GDP Growth and Chinese new year fireworks

Chinese new year dates

We will scrape wikipedia to extract and clean dates of Chinese New Year. The great rvest package will be our friends for that see example on r-blogger.

url <- "https://en.wikipedia.org/wiki/Chinese_New_Year"
table_xpath <- '//*[@id="mw-content-text"]/div/table[3]'
cny <- url %>%
  read_html() %>%
  html_nodes(xpath= table_xpath) %>%
  html_table(fill = TRUE)
## keep only what we want
cny <- cny[[1]]
cny <- cny[-1,1:4]
names(cny) <- c("year", "date","animal","weekday" )
cny$date <- paste(cny$date, cny$year)
cny$date <- as.character(strptime(cny$date, "%d %b %Y"))
head(cny)

##   year       date  animal   weekday
## 2 2001 2001-01-24   Snake Wednesday
## 3 2002 2002-02-12   Horse   Tuesday
## 4 2003 2003-02-01    Goat  Saturday
## 5 2004 2004-01-22  Monkey  Thursday
## 6 2005 2005-02-09 Rooster Wednesday
## 7 2006 2006-01-29     Dog    Sunday

## extract the New Year eve and night
cny$eve <- as.character(as.Date(cny$date)-1)

PM2.5

We will simply reuse directly the dataset generated in(our previous post)

load(file = "./data/aqi-1.Rda")
names(aqi)

##  [1] "country"               "city"                  "date.time"            
##  [4] "date"                  "year"                  "month"                
##  [7] "day"                   "hour"                  "measurement.available"
## [10] "pm2.5"                 "aqi"

# Remove Paris
aqi <- aqi[city != "Paris"]

The codebook for this dataset will be soon available. We only keep here data for chinese new year date and eve.

aqicny <- aqi[date %in% cny$date | date %in% cny$eve]
aqicny[ , time := hour ] # keep only hours
aqicny[ date %in% cny$date, time := time + 24 ] # add 24 for NY day
aqicny[ , time := time - 24] # re-center on mid-night
aqicny[ , year := as.factor(year) ]

Let’s visualise the evolution of particles during this period. We clearly notice a recurring pattern with peak of pollution around midnight !

g <- ggplot(data = aqicny) +
  theme_tufte() + theme(legend.position = "none") +
  geom_point(aes( x = time,y = pm2.5, col = city), size = 0.5) +
  geom_vline(xintercept = 0, color = "grey") +
  facet_grid(facets = city ~ year) +
  ggtitle("Peak of pm2.5 during chinese new year eve (fireworks)") +
  xlab("Hours from new year's midnight") + 
  theme(axis.text.x = element_text(angle = 90)) +
  theme(axis.title.x = element_text(margin = margin(t = 20, r = 0, b = 0, l = 0))) +
  ylab("pm2.5 in micro g/m3") +
  scale_color_brewer(palette="Dark2") #Dark2 as alternative
p <- plotly::ggplotly(g)

## Fix plotly label spacing
## thanks : 
## https://stackoverflow.com/questions/42763280/r-ggplot-and-plotly-axis-margin-wont-change
p[['x']][['layout']][['annotations']][[1]][['y']] <- -0.06
p[['x']][['layout']][['annotations']][[2]][['x']] <- -0.06
p %>% config(displayModeBar = F) 

#%>% layout(margin = list(l = 75))

GDP Growth

The best source we found is the official China statistic bureau. You can follow the following this link to download the data (Choose National Account / Gross Regional Product Index) - you will need to register to download the file.

## data.stats.gov.cn files tend to be formatted with similar patterns
dataFile <- "https://data.datapleth.io/ext/GDP/GDP_Growth_AnnualbyProvince.csv"
dataFile <- "https://datapleth.sfo2.digitaloceanspaces.com/ext/GDP/GDP_Growth_AnnualbyProvince.csv"

## Not very elegant here but fread returns strange characters which breaks Knitr
gdpGrowth <- as.data.table(
  read.csv(
    dataFile, header = TRUE, stringsAsFactors = FALSE,skip = 3
  )
)
## Remove NA values
gdpGrowth <- gdpGrowth[complete.cases(gdpGrowth)]
## Fix dF names
setnames(gdpGrowth, c("Region", gsub("^X", "", names(gdpGrowth)[-1])))
## get the proper province names
gdpGrowth[ , Region := gsub("^Inner Mongolia$", "Nei Menggu", Region) ]
gdpGrowth[ , Region := gsub("^Tibet$", "Xizang", Region) ]

## Melt as long format
gdpGrowth <- melt(gdpGrowth, id.vars = "Region")
## Get the GDP Growth in %
gdpGrowth[ , value := value - 100 ]
setnames(gdpGrowth, c("Region", "Year", "provincial.gdp.Growth"))
## Get the year in numeric instead of factor
gdpGrowth[ , Year := as.numeric(as.character(Year)) ]


## Get the GDP in percentage of growth
#gdpGrowth[ , provincial.gdp.Growth := (provincial.gdp.Growth - 100)]

## Let's keep only the province for which we have PM2.5 data
listOfProvince <- c("Shanghai", "Beijing", "Sichuan", "Guangdong", "Liaoning")
gdpGrowth <- gdpGrowth[ Region %in% listOfProvince ]
## add the province capital
gdpGrowth[ , city := Region ]
test <- gdpGrowth$Region == "Liaoning"
gdpGrowth[test, city := "Shenyang"]
test <- gdpGrowth$Region == "Guangdong"
gdpGrowth[test, city := "Guangzhou" ]
test <- gdpGrowth$Region == "Sichuan"
gdpGrowth[test, city := "Chengdu"]

g <- ggplot(data = gdpGrowth, aes( x = Year,y = provincial.gdp.Growth), group = city) +
  theme_tufte() + theme(legend.position = "none") +
  geom_segment( aes(x=Year, xend=Year, y=0, yend=provincial.gdp.Growth, group = city), color = "grey") +
  geom_point(aes(col = city)) +
  scale_color_brewer(palette="Dark2") +
  geom_hline(yintercept = 0, color = "grey") +
  facet_wrap(facets = city ~ ., nrow = 1) +
  ggtitle("GDP Growth of associated province") +
  xlab("Year") + 
  theme(axis.text.x = element_text(angle = 90)) +
  ylab("Provincial GDP growth in %")
p <- plotly::ggplotly(g)
## plotly label spacing
p[['x']][['layout']][['annotations']][[1]][['y']] <- -0.15
p[['x']][['layout']][['annotations']][[2]][['x']] <- -0.04
p %>% config(displayModeBar = F)