Looking to Class 3: How many 2022 Senate Retirements are Possible?

Author's Notes: While I am not a fan of (American) football, there will not be a post this coming Sunday, February 7th due to Super Bowl Sunday. Stay safe and healthy this coming weekend, regardless if you will be following the big game or not! 

County-Level Determinants and Support for the 2020 Mississippi Flag Referendum

County-Level Determinants and Support for the 2020 Mississippi Flag Referendum

Introduction: In addition to last year's presidential, congressional, and gubernatorial elections, several closely-watched referendum votes took place across the country on varying policy topics. One of the most high-profile referendum votes was the Mississippi Flag Referendum (or Measure 3). In the aftermath of George Floyd's death and intense discussion of past and present racism in America, pressure grew on Mississippi to change its flag, which still bore the Confederate battle flag. The state government demoted the old flag from public use and formed a special commission to create a new state flag design (Pettus, "Mississippi's new magnolia"). 

With broad support, a replacement flag design was chosen by the commission, with the adoption choice to be given to the state's voters through Measure 3; a majority "Yes" vote would support adopting the flag, while a majority "No" vote would have led to the new design not being adopted (Cineas). On Election Day, the referendum passed with almost 75% support from Mississippi voters, a wide swing compared to the similar 2001 flag referendum, which failed by double-digits (Avery). In early January, the state legislature voted on, and Governor Tate Reeves (R) signed legislation that made the new flag official (Pettus, "Mississippi governor"). Today's post seeks to understand which socio-demographic and political determinants were most associated with support for the referendum through analyzing county-level socio-demographic, socio-economic, and political data. In particular, the quantitative analysis portion of the post will further break down the data to determine the same levels of correlation among White Mississippians. 

Visual 1: Mississippi Measure 3 Vote by County

Note: On the scale, a county with a color corresponding to "1" indicates that the difference between the percent who voted "Yes" and the percent who voted "No" was 100%, vice-versa for "-1,". 

Election Results and Polling Analysis: Visual 1 above showcases the results of the referendum by county, showing its broad support. All but two counties (George and Greene) supported the referendum, and only two others (Perry and Tishomingo) supported it by less than a 10% margin. All of the counties containing major urban and suburban areas voted in favor of the referendum, as did all the counties that voted for President Joe Biden (D). Based off county results alone, support was highest in majority-Black counties (especially the "Black Belt" counties in the Mississippi River valley) and all of Mississippi's major urban/suburban counties. Opposition to Measure 3 was highest in the southeastern and northeastern portions of the state, which are rural and majority-White. 

I was only able to find one pre-election poll asking Mississippians about the flag referendum, which was released in October 2020, and polled 507 likely voters. 61% of those polled supported changing Mississippi's flag, including a vast majority of Democratic, Independent, Black, urban, college-educated, and young voters regardless of race. A smaller but fairly large net proportion of suburban, rural, and college-educated White voters also indicated they would vote "Yes," while White voters overall narrowly supported it. Opposition to the measure in the poll was mainly concentrated among Republican voters (50% opposing with 41% supporting), White voters over age 35, non-college-educated White voters, and Trump supporters, although opposition among those groups was not overwhelming ("Mississippi," 7).

Methodology: Table 1 below displays all of the variables that will be used in this post, and how each one is modeled. Election data came from CNN, while all raw socio-demographic and -economic data was transcribed from the Census Bureau's 2019 American Community Survey (ACS) 1-year estimates, available on the Census Bureau's public databank (previously FactFinder). Population density data was transcribed from the Census Bureau's COVID-19 data site, with the most recent data being 8 months past. Links to each source are posted below. 

Table 1: Variables 

Note: All variables refer to the values at the county-level. The dependent variable is notated by (DV), while independent variables are notated by (IV). 

Two different correlation models will be used, one including all of Mississippi's 84 counties, and the other examining counties that are at least 50% White to control for majority-Black counties that are socio-demographically similar to rural majority-White counties (e.g. population, population density, college education attainment rates, etc.). As indicated in Table 1, the dependent variable is "Yes" support for Measure 3, while 8 explanatory variables will be correlated, several of which will utilize data on White individuals per county specifically. 

