polling_error_states.rmd

```{r}
library(readr)
library(tidyverse)

mit_potus_data <- read_csv("1976-2020-president.csv")

state_potus_2020 <- mit_potus_data %>%
    filter(year == 2020) %>%
    rename(state_code = state_po) %>%
    group_by(state_code) %>%
    summarize(
        potus_votes_dem = sum(ifelse(party_detailed == "DEMOCRAT", candidatevotes, 0), na.rm=T),
        potus_votes_gop = sum(ifelse(party_detailed == "REPUBLICAN", candidatevotes, 0), na.rm=T)
    ) %>%
    ungroup() %>%
    mutate(
        vote_share_two_way_dem = potus_votes_dem / (potus_votes_dem + potus_votes_gop)
    )

fivethirtyeight_polls <- read_csv("https://raw.githubusercontent.com/fivethirtyeight/data/master/pollster-ratings/raw-polls.csv")

state_polls <- fivethirtyeight_polls %>%
    filter( 
        type_simple %in% c("Pres-G","Sen-G", "Gov-G")
        & !(methodology %in% c("Face-to-face", "Mail"))
        & location != "US"
        & year %in% c(2008, 2010, 2012, 2014, 2016, 2018, 2020, 2022)
    ) %>%
    mutate(
        state_code = substr(location,0,2),
        state_code = case_when(
            state_code %in% c("N1","N2") ~ "NE",
            state_code %in% c("M1","M2") ~ "ME",
            TRUE ~ state_code
        ),
        methodology = ifelse(grepl("/", methodology, fixed=TRUE), "Mixed Mode", methodology)
    ) %>%
    rename(
        sample_size = samplesize,
        margin_bias_dem = bias,
        district = location
    ) %>%
    select(
        year,
        state_code,
        district,
        type_simple,
        type_detail,
        partisan,
        methodology,
        sample_size,
        margin_bias_dem
    ) 

swings <- c("AZ","FL","GA","IA","ME","MI","MN","NE","NV","NH","NC","OH","PA","TX","WI")

state_polls <- state_polls %>%
    mutate(
        year = as.factor(year),
        is_swing = case_when(
            state_code %in% swings ~ "Swing State",
            TRUE ~ "Non-swing State"
        )
    )

state_polls_avg <- state_polls %>%
    group_by(year, state_code, methodology) %>%
    summarize(margin_bias_dem = mean(margin_bias_dem, na.rm = T)) %>%
    ungroup()

```


```{r}

plotting_data <- state_polls_avg %>%
    left_join(state_potus_2020, by = c("state_code"))

```

```{r}

ggplot(data = plotting_data, aes(x = vote_share_two_way_dem, y = margin_bias_dem, color = year)) +
    geom_hline(yintercept = 0) +
    geom_point(stroke = 0, alpha = .10) +
    geom_smooth(method = "lm", se = FALSE) + 
    theme_minimal() + 
    theme(legend.position="right") +
    labs(title = "Polling Margin Bias by Biden Vote Share", x = "Biden Two-Way Vote Share (2020)", y = "Mean Margin Bias") + 
    facet_wrap(~methodology)

ggsave("state_year_error_method.png")

ggplot(data = plotting_data, aes(x = vote_share_two_way_dem, y = margin_bias_dem)) +
    geom_hline(yintercept = 0) +
    geom_point(stroke = 0, alpha = .10) +
    geom_smooth(method = "lm", se = FALSE) + 
    theme_minimal() + 
    theme(legend.position="top") +
    labs(title = "Polling Margin Bias by Biden Vote Share", x = "Biden Two-Way Vote Share (2020)", y = "Mean Margin Bias") + 
    facet_wrap(~year)

ggsave("state_year_error.png") 

```