COBallotInitiative

How does education level influence support for ballot initiatives when the wording of the measure changes?

Kayla Bond

Amendment T (2016)

“Removal of the exception to the prohibition of slavery and involuntary servitude when used as punishment for persons duly convicted of a crime.”

Failed 50.32%–49.68% with 2,576,759 total votes

Amendment A (2018)

“Prohibits slavery and involuntary servitude in all circumstances, including as punishment for a crime.”

Passed 66.21%–33.79% with 2,416,132 total votes

Support increased by ~16 percentage points, despite 2018 turnout being only ~6 percentage points lower than 2016.

Education as a Proxy

Education (especially higher education) is used as a proxy for reading comprehension
Hypothesis:
- Precincts with lower education may see the largest increase in support once the wording is clarified
- If so, wording (not content) helps determine whether a measure passes or fails

Data and Methods

Precinct-level Election results from the CO Secretary of State - Amendment T (2016) - Amendment A (2018)
Precinct boundaries
- Shapefiles from UF Election Lab (2016 & 2018)
Demographics
- American Community Survey (ACS) 5-year estimates via {tidycensus}:
  - 2012–2016 ACS for the 2016 election
  - 2014–2018 ACS for the 2018 election

ACS Education Variables pulled from ACS Table B15003 (population ages 25+):

acs_vars <- c(
total_edu_25_plus   = "B15003_001",
high_school_diploma = "B15003_017",
GED_or_alt          = "B15003_018",
associates          = "B15003_021",
bachelors           = "B15003_022",
masters             = "B15003_023",
professional        = "B15003_024",
doctorate           = "B15003_025")

These are later collapsed into:

High school equivalent
Associate’s
Bachelor’s
Master’s or higher

Spatial Join

Centroid based join assigns each precinct the ACS values from the tract that contains its geometric center

Precinct geometries and Census tracts rarely align
Approximation only

Precinct Centroid Join (2016)

# get precinct centroids
    precinct_centroids <- merged_clean_2016 |> 
      st_centroid()
    
# join each precinct centroid to the tract it falls in
    precinct_tract_join <- st_join(
      precinct_centroids |> select(Precinct),
      acs_2016_wide,
      join = st_within)

# drop geometry to merge back
    precinct_tract_intersect <- st_drop_geometry(precinct_tract_join)

Precinct Level Support (2016)

Figure A. The amendment ultimately failed statewide.

Highest support in urban areas like Denver and Boulder (up to ~83%)
Much lower support in rural counties (13–16%).

Precinct Level Support (2018)

Figure B. Increased support, the amendment passed.

92–93 percent support in Denver & Boulder
Lowest support in Rio Blanco County (~7%)

Change in Support (2016-2018)

Figure C. Change in support from 2016 to 2018

Support increased across nearly all precincts
Largest gains appear in rural areas

Education and Ballot Support

Education levels come from the ACS for adults age 25 and older grouped into two categories:

Those without a post-secondary degree
(no HS completion, HS diploma or GED, some college but no degree)
Those with a higher degree
(Associate’s, Bachelor’s, Master’s, or above)

Hypothesis: Precincts that are more highly educated should have smaller changes in support.

Figure D.

Relationship between education levels and support for the amendments in 2016 and 2018.

Support is higher across all education levels in 2018 than in 2016.

Filtering Data For Regression Analysis

# --- Data Prep for Regression

cutoff <- 0.35   # 30% with a higher degree

right_2016 <- merged_final_2016 |>
  filter(percent_higher_degree >= cutoff)

# --- Trim extreme outcome outliers (1%–99%) ---
clean_2016 <- right_2016 |>
  filter(
    between(ballot_support_rate,
            quantile(ballot_support_rate, 0.01, na.rm = TRUE),
            quantile(ballot_support_rate, 0.99, na.rm = TRUE))
  )

Regression Results

In 2016, education was positively related to support for Amendment T.

coefficient: 0.426

Becomes roughly linear once at least 35% of adults hold a higher degree.

The 2018 relationship is similar, education remains a strong predictor of support

coefficient: 0.398

Support was higher in every precinct overall, not just those with lower education levels.

Discussion

Support rose across all education levels from 2016 to 2018.

Broad increase, meaning education alone is not a strong indicator of how voters interpreted the ballot language.

Limits:

Approximations due to centroid-based joins and shifting precinct boundaries between 2016 and 2018

Areas for Improvement:

Support stayed high in urban areas, but many rural precincts showed some of the biggest increases
Use of more precise spatial methods.

Conclusion

Education is a strong and stable predictor of support for these amendments, but it does not significantly explain the 16-point jump in 2018.

Clear, accessible language remains essential for ensuring voters (regardless of education level) can accurately interpret what a measure does.

References

Colorado Secretary of State. 2016 General Election – Amendment T Precinct Results.
https://historicalelectiondata.coloradosos.gov/contest/4441

Colorado Secretary of State. 2018 General Election – Amendment A Precinct Results.
https://historicalelectiondata.coloradosos.gov/contest/3934

UF Election Lab. Colorado 2016 Precinct-Level Election Results Dataset.
https://election.lab.ufl.edu/dataset/co-2016-precinct-level-election-results/

UF Election Lab. Colorado 2018 Precinct-Level Election Results Dataset.
https://election.lab.ufl.edu/dataset/co-2018-precinct-level-election-results/

U.S. Census Bureau. American Community Survey (ACS) 5-year Estimates, 2012-2016, 2014–2018. Retrieved using the {tidycensus} R package.