Link to Article

The article I have chosen to analyze was published on February 4, 2020 on FiveThirtyEight and discusses the impacts of the Iowa Democratic Party’s mishandling of the Iowa Caucuses on February 3. The article can be found here.

Article Summary

The author, Nate Silver, discusses the impact of the mistakes during the Iowa Caucuses on the rest of the Democratic nomination process. FiveThirtyEight had developed forecasting models for the Democratic primaries and the impact of Iowa on the rest of the process cannot be overstated. As the first state in a lengthy nomination process, the media attention that the winner of the Iowa Caucuses garners and the impact that can have on voters nation wide is highly valuable. Silver discusses a model FiveThirtyEight developed in order to understand the lift that Iowa could produce. However, this year the Iowa Caucuses happened the day after the Super Bowl, during an impeachment trial, and the day before the State of the Union. This resulted in a lack of media coverage specifically focused on the primary and early polls indicate a lower turnout. However, the main story that is coming out of Iowa is the fact that there are still no results (as I’m doing this assignment at 11:31AM on Tuesday). Silver predicts that this will result in the importance of winning Iowa being greatly reduced. Months of work and focus on this particular state could be wasted as the media narrative will focus largely on the mistakes that led to the delay, not the winner. To understand the potential impact of this, Silver reran the forecasting model as if the Iowa caucuses never happened. For some candidates this was helpful to their chances of winning the nomincation (Biden). For others, this was very damaging (Sanders and Buttigieg). Silver also points out, however, that none of the differences are as large as one might have anticipated. Silver ponders how this might have impacted other candidates who dropped out due to low polling in Iowa. He predicts this might have kept the field more diverse. The results of the forecasting model, however, still predict this will cause a blow to whatever candidate does win when results are released.

Areas of Application

The analysis that Silver does in this article primary revolves on forecasting models. He discusses the model that was built initially to predict the outcomes of the Democratic nomination process and how this model might change if the Iowa caucuses did not occur. It is extremely difficult to build models for primaries because it is hard to quantify the impact of one primary on the next. This makes the models have a high level of uncertainty. If anyone is curious and wants to read about how they made it here is the link.

Author Information

Nate Silver is the founder and editor in chief of FiveThirtyEight. His primary interests are sports and politics. He is well known for developing election forecasting models, like the one he discusses in this article.

Reaction to Article

I found this article extremely interesting because I have been struggling to understand the impacts of the mistakes by the Iowa Democratic Party. Silver’s analysis of the forecasting model with and without the caucuses was very insightful and helped me to better understand how Iowa might change the rest of the nomination process.

Analysis

I downloaded data that FiveThirtyEight published on presidential primary polls to get a sense of the Iowa voters.

library(tidyverse)
library(knitr)
poll_data <- read_csv("president_primary_polls.csv")
## Parsed with column specification:
## cols(
##   .default = col_character(),
##   question_id = col_double(),
##   poll_id = col_double(),
##   cycle = col_double(),
##   pollster_id = col_double(),
##   sponsor_ids = col_number(),
##   pollster_rating_id = col_double(),
##   sample_size = col_double(),
##   sponsor_candidate = col_logical(),
##   internal = col_logical(),
##   tracking = col_logical(),
##   nationwide_batch = col_logical(),
##   candidate_id = col_double(),
##   pct = col_double()
## )
## See spec(...) for full column specifications.

Extract the rows relevant to Iowa and top candidates.

iowa <- poll_data[poll_data$state == "Iowa",]
top <- c("Biden", "Buttigieg", "Klobuchar", "Sanders", "Warren")
candidates <- iowa[grep(paste(top, collapse="|"), iowa$candidate_name),]

See most popular candidate.

barplot(table(candidates$candidate_name), ylim=c(0,100), cex.names=0.7, col = "steelblue")

Figures

Results from Forecasting Model Without Iowa

Biden’s results improve while Sanders drops

Biden’s results improve while Sanders drops

The Person Iowa is Sleeping on

Warren 2020

Warren 2020