The Republican-controlled Supreme Court, empowered by a seat stolen when GOP senators blocked Merrick Garland’s nomination in 2016, has been working hard to give the Republican Party electoral wins, even before Anthony Kennedy's retirement. Most recently, this has meant enabling the purging of voter rolls in a decision that approved of an Ohio effort to suppress the votes of young people, low-income people, and people of color. Decisions like that serve to increase structural barriers in the American political system—from voter registration and felon disenfranchisement to the fact that many offices are decided in low-turnout off-cycle elections.
But despite broad suspicions that this may affect policy, there have been few attempts to measure it. Our think tank, Data for Progress, has designed a new model to do exactly that, and using it we’ve predicted how elections could change if voters fought through these barriers and cast ballots en masse.
Our model is based on a frequently used survey technique called multilevel regression and post-stratification. It’s a big phrase, but the concept is quite simple: Our model identifies the individual determinants (such as race and gender) as well as geographic determinants (for instance whether the geography is urban or rural) of support for a policy from a survey sample. It then “post-stratifies,” or estimates support in the entire population.
This framework makes efficient use of the sample data and allows us to get good-quality estimates of opinion at the state level from national polls. An intuitive way to think about MRP is that knowing the opinions of people in New York should make it easier for us to guess what people in California think about that topic and vice versa. MRP encodes this intuition into a model that integrates information from the respondents in every state and demographic category. This allows us to estimate the opinions of subgroups which make up only a small part of the survey sample.
Typically, polls are post-stratified to the voting-eligible population, but we can extend this method by post-stratifying to a population that reflects the electorate in a given year, or even to hypothetical electorates. We can ask our model not only to tell us what support for a policy in California is, but support would look like if, say, California was 5 percentage points whiter.
This gives us a sense of how varying turnout would shift support for a policy. For this piece, we examine full turnout, 2016-level turnout (a presidential election), and 2014-level turnout (a midterm election). To capture the inherent uncertainty in estimating public opinion, our model simulates a range of outcomes which are consistent with the survey data. In each simulation our model gives us the level of support for a policy in each state. We then count up the number of states where support is estimated to be above 50 percent. The histograms below show the results of 1,000 model simulations estimating support, with the mean estimate marked with a red bar. To put this in a presidential context, in a full turnout environment a $15 minimum wage would win 373 electoral college votes (on average across the 1,000 simulations), compared with 329 votes in a low-turnout environment.
We begin with Medicare for all, a policy which has broad support nationally. (We used data from the Kaiser Family Foundation.) Post-stratifying to the national voting eligible population and simulating full turnout, we find that Medicare for all has support in a supermajority of states. However, when we run a model using the 2014 midterm electorate, support drops by 3 points in the median state. In states with large Latinx and young populations, the dropoff is roughly 5 points. On average across 1,000 simulations, we estimate that in a full-turnout environment, Medicare for all would win 442 electoral college votes, but in a low-turnout environment would win on average 370 electoral college votes. As the chart shows, there is uncertainty in the estimates, but it is clear that lower turnout significantly damages support for of Medicare for all. This also ties into an increasing academic literature, recently explored in a Century Foundation report showing that barriers to turnout skew policy-making toward the healthy.
Next we turned to a $15 minimum wage, another popular progressive proposal, using data from the Pew Research Center. We find that in the median state, the $15 minimum wage loses 2 points moving from full turnout to 2014 turnout. In some states, such as New Mexico, Texas, Nevada, and Arizona—all with large young and Latinx populations—the shift is 4 or 5 points. Here too, we find that the drop in turnout is enough to reduce the the minimum wage below majority in five states, which could have a significant impact in the Senate.
Our results provide some evidence for why Republicans feel compelled to aggressively suppress the ballot: A smaller electorate helps to advance their economic policy priorities. On the other hand, progressives concerned about economic justice must understand that limited access to the ballot is an impediment to economic justice. However, in some states like New York, even Democratic politicians like Governor Andrew Cuomo have suppressed the ballot by doing nothing to advance policies like early voting, automatic voter registration, and same-day registration. We estimate that a low-turnout environment shaves roughly 3 percentage points off of support for Medicare for all in New York.
To be clear, turnout is not the only reason for moderate Democrats' hesitancy to endorse these policies. As we showed in our report with Justice Democrats, many Democratic senators representing states where the majority of their constituents support policies like ending mandatory minimums, a $15 minimum wage, and Medicare for all don’t follow their voters’ lead. Academic research suggests that Medicare for all and other policies might be hampered because politicians fundamentally misunderstand the electorate. Political scientists David Broockman and Christopher Skovron surveyed 3,765 politicians and compared their views to modeled support (also using MRP) for policies in their constituencies. They find that politicians from both parties dramatically overestimate the conservatism of the voters they represent. In another study done by political scientists Nicholas Carnes and Melody Crowder-Meyer, it was found that Democratic Party leaders were far more likely than Republicans to favor centrist candidates, and that leaders in both parties overestimated the conservatism of the electorate.
Our work further confirms what a plethora of academic research shows: Barriers to voting have skewed the electorate rightward, and this problem is particularly pronounced in midterm elections. Any progressive policy can be expected to lose between 1 to 3 points off of national toplines due to turnout. The fight for economic justice cannot be understood separately from the fight for voting access.
Sign up for our newsletter to get the best of VICE delivered to your inbox daily.
Colin McAuliffe is a co-founder of Data for Progress. Follow him on Twitter.