hawaii_population.Rmd

---
title: "Population Decline"
output: html_notebook
---

## Notebook Purpose

To study Hawaii's population decline. 

```{r setup}
library(tidycensus)
library(tidyverse)
library(ggplot2)

census_api_key(read_file("D:/keychain/census_key.txt"), install = TRUE, overwrite = TRUE)

# Reference for get_estimates() code
# https://walker-data.com/tidycensus/articles/other-datasets.html

# Reference for iteration code
# https://walker-data.com/census-r/wrangling-census-data-with-tidyverse-tools.html#preparing-time-series-acs-estimates

# Reference for Census popest API
# https://www.census.gov/data/developers/data-sets/popest-popproj/popest.Vintage_2021.html

# Reference for Census popest announcements
# https://www.census.gov/newsroom/press-releases/2021/2021-population-estimates.html

# Census glossary of popest terms
# https://www.census.gov/programs-surveys/popest/about/glossary.html

# Compare to DBEDT report on brain drain
# https://files.hawaii.gov/dbedt/economic/reports/Brain_Drain_Hawaii_Born_Population.pdf
# As featured in https://www.civilbeat.org/2022/01/hawaiis-population-drain-outpaces-most-states-again/
```

## Population Components

This section looks at births, deaths, and migration 2015-2021. 

```{r}
# 2020 and 2021 components not available through API, only FTP
recent_years <- read.csv("https://www2.census.gov/programs-surveys/popest/datasets/2020-2021/state/totals/NST-EST2021-alldata.csv", header = TRUE)
hi_recent_years <- filter(recent_years, recent_years$STATE == 15)

# Get rest of years through API
years <- 2015:2019
names(years) <- years

comp_by_year <- map_dfr(years, ~{
  get_estimates(
    geography = "state",
    product = "components",
    state = "HI",
    year = .x
  )
}, .id = "year")

## Merge all years together from both sources

# remove extra geographic columns
hi_recent_years <- hi_recent_years[6:30]
# pivot long, 1 row per variable
hi_recent_years <- pivot_longer(hi_recent_years, everything(), names_to = c("variable", "year"), names_sep = -4, values_to = "value")
# keep raw pop estimates separately in case I need them later
popest <- hi_recent_years[0:5,]
hi_recent_years <- hi_recent_years[-(0:5),]
# remove RESIDUAL, which is minor and does not appear in comp_by_year
hi_recent_years <- hi_recent_years[-(13:14),]
# reorder & row bind with comp by year
hi_recent_years <- hi_recent_years[c(2,1,3)]
comp_by_year <- comp_by_year[-(2:3)]
comp_by_year <- rbind(comp_by_year, hi_recent_years)
# pivot wide, 1 row per variable. remove rates.
comp_by_year <-  pivot_wider(comp_by_year, names_from = "year")
comp_rates <- comp_by_year[7:12,]
comp_by_year <- comp_by_year[0:6,]

# NOTE: 2020 data looks unreliable.

# Visualizations
# TODO line chart with B/D/M/Net over time
comp_by_year %>% ggplot(, aes(x = , y = BIRTHS) + 
                          geom_line())

```

```{r}
pop_by_year <- map_dfr(years, ~{
  get_estimates(
    geography = "state",
    product = "population",
    state = "HI",
    year = .x
  )
}, .id = "year")
```


## Age Group Characteristics

This section breaks down the population by age group to understand 

```{r}
# Age group characteristics only available to 2019
# Later datasets to be released here in 2022: https://www.census.gov/data/datasets/time-series/demo/popest/2020s-state-detail.html
char_by_year <- map_dfr(years, ~{
  get_estimates(
    geography = "state",
    product = "characteristics",
    state = "HI",
    year = .x,
    breakdown = "AGEGROUP", 
    breakdown_labels = TRUE
  )
}, .id = "year")

# TODO Pivot wide, 1 row per year

# Visualizations
# group by summarize ages: https://walker-data.com/census-r/wrangling-census-data-with-tidyverse-tools.html#preparing-time-series-acs-estimates
# TODO animated population pyramid by year

```
```{r}
# TODO could experiment with migration flows as well:
# https://walker-data.com/tidycensus/articles/other-datasets.html
```