016.

The question asked,

When it comes to undocumented immigrants currently in the U.S which comes closer to your view — even if neither is exactly right.

  • Undocumented immigrants living in the U.S. are more likely than American citizens to commit serious crimes [OR]

  • Undocumented immigrants living in the U.S. are NO more likely than American citizens to commit serious crimes

To begin, we find that these attitudes benefit from geography. We estimate that 27 percent of people believe immigrants are more violent nationally, but 29 percent of people believe that in the median state. This aligns with our previous research suggesting that anti-immigrant views benefit from political geography.


One thing that stands out looking at the map is that the states with the most anti-immigrant views are also the states where immigrants don’t tend to concentrate. To explore this, we use a variable from the Census Bureau that includes the share of each state that is accounted for by new immigrants arriving between 2010 and 2016.  The average share of new immigrants in states was 1.5. The state with the highest share of new immigrants was New York (where 23 percent of people think immigrants commit more crimes followed by DC (where 12 percent do), and the state with the lowest was West Virginia (where 39 percent of people think immigrants commit more crimes). As the chart below shows, this relationship is strong across the data.


The adjusted R-squared of the model is .62, indicating that the share of a state’s population that is new immigrants explains a good portion of the variation in anti-immigrant views. The coefficient indicates that a 1 point increase in the percent of new immigrants in a state is associated with a four point decrease in anti-immigrant attitudes. The states with the lowest share of new immigrants are the states where people are most likely to believe that undocumented immigrants are more prone to crime.

Anti-Muslim Stereotypes Are Associated With Support for Bombing Iran

Sean McElwee (@SeanMcElwee)

To explore the implications of Jon’s working paper for the present moment, I turn to the 2012 American National Elections (ANES) survey, which includes a question on its web survey asking respondents:

How well does the word 'violent' describe most [X]? [Extremely well, very well, moderately well, slightly well, or not at all / not at all, slightly well, moderately well, very well, or extremely well]?

where [X] includes a series of religious affiliations, including Muslims, Mormons, Catholics, Protestants and Atheists. The ANES also includes several questions about US policy towards Iran. I analyze the following question:

To try to prevent Iran from developing nuclear weapons, would you favor, oppose, or neither favor nor oppose the United States bombing Iran's nuclear development sites.

I recoded the variable so that 1 indicated agreement and 0 indicated either disagreement or neither. Thirty percent of respondents favored bombing Iran, and the rest indicated they were opposed to bombing Iran or did not have a preference either way.

To analyze how stereotypes influence attitudes, I subtracted each individual’s rating of the violence of Protestants from their rating of the violence of Muslims, and then rescaled the variable so 1 indicated a respondent believed the word violent described Muslims “extremely well” and Protestants “not at all well,” .5 indicated parity and 0 indicates an individual believed the word “violent” described Muslims “not at all well,” but Protestants “extremely well."

Below is the distribution of the scale by party. The mean score for all respondents was .61, indicating a bias towards believing Muslims to be more violent. The mean score for Republicans was .68 and the mean score for Democrats was .58 (indicating that Democrats also harbor anti-Muslims stereotypes, though not the extent Republicans do).


I then modeled the relationship between relative beliefs about religious group violence and attitudes toward Iran policy. As controls, I include binary variables indicating race (white and person of color), gender, education (college and non college), and Evangelicalism (born-again or non born-again); as well as 0 to 1 scales indicating age, income, ideology, and partisanship.

I find that stereotyping Muslims is a powerful predictor of support for bombing Iran. Moving from the lowest point in the scale to the highest, holding all other factors equal, increases the predicted probability of support for bombing Iran by 30 percentage points. Men are also a bit more than 10 percentage points more supportive of bombing Iran, and, unsurprisingly, conservatives are more supportive of bombing Iran than liberals.


Racism Benefits From Political Geography

Colin McAuliffe (@ColinJMcAuliffe) and Sean McElwee (@SeanMcElwee)

One of the central goals of Data for Progress is to build tools to understand the political geography of public opinion, and one tool we use is the national-median gap. The idea is quite simple: we compare support for a policy nationally to its support in the median state (a similar standard can be used to measure gerrymandering). What the national-median gap measures is how a policy polls across the nation to how it polls in the middle-most (median) state (i.e. half of states will have lower support and half of states will have higher support).

This measure is important for Democrats to use and understand. Let’s explore a simple example. In 2012, Obama won 52 percent of the national two-party vote. In the median state, he won 52 percent of the two-party vote, producing a national-median gap of less than a percentage point. By contrast, in 2016 Clinton won 51 percent of the national two-party vote, however, in the median state Clinton won 48 percent of the vote, for a national median gap of 3 points, a gap nearly eight times larger than Obama’s (.4). That is, support for Obama was more efficiently distributed throughout the country.

In a previous blog post, we showed that Trump benefited from political geography. Here, we analyze Trump’s signature policy priority: the border wall, which we previously mapped here. We estimate support for the border wall in the median state is 3.3 percentage points higher than its national support. To put that in context, a 3.3 percentage point boost is almost the same size geographic advantage that converted Donald Trump’s popular vote loss into an electoral college win.

Not every policy will have the same advantage or disadvantage due to geography, and each policy needs to be analyzed on a case by case basis. However our preliminary analysis of a number of anti-black and anti-immigrant views suggests that such opinions benefit substantially from the structure of American politics. Given that politics in the US is determined by states, we should mentally add a bit more than three points to national polling support for the border wall. As we saw in the 2016 election, that’s often enough to make a difference. White supremacy is built into the political geography of the United States, and this helps us understand how it plays out in politics.



Trump Is Benefiting From Political Geography

Colin McAuliffe (@ColinJMcAuliffe) and Sean McElwee (@SeanMcElwee)

A key project for Data for Progress is deepening the understanding of how geography interacts with politics. Geography is recognized as an important factor in electoral politics, but is often ignored in opinion polling which typically only estimate national popular support. State level polling is very expensive and rarely performed, particularly for public consumption. Data for Progress uses state of the art statistical methods to estimate state level opinion from national polls, which is cheap enough to make state level estimates routine.

We’re testing out a new way to summarize how opinions are distributed through the states by estimating the support in the median state. Half of states will have higher support than the median while the other half will be lower. That means that if there is majority support for a policy in the median state, then there is majority support in a majority of states. This is a useful way to determine how viable a policy is in the Senate.

Comparing the national popular support to the support in the median state is a simple way to determine whether a policy is helped or hindered by geography. If the national average is bigger than the median, that means that support for a policy is more concentrated in a smaller number of states, as opposed to being spread out evenly across all states. For a recent example of this dynamic, consider the 2016 election, where Hillary Clinton won the national popular vote but lost a majority of states and the electoral college. Based on our analysis of polling to provided us by the Center for American Progress (see their findings here), Trump’s geographic advantage still persists a year after his inauguration. Analyzing Trump's approval in January 2018 shows us that while Trump has low national approval, his support in the median state exceeds his national support by about 3 percentage points, which is the same boost he got from geography on election day 2016. 

Understanding political geography for national electoral politics is just the beginning. Our goal is the use these methods to help advance the most progressive agenda that is feasible at every level of government.