Skip to main content
SpringerLink
Log in

Building a Winning Coalition: Understanding County-Level Support for Donald Trump in the 2016 Election

  • Chapter
  • First Online:
From the Iowa Caucuses to the White House

Part of the book series: Palgrave Studies in US Elections ((PSUSE))

  • 196 Accesses

Abstract

In this chapter, I present a discussion of the academic literature oriented toward presidential elections, Iowa elections, and the 2016 election. The literature is used to theoretically develop models of vote choice using county-level data. The county-level modeling reveals that Trump was able to build a winning coalition between core constituencies of the Republican base (i.e., Republican and evangelical voters), and secure the votes of white, working-class voters across the state. County-level models also indicate that Trump’s improved performance over Mitt Romney in 2012 was driven by white, working-class voters as well, not by the core constituencies of the Republican base.

The big divide was that the Clinton campaign didn’t really show up in rural Iowa, didn’t have a message for rural Iowa, and basically let Trump do whatever he wanted in that area. Trump didn’t really have a ground game that could take advantage of that, but he did have a message that resonated. He had some specific issues that he promised to work on and change and do things differently if he were elected. And Hillary didn’t offer her version of that to those same voters.

—Jeff Link, Iowa Democratic Strategist

This is a preview of subscription content, log in via an institution to check access.

Access this chapter

Log in via an institution
Chapter
USD 29.95
Price excludes VAT (USA)
  • Available as PDF
  • Read on any device
  • Instant download
  • Own it forever
eBook
USD 39.99
Price excludes VAT (USA)
  • Available as EPUB and PDF
  • Read on any device
  • Instant download
  • Own it forever
Hardcover Book
USD 54.99
Price excludes VAT (USA)
  • Durable hardcover edition
  • Dispatched in 3 to 5 business days
  • Free shipping worldwide - see info

Tax calculation will be finalised at checkout

Purchases are for personal use only

Institutional subscriptions

Notes

  1. 1.

    Ideally, the modeling would have included the share of active Independent, or No Party voters as they are called in Iowa. The share of active No Party voters was excluded for two reasons. First, without individual-level data or a comprehensive study of Iowa’s No Party voters to cite, we cannot theoretically understand the linkage between the share of No Party voters and Trump’s share of the two-party vote. Iowa’s No Party voters could be centrists. They also could also be partisan leaners. Since we do not know, it was important to exclude the measure. Second, the association between the share of active Republican registrants and the share of active No Party registrants is negative and statistically significant (r = −0.68). What this demonstrates is that counties with large shares of active Republican voters are less likely to have a robust share of No Party voters. It is more likely, however, for counties to have robust shares of both active Democratic and No Party voters. When both variables (Republican and No Party) are modeled together as explanatory variables of Trump’s share of the two-party vote, the sign for both variables is positive. However, the correlation between the share of active No Party Registrants and Trump’s two-party vote is negative and statistically significant (r = −0.35). The regression result is spurious as controlling for counties with larger shares of Republican voters simply picks up what is going on in counties with higher shares of active Democratic voters, thus not providing a reliable and valid examination of the behaviors of Iowa’s No Party voters. Hoffman and Larimer (2015) found a similar pattern between active Democratic, Republican, and No Party voters in voter registration data from 2014.

  2. 2.

    For the purposes of this chapter, educational attainment (i.e., less than a college degree) is used as a proxy for defining white, working-class voters. While not a perfect proxy for class, data on educational attainment is readily available and is a good predictor of current employment and future job prospects. For a more detailed discussion of proxies for defining the working class, see Abramowitz and Teixeira (2009).

  3. 3.

    The Consumer Price Index (CPI) was used to inflate the measure from year 2000.

  4. 4.

    In Iowa, vote share by race was only calculated for white voters and Hispanic voters in 2016 due to the small proportion of African American and Asian voters in exit polls (Exit Polls 2016b). In 2012, it was only calculated for white voters for the same reason (Exit Polls 2012b).

  5. 5.

    Republican Party of Iowa Chairman Jeff Kaufmann (2019) noted in his interview that political observers from outside the state often believe evangelical Christians are the sole driving force behind the Republican Party of Iowa. Kaufmann believes the Party is much more diverse than that, including a significant group of voters who have populist preferences.

  6. 6.

    The full dataset and R scripts are available from the author upon request.

  7. 7.

    The educational attainment variable had three significant outliers. For all 99 counties, the average for the variable is 21.2% with a standard deviation of 6.9%. The three outliers were Johnson County (53.3%), Story County (48.7%), and Dallas County (48.4%). With the outliers removed, the average for the variable is 20.3% with a standard deviation of 4.7%. All three models were run with the educational attainment outliers removed to see if it substantially changed the findings. It did not. The only significant difference was for the model which predicts Trump’s share of the two-party vote. In the model, the coefficients remain stable, but the change in adjusted median household income measure falls out of statistical significance, most likely because Dallas County saw the largest growth in adjusted median household income and Johnson County saw growth as well. Otherwise, the results are the same. For the other two models, all coefficients remain relatively stable and statistically significant predictors are identical except for the educational attainment variable in the pivot counties model. While its coefficient is slightly smaller, the p-value went from 0.0059 in the model with all 99 counties to 0.0505 when the outliers were removed. As a result, the two regression models and the logistic regression model reported here include all 99 counties.

  8. 8.

    The model was also run using Romney’s share of the two-party vote in 2012 instead of the percentage of Republican registrants. The results were nearly identical, which is not surprising considering that the correlation between the share of Republican voters in 2016 and Trump’s share of the two-party vote is 0.77 and the correlation between Romney’s two-party vote share and Trump’s two-party vote share is 0.89. Several predictors saw minor changes in the weight of their coefficients which did not impact the p-value. However, the size of the coefficient for the evangelical population decreased by 0.016 which shifted the p-value from 0.031 to 0.133. As a result, I report the model with the share of Republican registrants for consistency of predictors across the three models.

  9. 9.

    I hesitate to draw a definitive conclusion here for two reasons. First, the pattern does not hold for counties with urban populations between 2500 and 20,000. The average overperformance was larger for those adjacent counties versus the not adjacent counties. Second, the sample sizes in several categories are relatively small. For counties with urban populations exceeding 20,000, there are only three counties that are adjacent to metro areas and five that are not. For counties with less than 2500 urban population, there are 9 adjacent and 11 that are not.

  10. 10.

    For consistency with the Trump overperformance model, the logistic regression model for pivot counties uses share of Republican voters instead of Romney’s share of the two-party vote in 2012. See endnote 8.

  11. 11.

    The latter explanation is the most likely. After rerunning the model without educational attainment in it, the coefficient for rural–urban continuum became positive, and although it did not reach statistical significance, its p-value was closer to the critical value of 0.05 (p = 0.299 versus p = 0.467).

References

Download references

Author information

Authors and Affiliations

  1. Department of Political Science, Central College, Pella, IA, USA

    Andrew D. Green

Authors
  1. Andrew D. Green
    View author publications

    You can also search for this author in PubMed Google Scholar

Corresponding author

Correspondence to Andrew D. Green .

Rights and permissions

Reprints and permissions

Copyright information

© 2020 The Author(s)

About this chapter

Check for updates. Verify currency and authenticity via CrossMark

Cite this chapter

Green, A.D. (2020). Building a Winning Coalition: Understanding County-Level Support for Donald Trump in the 2016 Election. In: From the Iowa Caucuses to the White House. Palgrave Studies in US Elections. Palgrave Pivot, Cham. https://doi.org/10.1007/978-3-030-22499-8_3

Download citation

Publish with us

Policies and ethics

Search

Navigation