Results:

Table 2: Correlation Results

Notes: Values in parenthesis refer to the p-value per correlation calculation. Values shaded in red are statistically-significant at the 5% level or lower. 

As shown on Table 1, the independent variables featuring the strongest correlation with the dependent variable were %White, %WhiteCollegeEduc, and the two presidential election candidate variables (Biden and Trump support). When narrowing the units of analysis to majority-White counties, the correlation strength of %CollegeEduc, %WhiteCollegeEduc, Population, and PopDensity all increase significantly. Between the two models, AverageAge (all and White) and MedianIncome show the weakest levels of correlation with referendum support, except for Population and PopDensity when all counties are included in the analysis. 

Analysis and Inferences: While the county-level quantitative analysis conducted above do not perfectly reveal electoral trends because such data aggregates all citizens per county (including those who can't and/or didn't vote), some inferences can be made from the correlation results relative to the pre-referendum poll discussed above. For example, the high level of positive correlation between counties with higher rates of college degree attainment (especially among majority-White areas and White citizens) and higher "Yes" support stands out. On the other hand, locales with higher percentages of White citizens are associated with declining support parallels the exit poll's finding that White voters' pre-election support for the referendum was heavily divided with significant pockets of opposition. This links up to the poll's findings that a majority of urban and suburban voters supported the measure, while rural voters were divided. A surprising result was the lack of significant correlation between median age (all and majority-White) and support for the referendum, as the poll showed that young voters (White and Nonwhite) heavily supported the measure, while older voters were more split (and older White voters opposing it). This is likely a reflection on the usage of county-level statistics, as the total range of median ages was relatively small in this data sample, thus obscuring the actual ages of both voters and those living in the state. Finally, despite the relatively weak level of positive correlation, counties with higher population and population density were associated with higher support, especially when examining majority-White areas. 

Data Sources:

"Average Household Size and Population Density." US Census Bureau COVID-19 Site, covid19.census.gov/datasets/average-household-size-and-population-density-county/data?geometry=-113.942%2C-52.677%2C114.573%2C85.376&orderBy=State. Accessed 27 Jan. 2021. 

"Explore Census Data." United States Census Bureau, data.census.gov/cedsci/. Accessed 27 Jan. 2021. 

"Mississippi Ballot Measure Results." CNN, last updated 16 Dec. 2020, www.cnn.com/election/2020/results/state/mississippi/ballot-measures/1. Accessed 27 Jan. 2021. 

"Mississippi President Results." CNN, last updated 16 Dec. 2020, www.cnn.com/election/2020/results/state/mississippi/president. Accessed 27 Jan. 2021. 

Works Cited:  

Avery, Dan. "Mississippi voters decide to replace Confederate-themed state flag." NBC News, 3 Nov. 2020, www.nbcnews.com/news/nbcblk/mississippi-voters-decide-replace-confederate-themed-state-flag-n1246244. Accessed 28 Jan. 2021. 

Cineas, Fabiola. "Live results for Mississippi’s new state flag ballot initiative." Vox, 4 Nov. 2020, www.vox.com/policy-and-politics/2020/11/3/21546895/live-results-mississippi-measure-3-state-flag-ballot-initiative. Accessed 28 Jan. 2021. 

"Mississippi Survey October 2020." Civiqs, 2020. civiqs.com/documents/Civiqs_DailyKos_MS_banner_book_2020_10_46eq84.pdf. Accessed 27 Jan. 2021. 

Pettus, Emily Wagster. "Mississippi governor signs law for new state flag without rebel sign." Clarion Ledger, 11 Jan. 2021, www.clarionledger.com/story/news/politics/2021/01/11/mississippi-governor-sign-law-new-state-flag-without-rebel-sign/6627621002/. Accessed 31 Jan. 2021. 

---. "Mississippi’s new magnolia flag starting to fly after vote." Associated Press, 4 Nov. 2020, apnews.com/article/election-2020-religion-race-and-ethnicity-mississippi-elections-3e31d01e0e0b8c062ea202b7d8424ecf. Accessed 30 Jan. 2021.