startup-age.qmd

---
title: "Platforms and Power"
subtitle: "Visualising the age to which online startups survive"
author: "James Goldie, 360info. Data: Paul McCarthy"
date: "2022-05-12"
code-fold: true
theme: style/article.scss
---

This analysis reproduces and refines [a graphic of the falling startup survival rates](https://github.com/behavioral-ds/online-diversity/blob/main/scripts/plot_fig6.R) from [McCarthy et al. 2020](https://github.com/behavioral-ds/online-diversity).

```{r}
library(tidyverse)
library(lubridate)
library(here)
library(themes360info)
library(ggtext)
```

This data comes "wide", with columns for each month's observation. We're going to pivot it long,
then calculate the age of startups at any given time.

```{r}
#| label: tidy

survival <-
  read_csv(here("data", "species-survival.csv")) %>%
  rename(birth_year = "...1") %>%
  pivot_longer(-birth_year, names_to = "time", values_to = "survived") %>%
  # calculate the age at any given moment
  mutate(
    birth = ymd(paste0(birth_year, "-01-01")),
    time = ymd(paste0(time, "-01")),
    age_yrs = (birth %--% time) / years(1)) %>%
  filter(!is.na(survived)) %>%
  select(birth_year, time, age_yrs, survived) %>%
  print()
```

Now, let's visualise it! To increase clarity, let's put the focus on the first year, 2005, and on the most current, 2019:

```{r}
#| label: vis

hl_years <- c(2005, 2008, 2018)

line_colours <-
  rep(colours_360("darkgrey"), length(unique(survival$birth_year))) %>%
  set_names(unique(survival$birth_year))
line_colours["2005"]  <- "#2166ac"
line_colours["2008"] <- "orange"
line_colours["2018"]  <- "#e41a1c"

survival %>%
  mutate(
    year_alpha = if_else(birth_year %in% hl_years, 1, 0.15),
    year_width = if_else(birth_year %in% hl_years, 1.5, 0.75),
    ) %>%
  {
    ggplot(.) +
      aes(x = age_yrs, y = survived, colour = factor(birth_year), group = factor(birth_year)) +
      geom_line(aes(alpha = year_alpha, size = year_width)) +
      geom_richtext(aes(label = paste0("Websites from<br>**", birth_year, "**")),
        data = . %>%
          filter(birth_year %in% hl_years) %>%
          group_by(birth_year) %>%
          arrange(desc(age_yrs)) %>%
          slice(1) %>%
          ungroup(),
        family = "Body 360info", fontface = "plain",
        hjust = "left", nudge_x = 0.2, label.colour = NA, fill = NA) +
      scale_x_continuous(expand = expansion(add = c(0, 2.5))) +
      scale_y_continuous(labels = scales::label_percent()) +
      scale_alpha_identity() +
      scale_size_identity() +
      scale_colour_manual(values = line_colours, guide = NULL) +
      theme_360() +
      theme(
        panel.grid.major.x = element_blank(),
        panel.grid.minor.x = element_blank(),
        panel.grid.minor.y = element_blank(),
        plot.subtitle = element_markdown(
          family = "Body 360info", face = "plain")) +
      labs(
        x = "Website age", y = "Survival rate",
        title = toupper("WEBSITE SURVIVAL"),
        subtitle = paste(
          "Around **half of the <span style=\"color:#2166ac;\">websites that began in 2005</span>** are still around 15 years later.",
          "But since then fewer and fewer websites have survived past their first few years.",
          "**Over 90% of the <span style=\"color:#e41a1c;\">websites that launched in 2018</span> failed** within two years.",
          sep = "<br>"),
        caption = paste(
          "**CHART:** James Goldie, 360info",
          "**ADAPTED** from McCarthy et al. 2020",
          "doi.org/10.1371/journal.pone.0249993", sep = "<br>"))
  } %>%
  save_360plot(here("out", "startup-survival.png"), shape = "sdtv-landscape") %>%
  save_360plot(here("out", "startup-survival.svg"), shape = "sdtv-landscape") 
```