Click the Original, Code and Reconstruction tabs to read about the issues and how they were fixed.

Original

Objective

The use of cash in Australia has been on the decline for some time. Back in 2007 almost 75% all in-person transactions were completed with cash by 2019 this had declined approximately 30% (RBA 2020). The emergence of contactless technologies such as PayPass and Apple Pay has seen these become the payment method for choice amongst many Australians for over the counter transactions.

The article “Australia: Not Dead yet, the Ups and Downs of Cash and other Payments” published on the website https://cashessentials.org looks at the issue of that the ‘death’ of cash payments has been exaggerated, the article uses several visualisations of various aspects of Australia’s cash economy such as cash circulating in terms of note denomination, ATM withdrawal figurers and cheque payments, the article also attempts to factor in the impact of the Covid-19 pandemic.

The website https://cashessentials.org does describes itself as:

      “an independent initiative offering a platform for debate about the payments and monetary ecosystem”

This would imply that the site’s content is aimed at individual and groups that operate within the payment function within the Australian financial ecosystem, this would cover a wide range of industries such as banks, cash transport, traders to name just a few.

The chosen visualisation appears in the article under section “The Value Of Atm Withdrawals Is Down, The Number Of Atm Withdrawals Have Dropped Even More” and aims to illustrates the value of ATM withdrawals in Australia over the past 24 months. The data source used would be considered very credible being the Reserve Bank of Australia.

While the chart does address the decreasing dollar value of transactions and this is confirmed by a very obvious decline in the dollar value of transactions from March 2020 onwards the issue of the actual number of transactions is not represented visually anywhere and only mentioned in the text of the article.

Type of chart

The stacked area chart is problematic when trying to identify values of the components that make up the value of ATM transactions. While the major value, in this case, “AU debit card withdrawals”, can be identified, it is difficult to say, for example, what the value of “Overseas Cardholders Withdrawals” may be. Considering that there is a high probability that professionals would be reading the article and assessing the visualisation the value of the individual components would be of interest.

Truncated Axis

The use of truncated axes is considered a deceptive practice in data visualisation, Pandey et all describe it as a visual distortion (Pandey et all, 2015). In this example where a truncated ‘Y’ axis is combined with a stacked area chart it is difficult to see the true ratios of the components that make up total ATM withdrawals. The truncated axis starts at 5,000 and it appears this has been done for visual impact as at first glance the reader would get the impression that ATM withdrawals decreased to almost zero in April 2020 when in reality that figure would be closer to $7 Billion.

Use of Colours

The visualisation uses a green colour scale with varying shades to differentiate between the components, the issue is that the shades of green that have been used are far too close together for the two middle components in the chart making it very difficult to differentiate between the two. Although this could be considered a deceptive practice, I believe this was just poor design by the original author, several other visualisations with the article employ the same colour scale.

References

Article/visualisation being critiqued Other references used

Code

The following code was used to fix the issues identified in the original.

Pre-processing code to bring in the data from the Reserve Bank of Australia.

#Required libraries
library(ggplot2)
library(tidyr)
library(dplyr) 
library(readxl)
library(timeDate)

#Download data from RBA and read in Excel file to data frame
download.file('https://www.rba.gov.au/statistics/tables/xls/c04-1-hist.xlsx','c04-1-hist.xlsx',mode="wb")
atmdata <- read_xlsx('c04-1-hist.xlsx',sheet = "Data")

#remove redundant columns and rename columns
atmdata <- atmdata[c(1,3,11,13,17)]
colnames(atmdata) <- c('Date',
                       'AU debit card withdrawls',
                       'AU credit card withdrawls',
                       'Overseas cardholder withdrawls',
                       "Withdrawls at IAD ATMS's") #Assign userfreindly field names

#remove redundant rows at top of dataframe
atmdata <- atmdata[-c(1:10),] 

#convert from string to intger
atmdata$Date <- as.integer(atmdata$Date) 

#convert from intger to date
atmdata$Date <- as.Date(atmdata$Date, origin = "1899-12-30")

#filter out data not required
atmdata <- atmdata %>% filter(atmdata$Date >= '2018-05-31' & atmdata$Date <= '2020-04-30') 

#use first day of month for month period
atmdata$Date <- as.Date(timeFirstDayInMonth(atmdata$Date))

#change layout of table to use with geom_area
atmdata <- atmdata %>% gather("AU debit card withdrawls",
                   "AU credit card withdrawls",
                   "Overseas cardholder withdrawls",
                   "Withdrawls at IAD ATMS's", key = "TranType", value = "Dollars")

#factor transaction type so as to order correctly
atmdata$TranType <- atmdata$TranType %>% factor(levels=c("Withdrawls at IAD ATMS's",
                                                         "Overseas cardholder withdrawls",
                                                         "AU credit card withdrawls",
                                                         "AU debit card withdrawls"), ordered = TRUE)
#convert dollar amount from string to double
atmdata$Dollars <- as.double(atmdata$Dollars) 

#sort data
attach(atmdata)
atmdata <- atmdata[order(Date, TranType),]

Code recreating the original chart.

# #create stacked area chart
originalPlot <- ggplot(atmdata, aes(x=Date, y=Dollars, fill=TranType)) + geom_area()+
  labs(title = "Australian Dollar Value of ATM Cash Withdrawals",x = "", y = "AUD Millions") +
  scale_x_date(date_breaks = "1 month",date_labels = "%b-%Y") +
  theme(axis.text.x = element_text(angle = 45,vjust=-0.01),
        legend.position="bottom",
        panel.background = element_rect(color = "black",fill = "white"),
        panel.grid.major.y = element_line(color = "black"),
        panel.grid.major.x = element_blank()) +
  guides(fill=guide_legend(title="",nrow = 2)) + 
  coord_cartesian(xlim = c(as.Date("2018-05-01"),as.Date("2020-04-01")),ylim = c(5000,15000),expand=0) +
  scale_fill_manual(values = c("#24E529","#33AA33","#159918","#0A660C"))

Code fixing the identified issues.

# #create stacked area chart - daceted

#re-order factors
atmdata$TranType <- atmdata$TranType %>% factor(levels=c("AU debit card withdrawls",
                                                                 "Withdrawls at IAD ATMS's",
                                                                 "AU credit card withdrawls",
                                                                 "Overseas cardholder withdrawls"), ordered = TRUE)

NewPlot <- ggplot(atmdata, aes(x=Date, y=Dollars, fill=TranType)) + geom_area()+
  labs(title = "Australian Dollar Value of ATM Cash Withdrawals",x = "", y = "AUD Millions") +
  scale_x_date(date_breaks = "6 month",date_labels = "%b-%Y") +
  scale_y_continuous(breaks = seq(0, 12000, by = 1000)) +
  theme(axis.text.x = element_text(angle = 90,vjust=-0.01),
        legend.position="none",
        panel.background = element_rect(color = "black",fill = "white"),
        panel.grid.major.y = element_line(color = "grey"),
        panel.grid.major.x = element_blank()) +
  guides(fill=guide_legend(title="")) + 
  coord_cartesian(xlim = c(as.Date("2018-05-01"),as.Date("2020-04-01")),ylim = c(0,12000),expand=0) +
  facet_grid(~ TranType,labeller = label_wrap_gen())

Data Reference

Reconstruction

The following plot recreates the original plot in R to test the the data downloaded from the RBA. This was done to create a ‘baseline’ and verify that that the data from the RBA lined up with the original visulaisation, it also acted as a starting point from which imporvements could be made.

The new visualistion with the issues identified having been rectified.