Discrimination against mobile EU citizens before and during the first Covid-19 lockdown: evidence from a conjoint experiment in Germany

Xavier Fernández i Marín, Carolin Rapp, Christian Adam, Oliver James, Anita Manatschal

2021-06-21 13:37:51

Introduction

This report presents the data analysis for “Discrimination against mobile EU citizens before and during the first Covid-19 lockdown: evidence from a conjoint experiment in Germany”, published in the European Union Politics.

Data arrangement and cleaning for covid infection rate (exposure)

source("load_packages.R")
library(readxl)

Generate a variable of exposure to COVID made up of the combination between:

# Original data can be retrieved from:
# https://www.destatis.de/DE/Themen/Laender-Regionen/Regionales/Gemeindeverzeichnis/Administrativ/02-bundeslaender.html
url.population <- "https://www.destatis.de/DE/Themen/Laender-Regionen/Regionales/Gemeindeverzeichnis/Administrativ/02-bundeslaender.xlsx"

#download.file(url.population, destfile = "02-bundeslaender.xlsx")

population <- read_excel("02-bundeslaender.xlsx", 
                         sheet = 2, skip = 7,
                         col_names = FALSE) %>%
  select(Region= `...1`,
         Population = `...4`) %>%
  filter(!is.na(Population)) %>%
  filter(!is.na(Region)) %>%
  mutate(Region = str_sub(Region, 5, -1)) %>%
  mutate(Population = as.integer(Population))
source("load_packages.R")
url.covid.rki <- "https://opendata.arcgis.com/datasets/dd4580c810204019a7b8eb3e0b329dd6_0.csv"
# Uncomment the following line if you need to download the file automatically
#download.file(url.covid.rki, destfile = "dd4580c810204019a7b8eb3e0b329dd6_0.csv")
# or better yet, download it manually, as it is very large as of June 2021 (>300
# Mb), and download.file() may not handle it.
covid.orig <- read.csv("dd4580c810204019a7b8eb3e0b329dd6_0.csv")
head(covid.orig)
##   ObjectId IdBundesland         Bundesland    Landkreis Altersgruppe Geschlecht
## 1        1            1 Schleswig-Holstein SK Flensburg      A00-A04          M
## 2        2            1 Schleswig-Holstein SK Flensburg      A00-A04          M
## 3        3            1 Schleswig-Holstein SK Flensburg      A00-A04          M
## 4        4            1 Schleswig-Holstein SK Flensburg      A00-A04          M
## 5        5            1 Schleswig-Holstein SK Flensburg      A00-A04          M
## 6        6            1 Schleswig-Holstein SK Flensburg      A00-A04          M
##   AnzahlFall AnzahlTodesfall             Meldedatum IdLandkreis
## 1          1               0 2020/09/30 00:00:00+00        1001
## 2          1               0 2020/10/29 00:00:00+00        1001
## 3          1               0 2020/11/03 00:00:00+00        1001
## 4          1               0 2020/11/20 00:00:00+00        1001
## 5          1               0 2020/11/23 00:00:00+00        1001
## 6          1               0 2020/12/18 00:00:00+00        1001
##              Datenstand NeuerFall NeuerTodesfall               Refdatum
## 1 18.06.2021, 00:00 Uhr         0             -9 2020/09/30 00:00:00+00
## 2 18.06.2021, 00:00 Uhr         0             -9 2020/10/29 00:00:00+00
## 3 18.06.2021, 00:00 Uhr         0             -9 2020/11/03 00:00:00+00
## 4 18.06.2021, 00:00 Uhr         0             -9 2020/11/19 00:00:00+00
## 5 18.06.2021, 00:00 Uhr         0             -9 2020/11/18 00:00:00+00
## 6 18.06.2021, 00:00 Uhr         0             -9 2020/12/14 00:00:00+00
##   NeuGenesen AnzahlGenesen IstErkrankungsbeginn     Altersgruppe2
## 1          0             1                    0 Nicht übermittelt
## 2          0             1                    0 Nicht übermittelt
## 3          0             1                    0 Nicht übermittelt
## 4          0             1                    1 Nicht übermittelt
## 5          0             1                    1 Nicht übermittelt
## 6          0             1                    1 Nicht übermittelt

Only work with the data up to the maximum date in which sampling was done. Calculate a 5-days moving average for infection rates on 100,000 inhabitants.

load("sample-eup.RData")
dates.main <- range(I$DateCompleted[I$Sample=="Main"])
dates.crisis <- range(I$DateCompleted[I$Sample=="Crisis"])

# Original data contains entries by age group and sex, and at the Landkreis (local) level
# We remove that and group by region and by date
covid <- covid.orig %>%
  tibble() %>%
  select(Region = Bundesland, 
         Date = Meldedatum,
         Infected = AnzahlFall,
         Death = AnzahlTodesfall) %>%
  mutate(Date = as.Date(str_sub(Date, start = 1, end = 10), 
                        format = "%Y/%m/%d")) %>%
  # Avoid days greater than one month after the last day of sampling
  filter(Date <= (max(I$DateCompleted) + 30)) %>%
  # Filter Infected and Death with -1
  filter(Infected >= 0 & Death >= 0) 

# Calculate total cases grouping age groups and sex
covid <- covid %>%
  group_by(Region, Date) %>%
  summarize(Infected = sum(Infected),
            Death = sum(Death)) %>%
  ungroup() %>%
  # Merge with population
  left_join(population)


# Calculate cumulative cases and infection rates
covid.cum <- covid %>%
  group_by(Region) %>%
  arrange(Region, Date) %>%
  mutate(Infected = cumsum(Infected),
         Death = cumsum(Death)) %>%
  mutate(`Infection rate` = Infected / Population * 1e5,
         `Death rate` = Death / Population * 1e5) %>%
  mutate(`Infection growth rate` = (Infected - lag(Infected)) / lag(Infected),
         `Death growth rate` = (Death - lag(Death)) / lag(Death)) %>%
  mutate(`Infection growth rate smoothed` = zoo::rollmean(`Infection growth rate`, k = 5, fill = NA),
         `Death growth rate smoothed` = zoo::rollmean(`Death growth rate`, k = 5, fill = NA))

Infection rate by 100,000 inhabitants. Figure 1: Infection rate by 100,000 inhabitants. 

covid.cum %>%
  ggplot(aes(x = Date, y = `Infection rate`, group = Region)) +
  geom_line()

Infection rate daily growth rate . Figure 2: Infection rate daily growth rate . 

covid.cum %>%
  ggplot(aes(x = Date, y = `Infection growth rate`, group = Region)) +
  geom_line()